Rac1 acts in conjunction with Nedd4 and dishevelled-1 to promote maturation of cell–cell contacts

Cell–cell adhesion is essential for tissue integrity and for the barrier function of epithelia and endothelia. Cellular contacts are formed by homotypic interactions of specialized cell-adhesion molecules such as the cadherins. Such contacts are subject to tight regulation and can be very dynamic (Dejana, 2004; Hartsock and Nelson, 2008; Watanabe et al., 2009). In addition, cell–cell contacts play an important role in the spatial organization of signalling proteins and are key determinants of cell polarity (Iden and Collard, 2008; Mertens et al., 2005). Conversely, loss of cell–cell contact is required for cell division, tissue remodelling and efficient cell motility. The (dys)regulation of junctional integrity therefore represents a key feature of cellular transformation and of metastatic capacity of tumour cells (Cavallaro and Christofori, 2004).

The formation of cell–cell contact between neighbouring cells requires the initiation of nascent intercellular adhesions that subsequently mature to strong cell–cell junctions. This process is controlled by the microtubule (MT) and actin cytoskeleton which govern vesicular traffic and internalisation as well as the anchoring of cell–matrix and cell–cell adhesion complexes (Harris and Tepass, 2010). A large number of regulatory proteins orchestrate MT and actin dynamics, of which the Rho-like GTPases are the most extensively studied.

Activation of RhoA, Rac1 and CDC42 can regulate loss as well as formation of cell–cell contacts and several downstream effectors such as Rho Kinase and IQGAP have been implicated in these events (Braga et al., 1997; Hordijk et al., 1997; Kuroda et al., 1997; Kuroda et al., 1998). The stimulation of actin polymerization by Rac1, through its activation of the Arp2/3 complex, is a key aspect of its capacity to initiate cell–cell contact as a result of membrane protrusion (Watanabe et al., 2009). However, specific molecular mechanisms that allow Rac1 to regulate intercellular adhesions have remained largely elusive.

In addition to post-translational modifications such as phosphorylation or acetylation, protein function and targeting are regulated by the conjugation of ubiquitin or ubiquitin-like proteins (Grabbe et al., 2011). Ubiquitylation generally occurs at specific lysine residues within the target proteins, is reversible and therefore, like phosphorylation, a dynamic and tightly regulated modification. Addition of a single ubiquitin (mono-ubiquitylation) can affect protein localisation (e.g. through internalization or through binding to other proteins), whereas addition of consecutive ubiquitins (poly-ubiquitylation) can act as a signal for proteasomal degradation (d’Azzo et al., 2005; Grabbe et al., 2011; Haglund et al., 2003; Pickart, 1997). Ras and RhoGTPases are subject to regulation by ubiquitylation (Nethe and Hordijk, 2010). Importantly, localised RhoA ubiquitylation by Smurf promotes cell polarization and directional migration (Boyer et al., 2006; Sahai et al., 2007; Wang et al., 2003). Activated Rac1 is also subject to ubiquitylation and degradation (Doye et al., 2002; Pop et al., 2004). Recently, our lab showed that poly-ubiquitylation of endogenous, active Rac1 is regulated by adhesion to fibronectin and by the Rac1-binding adapter protein caveolin-1 (Nethe et al., 2010).

In this study, we describe the identification of a newly identified Rac1-interacting protein, the ubiquitin ligase Nedd4 [neuronal precursor cell expressed and developmentally downregulated protein 4-1 (Kumar et al., 1992; Kumar et al., 1997)]. We found that loss of Nedd4 reduces epithelial cell–cell contact and transepithelial resistance. In addition, we could show that Nedd4, in conjunction with activated Rac1, promotes the ubiquitylation and degradation of the adapter protein dishevelled-1 (Dvl1), a negative regulator of epithelial cell–cell contact (Elbert et al., 2006). Together, our data identify a novel pathway downstream of Rac1, which enhances epithelial integrity through Nedd4 mediated degradation of Dvl1.

We and others previously showed that the hypervariable Rac1 C-terminal domain associates to protein and lipid kinases, as well as to the Rac1-GEF β-PIX and a series of adapter proteins (Modha et al., 2008; ten Klooster et al., 2006; ten Klooster et al., 2007; Tolias et al., 1998; van Duijn et al., 2010; van Hennik et al., 2003; de Kreuk et al., 2011; Nethe et al., 2010). Here we report our analysis of a novel Rac1 interactor identified in the same proteomic screen, the HECT (homologous to the E6-AP carboxyl terminus) ubiquitin ligase Nedd4 (Fig. 1A).

Following its initial identification, pull-down experiments using lysates from HeLa cells and peptides encoding the C-termini of a series of RhoGTPases showed that endogenous Nedd4 associated to the Rac1 C-terminus whereas binding to the C-termini of other RhoGTPases was either weak (e.g. Rac3) or undetectable (Fig. 1B). Use of peptides encoding mutations within the Rac1 C-terminal domain (van Hennik et al., 2003) subsequently showed that the polybasic region, rather than the proline stretch, mediates Rac1 C-terminal binding to Nedd4 (Fig. 1C). Conversely, a GST–Rac1ΔC fusion protein, lacking the Rac1 C-terminal domain, showed reduced binding to Nedd4 as compared to full-length GST–Rac1 (Fig. 1D). This shows that the Rac1 hypervariable C-terminus is both necessary and sufficient for its association to Nedd4. Confirmation of the association between endogenous Rac1 and endogenous Nedd4-1 by co-immunoprecipitation was unsuccessful, most likely due to low levels of the endogenous pools of either protein participating in the interaction and the fact that the Rac1 antibody recognizes the C-terminal hypervariable domain, which mediates the association to Nedd4.

Nedd4 harbours a calcium- and lipid-binding C2 domain in its N-terminus that mediates membrane targeting (Plant et al., 1997; Wang et al., 2010). Pull-down assays using lysates of HeLa cells, expressing wild-type (WT) Nedd4 or a mutant lacking the C2 domain (Nedd4ΔC2; Fig. 1E), showed that both bound the Rac1 C-terminus, indicating that the C2 domain is dispensable for this interaction. Subsequent experiments using full-length GST–Rac1 showed that the central region of Nedd4, encoding the four WW domains, is sufficient for Rac1 binding (Fig. 1F). Together, these data show that Nedd4 associates to Rac1 in a fashion that requires the Rac1 hypervariable C-terminal domain and the Nedd4 WW domains.

Immunoprecipitation of inactive (Rac1N17T) and active (Rac1Q61L) mutants, co-expressed in HeLa cells with HA-tagged Nedd4 showed that Nedd4 binds to both Rac1 mutants, albeit more efficiently to active Rac1 (Fig. 2A). Additional experiments in which Rac1 was expressed at levels below those of endogenous Rac1, confirmed these results, excluding artefactual binding due to Rac1 overexpression (supplementary material Fig. S1A). We were not able, however, to show ubiquitylation of Rac1 by Nedd4 (data not shown), suggesting that Rac1 is not a substrate for Nedd4 ubiquitin ligase activity. This result is in good agreement with two recent studies that identified HACE-1 as well as XIAP, but not Nedd4, as ubiquitin ligases for activated Rac1 (Oberoi et al., 2012; Torrino et al., 2011).

Subsequent analysis by confocal microscopy showed that endogenous Nedd4 colocalises with Rac1Q61L at cell–cell contacts and peripheral membrane ruffles (Fig. 2B). We could confirm such junctional colocalisation also with WT-Rac1, but not with Rac1T17N (supplementary material Fig. S1B). Activated Rac1 also localises to Focal Adhesions (FAs) (Nethe et al., 2010), but Nedd4 could not be detected at these sites (Fig. 2B). Interestingly, analysis of Nedd4 together with β-catenin showed that Nedd4 localised at mature, well-organized junctions but was absent from immature cell–cell contacts, which are characterised by a more diffuse band of β-catenin staining (Fig. 2C).

To test if Nedd4 contributes to the formation of epithelial cell–cell contacts, we used lentiviral shRNAs to deplete endogenous Nedd4 from HeLa cells and from H292 lung epithelial cells (Fig. 2D, and data not shown). Reduction of Nedd4 expression impaired the formation of discrete cell–cell contacts in HeLa cells (Fig. 2E) as well as in H292 cells (Fig. 2F). In both epithelial cell types, proper formation of adherens and tight junctions, marked by β- and γ-catenin and by ZO-I, respectively, was impaired upon reduction of Nedd4 expression (Fig. 2E–G, and data not shown). Similar effects on HeLa cell–cell contacts were found when cells were transfected with siRNA to Nedd4 (supplementary material Fig. S2A,B). Conversely, re-expression of shRNA-resistant murine Nedd4 rescued the formation of more mature junctions as detected by the linear distribution of β-catenin (Fig. 2H; supplementary material Fig. S2C).

To confirm the relevance of Nedd4 for epithelial integrity, H292 lung epithelial cells, transfected with shRNA to reduce Nedd4 expression, were seeded on gold electrodes, and transepithelial resistance (TER) was recorded by electrical cell-substrate impedance sensing (ECIS) (Lorenowicz et al., 2007). In this assay, cells are seeded in sufficient numbers to cover the electrode during cell spreading and TER is recorded continuously in a non-invasive fashion. The increase in TER in the first 2–10 hrs after seeding (depending on the cell type) reflects the speed and degree of cell spreading (Mitra et al., 1991). TER at later time-points represents junctional integrity. ShRNA-mediated loss of Nedd4 resulted in increased cell spreading, but TER in the second phase of the experiment was significantly lower as compared to the controls, indicative for reduced monolayer integrity (Fig. 3A), in line with the data in Fig. 2. Similarly, transfection of the cells with siRNA to Nedd4 also induced a significant reduction in TER (supplementary material Fig. S2D). In a complementary experiment, we expressed full-length human GFP–Nedd4 in HeLa cells, FACS-sorted the GFP-positive cells and measured TER following cell seeding by ECIS. Expression of GFP–Nedd4 increased the TER of the monolayer relative to the controls, indicative for the induction of strong cell–cell contacts and concomitant junctional maturation (Fig. 3B).

To analyse the functional relevance of Nedd4 further, we used a calcium-switch assay. EGTA-mediated depletion of extracellular calcium resulted in a loss of stable cell–cell contacts in H292 lung epithelial cells, as concluded from a broad band of β-catenin staining, and a loss of Nedd4 from intercellular junctions (Fig. 3C). One hour after EGTA washout and re-addition of calcium, β-catenin but not Nedd4, concentrated at a subset of intercellular contacts. At 5 hours after calcium re-addition, junctions were reformed and both β-catenin and Nedd4 concentrated at intercellular junctions. Together with the ECIS data in Fig. 3A,B, these experiments suggest that Nedd4 concentrates at mature rather than at immature cell–cell contacts acting, in conjunction with activated Rac1, as part of a positive feedback loop that promotes junctional maturation.

Because Nedd4 is a ubiquitin ligase, we subsequently screened potential Nedd4 substrates, identified by a PPXY or RXXQE motif (Persaud et al., 2009), that could function downstream of Nedd4 in the control of epithelial cell–cell contacts. We here focus on the adapter protein Dvl1, which negatively regulates cell–cell adhesion in epithelial MDCK cells (Elbert et al., 2006) and harbours a PPXY motif downstream of its DEP (dishevelled, Egl-10 and pleckstrin homology) domain (Fig. 4A). To confirm the negative effects of Dvl1 expression on cell–cell contact in HeLa cells, we expressed a Dvl1–GFP fusion protein and analysed the reformation of cell–cell contacts using a calcium-switch assay. Following loss of cell–cell contact by EGTA-mediated calcium depletion, restoration of junctional localisation by the re-addition of calcium, marked by the localisation of β-catenin, was impaired in cells that expressed Dvl1–GFP (Fig. 4B). This confirms the notion that Dvl1 expression negatively regulates epithelial integrity. In line with these findings, siRNA-mediated reduction of Dvl1 expression showed a small but significant increase in TER (supplementary material Fig. S3A,B).

To analyse the regulation of Dvl1 by Nedd4, we examined the levels of endogenous Dvl1 in GFP–Nedd4-transfected HeLa cells treated for 6 hrs with cycloheximide to block protein synthesis. Expression of Nedd4 reduced the levels of endogenous Dvl1, but not Dvl2 or Dvl3 (Fig. 4C). Conversely, cells depleted of Nedd4 showed an increase in endogenous Dvl1 without any effect on the levels of Dvl2 and Dvl3 (Fig. 4D). The shRNA-mediated loss of Nedd4 did not affect either the levels or nuclear translocation of β- and γ-catenin (Fig. 2E,F, Fig. 4D). This suggests that Nedd4-controlled stability of Dvl1 does not alter β-catenin signalling, as was shown for Dvl during canonical Wnt signalling (Gao and Chen, 2010).

Expression of Nedd4, but not of Nedd4-2 (Persaud et al., 2009), stimulated the ubiquitylation of Dvl1–HA (Fig. 4E). In contrast, ubiquitylation of Dvl1 remained unaffected upon inhibition of lysosomal degradation by chloroquine (Fig. 4E), a pathway that was recently implicated in the control Dvl2 degradation (Gao et al., 2010; Su et al., 2007). Together, these findings show that Nedd4 promotes the ubiquitylation and proteasome-dependent degradation of Dvl1.

To investigate the interaction between Nedd4 and Dvl1, we made GFP-tagged Nedd4 truncation constructs. GST–Dvl1, but not free GST, associated to the Nedd4 WW region (amino acids 196–505; Nedd4WW), but not the C2 domain (amino acids 1–134; Nedd4C2; Fig. 5A). In parallel, we analysed the localisation of these different Nedd4 domains (supplementary material Fig. S4). Although the C2 domain is important for Nedd4 membrane-targeting, the isolated GFP–C2 domain concentrated mostly at perinuclear vesicles. The WW domains, but not the HECT domain, localised also to perinuclear vesicles whereas both constructs were also found at the peripheral membrane, similar to the full-length protein (supplementary material Fig. S4).

To further map the association of Dvl1 with Nedd4 and test the relevance of the interaction for Dvl1 ubiquitylation, we generated FLAG-tagged Dvl1 constructs lacking the DIX domain (ΔDIX) and the PDZ domain (ΔPDZ), one lacking both the DEP domain and C-terminal region (ΔDEP-tail) and one lacking only the C-terminal region (Δ-tail). In addition, we generated two complementary Dvl1 protein fragments, the first comprising the C-terminus including the DEP domain (DEP-tail) and a second that extends from the PDZ domain to the C-terminus (PDZ-tail; Fig. 5B).

We first analysed the localisation of these Dvl1 constructs by confocal imaging (supplementary material Fig. S5, Fig. S6A). In line with published data (Schwarz-Romond et al., 2005), full-length Dvl1 localised to puncta that were dispersed in the cytosol (supplementary material Fig. S5). The different deletion and truncation constructs were found in various locations throughout the cell, including the perinuclear region, the nucleus, and to a varying extent associating with cytosolic puncta (supplementary material Fig. S5, Fig. S6A). Pull-down assays showed that GST–Nedd4 binds to full-length Dvl1, as well as the PDZ-tail and DEP-tail fragments of Dvl1 (Fig. 5B). However, Nedd4 did not associate to a Dvl1 protein that lacked the DEP domain and C-terminal region (ΔDEP-tail; Fig. 5B) indicating that this region of Dvl1 mediates its association with Nedd4.

We next analysed the ubiquitylation of Dvl1 by Nedd4. Dvl1 constructs lacking the DIX (ΔDIX) or PDZ (ΔPDZ) domains were ubiquitylated by Nedd4 (Fig. 5C). In contrast, removing the DEP domain and the adjacent C-terminal region (ΔDEP-tail) blocked ubiquitylation by Nedd4 (Fig. 5C). This is in line with our observation that this region is required for binding to Nedd4 (Fig. 5B). Analysis of the Dvl1Δtail protein, which does encode the DEP domain, but lacks the adjacent C-terminal region, showed that this protein was ubiquitylated by Nedd4 (Fig. 5D). This underscores the requirement of the DEP domain for ubiquitylation by Nedd4, likely because the DEP domain is required for Nedd4 binding. Nedd4 substrates are characterized by a PPXY motif. The Dvl1 C-terminus comprises the PPXY motif (Fig. 4A). Although in a co-transfection experiment, the Dvl1 C-terminal portion is dispensable for Dvl1 ubiquitylation (Fig. 5D), mutation of the PPXY motif does reduce efficient ubiquitylation of Dvl1 (Fig. 5E). This finding suggests that the PPXY motif is required, in conjunction with the DEP domain, for efficient Nedd4-dependent ubiquitylation of Dvl1.

Although we showed the importance of the DEP domain and adjacent C-terminal region for binding to Nedd4, this part of Dvl1 is not ubiquitylated by Nedd4 (Fig. 5D). In addition, neither the Dvl DIX nor PDZ domains encode the lysine residue(s) required for ubiquitylation (Fig. 5C). To look into this further, we assayed Dvl1 degradation induced by co-expression of Nedd4 (Fig. 5F). We co-expressed Nedd4 with either wtDvl1, the ΔDEP-tail protein that is not ubiquitylated, a Dvl1 mutant with a deletion of the N-terminal region (Dvl1 PDZ-tail) or full-length Dvl1 with mutations of the three lysine residues in between the DIX and PDZ domains (K118, K220, K225) to arginines (Dvl1-3KR). We found that, despite that fact that these Dvl1 proteins are expressed at different levels, Nedd4-induced loss in wt Dvl1 protein levels are not observed with any of these mutants (Fig. 5F). These data are in line with analysis of Nedd4-mediated ubiquitylation of these mutants (supplementary material Fig. S6B). The three lysines appeared to be redundant as individual mutation of these residues did not prevent Nedd4-mediated ubiquitylation (data not shown).

To further underscore the biological relevance of these lysine residues, we transfected cells with wt Dvl1 and the Dvl1 3KR mutant and analysed consequences for TER. The data in Fig. 5G show that wt Dvl1, but in particular the Dvl1 3KR mutant reduces junction maturation, in line with the data in Figs 2–f03,4 indicating that Nedd4-mediated degradation of Dvl1 serves to sustain and promote cell–cell adhesion.

In summary, these data show that Nedd4 binds, through its WW domains, to the DEP domain of Dvl1 and subsequently ubiquitylates a lysine-rich region between the DIX and PDZ domain, which results in Dvl1 degradation (Fig. 5G).

Expression of GFP-tagged Nedd4 together with Dvl1–HA showed that Dvl1 and Nedd4 colocalised in puncta that are detectable upon Dvl1 expression (Fig. 6A; supplementary material Fig. S5) and that colocalise with β-catenin at cell–cell contacts (Fig. 6A). Junctional colocalisation of Dvl1–HA and GFP–Nedd4 in areas lacking such puncta was also observed (supplementary material Fig. S7A). Endogenous Dvl1 was, like Dvl1–HA, found in puncta and on cell–cell contacts that colocalised with GFP–Nedd4 (Fig. 6B, Fig. 7C). Thus, similar to the Drosophila Dvl1 homologue Dsh, a portion of human Dvl1 localises at epithelial cell junctions (Bastock et al., 2003). Unfortunately, the available antibodies did not allow analysis of colocalisation of endogenous Dvl1 with endogenous Nedd4. As an additional control, we also co-transfected GFP–Nedd4-2 with Dvl1–HA. However, GFP–Nedd4-2 did not localise to the Dvl1–HA-induced puncta, resulting in minimal colocalisation, which is in agreement with the lack of Dvl1 ubiquitylation by Nedd4-2 (supplementary material Fig. S7B; Fig. 4E).

To test whether junctional localisation of Nedd4 is essential for Dvl1 ubiquitylation, we used the Nedd4 mutant lacking the C2 domain (ΔC2-Nedd4) that does not localise to cell–cell contacts, even in the presence of activated Rac1Q61L (Fig. 6C; supplementary material Fig. S7C) (Plant et al., 1997; Wang et al., 2010). The ΔC2-Nedd4 protein, similar to the catalytically inactive Nedd4C867S (Persaud et al., 2009), did not ubiquitylate Dvl1 (Fig. 6D). Additional studies showed that the ΔC2-Nedd4 protein did colocalise with Dvl1-positive puncta in the cytosol (supplementary material Fig. S7D), indicating that it is the targeting of Nedd4 to the plasma membrane, rather than to cytosolic puncta, which is required for Dvl1 ubiquitylation. This is in good agreement with our finding that Dvl1 ubiquitylation by Nedd4 requires the DEP domain (Fig. 5C), as this region mediates the association of Dvl1 with the plasma membrane (Wong et al., 2000).

Because Nedd4 and active Rac1 colocalise at cell–cell contacts (Fig. 2B), we tested whether Rac1 activity could stimulate Nedd4 activity. Co-expression of Rac1Q61L with GFP–Nedd4 promoted ubiquitylation of a series of endogenous proteins, the detection of which was further increased by blocking proteasomal degradation with MG132 (Fig. 7A). In addition, co-expression of an active (Rac1Q61L), but not an inactive (Rac1N17T) mutant of Rac1 with GFP–Nedd4, Dvl1–HA and His-tagged ubiquitin showed a clear increase in ubiquitylation of Dvl1–HA in the presence of GFP–Nedd4 (Fig. 7B). Conversely, the Rac1 inhibitor EHT1864 (Shutes et al., 2007) induces a reduction of Nedd4-mediated ubiquitylation of Dvl1 (supplementary material Fig. S8A). Similar to endogenous Nedd4, endogenous Rac1 colocalised with endogenous Dvl1 at the peripheral membrane of single cells (supplementary material Fig. S8B), at cell–cell junctions and at a fraction of Dvl1-positive puncta (Fig. 7C). Together, this data support a model in which active Rac1 stimulates Nedd4-mediated degradation of Dvl1 at intercellular junctions to promote epithelial cell–cell contact.

The current study describes a novel pathway by which Rac1 stimulates maturation of cell–cell contacts in epithelial cells. The pathway comprises Rac1-triggered translocation of the HECT ubiquitin ligase Nedd4 to cell–cell junction, and Nedd4-stimulated ubiquitylation and degradation of the scaffold protein Dvl1, a negative regulator of epithelial cell–cell contact (Elbert et al., 2006). These findings are summarised in a model, depicted in Fig. 8.

We and others previously identified a diverse array of proteins that bind to the hypervariable C-terminal domain in Rac1. These include the GEF β–PIX, adapter proteins such as Caveolin1, CD2AP, and PACSIN2, the nuclear oncogene SET/I2PP2A, PIP-5-Kinase, the Rac1-effector PRK and most recently mTOR (de Kreuk et al., 2011; Modha et al., 2008; Nethe et al., 2010; Saci et al., 2011; ten Klooster et al., 2006; ten Klooster et al., 2007; Tolias et al., 1998; van Duijn et al., 2010; van Hennik et al., 2003). In contrast to effector proteins that only bind to GTP-bound Rac1, most of these interactions are nucleotide-independent. Yet, activated Rac1 recruits Caveolin1 to focal adhesions and CD2AP to cell–cell contacts (Nethe et al., 2010; van Duijn et al., 2010). Nedd4 appears to respond in a similar way, accumulating at mature cell–cell contacts in cells expressing activated Rac1. Our present results indicate that Nedd4 in fact co-operates with Rac1 in junctional maturation since loss of Nedd4 resulted in immature cell–cell contacts and reduced TER.

Loss of Nedd4 ligases does not appear to affect epithelial cell–cell contacts in mouse embryos. Deficiency of Nedd4 leads to elevated levels of thrombospondin, an inhibitor of angiogenesis, and consequent heart defects and embryonic lethality (Fouladkou et al., 2010). In addition, loss of Nedd4 results in impaired formation of the neuromuscular junction, a specialized type of heterotypic cell–cell contact (Liu et al., 2009). Mice deficient for Nedd4-2 show a different phenotype. Here, loss of Nedd4-2 increases the expression of the sodium channel ENaC, resulting in impaired lung function and perinatal death (Boase et al., 2011). It might be that the epithelial phenotype of Nedd4 deficiency that we report here becomes more apparent after birth or that Nedd4 and Nedd4-2 are functionally redundant in the regulation of epithelial junctions in vivo. The latter notion is supported by a recent study (Van Campenhout et al., 2011).

Several mechanisms have been identified that regulate the activity of Nedd4 ligases. Nedd4 binds to adapter proteins, including Annexin XIIIb, Ndfip 1,2 and Grb10 that serve to recruit Nedd4 to specific membrane domains (e.g. lipid rafts), organelles (e.g. Golgi) or substrates (e.g. the IGF-1R) (Shearwin-Whyatt et al., 2006). In addition, Nedd4-2 can be serine phosphorylated, generating a binding site for 14-3-3 proteins that interfere with binding to ENaC, thereby regulating the expression of the sodium channel (Debonneville et al., 2001; Ichimura et al., 2005). The auto-inhibitory function of the C2 domain of Nedd4-2 is released upon binding to calcium, activating the ubiquitin ligase (Plant et al., 1997). The C2 domain of Nedd4 is required for targeting to cell–cell contacts (Fig. 6C), cooperating with Rac1 binding to the WW domains which occurs independent of the C2 domain (Fig. 1E,F). The unfolding of Nedd4 upon binding to junctional membranes may disrupt the interaction between the C2 and the catalytic HECT domain and activate Nedd4. We found that the C2 domain is required for the ubiquitylation of Dvl1 (Fig. 6D), in line with recent data underscoring its role in substrate specificity of the Smurf1 ubiquitin ligase (Lu et al., 2011). Thus, Rac1-stimulated accumulation of Nedd4 at intercellular junctions is a likely mechanism to enhance its ubiquitin ligase activity, in good agreement with the results in Fig. 7.

We found that Rac1 activity promotes Dvl1 ubiquitylation by Nedd4, but not Nedd4-2. This occurs likely through Lys48 poly-ubiquitylation, as Nedd4 expression results in proteasome-dependent degradation of Dvl1 protein. Nedd4 is not the only ubiquitin ligase for Dvl1 as the HECT ligase NEDL1 can ubiquitylate Dvl1 as well (Miyazaki et al., 2004), comparable to inversin and KLHL12 (Tauriello et al., 2010; Tauriello and Maurice, 2010). Dvl1 degradation at the plasma membrane is promoted by the adapter protein NKD2 (Hu et al., 2010), an antagonist of Wnt signalling. It is currently unclear whether NKD2 is involved in the Nedd4-mediated Dvl1 degradation. Conversely, we recently showed that the de-ubiquitinating enzyme CYLD mediates removal of Lys63-conjugated ubiquitin on Dvl1, which also affects Wnt signalling (Tauriello et al., 2010).

Although Dvl is an established regulator of cell polarity and cell–cell contact through its control of β-catenin-dependent, canonical Wnt signaling, there are several studies that link Dvl to cell–cell contact via additional signalling pathways (Schlessinger et al., 2007; Yamanaka and Nishida, 2007). Schlessinger and colleagues (Schlessinger et al., 2007) showed in rodent embryo fibroblasts that Dvl activity is downstream of the loss of cell–cell contact, induced following non-canonical Wnt5a-induced signaling. In epithelial MDCK cells, Dvl1 was found to negatively regulate cell–cell contact by reducing the association of E-cadherin with the actin cytoskeleton (Elbert et al., 2006). This effect was independent of canonical Wnt signalling. In our studies, Nedd4-mediated loss of Dvl1 did not correlate with any changes in expression levels of β-catenin, nor its translocation to the nucleus (Fig. 2E,F, Fig. 4D,E), suggesting that Nedd4 does not activate β-catenin-dependent Wnt signalling.

Although the notion that Rac1 activity enhances E-cadherin-based cell–cell adhesion dates back to work published by several groups already in 1997 (Braga et al., 1997; Hordijk et al., 1997; Takaishi et al., 1997), the underlying molecular mechanisms have not been addressed in much detail. The Rac1 effector IQGAP1 has been implicated as an important regulator, but its control of cell–cell adhesion is complicated as IQGAP1 can promote as well as inhibit E-cadherin function (Noritake et al., 2005). It is generally accepted that Rac1-stimulated, Arp2/3-mediated actin polymerization mediates Rac1- and E-cadherin dependent cell–cell contact (Noritake et al., 2005). In addition to actin, also MTs control formation and stabilization of cell–cell contact (Harris and Tepass, 2010). Dvl1 has previously been linked to the regulation of MT dynamics (Krylova et al., 2000) suggesting a potential functional connection. However, regulation of cell–cell contacts through the MT cytoskeleton is cell-type specific and this issue was not further addressed here. To which extent the actin cytoskeleton and MTs cooperate in the regulation of epithelial integrity downstream of the Rac1–Nedd4-1–Dvl1 pathway therefore remains to be established.

There is increasing evidence linking the cellular ubiquitylating machinery to both positive and negative control of cell–cell adhesion. The ubiquitin ligase Hakai is capable of triggering the ubiquitylation, internalization and degradation of E-cadherin, thereby weakening cell–cell contacts (Fujita et al., 2002). In mammary epithelial cells, the ubiquitin ligase Cbl regulates maintenance of adherens junctions by ubiquitylating EGFR-stimulated Vav2 (Duan et al., 2011), leading to a loss of cell–cell contact. In endothelial cells, the CCM2 (Cerebral Cavernous Malformation 2) protein has been suggested to regulate integrity and development of the vasculature through its stimulatory effect on the ubiquitylation of RhoA by Smurf1 (Crose et al., 2009; Whitehead et al., 2009). In good agreement with our current findings, a recent study showed that siRNA-mediated loss of Nedd4 ubiquitin ligases in epithelial cells results in impaired cell–cell contacts and discontinuous tight junctions (Van Campenhout et al., 2011). Thus, findings from several groups using different cell types support our current findings that local ubiquitylation and degradation of proteins represents a key regulatory mechanism in the control of intercellular junctions.

HeLa and H292 lung epithelial cells were maintained in Iscove’s Modified Dulbecco’s Medium (IMDM; BioWhittaker) containing 10% heat inactivated FCS (Bodinco), 2 mM L-glutamine and penicillin/streptomycin (all purchased from PAA Cell Culture Company) 37°C and 5% CO₂. Cells were passaged by trypsinization.

The following antibodies were used: anti-Nedd4 (07-049; Millipore) for IF, anti-Nedd4 (C5F5; Cell Signaling) for WB, anti-Dvl1 (AB5970; Millipore) for WB, anti-Dvl1 (D3320; Sigma) for IF, anti-Dvl2 (AB5972; Millipore), anti-Dvl3 (AB5974; Millipore), anti-β-catenin (610154; BD Bioscience), anti-γ-catenin (SC-7900; Santa Cruz), anti-ZOI (610966; BD Bioscience), anti-ubiquitin (MMs-257P; Covance), anti-GFP (JL-8; Clontech), anti-HA (H6908; Sigma) anti-c-myc (13–2500; Zymed). F-actin was stained with Rhodamine-labelled phalloidin (Invitrogen). The Rac1 inhibitor EHT1864 (E1657) was from Sigma.

Cells were transiently transfected with FuGene (Roche) as described (Nethe et al., 2010). The following constructs were used: GST–Rac, GST–RacΔC (a kind gift from R. Ahmadian, European Molecular and Cell Biology Laboratory, Heidelberg, Germany); GFP–Nedd4, originally from D. Rotin (The Hospital for Sick Children, Canada) (Pak et al., 2006), was kindly provided by J. Batt (University of Toronto, Canada); GFP–C867S-Nedd4 was generated by site-directed mutagenesis (Stratagene); ΔC2-Nedd4, the C2 domain and WW1-4 region of Nedd4 were generated by PCR and subsequently cloned into pEGFP (C1)-tagged constructs, using Kpn1 and EcoR1 restriction sites; Nedd4-2–GFP was kindly provided by C. P. Thomas (Itani et al., 2005) (University of Iowa, USA); 6xHis-myc-tagged ubiquitin expression plasmid, originally from R. R. Kopito (Ward et al., 1995) was kindly provided by J. Bertoglio (Inserm U749, France). GST–Dvl1 (Addgene); HA-tagged full-length Dvl1, ΔDIX, ΔPDZ and ΔDEP as previously described (Krylova et al., 2000), FLAG-tagged Dvl1 constructs were as previously reported (Tauriello et al., 2010), Dvl1 was subsequently cloned in pE-phiYFP (C1)-tagged constructs.

For immunostainings, cells were washed with ice-cold PBS, fixed with 3.7% paraformaldehyde for 20 min at RT and subsequently permeabilised with 1% Triton, 10% glycerol in PBS for 3 min at RT. Cells were immunostained with indicated antibodies and confocal images were captured with a Zeiss 510 Meta laser-scanning confocal microscope. Z-stacks and X/Z sections were generated and processed by means of LSM510 software. All data are representative for at least three or more experiments, unless indicated otherwise.

Lentiviral shRNA constructs from the TRC/Sigma Mission library were obtained from Sigma-Aldrich (St Louis, MO, USA). The human Nedd4-specific constructs used were: TRCN000007550 (#a) and -7551 (#b). The SHC002 scrambled shRNA construct (Sigma-Aldrich) was used as a negative control. All shRNA constructs were in the pLKO.1 vector backbone. shRNA-expressing lentiviral particles were prepared using HEK293T cells and virus was transduced as described previously (Nethe et al., 2010). Dvl1 siRNA (EHU-060361) and Nedd4 siRNA (EHU-132581) were obtained from Sigma-Aldrich (St Louis, MO, USA).

Confluent H292 cells were washed 2 times with PBS and incubated for 3 hrs with IMDM containing 2 mM EGTA. Cells were subsequently washed with PBS containing 1 mM Ca²⁺ and incubated for indicated times with MDM containing 10% FCS, 2 mM L-glutamine and 1 mM Ca²⁺. Next, the cells were fixed, stained and examined by confocal microscopy.

Peptide pull-down assays were performed as described previously (ten Klooster et al., 2006). In short, each assay was performed with 5 µg of indicated biotin-labelled peptide, 25 µl streptavidin-coated beads (Sigma-Aldrich) in NP-40 lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 10 mM MgCl₂, 10% glycerol, 1% NP-40). All peptides were fused to a protein transduction domain sequence: YARAAARQARA, which by itself was included as the control in the pull-down experiments (van Hennik et al., 2003). GST-fusion proteins were purified from BL21 bacteria as described (Nethe et al., 2010) and 100 µg of the indicated GST fusion protein was used per pull-down. Mass spectrometry analysis was performed as described (Kanters et al., 2008) and was used for the initial identification of Nedd4. Ubiquitylation of Dvl1 was assayed as described previously using HeLa cells, transfected with His-tagged ubiquitin (Nethe et al., 2010).

Peptides were synthesized on a peptide synthesizer (Syro II) using Fmoc solid phase chemistry. Peptides encoded a biotinylated protein transduction domain (Biotin-YARAAARQARAG) (Ho et al., 2001) followed by the 10 amino acids proceeding the CAAX domain for all used RhoGTPase peptides. The sequences of the Rac1 (P-A) and the Rac1 (PBQR) mutants are respectively: CAAAVKKRKRK and CPPPVKKAAAK.

ECIS-based cell spreading experiments were performed as previously described (ten Klooster et al., 2006). Briefly; ECIS electrodes (8W10E; Applied Biophysics) were coated with 10 µg/ml fibronectin (Sigma) in PBS for 1 h 37°C. 400,000 HeLa- or H292 cells were seeded per well in 400 µl IMDM containing 10% FCS, L-glutamine and penicillin/streptomycin. Cell spreading and monolayer formation were subsequently monitored by measuring the resistance at 4000 Hz.

The authors thank D. Rotin for GFP–Nedd4, C. P. Thomas for Nedd4-2–GFP, R. Ahmadian for GST–Rac1, J. Bertoglio for 6xHis–myc–ubiquitin and M. Fernandez-Borja for critical reading of the manuscript.