The two guanine nucleotide exchange factor domains of Trio ... - Nature

Oncogene (1998) 16, 147 ± 152  1998 Stockton Press All rights reserved 0950 ± 9232/98 $12.00

The two guanine nucleotide exchange factor domains of Trio link the Rac1 and the RhoA pathways in vivo Jean-Michel Bellanger, Jean-Bernard Lazaro, Sylvie Diriong, Anne Fernandez, Ned Lamb and Anne Debant CRBM-CNRS, ERS155, 1919 Route de Mende, BP5051, 34033 Montpellier CeÂdex 01, France

Trio contains two functional guanine nucleotide exchange factors (GEF) domains for the Rho-like GTPases and a serine/threonine kinase domain. In vitro, GEF domain 1(GEFD1) is speci®cally active on Rac1, while GEF domain 2 (GEFD2) targets RhoA. To determine whether Trio could activate Rac1 and RhoA in vivo, we measured the eect of Trio on Mitogen Activated Protein Kinase (MAPK) pathways and cytoskeletal rearrangments events mediated by the two GTPases. We show that: (i) the GEFD1 domain of Trio triggers the MAPK pathway leading to Jun kinase (JNK) activation and the production of membrane rues; (ii) co-expression of the TrioGEFD1 domain with a dominant-negative form of Rac blocked JNK induction, whereas a dominant-negative form of Cdc42 did not; (iii) a deletion mutant of TrioGEFD1 lacking a region important for exchange activity could not stimulate JNK activity; (iv) in contrast, the TrioGEFD2 domain does not stimulate JNK activity and induces the formation of stress ®bers, as does activated RhoA; (v) furthermore, co-expression of both GEF domains induces simultaneously the formation of rues and stress ®bers. Trio, therefore represents a unique member of the Rho-GEFs family possessing two functional domains of distinct speci®cities, that allow it to link Rho and Rac signaling pathway in vivo. Keywords: Trio; GDP/GTP exchange factor; actin; MAP kinases

Introduction The small GTPases Cdc42, Rac1 and RhoA have important regulatory roles in mediating various cytoskeletal rearrangments in many cell types (Hall, 1994) Guanine nucleotide exchange factors (GEFs) accelerate the exchange rate of GDP for GTP on these proteins and thereby activate them (Boguski and McCormick, 1993). In Swiss 3T3 cells, Cdc42 has been shown to stimulate the formation of ®lopodia, whereas Rac1 is required for lamellipodia formation and membrane ruing, and RhoA for the formation of actin stress ®bers (Nobes and Hall, 1995). In other cell types, these Rho-like GTPases seem to trigger tissuespeci®c responses. For example, in lymphoid cells, Rho is required for integrin-mediated aggregation (Laudanna et al., 1996; Tominaga et al., 1993). In neutrophils, Rac mediates the activation response, which includes an increase in pinocytosis and Correspondence: A Debant Received 1 May 1997; revised 1 September 1997; accepted 1 September 1997

stimulation of NADPH oxydase (Diekmann et al., 1994). More recently, Rac was also proposed to be involved in neuronal dierentiation (Luo et al., 1996). Finally, Cdc42 was proposed to participate in generating cell polarity towards antigen-presenting cells (Stowers et al., 1995). In addition, the GTPases of the Rho family have been implicated in cell cycle progression, since microinjection of activated Cdc42, Rac1 and RhoA induces G1 progression in quiescent Swiss 3T3 cells ®broblasts leading to S phase entry (Olson et al., 1995). Moreover, there is compelling evidence that Rac1 and RhoA are involved in Rasinduced malignant transformation (Qui et al., 1995). Consistent with these observations, the Rho family members have been linked to gene regulation: (i) Cdc42 and Rac1 but not RhoA, stimulate a Mitogen Activated Protein Kinase (MAPK) cascade, leading to the activation of JNK (for c-Jun N-terminal kinase also known as stress-activated protein kinases or SAPKs) (Coso et al., 1995; Minden et al., 1995); (ii) the three GTPases RhoA, Rac1 and Cdc42 potentiate the transcriptional activity of the serum response factor (SRF) (Hill et al., 1995), via a signaling pathway which remains to be elucidated. The ®rst GEF identi®ed for the Rho GTPase family was the dbl oncogene, which was shown to act on Cdc42Hs and RhoA (Hart et al., 1994). Subsequently, more than 15 proteins involved in cell signaling, including proteins with oncogenic capacity, were identi®ed as potential GEFs by sequence homology with the central portion of the dbl oncogene, the DH domain (dbl homology domain) (Cerione and Zheng, 1996; Olson, 1996). In addition, all the DH-containing proteins share a PH domain, which is also present in a variety of signaling molecules (Musacchio et al., 1993). However, sequence homology with the DH domain of Dbl does not necessarily confer GEF activity in vitro and in vivo. Only the DH-containing proteins Dbl, Vav, FGD1 were shown to stimulate JNK activity, consistent with the activation of Rac1 or Cdc42 (Minden et al., 1995; Olson et al., 1996; Zheng et al., 1996). Furthermore, some of the DH-family members display restricted speci®city, such as Lbc for RhoA, Tiam for Rac1 and Cdc24 and more recently FGD1 for Cdc42 (Michiels et al., 1995; Olson et al., 1996; Zheng et al., 1995, 1996). Trio is a multifunctional protein that was isolated by its capability to interact with the transmembrane tyrosine phosphatase LAR (Debant et al., 1996). Trio contains two functional GEF domains for the Rho-like GTPases and a serine/threonine kinase domain. We have shown that the ®rst GEF domain (GEFD1) has speci®c exchange activity in vitro towards Rac1, while the second GEF domain (GEFD2) has speci®c

Trio and GTPases signaling pathways in vivo J-M Bellanger et al

Table 1 Schematic diagram of the deletion mutants of Trio

The structure of the deletion mutants used in this study are shown here. Structural domains of Trio are abbreviated as follow: DH, Dbl homology; PH, Pleckstrin homology, Ig, immunoglobulin-like domain; pSK, protein serine kinase domain. Numbers refer to amino-acid residues. The deletion mutants were cloned as HA-fusion proteins in the pMT vector, as described in Materials and methods

TrioPSK

TrioGEFD2

TrioGEFD1

TrioGEFs

Control

a

HA-JNK

Western blot

RacV12

b

TrioGEFD1

Kinase assay

Cdc42V12

GST c-jun

Control

Western blot

c

RasV12

HA-JNK

RacV12

Kinase assay

TrioPSK

GST c-jun

TrioGEFD2

TrioGEFD1 and TrioGEFD2 activate in vitro the small GTPases Rac1 and RhoA respectively (Debant et al., 1996). To determine whether Rac1 is a target of the GEFD1 domain of Trio in vivo, we examined the eect of dierent domains of Trio on induction of JNK activity. Hamster CCL39 ®broblasts were co-transfected with expression vectors coding for deletion mutants of Trio and for a hemagglutinin (HA)-tagged JNK (Table 1). Twenty-four hours after transfection, HA-JNK was immunoprecipitated from cell lysates and tested for its activity on GST c-Jun (1 ± 79) in an in vitro kinase assay. Expression of the GEFD1 domain (TrioGEFD1, aa 1118 ± 1919), alone or in association with the GEFD2 domain (TrioGEFs, aa 1118 ± 2451), caused a marked stimulation of JNK activity (Figure 1a). In contrast, this was not seen with the GEFD2 alone (TrioGEFD2, aa 1849 ± 2451) or the serine kinase Cterm (TrioPSK, aa 2452 ± 2861) domains. In addition, the increase in JNK activity arising from the expression of TrioGEFD1 was in the same range as that caused by the expression of activated Rac1 (RacV12) or Cdc42 (Cdc42V12) (Figure 1b). Therefore, TrioGEFD1, but not TrioGEFD2, can eectively induce the activation of JNK, implying that Rac1 is a target in vivo of the TrioGEFD1 domain. Consistently, TrioGEFD1, but not TrioGEFD2, induces p38 MAPK activity (data not shown), which is another downstream target of Rac (Minden et al., 1995). The observation that the family of Rho GTPases failed to activate MAPK pathway leading to extracellular

TrioGEFD1

Results

signal-regulated kinases (ERKs) activation prompted us to examine whether the GEF domains of Trio cannot induce the ERK pathway. Dierent deletion mutants of Trio were co-expressed in CCL39 cells together with a (HA)-p44MAPK. Constitutively activated GTPases Rac1 and Ras (RacV12 and RasV12) were used in the experiment respectively as a negative and positive controls. After immunoprecipitation, HA-p44MAPK activation was tested by monitoring phosphorylation of myelin basic protein (MBP). RacV12 and all Trio deletion mutants tested failed to induce p44MAPK activity, whereas RasV12 strongly increased phosphorylation of MBP (Figure 1c). This observation indicates that TrioGEFD1 acts as activated Rac1. In order to determine whether the stimulation of JNK activity caused by the TrioGEFD1 domain is due to Rac activation exclusively, we co-

TrioGEFs

exchange activity towards RhoA. This observation suggests that Trio may act as a two-headed exchange factor by mediating the activation of both Rac and RhoA in vivo. To address this issue, we investigated whether the GEF domains of Trio were active in vivo by measuring the JNK activation and cytoskeletal rearrangements mediated by the small GTPases Rac1 and RhoA.

Control

148

MBP

Kinase assay

HA-MAPK

Western blot

Figure 1 TrioGEFD1, but not TrioGEFD2, speci®cally stimulates JNK activity. (a) CCL39 cells were transfected with plasmid pSRa3HA.JNK1 alone (lane 1) or together with plasmid pMT.HA.TrioGEFs (lane 2), pMT.HA.TrioGEFD1 (lane 3), pMT.HA.TrioGEFD2 (lane 4), pMT.HA.TrioPSK (lane 5). Forty hours after transfection, HA-JNK was immunoprecipitated from cell lysates and its activity assessed by phosphorylation of GST c-Jun (1 ± 79). The amount of HA-JNK protein in each sample was determined by Western blotting using the anti-HA antibody 12CA5. (b) CCL39 cells were transfected with plasmid pSRa3HA.JNK1 alone (lane 1) or together with plasmid pMT91.Myc.Rac1V12 (lane 2), pMT90.Myc.Cdc42V12 (lane 3), pMT.HA.TrioGEFD1 (lane 4). Cells were treated as described in a. (c) CCL39 cells were transfected with plasmid pECE.HA.p44.MAPK alone (lane 1) or together with plasmid pMT.HATrioGEFs (lane 2), pMT.HA.TrioGEFD1 (lane 3), pMT.HA.TrioGEFD2 (lane 4), pMT.HA.TrioPSK (lane 5), pMT91.Myc.Rac1V12 (lane 6), RasV12 (lane 7). Forty hours after transfection, HA-p44MAPK was immunoprecipitated from cell lysates and its activity was assessed using phosphorylation of MBP. The amount of HA-p44MAPK protein in each sample was determined by Western blotting using the anti-HAtag antibody 12CA5


(Ridley et al., 1992). Moreover, the expression of GEFD1 clearly induces the reduction of stress ®bers as was recently proposed for activated Rac1 and Cdc42, possibly via activation of PAK (Manser et al., 1997). On the contrary, Rhodamin-phalloidin staining of cells expressing TrioGEFD2 (Figure 4d) revealed the presence of stress ®bers, characteristic of activated RhoA (Ridley and Hall, 1992) but no membrane ruing. These date indicate that TrioGEFD1 and TrioGEFD2, respectively, mimic the eect of the GTPases Rac1 and RhoA on actin cytoskeleton rearrangments. In order to determine whether Trio may simultaneously activate both Rac and RhoA GTPases by its two GEFs domains, quiescent Swiss 3T3 were injected with a plasmid coding for both enzymatic domains (Trio GEFs, aa 1118 ± 2451) and stained for F-actin. Figure 4f clearly shows that expression of both Trio GEF domains caused an increase in both rues and stress ®bers formation. These results indicate that the GEFs domains of Trio concomitantly induce both Rac1 and RhoA pathways in vivo. Discussion Our observations provide evidence for a role for Trio in the activation of both the GTPases Rac1 and RhoA in vivo. We have shown here that: (i) the GEFD1 domain speci®cally activates Rac1 in vivo, by measuring two Rac1 targets in the signaling pathway induced by the GTPase, namely the stimulation of JNK activity and the formation of membrane rues; (ii) a dominant negative form of Rac1 (RacN17), but not of Cdc42 (Cdc42N17), completely inhibits the JNK pathway initiated by TrioGEFD1, consistent with the idea that JNK induction by TrioGEFD1 arises from Rac1 activation; (iii) a deletion mutant of TrioGEFD1 lacking the region important for the exchange activity (TrioDGEFD1) fails to stimulate either JNK activity or membrane ruing (data not shown), indicating that the catalytic activity of the GEFD1 domain is absolutely required for the induction of the JNK pathway; (iv) on the other hand, TrioGEFD2 does not

Kinase assay

HA-JNK

Western blot

Myc-GTPases

Western blot

GST c-jun HA-GEFD1/∆GEFD1

Figure 2 TrioGEFD1 induces JNK activity via Rac exclusively. CCL39 cells were transfected with plasmid pSRa3HA.JNK1 alone (lane 1) or together with plasmid pMT.HA.TrioGEFD1 (lanes 2 ± 8), in combination with increasing concentrations of plasmid pMT91.Myc.RacN17 (0.1 mg, lane 3; 0.2 mg, lane 4; 0.5 mg, lane 5) or with increasing concentrations of plasmid pMT90.Myc Cdc42N17 (0.1 mg, lane 6; 0.2 mg, lane 7; 0.5 mg, lane 8). Cells were then treated as described in Figure 1a. The amount of HAJNK and Myc-N17GTPases proteins in each sample were determined by Western blot analysis using the anti-HA and anti-Myc antibodies 12CA5 and 9E10, respectively

HA-JNK

Trio∆GEFD1

GST c-jun

TrioGEFD1

TrioGEFD1 RacN17 Cdc42N17 Control

Control

expressed dominant-negative forms of Cdc42 and Rac1 (Cdc42N17 and RacN17) with TrioGEFD1. In analogy to the activity of RasN17, these inhibitory mutants are thought to act by trapping their respective GEFs, and thus inhibiting the signal transduction pathway. We examined the eect of the mutated GTPases on JNK activation by TrioGEFD1. As shown in Figure 2, the increased expression of Rac1N17, and not Cdc42N17, progressively blocked the JNK activation elicited by TrioGEFD1. A dominant-negative form of RhoG (RhoGN17), another member of the Rho-GTPases family, did not block JNK activation elicited by TrioGEFD1 (data not shown). Therefore, we conclude that the pathway by which TrioGEFD1 induces the JNK activation includes Rac1, but not Cdc42 or RhoG. To con®rm that the enzymatic activity of the Trio GEFD1 domain was necessary for its induction on JNK activity, we designed a TrioGEFD1 deletion mutant (called TrioDGEFD1, which lacks the sequence LLLKELL, Table 1). It was shown for Dbl that the most highly conserved sequence among the various members of the Dbl family (LLLKELL in the case of Dbl) was absolutely required for the stimulation of GDP dissociation and transformation activity of Dbl (Hart et al., 1994). As shown in Figure 3, TrioDGEFD1 lacking this sequence failed to stimulate the JNK activity even though the mutant protein is well expressed. These data prove that the enzymatic activity of the exchange factor GEFD1 is absolutely required for its stimulatory eect on JNK activity. We then investigated whether both GEF domains of Trio could also activate Rac1- and RhoA-mediated cytoskeletal rearrangments, membrane ruing and formation of stress ®bers. For that purpose, immuno¯uorescence microscopy was performed on Swiss 3T3 cells microinjected with plasmids coding for TrioGEFD1 or TrioGEFD2. Cells expressing TrioGEFD1 produce intensive membrane ruing and lamellipodia, and the expressed protein is specially localized in these structures (Figure 4a). F-actin staining by Rhodaminphalloidin con®rm that the injected cells produce formation of membrane ruing and lamellipodia (Figure 4b), as already described for activated Rac1

Kinase assay Western blot Western blot

Figure 3 The exchange factor activity of TrioGEFD1 is required to stimulate JNK activity. CCL39 cells were transfected with plasmid pSRa3HA.JNK1 alone (lane 1) or together with plasmid pMT.HA.TrioGEFD1 (lane 2), pMT.HA.DTrioGEFD1 (lane 3). Forty hours after transfection, HA-JNK was immunoprecipitated from cell lysates and its activity was assessed by the phosphorylation of GST c-Jun (1 ± 79). The amount of HAJNK, HA-TrioGEFD1 and HA-TrioDGEFD1 proteins in each sample were determined by Western blotting using the anti-HA antibody 12CA5

149


150

stimulate the JNK activity but instead induces the formation of stress ®bers, indicating that RhoA is an in vivo target of TrioGEFD2; (v) none of the Trio domains induce the MAPK pathway leading to ERK

ANTI-HA

activation, nor do the Rho-family of GTPases (Coso et al., 1995; Minden et al., 1995). These in vivo data are entirely consistent with what we have determined in vitro; (vi) moreover, co-expression of the two GEF

PHALLOIDIN

Figure 4 The GEFs domains of Trio induce simultaneously both Rac and Rho-mediated actin cytoskeleton remodelling. Swiss 3T3 cells were microinjected with the plasmids pMT.HA.TrioGEFD1 (a,b), or pMT.HA.TrioGEFD2 (c,d), or pMT.HA.TrioGEFs (e,f), and 3 h later, immuno¯uorescence was performed as described in Materials and methods. The expression of the dierent Trio plasmids were revealed with the 12CA5 anti-HA antibody (a,c,e). The actin structures of the injected cells were visualized by Rhodamin-phalloidin staining (b,d,f). Bar, 20 mm


domains induces simultaneously the formation of rues and stress ®bers, showing clearly that the Trio GEF domains link both Rac1 and RhoA pathways in vivo. To our knowledge, this is the ®rst example of a member of the DH family that contains two functional DH domains, each activating a dierent GTPase in vivo. Few GEFs of the Dbl family have as well a restricted speci®city for the Rho GTPases. Lbc was shown to act as a GEF for RhoA in vitro, and produces RhoA-speci®c stress ®bers when expressed in Swiss 3T3 cells (Zheng et al., 1995). Tiam1, recently isolated in a screening for identifying invasionpromoting genes, acts as a Rac1, Cdc42 and Rho GEF in vitro, but Tiam induces a similar phenotype as V12Rac1 in vivo (Michiels et al., 1995). T-lymphoma cells expressing V12Rac become invasive, suggesting a role for Tiam and Rac in invasion and metastasis. More recently, it was shown that tyrosine-phosphorylated Vav, but not the non-phosphorylated protein, catalyzes GDP/GTP exchange on Rac in vivo (Crespo et al., 1997). This restricted speci®city does not agree with data obtained in Swiss 3T3 cells indicating that Vav could stimulate independently Rac1 and RhoA (Olson et al., 1996). The Saccharomyces cerevisiae protein Cdc24 has been shown both biochemically and genetically to have exchange activity on Cdc42 (Zheng et al., 1994). More recently, the DH-containing protein FGD1, responsible for the developmental disease faciogenital dysplasia (Aarskog ± Scott syndrome) was shown to function as a Cdc42Hs-speci®c GEF in vitro and to produce Cdc42-speci®c ®lopodia when expressed in Swiss 3T3 (Olson et al., 1996; Zheng et al., 1996). The expression of the GEFD1 domain of Trio clearly induces not only the formation of membrane ruing but also a decrease in stress ®bers formation. These data are consistent with the observation that constitutively activated aPAK caused loss of stress ®bers and dissolution of focal adhesions in Hela and Swiss 3T3 cells, as does activated Rac1 and Cdc42 (Manser et al., 1997). This recent proposal of stress ®bers loss after Rac1 or Cdc42 activation argue more for an apparent antagonism between Cdc42-Rac1 and RhoA pathways than the linear cascade initially proposed from Cdc42 via Rac1 to RhoA (Nobes and Hall, 1995). In this context, the simultaneous activation of both Rac1 and RhoA GTPases by the two GEFs domains of Trio strongly support a role for Trio as a two-headed exchange factor, capable of linking the pathways induced by both GTPases in vivo. Further studies will be required to determine the activation status of the downstream targets of Rac1 and RhoA after expression of both GEF domains of Trio. For example, activation of Rho-associated kinase a (ROKa), which acts downstream of RhoA to promote stress ®bers formation (Leung et al., 1996) and activation of PAK by Rac1, should have opposite eects on the formation of these structures. The next challenge will be to identify the molecular cross-talk between these two signaling cascades leading to the simultaneous formation of stress ®bers and rues induced by the GEF domains of Trio. Another good example supporting the crosstalk between the Rac1 and RhoA pathway, is the DH-family member, Ost. This protein has exchange factor activity towards Cdc42 and Rho in vitro, and binds Rac on its GTP

form, suggesting that Ost is an eector of Rac (Horii et al., 1994). Since Ost is only expressed in the brain, the crosstalk between Rac and Rho elicited by this exchange factor seems to be restricted to this tissue. The activation of Trio could be of importance in the crosstalk between Rac and RhoA GTPases during morphological organization involving actin cytoskeleton remodeling occuring in physiological processes such as cell motility or proliferation of ®broblasts. In other cell types, the crosstalk between the two GTPases pathways is less clear. Since there are some indications that distinct cell types may display dierent responses to a given Rho family member, the concomitant activation of Rac and Rho by Trio might amplify the tissue-speci®c response of these GTPases.

Materials and methods DNA constructs The TrioGEFD1 (aa 1118 ± 1919), the TrioGEFD2 (aa 849 ± 2451), the TrioGEFs (aa 1118 ± 2451), the TriopSK (aa 2452 ± 2861) were constructed by inserting appropriate cDNA fragments into the pMT.HAtag expression vector (Serra-Pages et al., 1995). The TrioDGEFD1 deletion mutant was obtained as follows. The nucleotides 4191 ± 4211 (aminoacids 1375 ± 1381) were deleted and replaced by a XbaI restriction site by PCR using two sets of oligonucleotides: (i) TCG AGT AAA AGT CTC CAG; TGC TCT AGA CTG ATA TTT CGT TCG; (ii) TGC TCT AGA ACG TGC TGT GAG GAA GGA; AGT GGT CCG CAC CAG ACA. The two PCR products were digested with EcoRV and XbaI for the ®rst product, and XbaI and NotI for the second product. Both digested DNAs were subcloned in Bluescript SK+ (Stratagene), and the resulting deleted fragment was cloned in the EcoRV and NotI sites of the pMT.HA.TrioGEFD1 vector. The construct was veri®ed by DNA sequencing. The pSRa3HA.JNK1 and the pECE.HA.p44MAPK plasmids were gifts from B Derijard and J Pouyssegur, respectively. The pMT91.MycRac1 (V12 and N17) and pMT90.Myc Cdc42 (V12 and N17) plasmids were gifts from P Chavrier. Cell transfections and microinjections CCL39 and Swiss 3T3 cells were maintained in DMEM (Biowhittaker) supplemented with 10% fetal calf serum (DAP), penicillin and streptomycin (Imperial). CCL39 cells in 35 mm dishes were transfected with 2 mg of total DNA with the lipofectamine reagent (Life Technologies). After 24 h of transfection, cells were serum-starved for 15 h and harvested in the Triton Lysis Buer (TLB). Swiss 3T3 cells were plated on glass coverslips. Five days later, cells were starved for 24 h and microinjected as described elsewhere (Lamb and Fernandez, 1997). Three hours after microinjection of the indicated plasmids, cells were ®xed and stained as described. Immuno¯uorescence Immuno¯uorescence was performed as follows: 3 h after microinjection, cells were ®xed 10 min in 3% formaldehyde and permeabilized for 30 s in 7208C acetone. The monoclonal antibodies anti-HA(12CA5), or anti-Myc (9E10), were incubated 60 min in PBS 0.5% BSA (dilution 1/500). The presence of the ®rst antibody was revealed by an FITC-conjugated goat-anti mouse antibody (Cappel, dilution 1/20). The actin structures were revealed by incubation with Rhodamin-phalloidin for 30 min.

151


152

Fluorescent images were directly photographed using a Kodak DCS420 professional digital color camera and acquired under Adobe Photoshop as TIFF format. The ®gures were then assembled completely under SGI Showcase 3.21 and printed directly as postscript ®les using a Kodak Colorease thermal sublimation printer. Immunoprecipitation and Western blotting Cells were lysed in (TLB): 20 mM Tris (pH 7.5), 137 mM NaCl, 2 mM EDTA, 1% Triton X-100, 25 mM bglycerophosphate, 1 mM sodium orthovanadate, 2 mM sodium pyrophosphate, 10% glycerol, 1 mM phenyl methyl sulfonyl ¯uoride (PMSF), 2 mg/ml leupeptin, 2 mg/ ml aprotinin. After centrifugation at 12 000 g for 20 min at 48C, HA-JNK or HA-p44MAPK were immunoprecipitated with the 12CA5 antibody. The immunoprecipitates were washed three times with TLB, and boiled in Laemmli buer. The proteins were resolved by 7.5%, 10% or 15% SDS ± PAGE and blotted on Immobilon-P membrane (Millipore). The expression of the HA-tagged and Myctagged proteins were detected using the 12CA5 and the 9E10 antibodies respectively. Immunocomplexes were visualized using enhanced chemiluminescence detection (Amersham).

Kinase assay The MAPK and JNK activities were determined as previously described (Bagrodia et al., 1995; Coso et al., 1995; DeÂrijard et al., 1994). The immunoprecipitates were washed three times with TLB, and twice with the kinase buer: 25 mM HEPES (pH 7.5), 25 mM MgCl2, 25 mM bglycerophosphate, 2 mM DTT, 0.1 mM sodium orthovanadate. Immunocomplex kinase assays were performed using 2 mg of substrate (Gst-c-Jun 1 ± 79 for JNK and MBP for MAPK assays) 50 mM ATP, 5 mCi of [g32P]ATP (3000 Ci/ mmole) for 20 min at 258C. The reactions were terminated by addition of Laemmli buer and analysed by SDS ± PAGE. Autoradiography was performed with the aid of an intensifying screen. Acknowledgements The authors are indebted to P Fort for fruitful discussions, R Hipskind for carefully reading the manuscript. This work was supported by Fondation pour la Recherche MeÂdicale, Association pour la Recherche contre le Cancer, FeÂdeÂration des Centres de Lutte contre le Cancer, Ligue Nationale Contre Le Cancer and by CNRS (grant `Biologie cellulaire'). J-B Lazaro was a recipient of Association FrancËaise coutre la Myopathie postdoctoral fellowship.

References Bagrodia S, DeÂrijard B, Davis RJ and Cerione RA. (1995). J. Biol. Chem., 270, 27995 ± 27998. Boguski MS and McCormick F. (1993). Nature, 366, 643 ± 653. Cerione RA and Zheng Y. (1996). Curr. Opin. Cell Biol., 6, 216 ± 222. Coso, OA, Chiariello M, Yu JC, Teramoto H, Crespo P, Xu NG, Miki T and Gutkind JS. (1995). Cell, 81, 1137 ± 1146. Crespo P, Schuebel KE, Ostrom AA, Gutkind JS and Bustelo XR. (1997). Nature, 385, 169 ± 172. Debant A, Serra-Pages C, Seipel K, O'Brien S, Tang M, Park SH and Streuli M. (1996). Proc. Natl. Acad. Sci. USA, 93, 5466 ± 5471. DeÂrijard B, Hibi M, Wu IH, Barrett T, Su B, Deng T, Karin M, Davis RJ. (1994). Cell., 76, 1025 ± 1037. Diekmann D, Abo A, Johnston C, Segal AW and Hall A. (1994). Science, 265, 531 ± 533. Hall A. (1994). Annu. Rev. Cell. Biol., 10, 31 ± 54. Hart MJ, Eva A, Zangrilli D, Aaronson SA, Evans T, Cerione RA and Zheng Y. (1994). J. Biol. Chem., 269, 62 ± 65. Hill CS, Wynne J and Treisman R. (1995). Cell, 81, 1159 ± 1170. Horii Y, Beeler JF, Sakaguchi K, Tachibana M and Miki T. (1994). EMBO J., 13, 4776 ± 4786. Lamb NJC and Fernandez A. (1997). Methods in Enzymol., 283, 72 ± 83. Laudanna C, Campbell JJ and Butcher EC. (1996). Science, 271, 981 ± 983. Leung T, Chen XQ, Manser E and Lim L. (1996). Mol. Cell. Biol., 16, 5313 ± 5327. Luo L, Hensch TK, Ackerman L, Barbel S, Jan LY and Jan YN. (1996). Nature, 379, 837 ± 840.

Manser E, Huang HY, Loo TH, Chen XQ, Dong JM, Leung T and Lim L. (1997). Mol. Cell. Biol., 17, 1129 ± 1143. Michiels F, Habets GG, Stam JC, van der Kammen RA and Collard JG. (1995). Nature, 375, 338 ± 340. Minden A, Lin A, Claret FX, Abo A and Karin M. (1995). Cell, 81, 1147 ± 1157. Musacchio A, Gibson T, Rice P, Thompson J and Saraste M. (1993). Trends Biochem. Sci., 18, 343 ± 348. Nobes CD and Hall A. (1995). Cell, 81, 53 ± 62. Olson MF. (1996). J. Mol. Med., 74, 563 ± 571. Olson MF, Ashworth A and Hall A. (1995). Science, 269, 1270 ± 1272. Olson MF, German Pasteris N, Gorski JL and Hall A. (1996). Curr. Biol., 6, 1628 ± 1633. Qiu RG, Chen J, Kirn D, McCormick F and Symons M. (1995). Nature, 374, 457 ± 459. Ridley AJ and Hall A. (1992). Cell, 70, 389 ± 399. Ridley AJ, Paterson HF, Johnston CL, Diekmann D and Hall A. (1992). Cell, 70, 401 ± 410. Stowers L, Yelon D, Berg LJ and Chant J. (1995). Proc. Natl. Acad. Sci., 92, 5027 ± 5031. Tominaga T, Sugie K, Hirata M, Morii N, Fukata J, Uchida A, Imura H and Narumiya S. (1993). J. Cell. Biol., 120, 1523 ± 1537. Zheng Y, Cerione RA and Bender A. (1994). J. Biol. Chem., 269, 2369 ± 2372. Zheng Y, Fischer DJ, Santos MF, Tigyi G, Pasteris NG, Gorski JL and Xu Y. (1996). J. Biol. Chem., 271, 33269 ± 33172. Zheng Y, Olson MF, Hall A, Cerione RA and Toksoz D. (1995). J. Biol. Chem., 270, 9031 ± 9034.