A Lack of SUMO Conjugation Affects cNLS-dependent Nuclear Protein Import in Yeast*

Yeast SUMO (Smt3) and its mammalian ortholog SUMO-1 are ubiquitin-like proteins that can reversibly be conjugated to other proteins. Among the substrates for SUMO modification in vertebrates are RanGAP1 and RanBP2/Nup358, two proteins previously implicated in nucleocytoplasmic transport. Sumoylated RanGAP1 binds to the nuclear pore complex via RanBP2/Nup358, a giant nucleoporin, which was recently reported to act as a SUMO E3 ligase on some nuclear substrates. However, no direct evidence for a role of the SUMO system in nuclear transport has been obtained so far. By the use of conditional yeast mutants, we examined nuclear protein importin vivo. We show here that cNLS-dependent protein import is impaired in mutants with defective Ulp1 and Uba2, two enzymes involved in the SUMO conjugation reaction. In contrast, other transport pathways such as rgNLS-mediated protein import and mRNA export are not affected. Furthermore, we find that the yeast importin-α subunit Srp1 accumulates in the nucleus ofulp1 and uba2 strains but not the importin-β subunit Kap95, indicating that a lack of Srp1 export might impair cNLS import. In summary, our results provide evidence that SUMO modification in yeast, as has been suspected for vertebrates, plays an important role in nucleocytoplasmic trafficking.

The nuclear pore complex (NPC) 1 can be considered as the gateway of bidirectional traffic between the nucleus and cytoplasm. Specific import and export pathways converge at the NPC and are defined by distinct interactions between individual proteins (nucleoporins or nups) and shuttling transport factors, the karyopherins (1). In addition, the small GTPase Ran and its effector proteins RanGAP1 (Ran GTPase-activating protein) and RanGEF (Ran guanine nucleotide exchange factor) are needed to confer directionality to transport processes (2). In higher eukaryotes, RanGAP1 is found in the cytoplasm and at the cytoplasmic face of the NPC where it forms a stable complex with the nucleoporin RanBP2/Nup358 (3,4). Binding of RanGAP1 to RanBP2/Nup358 depends on modification of RanGAP1 with SUMO-1, a small ubiquitin-like protein modifier (5). It was speculated that SUMO-dependent RanGAP1 localization at the NPC provides a means to temporally regulate nucleocytoplasmic trafficking (6). Apart from nuclear transport, sumoylation of proteins has been implicated in a number of essential cellular functions such as signal transduction, chromosome stability, and cell cycle regulation (7,8). It is an evolutionary conserved process, which is similar to protein ubiquitinylation but nevertheless functionally distinct (9). Unlike in the case of ubiquitin, SUMO modification does not lead to substrate degradation but rather has diverse effects on protein activity and intracellular localization (10). Three distinct steps can be distinguished in the SUMO modification pathway: 1) activation of SUMO, 2) transfer of SUMO to the conjugating enzyme, and 3) substrate modification. Since SUMO is synthesized as a precursor protein, a maturation step precedes the activation reaction. In yeast, C-terminal processing of the SUMO precursor is mediated by the processing protease Ulp1, which has an additional role in the deconjugation of SUMO-modified substrates (11). Mature SUMO is activated by SUMO-activating enzyme, an E1-like heterodimeric protein complex composed of Uba2 and Aos1 (12,13). Conditional uba2 mutants are deficient in SUMO conjugation and accumulate free SUMO (14,15). Following activation, the SUMO moiety is transferred to Ubc9, the E2-like-conjugating enzyme (13,16). Ubc9 then likely interacts with the recently identified E3-like proteins Siz1 and Siz2, which presumably mediate recognition and binding of substrate proteins (17,18). Since deletion of SIZ1 and SIZ2 is not lethal in yeast, other E3-like proteins specific for SUMO might exist (17). Post-translational SUMO modification is a reversible process, and it was found that the sumoylation status of proteins varies during the cell cycle (11,19). SUMO-deconjugating enzymes have been identified in yeast and mammals (6). The yeast Ulp1 and Ulp2 proteins are cysteine proteases that cleave SUMO from their substrates. Despite the fact that both enzymes are similar in their biochemical activity, they clearly differ in substrate specificity and function (11,15,20).
Information about the involvement of the SUMO modification system in nuclear trafficking is limited. So far, RanGAP1 is the only component of the transport apparatus that has been found to be SUMO conjugated in vivo (3). RanBP2/Nup358 can be sumoylated in vitro (21), and recently, it was shown that RanBP2/Nup358 acts as an E3 ligase on some nuclear substrates (22). Interestingly, mutations in the Drosophila UBC9 ortholog Semushi specifically block nuclear import of the transcriptional activator Bicoid (23). In budding yeast, the Ran-GAP1 ortholog Rna1 is not sumoylated (6) and other evidence for the involvement of the SUMO modification system in nuclear transport processes has not been reported. We have therefore started to investigate a putative role of the SUMO modi-fication system in nucleocytoplasmic trafficking and report here that in yeast SUMO conjugation is important for a specific nuclear protein import pathway.
In Vivo Nuclear Import Assay-The assay was performed as described (29). Essentially, cells were grown to mid-log phase in SD medium without uracil (SD/-Ura) and photographic images were taken using identical exposure times. To allow for redistribution of the reporter protein between nucleus and cytoplasm, cells were centrifuged, washed with water, and incubated for 45 min at 25°C with medium containing 10 mM sodium azide and 10 mM deoxyglucose. The GFP signal was monitored during this time and after pictures were taken, cells were washed in ice-cold medium without glucose and kept on ice. Before starting the re-import reaction, cells were incubated for 10 min at the non-permissive temperature of 37°C to induce the temperaturesensitive (ts) phenotype. Subsequently, glucose was added to a final concentration of 2% to start re-import. Photographs were taken after an additional 15 min at 37°C.
Localization of Proteins and mRNA-In vivo localization of GFPtagged proteins was done by examining cells directly with a Zeiss Axioplan 2 microscope using filter unit 13. Photographs were taken with an Axiocam camera using the Axiovision 3.0 software. Exposure times were identical for one set of experiments and images were assembled in Photoshop 5.0 (Adobe). Quantitation of Nup2-GFP fluorescence was done as described (26) using the NIH Image software. Indirect immunofluorescence analysis was as described (31) using a rabbit polyclonal anti-Srp1 antibody (1:1000, D. Görlich) and a goat anti-rabbit antibody coupled to rhodamine (1:1000, Dianova). Immunogold EM was performed as reported (31) using the anti-Srp1 antibody and a goat anti-rabbit antibody coupled to 12-nm gold (1:30, Dianova). In situ hybridization was done exactly as published (32) using a mouse monoclonal anti-digoxigenin antibody (1:75, Roche Molecular Biochemicals) and a FITC-conjugated goat anti-mouse antibody (1:500, Roche Molecular Biochemicals). For indirect immunofluorescence and in situ hybridization analyses filter units were Zeiss filter unit 01 (DAPI), 10 (FITC), and 15 (rhodamine).

RESULTS
Nuclear Import of a cNLS-GFP Reporter Protein Is Impaired in Mutants of the SUMO Conjugation Pathway-In higher eukaryotes, the SUMO conjugation system has been implicated in a variety of essential cellular activities including nucleocytoplasmic transport. In yeast, evidence for a role in nuclear transport has not been reported. To test the yeast SUMO modification pathway for defects in protein import, we employed ts mutants such as ulp1-333, uba2-1, and ulp2-1 (herein referred to as ulp1, uba2, and ulp2⌬, respectively). Since many mutations in nuclear transport factors do not lead to strong steady state cytoplasmic accumulation of otherwise nuclear proteins, we used an in vivo nuclear protein import assay, which is especially suited to investigate conditional transport mutants (33). This method was originally developed to determine nuclear import rates for proteins carrying a classical nuclear localization signal (cNLS) (29). As a reporter, GFP is fused to a cNLS. GFP alone is small enough to diffuse freely between nucleus and cytoplasm. However, due to the strong cNLS signal, the reporter protein concentrates in the nucleus of wild type cells at the permissive temperature (Fig. 1A, top row, start at 25°C). After poisoning the cells with sodium azide and deoxyglucose, cNLS-GFP equilibrates between nucleus and cytoplasm since active import is prevented but diffusion out of the nucleus is still possible (Fig. 1A, top row, azide at 25°C). After washing out the poison and addition of glucose, wild type cells immediately re-import the GFP reporter into the nucleus after a shift to higher temperature (Fig. 1A, top row, re-import at , and uba2 (JD90-1A) strains in the in vivo import assay. Cells were transformed with the cNLS-GFP reporter plasmid, grown in liquid SD/-Ura medium to 0.4 OD 600 at 25°C and photographed (start at 25°C). Cells were pelleted, washed into medium without glucose, and incubated for 45 min at 25°C in medium with deoxyglucose and azide to allow for equilibration of the cNLS-GFP reporter protein (azide at 25°C). Subsequently, cells were pelleted, washed into cold medium without glucose, and kept on ice until used for the re-import reaction. Re-import was initiated after a 10-min preshift to 37°C by adding glucose to the medium. Photographs were taken after 25 min at 37°C (re-import at 37°C). Exposure times were identical for the different experiments. B, cNLS-GFP import in the presence of mature SUMO. Wild type and ulp1 (MHY1488) cells were grown in SD/-Ura/-Trp medium at 25°C, and expression of mature SUMO from a plasmid borne CUP1 promoter (pIS50-T) was induced by adding copper sulfate to the growth medium for 5 h. The in vivo import assay was performed as in A.
37°C). In a temperature sensitive ulp1 strain, the GFP reporter protein partially localizes to the cytoplasm at the permissive temperature (Fig. 1A, 2nd row, start at 25°C). This localization pattern is even more pronounced after a 25 min shift to the non-permissive temperature (Fig. 1A, 2nd row, re-import at 37°C) indicating that mutations in ULP1 lead to cytoplasmic accumulation of proteins carrying a cNLS signal. In contrast, a ulp2⌬ mutant shows a wild type pattern of reporter protein localization (Fig. 1A, 3rd row, re-import at 37°C) indicating that in this strain nuclear import of the GFP reporter is not perturbed under the experimental conditions used. CNLS dependent import was also perturbed in a uba2 ts mutant: here, incomplete nuclear import of the reporter protein is visible at the re-import step after a 25-min incubation at the non-permissive temperature (Fig. 1A, bottom row, re-import at 37°C).
The Ulp1 enzyme has a dual role in the SUMO modification pathway since it is needed to process the SUMO precursor and deconjugates SUMO from substrate proteins (11). Hence, the ulp1 ts strain suffers from a lack of mature SUMO that can subsequently be activated for conjugation by the Uba2/ Aos1 dimer. In addition, ulp1 cells accumulate a distinct set of sumoylated proteins due to their impaired ability to deconjugate SUMO substrates (15,20). To test which function of Ulp1 was needed for nuclear protein import we transformed a plasmid-encoding mature SUMO into wild type and ulp1 strains and performed the import assay under the same conditions as above. As can be seen in Fig. 1B, bottom row, expression of mature SUMO does indeed cure the import defect in ulp1 cells indicating that this defect is primarily due to a lack in precursor processing and thus SUMO conjugation. Consistent with this assumption, expression of mature SUMO does not rescue the cNLS import defect in uba2 cells (not shown). Taken together, the results from the in vivo protein import experiments suggest that in yeast as has been suspected for vertebrates, the SUMO modification system is involved in nuclear transport processes. In addition, our results implicate that it is SUMO conjugation rather than precursor processing or substrate deconjugation that is needed for proper nuclear import.
RgNLS-GFP Reporter Protein Import and mRNA Export Are Not Impaired in ulp1 and uba2 Cells-To investigate whether the import defect in ulp1 and uba2 cells is specific for the cNLS pathway, we examined another protein import pathway which is used by proteins carrying a so-called rgNLS signal. The rgNLS is an arginine/glycine-rich nuclear import signal of the yeast Nab2 RNA binding protein (30,34,35) and is recognized by the importin ␤-like transport receptor Kap104 (36). To analyze rgNLS-dependent protein import, we used a rgNLS-GFP reporter protein in the in vivo protein localization assay. Fig.  2A shows the results of this experiment: as in wild type (top row, re-import at 37°C), ulp1 and uba2 cells show no significant cytoplasmic accumulation of the reporter protein after a 25-min shift to the non-permissive temperature ( Fig. 2A, middle and bottom rows, re-import at 37°C). This suggests that the import defect observed in ulp1 and uba2 cells is specific for the cNLS pathway. To rule out the possibility that this defect is due to a lack of mRNAs encoding proteins required for protein import, we performed in situ hybridization analysis to test for a general block in nuclear mRNA export. In wild type and ulp1 cells, we do not observe an accumulation of bulk poly(A)ϩ RNA in the nucleus after a shift to 37°C for 1 h (Fig. 2B, top and  bottom rows, FITC). This suggests that general mRNA export occurs normally in these strains. We also tested ulp2⌬ and uba2 mutants in this assay and observed no mRNA export defect (not shown).

Srp1 Accumulates in the Nucleus of ulp1 and uba2 Cells-
Since we observed a cytoplasmic accumulation of a cNLS-GFP reporter protein in ulp1 and uba2 strains, we were interested in identifying the components of this particular nuclear import pathway that would be affected by a lack of SUMO conjugation. In budding yeast, RanGAP1 (Rna1) is not sumoylated, and an apparent ortholog of RanBP2/Nup358 is lacking (37). These findings together with our data indicate that other additional targets within and specific for cNLS-dependent nuclear import are controlled by sumoylation. CNLS-dependent protein import is one of the best characterized nuclear transport pathways (38). The cNLS receptor is the importin ␣/␤ dimer (Srp1 and Kap95 in yeast). Srp1 binds to NLS signals of this class in the cytoplasm, associates with Kap95 and in this complex, travels through the NPC. In the nucleus, Kap95 binds to RanGTP thereby releasing Srp1 and its NLS cargo. To investigate whether the cNLS import defect in ulp1 and uba2 cells is caused by a mislocalization of Srp1, we used indirect immuno-FIG. 2. rgNLS-GFP import and mRNA export are not perturbed in ulp1 and uba2 cells. A, wild type, ulp1 (MHY1488) and uba2 (JD-90-1A) strains were transformed with a plasmid encoding the rgNLS-GFP reporter protein and selected on SD/-Ura plates. The in vivo protein import assay was performed as in the legend to Fig. 1. B, mRNA localization in wild type and ulp1 (MHY1488) strains. Cells were grown to 0.4 OD 600 at 25°C in liquid YPD medium, shifted to 37°C for 1 h, and prepared for in situ hybridization analysis as described under "Experimental Procedures." Bulk poly(A)ϩ RNA was detected using a digoxigenin-labeled oligo(dT) 50 probe, an anti-digoxigenin antibody, and a secondary antibody coupled to FITC. Nuclei were visualized with DAPI, and cell morphology was checked by Nomarski optics (DIC). fluorescence analysis on yeast strains, which had been shifted to the non-permissive temperature for 30 min. Under these conditions, Srp1 is found in the nucleus and cytoplasm of wild type cells (Fig. 3, 1st row, anti-Srp1). As a control, DAPI staining is used to reveal the position of nuclei (Fig. 3, 1st row, DAPI). However, in strains which either lack the nucleoporin Nup2 (nup2⌬) or carry a conditional allele of the yeast Ran-GAP1, rna1, the pattern is different; in these cells, Srp1 concentrates in the nucleus due to a lack in its nuclear export (Fig.  3, 2nd and 3rd row, anti-Srp1) (31, 39 -41). Remarkably, a similar pattern of nuclear Srp1 accumulation is also seen in the ulp1 and uba2 mutants (Fig. 3, 4th and 5th row, anti-Srp1). This suggests that the cNLS import defect observed in these strains is in fact due to a mislocalization of the cNLS receptor subunit Srp1 to the nucleus.
Kap95 Is Not Mislocalized in ulp1 and uba2 Strains-Once in the nucleus, the import complex composed of cNLS substrate, Srp1, and Kap95 is dissociated by the binding of RanGTP to Kap95. While Kap95 leaves the nucleus in a complex with RanGTP, nuclear export of Srp1 requires an exportin of the karyopherin-␤ class known as Cse1 (42)(43)(44). Srp1 only binds to Cse1 in an NLS-free state in the presence of RanGTP (44). After translocation into the cytoplasm, Srp1 and Kap95 are released from their respective export complexes by GTP hydrolysis on Ran. This step is mediated by the yeast proteins Rna1 (RanGAP1) and Yrb1 (RanBP1), two cytosolic proteins generally required for the termination of transport processes (Ref. 44 and references therein). It is obvious from this sequence of events that a lack of GTP hydrolysis in the cytoplasm must ultimately lead to severe perturbations of nuclear import and export processes. Since Srp1 accumulates in the nucleus of rna1 cells (Fig. 3) and RanGAP1 is a substrate for sumoylation in higher eukaryotes, we wanted to test whether it is a lack of GTP hydrolysis resulting in nuclear accumulation of Srp1 in ulp1 and uba2 cells. As a tool to study the function of the Ran machinery in living cells, we employed the importin-␤ subunit Kap95, which has been shown to change its localization in mutants of the yeast RanGAP1 RNA1 and the RanGEF PRP20 (41). When Kap95 is tagged with GFP in wild type cells, the protein localizes to the cytoplasm and nuclear rim at 25°C and 37°C, respectively (Fig. 4, WT). However, this localization pattern remarkably changes in strains carrying mutations in RNA1 and PRP20. When these cells are shifted to the nonpermissive temperature for 30 min, Kap95-GFP accumulates at the nuclear rim and in the nuclear interior due to a lack in cytoplasmic GTP hydrolysis in the case of rna1 and nuclear RanGTP depletion in the case of prp20 (Fig. 4, rna1 and prp20 at 37°C) (41). In ulp1 and uba2 mutants, Kap95-GFP localization is identical to that observed for wild type cells at both temperatures, indicating that in these strains the Ran machinery is functional. In summary, our experiments suggest that nuclear accumulation of Srp1 observed in ulp1 and uba2 mutants does not result from a compromised Ran machinery in these strains. In addition, since only Srp1 but not Kap95 is mislocalized to the nuclear interior it might in fact be nuclear export of Srp1, which is perturbed in strains lacking SUMO conjugation.
A Fraction of Nup2 Is Nuclear in ulp2⌬ Cells-During translocation, Srp1 interacts with a specific set of nucleoporins (45,46) including Nup1, Nup2, and Nup60, which are located at the nucleoplasmic side of the NPC (31, 47). Nup2 is unique among FIG. 3. Srp1 is nuclear in ulp1 and uba2 mutants. Indirect immunofluorescence analysis of wild type, nup2⌬, rna1, ulp1 (MHY1488), and uba2 (JD90-1A) strains. Cells were grown in liquid YPD medium to 0.4 OD 600 , shifted to 37°C for 30 min, and processed for immunofluorescence analysis as described under "Experimental Procedures." Srp1 was detected using an anti-Srp1 antibody and a secondary FITC-coupled antibody. DAPI staining reveals the position of nuclei.

FIG. 4. Kap95-GFP is not mislocalized in ulp1 and uba2 strains.
In vivo localization of Kap95-GFP. The chromosomal copy of KAP95 was C-terminally tagged with GFP in wild type, rna1, prp20, ulp1, and uba2 strains using a PCR-based integration strategy and selection on SD/-His. Yeast were grown to 0.4 OD 600 in the same medium, cultures were split and incubated at the permissive and non-permissive temperatures for 30 min before photographs were taken. the members of this group since it was found not to be permanently associated with the pore complex (26) and to interact with Srp1 during its export from the nucleus (31,39,40). Since Srp1 is nuclear in ulp1, uba2, and nup2⌬ strains (Fig. 3), we wanted to know whether a lack or mislocalization of Nup2 is responsible for Srp1 accumulation in SUMO modification pathway mutants. We therefore tagged Nup2 with GFP and examined its localization in living yeast. In wild type cells, Nup2-GFP is mainly associated with the NPC at the nuclear periphery (Fig. 5A, 1st row, GFP). However, a faint GFP signal is also observed in the nucleoplasm, consistent with Nup2's role as a mobile nucleoporin (26). In a nup60⌬ control strain, Nup2-GFP accumulates in the nuclear interior (Fig. 5A, 2nd row, GFP) since Nup60 is needed to anchor Nup2 at the NPC (48). In contrast, ulp1 cells show a wild type distribution of Nup2-GFP (Fig. 5A, 3rd row, GFP). Interestingly, in a ulp2⌬ strain, Nup2-GFP is found at the nuclear rim and in the nucleoplasm, suggesting that not all Nup2-GFP is NPC-associated (Fig. 5A,  4th row, GFP). We have quantified Nup2-GFP fluorescence along nuclear bisects for the strains tested and the results are shown in Fig. 5B: as can be seen from the profile plot for ulp2⌬ cells (Fig. 5B, 4th panel), the Nup2-GFP signal is shifted toward the nucleoplasm when compared with wild type cells (Fig.  5B, 1st panel) although not as strongly as in nup60⌬ cells (Fig.  5B, 2nd panel). In contrast, ulp1 cells show a wild type pattern of Nup2-GFP distribution (Fig. 5B, 3rd panel) and this is also true for uba2 cells (not shown). Taken together, these results indicate that the nuclear accumulation of Srp1 observed in ulp1 and uba2 cells is not due to a lack or mislocalization of Nup2 in these strains. Furthermore, since we have tested all strains at various temperatures and still observed no changes in Nup2-GFP localization at the nuclear rim (data not shown), we can exclude the possibility that nuclear membrane architecture is generally compromised in SUMO modification pathway mutants.
Srp1 Is Mislocalized at the NPC in ulp1 Cells-Although we did not observe changes in Nup2 localization in ulp1 and uba2 cells, we could not rule out the possibility of other structural alterations at the NPC, which might have lead to nuclear Srp1 accumulation and concomitantly a block in cNLS import. Ulp1 has been shown to localize to the nuclear envelope (20) and NPCs in yeast (15,49). Recently, human SENP2, one of several Ulp1 homologs in human cells, was found to localize to the nuclear basket structure of the NPC and it was speculated that SENP2 localization might be important for regulation of pore activity (50). Therefore, to investigate Srp1 localization at the NPC level, we performed immunogold EM with wild type, nup2⌬, and ulp1 strains. This technique allows the localization of nuclear pore components and soluble transport factors in the nanometer range thereby providing an excellent tool to detect even subtle changes in protein localization (31,51). Fig. 6A shows representative EM micrographs of yeast NPCs where Srp1 was localized by the use of an anti-Srp1 antibody and a secondary antibody coupled to 12-nm colloidal gold. The position of pore-associated gold particles was determined with respect to the central plane and the 8-fold symmetry axis of the FIG. 5. Nup2-GFP localization is not compromised in ulp1 and uba2 strains. A, Nup2-GFP localization. The chromosomal copy of NUP2 was C-terminally tagged with GFP in wild type, nup60⌬, ulp2⌬, and ulp1 (IS90-1B) strains as described for Kap95-GFP. Cells were grown at 25°C in liquid SD/-His medium to mid-log phase and photographed. Cell morphology was checked by Nomarski optics (DIC). B, quantitation of Nup2-GFP fluorescence. Plots of the fluorescence intensity across a nuclear bisect for each of the strains tested. 11-16 cells (n designates the number of cells) were examined for each strain. The thick line indicates the mean fluorescence distribution. 6. Srp1 is mislocalized at the NPC in ulp1 cells. A, representative EM micrographs of wild type yeast NPCs. Cells were grown in liquid YPD medium at 25°C to 0.4 OD 600 , harvested by centrifugation and subsequently prepared for immunogold EM. Ultrathin cryosections of whole cells were sequentially incubated with an anti-Srp1 antibody and a secondary antibody coupled to 12-nm colloidal gold. Individual gold particles are indicated by arrowheads. N and C designate the nuclear and cytoplasmic sides of the pore, respectively. The bar represents 100 nm. B, distribution of gold particles with respect to central and 8-fold symmetry axes at the NPC. Between 200 and 330 gold particles were localized for each strain. C, statistical analysis of Srp1 localization at the NPC. The histograms show the distribution of gold particles (% total gold) in terms of their vertical distance (nm) from the central plane of the NPC.

FIG.
NPC (31,52). Fig. 6B compiles the data of roughly 200 -300 individual gold particles at the NPCs of wild type, nup2⌬, and ulp1 strains. In wild type cells, we observe Srp1 at three distinct locations: close to the nuclear side of the central channel (0 to Ϫ50 nm, represented by gold particle 1 in Fig. 6A), in the center of the nuclear basket (Ϫ50 nm to Ϫ90 nm, gold particle 2) and at the distal ring of the nuclear basket (Ϫ90 nm to Ϫ120 nm, gold particle 3). A nup2⌬ control strain shows a similar but not identical pattern of Srp1 localization: here, the Srp1 signal clearly shifts toward the nuclear basket structure of the NPC consistent with a lack of Srp1 export in this strain (Fig. 6B,  nup2⌬). In the ulp1 mutant, Srp1 localization is similar to that observed for nup2⌬ cells (Fig. 6B, ulp1) indicating that like in nup2⌬, NPC structure might indeed be compromised. We have calculated the relative amounts of gold particles found at a specific distance from the NPC midplane. As can be seen from the histograms in Fig. 6C and Table I, the ulp1 and nup2⌬ mutants show an increased percentage of gold particles in the middle region (Ϫ50 to Ϫ90 nm, 26.8 and 22%, respectively) and terminal ring of the nuclear basket structure (Ϫ90 to Ϫ120 nm, 15 and 24.3%, respectively) as compared with wild type cells (same regions with 16.6% and 10.2%, respectively). Taken together, our results from indirect immunofluorescence analysis and immuno-EM show that in ulp1 cells a substantial fraction of Srp1 is trapped in the nuclear interior and nuclear basket structure of the NPC. Since overall Srp1 localization in ulp1 cells is very similar to that observed for nup2⌬ cells, this might indicate that Srp1 mislocalization is in fact due to a compromised NPC architecture in ulp1 cells. DISCUSSION In the present work, we have investigated nucleocytoplasmic transport in yeast mutants of the SUMO modification pathway in vivo. Our results show that in mutants deficient in SUMO conjugation such as ulp1 and uba2 cNLS-dependent protein import is impaired (Fig. 1A). However, other nuclear transport pathways are not generally affected since nuclear uptake of another NLS, the rgNLS, is not perturbed and general mRNA export is not blocked (Fig. 2, A and B). It is interesting to note, that the cNLS import defect in ulp1 cells can be rescued by overexpression of mature SUMO from a plasmid (Fig. 1B). In combination with data for the uba2 strain (Fig. 1A) this suggests that it is SUMO conjugation rather than SUMO precursor processing that is needed for protein import. Ulp1's other function as a deconjugating enzyme cannot be rescued by overexpression of mature SUMO. This implicates that SUMO deconjugation is not needed for cNLS-dependent protein import, which is consistent with our observation that a strain deleted for ULP2 (ulp2⌬), the other SUMO-deconjugating enzyme, does not show an import defect or mislocalization of Srp1 ( Fig.  1A and data not shown). What might be the reason for the observed cNLS import block? CNLS-dependent import is one of the best characterized nuclear transport pathways so far (38). In vertebrates, nuclear uptake of cNLS-containing transport cargos is mediated by the importin ␣/␤ dimer, or Srp1/Kap95 in yeast (53)(54)(55)(56)(57). In the cytosol, where RanGTP levels are low, Srp1 binds the cNLS substrate and subsequently associates with Kap95 to translocate through the NPC. In the nucleus, Kap95 binds to RanGTP thereby releasing Srp1 and the NLS substrate. Therefore, a high nuclear RanGTP concentration fulfils two purposes: first, import complexes are dissociated on the nuclear side of the pore by binding of RanGTP to the importin-␤ like transport receptor, and second, export complexes between RanGTP binding exportins such as Cse1 and their respective export cargos can be formed. A nuclear accumulation of Srp1 in ulp1 and uba2 strains might therefore reflect either a lack of dissociation of the import complex due to a low nuclear RanGTP concentration, a defect in export complex formation and/or a lack of export complex docking at the NPC. Since we observe that the importin-␤ subunit Kap95 is not mislocalized in ulp1 and uba2 strains (Fig. 4) as can be observed in Ran machinery mutants such as rna1, in which both importin subunits are mislocalized (Figs. 3 and 4), we can exclude the possibility that nuclear RanGTP concentrations significantly fluctuate in ulp1 and uba2 mutants thereby perturbing nuclear transport processes. In line with these findings is our observation that other Ran-dependent transport processes such as rgNLS-dependent protein import and mRNA export are functional in ulp1 and uba2 strains. Taken together, these results suggest that at least in yeast, a defect in SUMO conjugation has no significant impact on the Ran machinery with respect to its function in nuclear transport. Apart from the fact that in higher eukaryotes sumoylated RanGAP1 and RanBP2 are NPC-associated, they have also been found to co-localize with kinetochores and mitotic spindles which might suggest a role in the re-establishment of the Ran gradient in cells with an open mitosis (58). Genetic and biochemical data from yeast suggest that during translocation, Srp1 containing import and export complexes interact with a distinct subset of nucleoporins such as Nup1, Nup2, and Nup60 (45,46,59). These nups and another nucleoporin implicated in protein import, Nsp1 (60,61), belong to the so-called FXFG class of nups, which have been shown to reside in the nuclear basket structure and close to the central transporter in the NPC, respectively (31,47,51). To investigate a possible role of the NPC for Srp1 export in ulp1 and uba2 strains, we tagged Nup2, a nucleoporin previously implicated in Srp1 export (31,39,40) with GFP and investigated its localization in living yeast. We find that Nup2 is not mislocalized in ulp1 and uba2 cells at permissive and non-permissive temperatures, respectively ( Fig. 5 and data not shown). We also do not observe gross structural alterations of the nuclear membrane, such as membrane blebbing or, pore clustering, at any temperature tested (data not shown). Surprisingly, in the case of ulp2⌬, an intranuclear pool of Nup2 can be observed (Fig. 5, A and B). This result prompted us to investigate the stoichiometry of some of the nups known to interact with Srp1 during NPC translocation. Using Northern blot and immunoblot analysis, we find that mRNA and protein levels of Nup2 and Nsp1 are not altered in ulp1 and uba2 mutants suggesting that the stoichiometry of these nups is normal in these strains (data not shown). However, despite some intranuclear Nup2 protein in ulp2⌬ cells, we observe no transport defect in this strain (Fig. 1A). This indicates that Ulp2 and hence deconjugation of SUMO substrates has no significant role in nucleocytoplasmic transport. Changes in NPC composition and/or architecture are known to impair nucleocytoplasmic trafficking (62). Indeed, Srp1 is not only mislocalized to the nuclear interior in ulp1 cells but also at the NPC as we can show by immunofluorescence analysis and immunogold EM (Figs. 3 and 6). This finding might indicate that Srp1 is a target for SUMO modification and as a result can no longer efficiently bind at appropriate sites in the nuclear basket structure of the NPC in mutants compromised in SUMO conjugation. An alternative explanation is that another component of the transport machinery, possibly a nuclear pore protein involved in Srp1 export is somehow impaired in its function due to a lack of SUMO conjugation in these mutants. In immunoblot analysis using specific antibodies, we were not able to detect a SUMO modification on Srp1, the Srp1 export receptor Cse1 and nups involved in Srp1 export (data not shown). Extracts were prepared from ulp1, uba2, ulp2⌬, and wild type strains grown at permissive and non-permissive temperatures according to a protocol known to preserve SUMO conjugates (13). Although we were not able to identify a SUMO-conjugated protein probably responsible for the observed mislocalization of Srp1, our results demonstrate for the first time a role for the SUMO conjugation system in a specific nuclear protein import pathway in yeast. Nuclear Srp1 mislocalization in ulp1 and uba2 strains suggests an involvement of SUMO conjugation in Srp1 export. How the SUMO system might influence Srp1 export complex formation or possibly its interaction with the NPC is not clear at this point. However, there are intriguing parallels between yeast and higher eukaryotes since the nucleoporin RanBP2/Nup358 can be sumoylated in vitro (21). More recently, it was shown that RanBP2/Nup358 has E3 ligase activity toward some cNLScontaining substrates and it was speculated that sumoylation and nuclear import might be coupled processes (22). RanBP2/ Nup358 is a multifunctional nucleoporin since it has Ranbinding motifs, zinc-finger domains, and FG repeats (63) and has been implicated in the termination reaction of nuclear protein import (64). In addition, SENP2, an Ulp1 homolog in human cells has been shown to interact with Nup153, a nucleoporin in the nuclear basket structure of the NPC (50). Nup153 has no homolog in yeast but intriguingly, importin ␣/␤-mediated protein import is strongly reduced in nuclei that lack Nup153. This defect is specific since another protein import pathway, transportin-mediated M9-NLS import is not perturbed (65). We have tested the yeast nups known to reside in the nuclear basket structure of the NPC for a modification with SUMO. However, we were not able to detect sumoylated Nup1, Nup2, Nsp1, and Nup60 (data not shown). Interestingly, the enzymes of the SUMO modification pathway are either mostly nuclear (Uba2, Aos1, Ubc9, and Ulp2) or NPC-associated (Ulp1) (12,15,20,49,66,67) but whether mutations in these proteins might influence NPC structure and concomitantly nuclear transport is currently not known. Thus, the elucidation of the molecular details of SUMO's role in nuclear protein import and possibly NPC architecture remains a challenging task for future research.