Modification of Ran GTPase-activating Protein by the Small Ubiquitin-related Modifier SUMO-1 Requires Ubc9, an E2-type Ubiquitin-conjugating Enzyme Homologue*

Covalent modification of the Ran GTPase-activating protein RanGAP1 with the ubiquitin-related protein SUMO-1 promotes its association with Nup358, a component of the cytoplasmic fibrils emanating from the nuclear pore complex (1, 2). InXenopus egg extracts, Nup358 can be found in a complex with Ubc9 (3), a structural homologue of the E2-type ubiquitin-conjugating enzymes (UBCs). Here we show that a subset of the human homologue of Ubc9 (HsUbc9) colocalizes with RanGAP1 at the nuclear envelope. HsUbc9 forms thiolester conjugates with recombinant SUMO-1, but not with recombinant ubiquitin, indicating that it is functionally distinct from E2-type UBCs. Finally, HsUbc9 is required for the modification of RanGAP1 by SUMO-1. These results suggest that HsUbc9 is a component of a novel enzymatic cascade that modifies RanGAP1, and possibly other substrates, with SUMO-1.

The transport of selected proteins and nucleic acids across the nuclear pore complex (NPC) 1 is regulated by the small GTPase Ran/TC4 (4,5). Ran alternates between a GTP-bound state and a GDP-bound state, a transition facilitated by a nuclear GTP-exchange factor (RCC1) and a cytoplasmic GTPase-activating protein (RanGAP1) (reviewed in Refs. 6 and 7). The differential binding of Ran-GTP and Ran-GDP to carrier proteins (e.g. importins, transportin, and NTF2) and NPC proteins provides a mechanism that allows Ran to regulate the transport of cargo across the nuclear pore complex. In this system, the subcellular localization of RCC1 and RanGAP1 are major determinants of directed nuclear transport.
The localization of RanGAP1 to the nuclear pore complex is imparted by its specific association with Nup358/RanBP2 (1), a component of the cytoplasmic filaments emanating from the NPC (8,9). This interaction requires the covalent modification of RanGAP1 with a ubiquitin-related protein designated SUMO-1 (1,2). Recently, Ubc9, a structural homologue of the E2-type UBCs, was shown to coimmunoprecipitate with Nup358 in Xenopus egg extracts (3). Previous studies indicated that Ubc9 may be required for regulating the progression through the cell cycle in yeast (10,11). In addition, the human homologue of Ubc9 (HsUbc9) can functionally substitute for yeast Ubc9 (12,13). Although HsUbc9 has been proposed to mediate the ubiquitin-catalyzed degradation of mitotic cyclins (10 -12), no bona fide ubiquitin-conjugating activity has been demonstrated for HsUbc9.
We now show that HsUbc9 differs significantly from the E2 family of UBCs in that it can form a thiolester intermediate with SUMO-1, rather than ubiquitin. This enzymatic activity is ATP-dependent and is not observed in mutants of HsUbc9, which contain a single amino acid substitution at the putative catalytic site. Finally, HsUbc9 is required for the modification of RanGAP1 by SUMO-1.

MATERIALS AND METHODS
Preparation of Affinity Purified Antibodies-Rabbit antiserum raised against Escherichia coli-derived recombinant HsUbc9 was affinity purified by passage over a CNBr-activated Sepharose B matrix coupled to E. coli-derived recombinant HsUbc9 at a final antigen concentration of ϳ2-5 mg/ml, using methods described (14). Monoclonal antibodies to RanGAP1 (19C7) and SUMO-1 (21C7) were previously described (2). Rabbit anti-BSA antibody preparations were obtained from Calbiochem (San Diego, CA), and IgGs were purified by chromatography on protein G-Sepharose.
Immunofluorescence Cell Microscopy-HeLa cells were plated on coverslips and cultured for 2 days in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. When the culture reached 50 -75% confluence, the cells were fixed in 2% paraformaldehyde for 10 min at room temperature and methanol at Ϫ20°C for 10 min. The fixed cells were blocked for 1 h in 5% normal goat serum in PBS at room temperature and then incubated for 1 h at room temperature with a 1:3000 dilution of 1 mg/ml affinity purified polyclonal antiserum raised against HsUbc9 or with a 1:2000 dilution of monoclonal antibodies raised against RanGAP1 (19C7) in 2.5% normal goat serum. Cells were washed in PBS and then incubated with goat anti-rabbit IgG-rhodamine or with goat anti-mouse IgG-fluorescein diluted 1:400 in 2.5% normal goat serum. Cells were washed in PBS and mounted for immunofluorescence microscopy.
To prepare cell extracts for the RanGAP1 modification assay, ϳ1 ϫ 10 7 HeLa cells were washed and resuspended in 700 l of buffer A (20 mM HEPES, pH 7.3, 110 mM potassium acetate, 5 mM sodium acetate, 2 mM magnesium acetate, 1 mM EGTA, 2 mM DTT) containing 100 g/ml digitonin, 1 mM phenylmethylsulfonyl fluoride, and 1 g/ml aprotinin and left on ice for 15 min. Cells were centrifuged at 13,000 rpm for 15 min at 4°C, and the supernatant was transferred to a fresh tube.
In Vitro Translation and SUMO-1 Conjugation-HsUbc9 cDNA was cloned into the pSP65 expression vector. Mutants of HsUbc9 cDNA were generated by PCR-mediated site-directed mutagenesis (14). Linearized plasmids were transcribed in vitro using SP6 polymerase, and newly synthesized RNAs were briefly treated with DNase. RNAs were translated in vitro using rabbit reticulocyte lysate (Promega) and [ 35 S]methionine (10 mCi/ml). In vitro translated proteins were tested for their ability to form thiolester intermediates with ubiquitin (Sigma) or recombinant GST-SUMO-1. GST-SUMO-1 (SUMO1⌬C4) was synthesized with the four carboxyl-terminal amino acids (HSTV) deleted, exposing two glycines at the carboxyl terminus (15). To test for SUMO-1 conjugating activity, in vitro translated HsUbc9 was incubated with ϳ0.5 g of recombinant GST-SUMO-1 or ubiquitin in reaction buffer (50 mM Tris, pH 7.6, 5 mM MgCl 2 , 0.1 mM DTT, 4 mM ATP) for 1 h at 25°C. For the ATP depletion experiment, 10 units of hexokinase (Sigma) and 5 mM glucose were added to, and ATP omitted from, the reaction. After completion of the conjugation reaction, 5 mM dithiothreitol and sample buffer were added to the indicated sample, and the sample was boiled for 5 min. Samples which were not treated with dithiothreitol were incubated at room temperature for 20 min after addition of sample buffer. Reaction products were analyzed by SDS-PAGE followed by autoradiography.
Expression vectors encoding wild-type and mutant RanGAP1 constructs (16) were linearized and transcribed in vitro using T7 polymerase, and transcripts were translated using a rabbit reticulocyte lysate system as above. To create a RanGAP1 mutant that is unable to form a conjugate with SUMO-1, the codon corresponding to lysine at position 526 was mutated to encode arginine. The K526R (Lys3 Arg) RanGAP1 mutant is unable to form SUMO-1 conjugates in vitro or in vivo (15,16) Immunodepletion of HeLa Cell Extracts and RanGAP1 Modification Assay-HeLa cell extracts were incubated with protein A-Sepharose beads that were prebound to rabbit antiserum directed against HsUbc9, with preimmune serum, or with monoclonal antibody reactive against SUMO-1 (21C7) for 2 h on a rotator at 4°C. The immune complexes were precipitated, and the extracts were subject to an additional round of immunodepletion. The presence of the targeted protein in the immunodepleted extracts was below the limits of detection as indicated by immunoblot analysis (data not shown).
Wild-type or mutant RanGAP1 translated in vitro was incubated with 20-l HeLa digitonin extracts in the presence of an ATP regenerating system (50 mM Tris, pH 7.6, 1 mM ATP, 10 mM creatine phosphate, 10 g/ml creatine phosphokinase, 1 unit/ml yeast inorganic pyrophosphatase, 2 mM dithiothreitol, 5 mM MgCl 2 ) for 10 min at room temperature. Some reactions were supplemented with recombinant HsUbc9, HsUbcH5, or SUMO-1 (HSTV deletion mutant; Ref. 15). Reaction mixtures were boiled in sample buffer for 5 min and analyzed by SDS-PAGE followed by autoradiography.

RESULTS
Previous studies indicated that Ubc9 could be found in a complex with Nup358 (3), a large protein found at the cytoplasmic periphery of the NPC. Other studies indicated that Ran-GAP1 modified by the ubiquitin-related molecule SUMO-1 could also interact with Nup358 (1). Together, these results suggested the possibility that HsUbc9 could catalyze the modification of RanGAP1 with SUMO-1. To determine whether HsUbc9 and RanGAP1 might interact at the NPC in vivo, we compared the subcellular localization of these proteins in HeLa cells using immunofluorescence microscopy. As described previously, RanGAP1 was found to be distributed in a diffuse pattern throughout the cytoplasm and more densely at the rim of the nuclear envelope (Fig. 1B). Although HsUbc9 appears to be localized primarily to the nucleus and cytoplasm, a distinct subset of HsUbc9 also exhibits staining at the rim of the nuclear envelope (Fig. 1A). Interestingly, as was reported for RanGAP1 (2), HsUbc9 associates with the spindle apparatus in mitotic cells (not shown). The similarity in the subcellular distribution of RanGAP1 and HsUbc9, particularly at the rim of the nuclear envelope, supports the possibility that RanGAP1 may be a substrate for SUMO-1 modification by HsUbc9.
We further investigated the subcellular localization of Ubc9 by immunoblot analysis of rat liver homogenates using antibodies raised against HsUbc9. Previously, RanGAP1 was shown to be associated with the nuclear envelope fraction of rat liver homogenates (1,2). Equivalent numbers of cells were loaded in each lane ( Fig. 2A). A single protein that is recognized by antibodies reactive with HsUbc9 was found to be present in both the cytosolic and nuclear fractions ( Fig. 2A, lanes 4 and 5). Although the immunofluorescence microscopy data suggests that more of the protein resides in the nucleus than is indicated by the immunoblot, it is possible that some of the Ubc9 diffuses out of the nucleus during the subcellular fractionation due to its small size (18 kDa). DNase and RNase digestion of the nuclei, which releases many nuclear components into the supernatant, failed to liberate Ubc9 ( Fig. 2A, lane 6), indicating that a significant fraction of the nuclear Ubc9 is associated with the nuclear envelope. The insoluble nuclear envelope fraction that remained after two nuclease digests of nuclei also contained Ubc9 (Fig. 2A, lane 7). The nuclear envelope fraction represents ϳ0.3% of the total protein, yet contains an estimated 5-10% of the total Ubc9. This indicates that Ubc9 is concentrated at the nuclear envelope and is consistent with the immunofluorescence data shown in Fig. 1.
To test if RanGAP1 associates with Ubc9 at the NPC, the nuclear envelope fraction was solubilized and immunoprecipitated using protein A-Sephacryl in the absence or presence of antibodies reactive with BSA (as a specificity control), Ran-GAP1, or Ubc9. Immunoprecipitates were separated on SDSpolyacrylamide gels, transferred to nitrocellulose, and blotted with antibodies reactive with either Ubc9 (Fig. 2B, top panel) or RanGAP1 (Fig. 2B, bottom panel). Immunoprecipitates prepared using antibodies reactive with RanGAP1 included both Ubc9 (Fig. 2B, lane 3, top panel) and RanGAP1 (Fig. 2B, lane 3,  bottom panel), neither of which were detected in control precipitates (Fig. 2B, lanes 1 and 2, top and bottom panels). The two isoforms of RanGAP1 were previously shown to represent unmodified and SUMO-1 modified RanGAP1 (1, 2). In addition, RanGAP1 was present in immunoprecipitates prepared using antibodies reactive with Ubc9 (Fig. 2B, lane 7). Comparison of immunoprecipitates and supernatants after immunodepletion using control antibodies or antibodies reactive with RanGAP1 revealed that similar fractions of RanGAP1 and Ubc9 were depleted by the antibodies (Fig. 2B, lanes 1-6). Taken together, these data indicate that most, if not all, Ubc9 found at the nuclear envelope is present in a complex with RanGAP1. The preferential immunoprecipitation of unmodified RanGAP1 (compare the relative intensities of both RanGAP1 bands in lanes 3 and 6) is most likely due to antibody accessibility since modified but not unmodified RanGAP1 is complexed with Nup358 in the nuclear envelope extract (1).
If HsUbc9 functions like an E2-type UBC in the covalent addition of SUMO-1 to specific substrates, it should form a thiolester conjugate with SUMO-1 (17). We compared the electrophoretic migration of recombinant HsUbcH5 (a bona fide E2-type ubiquitin-conjugating enzyme that mediates the ubiquitination of p53 (18)) and HsUbc9 translated in the presence of [ 35 S]methionine in reticulocyte lysates supplemented with either ubiquitin or GST-SUMO-1 (Fig. 3A). GST-SUMO-1 was used in these experiments because a portion of in vitro translated HsUbc9 constitutively forms an ϳ36-kDa species that co-migrates at the same molecular mass as HsUbc9:SUMO-1 conjugates. In addition, GST-SUMO-1 was prepared with the four carboxyl-terminal amino acids (HSTV) deleted to facilitate conjugation (15). Addition of exogenous ubiquitin (Fig. 3A, lane  2), but not GST-SUMO-1 (Fig. 3A, lane 3), altered the migration of recombinant HsUbcH5 (from ϳ16 to ϳ24 kDa), consistent with the formation of the expected HsUbcH5:ubiquitin conjugate. The identity of the ϳ28-kDa species is unknown (Fig. 3A, lane 1), but its disappearance in the presence of ubiquitin suggests that it may be an adduct of UbcH5. In contrast, addition of exogenous GST-SUMO-1 (Fig. 3A, lane 6), but not ubiquitin (Fig. 3A, lane 5), altered the migration of recombinant HsUbc9, consistent with the formation of a HsUbc9:GST-SUMO-1 conjugate. Just as the formation of E2type UBC:ubiquitin conjugates is ATP-dependent, the putative HsUbc9:SUMO-1 conjugate was not observed in lysates depleted of ATP by hexokinase treatment (Fig. 3A, lane 7). Just as the thiolester bond between E2-type UBCs and ubiquitin is disrupted in the presence of reducing agents, the putative HsUbc9:GST-SUMO-1 conjugate was not observed in samples treated with dithiothreitol (Fig. 3A, lane 8). Interestingly, both ATP depletion and dithiothreitol treatment also abolished the presence of the 36-kDa HsUbc9 species (Fig. 3A, lanes 7 and 8), suggesting that this species represents a thiolester intermediate between HsUbc9 and endogenous SUMO-1. The apparent increase in the intensity of the 36-kDa band in the presence of ubiquitin (Fig. 3A, lane 5) may be due to a contaminant in the ubiquitin preparation used in the experiment, as the overall background signal is higher in those samples supplemented with ubiquitin (Fig. 3A, lanes 2 and 5).
Sequence comparison of HsUbc9 with known E2-type UBCs suggests that the cysteine residue at position 93 is the active site for thiolester conjugation (19), a conclusion that is supported by genetic reconstitution analysis of Ubc9 mutants (12,13). This prediction was tested by comparing the electrophoretic migration of 35 S-labeled recombinant wild-type (W) or mutant (M) HsUbc9 translated in reticulocyte lysates supplemented with buffer control (Fig. 3B, panel a), GST-SUMO-1 (Fig. 3B, panel b), or ubiquitin (Fig. 3B, panel c). The mutant HsUbc9 contains a cysteine to alanine substitution at position 93. Whereas the addition of ubiquitin had no effect on the  3 and 6), or anti-Ubc9 antibodies (lane 7) preadsorbed to protein A beads or with protein A beads only (lanes 1 and 4). Equivalent amounts of antibody-antigen complexes (lanes 1-3 and 7) and the immunodepleted supernatants (lanes 4 -6) were separated on a 15% SDS-polyacrylamide gel and immunoblotted with 1.5 g/ml anti-Ubc9 antibodies (top panel) or separated on an 8% SDS-polyacrylamide gel and immunoblotted with 0.5 g/ml anti-RanGAP antibodies (bottom panels). *, IgG heavy and light chains. migration of wild-type HsUbc9 (Fig. 3B, panel c), the addition of GST-SUMO-1 resulted in the appearance of a labeled protein migrating with a molecular mass of ϳ65 kDa, consistent with the formation of a HsUbc9:GST-SUMO-1 conjugate (Fig. 3B,  panel b, W). Substitution of cysteine with alanine at position 93 of HsUbc9 abolished its ability to form thiolester intermediates with GST-SUMO-1 (Fig. 3B, panel b, M), suggesting that the active site for thiolester conjugation to both SUMO-1 and ubiquitin are conserved between members of this structural family. Moreover, the 36-kDa species present in wild-type HsUbc9 translation reactions is not formed when mutant HsUbc9 is synthesized in vitro, indicating that newly synthesized wildtype HsUbc9 but not mutant HsUbc9 is able to utilize SUMO-1 found endogenously in the rabbit reticulocyte lysate (Fig. 3B,  panel A).
We used a similar approach to compare the modification of in vitro translated wild-type and mutant RanGAP1 with SUMO-1 (Fig. 4). In vitro translation of wild-type RanGAP1 produced the full-length recombinant protein (ϳ70 kDa) as well as the RanGAP1:SUMO-1 conjugate (migrating at ϳ90 kDa) (Fig. 4,  lane 3; Ref. 16). The 90-kDa species was not observed when a K526R RanGAP1 mutant (Lys3 Arg mutant), which is incapable of conjugating to SUMO-1, was translated in vitro (Fig. 4,  lane 1; Ref. 16). The conversion of the unmodified form of RanGAP1 to the SUMO-1 conjugated from rabbit reticulocyte lysates was incomplete (Fig. 4, lane 3). We therefore added HeLa cell extracts to in vitro translated RanGAP1, which was previously shown to contain all the necessary components for the modification reaction (1). The addition of HeLa cell extracts to the in vitro translation system facilitated the conversion of RanGAP1 to the RanGAP1:SUMO-1 conjugate (Fig. 4, lane 4), an effect that was not observed with the Lys3 Arg mutant (Fig.  4, lane 2). To determine whether HsUbc9 can catalyze the conjugation of RanGAP1 with SUMO-1, HeLa cell extracts were immunodepleted using pre-immune control sera (Fig. 4,  lane 10) or antibodies reactive with HsUbc9 (Fig. 4, lanes 5-7) or SUMO-1 (Fig. 4, lanes 8 and 9). Immunodepletion of either HsUbc9 (Fig. 4, lane 5) or SUMO-1 (Fig. 4, lane 8) specifically prevented the ability of HeLa extracts to support the conjugation of RanGAP1 with SUMO-1. Importantly, recombinant HsUbc9 (Fig. 4, lane 6), but not HsUbcH5 (Fig. 4, lane 7) reconstituted the conjugation reaction in HsUbc9-depleted ex-tracts. Similarly, recombinant SUMO-1 (HSTV deletion mutant) reconstituted the conjugation reaction in SUMO-1-depleted extracts (Fig. 4, lane 9). These results strongly support the requirement for HsUbc9 in the covalent conjugation of RanGAP1 with SUMO-1. DISCUSSION Our results suggest that while much of the total HsUbc9 present in the cell is freely soluble in the nucleus and cytoplasm, a significant fraction of HsUbc9 is also associated with the nuclear envelope, most likely as a part of a complex that includes RanGAP1. More importantly, we have shown that HsUbc9, which was previously thought to be an E2-type ubiquitin-conjugating enzyme, is required for the covalent addition of the ubiquitin-related molecule SUMO-1 to RanGAP1. Because we were unable to demonstrate the formation of HsUbc9ubiquitin intermediates in vitro, it is likely that HsUbc9 is a dedicated SUMO-1-conjugating enzyme and is the first characterized enzyme in this novel pathway. We initially identified HsUbc9 as a protein capable of interacting with the RRM type RNA-binding proteins TIA-1 and TIAR (20) in a yeast twohybrid screen. In recent years, several other proteins capable of interacting with Ubc9 in the yeast two-hybrid system have also been reported (21)(22)(23). Proteins shown to interact with both Ubc9 and SUMO-1 include the cytoplasmic tail of Fas/APO-1, nuclear PML protein, and Rad52 (24 -26). At present, it is unclear if HsUbc9 mediates the SUMO-1 modification of these other proteins or what effect SUMO-1 modification of these other proteins might have. Unlike ubiquitin modification, SUMO-1 modification of RanGAP1 is clearly not a signal for degradation. Rather, SUMO-1 modification of RanGAP1 has been shown to be a prerequisite for the interaction of RanGAP1 with Nup358 (1). SUMO-1 modification may therefore be a general means of regulating protein-protein interaction, function, and/or localization. For example, overexpression of SUMO-1 was shown to protect cells from Fas ligand-induced cell death, possibly by preventing the association of accessory molecules with the "death domain" of the SUMO-1 modified cytoplasmic tail of Fas (24). The majority of SUMO-1 modified proteins are localized to the nucleus (27). The localization of HsUbc9 and SUMO-1 (1,2) to the nucleus suggests that the SUMO-1 modification machinery may preferentially target nuclear proteins. It is interesting to note that many of the proteins identified as HsUbc9-interacting proteins in yeast twohybrid screens are also nuclear proteins (20 -23, 25, 26); moreover, HsUbc9 has also been localized to the synaptonemal complex of meiotic chromatin (23). This compartmentalization may ensure that certain interactions occur only in the nucleus. The localization of a subset of HsUbc9 to the periphery of the nuclear envelope may be due to its interaction with RanGAP1, SUMO-1, or Nup358.
In yeast, Ubc9 is required for viability (10,11). Repression of UBC9 synthesis prevents cell cycle progression at the G 2 or early M phase, and in ubc9 mutants, both CLB5, an S phase cyclin, and CLB2, an M phase cyclin, are stabilized (10). These findings led to the suggestion that Ubc9 is directly involved in the ubiquitin-mediated degradation of mitotic cyclins (10,11). However, our data as well as the findings of others indicate that Ubc9 does not function in ubiquitination but rather is required for modification of proteins with SUMO-1 in mammalian cells and with the SUMO-1 homologue Smt3p in yeast (28). We propose that Ubc9 function may therefore be required for regulating other nuclear events, including Ran-mediated nucleocytoplasmic transport. For example, mutations that abrogate Ubc9 activity in vivo may prevent the Ran-mediated import of factors into the nucleus, which are necessary to initiate specific events associated with the G 2 /M progression and/or FIG. 4. Modification of RanGAP1 requires HsUbc9 activity. HeLa cell extracts were immunodepleted with antibodies against HsUbc9 or SUMO-1 prior to use in in vitro modification assay. 35 Slabeled RanGAP1 translated in vitro in rabbit reticulocyte lysate was incubated with HeLa cell extracts for 10 min at room temperature and resolved on a 6% SDS-polyacrylamide gel and transferred to nitrocellulose prior to autoradiography. mitosis. Because the B-type cyclins are degraded at the onset of anaphase (29), mutations which act downstream during the cell cycle may appear to result in the stabilization of these cyclins and in the delay of G 2 /M progression. Although yeast Ubc9p was recently shown to form thiolester intermediates with Smt3p (28), Rna1p, the yeast homologue of RanGAP1, is apparently not modified with Smt3p (30,31). It therefore remains to be seen whether Smt3 is indeed required for nuclear import in yeast. However, other putative Smt3p-modified nuclear proteins may also play a direct or indirect role in influencing cell cycle progression and cyclin degradation. The identification of HsUbc9 as a component of the SUMO-1 conjugation pathway is therefore an important step in elucidating not only the function of this novel pathway, but also the means by which the activities of the substrates of SUMO-1/ Smt3p modification may be regulated and the effect that SUMO-1/Smt3p modification may have on cellular function.