Elsevier

Cellular Signalling

Volume 11, Issue 8, August 1999, Pages 545-554
Cellular Signalling

Topical review
Rho Guanine Dissociation Inhibitors: Pivotal Molecules in Cellular Signalling

https://doi.org/10.1016/S0898-6568(98)00063-1Get rights and content

Abstract

The small G proteins of the Ras family act as bimodal relays in the transfer of intracellular signals. This is a dynamic phenomenon involving a cascade of protein–protein interactions modulated by chemical modifications, structural rearrangements and intracellular relocalisations. Most of the small G proteins could be operationally defined as proteins having two conformational states, each of which interacts with different cellular partners. These two states are determined by the nature of the bound nucleotide, GDP or GTP. This capacity to cycle between a GDP-bound conformation and a GTP-bound conformation enables them to filter, to amplify or to temporise the upstream signals that they receive. Thus the control of this cycle is crucial. Membrane anchoring of the proteins in the Ras family is a prerequisite for their activity. Most of the proteins in the Rho/Rac and Rab subfamilies of Ras proteins cycle between cytosol and membranes. Then the control of membrane association/dissociation is an other important regulation level. This review will describe one family of crucial regulators acting on proteins in the Rho/Rac family—the Rho guanine nucleotide dissociation inhibitors, or RhoGDIs. As yet, only three RhoGDIs have been described: RhoGDI-1, RhoGDI-2 (or D4/Ly-GDI) and RhoGDI-3. RhoGDI 1 and 2 are cytosolic and participate in the regulation of both the GDP/GTP cycle and the membrane association/dissociation cycle of Rho/Rac proteins. The non-cytosolic RhoGDI-3 seems to act in a slightly different way.

Introduction

The Ras superfamily of GTP-binding proteins (also called small GTPases) function as molecular switches in regulating a multitude of biological processes, induced by extracellular signals, that include cell proliferation, apoptosis, differentiation, cytoskeletal reorganisations, membrane trafficking and nuclear–cytosolic transport. They can be subdivided into the Ras, Rho/Rac, Rab, Ran, Rad and Arf subfamilies according to sequence similarities, and these subdivisions are well correlated to their specific biological functions.

Ras-related proteins bind guanine nucleotides very tightly (binding constants ≅ 1011 M−1), and almost all of them cycle between an inactive GDP-bound conformation and an active GTP-bound conformation. In resting cells, most Ras-related proteins are found in the GDP-bound state in spite of the fact that GTP is more abundant in the cell. The cycling between the two conformations, which enables these proteins to act as binary switches, is tightly regulated by several proteins, as illustrated in Fig. 1. Guanine nucleotide exchange factors (GEFs) stimulate the replacement of GDP by GTP. GEFs exhibit their highest affinities towards the guanine-nucleotide-free state of the Ras-related proteins and are thought to promote GDP release by stabilising this intermediate transition state. The GTP-bound form of the proteins interacts with downstream targets or effectors. This interaction is terminated by hydrolysis of protein-bound GTP into GDP. The intrinsic GTPase activity of Ras-related proteins is very low. The GTPase activating proteins (GAPs) accelerate this intrinsic GTP hydrolytic activity and thereby downregulate the activity of the Ras-related proteins.

With the exception of the Ran proteins, all Ras-related proteins studied so far are post-translationally modified. Arf proteins are myristoylated at the N-terminus and Ras, Rho/Rac and Rab proteins contain in C-terminal cysteine-containing motif, which serve as attachment for one or two farnesyl or geranylgeranyl groups. The post-translational modification is necessary for membrane anchoring and activity of the concerned Ras-related proteins. Rab and Rho/Rac proteins are regulated by a third class of proteins, the guanine nucleotide dissociation inhibitors (GDIs), which inhibit the dissociation of the nucleotide bound to these proteins. GDIs have also been found to extract the membrane-associated, post-translationally modified GDP-bound proteins from the membrane by binding to the prenylated C-terminus, keeping the hydrophobic tail protected from aqueous solvent. Thus these GDI-complexed proteins serve as a cytosolic pool of inactive Rab and Rho/Rac proteins (see the schematic GDP/GTP cycle in Fig. 1). The dissociation of GDI from the protein is a prerequisite for membrane association and activation of these Ras-related proteins by GEFs.

This review will focus on the complex regulation of the proteins in the Rho/Rac family and treat just one of the crucial steps, which includes their interaction with the RhoGDIs.

Three GDIs have so far been reported to act on proteins in the Rho/Rac family:

  • RhoGDI-1 (also called RhoGDIα) was first purified from rabbit intestine and bovine brain cytosols 1, 2 as a protein inhibiting the dissociation of GDP from and the subsequent binding of GTP to RhoB, one of the Rho proteins. The corresponding bovine and human cDNAs were subsequently isolated 3, 4. The human RhoGDI-1 locus was assigned to 17q25.3-25-3. The ubiquitously expressed RhoGDI-1 protein contains 204 amino acids with a calculated Mr of 23,421,000. A RhoGDI protein was also characterised in yeast 5, 6 and was shown to be 36% identical with bovine RhoGDI-1.

  • RhoGDI-2 (also called D4/Ly-GDI or RhoGDIβ) was first found as a human protein by Leffers et al. [4]. The same year, this protein was characterised as a RhoGDI specifically expressed in human and murin hematopoietic tissues 7, 8, 9. GDI-2 is composed of 200 or 201 amino acids with a calculated Mr of 229,000.

  • RhoGDI-3 (also called RhoGDIγ) was identified by us as a protein preferentially expressed in murin brain tissues [10]. RhoGDI-3 contains 225 amino acids with a calculated Mr of 25,296,000. Subsequently, a human homologue of RhoGDI-3 was identified [11]. The human RhoGDI-3 locus was found on band 16p13 10, 11.

GDI-1 and GDI-2 show 73.6% identity in the 178 COOH-terminal residues, whereas the same region in GDI-3 shows 62.9% identity with GDI-1 and GDI-2. The most divergent sequences are found in the NH2-terminus, where GDI-3 has a putative amphipatic α-helix extension stabilised by a characteristic capping structure, LDXXEL [10] (Fig. 2).

The three RhoGDI species contain several patterns for serine/threonine kinases (red residues in Fig. 2) as the sequence TDDD, a strong pattern for casein kinase II found in all three GDIs. GDI-1 and GDI-3 contain, in addition, a strong pattern for protein kinase Ce (PKCe): KISFR and KITFK/KISFK, respectively. Finally, the GDI-1 and GDI-2 proteins show a strong pattern for the cGMP-dependent kinase: KKQS and KKDT/KKET, respectively. Phosphorylation of GDI-1 was shown to stabilise the RhoA-GDI-1 complex in the neutrophil cytosol [12]. In addition, the haematopoietic cell-specific GDI-2 was found to be phosphorylated on a threonine upon Jurkat T cell activation by phorbol esters [7]. More recently, stimulation of a human myelomonocytic cell line (U937) by phorbol esters was found to induce phosphorylation of GDI-2 in contrast with GDI-1, which remains unphosphorylated [13], underlining the differential regulation of the two GDIs in this haematopoietic system. Surprisingly, no PKC site can be clearly detected in GDI-2.

NMR spectroscopy studies on full-length RhoGDI-1 or on the protein deleted from the first 22 amino-terminal residues (GDIΔ22) showed that the 64 amino-terminal residues of the molecule were disordered in solution and that the remaining 140 residues constitute an ordered domain 14, 15. X-ray crystallography and NMR spectroscopy studies of the almost exclusively folded GDIΔ59 domain 14, 15 showed an immunoglobulin-like fold with nine β strands in two antiparallel sheets, forming a β sandwich, and one 2-turn 310 helix (Fig. 3, left-hand side). This structure contained a narrow hydrophobic pocket between the β sheets (Fig. 3, right-hand side) that was found to interact with an isoprenylated peptide 14, 15, suggesting that it may bind the carboxy-terminal isoprenyl group of the Rho-like proteins. Moreover, Ile 177 of GDI-1 (green arrow in Fig. 2), found to be important for the affinity towards the Rho-like protein Cdc42 [16], lies on the surface of the protein in the pocket (yellow in the space-filled representation in Fig. 3). Deletion of the most divergent 25 amino acids from the amino terminus of RhoGDI-1 (black arrow in Fig. 2) was shown to have no significant effect on the ability of GDI-1 to inhibit GDP dissociation from Cdc42 or on its ability to release Cdc42 from membrane bilayers [16]. However, deletion of as many as 42 residues from the amino terminus of RhoGDI eliminated both activities (red arrow in Fig. 2) [15]. Selective proteolysis of RhoGDI-1 with trypsin gave a fragment starting at residue 59 of GDI-1. GDIΔ59 was still found to bind GDP-Cdc42 in a high nanomolar range [15] but was unable to inhibit the dissociation of the bound nucleotide. Thus the ability of the GDI-1 to inhibit GDP dissociation from Cdc42 probably resides in residues 25–42 [15]. Addition of non-isoprenylated Rac1 [14] or Cdc42 [15] proteins to either full-length GDI-1 or the fully active GDIΔ22 led to specific changes in the chemical shifts of several residues. Residues 23–63 in the flexible amino-terminus, thought to contain the inhibitory activity, are the most affected, but several regions between residues 119 and 186 in the folded domain also are influenced. Thus combined biochemical and structural data suggested a model of RhoGDI-1 function in which the carboxy-terminal binding domain (GDIΔ59) targets the flexible amino-terminal inhibitory region to Rho proteins, resulting in membrane extraction and inhibition of nucleotide cycling.

There is no similarity between the structure of RhoGDI-1 and the structure of RabGDI [17], which performs the same function for the proteins in the Rab family. Hence, in spite of the very similar structure of Rho and Rab proteins, the structure of one of their regulators is unrelated.

It has been reported that, after the induction of apoptosis in T cells, the unstructured amino-terminal region of the GDI-1 homologue GDI-2 is specifically cleaved at two positions [residues 16–20 and 52–56 (green in Fig. 2)] by two different interleukin-1–β-converting enzyme (ICE)-like proteases 18, 19, 20. The structural study of GDI-1 suggests that a cleavage in the region 52–56 of GDI-2 will render this protein unable to regulate Rho-like proteins. It is interesting to note that the position of this region corresponds to that of the site cleaved by trypsin in GDI-1. Thus proteolysis touching this disordered amino-terminal region of GDI proteins could be a general mechanism for regulating their activity.

Proteins in the Rho/Rac family contain an insertion of 13 residues absent in the other Ras-like proteins. This inserted domain consists of two mobile and exposed α-helices in Rac1 and RhoA, as depicted in X-ray crystallography studies 21, 22. This region in Cdc42 was reported to be essential for the ability of GDI-1 to inhibit GDP dissociation from Cdc42 but not for the affinity of the two proteins [23]. One possible explanation could be that the amino-terminal region of GDI-1, containing the inhibitory activity, interacts with this exposed α-helix domain specific for Rho/Rac proteins.

Rho, Rac and Cdc42 (close to the Rac proteins) were shown to take part in the reorganisation of the actin cytoskeleton by regulating signal transduction pathways that link extracellular signals to the formation of actin stress fibre bundles and focal adhesions (RhoA), membrane ruffling activity or lamellipodia at the cell edge (Rac1) and filopodia or microspikes at the cell periphery (Cdc42) (reviewed in 24, 25). These proteins have additional roles: (1) Cdc42 and Rac1 are able to induce a cascade of sequential phosphorylations leading to activation of the Jun N-terminal kinase (JNK/SAPK) 26, 27, (2) RhoA is required for lysophosphatic acid- and serum-induced transcriptional activation by the serum response factor 28, 29, (3) Rac proteins take part in the activation of the superoxide-generating NADPH oxidase system in neutrophils and macrophages 30, 31, 32 and (4) all of them have a role in cell cycle control, because they are needed for progression through G1 [29].

RhoGDI-1 seems to have a broad action and is able to form 1:1 complexes with isoprenylated RhoA [33], RhoB [2], Rac1 and Rac2 34, 35, 36 and Cdc42 [37]. The binding and the inhibitory activity of GDI-1 towards all these proteins seems to be very similar, at least in vitro [38].

Several findings prove that, in resting cells, the Rho-like proteins are cytosolic and complexed in their GDP-bound conformation to GDI-1:

  • Rho/Rac proteins complexed to GDI-1 were isolated from the cytosol of several different resting cell types 6, 13, 30, 37, 39, 40, 41.

  • RhoA–GDI-1 complexes isolated from the cytosol of smooth muscle [42] and neutrophils [43] and the Rac1– or Rac2–GDI-1 complexes isolated from neutrophils and macrophages 31, 32 were found to contain GDP as the only bound nucleotide at a ratio of 1 mol of GDP per mole of complex.

  • In vitro, GDI-1 inhibits the intrinsic and the GEF-induced dissociation of GDP from all these proteins 23, 44, 45.

The reports concerning the affinity of GDI-1 for GDP-bound or GTP-bound forms of proteins in the Rho/Rac family appear somewhat contradictory. Concerning Rho and Rac proteins, GDI-1 is unable to extract the GTP-bound conformation from membranes 10, 46. In addition, GDI-1 binds, in vitro, to the GDP-bound forms of these proteins with a 10-fold higher affinity than for the GTP-bound forms [47]. However, even though GTP-bound RhoA has not yet been found complexed to GDI-1 in the cytosol, a cAMP-dependent phosphorylation of membrane-localised GTP-bound RhoA was found to enable RhoGDI-1 to extract GTP-RhoA from membranes in NK lymphocytes [48]. Moreover, in regard to Cdc42, GDI-1 catalysed the extraction from membranes of the GTP-bound conformation as well as the GDP-bound conformation [44]. More recently, GDI-1 was reported to bind, in vitro, to GDP-bound and GTP-bound conformations of Cdc42 equally well with an apparent Kd value, like that of GDP-bound forms of the other Rho proteins, in the nanomolar range (30 nM) [49]. Unfortunately, the nucleotide content of cytosolic Cdc42–GDI-1 complexes has not been described so far.

Considering these discrepancies, the possibility that some of the Cdc42 or RhoA proteins might be maintained in a GTP-bound form in the cytosol complexed to GDI-1 could not be excluded. Because GDI-1 is known to inhibit, at least in vitro, the intrinsic and GAP-stimulated GTP hydrolysis 50, 51, this complex could allow the transit of the activated form of Cdc42/RhoA between physically separated GEF and effector proteins in the cell.

In the course of cell activation, proteins of the Rho family are released from GDI-1. The release of GDI-1 from these complexes is an important step allowing the activation by guanine nucleotide exchange factors and membrane association of the GTP-bound Rho proteins through their isoprenyl group [52]. Initially, it was reported that no release of GDI-1 from Rac occurred in the absence of cell membranes [52]. Indeed, phospholipids have been shown, in vitro, to specifically enhance the release of Rac from GDI-1. Particularly effective were arachidonic acid, phosphatidic acid and phosphatidylinositols, and these lipids were active at concentrations of 0.5–50 μM [41].

Small GTP-binding proteins can act as foci for the formation of signalling complexes. RhoGDI-1 has been found in several of these multi-protein complexes, suggesting a more complicated role for GDI-1.

The use of constitutively activated RacG12V and RhoAG14V mutants showed that these proteins take part in the activation of the phosphatidylinositol phosphate (PtdInsP) 5-kinase, increasing in this way the phosphatidylinositol (PI)-4.5-P2 levels in platelets (Rac1) and fibroblasts (RhoA) 53, 54.

Interestingly, Rac1 and GDI-1 were recently isolated in association with a lipid kinase complex containing type I PtdInsP 5-kinase and diacylglycerol kinase (DGK) [55]. GDI-1 was found to associate with this complex in vitro and in vivo primarily through its interaction with Rac. Several findings suggest that the two lipid kinases are probably not directly activated by Rac:

  • To begin with, both Rac and RhoA proteins associate in vitro and in vivo with type I PtdInsP 5-kinase in a GTP-independent manner 56, 57. However, effector proteins, including serine/threonine and tyrosine kinases, a protein phosphatase, adapter molecules and phosphoinositide kinases (reviewed in 24, 25), are known to bind preferentially to the GTP-bound form of the Rho/Rac proteins. Thus PI-3-kinase binds to Cdc42 and Rac1 in a GTP-dependent manner 56, 58, 59.

  • Moreover, PtdInsP 5-kinase and DGK were found to interact as a complex with Rac1 [55]. Most target proteins require the effector domain (residues 30–40) of the Rho subfamily proteins for binding. However, a sequence within the last 50 carboxy-terminal residues of the GDP-bound Rac1, and thus outside the effector region, was sufficient for the binding of this lipid kinase complex [55].

  • Futhermore, DGK phosphorylates diacylglycerol producing phosphatidic acid (PA) known to stimulate specifically the type I PtdInsP 5-kinase [60].

A model was proposed by Tolias et al. [55] in which PtdInsP 5-kinase, DGK and GDP-bound Rac exist as a pre-formed cytosolic complex bound to GDI-1. When this GDI-Rac–lipid kinase complex is shuttled to the membrane, upon cell stimulation, synthesis of PA by DGK could activate the PtdInsP 5-kinase. PI-4,5-P2 produced in this way could stimulate the release of Rac from GDI-1, allowing Rac to be activated by nucleotide exchange factors (Fig. 4A). PI-4,5-P2 was also implicated in a number of Rho family functions. Indeed, PI-4,5-P2 binds to and regulates several actin regulatory proteins, including gelsolin, profilin, α-actinin and capZ (reviewed in [61].

RhoGDI-1 was also found to associate with a protein complex containing ezrin, radixin, and moesin (ERM) and their membrane-binding partner CD44 in BHK cells [62]. The ERM proteins participate in the actin filament/plasma membrane interaction as cross-linkers, and CD44 has been identified as one of the major membrane-binding partners for ERM proteins. The COOH-terminal part of ERM proteins interacts with actin, and the NH2-terminal part interacts with CD44. It has been postulated that, in the folded structure of native ERM proteins there is a mutual suppression between the NH2-terminal and COOH-terminal parts [63]. The presence of GTP-bound activated RhoA in cell lysates was found to be required for the binding of ERM proteins to CD44, and an in vitro binding assay revealed that PI-4,5-P2 enhanced the affinity of ERM proteins for CD44 [62]. These findings indicate that RhoA regulates the ERM/CD44 complex formation in vivo probably by up-regulating the PtdInsP 5-kinase, with subsequent enhancement of the PI-4,5-P2 level in plasma membranes. PI-4,5-P2 can then function as a modulator of the unfolding mechanism of ERM proteins, permitting their interaction with actin and CD44. However, it was also found, in MDCK cells, that microinjection of the guanosine 5′-(3-O-thio)triphosphate-bound activated form of RhoA did not induce an increase in the localisation of the ERM family at the plasma membrane, indicating that activation of RhoA itself is not sufficient for this relocalisation [64]. This finding suggests the intervention of other molecules. GDI-1, found in immunoprecipitated CD44/ERM complexes [62], could be a good candidate. Indeed, GDI-1 directly interacts with the NH2-terminal of ERM proteins in vitro, and the exogenous NH2-terminal part of ezrin, introduced into intact COS-7 cells, replaced the endogenous RhoA in the GDI-1 complex isolated from the cells [65]. However, the full-length ERM complex should first open its folded structure for the interaction with GDI-1. This unfolding mechanism remains to be clarified.

These findings nevertheless suggest that the ERM proteins dissociate GDP-bound RhoA from GDI-1, permitting the activation of RhoA by exchange factors, and that the formation of ERM–GDI-1 complexes could be important for the localisation of ERM proteins near the membranous CD44 (Fig. 4B).

Activation of the superoxide-generating NADPH oxidase of phagocytes requires the interaction of membrane-associated cytochrome b559 with three cytosolic components: p47-phox, p67-phox and a third component found to be Rac proteins complexed to GDI-1. Here again GDI-1 could be important for the regulation of the intracellular localisation of a multi-protein complex containing Rac, p47-phox and p67-phox 30, 31, 32, 39, 40.

A cytosolic Rac1–GDI-1 complex was found to regulate secretion in streptolysin-O permeabilised mast cells [66]. Both proteins were required for secretion in these permeabilised cells, whereas wild-type Rac alone had no effect in the assay, and introduction of exogenous Rho–GDI-1 inhibited exocytosis [67].

Taken together, these findings suggest that GDI-1 is important in regulating the membrane association/dissociation cycle not only of Rho/Rac proteins, but also of other proteins found associated with Rho family proteins in several signalling complexes.

On the other hand, the introduction of exogenous GDI-1 alone into living cells, leading to an excess of intracellular GDI-1, impairs cellular signalling dependent on the activation of one of the Rho/Rac proteins. RhoGDI-1 proteins have therefore often been microinjected into cells to identify the physiological functions of Rho/Rac proteins. Thus, microinjection of GDI-1 was shown to inhibit several cellular functions such as:

  • the NADPH oxidase activity in phagocytes [35]

  • the chemokinesis and the formation of stress fibres in Swiss 3T3 fibroblasts 68, 69

  • the sperm-induced cytoplasmic division of Xenopus embryos [70]

  • the hepatocyte growth factor (HGF)-induced motility of cultured mouse keratinocytes [71]

  • the insulin-, HGF- or 12-O-tetradecanoylphorbol-13-acetate (TPA)-induced membrane ruffling in KB cells (Insulin-induced ruffling was shown to involve Rac1, whereas HGF- or TPA-induced ruffling was shown to involve RhoA [72]

  • the exocytosis in preventing a partial disruption of the cortical actin network that accompanies exocytosis [73]

  • the transcription of muscle-specific genes [74]

The involvement of each protein in the Rho/Rac subfamily in these effects could then be established by a co-microinjection of the GTP-bound active form of one of the proteins.

GDI-2 is specifically expressed in haematopoietic tissues and predominately in B- and T-lymphocyte cell lines 7, 9. GDI-2 was initially found to act as an inhibitor of GDP dissociation from RhoA [7]. Rac1 and Cdc42 [9]. However, GDI-2 was found to be less efficient by a factor of 10 to 20 than GDI-1 in its capacity to inhibit GDP/GTP exchange or to promote membrane dissociation of all three proteins [9]. Later, the binding affinity of GDI-1 to Cdc42 was found to be 15-fold lower (Kd ~440 nM) than the binding affinity of GD-1 to the same molecule (Kd ~30 nM) [49]. The use of GDI-1/GDI-2 chimeras and point mutations demonstrated that a single residue (Ile 177 in GDI-1/Asp 174 in GDI-2) (see Figure 2, Figure 3) was responsible for much of the difference in the abilities of these proteins to act as a GDI for Cdc42 [16]. More recently, separation of GDI-1 and GDI-2 complexes from the cytosol of a human myelomonocytic cell line by microanalytical liquid chromatography demonstrated that the two GDIs associate with different Rho protein partners [13]. Indeed, GDI-1 formed complexes with Cdc42, RhoA, Rac1 and Rac2, whereas none of these proteins was found in the 45,000–50,000 Mr complex containing GDI-2. These findings demonstrated that GDI-2 is not able to form stable complexes in vivo with these proteins and suggested its association with some still uncharacterised Rho proteins.

The biochemical pathways regulated by GDI-2 are unknown as yet. Nevertheless, stimulation of T lymphocytes and myelomonocytic cells with phorbol esters leads to phosphorylation of GDI-2, as already described in this review, suggesting an involvement of GDI-2 in activation pathways of haematopoietic cells 7, 13. In addition, the expression of GDI-2 was dramatically increased during proliferation and differentiation in vitro of haematopoietic precursors from murine yolk sac [9], suggesting an important function of GDI-2 during maturation of these cells. Targeted gene disruption of GDI-2 showed a minimal phenotypic effect, indicating that its function may be partly complemented by other RhoGDIs. However, GDI-2 (−/−) macrophages, derived from in vitro embryonal stem cell differentiation, showed a slight but consistent reduction in their capacity to generate superoxide [75]. Moreover, GDI-2 deficiency in mice specifically decreased interleukin-2 withdrawal of apoptosis of lymph node cells [76], suggesting its implication in the regulation of lymphocyte survival.

Indeed, interesting results concern the capacity of GDI-2 to be proteolysed by caspase proteases. These enzymes have been shown to play key roles in inflammation and apoptosis, or programmed cell death, in mammalian cells. The caspases fall into two major subfamilies on the basis of sequence homology: the interleukin-1β-converting enzyme subfamily and the cell death-3 (CED-3) subfamily. Accumulating evidence indicates that members of the ICE subfamily predominantly play a role in inflammation, whereas members of the CED-3 subfamily largely participate in apoptosis. GDI-2 was found to be proteolysed by two enzymes: caspase-3/CPP32 in the CED-3 subfamily and caspase-1/ICE in the ICE subfamily 18, 19, 20. Apoptosis is important for the preservation of peripheral T cell homeostasis. In Fas(CD95)-induced apoptosis in Jurkat T cells, GDI-2 was rapidly truncated to a 23,000 Mr fragment by CPP32 (caspase-3) [18]. GDI-2 was cleaved between Asp 19 and Ser 20 in the cleavage site DELDS (green residues in Fig. 2). Negative selection by apoptosis, mediated by surface immunoglobulin M (IgM) signalling after encountering self antigen, eliminates also autoreactive B cells. Interestingly, GDI-2 was found to be cleaved by CPP32 in anti-IgM-mediated apoptosis in the human Burkitt lymphoma cell line, BL60 [20]. Cleavage of GDI-2 by caspase-1/ICE occurred at Asp 55 in the cleavage site LLGDG (green residues in Fig. 2), leaving a 19,000 Mr fragment 18, 19. As already stated in this review, the mutational and structural analysis of GDI-1 leads to the hypothesis that at least the proteolysis at Asp 55 should render GDI-2 unable to effectively regulate the unknown Rho-like protein complexed to this GDI.

We identified GDI-3 by using the yeast two-hybrid system to characterise proteins interacting with RhoB [10]. GDI-3 was found to interact, in this system, specifically with post-translationally processed RhoB and RhoG proteins, both of which show a growth-regulated expression in mammalian cells 77, 78. RhoB is more than 85% identical with RhoA, whereas RhoG is more closely related to the Rac and Cdc42 proteins of the Rho/Rac subfamily. We showed, however, that a deletion of the first 68 amino-terminal residues of GDI-3, corresponding to the majority of the disordered amino-terminus in GDI-1 (Fig. 2), prevented the interaction with RhoG in the yeast two-hybrid system. Moreover, no interaction was found with RhoA, RhoC or Rac1. In contrast, the human homologue of GDI-3 was found to interact, in vitro, with RhoA and Cdc42 but with less affinity compared with RhoGDI-1 [11].

We showed that purified GDI-3 was able to release GDP-bound but not GTP-bound RhoB from cell membranes and to inhibit the GDP/GTP exchange of RhoB [10]. GDI-3 showed less ability to inhibit the dissociation of pre-bound GTP. This finding could be due to a lower affinity of GDI-3 for the GTP-bound RhoB protein.

The most remarkable feature of GDI-3 is its association with particulate subcellular fractions, in contrast with all other RhoGDI and RabGDI proteins, which are cytosolic. Interestingly, GDI-3 was not extracted from this particulate fraction by detergents (Triton X100 or CHAPS) [10].

GDI-3 has 21 additional amino acids at the NH2-terminus compared with GDI-1 (see Fig. 2). This extension contains a strongly amphipatic α helix stabilised in its amino-terminal part by a characteristic capping structure, LDXXEL [10] (see Fig. 2). It is unlikely that this α helix, on the whole negatively charged, interacts directly with membranous phospholipids. This helix structure may instead be implicated in protein–protein interactions and permit GDI-3 association with an actin cytoskeletal subcellular fraction, as suggested by the association of GDI-3 with a Triton X100-insoluble fraction. Thus this distinct amino terminus in GDI-3, not seen in the other two GDIs, could provide a mechanism for localisation of GDI-3 to specific intracellular sites. Indeed, immunohistochemical studies showed a diffuse punctate distribution of exogenous GDI-3 in the cytoplasm with concentration around the nucleus in cytoplasmic vesicles [11]. In addition, truncation of this amino-terminal part prevented the association of GDI-3 to the particulate subcellular fraction (unpublished results).

Several observations lead to the hypothesis that GDI-3 preferentially acts on RhoB in vivo:

  • To begin with, GDI-3 shows an expression restricted to brain, lung and testis. Strikingly, RhoB is also mainly expressed in these organs.

  • Moreover, the intracellular localisation of GDI-3 is reminiscent of that found for endogenous RhoB [77]. Indeed, the endogenous RhoB protein was found, by immunofluorescence studies, to be associated to vesicular-like structures extending forward from a heavily stained juxta-nuclear compartment [77].

  • Futhermore, the non-cytosolic GDI-3 should not be able to form a cytosolic complex with Rho proteins in resting cells. Once again, RhoB is a good candidate for being a target for GDI-3. Indeed, the expression of RhoB, encoded by an immediate early gene, is induced by growth factors in cell activation. We have reported that the transient expression of endogenous RhoB is regulated during the cell cycle, contrasting with the permanent expression of the other Rho/Rac proteins. First detected at the G1/S phase transition, the level of the RhoB protein is maximal during S phase and disappears at the S/G2-M transition [77]. In addition, the epidermal-growth-factor– and cell-cycle–induced RhoB protein [77] as well as the transiently expressed epitope-tagged RhoB protein [79] have been found associated only with membranous subcellular fractions. Hence, RhoB seems never to cycle between cytosol and membranes and is not present in resting cells when the other Rho-like proteins are maintained in a soluble complex with GDI-1 or -2. This, in turn, implies that the release of the GDP-bound form of RhoB from HeLa cell membranes by the recombinant GDI-3 protein may not be indicative of the in vivo function of GDI-3.

These observations suggest a different role for GDI-3 compared with that of cytosolic GDI proteins. One could speculate that GDI-3 is able to link the vesicular-associated RhoB to a cytoskeletal compartment and thereby regulate the RhoB protein activity.

Section snippets

Concluding remarks

There are still many gaps in our knowledge of the mechanisms that shuttle Rho/Rac-GDI complexes to the membrane upon cell stimulation. Many questions have to be answered concerning the role and the specificity of RhoGDIs. Among the three RhoGDIs described as yet, only RhoGDI-3 seems to act specifically on a single Rho/Rac protein—namely, RhoB. The target(s) of RhoGDI-2 are still to be found. The most troublesome point is the broad specificity of the ubiquitously expressed RhoGDI-1 found

Acknowledgements

I thank Jacqueline Cherfils (LEBS, CNRS, Gif-sur Yvette, France), for expert assistance concerning Fig. 3, and Alain Sanson (CEA, Saclay, France), for critical reading of the manuscript. This work was supported in part by CNRS and grants from the Association pour la Recherche sur le Cancer and the Fondation pour la Recherche Médicale.

References (81)

  • N. Ohga et al.

    Biochem. Biophys. Res. Commun.

    (1989)
  • T. Ueda et al.

    J. Biol. Chem.

    (1990)
  • H. Leffers et al.

    Exp. Cell Res.

    (1993)
  • T. Masuda et al.

    Biol. Chem.

    (1994)
  • G. Zalcman et al.

    J. Biol. Chem.

    (1996)
  • N. Bourmeyster et al.

    Biochem. Biophy. Res. Commun.

    (1996)
  • J.P. Gorvel et al.

    FEBS Lett.

    (1998)
  • N.H. Keep et al.

    Structure

    (1997)
  • S.Q. Na et al.

    J. Biol. Chem.

    (1996)
  • W.J. Wu et al.

    J. Biol. Chem.

    (1997)
  • N. Tapon et al.

    Curr. Opin. Cell Biol.

    (1997)
  • O.A. Coso et al.

    Cell

    (1995)
  • A. Minden et al.

    Cell

    (1995)
  • C.S. Hill et al.

    Cell

    (1995)
  • M.T. Quinn et al.

    J. Biol. Chem.

    (1993)
  • Y. Bromberg et al.

    J. Biol. Chem.

    (1994)
  • K. Hiraoka et al.

    Biochem. Biophys. Res. Commun.

    (1992)
  • T. Mizuno et al.

    J. Biol. Chem.

    (1992)
  • S. Ando et al.

    J. Biol. Chem.

    (1992)
  • R. Regazzi et al.

    J. Biol. Chem.

    (1992)
  • T. Sasaki et al.

    Biochem. Biophys. Res. Commun.

    (1993)
  • T.H. Chuang et al.

    J. Biol. Chem.

    (1993)
  • S. Kuroda et al.

    Biochem. Biophys. Res. Commun.

    (1992)
  • N. Bourmeyster et al.

    Biochem. Biophys. Res. Commun.

    (1994)
  • D. Leonard et al.

    J. Biol. Chem.

    (1992)
  • A. Kikuchi et al.

    J. Biol. Chem.

    (1992)
  • T. Sasaki et al.

    J. Biol. Chem.

    (1993)
  • T.K. Nomanbhoy et al.

    J. Biol. Chem.

    (1996)
  • T.H. Chuang et al.

    J. Biol. Chem.

    (1993)
  • G.M. Bokoch et al.

    J. Biol. Chem.

    (1994)
  • J.H. Hartwig et al.

    Cell

    (1995)
  • L.D. Chong et al.

    Cell

    (1994)
  • K.F. Tolias et al.

    J. Biol. Chem.

    (1995)
  • Y. Zheng et al.

    J. Biol. Chem.

    (1994)
  • G.H. Jenkins et al.

    J. Biol. Chem.

    (1994)
  • K. Takahashi et al.

    J. Biol. Chem.

    (1997)
  • Y. Miura et al.

    J. Biol. Chem.

    (1993)
  • J.C. Guillemot et al.

    Blood

    (1996)
  • L. Yin et al.

    Mol. Immunol.

    (1997)
  • X. Huang et al.

    Adv. Appl. Math.

    (1991)
  • Cited by (0)

    View full text