Rho, Rac and Cdc42, three members of the Rho family of small GTPases, each control a signal transduction pathway linking membrane receptors to the assembly and disassembly of the actin cytoskeleton and of associated integrin adhesion complexes. Rho regulates stress fibre and focal adhesion assembly, Rac regulates the formation of lamellipodia protrusions and membrane ruffles, and Cdc42 triggers filopodial extensions at the cell periphery. These observations have led to the suggestion that wherever filamentous actin is used to drive a cellular process, Rho GTPases are likely to play an important regulatory role. Rho GTPases have also been reported to control other cellular activities, such as the JNK (c-Jun N-terminal kinase) and p38 MAPK (mitogen-activated protein kinase) cascades, an NADPH oxidase enzyme complex, the transcription factors NF-κB (nuclear factor κB) and SRF (serum-response factor), and progression through G1 of the cell cycle. Thus Rho, Rac and Cdc42 can regulate the actin cytoskeleton and gene transcription to promote co-ordinated changes in cell behaviour. We have been analysing the biochemical contributions of Rho GTPases in cell movement and have found that Rac controls cell protrusion, while Cdc42 controls cell polarity.
Regulation and biochemical activities
Rho GTPases belong to the Ras superfamily of small GTPases and are highly conserved throughout eukaryotes. To date, 20 genes encoding different members of the Rho family have been identified in the human genome, and it is assumed that each one acts as a molecular switch to control distinct biochemical pathways , although only Rho (three isoforms, RhoA, RhoB and RhoC), Rac (three isoforms, Rac1, Rac2 and Rac3) and Cdc42 have been studied in any great detail. Like all regulatory GTPases, these proteins exist in an inactive GDP-bound conformation and an active GTP-bound conformation (Figure 1). Their interconversion is tightly controlled by a large family (85 in mammals) of GEFs (guanine nucleotide-exchange factors) that increase GDP/GTP exchange rates. Many GEFs were originally identified as oncogenes after transfection of immortalized fibroblast cell lines with cDNA expression libraries and they all share a common feature: a DH (Dbl homology) domain adjacent to a PH (pleckstrin homology) domain. However, more recently, another, smaller, family of Rho GEFs has been identified related to the founding member DOCK180 (dedicator of cytokinesis 180) [2–4]. It is still very unclear how GEFs are themselves regulated, but it is quite likely that this will be central to understanding the mechanisms that underlie the spatial, as well as the temporal, activation of GTPases within a cell in response to outside influences. An equally large family of GAPs (GTPase-activating proteins) that stimulate the intrinsic GTPase activity of Rho GTPases has been identified . Although they are likely to down-regulate GTPase signalling, even less is known about how they are recruited and activated.
Using Swiss 3T3 cells as a model system, we showed in the early 1990s that Rho regulates the formation of contractile actin–myosin filaments in response to a variety of extracellular stimuli (e.g. lysophosphatidic acid or integrin engagement), Rac induces actin polymerization at the cell periphery to produce lamellipodia in response to PDGF (platelet-derived growth factor) or insulin, and Cdc42 promotes actin filament assembly and filopodia formation at the cell periphery in response to bradykinin and IL-1 (interleukin 1) [6–9]. Over the intervening years, however, many more activities of Rho GTPases have been demonstrated (Figure 2). One of the earliest to be characterized was the regulation of the activity of an NADPH enzyme complex by Rac in phagocytic cells, to produce reactive oxygen species . This is an important host response that accompanies actin-driven phagocytosis of bacterial pathogens, although Rac-mediated activation of the purified enzyme can be reconstituted in vitro in the complete absence of actin. From this and other observations, the idea has emerged that Rho GTPases co-ordinately regulate multiple biochemical pathways so as to promote complex changes in cell behaviour .
An indication that Rho GTPases can promote gene transcriptional changes came in 1995, when two groups showed that Rac and Cdc42 activate the JNK (c-Jun N-terminal kinase) and p38 MAPK (mitogen-activated protein kinase) cascades, and another group found that Rho stimulates the transcription factor, SRF (serum-response factor) [12–14]. The activation of MAPK pathways is independent of changes to the actin cytoskeleton; however, the activation of SRF appears to be linked directly to the levels of unpolymerized actin . Finally, Cdc42 was first identified as a cell cycle division mutant in Saccharomyces cerevisiae and was found to play a key role in polarized bud formation and in the morphological changes that accompanied pheromone signalling. It is now clear that this GTPase plays a widespread role in the establishment of cell polarity in all animal cells .
In an attempt to define the biochemical pathways activated by Rho GTPases, many groups have used yeast two-hybrid selection and affinity chromatography techniques to identify cellular targets of Rho, Rac and Cdc42. More than 50 potential targets have been identified to date, and a current major task is to define their individual roles . Much progress has been achieved in defining the pathways that link Rho, Rac and Cdc42 to the actin cytoskeleton (Figure 3). A major breakthrough here came with the observation that one of the downstream targets of Cdc42, WASP (Wiskott–Aldrich syndrome protein), interacts directly with the Arp2/3 (actin-related protein 2/3) complex, one of the two machines that promotes actin polymerization in all eukaryotic cells [18,19]. Later, Rac was shown to interact (indirectly) with a family of WASP-related proteins, WAVE (WASP family verprolin homology), which can also activate Arp2/3 [20,21]. Despite both Rac and Cdc42 activating Arp2/3, it is still not entirely clear why they lead to such morphologically distinct structures, i.e. lamellipodia and filopodia respectively, but in vitro reconstitution assays are helping to unravel these processes [22,23]. The ability of Rho to induce the assembly of actin–myosin filaments appears to depend on activation of two distinct targets, p160Rho kinase and mDia. p160Rho kinase has many substrates, but a key one is myosin light-chain phosphatase . Phosphorylation leads to inactivation of the phosphatase, which leads to an increase in the levels of phosphorylated myosin light chain and cross-linking of myosin II into actin filaments. MDia, on the other hand, belongs to the formin family of proteins, which constitute the other major mechanism for promoting actin polymerization in eukaryotic cells. Upon binding Rho GTP, mDia adopts an open conformation and binds to the barbed ends of actin filaments. It has the unusual property of allowing addition of actin monomers to filament ends, while preventing the binding of capping proteins, which would otherwise block elongation.
Changes to the actin cytoskeleton are a key feature of many important aspects of cell biology: shape and morphology, polarity, migration, cytokinesis and phagocytosis, to name a few. There is now overwhelming evidence that Rho GTPases play important regulatory roles in each of these situations, and we have focused much of our effort over the last 5 years in looking at cell migration. Initially, this was an attempt to understand how actin is organized in a migrating cell, since it has long been recognized that actin filament elongation and actin–myosin filament contraction provide the major driving forces for cellular movement. However, even using simple in vitro migration assays, it is clear that many other aspects of cell biology contribute to the migratory phenotype, such as microtubule organization and the secretory pathway. Finally, cells in vivo almost always move in a particular direction (i.e. not random) and in response to extracellular cues, often referred to as directional sensing. This important aspect of migration is one that we have focused on over recent years in our work.
In vitro, scratch-induced cell migration assay
To investigate the role of small GTPases in motility, we have used a simple in vitro migration assay in which small scratch-induced wounds are made in a confluent monolayer of primary cells (Figure 4). Cells can be transfected before making the monolayer, or microinjected after making the scratch, and the effects of various expression constructs are observed at the single-cell level by microscopy. In addition, by making multiple scratches (to the extent that around half of the cells are detached from the dish), attached cells can be recovered and a biochemical analysis can be performed (Figure 4). We have used both REFs (rat embryo fibroblasts) and rat astrocytes in this type of assay.
During a period of 5–6 h after scratching the monolayer, cells move forwards, co-ordinately as a sheet, to fill the gap without the necessity for cell division or DNA/protein synthesis. Within 30 min of wounding, rounded and retracted cells at the wound edge re-spread and extend filopodia and lamellipodia into the open space. After a further 90 min, cells in the front row have a polarized morphology, and lamellipodia are localized only at the front and not at the sides or rear edge of the cells, where they are in contact with their neighbours. Rhodamine–phalloidin staining reveals that cells in the monolayer contain abundant stress fibres, and, in addition, those at the front have actin-rich lamellipodial and filopodial extensions. Anti-phosphotyrosine antibodies reveal elongated focal adhesions in all cells, and, in addition, cells at the leading edge have small focal complexes. After approx. 5–6 h, opposing cells begin to make contact with each other, and time-lapse video microscopy reveals that ruffling stops within minutes of cell–cell contact.
To determine whether Rac is involved in wound-induced cell motility, groups of cells on opposite sides of a wound were injected with dominant-negative Rac (N17Rac) protein or a dominant-negative Rac DNA expression vector. Rhodamine–phalloidin staining revealed that inhibition of Rac blocks all lamellipodial activity and cell movement. Control cells injected with fluorescent dextran moved similarly to uninjected cells, and wounds closed within 5–6 h. We conclude that Rac is probably activated upon wounding and that Rac activity is essential for cell movement .
To determine the role of Cdc42 in cell migration, patches of cells on opposite sides of a wound were injected with dominant-negative Cdc42 (N17Cdc42) protein, or a dominant-negative Cdc42 DNA expression vector, and, in both cases, a small (50%) inhibition of cell movement was observed. The reason for this partial inhibition was examined more carefully. Migrating cells normally show a typical polarized morphology with membrane protrusions restricted to the front edge. Inhibition of Cdc42 destroys this polarity and protrusive lamellipodial activity is seen all around the cell periphery .
To characterize the signal transduction pathway through which Cdc42 promotes polarized actin polymerization, we looked at Cdc42-specific targets. The p65PAK family [PAKs (p21-activated kinases) 1, 2 and 3] of serine/threonine kinases has been linked to the formation of cell protrusions by others [25a], and so we first examined the effect of microinjecting an expression vector containing the AID (autoinhibitory domain) derived from PAK1. This domain interacts directly with the catalytic site of the three PAK isoforms and blocks kinase activity. When expressed in cells at the leading edge of a wounded REF monolayer, we found that actin-dependent protrusive activity was no longer restricted to the front of the cells, but became delocalized all around the periphery . We conclude that Cdc42 activates PAK at the leading edge of migrating cells and that its kinase activity is needed to restrict Rac activity to the front. Note, in these cells, Cdc42 and PAK are not needed to activate Rac, just to localize Rac activity. We are currently trying to identify the cellular substrate of PAK that is responsible for this activity.
The polarized morphology of migrating cells is also reflected in the reorganization of microtubules and the realignment of the Golgi and centrosome to face the direction of movement. To determine whether Cdc42 activity is required for this aspect of cell polarity, the extent of Golgi/centrosome reorientation was determined after injection of cells with either dominant-negative Cdc42 or with the Cdc42-binding domain derived from the protein, WASP. In both cases, inhibition of Cdc42 prevented Golgi/centrosome reorientation . Unlike the Cdc42-dependent polarization of actin protrusions, however, Cdc42-dependent polarization of the microtubules and centrosome are not dependent on PAK activity. To obtain further insights into how Cdc42 regulates microtubule polarity, we turned to primary astrocytes.
Astrocytes share many common features with fibroblasts with respect to Rho GTPases and migration (e.g. polarity in both cell types is controlled by Cdc42), but there are also important differences that have allowed us to make significant progress, particularly in uncovering how Cdc42 controls microtubule polarity. During migration, astrocytes undergo a significant morphological change and the cells elongate dramatically. In addition, they become highly polarized and this can be readily seen with a centrosome marker, which reorients to face the direction of migration. We have found that inhibition of Rac blocks both cell migration and cell elongation, yet the centrosome still relocates to face the front of the cell. This suggests that polarity is controlled by a Cdc42-dependent signal transduction pathway that can be uncoupled from physical movement or shape changes.
A significant boost to the analysis of this pathway came in 2000 with three reports that mPar6, the mammalian orthologue of Caenorhabditis elegans PAR-6, is a direct target of Cdc42 [27–29]. PAR-6 was first isolated as a gene that is essential for asymmetric cell division in the C. elegans zygote, raising the possibility that it might play a more general role in microtubule polarity in mammalian cells. Furthermore, during asymmetric cell division at least, PAR-6 is associated with an aPKC (atypical protein kinase C), which is also required for asymmetric distribution of spindle microtubules. Based on these ideas, and using microinjection experiments and small-molecule inhibitors in an astrocyte scratch-induced migration assay, we have uncovered a signal transduction pathway linking integrins, activated at the front of the cell, to the activation of Cdc42 and the mPar6–PKCζ complex and leading to the subsequent reorientation of the centrosome .
Genetic analyses in C. elegans have not revealed what the target of aPKC activity is during cell division, and so we used a screen of small-molecule inhibitors for a clue. In fact, we found that microtubule polarity and centrosome reorientation was inhibited by lithium, an inhibitor of, among other things, GSK3 (glycogen synthase kinase 3). Two other commercial inhibitors of GSK3 also blocked centrosome polarization, as did expression of a kinase-dead or a non-inhibitable (S9A) version of GSK3. Interestingly, in each case, cell elongation was not inhibited, but, just as with PKCζ inhibition, cell protrusions were no longer perpendicular to the leading edge of the wounded monolayer. GSK3 is believed to be constitutively active in cells, but, in response to upstream signals, it can be inhibited. The fact that both kinase-dead and non-inhibitable (i.e. constitutively active) GSK3 block polarity would suggest that localized inactivation of GSK3 (at the leading edge) is required for polarity establishment .
GSK3 is best known for its role in insulin and in Wnt signalling. We already knew that phosphoinositide 3-kinase and Akt, which participate in the insulin pathway, are not involved in microtubule polarity, and so we took a look at components of the Wnt pathway. One of these, APC (adenomatous polyposis coli), is particularly interesting; not only does it play a major role as a tumour suppressor in human cancer, but also, through its control of β-catenin stability, it is associated with the dynamic plus ends of microtubules. We went on to show that, during migration, APC strongly accumulates at the plus end of microtubules, but only at the leading edge of front-row cells. Furthermore, interfering with the association of APC and microtubules prevented centrosome reorientation. Our current model is that APC at microtubule plus ends associates or is captured by proteins in the plasma membrane at the front of the cell (Figure 5). This allows tension to be exerted through microtubules on the centrosome . We are currently trying to identify the nature of the membrane complex (X in Figure 5) that captures the microtubules. We are also very intrigued that the pathway activated by Cdc42–Par6–aPKC resembles very much the Wnt signal pathway. We are interested to know whether Wnt and Cdc42 can activate the same complex, but in different biological contexts.
Our analysis of cell migration, using the in vitro scratch assay, has revealed the importance of Cdc42 in controlling the polarity of both the actin and microtubule cytoskeletons. Furthermore, Cdc42 mediates these two polarity effects through distinct and separable signal transduction pathways. Most interestingly, however, Cdc42-dependent polarization of microtubules is mediated through the Par6–aPKC complex, which is required for the establishment of polarity in many other cellular contexts, including asymmetric cell division and both epithelial and neuronal morphogenesis. We believe this simple in vitro assay will allow us to explore the biochemical details that underlie these complex polarity pathways.
Abbreviations: APC, adenomatous polyposis coli; Arp2/3, actin-related protein 2/3; GEF, guanine nucleotide-exchange factor; GSK3, glycogen synthase kinase 3; MAPK, mitogen-activated protein kinase; PAK, p21-activated kinase; (a)PKC, (atypical) protein kinase C; REF, rat embryo fibroblast; SRF, serum-response factor; WASP, Wiskott–Aldrich syndrome protein
- © 2005 The Biochemical Society