Rho GTPases*

The mammalian Rho GTPase family currently consists of seven distinct proteins: Rho (A, B, and C isoforms), Rac (1 and 2 isoforms), Cdc42 (Cdc42Hs and G25K isoforms), RhoD, RhoG, RhoE, and TC10. Like other members of the Ras superfamily, Rho proteins act as molecular switches to control cellular processes by cycling between active, GTP-bound and inactive, GDPbound states (Fig. 1). Activation of the GTPase, through GDPGTP exchange, is promoted by guanine nucleotide exchange factors (GEFs), whereas inactivation (by an intrinsic GTPase activity) is stimulated by GTPase-activating proteins (GAPs). Rho guanine nucleotide dissociation inhibitors (Rho-GDIs) appear to stabilize the inactive, GDP-bound form of the protein. Activated Rho GTPases interact with cellular target proteins or effectors to trigger a wide variety of cellular responses, including the reorganization of the actin cytoskeleton and changes in gene transcription. This review aims to summarize current views of Rho GTPase signaling pathways; for reasons of space, however, this will be limited primarily to mammalian Rho, Rac, and Cdc42. More detailed information can be found elsewhere (1, 2).


Rho Proteins and the Actin Cytoskeleton
Rho proteins control the organization of the actin cytoskeleton in all eukaryotic cells. Activation of Rho in fibroblasts has been shown to cause the bundling of actin filaments into stress fibers and the clustering of integrins and associated proteins into focal adhesion complexes; activation of Rac promotes de novo actin polymerization at the cell periphery to form lamellipodial extensions and membrane ruffles, and activation of Cdc42 triggers actin polymerization to form filopodia or microspikes ( Fig. 2) (3)(4)(5)(6)(7). Actin filaments found in lamellipodia and filopodia are, like stress fibers, associated with integrin adhesion complexes, although the function of these complexes is not clear (5,7). Cross-talk between Rho proteins has been observed; in particular, Cdc42 is a strong activator of Rac, such that filopodial extensions are usually seen associated with lamellipodial protrusions (5,6).
The biological implications of these findings are wide ranging, and Rho GTPases are likely to play a regulatory role wherever filamentous actin is used to drive a cellular process. Already Rho, Rac, and Cdc42 have been implicated in cell movement, axonal guidance, cytokinesis, and morphogenetic processes involving changes in cell shape and polarity (1,2).

Rho Proteins and Gene Transcription
The JNK/stress-activated protein kinase (SAPK) and p38 MAP kinase cascades are known to control gene transcription in response to cellular stresses such as UV light and osmotic shock or challenge with inflammatory stimuli (8). Numerous groups have now reported that these two MAP kinase pathways can be activated by Rac and Cdc42, suggesting an analogous role to that played by Ras in the ERK MAP kinase cascade (Fig. 3) (9,10). The exact role of the Rho GTPases in MAP kinase activation is, however, far from clear; overexpression of constitutively activated Rac or Cdc42 leads to only modest activation of JNK reporter plasmids in cotransfection assays, and there are still very few examples where physiological activation of the JNK pathway has been shown to be dependent on endogenous Rac or Cdc42 activity. The genetic analysis of dorsal closure in Drosophila has, however, recently provided experimental support for a role for Rac in JNK regulation (11). Interestingly, in the Saccharomyces cerevisiae pheromone response pathway, where both Cdc42 and a JNK-like MAP kinase cascade are required, the GTPase is not required for activation of the kinase cascade per se but instead appears to be required for the correct cellular localization of the MAP kinase-containing signaling complex (12).
Rho, Rac, and Cdc42 have each been reported to activate serum response factor (SRF)-dependent transcription (via an as yet unidentified mechanism) and to activate the transcription factor, NFB (13,14) (Fig. 3). In the latter study, the authors suggested that the generation of reactive oxygen species by Rac might be the trigger for NFB activation, an interesting suggestion because it is known that Rac regulates an NADPH oxidase enzyme complex in phagocytes to produce superoxide (15). Rho, Rac, and Cdc42 activities are required during G 1 cell cycle progression, but whether this is because of their effects on the actin cytoskeleton and integrin adhesion complexes or whether it is because of more direct effects on gene transcription is not known (16 -20). It was shown in 1992 that oncogenic Ras is a potent activator of Rac, and mutational analysis of Ras has revealed that Rac activation is an essential downstream signal required for Ras-induced malignant transformation (4,21). Again it is not clear which Rac-regulated activity contributes to the transformed phenotype, but this has important implications in cancer biology (22).

Regulators of Rho Proteins
When considering biological responses such as directed cell migration or axonal guidance, it is clear that the activation of GTPases must be both temporally and spatially controlled; in this respect it seems likely that GEFs will play a major role. A surprisingly large family (Ͼ20) of Rho GEFs has been identified, each of which shares two common motifs, the Dbl homology domain, which in some cases at least has been shown to encode the catalytic nucleotide exchange activity, and a pleckstrin homology domain, whose function is unclear but might determine subcellular localization (1,23). Experimentally some GEFs appear to be specific for an individual GTPase, e.g. Lbc for Rho, Tiam1 for Rac, and FGD1 for Cdc42, whereas others seem to act on all three GTPases, e.g. Vav and Dbl (24 -26). It is currently unclear why the Rho GEF family is so large; only two mammalian exchange factors have been identified for Ras to date, and the major one of these, Sos, has even been shown to encode Rac GEF activity at its N terminus (27). There is also little data on the mechanism of activation of Rho GEFs by extracellular agonists. In fibroblasts, it appears that the ␣ subunits of the heterotrimeric G proteins, G 12 and G 13 , along with an unknown tyrosine kinase, are required to link the lysophosphatidic acid (LPA) receptor to Rho activation, whereas phosphatidylinositol 3-kinase is required to link the platelet-derived growth factor (PDGF) receptor to Rac activation (28,29). Whether GEFs are the proximal targets of these activities remains to be established.
The dissociation of the GTPase from a Rho-GDI complex ( Fig. 1) is likely to be an additional key feature of the activation mechanism. GEFs added to a GTPase-GDI complex in vitro are unable to stimulate nucleotide exchange, and so a dissociation signal would appear to be required. It could even be that this is the rate-limiting step for GTPase activation in vivo. A recent report suggests that members of the ezrin/radixin/moesin (ERM) family of proteins can induce dissociation of GDI from the complex (30). The role of GDI still remains somewhat of a mystery; Cdc42, RhoB, and, in some cells at least, Rac have been found predominantly bound to membranes and not cytosolic as would be predicted for a GDI complex (31,32). Furthermore, two groups have purified a Rac-GDI complex, which is fully functional in an in vitro NADPH oxidase activation assay and in a regulated secretion assay in permeabilized mast cells (33,34). This raises the possibility that GDI might reassociate with GTPases once they have been activated.

Effectors of Rho Proteins
The yeast two-hybrid selection system and biochemical affinity purification methods have been used extensively to identify Rho, Rac, and Cdc42 interacting proteins, and a large number of candidate effector proteins have been identified (1).
Rho-Although around eight targets for Rho have been identified, one of these, p160Rho kinase, has received much of the attention to date. The activity of this Ser/Thr kinase, which belongs to the myotonic dystrophy family of kinases, is enhanced (though not by much) after binding to Rho-GTP, and when overexpressed in cells, it has been reported to induce stress fibers independently of Rho (35)(36)(37). Furthermore, two substrates for the kinase have been identified, the myosin binding subunit of myosin light chain phosphatase and myosin light chain itself (1,38,39). Phosphorylation of these two substrates would be expected to lead to an increase in myosin light chain phosphorylation, myosin filament assembly, and F-actin bundling, thereby leading to stress fibers. It seems unlikely, however, that p160 Rho kinase will turn out to be the only Rho target required for the organized cytoskeletal changes induced by Rho; we have observed, for example, that a Rho mutant (with a single effector site amino acid substitution) no longer interacts directly with p160Rho kinase, yet this version of Rho induces cytoskeletal changes indistinguishable from wild type Rho. 2 The biological function of stress fibers is also 2 D. Drechsel and A. Hall, unpublished data.
FIG. 1. The Rho GTPase cycle. Rho GTPases are thought to exist in an inactive GDP-bound state complexed to one of the three known Rho-GDIs. In response to an incoming signal (see Fig. 2), Rho-GDI dissociates, and a GEF catalyzes nucleotide exchange and activation. It is thought that this is likely to occur at or close to the plasma membrane, because it is unlikely that isoprenylated GTPases will exist free in the cytosol. In its active state, the GTPase interacts with cellular targets or effectors (E1-x) to generate a cellular response. Currently around eight potential targets are known for each of the three GTPases, Rho, Rac, and Cdc42, but there is no reason to believe that this list is complete. Most of the targets are ubiquitously expressed, and it is unknown how a GTPase "chooses" its target under a particular set of circumstances. Finally, one of around 10 known Rho GAPs will interact with the GTPase-target complex and catalyze GTP hydrolysis (1). The GDP-bound form of the GTPase is then "extracted" from the membrane by Rho-GDI, and the cycle is complete. The exact role of Rho-GDI is unclear, and the rate-determining step in the cycle has not been determined. What is clear, however, is that Rho GTPases can, in their GTP-bound state, interact with numerous target proteins to induce coordinated signals.

FIG. 2. Rho, Rac, and Cdc42 and the actin cytoskeleton.
The activation of Rho, Rac, and Cdc42 by extracellular agonists has been best characterized in quiescent Swiss 3T3 fibroblasts (3)(4)(5)(6). In these cells, LPA (a major constituent of tissue culture serum) acting via a plasma membrane receptor activates Rho leading to the assembly of actin-myosin stress fibers and associated integrin adhesion complexes (focal adhesions). Rac can be activated by PDGF or by insulin, and this leads to actin polymerization at the cell periphery causing lamellipodial extensions and membrane ruffling activity. Lamellipodia are also associated with integrin adhesion complexes, although their function is not clear; they contain many of the same constituents as classical focal adhesions, but they are much smaller. They are not required for actin polymerization, but they may be required for cell movement (7). Bradykinin will activate Cdc42 in these cells to produce filopodia or microspikes and associated integrin complexes. There is much crosstalk within and between the Ras and Rho GTPase families. Activation of Cdc42 leads to a localized activation of Rac, and hence filopodia are intimately associated with lamellipodia; in fact in these cells it is difficult to see filopodia unless Rac activity is first inhibited. Quiescent Swiss cells have no detectable stress fibers, and activation of Rac under these conditions leads to a weak and delayed activation of Rho producing a few weak stress fibers. However, activation of Rac in cells growing in serum, which already have very pronounced and abundant stress fibers, leads to a loss of stress fibers. Finally, in addition to its well characterized effects on the ERK MAP kinase pathway, oncogenic Ras is a potent activator of Rac. not clear; it has even been suggested that they may represent an exaggerated arrangement of actin found only in cells growing in culture. However, the importance of actin-myosin filament assemblies is not in doubt, and in fact a recent paper has suggested that p160Rho kinase, acting on actin-myosin filaments, may play a key role in hypertension (40).
There has been one report that Rho can regulate the synthesis of phosphatidylinositol 4,5-bisphosphate (PIP 2 ), a lipid that has long been thought to have profound effects on the actin cytoskeleton (41). More recent work has suggested a role for PIP 2 in the assembly of focal adhesion complexes and in the activation of ERM proteins (42,43). Because functional ERM proteins are known to be essential for Rho (and Rac)-mediated actin changes, it is possible that a part of the function of Rho is to maintain PIP 2 levels (44).
The latest Rho target to be identified, p140mDia, has subsequently turned out to be the locus for a genetically inherited form of human deafness (45,46). p140mDia is related to the Drosophila protein diaphanous, which is involved in cytokinesis and in cell polarity establishment. These proteins can interact with the actin monomer binding protein, profilin, and they may, therefore, play a part in linking Rho to the actin cytoskeleton.
Rac and Cdc42-The two GTPases each have around eight known target proteins, although some are common to both. For example, the first Rac target to be identified, p65PAK, also interacts with Cdc42. The homology of p65PAK to an upstream regulator of the pheromone MAP kinase pathway in yeast (Ste20p) suggested that it might link Rac and Cdc42 to JNK activation in mammalian cells, and there is some experimental support for this (47)(48)(49)(50). Others, however, have failed to find a role for this kinase in JNK activation (51,52), and one group has even reported that p65PAK overexpression can (independently of its kinase activity) induce lamellipodia formation (53). A Rac target that is likely to play a central role in actin polymerization is phosphatidylinositol 4-phosphate 5-kinase.
As already mentioned, the product of this enzymatic activity, PIP 2 , is known to affect actin filament assembly, and it has been reported that in permeabilized platelets, both Rac and PIP 2 are essential for the release of capping proteins from the barbed ends of actin filaments, a prerequisite for actin polymerization in response to thrombin (54).
Unlike Rho and Rac, there have been no reports so far that Cdc42 affects phosphatidylinositol 4-phosphate 5-kinase activity, and in fact, activated Cdc42 can stimulate actin polymerization in a cell-free system independently of PIP 2 (55). In vitro assays such as this will undoubtedly have a major impact on the field in the near future. One Cdc42 target that has been reported to be essential for filopodia formation is N-WASP, a relative of the human Wiskott-Aldrich syndrome protein, WASP (56 -58). It appears that overexpression of N-WASP can potentiate, though not bypass, the ability of Cdc42 to induce filopodia (58). Another Cdc42 target, IQGAP, has been shown to interact directly with F-actin; it is not known if it plays any role in filopodia formation, but there is some data linking it to the assembly of actin filaments during cytokinesis (59).
It is clear then, that we are still some way off a biochemical description of the pathways linking Rac and Cdc42 to the JNK cascade and to actin polymerization. Amino acid substitutions in Rac have been identified which block JNK activation, but which do not block actin changes and vice versa (18 -20). This shows that these pathways can be biochemically separated and that Rac must interact with at least two distinct cellular targets to trigger the two responses. Despite these observations implying bifurcating pathways, the analysis of Cdc42 in yeast suggests that the two are likely to be coordinately controlled through the formation of a multimolecular signaling complex. So far there is no evidence for such a complex in mammalian cells, but this is likely to be a focus of attention in the future.

Conclusion
Rho GTPases play a central role in all eukaryotic cells, coordinately controlling the organization of the actin cytoskeleton with other cellular activities such as gene transcription, cell cycle progression, and adhesion. Elucidation of the underlying biochemical pathways will have a significant impact on understanding normal cell behavior as well as human diseases such as cancer and inflammation.

FIG. 3. Ras and Rho GTPases and gene transcription.
Three MAP kinase cascades have been characterized in mammalian cells. Ras is known to regulate the ERK pathway by interacting directly with the c-Raf MAP kinase kinase kinase (MKKK). In transfection assays, Rac and Cdc42 can activate the JNK and p38 MAP kinase pathways, but the mechanism is unclear. At least eight distinct MAP kinase kinase kinases capable of activating the JNK/p38 pathways have been described to date, including three (MLK, MEKK1, and MEKK4) that can interact directly with Cdc42 and Rac. It appears that the regulation of the JNK/p38 pathways is much more complex than that of the ERK cascade. In addition to regulating the activity of transcription factors via MAP kinase pathways, it has been reported that Rho, Rac, and Cdc42 can activate the SRF and NFB transcription factors.