Interactions of the amino acid residue at position 31 of the c-Ha-Ras protein with Raf-1 and RalGDS.

The Ras and Rap1A proteins can bind to the Raf and RalGDS families. Ras and Rap1A have Glu and Lys, respectively, at position 31. In the present study, we analyzed the effects of mutating the Glu at position 31 of the c-Ha-Ras protein to Asp, Ala, Arg, and Lys on the interactions with Raf-1 and RalGDS. The Ras-binding domain (RBD) of Raf-1 binds the E31R and E31K Ras mutants less tightly than the wild-type, E31A, and E31D Ras proteins; the introduction of the positively charged Lys or Arg residue at position 31 specifically impairs the binding of Ras with the Raf-1 RBD. On the other hand, the ability of the oncogenic RasG12V protein to activate Raf-1 in HEK293 cells was only partially reduced by the E31R mutation but was drastically impaired by the E31K mutation. Correspondingly, RasG12V(E31K) as well as Rap1A, but not RasG12V(E31R), exhibited abnormally tight binding with the cysteine-rich domain of Raf-1. On the other hand, the E31A, E31R, and E31K mutations, but not the E31D mutation, enhanced the RalGDS RBD-binding activity of Ras, indicating that the negative charge at position 31 of Ras is particularly unfavorable to the interaction with the RalGDS RBD. RasG12V(E31K), RasG12V(E31A), and Rap1A stimulate the RalGDS action more efficiently than the wild-type Ras in the liposome reconstitution assay. All of these results clearly show that the sharp contrast between the characteristics of Ras and Rap1A, with respect to the interactions with Raf-1 and RalGDS, depends on their residues at position 31.

The guanine nucleotide exchange factors for the Ral proteins, such as RalGDS and RGL, have been reported to associate with Ras in a GTP-dependent manner; the interaction was abolished by mutations in the effector region of Ras (8 -12). It has been shown that Ras enhances the guanine nucleotideexchange activity of RalGDS in COS cells (13). The Ras-binding domain of RalGDS has been identified (8 -12), and there is no apparent sequence homology between the RBDs of the Raf family and those of the RalGDS family. Recently, however, the RalGDS RBD was shown to have a ubiquitin-like fold similar to that of the Raf-1 RBD (14,15).
The Rap1A (or Krev-1) protein belongs to the Ras superfamily, and has the same amino acid sequence in the effector region as that of Ras (16 -18). However, at position 31, Rap1A has Lys, whereas Ras has Glu. The RalGDS RBD prefers Rap1A to Ras, but the Raf-1 RBD prefers Ras to Rap1A (19). The crystal structures of the complexes of the Raf-1 RBD with Rap1A⅐GMPPNP (the wild-type and the E30D/K31E mutant) have been solved by x-ray crystallography (20,21). In these two complex structures, Lys 84 of the Raf-1 RBD is not involved in the interaction with the wild-type Rap1A but forms strong and weak salt bridges with Glu 31 and Asp 33 , respectively, of the E30D/K31E mutant Rap1A. The replacement of Lys by Glu at position 31 of Rap1A increased the affinity for the Raf-1 RBD and decreased that for the RalGDS RBD (21). It was proposed, therefore, that the negative charge of Glu 31 creates a favorable complementary interface for the Ras-Raf interaction (21). We have found that the Rap1A-type E31K and D30E/E31K mutations impair the GAP-induced increase in the Ras GTPase activity and the ability to induce neurite outgrowth of PC12 cells (22). Other groups have also reported that the E31K and D30E/E31K mutations reduce the transforming activity of oncogenic Ras in NIH3T3 cells (23,24). In the present study, to analyze in more detail how the Glu residue at position 31 of Ras is involved in the interaction with the two different targets, we tested more mutations at position 31 for their effects on various Ras functions.

EXPERIMENTAL PROCEDURES
Plasmid Construction-Mutations were introduced into a synthetic human c-Ha-ras gene (25) by site-directed mutagenesis with two PCR steps. The ras genes with the mutations in the oncogenic G12V background were subcloned into pMAMneo (CLONTECH) (22) for induced expression in PC12 cells and into pCMV5 (26) for transient expression in HEK293 cells. To prepare the Raf-1 RBD as a GST fusion protein, the DNA fragment corresponding to amino acid residues 51-131 was amplified by PCR from the human full-length raf-1 gene (27,28) and was subcloned into pGEX-4T-3 (Amersham Pharmacia Biotech). The gene for the rat RalGDS RBD was prepared by PCR with the primers, GCG-CGGATCCTCACTGCCTCTCTACAACCAGCAGGTG and GCGCGTC-GACTTAGAAGATGCCTTTGGCAATCCTGAG, from a rat brain cDNA library (CLONTECH). To prepare the GST fusion form of the RalGDS RBD, the PCR product was digested with BamHI and SalI and was then inserted into the BamHI-SalI sites of pGEX-4T-1.
Protein Purification-The wild-type and mutant Ras proteins were expressed in Escherichia coli and were purified by chromatography on DEAE-Sephacel and Sephadex G-75 (Amersham Pharmacia Biotech) columns as described (22). To obtain the post-translationally modified forms of Ras, Ras G12V (E31A), Ras G12V (E31R), and Rap1A, each protein was produced in Sf9 cells infected with baculoviruses containing either the ras, the ras G12V (E31A), the ras G12V (E31R) or the rap1A gene and was purified as described previously (29,30). The GST fusion forms of the Ras binding domains, GST-Raf-1-(51-131) and GST-RalGDS-(769 -895), were each expressed in E. coli at 30°C and were purified using a glutathione-Sepharose 4B (Amersham Pharmacia Biotech) according to the manufacturer instructions. A recombinant Xenopus kinase-negative MAPK (KN-MAPK) in a GST fusion form and a histidine-tagged Xenopus MAPKK were purified from E. coli as described (31,32). For the liposome reconstitution assay, baculoviruses producing a GST-fused Ras (GST-Ras), GST-RalB, GST-Rap1A, or GST-Ras G12V (E31K) were generated as described (33,34). All procedures of passage, infection, and transfection of Sf9 cells and the isolation of recombinant baculoviruses were carried out as described (35). The post-translationally modified form of GST-Ras, GST-RalB, GST-Rap1A, and GST-Ras G12V (E31K) were purified from the membrane fraction of Sf9 cells as described (36). GST-RalGDSb was purified from E. coli as described (33,34).
In Vitro Binding Assays-One g of either GST-Raf-1-(51-131) or GST-RalGDS-(769 -895), in 150 l of phosphate-buffered saline containing 5 mM MgCl 2 and 0.5% Triton X-100, was mixed with 10 l of glutathione-Sepharose 4B beads suspended in phosphate-buffered saline. The mixture was incubated at 4°C for 1 h with various amounts of either wild-type or mutant Ras, which had been complexed with GTP␥S as described previously (37). After this incubation, the resin was washed with 20 mM Tris-HCl buffer (pH 7.5) containing 5 mM MgCl 2 and 150 mM NaCl. The bound proteins were eluted from the resin by boiling in Laemmli's buffer and were fractionated by SDS-PAGE. The Raf-1 CRD binding assay was performed as described previously (6). Briefly, MBP-Raf-1(136 -206), immobilized on the amylose resin, was incubated with 12 pmol of the wild-type or mutant Ras or Rap1A bound with GTP␥S in 20 mM Tris-HCl buffer (pH 7.5) containing 40 mM NaCl, 1 mM EDTA, 1 mM DTT, 5 mM MgCl 2 , and 0.1% Lubrol PX. After an incubation at 4°C for 2 h, the resin was washed, and the bound proteins were eluted and subjected to SDS-PAGE. Immunoblots were probed with the anti-Ras antibody RAS004 (38) or with an anti-Rap1A antibody (Santa Cruz Biotechnology) and were developed using the ECL immunodetection system (Amersham Pharmacia Biotech). Western blots were scanned and relative band intensities were determined with a BioImage densitometer (MilliGen).
After the incubation, the reaction was stopped by adding Laemmli's buffer and boiling. The samples were fractionated by SDS-PAGE, and the phosphorylation of KN-MAPK was measured with a Fuji BAS2000 Bio-imaging Analyzer.
Transfection of PC12 Cells-Each pMAMneo vector with a mutant ras gene was transfected into PC12 cells as described previously (39). Transfectants were selected in medium containing G418 (400 g/ml), and mutant Ras expression was induced by the addition of dexamethasone to a final concentration of 1 M in the culture medium. After 24 h, the number of cells with extended neurites was counted.
MAPK Kinase Assays-Cell extracts (20 g/assay) were electrophoresed in a 12% SDS-polyacrylamide gel, and were transferred to a nitrocellulose membrane for Western blot analysis using an anti-ERK antibody (Santa Cruz Biotechnology). The kinase detection assay on polyacrylamide gels was also performed as described previously (40). Briefly, the cell extracts were electrophoresed on SDS-polyacrylamide gels (12%) containing myelin basic protein (0.5 mg/ml) as the substrate for ERK. After denaturation and renaturation, the gels were incubated for the kinase reaction and were subjected to autoradiography to detect the phosphorylation of myelin basic protein.
RalGDS Assay Using Liposome Reconstitution System-RalGDS activity to stimulate the dissociation of GDP from Ral in the liposomes was measured as described (36). Briefly, liposomes were made by sonication of dried lipids containing phosphatidylserine, phosphatidylinositol, phosphatidylcholine, and phosphatidylethanolamine.

RESULTS AND DISCUSSION
To determine which properties of the amino acid residue at position 31 of Ras are necessary for the interactions with its targets, Raf-1 and RalGDS, we substituted the Glu 31 of the c-Ha-Ras protein by Asp, Ala, and Arg and analyzed the effects of these mutations, as well as those of the previously reported Lys mutation.
Binding to the Raf-1 RBD-First, we examined the effects of the E31D, E31A, E31R, and E31K mutations on the interaction of Ras with the Raf-1 RBD (amino acid residues 51-131 of Raf-1). Various concentrations of each mutant Ras protein in the GTP␥S-bound form were added to the GST-Raf-1(51-131) fusion protein mixed with glutathione-Sepharose, and the amounts of the Ras protein in the precipitates were analyzed by Western blotting (Fig. 1). The Ras(E31D) and Ras(E31A) mutant proteins were found to bind the Raf-1 RBD as strongly as the wild-type Ras (Fig. 1). In contrast, the E31R and E31K mutations reduced the extent of the Raf-1 RBD binding by about 5-fold (Fig. 1).
It has been reported that Glu 31 and Asp 33 of the Ras-type mutant of the Rap1A protein (E30D/K31E) make stronger and weaker salt bridges, respectively, with Lys 84 of the Raf-1 RBD (21) and that the K84A mutation drastically reduced the affinity of the Raf-1 RBD for Ras⅐GTP (42). Therefore, it has been proposed that the salt bridge between Lys 84 of Raf-1 and Glu 31 of Ras is also important for the Ras⅐Raf association (21). However, this study showed that the Ras(E31A) mutant can bind to the Raf-1 RBD as efficiently as the wild-type Ras. Therefore, the introduction of a basic residue at position 31 causes significant repulsion between Ras and the Raf-1 RBD, whereas the acidic residue characteristic of Ras does not appear to create any attractive interface for the Ras-Raf interaction. On the other hand, the D33A mutation of Ras impaired its ability to bind the Raf-1 RBD. 2 Consistently, it was reported that the D33N mutation of Ras causes a drastic reduction in the affinity for a Raf-1 fragment containing the RBD (4). Thus, for Ras binding with the Raf-1 RBD, the salt bridge from the Lys 84 of Raf-1 to the Asp 33 of Ras is much more important than that to the Glu 31 of Ras. This difference in the interactions of the Raf-1 RBD with Glu 31 and Asp 33 between the Ras and Rap1A backgrounds may be due to the structural differences at other positions within these proteins; Ras has Ile 21 and His 27 , corresponding to the Rap1A residues Val 21 and Ile 27 , respectively, which are involved in the interface with the Raf-1 RBD in the crystal structure (20).
Activation of Raf-1-Next, to examine the effect of these mutations on the ability of Ras to activate Raf-1, we transfected HEK293 cells with each mutant ras gene in the oncogenic G12V background. The anti-Raf-1 immunoprecipitate was incubated with [␥-32 P]ATP in the presence of MAPKK and the KN-MAPK, and the incorporation of 32 P into the KN-MAPK was quantitated ( Fig. 2A). The amounts of the mutant Ras proteins expressed in HEK293 cells were nearly the same (Fig.  2B). The E31A and E31D mutations had no effects on the stimulation of the MAPKK kinase activity of Raf-1. On the other hand, the Raf-1 activation activity of Ras was partially impaired by the E31R mutation and, furthermore, was drastically decreased by the E31K mutation.
Activities in PC12 Cells-To analyze the effects of these mutations on the activities of Ras in another cell type, pheochromocytoma (PC) 12 cells were stably transfected with pM-AMneo vectors that conditionally express the Ras G12V , Ras G12V (E31D), Ras G12V (E31A), Ras G12V (E31R), or Ras G12V (E31K) protein. The Ras G12V (E31K)-expressing cells did not extend neurites, as reported (22), whereas the Ras G12V (E31D)-, Ras G12V (E31A)-, and Ras G12V (E31R)-expressing cells extended neurites as well as the cells expressing Ras G12V (Fig. 3). The expression levels of Ras G12V , Ras G12V (E31D), Ras G12V (E31A), and Ras G12V (E31R) were nearly the same (Fig. 4A). The level of Ras G12V (E31K) was slightly lower than those of other mutants, probably because of its autoinhibitory activity (22), but was nevertheless sufficiently higher than that of the endogenous Ras (Fig. 4A). In conclusion, the neurite-inducing activity of Ras G12V was abolished by the E31K mutation but not affected by the E31D, E31A, and E31R mutations (Fig. 3). It should be emphasized that the E31K mutation impairs this function of Ras much more drastically than the E31R mutation. Furthermore, the abilities of these Ras G12V , Ras G12V (E31A), Ras G12V (E31D), Ras G12V (E31R), and Ras G12V (E31K) mutants to induce ERK activation in PC12 cells were examined from delayed mobility of the phosphorylated form on SDS-PAGE (Fig. 4B) and also by an in-gel kinase assay (data not shown). The ERK activities in the Ras G12V (E31A)-, Ras G12V (E31D)-, and Ras G12V (E31R)-expressing cells were elevated as well as in the Ras G12V -expressing cells (Fig. 4B). On the other hand, in the cell expressing Ras G12V (E31K), the ERK activity was the same as in the control cell. In summary, the abilities of the Ras mutants to activate Raf-1 correlate with their abilities to induce ERK activation and also with the induction of neurite outgrowth in PC12 cells. Therefore, the inability of Ras G12V (E31K) to induce neurite outgrowth is probably due to the inability to activate Raf.
Binding to Raf-1 CRD-Ras activation of Raf-1 in both HEK293 and PC12 cells was decreased more severely by the E31K mutation than by the E31R mutation, although Ras(E31R) and Ras(E31K) had nearly the same Raf-1 RBDbinding activities. Recently, we found that Ras binds to the Raf-1 CRD (amino acid residues 132-206) in a GTP-independent manner, which is necessary for the activation of Raf-1 (6) while another group reported a weak GTP-dependence of the binding of Ras with the Raf-1 CRD (5). Furthermore, we found that Ras G12V (E31K) as well as Rap1A has an abnormally enhanced ability to bind the Raf-1 CRD (30). It seems to be 2 M. Shirouzu, K. Hashimoto, and S. Yokoyama, unpublished data.

FIG. 1. Association of the Ras mutants with the Raf-1 RBD.
Various amounts of the E31D (E), E31A (q), E31R (Ⅺ), or E31K (f) mutant or the wild-type Ras (ϫ) were loaded with GTP␥S and tested for GST-Raf-1(51-131) binding, as described under "Experimental Procedures." The intensity of the Ras band was estimated by densitometry. When GST was used instead of GST-Raf-1(51-131), no bands were detected. The data represent the average of at least three experiments.

FIG. 2. Activation of Raf-1 in HEK293 cells by the Ras mutants.
A, HEK293 cells were transfected with 0.8 g of cDNA encoding the indicated Ras mutants. Sixteen h after transfection, the cells were deprived of serum for 24 h. Raf-1 was immunoprecipitated from the extracts by an anti-Raf-1 antibody and was examined for its ability to induce phosphorylation of KN-MAPK in the presence of MAPKK, as described "Experimental Procedures." The incorporation of 32 P into the KN-MAPK was quantitated with a Fuji BAS2000. Each value had the value of cells transfected with pCMV5 alone subtracted and then was normalized by the value of the cells expressing the Ras G12V protein. The data represent the average of at least five independent experiments. B, immunoblotting detection of cell lysates of Ras mutants with the anti-Ras antibody RAS004. through this abnormal interaction of Lys 31 with the Raf-1 CRD that Rap1A G12V and Ras G12V (E31K) dominantly inhibit the Raf-1 activation by Ras G12V , which is related to the Ras-antagonizing activity of Rap1A (30). In this study, we examined the ability of Ras G12V (E31R) to bind to the Raf-1 CRD. Since the binding of Ras to the CRD requires post-translational modification of the C terminus of Ras, we purified the modified form of Ras G12V (E31R) from Sf9 cells, and used it for the Raf-1 CRD binding assay. The Ras G12V (E31R) was shown to bind the Raf-1 CRD as strongly as Ras (the GTP␥S-bound forms, Fig. 5; and the GDP-bound forms, data not shown). Therefore, it is likely that the difference in the Raf-1 activation ability between the E31K and E31R mutants is due to the difference in the ability to bind the Raf-1 CRD. The reason why the extent of Raf-1 activation by Ras G12V (E31K) is much lower than that by Ras G12V (E31R) may be that Ras G12V (E31K) autoinhibits the Raf-1 activation. The present mutagenesis analyses indicate that the wild-type Glu 31 residue of Ras is not required for interaction with either RBD or CRD of Raf-1. In contrast, the Lys residue replacing Glu 31 appears to interact with the Raf-1 CRD. It is interesting that the Raf-1 CRD very specifically recognizes the amino acid at position 31 and prefers Lys to Arg.
Binding to the RalGDS RBD--As in the case of the Raf-1 RBD, we tested the Ras(E31D), Ras(E31A), Ras(E31R), and Ras(E31K) mutant proteins in the GTP␥S-bound form for their abilities to bind with a GST fusion protein of the rat RalGDS RBD, the C-terminal 127 residues corresponding to the mouse RalGDS RBD (10). Interestingly, the Ras(E31A), Ras(E31R), and Ras(E31K) mutants bound the RalGDS RBD more tightly than the wild-type Ras (Fig. 6). In contrast, the E31D mutation did not affect the interaction of Ras with the RalGDS RBD. This indicates that the negative charge at position 31 of Ras particularly weakens the interaction with the RalGDS RBD.
Binding Preference Switching-The affinities of Ras for the Raf-1 RBD and Rap1A for the RalGDS RBD are 100 times higher than those of Ras for the RalGDS RBD and Rap1A for the Raf-1 RBD, respectively (19). Furthermore, the K31E mutation of Rap1A has been reported to cause a 15-fold increase in the affinity of Rap1A for the Raf-1 RBD and a 20-fold decrease in that for the RalGDS RBD (21). Symmetrically, in this study, the E31K mutation of Ras was found to decrease the Raf-1 RBD-binding (about 5-fold) and to increase the RalGDS RBDbinding (about 5-fold). Therefore, not only in the Rap1A background (21) but also in the Ras background, the residue at position 31 serves as the determinant for the preference of either the Raf-1 RBD or the RalGDS RBD. On the other hand, the modes of Ras binding are similar between the Raf-1 and RalGDS RBDs, in that only one type of charge at position 31 of Ras is particularly unfavorable; binding is weaker with positively and negatively charged side chains, respectively, than with oppositely charged (negatively and positively, respectively) and neutral side chains, at position 31. Actually, it has recently been reported that the RalGDS RBD has a fold similar to that of the Raf-1 RBD (14,15).
Effects of Ras G12V (E31A) and Ras G12V (E31K) on the RalGDS Action-It has been reported that Ras, but not Rap1A, can activate the GDP/GTP exchange activity of RalGDS toward Ral in COS cells (13) and that RalGDS stimulated the dissociation of GDP from Ral more strongly in the presence of Rap1A⅐GTP␥S than Ras⅐GTP␥S in the reconstitution assay (36). The discrepancy of the action of Rap1A on RalGDS between intact cell and cell-free systems might be due to the different subcellular distribution of Rap1A and Ral. In this study, we used this recon-stitution system to analyze the RalGDS-binding abilities of Ras G12V (E31A) and Ras G12V (E31K). When the GTP␥S-bound form of Ras G12V (E31A) or Ras G12V (E31K) was incorporated with Ral in the liposomes, RalGDS stimulated the dissociation of GDP from Ral (Fig. 7). These Ras mutants, as well as Rap1A, exhibited higher RalGDS activation activities than that of the wild-type Ras. Thus, in the reconstitution assay system, the RalGDS-activation activity of Ras/Rap1A depends on the Ral-GDS RBD-binding activity.
All of the results of the present study demonstrate that the sharp contrast between Ras and Rap1A, in terms of the Raf-1 and RalGDS interactions, depends on the characteristics of their residues at position 31. Intriguingly, the mechanisms of these interactions appear to be different from the attraction of the charge at this position for the target domain.