RhoGDI-3 is a new GDP dissociation inhibitor (GDI). Identification of a non-cytosolic GDI protein interacting with the small GTP-binding proteins RhoB and RhoG.

RhoB is a small GTP-binding protein highly homologous to the RhoA protein. While RhoA is known to regulate the assembly of focal adhesions and stress fibers in response to growth factors, the function of RhoB remains unknown. We have reported that the transient expression of the endogenous RhoB protein is regulated during the cell cycle, contrasting with the permanent RhoA protein expression (1). Using the yeast two-hybrid system to characterize proteins interacting with RhoB, we identified a new mouse Rho GDP dissociation inhibitor, referenced as RhoGDI-3. The NH2-terminal α helix of RhoGDI-3 is strongly amphipatic and differs thus from that found in previously described bovine, human, and yeast RhoGDI proteins and mouse and human D4/Ly-GDIs. Contrary to the cytosolic localization of all known GDI proteins, acting on Rab or Rho, RhoGDI-3 is associated to a Triton X-100-insoluble membranous or cytoskeletal subcellular fraction. In the two-hybrid system, RhoGDI-3 interacts specifically with GDP- and GTP-bound forms of post-translationally processed RhoB and RhoG proteins, both of which show a growth-regulated expression in mammalian cells. No interaction is found with RhoA, RhoC, or Rac1 proteins. We show that GDI-3 is able to inhibit GDP/GTP exchange of RhoB and to release GDP-bound but not GTP-bound RhoB from cell membranes.

RhoA, Rac, and Cdc42Hs control also signal transduction pathways that are essential for cell growth. Activated Cdc42Hs and Rac, but not RhoA are efficient activators of a cascade leading from extracellular stimuli to c-Jun NH 2 -terminal kinases (JNKs) and p38/Mpk2 (18,19). The serine/threonine kinase p65 PAK (for p21-activated serine/threonine kinase), activated by interaction with Rac and Cdc42Hs, could initiate this cascade (20). RhoA, probably using a different pathway, is required for lysophosphatic acid-, serum-, and AlF 4 Ϫ (aluminum fluoride ion)-induced transcriptional activation by the serum response factor (21,22).
We have focused on the RhoB protein, encoded by an immediate early gene (23) and more than 85% identical to RhoA. In contrast to RhoA, little is known about the physiological functions of RhoB. We have reported the transient accumulation of RhoB through the cell cycle of HeLa cells. First detected at the G 1 /S phase transition, the level of the RhoB protein is maximal during S phase and declines at the S/G 2 -M transition (1). The intracellular distribution of the epidermal growth factor-or cell cycle-induced unstable RhoB protein differs from that of the endogenous RhoA protein that is found mainly cytosolic and, to a lesser extent, associated with membrane fractions (1). The endogenous RhoB protein, exclusively associated with total membranous fractions, was found, by immunofluorescence studies, to be associated to vesicular-like structures extending forward from a heavily stained juxta-nuclear compartment (1). Furthermore, using an epitope tagging approach coupled to micro-injection of living cells, Adamson et al. (24) reported an association of the epitope-tagged RhoB protein with early endosomes and a pre-lysosomal compartment.
Taken together, these data suggest a role of RhoB in cellular events during the G 1 /S transition and/or during the S phase of the cell cycle. This could implicate regulation of vesicular traffic or the translocation of components essential for cytoskeleton modelling during the S phase. Interestingly, RhoB was recently found to be weakly growth-activating in NIH3T3 cells and its activity was found essential for transformation of Rat1 fibroblasts by oncogenic Ras (25).
As a first attempt to precisely define the role of RhoB in intracellular pathways, we searched for effectors and regulatory proteins interacting with RhoB using the yeast two-hybrid system.
Like Ras, Rho proteins have two interconvertible forms: a GTP-bound active form and a GDP-bound inactive form. Cycling between the two conformations is tightly regulated by at least three classes of proteins. The GTPase-activating proteins (GAPs) 1 accelerate the intrinsic GTP hydrolytic activity of the Rho proteins and thereby down-regulate their activity, while the guanine nucleotide exchange factors (GEFs) stimulate the replacement of GDP by GTP and up-regulate the Rho proteins (26). Several of these regulatory proteins seems to act as "linkers" between different Rho proteins or between Rho and other GTPbinding proteins of the Ras superfamily. In particular, the Ost protein may link Cdc42Hs, Rac, and RhoA in certain cell types as it seems to be a target of GTP-bound Rac and a GEF for RhoA and Cdc42Hs (27). Likewise, several GAPs link different intracellular pathways. The p190 RhoGAP binds to p120 ras -GAP in growth factor-stimulated cells, thereby mediating communication between Ras and Rho (28). A Ral effector, RLIP76, acts as a GAP for Cdc42Hs and Rac1, thereby linking an other protein of the Ras-like family to the Rho pathway (29). Finally, the non-conventional myosin Myr5, which binds to actin, is a GAP for RhoA and Cdc42Hs in vitro, thereby linking directly Rho proteins to actin (30).
The third class of regulatory proteins is the guanine nucleotide dissociation inhibitors (GDIs) inhibiting the GDP to GTP exchange reaction (31). Two GDIs acting on Rho proteins were known until now: the ubiquitously expressed RhoGDI-1 (31,32), and the more recently identified D4/Ly-GDI (33-35) expressed only in hematopoietic tissues. RhoGDI has been shown to have several interesting biochemical activities. Micro-injection of RhoGDI inhibits several downstream functions of Rho (36 -38). In addition, the RhoGDI protein is capable of extracting the GDP-bound forms of Rho proteins from biological membranes by binding to the isoprenylated carboxyl terminus of the Rho proteins suggesting that Rho-like proteins cycle between the plasma membrane and the cytosol (39,40). In resting cells, GDP-forms of Rac, RhoA, and Cdc42Hs are maintained as inactive cytosolic complexes with RhoGDI, but are released and translocated to membranes during the course of cell activation (41)(42)(43)(44). RhoGDI can also bind to the GTP forms of Cdc42Hs, Rac, and RhoA, albeit with lower affinity to the GTP than the GDP forms of Rac and RhoA (32,45). It has been shown that the release of Rho proteins from RhoGDI does not occur in the absence of GTP or membranes and could be regulated by membrane-associated GEFs or by various lipid mediators in activated cells (46).
Here we report the isolation and the characterization of a third member of the RhoGDI family interacting with RhoB and RhoG proteins of the Rho family in the yeast two-hybrid system.
Plasmids-For LexA fusion protein expression, we used the pVJL10 plasmid. The bait of our screen was the Gly-14 3 Val mutant of RhoB (equivalent to Ras-G12V). RhoB-G14V was generated by oligonucleotide-directed mutagenesis using a double-strand mutagenesis kit (Stratagene), appropriate primers and a Bluescript SK ϩ -RhoB cDNA construct as template. RhoB-wt and RhoB-G14V cDNAs were subcloned into pVJL10 in frame with LexA. RhoB-G14V⌬C with a COOH-terminal deletion (18 amino acids) encompassing the cysteines at positions 189, 192, and 193 was obtained by digesting the RhoB-G14V cDNA by PstI before cloning into the pVJL10 plasmid.
The mouse BALB/c brain cDNA library, used in the two-hybrid screen, was constructed in fusion with GAL4AD in pGAD1318 plasmids (51).
PCR was performed to generate EcoRI and BamHI cloning sites allowing subcloning of the RhoGDI-3 coding region, contained in the 7.5 pGAD1318 clone, into the pMal-c2 procaryotic expression vector (New England Biolabs). The same coding region was also subcloned, after PCR generating adequate cloning sites, into the pCB6 eucaryotic expression vector in frame with a tagging sequence encoding a peptide (YTDIEMNRLGK) from vesicular stomatitis virus (VSV) G-protein.
Human RhoGDI-1 cDNA was obtained by PCR using a human placental ASV1 phage () cDNA library and primers designed from the submitted sequence of human RhoGDI-1 (accession number D13989). EcoRI and BamHI restriction sites were included in the primers in order to clone the PCR product into pMal-c2 procaryotic expression vector.
All PCR constructs and mutants were sequenced using the Sanger dideoxy-termination method (Sequenase kit, United States Biochemical Corp.).
Two-hybrid Screen-The yeast reporter strain L40, which contains the reporter genes LacZ and HIS3 downstream of the binding sequences for LexA, was sequentially transformed with the pLexA-RhoBG14V plasmid and with the mouse brain cDNA library in pGAD1318 plasmids using the lithium acetate method. Double transformants were plated to synthetic medium lacking histidine, leucine, tryptophan, uracil, and lysine. The plates were incubated at 30°C for 3 days. His ϩ colonies were patched on selective plates and assayed for ␤-galactosidase activity by a filter assay. pGAD1318 library plasmids from colonies displaying a His ϩ /LacZ ϩ phenotype were recovered using Escherichia coli HB101 cells plated on leucine-free medium.
Each plasmid was subsequently used to retransform L40 yeast cells containing on the one hand the pLexA-RhoBG14V hybrid and on the other hand the pLexA-lamin C construct as a negative control. This step allowed us to correlate the His ϩ /LacZ ϩ phenotype with a single library plasmid. The cDNA inserts in a sample of the specific clones were classified according to their sizes by PCR, and the abundance of each class, among all the specific clones identified, was estimated by dot-blot hybridization using 32 P-labeled cDNA inserts (Ready To Go kit, Pharmacia). The cDNA inserts from each class of specific clones were sequenced on both strands using the Sanger dideoxy termination method.
Sequenced pGAD1318 library plasmids were then tested for interaction with various Ras-like proteins in fusion with LexA by a mating technique (52). For this purpose, the MAT␣ yeast strain AMR 70 was transformed with pGAD1318 library plasmids, whereas the yeast strain MATa L40 was transformed with one of the pLexA constructs.
Northern Blot Analysis-Mouse tissue RNA blots were purchased from Clontech. Hybridization were performed with three randomprimed 32 P labeled probes representing the complete RhoGDI-3 cDNA purified from clone 7.5, the 5Ј deleted 8.7 cDNA spanning bases 247-880 (635 bp), and a 5Ј 197-bp fragment from clone 7.5, deleted in clone 8.7. Sequential probing of the same blots were performed as previously described (1). The quality and the amounts of RNA in each lane were assessed by sequential hybridization of filters with glyceraldehyde-3phosphate dehydrogenase and human ␤-actin probes. Washing was done at high stringency (0.1 ϫ SSC, 0.1% SDS) at 65°C.
Transfection and Fractionation of HeLa Cells-HeLa cells were transfected with 10 g of pCB6-VSV-RhoB or -RhoGDI-3 plasmid constructs/60-mm dish using a modified calcium phosphate co-precipitation method as described previously (1). After transfection, cells were grown for 48 h, after which 10 7 cells were harvested and lysed, and post-nuclear supernatants were centrifuged at 100,000 ϫ g (Beckman TL100 ultracentrifuge) for 15 min to generate cytosolic fractions and crude membrane pellets (1). Crude membrane fractions were then washed in membrane buffer (50 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 50 mM NaCl, 1 mM DTT), centrifuged at 100,000 ϫ g and treated on ice for 30 min in membrane buffer containing 1% Triton X-100 or 0.6% CHAPS and a mixture of protease inhibitors before a final centrifugation at 100,000 ϫ g to obtain soluble and insoluble subcellular fractions.
Western Blot Analysis-Protein concentrations were determined by the method of Bradford using the Bio-Rad protein assay (Bio-Rad, Mü nchen, Germany). The protein samples were resolved in 12.5% or 10% SDS-PAGE, electrotransferred to nitrocellulose sheets, stained with Ponceau-S to verify equal loading in each lane, and then submitted to immunoblot analysis.
Polyclonal anti-RhoGDI antibodies Sc K21 (Santa Cruz Biotechnology), raised against a peptide corresponding to amino acids 178 -198, a highly conserved region in RhoGDI proteins, recognize both the human RhoGDI-1 and the human hematopoietic D4/Ly-GDI. The polyclonal anti-RhoGDI antibody, Sc A20 (Santa Cruz Biotechnology), raised against a divergent NH 2 -terminal sequence in RhoGDI-1, recognizes only this RhoGDI protein. They both were used at 1/500 dilution.
LexA fusion protein expressions in yeast cells were analyzed using a polyclonal antibody directed against the LexA protein at 1/1000 dilution (a kind gift from P. Moreau, Gif-sur-Yvette, France). Bound antibodies were detected by peroxydase-labeled anti-rabbit and anti-mouse antibodies and the ECL chemiluminescence system (Amersham, Alesbury, UK) as described previously (1).
Release of the RhoB Protein from HeLa Cell Membranes-Crude membrane pellets from pCB6-VSV-RhoB-transfected HeLa cells were taken up in 50 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1 mM MgCl 2 , and 1 mM DTT. 50 l of membranes, containing 400 g of proteins, were pre-exchanged with GDP␤S or GTP␥S (Sigma) by adding 7 l of 0.1 M nucleotide, 1 l of 0.2 M DTT, 7 l of 30 mM pH 7.5 EDTA, and 4 l of a mixture of protease inhibitors and incubating at 30°C for 10 min. The reaction was stopped by adding 16 l of 0.2 M MgCl 2 . 17 l of exchanged membranes (136 g) were then incubated with 10 l of MBP-RhoGDI-3 (2 mg/ml), MBP-RhoGDI-1 (2 mg/ml), or BSA (2 mg/ml) at 30°C for 40 min. 13 l of 10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM MgCl 2 was then added and the reactions centrifuged at 100,000 rpm at 4°C for 15 min (54). Pellet (membrane fraction) and supernatant (soluble fraction) were then analyzed for RhoB content by SDS-PAGE and Western blotting with affinity-purified anti-RhoB antibodies.
Nucleotide Dissociation Inhibition Assays-Washed crude membrane pellets from pCB6-VSV-RhoB-transfected HeLa cells were treated on ice for 30 min in membrane buffer (50 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 50 mM NaCl, 1 mM DTT) containing 1% Triton X-100 and a mixture of protease inhibitors before centrifugation at 100,000 ϫ g. For each independent experiment using either [ 3 H]GDP or [ 3 H]GTP, 300 g of solubilized proteins from the supernatant were diluted 40-fold in immunoprecipitation buffer (50 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 150 mM NaCl, 1 mM DTT, 0.1% Triton X-100) containing protease inhibitors (final volume: 1.2 ml). The RhoB proteins were then immunoprecipitated with 9 l of P 5 D 4 ascitic fluid (mouse monoclonal anti-VSV glycoprotein antibody; Sigma) and 240 l of protein A/G Plus-Agarose (Santa Cruz Biotechnology) at 4°C for 12 h. The beads were collected by centrifugation and washed three times with 3 ml of ice-cold immunoprecipitation buffer and twice with 3 ml of ice-cold buffer containing low free magnesium concentration (50 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1 mM DTT, 1 mM MgCl 2 , 2.5 mM EDTA) supplemented with 0.1% Triton X-100. At this stage, two aliquots (40 l of beads) were withdrawn and consecutively analyzed for RhoB content by SDS-PAGE and Western blotting with affinity purified anti-RhoB antibodies. The remaining beads were diluted 8-fold in ice-cold buffer containing low free magnesium concentration, lacking Triton X-100, and containing a mixture of protease inhibitors. 5 Ci of [ 3 H]GDP or [ 3 H]GTP (Amersham) were added and the beads incubated at 30°C for 10 min with frequent mixing. They were then placed on ice and split into four equal aliquots, and 20 l of MBP-RhoGDI-3 (1 mg/ml) were added to half the tubes and 20 l of BSA (1 mg/ml) to the other half. The beads were gently resuspended, and after incubation for 10 min on ice one of the MBP-RhoGDI-3-treated samples and one sample treated with BSA were diluted in 1 ml of ice-cold buffer containing high free magnesium concentration (50 mM Tris-HCl, pH 7.5, 50 mM NaCl, and 25 mM MgCl 2 ) in order to estimate the maximal labeled nucleotide bound. The dissociation of the bound labeled nucleotide from the immunoprecipitated RhoB in the remaining tubes was started by adding immediately an excess of unlabeled GTP (2 mM) and incubating at 30°C for 5 min. The reaction was stopped by adding rapidly 1 ml of ice-cold buffer containing high free magnesium concentration. The amount of labeled nucleotide ([ 3 H]GDP or [ 3 H]GTP) bound to RhoB was quantified by vacuum filtration of the beads through nitrocellulose filters (BA85, Schleicher & Schuell) followed by three washes with 3 ml of ice-cold high magnesium buffer and liquid scintillation counting.
Chromosome Localization-The murine RhoGDI-3 cDNA (882 bp) was labeled with [ 3 H]dCTP (Amersham) by random priming and hybridized to human normal chromosomes as described previously (55). The same cDNA fragment labeled with [ 32 P]dCTP by random priming was hybridized to a Southern blot of EcoRI-digested DNA from a BIOS panel of somatic cell hybrids.

Identification of RhoB-interacting Proteins by Two-hybrid
Screening in Yeast-To identify proteins interacting with the GTP-bound form of RhoB, reporter L40 yeast cells were sequentially transformed with plasmids encoding the LexA DNA binding domain in fusion with RhoB-G14V and pGAD1318 plasmids containing mouse brain library cDNAs inserted 3Ј to the transcriptional activation domain of GAL4 (GAL4AD). The interaction between RhoB-G14V fused to the LexA DNA binding domain and a protein fused to GAL4AD allows the transcription of two reporter genes conferring histidine auxotrophy and ␤-galactosidase activity. The mouse brain cDNA library was used since we previously found that RhoB was highly expressed in mouse brain (personal data).
Approximately 2 ϫ 10 6 yeast transformants were screened. 186 colonies were found to grow on histidine-free plates; among them, 102 displayed ␤-galactosidase activity. Library plasmids were rescued from these 102 His ϩ /LacZ ϩ colonies. The L40 strain containing pLexA-RhoBV14 was then retransformed with 24 selected library plasmids in order to correlate the His ϩ /LacZ ϩ phenotype with a unique library plasmid. Simultaneously, the L40 strain containing pLexA-lamin C was transformed with the same 24 library plasmids and plated on histidine-free plates. 11 of these 24 clones interacted with lamin C and were therefore considered as nonspecifically interacting clones.
PCR reactions were performed to determine the size of the inserts contained in the 13 plasmids yielding a specific His ϩ / LacZ ϩ phenotype; 6 plasmids contained inserts of about 880 bp (class A), 5 plasmids contained inserts of about 650 bp (class B), and 2 contained 2-kb inserts (class C) (data not shown). To reduce the number of redundant remaining clones to be sequenced, the DNA from the 102 originally positive clones were hybridized with probes corresponding to each of the three species of inserts. The cDNA inserts of class A and B showed the same pattern of hybridization suggesting that they were parts of the same cDNA. The hybridization signals obtained with the 2-kb insert did not overlap with the former signals, indicating that this library insert represented an independent cDNA (data not shown).
To summarize, 91 clones gave a positive signal with class A and B inserts whereas 11 inserts were positive with the 2-kb class C probe.
The present paper will focus on the most represented RhoB partner in this mouse brain library.
Identification of a New RhoGDI Protein-Three clones were examined in more detail. The nucleotide sequences of clone 7.5 (882 bp) and 4.5 (880 bp) revealed a 678-bp open reading frame. The neighboring sequence of the ATG codon, in frame with LexA, was in good agreement with the Kozac consensus sequence. The nucleotide sequence of one of the partial class B cDNAs (clone 8.7; 635 bp) showed that this cDNA lacked 204 nucleotide coding for the first 68 NH 2 -terminal residues but contained identical 3Ј coding and non-coding sequences. The screening of a gt10 mouse embryo cDNA library allowed us to identify a stretch of 37 additional base pairs 5Ј to the already known sequence of clone 7.5 with an in-frame terminator codon 39 bp upstream the ATG codon. Therefore, it seemed highly probable that the coding sequence contained in clone 7.5 was complete. Fig. 1 shows the nucleotide and deduced amino acid sequence of the entire 921-bp cDNA. The open reading frame codes for 225 amino acids with a calculated molecular mass of 30 kDa.
A search in the GenBank™ data base revealed that this cDNA bore high homology to the human and bovine RhoGDI cDNAs (64% identity in 620 overlapping 3Ј coding sequences) and to the mouse and human D4/Ly-GDI cDNAs (62% identity in 700 overlapping 3Ј coding sequences). This cDNA could then encode for a new RhoGDI protein, which we named RhoGDI-3.
At the protein level, RhoGDI-3 showed 52% identity in the 178 COOH-terminal amino acids with the other RhoGDI proteins. Comparison of mouse RhoGDI-3 with mouse D4/Ly-GDI, human Ly-GDI, human RhoGDI, and bovine RhoGDI protein sequences is shown in Fig. 2.
The 47 NH 2 -terminal residues in RhoGDI-3 differed drastically from RhoGDI-1 and D4/Ly-GDI protein sequences, and RhoGDI-3 contained 21 additional amino acids in the NH 2terminal region compared to the previously described RhoGDI-1 proteins. The alignment of the more homologous COOH-terminal parts of the proteins, proposed by the CLUSTAL V program, introduced an artificial gap in RhoGDI-1 and D4/Ly-GDI protein sequences in an effort to align the NH 2 -terminal parts of these proteins with respect to RhoGDI-3.
It should be noted that the consensus N-glycosylation site (Asn-X-Ser/Thr), found twice in human D4/Ly-GDI (34) and once in mouse D4/Ly-GDI as well as in human and bovine RhoGDI-1, was not found in mouse RhoGDI-3.
We used four programs for secondary structure predictions of the five RhoGDI proteins. As shown in Fig. 2, the NH 2 -terminal ␣-loop-␣ structure of the other known RhoGDIs was replaced in RhoGDI-3 by a more complex helix structure containing at least three helix segments. Remarkably, the long NH 2 -terminal ␣ helix in RhoGDI-3 was highly amphipatic and thus strongly differed from the NH 2 -terminal helix in the other RhoGDI proteins analyzed in this paper.
Characterization of the Interaction between RhoGDI-3 and Rho Proteins-The interaction specificity of RhoGDI-3 with various Ras-like proteins was examined by yeast mating experiments. For this purpose, several LexA-fusion proteins, altered or not in their GTP hydrolysis capacity or in their processing, were used. LexA-lamin C fusion protein was used as negative control, and the interaction between LexA-Ras and GAL4AD-Raf fusion proteins was used as positive control (data not shown). All LexA fusion proteins were expressed at approximately the same level in transformed yeast cells, as determined by Western blot analysis with anti-LexA antibodies (data not shown). Furthermore, LexA-RhoC and -Rac1 hybrids were shown to interact in yeast with other specific partners.
As shown in Fig. 3, the RhoGDI-3 protein did not interact with proteins from the Ras branch (Ras, Rap2, and RalA) or with the Rab6-G22V, known to interact with RabGDI proteins (51). In the Rho/Rac branch, RhoGDI-3 interacted only with RhoB, from the Rho subfamily, and with RhoG, from the Rac subfamily. These interactions were also found with the activated GTP-bound conformations of RhoB and RhoG (RhoB-G14V and RhoG-G12V). This was in good agreement with previous observations that RhoGDIs are able to interact with both GDP-and GTP-bound forms of Rho proteins (45), contrasting with RabGDIs only able to interact with GDP-bound Rab proteins (56). Finally, the COOH-terminal deletion mutations RhoB-G14V⌬C and RhoG-G12V⌬C, which abolish post-translational processing, also completely abolished interaction with RhoGDI-3 (Fig. 3). This is consistent with the fact that RhoGDI-3, as all known GDIs, can only interact with isoprenylated forms of Rho proteins (39). Remarkably, RhoGDI-3 failed to interact with RhoA-G14V and RhoC-wt (88% identical to RhoB) or with Rac1-G12V (73% identical to RhoG) (Fig. 3).
As shown in Fig. 4, the incomplete RhoGDI-3 protein, lacking the first 68 NH 2 -terminal residues (clone 8.7), could only interact with the GTP-bound RhoB-G14V protein. These large region, corresponding to a ␣-loop-␣-loop secondary structure (Fig.  2), could be important for the specificity of the interaction between GDI-3 and Rho proteins possibly via the maintenance of an appropriate tertiary structure. A previous report (57) demonstrated that a truncation of the 25 NH 2 -terminal amino acids in RhoGDI-1 and D4/Ly-GDI did not impair the activity of these two proteins. The secondary structure predictions shown in Fig. 2 suggested that these 25 residues correspond to the first ␣ helix of the molecules.
RhoGDI-3 mRNA Shows a Tissue-specific Expression-To determine whether RhoGDI-3 mRNA was expressed in particular cell types, we probed a Northern membrane containing poly(A) ϩ RNA isolated from various mouse organs. Fig. 5 shows that the complete RhoGDI-3 cDNA (919-bp) and the 5Ј-deleted 8.7 cDNA (635 bp) revealed a 1.2-kb transcript only detected in brain, lung, kidney, and testis. The human ␤-actin probe showed that these variations did not reflect the variations in the amounts of loaded RNA. Moreover, the glyceraldehyde-3phosphate dehydrogenase mRNA, whose expression is known to be high in skeletal and cardiac muscles, had a different pattern of expression. The same transcript was detected by a 5Ј cDNA probe corresponding to the 5Ј untranslated region and the nucleotide sequences coding for 21 NH 2 -terminal amino acids, showing that these sequences was included in the same polyadenylated mRNA. Strikingly, RhoB mRNA was also detected mainly in brain, lung, and testis (data not shown).
In conclusion, the size (1.2 kb) and the expression pattern of the RhoGDI-3 mRNA in mouse tissues differed from that of the more ubiquitously expressed RhoGDI-1 mRNA (2.4 kb) and the Ly-GDI mRNA (1.4 kb) only expressed in hematopoietic tissues (35).

RhoGDI-3 Releases the GDP-bound Form of RhoB from
Membranes-RhoGDI-3 was expressed as a fusion protein with the maltose-binding protein in E. coli strain BL21 and purified on a maltose resin column. The MBP fusion protein was recognized both by Sc K21 polyclonal anti-RhoGDI and monoclonal anti-MBP antibodies as a protein migrating with an apparent molecular mass of 68 kDa (data not shown). The polyclonal Sc K21 anti-RhoGDI antibodies, directed toward a conserved COOH-terminal peptide, were previously reported to detect both RhoGDI-1 and D4/Ly-GDI proteins. The polyclonal Sc A20 antibodies, raised against a divergent NH 2 -terminal RhoGDI-1 sequence and previously shown not to detect D4/Ly-GDI, were unable to detect the recombinant RhoGDI-3 protein (data not shown).
The RhoGDI-1 protein is known to release GDP forms of Rac, RhoA, RhoB, and Cdc42Hs from biological membranes. To determine whether RhoGDI-3 is active, we tested the ability of purified recombinant MBP-RhoGDI-3 to extract RhoB from HeLa cell membranes.
Hela cells were transfected with the pCB6 expression vector containing VSV-tagged RhoB. The membrane-bound RhoB protein was pre-exchanged with GDP␤S or GTP␥S and incubated with MBP-RhoGDI-3, MBP-RhoGDI-1, or an equivalent amount of BSA.
As shown in Fig. 6, MBP-RhoGDI-3 and MBP-RhoGDI-1 were able to extract the RhoB protein from GDP-treated but not from GTP-treated membranes. This behavior was in agreement with that previously reported for RhoGDI-1, known to The secondary structure predicted for the COOH-terminal parts of the proteins, shown in the lower panel, is the same for all RhoGDI. Broken lines represent strong contradiction between the four predicting methods used. Spaces between secondary elements are related to the intrinsic difficulty to predict the exact limit of these elements. release only the GDP-bound RhoB protein from membranes (40), and confirmed the ability of RhoGDI-3 to function as a GDI protein for RhoB.
Assays for RhoGDI-3 Activity to Regulate Nucleotide Exchange Reactions of RhoB-We tested the ability of purified recombinant MBP-RhoGDI-3 to inhibit GDP/GTP exchange on RhoB. Previous results demonstrated that the RhoGDI-3 protein, as all known GDIs, can only interact with isoprenylated forms of Rho proteins. To obtain large amounts of processed proteins, wild-type RhoB protein with a NH 2 -terminal VSV-G epitope was transiently expressed in HeLa cells. As RhoB is exclusively associated to membranous compartments (1,24), the membrane-bound RhoB protein was extracted from crude membrane fractions of transfected HeLa cells by Triton X-100 (1%) and the RhoB proteins in the solubilized fraction were partially purified by immunoprecipitation with an anti-VSV monoclonal antibody and captured on protein A/G-agarose. Western blotting of the immunoprecipitated proteins demonstrated a single band, which was recognized by the polyclonal anti-RhoB antibodies (data not shown). After extensive washing, the immunoprecipitated RhoB proteins were bound to either [ 3 H]GDP or [ 3 H]GTP and split into incubations with or without recombinant RhoGDI-3 in the presence of an excess of unlabeled GTP. As shown in Fig. 7, recombinant MBP-RhoGDI-3 was found to inhibit the exchange of prebound [ 3 H]GDP with exogenous GTP at low free Mg 2ϩ concentrations, at which Rho proteins have a high intrinsic GDP dissociation rate. In the presence of RhoGDI-3, the amount of [ 3 H]-GDP still bound to RhoB after 5 min at 30°C was Ͼ80%, whereas in the presence of an equal amount of BSA it was Ͻ20%. MBP-RhoGDI-3 showed less ability to inhibit the dissociation of bound [ 3 H]GTP. RhoGDI-3 permitted only a reduction of about 20% in the extent of [ 3 H]GTP dissociation. It has been shown that bovine brain RhoGDI-1 produced only a weak inhibition of prebound GTP dissociation from CDC42Hs (32). However, and contrasting with other proteins of the Rho family (32), the intrinsic dissociation rate of [ 3 H]GTP bound to RhoB seemed not to be slower, in the absence of RhoGDI-3, than that found for preloaded [ 3 H]GDP (Fig. 7). total membranes and cytosol followed by immunoblotting with polyclonal Sc K21 anti-RhoGDI antibodies revealed a protein migrating with an apparent molecular mass of 34 kDa corresponding well to that expected for VSV-RhoGDI-3 (Fig. 8). The same protein was also revealed by the monoclonal anti-VSV P4D5 antibody (data not shown).

RhoGDI-3 Is Not Present in Cytosolic Subcellular
Contrasting with RhoGDI-1, known to be cytosolic, the transiently expressed RhoGDI-3 was only detected in membrane fractions, where it represented the only detected protein by this anti-RhoGDI antibody. Moreover, no endogenous RhoGDI-3 protein could be detected in non-transfected HeLa cells.
In addition, the polyclonal Sc K21 anti-RhoGDI antibody recognized a cytosolic protein, migrating with an apparent molecular mass of 28 kDa, corresponding well to that found for endogenous RhoGDI-1 and D4/Ly-GDI proteins (31,35). This protein, also found in non-transfected cells, probably represented the RhoGDI-1 protein, as D4/Ly-GDI was previously found not to be expressed in HeLa cells (35).
Finally, RhoGDI-3 was not extracted from the membrane fraction by detergents such as Triton X-100 (1%) or CHAPS (0.6%), suggesting its association to an integral membrane protein or to a cytoskeletal subcellular fraction (Fig. 8).
Chromosome Localization-In the 250 mitoses analyzed by in situ hybridization with the mouse RhoGDI-3 cDNA probe (882 bp), 54 silver grains (16, 7%) were detected on the human chromosome 16 and 28 out of them (51, 8%) were located on band 16p13 (Fig. 9). Although a rather high background was found, which may be due to interspecific hybridization, no other significant grain accumulation could be detected on any other chromosome. These results were confirmed by hybridization of a Southern blot from a Bios panel of somatic cell hybrids with the RhoGDI-3 cDNA probe (data not shown). DISCUSSION We have identified a cDNA encoding a new mouse GDI-like protein, RhoGDI-3, interacting specifically with the GDP and GTP-bound forms of RhoB and RhoG in the yeast two-hybrid system. We have also shown that this GDI is functional. The RhoGDI-3 protein, like RhoGDI-1 (40), stimulates the release of the GDP-bound but not the GTP-bound RhoB protein from HeLa cell membranes. The RhoGDI-3 protein also inhibits the GDP/GTP exchange of RhoB but shows less ability to inhibit the dissociation of prebound GTP. This could be due to a lower affinity of RhoGDI-3 for the GTP-bound RhoB protein as already found for the interactions of RhoGDI-1 and GTP forms of Rac and RhoA (58). Interestingly, RhoGDI-1 has recently been found to interact with GDP-bound and GTP-bound forms of Cdc42Hs with identical affinities (59).
The RhoG protein is more closely related to the Rac, Cdc42Hs, and TC10 members of the Rho family. Remarkably, while RhoB is encoded by an immediate-early gene, RhoG is encoded by a growth-regulated gene belonging to the late primary response genes transcribed in the late G 1 phase of the cell cycle (8).
Contrasting with the ubiquitously expressed RhoGDI-1, RhoGDI-3 shows a tissue specific expression, being only detected in mouse brain, lung, and testis. Strikingly, RhoB is also mainly expressed in these organs, whereas the more ubiquitously expressed RhoG reaches a particular high level in lung tissues (8).
The most remarkable feature of RhoGDI-3 is its association with membrane fractions, as opposed to all other RhoGDI and RabGDI proteins, which are cytosolic.
The NH 2 -terminal ␣ helix in RhoGDI-3, which differs drastically from that of known RhoGDI proteins, could be respon-  sible for this association. Unlike the NH 2 -terminal ␣ helix in the other RhoGDI, this helix structure is strongly amphipatic and stabilized in its NH 2 -terminal part by a characteristic capping structure LDXXEL (60).
It is unlikely that this ␣ helix, on the whole negatively charged, interacts directly with membranous phospholipids. This helix structure may rather be implicated in protein-protein interactions and permit RhoGDI-3 association to an integral membrane protein or to a cytoskeleton subcellular fraction, as suggested by the association of RhoGDI-3 with a Triton X-100-insoluble fraction.
This observation suggests a different role of RhoGDI-3 compared to that of cytosolic RhoGDI proteins. Unlike RhoA, Rac and CDC42Hs proteins, which cycle between the plasma membrane and the cytosol, the epidermal growth factor and cell cycle-induced RhoB protein (1) as well as the transiently expressed epitope-tagged RhoB protein (24) have only been found associated to membranous compartments. This in turn implies that the release of the GDP-bound form of RhoB from HeLa cell membranes by the recombinant RhoGDI-3 protein may not reflect the in vivo function of RhoGDI-3. In resting cells, while RhoB is not detected, RhoA, Rac, and CDC42 are largely cytosolic and their GDP-bound inactive forms are maintained in a stoichiometric complex with RhoGDI-1 (46,42). Translocations of these Rho proteins to the plasma membrane, induced by extracellular stimuli, is associated with their activation and their release from RhoGDI-1 (46).
One could speculate that RhoGDI-3 could link the vesicularassociated RhoB to an integral membrane protein or to a cytoskeletal compartment and thereby regulate the RhoB protein activity during the S phase of the cell cycle. We are currently attempting to identify the cellular compartment and/or protein to which RhoGDI-3 is associated.
In the yeast two-hybrid system, RhoGDI-3 interacts with the same intensity with the GDP-and GTP-bound conformations of RhoB and RhoG proteins. Complexes containing RhoGDI-1 and GTP-bound forms of Rho proteins have not yet been isolated in vivo. RhoGDI-1 has only been reported to form stable complexes with RhoA, Rac, and CdC42Hs proteins in the GTP states in vitro (45,58,59). Interestingly, the Rac-GTP-RhoGDI-1 and Cdc42Hs-GTP-RhoGDI-1 complexes were found resistant to the action of recombinant RhoGAP preventing GAP stimulated GTP hydrolysis (61,62). It is also interesting to note that Rho-GDP complexed with RhoGDI-1 is known to be unable to interact with target and GEF proteins (46,63).
Another remarkable feature of RhoGDI-3 is the fact that it can make a distinction between RhoB on the one hand and RhoA and RhoC on the other hand, as well as between RhoG and Rac proteins in opposition to the RhoGDI-1 protein known to interact with all Rho proteins of the Rho/Rac family. Little is known about the region(s) of the RhoGDI molecules that are involved in their interactions with Rho proteins. Recently published work (57) provides a first insight into the structural basis for the binding and function of the RhoGDI-1 and D4/Ly-GDI proteins. The authors reported that the deletion of the eight COOH-terminal amino acids of these two GDI molecules eliminated their ability to release the Cdc42Hs protein from membrane bilayers. Moreover, a limited region of nine amino acids from the COOH-terminal domain (positions 172-180) seemed to be important for the higher efficacity of RhoGDI-1, relative to D4/Ly-GDI, in triggering the release of Cdc42Hs from membranes. Within this region there are four differences between RhoGDI-1, D4/Ly-GDI, and RhoGDI-3. The RhoGDI-3 protein has an additional conservative change at position 199 (Arg instead of Lys) corresponding to position 178 in the other RhoGDI molecules. Additional work is needed to elucidate the precise domain(s) implicated in distinct functional specificities of the three RhoGDI proteins.
The identification of a third RhoGDI protein, interacting with the RhoB and RhoG proteins, both of which show a growth-regulated expression in mammalian cells, indicates that it could belong to a new family of RhoGDI proteins with restricted specificity, playing a role in the activity of these Rho proteins during the cell cycle.