|Home | About | Journals | Submit | Contact Us | Français|
The development of multicellular organisms is associated with extensive rearrangements of tissues and cell sheets. The driving force for these rearrangements is generated mostly by the actin cytoskeleton. In order to permit the reproducible development of a specific body plan, dynamic reorganization of the actin cytoskeleton must be precisely coordinated in space and time. GTP-exchange factors that activate small GTPases of the Rho family play an important role in this process. Here we review the role of this class of cytoskeletal regulators during important developmental processes such as epithelial morphogenesis, cytokinesis, cell migration, cell polarity, neuronal growth cone extension and phagocytosis in different model systems.
Cells in a developing organism are in a continuous state of change. They grow, divide, change their shape, move or die in a magnificently coordinated effort with the objective to reproduce the species-specific body plan of the organism they belong to. The ability of cells to achieve this task relies to a large extent on the actin cytoskeleton. Actin filaments occur dispersed throughout the cell, but are highly concentrated at the cell cortex where they are organized into bundles and networks of filaments. A large number of conserved proteins that cap, sever, cross-link, bundle, nucleate or move actin filaments have been identified.1 The activity of these actin-regulators needs to be precisely coordinated in space and time to ensure normal cell behavior and tissue development.
A broad range of cytoskeletal functions are controlled by small GTPases (Guanosine Triphosphatases) of the Rho family, which act as molecular switches that regulate cell shape, division, migration, polarity, and adhesion.2 The canonical members of this GTPase family, RhoA, Rac1 and Cdc42, have been identified in the early 1990's2–6 and have more recently also been implicated in the regulation of gene expression, cell cycle progression, microtubule dynamics and cell growth.2 Rho family GTPases relay extrinsic or intrinsic signals to a wide array of effectors many of which are kept in an auto-inhibitory state in the absence of a signal.7,8 Upon activation, the GTPase binds to its cognate effector and triggers a conformational change that activates the effector. Major effector pathways discussed in this review are summarized in Box 1.
Three classes of conserved Rho family GTPase regulators have been identified (Fig. 1); (1) Rho-guanine nucleotide exchange factors (RhoGEFs) catalyze the exchange of GDP for GTP, and thereby convert the GTPase into its active state; (2) GTPase activating proteins (RhoGAPs) accelerate the slow intrinsic GTPase activity of Rho family GTPases and convert the GTPase back to it's inactive state; (3) Rho-guanine nucleotide dissociation inhibitors (RhoGDIs) prevent spontaneous activation by sequestering the inactive GDP-bound form of the GTPase in the cytoplasm.
Among these regulators, RhoGEFs play a particularly important role in regulating GTPase signaling. RhoGEFs fall into one of two conserved protein families, the Dbl-GEFs and DHR2/CZH-GEFs, which differ in the conserved domains that mediate membrane attachment and catalyze nucleotide exchange on the cognate GTPase. The mechanism of nucleotide exchange is highly conserved within each family, but differs between families. Members of each group are present in plants and early eukaryotes, revealing an ancient evolutionary origin (Box 2).
Animal genomes encode multiple RhoGEFs and many are expressed in spatially and temporally restricted patterns during development. Analysis of the Drosophila and C. elegans genomes has revealed 26 fly and 20 worm genes that fall into the Dbl family, and 4 fly and 3 worm genes that belong to the DHR2/CZH family. The fish and mammalian genomes harbor approximately 70 Dbl-GEFs and 11 DHR2/CZH-GEFs. The human genome encodes 69 Dbl-GEFs and 11 DHR2/CZH-GEFs.9,10
The number of RhoGEFs encoded in the genome is much greater than the number of GTPases they regulate and this disparity has led to the hypothesis that individual RhoGEFs may provide functional specificity by channeling GTPase signaling through one or several of a range of possible effector pathways. Thus, signaling events upstream of Rho family GTPases which involve RhoGEFs and RhoGAPs may specify signaling downstream of Rho family GTPases.11,12 It is possible that RhoGEFs and RhoGAPs cooperate to achieve a distinct level, duration or subcellularly localized activation of Rho family GTPases, which may allow stimulation of specific downstream effector pathways.13,14 Several RhoGEFs are part of multi-protein complexes that include specific GTPase effector proteins, which could provide a mechanism for selective activation of downstream effector pathways.
Here we review recent advancements in characterizing the role of RhoGEFs during animal development. We use six examples of conserved cellular behaviors crucial for animal development such as apical constriction of epithelial cells, cytokinesis, cell migration, establishment of cell polarity, axonal morphogenesis and phagocytosis to illustrate emerging concepts and current directions in the field. In each case, conserved intracellular signaling networks involving RhoGEFs have been identified which impinge on the cytoskeletal machinery that generates the physical force driving the cellular process and, ultimately, the developmental process to which the cellular behavior contributes.
Epithelial cells that line cavities, tubes and the body surface15,16 exhibit polarity that regionalizes their plasma membrane into distinct apical and basolateral domains.15,17 The apical cell membrane is organized into a domain that faces the external or lumenal environment, and a subapical belt of adherens junctions (AJs) that provides a strong mechanical link between adjacent cells. During development, epithelial sheets play important roles in the sculpturing and compartmentalization of the embryo. Sheets of epithelial cells give rise to various three-dimensional structures, including shallow grooves, deep invaginations, plate-like placodes, small pits or hollow tubes by undergoing intricate changes in cellular morphology.18
The mechanisms of epithelial morphogenesis have been extensively investigated during the invagination of mesodermal primordia in the Drosophila embryo, which begins approximately three hours after egg laying.19–21 The prospective mesoderm invaginates from the ventral side of the embryo, forming a transient groove known as the ventral furrow. The ventral furrow encompasses a band of approximately 20 cells along the anterior-posterior axis of the embryo, excluding the terminal areas. The first morphologically distinguishable event in ventral furrow formation is the apical flattening of the ventral-most cells of the presumptive mesoderm. This is followed by rapid apical constriction, which converts the cells from a cuboidal to a wedge-shaped appearance. The change in cell shape bends the epithelium thereby forcing mesodermal cells to move inside the embryo.22
The Rho1-specific Dbl-GEF DRhoGEF2 has been placed at the heart of a signaling pathway that triggers apical constriction of ventral furrow cells. Embryos lacking DRhoGEF2 do not form a ventral furrow.23,24 DRhoGEF2 is the sole fly member of the regulator of G-protein signaling (RGS) domain-containing sub-family, which includes mammalian PDZ-RhoGEF, p115-RhoGEF and LARG,25–28 zebrafish Arhgef1129 and C. elegans CeRhoGEF.30 A large body of evidence supports that this RhoGEF sub-family promotes nucleotide exchange specifically on the GTPase RhoA (Rho1 in Drosophila).25,27,28,31–35
Mesodermal fate is established prior to gastrulation by two genes, twist (twi) and snail (sna), encoding zygotic transcription factors, whose expression is restricted to a band of ventral cells that defines the mesodermal primordium.21,36,37 twi regulates sna,36–38 and also activates two other genes required for efficient gastrulation, folded gastrulation (fog)39–41 and T48,42 which encode an apically secreted ligand and a cortically localized transmembrane protein, respectively. According to a recent model,18,42 Fog and T48 act in parallel to recruit DRhoGEF2 to the apical cell cortex of mesodermal cells (Fig. 2A). T48 directly binds to the PDZ domain of DRhoGEF2,42 whereas Fog acts in an autocrine fashion to activate the Gα subunit Concertina (Cta),40,43 presumably by activating an as yet unidentified G-protein coupled receptor. By analogy to mammalian DRhoGEF2 orthologs, Cta is thought to interact with the RGS domain of DRhoGEF2 to promote its apical enrichment and activation. Studies in Drosophila cell culture have suggested that this could be mediated by Cta-induced dissociation of DRhoGEF2 from microtubule tips.44
Once localized to the apical membrane, DRhoGEF2 catalyzes nucleotide exchange on apical Rho1 to induce downstream signaling.23,24 Activated Rho1 appears to signal to two Rho1 effectors, the kinase Rok (Rho kinase)41,45,46 and the formin Diaphanous (Dia)47,48 which may act in parallel to promote assembly and subsequent contraction of an apical actin-myosin network.49 In this process, Rok has a conserved function to increase the phosphorylation state of myosin regulatory light chain (MRLC),45,46 encoded by spaghetti squash (sqh) in Drosophila.50,51 Phosphorylation of Sqh promotes assembly of myosin II into force-generating filaments, which results in increased motor activity.52,53 Phosphorylation also appears to be essential for recruitment of myosin II to the apex of cells prior to gastrulation. Accordingly, apical myosin II accumulation is lost and cells fail to constrict in rok mutants41 or in embryos treated with the Rok inhibitor Y27632,54 a phenotype mimicked by loss of DRhoGEF2.41,55 In addition, a mutated version of myosin II that lacks the actin-binding domain failed to localize to the apical side of ventral furrow cells, providing strong evidence that myosin II localization in ventral furrow cells is dependent on actin binding and/or contractility.41 Together, these data suggest that apically recruited DRhoGEF2 activates Rok through Rho1 to trigger phosphorylation of Sqh, which, in turn, is a prerequisite for apical enrichment of myosin II. In accordance with this view, overexpression of DRhoGEF2-pathway components in Drosophila Schneider-2 (S2) cells promotes cortical myosin II accumulation and cell contraction that can be inhibited by Rok inactivation.44
The role of Dia in the generation of contractile force as well as its mechanistic link to DRhoGEF2 is less clear. Based on the established role of Diaphanous-related Formins (DRFs) in nucleation and polymerization of unbranched actin filaments, it seems likely that Dia participates in the organization or stabilization of apical actin. However, recent data has revealed a role for Dia in promoting myosin II activity and AJ stability in Drosophila.48,56 Dia may cooperate with the actin-binding non-receptor tyrosine kinase Abelson (Abl),57 which regulates apical constriction of ventral furrow cells, similar to Dia.58 The role of Abl in cell constriction is conserved, as double mutants for the two mouse abl orthologs, abl and arg, exhibit disrupted neural tube closure and defects in actin organization in constricting neuroepithelial cells.59 Interestingly, in Drosophila abl mutants many ventral furrow cells fail to constrict despite the proper localization of DRhoGEF2 and myosin II.58 Thus, Abl may regulate ordered apical actin assembly at the cell cortex in parallel to DRhoGEF2.57,58 By contrast to DRhoGEF2, Abl acts specifically by downregulating the actin anti-capping protein Enabled (Ena)57,60 and embryos mutant for abl exhibit ectopic Ena as well as ectopic actin accumulations in ventral furrow cells.58
It has been difficult to explain how activation of DRhoGEF2 establishes the force-generating mechanism that drives the observed changes in cellular morphology. Recent use of live imaging techniques and detailed measurements of apical cell constriction has shed new light on this important issue. In addition to a contractile actin-myosin network, apical constriction requires assembly of AJs at the apical cell cortex.41,61 According to a simple purse-string model, myosin II-dependent sliding of actin filaments past each other induces contraction of a circumferential microfilament network, which reduces the apical cell perimeter and causes cells constrict. This creates tension in the tissue, which is transmitted from cell to cell through AJs and forces the epithelium to bend. Recent findings suggest that this model for apical constriction may need to be revised. Using live recordings of ventral furrow cells, Martin and colleagues found that cells undergo repeated constriction-stabilization cycles, which occur asynchronously between neighboring cells.62 These pulsed constrictions appear to be driven by actin-myosin contractions at the medial apical cortex and not by contraction of a circumferential actin belt. The authors suggest a “ratchet model” for apical cell constriction in which pulses of subcellular contractions of a cortical actin-myosin network pull discrete AJ sites inwards to narrow the apical area incrementally (Fig. 2B). To understand how DRhoGEF2 signaling regulates this form of dynamic contraction at the medial apical cell cortex as well as the subsequent stabilization phase should be an important subject for future investigation.
Complex cellular behaviors require spatial and temporal coordination of distinct cytoskeletal functions such as actin polymerization, depolymerization, anchoring, crosslinking and movement. These processes, in turn, require the site-specific deployment of multiple effector pathways acting downstream of Rho family GTPases. The role of RhoGEFs in spatial regulation of different effector pathways has been intensely studied in the context of cytokinesis.
Cytokinesis requires the assembly of a membrane-tethered contractile actin-myosin ring, which is involved in physical separation of the two daughter cells that complete mitosis. During cytokinesis, myosin-dependent sliding of actin filaments diminishes the diameter of the contractile ring to produce a circumferential indentation that spans the cells equator and is referred to as the cytokinetic furrow. Continued contraction of the ring leads to separation of the nascent cells as the furrow ingresses between them.63
Contractile ring assembly requires the GTPase RhoA which was initially implicated in cytokinesis because injection of the RhoA-specific inhibitor C3 transferase, stops furrowing in sea urchin and frog embryos.64,65 The first evidence of a RhoGEF regulating cytokinesis emerged from a genetic screen in Drosophila that identified pebble (pbl) as essential for cell division.66 Analysis of the pbl mutant phenotype revealed that embryonic cells lacking pbl failed to assemble a contractile ring resulting in cytokinesis defects.67,68 Subsequently, it was shown that Pbl accumulates at the cell equator during cytokinesis where it acts as a Rho1-specific GEF.69 Drosophila Pbl is the sole fly ortholog of C. elegans and vertebrate ECT2, which are similarly required for contractile ring assembly upstream of RhoA.70–73 In addition to the characteristic tandem DH-PH domain, these RhoGEFs share two N-terminal BRCT (BRCA1 carboxyl terminus) domains (see Box 2), first identified in the familial breast and ovarian cancer susceptibility gene BRCA1. BRCT domains are protein-protein interaction modules and are essential for Pbl recruitment to the cell equator during cytokinesis.74 In the central region, Pbl harbors a PEST sequence, which often serves as a protein degradation signal, and a nuclear localization signal (NLS), which is also present in the human, mouse and worm orthologs.69,71 Similar to other Dbl family GEFs, N-terminally truncated mammalian ECT2 acts as an oncogene in cultured cells.75,76 In flies, overexpression of an analogous truncated form of Pbl, but not wild type Pbl, has a dominant effect.69,77 Together, the data suggest an auto-inhibitory model according to which the GEF activity of Pbl/ECT2 is under negative control by N-terminal domains. Upstream factors may disrupt intramolecular interactions to activate the enzymatic activity of the GEF by binding to such domains, thus triggering local Rho1/RhoA-activation.
Pbl/ECT2 is part of an evolutionary conserved signaling pathway that orchestrates contractile ring dynamics by promoting site-specific activation of Rho1/RhoA at the cell equator (Fig. 3). Following its activation, Rho1/RhoA promotes contractile ring assembly through the recruitment of several protein components.78 GTP-bound Rho1 guides localized production of parallel bundles of actin filaments via its effector Dia79,80, which interacts with the Drosophila Profilin ortholog Chickadee (Chic)81,82 to drive actin polymerization.83 In Drosophila, an alternative Pbl and Rho1-dependent mechanism, which delivers vesicles carrying F-actin that has been assembled elsewhere in the cell to the furrow, has been described.84 The molecular mechanism of this vesicle-mediated delivery of F-actin is currently unclear.
Activation of Rho1/RhoA also results in the recruitment of myosin II, another component of the contractile ring, to the cell equator. Myosin II recruitment depends on the phosphorylation of its MRLC subunit. In vitro studies have implicated at least two effector kinases downstream of Rho1/RhoA in MRLC phosphorylation: Citron kinase (Sticky (Sti) in Drosophila) and ROCK (Rok in Drosophila). However, evidence from Drosophila cell culture suggests that Rok but not Sti is essential for Sqh phosphorylation and localization of myosin II to the equatorial cortex during mitosis.85 Consistent with a conserved role for Rok/ROCK in cytokinesis, human and nematode cells depleted of ROCK activity exhibit defects in actin ring contractility.86–88 In cultured Drosophila cells, Rok depletion results in complete failure to form a furrow. Sti, by contrast, appears to regulate maintenance of the contractile ring during later stages of cytokinesis.89–91 In the mouse, Citron is essential only in a subset of tissues.92
Recent studies in Drosophila have identified another Pbl-regulated Rho1-effector pathway deployed during cytokinesis. In parallel to the activation of Dia, Rok and Sti, Rho1 directs the accumulation of the scaffolding protein Anillin at the future site of cleavage to promote furrow stability. Evidence suggests that Rho1 binds Anillin, which also binds several contractile ring components such as actin, myosin II and septins.93,94 It is believed that Anilin functions to dynamically link the contractile ring to the plasma membrane and to microtubules.95 Thus, several pathways downstream of Pbl/ECT2 cooperate in the establishment, stabilization and contraction of the actin ring during cytokinesis. Precisely how Pbl/ECT2 coordinates the activity of these different effector pathways is currently unknown.
Localized activation of Rho1/RhoA by Pbl/ECT2 at the cell equator appears to define the plane of division in animal cells. However, how is the zone of Pbl/ECT2-activity determined? Studying the distribution of active RhoA in echinoderm and amphibian embryonic cells has demonstrated that the position of RhoA-activation along the equator is dictated by the position of the mitotic spindle.96
The link between microtubules and Rho1/RhoA activation may be mediated by an evolutionary conserved microtubule-associated complex, the centralspindlin complex,97 that captures Pbl/ECT2 at the cell equator.74 Centraspindlin is a two-protein complex composed of a plus-end-directed kinesin-like motor protein, known as ZEN-4 in C. elegans,98,99 Pavarotti (Pav) in Drosophila100 and MKLP-1/2 (Mitotic Kinesin-Like Protein 1/2) in mammals,101,102 and a Rho family GAP, dubbed CYK-4 in worms,97 RacGAP50C/Tumbleweed (Tum) in flies74 and MgcRacGAP in vertebrates.103 The centralspindlin complex is required for the earliest steps of animal cytokinesis. Thus, loss of either component of the complex causes a failure in cytokinetic furrow formation or ingression in numerous systems.63 It is believed that the kinesin-like motor protein component ZEN-4/Pav/MKLP-1/2 moves centralspindlin towards the plus-ends of equatorial, astral and midzone microtubules during mitosis. Once properly localized at the cell equator, the centralspindlin complex recruits Pbl/ECT2 to the future cleavage site.74 This triggers local Rho1/RhoA activation, which directs assembly and constriction of the actin ring that powers cytokinesis.
Several lines of evidence support the centralspindlin-Pbl/ECT2 model for furrow positioning. Three-dimensional reconstructions of Pbl, RacGAP50C/Tum and Pav localization during cytokinesis in Drosophila embryonic epithelial cells have revealed a double ring arrangement in which a cortical microtubule-associated ring of centralspindlin complexes is juxtaposed to a ring of Pbl.74 Ectopically localized RacGAP50C/Tum induces ectopic furrowing,104 centralspindlin decorates microtubules at the site of cleavage prior to the onset of cytokinesis,105 and Pbl localization depends on RacGAP50C/Tum and Pav.106 These data are further supported by evidence of a direct interaction between Pbl and RacGAP50C/Tum, as well as RacGAP50C/Tum and Pav.74,107–110 Somers and Saint74 have shown that RacGAP50C/Tum binds to the extended N-terminal BRCT domain of Pbl through an N-terminal coiled-coil domain. This might activate the GEF-activity of Pbl by relieving the auto-inhibitory constraint that the N-terminus imposes on the Dbl domain.
In summary, it appears that an evolutionary conserved core mechanism regulates site-specific recruitment of Pbl/ECT2, which results in activation of a set of conserved Rho1/RhoA-effector pathways acting in concert to orchestrate cytokinesis.
Drosophila Pbl also illustrates the concept that individual RhoGEFs can be deployed in more than one process, as it has been found that Pbl controls the lateral migration of mesoderm cells during gastrulation in addition to its role in cytokinesis. Interestingly, in the mesoderm Pbl is regulated by different upstream signals and exhibits altered GTPase specificity.
Cell migration requires cells to generate a polarized pushing force that is coordinated in space and time. Recent in vivo data regarding the molecular mechanism of cell migration have emerged from studies using the lateral spreading of mesodermal cells in the early Drosophila embryo as a model. Mesoderm progenitors originate on the ventral side of the early fly embryo. During gastrulation, mesodermal cells invaginate, undergo epithelial-mesenchymal transition and then migrate laterally over the inner surface of the ectoderm.111 Migrating mesodermal cells adhere to the overlaying ectoderm and extend multiple cellular protrusions in the direction of migration. The signal to migrate is provided by two Fibroblast Growth Factor 8 (FGF8)-like ligands, Thisbe (Ths) and Pyramus (Pyr),112,113 which activate the FGF receptor Heartless (Htl) in a partially redundant fashion.114–118 Interestingly, pbl mutants have a mesoderm migration defect that is reminiscent of the htl or ths, pyr double mutant phenotypes.119,120 Whereas internalization of the mesoderm occurs normally in pbl mutants, subsequent lateral migration of mesodermal cells fails due to the inability of mesoderm cells to attach to the overlaying ectoderm and to support protrusive activity. Interestingly, C. elegans ECT2, has also been implicated in cell migration during development71 and it will be interesting to see if the role of Pbl/ECT2 proteins in cell migration represents a universally conserved mechanism.
Functional dissection of Pbl has shown that distinct protein domains mediate the roles in cytokinesis and mesodermal cell migration as expression of truncated versions of Pbl partially rescues mesodermal cell migration but not the cytokinesis defects.77,119 Both the cell migration and cytokinesis functions of Pbl require GEF activity. Intriguingly, van Impel and colleagues could show that Pbl may act downstream of Htl to activate Rac1 and its paralogs77 (collectively termed Rac GTPases) instead of Rho1 and that the C-terminal tail of Pbl is important for substrate specificity. C-terminally truncated Pbl interacts with Rac GTPases but not Rho1, which is consistent with the idea that the C-terminus may impose a negative constraint on the GEF activity of Pbl towards Rac GTPases. It is possible that FGF-signaling modifies the C-terminus of Pbl and thus converts Pbl from a Rho1-specific to a Rac GTPase-specific GEF (Fig. 4).
Similar to its fly ortholog Pbl, C. elegans ECT2 is not only required for cytokinesis but plays an essential role early in development during anterior-posterior (head-tail) polarity determination in the zygote. In C. elegans, anterior-posterior polarity is established at the one-cell stage by the site of sperm entry, which triggers asymmetric distribution of polarity determinants in the embryo. The sperm-derived centrosome settles near the sperm entry site where it makes contact with the oocyte cortex and nucleates microtubules. This zone becomes the future posterior pole of the embryo. Once localized, the centrosome transfers the initial asymmetry to the actin cytoskeleton resulting in a gradient of contractile actin-myosin directed towards the anterior pole that pulls the polarity determinants PAR-3, PAR-6 and atypical protein kinase C (aPKC, PKC-3 in C. elegans) towards the prospective anterior pole of the embryo. Contraction of the actin-myosin network is essential for anterior-posterior axis establishment and for anterior PAR-3, PAR-6 and PKC-3 localization in the zygote.121,122
Prior to fertilization, the cortical actin-myosin network of the oocyte is unpolarized and proteins associated with the cortex, such as the PAR-complexes, are uniformly distributed.123 Following fertilization, this network begins to contract uniformly. Contractions are triggered by ubiquitous cortical ECT2-dependent activation of RhoA (also known as Rho-1) and myosin II.124–126 Consistent with this, ECT2 co-localizes with myosin II and depletion of ECT2 or RhoA from the zygote causes defects in contractility and polarity, in addition to failure of cytokinesis.
In addition to the paternal centrosome, the sperm contributes a pool of the Rho family GAP CYK-4 to the oocyte.126 Using an antibody against CYK-4, Jenkins and co-workers showed that the sperm-donated CYK-4 localizes around the centrosomes and to the overlaying cortex (i.e., at the future posterior end) where it regulates early contractile events that are important for the segregation of polarity determinants. Thus, initial symmetry is broken as sperm-derived CYK-4 mediates a local down-regulation of RhoA activity at the cortex overlaying the centrosomes.126 At the same time, ECT2 becomes eliminated from this region of the embryo,124 reinforcing the asymmetry in RhoA activation along the anterior-posterior axis. The localized downregulation of RhoA activity results in a local loss of contractility and in partial collapse of the actin-myosin network at the future posterior pole. Consistent with this, the earliest detectable sign of polarity establishment is the cessation of contractions in a small region of the cortex overlying the centrosomes.121 The local collapse of the actin-myosin network is believed to promote a progressive spreading of the non-contractile domain throughout the future anterior half of the embryo in which the actin-myosin network continues to contract under the influence of ECT2 and RhoA. In this way, two domains are set up in the one-cell embryo; one domain containing a contractile actin-myosin network with associated anterior determinants; the other devoid of such a network and anterior determinants. Instead, the non-contractile cortical domain accumulates factors that confer posterior identity, such as PAR-1 and PAR-2.123 In conclusion, local accumulation of the Rho family GAP CYK-4 and exclusion of the RhoGEF ECT2, results in spatial differences in actin-myosin contractility in the early embryo, which direct the asymmetric localization of polarity determining factors along the anterior-posterior axis.
Complex cellular behaviors require the deployment of multiple effector pathways acting downstream of different Rho family GTPases. The complexity of these GTPase-signaling networks that control cellular morphology has been extensively studied in the context of neurogenesis. The development of neuronal progenitors to mature neurons relies on the combined and localized activation of several effectors downstream of RhoA, Rac1 and Cdc42 at different stages of cellular maturation, and on extensive crosstalk between these GTPases.127,128
Mature neurons are characterized by one long axon that sends signals, and multiple shorter dendrites that receive signals. Shape, polarity and size of neurons are determined during development and depend on the dynamic properties of an underlying dendritic and axonal cytoskeleton. During neuronal development, the axonal cytoskeleton undergoes reorganization in response to a variety of extracellular cues that can serve as either attractive or repulsive signals by virtue of their ability to induce extension or retraction of axonal growth cones.129 Extracellular cues are received by receptors at the cell surface, and transmitted by a signaling network to Rho family GTPases, which regulate the cytoskeletal machinery throughout the maturation process from neuronal progenitor to fully functional neuron. During growth cone morphogenesis, local activation of Cdc42 and Rac1 promote the extension of filopodia and lamellipodia, respectively, which is important for cone outgrowth, while activated RhoA generates contractile forces that may drive forward translocation of the growth cone body. Growth cone retraction is believed to result from RhoA-mediated contractility concurrent with decreases in Cdc42 and Rac1 activity. Recent progress has helped to understand how RhoGEFs coordinate the GTPase signaling network that directs growth cone specification and morphogenesis in vivo.
The signaling network that underlies growth cone morphology has been extensively studied in cultured mammalian primary hippocampal neurons.127 In this context, the exchange factors Ephexin1 and PDZ-RhoGEF have been found to trigger growth cone collapse through RhoA in response to the activation of Ephrin and PlexinB1 receptors on the growth cone surface of primary cells (Fig. 5A).130,131 Ephexin1 knockout mice exhibit no apparent defects in axon pathfinding during embryonic development,132 however, adult ephexin1-/- mice have defects in structural maturation and neurotransmission at neuromuscular junctions133 and shRNA-mediated knockdown of the single chick ortholog resulted in motor axon defects in the embryo.132 Expression of Ephexin1 in fibroblasts elicits morphological changes consistent with the activation of RhoA, Rac1, and Cdc42 suggesting that Ephexin1 has a broad GTPase specificity.130 Interestingly, following Ephrin receptor activation, Ephexin1 becomes phosphorylated on Tyr87 and this enhances the exchange activity of Ephexin1 towards RhoA but not Rac1 and Cdc42,132 which is reminiscent of the switch of GTPase activity that has been observed for Pbl (see above).
Other exchange factors that have been analyzed include the closely related Rac GTPase-specific GEFs Tiam1 (T-lymphoma invasion and metastasis 1)134,135 and STEF (Sif and Tiam1-like exchange factor, also called Tiam2)136, which are orthologs of Drosophila Still life.137,138 Tiam1 and STEF are expressed at high levels in the developing murine brain139,140 and both proteins have been implicated in axonal development.141–143 Overexpression of Tiam1 in primary cells induces the extension of multiple axon-like neurites whereas treatment with an antisense oligonucleotide resulted in a dramatic decrease in the number of cells displaying an axon-like neurite.142 However, tiam1 knockout mice show no clear embryonic phenotype, even though oncogenesis is reduced.144 This suggests that other GEFs capable of activating Rac1 in neural cells, most notably STEF, but also Dock4,145 Dock7,146 β-Pix147 and Vav sub-family GEFs148,149 might act redundantly. Recent studies have provided interesting insights into the upstream signals that regulate Tiam1 and STEF during axonal specification. In brain lysates, these GEFs form a complex with the polarity proteins Par3, Par6 and aPKC, which act downstream of Cdc42. Specifically, Tiam1 and STEF bind Par3. Cell culture experiments suggest that Cdc42 may act first to select which neurite should become the future axon, and then signals to Tiam1/STEF through the Par3/Par6/aPKC complex to activate Rac1, which promotes extension of the pre-selected axon through actin remodeling (Fig. 5B).150
Another well-studied Rac1 activator in the nervous system is Trio, which was originally identified as a binding partner for the human receptor-like tyrosin phosphatase LAR151 that conveys extracellular cues to the actin cytoskeleton during development. Trio belongs to a sub-family of Dbl-GEFs that includes the mammalian GEFs Duet152 and Kalirin153 and C. elegans UNC-73 (see Box 2).154 Trio-related proteins are unique in that they possess two Dbl domains arranged in tandem of which the N-terminal GEF domain (GEF1) has been shown to induce nucleotide exchange on Rac1 or the Rac1-like GTPase MIG-2155 while the C-terminal (GEF2) domain is specific for Rho1/RhoA.156 In addition, mammalian Trio-related proteins contain a C-terminal serine-threonine kinase domain.
The in vivo roles of Trio and Kalirin have been analyzed using knockout mice. Trio is essential in the embryo and mutant animals exhibit abnormal skeletal muscle and aberrant organization in several brain regions, including the hippocampus and olfactory bulb, late in development.157 Kalirin knockout mice are viable until adulthood but show behavioral phenotypes associated with reduced cortical Rac GTPase signaling.158,159 The lack of more severe phenotypes may reflect functional redundancy between several RhoGEFs.
Further evidence for a role of Trio in neuronal development comes from analyses in C. elegans and Drosophila. C. elegans unc-73 mutants exhibit strong axon outgrowth, axon guidance and cell migration defects.154,160 A detailed genetic analysis has uncovered that unc-73 is a complex locus encoding at least eight differentially expressed mRNAs. One of these transcripts encodes a protein with tandem Dbl domains, one encodes a protein with only the GEF1 domain and the remaining six encode proteins containing only the GEF2 domain.161 Phenotypic comparisons of animals with impaired GEF1 or GEF2 domains have shown that only the GEF1 domain is essential for axon guidance, supporting that UNC-73 acts on Rac GTPases to promote axonal growth. The GEF2 domain, however, is required to modulate synaptic neurotransmission upstream of RhoA.161,162 Different isoforms have also been identified for the mammalian Trio and Kalirin orthologs many of which are neuronal specific, suggesting that the regulatory mechanism may be conserved from worms to mammals.163,164
In Drosophila, photoreceptor cells and mushroom body neurons deficient for trio display axon guidance and growth defects.138,165,166 Removing trio from Drosophila embryos results in CNS and PNS defects, including axon stalling and guidance errors.13,165,167,168 The neural defects of trio mutant flies are reminiscent of defects observed after simultaneous removal of all three Drosophila Rac-like genes, and epistasis and biochemical experiments using the Trio GEF1 domain confirm that Trio acts through Rac1, Rac2 and the Rac1-like GTPase Mtl.13,166 Moreover, Trio has been functionally linked to the Rac GTPase-effector Pak, and genetic evidence indicates that the Trio/Rac1/Pak module regulates cytoskeletal dynamics in the axonal growth cone (Fig. 6A).13,138,166
The signals that regulate Trio activity in the axonal growth cone are not well understood. Studies in mouse, fly and worm have revealed physical and functional interactions between Trio sub-family members and attractive Netrin receptors, known as Deleted in Colorectal Cancer (DCC) in mammals, UNC-40 in C. elegans and Frazzled (Fra) in Drosophila. This suggests a conserved role for Trio proteins downstream of attractive Netrin receptors to regulate Rac GTPase-mediated axon outgrowth.167,169–171 In Drosophila, further dissection of the pathway suggested that Trio may cooperate with the non-receptor tyrosin kinase Abl and its target, the actin-binding protein Ena, downstream of Fra.167,169 Binding studies have revealed that Abl physically interacts with Fra, and since Trio also binds Fra, this may allow Abl to phosphorylate Trio, which could modulate its exchange factor activity.169 In conclusion, a network of factors, including Trio and Rac GTPases, may act at the interface of cell surface receptors and cytoskeletal regulators in the axonal growth cone. Notably, this network may also include other RhoGEFs such as the Drosophila Ras/RacGEF Sos that regulates Rac1-dependent cytoskeletal reorganizations downstream of the repulsive Slit receptor Roundabout (Robo)172 and the Rho1-specific activator GEF64C that counteracts Robo dependent repulsion.173 It is important to note that Trio proteins might engage with different upstream factors and downstream regulators in a context-dependent fashion. Thus, at a later stage of neuronal maturation, Trio has been linked to the Drosophila LAR receptor and to the Rho1-effector Dia to regulate presynaptic growth (Fig. 6A).174
Studies on C. elegans UNC-73 function have provided additional information on the role of this GEF in growth cone development.175 As in Drosophila, the GEF1 domain of UNC-73 has been positioned downstream of the Netrin receptor UNC-40,170 and upstream of Rac1 and Pak during axonal outgrowth.176 Intriguingly, UNC-73 can also act upstream of the netrin receptor and the Slit receptor SAX-3 (Robo) to direct axonal growth cone migration.177,178 This is surprising since it is generally assumed that Rho family GTPases operate downstream of guidance receptors to regulate cytoskeletal reorganization. Genetic evidence suggests a mechanism whereby UNC-73 acts together with the Rac1-related GTPase MIG-2 and the kinesin-related protein VAB-8 to specify the subcellular distribution of UNC-40 and SAX-3 receptors at the cell surface (Fig. 6B). Consistent with this, yeast two-hybrid analysis has confirmed physical interactions of UNC-73 with VAB-8 and SAX-3, in addition to UNC-40.177 A recent study has identified CRML-1 (the C. elegans homologue of the mammalian actin-uncapping protein CARMIL), which binds to UNC-73 and negatively regulates its ability to recruit SAX-3 to the cell membrane.179 Thus, VAB-8 may signal through UNC-73 to positively regulate receptor levels at the cell surface, whereas CRML-1 may signal through UNC-73 to counteract this effect. A role of Rho family GTPases in the intracellular trafficking of receptor complexes has also been described180 and it will be interesting to see if Trio proteins play a similar role in other systems.
During Drosophila oogenesis a cluster of 6–10 specialized somatic cells, termed border cells, located at the anterior end of the egg chamber delaminate from the monolayered follicular epithelium and move posteriorly as a coherent group of cells until they, after travelling a distance of approximately 100 µm in six hours, reach the oocyte.
The mechanism of border cell migration and the signaling pathways involved have been a subject of intense investigation, since the behavior of border cells is reminiscent of invasive cells.181,182 Time-lapse recordings of migrating border cell clusters have revealed two distinct phases.183 In an early phase, one or two highly polarized leading edge cells exhibit extensive and dynamic actin-rich protrusions184 that can extend several cell diameters in length.185 Cell behavior during this phase is reminiscent of solitary migrating cells in culture. However, at half way through the egg chamber, cell behavior changes and the border cells now round up, extend shorter protrusions and enter into a “shuffling” mode. This second phase is characterized by position changes within the cluster as the cells collectively move posteriorly.183
Mutant analyses have revealed that both phases require signaling downstream of the receptor tyrosine kinases (RTKs), EGFR and PVR.186,187 Moreover, several RTK ligands have been identified that act redundantly as chemo-attractants during this process186–189 and the atypical Rac GTPase-specific GEF Myoblast city (Mbc) has been positioned downstream of the RTKs187 (Fig. 7A). Mbc is a member of the DHR2/CZH family of RhoGEFs, which includes the mammalian genes Dock1 (also known as Dock180190) to Dock11, C. elegans CED-5191 and Drosophila Sponge (Spg, the fly Dock4 ortholog192,193).10,194 These proteins share a conserved catalytic Docker domain (DHR2), which mediates GDP-GTP exchange on Rac GTPases or Cdc42 (see Box 2).9,194 The Drosophila mbc gene was originally identified through its requirement for myoblast fusion195 and has subsequently been implicated in many other Rac1-dependent processes.184,196–198 Cell clusters simultaneously mutant for mbc and spg exhibit an early arrest in border cell migration.184 The same is true for cell clusters that lack the conserved adaptor protein and Mbc/Spg binding partner Elmo,184,199 or for cells expressing a dominant negative form of Rac1.200
Studies of mosaic border cell clusters (i.e., clusters in which some cells are wild-type and others are mutant for a particular gene) have provided additional insights into the molecular mechanisms of border cell migration. In order to examine the precise requirement of RTK-Mbc/Elmo-Rac1 signaling, Bianco and colleagues created mosaic clusters containing individual border cells devoid of Elmo function in an otherwise wild type environment.183 Strikingly, such elmo mutant cells exhibited defects in the early but not in the late phase of migration. Thus, during the early-phase, elmo mutant cells are located at the rear of the migrating cluster from where they appeared to piggyback on their wild type neighbors. Elmo is not required during the later phase of migration, which is controlled by the Ras/Raf/MAPK pathway and phospholipase Cγ.183
Mbc-related proteins have been shown to act together with the SH2 and SH3 domain-containing adapter protein Crk201 in several contexts and the Mbc ortholog Dock180 was originally identified in a screen for novel binding partners of vertebrate Crk.190 It is believed that binding to Crk proteins may concentrate Dock180 homologs at specific sites at the cell membrane. The Dock180-Crk interaction is conserved in flies202 and both genes have been shown to regulate closure of the adult thorax during metamorphosis.198 However, a role for Drosophila Crk in border cell migration has not been reported and evidence from myoblasts suggests that Crk may not be essential for Mbc function in all cell types.203 During myoblast fusion, binding of the conserved DHR1 domain to the membrane lipid-derived second messenger phosphatidyl-inositol 3,4,5-triphosphate (PIP3) appears to be important for Mbc function.203 PIP3 is known to accumulate at the front of chemotactic cells where it polarizes the cell along the axis of migration.204 The Mbc-PIP3 interaction is conserved between flies and mammals,205 and in migrating cells in culture the Dock180-PIP3 interaction has been shown to mediate the localization of this GEF and its binding partner Elmo to the leading edge.206 It will be interesting to see if Mbc-PIP3 interactions play a role during border cell migration.
The role of the Mbc-Elmo complex in cell motility upstream of Rac1 is likely conserved in other systems. In C. elegans, the Mbc ortholog CED-5 is essential for proper migration of two specific gonadal cells, the distal tip cells, during larval development.191 CED-5 acts in a pathway together with CED-2, CED-12 and CED-10 (orthologs of Drosophila Crk, Elmo and Rac1, respectively).207–210 The worm CED-2, CED-5 and CED-12 proteins interact with each other to form a ternary complex, and a similar complex has been documented in mammalian systems. Consistent with a conserved role for this complex in cytoskeletal reorganization, co-transfection of Elmo1, Dock180 and Crk in cultured fibroblasts induces extensive membrane ruffling that can be suppressed by co-transfection of dominant negative Rac1N17.208
The CED-2/CED-12/CED-5-CED-10 signaling casette regulating migration of distal tip cells is also deployed during engulfment of apoptotic cells by macrophages (hence the acronym CED, for Cell Death Abnormal) and reveals mechanistic links between cell migration and engulfment (Fig. 7B). Both processes involve actin remodeling and polarized extension of cell surfaces regulated by Rac1. The recognition of phosphatidylserine (PtdSer) on the surface of apoptotic cells triggers signaling during cell corpse clearance.211 Early in the apoptotic process, PtdSer is translocated from the inner to the outer leaflet of the lipid bilayer of the dying cell. Exposure to PtdSer stimulates the PtdSer receptor on the phagocyte cell, which has been reported to physically interact with CED-5 (Mbc/Dock180) and CED-12 through its intracellular domain.212 Alternatively, the Rac-like GTPase MIG-2 and its GEF UNC-73 can channel signaling from surface receptors to trigger CED-2/CED-12/CED-5-CED-10 pathway activation during cell engulfment.213 The role of the CED-2/CED-12/CED-5-CED-10-pathway in cell engulfment and the function of PtdSer as a recognition signal for phagocytosis214–216 is conserved between worms and mammals.217,218 However, whether PtdSer-receptors bind the mammalian CED-5 and CED-12 homologs DOCK180 and Elmo is still an open question. Together, these studies highlight the CED-2/Crk-CED-5/DOCK180-CED-12/Elmo tripartite complex as a conserved signaling module that links extracellular cues to localized Rac1 activation, which, in turn, induces cytoskeletal rearrangements at the cell membrane to drive cell motility and phagocytosis of apoptotic cells. The function of this RhoGEF-mediated signaling cassette upstream of Rac1 to regulate both cell migration and phagocytosis illustrates the deployment of the same intracellular signaling cassette to control distinct processes during development.
This review has presented several examples of RhoGEF function during development, highlighting how RhoGEFs coordinate different GTPase effector pathways and how individual RhoGEFs control different cell behaviors at different stages during the development of an organism. However, the mechanisms that connect RhoGEFs to upstream organizers of tissue morphogenesis and lead to RhoGEF activation and recruitment to specific subcellular sites are still not well understood and should be a particular focus of future efforts.
The examples of RhoGEF function presented in this review focus on genes that are particularly accessible to genetic analysis in model organisms since their mutation leads to interpretable phenotypes. However, these genes represent only a small fraction of RhoGEFs encoded in the genome and many RhoGEFs remain uncharacterized. There is increasing evidence that many RhoGEFs may act redundantly during tissue morphogenesis and future analysis should focus on the use of innovative approaches such as RNA interference to attempt to inactivate several RhoGEFs simultaneously during specific developmental processes.
Work in our laboratory is supported by grants from the Swedish Research Council and the Swedish Cancer Society to U.H.
Previously published online: www.landesbioscience.com/journals/smallgtpases/article/12672