F-actin-dependent Translocation of the Rap1 GDP/GTP Exchange Factor RasGRP2* □ S

RasGRPs constitute a new group of diacylglycerol-de-pendent GDP/GTP exchange factors that activate Ras subfamily GTPases. Despite a common structure, RasGRPs diverge in their GTPase specificity, subcellular distribution, and downstream biological effects. The more divergent family member is RasGRP2, a Rap1-spe-cific exchange factor with low affinity toward diacyl-glycerol. The regulation of RasGRP2 during signal transduction has remained elusive up to now. In this report, we show that the subcellular localization of RasGRP2 is highly dependent on actin dynamics. Thus, the induction of F-actin by cytoskeletal regulators such as Vav, Vav2, Dbl, and Rac1 leads to the shift of RasGRP2 from the cytosol to membrane ruffles and its co-localiza-tion with F-actin. Treatment of cells with cytoskeletal disrupting drugs abolishes this effect, leading to an abnormal localization of RasGRP2 in cytoplasmic clusters of actin. The use of Rac1 effector mutants indicates that the RasGRP2 translocation is linked exclu-sively to actin polymerization and is independent of other pathways such as p21-activated kinase JNK, or superoxide production. Biochemical experiments demonstrate that the translocation of RasGRP2 to membrane ruffles is mediated by the direct association of this protein with F-actin, a property contained within its 150 first amino acids. Finally, we show that the RasGRP2/F-actin interaction promotes the regionalized activation of Rap1 in juxtamembrane areas of the cell. These results reveal a novel function of the actin cytoskeleton in mediating the spatial activation of Ras subfamily GTPases through the selective recruitment of GDP/GTP exchange factors.

One frequent event in the signaling pathways engaged by most extracellular factors is the activation of GTPases of the Ras subfamily (1). This group of GTP-binding proteins is composed of Ras (K-RasA, K-RasB, N-Ras, and H-Ras), R-Ras (R-Ras, TC21/R-Ras2, and M-Ras/R-Ras3), Rap (Rap1A, Rap1B, Rap2A, and Rap2B), and Ral (RalA and RalB) proteins (2). These GTPases share a common tridimensional structure and a regulatory cycle strictly dependent on the binding of guanosine nucleotides (1,3). In nonactivated cells, these GT-Pases are found in an inactive, GDP-bound state. The stimulation of receptors by their respective ligands leads to the release of GDP and the incorporation of GTP molecules. This exchange of nucleotides induces a conformational change within the switch I and switch II of the GTPases that converts them in optimal docking sites for effector molecules such as c-Raf family proteins, Ral GDP dissociation stimulator (RalGDS), 1 and phosphatidylinositol 3-kinase (1,4). The concomitant activation of these effector proteins by either posttranslational modification (i.e. phosphorylation) or changes in their subcellular localization initiate the stimulation of several signaling cascades that determine, among other biological responses, the proliferative, developmental, and/or motility status of the cell that has received the extracellular stimulus (1,4). The GTPase signal is subsequently shut off at the end of the stimulation period by the hydrolysis of the bound GTP, a step mediated by GTPase-activating proteins (GAPs) (1,3). The importance of this cycle is underscored by the observation that point mutations affecting either the GTP hydrolysis or the GDP/GTP exchange originate GTPase proteins with high oncogenic potential (3).
Because of this activation/deactivation cycle, the main regulatory step in the modulation of the activity of these GTPases is the stimulus-dependent triggering of GDP/GTP exchange. This exchange is catalyzed by enzymes known indistinctly as guanosine nucleotide exchange factors (GEFs), guanosine nucleotide-releasing factors (GRFs), GDP-releasing proteins (GRPs), or GDP dissociation stimulators (5). GEF families described so far for Ras subfamily proteins include Sos proteins (Sos1 and Sos2), Ras GRFs (RasGRF1 and RasGRF2), cyclic nucleotide Ras GEF (CNRasGEF), and Ras GRPs (RasGRP1, RasGRP3, RasGRP4, and the N-myristoylatable and nonmyristoylatable RasGRP2 versions generated by the differential splicing of the rasgrp2 gene) (5). In addition to these families, there are other GEFs specific only for Rap proteins, such as C3G and Epac. Although these GEFs are highly divergent from a structural point of view, they have in common the presence of a Cdc25 domain responsible for stimulating nucleotide exchange on either Rap and/or Ras GTP hydrolases (5). Sos and RasGRF proteins also contain a Dbl homology domain involved in the activation of Rho/Rac GTPases (5). According to this, several reports have indicated that these GEFs can promote the activation of Rac1 (6,7).
Most GEF families have become specialized in the activation of Ras subfamily proteins in specific signaling contexts. Sos and C3G proteins are involved in connecting protein-tyrosine kinases with the stimulation of Ras subfamily GTPases (5). Ras-GRF, CNRasGEF, and Epac proteins are coupled to heterotrimeric G-proteins (5). In contrast to those GEFs, the different members of the RasGRP family are more difficult to assign to a common framework of biological functions and regulatory mechanisms. This is because these exchange factors diverge significantly in a number of biochemical, cellular, and biological properties. At the biochemical level, RasGRPs work catalytically on an overlapping, but not identical, subset of GTPases. Thus, RasGRP1 activates the exchange activity of TC21, M-Ras, and Ras proteins. Instead, it is inactive on Rap GTPases (8 -11). RasGRP2 acts mostly on Rap1 and, with lower affinity, on other Ras proteins except H-Ras (8,12). Interestingly, the catalytic specificity of RasGRP2 appears to be different in function of the splice variant used (13). RasGRP4 activates Ras but not Rap proteins (14). RasGRP3 is the more promiscuous member, being active on the majority of Ras subfamily GTPases tested so far (8,15). As a probable consequence of this biochemical heterogeneity, RasGRPs exhibit different biological properties, including differences in the activation of extracellular-regulated kinases (inhibited by RasGRP2 and stimulated by the other RasGRPs), the stimulation of JNK (stimulated preferentially by RasGRP1 and -3), the induction of neuronal differentiation (where RasGRP2 is inactive), and the promotion of anchorage-independent growth (induced by Ras-GRP1, -3, and -4) (12,14,15). RasGRP proteins are also highly divergent in their organ and tissue distribution (12, 14 -20). Finally, and perhaps more intriguingly from a regulatory point of view, RasGRPs show a different susceptibility toward upstream signals. The activation of RasGRP1, -3, and -4 is mediated via the phospholipase C-␥-dependent generation of diacylglycerol (DAG) (9,(21)(22)(23). This second messenger binds to a C-terminal ZF region present in those GEFs (see Fig. 5), making it possible the translocation of RasGRPs to membranes and their subsequent association with the target GTPases (5,21). In contrast, RasGRP2 shows a very poor response to DAG and, as a consequence, it does not undergo the characteristic rapid translocation of other RasGRPs to the plasma membrane and endomembranes when cells are treated with DAG agonists (13,24). The divergence among the RasGRP ZF regions also conditions the subcellular localization of these proteins. Thus, RasGRP1 and -3 localize preferentially in both the endoplasmic reticulum and Golgi apparatus in exponentially growing cells (24). In contrast, RasGRP2 is found in the cytoplasm or in both the cytosol and the plasma membrane depending on the splice variant analyzed (13,24). The above observations indicate that RasGRP2 is the most divergent member of the family, in terms of substrate specificity, subcellular localization, and regulation by upstream signals.
The different behavior or RasGRP2 led us to look for regulatory signals that could modulate its activity during cell stimulation. During our work with RasGRPs (24), we observed that the nonmyristoylatable RasGRP2 isoform could be found in exponentially growing cells in membrane areas rich in F-actin. Moreover, and unlike RasGRP1 and -3, the treatment of quiescent cells with either epidermal growth factor or phorbol esters did not induce the effective translocation of this protein toward the plasma membrane or the Golgi apparatus. Instead, a small enrichment of RasGRP2 in peripheral membrane ruffles was observed. These observations led us to investigate whether this idiosyncratic member of the RasGRP family could be regulated by signals promoting actin polymerization and cytoskeletal change. In this report, we show that different cytoskeletal regulators such as Vav, Vav2, Dbl, and Rac1 can promote the effective translocation of RasGRP2 toward juxtamembrane areas and membrane ruffles. Moreover, we demonstrate that this effect is because of the intrinsic property of RasGRP2 of associating with polymerized filaments of actin via its N-terminal domain.
Tissue Culture and DNA Transfections-COS1 and Jurkat cells were cultured at 37°C and a humidified 5% CO 2 atmosphere. COS1 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% calf serum plus 100 units⅐ml Ϫ1 penicillin and streptomycin. Jurkat cells were grown in RPMI medium supplemented with 10% fetal bovine serum plus 100 units⅐ml Ϫ1 penicillin and streptomycin. All tissue culture reagents were obtained from Invitrogen. For ectopic protein expression, COS1 cells were transfected with the appropriate plasmids using a liposomal carrier (FuGENE6, Roche Molecular Biochemicals) according to the instructions from the manufacturer. For standard immunofluorescence studies, COS1 cells were transfected with 1 g of each mammalian expression vector. For Rap1 GTP pull-down experiments, cells were transfected with 0.1 and 1 g of vectors encoding RasGRP2 and Vav (⌬1-186), respectively. For fluorescence resonance energy transfer (FRET) analyses, COS1 cells were transfected with 0.5 g of pRaichu-Rap1-404X, 0.5 g of the vector encoding EGFP-Vav (⌬1-186), and 1 g of the plasmid containing FLAG-tagged RasGRP2 in the appropriate combinations. Jurkat T cells were transfected by elec-troporation with 10 g of each indicated plasmid. Electroporations were performed in Opti-MEM medium (Invitrogen) with a Gene Pulser II apparatus (Bio-Rad) using 0.4-cm gap electroporation cuvettes with settings of 280 V and 1200 microfarads. In all cases, transfections were supplemented with empty vector to normalize the total amount of plasmid DNA used in each transfection.
F-actin Co-sedimentation Assays-Actin co-sedimentation assays were performed as reported (33). Briefly, F-actin was obtained by incubating G-actin (Sigma) in F-buffer (5 mM Tris-HCl (pH 7.8), 1 mM ATP, 0.5 mM dithiothreitol, 0.2 mM CaCl 2 , 0.2 mM MgCl 2 , 100 mM KCl) for 60 min at 37°C. GST-RasGRP2 (amino acids 1-150) was purified by standard GST fusion protein purification protocols. Purified proteins were incubated with various amounts of F-actin in F-buffer in a total volume of 100 l at 37°C for 45 min and then ultracentrifuged for 60 min at 100,000 ϫ g in a TLA-100 rotor (Beckman). After the centrifugation, supernatants (100 l) were taken off and diluted 1:1 with SDS-PAGE sample buffer. Pellets were resuspended in 100 l of SDS-PAGE sample buffer. Aliquots of the supernatants (30 l) and pellets (15 l) were denatured by boiling, separated by SDS-PAGE, transferred onto nitrocellulose filters, and immunoblotted with antibodies to GST (Santa Cruz Biotechnology, 1:1,000 dilution) and actin (Sigma, 1:5,000 dilution). Immunoreactive bands were visualized using a chemiluminescence detection system (ECL, Amersham Biosciences).
GST Pull-down Assays-For Rap1 activity assays, COS1 cells were lysed in a buffer containing 50 mM Tris-HCl (pH 7.6), 150 mM NaCl, 10 mM MgCl 2 , 1% IGEPAL (Sigma), 100 M Na 3 VO 4 (Sigma), 1 mM NaF (Sigma), and a mixture of protease inhibitors (Cømplete, Roche Molecular Biochemicals). Lysates were centrifuged at 14,000 ϫ g for 10 min at 4°C. The resulting supernatants were incubated for 60 min at 4°C with 10 g of GST protein fused to the Ras/Rap binding domain of RalGDS immobilized onto glutathione-Sepharose beads (Amersham Biosciences). After incubation, beads were collected and washed three times with lysis buffer. Proteins were then eluted from the beads using Laemmli's sample buffer, separated electrophoretically, and analyzed by immunoblotting using anti-Rap1 antibodies (Santa Cruz Biotechnology). Pull-down assays conducted with GST proteins containing either the N-terminal domain (residues 1-150) of RasGRP2 or the Rac1 binding domain of PAK1 were performed as described above. When appropriate, the proteins transferred to the nitrocellulose filter were visualized by staining with Ponceau solution (Sigma), followed by three washes with water.
Immunofluorescence Analysis-COS1 cells were grown onto glass coverslips and transfected as indicated above. 24 -48 h after transfection, cells were washed, fixed with formaldehyde (Sigma), and subjected to immunostaining. Immunological reagents used were anti-FLAG M2 (Sigma), anti-HA (Covance), and anti-Dbl (Santa Cruz Biotechnology). Cy3-and Cy2-labeled secondary antibodies were obtained from Jackson Immunolaboratories. Rhodamine-phalloidin and cytochalasin D were obtained from Molecular Probes and Sigma, respectively. For FRET imaging, COS-1 cells were grown on polylysine-coated glasses and transfected. 24 h after transfection, cells were serum-starved for 4 h, washed, fixed, and analyzed by confocal microscopy. Capture of FRET images was performed as described previously (32). Immunofluorescent signals were registered using a laser scanning confocal microscope (LSM510, Zeiss). Imports of confocal images were made using the LSM510 software (Zeiss). Final processing of images was done with the Adobe PhotoShop (Adobe Systems) program.
Integrin Activation-30 h after electroporation, Jurkat cells were purified away from cellular debri by centrifugation onto Ficoll-Hypaque medium gradients (Sigma), resuspended in RPMI 1640 supplemented with 10% fetal calf serum (Invitrogen), and allowed to settle for 30 min onto 12-mm coverslips coated with 50 g/ml poly-L-Lys (Sigma). Cells were then fixed in ice-cold 4% formaldehyde for 10 min and permeabilized by an incubation of 5 min with 0.5% Triton X-100 in phosphatebuffered saline solution. Detection of doubly transfected cells was carried out by microscopic detection of EGFP fusion proteins and staining with biotinylated anti-FLAG (M2 clone) and Alexa350-conjugated streptavidin. To detect activation of ␤ 1 integrins, we performed staining experiments with the HUTS-21 monoclonal antibody (34), whereas total ␤ 1 integrin was detected with the TS2/16 monoclonal antibody (34). In both cases, Alexa568-conjugated goat anti-mouse IgGs was employed. Quantification was performed by direct counting of doubly transfected cells and optical comparison with neighboring, nontransfected cells in a Leica DMR photomicroscope equipped with QFISH software (Leica, Mannheim, Germany). At least 300 cells/condition were counted in three separate, independent experiments.

Translocation of RasGRP2 upon Expression of Specific Subtypes of Rac1
GEFs-During our previous study of the subcellular localization of RasGRP family members (24), we observed that the nonmyristoylatable isoform of RasGRP2 (referred to hereafter as RasGRP2) shows a distinctive localization within the cell when compared with its close relatives RasGRP1 and RasGRP3. Although these two exchange factors were seen in endomembranes such as the endoplasmic reticulum and the Golgi apparatus, RasGRP2 always displayed a dispersed localization in the cytoplasm with occasional detection in membrane ruffles (i.e. see panel A in Figs. 1 and 2). This localization did not change when cells were treated with phorbol esters (see Fig. S1 in the Supplemental Materials available in the on-line version of this article). These observations indicated to us that the intracellular dynamics of RasGRP2 during signal transduction could be dependent on specific cytoskeletal structures. As a first step to investigate this possibility, we analyzed the effect of overexpressing a number of cytoskeletal regulators on the subcellular localization of RasGRP2. To this end, we compared by confocal microscopy the subcellular localization of RasGRP2 when expressed in COS1 cells either alone or in the presence of oncogenic Vav (⌬1-186), oncogenic Vav2 (⌬1-187), oncogenic Dbl, oncogenic Lbc, full-length Sos1, and full-length RasGRF2. Vav and Vav2 are GEFs specific for Rac1, RhoA, and RhoG (35). Dbl is a GEF for most Rho/Rac proteins (36). Lbc is a GDP/GTP exchange factor for RhoA subfamily members (37). Sos1 and RasGRF2 are GEFs with dual specificity for both Ras and Rac1 proteins (6,7), an activity that is mediated by their Cdc25 and DH domains, respectively. All these proteins are constitutively active, as inferred from GTPase pull-down and focus formation assays in COS1 and NIH3T3 cells, respectively. 2 When expressed alone, Ras GRP2 shows a uniform cytoplasmic distribution, with only a minority of the protein located in peripheral ruffles ( Fig. 1, panel A). Upon co-expression of either oncogenic Vav or Vav2 proteins in COS1 cells, RasGRP2 redistributes to both juxtamembrane areas and membrane ruffles ( Fig. 1, panels B-D and E-G, respectively). This translocation event is further confirmed when the localization of RasGRP2 is studied by stacking several confocal sections to build a complete topto-bottom cross-section through a specific cell area (Fig. 2, compare panels A and B). Further analysis of the transfected cells by confocal immunofluorescence indicated that Vav family proteins and RasGRP2 usually co-localize within the cell ( Fig.  1, panels D and G, yellow areas). However, we could also observe transfected cells showing effective RasGRP2 translocation without having Vav oncoproteins at the plasma membrane (i.e. Fig. 2, panel B), 2 suggesting that the RasGRP2 shuttling is mediated by a downstream biochemical effect induced by the Vav oncoproteins rather than by the physical interaction between the two molecules. RasGRP2 also translocates to membrane ruffles when co-transfected with oncogenic Dbl (Fig. 1, panels H-J). As above, co-localization of Dbl and RasGRP2 could be observed under those culture conditions (Fig. 1, panel J, yellow areas). Staining of the intracellular F-actin with rhodamine-labeled phalloidin confirmed that the areas where RasGRP2 is present in Vav, Vav2, and Dbl-expressing cells are highly enriched in polymerized actin (see Fig.  6A). In contrast to the above results, RasGRP2 does not change its cytoplasmic distribution when co-expressed with Lbc, Sos1, or RasGRF2 proteins (Fig. 1, panels K-M, N-P, and Q-S, respectively). Interestingly, neither Sos1 nor RasGRF2 proteins induce any membrane ruffling by themselves, despite the fact that they have been described as putative exchange factors for Rac1 (Fig. 1, panels N and Q, respectively). This is not because of lack of activity, because both proteins can trigger high levels of Ras activation in those cells, as assessed by both Ras-GTP pull-down experiments and Erk activity assays. 2 Taken together, these results indicate that the subcellular localization of RasGRP2 can be modulated by membrane signals engaged by specific subtypes of GEFs that are active on Rac1. Instead, no effects are seen when exchange factors specific for RhoA and Ras are used.
Translocation of RasGRP2 Is Mediated by Rac1-dependent Signals-Because Vav is a multidomain protein that engages both GTPase-dependent and independent pathways (38), we used a mutagenesis approach to narrow down the minimal structural requirements needed for their effect on RasGRP2. To this end, we studied the subcellular localization of RasGRP2 when co-expressed with a collection of Vav mutants in which the DH, PH, ZF, and SH3-SH2-SH3 domains had been inactivated by either missense mutation (DH, PH, ZF) or complete deletion of the C terminus (SH3-SH2-SH3) (Fig. 3A). Previous studies had revealed that the individual mutation of the DH, PH, and ZF in both the wild type and oncogenic versions of Vav resulted in the complete inhibition of their biological activities (27). This appears to be a result of the lack of proper activation of Rac1 in vivo by all these mutant proteins (27). In contrast, the oncogenic versions of Vav lacking the C-terminal SH3-SH2-SH3 domains conserve their biological activities intact, including the activation of Rac1, promotion of cytoskeletal change, stimulation of Rac1 effectors such as JNK, and transforming activity (27). When these mutant Vav proteins were tested, we found that Vav (⌬1-186ϩ⌬608 -845, Fig. 3A, protein version B) could induce levels of translocation of RasGRP2 to membrane ruffles similar to those found with the oncogenic version of Vav (Fig. 3B, compare panels A and B). In contrast, the inactivation of either the DH, PH, or ZF domains totally abrogates the effect of Vav oncoprotein on RasGRP2 (Fig. 3B, compare panel A with panels C, D, and E, respectively). Because these mutants cannot activate Rac1 in vivo (27), the present results suggest that the action of Vav on the subcellular localization of RasGRP2 is dependent on the catalytic activity of oncogenic Vav on Rho/Rac family proteins.
To test this idea further, we investigated whether the effect of Vav oncoproteins on RasGRP2 could be mimicked by the expression of some of their GTPase substrates. To this end, we expressed RasGRP2 in COS1 cells in the absence or presence of the constitutively active versions of Rac1 (Q61L mutant), RhoA (Q63L mutant), and Cdc42 (Q61L mutant). Upon transfection, the subcellular localization of RasGRP2 in each condition was studied 24 h later in fixed cells using confocal microscopy. In agreement with our previous results (see Fig. 3), we found that Rac1 Q61L was as effective as the Vav oncoprotein in promoting the translocation of RasGRP2 to juxtamembrane areas of the cell (Fig. 4, panels B and C). Staining of cells with rhodaminelabeled phalloidin confirmed that those areas are highly enriched in F-actin (see Fig. S2, panels E-H, in the Supplemental Material on-line). By contrast, the expression of RhoA Q63L or Cdc42 Q61L does not induce any detectable translocation when compared with cells expressing RasGRP2 alone (Fig. 4, panels D and E and panels F and G; Fig. S1, panels I-L and M-P, respectively), suggesting that this biological effect is specific for Rac1. Because Rac1 can activate several signaling pathways (39), we next utilized a collection of GTPase mutants (Rac1 Q61LϩF37A , Rac1 Q61LϩY40C , and Rac1 G12Vϩ⌬Ins ) impaired in the activation of specific downstream molecules to identify the route involved in RasGRP2 translocation. Rac1 Q61LϩF37A activates PAK and JNK but cannot trigger transformation or cytoskeletal change. Rac1 Q61LϩY40C can induce cytoskeletal changes and tumorigenesis without detectable activation of JNK or PAK kinases. Finally, Rac1 Q61Lϩ⌬Ins (a mutant protein lacking the ␣3Ј insert region) can activate JNK, PAK, and cytoskeletal change but cannot trigger superoxide production in cells (25,40). When these Rac1 mutants were tested, we found that both the Y40C and ⌬␣3Ј insert mutants of Rac1 were as active as Rac1 Q61L in promoting RasGRP2 translocation (Fig. 4, panels J and K and panels L and M). By contrast, the F37A mutant was much less efficient than the rest of

RasGRP2 Tethering to the Plasma Membrane via F-actin
Rac1 Q61L proteins in this particular biological activity (Fig. 4, panels H and I). The effect of Rac1 mutants on RasGRP2 correlated well with their ability to induce F-actin polymerization in COS1 cells (see Fig. S2, panels Q-Z 3 ). Thus, the Rac1dependent translocation of RasGRP2 to areas of membrane ruffling seems to be linked to cytoskeletal events and uncoupled from other intracellular pathways that are dependent on the stimulation of PAK, JNK, and superoxide molecules.
The Translocation of RasGRP2 Is Dependent on Its N-terminal Region-We next used an extensive collection of RasGRP2 mutants to identify the region of this exchange factor involved on its translocation toward membrane ruffles (Fig. 5A). These mutants were co-expressed in COS1 cells with an EGFP fused to the Vav (⌬1-186) oncoprotein and subsequently analyzed by confocal microscopy to evaluate their specific subcellular localizations. As shown in Fig. 5B (panels D-F, G-I, and J-L), EGFP-Vav (⌬1-186) could promote the effective translocation of the RasGRP2 mutants containing an intact N-terminal region (amino acids 1-150). In contrast, we observed that EGFP-Vav (⌬1-186) could not promote the translocation of a truncated version of RasGRP2 lacking those residues even when the ZF region was present (Fig. 5B, panels P-R). The implication of the N-terminal region of RasGRP2 in this translocation event was further demonstrated by the observation that the Vav oncoprotein can promote the translocation of this region of RasGRP2 when expressed in COS1 cells (Fig. 5B, panels M-O).
No translocation of any of the RasGRP2 mutants is observed in the absence of Vav, 2 indicating that this effect is the result of the biological activity of Vav and not the consequence of the deletion of the C-terminal regions of these mutant proteins. Taken collectively, these results indicate that the structural determinants for the translocation of RasGRP2 to membrane ruffles are concentrated in the N-terminal domain encompassing the Ras exchange motif and putative nuclear localization signal regions.
RasGRP2 Translocation Is Dependent on F-actin Polymerization-Given our previous results showing that the subcellu-lar localization of RasGRP2 is strictly dependent on regulators of F-actin dynamics (see Figs. 1-4), we evaluated the possibility that the translocation of RasGRP2 could be the consequence of the direct interaction of the N-terminal region of this protein with cytoskeletal components. To test this possibility, we first investigated the effect of a F-actin disrupting agent (cytochalasin D) in the subcellular localization of RasGRP2 in cells expressing the Vav oncoprotein. The incubation of COS1 cells with this drug for short periods of time induces the disruption of the peripheral F-actin structures into intermediary cytoplasmic bundles and clusters of actin (Fig. 6A, compare upper and lower middle panels). Interestingly, it was observed that Ras-GRP2 followed similar alterations that F-actin in its subcellular localization under these experimental conditions (Fig. 6A, compare upper and lower left panels), an indication of a direct connection between RasGRP2 and polymerized actin. This observation is further confirmed by the demonstration that Ras-GRP2 and F-actin keep their co-localization under those culture conditions (Fig. 6A, lower panel on the right, yellow areas).
Next, we decided to evaluate biochemically the possible interaction of the N-terminal region of RasGRP2 with F-actin in the absence of other cellular components. To this end, a GST-RasGRP2 (amino acids 1-150) fusion protein was purified from bacterial lysates by affinity chromatography onto glutathione-Sepharose beads. This resulted in the isolation of three proteins representing the entire RasGRP2 (residues 1-150) protein and two smaller, abrupt translational termination fragments (Fig. 6B, top panel on the left). After incubation of these proteins with the indicated amounts of in vitro polymerized F-actin, the mixtures were subjected to ultracentrifugation to evaluate the amount of the GST-RasGRP2 that bound (the fraction present in the pellet) or did not bind (the fraction present in the supernatant) to F-actin. We found that the N-terminal region of RasGRP2 does associate with the in vitro polymerized actin in a concentration-dependent manner (Fig.  6B, upper panel on the right). Interestingly, the two smaller GST-RasGRP2 fragments also associate with F-actin under these conditions (Fig. 6B, upper panel on the right), suggesting that at least the last 30 -40 amino acids of the RasGRP2 N-terminal region are dispensable for the interaction. In agree-ment with this coprecipitation, we observed a concomitant reduction of the three RasGRP2 fragments in the supernatant fractions of the centrifugation (Fig. 6B, upper panel on the left). Only marginal precipitation of the GST-RasGRP2 proteins is observed when ultracentrifuged in the absence of F-actin (Fig.  6B, upper panel on the right, first lane). As an additional control for the specificity of this interaction, we observed that F-actin does not bring down the nonchimeric GST protein (Fig.  6B, third right panel from top), demonstrating that the observed heteromolecular interaction between F-actin and Ras-GRP2 is specific. Immunoblot analysis with anti-actin antibodies confirmed the presence of F-actin in the ultracentrifugation pellets (Fig. 6B, second and fourth panels on the right). Finally, we also analyzed whether the N-terminal region of RasGRP2 could bind to Rac1 Q61L to exclude the possibility that the translocation of RasGRP2 could be mediated by additional interactions with the activated version of this GTPase. To this end, cell lysates expressing the AU5-tagged version of this GTPase were subjected to pull-down experiments with either GST or GST-RasGRP2 (1-150). As positive control, we performed similar pull-down experiments with a GST protein fused to the GTPase binding domain of a bona fine Rac1 downstream element (PAK1). No detectable binding of Rac1 Q61L to the GST or the GST-Ras GRP2 fusion protein was detected under these experimental conditions (see Supplemental Material, Fig. S3, upper  panel, lanes 1 and 2). In contrast, a strong association was observed between Rac1 Q61L and the GST-PAK1 fusion protein (Fig. S3, upper panel, lane 3). Taken together, these results indicate that the active versions of Rac1 and Rac1-specific GEFs promote the translocation of RasGRP2 proteins to the cell periphery as a direct consequence of their action on the actin cytoskeleton.
The Interaction between RasGRP2 and the Rac1 Pathway Promotes Spatially Restricted Biological Effects-To shed light into the physiological role of the Vav/Rac1-RasGRP2 interaction, we decided to study the effect that the co-expression of the Vav oncoprotein and RasGRP2 has on the stimulation of the endogenous Rap1 protein, one of the main GTPase substrates of RasGRP2. For this purpose, cell lysates obtained from starved COS1 cells expressing the indicated combinations of proteins were subjected to pull-down experiments using a GST protein fused to the Ras/Rap binding domain of RalGDS. This fragment binds to the activated, GTP-bound forms of Ras and Rap, thus being an adequate tool to determine the activation status of those GTPases in vivo. The expression of RasGRP2 alone in those cells induces a significant increase in the levels of GTP-bound Rap1 when compared with mock-transfected cells (Fig. 7A, upper panel, compare lanes 1 and 3). The expression of the Vav oncoprotein induces only a marginal, although detectable, increase in the levels of GTP-bound Rap1 (Fig. 7A,  upper panel, compare lanes 1 and 2). However, the co-expression of the Vav oncoprotein with RasGRP2 does not result in higher levels of activation of Rap1 when compared with the levels observed when RasGRP2 is expressed alone (Fig. 7A,  upper panel, compare lanes 1, 3, and 4). Immunoblot analysis confirmed that all proteins used in this analysis were expressed properly (Fig. 7A, second, third, and fourth panels from  top). Identical results were obtained when the activation status of a ectopically expressed Rap1 protein (HA-tagged) was measured with the same technique. 2 These observations indicate that the change in the subcellular localization of RasGRP2 induced by the Rac1 cascade does not translate into higher levels of its exchange activity toward Rap1 when measured at the whole cell scale.
Next, we investigated whether the Vav/RasGRP2 interconnection could result in the activation of Rap1 in a spatially

RasGRP2 Tethering to the Plasma Membrane via F-actin
restricted manner within the cell. To monitor the activation of Rap1 at the subcellular level, we made use of a Raichu-Rap1 probe coupled to the utilization of FRET techniques (32). The Raichu-Rap1 probe contains in the same molecule several protein units, including wild type Rap1, the Ras-binding domain of Raf (RBD), and a yellow (YFP) and cyan (CFP) fluorescent protein (Fig. 7B). Under conditions of nonstimulation, the YFP and CFP proteins are pulled apart from each other, resulting in lack of FRET between them (Fig. 7B). However, under conditions of stimulation of GDP/GTP exchange on Rap1, the binding of GTP-bound Rap1 to the RBD moiety of the same Raichu-Rap1 molecule brings together the two fluorescent proteins, thereby increasing the FRET between them (Fig. 7B). In this case, the level of activation of Rap1 in a specific experimental context can be quantitatively estimated by measuring the intensity of the FRET emitted by the cell in those particular conditions (32). Using this method, we found that the expression of Raichu-Rap1 alone does not result in any significant activation of this chimeric molecule in COS1 cells (Fig. 7C,  upper panel on the left). The expression of Raichu-Rap1 and RasGRP2 in the absence of the Vav oncoprotein resulted in the activation of the reporter protein in the cytosol and perinuclear areas (Fig. 7C, lower panel on the left). This is consistent with previous reports on the effect of this exchange factor on Raichu-Rap1 (41). By contrast, the co-expression of Raichu-Rap1 with the Vav oncoprotein resulted in a low stimulation of this re- porter molecule in peripheral membrane ruffles, without any detectable effect on the cytosol (Fig. 7C, upper middle panel  and upper panel on the right). When Raichu-Rap1 was co-expressed with the two exchange factors simultaneously, we found an additive effect, with cells showing a uniform activation of the fluorescent protein in the cytosol and a much higher stimulation in the peripheral membrane ruffles (Fig. 7C, lower middle panel and lower panel on the right). This result indicates that the induction of F-actin polymerization favors the specific activation of Rap1 in juxtamembrane areas and membrane ruffles.
To further evaluate the effect of the Vav/Rac1 pathway on the activity of RasGRP2 in vivo, we studied the effect of the co-expression of Vav and RasGRP2 in the activation of integrin signaling, a biological effect previously assigned to the Rap1 pathway (42,43). To this end, we electroporated in Jurkat cells a mammalian expression vector encoding RasGRP2 either alone or in the presence of expression vectors encoding the EGFP fused to the Vav oncoprotein. After electroporation, cells were cultured for 30 h and the level of integrin activation was estimated by staining cells with HUTS-21, a monoclonal antibody that specifically recognizes activated integrin complexes containing ␤ 1 chains (34). As positive control, we used Jurkat cells expressing transiently the constitutively active version of Rap1 (G12V mutant) (42,43). As negative control, we utilized Jurkat cells expressing the EGFP alone. As shown in Fig. 7D, both RasGRP2 and Vav induce a detectable activation (3-4fold) of integrins in Jurkat cells when compared with the EGFP control. Co-expression of RasGRP2 and Vav induces a synergistic effect, resulting in a 10-fold activation of integrins when compared with control cells. As expected, the expression of Rap1 G12V induces also a robust activation of integrin complexes, as assessed by the increased number of HUTS21-expressing cells (Fig. 7D). These results indicate that the crosstalk between Vav and RasGRP2 results in a regionalized activation of the Rap1 pathway. DISCUSSION The traditional view of the actin cytoskeleton as a passive entity destined to maintain the shape of the cell and promote its motility has changed significantly during recent years. It has now became clear that, in addition to those structural functions, the actin cytoskeleton plays active roles in a number of processes affecting the biological status of the cell. Those include, among others, the formation of membrane domains that establish cell polarity (44), the creation of synaptic clusters contributing to the signaling output of stimulated cells (45), the stabilization of signaling platforms that engage different intracellular pathways (i.e. protein kinase C) (45), and the induction of specific, F-actin-linked transcriptional responses (46). Given the central role of the Ras superfamily in signal transduction, it is not surprising that these GTPases are also downstream targets of the cell cytoskeleton. For instance, several reports have shown that the function of F-actin-binding proteins such as merlin, hamartin, and Eps8 can affect either negatively or positively the activation state of Rho/Rac GTPases and/or its regulators (i.e. RhoGDIs, Sos1) (33,44).
Our results show a new version of the interaction between the cytoskeleton and GTPases of the Ras superfamily: the interference with the subcellular localization of the GDP/GTP exchange factors of the RasGRP family. Thus, we have shown a close co-localization between RasGRP2 and F-actin using different confocal immunofluorescence techniques. This property of RasGRP2 could be observed in exponentially growing cells and, more dramatically, in cells exhibiting robust membrane ruffling upon the transfection of cytoskeletal regulators such as the exchange factors Vav and Dbl or the GTPase Rac1. Using different mutant proteins for Vav or Rac1, we could also demonstrate that this effect is strictly dependent on the activity of the Rac1 pathway and, within this route, linked to the process of actin polymerization. Instead, the translocation of RasGRP2 is independent of other Rac1-dependent pathways such as those mediated by PAK, JNK, and/or superoxide production. The close relationship between RasGRP2 and F-actin was further substantiated by two additional experimental observations. First, we have demonstrated that RasGRP2 follows the same changes in subcellular localization that F-actin when cells are treated with cytochalasin D, a known inhibitor of actin polymerization. This drug causes the disruption of the F-actin meshwork, resulting in the disassembly of membrane ruffles and the scattered accumulation of F-actin and RasGRP2 in short bundles of cytoplasmic aggregates. Second, we could demonstrate using ultracentrifugation experiments that RasGRP2 can interact physically with F-actin in vitro. Such interaction is based on a direct physical contact between these two proteins, because it is observed using purified regions of RasGRP2 and F-actin in the absence of other cellular components. Moreover, we have identified the RasGRP2 N terminus as the minimal region for the binding to F-actin and for the localization of this exchange factor in F-actin-containing structures. Although we have only focused in this study on the behavior of the nonmyristoylated isoform of RasGRP2, it is possible that this regulatory mechanism could be also applicable to the second splice variant of RasGRP2 (Myr-RasGRP2) that can be modified by both myristoylation and palmitoylation at its N terminus. In this regard, it has been noted that the efficiency of those posttranslational modifications is low in vivo, resulting in the segregation of Myr-RasGRP2 between both the plasma membrane and the cytoplasm. This has been attributed to the atypical nature of the myristoylation signal present in Myr-RasGRP2 (13). Accordingly, we can surmise that the mechanism described here could represent also a good signaling stratagem to facilitate the activation of the nonmyristoylated fraction of Myr-RasGRP2 that remains, as its other splice variant counterpart, located in the cytoplasm (13).
We have recently observed that Vav and Rac1 can mediate the activation of phospholipase C-␥1 in stimulated chicken DT40 B-cells (48) and human Jurkat T-cells. 3 One of the consequences of this cross-talk is the stimulation of RasGRP1, a RasGRP2 relative that shows high affinity toward DAG and phorbol esters (48). Despite this evidence, we believe that the effect of the Vav/Rac1 pathway on RasGRP2 must be largely independent of phospholipase C-␥1 activity. This is a result of the fact that the affinity of RasGRP2 for DAG and its chemical analogs is very low. In agreement with this, we could not observe any increase in the exchange activity of RasGRP2 toward Rap1 upon treatment of cells with PMA. 2 Moreover, RasGRP2 does not translocate to the plasma membrane or Golgi apparatus upon exposure of cells to PMA (24). By contrast, RasGRP1 does show high levels of activity in both experimental assays (24). We have also observed that the interference with phospholipase C-␥1 activity using specific inhibitors (U73122) in vivo does not affect the Vav/Rac1-mediated trans-FIG. 6. A, effect of F-actin disruption in the subcellular localization of RasGRP2. COS1 cells co-expressing FLAG-Ras-GRP2 and untagged Vav (⌬1-186) were incubated for 30 min with the indicated amount of cytochalasin D. Cells were then fixed, incubated with mouse anti-FLAG antibodies followed by Cy2-labeled antibodies to mouse IgGs, stained with rhodamine-phalloidin, and analyzed by confocal immunofluorescence analysis. Images show in green and red the localization of FLAG-RasGRP2 (left panels) and F-actin (middle panels), respectively. The areas of overlap between those two proteins are shown in yellow (right panels). B, interaction of F-actin with the bacterially expressed RasGRP2 N terminus. 0.5 g of either the GST-RasGRP2 (1-150) fragments or GST alone were mixed with the indicated amounts of in vitro polymerized F-actin and analyzed by ultracentrifugation as described under "Materials and Methods." Equivalent amounts of the ultracentrifugation supernatants (left panels) and pellets (right panels) were then analyzed by immunoblot analysis using anti-GST antibodies to visualize the GST-RasGRP2 (upper panels) and nonchimeric GST protein (third row of panels from top). The filters were subsequently immunoblotted with antiactin antibodies to visualize the amount of unpolymerized (second and fourth panels on the left) and polymerized (second and fourth panels on the right) actin that is present in the supernatant and pellet fractions, respectively. The antibodies used in the immunoblots are shown at the right. WB, Western blot. location of RasGRP2 to membrane ruffles. 2 Taken collectively, these results suggest that the effect of those cytoskeletal regulators on RasGRP2 is primarily dependent on their direct effect on the actin cytoskeleton.
Our experimental evidence suggests that the exclusive functional end point of this interaction is favoring the translocation of RasGRP2 to juxtamembrane areas of the cell. This statement is based on our observations indicating that the association between RasGRP2 and F-actin does not have any effect on the intrinsic biochemical properties of each component of the complex. For instance, we have observed that RasGRP2, unlike other F-actin proteins, does not show any actin-polymerizing activity either in vivo or in vitro. 2 As a consequence, the Ras-GRP2/F-actin interaction should not result in the stabilization of the associated F-actin bundles or in an increase in the polymerization rates of actin. This behavior is similar to other F-actin-binding proteins such as Eps8, S-nexilin, or afadin that do not display detectable actin-bundling activities (33,49,50). Conversely, we have observed that the interaction between F-actin and RasGRP2 does not change the basal enzyme activity of this exchange factor. Thus, our Rap1-GTP pull-down experiments indicate that the enzyme activity of RasGRP2 in vivo is similar in the absence or presence of the Vav oncoprotein despite the fact that, under the latter conditions, the polymerization of F-actin is highly increased within the cell (see Fig.  7A). The visualization of the local activation of Rap1 using the Raichu-Rap1 probe and FRET methods also confirms this view, because we can observe areas of high RasGRP2 activity without a concomitant enrichment in F-actin (i.e. cytoplasm and perinuclear areas). The observation that the exchange activity of RasGRP2 on Rap1 does not fluctuate in the presence of cytoskeletal regulators such as Vav or Rac1 indicates that the main functional role of its interaction with F-actin is the regionalization of RasGRP2 activity. This idea has been further verified by our FRET experiments utilizing the Raichu-Rap1 reporter, because they showed that the co-expression of the Vav oncoprotein promotes an increase in the pool of activated Rap1 located in membrane ruffles. Interestingly, we have observed that the expression of the vav oncogene product in the absence of RasGRP2 induces also a small, although detectable, local activation of the Raichu-Rap1 probe in those areas, suggesting that Vav may be acting on the endogenous RasGRP2 or, alternatively, on additional targets involved in Rap1 regulation.
The reason for the interaction of RasGRP2 with F-actin during cell signaling remains to be fully explored. Because it has been reported that Rap1 translocates to the cytoskeleton in certain cell types (51)(52)(53)(54), this regulatory mechanism can represent a strategy to ensure the close proximity of Rap1 with one of its GEFs. Given that Rap1 can act as a Ras antagonist (47), it can be also argued that the concentration of Rap1-GTP in F-actin-rich areas may also ensure the activation of its pathway while maintaining the Ras route active in other cellular locations. Finally, the reported mechanism may facilitate the segregation of activated Rap1 in F-actin-dependent and independent pools that could help the engagement of specific signal transduction pathways in different areas of the cell. In any case, our results showing that Vav increases the level of activation of integrin complexes by RasGRP2 suggest that this interaction could play important roles in biological processes linked to cell adhesion.
Although we have focused our attention on the translocation of RasGRP2 induced by cytoskeletal regulators, we cannot exclude the possibility that other factors could contribute to the compartmentalized activation of Rap1 within the cell. For instance, the overall levels of Rap1 activation could be influenced by the presence of GAPs some of which are present in the cytoskeleton (52). This part of the equation may be important because recent results published by Matsuda and co-workers (41) suggest that the pattern of GAP activity "predetermines" in certain ways the areas of optimal stimulation for specific GTPases of the Ras superfamily, including the Rap GTPases. The signaling output of Rap1 in areas enriched in F-actin may be also influenced by the distribution of its effectors in those structures. Finally, the stability of the activation cycle of Rap1 may be negatively affected by the disassembly of the membrane ruffle by the arrival of F-actin disrupting molecules. The separate analysis of all these possible regulatory steps should give us in the future an exact picture of the dynamic forces affecting the signaling output of these GTPases during signal transduction under both physiological and pathological conditions.