p120 Ras GTPase-activating Protein Interacts with Ras-GTP through Specific Conserved Residues*

Previous structural studies of RasGAP have failed to clearly localize sites of Ras interaction to individual amino acids. Hypothesizing that sites of interaction with Ras-GTP would be conserved, 11 of the most highly conserved amino acid residues of RasGAP were changed by mutation. Each mutant protein was purified as a glutathione S -transferase catalytic domain fusion and analyzed for protein stability, Ras GTPase stimulating activity, affinity for Ras-GTP, and when possible, sec- ondary structure. The majority of conserved positions were found to be important structurally but with no direct role in Ras interactions. However, Arg 786 , Lys 831 , and Arg 925 were observed to be essential for binding to Ras-GTP but not for protein structure. RasGAP residues 890–902 (block 3A) were observed to be homologous to residues 1540–1552 of the yeast adenylyl cyclase with amino acid substitutions in both regions resulting in increased affinity for Ras. This is the first example of a conserved Ras interaction motif in distinct Ras effector proteins. Our data are supportive of a model for GAP/ Ras-GTP association in which the conserved, positively charged Arg 786 , Lys 831 , and Arg 925 residues form salt bridges with the conserved, negatively charged residues in the Ras effector loop.

Ras is a ubiquitous GTPase that plays a critical role in cell growth, division, and differentiation as a signal transducer (1,2). As a GTPase, the GTP bound form of Ras is active and able to transduce signals from surface transmembrane receptors to the Raf protein kinase, resulting in the activation of mitogenactivated protein kinase pathway (1,3). Ras possesses a low intrinsic GTPase activity (K cat ϭ 1.2 ϫ 10 Ϫ4 s Ϫ1 ), which can slowly hydrolyze the bound GTP into GDP. The receptor-initiated signal is terminated through the action of the Ras GTPase-activating protein (RasGAP), 1 which stimulates a 10 5fold rate increase of the Ras GTPase, resulting in rapid formation of a Ras-GDP complex (4). Ras interacts with GAP through switch I, switch II, and probably a small region centered around residue Asp 92 (2,5). The switch I and II regions of Ras are affected conformationally by the presence of GTP (6,7). Switch I, also known as the effector region, includes Ras residues 32-38. This region also mediates binding to the Raf pro-tein kinase. Mutations in switch I can have varied effects on Ras function, including loss of biological function, insensitivity to RasGAP stimulation, and loss of binding to the Raf and RasGAP proteins (8 -11). The switch II region includes residues 60 -76 and has a more flexible conformation than switch I. It has been proposed that the switch II region is responsible for Ras GTPase catalysis (12,13). Mutations in this region often result in a loss of Ras GTPase activity and GAP sensitivity but retain normal or higher binding affinity for RasGAP (4). The region around Ras residue aspartate 92 has also been identified to be important for RasGAP binding (5).
RasGAP is a 120-kDa protein with multiple functional domains (14). The carboxyl-terminal 343 amino acids (702-1044) of the protein constitute the catalytic domain. This domain can bind to Ras-GTP and stimulate Ras GTPase activity, although not as effectively as full-length RasGAP (4,15). Other proteins capable of stimulating the Ras GTPase have been identified in vertebrates and lower eukaryotes (16). These proteins all have high sequence homology to the RasGAP catalytic domain. Based upon the similar sequences of these RasGAP homologues, four highly conserved regions have been identified in this domain: block 1, block 2, block 3A, and block 3B (17,18). Most research intended to identify sites of Ras interaction on RasGAP or its homologues has been focused on block 3A and 3B because of the identification of mutations in these regions of neurofibromin in tumor cells (19). Mutations affecting lysine 1423 (lysine 932 in RasGAP) in block 3B of neurofibromin were found to be responsible for some cases of neurofibromatosis (19). This position is conserved among RasGAP family members and has been reported to be required for both protein stability and Ras protein association (19 -21).
Mutagenesis studies of Ras have already uncovered many important residues involved in RasGAP binding. However, the corresponding residues on RasGAP needed for Ras binding and the mechanism of this interaction remain unclear. Prior mutagenic studies of RasGAP and neurofibromin have demonstrated the functional importance of blocks 1, 3A, and 3B while failing to identify a distinct region for Ras interaction (17,18). Alignment of all known RasGAP-related protein sequences reveals approximately 17 residues that are highly conserved (17,18,22,23). Although some of them have already been analyzed for functional importance, the majority have not. It was hypothesized that sites of protein interaction would remain conserved. In this study, we mutated 11 of the conserved residues in the RasGAP catalytic domain and examined each mutant protein for changes in secondary structure, catalytic properties, and Ras binding. Additionally, block 3A of RasGAP was found to be similar in sequence to a site previously described in yeast adenylyl cyclase as affecting Ras binding (24). A mutation analogous to the adenylyl cyclase SSR2-1 mutation was introduced into RasGAP and found to have a similar functional effect. The biochemical analysis of these mutants and consid-eration of previously characterized Ras mutations suggests that discrete sites of interaction exist on RasGAP that interact separately with the Ras switch I and II regions, affecting both protein binding and GTPase activation.
Site-directed Mutagenesis-Mutations were made in the full-length cDNA of bovine RasGAP using the Promega Altered Sites in vitro mutagenesis system (25,26). Mutations and clone integrity were confirmed by DNA sequencing with Sequenase (U. S. Biochemical Corp.).
Expression and Purification of GST-GAP Fusion Proteins in Escherichia coli RR1lac iq -The 1-kilobase pair BamHI-EcoRI fragments encoding wild type and mutant bovine RasGAP[702-1044] (known as GAP-C) were cloned into the corresponding sites of the pGEX-2T expression vector. The mutations were reconfirmed by DNA sequencing. Using these constructs, the wild type and mutant GAP-C proteins were expressed in the RR1lac iq E. coli strain as glutathione S-transferase (GST) fusion proteins. The freshly transformed cells were grown in 1 liter of Luria Broth with 125 g/ml ampicillin (LBA) at 37°C to an A 600 of 0.4 -0.8 and induced with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside. The induced cells were grown at 30°C for an additional 3 h. Cell were pelleted, washed once with TED buffer (50 mM Tris-HCl, pH 7.5, 1 mM EGTA, 5 mM dithiothreitol) and frozen in Ϫ80°C. The purification of the GST fusion proteins was as described previously (10). Protein concentrations were determined by Bradford assays and verified by band intensities on SDS-polyacrylamide gels stained with Coomassie Blue.
Ras Protein Expression and Purification-As described previously (10), wild type human H-Ras, H-Ras[⌬34A38], and H-Ras[L61] were expressed in RR1lac iq E. coli. Ras proteins were purified from the cell lysate by fast protein liquid chromatography over a DEAE-Sepharose column followed by G-75 superdex column (Pharmacia). The protein concentration was determined by Bradford assay and band intensities on SDS-polyacrylamide gels stained with Coomassie Blue.
Purification of Nonfusion GAP-C by Thrombin Cleavage of GST Fusion Proteins-In order to purify the GAP-C catalytic fragments, GST-GAP-C fusion proteins were expressed as described above. The cells were lysed, and the GST fusion proteins were bound to glutathione-Sepharose 4B. The beads were washed once with a 10ϫ bed volume of high salt TED buffer (TED buffer with 1 M NaCl) and three times with a 10ϫ bed volume of TED buffer. The beads were then incubated at room temperature overnight with an equal volume of TED buffer plus 5 mM CaCl 2 and thrombin (60 units). When the cleavage reaction was completed, the reaction mixture was centrifuged at 500 ϫ g for 5 min, and the supernatant was collected. The beads were washed by an equal volume of TED buffer three times and high salt buffer twice. All the supernatants were collected and analyzed by SDS-PAGE. All fractions containing the free 39-kDa GAP-C fragment were pooled and concentrated with a Centricon-10 microconcentrator. The protein concentrations were determined by Bradford assay and Coomassie Bluestained SDS-polyacrylamide gels.
H-Ras GTP Charging Reactions-The method used for charging H-Ras with GTP has been previously described (10). When charging Ras with [␥-32 P]GTP, 1.33 nM H-Ras and 88.9 nM [␥-32 P]GTP were incubated in 100 l of 20 mM NaHEPES, pH 7.5 with 1 mM MgCl 2 , 2 mM EDTA, 2 mM dithiothreitol, 100 g/ml BSA in 30°C for 20 min. The reaction was quenched by adding 300 l of ice-cold 20 mM NaHEPES buffer, pH 7.5, 1 mM MgCl 2 . The charged Ras proteins were separated from free GTP on a PD-10 column (Pharmacia) and kept on ice for GAP assays for no more than 8 h. When charging Ras with nonradioactive GTP, wild type H-Ras, H-Ras[⌬34A38], or H-Ras[L61] were incubated at 30°C for 20 min with 2 mM GTP, 1 mM dithiothreitol, 2.2 mM EDTA in 50 mM Tris-HCl buffer, pH 7.5. The reaction mix was then chilled on ice and concentrated in Centricon-10 concentrators. Free guanine nucleotide was removed by three additional concentration steps with a Centricon-10 with 20 mM NaHEPES, pH 7.5, 1 mM MgCl 2 . The concentrations of the charged Ras proteins were determined by Bradford assay and verified by Coomassie Blue-stained polyacrylamide gels. All charged Ras proteins were stored at Ϫ80°C and thawed prior to each experiment.
Serial Dilution Assays of Ras GTPase Stimulating Activities Using Purified GAP Proteins-The GAP activity of each purified GST-GAP fusion protein or GAP-C protein was determined by serial dilution assay. 100 l of [␥-32 P]GTP charged H-Ras protein (less than 0.1 nM Ras-GTP final concentration) were mixed with 420 l of 2ϫ reaction buffer (40 mM NaHEPES, 2 mM MgCl 2 ), 100 l of BSA (10 mg/ml), and 180 l of double-distilled H 2 O. Serial 1:2 dilutions of each GAP protein were made and kept on ice. 45 l of Ras mix were prewarmed at 30°C for 2 min followed by the addition of 5 l of diluted GAP proteins. The reactions were incubated at 30°C for 10 min and quenched by ice-cold 5% charcoal in 50 mM phosphate buffer, pH 8.0. The reaction mixtures were centrifuged briefly to pellet the charcoal, the supernatants were collected, and the hydrolyzed [ 32 P]phosphate was detected by Cherenkof counting in a Beckman LS 3801 scintillation counter. GAP Assays Using Cell Extracts-For mutants with very low expression levels, GAP assays were performed using crude cell extracts. Freshly transformed RR1lac iq E. coli were grown in 5 ml of LBA at 37°C to an A 560 of 0.6. Protein expression was induced by the addition of 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside and incubation at 30°C for an additional 3 h. The cells were pelleted and frozen at Ϫ80°C. The cells were then lysed by sonication in 260 l of lysis buffer (as above) followed by the addition of 0.1% triton X-100,10 g/ml DNase I, and 10 mM MgCl 2 . The cell lysates were then centrifuged at 14,000 rpm for 10 min. The supernatants were collected and kept on ice for GAP assays and Western blotting. GAP assays were performed as above except that 10 l of cell extract were added to 40 l of prewarmed Ras mix.
Kinetic Assays for the GST-GAP Mutants with Altered Activity-Kinetic assays were used to determine the catalytic activity (K cat ) and binding affinity (K m ) for Ras-GTP of wild type GST-GAP-C, and the K831, Q935 and V897 mutant proteins. In these assays, Ras-GTP was used as the substrate and GST-GAP-C as the enzyme. Different amounts of nonradioactive Ras-GTP (from 0.012 to 123.4 M) were mixed with [␥-32 P]GTP charged Ras (about 0.8 nM) in order to lower the specific activity of radioactive RasGTP and render a higher concentration of substrate over enzyme. Ras-GTP substrate was prepared in 20 mM NaHEPES buffer, pH 7.5, 1 mM MgCl 2 prewarmed at 30°C for 2 min followed by addition of GST-GAP-C proteins. The reactions were allowed to proceed at 30°C. Samples of each reaction were taken at 30 s, 1 min, 2 min, and 5 min and quenched with 5% ice-cold charcoal in 50 mM phosphate buffer, pH 8.0. The reaction mixtures were centrifuged briefly to pellet the charcoal, the supernatants were collected, and the hydrolyzed [ 32 P]phosphate was detected by Cherenkof counting.
Chemical Coupling of Glutathione to BSA-To a 10 mg/ml solution of BSA (Sigma) in degassed 50 mM sodium phosphate, pH 8.0, was added 12 mg/ml m-maleimidobenzoyl-N-hydroxysulfosuccinimide ester (Pierce). The pH was readjusted to 8.0 with 1 N NaOH, and the solution was allowed to stand at room temperature for 1 h. The reaction solution was buffer exchanged into degassed 50 mM sodium phosphate, pH 6.5, over a column of Sephadex G-25. 12 mg/ml of reduced glutathione was added, and the pH was readjusted to 6.5 with 1 N NaOH. The reaction mixture was allowed to stand in the dark at room temperature for 16 h. The solution was again buffer exchanged into 50 mM sodium phosphate buffer, pH 6.5, over a Sephadex G-25 column. The final pH was adjusted to 7.0, filter sterilized, and stored at 4°C.
Detection of Binding between GST-GAP-C and Ras-GTP by ELISA-Immunochemistry was used to measure stable complex formation between GST-GAP-C and Ras-GTP. Immulon 4 microtiter wells (Dynatech) were coated with approximately 10 g of glutathione-conjugated BSA in Dulbecco's phosphate-buffered saline (Pierce) at 37°C for 2 h. The uncoupled reactive residues on the surface of the wells were blocked with 250 l of a 5% Carnation powdered milk in TBS buffer (0.05% Tween-20, 50 mM Tris, pH 7.5, 150 mM NaCl, 0.01% Thimerosal) for 1 h at 37°C followed by three washes with TBS buffer. GST-GAP-C proteins (84 g/ml) or GST (35 g/ml) in TBS buffer with 1% ELISA grade BSA (Sigma) were added and incubated at 37°C for 1 h. After binding GST-GAP-C to the glutathione on the surface, the wells were washed with HEPES buffer (20 mM NaHEPES, pH 7.5, 0.05% Tween-20, 1 mM MgCl 2 , 0.01% Thimerosal) and incubated with 50 g/ml Ras[L61]GTP or 200 g/ml Ras[⌬34A38]GTP in HEPES buffer with 0.1% ELISA grade BSA at room temperature for 30 min. The wells were washed with HEPES buffer and incubated with a 1:1000 dilution of Pan Ras antibody (Oncogene Sciences) for 1 h at 37°C. After washing with HEPES buffer, 100 l of a 1:500 dilution of anti-mouse horseradish peroxidase (Amersham Corp.) was added for 1 h at 37°C. The wells were washed with HEPES buffer, and 100 l of 1-step TM ABTS substrate (Pierce) was added to each well and allowed to react at room temperature for 30 min. Color reactions were stopped by the addition of 50 l of 1% SDS followed by measuring the absorbance at 405 nM. The amount of solid phase GST-GAP-C protein used in the assay was confirmed by removing the bound protein from parallel wells with boiling SDS-PAGE sample buffer and quantitation by Coomassie Bluestained SDS-PAGE gels.
Circular Dichroism Spectropolarimetry-Circular dichroism analyses were performed to study the secondary structure of wild type and mutant GAP catalytic domains. GAP-C proteins purified after thrombin cleavage were concentrated and exchanged into 25 mM phosphate buffer, pH 7.4, 1 mM MgCl 2 using Centricon-10 concentrators (Amicon). The protein concentrations were determined by Bradford assays and SDS-polyacrylamide gels stained with Coomassie Blue. A Jasco J-720 spectropolarimeter (Japan Spectroscopic Co.) was used to measure the 190 -250-nm spectra of 0.1 mg/ml GAP-C protein samples in 0.2-cm path length cuvettes at room temperature. The scanning speed was 20 nm/min. Each spectrum was an average of three scans. The resolution of the scans was 0.5 nm, and the sensitivity was 200 mdeg.
Data Processing-For GAP kinetic assays, the time course at each substrate concentration was used to calculate the initial velocity of the reaction. The initial velocities (V i ) and their corresponding substrate concentrations were used to estimate the Michaelis-Menton constant (K m ) and the maximal velocity of the reaction (V m ) by fitting the data to the Michaelis-Menton equation using the Enzfitter progran (Elsevier Biosoft). The catalysis constant (K cat ) was calculated by dividing V m with the GST-GAP-C protein concentration used in the reaction. For the circular dichroism studies, the percentages of secondary structure in the GAP-C proteins were estimated using the SSE338 program provided with the J700 system software (Japan Spectroscopic Co.) based on the reference spectra of Yang et al. (27). The data were plotted as mean residue ellipticity verses wavelength. Mean residue ellipticity is the ellipticity generated by one molar amino acid residues/dm pathlength.
One indication of structural instability introduced by the loss or substitution of residues important for folding is a large decrease in soluble polypeptide upon expression in E. coli. The level of soluble expression of each mutant RasGAP protein was measured by growing small volumes of transformed cell cultures under identical conditions, followed by purification of GST-GAP-C fusion protein on glutathione-Sepharose 4B beads. The captured proteins were stripped from the beads by boiling in SDS-PAGE sample buffer and visualized by SDS-PAGE and Coomassie Blue staining. The results of one such experiment is shown in Fig. 2. Compared with wild type GST-GAP-C, four mutants (R786Q, K831Q, Q935H, and K958Q) showed the same level of expression and were considered structurally sta-ble. Three mutants (L811T, R925E, and F757A) were expressed about 5-fold less than wild type GST-GAP-C, suggestive of a minor effect on protein stability. Four mutants (E774K, E826K, W882M, and F898A) were expressed at very low levels suggesting gross destabilization of the RasGAP catalytic fragment.
RasGAP Activities of GST-GAP-C and Nonfusion GAP-C Mutant Proteins-Using purified proteins, the GAP activity profile of the seven most stable GST-GAP-C mutants was compared with wild type GST-GAP-C. Fig. 3A shows that the F757A, L811T, and K958Q mutants had GAP activity similar to wild type GST-GAP-C, the K831Q and Q935H mutants had reduced activities, and the R786Q and R925E mutants had no detectable GAP activity. To confirm that the alterations in GAP activity were independent of the GST fusion, GST carrier protein was removed by proteolytic cleavage, permitting purification of free GAP-C proteins. GAP activities of the three most stable mutants, R786Q, K831Q, and Q935H, were compared with wild type nonfusion GAP-C (Fig. 3B). The results were similar to those obtained with the GST fusion proteins.
The expression levels of the E774K, E826K, W882M, and F898A mutants were too low to obtain sufficiently purified protein for assay. Instead, crude extracts of cells expressing these mutants were assayed for GAP activity and compared with wild type GST-GAP-C and GST alone. No GAP activity was detected for any of the mutants (data not shown). However, because of low levels of expression, no definite conclusion concerning relative GAP activities can be made for these unstable mutants.
K m and K cat Determinations for Reduced Activity, Stable GAP-C Mutants-Either an inability to bind Ras-GTP or a defective catalytic mechanism could explain the lack of Ras GTPase activating function for the impaired mutant GAP proteins. To distinguish between these two possiblities, kinetic assays were performed with the stable K831Q and Q935H mutants using Ras-GTP as the substrate. From these assays the K m for Ras-GTP and K cat for GTPase stimulation were calculated (Table I). In the case of K831Q, Ras affinity was reduced with the calculated K m three times greater than that of wild type GST-GAP-C.The Q935H mutant was observed to be catalytically impaired with a 39-fold reduction in K cat , although the K m of this mutant for Ras-GTP was actually reduced.
Detection of Binding Defects between Mutant GST-GAP-C Proteins and Ras by ELISA-Kinetic assays were not possible for the R786Q and R925E mutants due to a complete lack of detectable GTPase stimulating activity. As an alternative method for measuring Ras-GTP association, an ELISA assay was developed. In essence, ELISA plates were coated with GST-GAP-C proteins and allowed to associate with the GAPinsensitive, high affinity Ras[L61]-GTP protein. The assay was GAP-specific and was able to clearly distinguish between known binding and nonbinding forms of the Ras protein (Fig.  4). As predicted by the kinetic assays, the K831Q mutant bound Ras[L61]-GTP more poorly and the Q935H mutant bound Ras[L61]-GTP better than wild type GST-GAP-C. The R786Q and R925E mutants were severely impaired for Ras association, consistent with their inability to stimulate the Ras GTPase.
Mutant Protein Secondary Structure Estimations by Circular Dichroism Spectropolarimetry-Of the fully stable GAP proteins analyzed, the R786Q and K831Q mutants appeared to be defective for Ras-GTP binding, whereas the Q935H mutant was catalytically impaired. CD spectropolarimetry was used to measure the total secondary structure motifs in the nonfusion form of these GAP mutants as a gauge of subtle alterations of structure (28). When plotted as mean residue ellipticity versus scanned wavelength, the spectra of these three mutants were indistinguishable from the wild type spectrum. Boiled wild type GAP-C and GAP-C[R786Q] had significantly different spectra, indicating little secondary structure (Fig. 5). Analysis of the spectral data provided estimations of the total secondary structure (Table II). All four GAP-C proteins had a high percentage of ␣-helices and random coils and a low percentage of ␤-strands and ␤-turns. These data are supported by the Garnier secondary structure prediction algorithm (Table II)

and K cat values of GST-GAP-C fusion proteins
The initial velocity at each Ras-GTP concentration was calculated from the time course as described under "Experimental Procedures." K m and V m values were estimated by fitting the initial velocities and their corresponding substrate concentrations from GAP kinetic assays with Enzfitter program (Elsevier Biosoft). K cat was calculated by dividing V m with the enzyme concentration in the reaction. The GST-GAP-C enzyme concentrations used in the reactions were: wild type, 78.8 nM; K831Q, 342 nM; Q935H, 453.8 nM; and V897Y, 123.8 nM. ND, not determined. GST  No significant differences in total secondary structural elements were observed between wild type GAP-C and the three mutants, suggesting that the functional differences observed between these mutants and wild type GAP-C were not due to gross alterations in protein structure.
RasGAP Residues 891-903 Share Conserved Sequence and Function with Yeast Adenylyl Cyclase Residues 1528 -1540 -Previously, one of us identified a mutation in the yeast RASresponsive adenylyl cyclase gene, CYR1, which suppressed inactive RAS effector mutants (24). The mutation, SSR2-1, substituted a tyrosine for aspartate at CYR1p residue 1574, which maps in one of two separate sites of RAS interaction on the adenylyl cyclase protein (30,31). This substitution increased enzyme responsiveness to RAS, presumably by increasing affinity for RAS proteins. Examination of the sequence flanking CYR1p residue 1574 revealed an intriguing similarity with the highly conserved block 3A region of RasGAP (see Fig.  6A) with valine 897 analogous to CYR1p aspartate 1547. To determine if these regions of CYR1p and RasGAP also shared a conserved function, valine 897 of RasGAP was substituted with a tyrosine, analogous to the SSR2-1 mutation. The resulting mutant GST-GAP-C protein was expressed at moderate levels (see Fig. 2), purified, and analyzed kinetically. Fig. 6B shows that the V897Y mutant had significantly increased Ras GTPase stimulating activity when compared with wild type GST-GAP-C. Kinetic analysis revealed that the K m value of the V897Y mutant for Ras-GTP was 69 M compared with 115 M for wild type GST-GAP-C explaining the increase in activity (Table I). DISCUSSION Regulation of Ras activity is an important physiological role of the Ras GTPase-activating proteins. Extensive mutagenesis of the Ras proteins has identified regions of the protein critical for binding and GTPase activation by RasGAP. Similar studies of RasGAP have been less revealing and have failed to identify a discrete site of Ras binding analogous to the Ras effector region. Peptides based on the sequence of GAP block 3A inhibited Ras GTPase activation (32), although mutagenesis of the conserved residues in block 3A impaired catalytic function but not Ras binding (17,18). Mutation of the conserved lysine 1423 of neurofibromin (lysine 932 in RasGAP) and valine 853 of FIG. 5. Circular dichroism spectra of wild type and mutant RasGAP catalytic fragments. The secondary structure composition of the wild type and mutant RasGAP catalytic fragments was estimated using circular dichroism spectropolarimetry. Each spectra was the average of three consecutive scans of 0.1 mg/ml protein solutions from 190 to 250 nm. The spectra shown are expressed as mean residue ellipticities versus wavelength. Short dashes, wild type GAP-C; long dashes, GAP-C-R786Q; dashes with one dot, GAP-C-K831Q; dashes with two dots, GAP-C-Q935H; solid line, boiled wild type GAP-C; dotted line, boiled GAP-C-R786Q.

TABLE II
Estimations of secondary structure percentages for wild type and mutant GAP catalytic fragments The CD spectra data were used for the estimation of the secondary structure compositions of wild type and mutant GAP-C fragments using the SSE338 program (Japan Spectroscopic Co.). The primary amino acid sequence of bovine RasGAP was used for secondary structure prediction by the Garnier method (29).  6. A, sequence alignment of adenylyl cyclase in S. cerevisiae and bovine RasGAP block 3A region. The locations of the SSR2 mutation in adenylyl cyclase and the V897Y mutation in RasGAP are shown. B, GAP activities of GST-GAP-C-V897Y compared with wild type GST-GAP-C and GST-GAP-C-R925E. GAP activity is shown as the percentage of saturated wild type activity. Ⅺ, wild type GST-GAP-C; µ, GST-GAP-C-V897Y; m, GST-GAP-C-R925E.
RasGAP were found to impair Ras binding (21). However, substitution of either of these residues also had a severe destabilizing effect upon the protein, which might suggest that they are not directly involved in Ras binding but rather alter the Ras binding site through gross structural changes (20).
The eight members of the RasGAP family share significant homology with the p120 RasGAP catalytic domain, retaining ten amino acid positions that are completely conserved and seven that are moderately conserved (17,18,22,23). In a previous study we have reported that residues involved in the association of the Raf-1 and Ras proteins were highly conserved during evolution and that many had charged side chains (34). These observations suggest that the Ras binding residues of the RasGAP family of proteins may also be conserved and involve charged residues. A biochemical approach was used in order to test this hypothesis. Eleven of the conserved and semi-conserved residues within the RasGAP catalytic domain were mutated semi-conservatively and tested for effects on protein structure and Ras interaction.
Mutation of amino acid residues involved in protein folding and tertiary structure frequently results in poor solubility or rapid degradation when expressed recombinantly in E. coli (18,33). Of the 11 mutations introduced into RasGAP, four had severe consequences for the apparent stability of the protein upon expression in E. coli. This result suggests that glutamate 774, glutamate 826, tryptophan 882, and phenylalanine 898 are important for the folding or tertiary structure of RasGAP and probably are not involved directly in Ras binding or GTPase activation. Mutation of leucine 811, arginine 925, and phenylalanine 757 moderately reduced soluble protein expression, indicating a possible role in protein structure. The substitutions at positions arginine 786, lysine 831, glutamine 935, and lysine 958 did not affect protein stability, suggesting that these residues may be located on the surface of the protein.
Based on these results, analysis was focused on phenylalanine 757, arginine 786, leucine 811, lysine 831, arginine 925, glutamine 935, and lysine 958 to determine their roles in Ras binding and GTPase stimulation.
We observed that the F757A, L811T, and K958Q mutants had GAP activity profiles similar to wild type RasGAP, K831Q, and Q935H had greatly reduced GAP activities and that the R786Q and R925E mutants had no detectable GAP activity. Kinetic assays revealed that the lower GAP activity of K831Q was primarily a result of its lower affinity for Ras-GTP. In contrast, the Q935H mutant had increased affinity for Ras-GTP. Reduction in the GAP activity of Q935H was the result of a catalytic activity that was nearly 39-fold lower than wild type GST-GAP-C. By ELISA we demonstrated that the complete loss in GAP activity observed for the R786Q and R925E mutants was primarily due to an inability to bind to Ras-GTP. Additionally, arginine 786, lysine 831, and lysine 935 were found not to be required for protein structure based upon circular dichroism spectropolarimetry, suggesting that each of these residues has a direct role in Ras-GTP binding and catalysis.
The kinetic assays performed in this study found the K m of wild type GST-GAP-C for wild type Ras-GTP to be 115 M. This value is larger than the reported 19 M for GAP-C for two possible reasons (4). First, the GST fusion has been shown to partially reduce binding affinity of GAP-C for Ras-GTP (18). In fact, as shown in Fig. 3 (A and B), nonfusion GAP-C proteins have higher GAP activity than the corresponding GST fusion proteins. Second, the GAP-C used in previous kinetic studies was expressed in insect cells, whereas we used GAP proteins expressed in bacteria. Although not formally reported, RasGAP expressed in insect, bacterial, or mammalian cells all differ severalfold in their relative binding affinities and catalytic rates. 2 The K cat of wild type GST-GAP-C determined from this study is 4.64 s Ϫ1 , which is similar to the reported 4.2 s Ϫ1 (4).
Homology mapping of the members of the RasGAP family of proteins has identified four highly conserved regions in the RasGAP catalytic domain designated blocks 1, 2, 3A, and 3B. Within the 27-amino acid block 1, we have identified two conserved residues important for both protein structure (glutamate 774) and Ras binding (arginine 786). The residue equivalent to glutamate 774 in the GAP-related domain of neurofibromin (glutamate 1264) was also mutated by Gutmann et al. in their study (17). The substitution of this glutamate by tyrosine suppressed the heat shock sensitivity of a yeast ira Ϫ strain, suggesting retention of GAP function in vivo, although no activity was detected in crude lysates. Substitution of Ras-GAP tyrosine 798 in block 1 with histidine also resulted in a highly unstable protein with poor Ras binding in vitro (18). Residue Arg 786 was shown in this study to play a direct role in binding Ras-GTP. This residue might also play a role in catalysis, because the R786Q mutant had no detectable GAP activity even though it did associate weakly with Ras. Based on these data, block 1 is probably most important for the structural integrity of the GAP catalytic domain with arginine 786 playing an individual role in the direct binding of RasGTP.
The block 2 homology is the longest conserved region in the GAP catalytic domain containing about 37 amino acids. A previous study showed this region may not be essential for binding RasGTP or catalysis, because a functionally active splicing variant of neurofibromin contains a 21-amino acid insertion in block 2 (34). We observed that the W882M mutation in this block resulted in a very unstable protein. Previously we mutated seven residues in block 2, including two highly conserved positions. These mutations resulted in mostly extremely unstable, active proteins but not in stable, binding-defective mutants (18). Therefore, we conclude that block 2 is a structurally important region with no direct role in Ras binding or catalysis.
Block 3A is the most highly conserved homology block, containing six invariant residues out of 23. For this reason block 3A has been assumed to have an important role in GAP function and has been extensively studied. Neurofibromin mutants, with substitutions corresponding to the conserved GAP residues phenylalanine 898, proline 904, alanine 905, and proline 909 had normal GAP activity in vivo or in vitro (17,21). Mutation of RasGAP residues leucine 902, arginine 903, and isoleucine 906 resulted in proteins that still bound Ras yet had impaired GTPase stimulating activity (18,35). In this study, we changed the final two unstudied conserved residues in block 3A. We found that the F898A mutant was extremely unstable, preventing any conclusion about a functional role for this residue other than structural. The other mutant analyzed was V897Y, which resulted in a 2-fold increase in affinity for Ras and a corresponding increase in GAP activity. This particular substitution was chosen because of a curious sequence similarity between block 3A in RasGAP and a Ras-binding site in the yeast adenylyl cyclase. Our data suggest that these sequences are conserved functionally as well. Mutation of adenylyl cyclase aspartate 1547 to tyrosine, which positionally corresponds to GAP valine 897, resulted in an enzyme that could be stimulated by Ras effector mutants (24). The phentotype of the yeast mutant was suggestive of an increased affinity for Ras proteins, as we observed with the RasGAP V897Y mutant. Additionally, a synthetic peptide corresponding to a sequence within block 3A (residue 891-906) inhibited GAP activity of wild type RasGAP in an in vitro assay, indicating the region 891-906 has the ability to bind to Ras-GTP (32). Based on the available data, we conclude that block 3A is not a high affinity binding site for Ras but rather a secondary site. This is possibly the only example of a conserved Ras binding motif found in functionally distinct Ras effector proteins.
Mutation of multiple conserved residues in RasGAP block 3B suggests that this region is a site of Ras interaction as proposed by Poullet et al. (17,21). The R925E mutant demonstrated a complete loss of binding to Ras-GTP, whereas the Q935H mutant had increased affinity for Ras-GTP and a dramatic decrease in catalytic activity. Numerous studies of neurofibromin have indicated that lysine 1423 in block 3B (lysine 932 in GAP) also has an essential role in GAP function (19 -21). Poullet et al. showed that lysine was the only functional amino acid at that position (21). Block 3B is the only region of the GAP/neurofibromin catalytic domain to harbor multiple residues found to be important for both high affinity Ras binding and GTPase enhancement.
The only conserved residue in the RasGAP catalytic domain found to be important for Ras binding and located outside of the homology blocks was lysine 831. This residue is located between block 1 and block 2. Substitution of lysine 831 with glutamine resulted in a very stable protein with greatly reduced affinity for Ras-GTP.
In this study, the amino acid residues found to have the greatest contribution to Ras binding were arginine 786, lysine 831, arginine 925, and glutamine 935. Circular dichroism spectropolarimetry showed that the R786Q, K831Q, and Q935H mutants had wild type secondary structure content, indicating that changes in affinity for Ras were not due to gross structural changes. It is likely that these residues are located on the surface of the protein and directly interact with residues on Ras protein. Interestingly, three out of these four residues are basic, drawing a parallel with the Ras binding domain of Raf-1, which utilizes multiple salt bridges between conserved basic residues and the Ras effector region (33,36). This hypothesis, that Ras-GAP association is mediated by ionic interactions, is supported by the fact that GAP binding is very sensitive to salt concentration (37). The CD spectra secondary structure estimation of the GAP catalytic domain indicated a much higher percentage of ␣-helices than the Ras binding domain of Raf-1, which contains mostly ␤-sheets (36). This might explain the lack of sequence similarity between the Ras binding domains of RasGAP and Raf-1 as well as their different affinities for Ras-GTP (4,38).
The known GAP binding regions on Ras include switch I, switch II, and an undefined region around aspartate 92 (2,5,39). Many of the mutations in the switch I region reduce affinity for RasGAP or neurofibromin and are insensitive to GAP activity (11). In contrast, mutations in the switch II region often lead to impairment of GTPase activity and even an increase in binding affinity for GAP (4). Substitution of glutamate 99 in yeast RAS2 (aspartate 92 in H-Ras) with lysine greatly increased affinity for RasGAP (5).
Integration of the available information about the respective regions of association of the Ras and RasGAP proteins suggests a model for binding and GTPase activation. We propose that the conserved basic RasGAP residues, Arg 786 , Lys 831 , and possibly Arg 925 interact directly with conserved acidic residues in the Ras effector region and possibly aspartate 92 to promote association with the GTP-bound form of Ras. The interactions between these respective protein regions could be largely electrostatic in nature. According to this model, block 3A, and to some extent 3B, constitute a secondary site of interaction through binding to the Ras switch II region. Interactions between block 3A/B and switch II would then lead to activation of the Ras-GTPase, possibly by stabilizing the transition state of GTP hydrolysis reaction or assisting the formation of an optimal microenvironment for the catalytic site (40,41). An interaction between block 3A and Ras switch II is supported by the similarity of phenotypes caused by mutations in each region. Specifically, most mutations in either region result in a loss of GTPase stimulation with no decrease in the affinity of binding. Mutations in RasGAP block 3B and Ras switch II often significantly increase the affinity of RasGAP/Ras association. The interaction between block 3A and the Ras switch II region is also supported by the observation that a block 3A oligopeptide can bind to Ras-GTP but not Rap1A-GTP (32). Rap1A and Ras share the same switch I region, but their switch II regions are different. The ability of this peptide to distinguish Ras from Rap1A implies that block 3A is involved in a direct interaction with the Ras switch II region.