Identification and Characterization of Potential Effector Molecules of the Ras-related GTPase Rap2*

In search for effectors of the Ras-related GTPase Rap2, we used the yeast two-hybrid method and identified the C-terminal Ras/Rap interaction domain of the Ral exchange factors (RalGEFs) Ral GDP dissociation stimulator (RalGDS), RalGDS-like (RGL), and RalGDS-like factor (Rlf). These proteins, which also interact with activated Ras and Rap1, are effectors of Ras and mediate the activation of Ral in response to the activation of Ras. Here we show that the full-length RalGEFs interact with the GTP-bound form of Rap2 in the two-hybrid system as well as in vitro. When co-transfected in HeLa cells, an activated Rap2 mutant (Rap2Val-12) but not an inactive protein (Rap2Ala-35) co-immunoprecipitates with RalGDS and Rlf; moreover, Rap2-RalGEF complexes can be isolated from the particulate fraction of transfected cells and were localized by confocal microscopy to the resident compartment of Rap2,i.e. the endoplasmic reticulum. However, the overexpression of activated Rap2 neither leads to the activation of the Ral GTPase via RalGEFs nor inhibits Ras-dependent Ral activation in vivo. Several hypotheses that could explain these results, including compartmentalization of proteins involved in signal transduction, are discussed. Our results suggest that in cells, the interaction of Rap2 with RalGEFs might trigger other cellular responses than activation of the Ral GTPase.

Ras proteins are monomeric GTPases that play a pivotal role in the control of cell proliferation; they function as binary switches by cycling between an inactive form bound to GDP and the active GTP-bound state (1). Activation, through the dissociation of bound GDP and subsequent binding of GTP, is catalyzed by GEFs, 1 such as CDC25/Ras-GRF and Sos (2)(3)(4). Return to the inactive state is ensured via stimulation of the low intrinsic GTPase activity of Ras by GTPase-activating proteins, such as p120-GTPase-activating protein, neurofibromin, and Gap IP4BP (2,5). In the active GTP-bound state, Ras exerts its biological effects by turning on several effectors that activate downstream pathways. Soluble serine/threonine kinases B-Raf and c-Raf, once activated by Ras-GTP through a mechanism that is not fully understood, trigger a cascade of protein kinases that results in the activation of mitogen-activated protein kinases extracellular signal-regulated kinases 1 and 2 (6). Another target of Ras is the catalytic subunit of phosphatidylinositol 3-OH kinase (PI3K) (7); this pathway leads to the activation of the protein kinase Akt (8,9), as well as the activation of Ras-related proteins of the Rho/Rac/Cdc42 family involved in controlling the polymerization state of the actin cytoskeleton, cell adhesion, and gene transcription (10,11). Ras is also able to activate the related GTPase Ral through Ral-GEFs that constitute direct effectors of Ras (12)(13)(14)(15); three such proteins, namely RalGDS (16), RGL (17), and Rlf (18) have been extensively characterized, and the isolation of another member of this family has been recently reported (19). Elegant genetic experiments have shown that in most cellular systems, the complementary action of at least two of these three pathways (Raf, PI3K, and RalGEFs) is necessary for Ras to transform murine fibroblasts in culture (11, 13, 20 -22). Other Ras effectors, such as the isoform of protein kinase C (23), Rin (24), and AF-6/Canoe (25) have been identified; their involvement in Ras action has not yet been documented.
The Rap group of Ras-related proteins is composed of the two closely related Rap1A and Rap1B (which are 95% identical) and the two 90% identical Rap2A and Rap2B proteins (26 -28); overall, Rap1 and Rap2 share 60% identity. They constitute the closest Ras relatives because they share more than 50% overall identity with Ras proteins and exhibit very similar three-dimensional structures (29,30). Moreover, Rap1 contains the same effector domain as Ras, which has prompted speculations that Rap proteins may behave as Ras antagonists. Indeed, the Krev-1 gene encoding Rap1A had been isolated on the basis that its overexpression was able to revert the phenotype of Ras-transformed NIH 3T3 fibroblasts (31); since then, several laboratories have provided evidence that overexpression of Rap1 can indeed interfere with Ras function. Molecular basis for such findings resides in the fact that Rap1 is capable of binding with affinities similar to or in some cases even higher than Ras with Ras effectors, such as Raf-1 kinase, PI3K, and RalGEFs (7,18,32). Yet the physiological role of Rap1 appears to vary according to the biological model studied. Rap1 seems to promote cAMP-dependent as well as NGF-stimulated B-Raf and mitogen-activated protein kinase activation in PC12 cells (33,34). Conversely, it has also been shown to participate in the maintenance of T cell anergy by acting as a negative regulator of T cell receptor (TCR)-mediated interleukin-2 gene transcription (35) and to inhibit Raf kinase by forming a catalytically inactive complex in quiescent Chinese hamster ovary cells that is reversed upon insulin stimulation (71). It is also possible that the function of Rap1 is independent of regulating Ras signaling, because activation of endogenous Rap1 by extracellular signals fails to interfere with Ras effector signaling in fibroblasts (72).
In contrast with Rap1, no function has yet been attributed to Rap2. Although it also contains the same effector domain as Ras, except for a single substitution of a serine to phenylalanine at position 39 (a similar substitution in Ras only moderately affects its transforming potential), its overexpression does not antagonize Ras signaling (36). In an effort to uncover the function of Rap2, we searched for potential effectors by using the yeast two-hybrid system. This enabled us to identify a novel protein, RPIP8, that specifically interacts with Rap2 and is principally expressed in brain (37). As described in this paper, we also isolated partial cDNAs encoding the C-terminal region of the RalGEFs RalGDS, RGL, and Rlf. These three related proteins, which constitute effectors of Ras, are capable of inducing activation of the Ras-related Ral GTPase, i.e. nucleotide exchange leading to the formation of active Ral-GTP complexes (12,38,39). Although Ras and Rap1 can both interact with RalGDS and Rlf in cells, only Ras is capable of inducing activation of the GTPase Ral in vivo (12,14). By themselves, RalGEFs exhibit little biological activity, only slightly stimulating transcription from the c-fos promoter; however, upon co-expression with activated Raf, RalGDS greatly synergizes to activate c-fos promoter activity, as well as cell proliferation and morphological transformation (13,40). Moreover, targeting Rlf to the plasma membrane constitutively activates the protein, which is then able to stimulate gene induction and cell growth (38). RalGEFs exhibit considerable homology among each other in their 130 most C-terminal residues, which constitute the Ras and Rap1 interaction domain (RID) (18). They contain a conserved central domain homologous to the RasGEF CDC25 that is responsible for their exchange factor activity toward Ral as well as their stimulating effects on cell growth and gene induction (16,38).
In this study, we show that Rap2 binds to full-length Ral-GEFs in vitro as well as in vivo. This interaction only occurs with active Rap2; Rap2-RalGEF complexes are found in the particulate fraction of transfected cells, and active Rap2 is capable of recruiting RalGDS and Rlf to its resident compartment, the endoplasmic reticulum, suggesting that RalGEFs may indeed constitute effectors of Rap2 function. However, ectopic expression of activated Rap2 does not lead to the activation of the GTPase Ral, nor does it interfere with the ability of Ras to activate Ral. These results suggest that RalGEFs could also serve a function other than activating Ral in cells and that this novel function could be regulated by their interaction with the GTPase Rap2.

EXPERIMENTAL PROCEDURES
Two-hybrid Screens-Screening of a mouse brain cDNA library with the first 168 residues of Rap2A containing a Gly-12 to Val substitution fused to the C terminus of the bacterial LexA protein has already been described (37). The coding sequence for the full-length Rap2B protein was obtained by reverse transcription-polymerase chain reaction from total RNA of the human pro-erythrocyte line HEL (a generous gift from Dr. Dominique Dumesnil) using oligonucleotides 5Ј-ACT GGG ATC CAC CAT GAG AGA GTA CAA AGT GGT GGTG-3Ј and 5Ј-CCC TCG TCG ACG GAC TAC GCC GCG TAG TTC ATC TGC CGC AC-3Ј as forward and reverse amplimers respectively and Pfu polymerase (Stratagene); the resulting product was digested with BamHI and SalI and cloned into pGBT10 (3) 3 -CYCI-lacZ) and used to screen a cDNA library from human Jurkat cells as described (41).
In Vitro Interactions-Ras family proteins were expressed as GST fusions, purified on glutathione-Sepharose beads, and loaded with GDP or Gpp(NH)p prior to binding reactions as described previously (37). Potential partners were transcribed and translated in vitro in the presence of [ 35 S]methionine in a coupled reticulocyte lysate using the bacteriophage T7 RNA polymerase (TNT, Promega) from templates obtained as follows. The RIDs of RalGEFs (obtained from the twohybrid screens described in this study) were amplified with polymerase chain reaction with Pfu polymerase and subcloned at the 5Ј EcoRI site of the pGEMMyc4 vector (a generous gift from Harald Stenmark and Marino Zerial) in frame with a Myc epitope and downstream of the bacteriophage T7 promoter. The 5Ј and 3Ј oligonucleotide primers used to amplify these sequences were as follows: for RalGDS, 5Ј-TGTG GAA TTC GCC TCC ACC ACG CCC GTG GCT GCC-3Ј and 5Ј-CCTTG CTC GAG TCA GAA GAT GCC CTT GGC AAA TCTT-3Ј; for RGL, 5Ј-TGTG GAA TTC AAC AAT CCT AAA ATC CAC AAG CGC-3Ј and 5Ј-CCTTG CTC GAG TCA GAG GGT GAT CTT GCT GTG CCT-3Ј; and for Rlf, 5Ј-TGTC GAA TTC TCC CCT AGG CCT TCT CGG GGT-3Ј and 5Ј-GTCT GTC GACTCAG AAC AGT GCC CGTGC-3Ј. The RBD of c-Raf, obtained as a GST fusion from A. Wittinghofer (44), was excised with BamHI and EcoRI, subcloned into the pMalC2 vector (New England Biolabs), amplified with oligonucleotides homologous to vector sequences providing a 3Ј XhoI site, and cloned as above in pGEMMyc4. The absence of mutations in all clones was assessed by DNA sequencing. RPIP8 was transcribed and translated from pBS-RPIP8 as described (37). Full-length RalGDS and RGL coding sequences were excised from pCEP4RalGDS and pCEP4RGL, respectively (generous gifts from M. White (13)) with BamHI and inserted into the BamHI site of pGEMMyc4; full-length Rlf was excised from pPC86-Rlf with SalI and NotI, blunted with Klenow at its 3Ј extremity, and inserted into the 5Ј SalI and 3Ј PstI (blunted with T4 DNA polymerase) sites of pGEMMyc4.
For binding experiments, 10 l of glutathione-Sepharose 4B beads bound to 250 ng of GST, GST-Rap2A, GST-Ha-Ras, GST-Rap1A, and GST-RalA proteins were washed three times in ice-cold exchange buffer (25 mM Tris-HCl, pH 7.5, 2 mM EDTA, and 5 mM dithiothreitol) and incubated for 30 min at 37°C in 20 l of exchange buffer containing 150 mM of Gpp(NH)p or GDP. The beads were then diluted in 180 l of interaction buffer (25 mM Tris-HCl, pH 7.5, 1 mM EDTA, 5 mM MgCl 2 , and 5 mM dithiothreitol) containing 200 M of the appropriate nucleotide and incubated for 3 h at 4°C with 1 l of in vitro translated [ 35 S]methionine-labeled potential effector. Beads were washed four times with 1 ml of interaction buffer and then boiled in SDS gel sample buffer to recover bound proteins. Samples were analyzed by SDS-polyacrylamide gel electrophoresis on 10% polyacrylamide gels; after staining with Coomassie Blue to detect the GST and GST fusion proteins, gels were treated with Amplify (Amersham Pharmacia Biotech), dried, and exposed to film.
In Vivo Interactions-The cDNAs encoding Ras and Rap2 GTPases as well as RalGEFs were subcloned into mammalian expression vectors under the control of the cytomegalovirus promoter as follows. The coding sequences for Rap2 proteins carrying a Gly to Val substitution at position 12 (Rap2Val-12) and a Thr to Ala substitution at position 35 (Rap2Ala-35) (45) were amplified by polymerase chain reaction with Pfu using primers 5Ј-GTGT GGA TCC ACC ATG CGC GAG TAC AAA GTG GTG GTG-3Ј and 5Ј-TCTT CTC GAGC CTA TTG TAT GTT ACA TGC AGA ACA-3Ј and inserted into the BamHI and XhoI sites of pcDNA3 (Invitrogen). Full-length RalGDS and RGL sequences excised as indicated above were inserted into the BamHI site of a pRK5 vector engineered to encode an N-terminal Myc epitope fused in frame to the N terminus of the protein of interest (a generous gift from Dr. Alan Hall).
HeLa cells were co-transfected with 4 g of Rap2 and 8 g of RalGDS or Rlf expression constructs (or the relevant empty vector) per 8.5-cm dish with calcium phosphate; 40 h after transfection, cells were washed twice with phosphate-buffered saline and processed as follows. For experiments performed with total cell extracts, cells were lysed on ice in 25 mM Hepes buffer, pH 7.5, containing 0.1 M NaCl, 1% Nonidet P-40, and protease inhibitors, and debris were removed by a 15-min centrifugation at 14,000 ϫ g. In order to prepare membranes, cells were swollen on ice in hypotonic buffer (25 mM Hepes, pH 7.5, containing protease inhibitors) and lysed by 100 strokes of the tight-fitted pestle of a Dounce homogeneizer. The postnuclear supernatant obtained after centrifugation at 1000 ϫ g for 3 min was submitted to further centrifugation at 45,000 rpm for 30 min in a Beckman TLA 45 rotor; membranes were resuspended in hypotonic buffer containing 0.1 M NaCl and recentrifuged as described above. They were solubilized in the same buffer containing 1% Nonidet P-40 (30 min on ice), and insoluble material was eliminated by a final centrifugation as described above.
Solubilized extracts were precleared with protein A-Sepharose and immunoprecipitated with 5 g of anti-Myc 9E10 antibody (Boehringer Mannheim) followed by protein A-Sepharose as described previously (46); the presence of Rap2 in immune complexes was revealed by Western blotting with affinity-purified polyclonal anti-Rap2 antibodies (47) and visualized by ECL (Amersham Pharmacia Biotech).
Intracellular Localization of Co-expressed GTPases and RalGEFs-5 ϫ 10 6 exponentially growing HeLa cells were electroporated in a final volume of 200 l of culture medium supplemented with 15 mM Hepes buffer, pH 7.5, at 960 microfarads and 240 V in the 0.4-cm electrode gap cuvettes of a Bio-Rad Gene Pulser with 6 g of each expression construct for Myc-tagged RalGDS or Rlf, and Ras or Rap2 GTPase as indicated. After electroporation, cells were washed once and plated in four 35-mm dishes containing glass coverslips. 24 h later, they were washed twice with phosphate-buffered saline, fixed for 6 min at Ϫ20°C in methanol, and simultaneously incubated with affinity-purified anti-Rap2 or anti-Ha-Ras (Santa-Cruz, catalog no. sc-520) and 9E10 anti-Myc (1 g/ml, Boehringer Mannheim) antibodies as described previously (47). Complexes were stained with fluorescein isothiocyanatecoupled anti-mouse and tetramethylrhodamine isothiocyanate-coupled anti-rabbit antibodies (Jackson ImmunoResearch Laboratories, Inc.) and visualized with a Leica scanning laser confocal microscope.
Activation of Ral by Rap2 in Vivo-In order to assess whether Rap2 could activate Ral via RalGDS or Rlf, COS-7 cells grown in 5-cm dishes were transfected with 1.5 g of pMT2-HA-Ral together with 2 g of expression vector for RasVal-12 or Rap2Val-12 and/or 1 g of pcDNA3-Rlf or pcDNA-Myc-RalGDS (a generous gift from Thomas Linnemann and Alfred Wittinghoger) as indicated. After metabolic labeling of cells with [ 32 P]orthophosphate, HA-Ral bound nucleotides were analyzed and quantitated as described (38).
Ras-dependent Ral activation in the response to insulin, was measured in A14 cells, which are NIH 3T3 fibroblasts expressing the human insulin receptor (48), grown in 5-cm dishes, and transfected with 1 g of pMT2HA-Ral together with 1 g of pcDNA3 or 1 g of pcDNA3-Rap2-Val-12. After serum starvation for 16 h, the cells were stimulated for 5 min with 1 M of insulin and lysed. Ral-GTP levels were determined by trapping the active complex on beads covered with GST-RalBD fusion proteins (GST fused to the Ral binding domain of the Ral effector RLIP76 (49)) as described (15). Rap2 and Ral proteins were detected on Western blots using monoclonal antibodies (Transduction Laboratories).

Rap2-GTP Interacts with the C-terminal Ras
Interaction Domains of RalGEFs-In order to search for potential effectors of Rap2, through which it may exert its biological effects, we performed two independent screens using the two-hybrid method in the yeast S. cerevisiae. We screened a mouse brain cDNA library with residues 1-168 of Rap2A carrying a Gly to Val substitution at position 12 (Rap2AVal-12) fused to the C terminus of the DNA binding domain of the bacterial transcription activator LexA as a bait, and a human Jurkat T lymphoma cDNA library with the full-length Rap2B protein fused to the C terminus of the DNA binding domain of the yeast transcription activator GAL4 as a bait. Among potential effectors, we isolated the C-terminal regions of the RalGEFs RalGDS (residues 702-852), RGL (residues 618 -768), and Rlf (residues 607-778); Rlf was isolated in both screens, whereas RGL and Ral-GDS were obtained from the mouse brain cDNA library and the Jurkat library, respectively. Because these molecules are known effectors of Ras and have been shown to bind Rap1, we proceeded to investigate the biochemical and biological signif-icance of their interaction with Rap2.
The interaction spectrum of RalGEFs with Ras family proteins was assessed by mating derivatives of yeast strain HF7c (mat a) expressing GTPases fused to the C terminus of the DNA binding domain of GAL4, with derivatives of yeast strain Y187 (mat ␣) expressing the various effector proteins fused to the C terminus of the activation domain of GAL4 and testing the resultant diploid strains for growth in the absence of histidine and ␤-galactosidase activity. As shown in Table I, the C-terminal domains of all three RalGEFs (RalGDS, RGL, and Rlf) interacted as strongly with Rap2A and Rap2B as they did with Ras and Rap1A; among other Ras-related proteins, they interacted less well with R-Ras and not at all with Ral. In contrast, Raf-1 kinase interacted strongly with Ras and poorly with Rap1 and Rap2, whereas RPIP8 (a potential Rap2 effector expressed in brain (37)) bound specifically to Rap2 (A and B). The same pattern of interaction was observed irrespective of whether Rap1A and Rap2B proteins were complete or truncated of their 18 C-terminal residues. No interaction occurred when Rap1A and Rap2 (A and B) proteins carried a Ser to Asn mutation at position 17, and they had therefore lost their high affinity for GTP (data not shown); such a pattern of interaction is characteristic of potential effectors of Ras-related proteins.
An in vitro binding assay was devised in order to investigate these interactions by a more direct and independent method. It consisted in assessing the ability of potential effectors, transcribed and translated in vitro in the presence of [ 35 S]methionine, to bind to the various Ras-related proteins fused to glutathione S-transferase, immobilized on glutathione-Sepharose beads, and loaded either with GDP or with a nonhydrolyzable analogue of GTP, Gpp(NH)p. Fig. 1 shows that this assay displayed results similar to those obtained by the yeast two-hybrid method: RPIP8 and the RBD of Raf-1 interacted specifically with Rap2 and Ras, respectively, whereas the Cterminal domains of all three RalGEFs bound to Ras, Rap1, Rap2A, and Rap2B indiscriminately. In all cases, the interaction only occurred with the GTP-bound form of the GTPases. Hence, RalGEFs, through the direct binding of their C-terminal domain to Rap2-GTP complexes, constitute potential effectors of Rap2A and Rap2B GTPases.
Interactions with Full-length RalGEFs-Because the RIDs of RalGEFs only represent the C-terminal one-third of these molecules, we investigated whether the full-length proteins might contain regulatory regions that would introduce some differential specificity in their interactions with Ras, Rap1A, Rap2A, and Rap2B. As shown in Table I, using the yeast two-hybrid system, full-length RalGDS and Rlf (we were unable to express a functional RGL protein in yeast) interacted with all four GTPases, albeit slightly less efficiently with Rap1A than with Ras and Rap2 (A and B). As with their RIDs, the interaction of full-length RalGEFs with Rap1 and Rap2 was unaffected by the presence or absence of the 18 C-terminal residues of these GTPases (not shown). It is noticeable that, in contrast with the results obtained with their RIDs, full-length RalGDS and Rlf could no longer interact with R-Ras, suggesting that RalGEFs do not constitute physiological effectors of the R-Ras GTPase.
Using the same in vitro binding assay as above, we extended the results obtained with the yeast two-hybrid system: Fig. 2 shows that full-length RalGDS, RGL, and Rlf interacted with Ras, Rap1, Rap2A, and Rap2B in their active GTP-bound form.
In this assay, Rlf interacted better with Ras than with Rap1 and Rap2, in agreement with the reported high affinity of the Ras-GTP-Rlf interaction (18). Although some interaction of RalGDS and RGL occurred in certain experiments with the GDP-forms of Rap GTPases, it was always much weaker than that observed with their GTP-bound form. In summary, the data obtained with the yeast two-hybrid system as well as with an in vitro binding assay suggest that RalGEFs could indeed constitute effectors of Rap2A and Rap2B GTPases.
Interaction of Rap2 with RalGEFs in Mammalian Cells-In order to assess whether such interactions between RalGEFs and Rap2 could also occur in mammalian cells, we co-transfected HeLa (Figs. 3 and 4) or HEK 293 cells (not shown) with expression vectors for the constitutively active (carrying a Gly to Val substitution at position 12) and inactive (carrying a Thr to Ala substitution at position 35 (45)) mutants of Rap2A together with constructs encoding full-length RalGDS or Rlf tagged at their N terminus with an exogenous Myc epitope.  Fig. 3. 40 h after transfection, cells were mechanically lysed in hypotonic buffer, and the particulate fraction of cells was isolated by ultracentrifugation, solubilized with 1% Nonidet P-40, and submitted to immunoprecipitation and Western blotting as in Fig. 3. In the middle panels, the particulate fraction was prepared from cells transfected with Rap2 or Myc-Rlf vectors only; these fractions were then mixed, solubilized, and immunoprecipitated as above. Similar results were observed after transfection of HEK 293 cells.

FIG. 4. A complex between Rap2 and RalGEFs is formed in the particulate fraction of cells. HeLa cells were co-transfected with expression vectors for Rap2 and Myc-tagged RalGEFs as in
Under these conditions, Rap2Val-12 was present in complexes immunoprecipitated from whole cell lysates with anti-Myc antibodies (Fig. 3); such was not the case when the inactive Rap2Ala-35 mutant was expressed, showing that RalGDS and Rlf only associated in mammalian cells with the active form of Rap2. These results are similar to those obtained in control experiments with Ras (not shown).
Although they are synthesized as soluble precursors, mature Ras family GTPases are bound to cellular membranes, Ras to the plasma membrane and Rap2 to the endoplasmic reticulum, following a series of posttranslational modifications that involve prenylation and palmitoylation in the case of Ras and Rap2 (47,50). In order to establish that the complexes between RalGEFs and Rap2 did not involve unprocessed cytosolic precursors of the GTPase but bona fide processed and membraneassociated protein, the particulate fraction of transfected cells was isolated by ultracentrifugation prior to the immunoprecipitation of RalGEFs with anti-Myc antibodies as above (Fig. 4). As previously reported (38), a significant proportion of transfected RalGDS and Rlf (10 -20%) was associated with the particulate fraction of cells; this proportion did not vary with the co-expression of Ras (not shown) or Rap2 (Fig. 4) proteins and represented nonspecific association with cellular membranes (see below and Fig. 6). As with whole cell extracts, the active Rap2 protein (but not the inactive Ala-35 mutant, not shown) was co-immunoprecipitated with RalGDS and Rlf from solubilized membranes of co-transfected cells. In contrast, when membranes prepared from cells only transfected with Rap2 or RalGEFs were mixed, solubilized, and submitted to immunoprecipitation, only a very minor amount of Rap2 was recovered in the immunoprecipitates (Fig. 4); similar results were obtained after transfection in HEK 293 cells, as well as in control experiments performed with Ras (not shown). Hence, these complexes between Rap2 and RalGEFs were formed on membrane structures prior to cell lysis, suggesting that the interaction between active Rap2 and RalGEFs may indeed occur in mammalian cells.
Rap2 Does Not Lead to Activation of the Ral GTPase-In order to assess whether the observed interaction between Rap2 and members of the RalGEF family resulted in activation of the Ral GTPase, we measured the levels of Ral-GTP in transfected COS-7 cells after labeling of the nucleotide pools with [ 32 P]orthophosphate. Ectopic expression of either RalGDS or Rlf enhanced the level of Ral-GTP from 4 -6% (relative to the total amount of Ral-GDP ϩ Ral-GTP) to 15-20% (Fig. 5, A and  B); co-expression of activated Ras further stimulated Ral-GTP formation up to 30 -50% ( Fig. 5A and Refs. 38 and 39). However, in contrast to activated Ras, overexpression of activated Rap2 did not further enhance the level of Ral-GTP induced by RalGDS (Fig. 5A) or Rlf (Fig. 5B). These results demonstrate that despite the ability of active Rap2 to form complexes with RalGEFs in vivo, active Rap2 does not stimulate the ability of RalGDS or Rlf to activate Ral under these circumstances. Therefore, it is unlikely that the Ral GTPase is involved in signal transduction downstream of Rap2.
Rap2 Recruits RalGEFs to Its Resident Compartment, Which Is Different from That of Ras-As a possible hint to the actual inability of Rap2 to activate Ral despite its interaction in vivo with RalGEFs, we examined by confocal microscopy whether the subcellular localization of RalGDS and Rlf was similar in cells expressing activated forms of Ras and Rap2. Immunofluorescence experiments were performed on HeLa cells co-transfected with expression constructs for Myc-tagged RalGEFs and constitutively inactive (Fig. 6A) or activated (Fig. 6B) GTPases, which were fixed with cold methanol, conditions that eliminate cytosolic proteins and enable to visualize only molecules present on structures such as the cytoskeleton and cellular membranes. In the absence of co-expressed activated Ras or Rap2, a faint and diffuse localization of ectopically expressed RalGDS and Rlf to intracellular membranes was observed in transfected cells (Fig. 6A, a and b). Upon co-expression with activated Ras, the membrane-associated fraction of RalGDS (Fig.  6B, a and b), as well as Rlf (Fig. 6B, c and d), was localized with Ras at the plasma membrane. In contrast, when they were co-expressed with the active form of Rap2, the membraneassociated fraction RalGDS, as well as Rlf, co-localized with this GTPase at the endoplasmic reticulum (Fig. 6B, e-h). In control experiments, co-expression of Ral GEFs with the inac-  and h), and Rlf and Rap2Ala-35 (i and j). 24 h later, cells were processed for double immunofluorescence as described under "Experimental Procedures" in order to simultaneously visualize RalGEFs, stained in green, with monoclonal anti-Myc followed by fluorescein isothiocyanatecoupled anti-mouse antibodies (c, e, g, and i), and GTPases, stained in red, with rabbit polyclonal anti-Ras or anti-Rap2 antibodies followed by tetramethylrhodamine isothiocyanate-coupled anti-rabbit antibodies (d, f, h, and j) in co-transfected cells. B, HeLa cells were co-electroporated as in A with the following expression constructs for Myc-tagged RalGEFs and constitutively activated GTPase mutants: RalGDS and RasVal-12 (a and b), Rlf and RasVal-12 (c and d), RalGDS and Rap2Val-12 (e and f), Rlf and Rap2Val-12 (g and h). They were processed for double tive RasN17 (Fig. 6A, c-f) and Rap2Ala-35 (Fig. 6A, g-j) mutants did not cause them to co-localize with the GTPases but rather to maintain the same faint and nonspecific membrane labeling as when they were expressed alone. It is noteworthy that RasN17 was localized at the plasma membrane similarly to RasVal-12, whereas inactive Rap2Ala-35 (as well as Rap2AAsn-17, not shown) did not specifically label the endoplasmic reticulum as transfected Rap2Val-12 (Fig. 6B, f and h) or endogenous protein (47), but it did label various cellular membranes, including the plasma membrane. Hence, the colocalization of ectopically expressed RalGEFs with the active GTPase co-expressed in transfected cells represents recruitment of RalGDS and Rlf by active Ras and Rap2 at their resident compartment, i.e. the plasma membrane for Ras and the endoplasmic reticulum for Rap2. Such a recruitment further hints that RalGEFs may indeed act as effectors of Rap2 function and also provides an explanation for why active Rap2 does not lead to the activation of Ral.
Rap2 Does Not Interfere with Ras-mediated Ral Activation-Given that the overexpression of activated Rap2 can recruit RalGEFs yet does not lead to Ral activation, we wished to test the hypothesis that in such a manner, activation of Rap2 could regulate Ras signaling to Ral, possibly by sequestering Ral-GEFs. To this end, we used a cell line derived from NIH 3T3 overexpressing the insulin receptor (A14), in which insulin stimulation leads to Ras-dependent Ral activation (15), and investigated whether overexpression of activated Rap2 was able to interfere with this response (Fig. 7). In these experiments, Ral activation was qualitatively measured by trapping the active Ral-GTP complex on a GST-RalBD fusion protein as in (15). When A14 cells were transfected with HA-tagged Ral alone, insulin stimulated both the activation of endogenous Ral, representing the whole cell population, and HA-Ral ectopically expressed in transfected cells, confirming that this method may indeed be used to assess Ras to Ral signaling in transfected cells. Transfection of Rap2 expression vector caused a vast overexpression of the protein but did not affect the expression level of co-transfected HA-Ral as compared with control cells transfected with empty vector. Under basal conditions, overexpression of Rap2 was associated with a slight activation of HA-Ral. Quite remarkably, the level of HA-Ral activation in response to insulin stimulation was similar in cells overexpressing or not overexpressing activated Rap2; further enhancement of Rap2 expression level had no noticeable effect on insulin-stimulated Ral activation as well (not shown). It should be noted that after insulin stimulation, Ras becomes strongly activated, so GTP-bound Ras and Rap2Val-12 would be expected to compete for RalGEFs binding. Our results indicate that, under these conditions, and despite its vast overexpression, Rap2 is unable to sequester the entire endogenous pool of RalGEFs, and Ras-RalGEF complexes are formed as attested by the activation of Ral in response to insulin. In conclusion, we have shown that although activated Rap2 is able to interact with RalGEFs in cells, neither does it lead to Ral activation, nor can it interfere with signaling from Ras to Ral.

DISCUSSION
In this study, we show that the Ras-related GTPase Rap2 is capable of binding to proteins, the normal function of which is to activate another Ras-related GTPase, Ral, yet this interaction, observed in cells overexpressing Rap2 and RalGEFs, does not lead to formation of the active Ral-GTP complex. This situation is reminiscent of what is observed with another Rasrelated GTPase, Rap1, which binds RalGDS under similar experimental set-ups but does not either lead to Ral activation (12,14). This latter observation was somewhat surprising in view of the fact that biochemical studies had revealed that the C-terminal RID of RalGDS binds Rap1 with a 10-fold higher affinity than Ras, which had led to speculate that RalGDS might well act as an effector of Rap1 function (32). The two qualitative yeast two-hybrid and in vitro assays described here that we performed with the C-terminal RIDs and with the full-length RalGEFs suggest that there is little differential specificity of the Ras, Rap1, and Rap2 GTPases for the three RalGEFs, RalGDS, RGL and Rlf, respectively. This observed promiscuity is not an artifact of the methods used, because Ras and Rap2 interacted specifically with Raf-1 and RPIP8, respectively, in both assays; moreover, RalGEFs did not behave as potential effectors of another Ras-related GTPase, R-Ras (Table I). Besides the near identity of their effector regions, Ras, Rap1, and Rap2 present a high degree of overall sequence conservation (45-55%); moreover, comparison of the three-dimensional structures of Ras and Rap1 (29,51) with the recently established one of Rap2 (30) shows that all three proteins, in their GTP-bound forms, are very similar, especially in their switch I, switch II, and effector regions. Therefore, the apparent molecular promiscuity demonstrated above reflects the biochemical possibility of all three RalGEFs to interact with the GTP-bound form of the three GTPases, Ras, Rap1, and Rap2.
One can, however, formulate several hypotheses as to why Rap proteins are apparently able to bind RalGEFs yet cannot lead to the activation of Ral. The first one is that Rap proteins and RalGEFs may never actually interact in cells. This would be surprising because RalGEFs are principally cytosolic (although we and others consistently find that 10 -20% transfected RalGDS and Rlf are associated with cellular membranes; see Fig. 4 and 6) (38,52) and should therefore be able to access Rap proteins that, similarly to Ras, are bound to membranes via their C-terminal extremities and remain on the cytoplasmic FIG. 7. Overexpression of Rap2-Val-12 fails to interfere in insulin-induced Ral activation. Subconfluent serum-starved A14 cells, transfected with 1 g of pMT2HA-Ral together with either 1 g of pcDNA3 (left two lanes) or 1 g of pcDNA3-Rap2-Val-12 (right two lanes), were stimulated with insulin (1 M for 5 min), which activates Ral in a Ras-dependent manner (15). Activated Ral was isolated from the cell extracts by GST-RalBD precoupled to glutathione beads and detected using a monoclonal RalA antibody (upper panel). As a consequence, activation of both transfected HA-Ral (upper band) and endogenous Ral (lower band) was detected. The expression of transfected HA-Ral and Rap2-Val-12 in total lysates was analyzed by Western blotting using anti-RalA and anti-Rap2 monoclonal antibodies (lower panel).
immunofluorescence as in A in order to simultaneously stain RalGEFs (a, c, e, and g) with fluorescein isothiocyanate and GTPases (b, d, f, and h) with tetramethylrhodamine isothiocyanate-coupled secondary antibodies in co-transfected cells. Images were obtained by confocal immunofluorescence microscopy; similar results were observed in several experiments. face of their respective compartments (47,53,54). In fact, in the experiment that attempts to uncover Rap-stimulated Ral activation, Rap2 and Rlf are overexpressed and can be co-immunoprecipitated attesting of their effective interaction (not shown).
The logical second hypothesis is that Rap-RalGEF interactions are unproductive, i.e. that contrarily to Ras, Rap proteins cannot induce activation of the GEF activity of RalGDS and Rlf on Ral. Yet when posttranslationally modified forms of Rap1 and Ral were incorporated into liposomes, Rap1 was able to stimulate through RalGDS the dissociation of GDP from Ral (14). Moreover, ectopic overexpression of Rap1 and Rlf in COS cells resulted in a 1.5-2-fold activation of coexpressed Ral (72). However, in a similar situation, Rap2 did not lead to Ral activation through RalGDS or Rlf (Fig. 5), although a vast overexpression of activated Rap2 alone was associated with a very modest increase in Ral-GTP (Fig. 7). Therefore, at the biochemical level, Rap1 is able to activate the exchange factor activity of RalGDS and Rlf toward Ral. Our data hint that Rap2 may not be able to do the same, and biochemical experiments performed by reconstituting posttranslationally modified Rap2 and Ral in micelles together with recombinant RalGEFs would be required to formally address this question. As in the case of the interaction of Ras with Raf, it is not yet understood whether targeting effectors to the membrane is sufficient to cause their activation as suggested by grafting membrane targeting sequences to the C terminus of Raf-1 (55,56), PI3K (57), and Rlf (38) and the inability of non-prenylated Ras mutants to activate Rlf in cells (38), or whether conformational changes in the effector induced by interaction with the GTP-bound GTPase also play a role in the activation mechanism as suggested in transfection experiments with Raf-1 (58,59) and NMR spectroscopy experiments with the RID of Ral-GDS (60) and Rlf (61).
There are structural differences in the effector and so-called "extended effector" regions that distinguish Rap2 from Ras and Rap1 that may impair the ability of Rap2 to activate RalGEFs. Rap2 contains a phenylalanine at position 39, in the effector region, instead of the serine found in Ras and Rap1. Although this position does not appear critical from the three-dimensional structure of Rap2 (30), it is involved in the interaction of Rap1A with the Raf-1 RBD (29), and a replacement of serine 39 by phenylalanine reduces the transforming ability of Ras by 3-10-fold (36); one should examine whether this effect is due to a reduced ability of Ras to activate one or several of its effectors, such as Raf-1, RalGEFs, or PI3K. There are also nonconservative substitutions at positions 25 and 43 that exhibit glutamine to threonine and glutamine to glutamic acid substitutions, respectively, in Rap2 as compared with Ras and Rap1, as well as the conservative replacement of valine 44 in Ras and Rap1 by isoleucine in Rap2. It is noteworthy that residues 25 and 43 do not appear to make direct contact with the Raf RBD in its crystal structure complexed to Rap1 (29); however, they could be involved in the interaction of Ras and Rap1 with the cysteine rich domain of Raf-1 (62,63), which is necessary for optimal Ras binding and Raf-1 activation in cells (64,65). The possibility that such residues are also necessary for Ras and Rap1 to activate RalGEFs in cells could be investigated by making the appropriate substitutions in Ras/Rap1 and Rap2 and assessing by transfection their ability to activate Ral via RalGEFs.
Finally, and perhaps most importantly, Rap2 is able to recruit RalGEFs to its resident compartment, the endoplasmic reticulum, as Ras recruits them to the plasma membrane. Yet under these circumstances, Rap-RalGEF complexes might not be able to meet Ral, the subcellular localization of which is quite diverse, because it has been found in plasma membrane fractions and cytoplasmic vesicles, including clathrin-coated and secretory vesicles (66,67). Therefore even if a Rap2-Ral-GEF interaction were to occur under physiological conditions, the different subcellular localizations of the Ras-related proteins, Ras at the plasma membrane and Rap2 at the endoplasmic reticulum (47), would ensure that Rap2 and Ral do not act in the same transduction pathways, whereas activation of Ras leads to the activation of Ral.
Because overexpressed activated Rap2 is able to recruit overexpressed RalGDS and Rlf to the endoplasmic reticulum, one could have expected overexpression of activated Rap2 to sequester RalGEFs away from the plasma membrane and therefore inhibit Ras-dependent Ral activation. This could have represented a mechanism for Rap2, as already suggested for Rap1, to control signaling downstream from Ras. However, we have shown that in a cell line overexpressing the insulin receptor, where Ral is activated in the response to insulin stimulation via endogenous Ras and RalGEFs, the overexpression of Rap2 is unable to interfere with Ras-dependent Ral activation. This is in line with previous observations that overexpression of Rap2 has no effect on the growth-promoting effects of Ras (36). Several plausible explanations include the possibility that a pool of membrane-associated RalGEFs remains in the vicinity of Ras and Ral, due to their association with a membrane microdomain or molecular scaffold, mechanisms that have been suggested to increase the efficiency of signal transduction in mammalian cells (68 -70). Our results, showing that the biochemical interaction promiscuity of Ras, Rap1, and Rap2 GTPases with RalGEFs does not lead to functional promiscuity in cells, suggest that compartmentalization of signaling proteins is of the greatest importance to ensure the functional specificity of signaling pathways.
The possibility of a physiological Rap2-RalGEF interaction raises the question of whether RalGEFs might have another physiological role in addition to stimulating the activation of Ral. In fact, because activated Ral mutants cannot substitute for RalGDS or Rlf to transform cells or activate transcription from the c-fos promoter (13,38,40), RalGEFs probably also exert a Ral-independent function in the Ras signaling pathway. In the case of Rap2 signaling, a yet to be identified partner of RalGEFs, which might be specifically present at the surface of the endoplasmic reticulum, could serve as a target of active Rap2-RalGEF complexes. Whether RalGEFs play a role downstream of Rap2 by activating cellular pathways other than those involving Ral is currently under investigation.