Interaction between Ran and Mog1 Is Required for Efficient Nuclear Protein Import*

Mog1 is a nuclear protein that interacts with Ran, the Ras family GTPase that confers directionality to nuclear import and export pathways. Deletion of MOG1 in Saccharomyces cerevisiae ( (cid:1) mog1 ) causes temperature-sen-sitive growth and defects in nuclear protein import. Mog1 has previously been shown to stimulate GTP release from Ran and we demonstrate here that addition of Mog1 to either Ran-GTP or Ran-GDP results in nucleotide release and formation of a stable complex between Mog1 and nucleotide-free Ran. Moreover, MOG1 shows synthetic lethality with PRP20 , the Ran guanine nucleotide exchange factor (RanGEF) that also binds nucle-otide-free Ran. To probe the functional role of the Mog1-Ran interaction, we engineered mutants of yeast Mog1 and Ran that specifically disrupt their interaction both in vitro and in vivo . These mutants indicate that the interaction interface involves conserved Mog1p residues Asp 62 and Glu 65 , and residue Lys 136 in yeast Ran. Mutations at these residues decrease the ability of Mog1 to bind and release nucleotide from Ran. Furthermore, the E65K-Mog1 and K136E-Ran mutations in yeast cause temperature sensitivity and mislocalization of a nuclear import reporter protein, similar to the phenotype observed for (cid:2) -Myc antibody (1:2000 dilution) or (cid:2) -GST antibody (1:500 dilution). To analyze the mutant proteins and confirm their expression at equivalent levels to wild-type protein, equal concentrations of lysates (generally (cid:4) 10 (cid:3) g of total protein) were resolved by SDS-PAGE and immunoblotted with either monoclonal (cid:2) -Myc antibody (1:2000 dilution) or a polyclonal (cid:2) -Mog1p antibody (1:5000 dilution). Nuclear Protein Localization— An NLS-GAL4AD-GFP reporter plasmid (pAC697) (29) was transformed into wild-type (ACY193), (cid:1) mog1 (ACY547) , E65K-Mog1 (ACY548), and K136E-Gsp1 (ACY570) yeast cells. Cells were grown to log phase at 25 °C and shifted to 37 °C for 4 h. For all experiments, cells were stained with 4 (cid:2) ,6-diamidino-2-phenylin-dole to visualize the DNA and confirm the location of the nucleus (data not shown). The localization of the NLS reporter was monitored by directly viewing the GFP signal in living cells through a GFP optimized filter (Chroma Technology) using an Olympus BX60 epifluorescence microscope equipped with a Photometrics Quantix digital camera.

The nucleotide state of the Ras family GTPase, Ran, governs the interactions between nuclear transport receptors and protein substrates. Import receptors bind cargo in the absence of Ran-GTP whereas export receptors bind cargo in a trimeric complex with Ran-GTP (reviewed in Ref. 1). The directionality of transport is based on nuclear Ran being primarily GTPbound and cytoplasmic Ran mainly GDP-bound. Ran-GTP is generated in the nucleus by the Ran guanine nucleotide exchange factor (RanGEF), 1 RCC1 (2, 3). Conversely, the primary Ran GTPase-activating protein, RanGAP1 (4), is localized in the cytoplasm (5,6) and is believed to maintain cytoplasmic Ran in the GDP-bound state. NTF2 recycles Ran-GDP from the cytoplasm to the nucleus (7,8) where Ran-GTP is regenerated by RCC1. Thus, the asymmetric distribution of the two nucleotide states of Ran, vital to the directionality of nuclear transport, is maintained through the concerted efforts of several Ran-binding proteins, including RCC1, RanGAP1, and NTF2.
Mog1, a recently identified Ran-binding protein, appears to have a novel role in regulating the Ran GTPase cycle. Mog1p was first isolated in yeast as a suppressor of temperaturesensitive alleles of the Saccharomyces cerevisiae Ran homologue, Gsp1p (9). Mog1 is predominantly nuclear and both Mog1 and its interaction with Ran are evolutionarily conserved (9 -12). The yeast ⌬mog1 strain is temperature-sensitive for growth and shows defects in signal-mediated protein import, suggesting a role for Mog1p in nuclear protein import (9). The temperature sensitivity of ⌬mog1 is suppressed by overexpression of either GSP1 or NTF2 and many of the temperaturesensitive gsp1 mutants suppressed by MOG1 are also suppressed by NTF2 (9). The genetic interaction between MOG1 and NTF2 implies that the functions of these two proteins may at least partially overlap. Recently, Mog1 has been reported to stimulate nucleotide release from Ran-GTP (11,13). Although these observations suggest an involvement of Mog1 in regulating the nucleotide state of Ran, the precise function of Mog1 in Ran-coordinated nuclear transport has not been defined.
We previously described the 1.9-Å resolution crystal structure of S. cerevisiae Mog1p and identified a highly conserved cluster of surface residues as a putative Ran-binding site (10).
Here we have used structure-based targeted mutagenesis of yeast Mog1p and Gsp1p to probe the interaction interface of the protein complex and the cellular function of this interaction. We show that the conserved acidic surface residues Asp 62 and Glu 65 of Mog1p are important in maintaining the Mog1-Ran interaction. Although Mog1p and Ntf2p cannot bind Gsp1p simultaneously (10), they do not appear to bind to the same site. Hence, mutations in the switch I and II loop regions of Ran that disrupt NTF2 binding (14) do not prevent Mog1 binding and conversely, residue Lys 136 in Gsp1p is required for interaction with Mog1p but not with Ntf2p. The E65K-Mog1p and K136E-Gsp1p mutations decrease the strength of the Mog1-Ran interaction and cause defects in growth and nuclear protein import in yeast, indicating that a direct Mog1-Ran interaction is important for these functions in vivo.

Recombinant Protein Expression and Purification-Yeast
Mog1p was expressed in Escherichia coli and purified as described (15). His-tagged human Mog1, which is 29% identical and 47% similar to yeast Mog1p, was expressed and purified as described (12). Yeast Ntf2p was prepared as previously described for vertebrate NTF2 (16) and the yeast homologue of importin-␤, Kap95p, was purified according to the protocol used for vertebrate importin-␤ (17).
The open reading frame of GSP1 was amplified by PCR and cloned into the NdeI/BamHI sites of pET15b (Novagen) to generate an NH 2terminal His-tag. His-tagged Gsp1p was expressed in E. coli strain BL21 (DE3) and purified by Ni-NTA affinity and gel filtration chromatography (12). The nucleotide state of Gsp1p was modulated by incubating the bacterially expressed protein with GDP, GTP, mantGDP, or mantGTP at 100-fold molar excess in the presence of 20 mM EDTA for 30 min at room temperature. Nucleotide exchange was completed by the addition of 40 mM MgCl 2 . Unbound nucleotide was removed and the Gsp1p exchanged into the desired buffer using a PD-10 gel filtration column (Amersham Pharmacia Biotech). Wild-type canine Ran and Ran mutants were expressed as described (14), purified by ion exchange chromatography on a Q-Sepharose column (Amersham Pharmacia Biotech), and charged with GDP or GTP as described above.
Mutagenesis of Mog1p and Gsp1p-To improve protein expression and stability of recombinant Mog1p, we created an NH 2 -terminal truncation construct that uses Met 31 as the translation start site. Aminoterminal truncated Mog1p retained the ability to bind Gsp1p in vitro and complemented the temperature-sensitive phenotype of ⌬mog1 (data not shown). Amino acid substitutions were introduced into Mog1p and Gsp1p expression constructs by PCR-based site-directed mutagenesis (QuikChange Site-directed Mutagenesis Kit, Stratagene). For expression in yeast, MOG1 and GSP1 open reading frames cloned in pBluescript (Stratagene) were mutagenized and then subcloned into yeast plasmids. For each mutant, the presence of the desired mutations and absence of extraneous mutations was confirmed by sequencing the open reading frame. Gsp1p mutants were screened for binding to Mog1p-Sepharose and Ntf2p-Sepharose beads as described (18).
His-tagged Protein Binding Assays-His-tagged proteins were prebound to Ni-NTA agarose (Qiagen) in phosphate-buffered saline supplemented with 5 mM MgCl 2 . Binding to the immobilized His-tagged protein was assayed in reactions containing an excess of the second protein in 0.5 ml phosphate-buffered saline/MgCl 2 . Following incubation for 1 h at 4°C, unbound material was removed by washing the beads extensively with buffer. Bound proteins were eluted from the beads by boiling in SDS sample buffer and then examined by SDS-PAGE.
X-ray Crystallography-Crystals of E65A-Mog1p and E65K-Mog1p were grown by hanging drop vapor diffusion using the conditions employed for the wild-type protein (15) and had P2 1 2 1 2 1 symmetry. A 1.9-Å resolution data set was collected on beamline EH14 -2 (ESRF, Grenoble) using 0.933-Å wavelength radiation (Table I) and processed using MOSFLM and SCALA (CCP4). The E65A-Mog1p structure was solved by molecular replacement using AMORE (CCP4) and the wildtype structure (10). The structural model obtained after refinement using CNS (19) had excellent statistics (Table I) and very closely resembled the wild-type structure (C ␣ room mean square deviation of 0.31 Å).
Fluorescent MantGDP/GTP Release Assay-Gsp1p (20 M) charged with either mantGDP or mantGTP (synthesized as described previously (20)) was mixed with an equimolar quantity of wild-type or mutant Mog1p or, as a control, with buffer alone (20 mM Tris-HCl, pH 7.4, 100 mM NaCl, 5 mM MgCl 2 ). Reactions were incubated for 1 h at room temperature, followed by gel filtration chromatography using a calibrated Superdex 75 FPLC column (Amersham Pharmacia Biotech) at a flow rate of 0.5 ml/min. Fractions were analyzed for protein content by measuring absorbance at 280 nm and for mant nucleotide content by exciting at 366 nm and measuring fluorescence emission at 450 nm. Peak fractions were also concentrated and examined by SDS-PAGE to determine which fractions contained Gsp1p, Mog1p, or both (data not shown).
Construction of Yeast Mutants-A complete deletion of the MOG1 open reading frame was created using a previously described PCRbased strategy (21) in the wild-type diploid ACY247 (22). The heterozygous diploid was subsequently sporulated and tetrads were dissected to generate the haploid ⌬mog1 strain (ACY547). The prp20 -1⌬mog1 double mutant (ACY495) was created by deleting the MOG1 open reading frame in the haploid prp20-1 mutant (23). Deletion of MOG1 was confirmed by PCR. The E65K mutation was introduced into the MOG1 open reading frame cloned in the HIS3 integrating plasmid, pRS303 (24). E65K-Mog1 was then integrated at the endogenous MOG1 locus by linearization of E65K-Mog1-pRS303 (pAC868) and transformation into the wild-type haploid ACY193 (25) to create ACY548. Transformants that grew on plates lacking histidine were selected and the presence of the E65K-Mog1 mutation was confirmed by PCR and sequencing. To generate cells carrying K136E-Gsp1p or Myc-tagged Gsp1p plasmids as the sole source of Gsp1p, ⌬gsp1⌬gsp2 cells (ACY212) (26) were transformed with centromeric plasmids encoding K136E-Gsp1p (pAC838), myc-Gsp1p (pAC627), or myc-K136E-Gsp1p (pAC839) and streaked on 5-fluoroorotic acid to remove the wild-type GSP1 plasmid (27).
In Vivo Analysis-For growth analysis, yeast cells were grown to saturation in selective media at 25°C. Cells were diluted to 5 ϫ 10 6 cells/ml and shifted to 37°C. Growth was monitored by measuring the optical density (A 600 nm) at 2-h intervals. Growth on plates was assessed by growing cultures to saturation as above, serial diluting them (1:10), and spotting onto plates. Plates were incubated at either 25 or 37°C for 5 days.
GST Binding Experiments/Immunoblotting-Yeast expressing myc-Gsp1p or myc-K136E-Gsp1p as the only source of Gsp1p were transformed with 2 galactose inducible plasmids (28) encoding GST alone (pAC403), GST-Mog1p (pAC590), GST-D62K-Mog1p (pAC877), or GST-E65K-Mog1p (pAC878). GST binding experiments were performed as described previously (12) except that cells were induced with galactose for 4 h at room temperature. One g of total protein from lysate and unbound samples and 30 l of the bound sample were resolved on a 12% SDS-PAGE gel, then transferred to nitrocellulose and probed with monoclonal ␣-Myc antibody (1:2000 dilution) or ␣-GST antibody (1:500 dilution). To analyze the mutant proteins and confirm their expression at equivalent levels to wild-type protein, equal concentrations of lysates (generally ϳ10 g of total protein) were resolved by SDS-PAGE and immunoblotted with either monoclonal ␣-Myc antibody (1:2000 dilution) or a polyclonal ␣-Mog1p antibody (1:5000 dilution).
Nuclear Protein Localization-An NLS-GAL4AD-GFP reporter plasmid (pAC697) (29) was transformed into wild-type (ACY193), ⌬mog1 (ACY547), E65K-Mog1(ACY548), and K136E-Gsp1 (ACY570) yeast cells. Cells were grown to log phase at 25°C and shifted to 37°C for 4 h. For all experiments, cells were stained with 4Ј,6-diamidino-2-phenylindole to visualize the DNA and confirm the location of the nucleus (data not shown). The localization of the NLS reporter was monitored by directly viewing the GFP signal in living cells through a GFP optimized filter (Chroma Technology) using an Olympus BX60 epifluorescence microscope equipped with a Photometrics Quantix digital camera.

Engineered Mog1p Mutants with Decreased Affinity for
Ran-The crystal structure of Mog1p showed a cluster of highly conserved residues (Fig. 1A) on a ϳ25-Å diameter surface patch (10). Many of these residues are charged and could potentially form intermolecular salt bridges with oppositely charged residues on Ran. We targeted these charged residues on Mog1p for mutagenesis and compared the mutant proteins to wild-type Mog1p for their ability to bind Gsp1p-GDP and Gsp1p-GTP. Wild-type Mog1p bound to Gsp1p in either nucleotide state (Fig. 1B), although the amount of Gsp1p-GTP bound was greater than Gsp1p-GDP. The binding of Gsp1p in both nucleotide states to Mog1p beads is consistent with recent reports that Mog1 binds not only Ran-GTP as initially suggested (9) but Ran-GDP as well (10,12). Although mutations at Arg 58 (R58A or R58E) did not reduce Gsp1p binding (data not shown), mutations at either Asp 62 (D62K) or Glu 65 (E65K and E65A) significantly reduced the affinity of Mog1p for both nucleotide states of Gsp1p in vitro (Fig. 1B). The effect of the Mog1p mutations was even more pronounced in vivo since the amount of Gsp1p that co-purified with GST-tagged Mog1p from yeast cells was decreased to undetectable levels for both D62K-Mog1p and E65K-Mog1p (Fig. 1C).
Since residues Asp 62 and Glu 65 are on the surface of Mog1p, it is unlikely that mutating them to lysine would cause an overall conformational change in the molecule. To investigate the effect of mutations in these residues, we used crystallography and CD spectroscopy. We solved the crystal structure of the E65A-Mog1p mutant by molecular replacement to 2.0-Å reso-lution (Table I). The C ␣ root mean square deviation between the E65A-Mog1p and wild-type Mog1p crystal structures was 0.31 Å indicating that negligible conformational changes had been introduced by the mutation. Similar crystals were also obtained for the D62K-Mog1p and E65K-Mog1p mutants, suggesting that they too retained the same three-dimensional structure as wild-type Mog1p, although a severe twinning problem prevented us from obtaining high-resolution crystal structures for these two mutants. The CD spectra for wild-type Mog1p and the mutants were identical (data not shown), consistent with the mutants having the same conformation as wild-type Mog1p.
Mog1p-stimulated Nucleotide Release from Ran-GDP and Ran-GTP-Our binding experiments (Fig. 1B) suggested that, in contrast to previous reports (11,13), Mog1 might bind and stimulate nucleotide release from both the GDP-and GTPbound states of Ran. To investigate this possibility in more detail, we developed a guanine nucleotide release assay using the fluorescent nucleotide analogues, mantGDP and mantGTP, which are comparable to unmodified nucleotides in their affinity for Ran (30). Following incubation of Mog1p with Gsp1p-mantGDP or Gsp1p-mantGTP, samples were resolved by gel filtration and fractions were examined for protein absorbance and fluorescence emission. This allowed the relative amounts of Mog1p-Gsp1p complex, free protein monomers, and free nucleotide to be compared. This assay demonstrated that addition of Mog1p to either nucleotide state of Gsp1p generated a 1:1 complex between wild-type Mog1p and nucleotide-free Gsp1p with the concomitant release of either GDP or GTP, whereas in the absence of Mog1p, the fluorescent nucleotide remained stably bound to Gsp1p (Fig. 2, A and B). The Mog1p mutants were also assayed for their ability to bind Gsp1p and stimulate nucleotide release. Compared with wild-type Mog1p, significantly reduced levels of complex formation and nucleotide release were observed when the Mog1p mutants were incubated with either nucleotide state of Gsp1p (Fig. 2, A and B).
In Vivo Analysis of Mog1p Mutants-Because E65K-Mog1p displayed a more dramatic decrease in Ran binding than D62K-Mog1p in vitro, we chose to examine the effects of the E65K mutation in S. cerevisiae. To assess the in vivo function of E65K-Mog1p, we integrated E65K-Mog1 at the endogenous MOG1 locus making E65K-Mog1p the only form of Mog1p expressed in these cells. Cells lacking MOG1 are temperature sensitive for growth at 37°C (9). To determine if the E65K-Mog1p mutant also conferred a temperature-sensitive growth phenotype, we compared the growth of E65K-Mog1 cells to wild-type and ⌬mog1 cells (Fig. 3). We found that E65K-Mog1 cells grew more slowly than wild-type cells at 37°C. The phenotype observed was not as profound as in cells lacking Mog1p (Fig. 3B), suggesting that either the Mog1p-Gsp1p interaction was not completely disrupted by the E65K-Mog1p mutation in vivo or that Mog1p performs additional functions in the cell that do not involve its interaction with Gsp1p. Mog1p levels in wild-type and E65K-Mog1 cells were comparable, confirming that the growth phenotype observed for E65K-Mog1 was not due to a decreased level of E65K-Mog1p (data not shown).
Identification of a Putative Mog1-binding Site on Ran-Since Gsp1p-GDP binding by Mog1p and Ntf2p is mutually exclusive, the two proteins may have a common or at least overlapping binding site on Ran (10). The NTF2-Ran interaction is mediated by residues in the switch I and II loops of Ran (14,31,32), the regions of Ras family GTPases that change conformation in response to nucleotide state (33,34). To determine if the same residues were involved in the Mog1-Ran interaction, we tested the Ran switch I mutant T42A and switch II mutants Q69L, F72W, and R76E for binding to His-tagged human Mog1. Con-  In reactions with wild-type Mog1p and Gsp1p in either nucleotide state, the major peak of protein absorbance elutes at fraction 20 with an apparent molecular mass of 50 kDa, which corresponds to a 1:1 complex between Mog1p and Gsp1p. Minor peaks of absorbance at 280 nm are also observed, corresponding to free protein monomers and free nucleotide. No fluorescence emission is associated with fractions containing the Mog1p-Gsp1p complex. Instead, the main peak of fluorescence elutes in fractions containing free nucleotide, with a second minor peak co-eluting with the Gsp1p monomers. The protein absorbance and fluorescence emission profiles for reactions between Gsp1p-mantGDP and the D62K-Mog1p or E65K-Mog1p mutants are similar to those of the control reaction without Mog1p. Complex formation and nucleotide release is observed for reactions between Gsp1p-mantGTP and the Mog1p mutants, but at reduced levels compared with reactions with wild-type Mog1p. sistent with the yeast Mog1p binding assays (see Fig. 1B), the amount of Ran-GDP bound to human Mog1 was lower than observed with Ran-GTP (Fig. 4A). Importantly, the Ran switch loop mutants in both nucleotide states displayed wild-type levels of binding to Mog1, with the possible exception of Q69L-GDP, which showed a slightly lower level of binding (Fig. 4A). In contrast, the NTF2 control displayed significantly reduced binding to the switch II mutants Q69L, F72W, and R76E compared with wild-type Ran-GDP (Fig. 4A). These results indicate that the switch loops do not play a major role in the formation of a Mog1-Ran complex.
To obtain Ran mutants with decreased affinity for Mog1, we performed a mutagenesis screen to identify charged surface residues on Gsp1p that are important for the interaction with Mog1p. Gsp1p mutants (R97E, K101E, R108E, R112E, E115K, K129E, E130K, R131E, K132E, K134E, and K136E) were assayed for binding to Mog1p-Sepharose and, as a control, to Ntf2p-Sepharose. Mutation of Lys 136 , which is located ϳ25-30 Å away from the switch loop region of Ran (Fig. 4B) reduced Mog1p binding without decreasing the interaction with Ntf2p (see below). Mutations at two other residues, Arg 108 and Arg 112 , close to the NTF2-binding site on Ran (Fig. 4B) disrupted Ntf2p binding without reducing Mog1p binding (data not shown).
In Vitro and In Vivo Analysis of K136E-Gsp1p-Mog1p was assayed for its ability to interact with His-tagged K136E-Gsp1p in either nucleotide state. Compared with wild-type Gsp1p, only a weak residual interaction with K136E-Gsp1p was observed for wild-type Mog1p (Fig. 5A, lanes 1-2 and 7-8). A similar low level of binding to K136E-Gsp1p was detected for the D62K-Mog1p and E65K-Mog1p mutants, even when pres-ent in 5-fold excess over the level of wild-type Mog1p (Fig. 5A,  lanes 3-4 and 9 -10). In contrast, neither Ntf2p nor Kap95p binding to K136E-Gsp1p was decreased compared with wildtype Gsp1p (Fig. 5A, lanes 5-6 and 11-12) indicating that the K136E mutation did not introduce a major conformational change. The effect of the K136E substitution on the strength of the Mog1-Ran interaction in yeast was also examined. Unlike wild-type Gsp1p, which co-purified with GST-tagged Mog1p, we were unable to detect any K136E-Gsp1p associated with GST-tagged Mog1p (Fig. 5B).
To analyze the function of K136E-Gsp1p in vivo, we expressed this mutant as the only source of Gsp1p in yeast. The growth of wild-type, ⌬mog1, and K136E-Gsp1 cells was monitored by A 600 (Fig. 5C). This analysis showed that K136E-Gsp1 cells were viable, indicating that K136E-Gsp1p can functionally replace wild-type Gsp1p. However, the growth of K136E-Gsp1 was slower than wild-type cells and similar to ⌬mog1 The growth of E65K-Mog1 cells was compared with wild-type (E65K-Mog1 ϩ pMOG1) and ⌬mog1 cells at 37°C. A, cells were grown to saturation, serial diluted (1:10) and spotted onto plates. B, saturated cultures were diluted to 5 ϫ 10 6 cells/ml before shifting to 37°C. Growth was monitored at A 600 at 2-h intervals. Cells expressing E65K-Mog1p (OE) grow more slowly than wild-type cells (q). Growth of ⌬mog1 (f) is also shown.

FIG. 4. Identification of a putative Mog1-binding site on Ran.
A, His-tagged human Mog1 immobilized on Ni-NTA-agarose was assayed for binding to wild-type vertebrate Ran, T42A, Q69L, F72W, or R76E charged with GDP or GTP. As a control, GDP-bound wild-type Ran and mutants were also assayed for binding to NTF2-Sepharose beads. B, schematic representation of the Ran-GDP structure showing the positions of residues targeted for mutagenesis. Lys 134 of Ran (red), located on helix ␣4 (dark pink) is equivalent to Lys 136 in Gsp1p, which is required for efficient Mog1p binding. Mutation of two residues (green) decrease binding of Gsp1p to Ntf2p without affecting the Mog1-Ran interaction. The positions of the switch I loop (yellow), switch II loop (orange), and guanine nucleotide (indigo) are indicated. Conserved residues Lys 123 and Asp 125 (violet) stabilize binding of the guanine base. The COOH-terminal helix ␣6 (light pink) packs against helix ␣4 in the Ran-GDP structure. cells (Fig. 5C), consistent with a loss of the Mog1p-Gsp1p interaction. Expression levels of K136E-Gsp1p were comparable to those for wild-type Gsp1p, indicating that the phenotype observed was not due to decreased expression of the mutant protein (data not shown).
E65K-Mog1 and K136E-Gsp1 Are Defective in Nuclear Protein Import-Cells lacking Mog1p show defects in nuclear protein import (9). To determine if the E65K-Mog1p and K136E-Gsp1p mutations cause protein import defects, we utilized a NLS-GAL4AD-GFP reporter protein (29). In wild-type cells, the NLS-GFP reporter protein was localized to the nucleus (Fig. 6) whereas in ⌬mog1 cells it was mislocalized to the cytoplasm, consistent with a previous report (9). Both E65K-Mog1 and K136E-Gsp1 cells also displayed defects in nuclear protein import (Fig. 6). The cytoplasmic mislocalization of the NLS-GFP reporter protein was more pronounced for K136E-Gsp1 than for ⌬mog1 and E65K-Mog1 (Fig. 6), which may indicate that other functions of Gsp1p, in addition to Mog1p binding, were affected by the K136E-Gsp1p mutation.
Genetic Interaction between MOG1 and PRP20 -Mog1p binds tightly to the nucleotide-free form of Gsp1p (11,13). The only other protein that binds to the nucleotide-free form of Ran is the RanGEF, RCC1 (2) in vertebrates or Prp20p (35) in yeast. As a genetic test for functional overlap between MOG1 and PRP20, we tested for any synthetic growth defects in a double mutant. Since PRP20 is essential, we used a well characterized temperature-sensitive allele, prp20-1 (36,37), which shows reduced binding to Gsp1p (38), suggesting that nucleotide exchange by this mutant protein is also reduced. If Mog1p is involved in modulating the nucleotide state of Gsp1p within the nucleus, as has been demonstrated for the mammalian exchange factor, RCC1 (2, 3), cells with mutations in both MOG1 and PRP20 might be more growth compromised than either of the single mutants. We therefore deleted MOG1 in the prp20-1 mutant to generate the prp20-1⌬mog1 double mutant. The prp20-1⌬mog1 double mutant was unable to grow under conditions that supported the growth of each single mutant (Fig.  7), demonstrating a genetic interaction between MOG1 and PRP20.  11 and 12). The D62K-Mog1p and E65K-Mog1p mutants were present at 5-fold molar excess compared with wild-type Mog1p. B, yeast expressing myc-Gsp1p or myc-K136E-Gsp1p as the only functional copy of Gsp1p were transformed with plasmids encoding GST or GST-Mog1p. Binding assays were carried out as described under "Experimental Procedures." An immunoblot probed with an anti-Myc antibody to detect Myc-tagged Gsp1p is shown. Lanes are designated lysate (L), unbound (U), or bound (B). The expression of wild-type Gsp1p or K136E-Gsp1p was similar and the amount of GST fusion protein in the bound (B) sample was approximately equal for each sample (data not shown). The negative control GST alone does not interact with Gsp1p or K136E-Gsp1p. C, wild-type Gsp1p, K136E-Gsp1p, or an empty vector were expressed in ⌬gsp1⌬gsp2 yeast cells that were maintained by a wild-type Gsp1p URA plasmid (see "Experimental Procedures"). K136E-Gsp1p function was analyzed by streaking the transformants on 5-fluoroorotic acid to remove the wild-type GSP1 plasmid. The growth of cells expressing K136E-Gsp1p (OE) was compared with wild-type (q), and ⌬mog1 (s) at 37°C by measuring A 600 at 2-h intervals as described under "Experimental Procedures."

DISCUSSION
The Ran-binding protein, Mog1, has emerged as a novel factor involved in regulating the nucleotide state of Ran. Previous reports showed that Mog1 is a guanine nucleotide release factor for Ran-GTP (11,13) and we have demonstrated that Mog1 also stimulates the release of nucleotide from Ran-GDP. Thus, an interaction between Mog1 and either nucleotide state of Ran results in nucleotide release and formation of a stable complex between Mog1 and nucleotide-free Ran. The RanGEF, RCC1, can also catalyze nucleotide release from either nucleotide state of Ran (30) but unlike RCC1, Mog1 does not promote rebinding of nucleotide (11). Therefore, Mog1 binding may cause Ran to adopt a unique conformation that releases the nucleotide and stabilizes the molecule in this nucleotide-free state. We have used structure-based mutagenesis of Mog1 and Ran to identify residues required for formation of this complex and to assess the importance of the Mog1-Ran interaction in vivo.
From the crystal structure of S. cerevisiae Mog1p, we predicted that a highly conserved cluster of surface residues might constitute a binding site for Ran. Mutagenesis of this conserved surface patch revealed that residues Asp 62 and Glu 65 are important for the Mog1-Ran interaction. Replacement of wildtype Mog1p in yeast with the E65K-Mog1p mutant caused temperature-sensitive growth and defects in nuclear protein import, similar to the phenotype observed for ⌬mog1 cells. These results suggest that a primary function of Mog1p in the cell relies on its ability to bind Ran.
The switch loops of Ran (involved in the NTF2-Ran interaction) were possible candidates for a Mog1-binding site because Mog1p and Ntf2p cannot bind Gsp1p simultaneously (10). However, several switch I and II loop mutants retained the ability to bind Mog1, suggesting that this region of Ran is not involved directly in the interaction with Mog1. Although reduced binding to Mog1 was previously reported for the switch II mutant, Q69L-Ran-GTP (11), we found that Mog1 had a wild-type level of affinity for Q69L-Ran-GTP. Because the Q69L-Ran mutant is prone to denaturation, we always carried out our assays using freshly prepared protein.
In a mutagenesis screen of Gsp1p charged surface residues, we observed reduced Mog1p binding to the K136E-Gsp1p mutant. The conformation of Gsp1p was not likely to have been perturbed by this mutation because wild-type levels of binding to both Ntf2p and Kap95p were observed for K136E-Gsp1p. Like ⌬mog1 and E65K-Mog1, the K136E-Gsp1 mutant showed protein import defects, consistent with Lys 136 being involved in the Mog1-Ran interaction. These observations support the hy-pothesis that the Mog1-Ran interaction is required for efficient nuclear protein import.
Lys 136 in Gsp1p corresponds to Lys 134 in vertebrate Ran, which is located on helix ␣4 in the Ran-GDP structure (39). Because the COOH-terminal helix ␣6 is juxtaposed against helix ␣4 (Fig. 4B), Mog1 would likely have to displace the COOH-terminal helix to bind and release nucleotide from Ran-GDP. This might explain the higher affinity of Mog1 for Ran-GTP compared with Ran-GDP, since the COOH-terminal helix is already displaced from this region in Ran-GTP (40 -42). Consistent with this hypothesis, deletion of the COOH-terminal DEDDDL motif, which stabilizes the conformation of helix ␣6 in Ran-GDP (40), increases the affinity of Mog1 for Ran (13). Deletion of the Ran-DEDDDL motif also stimulates the formation of complexes between importin-␤ and Ran (43,44) and between RCC1 and Ran (45). Since both importin-␤ and RCC1 make contacts with Lys 134 and other residues in the helix ␣4 region of Ran (41,42,46), the binding sites for importin-␤ and RCC1 on Ran partially overlap with the putative Mog1-binding site. Overlapping binding sites for Mog1 and importin-␤ on Ran would explain our inability to detect a trimeric complex between Mog1, Ran, and importin-␤. 2 Although the Mog1-and NTF2-binding sites on Ran do not appear to overlap, simultaneous binding is likely prevented because the two molecules bind different conformational states of Ran. NTF2 binds Ran in the GDP-bound state in which the COOH-terminal helix is stabilized against the core of the molecule, whereas NTF2 is prevented from binding Ran-GTP by a steric clash with the COOH-terminal linker in its displaced conformation (40). Conversely, our model suggests that Mog1 binding to Ran requires displacement of the COOH terminus, which may explain why Mog1 and NTF2 binding to Ran is mutually exclusive.
The putative Mog1-binding site in the Lys 134 region is adjacent to a loop containing residues Lys 123 and Asp 125 which are part of the highly conserved NKxD nucleotide-binding motif (Fig. 4B). These residues interact with the guanine base in both nucleotide states of Ran (39,40). Therefore, Mog1 binding to helix ␣4 may induce a conformational change in this loop that disrupts binding of Lys 123 and Asp 125 to the guanine base promoting nucleotide release from Ran-GDP or Ran-GTP. Like Mog1, RCC1 also interacts with Ran residues in helix ␣4 and large conformational changes are observed for helix ␣4 and the NKxD base-binding motif of Ran in the RCC1-Ran structure (46).
In summary, we have defined residues on Mog1 and Ran important for their interaction in vitro and in vivo and have demonstrated the importance of the Mog1-Ran interaction in nuclear protein import. We have also shown that Mog1 binds and stimulates nucleotide release from both Ran-GDP and Ran-GTP in vitro, suggesting that the function of Mog1 in nuclear protein import may involve its modulation of the nucleotide state of Ran. The genetic interaction we detected between MOG1 and PRP20, the yeast RanGEF, raises the possibility that these two proteins may act synergistically to generate Ran-GTP in the nucleus. FIG. 7. ⌬mog1 and prp20-1 mutants are synthetically lethal. The ⌬mog1 and prp20-1 mutants were combined by deleting the MOG1 gene in a prp20-1 background as described under "Experimental Procedures." The double mutant cells were maintained by a PRP20 plasmid. The growth of prp20-1⌬mog1 was compared when transformed with both PRP20 and MOG1 (wild-type), MOG1 (prp20-1), PRP20 (⌬mog1), or empty vector (prp20-1⌬mog1). For analysis of the double mutant, the wild-type PRP20 maintenance plasmid was expelled using 5-fluoroorotic acid. Cells were grown to saturation in synthetic media before serial dilution (1:10) and spotting. The plate was incubated at 25°C for 5 days. The wild-type, ⌬mog1, and prp20-1 cells all grow at similar rates. prp20-1⌬mog1 cells are unable to grow, demonstrating a synthetic lethal interaction between these two mutations.