Point Mutants of c-Raf-1 RBD with Elevated Binding to v-Ha-Ras*

A mutational analysis of the Ras-binding domain (RBD) of c-Raf-1 identified three amino acid positions (Asn64, Ala85, and Val88) where amino acid substitution with basic residues increases the binding of RBD to recombinant v-Ha-Ras. The greatest increase in binding (6–9-fold) was observed with the A85K-RBD mutant. The elevated binding for the A85K-RBD and V88R-RBD mutants was also detected with Ras expressed in cultured mammalian cells, namely NIH-3T3 and BAF cells. None of the wild type residues in RBD positions Asn64, Ala85, and Val88 have been previously implicated in the interaction with Ras (Block, C., Janknecht, R., Herrmann, C., Nassar, N., and Wittinghofer, A. (1996)Nat. Struct. Biol. 3, 244–251; Nassar, N., Horn, G., Herrmann, C., Scherer, A., McCormick, F., and Wittinghofer, A. (1995)Nature 375, 554–560). The discovery of elevated binding among the mutants in these positions implies that additional RBD residues can be used to generate the Ras·RBD complex. These findings are of particular significance in the design of Ras antagonists based on the RBD prototype. The A85K-RBD mutant can be used to develop an assay for measuring the level of activated Ras in cultured cells; Sepharose-linked A85K-RBD·GST fusion protein served as an activation-specific probe to precipitate Ras·GTP but not Ras·GDP from epidermal growth factor-stimulated cells. A85K-RBD precipitates up to 5-fold more Ras·GTP from mammalian cells than wild type RBD.

Ras is a small G protein that functions as a molecular switch for the transduction of signals generated by a variety of activated cell surface receptors. In resting cells Ras exists in its inactive GDP-bound form. Upon receptor activation, guanine nucleotide exchange factor proteins catalyze the replacement of GDP by GTP (3)(4)(5). The association with GTP converts Ras to an active form capable of relaying signals to an array of downstream effectors (6,7). c-Raf-1 is an effector of Ras (6, 8 -10). Raf is a serine-threonine kinase (11), which transmits incoming signals to the mitogen-activated protein kinase pathway (12)(13)(14). This pathway has been implicated in a variety of cellular responses, including cell proliferation, survival, and differentiation (6,7). The N-terminal Ras-binding domain (RBD) 1 of Raf has been narrowed to a region encompassed by Raf residues 51-131 (15,16). RBD binds to the Ras⅐GTP complex directly in vitro (15,16). Bacterially expressed recombinant RBD adopts a stable folded conformation in vitro (16). RBD expression in v-Ha-Ras-transformed NIH-3T3 fibroblasts leads to the suppression of the transformed phenotype (17). The ability of RBD to reverse v-Ha-Ras-induced cell transformation makes the Ras⅐RBD interface a target for the development of Ras antagonists.
The details of the RBD interaction with Ras revealed in part by a high resolution x-ray crystal structure of the complex between the RBD and another Ras family member, Rap1A (2). A computer model of the RBD⅐Ras complex has been generated on the basis of the Rap1A⅐RBD crystal complex (18). Several RBD residues that are in direct contact with Ras have been inferred from the crystal structure (2) as well as from a mutational analysis of the proposed binding surface of RBD (1,19). RBD residues Gln 66 , Arg 67 , Lys 84 , Arg 89 , Arg 59 , Gln 64 , and Thr 68 are in direct contact with Ras (1,2). A mutational analysis of the RBD binding site showed that Gln 66 , Lys 84 , and Arg 89 are the major contributors to the binding affinity between Ras and Raf (1). Genetic studies on Drosophila melanogaster have also demonstrated that Arg 89 is important to the interaction with Ras both in vivo and in vitro (20). Even a highly conservative substitution of Arg 89 -RBD with lysine abolished the binding between v-Ha-Ras and RBD (21). Lys 84 is thought to be responsible for effector specificity, favoring the formation of the Ras⅐Raf complex in preference to the Rap1A⅐Raf complex (1,22,23). Arg 59 represents a point of A-Raf and c-Raf-1 isozyme discrimination; it is one of the residues that determine the higher affinity of binding for the c-Raf-1⅐v-Ha-Ras complex compared with the A-Raf⅐v-Ha-Ras complex (24).
The measurement Ras activation is critical to the analysis of Ras function. If a significant change in the Ras⅐GDP:Ras⅐GTP ratio is detected between unstimulated and stimulated cells, it indicates that, at least in part, the signal from the ligand of interest is being mediated by Ras (25). The binding domains from effector proteins of small GTPases have been used previously as activation-specific probes to discriminate between GTP-and GDP-bound forms of G-proteins, with differences in binding affinities of several orders of magnitude (26 -28). The binding domain of RalGDS was used to measure Rap1 activation in platelets (26), the Ral-binding domain of the putative Ral effector (RLIP76) isolated GTP-bound Ral (27), and the minimal RBD of c-Raf-1 was used as an activation-specific probe for Ras (28).
The affinity of wild-type (wt) RBD for Ras⅐GDP is 100-fold lower than for Ras⅐GTP (29). GST⅐RBD fusion protein linked to Sepharose beads has been used to precipitate oncogenic Leu 61 -Ha-Ras, which is bound constitutively with GTP (28). The dominant negative Asn 17 -Ha-Ras, which is in the constitutively GDP-bound form (30), does not associate with the RBD⅐GST fusion protein (28). Recently, another activation-specific probe for Ras has been described, in the form of a Fab antibody fragment that exclusively binds Ras⅐GTP but not Ras⅐GDP (31).
Previous mutational studies of RBD described a reduced or unaltered affinity for Ras (1,19,24). The present work has identified several RBD mutants with increased binding to Ras. These mutants provide new tools to analyze the Ras-Raf interaction and can be used as more sensitive probes than wt-RBD for GTP-bound Ras.

EXPERIMENTAL PROCEDURES
Construction of RBD (Residues 51-131) Mutants-The mutants were constructed by polymerase chain reaction site-directed mutagenesis (33), using c-Raf-1 cDNA (17) as a template. Oligonucleotides were synthesized incorporating the desired mutations as well as EcoRI restriction sites at the 5Ј-end to facilitate subcloning into the pGEX-2TH vector linearized with EcoRI. Mutations were verified by nucleotide sequencing.
Protein Expression and Purification-The procedure used for the expression and affinity purification of the glutathione S-transferase⅐ RBD fusion proteins, using glutathione-coated Sepharose beads (Center for Protein and Enzyme Technology, LaTrobe University), was described by J. Frangioni and B. Neel (34). Sepharose-linked GST⅐RBD fusion proteins were suspended in storage buffer (34) as a 1:1 suspension and then stored in 1-ml aliquots at Ϫ20°C. v-Ha-Ras was expressed in NM522 cells (17) using the pUC8 expression vector (35), as described previously (36). v-Ha-Ras was isolated using the procedure of Johnson and Hecht (37). The proteins released from bacterial cells were resolved on a 15 ϫ 3 cm DEAE-Sephacel (Amersham Pharmacia Biotech) column with a 120-ml linear gradient of 0 -0.4 M NaCl in buffer A (38). Fractions containing v-Ha-Ras (determined by SDS-PAGE) were pooled and stored at Ϫ20°C.
Measurement of Protein Concentrations-The concentrations of proteins linked to Sepharose beads (GST⅐RBD as well as RBD-bound Ras) were determined after scanning Coomassie Blue-stained SDS-PAGE by densitometry (Molecular Dynamics 300 series computing densitometer) and measuring band densities relative to bovine serum albumin standards with ImageQuant 3.3 (Molecular Dynamics). The concentration of purified Ras was determined using the Bio-Rad protein assay dye reagent (Bio-Rad).
In Vitro Ras-binding Assays-[␥-32 P]GTP or GTP␥S labeling of v-Ha-Ras and its binding to GST⅐RBD fusion proteins were performed as described previously (17). The labeled v-Ha-Ras was added to GST⅐RBD constructs linked to glutathione-coated Sepharose beads (Center for Protein and Enzyme Technology, LaTrobe University). Dilutions of the beads were carried out with 200-l pipette tips that had been cut with a scalpel to allow unobstructed movement of the beads. Each sample contained 0.9 M RBD⅐GST and 1 M [␥-32 P]GTP labeled v-Ha-Ras. The reactions were carried out in 200 l of binding buffer (50 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 0.5 mg/ml bovine serum albumin, 0.5 mM dithiothreitol, 0.1 M NaCl), on a rotating wheel, at room temperature for 15-30 min. The beads were washed four times in binding buffer by centrifugation. The radioactivity associated with the beads was measured using a Tri-Carb R liquid scintillation analyzer (Packard United Technologies). The binding between v-Ha-Ras and the different RBD mutants was expressed as a percentage of the radioactivity associated with wt-RBD.
Conditions for the Ras-binding Assay Visualized by SDS-PAGE-Reaction mixtures contained 6.0 M GST⅐RBD linked to Sepharose beads as well as 1.2 M v-Ha-Ras⅐GTP or v-Ha-Ras⅐GDP in 200 l of binding buffer. The reactions were shaken for 15 min at room temperature. The beads were washed three times with ice-cold binding buffer, suspended in 100 l of sample buffer, and heated (95°C for 5 min). Aliquots (50 l) of the supernatants were resolved by 12% SDS-PAGE and stained with Coomassie Blue.
Nucleotide Dissociation Assays-GTP␥S-labeled v-Ha-Ras (2.3 M) was incubated with 1.1 M wt-RBD⅐GST or A85K-RBD⅐GST linked to glutathione-coated Sepharose beads. After a 75-min incubation period, the Sepharose beads were washed three times by centrifugation. Dissociation of the Ras⅐RBD complexes was commenced by adding 287 M GTP␥S. At given intervals, aliquots (60 l) were taken from the dissociation mixture (500 l starting volume) and washed four times by centrifugation. The radioactivity associated with the beads was determined using a Tri-CarbR liquid scintillation analyzer (Packard United Technologies).
Ras Activation in BAF/EGFR Cells-BAF cells, which had been transfected with a human EGF receptor or a kinase negative EGF receptor mutant expression vector (39), were starved of serum and interleukin-3 for 5 h in minimal medium (RPMI). The cells (10 7 cells/ sample) were collected as a pellet after centrifugation (1000 rpm, 5 min, Sigma 4K15 centrifuge), transferred to a 1.5-ml Eppendorf tube, and suspended in 1 ml of serum-free medium. The cells were stimulated with 100 nm/ml EGF (40,41) at room temperature for 3 min. The cells were centrifuged, and the cell pellets were snap frozen on dry ice and stored at Ϫ70°C overnight. The freezing step facilitated cell lysis.
Precipitation of Cellular Ras⅐GTP-A modification of the method described by de Rooij and Bos (28) was used. Frozen NIH-3T3 cells, which had been transfected with v-Ha-Ras (17), or activated BAF/ EGFR cells were lysed in 200 l/sample of ice-cold lysis buffer (0.5% Triton X-100, 0.5% deoxycholate, 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Trasylol, and 1 mM phenylmethylsulfonyl fluoride). Insoluble cellular debris were removed after centrifugation (5 min, 4°C). The supernatant was then added to tubes containing 1 ml of cold binding buffer and 100 g of GST⅐RBD fusion protein linked to Sepharose beads. In positive control samples, total cellular Ras was immunoprecipitated with the anti-Ras Y13-259 antibody (American Type Culture Collection) prebound to protein-G-coated Sepharose beads (20 l). All assay samples were incubated on a rotating wheel (4°C, 30 min). The beads were then washed three times in binding buffer by centrifugation. The samples were subjected to Western blot analysis. Proteins resolved by 15% SDS-PAGE were transferred electrophoretically to a polyvinylidene difluoride membrane (Immobilon, Millipore). RBD-bound Ras was visualized using the Pan-ras (AB-2) antibody (Calbiochem), horseradish peroxidase-linked goat anti-mouse IgG (Bio-Rad) antibody, the enhanced chemiluminescence (ECL) reagent (Amersham Pharmacia Biotech) and Eastman Kodak Co. X-AR x-ray film. Ras signals were quantified by scanning densitometry (Molecular Dynamics 300 Series computing densitometer) with ImageQuant 3.3 software (Molecular Dynamics).
Computer Modeling of the Complexes between the c-Raf-1 RBD Mutants and Ras-The structure of the wild type Ras⅐RBD complex (18) was used to model the complexes between RBD mutants and Ras. The procedure involved two steps. First, Ala 85 was replaced with Lys 85 , Val 88 with Arg 88 , and Asn 64 with Lys 64 . Second, the complexes were energy minimized with the CVFF force field, as implemented in DISCOVER (INSIGHTII, 1999). A 500-step conjugate gradient energy minimization was carried out to optimize position and orientation if side chains of the mutants with the rest of Raf and Ras protein fixed. The nonbonded (electrostatic and van der Waals) interactions were truncated at 10 and a distance-dependent dielectric constant (4r) was used to mimic solvent.

RESULTS
A mutational scan of the c-Raf-1 RBD involved the measurement of in vitro v-Ha-Ras binding to the point mutants of RBD, which were expressed as GST fusion proteins and linked to Sepharose beads. The results of the scan are summarized in Fig. 1. The substitution of alanine in RBD position 85 with arginine resulted in a 2.5-fold increase in binding to v-Ha-Ras (Fig. 1B). Ala 85 is part of an ␣-helix in RBD (Fig. 2). Residues Arg 84 and Arg 89 , which are adjacent to Ala 85 within the helix (Fig. 4A), are part of the Ras/Raf binding interface (1); however, Ala 85 had not been previously implicated in the formation of the Ras⅐Raf complex.
The purity of GST⅐RBD fusion proteins was analyzed by SDS-PAGE and Coomassie Blue staining (Fig. 1A). In some instances the mutations destabilized the structure of RBD making it more susceptible to degradation by Escherichia coli proteases. The degradation products are clearly visible in some of the mutant preparations and are most prominent in the R98V-RBD preparation. The degradation products were excluded from the estimates of GST⅐RBD concentrations, because the protein estimates were based on the densities of the discrete GST⅐RBD bands (see "Experimental Procedures").
The remaining RBD mutants in the binding scan showed reduced binding to v-Ha-Ras (Fig. 1B). Residues Arg 59 , Arg 67 , Lys 84 , and Arg 89 are part of the binding interface between Raf and Ras (1). In previous investigations, substitution of the basic residues in these positions with neutral amino acids similarly reduced the affinity of the RBD for Ras (1,2,42). The substitution of arginine with leucine in position 89 abolished the interaction between Ras and Raf both in vivo and in vitro (see Fig. 1B (1,20)). The affinity of the K84A mutant was reduced by more than 100-fold (1), the binding of R59A-RBD was 3.4% relative to wt-RBD, whereas the binding of R67A was 6.2%. In our hands, mutation of the known interactive RBD residues led to less dramatic reductions in binding than in previous studies (1,19). In Fig. 1B, the binding relative to wt-RBD was 8.6% for K84L-RBD, 43% for R59L-RBD, and 42% for R67L-RBD. The discrepancy in the binding ratios could arise from the differences in the particular neutral residues used. Leu has a longer hydrophobic side chain than Ala, which can alter the RBD-solvent interactions. Furthermore, in the present investigation, the assays were carried out at high protein concentrations, which should minimize the difference between high and low affinity constructs.
Arginine substitutions at positions 91 and 98, which appear to be adjacent to the binding interface, lead to a dramatic loss of binding (Fig. 1B). According to a computer simulation of the Ras⅐RBD complex (18), RBD Leu 91 -RBD forms hydrogen bonds with Arg 41 -Ras and Lys 42 -Ras. The L91R-RBD mutant disrupts this interaction. The side chain of Val 98 -RBD is located within the hydrophobic core of the RBD molecule (43). The V98R-RBD mutant is likely to abolish RBD/Ras binding as a consequence of distorted RBD folding. The Ala 103-111 cluster mutant, in which a 9-residue stretch of random coil was replaced with alanines, also displayed reduced binding to v-Ha-Ras. It must be noted that no steps were taken to monitor correct folding of the RBD mutants. Therefore, misfolding cannot be ruled out as the cause of reduced binding. An attempt was made to purify other arginine RBD mutants, V72R, C81R, L86R, L121R, and V128R, but the protein yields of these constructs were too low to detect in our binding assay.
The Importance of Charge and Molecular Packing in Position 85 of RBD-The replacement of the small, nonpolar alanine side chain in position 85 of RBD with the positively charged, large arginine side chain leads to an increase in binding to the v-Ha-Ras⅐GTP complex (Fig. 1B). Residue Ala 85 is part of the A1 ␣-helix within wt-RBD ( Fig. 2 and 4A). There are no charged residues in this position in the native Raf primary sequences (Fig. 2). No interaction between Ala 85 and Ras residues has been reported to date, but the neighboring residues Lys 84 and Arg 89 appear to be major contributors to the high affinity binding between Ras and RBD (23). It is plausible that an additional positive charge in this critical binding region would enhance the strength of its coulombic interaction with Ras.
Several point mutants in RBD were expressed to test the importance of charge and molecular packing in position 85 to the binding of RBD to Ras. Isoleucine, lysine, arginine, and aspartic acid were used to make a set of position 85 mutants. Again, 12.5% SDS-PAGE and Coomassie Blue staining ( purified mutant RBDs by densitometry. The binding of these mutants to v-Ha-Ras⅐GTP is shown in Fig. 3B. Instead of GST, a truncation mutant Raf-(66 -131)-GST was used as a negative control. This derivative has no binding above background (17). The binding of the A85R-RBD mutant to Ras was 290% of wt-RBD, a similar result to that shown in Fig. 1B. The substitution of another positively charged residue, lysine, into position 85 led to an increase in binding in excess of 7-fold relative to wt-RBD (Fig. 3B). Mutation to a negatively charged aspartic acid interfered with binding, reducing Ras⅐RBD complex formation to 3% of wt-RBD. The substitution of alanine with isoleucine reduced the binding to 18%.
According to a computer generated model of the Ras⅐RBD complex (18), the lysine side chain at position 85 could form a salt bridge with Asp 38 of Ras (Fig. 3C), which also binds Arg 89 of wt-RBD (2). A similar explanation would apply to the effect of the arginine substitution at position 85 of RBD. The longer side chain of arginine may account for the lesser binding of A85R-RBD compared with A85K-RBD. The longer side chain could bring the positive charges from Arg 85 in the vicinity of Arg 89 . If this were to happen, the repulsion between the two adjacent positive charges would distort the orientation of the Arg 89 side chain and, hence, reduce the strength of the critical bond between Arg 89 -RBD and Asp 38 -Ras. The reduction in binding between v-Ha-Ras and A85I-RBD (Fig. 3B) indicates that a large side chain is of itself disruptive in position 85. The loss of binding in the A85D-RBD mutant (Fig. 3B) is also consistent with the computer model of the RBD⅐Ras complex (18), which suggests that a negatively charged residue in position 85 would undermine the bond between Arg 89 -RBD and Asp 38 -Ras through the coulombic repulsion of Asp 38 -Ras.
Point Mutations in the Region of the RBD ␣-Helix-Mutation of Ala 85 in RBD to a basic residue led to increased binding to Ras (Fig. 3A). We examined whether a similar trend could extend to other amino acids in this region of RBD. Ala 85 is part of the A1 ␣-helix (Fig. 2), which is illustrated schematically in Fig. 4A. The spiraling structure of the helix brings together residues that are separated within the primary sequence. Thus Cys 81 , Leu 82 , Lys 84 , Leu 86 , and Val 88 are adjacent to Ala 85 . The diagram of the RBD ␣-helix (Fig. 4A) suggests that Val 88 is ideally placed to mimic the increased Ras binding of Ala 85 mutants. Just like Ala 85 , Val 88 is a small nonpolar residue and does not appear to be directly involved in binding Ras. Its closest neighbors are in positions 85 and 84, which, in mutated and native forms, respectively, formed part of the RBD binding interface. The substitution of valine at position 88 with arginine results in a 4-fold increase in binding to v-Ha-Ras (Fig.  4C). Other basic residues at position 88 produce a lower, though still significant increases of 71% for V88K and 166% for V88H. As expected, a negative charge in this position interferes with complex formation, just as it does at position 85 (Fig. 3C); the binding of V88D-RBD is less than 10% the wt-RBD binding (Fig. 4C). A computer simulation of the RBD⅐Ras complex (18), based on the Rap1A⅐RBD crystal structure (2), suggested that the side chains of residues 88 and 85 point in different directions. Asp 30 and Glu 31 in Ras are the most likely candidates for binding Arg 88 -RBD (Fig. 4D); the same residues that may form salt bridges with Lys 84 in wt-RBD (23).
We also constructed, expressed, and purified a double mutant, A85K/V88R-RBD. The double mutant incorporated both mutations, which individually elevate RBD binding to Ras. Fig.  4C shows that the effect of the two mutations was not cumulative. The increase in binding was greater than for the V88R mutation alone, but less than for the A85K mutation (Fig. 4C). The effect of simultaneous substitutions in positions 88 and 85 is difficult to interpret from the computer model of the Ras⅐RBD complex (18).
Mutation of other positions of the ␣-helix to basic residues either reduced or had no effect on the binding of RBD to Ras (Fig. 4C). Lys 84 is one of three RBD residues responsible for high affinity binding with Ras (1). In this position, not only the positive charge, but also the lysine side chain per se is important to the interaction with Ras. The ⑀-amino group forms salt bridges with two Ras residues, Glu 31 and Asp 33 (44). A conservative substitution to arginine led to a 2-fold decline in binding (49%). In contrast, the same change in position 87 had no effect; no difference was detected in the binding of Lys 87 -RBD and K87R-RBD. Deletion of the positive charge from position 87 reduced binding to 30% in K87Q-RBD. Reversing the charge in the D80K-RBD mutant had a relatively small impact, retaining 57% of binding. This residue is exposed to the solvent and it forms no direct contacts with Ras (18).
Arginine mutations in two positions adjoining Ala 85 , Cys 81 , and Leu 86 , resulted in low protein yields, indicating that the mutant proteins were either unstable or misfolded, presumably making them more sensitive to bacterial proteolytic enzymes. The side chains of both Cys 81 and Leu 86 are buried in the hydrophobic interior of RBD (18), and charged substitutions in these positions are likely to interfere with the folding of RBD.
Positive charges just outside the helix, at position 90, also interfered with binding. The activity of the G90K mutant was 29% of wt-RBD (Figs. 3 and 4C). In the Ras⅐RBD complex, Gly 90 is located in the vicinity of three basic Ras residues Arg 41 -Ras, Lys 42 -Ras, and His 27 -Ras. The insertion of another positive charge into this environment from G90K-RBD would be expected to repel the Ras residues and distort their orientation. Negatively charged G90D-RBD had the least impact (80%), whereas the binding of G90Q-RBD was 40% of wt-RBD.
Mutations in B1 and B2 ␤-Sheets of RBD-The residues of RBD that interact with Ras (see Fig. 2) are located in the A1 ␣-helix and the B1 and B2 ␤-sheets (2). Our results in Figs Fig. 5A. The substitution of asparagine in position 64 by lysine increased binding to Ras 3.8-fold (Fig. 5B).
Asn 64 -RBD is part of the Ras binding site, forming three hydrogen bonds with Arg 41 -Ras (43). One bond is formed between the carbonyl oxygen of Asn 65 -RBD and the N⑀ of the Arg 41 -Ras, the second bond is between the same Ras nitrogen and the O␦ of Asn 64 -RBD, and the third is between the O␦ of Asn 64 -RBD and the guanidium group of Arg 41 -Ras (2,23,44). However, the association is relatively weak, as these hydrogen bonds are solvent exposed (23). A mutation to N64A-RBD was reported to have no effect on the Ras-RBD interaction (19). Another study (1) reported that under physiological ionic conditions over 20% of wt-RBD binding was retained by the N64A-RBD mutant.
The N64K-RBD mutant might be expected to disrupt the interaction with Arg 41 -Ras. However, the binding between v-Ha-Ras and N64K-RBD is increased (Fig. 5B). Other Ras residues, such as Glu 3 and Asp 54 of Ras (18), could form a direct salt bridge N64K-RBD (Fig. 5C).
The insertion of the negatively charged aspartic acid into position 64 reduced binding by 50% in our scan ( Fig. 3 and 6B). The same mutation has been reported to reduce the affinity between Ras and RBD by a factor of 14 (1). The results from the two studies are not directly comparable because the relative binding was measured at a high concentration of Raf (0.9 m) in our scan and cannot be compared directly to the measure of affinity reported by Block et al. (1). Furthermore, the reactions in Block's study were carried out under conditions of low ionic strength, which increases the affinity of RBD for Ras (1).
Several other RBD mutants were included in the binding scan in Fig. 5. Their association with v-Ha-Ras was as follows: 9% for Q66K, 11% for T68K, 29% for V69R, 34% for N71E, and 60% for S77R. Residues Gln 66 , Thr 68 , and Val 69 are part of the Ras binding site (1,2). Alanine point mutants of these residues were also made in an earlier study (1), which reported reductions in affinity by factors of 55, 10, and 7, respectively. Once again, the figures from the two studies are not directly comparable because of the difference in the experimental conditions likely salt bridge formed between the V88R-RBD mutant and Glu 31 -Ras. The mutant side chain is shown in yellow.  (43). B, RBD protein preparations resolved on a 12.5% SDS-PAGE and stained with Coomassie Blue. The position for RBD⅐GST is marked. C, the binding of Raf-RBD mutants to v-Ha-Ras⅐GTP, expressed as a percentage of wt-RBD binding. GST was used as a negative control, and A85K-RBD set an upper limit for Ras binding, ensuring that bindable Ras⅐GTP was not depleted below that level. The mutants with increased binding to Ras were assayed at least three times. D, illustration of the during the measurements of relative binding and affinity.
Ras⅐GDP Binding and GTP Exchange-Wt-RBD preferentially binds active, GTP-bound Ras (29). We tested whether two of the mutants with elevated binding to Ras, V88R-RBD and A85K-RBD, retained the preference for GTP-bound Ras. The association between RBD and Ras⅐GDP or Ras⅐GTP was visualized by SDS-PAGE (Fig. 6A). GTP or GDP-bound v-Ha-Ras was prepared in a nucleotide exchange reaction (see "Experimental Procedures") containing 4.8 mM GTP or GDP in 1.2 ml of exchange buffer. The results (Fig. 6A) clearly show that wt-RBD, V88R-RBD, and A85K-RBD show preferential binding to GTP-bound v-Ha-Ras. No v-Ha-Ras⅐GDP binding was detected. The highest amount of v-Ha-Ras⅐GTP was precipitated with the Ala 85 -RBD construct, less with V88R-RBD, less still with wt-RBD, and none with R89L-RBD. The order of binding corresponds to the results of the binding scan (Fig. 3B). The other bands visible on the gel are the degradation products from RBD⅐GST.
To allay concerns about different rates of GTP dissociation from v-Ha-Ras in the presence of wt-RBD or RBD mutants with elevated binding, we compared the effect of wt-RBD and A85K-RBD on the dissociation of GTP␥S (a nonhydrolyzable analogue of GTP) from v-Ha-Ras. GTP␥S dissociation from the v-Ha-Ras⅐GTP␥S complex in the presence of excess unlabeled GTP␥S is shown in Fig. 6B. There is no difference in GTP␥S exchange between the samples containing wt-RBD or A85K-RBD, indicating that the increased binding of the A85K-RBD mutant to Ras is not a consequence of reduced GTP dissociation.
Binding of RBD Mutants to Eukaryotically Expressed v-Ha-Ras-One concern about the binding scan, which uses recombinant Ras, is its relevance to physiological binding. In bacteria, Ras proteins are not subjected to post-translational processing, whereas in mammalian cells post-translational modifications appear to be essential for the biological activity of Ras (45)(46)(47). These modifications involve farnesylation (48), palmitoylation (49,50), cysteine methylation (51,52), and proteolysis of the AAX amino acids from the C-terminal CAAX motif (51, 53) (C, cysteine; A, aliphatic; X, any amino acid). The post-translational modifications are responsible for the localization of Ras at the plasma membrane (51,54), where it associates with an array of downstream effectors (6,7).
We tested the binding of some of the RBD mutants to Ras produced in mammalian cells. RBD⅐GST fusion protein Sepharose beads were used to capture Ras from the lysates of v-Ha-Ras-transfected NIH-3T3 cells. v-Ha-Ras carries two oncogenic mutations, G12V and A59T, which reduce the GTPase activity of v-Ha-Ras (55) and render it resistant to stimulation by FIG. 5. Mutations in B1 and B2 ␤-sheets of RBD. A, RBD protein preparations resolved on a reducing 12.5% SDS-PAGE and stained with Coomassie Blue. The positions for RBD⅐GST and GST are marked. Protein concentrations were estimated by densitometry relative to standard bovine serum albumin concentrations. B, the binding of RBD mutants to v-Ha-Ras⅐GTP, expressed as a percentage of wt-RBD binding. GST was used as a negative control, and A85K-RBD set an upper limit for Ras binding, ensuring that bindable Ras⅐GTP was not depleted below that level. The mutants with increased binding to Ras were assayed at least twice. C, illustration of the likely salt bridge formed between the N64K-RBD mutant and Glu 3 -Ras. The mutant side chain is shown in blue. Arg 41 -Ras also forms salt bridges to Glu 3 -Ras and Asp 54 -Ras. See "Results" for details. GTPase-activating protein (56). In addition, the A59T mutation increases the exchange of GDP for GTP (57,58). As a result, v-Ha-Ras exists predominantly in the active GTP-bound form, which is able to bind RBD. The binding of RBD mutants to v-Ha-Ras expressed in NIH-3T3 cells is shown in Fig. 7A. Following incubation of cell lysates with the Sepharose-linked GST⅐RBD fusion proteins, bound Ras was visualized by Western blotting. A85K-RBD bound nearly 6-fold more v-Ha-Ras than wt-RBD, whereas the increase for V88R-RBD was over 4-fold. No Ras associated with the inactive form of RBD, R89L-RBD (20). Fig. 7B shows the relative amounts of v-Ha-Ras associated with RBD-coated Sepharose beads or remaining in the cell extracts after incubation with the beads. In comparison with the negative control (R89L-RBD), most of the Ras is removed from the samples after incubation with A85K-RBD; less is removed from the extracts by wt-RBD-coated beads. The total amounts of v-Ha-Ras recovered with the wt-and A85K-RBD-coated beads are less than the amounts of Ras lost from the corresponding extracts. It is likely that some of the v-Ha-Ras dissociated from RBD during the washing steps.
The Use of A85K-RBD as an Activation Specific Probe for Ras-The elevated binding between Ras and A85K-RBD suggested that the mutant protein would be more effective than wt-RBD as a probe for detecting and quantifying Ras activation in cells (28). The effectiveness of A85K-RBD as an activation specific probe for Ras⅐GTP was tested in BAF/EGFR cells that had been transfected with the wild type EGF receptor (wt) or an inactive, kinase negative mutant of the EGF receptor (K721R) (39). Acting through a cascade of intermediary signaling proteins, the binding between EGF and its receptors indirectly leads to a transient association of Ras with GTP (59 -61). Serum-starved cells were stimulated with EGF, lysed, and incubated with the different forms of the RBD⅐GST fusion protein linked to Sepharose beads. Association of Ras with the beads was analyzed by Western blotting. The results in Fig. 8 show that A85K-RBD was 5-fold more effective than wt-RBD for detecting Ras activation. (lanes 3 and 4). No bands corresponding to activated Ras were detected in the serum-starved cells, in the absence of EGF stimulation (lanes 1 and 2), indicating that neither wt-RBD nor K85A-RBD bind to Ras in its inactive GDP form. In quiescent cells Ras remains in a complex with GDP (41), incapable of binding wt-RBD (8,9,15,62). Activated Ras was barely detectable in control BAF/EGFR cells that had been transfected with K721R-EGFR, the kinase negative EGF receptor (lanes [5][6][7][8]. The difference in the signals from wt and K721R-transfected cells served to confirm that Ras activation was largely the result of stimulating the EGF receptor tyrosine kinase. DISCUSSION A mutational analysis of the Ras-binding domain of c-Raf-1 identified three amino acid positions (Asn 64 , Ala 85 , and Val 88 ) where substitution with basic residues increases the binding between RBD and bacterial recombinant v-Ha-Ras. The greatest increase in binding was observed with the A85K-RBD mutant (Figs. [3][4][5]. The elevated binding to the A85K-RBD and V88R-RBD mutants occurs with Ras expressed in mammalian cells, namely NIH-3T3 (Fig. 7) or BAF cells (Fig. 8).
Like wt-RBD, the mutants associate preferentially with GTP-bound Ras. There was no binding to Ras⅐GDP for any of the RBD constructs (Fig. 6A) nor was any Ras precipitated from unstimulated BAF cells (Fig. 8). Whereas Moodie et al. (63) described a mutation of Ras, Q61L, which rendered the Q61L-Ras⅐GDP complex capable of binding c-Raf-1, B-Raf, and phosphatidylinositol 3-kinase, there have been no previous reports of Ras effector mutants associating with the Ras⅐GDP complex.
The increase in the binding of RBD mutants to v-Ha-Ras is not related to GTP exchange or hydrolysis. The rate of GTP exchange is the same in the presence of wt-RBD or A85K-RBD (Fig. 6B). There is contradictory evidence in the literature about the GTPase activity of c-Raf-1. Warne et al. (9) reported a weak increase of about 20% in the GTPase stimulating activity of Ras in the presence of an N-terminal Raf fragment (residues 1-257). Subsequent studies showed that c-Raf-1 Nterminal truncation mutants (including wt-RBD) have no impact on the GTPase activity of Ras (8,15,29,64). In our assays, a 20% reduction in GTP hydrolysis would not account for the 200 -900% increases in the binding of RBD mutants to Ras. Furthermore, because v-Ha-Ras is an oncogenic variant of Ras that is resistant to GTPase-activating protein activation (65), GTP hydrolysis is unlikely to play a role in the binding scans or in Ras precipitation from v-Ha-Ras-transfected NIH-3T3 cells. In addition, the increased binding of the A85K-RBD mutant was subsequently confirmed using a nonhydrolyzable GTP analogue, GTP␥S. 2 The binding between the remaining RBD mutants and v-Ha-Ras was reduced or unaltered. Mutation of residues that form the binding interface with Ras resulted in reduced binding, in agreement with previous observations (1,2,42). In the Ala 103-111 cluster mutant a random coil segment was replaced with alanines. A shorter stretch of sequence was previously mutated in the same region (19); the substitution with alanines of amino acids 104 -106 and 108 -111 resulted in 76 and 56% binding, respectively. Our cluster mutant covered two additional amino acids, His 103 and Gly 107 . His 103 is not likely to be of significance because it is present in only two known Raf sequences (Fig. 2). In contrast, Gly 107 is highly conserved. Glycine can adopt a wider range of conformations than other residues allowing the main chain of proteins to bend more freely. The twist in the loop joining RBD ␤-sheets 3 and 4 on the computer that generated a model of the RBD⅐Ras complex (18) suggests that Gly 107 may form a glycine turn.
It is likely that more RBD positions could be found where the substitution of wild type residues would enhance the binding to Ras. A random mutagenesis approach may help in identifying such positions. A combination of such mutations in a single RBD molecule could possibly lead to further increases in Ras binding. The binding of one double mutant, A85K/V88R-RBD is shown in Fig. 4C. In this instance, no cumulative increase in binding was evident. The lack of cumulative binding in the A85K/V88R-RBD protein could be because of the repulsion between the adjacent positively charged mutant residues destabilizing the ␣-helix or because of unfavorable interactions between the combination of these residues and Ras. It would be interesting to try a combination of mutants where the positively charged side chains are separated by a greater distance e.g. N64K/A85K-RBD or N64K/V88R-RBD.
It is interesting to note that in every RBD mutant with elevated binding to Ras, the residues responsible for the increases in binding are absent from the native Raf orthologues (Fig. 2). The RBD sequence alignment of in Fig. 2 shows only neutral and predominantly small amino acids corresponding with position 85 of c-Raf-1; alanines predominate, but there is also a serine in Xenopus c-Raf-1 and a leucine in LIN-45 from Caenorhabditis elegans. The conservation of uncharged amino acids in position 85 suggests a functional significance for these residues, even though Ala 85 has not been implicated in the RBD binding site (1,2). Clearly, these neutral residues are not required for the maintenance of the RBD structure, because their replacement with lysine or arginine leads to elevated binding (Fig. 3C). Despite their positive contribution to binding, there is no instance of a basic residue in the RBD positions aligned with Ala 85 of human c-Raf-1 (see Fig. 2). Similarly, there is no instance of a basic amino acid in position 64 of c-Raf (Fig. 2). Asparagine is highly conserved in all cases, with the exception of Lin-45 from C. elegans, where it is replaced by phenylalanine, a nonpolar aromatic residue.
Similarly, with the exception of the C. elegans Raf analogue, Lin-45, there are no positively charged residues aligned with position 88 of c-Raf. Lin-45 contains lysine in the aligned position. Of the three basic residues tested, lysine had the least impact in position 88, increasing RBD binding to Ras to only 170%, compared with 270% for histidine and 410% for arginine. Furthermore, the Ras binding site of Lin-45 may not be identical to RBD, because out of eight highly conserved interactive residues (Arg 59 , Asn 64 (1)) are preserved in Lin-45 (Fig. 3A). The Lin-45 lysine aligned with c-Raf position 88 might be expected to compensate for the possible absence of the other favorable interactive sites usually provided by positions equivalent to 59, 64, 67, 68, and 69 of c-Raf-1.
It is interesting to speculate that evolutionary selection may be responsible for the absence of basic residues from the RBD positions where such residues would result in elevated Ras binding. If such an evolutionary pressure did exist, it would imply that elevated binding to Ras is in some way suboptimal for biological function. Geyer et al. (66) and Ito et al. (67) showed that Ras exists in the form of multiple conformational isomers. Both reports postulated that different Ras isomers may function to discriminate among multiple Ras effectors for binding.
In the present experiments the concentrations of Ras and RBD are an order of magnitude higher than the reported K d for the interaction (15-160 nM (1, 29, 64, 68 -72)). Therefore any difference in binding is not expected to reflect a difference in affinity. A detailed kinetic analysis of the v-Ha-Ras-RBD interaction is presented in a separate paper, 2 where we show v-Ha-Ras exists as a population of isomeric conformers and that A85K-RBD binds a greater proportion of these conformers than wt-RBD, giving rise to the difference in Ras saturation levels.
The ability of Raf mutants to bind a wide range of conformations may impede appropriate interaction with other effectors (6,7), resulting in evolutionary pressures against such mutants. The finding that the binding interface of Ras can accommodate several more favorable interactions than are found in wild type RBD is not entirely surprising. Two mutants of another Ras-binding protein, NF1, were found to bind 5-10-fold more Ras than wild type NF1 (73). Furthermore, random mutagenesis of yeast Ras2 uncovered 57 point mutants that displayed increased binding to Raf (74). Until now, no corresponding Raf mutants have been described.
The most immediate application for the A85K-RBD mutant is as an activation-specific probe for Ras⅐GTP. GST⅐RBD fusion protein has been used as a probe for endogenous activated Ras in insulin-stimulated A14 cells and glial-derived neurotrophic factor-stimulated SK-P2 cells (28). A85K-RBD increases the sensitivity of the assay 4 -6-fold. De Rooij and Bos (28) speculated that such probes have potential applications in research and clinical diagnosis. Ras mediates signals from a myriad of cellular stimuli, and a sensitive simple method for measuring FIG. 8. Ras activation. Western blot analysis showing Ras activation in EGF stimulated (ϩ) or unstimulated (Ϫ) BAF/ EGFR cells (39). The cells had been transfected with the wild type EGF receptor (BAF-WT) or a kinase negative EGF receptor mutant (BAF-K721R).
its activation state would help clarify its role in tumor cells or signal transduction.
Around 30% of human cancers carry oncogenic mutations in Ras (32,75,76). However, the aberrant Ras activation in the incidence of cancer could be even more pervasive. Activating mutations in signaling proteins upstream of Ras may elevate the level of Ras⅐GTP even in the absence of oncogenic mutations in the Ras molecule itself. It may be possible to use an A85K⅐RBD fusion protein as an activation-specific probe to test tumor samples for the presence of activated Ras⅐GTP.
In vivo, RBD has been shown to suppress the transformed phenotype in v-Ha-Ras-transfected NIH-3T3 cells (17), suggesting that this molecule acts as a Ras antagonist. In the present report we have described six RBD mutants with elevated binding to recombinant and cellular Ras. The increased binding of RBD mutants to Ras⅐GTP could increase their potential as Ras antagonists, hence as anti-cancer drugs. Such molecules need to be effective not only in their role as Ras antagonists but also in their ability to penetrate cells and resist degradation or excretion. The additional interactions generated by the Arg 64 , Lys 85 , or Arg 88 mutations would expand the range of possible molecules that fulfill these criteria and increase the chance of finding a useful therapeutic drug based on the RBD prototype.