Perturbation of SUMOlation Enzyme Ubc9 by Distinct Domain within Nucleoporin RanBP2/Nup358*

Ubc9, a conjugation enzyme for the ubiquitin-related modifier SUMO, is present predominantly in the nucleus and at the nuclear pore complex. The functional significance of its subcellular compartmentalization, however, remains to be elucidated. Here, we define a Pro-Glu-Asp-Ser-Thr-rich element containing 129 amino acid residues, designated IR1+2, on the human nucleoporin RanBP2/Nup358, which binds directly to Ubc9 with high affinity both in vitro and in vivo. When IR1+2 tagged with green fluorescence protein at its amino terminus (GFP-IR1+2) was transfected into COS-7 cells, we found that ∼90% of the nuclear Ubc9 was sequestered in the cytoplasm. We also observed that both SUMO-1 and SUMO-2/3 were mislocalized, and promyelocytic leukemia protein PML formed an enlarged aggregate in the nucleus. Moreover, the homologous recombination protein Rad51 mislocalized to the cytoplasm, and Rad51 foci, a hallmark of functional association of Rad51 with damaged DNA, did not form efficiently even in the presence of a DNA strand breaker. These findings emphasize that the IR1+2 domain is a useful tool for manipulating the nuclear localization of Ubc9 and perturbing the subcellular localization of SUMOs and/or SUMOlated proteins, and they emphasize the important role of nuclear Ubc9 in the Rad51-mediated homologous recombination pathway, possibly by modulating intracellular trafficking of Rad51.

ferent species have been identified as targets for SUMO conjugation (SUMOlation), suggesting that the SUMO pathway plays important roles in many vital cellular processes (4).
The first component of the SUMO pathway identified in budding yeast and fission yeast was Ubc9/Hus5 (5,6). This protein primarily mediates SUMOlation but not ubiquitinylation (7,8). In budding yeast, blocking Ubc9 expression halts cell cycle progression at the G 2 /M phase and causes an aberrant mitotic morphology (5). In fission yeast, a mutant lacking Hus5 (a budding yeast homologue of Ubc9) grows slowly, undergoes abortive mitosis, and frequently loses mini-chromosomes, and a small percentage of these cells displays a "cut" morphology (6). It has been reported that Ubc9 enhances the nuclear localization of Ubc9-␤-galactosidase fusion protein in budding yeast (5). In both invertebrate and vertebrate systems, biochemical fractionation and indirect immunofluorescence analyses indicate that Ubc9 becomes enriched in the nucleus (4, 9 -13). Interestingly, in some cultured mammalian cells, Ubc9 can be detected at the nuclear pore complex (NPC) (4,11). However, it remains to be determined whether the subcellular localization of Ubc9 in the nucleus and/or the NPC is important for the function of Ubc9 in cell cycle control and maintenance of chromosome stability.
More than 30 different vertebrate proteins have been identified in yeast two-hybrid assays as potential candidates for SUMOlation by Ubc9 (4). Although there is little homology between these Ubc9-interacting proteins, many of them contain a PEST sequence, a stretch of at least 12 amino acids rich in proline, glutamate, aspartate, serine, and threonine residues (4,14,15). It should also be noted that many of these Ubc9interacting proteins have been implicated in the maintenance of vertebrate genome integrity, including the tumor suppressor protein p53 (16), the nuclear protein kinase HIPK2 (14), the promyelocytic leukemia protein (PML) (17), the homologous recombination proteins Rad51/52 (16,18), the poly(ADP-ribose) polymerase PARP (19), the components of the synaptonemal complex proteins Cor1 and Syn1 (20), the Werner's syndrome gene product Wrn (21) and the DNA topois II␤ (22). However, the biological significance of these interactions remains obscure.
In our previous studies, we documented that Ran-binding protein 2/358-kDa nucleoporin (RanBP2/Nup358), a component in the NPC, forms a tight complex with Ubc9 in Xenopus laevis egg extracts (23,24). In this report, we further characterize the interaction between RanBP2 and Ubc9 and provide evidence that Ubc9's distribution across the nuclear membrane, nuclear versus cytoplasmic, is important in controlling the subcellular localization of SUMOs and/or SUMOlated proteins, the assembly of PML bodies, and the homologous recombination process mediated by Rad51.
GST Pull-down Assay-Expression and purification of glutathione S-transferase (GST) fusion protein were carried out as described previously (24). X. laevis egg extracts were prepared as described previously (23). COS-7 cell lysates were prepared as described below. Standard binding conditions were as follows: 300 l of egg/cell extract was incubated with gentle rocking for 20 min at 4°C with 20 l of the glutathione-Sepharose beads (Amersham Biosciences, Inc.) pre-absorbed with appropriate GST fusion protein. The beads and proteins associated with them were precipitated by low speed centrifugation. After washing the beads four times with 700 l of ice-cold phosphatebuffered saline (PBS) containing 0.5% Triton X-100 and 0.05% Tween 20, 50 l of SDS sample buffer (50 mM Tris-HCl, pH 6.8, 100 mM dithiothreitol, 2% SDS, 0.1% bromphenol blue, 10% glycerol) was added and incubated for 5 min at 90°C. The eluted proteins were separated on a 4-20% gradient polyacrylamide gel electrophoresis (SDS-PAGE, Invitrogen, Carlsbad, CA) and visualized by silver or Coomassie Brilliant Blue staining. Immunoblot analyses were carried out as described previously (1).
Immunoprecipitation Analysis-Approximately 10 8 of COS-7 or BHK cells were harvested by gentle trypsin treatment and lysed with gentle homogenization in 2 ml of PBS containing 0.005% digitonin (Sigma-Aldrich, St. Louis, MO) and a protease inhibitor mixture (Sigma-Aldrich). After centrifugation, the soluble fraction was incubated with 2 g of antibody and 20 l of protein A-Sepharose beads (Amersham Biosciences, Inc.) at 4°C for 1 h. The collected beads were washed four times with PBS containing 0.1% Triton X-100, then proteins associated with the beads were analyzed on 4 -20% SDS-PAGE gels and visualized by silver staining.
Indirect Immunofluorescence Analysis-COS-7 or BHK cells were grown on coverslips and transfected with SuperFect transfection reagent (Qiagen, Valencia, CA). In Fig. 6 (see below), ϳ48 h after transfection, 5 g/ml mitomycin C, 5 g/ml etoposide, 5 g/ml aphidicolin, and 50 g/ml ␣-amanitin (Calbiochem, La Jolla, CA) were added to the culture medium and incubated for 2 h, respectively. Cells were fixed with 3% paraformaldehyde in PBS for 10 min at room temperature and methanol at Ϫ20°C for 2 min. For the detection of NPC-associated proteins, cells were permeabilized in 0.005% digitonin in PBS on ice for 5 min and washed twice in ice-cold PBS before fixation with paraformaldehyde and methanol. Conditions of the primary-and secondaryantibody incubations were described previously (1). Fluorescence images were analyzed with an Olympus system microscope Model BX60 (Olympus, Melville, NY) using the MetaMorph imaging system (Universal Imaging Corp., West Chester, PA). Quantitation of fluorescence described below (see Fig. 2E) was also performed using the MetaMorph system. Micrographs were arranged with Adobe Photoshop, version 3.0 or 5.0 (Adobe Systems, Mountainview, CA).

IR1ϩ2
Interacts with Ubc9 Both in Vitro and in Vivo-We reported previously that the nucleoporin RanBP2 interacts with Ubc9 via a 241-amino acid region (amino acids 2596 -2836) on RanBP2 (24). This domain was referred to as an IR domain by Yokoyama et al. (25) who originally identified two internal repeat units within this region (Figs. 1, A and B). To further dissect the IR domain, GST pull-down assays were carried out using five GST fusion protein constructs. These constructs contained internal repeat unit 1 (GST-IR1), internal repeat unit 2 (GST-IR2), the region between IR1 and IR2 (GST-M), the region containing both internal repeats (GST-IR1ϩ2) and the entire IR domain (GST-IR), respectively (Fig. 1A). GST protein was used as a negative control. The GST fusion proteins were incubated with Xenopus egg extracts or COS-7 cell lysates followed by pull down and immunoblot assays to detect association with Ubc9. We found that Ubc9 was abundant in both the GST-IR and GST-IR1ϩ2 pull-down fractions, but very little was in the GST-IR1, GST-IR2, and GST-M fractions (Fig.  1, A and C, and data not shown). Interestingly, we observed the SUMOlated form of RanGAP1 in the GST-IR fraction but not in the GST-IR1ϩ2 fraction (Fig. 1, A and C, and data not shown), suggesting that the binding domain for Ubc9 can be dissected from the binding domain for SUMOlated RanGAP1 on RanBP2. Taken together, these results indicate that IR1ϩ2 might contain the binding module specific for Ubc9.
To address whether the interaction between IR1ϩ2 and Ubc9 is direct, we performed GST pull-down assays using GST-IR1ϩ2 and purified recombinant Ubc9. As shown in Fig. 1D, Ubc9 co-precipitated with GST-IR1ϩ2 but not with GST, indicating that Ubc9 can bind directly to IR1ϩ2 without requiring additional cellular factors.
Next, we made a mammalian expression construct of IR1ϩ2 tagged at its N terminus with GFP (GFP-IR1ϩ2) to test whether IR1ϩ2 interacts with Ubc9 in vivo. COS-7 cells were transiently transfected with either the GFP-IR1ϩ2 or the GFP construct, and the cells were cultured for 48 h. Under our experimental conditions, equivalent amounts of GFP-IR1ϩ2 and GFP were expressed (Fig. 1E). After harvesting the cells, total cell lysates were subjected to immunoprecipitation analysis using an anti-GFP antibody (Fig. 1F). A GFP-IR1ϩ2 band (ϳ45 kDa) and a co-precipitated 18 kDa band, which was recognized by anti-Ubc9 antibody, were detected in the GFP-IR1ϩ2-transfected cell extracts, whereas Ubc9 was not detected in the GFP-transfected cell extracts. These results indicate that a significant amount of endogenous Ubc9 interacts very tightly with the exogenously expressed GFP-IR1ϩ2 via the IR1ϩ2 domain in vivo.
IR1ϩ2 Contains PEST Sequences-Yeast two-hybrid assays identified more than 30 different proteins that can interact with Ubc9. Many of these proteins contain PEST sequences (4,14). Therefore, we decided to look for PEST sequences in IR1ϩ2. Using the PEST-Find algorithm (see "Experimental Procedures"), we found a region containing a PEST element in internal repeat 1 of IR1ϩ2 (amino acids 2664 -2681) (Fig. 1B).
These results indicate that IR1ϩ2 might have biochemical properties similar to those of PEST-containing Ubc9-interacting proteins.
Some E2 ubiquitin carrier proteins interact with E3 ubiquitin-protein isopeptide ligases via a RING (really interesting new gene) finger motif (28) or a HECT (homologous to E6-AP C terminus) domain (29). In the case of Ubr1, which is a unique E3 ubiquitin ligase in the amino-terminal end rule pathway, a BRR (basic residue-rich) region is involved in the interaction with an E2-ubiquitin carrier protein (30). Thus, we were interested to learn whether IR1ϩ2 contains such motifs/domains. Neither RING-finger, HECT domain, nor BRR regions were detected within the IR1ϩ2 domain, suggesting that the mechanism by which RanBP2 interacts with Ubc9 might be distinct from previously characterized E2⅐E3 interactions in the ubiquitin and amino-terminal end rule pathways.
GFP-IR1ϩ2 Sequesters the Nuclear Pool of Ubc9 in the Cytoplasm-The immunoprecipitation experiment shown in Fig.  1F suggests that a significant amount of the endogenous Ubc9 forms a stable complex with exogenously expressed GFP-IR1ϩ2 in vivo. To investigate the subcellular localization of the GFP-IR1ϩ2⅐Ubc9 complex, we performed indirect immunofluorescence analyses. As shown in Fig. 2D, both GFP-IR1ϩ2 and Ubc9 were detected abundantly in the cytoplasm but were barely detected in the nucleus, suggesting that the complex is primarily cytoplasmic. Because immunofluorescence microscopy on the non-transfected COS-7 cells detected the endogenous Ubc9 predominantly in the nucleus ( Fig. 2A, panel a), the formation of the GFP-IR1ϩ2⅐Ubc9 complex in the cytoplasm might result in sequestration of the nuclear Ubc9 in the cytoplasm. The nuclear Ubc9 content was quantified by measuring the level of fluorescence produced by the anti-Ubc9 antibody. The pixels within the nucleus (Fig. 2D, panel b) were counted, and the integrated signal intensities of the anti-Ubc9 antibody were measured. The measurements revealed that ϳ90% of the nuclear Ubc9 was sequestered in the cytoplasm in the GFP-IR1ϩ2-expressing cells (Fig. 2E).
GFP-IR1ϩ2 Expression Has Little Effect on the NPC-associated Ubc9 -It had been reported that Ubc9 is concentrated not only inside of the nucleus but also at the NPC in vertebrate cells (4, 10 -13). In support of these observations, our immunofluorescence analysis of non-transfected COS-7 cells revealed that Ubc9 exists predominantly in the nucleus and also at the nuclear rim/NPC ( Fig. 2A, panel a). As shown in Fig. 2A (panel  b), the NPC staining was most apparent when the cells were permeabilized with 0.005% digitonin before fixation and labeling. Because digitonin preferentially solubilizes the cholester-The purified recombinant Ubc9 was incubated with the beads containing GST or GST-IR1ϩ2. The proteins associated with beads were analyzed by 4 -20% SDS-PAGE and visualized by Coomassie Blue staining.

FIG. 1. Ubc9 interacts with the IR1؉2 domain on RanBP2 in vitro and in vivo.
A, schematic representation of RanBP2 and summary of GST pull-down assay. RanBP2 is a large nucleoporin with a leucine-rich region, four RanBP1-homologous domains (RBD1-4), eight zinc-finger motifs, a cyclophilin homologous domain (CH), and an internal repeat domain (IR). Numbers indicated above the GST-IR construct correspond to amino acid residues in RanBP2. The table on the right summarizes the interactions of each GST fusion protein with Ubc9, SUMO-1, SUMO-2/3, and SUMO-1 modified RanGAP1 in the GST pull-down assays using Xenopus egg extracts (see "Experimental Procedures" for details). B, amino acid sequence of the IR1ϩ2 domain. The internal repeat units 1 and 2 are underlined. Dashes within the sequence indicate spaces inserted to make an optimal match. Identical, conserved, and similar amino acid residues between the internal repeats 1 and 2 are indicated by vertical lines, double dots, and single dots, respectively. An amino acid sequence similar to a PEST sequence detected in IR1 is indicated by ϩ. C, Ubc9 is a major IR1ϩ2-interacting protein in COS-7 cell lysate. GST pull-down assays were carried out using COS-7 cell lysates (see "Experimental Procedures" for details). Proteins associated with the GST fusion proteins were separated by 4 -20% SDS-PAGEs. One gel was silver-stained (left panel), and the other was subjected to an immunoblot analysis using anti- ol-rich plasma membrane but does not extract the nuclear membrane, many soluble cytoplasmic and nucleoplasmic proteins, including the nucleoplasmic Ubc9, are extracted efficiently with the digitonin treatment, whereas the proteins associated with insoluble materials such as the nuclear membrane/NPC remains intact. Furthermore, the nuclear and NPC localization of Ubc9 was also demonstrated by the transient expression of Ubc9 tagged at its C terminus with GFP (Ubc9-GFP) in COS-7 cells. As shown in Fig. 2B, when cells were transfected with the Ubc9-GFP construct, the green fluorescence was detected both in the nucleus and at the NPC. Taken together, these data support the previous observations that Ubc9 localizes both at the NPC and within the nucleus.
Interestingly, we found that, in contrast to the marked mislocalization of the nuclear pool of Ubc9 (Fig. 2, D (panels a and  b) and E), the localization of NPC-associated Ubc9 was not dramatically altered in the GFP-IR1ϩ2-expressing cells (Fig. 2,  D (panels c and d) and E). This suggests that GFP-IR1ϩ2 sequesters the nuclear pool of Ubc9 more efficiently than the NPC-associated pool of Ubc9.
Implications for Nuclear Transport in the GFP-IR1ϩ2-ex-pressing Cells-Based on the high affinity interaction of Ubc9 with IR1ϩ2, it appears that the formation of the GFP-IR1ϩ2⅐Ubc9 complex induces the mislocalization of the nuclear Ubc9 to the cytoplasm (Figs. 1 and 2). However, one could argue that the mislocalization of Ubc9 is an indirect consequence of the inhibition of protein/RNA transport, which could be induced somehow by the cytoplasmic overexpression of GFP-IR1ϩ2. To assess this possibility, the subcellular localization of several proteins involved in the regulation of nuclear trafficking and RNA processing were examined in the GFP-IR1ϩ2expressing cells. The proteins that we investigated included: Ran (31), a nuclear small GTP-binding protein controlling directionality of nuclear trafficking (Fig. 3A); Karyopherin/Importin ␤ (31), the best-characterized member of the nuclear transport receptor family (B); RCC1 (31), a guanine nucleotide exchange factor for Ran that binds to chromatin (C); FXFGcontaining nucleoporins (31), components of the NPC (D); hnRNP-C1/2 (32), one of the most abundant eukaryotic pre-mRNA-binding proteins (E); and SC-35 (33), an RNA splicing factor predominantly localized in the nuclear speckle (F). RanGAP1 (34,35), an essential nuclear transport regulator In c and d, the cells were treated with 0.005% digitonin before fixation. The cell expressing the GFP-IR1ϩ2 fusion protein was detected in the green fluorescence channel (a and c). Because GFP-IR1ϩ2 can be extracted efficiently by the digitonin, the residual green fluorescence is barely seen after the digitonin treatment (c). In the upper panels (a and b), there are three cells in the field; one is transfected and expressing GFP-IR1ϩ2 (bottom right), and the others are non-transfected cells. In the lower panels (c and d), there are two cells; one is transfected and expressing GFP-IR1ϩ2 (bottom right), and the other is a non-transfected cell (top left). E, the nuclear Ubc9 is sequestered in the cytoplasm in GFP-IR1ϩ2-expressing cells. The nuclear and the NPC-associated Ubc9 contents were quantified by measuring the level of fluorescence produced by an anti-Ubc9 antibody. The pixels within the nucleus (left side) and around the NPC (right side) were counted, and the integrated signal intensities produced by the anti-Ubc9 antibody were measured. f.u., fluorescence units. and a major SUMOlation substrate, was also visualized (Fig.  5A). We found that the subcellular localizations of these factors were not altered dramatically in the GFP-IR1ϩ2-expressing cells, suggesting that the general machinery for nuclear transport is not perturbed in the GFP-IR1ϩ2-expressing cells. In addition, it should be noted that no remarkable difference in nuclear morphology was observed between the GFP-IR1ϩ2expressing cells, the GFP-expressing cells, and the non-transfected cells during the cultured period (up to 3 days) ( Fig. 2C and data not shown). These results also indicate that GFP-IR1ϩ2 expression does not perturb the nuclear transport system in general.
SUMOs and/or SUMOlated Proteins Mislocalize in the GFP-IR1ϩ2-expressing Cells-Given that Ubc9 is a SUMOlation enzyme, we expected that the mislocalization of Ubc9 would have an effect on the subcellular localization of SUMOs and/or SUMOlated proteins. To investigate this, we performed indirect immunofluorescence analyses using anti-SUMO-1 and anti-SUMO-2/3 antibodies. First, we found that both anti-SUMO-1 and anti-SUMO-2/3 antibodies showed decreased fluorescence in the nucleus and significantly increased fluorescence in the cytoplasm of GFP-IR1ϩ2-transfected COS-7 cells compared with non-transfected cells (Fig. 4A). These observations indicate that SUMOs and/or SUMOlated proteins mislocalize in the GFP-IR1ϩ2-expressing cells. Second, we observed that the number of nuclear dots detected by anti-SUMO-1 and anti-SUMO-2/3 antibodies was significantly lower in the GFP-IR1ϩ2-expressing cells (Fig. 4A and Table I). In the nuclei of the non-transfected cells, multiple (up to 20) nuclear dots were usually observed with the anti-SUMO-1 (Fig. 4A (panel b) and Table I) and by the anti-SUMO-2/3 antibodies (Fig. 4A, panel  d). In contrast, neither the anti-SUMO-1 antibody nor the anti-SUMO-2/3 antibody detected multiple foci in the nuclei of GFP-IR1ϩ2-expressing cells, suggesting that the assembly of discrete foci enriched with SUMOs is perturbed in the nucleus by the overexpression of GFP-IR1ϩ2.

Inappropriate Assembly of PML in the GFP-IR-1ϩ2-express-
ing Cells-Nuclear foci detected by anti-SUMO-1 antibody have been colocalized with the nuclear scaffold-associated large mutiprotein complex, known as the PML bodies, PML oncogenic domains or ND10 (36 -42). To investigate the assembly of PML, indirect immunofluorescence analysis was performed in the GFP-IR1ϩ2-expressing cells using an anti-PML mouse monoclonal antibody, 5E10 (27). As shown in Fig. 4 (B and C) and Table II, we found that most of these cells contained an enlarged PML dot compared with normal PML bodies detected in the non-transfected cells. For example, ϳ60% of the GFP-IR1ϩ2-expressing cells contained a single enlarged focus (Fig.  4C (panels a and b) and Table II) and about 30% had two enlarged foci (Fig. 4C (panels c and d) and Table II). The remaining cells usually contained a single, enlarged focus plus several (two or three) small foci (Table II and data not shown). In contrast, anti-PML antibody detected three to ten dots in most (ϳ70%) of the non-transfected cells (Fig. 4B and Table II). These results indicate that the overexpression of GFP-IR1ϩ2 alters the size and number of PML bodies.
Given that SUMOlation of PML is required for proper assembly of the PML bodies (32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42), we suspected that the decreased level of Ubc9 in the nucleus of the GFP-IR1ϩ2expressing cells might inhibit the SUMOlation on PML and, therefore, perturb the formation/maintenance of PML bodies. To see whether the SUMOlation on PML was affected, an HA-tagged PML construct was co-transfected with either the GFP-IR1ϩ2 or GFP construct, and the transfected cells were cultured for 48 h. After harvesting the cells, total lysates were subjected to immunoprecipitation analysis using an anti-HA antibody followed by immunoblot analysis using an anti-SUMO-1 antibody. We found that the level of SUMOlated HA-PML in the GFP-IR1ϩ2-expressing cells was not reduced significantly compared with the level in the GFP-transfected cells (data not shown), suggesting that the GFP-IR1ϩ2 expression has little effect on SUMOlation of the exogenously overexpressed PML. Because exogenously expressed HA-PML may not represent a pool of endogenous PML, we next investigated endogenous PML and its modified form using anti-PML antibody. However, our immunoblot analysis was not sensitive enough to detect either endogenous PML or the modified form of PML (data not shown). Therefore, it is currently unknown whether the decreased level of Ubc9 in the nucleus of the GFP-IR1ϩ2-expressing cells inhibits SUMOlation of endogenous PML.

Rad51
Mislocalizes in the GFP-IR1ϩ2-expressed Cells-To investigate the effects of GFP-IR1ϩ2 expression on the Ubc9/ SUMO pathway, we performed indirect immunofluorescence analyses with antibodies either specific for proteins associated with Ubc9 or known to be SUMOlated. As with SUMOs and PML, we expected to see the mislocalization of these factors in the GFP-IR1ϩ2-expressing cells. The proteins that we investigated were as follows: RanGAP1, p53, Rad51, DNA topoisomerase I, DNA topoisomerase II, and poly-(ADP-ribose)polymerase PARP. Rad51 (16,18,43) and PARP (19) have been shown to interact with Ubc9 in a yeast two hybrid system. RanGAP1 (34,35) and DNA topoisomerase I (44) are SUMOlated in vivo. p53 (16,(45)(46)(47) and DNA topoisomerase II (22) interact with Ubc9 and can be SUMOlated. When the immunofluorescence patterns of these proteins were observed, we found that Rad51 was clearly different between the GFP-IR1ϩ2-expressing and the non-transfected cells (Fig. 5C); i.e. Rad51 presents predominantly in the nucleus in the non-transfected cells, but diffuses throughout the cell with some bright foci in the cytoplasm in the GFP-IR1ϩ2-expressing cells. The anti-RanGAP1, -p53, -DNA topoisomerase I, -DNA topoisomerase II, and -PARP antibodies, however, did not show significant differences between transfected and non-transfected cells (Fig. 5, A, B, D-F). These data indicate that the subcellular localization of Rad51 is very sensitive to the expression of GFP-IR1ϩ2.
We then performed indirect immunofluorescence with 35 different antibodies to identify other proteins that might mislocalize in the GFP-IR1ϩ2-expressing cells. These included antibodies to DNA recombination/repair proteins and cell cycle regulators (the list of the antibodies is under "Experimental Procedures"). Unlike SUMO-1, SUMO-2/3, PML, or Rad51, we did not find altered immunofluorescence patterns between the non-transfected and the GFP-IR1ϩ2-expressing cells (data not shown). However, if a small percentage of these proteins mislocalized, we might not be able to detect them with our immunofluorescence technique.
Rad51 Foci Are Poorly Induced by DNA Strand Breakers in the GFP-IR1ϩ2-expressing Cells-It has been reported that DNA damage from ␥ or ultraviolet irradiation, methanesulfonate, or DNA-strand breakers induces a very rapid and dramatic increase in the percentage of cells with multiple discrete foci containing Rad51, known as Rad51 foci, in the nucleus (48). To test whether Rad51 foci were induced in the GFP-IR1ϩ2-expressing cells following DNA damage, we treated the cells with either mitomycin C (MMC) or etoposide prior to indirect immunofluorescence. MMC cross-links DNA at guanine and adenine residues and disrupts base pairing (49). Etoposide interferes with DNA topoisomerase II-mediated breakage and reunion cycle by trapping the enzyme in the cleaved complex, which results in irreparable DNA double-strand breaks (50). Following 2-h exposure to MMC or etoposide, the percentage of cells with Rad51 foci in both GFP-expressing and non-transfected cells had increased (Fig. 6, A and B). However, in the GFP-IR1ϩ2-expressing cells, we did not observe a significant increase in the percentage of the cells containing Rad51 foci. Moreover, the staining was predominantly cytoplasmic (Fig.  6A), suggesting that the targeting of Rad51 to region of DNA  damage is restricted, because the nuclear translocation/retention of Rad51 is perturbed. To demonstrate the specificity of our assay, the cells were exposed to the replication inhibitor aphidicolin (AP) and the transcription inhibitor ␣-amanitin. These two chemical agents did not cause a significant increase in the percentage of cells with Rad51 foci in either the GFP-IR1ϩ2-expressing cells or the control cells (Fig. 6B). Taken together, these results indicate that the induction of Rad51 foci by DNA-strand breakers is markedly inhibited in the GFP-IR1ϩ2-expressing cells probably due to changes in the intracellular trafficking of Rad51.

DISCUSSION
RanBP2/Nup358 is a 358-kDa nucleoporin that interacts with the 18-kDa SUMOlation enzyme Ubc9 through a 129amino acid domain that we have designated IR1ϩ2. The domain consists of three distinct parts; 53 amino acids of internal repeat 1 (IR1), 51 amino acids of internal repeat 2 (IR2), and 25 amino acids between the two repeats (Fig. 1B). We found a strong homology to PEST sequences ( 2664 PDYVSEEEEDDED-FETAV 2681 ) in IR1 but not in IR2. Although the IR1-PEST sequence appears to contribute to the interaction of IR1ϩ2 with Ubc9, IR1 itself binds poorly to Ubc9 compared with IR1ϩ2 (Fig. 1A). Thus, the IR1-PEST sequence is not sufficient to confer full binding activity with Ubc9. It is known that PEST sequences are predicted to regulate the instability of proteins in general (15). However, it is unlikely that the PEST element in IR1 confers instability upon IR1ϩ2, because GFP-IR1ϩ2 appears to be expressed stably in COS-7 and BHK cells ( Fig. 1 and data not shown). The PEST sequence in IR1ϩ2, therefore, might simply be similar to a PEST element but is really a component of a larger binding region that interacts with Ubc9. We are currently dissecting the IR1ϩ2 region of RanBP2 to define the minimum and essential polypeptide sequence required for binding to Ubc9.
In COS-7 cells endogenous Ubc9 localizes not only in the nucleus but also at the NPC ( Fig. 2A). In addition, we demonstrate that exogenously expressed Ubc9-GFP is targeted to both the nucleus and the NPC (Fig. 2B). Our discovery of the high affinity interaction between the IR1ϩ2 domain of RanBP2 and Ubc9 might explain at least how Ubc9 is targeted to the NPC. However, we believe that there must be factors in addition to RanBP2 that enhance the association of Ubc9 with the NPC; in contrast to the efficient sequestration of Ubc9 from the nucleus, NPC-associated Ubc9 is not depleted by the cytoplasmic overexpression of GFP-IR1ϩ2 (Fig. 2, D and E). This observation indicates either the presence of another NPC component(s) with a higher affinity for Ubc9, or implies additional factor(s) that increases the affinity of IR1ϩ2 for Ubc9 at the NPC. We are currently searching for proteins that could do this. One candidate may be SUMOlated RanGAP1, which interacts with RanBP2 and may increase the affinity of Ubc9 for RanBP2 at the NPC (51).
Co-immunoprecipitation of transiently expressed GFP-IR1ϩ2 with Ubc9 provides a compelling evidence for their tight  5. Rad51 mislocalizes in the GFP-IR1؉2-expressing cells. COS-7 cells were transfected with the GFP-IR1ϩ2 construct and grown on coverslips for 48 h. The cells on the coverslips were fixed and subjected to immunofluorescence analyses using the primary antibodies indicated followed by rhodamineconjugated secondary antibodies. Cells expressing the GFP-IR1ϩ2 fusion protein were detected under the green fluorescence channel (left panels). Antibody staining was detected in the rhodamine channel (right panels). Anti-RanGAP1 (A), anti-p53 (B), anti-Rad51 (C), anti-DNA topoisomerase I (D), anti-DNA topoisomerase II (E), and anti-PARP (F) antibodies were used as primary antibodies. and specific association in vivo (Fig. 1F). The interaction appears to induce relocalization of the nuclear Ubc9 to the cytoplasm because both GFP-IR1ϩ2 and Ubc9 are detected in the cytoplasm (Fig. 2D). Such alteration of Ubc9 concentration across the nuclear membrane appears to deleteriously affect the SUMO pathway, because we see the marked mislocalization of SUMOs and/or SUMOlated proteins in the GFP-IR1ϩ2expressing cells (Fig. 4A). Although the experiments documented here do not specifically address how sequestration of Ubc9 in the cytoplasm affects subcellular localizations of SUMOs and/or SUMOlated proteins, we propose three possible mechanisms. First, the nuclear depletion of Ubc9 resulting from the GFP-IR1ϩ2 expression may disrupt interactions between Ubc9 and nuclear SUMOlation substrates, causing SUMOlation in the nucleus to be down-regulated and producing a pool of unconjugated SUMOs free to diffuse in and out of the nucleus. Loss of nuclear dots detected by anti-SUMO-1 and -SUMO-2/3 antibodies and the appearance of enlarged nuclear dots detected by anti-PML antibody in the GFP-IR1ϩ2-expressing cells may support this possibility (Fig. 4, A-C). Second, the GFP-IR1ϩ2⅐Ubc9 complex in the cytoplasm may upregulate SUMOlation in the cytoplasm, including the cytoplasmic SUMOlation of proteins such as RanGAP1. However, neither altered localization or augmentation of SUMOlation of transiently transfected RanGAP1 has been detected so far in GFP-IR1ϩ2-expressing cells (Fig. 5A and data not shown), indicating that this second scenario is unlikely at least in terms of RanGAP1's SUMOlation. Third, Ubc9 may interact with conjugation-free SUMOs and/or SUMOlated proteins via a non-covalent interaction that is independent of SUMOlation, and the interaction may play a critical role in the determining subcellular localization of SUMOs/SUMOlated proteins. It is possible that nuclear translocation of a subset of SUMOs/ SUMOlated proteins may require such non-covalent interaction with Ubc9. Currently, it has been reported that Importin 13 binds Ubc9 efficiently and in a Ran-sensitive manner, suggesting that the nuclear translocation of Ubc9 through the NPC is mediated by Importin 13 (52). In future experiments, it will be interesting to see whether Importin 13 regulates nuclear translocation of not only Ubc9 but also SUMOs/SUMOlated proteins that associate non-covalently with Ubc9.
We were surprised to find that the mislocalization of Ubc9 coincides with the mislocalization of Rad51 in the GFP-IR1ϩ2expressing cells regardless of the presence of DNA damage (Figs. 5 and 6). Although it has been reported that Rad51 interacts with Ubc9 in the yeast two-hybrid system (16,18), we have thus far not detected a SUMOlated Rad51 in either nontransfected or GFP-IR1ϩ2-expressing cells (16,18,43, and data not shown), indicating that Rad51 is not a SUMOlation substrate. How Rad51 is mislocalized is not clear, but one might note that Li et al. (43) have recently shown a complex of Rad51 with Rad52 and SUMO-1, in which SUMO-1 is not covalently conjugated to either Rad51 or Rad52. It is possible that Ubc9 is involved in the proper targeting of the Rad51⅐Rad52⅐SUMO (or SUMOlated factor X) complex to the nucleus through a non-covalent interaction that is independent of SUMOlation. Intriguingly, it has been reported that perturbation of the Ubc9/SUMO pathway in Drosophila and zebrafish causes the mislocalization of the nuclear factors: Dorsal, a member of the Rel family of transcription factors (53), Bicoid, the anteroposterior patterning morphogen (54), and Vsx-1, a paired-like CVC homeobox gene product (55). These reports are FIG. 6. Rad51 foci are not efficiently formed in the GFP-IR1؉2-expressing cells even in the presence of the DNA strand breakers, mitomycin C and etoposide. A, poor induction of Rad51 foci in the GFP-IR1ϩ2-expressing cells. COS-7 cells were transfected with the GFP-IR1ϩ2 construct and grown on coverslips for 48 h. After addition of mitomycin C (MMC; a, b, c, and d) or etoposide (e, f, g, and h) to the culture media, cells were incubated for 2 h. Cells on the coverslips were fixed and subjected to an immunofluorescence analysis using anti-Rad51 primary and rhodamine-conjugated secondary antibodies. Cell expressing the GFP-IR1ϩ2 fusion protein were detected in the green fluorescence channel (a, c, e, and g). The antibody staining was detected in the rhodamine channel (b, d, f, and h). B, reduced percentage of nuclei containing Rad51 foci in the GFP-IR1ϩ2transfected cells. The non-transfected, GFP-transfected, or GFP-IR1ϩ2 transfected cells were treated with aphidicolin (AP), ␣-amanitin (AM), mitomycin C (MMC), or etoposide (ETO). After staining with anti-Rad51 antibody, at least 120 nuclei were analyzed for each cell culture to calculate the percentage of nuclei containing Rad51 foci. Nuclei containing at least 20 foci were considered to be positive for Rad51 foci. reminiscent of our result and may indicate the function of Ubc9 that is distinct from its enzymatic activity.
In conclusion, this study demonstrates that a high affinity association between the IR1ϩ2 domain of RanBP2 and Ubc9 can be used to manipulate Ubc9 subcellular localization as well as subcellular localization of SUMOs and/or SUMOlated proteins. Furthermore, our findings in GFP-IR1ϩ2-expressing cells, that Rad51 is mislocalized and that there are no Rad51 foci, may argue that the previously suspected connection between the Ubc9/SUMO pathway and the Rad51-mediated homologous recombination process may be a result of the former pathway's ability to modulate intracellular trafficking of Rad51.