Formation of the Ras Dimer Is Essential for Raf-1 Activation*

Although it is well established that Ras requires membrane localization for activation of its target molecule, Raf-1, the reason for this requirement is not fully understood. In this study, we found that modified Ras, which is purified from Sf9 cells, could activate Raf-1 in a cell-free system, when incorporated into liposome. Using a bifunctional cross-linker and a protein-fragmentation complementation assay, we detected dimer formation of Ras in the liposome and in the intact cells, respectively. These results suggest that dimerization of Ras in the lipid membrane is essential for activation of Raf-1. To support this, we found that, when fused to glutathione S-transferase (GST), unprocessed Ras expressed in Escherichia coli could bypass the requirement for liposome. A Ras-dependent Raf-1 activator, which we previously reported (Mizutani, S., Koide, H., and Kaziro, Y. (1998)Oncogene 16, 2781–2786), was still required for Raf-1 activation by GST-Ras. Furthermore, an enforced dimerization of unmodified oncogenic Ras mutant in human embryonic kidney (HEK) 293 cells, using a portion of gyrase B or estrogen receptor, also resulted in activation of Raf-1. From these results, we conclude that membrane localization allows Ras to form a dimer, which is essential, although not sufficient, for Raf-1 activation.

Ras GTPases (Ha-, Ki-, and N-Ras) are the key proteins in eukaryotic signal transduction directed toward cellular proliferation and differentiation (1)(2)(3). The biological activity of Ras is controlled by a regulated GDP/GTP cycle. Guanine-nucleotide exchange factors (Ras-GRF1/2, mSos1/2) induce dissociation of GDP from Ras⅐GDP to form an active, GTP-bound form of Ras. On the other hand, GTPase-activating proteins (p120GAP, NF1) accelerate the intrinsic GTP hydrolytic activity of Ras to promote the formation of an inactive, GDP-bound form of Ras. Upon binding of GTP, Ras alters its conformation to interact with multiple downstream effectors. One of the well characterized effectors is a serine/threonine kinase Raf-1 (4,5), which induces activation of a dual specificity kinase MEK. 1 Activated MEK in turn activates a serine/threonine kinase ERK, which phosphorylates a variety of proteins including protein kinases, transcription factors, and cytoskeletal proteins (6).
Although it has been demonstrated that Ras binds to Raf-1 directly, the precise mechanism of Raf-1 activation by Ras is not fully understood. According to the current model (5), when Ras is activated, the effector region of Ras interacts with the Ras binding domain of Raf-1, which leads to the binding of cysteine-rich domain of Raf-1 to Ras. These interactions relieve the masking of the C-terminal catalytic domain of Raf-1 by the N-terminal regulatory domain and allow Raf-1 to interact with the downstream kinase, MEK. However, since direct interaction of Ras with Raf-1 is insufficient for Raf-1 activation (5,7), an additional molecule(s) has been expected to be involved in this activation. In fact, using a cell-free system, we have found a Ras-dependent Raf-1 activator in the cytosolic fraction (8).
Ras is known to undergo a series of post-translational modifications at the CAAX (where A indicates an aliphatic amino acid) motif in its C terminus, such as farnesylation at Cys-186 and palmitoylation of Cys-181 and Cys-184. The modifications are necessary to localize Ras on the plasma membrane (3). Membrane localization is essential for Ras function, because oncogenic Ras mutant proteins lacking Cys-186, including Ras[G12V,C186S], lose their ability to activate Raf-1 (9). On the other hand, it has been reported that Raf-1 can be activated when targeted to the plasma membrane by addition of the CAAX motif of Ras (10,11). These observations suggest that a certain event(s) necessary for Raf-1 activation occurs in the plasma membrane. The results described in this paper indicate that the dimer formation of Ras is an essential event in the plasma membrane for Raf-1 activation.
Protein Preparation-Modified Ki-Ras was purified from Sf9 cells infected with baculovirus encoding Ki-Ras as described in Ref. 15.
* This work was supported by Grants-in-aid for Scientific Research on Priority Areas 10680666 and 11160204 from the Ministry of Education, Science, Sports, and Culture of Japan and by Core Research for Evolutional Science and Technology (CREST) of Japan Science and Technology Corp. (JST). Our laboratory at Tokyo Institute of Technology is supported by funds donated by Schering-Plough Corp. 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.
Unmodified Ha-Ras was prepared as described previously (16). GSTfused Ha-Ras and Ki-Ras were produced using pGEX-Ha-ras and pGEX-Ki-ras, respectively, and purified according to the manufacturer's instructions. To obtain unmodified Ki-Ras, GST-fused Ki-Ras was digested with thrombin, and thrombin and GST were removed using benzamidine-and glutathione-Sepharose (Amersham Pharmacia Biotech). Preparation of the GTP-and GDP-bound forms of Ras was carried out as described previously (17), except that samples were incubated for 30 min instead of 10 min.
In the experiments with liposome, the cytosolic fraction was used as the Raf fraction and incubated with Ki-Ras-containing liposome at 16°C for 30 min. For the experiments using GST-Ras, the cytosolic fraction was subjected to a Resource Q column (Amersham Pharmacia Biotech), and 0 -250 mM NH 4 Cl fraction was saved and used as the Raf fraction. The fraction was incubated with GST-Ras for 2 h at 4°C.

Modified Ras Incorporated into Liposome
Can Activate Raf-1-A possible explanation for the requirement of the plasma membrane was that the plasma membrane may contain a protein molecule(s) indispensable for Ras-induced Raf-1 activation. We examined this possibility using a cell-free system, which is based on the activation of RafFH (Raf-1 with FLAG epitope tag and six histidine residues in its C terminus) in the cytosolic fraction prepared from HEK293 cells expressing RafFH (the Raf fraction) by the membrane fraction from Sf9 infected with Ha-Ras[G12V] baculovirus (the Ras membrane) (8). If a certain protein(s) in the membrane fraction is required for Raf-1 activation, it is expected that liposome containing only the purified Ras will not be able to substitute for the Ras membrane. However, when purified Ki-Ras⅐GTP was incorporated into liposome and incubated with the Raf fraction, RafFH was strongly activated (Fig. 1), whereas liposome either alone or with Ki-Ras⅐GDP showed little activity. Similar results were obtained with partially purified Ha-Ras (data not shown). On the other hand, RafFH activation was not observed when the Raf fraction was incubated with Ki-Ras⅐GTP, which was not incorporated into liposome. These results suggest that protein molecules in the membrane fraction are dispensable, while liposome containing Ras is absolutely required for Raf-1 activation.
Detection of a Ras Dimer in Lipid Membrane-Since it has been reported that artificial dimerization of Raf-1 results in Raf-1 activation (18,19), we thought that Ras might require liposome for its dimerization. To test this possibility, we incubated Ki-Ras⅐GTP-containing liposome with a homobifunctional amine-reactive cross-linker, EGS. In the absence of EGS, we observed only a single band of monomeric Ras (21 kDa) ( Fig.   2A). On the other hand, an additional band appeared in the presence of EGS, whose mobility corresponded to the molecular weight of dimerized Ras. This band was not observed when unmodified Ki-Ras⅐GTP was incubated with EGS in the presence of liposome. These results suggest that modified Ras can dimerize when reconstituted in liposome.
To examine if Ras dimerization occurs also in intact cells, we performed a protein-fragmentation complementation assay using ␤-galactosidase (␤-gal) (20). In this assay, two deletion mutants of ␤-gal, ⌬␣ and ⌬, show ␤-gal activity only when they are forced to interact with each other. We fused the two deletion mutants to the N terminus of Ha-Ras and transiently expressed them in HEK293 cells. When either ⌬␣-Ha-Ras[G12V] or ⌬-Ha-Ras[G12V] alone was expressed, no bluestained (i.e. ␤-gal-positive) cells were observed (data not shown). However, when both molecules were expressed together, a large number of blue-stained cells appeared (Fig. 2B). On the other hand, only a few cells were stained when unfused FLAG-tagged ⌬␣ was expressed in place of ⌬␣-Ha-Ras[G12V], although expression levels were approximately the same between ⌬␣-Ha-Ras[G12V] and unfused ⌬␣, as judged from Western blotting analysis (data not shown). These data suggest that Ras forms a dimer in HEK293 cells with association of ⌬␣ and ⌬ fragments to restore the ␤-gal activity.
Artificial Dimerization of Ras Leads to Raf-1 Activation in a Cell-free System-If Ras dimerization in lipid membrane is an essential process for Raf-1 activation, bacterially produced, unmodified Ras protein may restore its Raf-1 activating ability by synthetic dimerization. To explore this possibility, we produced dimerized Ha-Ras by fusion with glutathione S-transferase (GST) in Escherichia coli. Dimerization of GST-fused Ha-Ras (GST-Ras) was confirmed by gel filtration (data not shown). When bacterially produced, unmodified Ha-Ras was incubated with the Raf fraction, Ha-Ras could not activate RafFH even in the GTP-bound form (Fig. 3A). On the other hand, GST-Ras induced RafFH activation in a GTP-dependent manner. The activation was not observed when unfused Ha-Ras and GST were incubated together with the Raf fraction. These results suggest that dimerization of Ras by fusion with GST can bypass incorporation of Ras into liposome for Raf-1 activation.
Since the Raf fraction contains a Ras-dependent Raf-1 activator (8), we examined whether dimerized Ras still requires the activator for Raf-1 activation. After RafFH was immunoprecipitated from the Raf fraction by anti-FLAG antibody, the precipitate was washed extensively to remove the activator. When the immunoprecipitated RafFH was incubated with GST-Ras, no activation of RafFH was observed (Fig. 3B). On the other hand, RafFH was activated when the activator, which was partially purified from rat brain cytosols (8), was added to the reaction mixture. These results indicate that dimerization of Ras is not sufficient for Raf-1 activation and that the activator is still necessary for this activation.
Unmodified Ras Activates Raf-1 in a Dimerization-dependent Manner in Intact Cells-As described above, dimerization of Ras leads to Raf-1 activation in the cell-free system. To determine whether the same is true in intact cells, we fused the N terminus of Ha-Ras[G12V,C186S] to the N-terminal domain of the B subunit of E. coli gyrase (GyrB) or the hormone binding domain of murine estrogen receptor (ER). Ras[G12V,C186S] has been reported to bind with Raf-1, but not to activate it because of the lack of its farnesylation site (9). GyrB and ER are known to form a homodimer in response to their specific ligands, coumermycin and 4-hydroxytamoxifen (4-HT), respectively (18,21). When HEK293 cells were transfected with GyrB-Ras[G12V,C186S], treatment with coumermycin efficiently induced RafFH activation (Fig. 4A). In contrast, novobiocin, a monovalent analogue of coumermycin that cannot induce GyrB dimerization, failed to stimulate RafFH activation (data not shown). ER-Ras[G12V,C186S] also activated RafFH in a 4-HT-dependent manner (Fig. 4B). These results suggest that enforced dimerization of Ras induces Raf-1 activation in intact cells, as well as in the cell-free system.

DISCUSSION
Although activation of Raf-1 is initiated by association with Ras⅐GTP on the cytoplasmic membrane, binding of Ras⅐GTP to Raf-1 does not induce Raf-1 activation in vitro. Several studies suggested that an additional factor(s) in the plasma membrane is required for Ras-induced activation of Raf-1 (10, 11). One of the candidate molecules is a membrane-bound tyrosine kinase Src, since Ras and Src synergistically activate Raf-1 (22, 23).
However, our present data indicate that purified Ras incorporated into liposome can activate Raf-1 in the absence of other membrane proteins. Thus, it is likely that, although other membrane molecules including Src may augment Ras-dependent Raf-1 activation in vivo, they are dispensable in the presently used cell-free system.
Using a cell-free system, Stokoe et al. (24) have also shown that Ras is the only required protein in the membrane for Raf-1 activation. However, in their system, modified Ras did not require liposome for Raf-1 activation. The reason for the discrepancy between their results and ours is presently unknown. Since Raf-1 interacts with phosphatidylserine or phosphatidic acid (25,26), it was also possible that phospholipid in the liposome might enhance the Raf-1 activation. However, this is unlikely since GST-Ras could induce Raf-1 activation even in the absence of liposome (Fig. 3A). The association of phospholipids with Raf-1 may not be absolutely required for Raf-1 activation, although it may facilitate the translocation and/or activation of Raf-1 in vivo.
Experiments using EGS and a ␤-gal system (Fig. 2) suggest that Ras forms a homodimer when modified and localized on the plasma membrane. However, it remains unclear how Ras dimerizes on the membrane. Since no difference was observed in cross-linking efficiency between Ras⅐GTP and Ras⅐GDP (data not shown), it is possible that Ras forms a dimer constitutively. Furthermore, it seems that the affinity of the protein moiety of Ras is not high enough for a dimer formation, because bacterially produced (i.e. unmodified) Ras is eluted as a monomer on gel filtration chromatography (27), and we could neither cross-link unmodified Ras ( Fig. 2A) nor immunoprecipitate a Ras dimer from cell lysate (data not shown). Thus, it appears that not protein, but lipid moiety, may be responsible for the dimer formation. An alternate possibility is that, although the affinity between Ras proteins is too low to form a stable dimer, membrane localization increases accessibility of the protein moieties of Ras.
Previously, we found a Ras-dependent Raf-1 activator in the cytosolic fraction (8). The importance of the dimer formation of Ras in Raf-1 activation suggested the possibility that the activator may be an inducer of Ras dimerization. However, GST-Ras failed to activate Raf-1 in the absence of the activator (Fig.  3B). Requirement of the activator was also observed in the experiments using Ras-containing liposome (data not shown). Thus, it is unlikely that the role of the activator in Raf-1 activation is to induce Ras dimerization.
In conclusion, we have demonstrated that Ras requires the lipid membrane for Raf-1 activation and that this requirement can be bypassed by the enforced dimerization of Ras both in vivo and in vitro. Furthermore, we found that Ras dimerization occurs in the membrane. These results strongly suggest that Ras-Ras interaction in the plasma membrane is an essential step for Ras-induced activation of Raf-1. Subsequent association of two Raf-1 molecules on the dimeric Ras may facilitate the increase of Raf-1 kinase activity. The present results raise the intriguing possibility that other low molecular weight GTPbinding proteins may also function as dimers. Such a possibility is currently being investigated in our laboratory.