Noncovalent Binding of Small Ubiquitin-related Modifier (SUMO) Protease to SUMO Is Necessary for Enzymatic Activities and Cell Growth*

SUMO proteases possess two enzymatic activities to hydrolyze the C-terminal region of SUMOs (hydrolase activity) and to remove SUMO from SUMO-conjugated substrates (isopeptidase activity). SUMO proteases bind to SUMOs noncovalently, but the physiological roles of the binding in the functions of SUMO proteases are not well understood. In this study we found that SUMO proteases (Axam, SENP1, and yeast Ulp1) show different preferences for noncovalent binding to various SUMOs (SUMO-1, -2, -3, and yeast Smt3) and that the hydrolase and isopeptidase activities of SUMO proteases are dependent on their binding to SUMOs through salt bridge. Expression of Smt3 suppressed the phenotype of yeast mutant lacking smt3, which exhibits growth arrest, and the binding of Ulp1 to Smt3 was essential for this rescue activity. Although expression of an Smt3 mutant (smt3R64E(GG)), which conjugates to substrate but loses the ability to bind to Ulp1, rescued the phenotype of yeast lacking smt3 partially, the mutant cells showed an increment in the doubling time and a delay of desumoylation of Smt3-conjugated Cdc3. These results indicate that the noncovalent binding of SUMO protease to SUMO through salt bridge is essential for the enzymatic activities and that the balance between sumoylation and desumoylation is important for cell growth control.

Small ubiquitin-related modifier (SUMO) 3 family proteins function by covalent attachment to other proteins as posttranslational modifications (1,2). Mammals contain at least three SUMO family members SUMO-1, SUMO-2, and SUMO-3. SUMO-2 and SUMO-3 share 43 and 42% identity to SUMO-1, respectively, but share greater than 96% sequence identity to each other. Saccharomyces cerevisiae (S. cerevisiae) Smt3 shares 43% amino acid identity with mammalian SUMO-1, 39% identity with SUMO-2, and 40% identity with SUMO-3. SUMO modifies many proteins that participate in diverse cellular processes, including transcriptional regulation, nuclear transport, maintenance of genome integrity, and signal transduction (1,2). Covalent interaction between SUMO and its targets is achieved by formation of an isopeptide bond between the C-terminal glycine of SUMO and the ⑀-amino group of a lysine in the acceptor protein (3). This reaction is ATP-dependent and requires the SUMO-activating E1 enzyme Aos1/Uba2, the SUMO-conjugating E2 enzyme Ubc9, and a SUMO E3 ligase (4 -6).
The pattern of SUMO conjugates changes dynamically during the cell cycle and in response to various stimuli (7). SUMO proteases have at least two functions in this process. They remove SUMO from proteins, making the modification reversible, and they also provide a source of free SUMO to be used for conjugation. Free SUMO is generated both from newly synthesized SUMO, which is cleaved to remove a short C-terminal peptide and expose glycine at the C terminus, and from desumoylation of existing conjugates. All known SUMO proteases belong to the ubiquitin-like protease 1 (Ulp1) cysteine protease family, which was originally found in S. cerevisiae (7). Ulp1 proteases are characterized by a C-terminal 200-amino acid core domain that contains the catalytic triad Cys-His-Asp. In addition to the catalytic domain, Ulp1 proteases contain N-terminal domains that are variable in both size and sequence. These domains are critical for intracellular localization and may also have roles in target specificity or regulation of activity (8).
Two SUMO proteases with distinct functions have been described in S. cerevisiae. Ulp1 localizes to the nuclear pore complex (NPC) and is required for cleaving both the Smt3 precursor and Smt3 conjugates to other proteins; whereas Ulp2/ Smt4 localizes to the nucleus, does not cleave the precursor, and appears to desumoylate a distinct set of conjugates (7,9,10). At least seven genes in mammalian genomes encode proteins with the Ulp domain, which constitute the so-called Sentrin-specific protease (SENP) family (11). Among them, SENP1 mainly localizes to foci in the nucleus and the nuclear rim (12), while SENP2 (also called Axam) binds to the nucleoplasmic side of the NPC (13). Although it has been suggested that different members of the SENP family differ in their catalytic properties, the details have not yet been systematically studied.
SUMO proteases are also known to bind to SUMOs noncovalently. SMT3IP1 (SENP3) was originally identified as a Smt3binding protein (14). Furthermore, analyses of the x-ray structures of the complexes between Ulp1 and Smt3, between SENP2 and SUMO-1, and between SENP1 and SUMO-2 have identified some interfaces between SUMO proteases and SUMO proteins (15)(16)(17). It has been shown that six motifs of SUMO proteases recognize the SUMO/Smt3 ␤-sheet and C-terminal strand. Among them, the region of SENP1 that interacts with the C-terminal strand of SUMO is important for the enzymatic activity of SENP-1, whereas the region of SENP1 that recognizes the SUMO ␤-sheet through salt bridge plays less critical roles (17). It has been shown that mutation of the amino acid involved in the salt bridge in Ulp1 cannot rescue the ulp1 phenotype (15), indicating that the noncovalent binding is necessary for biological functions of SUMO proteases. However, the role of the noncovalent binding between SUMO proteases and SUMOs through salt bridge is not clear. In this study we analyzed the physiological significance of the salt bridge between SUMO proteases and SUMOs. Here we show that their noncovalent binding is important for the hydrolase and isopeptidase activities of SUMO proteases and is essential for cell growth.
Strains and Genetic Manipulations-S. cerevisiae strains used in this study are listed in Table 1. W303 was used as a wild-type strain. To generate YHK001, YHK002 was transformed with the ulp1::HIS3 fragment and selected a clone viable on SD-His-Ura plates and non-viable on SD-His ϩ 5-FOA plates. The ulp1::HIS3 fragment was amplified with the primers 5Ј-ATTAAACAGCTTTCAGCAGGTGCCG-3Ј and 5Ј-CCA- CATAAGGTCGAAGACAATCGTC-3Ј. MHY1321 genomic DNA was used as a template for PCR to amplify the ulp1::HIS3 fragment. To generate YHK003, YCp-CDC3-HA::HIS3 was cut with NcoI and introduced into W303, and His ϩ transformants were confirmed by PCR and Western blotting. Preparation of rich (yeast extract-peptone-dextrose, YPD) and synthetic complete (SC) media and standard genetic methods were as described (23).
Enzyme Assays in Vitro-To examine the hydrolase activity of SUMO proteases in vitro, 2 M GST-SUMOs-Myc were incubated with the indicated concentrations of GST-Axam, GST-SENP1, or GST-Flag-Ulp1 in 25 l of reaction mixture (100 mM Tris-HCl, pH 8.0, 2 mM DTT, 1 mM EDTA, and 5% glycerol) at 30°C for 10 min as described previously (21). After the incubation, the mixtures were subjected to SDS-polyacrylamide gel electrophoresis (PAGE) followed by Coomassie Brilliant Blue staining. Band intensities were quantified using NIH Image (version 1.63). The K m value was calculated by averaging the values derived from the fits for each data set.
For analysis of cleavage of SUMO-Tcf-4 by SUMO proteases in vitro, 1 M SUMOs-Tcf-4 were incubated with the indicated concentrations of GST-Axam, GST-SENP1, or their mutants in 25 l of reaction mixture (50 mM Tris-HCl, pH 7.5, 1 mM DTT, 20 mM EDTA, and 150 mM NaCl) for 30 min at 30°C. After the incubation, the mixtures were subjected to SDS-PAGE followed by Coomassie Brilliant Blue staining.
Fluorescence-activated Cell Sorting Analyses-W303 was grown in YPD medium and smt3⌬ background strains were grown in SG-Leu-Ura medium at 28°C. All strains were resuspended in YPD medium to 1 ϫ 10 6 cells/ml and incubated at 28°C. At 0 and 12 h, 1 ϫ 10 7 cells were collected by centrifugation, suspended in 0.2 M Tris-HCl, pH 7.5 and fixed with 70% (v/v) ethanol and stored overnight at Ϫ20°C. The fixed cells were washed with 0.2 M Tris-HCl, pH 7.5 and suspended in 200 l of the same buffer. After a brief sonication (5 s), 100 l of the same buffer containing RNase A (1 mg/ml) was added to the sample, and the incubation was continued for 4 h at 30°C. The RNase A-treated cells were stained with 100 l of propidium iodide (PI) solution (50 g/ml) in 4 mM sodium citrate, 10 mM NaCl, and 0.1% Nonidet P-40 for 15 min on ice. The fluorescence intensities of stained cells were analyzed with a FACSCalibur (Becton Dickinson) and FlowJo (Tree Star). The morphology and nuclear localization of PI-stained cells were observed with an inverted microscope (IX-FLA; OLYMPUS, Tokyo, Japan).
To examine the desumoylation of the Smt3 modification of Cdc3-HA in intact yeast cells, the T-9 strain was arrested at G 2 /M with 15 M nocodazole for 4 h. At the indicated time after nocodazole was removed, cells were lysed in lysis buffer (50 mM Tris-HCl, pH 8.0, 300 mM mannitol, 100 mM KCl, and 1 mM EGTA) containing protease inhibitor mixture (20 g/ml aprotinin and leupeptin, 10 g/ml pepstatin A, and 10 mM PMSF) and 10 mM NEM. After the lysates (200 g of protein) were immunoprecipitated with anti-HA antibody, the precipitates and the samples were further incubated with the indicated amounts of GST-Flag-Ulp1, GST-Flag-ulp1C580S, or GST-Flag-ulp1D451N in 25 l of reaction mixture for 30 min at 30°C. After incubation, the mixtures were subjected to SDS-PAGE and probed with the anti-HA or anti-GST antibody. To quantify the sumoylation level of Cdc3, the intensity of Smt3conjugated Cdc3-HA, which was recognized by anti-HA antibody, was measured by the NIH Image, and normalized with the intensity of tubulin. The sumoylation level of Cdc3-HA at each time was expressed as the percentage of the sumoylated Cdc3-HA before the release from nocodazole arrest.
Other Assays-The immunofluorescence study was done as described (22,24).

Interaction of SUMO Proteases with SUMOs-
The nomenclatures of SUMO-2 and SUMO-3 are confusing at present. According to a previous report (1), throughout this study we will refer to SUMO-2 and SUMO-3 as SUMO proteins that have 11 amino acids and 2 amino acids, respectively, after the C-terminal diglycine motif (Fig. 1A).
Using yeast two-hybrid screening, we isolated SUMO-3 as one of the Axam-binding proteins (data not shown). To clarify the physiological significance of the binding of SUMO-3 and Axam, the specificity of the binding between SUMO proteases and SUMOs was examined biochemically. When the lysates of COS cells expressing GFP-Axam were incubated with GST-SUMOs, GFP-Axam was precipitated with GST-SUMO-3 and GST-SUMO-2 but not with GST-SUMO-1 (Fig. 1B, left panel). GST-SUMO-1, -2, and -3 all precipitated Flag-SENP1, another SUMO protease (Fig. 1B, right panel). Ulp1-Smt3 crystal structural analyses showed the formation of a salt bridge between Arg 64 of Smt3 and Asp 451 of Ulp1 (15). These amino residues are conserved in SUMO-1 (Arg 63 ), SUMO-2 (Arg 58 ), SUMO-3 (Arg 59 ), Axam (Asp 412 ), and SENP1 (Asp 464 ). None of the SUMO mutants in which Arg is changed to Glu could precipitate Axam or SENP1 (Fig. 1B). AxamD412N and SENP1D464N, in which Asp is changed to Asn, did not form a complex with GST-SUMO-3 (Fig. 1C). Furthermore, AxamD412N did not bind to GST-SUMO-2 either, and SENP1D464N interacted with neither GST-SUMO-1 nor GST-SUMO-2 (data not shown). To examine whether the C-terminal region of SUMO is required for the binding to SUMO protease, we generated SUMO mutants in which the C-terminal region after the diglycine motif (GG) or the C-terminal region, including the diglycine motif (⌬GG), is deleted. Axam bound to GST-SUMO-3 (GG) and GST-SUMO-3 (⌬GG) with the similar efficiency to its binding with GST-SUMO-3 (WT) (Fig. 1D, left panel). SENP1 interacted with GST-SUMO-1(GG) or GST-SUMO-3(GG) but not with GST-SUMO-1(⌬GG) or GST-SUMO-3(⌬GG) (Fig. 1D, right panel). These results indicate that the specificity of the binding between SUMO protease and SUMO is dependent on the salt bridge and that the recognition of the SUMO diglycine motif by SUMO protease may be specific.

The Binding of SUMO Proteases to SUMOs Regulates Sumoylation Status in Intact Cells-To show how the salt bridge of SUMO proteases and SUMOs affects sumoylation in intact cells, we expressed various SUMO-1 mutants in HeLa S3 cells and observed their intracellular localization
To see whether these SUMO mutants were present as monomer forms or protein-conjugated forms, lysates of cells expressing various SUMO-1 mutants were analyzed by Western blotting (Fig. 4B). When HA-SUMO-1 was expressed, both the monomer form of SUMO-1 and SUMO-conjugated proteins were observed. Expression of HA-SUMO-1(GG) and HA-SUMO-1(⌬GG) mainly produced SUMO-conjugated proteins and the monomer form of SUMO-1, respectively. Although the monomer form was observed in lysates of cells by expressing HA-SUMO-1R63E, SUMO-conjugated proteins were mainly detected by the expression of HA-SUMO-1R63E(GG). Consistent with these results, Tcf-4 was sumoylated efficiently when SUMO-1R63E(GG) was expressed (Fig. 4C).
Taken together with the data from the biochemical analyses, these results indicate that the SUMO Arg mutants in which the C-terminal region is cleaved are used for sumoylation and that the SUMO Arg mutant-conjugated proteins are not substrates for SUMO proteases in intact cells. Sumoylation of proteins is critical for various cell functions, but it is not clear whether desumoylation itself is necessary. We thought that SUMO-1R63E(GG), SUMO-2R58E(GG), and SUMO-3R59E(GG) would be good tools for analyzing the roles of the salt bridge between SUMO proteases and SUMOs in this point. However, it was hard to effectively knockdown SUMO-1, SUMO-2, SUMO-3, SENP1, and Axam by RNA interference in mammalian cells. Therefore, we used smt3 and/or ulp1 mutant of yeasts.
The Binding of Smt3 and Ulp1 Is Essential for Cell Growth-To test the importance of the salt bridge between Ulp1 and Smt3 in yeast biochemically, the lysates of yeast expressing GFP-Flag-Ulp1 were incubated with various GST-SUMOs. GFP-Flag-Ulp1 formed a complex with GST-Smt3 but not with GST-smt3R64E (Fig. 5A). Furthermore, it formed a complex with GST-SUMO-1 with lower efficiency as compared with GST-Smt3, and did not bind to GST-SUMO-2 or GST-SUMO-3 (Fig. 5A). GFP-Flag-ulp1C580S but not GFP-Flag-ulp1D451N bound to GST-Smt3 (Fig. 5A). GFP-SENP1 but not GFP-Axam showed binding activity for GST-Smt3, and GFP-SENP1 did not bind to GST-smt3R64E (data not shown). GST-Smt3 and GST-smt3(GG) bound to GFP-Flag-Ulp1 at the similar efficiency, but GST-smt3R64E(GG) did not (Fig. 5B). As is the case with the relationship between mammalian SUMO proteases and SUMOs, Ulp1 hydrolyzed the C-terminal region of Smt3 but not that of smt3R64E (Fig. 5C, left panel). In addition to Smt3, Ulp1 cleaved the C-terminal region of SUMO-1 less efficiently, and did not hydrolyze SUMO-2 or SUMO-3 (Fig. 5C,  left panel). ulp1D451N did not show hydrolase activity as well as ulp1C580S (Fig. 5C, right panel). Furthermore, SENP1 but not Axam exhibited hydrolase activity for smt3R64E (data not shown).
We examined the role of the salt bridge between Smt3 and Ulp1 in isopeptidase activity. Smt3 is attached to the septins Cdc3, Cdc11, and Sbs1/Sep7 during mitosis, with conjugates appearing shortly before anaphase onset and disappearing abruptly at cytokinesis (25). GST-Flag-Ulp1 catalyzed the deconjugation of Smt3 from Smt3conjugated Cdc3-HA, which were immunoprecipitated from the T-9 (CDC3-HA) strain arrested with nocodazole in a dose-dependent manner (Fig. 5D). However, GST-Flag-ulp1D451N catalyzed the deconjugation less efficiently and GST-Flag-ulp1C580S did not show isopeptidase activity for Smt3-conjugated Cdc3-HA (Fig.  5D). These results indicate that the hydrolase and isopeptidase activities of Ulp1 are dependent on its salt bridge with Smt3.
To gain further insights into the biological significance of the salt bridge between SUMO proteases and SUMOs, we constructed yeast strains in which the authentic SMT3 gene is deleted and SMT3 (YHK004) or its mutants or human SUMOs (YHK005 and YHK006) are instead expressed under the control of the GAL1 promoter. Consistent with the previous observations (18), the smt3-null strain expressing SMT3 under the GAL1 promoter (T-2) grew on a galactose-containing plate but not on a glucose-containing plate with 5-fluoro-orotic acid (5-FOA), which selects for cells that have lost the URA3marked plasmid (pYESHA-SMT3-URA3) (Fig. 6A). Various smt3 mutants were expressed in the T-2 strain, and the growth on the glucose-containing plate was examined. Although multicopies of smt3R64E and smt3(⌬GG) both failed to rescue the smt3-null strain from the lethality, multicopies of smt3R64E(GG) and smt3(GG) were able to restore the cell growth of the smt3-null mutants comparably to SMT3 (Fig.  6A). Consistent with the results that Ulp1 hydrolyzes SUMO-1 but not SUMO-3 in vitro, overexpressed SUMO-1 but not SUMO-3 could suppress the lethality of the T-2 strain on 5-FOA containing plates (data not shown). Therefore, SUMO-1 is able to substitute for Smt3 in yeast. The ulp1-null mutant has also been reported to be inviable (7). To test whether SENP1 and Axam are able to rescue the ulp1-null mutants from the lethality with a URA3-marked plasmid (YCp50-ULP1) on a 5-FOA containing plates, we constructed plasmids which could express SENP1 or Axam under the GAL1 promoter. SENP1 but not Axam restored cell growth under the conditions in which ULP1 restored cell growth of the ulp1 strains (Fig. 6B). However, none of ulp1D451N, SENP1D464N, and AxamD412N could rescue these strains (Fig. 6B). Therefore, the specificity of the salt bridge between SUMO proteases and SUMOs determines their ability to restore the cell growth in the smt3-or ulp1-null mutants.
Failure of Desumoylation Delays Cell Growth-The smt3null cells expressing smt3 mutants were grown for 12 h in liquid culture containing glucose. The smt3-null cells transformed with vector alone exhibited slow growth, with an increase of the doubling time to 6.5 h (Fig. 7, A and B). The doubling time was recovered to normal by the expression of SMT3 or smt3(GG) in the smt3-null mutants. Expression of smt3R64E did not suppress the growth defect (doubling time was 4.9 h), while smt3R64E(GG) partially suppressed it (doubling time was 2.8 h).
The cell morphology of each mutant was examined after 12 h of growth in liquid culture (Fig. 7, A and B). More than half of the smt3-null mutant cells showed large-or elongated-budded morphology. Similarly to the wild type, about 34 ϳ 40% of the smt3-null cells expressing SMT3 or smt3(GG) were in G 1 phase, as judged by unbudded morphology, and 9 ϳ 11% of the cells were in G 2 /M phase. However, ϳ43% of the smt3-null cells expressing smt3R64E exhibited large-or elongated-budded morphology and remained in G 2 /M phase. Staining of nuclei with propidium iodide (PI) revealed a single nucleus near the bud neck in most cells, indicating that the cells had accumulated at a stage before anaphase. The smt3null cells expressing smt3R64E(GG) showed intermediate phenotypes: the proportions of cells in G 1 phase was decreased to 28.9% and that in G 2 /M phase was increased to 29.4%. One remarkable aspect of the phenotype was the increase in the cells in which a dividing nucleus was detected in the mother-daughter neck of the dividing cells.
Cells were also analyzed by fluorescence-activated cell sorting (FACS). There was no significant difference in DNA content between the wild-type and smt3-null cells expressing SMT3 or smt3(GG) after 12 h of growth in liquid culture containing glucose, whereas the smt3null cells expressing smt3R64E or smt3R64E(GG) exhibited an enrichment of cells with 2N DNA content, which is consistent with the observations that these cells arrest in G 2 /M phase. (Fig. 7C). Furthermore, the smt3R64E(GG)-expressing cells showed an increase of the cells with 1N DNA as compared with the smt3null cells.
Cdc3 is known to be modified with Smt3 during mitosis (25). We expressed CDC3-HA with SMT3 or smt3R64E(GG) in the smt3-null cells and tested the sumoylation status of Cdc3-HA. The cells were arrested at the G 2 /M phase by nocodazole and then released to allow them to enter M phase. At 0 time, Cdc3-HA was modified with Smt3 and smt3R64E at similar efficiency (Fig.  7D). At 60 min after release, the Smt3-conjugated form of Cdc3-HA decreased, and it almost disappeared from the cells expressing SMT3 at 100 min. But, the smt3R64E(GG)-conjugated form of Cdc3-HA was still observed at 100 min after release in the cells expressing smt3R64E(GG). We quantified Smt3-conjugated and smt3R64E(GG)-conjugated Cdc3-HA (Fig. 7D). Although the difference between desumoylation from Smt3-conjugated Cdc3 and smt3R64E(GG)-conjugated Cdc3 was small, we reproducibly obtained the same results showing that desumoylation of Cdc3 occurs more efficiently in the cells expressing SMT3 than those expressing smt3R64E(GG). Therefore, desumoylation is impaired in yeasts by expressing smt3R64E(GG). respectively (15)(16)(17). Structural analyses predicted that Asp 451 in Ulp1, Asp 412 in Axam, and Asp 464 in SENP1 participate in salt bridges with Arg 64 in Smt3, Arg 63 in SUMO-1, and Arg 58 in SUMO-2, respectively. Mutations in either the Asp residue in SUMO proteases or the Arg residue in SUMOs indeed disrupted the SUMO protease-SUMO binding when biochemical binding assays were performed at the submicromolar concentrations. Thus, these amino acids are necessary for noncovalent binding between SUMO proteases and SUMOs. However, in the case of Axam and SUMO-1, or Ulp1 and SUMO-2 or SUMO-3, these amino acids were not sufficient for the interaction. Therefore, our results are not consistent with the findings about the crystal structures of SENP2 (Axam) and SUMO-1 (16). This discrepancy might be due to the differences of the concentrations used between the two studies. Crystallization requires more than millimolar concentrations, while our binding assay was done at less than micromolar concentrations. The importance of the salt bridge between SUMO proteases and SUMOs have not yet been studied systematically, rather it has been shown that the interaction plays a less critical role in hydrolase and isopeptidase activities in SENP1 (17). Therefore, we analyzed the significance of the salt bridge in enzymatic activities. Axam did not bind to or hydrolyze SUMO-1, and Ulp1 did not interact with or hydrolyze SUMO-2 and SUMO-3. Taken together with the observation that disruption of the salt bridge causes loss of the hydrolysis activity of SUMO proteases for SUMOs, the noncovalent binding through the salt bridge is essential for removing a short C-terminal peptide of SUMOs. Several groups, including ours, showed that Axam hydrolyzes SUMO-1 (21,26). When a large amount of Axam was incubated with SUMO-1 for a long time, hydrolysis of SUMO-1 was indeed observed in vitro. It is possible that the salt bridge between SUMO proteases and SUMOs is related to affinity. Therefore, under physiological conditions in intact cells, Axam would neither bind to nor hydrolyze SUMO-1. It has been reported that the amino acid sequence after the C-terminal diglycine motif determines the specificity of hydrolysis of SUMOs by SUMO proteases, as indicated by the finding that replacement of the C-terminal fragment of SUMOs changed the efficiency of hydrolysis of SUMOs by SENP1 and Axam (16,27). The reason why Axam cannot hydrolyze SUMO-2 irrespective of their binding is thought to be due to the C-terminal sequence of SUMO-2. It is interesting to speculate that SUMOs undergo local rearrangements in response to the binding to SUMO proteases, resulting in the interaction of the catalytic domain of SUMO proteases with the SUMO C-terminal sequence.

DISCUSSION
Our results also showed that the salt bridge is important for the isopeptidase activity of SUMO proteases. We used SUMOconjugated Tcf-4 as a substrate. As expected, SENP1 deconjugated SUMO-1, SUMO-2, and SUMO-3 from the substrates in vitro. As a result with mutation of either the Arg residues in SUMOs or the Asp residue in SENP1, the deconjugation reaction was completely suppressed. Consistent with the results of these in vitro studies, Tcf-4 was highly sumoylated by expressing SUMO-1R63E(GG) in intact cells. Therefore, SUMO Arg mutants may be a good tool to modify substrates with SUMO efficiently. Although Axam could not show hydrolysis activity for SUMO-1, it deconjugated SUMO-1 from SUMO-1-Tcf-4. The isopeptidase activity was completely dependent on the salt bridge between Axam and SUMO-1-Tcf-4, which was demonstrated by the biochemical binding assay. Axam also showed the isopeptidase activity for SUMO-1-p53 (data not shown). Therefore, conjugation with substrates may change the structure of SUMO-1 or increase the affinity of SUMO-1 to Axam, resulting in its interaction with Axam through the salt bridge. It is notable that 10-fold lower concentrations of SENP1 and Axam were required for the removal of SUMOs from SUMOconjugated Tcf-4 than the hydrolysis of the C-terminal region of SUMO, suggesting that both proteases interact more readily with SUMO-conjugated substrates. However, the apparent lack of substrate specificity toward SUMO conjugates of Axam and SENP1 compared with SUMO monomer proteins suggests some difference of the mechanisms between the recognition of a monomer form of SUMO and a protein-conjugated SUMO by SUMO proteases.
To show that our in vitro findings using purified proteins are important in intact cells, we used a yeast genetics approach. The smt3-null mutant showed the phenotypes of the ulp1-null mutant, producing elongated cells and failing to undergo cytokinesis. Consistent with the previously reported observations (15), ulp1D451N, which cannot hydrolyze Smt3, did not suppress the lethality of the ulp1-null mutants. Furthermore, SUMO-1 but not SUMO-3 recovered the growth of the smt3null mutants, and SENP1 but not Axam restored the growth of the ulp1-null mutant, consistent with the in vitro results that Ulp1 binds to and hydrolyzes SUMO-1 but not SUMO-3, and that SENP1 but not Axam binds to and hydrolyzes Smt3. These results clearly indicate that the SUMO-cleaving activity of SUMO proteases based on the salt bridge between SUMO proteases and SUMOs correlates with their ability to regulate cell growth. The importance of this noncovalent binding is conserved in other ubiquitin-like modifier and protease systems. Atg8, a member of a novel ubiquitin-like protein family, is an essential component of the autophagic machinery in yeast and is recognized by Atg4, a cysteine protease (28). It has been reported that the conserved Phe residue in the same region as Smt3R64 is essential for the binding of Atg8 to Atg4, cleavage by Atg4 activity, and autophagic pathways (29). Smt3R64E(GG) indeed overcame the cell cycle abnormality in the smt3-null cells, but the cells still exhibited some defects. The ratio of the cells in which a nucleus was observed on both sides of the bud neck increased in the smt3R64E(GG)-expressing cells compared with the smt3R64E-expressing cells. Although we did not examine all of the sumoylated proteins, at least desumoylation of Cdc3 was suppressed in the smt3R64E(GG)-expressing cells compared with wild-type cells. Therefore, impairment of desumoylation of proteins including Cdc3 may be involved in the delays of cytokinesis. Thus, the balance between the sumoylation and desumoylation of the SUMO target proteins is important for cell cycle control. However, inhibition of desumoylation of smt3R64E(GG)-conjugated Cdc3 was less prominent than expected from the results with in vitro experiments. It has been shown that SUMO-1modified RanGAP1 is protected from SENP2 activity when complexed with Nup358 and Ubc9 (13). By analogy with this observation, Smt3-conjugated Cdc3 may bind to protein(s) in yeast, thereby it is not easily accessible to endogenous Ulp1. This might be the reason why the difference between desumoylation of Smt3-conjugated Cdc3 and Smt3R64E(GG)-conjugated Cdc3 is small.
Although PIAS family members, SUMO E3 ligases, are also known to bind to SUMO or SUMO-conjugated proteins (24,30), the physiological roles of such binding are not clear. Clarifying the biological significance of the noncovalent interaction between SUMOs and enzymes that regulate sumoylation would be necessary to understand the molecular mechanisms of the determination of the substrate specificity of sumoylation and the subcellular localization of SUMO-conjugated proteins.