Cycling around cell–cell adhesion with Rho GTPase regulators

Epithelial cells are essential for the homeostasis and morphogenesis of different tissues. A key factor that underpins appropriate epithelial function is the ability of cells to adhere tightly to their neighbours and polarise, showing distinct molecular platforms and signalling properties at the apical or basolateral domain (Green et al., 2010). Indeed, distinct pathologies are caused by inappropriate regulation of signalling that interferes with epithelial adhesion.

A hallmark of epithelia is the presence of specialised adhesive structures that provide strong cohesion within a monolayer of cells. There are three main classes of epithelial adhesive structures that have different molecular components, cytoskeletal interactions and functions. Adherens junctions (AJs) are formed by classical cadherins receptors (E- or P-cadherin), which associate indirectly with the cortical actin cytoskeleton (Niessen et al., 2011). Desmosomal cadherins (desmogleins or desmocollins) assemble desmosomes, and their interaction with intermediate filaments provides mechanical strength to epithelial sheets (Green et al., 2010). By contrast, tight junctions (TJs) are formed by claudins and occludins in association with actin filaments to regulate the passage of ions and small solutes across epithelial cells (Terry et al., 2010). Additional transmembrane proteins, such as nectins, cooperate with AJs and TJs to mediate cell–cell adhesion (for reviews see Citi et al., 2009; Green et al., 2010).

Epithelial junction assembly and maintenance occur through a number of sequential cellular processes (Fig. 1) that are modulated by distinct signalling pathways. Over the past years, research focused on three main points: first, how signalling platforms integrate intracellular trafficking and cytoskeletal dynamics to maintain strong cohesion and appropriate levels of receptors at junctions (Green et al., 2010); second, how junction formation leads to local and global rearrangements of cytoskeletal architecture that reinforce contacts and underpin epithelial polarisation (Mège et al., 2006; Gomez et al., 2011; Niessen et al., 2011); and third, how the initial cell–cell contact triggers the formation of distinctive adhesive structures. Earlier work supports the view of a hierarchical process, in which cadherin complexes are the first to engage and are the main drivers of epithelial adhesion.

Newly formed contacts are characterised by intermingling of molecular components of AJs with TJs or desmosomes (Green et al., 2010). Yet, as junctions mature, their distribution is resolved into distinct adhesive complexes along the lateral domain (Green et al., 2010). Although formation and maturation of AJs, TJs and desmosomes are coordinated spatially and temporally, evidence suggests that each of these complexes is regulated in its own right through a distinct network of signalling molecules. Specific pathways might impinge separately on each adhesive structure at mature junctions; however, some level of coordination towards the precise morphology, organisation and function of cell–cell contacts must be in place.

Most of our knowledge on signalling that regulates junctions is derived from the three best-characterised members of the Rho small GTPases: RhoA, Rac1 and Cdc42 (for recent reviews see Citi et al., 2009; Niessen et al., 2011). Rac1 is required for the recruitment of F-actin to clustered cadherin receptors, TJ assembly and epithelial morphogenesis (Braga and Yap, 2005; Citi et al., 2009); RhoA is necessary for the formation and function of adhesive complexes (Braga and Yap, 2005; Citi et al., 2009). However, conflicting results show activation and inactivation of RhoA by junctions (Calautti et al., 2002; Noren et al., 2003; Yamada and Nelson, 2007), and hyperactivation of RhoA signalling is often found in malignant cells (Karlsson et al., 2009). A potential RhoA-dependent function that is relevant for junctions is contraction, which is necessary for tension and remodelling of circumferential actin filaments (Gomez et al., 2011). These events regulate the mobility of clustered receptors at junctions, size of cell–cell contacts and resistance to tension that is generated by polarisation, morphogenetic events, cell movement or cell division, for example (Gomez et al., 2010; Liu et al., 2010; Borghi et al., 2012). By contrast, Cdc42 is required for cadherin adhesion in some, but not all, cell types (Braga and Yap, 2005; Erasmus et al., 2010). The main function of Cdc42 appears to be in epithelial morphogenesis, the localisation of polarity complexes to junctions, and the assembly and function of TJs (Terry et al., 2010).

When analysing signalling events, it is important to consider three points. First, signalling pathways must be interpreted with respect to the adhesive receptor that mediates adhesion, the cell type and the tissue of origin (i.e. mesenchymal, neuronal, epithelial or tumour cell lines). The latter is particularly relevant, as expression of an oncogene can substantially rewire intracellular signalling and provide further levels of complexity. Second, the hierarchy and cooperation among different adhesive systems must be considered. For example, to claim specificity of regulation of TJs, one has to show that AJs are not perturbed. Third, functional studies on cell–cell adhesion using different experimental models might not be directly comparable (Fig. 2). Distinct cellular processes are involved depending on the model used, and the signalling events that mediate the formation, reinforcement or stabilisation of junctions might not be the same. Importantly, although the same GTPase can participate in junction assembly and destabilisation, the regulators, effectors and cellular processes involved might differ in each case (Fig. 2).

In this Commentary, we focus on our current understanding of the different Rho regulators that participate in the assembly and stabilisation of AJs and TJs, the distinct mechanisms by which they localise to cell–cell contacts and the surprisingly disparate cellular functions they regulate to facilitate cohesiveness among epithelial cells. A new perspective is unravelled, in which cell–cell contacts can also influence the function of Rho regulators to control other cellular processes that are relevant for epithelial homeostasis and function.

With few exceptions, members of the Rho GTPase subfamily cycle between an active and inactive state (Box 1). The continuous cycling between these two states is tightly controlled by a number of regulatory proteins that specify where and for how long the signal is ‘on’ and which cellular function is modulated. Rac1 is activated by newly formed contacts in a number of cell types, but its activity must be carefully fine-tuned to be compatible with cell–cell contacts (Braga and Yap, 2005). By contrast, RhoA and Cdc42 have been shown to be either activated or inactivated by junction assembly in epithelia (Braga and Yap, 2005; Erasmus et al., 2010). One argument is that for each cellular process that is triggered by cell–cell contacts (Fig. 1), multiple GTPases must be dynamically turned ‘on’ or ‘off’. The latter is facilitated by guanine-nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs), respectively (Box 1) (Rossman et al., 2005; Tcherkezian and Lamarche-Vane, 2007).

So far, several GEFs and GAPs have been described to modulate the organisation, molecular composition and function of adhesive complexes in different cell types (Table 1). The strength of supporting evidence for their role in cell–cell adhesion varies. Clearly, it is essential to demonstrate the requirement of the catalytic activity or ‘rescue’ by the putative GTPase substrate. Because a GEF or GAP usually has more than one in vitro GTPase substrate (Table 1), it is also important to define which small GTPase regulates junctions in vivo. Of note, although the local activation of small GTPases is paramount, the presence of a GAP or GEF at junctions does not necessarily imply a functional regulation of cell–cell contacts. Indeed, a number of GEFs and GAPs are inactivated at junctions (see below; Fig. 3).

Below, we discuss the specific regulators that participate at different steps of AJ and TJ biogenesis and maintenance. For the sake of simplicity and specificity, GAPs and GEFs are highlighted in the context of the cellular process regulated, as inferred by the experimental model used (Fig. 2) and controls provided. Furthermore, we discuss a variety of distinct mechanisms used to confine GEF or GAP activity to particular intracellular locations, but that impact on junctions (Fig. 3).

Very few regulators have been shown to primarily modulate assembly of AJs. Depletion of the GAP SH3 domain-binding protein (SH3BP1) in Caco-2 and A431 cells strongly perturbs the recruitment of E-cadherin to newly formed junctions (Fig. 2) (Elbediwy et al., 2012). As expected from the strong defects in AJs in the absence of SH3BP1, mature junctions are also perturbed and have reduced levels of the TJ component zonula occludens-1 (ZO1, also known as TJP1) and transepithelial resistance (TER). In the early stages of junction formation, SH3BP1 is recruited to cell–cell contacts as part of a multi-protein complex containing paracingulin and CD2-associated protein (CD2AP) (Fig. 3). SH3BP1 has in vitro GAP activity on Rac1 and Cdc42; however, depletion of SH3BP1 leads to an overall increase in active Cdc42, but not Rac1, in whole-cell lysates. Importantly, reduced levels of active Cdc42 are found at junctions in the absence of SH3BP1, indicating that this GAP is required for confining Cdc42 activity to cell–cell contacts (Elbediwy et al., 2012).

Instead of preventing initial assembly of cell–cell contacts, a subset of regulators modulate cellular processes that reinforce the stability of pre-formed junctions (Fig. 1), such as delivery of cadherin receptors to the cell surface, F-actin recruitment or transcription events.

Several GAPs regulate the maintenance of cell–cell adhesion. The first one, DLC1 (also known as DLEC1), is the prototypical member of the deleted in liver cancer family of proteins. DLC1 inactivates RhoA and Cdc42, and has been shown to regulate migration and proliferation. As the name suggests, DLC1 is downregulated in numerous cancers (Liao and Lo, 2008). Re-expression of DLC1 in various prostate cancer cell lines stabilises AJs (Tripathi et al., 2012) through an interaction with α-catenin. DLC1 also inhibits cell growth through GAP-dependent and -independent mechanisms (Goodison et al., 2005; Wong et al., 2005; Guan et al., 2008). DLC1 activity and its ability to bind α-catenin are both necessary for the strong tumour-suppressive activity shown by DLC1 expression (Tripathi et al., 2012).

PX-RICS is a splice variant of RhoGAP involved in the β-catenin–N-cadherin and N-Methyl-D-aspartate receptor signalling (RICS, also known as ARHGAP32), which has a low in vitro GAP activity towards Rac1 and Cdc42 (Hayashi et al., 2007). In the absence of PX-RICS, N-cadherin and β-catenin are lost from cell–cell contacts and accumulate at the perinuclear region of epithelial cell lines and PX-RICS-null mouse embryonic fibroblasts (MEFs) (Nakamura et al., 2008). Although PX-RICS interacts with β-catenin, interestingly it does not localise to junctions, but instead to the Golgi complex (Fig. 3) (Nakamura et al., 2008). β-Catenin associates with cadherins as soon as the latter is synthesised (Chen et al., 1999) and thus PX-RICS is well placed to regulate the transport of N-cadherin–β-catenin to the surface. Indeed, PX-RICS expression in PX-RICS-null MEFs restores the localisation of N-cadherin and β-catenin, whilst a GAP-deficient mutant does not. Although the requirement of Cdc42 or Rac1 has not been formally demonstrated in the mislocalisation of N-cadherin, a potential mechanism is the indirect coupling of PX-RICS to the dynein–dynactin motor protein to enable ER-to-Golgi transport of N-cadherin and β-catenin (Nakamura et al., 2010).

The GAPs p190A (also known as ARHGAP35) and p190B (also known as ARHGAP5) are structurally similar but with distinct phenotypes in knockout mice; p190A^−/− mice die in utero, whereas p190B-null mice are viable, but have perturbed growth and breast ductal morphogenesis (Chakravarty et al., 2003). In different cell types, p190 family members inactivate RhoA activity downstream of C-cadherin (Noren et al., 2003), N-cadherin (Wildenberg et al., 2006) and E-cadherin (Ratheesh et al., 2012). Interestingly, the localisation of p190A and p190B to junctions in fibroblasts and epithelial breast cancer cells requires active Rac1 (Wildenberg et al., 2006; Ratheesh et al., 2012). p190B binds directly to active Rac1 in a region outside of the GAP domain (Bustos et al., 2008; Ratheesh et al., 2012), whereas recruitment of p190A might be mediated by its interaction with p120^CTN (Wildenberg et al., 2006).

Myosin IXA and IXB (Myo9A and Myo9B, respectively) are RhoGAPs containing two myosin head-like domains and are found at sites of cell–cell contacts in epithelial cells (Abouhamed et al., 2009; Chandhoke and Mooseker, 2012; Omelchenko and Hall, 2012). Myo9A knockout mice develop hydrocephalus as a result of defects in cell–cell adhesion of the ependymal cells lining the ventral caudal third ventricle (Abouhamed et al., 2009). In bronchial epithelial cells, depletion of Myo9A does not prevent initial junction assembly, but these contacts are unstable (Omelchenko and Hall, 2012). Consistent with this defect, the absence of Myo9A reduces the localisation of AJ and TJ proteins to mature junctions (Abouhamed et al., 2009). By contrast, Myo9B RNA interference (RNAi) does not perturb AJs, but abolishes assembly of TJ complexes (Chandhoke and Mooseker, 2012). These differences are intriguing, considering that depletion of either RhoGAP leads to hyperactivation of RhoA. Nevertheless, in addition to defective junction organisation, Myo9A or Myo9B signalling impacts on motile events, such as epithelial wound healing and collective cell migration (Chandhoke and Mooseker, 2012; Omelchenko and Hall, 2012).

Two distinct GEFs have been shown to participate in the regulation of tension at filaments adjacent to junctions, which is important for the reinforcement of receptors at adhesion sites and junction morphology. First, the Cdc42 GEF Tuba regulates the shape of cell–cell contacts and appropriate organisation of adhesive receptors along newly formed and mature junctions (Fig. 2) (Otani et al., 2006). In the absence of Tuba, the distribution of cadherin and F-actin is distorted in the lower lateral regions of mature junctions, and AJs and TJs are warped (Otani et al., 2006). This phenotype depends on Cdc42 and neural Wiskott–Aldrich syndrome protein (N-WASP), a Cdc42 effector. Although adhesive structures are present in Tuba-depleted cells, they do not function properly; defective lumen formation in epithelial cysts results from abnormal formation of a transient apical complex and/or the orientation of the mitotic spindle (Bryant et al., 2010; Qin et al., 2010). Disruption of Tuba function also has important consequences for bacterial uptake (see below).

The second is the RhoA, Rac1 and Cdc42 GEF epithelial cell transforming sequence 2 oncogene (ECT2) (Tatsumoto et al., 1999), which localises to the nucleus, cell–cell contacts and, in mitotic cells, the spindle and contractile ring (Yüce et al., 2005). ECT2 interacts with polarity complexes, actin and microtubule-binding proteins, co-precipitates with cadherin complexes and activates RhoA and Cdc42 in vivo (Hara et al., 2004; Liu et al., 2004; Ratheesh et al., 2012). In Madin-Darby canine kidney (MDCK) and MCF-7 human breast cancer cells, depletion of ECT2 does not prevent the assembly of AJs or TJs (Liu et al., 2004; Ratheesh et al., 2012). However, the lateral distribution of E-cadherin is perturbed, and levels of active RhoA and Myosin IIA at AJs are reduced (Ratheesh et al., 2012); junction-dependent regulation of Rac1 and Cdc42 is not reported. Thus, ECT2 regulates RhoA signalling to stabilise E-cadherin at the correct apical position (Shewan et al., 2005; Smutny et al., 2010; Ratheesh et al., 2012). Interestingly, ECT2 localisation and function is tightly coupled with two GAPs, MgcRacGAP (also known as RACGAP1) and p190B (see Box 2).

Controversial roles and interesting properties have been reported for APC-stimulated exchange factor 1 (ASEF, also known as ARHGEF4). ASEF directly interacts with the tumour suppressor adenomatous polyposis coli (APC) (Kawasaki et al., 2000), which releases autoinhibition of ASEF to activate Rac1 or Cdc42 (Kawasaki et al., 2003; Mitin et al., 2007). Overexpression of constitutively active ASEF has been shown to either increase (Muroya et al., 2007) or reduce (Kawasaki et al., 2003) E-cadherin levels at mature junctions in a GEF-dependent manner. The reasons for these observed discrepancies are not clear. As APC is known to regulate the localisation and activity of ASEF, tumour cell lines containing different mutations in APC might show altered cellular responses to ASEF depletion or overexpression (Kawasaki et al., 2003; Mitin et al., 2007).

In contrast to mediating junction stabilisation by affecting the cytoskeleton or the trafficking of E-cadherin as described above, the Rho GEF triple functioning domain protein (Trio) employs a different strategy, the regulation of E-cadherin transcription. Trio is a multi-domain protein that activates RhoG, Rac1 and RhoA, and is required for several cadherin-dependent developmental processes involving the activation of Rac1, such as defining cadherin-11 clusters in neurons, neural crest cell migration in Xenopus and M-cadherin-driven myoblast fusion (Backer et al., 2007; Charrasse et al., 2007; Kashef et al., 2009; Bach et al., 2010). Trio binds to the cytoskeletal protein Trio-associated repeat on actin (TARA, also known as TRIOBP) (Seipel et al., 2001) and, in epithelial cells, the constitutive inhibition of Trio through its interaction with TARA is necessary to maintain E-cadherin levels (Yano et al., 2011). In mature junctions in MDCK cells, TARA depletion relieves Trio inhibition, leading to an increase in Rac1 activation and the subsequent loss of E-cadherin from AJs (Yano et al., 2011). Instead of a local perturbation of actin filaments to destabilise adhesion, the TARA–Trio axis regulates E-cadherin at the transcriptional level, through the p38 mitogen-activated protein kinase pathway that activates the transcriptional repressor Tbx3 (Yano et al., 2011).

Another subset of GTPase regulators is involved specifically during the disassembly of junctions, which might or might not participate in junction dynamics at steady state. Instead, these regulators might be necessary in the context of the signalling cascades that are induced by oncogenes, treatment with growth factors or invasion of bacterial pathogens.

A number of intracellular pathogens interact with cell adhesion receptors for their uptake and subvert small GTPase signalling to enable invasion (Bonazzi and Cossart, 2011). Among them is Listeria monocytogenes, which binds directly to the extracellular domain of E-cadherin and triggers substantial modifications of actin filaments at junctions and in the cytoplasm to enable its engulfment and intracellular movements, respectively (Bonazzi and Cossart, 2011). The dynamic rearrangement of the actin cytoskeleton at AJs might occur by distinct mechanisms. First, the Listeria virulence protein InlC interacts directly with the Cdc42 GEF Tuba and prevents its association with N-WASP, thereby inhibiting the function of Tuba in reinforcing the tension at the sub-cortical cytoskeleton at cell–cell contacts, and thus facilitating cell-to-cell spread of Listeria (Rajabian et al., 2009). A second mechanism is through ARHGAP21 (formely known as ARHGAP10), which is a Cdc42, and RhoA GAP that interacts with α-catenin and localises to junctions in Caco-2 and JEG-3 cells (Sousa et al., 2005). Depletion of ARHGAP21 decreases the levels of α-catenin and prevents uptake of Listeria. Although ARHGAP21 does not appear to regulate the localisation of E-cadherin to junctions, it is likely that Listeria is able to use ARHGAP21 to perturb cadherin dynamics and facilitate pathogen entry. Together, these data suggest that ARHGAP21 is specifically required for engulfment of Listeria through GAP-dependent regulation of α-catenin localisation and cytoskeleton rearrangements.

The RhoA GEF neuroepithelial transforming proteins 1 isoform 1 (Net1) and isoform 2 (Net1A) can destabilise junctions depending on their localisation and protein levels (Carr et al., 2009; Papadimitriou et al., 2012). Both Net1 and Net1A are stabilised by cadherin-based adhesion through interaction with the protein Disc Large (DLG) in MCF-7 cells (Fig. 3) (García-Mata et al., 2007; Carr et al., 2009). This stabilisation leads to the nuclear translocation of Net1 or Net1A, thus preventing their proteosomal degradation (Carr et al., 2009). Depletion of Net1A in HaCaT cells perturbs the levels of E-cadherin and ZO1 at junctions and reduces RhoA activity. Transforming growth factor-β (TGF-β) is a tumour suppressor, but can also promote epithelial–mesenchymal transition (EMT) (Ikushima and Miyazono, 2010). Short-term exposure of HaCaT to TGF-β increases Net1A protein levels, which in turn activates RhoA and stabilises epithelial junctions under these experimental conditions (Papadimitriou et al., 2012). By contrast, prolonged exposure to TGF-β leads to degradation of cytoplasmic Net1 or Net1A, and disassembly of cell–cell contacts (Carr et al., 2009; Papadimitriou et al., 2012). Thus, depending on the level and duration of TGF-β treatment, TGF-β signalling can modulate the function of Net1 and Net1A to either stabilise cell–cell contacts or promote junction disassembly (Nakaya et al., 2008).

Distinct GAPs and GEFs have been shown to interfere with TJs, independently of their possible effect on E-cadherin-dependent adhesion. The appropriate function of these regulators at TJs requires a repertoire of interacting proteins that modulate their localisation or activity (Fig. 3).

The GAPs Cdc42 GTPase activating protein (CdGAP) and RhoGAP interacting with CIP4 homologs protein 1 (Rich1, also known as ARHGAP17) have been shown to regulate the maintenance of TJs by displacing polarity proteins through two distinct mechanisms (Wells et al., 2006; Togawa et al., 2010). CdGAP regulates TJ disruption in the context of hepatocyte growth factor (HGF)-stimulated scattering (Togawa et al., 2010). In MDCK cells, HGF treatment activates the extracellular-signal-regulated kinase (ERK) pathway, which phosphorylates and inactivates CdGAP (Tcherkezian et al., 2005), thereby increasing the available pool of active Cdc42. As active Cdc42 interacts with Par6, the imbalance in levels of active Cdc42 leads to the removal of Par6 from TJs. Mislocalisation of Par6 is rescued by expression of a non-phosphorylatable CdGAP mutant even in the presence of HGF (Togawa et al., 2010). Together with the fact that CdGAP levels are upregulated in transformed mammary cells (He et al., 2011), the results indicate that CdGAP can positively contribute to the invasive phenotype.

Rich1 is a GAP for Rac1 and Cdc42 that co-localises with TJ components at cell–cell contacts through an interaction with the scaffold protein Amot (Wells et al., 2006). Depletion of Rich1 leads to the mislocalisation of ZO1 and the polarity complex proteins Par3 (also known as PARD3) and Pals1 (also known as MPP5), but has no effect on the distribution of E-cadherin (Wells et al., 2006). Expression of GAP-deficient, but not wild-type Rich1, results in cells with reduced TER, indicating that Cdc42 inactivation is necessary for the phenotype. Interestingly, expression of Amot inhibits Rich1 GAP activity against Cdc42. Thus, Rich1 is required to regulate active Cdc42 levels in order to maintain TJ organisation and barrier integrity. When this regulation is lost (either by expression of Amot or depletion of Rich1), the molecular organisation of TJs and TER are severely compromised (Fig. 3).

GEF-H1 (also known as ARHGEF2), a RhoA and Rac1 GEF, is found inactivated at cell–cell contacts (Aijaz et al., 2005; Terry et al., 2011) through an interaction with cingulin and paracingulin (Aijaz et al., 2005; Guillemot et al., 2008). GEF-H1 interacts with microtubules and regulates cytokinesis and cell migration (Birkenfeld et al., 2008). In cells with mature junctions, GEF-H1 depletion does not perturb TJ organisation or morphology (Benais-Pont et al., 2003; Samarin et al., 2007) but, upon removal of Ca²⁺, GEF-H1 is required for disassembly of both AJ and TJ structures in SK-CO15 cells (Samarin et al., 2007). Interestingly, in spite of the lack of defects in the molecular and ultra-structure organisation of TJs, overexpression of GEF-H1 can selectively increase paracellular permeability to small molecular weight tracers but does not perturb TER (Benais-Pont et al., 2003). This regulation of paracellular permeability by GEF-H1 is likely to involve RhoA, as Rac1 is not activated by expression of GEF-H1 in MDCK cells (Benais-Pont et al., 2003).

Two distinct RhoA GEFs have been implicated in the organisation and contractility of peripheral circumferential filaments that are necessary for TJ function; p114RhoGEF (also known as ARHGEF18) and ARHGEF11 (also known as PDZ-RhoGEF) (Fukuhara et al., 1999). TJ maturation and barrier function are regulated by p114RhoGEF in Caco2 cells (Terry et al., 2011), but not in Eph4 cells (Itoh et al., 2012). Targeting of p114RhoGEF to mature junctions depends on its binding to the junctional protein cingulin and the polarity complex proteins Par3 and PatJ (Nakajima and Tanoue, 2011; Terry et al., 2011). p114RhoGEF also associates with the FERM-domain protein Lulu2, although this interaction is not required for its localisation at mature cell–cell contacts (Nakajima and Tanoue, 2011). Interestingly, association of Lulu2 increases the GEF activity of p114RhoGEF by an unknown mechanism (Fig. 3). Upon p114RhoGEF depletion, ZO1 staining appears warped and correlates with a loss of RhoA activation and myosin phosphorylation at cell–cell contacts (Terry et al., 2011). This suggests that p114RhoGEF is required to regulate RhoA activation at contacts. The latter in turn generates the necessary level of contraction for the maintenance of tension without which proper epithelial cell shape cannot persist (Terry et al., 2010; Nakajima and Tanoue, 2011).

ARHGEF11 interacts with ZO1 but is not crucial for the maintenance of mature junction architecture (although a minor effect is observed as increased cadherin levels at junctions and in detergent-soluble pools) (Itoh et al., 2012). ZO1-dependent ARHGEF11 positioning to cell–cell contacts and incorporation into primordial junctions (nascent AJs and TJs) is important for orchestrating the reorganisation of the actomyosin cytoskeleton as junctions mature. EpH4 cells without ARHGEF11 form junctions more slowly than control cells, with a delay in remodelling of Myosin IIb and reduced TER after 36 hours (Itoh et al., 2012). Taken together, these results suggest at least two distinct mechanisms to regulate RhoA activity that impinge on the function of TJs; positioning of p114RhoGEF through the binding of cingulin and of ARHGEF11 through the binding of ZO1 (Terry et al., 2011; Itoh et al., 2012). Both pathways control tension and TJ-dependent remodelling of peripheral filaments during TJ maturation to enable appropriate regulation of epithelial permeability.

The Rac1 GEF T-cell lymphoma invasion and metastasis 1 (Tiam1) regulates a variety of processes that are necessary for the maturation and stabilisation of both AJs and TJs. This property is counterintuitive, and even considering variations in experimental cell conditions suggests that a single GEF operates downstream of distinct adhesive structures.

Tiam1 is not essential for maintenance of AJs, as Tiam1-null mice are viable and epidermal morphogenesis occurs normally (Mertens et al., 2005). Consistent with a role in reinforcement of cadherin adhesion, depletion of Tiam1 can reduce the levels of E-cadherin at MDCK junctions (Malliri et al., 2004). However, expression of Tiam1 blocks the disruption of junctions that is induced by oncogenic signals (Fig. 2) downstream of H-Ras transformation (Hordijk et al., 1997) or HGF-induced scattering of MDCK cells (Sander et al., 1998), and is required for E1A-induced mesenchymal–epithelial transition in fibroblasts (Malliri et al., 2004). These data suggest that appropriate Rac1 signalling is a key component of the dedifferentiation programme during malignancy. Another example is the activation of the non-receptor tyrosine kinase Src, which potently perturbs cell–cell adhesion in a Tiam1- and Rac1-dependent manner (Palovuori et al., 2003; Woodcock et al., 2009). Under these conditions, phosphorylated Tiam1 is specifically targeted for degradation, implying that Tiam1 counteracts the destabilisation of contacts that is induced by Src activation (Woodcock et al., 2009). These roles of Tiam1 illustrate that the activity of a single Rho GEF can prevent the disruption of cell–cell adhesion by different stimuli.

The requirement for Tiam1 at TJs is somewhat controversial, with results showing that Tiam1 promotes or inhibits TJ formation and function. Epidermal keratinocytes or MDCK cells depleted of Tiam1 are defective in Rac1 activation and TER development following junction assembly (Mertens et al., 2005; Guillemot et al., 2008). A potential mechanism is that Tiam1–Rac1 signalling activates Par3 to drive TJ biogenesis (Mertens et al., 2005). However, other studies also in MDCK cells, suggest that Par3 downregulates Tiam1 activity and thus negatively regulates TJ assembly (Chen and Macara, 2005; Mack et al., 2012).

A solution for these discrepancies might be found in the binding partners that localise Tiam1 at cell–cell contacts. Paracingulin (Guillemot et al., 2008) and β2-syntrophin (Mack et al., 2012) interact with and promote the activation of Tiam1, whereas the interaction with Par3 inhibits the GEF activity of Tiam1 (Chen and Macara, 2005). Interestingly, Par3 and β2-syntrophin are differentially localised along the apico-basal axis of the lateral domains and thus provide a gradient of Rac activity at epithelial junctions (Mack et al., 2012). When apico-basal positioning of Tiam1-dependent Rac activation is lost, increased apical Rac activation leads to impaired development of TER in MDCKII cells (Mack et al., 2012). The Tiam1-mediated gradient of Rac1 activation can potentially also explain the differential effects on AJs or TJs, depending on the degree of segregation of these adhesive structures during junction biogenesis.

In addition to triggering signalling that feeds back into the reinforcement and maintenance of cell–cell contacts, adhesive receptors might influence the localisation of GEFs and GAPs and how they regulate other cellular processes that are important for epithelial growth, survival and differentiation. Following junction dissociation in embryonic stem cells, active BCR-related protein (ABR) is activated to regulate cell rounding, blebbing and apoptosis. These processes require both its GAP and GEF catalytic domains to balance the levels of active Rac1 and RhoA, respectively (Ohgushi et al., 2010). In addition to its regulation of TJ function, GEF-H1 participates in the inhibition of proliferation that is driven by junctions and confluence (Aijaz et al., 2005; Birkenfeld et al., 2008). GEF-H1 is sequestered to and inactivated at mature cell–cell contacts by binding to cingulin (Fig. 3), thereby reducing the levels of active RhoA (Benais-Pont et al., 2003; Aijaz et al., 2005; Citi et al., 2009). As active RhoA promotes proliferation in different cell types, GEF-H1 is an excellent example of how a specific intracellular localisation does not necessarily imply the direct activation of Rho GTPases and the local regulation of a process.

Attachment and spreading of cells onto plates that are coated with the extracellular domains of E-cadherin or Nectin can be used as a simplified model for initial adhesion that mimics transinteraction of adhesion receptors (Fig. 2). In this context, it has been shown that E-cadherin- or Nectin-dependent cell attachment and spreading triggers the activation of Src, which in turn phosphorylates and activates VAV2 (Fig. 3) (Kawakatsu et al., 2005; Fukuyama et al., 2006), a GEF for RhoA, Rac1 and Cdc42. Activation of VAV2 increases the activation of Rac and formation of lamellipodia as cells spread onto the E-cadherin substratum. However, direct evidence that either overexpression of VAV2 or its depletion interferes with E-cadherin stability at junctions is lacking. Nevertheless, earlier studies using non-epithelial cells show that VAV2 interacts with p120^CTN (Noren et al., 2000; Chauvet et al., 2003), suggesting that VAV2 is indirectly recruited to E-cadherin-mediated adhesion.

A cDNA expression screen identified ARHGAP12 as a component of cell–cell junctions in epithelial cells in different tissues (Matsuda et al., 2008). ARHGAP12 is a Rac1-specific GAP (Gentile et al., 2008) and is important for cell scattering, a process that requires Rac1. ARHGAP12 depletion enhances cell scattering triggered by HGF, and conversely, ARHGAP12 overexpression inhibits it (Gentile et al., 2008). Mechanistically, ARHGAP12 is transcriptionally repressed in response to HGF, providing an example of downregulation at the mRNA level. Interestingly, overexpression of ARHGAP12 does not interfere with cell–cell adhesion in MDCK (Gentile et al., 2008) or Eph4 epithelial cells (Matsuda et al., 2008), suggesting that ARHGAP12 regulates levels of active Rac1 that are required for migration and invasion, rather than in junction maintenance.

Much progress has been made since Rho GTPases were first implicated in the regulation of cadherin-mediated cell–cell adhesion. Yet, the evidence generated to date only scratches the surface of how Rho proteins are regulated by GEFs and GAPs at cell–cell contacts. An exciting route that is now being taken is to carefully dissect the precise cellular process in which a GEF or GAP operates, how it is recruited and modulated at sites of cell–cell contacts, and how its activity can be integrated in the context of macromolecular organisation, junction biogenesis and stabilisation. Recent developments add new challenges to this journey, such as the existence of alternative mechanisms to spatially fine-tune the function of a GEF or GAP, the GTPase cascades and the involvement of cytoskeletal proteins to recruit, retain and/or modulate the activity of different regulators at junctions. Finally, although stability of cell–cell contacts relies heavily on the appropriate regulation of GEF and GAP functions, conversely, junctions modulate the activities of GEFs and GAPs to drive contact-dependent cellular events, such as inhibition of proliferation and apoptosis, thus fully connecting cell–cell adhesion to cell specification and function. The multiple levels underlying this complexity are yet to be fully unravelled, but they do make for a trip that is well worth taking.

The authors declare no competing interest. Due to space restrictions, we apologise for not being able to cite all relevant literature.