The Rho-family proteins make up a major branch of the Ras superfamily of small GTPases. To date, 22 human genes encoding at least 25 proteins have been described. The best known `classical' members are RhoA, Rac1 and Cdc42. Highly related isoforms of these three proteins have not been studied as intensively, in part because it has been assumed that they are functionally identical to their better-studied counterparts. This now appears not to be the case. Variations in C-terminal-signaled modifications and subcellular targeting cause otherwise highly biochemically related isoforms (e.g. RhoA, RhoB and RhoC) to exhibit surprisingly divergent biological activities. Whereas the classical Rho GTPases are regulated by GDP/GTP cycling, other Rho GTPases are also regulated by other mechanisms, particularly by transcriptional regulation. Newer members of the family possess additional sequence elements beyond the GTPase domain, which suggests they exhibit yet other mechanisms of regulation.
The Rho-family proteins are defined by the presence of a Rho-type GTPase-like domain. A structural feature that distinguishes the Rho proteins from other small GTPases is the so-called Rho insert domain located between the fifth β strand and the fourth α helix in the small GTPase domain (Valencia et al., 1991). Most typical Rho proteins are small (190-250 residues) and consist only of the GTPase domain and short N- and C-terminal extensions. However, some of the more atypical family members contain additional well-defined domains and can be >700 amino acids long. Within their GTPase domains, they share approximately 30% amino acid identity with the Ras proteins and 40-95% identity within the family. All members contain the sequence motifs characteristic of all GTP-binding proteins, bind to GDP and GTP with high affinity, and are thought to cycle between active, GTP-bound and inactive, GDP-bound states. In addition, the majority of members undergo C-terminal post-translational modification by isoprenoid lipids (Fig. 1). Together with other C-terminal modifications or sequences, isoprenoid addition facilitates their subcellular location and association with specific membranes, which is crucial for their functions. However, the Chp, RhoBTB and Miro proteins lack this mode of membrane targeting.
Most of the functional information on Rho-family proteins has come from studies of RhoA, Rac1 and Cdc42. All share common growth-promoting and anti-apoptotic functions, as well as regulation of gene expression, through activation of signaling molecules such as serum response factor, NF-κB, the stress-activated protein kinases and cyclin D1 (Pruitt and Der, 2001; Van Aelst and D'Souza-Schorey, 1997). All three promote actin cytoskeleton reorganization, but have distinct effects on cell shape and movement (Hall, 1998; Schmitz et al., 2000). RhoA promotes actin-myosin contractility and, thereby, the formation of stress fibers and focal adhesions, regulating cell shape, attachment and motility. Rac1 promotes actin polymerization and the formation of lamellipodia, which are curtain-like extensions that consist of thin protrusive actin sheets at the leading edge of migrating cells. Cdc42 causes formation of filopodia, which are thin, finger-like cytoplasmic extensions that contain tight actin bundles and might be involved in the recognition of the extracellular environment.
In common with Ras and other small GTPases, the GDP- and GTP-bound states of the classical Rho GTPases differ in conformation primarily in two localized regions, switch I and II, and this feature is probably shared by all Rho GTPases (Vetter and Wittinghofer, 2001). The conformation of the GTP-bound protein results in increased binding affinity for downstream effector proteins. Each Rho-family protein recognizes multiple effectors, and some effectors are recognized by multiple family members. Interaction with and activation of the effector function stimulates signaling pathways that mediate the diverse functions of Rho-family proteins. The multitude of effectors identified for RhoA, Rac1, Cdc42 and other family members (Aspenström, 1999; Bishop and Hall, 2000) reflects the complex and diverse functional properties of these proteins.
The Rho-family of proteins
On the basis of primary amino acid sequence identity, structural motifs and biological function, the Rho family can be divided into six subfamilies that exhibit similar, but not identical, properties (Figs 2, 3). These are: the RhoA-related subfamily (RhoA, RhoB and RhoC); the Rac1-related subfamily [Rac1 (and its splice variant Rac1b), Rac2, Rac3 and RhoG]; the Cdc42-related subfamily [Cdc42 (and its brain-specific C-terminal splice variant G25K), TC10, TCL, Chp/Wrch-2 and Wrch-1]; the Rnd subfamily (Rnd1, Rnd2, and RhoE/Rnd3); the RhoBTB subfamily; and the recently described Miro subfamily. In addition, RhoD, Rif and TTF/RhoH do not obviously fall into any of these subfamilies.
RhoA, RhoB and RhoC exhibit significant amino acid sequence identity (∼85%), all stimulate actin-myosin contractility, and all are thought to interact with the same guanine nucleotide exchange factors (GEFs) and effectors. However, they clearly have some functional differences (see below). In large part, these are likely to be a consequence of divergence in their C-terminal 15 amino acids (where the highest degree of difference is found; Fig. 2), which dictate distinct subcellular locations. When overexpressed in cells as epitope-tagged proteins, RhoA is cytosolic and to a certain degree bound to the plasma membrane, RhoB is associated with the plasma membrane and endomembrane vesicles, and RhoC is cytosolic and associated with undefined perinuclear structures (Adamson et al., 1992a; Michaelson et al., 2001; Wang et al., 2003).
RhoB (Chardin et al., 1988) contrasts with RhoA and is also transcriptionally regulated. Both the RhoB mRNA and protein are unstable molecules and are upregulated by growth factors and during the G1 and S phases of the cell cycle (Jahner and Hunter, 1991; Zalcman et al., 1995), which indicates that RhoB has a role in cell proliferation. RhoB can be prenylated by either a geranylgeranyl (GG), like most other Rho proteins, or a farnesyl (F) isoprenoid group, and is modified additionally by the palmitate fatty acid (Adamson et al., 1992b) (Fig. 1). As opposed to RhoA and RhoC, RhoB seems to have a growth-inhibitory effect (Chen et al., 2000; Du and Prendergast, 1999). Expression of RhoB has been reported to be downregulated in tumors (Adnane et al., 2002). RhoB-deficient mice, which develop normally, show enhanced carcinogen-induced skin tumor formation (Liu et al., 2001). However, cell-type differences are seen and, like constitutively activated RhoA, constitutively activated RhoB can transform rodent fibroblasts (Prendergast et al., 1995; Wang et al., 2003). In addition, RhoB, but not RhoA or RhoC, can regulate the transport of late endosomes (Gampel et al., 1999; Mellor et al., 1998). Finally, RhoB might be targeted by farnesyltransferase inhibitors (FTIs). Although developed originally as anti-Ras drugs, the FTIs might exert their anti-tumor activity by forcing a shift of RhoB from the F to the GG form (Prendergast, 2000).
RhoC (Chardin et al., 1988) has also been connected to cancer development. It is upregulated in malignant pancreatic ductal carcinomas (Suwa et al., 1998), inflammatory breast cancer tumors (van Golen et al., 2000a) and highly metastatic melanomas (Clark et al., 2000). Ectopic overexpression of RhoC increases the tumorigenic and metastatic properties of tumor progenitor cells (Clark et al., 2000; van Golen et al., 2000a). In addition, RhoC induces the expression of angiogenic factors in human mammary epithelial cells (van Golen et al., 2000b), possibly facilitating the vascularization of tumors in which it is expressed. By contrast, activated RhoC does not transform NIH 3T3 mouse fibroblasts (Wang et al., 2003). Interestingly, we have found a GEF that interacts with RhoA and RhoB but not with RhoC (Arthur et al., 2002), and Sahai and Marshall have reported that RhoC binds more efficiently to Rho kinase, a major effector for the RhoA-like proteins, than does RhoA (Sahai and Marshall, 2002).
All the Rac-related proteins stimulate the formation of lamellipodia and membrane ruffles, presumably through interaction with the PIR121-Nap125-HSPC300-WAVE complex (Eden et al., 2002). Within the subfamily, Rac1, Rac2 and Rac3 share significant sequence identity (∼88%). These three diverge primarily in the C-terminal 15 residues, whereas RhoG is somewhat more divergent overall (72% identical to Rac1). A splice variant of Rac1, Rac1b, generated by alternative exon usage, contains an additional 19-residue insert directly C-terminal to the switch II region. Rac1b is expressed preferentially in breast and colon cancers (Jordan et al., 1999; Schnelzer et al., 2000), possesses an increased intrinsic guanine nucleotide exchange rate and decreased intrinsic GTPase activity (Fiegen et al., 2004; Matos et al., 2003). Rac1b is also unable to interact with Rho guanine nucleotide dissociation inhibitor (RhoGDI) and, thus, it exhibits enhanced association with the plasma membrane (Matos et al., 2003). Consequently, it exhibits properties consistent with a constitutively activated protein. However, Rac1b also has differential signaling properties. It is unable to activate the JNK stress-activated kinase, but retains the ability to activate the NF-κB transcription factor (Matos et al., 2003). This might be explained by a differential binding to effectors, as highlighted by its strongly reduced affinity for the PAK serine/threonine kinase (Fiegen et al., 2003; Matos et al., 2003).
Rac1 expression is ubiquitous, whereas Rac2 (Didsbury et al., 1989) is expressed only in hematopoietic cells, where it seems to have specialized functions. A lack of Rac1 results in embryonic lethality (Sugihara et al., 1998), whereas a Rac2 deficiency allows normal development but results in mice that have hematopoietic cell defects (Roberts et al., 1999). Rac2 downregulation, deletion or inactivation correlates with several neutrophilic, phagocytic and lymphocytic defects (Ambruso et al., 2000; Kasper et al., 2000; Li et al., 2000; Roberts et al., 1999; Williams et al., 2000). Many of these defects are probably caused by the Rac2-specific activation of NADPH oxidase (Dorseuil et al., 1996; Mizuno et al., 1992) and generation of reactive oxygen species (ROS) in hematopoietic cells (Werner, 2004). ROS production is important both in host defense mechanisms (Roberts et al., 1999) and in blood cell differentiation signaling (Li et al., 2000). The additional defects in hematopoietic cell function observed with a combined Rac1 and Rac2 deficiency further emphasize the overlapping and distinct roles of these two related GTPases in hematopoietic cells (Gu et al., 2003).
Rac3 is most highly expressed in brain but is upregulated upon serum stimulation of fibroblasts (Haataja et al., 1997). It is localized to the membrane to a higher degree than is Rac1 (Mira et al., 2000). Furthermore, it is hyperactive in breast cancers (Mira et al., 2000) and the Rac3 gene is close to a region of chromosome 17 that is often deleted in breast cancers (Morris et al., 2000), which might deregulate its expression.
Vincent et al. first described RhoG as a late-response gene induced after serum stimulation of starved fibroblasts, suggesting that it is a transcriptionally regulated gene whose product is involved in regulation of the cell cycle (Vincent et al., 1992). Activated mutants of RhoG induce lamellipodia formation, membrane ruffling and to some extent formation of filopodia (Gauthier-Rouviere et al., 1998). How it signals is controversial; some reports say that it mediates its cellular effects by activating Rac1 and Cdc42 (Gauthier-Rouviere et al., 1998; Katoh and Negishi, 2003; Katoh et al., 2000), whereas others argue that it signals independently of them (Bellanger et al., 2000; Movilla and Bustelo, 1999; Prieto-Sanchez and Bustelo, 2003; Schuebel et al., 1998; Wennerberg et al., 2002). However, a recent report provides a mechanism for how RhoG can activate Rac1, implicating the DOCK180 RhoGEF (Katoh and Negishi, 2003). Katoh et al. have also reported that RhoG plays a specific role in nerve growth factor (NGF)- and/or Ras-mediated neurite outgrowth in PC12 pheochromocytoma cells (Katoh et al., 2000).
The Cdc42-related proteins all stimulate the formation of filopodia, probably through their association with Wiscott-Aldrich syndrome protein (WASP) or N-WASP (Machesky and Insall, 1998; Miki et al., 1998); the possible exception to this is Wrch-1, which does not bind either WASP or N-WASP (Aspenström et al., 2004). Two isoforms of human (and mouse) Cdc42 arise from alternative exon splicing of the same gene (Marks and Kwiatkowski, 1996; Nicole et al., 1999). Both are 191-residue proteins that differ at residue 163 and in their C-terminal 10 amino acids. The most commonly studied form (sometimes referred to as placental Cdc42p, Cdc42Hs or Cdc42a) is expressed ubiquitously, whereas the other form (brain Cdc42p, Cdc42b or G25K) is restricted to the brain.
TC10 (Drivas et al., 1990) and TCL (also known as TC10β and RhoT; Vignal et al., 2000) are very closely related and have functions highly similar to that of Cdc42. TC10- and TCL-induced filopodia are longer than those promoted by activated Cdc42 (Murphy et al., 1999; Neudauer et al., 1998). TC10 and TCL interact with most Cdc42 effectors, but differences clearly exist (Murphy et al., 1999; Neudauer et al., 1998). In contrast to the exclusively geranylgeranylated C-terminus of Cdc42, the C-terminal CAAX motifs of TC10 and TCL suggest that they can be either farnesylated or geranylgeranylated. TC10 is also modified by palmitate, which prevents its recognition by RhoGDI-1 (Michaelson et al., 2001). The activities of TC10 and TCL seem to be regulated by extracellular stimuli different from those that activate Cdc42. TC10 and TCL have yet to be to found to be regulated by GEFs of the Dbl family (Schmidt and Hall, 2002), which are thought to be the most common activators of Rho-family proteins. Instead, Chiang et al. has suggested that TC10 is activated by the Rap GEF C3G upon insulin stimulation of adipocytes (Chiang et al., 2002). TC10 activation then mediates GLUT-4 translocation to the plasma membrane and is necessary for adipocyte differentiation (Nishizuka et al., 2003). In addition, TC10 expression and TCL expression are upregulated during nerve regeneration (Tanabe et al., 2000) and neurite differentiation (Abe et al., 2003), and activation of these GTPases stimulates neurite extension. Two different splice isoforms of TCL have been described (Chiang et al., 2002). One has a 10 amino acid extension at the N-terminus.
Aronheim et al. identified the rat protein Chp (or, as its human version is called, Wrch-2) as a Pak2-serine/threonine-kinase-interacting protein in a yeast two-hybrid screen (Aronheim et al., 1998). In common with activated versions of Cdc42, activated Chp binds Pak and N-WASP and stimulates formation of filopodia (Aronheim et al., 1998). Chp differs from other Rho-family members in that it lacks a conventional C-terminal CAAX prenylation sequence. Nevertheless, the C-terminus is needed for its biological effects (Aronheim et al., 1998).
The gene encoding Wrch-1 is upregulated by the Wnt-1 signaling pathway (Tao et al., 2001). Wnt-1 stimulates the Frizzled cell-surface receptor, leading to inactivation of the APC tumor suppressor and activation of β-catenin transcriptional activity and transformation. Expression of activated Wrch-1 in mammary epithelial cells mimics the morphological effects of Wnt-1 transformation (Tao et al., 2001), indicating that Wrch-1 is a crucial signaling component in Wnt-dependent oncogenesis. To date, it has not been determined whether known RhoGEFs, GTPase-activating proteins (GAPs) or GDIs are regulators of Chp and Wrch-1. However, the ability of a dominant-negative S45N version of Chp (based on analogous mutants of Ras and Rho GTPases that block GEF function) to block tumor necrosis factor α (TNF-α)-mediated signaling argues that Chp GEFs do exist.
Both Chp and Wrch-1 possess additional N-terminal and C-terminal sequences that flank their GTPase domains (Fig. 3). Interestingly, their N-terminal sequences contain proline-rich sequences that might serve as Src-homology 3 (SH3) domain recognition sequences. This suggests that SH3-domain-containing proteins could recognize and regulate their functions. Similarly, their additional C-terminal sequences might distinguish their functions and regulation from those of the more classical Cdc42-related proteins.
The Rnd subfamily
The three Rnd-family members show a fairly high degree of sequence identity to RhoA (45-49%) but nevertheless exhibit biochemical and functional features distinct from those of RhoA. First, they are considered to be GTPase deficient, they are not stable in either a GDP-bound or nucleotide-free form (Nobes et al., 1998), and RhoE/Rnd3 (also known as Rho8) is constitutively GTP bound in vivo (Foster et al., 1996). They also contain a substitution leading to low affinity for nucleotides. Second, they are farnesylated instead of geranylgeranylated (Foster et al., 1996; Nobes et al., 1998). Third, the striking functional feature of at least Rnd1 (also known as Rho6) and RhoE/Rnd3 is that, in contrast to the stimulation of contractility and the subsequent formation of actin stress fibers and focal adhesions caused by the related RhoA GTPase, they inhibit these effects and cause cell rounding (Guasch et al., 1998; Nobes et al., 1998). The mechanism for this seems to be a combination of both binding and activating p190RhoGAP (Wennerberg et al., 2003) and, at least in the case of RhoE/Rnd3, binding and inactivating the RhoA effector ROCK1 (Riento et al., 2003). Another Rnd effector that could be involved in this effect is Socius (Katoh et al., 2002), but it is unclear how this contributes to the rounding activity of these proteins.
Rnd1 (Nobes et al., 1998) is predominantly expressed in adult brain and liver. However, in fetal tissue, it has a more ubiquitous expression pattern (Nobes et al., 1998) and, in Xenopus embryos, it is highly expressed in tissues that undergo extensive morphogenetic changes (Wunnenberg-Stapleton et al., 1999). Ectopic overexpression of Rnd1 in PC12 cells stimulates neurite extension (Aoki et al., 2000) and inhibits Ca2+ sensitization of smooth muscle cells (Loirand et al., 1999), but it is unclear whether these are physiologically significant activities of Rnd1.
Rnd2 (also known as Rho7 and RhoN) (Nobes et al., 1998) is most highly expressed in testis, brain and liver (Nishi et al., 1999; Nobes et al., 1998). Little is known of its function, but it has been shown to bind to three proteins: Vps4-A, a protein involved in endosome sorting (Tanaka et al., 2002); rapostlin, a protein that induces neurite branching (Fujita et al., 2002); and MgcRacGAP, a RhoGAP that has been implicated in cytokinesis (Naud et al., 2003).
Foster et al. initially identified RhoE/Rnd3 as a protein that interacts with p190RhoGAP (Foster et al., 1996), but the significance of this interaction was only clarified recently (Wennerberg et al., 2003) by the finding that p190RhoGAP is an effector in Rnd-mediated inhibition of RhoA function. RhoE/Rnd3 mRNA is ubiquitously expressed, but the protein seems to be sparsely present (Foster et al., 1996). Raf transformation of Madin-Darby canine kidney (MDCK) cells causes upregulation of RhoE/Rnd3, which mediates the stress fiber loss and multilayer growth that accompanies this transformation (Hansen et al., 2000). Similarly, in human epithelial cells, Ras transformation causes an upregulation of RhoE/Rnd3 protein levels (X. Singh and C.J.D., unpublished). Platelet-derived growth factor (PDGF)-stimulation of fibroblasts transiently stimulates synthesis of RhoE/Rnd3 protein, which accompanies the observed transient loss of stress fibers and rounding of the cell body (Riento et al., 2003). Thus, rather than depending on GDP/GTP cycling, RhoE/Rnd3 might be regulated at the level of transcription or translation.
RhoD and Rif
Although RhoD (also called RhoHP1) and Rif are structurally related (48% sequence similarity) and possess additional N-terminal sequences, initial studies indicated that they are functionally divergent with regards to effects on actin organization (Murphy et al., 1996; Ellis and Mellor, 2000). However, more-recent evidence suggests that they have common functions in the regulation of actin organization (Aspenström et al., 2004). RhoD has been implicated in regulating two distinct actin-associated cellular processes. First, it regulates the actin-dependent transport of early endosomes (Murphy et al., 1996) through binding to the Diaphanous-related formin protein hDia2C and subsequently activating Src (Gasman et al., 2003). Second, it promotes the disassembly of actin stress fibers and disruption of focal adhesions and inhibits cell motility (Murphy et al., 2001; Tsubakimoto et al., 1999). At this point, it is unclear whether the two processes are related. Another study showed that transient expression of a GTPase-deficient mutant of RhoD induces formation of filopodia, an activity similar to that seen with GTPase-deficient Rif in endothelial cells (Aspenström et al., 2004). This suggests that regulation of actin organization by RhoD depends on the cellular context.
Rif is widely expressed and promotes the formation of filopodia (Ellis and Mellor, 2000) even though it is only remotely related to Cdc42 (43% identity) and, on the basis of its effector domain sequence, one would not expect it to interact with Cdc42 effectors involved in actin reorganization (N-WASP and WASP). Interestingly, Rif-dependent filopodia formation is independent of Cdc42 and vice versa (Aspenström et al., 2004; Ellis and Mellor, 2000), and instead the two GTPases can cooperate in the generation of these structures.
RhoH/TTF (`translocation three four') was identified originally as a chimeric transcript from a chromosome 3:4 translocation (Dallery et al., 1995) and subsequently renamed RhoH because of its hematopoietic-cell-specific expression. The RhoH/TTF gene is rearranged and deregulated or fused to other genes in both non-Hodgkin's lymphoma and multiple myeloma (Preudhomme et al., 2000). RhoH/TTF possesses a Rho insert sequence, albeit shorter than that seen in other Rho GTPases. Like the Rnd proteins, it lacks the conserved residues corresponding to G12 and Q61 found in other Rho GTPases and consequently is likely to be GTPase deficient and hence not regulated by GDP/GTP cycling (Li et al., 2002). Instead, the downregulation of RhoH/TTF expression by phorbol myristic acid suggests it is transcriptionally regulated. RhoH/TTF also interacts with all three RhoGDIs, which suggests that regulation of subcellular location is also another important mode of regulation. As opposed to most other Rho proteins, ectopic overexpression of activated RhoH/TTF has no effects on the actin cytoskeleton (Aspenström et al., 2004) but, instead, inhibits Rac1, RhoA and Cd42 signaling in T cells (Li et al., 2002).
The RhoBTB subfamily
Rivero et al. initially identified the RhoBTB subfamily of proteins in a database search for Rho-related GTPases in Dictyostelium, finding orthologs in Drosophila and vertebrates (Rivero et al., 2001). These atypical Rho GTPases are structurally different from the other Rho-family members and possess significant additional sequences following their Rho GTPase domain (Fig. 1). The additional C-terminal sequences include a tandem repeat of BTB domains and a lack of C-terminal CAAX prenylation signals. The BTB domain was identified initially in the Drosophila transcriptional repressors Broad complex, Tramtrack and Bric-a-brac, and can be found in ∼200 human proteins (Collins et al., 2001). This family consists of three members in mammals: RhoBTB-1, RhoBTB-2 and RhoBTB-3. RhoBTB-1 and RhoBTB-2 possess the Rho insert sequence that distinguishes Rho GTPases from all other Ras superfamily proteins. However, the GTPase domain of RhoBTB-3 is poorly conserved and not a typical Rho domain, and RhoBTB-3 should therefore not be considered a Rho-family protein.
RhoBTB2 was identified independently as a gene (DBC2) that is deleted, mutated or not expressed in many breast cancer cell lines and tissue samples (Hamaguchi et al., 2002). Expression of the wild-type protein, but not of mutant forms found in breast cancers, inhibits proliferation of T47D human breast cancer cells (Hamaguchi et al., 2002). Ectopic expression of RhoBTB-1 or RhoBTB-2, unlike most other Rho GTPases, does not cause alterations in actin organization or cell morphology (Aspenström et al., 2004).
The Miro subfamily
The Miro subfamily consists of two proteins: Miro-1 and Miro-2 (Fransson et al., 2003). These proteins contain two putative GTPase domains and two EF-hand motifs. The N-terminal GTPase domain bears most sequence similarity to other Rho GTPases. Because these proteins were found in a database search for proteins containing a Rho consensus domain, they have been referred to as atypical Rho proteins (Fransson et al., 2003). However, given their overall sequence divergence from other Rho GTPases, together with the lack of the Rho-specific insert sequences (Fig. 3), perhaps these should not be considered Rho proteins. Instead, they could be seen as a new family of Ras superfamily GTPases. The Miro GTPases localize to mitochondria and might promote apoptosis (Fransson et al., 2003). They do not have any effects on the actin cytoskeleton or cell morphology (Aspenström et al., 2004).
Regulation of Rho protein activity
Traditionally, the activities of Rho GTPases are considered to be regulated by the switching between an inactive, GDP-bound form and an active, GTP-bound form. However, recent findings have indicated that their activities are regulated in several additional ways. These alternative modes of regulation seem to be especially important for many of the less studied Rho GTPases.
Regulation by nucleotide binding
The GDP/GTP cycling of Rho-family proteins is controlled mainly by three distinct functional classes of regulatory protein: (1) GEFs, which stimulate the weak intrinsic exchange activity of Rho-family proteins to promote an exchange of the bound GDP for GTP and thus formation of active Rho-GTP (Schmidt and Hall, 2002) and association with downstream effectors; (2) GAPs, which stimulate the intrinsic GTP hydrolysis activity of Rho-family proteins and thereby promote formation of the inactive GDP-bound protein (Moon and Zheng, 2003); and (3) GDIs, which inhibit Rho proteins by blocking nucleotide exchange, and thus the binding of effectors and GAPs to GTP-bound Rho GTPases (Olofsson, 1999). RhoGDIs also regulate association of Rho GTPases with membranes.
With a few exceptions, the regulation of these three families is unclear. The RhoGEF family is large, consisting of ∼60 members belonging to the Dbl family of proteins (Schmidt and Hall, 2002) and ∼10 members belonging to the more recently discovered Dock family (Côté and Vuori, 2002). The RhoGAP family is also vast: ∼80 distinct members are encoded in the human genome, and their modes of regulation are defined even less well than those of the RhoGEFs. Outside their RhoGEF or RhoGAP domains, these proteins are highly diverse in sequence and domain structure. This diversity probably reflects the fact that Rho GTPase activity can be regulated in a multitude of fashions (Moon and Zheng, 2003; Schmidt and Hall, 2002).
Regulation by lipid modification and subcellular localization
Rho GTPase function is critically dependent on association of the GTPases with membranes and their subcellular locations (Fig. 4). These properties are influenced by C-terminal lipid modifications and sequences (Fig. 1), as well as by interaction with RhoGDIs. The diversity of the cellular distributions of Rho GTPases reflects the important role that location has in their function.
As opposed to RhoGEFs and RhoGAPs, there are only a few known RhoGDIs. Three well-defined RhoGDIs have been identified: RhoGDI-1 (α), RHoGDI-2 (D4/LyGDI/β), and RhoGDI-3 (γ), and a few other proteins are suggested to have GDI activity towards Rho proteins (Anastasiadis et al., 2000; Degani et al., 2002). The three conventional GDIs seem to have different Rho-binding specificites. Whereas RhoGDI-1 binds well to RhoA, Rac1 and Cdc42, RhoGDI-3 binds well to RhoB and RhoG, and seemingly with a lower affinity to RhoA and Cdc42, but not to Rac1 and Rac2 (Adra et al., 1997; Zalcman et al., 1996). The targets for RhoGDI-2 remain unclear, but it does not bind well to RhoA, Rac1 or Cdc42 (Olofsson, 1999). Interestingly, the RhoGDI-1 concentration in the cell compares with that of the target GTPases (Michaelson et al., 2001). This stoichiometry indicates that a large fraction of Rho proteins in a cell are bound and inactivated by GDIs and therefore that binding of Rho to Rho-GDI is a potent regulatory mechanism. Several proteins bind to RhoGDIs and might thereby regulate their association with Rho proteins (Takahashi et al., 1997; Yamashita and Tohyama, 2003), for example, by releasing Rho proteins at membrane sites where they are needed. Changes in RhoGDI expression have been reported in some cancers (Gildea et al., 2002; Jones et al., 2002).
The majority of Rho GTPases terminate with CAAX tetrapeptide sequences that are isoprenylated. Some Rho GTPases are modified by the C15 F isoprenoid, and others are modified by the more hydrophobic C20 GG isoprenoid modification (Ghomashchi et al., 1995; Silvius and l'Heureux, 1994) (Fig. 1). Despite the fact that the GG modification promotes stronger membrane association, the type of isoprenoid does not seem to contribute to functional distinctions, except perhaps in the case RhoB (Allal et al., 2000; Joyce and Cox, 2003; Michaelson et al., 2001; Solski et al., 2002). Isoprenylation alone is not sufficient to determine the subcellular location of Rho GTPases. As in the case of Ras proteins, a second C-terminal signal is required. Some members are modified additionally by palmitate at C-terminal cysteine residues immediately upstream of the CAAX motif, whereas some contain lysine/arginine-rich sequences, and others contain both sequence elements. Palmitoylation can also influence RhoGDI recognition (Michaelson et al., 2001). For example, the palmitoylation of RhoB prevents RhoGDI-1 recognition and hence further influences subcellular localization (Michaelson et al., 2001). Some Rho GTPases lack conventional CAAX motifs (Chp, RhoBTB and Miro); whether their C-terminal sequences are still important for subcellular localization is not known.
This diversity in C-terminal sequences contributes to the diversity of functions seen with otherwise highly related Rho GTPases. Studies of the subcellular localization of green fluorescent protein (GFP)-tagged Rho proteins in living cells (Michaelson et al., 2001) concluded that even closely related Rho proteins could be localized to very different subcellular compartments in the cell. For example, whereas Rac1 is largely located at the plasma membrane, its close relative Rac2 (92% identity) mainly localizes to endomembranes. Distinct subcellular localization might dictate distinct upstream activation/inactivation and the downstream effector signaling. The same is true for the highly related RhoA, RhoB and RhoC proteins. A recent study developed an innovative approach to demonstrate this diversity of targeting. Zheng and colleagues showed that the GAP domain of the RhoA, RhoB and RhoC GAP p190 RhoGAP, when fused to the C-terminus of either RhoA, RhoB or RhoC, can specifically inhibit the function of each of the three GTPases, respectively (Wang et al., 2003). This, together with their association with distinct membrane compartments, emphasizes that these structurally and biochemically related Rho isoforms have distinct cellular functions. Further evidence came from the recent work of Hordijk and colleagues. They showed that peptides corresponding to the polybasic regions of RhoA, Cdc42, Rac1 and Rac2 could specifically inhibit the corresponding GTPase in cells, concluding that the these motifs bind to saturatable sites and that this is important for function (van Hennik et al., 2003).
In addition, RhoGDIs play an important role in regulating subcellular localization of Rho. For example, Del Pozo et al. have shown that RhoGDI delivers activated Rac1 and Cdc42 at new sites of integrin adhesion to allow protrusion at the leading edge of a migrating cell (Del Pozo et al., 2002). RhoG is localized to the Golgi complex by RhoGDI-3 (Brunet et al., 2002). RhoA and RhoB localize very differently in the cell (Gampel et al., 1999; Michaelson et al., 2001), potentially because of their differential GDI binding. Binding of RhoA to GDI can be regulated through phosphorylation of a C-terminal serine in RhoA by protein kinase A (Lang et al., 1996) and binding of other Rho proteins to GDIs might be regulated in similar ways (Ellerbroek et al., 2003). RhoGDI-2 is downregulated in many metastatic tumor samples, and re-expression of RhoGDI-2 in metastatic cell lines that do not express it can inhibit their metastatic and migratory behavior (Gildea et al., 2002). However, as the GTPase(s) targeted by RhoGDI-2 is not known, which RhoGTPases the downregulation affects is unclear.
Regulation by gene expression
In addition to being regulated by GEFs, GAPs and GDIs, many Rho GTPases seem to be highly regulated at the level of their expression (Fig. 1). Whereas the mRNA and protein levels of Rho GTPases such as RhoA, Rac1 and Cdc42 are relatively stable in essentially all tissues, those of the majority are not. Several Rho proteins are restricted to certain tissues, which indicates that they perform specialized functions in these cells. Rac2, for example, has a specialized role in the production of oxygen radicals in host defense mechanisms in hematopoietic cells (Roberts et al., 1999). Furthermore, several of the less studied Rho GTPases seem to be transcriptionally regulated, which suggests that their genes are only transcribed when needed (see above). For example, RhoG is a growth-stimulus-induced early response gene (Vincent et al., 1992). Since members of the Rnd subfamily do not seem to be regulated by GTP/GDP loading (see above), and instead their levels are upregulated by mitogenic stimuli (Hansen et al., 2000; Riento et al., 2003), these proteins might also be controlled at the level of transcription or translation. The importance of transcriptional regulation is further strengthened by the fact that some Rho GTPases (e.g. RhoB) are unstable and therefore have short half-lifes in the cell (Lebowitz et al., 1995).
Although 22 distinct genes encode at least 25 human Rho-family GTPases, our current knowledge of Rho GTPase function is based predominantly on what we know about RhoA, Rac1 and Cdc42. The emphasis on these three proteins is in part because of the availability of reagents to study their function and perhaps in part because of their crucial roles in cell physiology. Nevertheless, as the other Rho GTPases are further studied, some are likely to increase in stature in cell biology. Some of these clearly have functions distinct from the three `classical' Rho proteins. Others have signaling properties and effects on the cytoskeleton that seem to overlap those of RhoA, Rac1 and Cdc42. Furthermore, they might be involved in many biological phenomena in which it had been concluded that RhoA, Rac1 or Cdc42 might play a role, solely because these three were the only Rho proteins considered and the tools used, such as dominant-negative mutants or protein overexpression, were not truly specific. This situation is further complicated because many of the commercially available antibodies towards Rho proteins crossreact with several family members. Other less-studied Rho proteins will probably also be shown to play roles in processes where Rho-family members have not been implicated before. Future studies with awareness of the large number of similar Rho proteins, their cellular expression patterns and subcellular localizations, together with the use of more-specific reagents, will hopefully clarify the individual roles of each Rho protein.
We apologize for not being able to cite original work of many colleagues owing to space constraints. We thank Kent Rossman for many valuable discussions and suggestions. Our studies were supported by grants from the National Institutes of Health (CA63071, CA67771 and CA92240). K.W. was supported by a fellowship from the Swedish Foundation for International Cooperation in Research and Higher Education.