Oligomerization Regulates the Localization of Cdc24, the Cdc42 Activator in Saccharomyces cerevisiae*

Guanine nucleotide exchange factor activation of Rho G-proteins is critical for cytoskeletal reorganization. In the yeast Saccharomyces cerevisiae, the sole guanine nucleotide exchange factor for the Rho G-protein Cdc42p, Cdc24p, is essential for its site-specific activation. Several mammalian exchange factors have been shown to oligomerize; however, the function of this homotypic interaction is unclear. Here we show that Cdc24p forms oligomers in yeast via its catalytic Dbl homology domain. Mutation of residues critical for Cdc24p oligomerization also perturbs the localization of this exchange factor yet does not alter its catalytic activity in vitro. Chemically induced oligomerization of one of these oligomerization-defective mutants partially restored its localization to the bud tip and nucleus. Furthermore, chemically induced oligomerization of wild-type Cdc24p does not affect in vitro exchange factor activity, yet it results in a decrease of activated Cdc42p in vivo and the presence of Cdc24p in the nucleus at all cell cycle stages. Together, our results suggest that Cdc24p oligomerization regulates Cdc42p activation via its localization.

Cdc24p localization varies throughout the cell cycle and in response to mating pheromone (4 -6). This GEF localizes to sites of polarized growth, i.e. at the incipient bud site during G 1 phase, the bud tip at S and G 2 /M phases, and the motherdaughter bud neck during M phase. In addition, Cdc24p shuttles between these sites of growth and the nucleus, and this exchange factor accumulates in the nucleus during late M phase and early G 1 phase via an association with the cyclin-dependent kinase inhibitor Far1p. It is likely that Cdc24p, which is sequestered in the nucleus, is unable to activate Cdc42p (4,5). Conversely, Cdc24p, which is exported from the nucleus upon entry into the cell cycle, or upon exposure to mating pheromone, is recruited to the sites where Cdc42p is present and activates this G-protein. Cdc24p phosphorylation also varies throughout the cell cycle with phosphorylated forms accumulating at bud emergence (4,5). Phosphorylation is thought to trigger the release of Cdc24p from the Bem1 scaffold protein (7); however, this proposal is controversial (8), and the functional significance of Cdc24p phosphorylation remains unclear.
A number of Dbl family GEFs, including Cdc24p and Dbl, have been shown to self-associate, either via intra-or intermolecular interactions (18,(22)(23)(24)(25)(26)(27). For example, for the DH domain, amino-terminal and carboxyl-terminal regions can bind one another and inhibit in vitro or in vivo activity (18,23,24). These associations can occur intermolecularly, yet have been suggested to occur intramolecularly. Homotypic intermolecular interactions appear to be an additional regulatory mechanism for mammalian GEFs and occur via different domains, depending on the GEF. For example, Dbl oligomerizes via its DH domain (27), whereas LARG (leukemia-associated RhoGEF) and p115 RhoGEF oligomerize via their carboxyl-terminal regions (25). This has raised the question whether in yeast Cdc24p oligomerizes and if so what is its regulatory function.
In this study, we show that Cdc24p oligomerizes in vivo and that the DH domain is necessary and sufficient for this homotypic interaction. Analyses of site-specific mutants suggest that the conserved region 2 (CR2) of the DH domain is necessary for Cdc24p oligomerization and localization to the bud tip and nucleus. In addition, we show that the level of Cdc24p oligomers is unaffected by G 1 cell cycle arrest, wherein the majority of Cdc24p accumulates in the nucleus. Furthermore, chemically induced Cdc24p oligomerization results in a decrease in activated Cdc42p and a delay in nuclear export during the cell cycle in vivo, yet does not affect guanine nucleotide exchange factor activity in vitro. Taken together our results indicate that Cdc24p oligomerization is important for GEF localization and that oligomer dissociation is required for efficient nuclear export, suggesting that GEF homo-oligomerization via its DH domain may provide a mechanism to down-regulate exchange factor activity in vivo.

EXPERIMENTAL PROCEDURES
Strains and Plasmids-Standard techniques and media were used for yeast growth and genetic manipulation (28). Strains used in this study are described in Table 1. A 3xHA epitopetagged Cdc24p fusion was constructed using an AgeI-linearized pDM1-Cdc24-3xHA gene replacement cassette in either RAY819 or RAY1599. pDM1-Cdc24-3xHA (a gift from D. McCusker) was constructed by cloning a BamHI/SalI fragment from pGEX4TCDC24 that contained the 382 carboxyl-terminal amino acids of Cdc24p into YIplac211-3xHA resulting in the carboxyl-terminal portion of Cdc24p fused to 3xHA followed by a URA3. Cdc24 mutants and FKBP fusions were transformed in RAY950, which were then grown in the presence of glucose to repress GalHis 6 Cdc24, and the loss of this pRS416 URA3 plasmid was confirmed by its inability to grow on media lacking uracil and by PCR.
DH domain point mutations were generated by site-directed mutagenesis and the addition of a silent restriction site (Table 3, primers 9 -22). All mutations generated were confirmed by sequencing (ABI PRISM big-dye terminator cycle sequencing kit). For Cdc24 DH mutant localization, DH domain mutations DH3, DH5, and DH8 were subcloned using KpnI/SalI into p414Cdc24mycGFP, and similarly the DH3 mutation was subcloned into p414Cdc24mycCFKBPmGFP.
For amino-terminal FKBP fusions, two unique restriction sites (BlpI and SphI) were inserted 5Ј of the CDC24 ATG codon  (31), yielding p414Cdc24HACmGFP and p414Cdc24mycCmGFP. Subsequently, FKBP was PCR-amplified from pC4Mfv2E using primer pairs 32 and 35, and the ClaI-digested PCR product was cloned into appropriately digested p414Cdc24HACmGFP and p414-Cdc24mycCmGFP, yielding p414Cdc24HACFKBPmGFP and p414Cdc24mycCFKBPmGFP. All mutations generated were confirmed by sequencing (ABI PRISM big-dye terminator cycle sequencing kit). Sequence Alignment and Structure Modeling-Sequence alignments were carried out using the BLAST algorithm (32). Structural model threading was carried out using a web-based Modeler program (33,34) with the Tiam1, Rac1 Protein Data Bank file 1FOE (35).
Protein Binding Assays-Pulldowns were carried out as described previously (11) with the following modifications. Yeast cells (ϳ2 ϫ 10 8 ) from logarithmically growing cultures were harvested by centrifugation and lysed by agitation with glass beads (Sigma) in buffer A (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.1% Triton X-100, 1 mM PMSF, and protease inhibitor mixture; Roche Applied Science) with a Hybaid Ribolyser. Cell extracts were clarified by two centrifugations (10,000 ϫ g for 10 min). Supernatants, which contained the majority of the tagged proteins, were incubated with anti-HA resin (either from 12CA5 hybridoma supernatant and protein A-Sepharose (GE Healthcare) cross-linked with dimethyl pimelimidate (Sigma) or anti-HA resin from Vector Laboratories) in buffer A for 3 h. Resin was washed five times in buffer A and analyzed by SDS-PAGE, transferred to a nitrocellulose membrane (Amersham Biosciences), and probed with mouse monoclonal antibody HA11 against HA (1:1000 dilution; Covance) or mouse monoclonal antibody 9E10 against myc (1:1000 dilution; Covance). Immunoblots were visualized by enhanced chemiluminescence (luminol-coumaric acid) on a Fuji-Las3000. Equal amounts of cells were used in each experiment, and equal amounts of protein from the 10,000 ϫ g supernatants was confirmed by Ponceau S staining of nitrocellulose membranes. Pulldowns of cells expressing Cdc24FKBP fusions were carried out as described above except that extracts were prepared with 0.5ϫ yeast buster reagent (Calbiochem) following the manufacturer's instructions. For cells grown in the presence of AP20187 (1 M), all subsequent steps were carried out with buffers containing 0.1 M AP20187. To examine the effect of induced oligomerization in vitro, cell extract 10,000 ϫ g supernatants were treated with 0.1 M AP20187 for 15 min at 4°C prior to pulldowns.
In vitro coupled transcription-translation was carried out using a Quick TNT rabbit reticulocyte lysate system (Promega) and 5 Ci of [ 35 S]methionine (ICN 1175 Ci/mmol) using either pT7CDC24 (36) or T7-luciferase, following the manufacturer's instructions. GST and MBP fusion proteins were expressed in E. coli BL21 cells grown at 30°C and induced with isopropyl 1-thio-␤-D-galactopyranoside for 4 h. Cells expressing GST or MBP fusions were lysed in buffer B (phosphate-buffered saline, protease inhibitor mixture (Roche Applied Science), 1 mM PMSF, 0.1% Triton X-100, and 1 mM EDTA) by sonication followed by snap freezing and thawing. Extracts were clarified by two centrifugations (10,000 ϫ g for 10 min), and fusion proteins were isolated using glutathione-Sepharose 4B resin (Amersham Biosciences) or amylose resin (New England Biolabs). Protein concentrations were estimated by comparing intensities of bands on Coomassie-stained SDS-polyacrylamide gels with bovine serum albumin (Sigma) as a standard. GSTCdc24 resin was washed with the same buffer containing 250 mM and then 500 mM NaCl. GSTCdc24 was eluted with buffer C (10 mM Tris-Cl, pH 8, 10 mM glutathione). After extensive dialysis using a Centricon YM30 (Amicon), protein-containing fractions were rebound to the respective resin.
For [ 35 S]Cdc24 binding to GSTCdc24 or GST GSH-Sepharose resin-bound proteins (ϳ200 ng of protein) were incubated with radiolabeled protein in buffer D (50 mM HEPES, pH 7.6, 50 mM NaCl, 5 mM MgCl 2 , 10% glycerol, 0.1% Triton X-100) for 3 h at 4°C. Resin samples were then washed five times with 100 ml of buffer E (50 mM Tris-Cl, pH 8, 150 mM NaCl, 1 mM PMSF, and protease inhibitor mixture; Roche Applied Science) at 4°C. Proteins were eluted with SDS-PAGE sample buffer and analyzed by SDS-PAGE followed by Coomassie Blue staining and autoradiography. A Fuji BAS1000 PhosphorImager was used for quantification based upon two input amounts.
For Cdc24DHPH binding to Cdc24DH expressed in yeast, supernatants (10,000 ϫ g) of extracts from yeast cells (ϳ2 ϫ 10 8 ) expressing 3xmycCdc24DH-3xmyc in buffer A were incubated with either MBP or MBPCdc24DHPH (ϳ200 ng of protein) bound to amylose resin for 2 h at 4°C. For binding of bacterially expressed His 6 Flag2xmyc-Cdc24DH to GSTCdc24DHPH and GSTCdc24DH, 10,000 ϫ g supernatants of bacterial cells expressing His 6 FLAG2xmyc-Cdc24DH in buffer F (phosphate-buffered saline, protease inhibitor mixture (Roche Applied Science), 1 mM PMSF, and 0.1% Triton X-100) were incubated with GST, GSTCdc24DHPH, or GSTCdc24DH bound to GSH-agarose (ϳ200 ng of protein) for 2 h at 4°C. Resin samples were then washed five times with 10 volumes of the respective buffer. Proteins were eluted with SDS-PAGE sample buffer, analyzed by SDS-PAGE, transferred to nitrocellulose, probed with mouse monoclonal antibody 9E10 against myc (1:1000 dilution; Covance) for experiments with 3xmycCdc24DH-3xmyc or rabbit polyclonal sera against the FLAG epitope (1:1000; Santa Cruz) for experiments with His 6 Flag2xmyc-Cdc24DH and visualized by enhanced chemiluminescence (luminol-coumaric acid) on a Fuji-Las3000. Equal amounts of cells were used in each experiment, and equal amounts of protein from the 10,000 ϫ g supernatants was confirmed by Ponceau S staining of nitrocellulose membranes.
GSTCRIB pulldown experiments were carried out essentially as described (37) with GSTCRIB-GSH-agarose (ϳ400 ng of protein). Cell extract 10,000 ϫ g supernatants were also incubated with 0.1 M AP20187 at 4°C for 1 h prior to incubation with GST-CRIB resin. Cell extract 10,000 ϫ g supernatants were loaded with GTP␥S to determine the maximum amount of Cdc42⅐GTP. EDTA was added (20 mM final concentration) to the 10,000 ϫ g supernatants, and subsequently GTP␥S (190 M final concentration) was added, and the mixture was incubated at room temperature for 15 min. MgCl 2 was then added (107 mM final concentration), and the samples were chilled to 4°C. Proteins were eluted from GSTCRIB resin in SDS-PAGE sample buffer, analyzed by SDS-PAGE, transferred to nitrocellulose, probed with rabbit anti-Cdc42 (36) (1:1000 dilution), and visualized by enhanced chemiluminescence (luminol-coumaric acid) on a Fuji-Las3000. Equal amounts of cells were used in each experiment, and equal amounts of protein from 10,000 ϫ g supernatants was confirmed by Ponceau S staining of nitrocellulose membranes.
Guanine Nucleotide Exchange Factor Assays-GST and MBP fusion proteins were expressed in E. coli BL21 cells grown at 30°C and induced with isopropyl 1-thio-␤-D-galactopyranoside for 4 h. Cells expressing MBP-Cdc42 and GST fusions were lysed in buffer G (20 mM Tris-Cl, pH 7.5, 100 mM NaCl, 2.5 mM MgCl 2 , 1 mM DTT, 10 M GDP, and 1 mM PMSF) (38) and buffer H (50 mM Tris-Cl, pH 7.5, 100 mM NaCl, 2.5 mM MgCl 2 , 1 mM DTT, 1 mM PMSF, 10% glycerol, and Boehringer protease inhibitor mixture), respectively, by sonication followed by snap freezing and thawing. Extracts were clarified by centrifugation (4,000 ϫ g for 10 min), and fusion proteins were isolated using amylose resin or glutathione-Sepharose 4B resin. MBP-Cdc42 resin was washed extensively with buffer G and MBP-Cdc42 was eluted with same buffer G containing 10 mM maltose. GST-DHPH and GST-DHPH-FKBP resins were washed extensively in buffer H and successively with buffer H containing 250 mM potassium acetate and 500 mM potassium acetate. After an additional wash in buffer H, fusion proteins were eluted with buffer I (50 mM Tris-Cl, pH 8, 100 mM NaCl, 2.5 mM MgCl 2 , 1 mM DTT, 1 mM PMSF, 10% glycerol, and 10 mM glutathione). Protein concentrations were estimated by comparing intensities of bands on Coomassie-stained SDS-polyacrylamide gels and by Bradford protein assays with bovine serum albumin as a standard (39). To remove GST, DHPH and DHPH-FKBP fusions were incubated with PreScission protease (Amersham Biosciences, 20 units) for 24 h at 24°C. Subsequently, proteins were concentrated using a Centricon YM30 (Millipore).
Guanine nucleotide exchange factor assays were carried out in a 25-l volume with 18 l of buffer L1 (25 mM Tris-Cl, pH 7.5, 200 mM (NH4) 2 SO 4 , 5 mM MgCl 2 , 2 mM DTT, 1 mM EDTA, 5 M GTP, Boehringer protease inhibitor mixture), which con-  A, supernatants (10,000 ϫ g) from yeast expressing indicated epitope-tagged Cdc24p were either analyzed by SDS-PAGE (Input, 2% anti-HA and 1% anti-myc blot) directly or following anti-HA pulldown (upper panel). Immunoblots were probed with the indicated sera. Supernatants (10,000 ϫ g) from yeast expressing HA epitopetagged Cdc24p and myc-tagged Sph1p were either analyzed by SDS-PAGE (Input, 2%) directly or following anti-HA pulldown and probing with anti-myc sera (lower panel). B, partially purified GSTCdc24 or GST from E. coli was used for binding experiments with [ 35 S]luciferase (Luc) or [ 35 S]Cdc24 (C24) generated in coupled transcription-translation rabbit reticulocyte lysates. Samples were analyzed by SDS-PAGE stained with Coomassie Blue (upper panel), and radioactivity was visualized by PhosphorImager (lower panel). The first three lanes from the left are 10, 20, and 5% of indicated radioactive protein input, respectively. Samples were then filtered on prewetted 0.45-m HA filters (Millipore) that were then washed with five times with 2 ml of buffer L3 (25 mM Tris-Cl, pH 7.5, 5 mM MgCl 2 , 2 mM DTT, 1 mM EDTA, and 10 M GTP). Filters were dried, and radioactivity was counted with Biofluor scintillation fluid (PerkinElmer Life Sciences) in a Beckman LS 6500 scintillation counter.
Localization of Cdc24p-Strains with Cdc24HAGFP, Cdc24HAFKBPGFP, or Cdc24mycFKBPGFP as the sole copy or wild-type strains with Cdc24mycGFP, Cdc24-DH3mycGFP, Cdc24-DH5mycGFP, or Cdc24-DH8mycGFP were imaged on a Zeiss LSM510 meta inverted confocal microscope using a Plan-Apo 63ϫ NA 1.4 objective and 488-nm LASER excitation as described (4). Both differential interference contrast and fluorescence images of random field of views were captured, and cells with GFP fluorescence localized to the bud, motherdaughter bud neck, or nucleus were counted. Based upon differential interference contrast images, cells with GFP localized fluorescence were counted in the following categories: stage 1a, unbudded; stage 1b, very small budded (Ͻ1/8 diameter of mother cell); stage 2, small budded (Յ 1 ⁄ 2 radius of mother cell); and stage 3, dividing cells with mother and daughter cells equal size.

RESULTS
Cdc24p Forms Oligomers-To determine whether Cdc24p binds Cdc24p in vivo, we used a strain that expressed two different epitope-tagged Cdc24 proteins, Cdc24-3xHAp and 3xmyc-Cdc24p. As a control we used a strain in which only triple myc-tagged Cdc24p was expressed. Both tagged versions of Cdc24p were expressed via the CDC24 promoter with the HA fusion replacing the genomic copy and the myc fusion on a centromeric plasmid. The HA-tagged Cdc24p was pulled down using an anti-HA antibody, and bound 3xmyc-Cdc24p was detected by immunoblot probed with an anti-myc antibody. As illustrated in Fig. 1A, 3xmyc-Cdc24p bound Cdc24-3xHAp. We also observed lower molecular weight bands in the ␣-myc probed blot that were likely to be 3xmyc-Cdc24p degradation products, as they were only observed in strains expressing 3xmyc-Cdc24p. Their apparent sizes (ϳ80, ϳ65, and ϳ60 kDa) indicated that these fragments contain the DH domain. These degradation fragments bound Cdc24-3xHAp suggesting that the carboxyl terminus of Cdc24p is not necessary for this binding. We examined if this binding was specific using as a control another yeast protein Sph1-mycp. Fig. 1A shows that Sph1mycp does not bind Cdc24p, confirming the specificity of the Cdc24p-Cdc24p interaction. An average of 3 Ϯ 1% (n ϭ 3) of the total 3xmyc-Cdc24p bound to Cdc24-3xHAp; however, this value is an underestimate as each epitope-tagged protein can associate with itself or the other tagged fusion. This level of oligomers is consistent with the notion that Cdc24p oligomers are dynamic. To determine whether this association reflects direct binding as opposed to a protein complex containing many Cdc24p molecules, we examined whether Cdc24p expressed in bacteria could bind to rabbit reticulocyte translated [ 35 S]Cdc24p. Immobilized GST-Cdc24, but not GST alone can bind [ 35 S]Cdc24p (Fig. 1B). This interaction was specific as immobilized GST-Cdc24 did not bind [ 35 S]luciferase. Together these data establish that Cdc24p oligomerizes.
The Cdc24p Dbl Homology Domain Is Necessary and Sufficient for Oligomerization-To determine the region of Cdc24p that is required for oligomerization, we generated amino-terminal and carboxyl-terminal deletions of 3xmyc-Cdc24p ( Fig.  2A) and examined pulldowns with Cdc24-3xHAp. Deletion of the amino-terminal 280 amino acids of Cdc24p, including the CH domain, did not abolish Cdc24p oligomerization (Fig. 2B). Similarly, deletion of the PH domain in the full-length Cdc24p or in a construct lacking the amino-terminal 280 amino acids had little to no effect on Cdc24p binding (Fig. 2, B and D). Indeed the amino-terminal 470 amino acids of Cdc24p, which include the DH domain, bound to Cdc24p, indicating that the PB1 domain at the carboxyl terminus is not required for oligomerization (Fig. 2C). The DH domain is the only region, which is common to these constructs. Cdc24p lacking the DH domain did not bind Cdc24p, and similarly a carboxyl-terminal 375 amino acid fragment lacking the CH and DH domains did not bind Cdc24p (Fig. 2, E and F). These results also confirm the specificity of the Cdc24p-Cdc24p interaction. Furthermore, the DH domain was sufficient to bind Cdc24-3xHAp in yeast (Fig.  2F), suggesting that the DH domain alone is sufficient for oligomerization. Consistent with this result, the DH domain expressed in yeast bound an MBP-DHPH domain fusion (Fig.  3A). The DH domain expressed in bacteria bound both GST-DH and GST-DHPH domain fusions (Fig. 3B), with the binding of the latter being somewhat reduced. Together these results indicate that the DH domain is necessary and sufficient for Cdc24p oligomerization.
We next examined if conserved residues required for catalytic activity and G-protein binding are necessary for exchange factor oligomerization. The Cdc24p DH domain, modeled onto the structure of the Tiam1 DH domain, has a flattened elongated helical bundle with six major helical axes (Fig. 4A) (35,40) similar to the Dbs DH domain (40). DH domains are characterized by three conserved regions (CR1-3), which range in length A, supernatant (10,000 ϫ g) from yeast expressing Cdc24 3xmycDH-3xmyc was either analyzed by SDS-PAGE (Input, 1%) directly or following incubation with amylose resin-bound MBP or MBPDHPH (Pull down). Immunoblots were probed with anti-myc sera. B, supernatants (10,000 ϫ g) from bacteria expressing Cdc24 His 6 Flag2xmycDH were either analyzed by SDS-PAGE (Input) directly or following incubation with indicated GST-agarose (Pull down). Immunoblots were probed with anti-FLAG sera. from 21 to 28 amino acids (Fig. 4B). CR1, which is composed principally of ␣-helix 1 and CR3, which includes the ␣-helix 2 together with the carboxyl-terminal end of ␣-helix 3 and ␣-helix 6, forms an exposed surface, which constitutes the major contact sites to the GTPase (35,40). CR2, which is composed principally of ␣-helix 5, is thought to stabilize the helical bundle and is on the opposite side of the bundle with respect to CR1 and CR3. We mutagenized residues that are conserved in most RhoGEFs as follows: E285A (DH1) in CR1; F322A (DH3), Q332G (DH4), and L339A/E340A (DH5) in CR2; Y417A (DH6) and K422G/E423G (DH7) in CR3; and N452G/E453G (DH8) in the 6th helix. The crystal structure of Dbs-Cdc42 suggests that equivalent residues mutated in Cdc24p DH1, DH7, and DH8 make contacts with Cdc42p (40). The Dbl mutations equivalent to DH3 and DH8 substantially affected GEF activity (41).
Strikingly, DH1, DH3, DH4, DH6, and DH7 mutants were functional in vivo, whereas only DH5 and DH8 were inviable (Fig. 4D). DH1 and DH6 were unaffected in their ability to oligomerize with wild-type Cdc24p (Fig. 4C). The oligomerization defects of the other DH domain mutants could be divided into three classes with DH4 and DH8 showing a reduction in oligomerization of ϳ20%, DH3 and DH7 exhibiting ϳ2-fold reduction in oligomerization, and DH5 showing a substantial defect with an ϳ10-fold reduction (Fig. 4C). We also carried out control pulldowns with anti-HA antibody using strains, which expressed only the myc-tagged Cdc24p DH mutants, and no binding was observed with any mutant. In contrast to Cdc24p DH4, the equivalent Dbl mutant is oligomerization-deficient, whereas the Dbl mutant equivalent to DH5 had little effect on oligomerization (27). As both mutations that substantially reduced oligomerization (DH3 and DH5) are in the CR2, our results are consistent with studies on Dbl oligomerization, in which the CR2 is important for DH-DH interactions.
Cdc24p Oligomerization Is Necessary for Its Nuclear Accumulation-As Cdc24p shuttles between the sites of growth and nucleus (4 -6, 42), we examined whether the level of Cdc24p oligomers is altered when this GEF accumulates in the nucleus. The level of Cdc24p oligomers in a cdc28 temperature-sensitive strain was unaffected at the nonpermissive temperature in which Cdc24p accumulates in the nucleus as cells arrest at G 1 (4) (a 5-fold decrease in the percentage of budded cells) (Fig. 5). These results are consistent with Cdc24p oligomers being present in the nucleus.
We examined whether the DH mutants representing the three classes of oligomerization defects could localize to the nucleus. To determine the localization of Cdc24p mutants DH3, DH5, and DH8, we examined wild-type cells expressing GFP fusions as DH5 and DH8 mutants were inviable as the sole copy. These strains grew similarly to the wild type, and localization of DH8 GFP fusion was indistinguishable from the wildtype Cdc24p GFP fusion (Fig. 6, A and B). In wild-type cells prior to bud emergence (Fig. 6B, stage 1a), essentially all cells have Cdc24-GFP localized to the nucleus (Fig. 6, A and B). Strikingly, the percentage of cells with Cdc24p observed in the nucleus was reduced by ϳ30% with the DH3 GFP fusion and ϳ50% with the DH5 GFP fusion (Fig. 6, A and B). All three mutants exhibited a decrease in the number of cells with Cdc24p in the nucleus following bud emergence (Fig. 6B, stage  1b). The DH3 fusion was observed in the nucleus in dividing cells similar to the wild type, whereas the DH5 fusion did not accumulate in the nucleus of dividing cells. Quantitation of the percentage of cells with the different GFP fusions localized to the bud tip revealed that both DH3 and DH5 mutants were somewhat defective in localization to small and large buds. These results are consistent with the idea that Cdc24p oligomerization is critical for its localization, both to the nucleus and the bud tip.
Given that the Cdc24p DH3 mutant was viable as the sole Cdc24p copy and could localize to the nucleus and bud tip, albeit with reduced efficiency, we expected this mutant to be able to activate Cdc42p in vivo. We used the CRIB pulldown method to measure the level of Cdc42⅐GTP in cells expressing Cdc24-DH3 as the sole Cdc24p copy. Fig. 6C shows that the levels of Cdc42⅐GTP were similar in cdc24-DH3 and wild-type cells suggesting that a 2-fold decrease in oligomerization does not substantially alter Cdc42 activation in vivo.
Cdc24p Oligomerization Regulates Cdc42p Activation in Vivo-To investigate the effect of Cdc24p oligomerization, we used a regulated oligomerization system, which enables druginduced multimerization (43). We generated fusions of epitope-tagged Cdc24p with two modified FKBP domains in which oligomerization can be induced by a synthetic, cell-permeant, small bivalent molecule. Two copies of the FKBP domain carrying an F36V mutant were used to favor the formation of higher order oligomers with the synthetic oligomerization inducer AP20187, which binds specifically to this modified FKBP domain without substantially affecting the yeast FK506-  (32). Residues mutated in CR1 (red), CR2 (blue), and CR3 (green) are indicated. C, relative binding of indicated 3xmyc-Cdc24p with Cdc24-3xHAp. Binding of indicated wild type (WT), DH3, and DH5 3xmyc-Cdc24p mutants with Cdc24-3xHAp was carried out as described in Fig. 1A (left panel) and average of 3-4 experiments quantitated relative to wild type 3xmyc-Cdc24p (right panel). D, in vivo function of 3xmyc-Cdc24p mutants. Indicated Cdc24p mutants were assessed for their ability to support growth of a cdc24 strain when the Cdc24p rescuing plasmid expression was repressed (Gluc.). IP, immunoprecipitation. FIGURE 5. Cdc24p oligomers can accumulate in the nucleus. Yeast strains were grown at either 25 or 30°C, and cdc28 cells were then shifted to 37°C for 105 min. The percentage of budded cells was determined from an aliquot of fixed cells. Pulldown assays with indicated strains were carried out as described in Fig. 1A, and similar results were observed in two independent experiments. WT, wild type; IP, immunoprecipitation.
sensitive proline rotamase (Fpr1p). This regulated oligomerization system has been used successfully in S. cerevisiae to induce Apg16p oligomerization and restore autophagy (44). Both Cdc24-HA-FKBP-HA and Cdc24-myc-FKBP-myc fusions were functional as the sole copy of Cdc24p in yeast, and these strains grew similar to wild-type cells at 25, 30, and 37°C (data not shown). When yeast cells were grown in the presence of 1 M AP20187, there was a substantial increase in the amount of Cdc24-myc-FKBP-myc bound (Fig. 7A), indicating that the majority of Cdc24p was oligomeric (Ͼ90% by quantitation). Incubation of cell extracts with AP20187 for 15 min at 4°C did not increase the amount of Cdc24p oligomers, suggesting that the FKBP domains may not be accessible to AP20187 during this short incubation at 4°C. AP20187 did not affect cell growth in strains in which Cdc24-FKBP was the sole copy (RAY1854) in liquid or solid media, bud site selection, nor actin cytoskeleton organization (data not shown).
We next examined if induced oligomerization of Cdc24p affected Cdc42p activation in vivo using the CRIB pulldown method to measure the level of Cdc42⅐GTP. Cells expressing Cdc24-HA-FKBP-HA as the sole Cdc24p copy were grown either in the presence or absence of 1 M AP20187, and Fig. 7B shows a specific decrease in the amount of Cdc42⅐GTP when cells were grown in the presence of AP20187. Quantitation of four independent experiments, normalized to the maximum level of activated Cdc42p, indicated that ϳ30% of the Cdc42p was activated in the absence of AP20187 with little effect of AP20187 added to cell extracts. In contrast, a 2-fold reduction in the level of activated Cdc42p was observed when cells were grown in the presence of AP20187. It is unlikely that this effect is because of an altered accessibility of the G-protein to the oligomerized exchange factor as similar results were observed with amino-terminal FKBP Cdc24p fusions (data not shown). These results suggest that oligomerization may regulate Cdc24p GEF activity, either by altering its distribution or affecting its catalytic activity.
To determine whether oligomerization affects Cdc24 GEF activity, we carried out exchange factor assays in vitro with wild-type and mutant DHPH domains in addition to a DHPH-FKBP fusion in which oligomerization could be induced. MBP-Cdc42, GST-DHPH, and GST-DHPH-FKBP were expressed and purified from bacteria (Fig. 8A, lanes 1-3). MBP-Cdc42 was purified in the presence of GDP and was able to bind [␣-32 P]GTP (Fig. 8B, gray diamond). Addition of GST-DHPH or GST-DHPH-FKBP resulted in an increase in GDP-GTP exchange by 2-3-fold, which was most evident at the 2-and 5-min incubation times (Fig. 8B, black triangle and gray box). These results are consistent with a previous report of Cdc24 exchange factor activity stimulation of ϳ3-fold (with [ 35 S]GTP␥S) using GST-Cdc24 and GST-Cdc42 from baculovirus (45). The GEF activity we observed required both Cdc24p and Cdc42p, as no significant GTP binding was detected when MBP replaced MBP-Cdc42, and no stimulation of GTP binding was observed when GST replaced GST-DHPH. In addition, we did not observe an increase in GDP-GTP exchange when heatinactivated GST-DHPH was used. To rule out interfering effects of the adjacent GST domain, we cleaved off the GST moiety by digestion with PreScission protease, resulting in DHPH and DHPH-FKBP proteins (Fig. 8A, lanes 4 and 5).
We examined the ability of the DH mutants DH3, DH5, and DH8 to stimulate Cdc42p GDP-GTP exchange. Fig. 8C shows that each mutant protein stimulated GDP-GTP exchange to a similar extent as the wild-type DHPH protein, with a 2-3-fold increase in GTP binding for mutant and wild-type DHPH compared with MBP-Cdc42 alone. These results suggest that the inviability of Cdc24p DH5 and DH8 mutants when present as the sole copy is not because of a substantial perturbation of guanine nucleotide exchange factor activity, but in the case of Cdc24p DH5 is likely to be due to its inability to localize correctly. These results suggest that GEF activity in vitro and oligomerization are independent functions. We next investigated if oligomerization affected Cdc24p DHPH guanine nucleotide exchange factor activity. Fig. 8D shows that preincubation of DHPH-FKBP with AP20187 did not affect the stimulation of Cdc42p GDP-GTP exchange with a 2-3-fold increase in GTP binding in the presence or absence of AP20187 compared with MBP-Cdc42 alone. Together our results indicate that oligomerization does not directly affect Cdc24p GEF activity.
To determine whether oligomerization affects Cdc24p localization, we examined cells expressing Cdc24-HA-GFP or Cdc24-HA-FKBP-GFP as the sole copy. These strains grew similarly, and the localization of these two Cdc24p fusions was indistinguishable (Fig. 9A). Cdc24p was localized to the nucleus prior to bud emergence (Fig. 9B,  stage 1a), to the bud tip upon bud emergence (Fig. 9B, stage 1b), at the bud plasma membrane as the bud increased in size (Fig. 9B, stage 2), and to the septum and both nuclei during cell division (Fig. 9B, stage 3). However, this GEF was observed in the nucleus at all cell cycle stages when Cdc24-HA-FKBP-GFP cells were grown in the presence of AP20187 (Fig. 9, B and C). After bud emergence, 75% of the cells grown in the presence of AP20187 had Cdc24p localized to their nuclei compared with 25% in its absence. At this stage, essentially all the cells had Cdc24p (in the presence and absence of AP20187) localized to the bud tip. Strikingly, in cells with larger buds (stage 2) in presence of AP20187, 40% of cells had Cdc24p localized to the nucleus, whereas in the absence of AP20187, Cdc24p was not observed in the nucleus. At stage 2, ϳ90% of the cells had Cdc24p localized to the bud plasma membrane in the presence or absence of AP20187. In cells undergoing cytokinesis, there was no difference in the distribution of Cdc24p in the presence or absence of AP20187 with similar percentages of cells having Cdc24p localized to the nucleus and cell division sites. These results suggest that oligomeric Cdc24p can undergo nucleocytoplasmic shuttling; however, export of nuclear oligomeric Cdc24p is reduced, resulting in nuclear Cdc24p at all cell cycle stages. This reduction in nuclear export is likely to account for the decrease in active Cdc42p.
To determine whether chemically induced oligomerization could restore the localization defect of the Cdc24p DH3 mutant, we fused FKBP to the carboxyl terminus to Cdc24p in Cdc24-DH3-GFP and examined cells expressing this fusion as the sole Cdc24p copy. This strain grew normally, and the localization of Cdc24-DH3-FKBP-GFP, in the absence of AP20187, was similar to that of Cdc24-DH3-GFP in wild-type cells (Fig.   FIGURE 7. Chemically induced Cdc24p oligomerization results in a decrease in Cdc42⅐GTP levels. A, yeast expressing Cdc24-HA-FKBP-HA and Cdc24-myc-FKBP-myc were grown in the absence (Ϫ) or presence of AP20187 (ϩ O/N) and supernatants (10,000 ϫ g) were incubated with (ϩ Ext) or without (Ϫ) AP20187. Samples were analyzed as described in Fig. 1A. B, yeast strains expressing as sole copy Cdc24-HA-FKBP-HA were grown with AP20187 as described in A, and samples were analyzed by SDS-PAGE prior to (Extracts) and following incubation with GST-CRIB resin (Pull downs), and immunoblots were probed with anti-Cdc42 sera. The average level of activated Cdc42p (n ϭ 4) is shown below, relative to the maximum level of activated Cdc42p in extracts following incubation with GTP␥S. Student's t test p value Ͻ0.02 (*). A and B, and Fig. 6, A and B). Strikingly, we observed a partial recovery of Cdc24-DH3-FKBP-GFP localization defect when cells were grown in the presence of AP20187 (Fig. 10, A  and B) with an increase in the number of cell with Cdc24p localized to the nucleus and bud tip. There was an ϳ2-fold increase in the percentage of cells grown in the presence of AP20187 with Cdc24p localized to the nucleus in cell cycle phases G 1 (stage 1a), G 1 /S (stage 1b), and G 2 /M (stage 2). Furthermore, there was an ϳ50% increase in the percentage of cells grown in the presence of AP20187 with Cdc24p localized to the bud tip in cell cycle phases G 1 /S (stage 1b) and an ϳ3-fold increase in G 2 /M (stage 2) cells. These results indicate that the localization defect of the Cdc24-DH3 mutant is due to a decreased ability to oligomerize.

DISCUSSION
We show that Cdc24p forms oligomers in vivo via the catalytic DH domain. This homotypic interaction requires the conserved region 2 of the catalytic DH domain, and mutation of two residues in CR2 (DH5), which dramatically reduces Cdc24p oligomerization, renders this exchange factor nonfunctional in vivo. Neither these DH5 mutations nor the DH3 mutation altered Cdc24p in vitro guanine nucleotide exchange factor activity. G 1 cell cycle arrest, which results in nuclear accumulation of this exchange factor, did not alter Cdc24p oligomerization levels. Both the DH3 and DH5 mutations, which reduced Cdc24p oligomerization, also resulted in a reduction in Cdc24p localization to the nucleus and the bud tip. The localization of the Cdc24p DH3 mutant to the nucleus and bud tip was partially restored upon chemically induced oligomerization. Chemically induced oligomerization of wild-type Cdc24p resulted in a decrease in Cdc42p activation and the accumulation of Cdc24p in the nucleus at all cell cycle stages. In vitro, induced oligomerization of the DHPH domain did not affect guanine nucleotide exchange factor activity. Together these results suggest that Cdc24p oligomerization is important for the correct localization of this exchange factor and may downregulate this GEF by increasing nuclear sequestration.
Using a pulldown assay with yeast cells expressing different epitope-tagged versions of Cdc24p and an in vitro binding assay, we have demonstrated that full-length Cdc24p forms oligomers. In independent experiments, ϳ3% of the total 3xmyc-Cdc24p associated with Cdc24-3xHAp, and if we assume that 3xmyc-Cdc24p associates equivalently with either 3xmyc-Cdc24p or Cdc24-3xHAp and vice versa, then minimally 6% of Cdc24p is in an oligomeric form in vivo. It is possible that Cdc24p oligomers are dynamic with their formation or stability dependent on the cell cycle stage, Cdc24p localization, Cdc24p association with additional proteins such as Cdc42p, Bem1p, or Rsr1p, or phosphorylation state. Blocking cells in G 1 phase in a cdc28 temperature-sensitive mutant did not dramatically alter the level of oligomers, suggesting that Cdc24p oligomers are present in G 1 cells. We consider it unlikely that oligomer formation is strictly dependent on association with Bem1p or Rsr1p as truncation mutants lacking domains, which are necessary for binding these proteins, still oligomerized. Furthermore, Cdc24p truncations lacking the amino-terminal 280 amino acids, which do not accumulate in the nucleus (4 -6), nonetheless oligomerized, suggesting that Cdc24p does not need to localize to the nucleus to form oligomers. A Cdc24p truncation, which lacks the carboxyl-terminal 380 amino acids, including the PB1 and PH domains, was able to bind wild-type Cdc24, albeit at a lower level compared with wild-type Cdc24p oligomerization. Such a carboxyl-terminal truncation does not localize to bud tips or sites of cell division (4 -6), suggesting that Cdc24p oligomer stability requires its cortical localization. Cdc24p oligomerization was also observed in vitro, indicating that other yeast proteins and phosphorylation are not critical for this multimeric state.
Similar to Dbl and RasGRF (22,27), the DH domain of Cdc24p is necessary and sufficient for oligomerization. This contrasts the oligomerization observed with the exchange factors ␣-Pix, p115 RhoGEF, PDZ-RhoGEF (PRG), and LARG, which requires the carboxyl-terminal region of the respective GEF (25,26,46). Previously, genetic and two-hybrid studies have suggested that the PB1 domain and an amino-terminal PB1-like region (18) interact intramolecularly, yet such interactions are not important for oligomerization as mutants lacking these regions still oligomerize. Nonetheless, the relationship between Cdc24p intra-and intermolecular interactions remains unclear, and it will be necessary to observe such interactions in intact cells to assess how they affect one another.
Our analyses of the importance of the different conserved regions in the DH domain indicate that CR2 is critical for oligomerization. Crystallographic studies with Dbs revealed that the equivalent residues mutated in Cdc24p CR1 (DH1) and CR3 (DH7) make contacts with Cdc42p (40). The former mutant oligomerizes similar to wild-type Cdc24p, whereas the latter is ϳ2-fold reduced. Surprisingly, the Cdc24p DH4 mutant (Dbl equivalent H556A) is hardly affected in its oligomerization, whereas this Dbl mutant is oligomerization-deficient (27). The , were analyzed by SDS-PAGE followed by Coomassie Blue staining. B, guanine nucleotide exchange factor activity of GST-DHPH and GST-DHPH-FKBP. Indicated proteins, including MBP-Cdc42 in the GDP-bound state, were incubated with [␣-32 P]GTP for the indicated times, quenched with chilled L2 buffer containing 1 mM GTP, and filtered through nitrocellulose membranes. Radioactivity retained on filters was determined by scintillation counting. Each determination was carried out in duplicate with standard deviation indicated. C, guanine nucleotide exchange factor activity of wild-type DHPH and DHPH mutants. Following cleavage of GST moiety using PreScission protease, the indicated DHPH proteins were assayed for guanine nucleotide exchange factor activity as described above. The amount of MBP-Cdc42⅐GTP after 10 min of incubation with DHPH was set to 100% in each experiment (3-9 pmol of Cdc42⅐GTP). Each determination was done in triplicate, and averages are shown. Experiments carried out with MBP alone had less than 5% Cdc42⅐GTP. Similar results were observed in two independent experiments. D, guanine nucleotide exchange factor activity of DHPH-FKBP in the presence and absence of AP20187. Indicated proteins were incubated either with or without 20 M AP20187 for 30 min at 4°C prior to guanine nucleotide exchange factor assay (as described above). Each determination was in duplicate with standard deviations indicated. Similar results were observed in two independent experiments. We also observed similar results in two independent experiments in which the GST moiety of GST-DHPH-FKBP was not cleaved.
Cdc24p DH3 mutant (Dbl equivalent F546A) is somewhat defective in oligomerization, whereas the DH5 mutant (Dbl equivalent E565A) is substantially defective in oligomerization. In Dbl, the former mutant (F546A) is defective in oligomerization, whereas the latter Dbl mutant (E565A) had little effect (27). This Cdc24p DH5 mutant is also nonfunctional in vivo, raising the attractive possibility that oligomerization is necessary for in vivo function. In vitro GEF assays indicate that Cdc24p DH3 and DH5 are both able to stimulate Cdc42p GDP-GTP exchange. Strikingly, Cdc24p DH3 and DH5 were defective in localizing to both the nucleus and the bud tip. Furthermore, chemically induced oligomerization of Cdc24p DH3 mutant partially restored localization to both the nucleus and the bud tip, indicating that the DH3 localization defect is in part due to a reduction in oligomerization. It is difficult to relate the levels of Cdc24p oligomer observed by co-immunopurification  to the localization defect observed in Cdc24p DH3 and DH5 cells. It is possible that the level of Cdc24p oligomers determined by co-immunopurification substantially underestimates the amount of oligomer in the cell, that these DH domain mutants affect other functions in addition to oligomerization, or that the localization of nonoligomeric Cdc24p is dependent on Cdc24p oligomers. Presently we are unable to distinguish between these possibilities. Irrespectively, truncation mutants, which do not localize to the nucleus or bud tip, can nonetheless oligomerize. Taken together these results suggest that Cdc24p oligomerization is necessary for its localization. Furthermore, the Cdc24-m1p mutant, which is defective in Far1 binding and nuclear localization, still forms oligomers. 4 Our results are consistent with studies on Dbl oligomerization, in which the CR2 region, which is opposite the Rho G-protein interaction site, is important for DH domain-DH domain interactions.
Our studies do not address the size of Cdc24p oligomers, specifically whether dimers, trimers, tetramers, or higher order multimers of Cdc24p exist. Certainly the existence of higher order Cdc24p multimers, if correctly localized, could result in increased local concentrations of activated Cdc42p. We examined the effect of constitutive Cdc24p oligomerization using two carboxyl-terminal fused FKBP domains, and we observed a decrease in the level of activated Cdc42p when cells were grown in the presence of the homo bi-functional dimerizer AP20187. Examination of the cellular distribution of these chemically induced Cdc24p oligomers revealed an increase in the number of cells with nuclear localized Cdc24p, most apparent in cells with small and medium sized buds. These results suggest that in wild-type cells Cdc24p can associate to form oligomers, which then dissociate for efficient nuclear export.
Recently Baumeister et al. (42) have shown that chemically induced dimerization of the Dbs DHPH domain is sufficient to alter its subcellular distribution and drive it to the plasma membrane. This effect is presumably because of an increase in avidity of the PH domain interactions with the phosphoinositide phosphatidylinositol 4,5-bisphosphate. However the authors were unable to detect an increase in Cdc42p activation upon dimerization of Dbs DHPH domain. Similarly, a Dbl mutant that is defective for oligomerization does not affect in vitro GEF activity (27). Our observations that mutants that are oligomerization-defective and chemically induced Cdc24p oligomerization both affect the localization of this exchange factor are consistent with this recent study on Dbs (42). Oligomerization-defective or constitutively oligomerized Cdc24p mutants catalyzed Cdc42p GDP-GTP exchange in vitro similar to wild-type Cdc24p. Therefore, we consider it unlikely that the decrease in activated Cdc42p observed in vivo is because of changes in exchange factor activity upon oligomerization, but rather nuclear sequestration.
Interestingly, two ring finger domain-containing proteins, which associate with Cdc24p, also appear to oligomerize. Ste5p has been shown to oligomerize by interallelic complementation, two-hybrid analysis, and pulldowns (47)(48)(49)(50)(51). The majority of both Ste5p (48) and Cdc24p appears to be monomeric in cells. Ste5p oligomerization affects its nuclear shuttling, recruitment to the plasma membrane, and activation (48). In contrast to our results with Cdc24p oligomers, which are more efficiently retained in the nucleus, Ste5p oligomers are more efficiently exported from the nucleus and more efficiently recruited to the plasma membrane (48). It is likely that Cdc24p binds oligomeric Ste5p (14). We favor the possibility that oligomeric Cdc24p binds Far1p, and this complex is retained in the nucleus, whereas monomeric Cdc24p binds oligomeric Ste5p, and this complex is efficiently exported from the nucleus. It also possible that Far1p oligomerizes, as artificial dimerization of Far1p can activate Cdc24p. However, Far1p dimers have not been detected in vivo (19). Further work will be needed to determine whether Cdc24p, Ste5p, and Far1p oligomerization regulates the association of these three proteins.
Oligomerization of Dbl family exchange factors is conserved between yeast and mammalian GEFs. Both yeast and mammalian exchange factors such as Dbl associate via their catalytic DH domain, and the conserved region 2 of the DH domain is required for this homotypic interaction. Our results suggest that oligomerization of Dbl family exchange factors may be important for regulating their localization and hence activity.