PAK is required for the disruption of E-cadherin adhesion by the small GTPase Rac

E-cadherin is one of the most important cell-cell adhesion receptors involved in tissue morphogenesis and maintenance of epithelial tissue integrity (Gumbiner, 2005). Loss of function of E-cadherin has emerged as an important event for local invasion and metastasis (Christofori and Semb, 1999). Furthermore, E-cadherin expression becomes down-regulated when epithelial cells acquire motile and invasive characteristics (i.e. during epithelial wound healing and embryonic development) (Thiery, 2003).

The small GTPase Rac1 is an important modulator of the actin cytoskeleton and a key regulator of E-cadherin function (Gumbiner, 2005). Activation of Rac1 perturbs the distribution of E-cadherin at junctions in different cell types (Akhtar and Hotchin, 2001; Braga, 2000; Edme et al., 2002; Keely et al., 1997; Quinlan, 1999; Shintani et al., 2006; Yagi et al., 2007). This effect is specific because other surface receptors, such as integrins, are not removed from cell-cell contacts with the same time frame as cadherins (Braga et al., 2000).

In vivo, Rac1-dependent disruption of junctions has been shown during intestinal epithelia differentiation (Stappenbeck and Gordon, 2000), salivary gland morphogenesis (Pirraglia et al., 2006) and tracheal tubulogenesis (Chihara et al., 2003). Furthermore, Rac activation is also required to perturb E-cadherin junctions after Ras activation (Braga et al., 2000; Edme et al., 2002), Csk kinase expression (Yagi et al., 2007) and HGF/scatter factor treatment (Ridley et al., 1995; Takaishi et al., 1994). Yet, these effects may be cell-type specific (Braga et al., 1999; Lozano et al., 2003).

Upregulation of Rac1 protein and mRNA levels has been demonstrated in human tumours of epithelial origin, implicating Rac1 in cellular transformation (Lozano et al., 2003; Sahai and Marshall, 2002a). Two other family members, Rac3 and Rac1b, have also been found overexpressed or activated in several epithelial tumours. Rac3 is divergent from Rac1 at the C-terminal region (Haataja et al., 1997; Mira et al., 2000). Rac1b is an alternative splice variant of Rac1 (Jordan et al., 1999; Schnelzer et al., 2000) with an in-frame insertion of 19 amino acids in a region important for Rac1 interaction with regulators and effectors. Rac1b can neither induce lamellipodia nor activate the two known Rac1 effectors p21 activated kinase (PAK) and the downstream signaling pathway of Jun N-terminal kinase 1 (JNK1) (Fiegen et al., 2004; Matos et al., 2003). Yet, the functional consequences of Rac1, Rac3 and Rac1b activation are only now beginning to be unravelled (Chan et al., 2005; Matos et al., 2003). Furthermore, it remains to be tested whether Rac3 and Rac1b have similar functions as Rac1 on cell-cell adhesion.

The precise mechanisms by which E-cadherin stability at junctions is perturbed by Rac1 are still unclear. Here, we set out to address whether activation of Rac3 or Rac1b can disrupt cadherin-dependent adhesion and identify specific effectors that can mediate this effect. We found that, in contrast to Rac1 and Rac3, Rac1b is not able to induce any substantial changes in E-cadherin cell-cell adhesion. This striking Rac1b phenotype indicates that differences in activation of signaling pathways by Rac1 and Rac1b may provide clues on the mechanism involved. Interestingly, PAK1 – a target that is not activated by Rac1b – is essential for Rac1 to perturb keratinocyte junctions. As PAK is also involved in epithelial morphogenesis and lamella formation (Sells et al., 1997; Tang et al., 1997), our results suggest that PAK1 is a pivotal Rac target that coordinates junction disassembly and increased motility.

We addressed whether, similarly to Rac1, Rac3 and Rac1b can also destabilize cadherins from keratinocyte cell-cell contacts. From four hours onwards after expression of the constitutively active Rac1 mutant Rac1^Q61L, E-cadherin localization at junctions was discontinuous with thickening of E-cadherin staining at the central region of cell-cell contacts. At later time points, cells partially detached from each other (Fig. 1a). A similar pattern and kinetics of junction disassembly was observed when constitutively active Rac3 (Rac3^G12V) was expressed and compared with Rac1^Q61L (Fig. 1b). In contrast to Rac1 and Rac3, prolonged expression of constitutively active Rac1b (Rac1b^Q61L) or wild-type Rac1b (data not shown) did not result in any substantial alteration of E-cadherin distribution (Fig. 2a). Quantitatively, Rac1b-expressing cells showed high levels of intact junctions as opposed to Rac1 and Rac3 (Fig. 2b). These results are interesting, considering that Rac1b is identical to Rac1 apart from an insert of 19 amino acids (Jordan et al., 1999).

The inability of Rac1b to remove E-cadherin from junctions could be due to degradation of Rac1b protein. However, exogenous Rac1b protein was stable in keratinocytes. First, exogenous Rac1b was clearly expressed at levels comparable with those of activated Rac1 (Fig. 2c) when detected by an anti-Myc antibody (recognizes the Myc epitope in exogenous Rac1 and Rac1b) or an antibody against Rac1 (recognizing endogenous and exogenous Rac1 and Rac1b). Second, transfection experiments showed that Rac1b was not degraded (Fig. 2d). Although Rac1b mRNA is expressed in keratinocytes (data not shown), endogenous Rac1b was not detected consistent with its low levels of expression and high turnover.

Furthermore, Rac1 is the main Rac family member expressed in normal keratinocytes. Considering the transfection efficiency (around 30-40% in our hands), levels of exogenous Rac1 are about threefold higher than those of endogenous Rac levels (Fig. 2d), which is in the lower range of expression of exogenous proteins. Taken together with the lack of effect of overexpressed Rac1b, our data argue that disassembly of cadherin adhesion by Rac1 did not result from overexpression per se [for additional controls and specificity see Braga et al. (Braga et al., 2000)].

One of the mechanisms known to participate in the disruption of cell-cell contacts is increased contractility (Sahai and Marshall, 2002b; Zhong et al., 1997). Rac activation can also participate in contractile events by counterbalancing Rho activity and other mechanisms (Bokoch, 2003). However, inhibition of contractility was ineffective in preventing junction disruption by Rac1 (supplementary material Fig. S1). We reasoned that differences in the phenotypes observed between Rac1b and Rac1/Rac3 may be due to distinct signaling pathways that are activated by each GTPase. As Rac1b can not transduce signals through JNK1 and PAK1 activation (Fiegen et al., 2004; Matos et al., 2003; Singh et al., 2004), Rac1 and Rac3 could mediate junction disassembly through activation of one or both of these kinases. However, JNK1-dependent pathways were not necessary for junction breakdown induced by active Rac1 (supplementary material Fig. S2).

PAK is a serine/threonine kinase and one of the best characterized Rac targets: it participates in cell motility, cytoskeletal rearrangements and is able to phosphorylate many different cytoskeletal and cytosolic proteins (Bokoch, 2003). We inhibited PAK function using three different constructs: a kinase-dead version of PAK1 (PAK1^K299R) (Tang et al., 1997), a kinase-dead PAK1 that does not bind Rac (PAK^{H83L,H86L,K299R}) (Sells et al., 1997; Tang et al., 1997) and the auto-inhibitory domain of PAK1, which does not interact with active Rac (amino acid residues 83-149; PAK^AID). PAK^AID interacts in trans with the kinase domain of PAK1, PAK2 and PAK3, therefore blocking their ability to phosphorylate downstream targets (Zhao et al., 1998).

Inhibition of PAK alone did not disrupt junctions, indicating that PAK function per se is not essential for cadherin-dependent adhesion (Fig. 3a,b). Co-expression of active Rac1 with any of these inhibitory constructs effectively blocked junction disruption. By contrast, PAK inhibition was unable to rescue junction disassembly induced by active Arf6, suggesting the specificity of this effect for Rac1 (our unpublished data). Furthermore, it is unlikely that the effects observed by blocking PAK function resulted from unbalancing other Rac1-dependent pathways (Fig. 3b): (1) only PAK1 is expressed in keratinocytes (Fig. 3c) and, (2) inhibition of other signaling pathways triggered by Rac1 did not interfere with junction disassembly (supplementary material Figs S1 and S2, and data not shown).

To confirm that endogenous PAK is required downstream of Rac1 to disrupt cell-cell contacts, we performed RNA interference (RNAi) to deplete PAK1 using three different oligonucleotides (Fig. 4a-c). Optimization experiments showed that levels of PAK1 were effectively reduced (70-80%), but not E-cadherin protein levels (Fig. 4b). When constitutively active Rac1 was expressed in PAK1-siRNA-treated keratinocytes (Fig. 4a), cadherin staining along junctions was significantly restored (60%) when compared with control siRNA (25%, Fig. 4c). The rescue effect observed with PAK1 siRNA was smaller than when dominant-negative PAK constructs were used (90%, Fig. 3). The most possible explanation is the population-based quantification used in the RNAi experiments, in which all Rac1-expressing cells were considered. This criterion reduces the level of the effects quantified: some keratinocytes would not show rescue of junction disruption as they would still have some endogenous PAK. Alternatively, although unlikely, we cannot formally exclude the possibility that the three dominant-negative PAK constructs used inhibit additional pathways than the one leading to PAK activation. Nevertheless, our data indicate that endogenous PAK is necessary for disruption of cell-cell contacts induced by Rac1.

How can PAK1 participate in the Rac1-dependent downregulation of cadherin from junctions? We excluded the participation of lamella formation and increased contractility as the cause of junction disassembly induced by Rac (supplementary material Fig. S1) (Braga et al., 2000). An attractive possibility involves PAK1-mediated phosphorylation of proteins necessary for junction stability. Alternatively, PAK1 could potentially participate in the internalization of membranes containing E-cadherin receptors, because PAK has been involved in pinocytosis (Bokoch, 2003; Kumar and Vadlamudi, 2002). Indeed, our results suggest that E-cadherin partially colocalized with active Rac1 in intracellular vesicles (E.L., unpublished), in agreement with previous observations (Akhtar and Hotchin, 2001; Akhtar et al., 2000).

The potential mechanisms described above are not mutually exclusive, because PAK1-dependent phosphorylation of junction-associated proteins may signal to induce internalization (Gumbiner, 2005). A number of reports suggest the importance of cadherin internalization (Le et al., 1999; Lock and Stow, 2005) and phosphorylation of cadherin-associated proteins in cadherin turnover (Fujita et al., 2002; Gavard and Gutkind, 2006). Overall, whether the reduced presence of E-cadherin at keratinocyte junctions following Rac activation results from defective recycling and/or increased degradation is unclear at the moment and requires further work.

Rac1 has been shown to participate downstream of signaling pathways known to control cadherin stability at junctions in physiological or pathological conditions, i.e. downstream of growth factor receptors (Lozano et al., 2003; Sahai and Marshall, 2002a). In particular, stimulation of endothelial cells with VEGF (Gavard and Gutkind, 2006) or TNFα (Nwariaku et al., 2004) requires Rac and/or PAK signaling to promote internalization of VE-cadherin. However, the fine mechanistic details downstream of PAK may vary according to cell type and stimuli (this work) (Braga et al., 1999; Gavard and Gutkind, 2006; Nwariaku et al., 2004). Taken together, these data point out that PAK1 might be a pivotal Rac target in the regulation of cadherin stability following different stimuli that destabilize junctions. As such, our work opens up exciting new avenues to investigate the mechanisms through which PAK influences adhesive and morphogenetic processes.

PAK has an emerging role in epithelial morphogenesis, differentiation and cell-cell contact inhibition of motility (Wang et al., 2003; Zegers et al., 2003). PAK1 function per se is not necessary for the maintenance of cadherin adhesion in keratinocytes (Figs 3, 4). Consistent with our results, constitutive activation of Mbt, the Drosophila homologue of PAK4, a class II PAK-family member, disrupts adherens junctions during eye development (Menzel et al., 2007). To our knowledge, we present the first evidence that PAK1 participates in the destabilization of E-cadherin following Rac1 activation. PAK1 activation correlates with transformation, increased invasion and hyperplasia (Kumar and Vadlamudi, 2002; Tang et al., 1997). Owing to its established role in lamellae protrusion (Bokoch, 2003), PAK might be an ideal Rac target to participate in junction disruption and, at the same time, stimulate migration in epithelial cells.

Normal human keratinocytes (strains Kb and Sf, passages 3-7) were cultured on a mitomycin-C-treated monolayer of 3T3 J2 fibroblasts at 37°C and 5% CO₂ as described (Rheinwald, 1989). For RNAi experiments, keratinocytes were seeded in normal medium, transferred to low-Ca²⁺ medium and grown until confluence (Braga et al., 1997).

Antibodies used were: anti-Myc epitope (9E10, CR-UK and JAC6, Serotec); anti-HA epitope (262K, Cell Signaling and SG77, Zymed); anti-E-cadherin (HECD1 and ECCD-2); anti-FLAG epitope (M2, Sigma); anti-Rac (23A8, Upstate); anti-Rac1b (gift from Peter Jordan, Centro de Genetica Humana, Portugal); anti-PAK (N20, Santa Cruz); anti-actin (C4, ICN Biomedicals) and anti-α-catenin (VB1) (Braga et al., 1995). Anti-Rac3 antibody was prepared in rabbits using amino acids 182-192 of human Rac3 as antigen. The affinity-purified antibody does not recognize purified or overexpressed Rac1 or Rac2. Secondary antibodies were from Jackson Immuno Research Laboratories. Y-27632, ROCK inhibitor, was a gift from A. Yoshimura, Yoshitomi Pharmaceuticals, Japan).

Full-length constitutively active Rac1^Q61L, Rac1b^Q61L and Rac3^G12V mutants were subcloned into the pCS2-myc vector. Rac1^Q61L was subcloned into pRK5-myc and pRK5-FLAG. PAK auto-inhibitory domain (PAK^AID) and ROCKα kinase-dead in pRK5-myc were a gift from Alan Hall (Memorial Sloane-Kettering Cancer Center, New York, NY). PAK1^K299R and PAK1^{H83L,H86L,K299R} were provided by Jonathan Chernoff (Fox Chase Cancer Centre, Philadelphia, PA). JNK1(APF) was a gift of Roger Davis (University of Massachusetts Medical School, Boston, MA) and CrkII^R38K and CrkII^W169K mutants from Hidesaburo Hanafusa (Osaka Bioscience Institute, Osaka, Japan).

Unless specified, cDNA constructs were injected at 0.1 mg/ml into the nucleus of 50-100 cells per coverslip as described previously (Braga et al., 1997). For ROCK inhibition experiments, cells injected with activated Rac1 cDNA plasmid were incubated for 5 hours with 25 μM Y-27632. Microinjected keratinocytes were fixed, stained and analyzed as previously described (Braga et al., 1997).

Keratinocytes grown in low-Ca²⁺ medium were transfected with 50 nM of siRNA targeting PAK1 [5′-AAUCUGUAUACACACGGUCUG-3′, D-003521-07 and D-003521-01 (Dharmacon)] or siCONTROL non-targeting siRNA (Dharmacon) using RNAiFect (Qiagen)] according to manufacturer's instructions. Around 66 hours post transfection, cell-cell contacts were induced by addition of Ca²⁺ for 1 hour. Cells were microinjected with pRK5-FLAG-Rac^Q61L and incubated for 7 hours. PAK1 knockdown was confirmed by western blotting for each experiment.

Keratinocytes grown in standard medium were transfected with 2 μg of plasmids encoding Myc-tagged, activated forms of Rac1, Rac1b and Rac3 and TransIT®-Keratinocyte Transfection Reagent (Cambridge Biosciences) according to the manufacturer's protocol. After overnight incubation cells were lysed in 50 mM Tris-HCl pH 7.5, 1% Triton X-100, 150 mM NaCl and protease inhibitors.

RNA was extracted from keratinocytes and HNP-1 monocytes (gift from Emmanuele Caron, Imperial College London, UK) using RNAzolB (Biogenesis) and in SuperScript™ II reverse transcriptase reactions according to the manufacturer's protocol (Invitrogen). Specific primers were designed for PAK1 5′-ATACTGGATGGCACCAGAGG-3′ and 5′-CTGGGCAGTTGAGTCACAGA-3′; PAK2 5′-AATTGACCCTGTTCCTGCAC-3′ and 5′-CCCGAAATATTGGGGAAAGT-3′; PAK3 5′-CAAGGGGCATCAGGTACTGT-3′ and 5′-ATCAGAGGAGTCAGGCTGGA-3′. PCR products were resolved on 1% agarose gel and visualized under UV light.

Only junctions between two expressing cells were considered for quantification, which was performed in different ways. First, to determine the disruption of junctions by Rac family members and rescue experiments, length measurements were performed for each junction (Leica LCS-Lite): the length of E-cadherin staining and the length of the cell-cell contact (corner to corner of neighbouring cells) for each contact. Their ratio was calculated; controls were set to 100% (i.e. cadherin staining = length of the cell-cell contact). In RNAi experiments, all Rac1-expressing cells were considered in the analysis becsause of the poor performance of anti-PAK1 antibodies regarding immunofluorescence. Statistical analysis was performed with Student's t-test; error bars give the standard error of the mean (± s.e.).

We thank E. Caron, J. Chernoff and A. G. Sáez for critical reading of the manuscript and all researchers listed therein for kindly providing reagents. We also thank J. Erasmus for technical assistance. E.L. was supported by the Ministerio de Ciencia y Tecnología, Spain, and Cancer Research UK; M.A.M.F. is a recipient of a PhD studentship from Association for International Cancer Research. V.M.M.B.'s lab is funded by the Medical Research Council and Cancer Research UK.