Bridging Ral GTPase to Rho pathways. RLIP76, a Ral effector with CDC42/Rac GTPase-activating protein activity.

RalA and RalB are GTPases of unknown function and are activated by proteins, RalGDS, that interact with the active form of another GTPase, Ras. To elucidate Ral function, we have searched for proteins interacting with an activated form of RalA using the two-hybrid method and a Jurkat cell library. We have identified a partial cDNA encoding a protein, RLIP1, which binds to activated RalA and this binding requires an intact effector domain of RalA. Biochemical data with purified RalA confirm the genetic results. This protein also bears a region of homology with GTPase-activating protein (GAP) domains that are involved in the regulation of GTPases of the Rho family and, indeed, RLIP1 displays a GAP activity acting upon Rac1 and CDC42, but not RhoA. This GAP region is not required for RLIP1 binding to Ral. The whole cDNA was cloned, and it encodes a 76-kDa polypeptide, RLIP76, which also binds RalA. The Rho pathway is involved in membrane and cytoskeleton modifications after mitogenic stimulation and acts in parallel to and synergistically with the Ras pathway. We propose that these pathways are linked through a cascade composed of Ras RalGDS Ral RLIP76 CDC42/Rac1/Rho, allowing modulation of the Rho pathway by the Ras pathway.

Ral proteins are biochemically well characterized GTPases whose functions have long remained elusive (1,2). A potential clue was provided by the finding that RalGDS 1 and a RalGDSlike protein, which are activators of RalA and RalB (3), interacts with the activated form of Ras and that this interaction requires the integrity of the Ras effector domain (4 -6). Thus RalGDS, and therefore Ral proteins, might be involved in transducing pathways that signal through Ras.
In order to decipher Ral function, we have searched for proteins that interact with the activated form of RalA. Using a two-hybrid method and a mutant of RalA deficient in its intrinsic GTPase activity (RalAV23), we have isolated a partial cDNA encoding a protein (RLIP1, Ral interacting protein 1) that has characteristics of a Ral effector protein.
The whole cDNA was isolated and sequenced; it contains an ORF encoding a predicted 76-kDa protein (RLIP76) that binds RalA.
Out of the Ral binding region, RLIP76 contains a GAP region related to RhoGAP domains and this structural homology reflects a functional homology with a GAP activity acting upon CDC42HS and Rac1.

EXPERIMENTAL PROCEDURES
Two-hybrid Screen-The two-hybrid system and the Jurkat cells library used in this study have already been described (7)(8)(9). For LexA fusion protein expression, we have used plasmids pBTM116 or derivatives with modified polylinkers. When necessary, PCR using Pfu polymerase was performed to generate adequate cloning sites. The bait of our screen was a fusion between LexA and a Val-23 mutant of RalA (equivalent to RasV12) deleted of its 27 C-terminal amino acids (RalAV23⌬CT) supposed to be involved in post-translational modifications and membrane localization. This fusion protein was expressed in yeast from plasmid pLRTA. LexA fusion proteins expression was checked on Western blots with anti-LexA antibodies (a gift from P. Moreau, Gif sur Yvette, France).
Yeast and the two-hybrid procedures were handled according to published methods (9,10). Library plasmids from transformed yeast colonies were recovered using HB101 as a recipient strain, selected on M9 medium lacking leucine. When mating was used for two-hybrid tests, strain L40 was mated with strain AMR70 (MAT␣, leu2, trp1, his3, ade2, URA3::lexAop-lacZ) (a gift from S. Fields). When two-hybrid results are presented, we are showing the results of ␤-galactosidase test on filter paper. There was no discrepancy between the His auxotrophy test and the ␤-galactosidase test.
When required, point mutations were introduced using the Transformer site-directed mutagenesis kit (Clontech). Any DNA fragment submitted to mutagenesis and all PCR products were sequenced.
Gene Expression-Gene expression was analyzed on a multiple tissue Northern blot (Clontech) where mRNAs from heart, brain, placenta, lung, liver, skeletal muscle, kidney, and pancreas were represented. A ␤-actin cDNA was used as a control.
cDNA Cloning-5Ј-RACE was carried out using 5Ј-RACE-ready cDNA from Clontech and following manufacturer's instructions. Isolation of cDNA from phage was performed according to usual techniques (11).
In Vitro Transcription/Translation of RLIP1-A PCR reaction was carried out with pRLIP1 as a template and adequate primers. The 5Ј primer contains a T3 RNA polymerase recognition site upstream of an ATG initiation codon in frame with RLIP1 (CGAATTAACCCTCACTA-AAGAAGGATGGAGATCCTAGAACTAGTCGG) (12). The 3Ј primer is downstream of the stop codon of the amplified fragment (GTAAAAC-* This work was supported in part by grants from Association pour la Recherche contre le Cancer (ARC), Ligue Nationale contre le Cancer, Ligue contre le Cancer (Comité de Paris), and Groupement de Recherche et d'Etudes sur les Génomes. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EMBL Data Bank with accession number(s) L42542. Transcription and translation in presence of [ 35 S]methionine were performed using 1 g of PCR product, and, sequentially, a mRNA capping kit and an in vitro translation kit (Stratagene).
Preparation of Proteins from Escherichia coli and in Vitro Binding Experiments-GST and of GST-Ral proteins from E. coli transformed with native or recombinant plasmid pGEX-4T1 were prepared following classical methods. All buffers contained 5 mM MgCl 2 . For in vitro binding studies, 5 g of glutathione-Sepharose 4B-bound proteins were washed twice in ice-cold binding buffer (20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 5 mM MgCl 2 , 1 mM Pefabloc, and 0.5% Nonidet P-40) and incubated overnight at 4°C with 1 l of in vitro translated 35 S-RLIP1 in 50 l of binding buffer containing 0.5 mM GTP (experiments with GST and GST-RalAV23), or GDP (in the cases of GST-Ral). After sedimentation of the beads, the supernatant ("the unbound fraction") was removed and the beads were washed three times with binding buffer containing 1 mM dithiothreitol. "Bound proteins" were recovered by boiling the beads in sample buffer. Unbound and bound fractions were subjected to SDS-PAGE on a 10% acrylamide gel. After staining with Coomassie Blue to detect the GST and GST-Ral proteins, gels were treated with Amplify (Amersham Corp.) and dried, and the presence of 35 S-RLIP1 was detected by autoradiography.
In vitro binding studies after guanine nucleotide exchange were performed as described above with the following alterations; 10 g of glutathione-Sepharose 4B-bound proteins were washed twice in ice-cold exchange buffer (20 mM Tris-HCl, pH 7.5, 10 mM EDTA, 5 mM MgCl 2 , 1 mM dithiothreitol, 1 mM Pefabloc), and incubated for 3 h at 4°C in 50 l of exchange buffer containing 90 M GTP or GDP. The nucleotide exchange reaction was stopped by adding MgCl 2 to 20 mM followed by two washes with binding buffer containing 20 mM MgCl 2 .
In all cases, protein concentrations were estimated by Coomassie Blue staining of SDS-PAGE gels and adjustments were made for the same amount to be used in all experiments.
Protein Purification and GTP Hydrolysis Assay-RLIP1 was expressed as a MBP fusion protein from vector pMal-c2 (New England Biolabs). Rac1 and Bcr-GAP proteins were expressed in E. coli and purified as GST fusion proteins, then clipped off with thrombin. CDC42 was a gift from P. Boquet. [␥-32 P]GTP-bound Rac1, CDC42, and Rap2A were prepared by incubating 200 nM protein with 25 mM Tris, pH 7.5, 5 mM EDTA, 0.2 mM MgCl 2 , 0.1 mg/ml bovine serum albumin, 1 mM dithiothreitol, 10 mM [␥-32 P]GTP (2 mCi to 30 Ci/mmol, DuPont NEN) in a 50-l volume for 15 min at room temperature. GTP hydrolysis was initiated by raising MgCl 2 and GTP to final concentrations of 20 mM and 200 M, respectively. GTP hydrolysis was stopped at different time points by addition of 2 ml of 50 mM ice-cold Tris, pH 8, 35 mM MgCl 2 , 1 mM dithiothreitol, 150 mM NaCl, then quantitated by rapid vacuum filtration on BA 85 nitrocellulose filter and radioactivity counting (derived from Ref. 13). The GTP hydrolysis was conducted as described above, but in the presence or absence of Bcr-GAP (400 nM), or MBP-RLIP1 (800 nM), and stopped after a 10-min incubation at room temperature.
FISH Analysis-Fluorescence in situ hybridization (FISH) to metaphase chromosomes prepared from a normal male was carried out according to an usual technique (14). The probe was a 1.6-kilobase pair DNA fragment from one of the inserts isolated during the two-hybrid screen.
Sequence Analysis-We used a Sequenase sequencing kit (Amersham Corp.) or a deaza-T7-sequencing kit (Pharmacia Biotech Inc.) for classical sequencing, or a Dye-deoxy terminator kit (Perkin Elmer) and an Applied Biosystems model 373A automatic sequencer. Sequence analysis was performed using computer facilities provided by the Centre Interuniversitaire de Traitement Informatique (CITI2) (15).

RESULTS
Two-hybrid Screen-Around 1,000,000 colonies were screened with RalAV23⌬CT as a "bait" and a Jurkat cells library. Two library plasmids, pRLIP1 and pRLIP2, were recovered that contained partial cDNAs expressing proteins RLIP1 and RLIP2 (for RRall interacting protein) fused to GAL4 activation domain, respectively. RLIP1 and RLIP2 interact specifically with RalAV23⌬CT, as opposed to several irrelevant proteins (lamin, stathmin (Ref. 16), hSos1 (Ref. 17)) (data not shown). RLIP2 is a C-terminal part of RLIP1.
Ral allele dependence of the interaction was checked using pRLIP1. RalAV23, RalAV23A46 (an effector domain mutant), RalAA26 (which mimics a Ras GDP-blocked mutant), and RalAV23A26 were cloned in pBTM116. Fig. 1A shows the signals displayed in a ␤-galactosidase test. First, RLIP1 is able to interact not only with RalAV23⌬CT (data not shown) but also with RalAV23 and RalAwt. Second, RLIP1 interaction with RalAA26 or with RalAV23A26 is undetectable. Since a G26A mutation is supposed to block RalA in a GDP-bound state, this result suggests that RLIP1 binds to RalA only when this latter is bound to GTP and not to GDP. Third, RLIP1 is unable to interact with a RalAA46 mutant. Based on sequence and structure similarities with c-Ha-Ras, a T46A RalA mutant would have an impaired effector domain. This result suggests that RalA requires an intact effector domain to bind to RLIP1. It also suggests that, in yeast, LexA-Ralwt is, at least in part, in the GTP-bound conformation, as is the case for other GTPases expressed as LexA fusions (9).
From these data, it emerges that RLIP1 is a good candidate to be an "effector" of RalA function.
In Vitro Binding-In order to confirm the genetic data, we have tested in vitro RLIP1 binding to GST-RalA and GST-RalAV23 proteins prepared from E. coli. Around 30% of both RalA proteins bind GDP and GTP (data not shown). RLIP1 cDNA was amplified by PCR. The PCR product was transcribed from a T3 promoter sequence included at the 5Ј end of the 5Ј PCR primer, and translated in vitro. A 35 S-labeled protein of apparent molecular mass of 66 kDa was produced ( Fig. 2A). Fig. 2B shows that RLIP1 did not bind to GST or to GST-RalA but does bind to GST-RalAV23.
When in vitro guanine nucleotide exchange was performed prior to RLIP1 binding, RLIP1 again did not bind to GST. It did not bind GST-RalA or GST-RalAV23 loaded with GDP. It bound GST-RalA and, even better, GST-RalAV23 loaded with GTP (Fig. 2C). These data show that RLIP1-RalA interaction is not mediated by a yeast protein and are consistent with the genetic results, i.e. RLIP1 interacts directly with the GTPbound form of RalA whose effector domain is required.

RLIP1 Binds Ral but No Other GTPase except Rac-
We addressed the question whether RLIP1 discriminates among GTPases. Table I summarizes the results obtained with the two-hybrid method, using different GTPases and pRLIP1. It also gives the positive controls used in each case.
In the Ras superfamily, within the Ras branch (to which RalA belongs; Ref. 18), RLIP1 was not able to interact with c-Ha-Ras or with Rap1A, Rap2A, or Rap2B. However, it does interact with RalB.
In the Rab branch, RLIP1 was not able to interact with Rab5, Rab6, Rab7, or Rab13.
Finally, in the Rho/Rac branch, we were not able to detect any interaction with RhoA, RhoB or RhoG, but RLIP1 is able to interact with Rac1 (Fig. 1B).
This latter interaction was further investigated using Rac1 alleles. Fig. 1B shows that RLIP1 is able to interact with Rac1V12S189 but not with Rac1V12N17S189, a dominant negative mutant blocked in the GDP-bound form, or Rac1V12A35S189 and Rac1V12A38S189, two effector domain mutants (19). These results suggest that Rac1 bound to GTP is able to interact with RLIP1 through its effector domain.
The GAP-like Region Displays a GAP Activity-Sequence analysis of RLIP1 (see below) has revealed a region highly homologous to GAP regions acting upon GTPases of the Rho family. The question whether the region of RLIP that looks like a GAP region is a GAP was addressed. RLIP1 was expressed fused to MBP from plasmid pMAL-c2. The fused protein was purified and assayed for stimulation of the GTPase activity of purified CDC42 and Rac1, and Rap2A as a control. Fig. 3 shows that there is no effect on Rap2A (specificity control), a significant and reproducible effect on Rac1 and a stronger effect upon CDC42. It also shows that this effect is weaker that the one obtained with Bcr-GAP protein (positive control) (20). RLIP1 exhibited no GAP activity upon RhoA (data not shown).
Molecular Biology of the cDNA-Northern blot analysis re-vealed that RLIP1 is expressed in all tested tissues as a 4-kilobase mRNA of low abundance (data not shown). Two consecutive rounds of 5Ј-RACE were required to obtain the full-length cDNA that was also recovered from a skeletal muscle cDNA library in gt10 and from a placenta cDNA library in EXlox. The sequence of the full-length cDNA was

TABLE I
Interaction of RLIP1 with different GTPases L40 was cotransformed with pairwise combinations of plasmids, one expressing the GTPases listed above as LexA-fusion proteins, the other expressing RLIP1 from plasmid pRLIP1 isolated during our two-hybrid screen for Ra1A interacting proteins. Column "Protein expression" refers to verification of expression of the fused protein by Western blotting. This verification was carried out during this work or by the referred person. ND, not determined. Column "Control by two-hybrid assay" means that the construct used to express the LexA-GTPase fusion protein has permitted to detect a specific partner using L40 and a two-hybrid assay. Usually this positive control was assessed by the referred person who was the kind donator of the plasmid. There is one main reading frame (ORF), from base 224 to base 2188, preceded by a correct translation initiation sequence (21). Two short ORFs (13 and 3 codons) are found within the 5Ј end of this cDNA but none of them is preceded by a correct translation initiation sequence. A 1664-base pair non-coding sequence is found 3Ј to the ORF.
This ORF encodes a protein made of 655 amino acids and of predicted molecular mass 76 kDa that we named RLIP76 (Fig.  4A). RLIP1 in plasmid pRLIP1 starts at amino acid 185, and RLIP2 starts at amino acid 403.
Data bank comparison revealed that the region extending FIG. 3. GTPase activating activity of RLIP1 upon Rac1 and CDC42. Purified Rac1, CDC42, and Rap2A proteins were loaded with [␥-32 P]GTP, and GTPase activity was assessed by a filter binding assay (30). 100% refers to the radioactivity bound to the protein at time 0. For each protein, the first column reflects the intrinsic GTPase activity, the second the GTPase activity in presence of RLIP1. For CDC42 and Rac1, the third column reflects the GTPase activity in presence of the GAP region of Bcr, which harbors a powerful GAP activity upon Rac1 and CDC42.

FIG. 4. Sequence and analysis of RLIP76.
A, amino acid sequence of RLIP76. B, homology between RLIP76 and GAP domains Both a Blitz homology search program (EMBL) and a BLAST homology search program (NCBI) aligned a region of RLIP76 with domains of proteins that display a CDC42/Rac/Rho GAP activity. The two best scores were obtained with Bcr-GAP region and n-chimaerin, and these alignments are shown here. Identical residues are bold, conserved residues are indicated by a ϩ on the consensus line. Blocks 1, 2, and 3 refer to three blocks conserved among proteins of this class (22). C, deletion analysis. RLIP76 coding region was inserted in plasmid pGAD1318, which allows expression of GAL4AD fusion proteins. From pRLIP1 isolated during the screen, different regions were deleted. For each construct, the junction region was sequenced. These plasmids were tested for their ability to elicit a positive signal in the twohybrid system in presence of a plasmid expressing a LexA-RalA protein or, when indicated, a LexA-Rac1 protein. D, functional regions. Based on the previous analysis (B and C), two regions can be defined functionally: a GAP region from residue 210 to 353, and a RalA binding region from residue 403 to 499. Compilation of results from three different programs predicting secondary structures (31)(32)(33) led to this scheme. The two regions represented by boxes are predicted by at least two of the programs to be composed mainly of ␣-helices. In addition, part of the second region is predicted to be superfolded as a coiled-coil structure (32). from amino acid 210 to amino acid 353 shares significant homology to regions of proteins bearing a CDC42/Rho/Rac-GAP activity, like Bcr, chimaerins, Drosophila rotund, and the Ras-GAP-binding protein p190 (22). Fig. 4B shows this striking homology with Bcr and n-chimaerin. Fig. 4C gives the results obtained with two-hybrid plasmids expressing different parts of RLIP76. These results allow definition of the maximum size of the region required for RalA binding. Together with secondary structure predictions (Fig.  4D), the overall structure of RLIP76 can be depicted schematically as composed of four regions: an N-terminal region where amino acids 65-170 are predicted to be structured in ␣-helices, the GAP-like region (aa 210 -353), the Ral binding region (aa 403-499) predicted to be composed in part of ␣-helices, and a C-terminal region (aa 499 -655). Part of this latter region and part of the Ral binding region (aa 440 -610) are predicted to be able to form a coiled-coil structure.
By FISH analysis of 19 R-banded metaphase cells, RLIP1 gene was localized on band 18p11 (25 chromosomes positive on both chromatid out of 38). A minor localization on band 3q26 was also detected (9 out of 38) that might suggest the existence of a related gene (data not shown). DISCUSSION We have identified a cDNA encoding a protein, RLIP1, that is able to interact with RalA and RalB, and which has the characteristics of a Ral effector; biochemical data and genetics suggest that RLIP1 binds better to Ral-GTP than to Ral-GDP and that this interaction requires a functional effector domain. According to Northern blot analysis, RLIP1 is ubiquitously (but at low levels) expressed, as are Ral and RalGDS, a Ral activator (3).
Although able to discriminate Ral from other GTPases, RLIP1 also binds to the active form of Rac1. We suppose that the molecular avatar of this binding is a GAP-like region whose absence impairs interaction with Rac1 but not with RalA; domains involved in Rac binding and in Ral binding are physically distinct. We also show that the structural homology with GAP-like regions reflects a functional homology. RLIP1 is able to activate specifically hydrolysis of GTP bound to Rac1 and to CDC42, but not, as expected, to Rap2A.
The whole cDNA was cloned; it encodes a 76-kDa protein, RLIP76, able to bind to RalA as based on a two-hybrid assay.
These findings raise several questions. Our results allow us to conclude that the GAP-like region of RLIP76 displays a bona fide GAP activity acting upon CDC42 and Rac1. However, this GAP activity is rather weak when compared to the GAP activity of Bcr tested in parallel. This could be due either to technical problems (we do not know how much of purified RLIP is active), to structural problems (either a larger part or a smaller part of RLIP76 could do better) or to biological constraints (a companion protein might increase this activity). These considerations lead to more questions. What is Ral doing to RLIP76? Is it localizing RLIP76 in the vicinity of its target, as happens to be the case for other GTPases involved in the subcellular localization of certain effectors (23,24), and/or is it modulating RLIP76-GAP activity?
Ras and Rho pathways are both activated during mitogenic signaling through transmembrane receptors. It is unclear if their activation is sequential or parallel, but they seem to work synergistically (19,(25)(26)(27)(28)(29). The Rho pathway, a cascade of GTPases, from CDC42 to Rho passing by Rac, acts upon structures involved in cell shape plasticity. Activation of Ras leads to several cytoplasmic and nuclear phenomena as well as membrane modifications. And Ral proteins are potentially switched to their active form through interaction of activated Ras with Ral activators. We propose that RLIP76 participates in the cross-talk between these GTPase cascades, modulating the state of activity of the CDC42/Rac/Rho pathway in response to Ras activation.
Finally, regions of RLIP76 seem a priori not to participate in the above functions. The ␣-helix-rich regions, especially the coiled-coil region, might be involved in interactions with other proteins. Alternatively, the coiled-coil region might participate in the homodimerization of RLIP76. After Ras and subsequent Ral activation, Ral binding to RLIP76 could separate the monomers and render the GAP catalytic region accessible to its target.
It will be of great interest in future RLIP studies to analyze the regulation and interplay of the various separate functional domains.