Effector Recognition by the Small GTP-binding Proteins Ras and Ral*

The Ral effector protein RLIP76 (also called RIP/RalBP1) binds to Ral·GTP via a region that shares no sequence homology with the Ras-binding domains of the Ser/Thr kinase c-Raf-1 and the Ral-specific guanine nucleotide exchange factors. Whereas the Ras-binding domains have a similar ubiquitin-like structure, the Ral-binding domain of RLIP was predicted to comprise a coiled-coil region. In order to obtain more information about the specificity and the structural mode of the interaction between Ral and RLIP, we have performed a sequence space and a mutational analysis. The sequence space analysis of a comprehensive nonredundant assembly of Ras-like proteins strongly indicated that positions 36 and 37 in the core of the effector region are tree-determinant positions for all subfamilies of Ras-like proteins and dictate the specificity of the interaction of these GTPases with their effector proteins. Indeed, we could convert the specific interaction with Ras effectors and RLIP by mutating these residues in Ras and Ral. We therefore conclude that positions 36 and 37 are critical for the discrimination between Ras and Ral effectors and that, despite the absence of sequence homology between the Ral-binding and the Ras-binding domains, their mode of interaction is most probably similar.

Ral is a small GTP-binding protein belonging to the subfamily of Ras proteins (1,2), which function as molecular switches in signal transduction pathways by alternating between an active GTP-bound and an inactive GDP-bound conformation. The ratio between the active and inactive form of the Ras proteins is regulated by the action of two types of proteins (3): GTPase activating proteins (GAPs), 1 which inactivate GTPases by stimulating the slow GTPase reaction, and guanine nucleotide exchange factors (GEFs), which activate GTPases by stim-ulating the slow GDP dissociation rate, allowing the protein to rapidly come at equilibrium with the cellular pool of guanine nucleotide (4). The guanine nucleotide affinities and the intrinsic GTPase activity of Ral are very similar to that of Ras (5). Ral is ubiquitously expressed, but it is especially abundant in brain, testis, and platelets (6 -10). The protein was found both in the plasma membrane and in cytoplasmic vesicles (11,12).
The cellular function of Ral remained elusive for a long time. Recently, however, progress was made via the discovery of Ral-specific regulatory activities. Multiple Ral-specific guanine nucleotide exchange factors (RalGEFs) have been isolated (13)(14)(15)(16)(17)(18). Furthermore, although so far the corresponding genes have not been isolated, Ral-specific GAP activities were identified in brain, testis, and platelets (19 -20). Strikingly, in addition to a Ral-specific GEF domain, the RalGEFs contain a C-terminally located Ras-binding domain (RBD), which is able to bind the GTP-bound forms of Ras and Rap proteins, so that the RalGEFs were recognized as one of the new families of Ras protein effectors (Ref. 21 and references therein). The RBDs of RalGDS (22,23) and Rlf (24) were shown to be structurally similar to the RBD of the Ser/Thr protein kinase c-Raf-1 (25)(26)(27), a well characterized Ras-effector protein. Interestingly, a Ras-dependent stimulation of Ral was proposed to function parallel to the Ras-Raf-Mek-Erk pathway in several types of cells (28 -35), but in platelets, Ral appears to be stimulated in a fashion similar to Rap1A, suggesting that also Rap1A can function upstream of Ral (36). In addition to the regulatory proteins, a putative Ral effector protein was identified: RLIP76, also named RalBP1 or RIP (37)(38)(39). Its Ral-binding domain (RalBD) was identified by deletion studies (37) and was used successfully as a probe in a pull-down assay to measure the activation of endogenous Ral upon cellular stimulation (35,36). Finally, Ral was shown to be involved in the phorbol ester-stimulated phospholipase D activity, apparently through a direct interaction of Ral with phospholipase D 1 (28, 30, 40 -42). Additional complexity in the Ral pathway(s) emerged with the isolation of the RLIP76-binding protein Reps1, which can also bind to the adaptor proteins Crk and Grb2 (43), and with the observation that calmodulin binds to RalA (44) and that both Ral and Arf are needed for phospholipase D activation (45).
Even though Ral interacts with RLIP76, it is largely unknown how this interaction takes place. Secondary structure prediction of RLIP76 indicated that the structure of RalBD comprises a coiled-coil region (37) and thus differs from the ubiquitin-like fold of the RBDs. In this work, we have investigated the effects of several mutations on the interaction between Ral and RLIP76 by double hybrid analysis and biochemical methods. Moreover, the specificity of the interaction of Ras and Ral proteins with their effectors was studied by sequence space analysis of the Ras-like protein sequences. The foremost result is that Lys-47 and Ala-48 (corresponding to Ras residues Ile-36 and Glu-37) in the effector region enable Ral to discriminate between RLIP76 and Ras effector molecules. Ral(K47I) is able to interact almost as potently as Ras with RalGEF-RBDs, whereas mutation E37A in Ras is sufficient for a significantly interaction with RLIP76. We thus conclude that, as opposed to the other members of the Ras-subfamily, Ral is unable to interact with Ras effector molecules due to critical differences in the effector region. Furthermore, introduction of mutations in the effector region may induce erroneous binding to effectors specific for other GTP-binding proteins. Finally, the fact that mutation of only two residues changes binding specificity suggests that the modes of protein-protein interaction in the Ral⅐RLIP and in the Ras⅐RBD complexes are similar.
Ral and Ras proteins were loaded with Gpp(NH)p or mGpp(NH)p (the derivative of 5Ј-guanylylimidodiphosphate, carrying the fluorescent N-methylanthraniloyl-group at the 2Ј-or 3Ј-position of the ribose (mixture of both isoforms)) by incubation with a 4-fold excess of (m)Gpp(NH)p in 50 mM Tris-HCl, pH 7.6, 200 mM (NH 4 ) 2 SO 4 , 10 mM EDTA, 5 mM dithioerythritol, and 5 units of alkaline phosphatase per mg of Ral or Ras for 1 h at room temperature. Subsequently, separation of the Ral or Ras protein from unbound nucleotides was obtained by gel filtration in Buffer B. Thereafter, the protein was concentrated using Vivaspin vials, and the concentration of Ral⅐(m)Gpp(NH)p was determined by high pressure liquid chromatography analysis.
Two-hybrid Analysis-The wild type and mutated Ral and Ras genes were subcloned by BamHI/SalI fragments into pVJL10, a derivative of pBTM116 (47). Fragments containing the RalBD or the RBD of c-Raf-1 (amino acids 51-131) were subcloned by PCR in the pGAD3S2X vector. The Ras-binding domain of mRalGDS (amino acids 702-852) or mRlf (amino acids 607-778) was expressed from pGAD1318. Plasmids pGAD3S2X and pGAD1318 are derivatives of the two-hybrid vector pGAD-GH.
2H assays were performed as described (37). L40 yeast cells transformed with plasmids allowing expression of LexA-fused proteins were mated with AMR70 cells transformed with plasmids allowing expression of proteins fused to the GAL4 activation domain. Diploids were tested for histidine prototrophy and LacZ expression. Quantification of ␤-galactosidase activity was performed as described (48).
GST Pull-down Experiment-400 g of GST-RalBD was bound to 400 l of GSH-Sepharose beads in 3 ml of Buffer C. Beads were collected by centrifugation and washed with the same buffer. For each assay 5 M of sRalA wt, hRalB wt, or hRalB mutants, loaded with GDP or Gpp(NH)p, were incubated in 1 ml with 3.5 M GST-RalBD bound to GSH-Sepharose beads for 1 h at 4°C. Beads were collected by centrifugation, washed three times with Buffer C, and resuspended with 15 l of SDS-sample buffer. The proteins were separated by 15% SDS-polyacrylamide gel electrophoresis and detected by Coomassie Brilliant Blue staining.
Fluorescence Measurements-The interaction of hRalB wt and the hRalB(K47I/A48E) mutant with the Rlf-, Rgl-, and RalGDS-RBDs was characterized with a Perkin-Elmer fluorescence spectrometer LS50B as described before (24,49). Measurements were performed in Buffer B containing 5% glycerol at 37°C; excitation wavelength was 366 nm, and emission wavelength was 450 nm. The determined dissociation constants were corrected with the percentage of the activity of the effector, which was specified by an active site titration as described (49).
Sequence and Structural Analysis-An assembly of 476 Ras sequences was collected from different data bases with the help of the Genequiz software (50). A comprehensive nonredundant alignment of these sequences was build with CLUSTALW (51) using the BLOSUM62 matrix (52). Only sequences with maximal 80% similarity were accepted, with the exception of Ral: only 8 Ral sequences were found in the data bases, all of which were accepted in the alignment. The final multiple sequence alignment that was obtained after some hand editing included 179 sequences, corresponding to 98 Rab, 36 Rho, 37 Ras, and 8 Ral sequences. This alignment can be found at http://www.cnb.uam. es/ϳcnbprot/ral.dir.
The detection of those residues that allot specificity to a subfamily, i.e. those residues that are conserved within a given subfamily and differ from the other subfamilies, which are named tree-determinant residues, was carried out with SequenceSpace (53). These tree-determinant residues are likely to be responsible for the functional differences between protein subfamilies. This approach was used in different systems, including the ␣-subunits of the trimeric G-proteins (54), SH2 domains (55), protein kinases, 2 and alcohol dehydrogenases (56), among others. In all these cases, the positions of the tree-determinant residues fit nicely with known protein interaction sites, e.g. the peptide substrate binding sites in the SH2 domains.
Surface accessibility of residues in the GDP-bound conformation of sRalA, 3 and the GDP-and the GTP-bound conformation of Ras (Protein Data Bank codes 4q21 and 5p21, respectively) was calculated with the program DSSP (57).

Isolation of the Complex of sRalA⅐Gpp(NH)p with RLIP76
The first trials to isolate the complex by incubating the purified proteins sRalA⅐Gpp(NH)p and RalBD, and subsequent purification over gel filtration column in 50 mM Tris-HCl, pH 7.6, 10 mM MgCl 2 , 150 mM NaCl, 5 mM dithioerythritol, were unsuccessful. Under these conditions, Ral behaved as a dimer and RalBD as a multimer (not shown). Consequently, a proper separation of the complex from the single proteins was not possible, and we could not be certain that a complex was formed.
Therefore, we modified our strategy. We bound GST-RalBD to a GSH-Sepharose column, loaded sRalA⅐Gpp(NH)p to the column, and, after washing in order to remove unbound Ral, cleaved with thrombin and eluted the complex from the column. Only with sRalA⅐Gpp(NH)p could a complex be eluted (Fig. 1); one could not be eluted with sRalA⅐GDP (not shown), demonstrating that the interaction between Ral and RalBD is GTP-specific, consistent with earlier findings with different techniques (37)(38)(39).
The complex-containing fractions of the GSH-Sepharose elution were further purified by gel filtration to remove noncomplexed proteins and to determine the ratio in which the com-plex is build. A partial separation of noncomplexed RalBD, which here behaved as a dimer, from the complex could be obtained by a first Superdex-75 chromatography step (Fig. 1A). The pooled fractions containing the complex (fractions 18 -22) were subjected to a second gel filtration, after which a nearly pure complex (fractions 20 -22) could be obtained (Fig. 1B). The complex eluted as a 33-kDa protein, showing it to be a 1:1 complex of the Ral protein (20.4 kDa) and RalBD (14.7 kDa). It was noted that the nonfused form of RalBD runs with a too high apparent molecular mass (Fig. 1), whereas GST-fused RalBD runs normally on SDS-polyacrylamide gel electrophoresis ( Fig. 2) (molecular mass, 40.9 kDa).

In Vitro Characterization of the Ral-RLIP76 Interaction
It was shown by fluorescence measurements for several effector molecules that they can inhibit the intrinsic dissociation rate of the GTP-bound complex of Ras-like proteins in a concentration-dependent manner (15,49,58). However, addition of up to 10 M RalBD did not lead to any effect on the intrinsic dissociation rate of sRalA⅐mGpp(NH)p or hRalB⅐mGpp(NH)p (not shown), nor to a change in emission spectrum.
Because fluorescence measurements did not allow the characterization of the interaction, we used a pull-down assay to analyze the binding of Ral to RLIP in vitro. We used a GSTfused RalBD and full-length RLIP76 as baits and tested the binding to Ral preloaded with either GDP or Gpp(NH)p. Partially purified GST-coupled full-length Rlip76 was able to specifically bind the Gpp(NH)p-bound forms of sRalA and hRalB (not shown). Also, the truncated form, RalBD, can interact specifically with the Gpp(NH)p-bound forms of sRalA and hRalB (Fig. 2). In comparison, Ras does not bind to RalBD.

What Makes the Specificity of the Interaction with Effector Molecules A Theoretical Approach
The complete functional interchange between a Ras protein and a Ral protein requires a large number of changes equal to the number of differences between the sequences. In order to select among them those residues that will affect the specificity of the interaction of Ras and Ral with their regulating proteins, we have chosen a general approach described before (53). After building a representative alignment of Ras-like proteins (see under "Experimental Procedures"), we could define those positions that correspond to good tree-determinants, i.e. that are conserved within a subfamily and differ from the other subfamilies. In addition, tree-determinants that are in the protein interior were eliminated as putative participants in the binding site. Finally, because we were mostly interested in the specific interaction with effector proteins, the exposed tree-determi-2 P. Bork, C. Sander, and A. Valencia, unpublished results. 3 I. R. Vetter, manuscript in preparation. nants positions that do not change their surface exposition between the GTP-and GDP-bound states were eliminated.
Sequence space analysis was performed as described (53). A comparison of the Ras, Rho, and Rab subfamilies and of the Ras ϪRal (comprising the Ras and Rap proteins, but not Ral), Ral, and Rho subfamilies is depicted in Fig. 3. Only 14 of the completely conserved residues in the eight known Ral sequences are different in the other subfamilies (Ral-tree-determinants), namely positions 7, 24,25,33,36,37,43,46,53,67,70,92,93, and 160 ( Fig. 3A; note that we used the Ras numbering in these comparisons). When comparing the Ras ϪRal , Ral, Rho, and Rab subfamilies, only position 37 is conserved but different in each subfamily (Fig. 3): Glu in Ras ϪRal (approximately 90% conserved), Ala in Ral (100%), Phe in Rho (nearly 100%), and Gly in Rab (nearly 100%). Position 36 is also conserved in Ral (Lys; 100%) and Rho (Val; Ͼ95%), and conserved but shared in Ras ϪRal and Rab (Ile; Ͼ 95% conserved). These two positions thus are the best tree-determinants for each of these subfamilies. When comparing the Ral, Ras ϪRal , and Rho subfamily, other tree-determinants specific of different subsets of families can be detected, e.g. position 92, which is conserved but different in each subfamily: Glu/Asp in Ras ϪRal (approximately 70% conserved), Ala in Ral (100%), and Asn in Rho (approximately 80%). A residue that is shared by the Ras ϪRal and Rho subfamily and different from Ral, is position 53: Leu in Ras ϪRal and Rho (nearly 100% conserved) and Ile in Ral The analysis of the structural characteristics (surface exposition and change of solvent accessibility upon the transition of the GDP-to GTP-bound conformation) extends the tree-determinant analysis. Five of the 14 Ral tree-determinant residues (positions 7, 24, 46, 53, and 160) represent conservative substitutions between apolar residues in different subfamilies, and all of them are part of the protein interior. Position 93 is occupied mostly by small apolar residues in all the subfamilies, except Ral, in which position 93 is a Thr. However, it does not seem to be a good candidate because it is neither exposed in Ral⅐GDP nor in the GDP-and GTP-bound conformation of Ras. Position 92, which is an Ala in Ral but a polar residue in the other families, is also not very exposed in Ral or Ras. The solvent accessibility of most of the remaining Ral tree-determinant positions that are exposed is affected by the change between the GDP and GTP states. Position 25 is exposed but has a similar solvent accessibility in the GDP-and the GTP-bound state. Moreover, position 25 is occupied by a polar residue in all subfamilies. These characteristics make the participation of position 25 in the specific interaction with the effector less likely. Position 43 has a larger solvent accessibility in the Ras⅐GDP conformation, whereas positions 33, 36, 37, 67, and 70 have a larger solvent accessibility in the Ras⅐GTP conformation. The latter positions thus appear to offer more interacting surface in the GTP-bound state than in the GDP-bound conformation and are consequently, based on the sequence and structure analysis, good candidates for the interaction with the effector.
Taken together, five positions fit to our criteria: positions 33, 36, 37, 67, and 70. When limiting our view to Ras ϪRal and Ral, positions 33 (Asp in Ras ϪRal and Glu in Ral), 67 (Met in Ras ϪRal and Ile in Ral), and 70 (Gln in Ras ϪRal and Asn in Ral) contain conservative changes between Ras ϪRal and Ral, whereas positions 36 (Ile in Ras ϪRal and Lys in Ral) and 37 (Glu in Ras ϪRal and Ala in Ral) differ dramatically. Thus, the latter two residues are our best candidates for the functional conversion between Ras and Ral. Both residues 36 and 37 are exposed and change substantially between the GDP-and GTP-bound conformations of Ras and are part of the interaction between Ras and the Raf-RBD (27). Remarkably, these positions also appear to be the best tree-determinant positions for all subfamilies.

A Practical Approach
Mutational Analysis of the Ral-RalBD Interaction-A series of hRalB variants carrying mutations in the effector region was created in order to analyze the interaction between Ral and RLIP. Fig. 3C shows the alignment between the effector regions of the diverse subfamilies of small GTP-binding proteins and indicates which mutations were introduced in Ral. The effects of the mutations were tested by 2H analysis and by the pull-down assay.
As became clear from the 2H analysis shown in Fig. 4, mutations in Ral residues 41, 42, 43, and 47 did not affect significantly the interaction of Ral with RalBD. Weaker binding to RalBD was observed with mutants hRalB(A48E) and hRalB(D49N), whereas double mutant hRalB(K47I/A48E) did not show any binding to RalBD. Similarly, in our pull-down  (53). The sequence of each subfamily is represented as a vector point in a multidimensional space (sequence space), with residue positions and types as the basic dimensions. Directions in sequence space represent specific sequence patterns and are depicted as planar projections. The tree-determinant residues for each subfamily can be recognized as the extreme positions on the subfamily axis and are here indicated by circled clusters. In addition, the positions of the treedeterminant residues are numbered near these clusters. Positions 53 and 130 are conserved but shared by the Rho and Ras ϪRal subfamilies and are located between these subfamily axes. Positions shared with Ral were not considered due to the low number of Ral sequences. The positions depicted in boldface are tree-determinant residues for each subfamily or shared by two subfamilies. B, sequence space analysis of the Ras, Rho, and Rab subfamilies. Note that whereas position 37 is conserved for all subfamilies and located within the tree-determinant cluster of each subfamily, position 36 is conserved for all subfamiles but shared between the Ras and Rab subfamilies. Consequently, it is located between the clusters of these subfamilies. C, sequence alignment of four Ras subfamilies (Rab, Rho, Ras ϪRal , and Ral). The number of protein sequences used for the analysis are indicated in parentheses. The numbering of the Ras residues is indicated above the alignment. Positions conserved in each subfamily are represented by the corresponding letter(s). The two best tree-determinants in positions 36 and 37 are framed according to chemical similarity. X indicates a variable position. The mutated positions described in this work are indicated by arrows.
assay, we observed that mutation A48E strongly affects the interaction, whereas the double mutation K47I/A48E practically abrogates the binding of hRalB to RalBD (Fig. 2).
As expected, the dominant-negative mutant hRalB(S28N), which is homologous to Ras(S17N), did not show any interaction with the effector molecule RLIP in the 2H analysis (Fig. 4).
The Nature of Ral Residues 47 and 48 Determines the Interaction with Effector Molecules-Our 2H analysis of the effector mutants of Ral shows that the double mutation K47I/A48E abrogates the interaction of Ral with RalBD. In addition, the double mutation enables Ral to bind RBDs of several Rasspecific effector molecules, e.g. the Ser/Thr kinase c-Raf-1 and the Ral-specific guanine nucleotide exchange factors RalGDS and Rlf (Fig. 4). Apparently, these residues are essential to hinder the interaction of Ral with Ras effectors. As can be seen in Fig. 3, hRalB(Lys-47) and hRalB(Ala-48) are homologous to Ras(Ile-36) and Ras(Glu-37), respectively. Thus, the double mutation K47I/A48E turns Ral into a Ras-like protein with respect to the amino acid sequence in the effector region and to the interaction with effector molecules. Interestingly, mutation K47I is sufficient to enable Ral to interact with the RalGEF-RBDs, whereas the extra mutation A48E is necessary in order to induce interaction with Raf-RBD.
We have determined the affinity of the mutant protein hRalB(K47I/A48E) for the RBDs of RalGDS, Rgl, and Rlf by a fluorescence assay as described (15,49,58). This fluorescence assay is based on the observation that RBDs are capable to decrease the dissociation rate of the GTP-bound form of Ras proteins and uses the fluorescently labeled GTP analogue mGpp(NH)p (Fig. 5). Whereas RalGDS-and Rgl-RBDs bind to the mutant with a low affinity, as shown by the dissociation constants of 7.0 and 15.2 M, respectively, Rlf-RBD shows an affinity comparable to that of Ras and Rap1A wild type (Table  I). This underlines the observation that the Ras-binding domain of Rlf differs in a number of aspects from that of the other two RalGEF proteins (24).
We undertook the reciprocal experiment and tried to turn a Ras in a Ral. An identical 2H assay to test interaction between RalBD or Ras effectors on one hand and Ras wt, Ras(I36K), Ras(E37A), or Ras(I36K/E37A) on the other hand showed only weak interaction (not shown). Therefore, we decided to measure these interactions with the more quantitative ␤-galactosidase assay in solution (59), as well with histidine prototrophy. As can be seen in Fig. 6, in this assay, the effects of mutations K47I, A48E, and K47I/A48E in hRalB on the interactions with RalBD and Ras effectors are similar, as determined with the 2H and pull-down assays (Figs. 2 and 4). The reversed mutations I36K, E37A, and I36K/E37A in Ras cause, to a large extent, the expected, opposite effects. Importantly, mutation E37A is apparently enough to induce a weak but significant interaction between Ras and RLIP. This is in agreement with the observation that mutation A48E in Ral strongly reduces the interaction with RLIP. Furthermore, Lys47 in Ral appears not to be essential for RLIP binding, but to be a requisite to prevent interaction with Ras effectors. Similarly, mutation I36K reduces the binding of Ras to RalGEFs (and to some extent to Raf), but does not induce binding of Ras to RLIP. The combination of both mutations induces in Ral a full shift in affinity from RLIP toward the Ras effectors, and in Ras an abrogation of the interaction with the Ras effectors and a weak but significant induction of binding to RLIP. These results are confirmed by the selection on His Ϫ plates (Fig. 7). DISCUSSION For several years after the discovery of Ral, overexpression of putatively activated or dominant-negative Ral variants did not lead to any obvious cellular phenotype, and thus the cellular function of Ral remained elusive. This situation changed upon the discovery that RalGEFs were potential Ras effector proteins (60,61). Soon, it became clear that Ral was involved in Ras-dependent cellular transformation parallel to the Ras-Raf pathway (28 -35). Direct detection of the GTP-form of endogenous Ral was made possible by pull-down assays using the GST-coupled form of RalBD, the Ral-binding domain of the Ral-specific effector protein RLIP76. This way, it could be demonstrated that Ral is activated by a Ras-dependent pathway in Chinese hamster ovary cells (35) and possibly by a Rap1-dependent pathway in platelets (36).
In this work, we have analyzed the interaction between Ral and the effector protein RLIP76. For the first time, we were able to isolate and purify the complex between Ral and the RalBD. Our results suggest that RalBD occurs as a homodimer that dissociates upon binding to Ral. Thus, binding to Ral induces a monomerization of RalBD. This may have a physiological function, e.g. to render the GAP region accessible to its substrate as proposed before (37). On the other hand, in vitro experiments suggested that the binding of Ral to RalBD does not affect the GAP activity of RLIP76 on Rac (62).
Earlier, it was shown that mutation T46A (for comparison, the homologous residue in Ras will be given in parentheses) (Ras Thr-35) inhibits the interaction between Ral and RLIP (37). We have further dissected the interaction between Ral and RLIP76 by testing a series of Ral mutants in a 2H assay, as shown in Fig. 4. Mutations E41K (Ras Asp-30) and D42K (Ras Glu-31) do not seem to affect the interaction of Ral with RLIP. This is different from the homologous mutations in Ras, because position 31 is of crucial importance for the interaction with Ras effectors: the wild type residue Glu-31 confers a high affinity for c-Raf-1, whereas Ras(E31K), which mimics Rap, FIG. 4. Double hybrid analysis of Ral effector mutants. L40/ AMR70 diploid cells expressing various combinations of LexA-fused proteins (rows) and GAL4 activation domain-fused proteins (columns) were tested for ␤-galactosidase activity using a filter method.
Interestingly, mutation K47I (Ras Ile-36) hardly affects the interaction of Ral with RLIP but enables Ral to interact with Ras effector proteins RalGDS and Rlf (Fig. 4). In combination with mutation A48E, which strongly reduces the Ral-RLIP interaction, double mutant hRalB(K47I/A48E) has practically no affinity for RLIP (Figs. 2, 4, 6, and 7). At the same time, this double mutant is able to interact with the Ras-specific effector molecules, of which at least the RalGEFs are bound with affinities that are comparable to those of Ras (Table I). On the other hand, introduction of mutations I36K and E37A abrogates the interaction of Ras with its effector molecules but enables Ras to interact with RLIP, even though the interaction is still weak ( Fig. 6 and 7).
Sequence space analysis shows that Ras positions 36 and 37 (Ral residues 47 and 48) are tree-determinant positions for the Ras, Ral, Rho, and Rab subfamilies because they are conserved in each subfamily but differ between the subfamilies (position 36 is shared by the Rab and Ras subfamilies) (Fig. 3b). Thus, in agreement with our experimental data, this analysis indicates these positions to be the most likely candidates that determine the specificity of the interaction of Ras and Ral with their effector(s).
The crystal structure of sRalA⅐GDP, which will be described in detail elsewhere, 3 shows that the three-dimensional structures of Ras and Ral are quite similar, as expected from the high sequence similarity. The main structural differences between Ral and Ras are located in the switch II region, which shows a shift of the ␣-helix comprising residues 78 -85 of sRalA (Ras 67-74). In contrast, the switch I regions of Ral and Ras superimpose very well, with a root mean square deviation of 0.3 143 for 11 C ␣ -atoms. In Fig. 8, we have depicted the electrostatic potentials of the surfaces of Ha-Ras (in the GDP-and GTP-bound conformation) with that of sRalA and the modeled structure of the double mutant RalA(K47I/A48E), with the switch I region toward the reader. When comparing the GDPbound structures of Ras and Ral, it becomes evident that the positive charge of Lys-47 in Ral and the negative charge of Glu-37 in Ras represents the main difference in the switch I region between these proteins, in accordance with our mutational analysis.
Remarkably, Ral and its partners RalGDS and RLIP do not exist in Saccharomyces cerevisiae or in any other unicellular eukaryotes (as far as genomic sequences are known), an evident discrepancy with Ras and the Ras-MAP kinase signaling modules. The Ras-Ral signaling module is thus of late appearance in evolution, and one is tempted to correlate the appearance of this Ras pathway with the emergence of multicellularity-perhaps at the same time as the receptor tyrosine kinase appearance in evolution. This aspect underlines the uniqueness of Ral within the Ras subfamily, as shown in this work by theoretical and experimental means.
In conclusion, our analysis shows that Lys-47 in Ral, even though not essential for interaction with RLIP76, is a requisite to prevent interaction of Ral with Ras effector proteins. The neighboring residue, Ala-48, in its turn, is important for the  6. Histogram of ␤-galactosidase activity as a tool to measure the interaction between Ras and Ral mutants with Ras and Ral effector proteins. Student's t test was performed to determine the significance of the interaction of Ras(E37A) and Ras(I36K/E37A) with RLIP and Raf relative to the signal observed with the empty vector. The resulting data are indicated in the histograms (Ͻ0.05, significant; Ͼ0.05, insignificant). FIG. 7. Double hybrid analysis of Ras mutants: His ؊ selection. L40/AMR70 diploid cells expressing various combinations of LexAfused proteins (rows) and GAL4 activation domain-fused proteins (columns) were patched on DO-WL (drop-out medium depleted of Trp and Leu) plates and replica-plated on DO-WLH (drop-out medium depleted of Trp, Leu, and His) and DO-WL plates to be tested for histidine prototrophy. All patches grew on DO-WL (data not shown). Pictures were taken of after 3 days at 30°C. interaction with RLIP. Mutation of these residues to the Raslike amino acids enables Ral to interact with the Ras effector proteins Raf and RalGEFs, whereas the opposite mutations enables Ras to interact with RLIP. Our results thus indicate that effects of mutations in the effector region of small GTPases should be interpreted with care, because interactions with other effector proteins may be induced. In this light, it seems possible that the remarkable dominant negative effects of Rac(Q61L/F37A) may not be caused by unproductive interactions with Rac effectors (69) but by productive or unproductive interactions with effectors of other small GTPases. Last but not least, despite a differently predicted secondary structure for RLIP, our results suggest that the mode of interaction between RLIP76 and Ral is similar to the mode of the interaction between RBD and Ras.