Coassociation of Rap1A and Ha-Ras with Raf-1 N-terminal region interferes with ras-dependent activation of Raf-1.

Raf-1 is a major downstream effector of mammalian Ras. Binding of the effector domain of Ras to the Ras-binding domain of Raf-1 is essential for Ras-dependent Raf-1 activation. However, Rap1A, which has an identical effector domain to that of Ras, cannot activate Raf-1 and even antagonizes several Ras functions in vivo. Recently, we identified the cysteine-rich region (CRR) of Raf-1 as another Ras-binding domain. Ha-Ras proteins carrying mutations N26G and V45E, which failed to bind to CRR, also failed to activate Raf-1. Since these mutations replace Ras residues with those of Rap1A, we examined if Rap1A lacks the ability to bind to CRR. Contrary to the expectation, Rap1A exhibited a greatly enhanced binding to CRR compared with Ha-Ras. Enhanced CRR binding was also found with Ha-Ras carrying another Rap1A-type mutation E31K. Both Rap1A and Ha-Ras(E31K) mutant failed to activate Raf-1 and interfered with Ha-Ras-dependent activation of Raf-1 in Sf9 cells. Enhanced binding of Rap1A to CRR led to co-association of Rap1A and Ha-Ras with Raf-1 N-terminal region through binding to CRR and Ras-binding domain, respectively. These results suggest that Rap1A interferes with Ras-dependent Raf-1 activation by inhibiting binding of Ras to Raf-1 CRR.

Ras belongs to a family of small GTP-binding proteins playing essential roles in cell proliferation and differentiation. Mammalian ras genes carrying activating mutations are found in many types of neoplastic tissue and are able to induce morphological transformation in vitro when transfected into fibroblast cell lines. However, the rap1A gene (1), encoding a 21-kDa GTP-binding protein with high homology to Ras, has been shown to to induce reversion of the transformed phenotype in Ki-ras-transformed NIH3T3 cells (2). In addition to the overall structural homology, Rap1A shares two important structural features with Ras. One is that Rap1A has an identical effector domain (amino acids 32-40) to that of Ras. The effector domain of Ras is essential for the association with and activation of its effectors (3). The other is that Rap1A undergoes similar post-translational modification to Ras at its C terminus except that Ras is farnesylated and Rap1A is geranylgeranylated (4). This modification is essential for the function of Rap1A as observed for Ras (5,6).
Raf-1, a serine/threonine kinase regulating the mitogen-activated protein kinase cascade, is a major mammalian Ras effector and is thought to play a key role in Ras-induced cellular transformation (7). Although the precise mechanism of Rasdependent Raf-1 activation remains unclear, it is known that the effector domain of Ras interacts with the N-terminal RBD 1 (amino acids 51-131) of Raf-1 and that this interaction is essential for physical association between these proteins as well as for the activation of Raf-1 (7). Rap1A, too, has been shown to associate with Raf-1 N-terminal fragment in vivo (8), and a recent x-ray diffraction study of the crystal of the complex between Rap1A and Raf-1 RBD has provided evidence for this association at the atomic level (9). These studies suggest the possibility that the suppression of Ras function by Rap1A is due to the competitive inhibition of Ras-RBD interaction (10), although it is unclear why Rap1A cannot activate Raf-1.
We have recently identified Raf-1 CRR (amino acids 152-184) as another Ras-binding domain and demonstrated that interaction of Ras with both RBD and CRR is necessary for the activation of Raf-1 (11). Two mutations, N26G and V45E, were found to abolish the interaction of Ha-Ras with CRR and attenuate the activation of Raf-1 by Ha-Ras. The fact that both of these mutations replaced Ha-Ras residues with corresponding Rap1A residues prompted us to examine the possibility that the inability of Rap1A to activate Raf-1 is due to its failure to interact with Raf-1 CRR. Contrary to the expectation, we found that Rap1A exhibited a greatly enhanced ability to bind to CRR.

EXPERIMENTAL PROCEDURES
Expression and Purification of Rap1A and Ha-Ras Proteins-Rap1A cDNA was amplified from a human lung fibroblast cDNA library by polymerase chain reaction (12) using a pair of primers, 5Ј-CGGGATC-CGATATGCGTGAGTACAAGCTAG-3Ј and 5Ј-AACTGCAGCAGCTA-GAGCAGCAGACATGATTTC-3Ј. After cleavage with BamHI and PstI in the primer sequences, it was cloned into matching cleavage sites of the baculovirus transfer vector pBlueBac III (Invitrogen Inc., San Diego, CA). The cDNA for an activated Rap1A, Rap1A V12 , was prepared by oligonucleotide-directed mutagenesis (13) and cloned into pBlueBac III as for the wild-type cDNA. pV-IKS, another baculovirus transfer vector for expressing proteins as GST fusions, was provided by Dr. D. Midra (University of California, San Francisco, CA) through Dr. A. Kikuchi (Hiroshima University, Hiroshima, Japan) (14). For expression of Ha-Ras fused to GST, Ha-Ras cDNA was amplified by polymerase chain reaction using a pair of primers, 5Ј-CGCGTCTAGAATGACGG-AATATAAGCTGGTG-3Ј and 5Ј-GCCGGAATTCTCAGGAGAGCACA-* This investigation was supported by grants-in-aids for cancer research and for scientific research from the Ministry of Education, Science, and Culture of Japan and by grants from the Ciba-Geigy Foundation (Japan) for the Promotion of Science and from the Suntory Institute for Bioorganic Research. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Assay for Rap1A and Ha-Ras Binding-MBP fusion proteins of Raf-1 N-terminal fragments were expressed in Escherichia coli and immobilized on amylose resin as described (11). Binding reaction was carried out by incubating 20 l of the resin carrying various amounts of MBP-Raf-1 proteins with various amounts of GTP␥Sor GDP-bound Ha-Ras or Rap1A in a total volume of 100 l of buffer A (20 mM Tris/HCl, pH 7.4, 40 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, 5 mM MgCl 2 , and 0.1% Lubrol PX) as described (11). After incubation at 4°C for 2 h, the resin was washed, and the bound proteins were eluted with buffer A containing 10 mM maltose and subjected to SDS-PAGE followed by Western immunoblot detection with anti-Ras monoclonal antibody Y13-259 (Oncogene Science Inc., Manhasset, NY) or anti-Rap1A polyclonal antibody (Santa Cruz Biotechnology Inc., Santa Cruz, California). Both anti-Ha-Ras and anti-Rap1A antibodies exhibited little cross-reactivity to Rap1A and Ha-Ras, respectively (data not shown). The assay for competitive inhibition of Ras binding to the MBP-Raf-1 fusion proteins by Rap1A were carried out by including a fixed amount of Ha-Ras and various amounts of Rap1A in the same binding reaction. For the in vitro co-association of Rap1A with GST-Ha-Ras, GST-Ha-Ras in Sf9 cell lysate was first immobilized on glutathione-Sepharose and then loaded with GTP␥S. The resin was then incubated with GTP␥S-bound Rap1A in the absence or the presence of purified MBP-Raf-1(51-131) or MBP-Raf-1(48 -206). The binding condition was the same as described above except that the bound proteins were eluted with 10 mM glutathione in buffer A.

Rap1A
Has Enhanced Binding Activity to Raf-1 CRR-In the previous study, we have shown that Ha-Ras binds to immobilized MBP-Raf-1(132-206), representing CRR, demonstrating that Raf-1 CRR acts as another Ras-binding domain independently of RBD (11). This binding was GTP-independent in contrast to the GTP-dependent binding to MBP-Raf-1(50 -131), representing RBD. Ha-Ras mutations N26G and V45E abolished binding to CRR without affecting binding to RBD. The fact that these two mutations replaced Ha-Ras residues with those of Rap1A prompted us to examine if Rap1A lacks the ability to bind to Raf-1 CRR. In the same in vitro binding assay, Rap1A bound to RBD in a GTP-dependent manner (Fig. 1A,  lanes 1 and 2), although the GTP dependence of this binding was less clear compared with that of Ha-Ras (11). Unexpectedly, Rap1A bound efficiently to CRR (lanes 3 and 4). As observed for Ha-Ras, this binding was GTP-independent (lanes 3 and 4) and required an intact zinc finger structure of CRR because Rap1A bound very poorly to MBP-Raf-1(132-206, C168S) (Fig. 1B). Binding of Rap1A to MBP-Raf-1(48 -206) containing both RBD and CRR was severalfold stronger and less GTP-dependent than that to RBD (Fig. 1A, lanes 5 and 6).
These results suggested that CRR is necessary for efficient association between Rap1A and Raf-1 and that CRR rather than RBD plays a major role in this association.
Comparison of the binding properties between Ha-Ras and Rap1A is shown in Fig. 1C. Ha-Ras bound to CRR yielded roughly 10-fold less signal than that to RBD even though the amounts of CRR in the binding reaction were doubled, indicating that the ability of Ha-Ras to bind to CRR was roughly 20-fold less than that to RBD. This estimation by immunoblot was consistent with our kinetic measurement of the affinity of Ha-Ras for RBD and CRR using a competitive inhibition of Ha-Ras-dependent activation of Saccharomyces cerevisiae adenylyl cyclase (15,16). 2 In a striking contrast, Rap1A bound to CRR yielded stronger signal than that to RBD under the same condition, indicating that it has the ability to bind equally to RBD and to CRR (see also Fig. 1A). After taking account of the observation that Rap1A yielded twice as much signal as Ha-Ras on the equimolar basis (Fig. 1D), the ability of Rap1A to bind to CRR was roughly 10-fold greater than that of Ha-Ras. In addition, the ability of Rap1A to bind to RBD was about one-half of that of Ha-Ras after the same normalization. These results indicated that Rap1A has greatly enhanced rather than impaired activity to bind to CRR.
In our previous report, we found that the binding of Ha-Ras to CRR required post-translational modification (11). To test if binding of Rap1A to CRR was also dependent on its posttranslational modification, we incubated RBD and CRR with a lysate of Sf9 cells expressing Rap1A. As shown in Fig. 1E, RBD bound both modified and unmodified Rap1A represented by the faster and slower migrating bands, respectively, on the Western immunoblot as described before (17

Mechanism of Suppression of Ras by Rap1A 11703
that binding of Rap1A to CRR also requires its modification.

Substitution of Lysine for Glutamate at Position 31 Makes Ras a Typical
Rap1A-type Mutant-The above data are somewhat contradictory to our previous finding that binding to CRR is necessary for Raf-1 activation, because Rap1A can bind to CRR better than Ha-Ras and still cannot activate Raf-1. One possible explanation is that enhanced binding to CRR is detrimental to the activation of Raf-1. To test this possibility, we first screened for a Ha-Ras mutant whose binding to CRR was abnormally enhanced. After testing more than 40 mutants described previously (16), the E31K mutant was found to possess roughly 10-fold higher activity to bind to CRR compared with wild type (Fig. 1C). Next, we co-expressed Raf-1 with either Rap1A V12 or Ha-Ras V12 (E31K) mutant in Sf9 cells. The Raf-1 immunoprecipitates were examined for the activity to induce phosphorylation of KNERK2 in the presence of MEK (Fig. 2C). The results showed that Rap1A V12 and Ha-Ras V12 (E31K) indeed could not activate Raf-1 (lanes 1-4). Further, triple expression of either of them along with Ha-Ras V12 and Raf-1 was found to suppress the activation of Raf-1 by Ha-Ras V12 (lanes 5 and 6). These data suggested that the enhanced binding to CRR is indeed detrimental to the activation of Raf-1. They also suggested the involvement of this enhanced binding in the suppressive action of Rap1A in Rasdependent Raf-1 activation, which is examined in the following section.
Rap1A and Ha-Ras Co-associate with Raf-1 N-terminal Region Containing Both RBD and CRR-We reasoned that the greatly enhanced ability of Rap1A to bind to CRR found here may also be involved in competitive inhibition of Ras-Raf-1 association by Rap1A. Rap1A would inhibit association of Ras with a Raf-1 N-terminal region containing both RBD and CRR much more efficiently than that with RBD alone. To test this idea, we first incubated MBP-Raf-1(50 -131) with a fixed amount of Ha-Ras and increasing amounts of Rap1A. As expected, the amount of bound Ha-Ras was found to be reduced by increasing amounts of Rap1A, whereas that of bound Rap1A was increased (Fig. 3A). We then tested MBP-Raf-1(48 -206) in the same experiment. To our surprise, the binding of Ha-Ras to MBP-Raf-1(48 -206) was found enhanced to some extent even when the amounts of bound Rap1A were increased (Fig. 3B).
The only possible explanation for these results would be that Rap1A and Ha-Ras co-associate with the same Raf-1 N-terminal molecule through their independent binding to CRR and RBD, respectively, and that these associations mutually stabilize each other. To test this possibility, we immobilized GST-Ha-Ras fusion protein onto glutathione-Sepharose and incubated it with Rap1A in the absence or the presence of MBP-Raf-1 fusion proteins. As shown in Fig. 3C, no Rap1A was found to be associated with GST-Ha-Ras when they were incubated in the absence of MBP

DISCUSSION
The current model of Ras-suppressive action of Rap1A involves competitive inhibition by Rap1A of the interaction between Ras and Raf-1 RBD (10). However, we found in this study that Rap1A possessed greatly enhanced ability to bind to CRR, another Ras-binding domain identified by us (11) and

Mechanism of Suppression of Ras by Rap1A 11704
others (18 -20). This strong binding even resulted in the coassociation of Rap1A and Ha-Ras with Raf-1 N-terminal region through their independent binding to CRR and RBD, respectively, and these bindings might mutually stabilize each other. Because we have previously shown that binding of Ras to both CRR and RBD is necessary for Raf-1 activation (11), it is very likely that this triple complex formation will result in impairment of Raf-1 activation due to the failure of Ha-Ras to bind to CRR. The previous observations that unmodified Rap1A cannot suppress Ras function (5,6) is also consistent with this, because we found that modification of Rap1A is essential for its binding to CRR.
A study with Ras/Rap1A chimera indicated that transforming potential of Ras requires both of the two regions, residues 21-31 and 45-54. On the other hand, Rap1A anti-oncogenicity requires mainly residues 21-31, although residues 45-54 is also required for full activity to suppress transformation (21). Disregarding residues that are changed conservatively, that are variable among the Ras family, and that are not exposed on the protein surface, three residues, 26, 31, and 45, have been postulated to determine whether the protein is oncogenic or anti-oncogenic (3,(21)(22)(23)(24). In fact, replacements of residues 26 (or 26 plus 27), 31 (or 30 plus 31), or 45 of activated Ras with those of Rap1A resulted in attenuation of transforming activity (22,24). However, it remained unclear which of these residues plays the most critical role.
Residues 26 -28 (including 26) and 42-49 (including 45) have been proposed to constitute a contiguous domain on the surface of Ras protein (3,25). The domain, termed "activator domain," was suggested to play an important role for activation of effectors through some physical interaction with them. In the previous study, we found that Ha-Ras proteins carrying mutations N26G and V45E failed to bind to Raf-1 CRR (11). These mutants were also incapable of activating Raf-1 in Sf9 cells. Based on these findings, we proposed that the activator domain interacts with Raf-1 CRR and that this interaction is essential for Raf-1 activation. This predicts that the inability of Rap1A to activate Raf-1 may result from its failure to bind to CRR, because Rap1A contains residues Gly 26 and Glu 45 , which abolished binding to CRR in the context of Ha-Ras. Contrary to this expectation, we found here that Rap1A exhibited greatly enhanced rather than reduced binding to CRR.
One explanation for this apparent discrepancy is that side chains of the activator domain residues such as 26 and 45, which are divergent between Ras and Rap1A, might not interact directly with CRR. Instead, the activator domain residues such as Phe 28 and Lys 42 , which are conserved between Ras and Rap1A, might participate in direct interaction with CRR. Consistent with this idea, F28A and K42A mutations have been shown to impair activities of Ras (3,25). In the context of Ha-Ras, both Asn 26 and Val 45 might be necessary for these interacting residues to assume their functional conformation, which are destroyed by the N26G and V45E mutations. On the other hand, in the context of Rap1A, such effects of Gly 26 and Glu 45 might be masked by conformational effects of other flanking residues that are not conserved with Ha-Ras. In this regard, residue Lys 31 of Rap1A might play a critical role because we found here that E31K mutation alone enhanced CRR binding in the context of Ha-Ras.
A support for the conformational role of the residue 31 over the activator domain also comes from a study of Rap1A mutants. Nassar et al. recently solved the structure of Rap1A(E30D,K31E) double mutant complexed with RBD by x-ray crystallography (26). Comparison of this structure with that of wild-type Rap1A complexed with RBD indicated that the mutation led to a movement by more than 1.2 Å of a loop containing residues 44 -50, which overlapped with the activator domain. Thus, it is likely that residues 30 and 31 of Ras influence the conformation of the activator domain residues. In addition, Glu 31 of Rap1A(E30D,K31E) was found to interact directly with Lys 84 of Raf-1 RBD, suggesting that E31 of Ras takes part in the same ionic interaction. Nassar et al. also showed that Rap1A(E30D,K31E) and Rap1A(K31E) acquired an activity to stimulate transcription from a Ras-dependent promoter in vivo, albeit to a small extent, and argued that this is accounted for by a large increase in the affinity of Rap1A for RBD due to the newly created ionic interaction. However, the result is also consistent with our finding that the identity of the residue 31 affects CRR binding. The observed conformational change of the activator domain of Rap1A might have brought about a reduction in the affinity of the Rap1A activator domain for CRR to a level appropriate for Raf-1 activation, contributing to the acquisition of the Ras-like activity. Although the solution structure of CRR has been solved (27), further understanding of the mechanism of CRR binding should await the structural analysis of the complex between CRR and Ras or Rap1A.