Analysis of the Role of the Hypervariable Region of Yeast Ras2p and Its Farnesylation in the Interaction with Exchange Factors and Adenylyl Cyclase*

Ras proteins from Saccharomyces cerevisiae differ from mammalian Ha-Ras in their extended C-terminal hypervariable region. We have analyzed the function of this region and the effect of its farnesylation with respect to the action of the GDP/GTP exchange factors (GEFs) Cdc25p and Sdc25p and the target adenylyl cyclase. Whereas Ras2p farnesylation had no effect on the interaction with purified GEFs from the Cdc25 family, this modification became a strict requirement for stimulation of the nucleotide exchange on Ras using reconstituted cell-free systems with GEFs bound to the cell membrane. Determination of GEF effects showed that in cell membrane the Cdc25p dependent activity on Ras2p was predominant over that of Sdc25p. In contrast to full-length GEFs, a membrane-bound C-terminal region containing the catalytic domain of Cdc25p was still able to react productively with unfarnesylated Ras2p. These results indicate that in membrane-bound full-length GEF the N-terminal moiety regulates the interaction between catalytic domain and farnesylated Ras2p·GDP. Differently from GEF, full activation of adenylyl cyclase did not require farnesylation of Ras2p·GTP, even if this step of maturation was found to facilitate the interaction. The use of Ha-Ras/Ras2p chimaeras of different length emphasized the key role of the hypervariable region of Ras2p in inducing maximum activation of adenylyl cyclase and for a productive interaction with membrane-bound GEF.

Ras proteins are GTPases cycling between the active GTPbound state and the inactive GDP-bound state. They transmit extracellular signals that regulate cell growth and differentiation (1). The level of activated Ras is controlled by the GTPaseactivating protein and the GDP/GTP exchange factor (GEF) 1 which in the case of Saccharomyces cerevisiae are Ira1p/Ira2p (2,3) and Cdc25p, respectively (4). This organism harbors a second RasGEF (Sdc25p, Ref. 5) of unclear functions, that can complement Cdc25p (6 -8). Ras1p and Ras2p regulate the activity of adenylyl cyclase and cAMP-dependent protein kinases (9). One major difference between yeast and mammalian Ras proteins lies in their C-terminal hypervariable region which in the case of Ras from the former organism is much more extended (ϳ120 versus ϳ20 aa residues). The function of this overextended C-terminal region is as yet unclear. Association with the cell membrane is an essential condition for the function of Ras proteins. Translocation of Ras to the inner surface of the membrane is promoted by sequential post-translational modifications of the C-terminal CAAX consensus box (10). The first step, the farnesylation of cysteine, is followed by proteolytic cleavage of the AAX peptide, methyl-esterification of the exposed isoprenylated cysteine and in the case of human N-Ras, Ha-Ras, and S. cerevisiae Ras1p and Ras2p, palmitoylation of one or two cysteines located upstream to the CAAX motif (11). After farnesylation, AAX proteolysis and methylation, Ras proteins are still mainly cytosolic; their tight association with the plasma membrane requires palmitoylation (12)(13)(14) or for K-Ras a signal composed of a polybasic domain (14). In mammalians, farnesylation was reported to be essential for the action of the ubiquitary exchange factor SOS (15); it targets Raf to the cell membrane (16 -19) and is necessary for transformation (20). In yeast farnesylation of Ras2p was found to be important for the interaction with the adenylyl cyclase-CAP complex (21,22). Information on the role of farnesylation in the activity of yeast Cdc25p and Sdc25p is so far limited to the observation that the isolated catalytic domain of Cdc25p promoted the nucleotide exchange on prenylated Ras and even more strongly on unprocessed Ras (15).
In this work we have analyzed the role of the C-terminal hypervariable region of Ras proteins and its farnesylation in the interaction with both full-length exchange factors Cdc25p and Sdc25p, and in the activation of adenylyl cyclase. As methodological approach, a well defined reconstituted in vitro system using membrane preparations from isogenic yeast strains was utilized in order to mimic in vivo conditions and compensate for the fact that the isolated full-length Cdc25p and Sdc25p are not yet available despite considerable efforts. In this context the activities of these two GEFs as components bound to the cell membrane were characterized and compared with the activity of membrane-bound GEF C-terminal region. The obtained results have further enlighted the regulatory role of the N-terminal region of GEF on the C-terminal catalytic domain and demonstrated the absolute requirement of Ras2p farnesylation for a productive interaction. The construction of Ha-Ras/Ras2p chimaeras has selectively defined the impor-tance of the hypervariable region of yeast Ras for the activation of adenylyl cyclase.

EXPERIMENTAL PROCEDURES
Media, Plasmids, and Yeast Methods-The standard rich medium used was YEPD (2% bacto-peptone, 1% yeast extract, and 2% dextrose). Selective synthetic media contained 0.67% yeast nitrogen base without amino acids (Difco) supplemented with all the auxotrophic requirements as described (23) and 2% dextrose or 3% glycerol or 2% raffinose plus 4% galactose. pFC1 (8), pYEDP1/8/2 (24), and pIND25-1 (25) are yeast vectors for the expression of full-length SDC25, CDC25 genes, and the 3Ј CDC25 terminal fragment (residues 877-1589), respectively. pYACE1 (26) is a vector overproducing wild-type adenylyl cyclase CYR1 gene product. The dimethyl sulfoxide-modified version (27) of the Liacetate method (28) was carried out for yeast transformation with either plasmid DNA or purified DNA fragments (10 -20 g) using 0.1 M NaCl/TE instead of TE alone for the preparation of yeast competent cells. The vectors pYEDP1/8/2, pIND25-1, and pYACE1 allowed expression in yeast under the control of the galactose inducible GAL10-CYC1 hybrid promoter. Yeast strains transformed by these plasmids were grown at 30°C on minimal selective medium using 2% raffinose as a carbon source to a cell density of 0.15 A 600 units and then induced with galactose (final concentration: 4%). Total yeast DNA was prepared as described in Ref. 29.
Yeast Strains-The genotypes of the yeast strains used are listed in Table I. The sdc25 Ϫ CDC25 ras1 Ϫ ras2 Ϫ CRI4 (AAT3B-⌬S25) and sdc25 Ϫ cdc25 Ϫ ras1 Ϫ ras2 Ϫ CRI4 (AAT3B-⌬2) strains were engineered by replacing the SDC25 genes of strains AAT3B and AAT3B-⌬1, respectively, with a disrupted sdc25::HIS3 allele. For this, the linearized pGEX2T SDC25 plasmid (32) carrying a 424-base pair BglII-BglII deletion of the 3Ј-SDC25 open reading frame was purified and ligated with a 1.25-kb BamHI-BamHI fragment derived from yDp-H (33), containing the HIS3 marker. The unique EcoRI site located at the C terminus of SDC25 was replaced by a SalI site. The purified 2.7kilobase XbaI-SalI fragment (10 -20 g) from this vector was used to transform AAT3B and AAT3B-⌬1 competent cells by homologous recombination. Transformants were selected on synthetic medium by histidine or leucine and histidine prototrophy, respectively. The integration events were mapped by Southern blot analysis on HindIII-EcoRI-digested genomic DNA using a 1124-base pair HaeIII fragment spanning the 3Ј terminal region of the gene as a probe.
The replacement of the ras2::URA3 chromosomal gene in strain AAT3B-⌬2 by the ras2::HIS3 allele gave rise to strain AAT3B-⌬2R2H. This replacement of the original disruption marker was required to facilitate further selection for URA ϩ transformants with pYEDP1/8/2, pIND25-1, or pFC1 plasmids. Conversion of URA3 with the HIS3 disruption marker was carried out with the one-step strategy (34) using the pUH7 marker swap plasmid. AAT3B-⌬2 competent cells were transformed with linearized SmaI-pUH7 fragment containing an URA3 gene disrupted with the HIS3 marker. HIS ϩ transformants were counter-selected for the marker URA3 using 5-fluoroorotic acid (35). The replacement of ras2::URA3 by ras2::HIS3 was confirmed by Southern blot analysis of HindIII-digested genomic DNA transformants using a EcoRI-HindIII 32 P-labeled probe containing the genomic RAS2 wildtype gene and its flanking sequences.
Purification of Farnesyl-Protein Transferase-E. coli JM 101 containing the plasmid pGP14-2/1/2 (39) was used to express the coupled S. cerevisiae RAM1/RAM2 gene products encoding farnesyl transferase (FTase). The transformed strain was grown at 24°C in 4 liters of LB-rich medium with 50 g ml Ϫ1 ampicillin and induced at a cell density of 0.5 A 600 with 0.1 mM isopropyl-␤-D-thiogalactopyranoside. After 12-15 h, the cells were collected by centrifugation, washed, and sonicated four times for 30 s at 4°C in 100 ml of buffer A (25 mM Tris-HCl, pH 7.8, 10 M ZnCl 2 , 1 mM MgCl 2 , 1 mM dithiothreitol) containing 80 mM NaCl, 1 mg ml Ϫ1 lysozyme, 100 g ml Ϫ1 DNase (Roche Molecular Biochemicals), and protease inhibitors (2 mM Pefablock S-C, 1.7 g ml Ϫ1 pepstatin, 2.5 g ml Ϫ1 aprotinin, 1 g ml Ϫ1 leupeptin; Roche Molecular Biochemicals). Supernatant from 2 h centrifugation at 140,000 ϫ g was loaded on a HiPrep 16/10 Source 30Q column (Amersham Pharmacia Biotech). The most active fractions eluted between 180 and 250 mM NaCl in buffer A were applied to an affinity chromatographic support carrying the hexapeptide TKCVIM (corresponding to the C-terminal residues of K-RasB) and purified (40). The collected active eluted fractions (ϳ50% pure) were concentrated by ultrafiltration and stored in buffer A with 50% glycerol at Ϫ20°C. Preparation of in Vitro Farnesylated Ras-GDP, GTP, or GTP␥Sbound Ras products (4 M) were farnesylated by a 45-min incubation at 30°C in farnesylation buffer with 0.5 M purified FTase, 100 M cold Fpp (Isotopchim) and in the presence of a protease inhibitors mixture (2 g ml Ϫ1 aprotinin, 1 g ml Ϫ1 leupeptin, 60 g ml Ϫ1 antipain, 2 mM Pefablock S-C, Roche Molecular Biochemicals). Level of farnesylation of the various Ras products using [ 3 H]Fpp as substrate was analyzed by electrotransfer to Nytran-N membrane (Schleicher & Schuell) from a 12.5% SDS-PAGE followed by autoradiography of the membrane pretreated with intensifier EN 3 HANCE (NEN Life Science Products Inc.).  washed twice with 3 ml of ice-cold 50 mM Tris-HCl, pH 7.5, 100 mM NH 4 Cl, 10 mM MgCl 2 , 7 mM ␤-mercaptoethanol. The filters were then counted for radioactivity. Adenylyl Cyclase Assay-The adenylyl cyclase assay was carried out as described (41). Yeast membranes were prepared from cells grown in their respective selective medium (36) and collected at a cell density of 1.5 A 600 units and used as source of membrane-associated adenylyl cyclase and GEF (Cdc25p, Sdc25p, or both factors). The cAMP production was determined after 18 min at 30°C at which time the reaction was linear. The 100-l reaction mixture, with the indicated concentrations of farnesylated or unfarnesylated Ras proteins in their preformed GDP, GTP, or GTP␥S complex, contained either 30 -40 g of membrane preparation for the yeast strains expressing the CRI4-encoded adenylyl cyclase (CRI4-adenyl cyclase) gene or 3.5 g of membrane preparation for the yeast strain TS1-6 which harbors an overexpression vector for the wild-type adenylyl cyclase gene (CYR1). These amounts of membranes in the assay gave similar levels of Ras-uncoupled adenylyl cyclase activity as determined in the presence of 1.5 mM MnCl 2 . Preformed Ras⅐GDP, GTP, or GTP␥S complexes were obtained in the presence of 0.5 mM of the corresponding unlabeled guanine nucleotide and farnesylated as described above. Unprenylated complexes were treated identically but omitting the FTase. The reaction was started with a mixture containing 50 mM MES, pH 6.2, 5 mM MgCl 2 , GTP, or GTP␥S (0.5 mM), cAMP (0.5 mM), [␣-32 P]ATP (0.3 mM, 5 GBq mmol Ϫ1 ), theophylline, creatine phosphate, and creatine kinase.

Determination of Ras⅐[ 3 H]GDP Dissociation Rates-Preformed
Other Methods-Protein concentration were measured by the Bio-Rad protein assay or in the case of yeast membranes by the Lowry method (42), using bovine serum albumin as a standard. SDS-PAGE was carried out using a 12.5% acrylamide separating gel. DNA probes were 32 P-labeled with the Megaprime DNA labeling system from Amersham Pharmacia Biotech.

Properties of the Biological Components-
The properties of the biological components used in this work were carefully characterized. The experiments involving yeast membranes were carried out with different membrane preparations in order to assure reproducibility of the results. Fig. 1 illustrates the forms and constructs of Ras used for our experiments. They were at least 90% pure on Coomassie Blue-stained SDS-PAGE ( Fig. 2A) and stable for at least several months when kept at Ϫ20°C in storage buffer (36). It is important to emphasize that all purified Ras species displayed a molar stoichiometric GDP or GTP binding close to one. Fig. 2B illustrates the farnesylation of the various Ras species after electro-transfer using [ 3 H]Fpp as substrate. The FTase assay showed that farnesylation of the Ras species was complete; 1 mol of farnesyl group was found to be bound per mole of intact Ras product.
The catalytic domains of Sdc25p (C-Sdc25p, 550 aa) and CDC25 Mm (C-CDC25 Mm , 285 aa) were pure, and Cdc25p (C-Cdc25p, 509 aa) Ͼ50% pure. The purified full-length CDC25 Mm (1262 aa), obtained as N-terminal fusion with the maltosebinding protein was at least Ͼ50% pure, the contamination consisting of its C-terminal truncated forms (38). All of these GEFs were stable for several months when conserved under the same conditions as the various Ras species.
In Vitro Farnesylation of Ras2p Does Not Influence the Cdc25 GEF dependent GDP Dissociation Rate-Farnesylation in vitro allows a detailed analysis of partially processed Ras avoiding the introduction of mutations into the site for palmi-toylation of Ras2p and the use of detergents for solubilization of processed Ras proteins from cell extract. Both these procedures could affect the interaction with Ras ligands.
At first, we examined whether in vitro farnesylation of Ras2p affected the intrinsic interaction with GDP and the GDP dissociation rate mediated by Cdc25p and Sdc25p catalytic domains but no effect was found (Table II), differently from the observations of other authors using the catalytic domain of Cdc25p (15). Because the specific activities of these various GEF catalytic domains are different (43), their concentration in the assays was adjusted to give a comparable stimulation on the Ras2p⅐GDP dissociation rate. Full-length Cdc25p and Sdc25p could not be tested because their isolation has as yet to be achieved. However, differently from a report on full-length SOS (15) we were unable to see any enhancement by farnesylation of the activity of full-length, purified mouse CDC25 Mm or of its catalytic domain C-CDC25 Mm (Table II).
Farnesylation of Ras2p Is Strictly Required for the Nucleotide Exchange Activity Mediated by Membrane-bound Cdc25p or Sdc25p-The complete disruption of both RAS1 and RAS2 genes or of the CDC25 gene is lethal (6,44,45). However, introduction of the CRI4 mutation (T1651I, Ref. 46) into adenylyl cyclase gene bypasses the requirement for both RAS and CDC25 genes via the constitutive production of low levels of cAMP. CRI4-adenylyl cyclase activity is still strongly stimulated by Ras⅐GTP proteins in vivo and in vitro (26,46). To take advantage of these properties, we have used a set of isogenic yeast strains in a CRI4 and ras1 Ϫ , ras2 Ϫ background. Starting from strain AAT3B, strains were constructed, in which CDC25 (AAT3B-⌬1), SDC25 (AAT3B-⌬S25) or both these genes (AAT3B-⌬2) were disrupted. It is essential to stress that measurement of Ras nucleotide exchange dependent on membranebound Cdc25p or Sdc25p cannot be directly determined by the classical methods on nitrocellulose or gel filtration, as a likely consequence of the inherent properties of the association between exchange factors and cell membrane. For this reason, in most our experiments the exchange activity on Ras was followed indirectly in a reconstituted adenylyl cyclase assay. The validity of this method was proved in previous work (41). Membrane preparations from AAT3B, AAT3B-⌬1, AAT3B-⌬S25, and AAT3B-⌬2 were used as a source of adenylyl cyclase in combination with either Cdc25p or Sdc25p, or both GEFs for in vitro assays in which purified intact Ras2p was added exogenously. This hybrid in vitro system reproduces in vivo conditions in which Cdc25p or Sdc25p are anchored to the membrane. We could so analyze selectively the effect of the membrane-associated GEF(s) on increasing concentrations of unfarnesylated or farnesylated Ras2p⅐GDP via the extent to which the generated Ras⅐GTP could activate adenylyl cyclase. Fig. 3A confirms that unfarnesylated Ras2p⅐GDP was unable to activate adenylyl cyclase whatever membrane preparation was used and shows that restoration of adenylyl cyclase is strictly dependent on farnesylation of Ras2p⅐GDP. The stimulation was slightly reduced when only Cdc25p was present as compared with membranes carrying both Cdc25 and Sdc25 products. The presence of Sdc25p alone stimulated markedly less than Cdc25p. The low but relevant residual cAMP production observed in the absence of GEF corresponds to the intrinsic regeneration of prenylated active complex. In agreement with this is the observation that preformed prenylated Ras2p⅐GTP␥S complex interacts more efficiently with adenylyl cyclase than the unprenylated Ras2p⅐GTP␥S complex (Fig. 3B). Therefore, the possibility of the existence of another low-expressed membrane-bound exchange activity sounds rather unprobable. Taking the extent of activation as a measure for the productive interaction between Ras2p and adenylyl cyclase, the concentrations inducing half-maximum activation (K a ), were calculated to be 7 nM for prenylated Ras2p and 100 nM for the unprocessed form whatever the source of yeast strain membranes. This indicates that neither Cdc25p nor Sdc25p can influence the interaction between Ras2p⅐GTP and the target adenylyl cyclase.
Differently from Intact Yeast GEF, Farnesylation of Ras2p Is Not Required for the Exchange Activity Dependent on Membrane-coupled C-terminal Region of Cdc25p-For a more detailed investigation of the specific effects of Ras2p farnesylation on the response to GEF, we have transformed a yeast strain depleted of genomic CDC25 and SDC25 with pYEDP1/ 8/2, pFC1, or pIND25-1 overexpressing full-length Cdc25p and Sdc25p, and the C-terminal region of the former (C-Cdc25p 877-1589), respectively. Figs. 4, A and B, confirm the strict dependence on farnesylation of Ras2p⅐GDP for the regeneration of the active complex mediated by overexpressed membrane-associated Cdc25p or Sdc25p. In both conditions, saturation curves were similar showing that farnesylated Ras2p can react with the same efficiency with either exchange factor.
Differently from membrane-bound intact GEF, farnesylation is not required for the exchange activity if a membrane-associated GEF lacking the N-terminal moiety (C-Cdc25p 877-1589) is used (Fig. 4C). In this case, Ras2p⅐GDP can rapidly be converted to its activated form even if unprenylated. With prenylated Ras2p, the saturation curve was similar to that observed with membrane-bound full-length Cdc25p. In conclusion, the requirement of Ras2p farnesylation for activation by membrane-associated intact GEF appears not to be the consequence of membrane targeting of Ras2p, since the C-terminal region of Cdc25p can react to the same extent with unfarnesylated or farnesylated form of Ras2p. The dependence on farnesylation is a selective property inherent in the membraneassociated full-length Cdc25p, indicating that the N-terminal moiety specifically controls the interaction with farnesylated Ras2p. The same is probably valid also for Sdc25p.
Comparison of the Activation of Adenylyl Cyclase by Ras2p and Ha-Ras-The reconstituted cell-free system described in the previous sections would make possible a precise analysis of the impact of farnesylation of Ras2p and Ha-Ras, if fully farnesylated Ha-Ras p21 were available. Unfortunately, in contrast to Ras2p, all our attempts to farnesylate in vitro a high percentage of recombinant Ha-Ras were unsuccessful, at most 10% Ha-Ras p21 being farnesylated. This low level of farnesylation is a probable consequence of a partial degradation of the C-terminal extremity of Ha-Ras. To overcome this handicap, we constructed a modified Ha-Ras in which the last 16 Cterminal residues were replaced by the last 16 residues of Ras2p. Fig. 2B shows that this construct designated Ha-Ras 1-173/Ras2p 307-322 can be fully farnesylated. Ha-Ras 1-173/ Ras2p 307-322 can activate CRI4-adenylyl cyclase and its affinity for adenylyl cyclase is increased by prenylation, like Ras2p (Fig. 5A). However, the maximal efficiency (V max ) of Ha-Ras 1-173/Ras2p 307-322, as measured by determining the cAMP synthesized as a function of increasing concentrations of the various Ras proteins, prenylated or unprenylated, was different from Ras2p, the latter being 2.4-fold more efficient. The K a values show that the affinity of prenylated Ras2p for CRI4adenylyl cyclase is 8 times stronger than that of prenylated Ha-Ras 1-173/Ras2p 307-322 (6.8 versus 55 nM; Table III), whereas the affinities of the corresponding unprenylated proteins are not very different (100 versus 178 nM, respectively). The requirement for farnesylation is even more evident in the case of activation of the wild-type CYR1 gene product (Fig. 5B) that showed a maximal level of activation 6-fold higher with farnesylated Ras2p than with farnesylated Ha-Ras. The de-   Table III. duced K a values were 26 and 100 nM, respectively (Table III). The ability of unprenylated Ras2p to interact with CYR1 gene product was very low (K a Ͼ 800 nM, These two constructs, which binds stoichiometrically GDP and GTP and are fully farnesylated, showed the same intrinsic dissociation rate values for GDP (Table IV), i.e. slower than that of Ras2p (1.1 ϫ 10 Ϫ2 min Ϫ1 versus 1.83 ϫ 10 Ϫ2 min Ϫ1 ) and close to that of Ha-Ras p21; also their response to C-Cdc25p (509 aa) being similar. Therefore, these two constructs conserved the inherent properties of Ha-Ras. Concerning the cAMP production as a function of unprenylated Ras⅐GTP␥S concentrations, Fig. 6 shows that the presence of the two Cterminal Ras2p domains can both increase V max and affinity of Ha-Ras for CRI4 adenylate cyclase to the levels observed with Ras2p (Table III).
Experiments were carried out to compare the profiles of activation of CRI4 (Fig. 7, panel A) and CYR1 (Fig. 7, panel B) gene products with increasing concentrations of the various prenylated Ras forms. With both farnesylated chimaeras, the saturation levels were comparable to that with farnesylated Ras2p. This result emphasizes the role of the extended Cterminal region of Ras2p for maximum activation of adenylyl cyclase. The extent of activation is particularly evident with the CYR1 product, since the two fused domains of Ras2p increased by 5-6 times the value of V max observed with farnesylated Ha-Ras. The affinity constants estimated from inverse plots of these experiments are summarized in Table III. Farnesylated Ha-Ras/Ras2p chimaeras show affinity constants identical to those of farnesylated intact Ras2p for the CRI4 (7-8 nM) and CYR1 gene (26 -38 nM) products. Comparison of the V max and K a values of prenylated and unprenylated Ras strongly suggests that the C-terminal region of Ras2p encompassing residues 173-307 includes specific structures important for the activation of adenylyl cyclase, whereas farnesylation is mainly involved in promoting the efficiency of the interaction.
Elements Determining the Sensitivity of Ras to the Membrane-associated Catalytic Domain and Full-length Cdc25p-We have also analyzed the sensitivity of Ha-Ras/ Ras2p chimaeras to membrane-bound Cdc25p or C-Cdc25p 877-1589 using the adenylyl cyclase reconstitution assay. Increasing concentrations of the different Ras⅐GDP constructs were used in their unprenylated and prenylated form. Fig. 8A shows that full-length Cdc25p can fulfill its exchange activity  Table III. not only on Ras2p but also on the various Ha-Ras/Ras2p chimaeras with high efficiency, provided that these products were farnesylated. Differently from unprenylated Ras2p, unprenylated Ha-Ras 1-173/Ras2p 307-322 was unable to respond to membrane associated C-Cdc25p (Fig. 8B). However, the fusion to the different N-terminal moieties of Ha-Ras with the complementing C-terminal region of Ras2p, as shown with the two chimaeras Ha-Ras 1-81/Ras2p 89 -322 and Ras 1-173/Ras2p 182-322, restored the sensitivity with the same efficiency as for unfarnesylated Ras2p. The results indicate that structural elements of the C-terminal hypervariable region of Ras2p are required for a productive interaction with membrane-bound C-Cdc25p 877-1589. DISCUSSION The reconstituted cell-free system used in this study reproduces in vitro the physiological conditions of the interaction between Ras and GEFs or adenylyl cyclase as cell membranebound components. Our observation that farnesylation of Ras2p did not affect the GDP dissociation mediated by purified catalytic domains of yeast GEFs or full-length CDC25 Mm emphasizes the need for a system in which these Ras ligands are associated with the cell membrane to study physiologically relevant interactions. The use of CRI4 adenylyl cyclase strains allowed the isolation of Ras-and GEF-free membranes, and enabled the selective study of the effect of Ras2p farnesylation on Cdc25p or Sdc25p interaction by bypassing the lethality of the deletions. In this system the activity of membrane-associated full-length GEFs was strictly dependent on farnesylation of Ras2p. A relevant residual exchange activity was observed in membranes from strains with disrupted CDC25 gene, that was also strictly dependent on farnesylation of Ras2p and could be attributed to Sdc25p. For the first time, it was so possible to define the extent of cell-membrane exchange activity dependent on Sdc25p acting as a second yeast GEF. Genetic analysis has shown that the SDC25 gene can functionally complement a cdc25 mutation and that the SDC25 and CDC25 genes are differently transcribed, the former being expressed late during growth (8). The same K m and V max of membranes with over-    Table III. produced Cdc25p or Sdc25p strongly suggests that the weaker activity of Sdc25p versus Cdc25p in native yeast membranes is related to a low level of expression.
Sdc25p and Cdc25p are associated with the membrane via hydrophobic sequences located in the C-terminal region (7,48). Membrane-associated overexpressed C-Cdc25p 877-1589 was found in this work to convert farnesylated Ras2p⅐GDP to the active state, as do membranes with intact Cdc25p or Sdc25p, but differently from these, it did not require farnesylation of Ras2p to stimulate the exchange activity. This shows that farnesylation is not involved in increasing the association of Ras2p with the cell membrane but favors specific proteinprotein interactions. In fact, it is known that farnesylation is not sufficient for a stable anchoring of Ras to the plasma membrane, palmitoylation of the upstream cysteine being required (12)(13)(14). Our observations are suggestive for a function of the N-terminal domain regulating the activity of the catalytic domain as proposed in Ref. 49. Farnesylation of Ras could induce a topological orientation allowing the accessibility to the catalytic domain of Cdc25p, where the region 1374 -1444 is essential for this interaction (50). Evidence in vitro and in vivo has indicated that the N-terminal moiety of mammalian Ras-GEFs CDC25 Mm (38) and SOS (51) down-regulates the activity of the catalytic region, despite a modular organization fully different from the N-terminal region of Cdc25p or Sdc25p.
Another relevant aspect of this work is the effect of Ha-Ras farnesylation on the response to GEF and adenylyl cyclase. As for Ras2p, farnesylation of Ha-Ras is required for a response to membrane-associated Cdc25p, but differently from Ras2p, unprenylated Ha-Ras is insensitive to membrane-associated C-Cdc25p 877-1589 and becomes as responsive as Ras2p only after fusion with the hypervariable region of Ras2p (residues 182-322). This demonstrates the essential role of this region for a productive interaction with the C-terminal domain of Cdc25p and disagrees with a suggested negative regulatory role of the C-terminal portion of Ras2p, which was proposed from the observation that Ras proteins lacking the C-terminal domain can bypass cdc25 mutations (52). We observed that purified C-Cdc25p can exert a GDP dissociating activity on Ha-Ras or Ha-Ras 1-173/Ras2p 307-322 even better than on Ras2p. These results show that with membrane-associated components protein interaction becomes subject to more selective constraints than with soluble components. Differently from region 8 -181 (homologous to region 1-173 in Ha-ras) which is acidic, the hypervariable region 182-322 displays a highly positive net charge, that together with the presence of extensive hydrophobic stretches is likely essential for binding to membrane-associated GEF.
Ha-Ras p21 was reported to be able to substitute for yeast Ras proteins in sustaining growth and adenylyl cyclase activation, but complementation of the defect of yeast Ras genes was not efficient (53,54). In our hands unprenylated recombinant Ha-Ras was practically unable to stimulate adenylyl cyclase and only its farnesylation enabled some stimulation. Even with farnesylated Ha-Ras maximum activation of adenylyl cyclase was much lower than with Ras2p, an effect that was more pronounced with wild-type adenylyl cyclase than with its CRI4 mutant. However, farnesylated Ha-Ras showed a strong affinity for both CRI4 and CYR1 gene products, almost as high as that of farnesylated Ras2p. In order to clarify the reasons for the functional differences, we extended the observations that an activated Ha-Ras Val-12/Ras2p chimaera containing the first 73 amino acids of Ha-Ras Val-12 stimulates adenylyl cyclase more efficiently than Ha-Ras (53). For this purpose, we constructed two N-terminal Ha-Ras/C-ter Ras2p chimaeras including C-terminal regions of Ras2p of different length (Ha-Ras 1-81/Ras2p 89 -322 and Ha-Ras 1-173/Ras2p 182-322) comprising the extended hypervariable region of Ras2p. Both constructs revealed the same profile of adenylyl cyclase activation as Ras2p, thereby defining the hypervariable domain as an additional important element for full reconstitution of the activity. Other major determinants for the interaction with adenylyl cyclase are the effector region (residues 32-40 in Ha-Ras and 39 -47 in Ras2p) (31,(55)(56)(57)(58)(59), its flanking residues and the switch 2 region (31,58). By means of the CRI4-encoded product, which is highly sensitive to Ras2p even if unfarnesylated, we showed that maximum stimulation of adenylyl cyclase by unprenylated Ras2p or Ha-Ras/Ras2p chimaeras is identical to that obtained with prenylated Ras2p. This emphasizes the crucial role of the hypervariable region for maximum activation and shows that the major effect of farnesylation is to increase the affinity between Ras2p and adenylyl cyclase rather than to stimulate adenylyl cyclase activity.
Even if there is still some uncertainty about a direct interaction between Ras proteins and adenylyl cyclase, due to the need of the adenylyl cyclase tightly associated protein CAP (60, 61) for a proper response to post-translationally modified Ha-Ras (22), it is possible that the basic nature of the C-terminal region of Ras2p could favor an efficient interaction with the leucine repeat-rich region following the N-terminal region of adenylyl cyclase. This interaction could induce a suitable conformation for maximum activation of the catalytic domain. Leucine-rich repeat regions (61) are critical in mediating protein-protein interaction (62), as has been recently shown in SUR-8, a conserved Ras-binding protein that contains this core consensus and positively regulates Ras-mediated signaling in Caenorhabditis elegans (63).
Farnesylation of Ras2p is not required for adenylyl cyclase activation, as has been tested with CRI4-adenylyl cyclase. However, as already reported for the wild-type CYR1 product (21), farnesylation of Ras2⅐GTP increases the affinity for adenylyl cyclase. Compared with the wild-type one, CRI4-adenylyl cyclase shows not only an increased sensitivity to Ras2p (26, 46) but also a higher affinity for farnesylated Ras. One should stress that the differences between these two adenylyl cyclases are only quantitative. Adenylyl cyclase is not as strongly associated with the plasma membrane as Cdc25p and Sdc25p, does not contain a hydrophobic region resembling a membrane-spanning domain (64) and in ras1ras2bcy1 cells it is located in the soluble fraction (65). Overexpression of Cdc25p has been reported to translocate adenylyl cyclase to the membrane fraction (65), while disruption of IRA1 gene dislocates it from the membrane (66), indicating that both Cdc25p and Ira1p are involved in anchoring adenylyl cyclase to the membrane. A complex between the Cdc25p SH3 domain and adenylyl cyclase not mediated by CAP has been demonstrated (67). Taken together these data suggest the existence in the cell of a large oligomeric complex including Cdc25p, Ira1p, and adenylyl cyclase. The association of Ras2p, Cdc25p, and adenylyl cyclase (68) is further supported by this work highlighting the common function of the C-terminal hypervariable region of Ras2p and its farnesylation in promoting the interaction with and/or activation of membrane-bound Cdc25p and adenylyl cyclase. However, neither Cdc25p nor Sdc25p, even when overexpressed, are able to increase the affinity of farnesylated Ras2p for adenylyl cyclase.
In conclusion, this work shows that the cellular localization of Ras2p, its regulators Cdc25p and Sdc25p and target adenylyl cyclase requires structural modifications that are dispensable under conditions of soluble purified components. We have highlighted the essential role of Ras2p farnesylation for GEF responsiveness and the involvement of its C-terminal hypervariable region in the interaction with Cdc25p. Compared with Ha-Ras, this region has been found to contain structural elements essential for the activation of adenylyl cyclase, the farnesylation facilitating this interaction.