Regulation of small GTPases at epithelial cell-cell junctions

Abstract

Small GTPases of the Rho family (RhoA, Rac1, and Cdc42) and the Ras family GTPase Rap1 are essential for the assembly and function of epithelial cell-cell junctions. Through their downstream effectors, small GTPases modulate junction formation and stability, primarily by orchestrating the polymerization and contractility of the actomyosin cytoskeleton. The major upstream regulators of small GTPases are guanine nucleotide exchange factors (GEFs) and GTPase activating proteins (GAPs). Several GEFs and a few GAPs have been localized at epithelial junctions, and bind to specific junctional proteins. Thus, junctional proteins can regulate small GTPases at junctions, through their interactions with GEFs and GAPs. Here we review the current knowledge about the mechanisms of regulation of small GTPases by junctional proteins. Understanding these mechanisms will help to clarify at the molecular level how small GTPases control the morphogenesis and physiology of epithelial tissues, and how they are disregulated in disease.

Keywords::

• Adherens junction
• tight junctions
• cytoskeleton
• epithelial cell

Introduction

Small GTPases are molecular switches that control a broad range of processes in eukaryotic cells. Rho family GTPases, a subset of small GTPases, comprise over 20 isoforms in mammalian organisms, whereas only 5, 10 and 11 isoforms have been described in yeast, Caenorhabditis elegans and Drosophila melanogaster, respectively (Jaffe and Hall 2005). This suggests that the evolution of vertebrate species correlates with an increasing complexity of small GTPases. Among the many targets of Rho family GTPases, cell-cell junctions are of critical importance for the morphogenesis of epithelial tissues. Rho family GTPases modulate cell-cell junctions mainly by controlling the assembly and contractility of the actin cytoskeleton, an essential structural scaffold of junctions. Specific proteins recruit the regulators of these GTPases (GEFs and GAPs) to junctions, thereby modulating their function at spatially defined sites. The dynamic interplay between these molecules ultimately orchestrates cell shape, cell motility and tissue morphogenesis.

In this review we focus on the molecular mechanisms that control the activity of Rho family GTPases (RhoA, Rac1, and Cdc42), and the Ras family GTPase Rap1 at vertebrate epithelial cell-cell junctions. This subject is introduced by a brief discussion of junctions, small GTPases, and their regulation. The reader is referred to excellent review articles, for a more comprehensive analysis of these areas.

The epithelial junctional complex

Epithelial cells form single- and multi-layered sheets that cover all body surfaces and cavities, and constitute the parenchyma of glandular organs. A key feature of epithelial cells is the presence, at their apicolateral surfaces, of a specialized junctional complex, which comprises tight junctions (TJ), zonula adherens (also denoted as adherens junctions [AJ]) and desmosomes (Farquhar and Palade 1963). Since most published work identifies TJ and AJ as targets and regulators of Rho family GTPases, we will focus here on TJ and AJ, and refer the reader to recent reviews about desmosomes (Garrod and Chidgey 2008, Green et al. 2010) for additional information in this area.

The functions of TJ and AJ

TJ and AJ are of fundamental importance in the development and physiology of epithelial cells and tissues, since they allow cells to adhere to each other, to form barriers between body compartments, and to receive and transmit extracellular and intracellular signals. TJ form a selective paracellular permeability barrier to the movement of water, ions, and other solutes (Van Itallie and Anderson 2006), and act as anatomical fences separating the apical from the basolateral domains of the plasma membrane (Diamond 1977). AJ are responsible for cell-cell adhesion, and play a key role in tissue sorting during development, through the expression of specific cadherin isoforms (Wheelock and Johnson 2003, Gumbiner 2005, Stepniak et al. 2009). Moreover, TJ and AJ are focal sites of anchoring and organization of the cytoskeleton (Mege et al. 2006, Meng and Takeichi 2009), and function as signaling hubs in the control of gene expression and cell proliferation, through the interaction of several of their components with transcription factors and other signaling molecules (Heuberger and Birchmeier 2010, McCrea et al. 2009).

The molecular structure of TJ and AJ

The architectural organization of TJ and AJ comprises transmembrane proteins that function either as adhesion receptors or as paracellular channels, and cytoplasmic proteins, that associate with the intracellular domains of transmembrane proteins, and link them to the cytoskeleton and to adaptor and signaling proteins. The molecular constituents of the membrane domain of TJ comprise four-pass membrane proteins and Ig-like adhesion molecules. The four-pass membrane proteins of TJ include occludin, tricellulin and claudins, the latter playing a pivotal role in defining the ionic paracellular permeability properties of TJ (Tsukita and Furuse 2002, Van Itallie and Anderson 2006, Furuse 2009). The Ig-like adhesion molecules of TJ include JAM-1, the cocksackie adenovirus receptor (CAR) (Hirabayashi and Hata 2006) and LSR (Masuda et al. 2011). The cytoplasmic domain of TJ contains several proteins with PDZ domains, such as ZO-1, ZO-2, ZO-3, MUPP1, and MAGI1 (Gonzalez-Mariscal et al. 2003, Guillemot et al. 2008b, Fanning and Anderson 2009). These proteins form complexes that act as docking platforms for membrane proteins, and recruit signaling and adaptor proteins, such as cingulin, paracingulin, ZONAB and symplekin (Gonzalez-Mariscal et al. 2003, Schneeberger and Lynch 2004, Guillemot et al. 2008b). Two major polarity complexes involved in the establishment and maintenance of apico-basal polarity are also associated with TJ. The aPKC-Par3-Par6 complex is linked to TJ through interactions of Par3 with JAM-A, whereas the Crumbs-PALS1-PATJ polarity complex is linked to TJ through interactions of PATJ with ZO-3 and claudins (Shin et al. 2006). At AJ, the Ig-like adhesion molecule nectin is associated with the cytoplasmic proteins afadin and ponsin (Takai and Nakanishi 2003), whereas the calcium-dependent adhesion receptor E-cadherin interacts with several intracellular proteins, including p120-catenin (p120ctn), β-catenin and α-catenin (Meng and Takeichi 2009). These cytoplasmic complexes organize the anchoring of AJ to the actin and microtubule cytoskeletons. Indeed, α-catenin, together with α-actinin, vinculin and eplin, participates in linking E-cadherin to actin filaments (Drees et al. 2005, Meng and Takeichi 2009), whereas the afadin-ponsin complex links nectin to actin filaments (Takai and Nakanishi 2003). p120ctn not only regulates E-cadherin clustering and stability (Reynolds 2007), but also connects E-cadherin to microtubules, through the PLEKHA7-Nezha complex (Meng and Takeichi 2009). At TJ, the link to the actin cytoskeleton is provided primarily through the ZO proteins ZO-1, ZO-2 and ZO-3 (Fanning et al. 1998, Wittchen et al. 1999). Importantly, the cytoskeletal networks associated with TJ and AJ undergo a profound reorganization and reorientation upon junction assembly, disassembly and maturation (Mege et al. 2006), and small GTPases play a key role in such cytoskeletal remodeling.

Rho family GTPase: Targets and effectors

In mammalian organisms, the Rho family of GTPases can be divided into six major groups: the Rho proteins (RhoA, RhoB and RhoC), the Rac proteins (Rac 1, 2, 3, and RhoG), the Cdc42-like proteins (Cdc42, TC10, TC10, TCL, Chp/Wrch2 and Wrch1), the Rnd proteins (Rnd1, Rnd2, Rnd3/RhoE), the RhoBTB proteins (RhoBTB1, RhoBTB2, RhoBTB3) and the Miro (mitochondrial Rho) proteins (Miro1, Miro2) (Wennerberg and Der, 2004). RhoA, Rac1 and Cdc42 are the best studied among the Rho family GTPases, and will be the focus of this review. When activated by binding to GTP, Rho family GTPases activate effectors, including protein and lipid kinases, which affect multiple cellular processes: cytoskeletal organization, membrane trafficking, cell survival, apoptosis, and gene expression (Fukata and Kaibuchi 2001, Ridley 2001, Benitah et al. 2004, Jaffe and Hall 2005). Through these targets, Rho family GTPases ultimately control cell-cell adhesion, cell motility, cell proliferation and morphogenesis.

Rho family GTPases have a major impact on the organization and contractility of the cytoskeleton (Figure 1). Rho promotes the formation of actin stress fibers, and enhances the contractility of actomyosin bundles. This is achieved through several mechanisms, including ROCK kinase-mediated phosphorylation and inhibition of the myosin light chain phosphatase (MLCP), and phosphorylation and activation of the myosin light chain kinase (MLCK), both of which increase myosin light chain phosphorylation, and hence actomyosin contractility (Amano et al. 2010). Moreover, actin filaments are stabilized by inhibition of the actin-severing protein cofilin, through ROCK kinase-mediated phosphorylation and activation of LIM kinase (Figure 1) (Maekawa et al. 1999). Furthermore, RhoA, Rac1 and Cdc42 regulate actin polymerization, albeit through different mechanisms. While RhoA activation stimulates unbranched actin polymerization through Diaphanous (Dia) and formins, Rac1 and Cdc42 promote polymerization and branching of actin filaments through the effectors WAVE and WASP, both of which activate the Arp2/3 complex to nucleate and polymerize actin filaments (Figure 1). Rac1 and Cdc42 also stimulate, through PAK kinases, the activity of LIM kinase (Edwards et al. 1999, Jaffe and Hall 2005). Moreover, activation of Rac1 and Cdc42 inhibits the association of IQGAP with β-catenin, thereby promoting cadherin-mediated cell-cell adhesion (Kuroda et al. 1998) (Figure 1).

Figure 1. Cytoskeletal effectors of Rho family GTPases. Schematic diagram representing the Rho family GTPases RhoA, Rac1 and Cdc42, their activities on major downstream effectors and targets, and their effects on cytoskeletal and junction organization. Green lines with arrows indicate activation, red lines with bar indicate inhibition, black lines with arrows indicate functional target/effect. Dia, diaphanous; ROCK, Rho kinase; PAK, p21-activated kinase; IQGAP, IQ motifs and GAP-related-domain containing protein; WAVE, WASP family Verprolin homologous protein; WASP, Wiskott-Aldrich syndrome protein; MLCK, myosin light chain kinase; MLCP, myosin light chain phosphatase; CRMP-2, Collapsin response mediator protein-2; LIMK, LIM domain kinase; CLIP-170, CAP-GLY domain containing linker protein; Arp, actin-related protein. See text for additional details. This Figure is reproduced in colour in Molecular Membrane Biology online.

Microtubule dynamics is also regulated by Rho family GTPases (Fukata et al. 2002) (Figure 1). In cancer cells, Cdc42- and Rac1-dependent activation of PAK results in phosphorylation of stathmin, resulting in microtubule stabilization (Daub et al. 2001, Wittmann et al. 2004). In neurons, Rho-dependent activation of ROCK inactivates CRMP-2, leading to microtubule disassembly and growth cone collapse (Arimura et al. 2005), and in Hela cells the Rho effector mDia coordinates the alignment of actin and microtubules along the long axis of the cell (Ishizaki et al. 2001). Moreover, Rac1 and Cdc42 promote the capture of microtubule plus-end at the leading edge of migrating cells, by promoting the interaction of CLIP-170 with the Rac effector IQGAP (Kuroda et al. 1998, Fukata et al. 1999). Finally, Rho family GTPases also regulate the interaction of the plus-end binding protein EB1 with the adenomatous polyposis coli (APC) tumor suppressor protein (Jaffe and Hall 2005).

Besides being master regulators of cytoskeletal organization, Rho family GTPases influence signaling pathways that regulate gene expression, for example through the serum-response factor (SRF), a coactivator of promoters containing the serum-response element (SRE), and through YAP/TAZ (Dupont et al. 2011). SRF translocation to the nuleus, together with the coactivator MAL, is promoted by Rho family GTPase-dependent actin remodeling, and depends on E-cadherin (Miralles et al. 2003, Busche et al. 2010). Recently it was shown that RhoA activity is required to allow the transcription factors YAP and TAZ to act as nuclear relays of mechanical signals exerted by extracellular matrix rigidity and cell shape, to stimulate cell proliferation and growth (Dupont et al. 2011). Additional mechanisms of regulation of gene expression by Rho family GTPases include modulation of kinases and scaffold proteins of the JNK and p38 MAP kinase pathways (Benitah et al. 2004).

Finally, the activities of Rho family GTPases influence many other cellular processes, such as the cell cycle, membrane trafficking, and apoptosis (Ridley 2001, Coleman et al. 2004, Jaffe and Hall 2005). In view of their multiple roles in regulating such fundamental cellular processes, it is not surprising that deregulation of Rho family GTPases has been implicated in cancer development and metastasis (Sahai and Marshall 2002a). This appears to depend primarily on deregulated expression and mutation of GEFs and GAPs and deregulated GTPase expression, but not on GTPase mutations (Vigil et al. 2010).

Regulation of Rho family GTPases

Small GTPases are active when bound to GTP, and inactive when bound to GDP. The hydrolysis of GTP, and thus the inactivation of GTPases, is promoted by GTPase activating proteins (GAPs). The release of GDP, and thus activation of the GTPase by binding to GTP, is stimulated by guanine exchange factors (GEFs). A third class of Rho GTPase regulatory proteins consists of guanine nucleotide dissociation inhibitors (GDIs), which bind to and stabilize the GDP-bound, inactive forms of GTPases. Regulation can also occur through pathways that regulate the stability and the localization of small GTPases. For example, the E3 ubiquitin ligase Smurf regulates RhoA degradation downstream of the Par-6/Par-3/aPKC polarity complex (Wang et al. 2003), and plakophilin targets RhoA to epithelial junctions (Godsel et al. 2010).

Guanine nucleotide exhange factors (GEFs)

The human genome contains over 70 genes coding for GEFs, many of which have not been characterized in detail (Schmidt and Hall 2002, Rossman et al. 2005). The same GTPase can be activated by different GEFs, and many GEFs activate more than one GTPase. The common structural feature of GEFs is the presence of two Dbl-homology (DH) domains that are crucial for their catalytic activity and one Pleckstrin Homology (PH) domain, responsible for binding phospholipids, and thus potentially important for plasma membrane targeting. A small number of GEFs, related to Dock180, lack the DH domain, and contain instead Dock Homology Regions (DHR), which promote guanine nucleotide exchange (Meller et al. 2005). A tight regulation of GEFs activity is crucial, and is carried out by different mechanisms, including protein-protein interactions (Schmidt and Hall 2002, Rossman et al. 2005), targeting to defined intracellular compartments by interaction with GTPases and adaptor proteins (Vaughan et al. 2011), relief of intramolecular inhibitory sequences (Katzav et al. 1991), micro-RNA dependent modulation of their levels of expression (Cottonham et al. 2010), and signal-mediated phosphorylation and degradation (Woodcock et al. 2009).

GTPase Activating Proteins (GAPs)

The human genome contains about 80 genes coding for GAPs, which act as inhibitors of small GTPases, by stimulating the low hydrolytic activity of RhoGTPases (Moon and Zheng 2003). Members of this family are multi-domains proteins characterized by the presence of a conserved RhoGAP domain. Two GAP proteins, Abr and Bcr, also have GEF activity, and the characteristic DH domain of GEFs (Tan et al. 1993, Cho et al. 2007). Many other GAPs have additional domains involved in protein-protein and protein-lipid interactions, such as PH and SH3 domains (Moon and Zheng 2003). A large numbers of GAPs show a weak selectivity towards RhoGTPases, while others are specific to a defined partner (Ridley et al. 1993, Moon and Zheng 2003, Tcherkezian and Lamarche-Vane 2007). GAPs are regulated by several mechanisms, including protein-protein interactions, post-translational modifications, and binding of second messenger molecules (Moon and Zheng 2003, Bernards and Settleman 2005).

Guanine nucleotide dissociation inhibitors (GDIs)

There are only three RhoGDIs so far described in the human genome (Olofsson 1999). RhoGDI1 is the most abundantly represented, and interacts with a large number of Rho family members (Fukumoto et al. 1990), whereas RhoGDI2 is specific of hematopoietic cells, and show a higher specificity for the Rac1 family (Scherle et al. 1993). RhoGDI3 is found in low abundance and is specifically expressed in lung, brain and testis (Zalcman et al. 1996). GDIs interfere with Rho GTPase signaling pathway by various mechanisms, including inhibition of GDP dissociation, inhibition of GAP-mediated GTP hydrolysis (Van Aelst and D'Souza-Schorey 1997), and cytoplasmic sequestration of RhoGTPases (DerMardirossian and Bokoch 2005). Interestingly, several RhoGTPases do not effectively interact with GDIs, suggesting that this inhibitory mechanism is restricted to a subset of GTPases (Olofsson 1999, DerMardirossian and Bokoch 2005).

The role of Rho family GTPases and Rap1 in the formation and maintenance of epithelial junctions

The formation, maintenance and remodeling of junctions, and hence epithelial morphogenesis and differentiation, require a precisely orchestrated mechanism of activation and inactivation of Rho family GTPases in space and time (Braga and Yap 2005, Samarin and Nusrat 2009, Watanabe et al. 2009). In addition to Rho family GTPases, the Ras-like protein Rap1 also regulates cadherin-based adhesion, JAM-A, and TJ-mediated barrier regulation (Kooistra et al. 2007, Boettner and Van Aelst 2009, Severson et al. 2009).

RhoA

The first evidence for the role of RhoA in the control of epithelial junctions came from the observation that inhibition of RhoA activity, either by expression of the Rho-inactivating enzyme C3 transferase, or by expression of dominant-negative mutants of RhoA, resulted in the thinning of the junction-associated actin filament belt, and the disassembly of both TJ and AJ (Nusrat et al. 1995, Braga et al. 1997, Takaishi et al. 1997, Gopalakrishnan et al. 1998, Jou et al. 1998, Bruewer et al. 2004). Conversely, the expression of constitutively active mutants of RhoA induced a thickening of actin filaments, which stabilized the AJ (Nusrat et al. 1995, Bruewer et al. 2004). Both ROCK and Diaphanous (Dia) have been identified as downstream RhoA effectors involved in the maintenance of the structure and function of TJ and AJ (Hirase et al. 2001, Walsh et al. 2001, Sahai and Marshall 2002b, Shewan et al. 2005) (Figure 1). In addition, the PRK2 effector is also involved in the Rho-dependent assembly of keratinocyte junctions, through the Fyn tyrosine kinase (Calautti et al. 2002). RhoA is also implicated in the modulation and disassembly of AJ in response to pathological or experimental stimuli, such as inflammatory cytokines or calcium depletion. The mechanisms include translocation of the guanine-exchange factor GEF-H1 to the AJ actomyosin contractile ring, and activation of ROCK kinase (Utech et al. 2005, Samarin et al. 2007).

Rac1

Rac1 plays a key role in the formation and the remodeling of AJ, but does not appear to act directly on TJ (Nakagawa et al. 2001). By promoting the formation of lamellipodia, Rac1 initiates the assembly of the primordial epithelial cell-cell contacts and allows the subsequent accumulation of E-cadherin at these sites. This leads to a positive feedback loop of spreading cell-cell adhesion, through the E-cadherin-dependent activation of phosphatidylinosotol-3-kinase (PI3-K) (Pece et al. 1999), the junctional recruitment of the Rac1 activator Tiam1, and the further activation of Rac1 (Nakagawa et al. 2001, Kovacs et al. 2002). Indeed, Tiam-1-mediated Rac1 activation promotes E-cadherin adhesion, whereas Tiam1 depletion inhibits the formation of AJ (Hordijk et al. 1997, Malliri et al. 2004, Mertens et al. 2005). Although it has been proposed that Tiam1 is recruited to membranes through its interaction with phospholipids generated by PI3-K activation (Michiels et al. 1997, Sander et al. 1998), some studies have questioned the role of PI3-K in the initial phases of E-cadherin-induced Rac1 activation (Nakagawa et al. 2001, Ehrlich et al. 2002, Lambert et al. 2002, Goodwin et al. 2003), and additional mechanisms that contribute to Tiam1 recruitment to membranes have been characterized (Bourguignon et al. 2000, Schmidt and Hall 2002, Guillemot et al. 2008a). Importantly, a basal Rac1 activity is necessary for the maintenance of AJ, since expression of both constitutively active and dominant negative isoforms of Rac1 leads to disruption of AJ (Braga et al. 1997, Takaishi et al. 1997, Jou and Nelson 1998, Bruewer et al. 2004). Rac1 stabilizes AJ by promoting actin polymerization through WAVE and Arp2/3, and by negatively regulating the interaction of IQGAP with β-catenin, which destabilizes the E-cadherin-based cell-cell contacts (Kuroda et al. 1998, Fukata et al. 1999) (Figure 1). Interestingly, activation of Rac1 can have different consequences on cell-cell adhesion and cell migration, depending on experimental context and environmental signals (Sander et al. 1998). For example, growth-factor-induced activation of Rac1 results in disruption of AJ, lamellipodia formation and scattering of epithelial cells (Ridley et al. 1995, Potempa and Ridley 1998). In contrast, in other model systems Rac1 protects AJ from disassembly, prevents cell scattering, and promotes the epithelial phenotype (Hordijk et al. 1997, Gopalakrishnan et al. 2002).

Cdc42

The role of Cdc42 in the assembly and maintenance of epithelial junctions is controversial. Nectin- and E-cadherin-mediated adhesion promote the activation of Cdc42 (Kim et al. 2000, Kawakatsu et al. 2002, Fukuhara et al. 2003, 2004), and depletion of the Cdc42 GEF Tuba results in thinning of lateral actin filaments, and changes in the junctional architecture in epithelial cells (Otani et al. 2006), suggesting that Cdc42 controls actin dynamics at junctions. However, expression of dominant negative Cdc42 does not induce significant changes in the perijunctional actin cytoskeleton (Kodama et al. 1999, Rojas et al. 2001). The stimulation of filopodia formation could be a mechanism through which Cdc42 activation promotes the formation of cadherin-based contacts (Vasioukhin et al. 2000, Otani et al. 2006). However, in keratinocytes Cdc42 activation and filopodia formation do not occur upon new junction formation (Erasmus et al. 2010). Overexpression of constitutively active Cdc42 disrupts both TJ and AJ, and Cdc42 activation is implicated in epithelial-mesenchymal transition, and in the Src-mediated phosphorylation and subsequent ubiquitination and degradation of E-cadherin (Kroschewski et al. 1999, Patel et al. 2005: Chen et al. 2006. Shen et al. 2008b). The role of Cdc42 in regulating the establishment of apico-basal epithelial polarity is unique, and distinguishes Cdc42 from Rac1. Cdc42 activation is essential in the formation of the Par-3/Par-6/aPKC polarity complex and in the establishment of epithelial apico-basal polarity (Yamanaka et al. 2001). Furthermore, Cdc42 and its GEF Tuba control lumen formation in epithelial cells in 3-D culture, by regulating mitotic spindle orientation and apical membrane exocytosis (Bryant et al. 2010, Qin et al. 2010).

Rap1

The role of Rap1 in cell-cell junction formation was first demonstrated in D. melanogaster, where Rap1 maintains the circumferential distribution of AJ (Asha et al. 1999, Knox and Brown 2002). In mammalian cells, Rap1 is required for the homotypic interaction of E-cadherin, rescues the junction disruption phenotype of cells expressing oncogenic Ras, and promotes AJ maturation (Hogan et al. 2004, Price et al. 2004). Rap1 is activated upon cadherin- and nectin-mediated cell-cell contact, through different GEFs, including Epac in endothelial cells, and C3G, PDZ-GEF1 and PDZ-GEF2 in epithelial cells (Kooistra et al. 2007, Boettner and Van Aelst 2009, Severson et al. 2009). The downstream effectors of Rap1 in epithelial cells include afadin/AF-6 (Boettner et al. 2000), and Rac1/Cdc42, both of which are activated following Rap1 activation (Sato et al. 2006). The activation of these latter GTPases has in turn been linked to the junctional recruitment and/or activation of Vav2, Tiam1 and FRG, following cadherin-dependent adhesion (Fukuyama et al. 2006). Thus, Rap1 is part of a positive feedback loop that involves Rho family GTPases, to promote the maturation of nectin- and cadherin-based cell-cell adhesion. However, Rap1 does not seem to be important for maintenance of AJ, once they have been assembled (Hogan et al. 2004, Fukuyama et al. 2005, Dube et al. 2008).

The role of junctional molecules in the regulation of Rho family GTPases

p120catenin

p120ctn is a Src kinase substrate and an armadillo domain family protein, which associates with the juxtamembrane domain of E-cadherin (Reynolds et al. 1994). In mice, p120ctn knockout leads to embryonic death, and its conditional inactivation leads to abnormal signaling, proliferation and growth (Davis and Reynolds 2006, Perez-Moreno et al. 2006, Schlegelmilch et al. 2011, Silvis et al. 2011). In fibroblasts, p120ctn overexpression inhibits RhoA activity and stimulates Rac1 and Cdc42 activities, resulting in a migratory phenotype, loss of stress fibers, and increased lamellipodia and filopodia formation (Anastasiadis et al. 2000, Noren et al. 2000, Grosheva et al. 2001). It has been proposed that cytoplasmic p120ctn binds to and inhibits RhoA, acting as a GDP dissociation inhibitor, whereas cadherin clustering promotes RhoA activation, by recruiting cytoplasmic p120ctn to junctions (Anastasiadis et al. 2000, Noren et al. 2000, Grosheva et al. 2001) (Figure 2A). In fibroblasts, p120ctn is present in a complex with the Rac1 GEF Vav2 (Noren et al. 2000), suggesting that p120ctn may stimulate Rac1 activity through Vav2, and thus indirectly inhibit RhoA through activation of the antagonistic Rac1 pathway (Nimnual et al. 2003) (Figure 2B). Another possible mechanism is indicated by the observation that Rac1 activation through receptor-mediated signaling promotes the association of the Rho GAP p190 RhoGAP with cadherin-bound p120ctn, resulting in localized inhibition of RhoA (Wildenberg et al. 2006) (Figure 2A). Since these regulatory interactions have been characterized in cultured fibroblasts, it is not clear to what extent they are applicable to epithelial cells (Anastasiadis and Reynolds 2001). In fact, studies on epithelial model systems have not provided a definitive answer about the role of p120ctn in RhoA regulation in normal polarized columnar epithelial cells. RhoA activity was reported to be increased upon depletion of p120ctn in serum-starved MDCK monolayers (Dohn et al. 2009) and breast cancer cells (Yanagisawa and Anastasiadis 2006, Yanagisawa et al. 2008). Moreover, the epidermis and primary keratinocytes isolated from p120ctn knockout mice show increased RhoA activity, increased stress fibers and reduced lamellipodia (Perez-Moreno et al. 2006). However, in another study, adenovirus-mediated overexpression of p120ctn resulted increased RhoA activity in primary keratinocytes, and decreased RhoA activity in fibroblasts, suggesting a cell-context-dependent function of p120ctn (Cozzolino et al. 2003). Furthermore, RhoA activity was significantly decreased upon p120ctn depletion in MDCK cells gown as cysts in 3-D culture (Marciano et al. 2011), supporting the notion that in epithelial cells junctional p120ctn stimulates, rather than inhibits, RhoA activity (Figure 2A). It should also be noted that the p120ctn isoform 1A, which associates with p190RhoGAP, is expressed in significant amounts in mesenchymal cells, but in very low amounts in epithelial cells (Marciano et al. 2011). Therefore the possible involvement of p190RhoGAP in down-regulating RhoA activity at epithelial junctions remains to be clarified (Figure 2A).

Figure 2. Regulation of small GTPases by junctional proteins. Schematic diagrams illustrating the major proteins and pathways involved in the regulation of RhoA (A), Rac1 (B), Rap1 (C), and Cdc42 (D) at junctions of mammalian epithelial cells. (E) is a graphic legend for panels A, B, C, D. Arrows indicate either physical/functional interaction, activation, or promotion of junctional recruitment. Lines with bar indicate inhibition. Dotted lines indicate indirect/unknown interaction. α = α-catenin; PL7 = PLEKHA7; p114RG = p114 RhoGEF; β-cat. = β-catenin. See text for additional details. This Figure is reproduced in colour in Molecular Membrane Biology online.

MAGI1

MAGI1 is a member of the membrane-associated-guanylate kinase (MAGUK) family of proteins, which are characterized by the presence of PDZ, SH3 and GUK domains. MAGI1 is localized in the cytoplasmic region of TJ and AJ, where it has been reported to associate with cadherin and nectin/JAM-A-based junctional protein complexes, and β-catenin (Ide et al. 1999, Dobrosotskaya and James 2000). In endothelial cells, MAGI1 is required for the activation of the small GTPase Rap1, by acting as a scaffolding protein for the Rap1 activator PDZ-GEF1 (Sakurai et al. 2006) (Figure 2C).

JAM-A, nectins and afadin

Both JAMs (Junctional Adhesion Molecules) and nectins are members of the family of immunoglobulin-like adhesion molecules, and interact with the cytoplasmic scaffold proteins afadin and ZO-1 at epithelial junctions (Mandai et al. 1997, Martin-Padura et al. 1998, Ebnet et al. 2000, Takai et al. 2008). In addition, JAM-A interacts with the polarity complex protein Par-3 (Martin-Padura et al. 1998, Ebnet et al. 2001, Itoh et al. 2001) (Figure 2B). Afadin is a multidomain protein containing PDZ, Ras- and actin-binding domains (Kuriyama et al. 1996, Mandai et al. 1997). JAM-A knockout mice show altered intestinal function, with increased permeability, proliferation and inflammation (Laukoetter et al. 2007), indicating multiple roles of JAM-A in TJ barrier regulation and signaling, in agreement with studies on cultured cells (Liu et al. 2000, Mandell et al. 2005). Conversely, knockout of afadin leads to early embryonic death, with loss of epithelial polarity (Ikeda et al. 1999, Zhadanov et al. 1999), indicating an essential role in morphogenesis. Interestingly, conditional knockout of afadin in intestinal epithelial cells after birth phenocopies the effect of JAM-A knockout in the intestine (Tanaka-Okamoto et al. 2011). JAM-A and nectins stimulate the activity of Rap1, and thus regulate junction integrity and cell migration, through afadin, which binds to Rap1 and functions as a Rap1 effector (Boettner et al. 2000, Mandell et al. 2005, Severson et al. 2009) (Figure 2C). One characterized mechanism is through the formation of a complex between JAM-A, afadin and the Rap1 GEF PDZ-GEF2 (Dube et al. 2008) at junctions (Severson et al. 2009). Afadin also promotes E-cadherin stabilization through its interaction with p120ctn, and has also been reported to bind to the Rap1 GAP SPA-1, and thus influence integrin-mediated adhesion (Su et al. 2003) (Figure 2C). Another putative Rap1 GAP, SPAL-3, has been localized at junctions (Matsuda et al. 2008), but its interaction with junctional proteins is not known. Finally, nectins are involved in Rac1 and Cdc42 activation upon the formation of the primordial cell-cell contacts, through mechanisms involving Rap1 (Fukuyama et al. 2005), c-Src, the Cdc42 GEF FRG, and the Rac1 GEF Vav2 (Kim et al. 2000, Kawakatsu et al. 2002, Fukuhara et al. 2003, 2004) (Figure 2–D).

Par-6 and Par-3

Par-3 and Par-6 are PDZ-domain-containing proteins, which form a complex with atypical protein kinase C (aPKC) and participate in the biogenesis of epithelial TJ (Izumi et al. 1998, Joberty et al. 2000, Johansson et al. 2000, Lin et al. 2000). The Par-3/Par-6/aPKC complex is involved in the establishment of apico-basal and antero-posterior polarity in a variety of invertebrate and vertebrate cell types (Kemphues et al. 1988, Ohno 2001), and is a signaling module that acts both as an upstream regulator and a downstream target of the small GTPases (Figure 2B–D). Binding of activated Cdc42 to Par-6 is required for the association of the Par-6/aPKC complex with Par-3, leading to TJ formation and establishment of apico-basal polarity (Joberty et al. 2000, Johansson et al. 2000, Lin et al. 2000) (Figure 2D). Par-6 also interacts with the transforming Cdc42-like GTPase Wrch-1, but activation of Wrch-1 results in the disruption of TJ (Brady et al. 2009). Par-3 is both a target and a regulator of Rac1 activity. Rac1 activation is required to recruit Par-3 to junctions, since in keratinocytes lacking Tiam1 junction assembly and localization of Par-3 to TJ are delayed (Mertens et al. 2005) (Figure 2B). However, in MDCK cells depletion of Par-3 results in a Tiam-1-dependent increase in Rac1 activity, suggesting that Par-3 recruits and negatively regulates Tiam-1 function at junctions (Chen and Macara 2005) (Figure 2B). In neuronal cells and fibroblasts, the Par-3/Par-6 complex promotes the recruitment of Tiam-1 to the growing cone and lamellipodia, respectively, but in this case Rac1 activity is stimulated, and not inhibited (Nishimura et al. 2005). In this context, RhoA-ROCK-kinase-mediated phosphorylation of Par-3 acts as an inhibitory switch, resulting in Rac1 inactivation, and disruption of the Par complex (Nakayama et al. 2008). In neuronal cells, Par-6 also controls RhoA activity through p190RhoGAP (Zhang and Macara 2008). However, such regulation has not been demonstrated in epithelial cells.

ZO proteins

ZO proteins (ZO-1, ZO-2 and ZO-3) are MAGUK family proteins localized in the cytoplasmic domain of TJ (Mitic and Anderson 1998), where they play redundant and non-redundant roles as scaffolds of TJ membrane proteins and anchors for the actin cytoskeleton (Umeda et al. 2006, Van Itallie et al. 2009, Xu et al. 2009). Mice knockout for either ZO-1 or ZO-2 have an embryonic lethal phenotype, whereas mice lacking ZO-3 have an apparently normal phenotype (Katsuno et al. 2008, Xu et al. 2008).

Several lines of evidence indicate that ZO proteins are involved in the regulation of RhoA, Cdc42 and Rac1. MDCK cells overexpressing the N-terminal part of ZO-3 (NZO-3) show reduced RhoA activity, increased migratory behaviour, and decreased stress fibers (Wittchen et al. 2003). Since the C-terminal region of ZO-3 binds both to NZO-3 and p120ctn, it was proposed that NZO-3 competes with p120ctn for binding to the C-terminus of endogenous ZO-3, and thus renders p120ctn available to interact with other proteins, such as Vav-2 or Rho, resulting in the inhibition of RhoA activity (Wittchen et al. 2003) (Figure 2A). However, a direct role of ZO-3 in regulating RhoA activity in epithelial cells has not been demonstrated. The Cdc42 GEF Tuba forms a complex with ZO-1, and requires ZO-1 in order to be recruited to junctions of colonic epithelial cells (Otani et al. 2006) (Figure 2B). MDCK cells depleted of Tuba show delayed AJ formation and reduced activation of Cdc42 during the calcium switch assay (Otani et al. 2006). Similarly, MDCK cells lacking ZO-1 and depleted of ZO-2 show delayed AJ assembly, and fail to activate Rac1 during the calcium switch (Ikenouchi et al. 2007). Taken together, these data suggest that the ZO protein complex is required to achieve efficient activation of Rac1 and Cdc42 during junction assembly, possibly through the direct or indirect recruitment of Tuba to junctions (Figure 2B–D).

E-cadherin, β-catenin and α-catenin

E-cadherin is the major calcium-dependent adhesion molecule of epithelial AJ, and its cytoplasmic domain interacts with the armadillo family proteins p120ctn and β-catenin. In turn, β-catenin binds α-catenin, which plays a key role in anchoring AJ to the actin cytoskeleton (Meng and Takeichi 2009). The importance of E-cadherin and associated proteins for cell-cell adhesion and morphogenesis is highlighted by the observation that the targeted disruption of these proteins gives an embryonic lethal phenotype in mice (Larue et al. 1994, Haegel et al. 1995, Torres et al. 1997).

E-cadherin is involved in the activation of both Rac1/Cdc42 and Rap1. Phosphatidylinositol-triphosphates (PIP3) generated at the membrane upon cadherin-mediated activation of PI-3-kinase (PI3-K) have been implicated in the binding to the PH domain of Rac1/Cdc42 GEFs, such as Tiam1 and Asef, leading to their recruitment to junctions (Sander et al. 1998, Pece et al. 1999, Kovacs et al. 2002, Malliri et al. 2004, Muroya et al. 2007) (see also Rac1 section). Furthermore, E-cadherin stimulates the activation of Rap1 during the early phases of junction assembly, through the transient interaction of its cytoplasmic domain with the Rap1 GEF C3G (Hogan et al. 2004) (Figure 2C). The ability of E-cadherin to signal to small GTPases may depend on its phosphorylation, as suggested by the observation that depletion of the cadherin-associated protein tyrosine phosphatase PTP-PEST affects RhoA and Rac1 activities, and enhances migratory activity of colon carcinoma cells (Espejo et al. 2010).

β-catenin contributes to the local activation of Rap1, by interacting with the Rap1 GEF, PDZ-GEF1, directly, and through its interacting partners MAGI-1 and MAGI-2 (Mino et al. 2000, Sakurai et al. 2006) (Figure 2C).

With regards to α-catenin, its possible involvement as a target of RhoA is indicated by its interaction with the RhoA GAP ARHGAP10 (Sousa et al. 2005). The activity of ARHGAP10 favors the association of α-catenin with cell-cell contacts, as well as enabling invasion by the bacterial pathogen Listeria (Sousa et al. 2005) (Figure 2A). Another Rho GAP, ARHGAP12, has been localized at epithelial junctions, but its interaction with junctional proteins has not been characterized (Matsuda et al. 2008).

Angiomotin

Angiomotin (Amot) belongs to a family of proteins characterized by a glutamine-rich domain, a coiled-coil domain, and a C-terminal PDZ-binding domain (Bratt et al. 2002, Sugihara-Mizuno et al. 2007). Amot interacts with the polarity complex proteins PALS1, PATJ and Par-3, and recruits the Cdc42 GTPase activating protein Rich1 to TJ in MDCK cells (Wells et al. 2006) (Figure 2D). The activity of Rich1 does not affect AJ, but is required to maintain the organization and barrier function of TJ, probably by allowing the cycling of Cdc42 between the GDP- and GTP-bound states, which is necessary to maintain the stability of the TJ polarity complexes (Wells et al. 2006). Amot has also been implicated in promoting cell migration and the breakdown of junctions in endothelial cells (Troyanovsky et al. 2001, Bratt et al. 2005). In fact, Amot interacts with the RhoGEF Syx through a ternary complex with Pals1 and PatJ, thereby controlling the spatially restricted activation of RhoA at lamellipodia of migrating endothelial cells (Ernkvist et al. 2009). In summary, Amot is a dual regulator of small GTPase activity: during endothelial migration it promotes local RhoA activation through Syx, whereas in epithelial cells it contributes to the maintenance of TJ by negatively regulating Cdc42 activity through Rich1. Studies on knockout mice show that Amot has a critical role during vascular patterning and endothelial polarization (Aase et al. 2007).

Cingulin and paracingulin

Cingulin and paracingulin are structurally similar proteins, comprising globular N-terminal domains, a central coiled-coil domain and small globular tails (Citi et al. 1988, Ohnishi et al. 2004, Guillemot and Citi 2006b). Both proteins are localized at TJ, where they are recruited by interaction with ZO-1 and other proteins (Citi et al. 1989, Cordenonsi et al. 1999, Ohnishi et al. 2004, Pulimeno et al. 2011). However, paracingulin is also localized at AJ (Ohnishi et al. 2004), where it forms a complex with E-cadherin, p120ctn and PLEKHA7 (Paschoud et al. 2010, Pulimeno et al. 2011). Both cingulin and paracingulin form a complex with and interact directly with the Rho GEF Lfc/GEF-H1 (Ren et al. 1998, Krendel et al. 2002, Benais-Pont et al. 2003), providing a mechanism to down-regulate Rho signaling and G1/S phase transition, as cell density increases (Aijaz et al. 2005, Guillemot et al. 2008a, Citi et al. 2009) (Figure 2A). Interestingly in fibroblasts, which lack TJ, GEF-H1 is sequestered by microtubules, and microtubule depolymerization leads to RhoA activation (Liu et al. 1998, Ren et al. 1998, Glaven et al. 1999, Krendel et al. 2002). Cingulin is also implicated in the recruitment to TJ of the Rho GEF p114RhoGEF, which has been proposed to promote spatially restricted, apical RhoA activation during junction assembly of corneal epithelial cells, while GEF-H1 would act on basally localized RhoA (Terry et al. 2011) (Figure 2A). However, cingulin depletion from columnar epithelial cells (Guillemot and Citi 2006a), and cingulin knockout in epithelial endodermal cells of embryoid bodies (Guillemot et al. 2004) do not have detectable inhibitory effects on TJ assembly, suggesting that the p114RhoGEF-mediated regulation of TJ assembly by cingulin may be restricted to specific cell types. The spatial regulation of RhoA activity at apical junctions versus basal stress fibers of epithelial cells is an example of how the subcellular distribution of adaptors for different GEFs and GAPs can modulate cell differentiation (Terry et al. 2011). Interestingly, modulation of RhoA activity is one of the mechanisms through which cingulin and paracingulin regulate gene expression (Guillemot et al. 2004, 2008a, Guillemot and Citi 2006b, Citi et al. 2009).

A unique feature of paracingulin, that distinguishes it from cingulin, is its role in the regulation of Rac1 activation in epithelial cells. The junctional localization of the Rac1 GEF Tiam1 is reduced in paracingulin-depleted cells, and Tiam1 forms a complex with and interacts directly with paracingulin (Guillemot et al. 2008a). As a result, depletion of paracingulin results in decreased Rac1 activation during junction assembly, with a delay in the formation of both AJ and TJ (Guillemot et al. 2008a). In summary, paracingulin is a dual regulator of small GTPase activity: (1) During junction assembly (e.g., low confluence cells), it promotes Rac1 activation, by recruiting Tiam1 to junctions; and (2) when junctions are mature (e.g., high confluency monolayers), it contributes to down-regulating RhoA activity, by sequestering GEF-H1 at junctions (Figure 2A).

Plakophilin2

Plakophilin-2 is a member of the armadillo family of proteins, and links desmosomal cadherins to the cytoskeleton, through its interaction with desmoplakin, plakoglobin, and intermediate filaments (Bass-Zubek et al. 2009). Desmosomes provide strong adhesion between epithelial cells, and also regulate differentiation and homeostasis (Garrod and Chidgey 2008, Green et al. 2010). Although desmosomes are associated with intermediate filaments, desmosome assembly requires the integrity of actin microfilaments (Pasdar and Li 1993). Interestingly, loss of plakophilin results in changes in cortical actin organization during junction assembly, with increased stress fiber formation (Godsel et al. 2010). Overall RhoA activity is increased in plakophilin-2-depleted cells, but active RhoA fails to accumulate at cadherin-based junctions (Godsel et al. 2010). However, it is not clear whether plakophilin regulates the activity and localization of RhoA directly, by interacting with it, or indirectly, by influencing the localization or activity of either a Rho GEF or a Rho GAP.

Dynamic regulation of small GTPases at junctions

Studies on the dynamics of localization and activation of Rho family GTPases in living cells have greatly improved our understanding of the spatio-temporal pattern of their regulation. The notion that Cdc42 activation occurs upon formation of new junctions was first demonstrated using fluorescently tagged effector domains that bind to activated Cdc42 (Kim et al. 2000). In epithelial cells, activation of Cdc42 is required to direct the establishment of apico-basal polarity, through the Par-3/Par-6/aPKC polarity complex (Par-6 and Par-3 section) (Figure 2D). Cdc42 activation is regulated by the GEF Tuba and the GAP Rich1, which associate with cytoplasmic protein complexes of TJ, including ZO-1 and PATJ/Amot, and by the GEF FRG, which is activated in a src-dependent manner upon nectin engagement (JAM-A, nectins and afadin, ZO proteins and Angiomotin sections) (Figure 2D).

Experiments using GFP-tagged Rac1 and FRET-based biosensors were crucial to clarify how Rac1 is activated at junctions upon formation of new cadherin-based cell-cell contacts (Ehrlich et al. 2002, Yamada and Nelson 2007). Cytoplasmic Rac1 transiently accumulates at the tips of lamellipodia in isolated cells, and its association with the membrane is stabilized by the formation of new junctions. As the region of contact ‘matures’, Rac1-GFP signal progressively decreases at these sites, whereas it increases along the expanding, newest sites of cell-cell contact (Ehrlich et al. 2002). Significantly, both the Arp2/3 complex, an effector of Rac1, and paracingulin, a Tiam-1-interacting protein, are distributed in a manner similar to Rac1 during junction expansion (Yamada and Nelson 2007, Paschoud et al. 2010). The relocalization of Rac1 to the edges of the cell-cell contact area, where it promotes further E-cadherin accumulation, generates a positive feedback loop that expands the area of cell-cell contact. The major Rac1 activator in epithelial cells is Tiam1, and it is recruited to junctions and regulated by paracingulin, Par-3 and phosphatidylinositol-3-phosphates (Rac1, Par-6 and Par-3 and Cingulin and paracingulin sections) (Figure 2B). Interestingly, although new junctional GAPs are being identified (Matsuda et al. 2008), it is unclear which Rac1-specific GAP is operating at epithelial junctions. New junction formation also requires the activation of Rap1, which is regulated by GEFs and GAPs that associate with E-cadherin, JAM-A, afadin, β-catenin and MAGI (MAGI1, JAM-A, nectins and afadin and E-cadherin, β-catenin and α-catenin sections) (Figure 2C).

Activated RhoA specifically accumulates at the expanding edges of cell-cell contact, and not in the center of these areas, and drives the myosinII-dependent, lateral expansion of the junction (Yamada and Nelson 2007). Accordingly, cycling of RhoA between active and inactive forms, which promotes the contractility of actomyosin filaments, must occur at these sites. Significantly, cingulin and paracingulin, adaptor molecules for the Rho GEFs, GEF-H1 and p114RhoGEF (Figure 2A), accumulate at the edges of the expanding regions of cell-cell contact during junction formation in living cells (Paschoud et al. 2010). Concerning the mechanisms through which RhoA is inactivated at junctions, a role could be played by the cytoplasmic protein complex associated with E-cadherin, since the Rho GAPs p190 and ARHGAP associate with p120ctn and α-catenin, respectively (Figure 2A). However, whether p120ctn stimulates or inhibits RhoA activity in epithelial cells is not clear, since the outcome of p120ctn depletion or overexpression experiments appears to be cell-context dependent (p120catenin and E-cadherin, β-catenin and α-catenin sections) (Figure 2A).

Conclusions and perspectives

Regulation of Rho family GTPases is of paramount importance in coordinating fundamental cellular activities, such as cytoskeletal organization, cell migration, and cell adhesion and proliferation (Figure 1). This regulation is achieved through a complex network of activators and inhibitors, which bind to specific junctional adaptors. Thus, the expression, modulation and subcellular localization of such regulatory molecules is critical in shaping the cell-context-dependent signaling output.

Epithelial junctions constitute specific, spatially restricted signaling modules for Rho family GTPases, since several transmembrane and cytoplasmic proteins belonging to TJ and AJ act either as regulators or targets of Rho family GTPases (Figure 2). For most of these proteins, specific interactions with GTPases, GEFs and GAPs have been demonstrated, providing an insight onto the molecular mechanisms of GTPase regulation. However, additional antibody and live imaging experimental tools to clarify the localization of GEFs and GAPs in fixed and living cells will be required to clarify the spatio-temporal dynamics of their interaction with junctional proteins.

Another important challenge for future studies will be to determine the relevance of GEFs, GAPs and their interacting junctional proteins not only in cultured cell model systems, but also at the whole organism level. Considering the large number of GEFs and GAPs that have been identified by genome sequence analysis, functional redundancies must be carefully examined, before definitive conclusions are drawn on the physiological roles of GEFs, GAPs, and their adaptor molecules.

Another key question concerns the mechanisms underlying the interactions between GEFs, GAPs and their adaptor molecules. How do junctional proteins, most of which exist in an equilibrium between junction-associated and cytoplasmic soluble pools (Shen et al. 2008a), activate or inactivate GEFs and GAPs? Do they bind to GEFs and GAPs only when they are associated with their partner molecules at junctions? Do they modulate the function of GEFs and GAPs by affecting their conformation or phosphorylation status? Or do they function mainly by sequestering GEFs and GAPs to a specific subcellular compartment?

Finally, a critical issue for future studies is how the mutually antagonistic cross-talk of GTPases is generated at junctions. In the Xenopus laevis cell wounding model, segregation of the zones of RhoA and Cdc42 activity depends on Abr, a Rho GEF that also contains a Cdc42 GAP domain (Benink and Bement 2005, Vaughan et al. 2011). However, such a GEF is unusual, and has not been localized at mammalian junctions. Thus, additional studies on the functional cross-talk of GEFs and GAPs, and the role of adaptor proteins in this cross-talk, are necessary to address these questions, and provide insights on the molecular basis mechanisms of small GTPase regulation at junctions.

Acknowledgements

We are grateful to the State of Geneva and the Swiss National Funds for supporting our research and to Nicolas Roggli for graphics help and advice. We apologize for not being able to cite several original articles, due to space limitations.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.