Tetraspanin CD151 regulates RhoA activation and the dynamic stability of carcinoma cell-cell contacts

Tetraspanins, a family of proteins with 33 members in mammals, play important roles in gamete fusion, brain development and function, immunity, photoreceptor morphology, and the maintenance of epidermal integrity (Hemler, 2005; Levy and Shoham, 2005). In addition to these physiological roles, tetraspanins are also implicated in a variety of pathological settings, notably tumor cell metastasis. For example, whereas tetraspanins CD9 and CD82 are potential metastasis suppressors (Boucheix et al., 2001; Lazo, 2007; Liu and Zhang, 2006; Tonoli and Barrett, 2005), tetraspanin CD151 has been linked to enhanced metastasis of colon, prostate and lung cancer (Ang et al., 2004; Hashida et al., 2003; Tokuhara et al., 2001). The mechanisms whereby tetraspanins might regulate metastasis can be grouped into several classes, including regulation of (i) pericellular matrix proteolysis (Gesierich et al., 2006; Hasegawa et al., 2007; Hong et al., 2006; Hong et al., 2005; Huang et al., 2007; Saito et al., 2006; Sugiura and Berditchevski, 1999), (ii) integrin cell-surface expression and trafficking (Berditchevski and Odintsova, 2007; Liu et al., 2007; Winterwood et al., 2006), (iii) integrin-dependent migration, invasion and signaling (Berditchevski, 2001; Hemler, 2005), (iv) integrin crosstalk with growth factor receptors (Sridhar and Miranti, 2006) and (v) pathological angiogenesis (Gesierich et al., 2006; Takeda et al., 2007). Most of these proposed tetraspanin functions center around regulation of cell-extracellular matrix interactions and downstream signaling events.

In addition to alterations in cell-matrix interactions, the metastasis of carcinomas, which arise in epithelial layers, often entails changes cell-cell adhesion. A prevailing view is that carcinoma cell metastasis might be triggered by aberrant reactivation of a developmental switch that includes a downregulation of E-cadherin-dependent cell-cell adhesion, similar to the epithelial-mesenchymal transitions of embryogenesis (Brabletz et al., 2005; Hugo et al., 2007; Thiery and Sleeman, 2006). However, the invasive capacity of certain E-cadherin-positive carcinomas (Cowin et al., 2005; Cowin and Welch, 2007; Jang et al., 2007) indicates that additional regulatory mechanisms, beyond simply extinguishing E-cadherin expression, are likely to come into play in some settings.

Several lines of research suggest that regulation of cell-cell interactions could be an important additional mechanism by which CD151 might influence metastasis. Humans and mice lacking CD151 show loss of epithelial integrity in the skin and kidney (Karamatic Crew et al., 2004; Sachs et al., 2006), and CD151-null mice also show defective wound healing (Cowin et al., 2006). In kidney epithelial cells from mice lacking α3β1 integrin (a major CD151 partner), E-cadherin localization and function appeared perturbed, and association of α3β1 integrin with CD151 might be important for the ability of α3β1 integrin to regulate E-cadherin in this system (Chattopadhyay et al., 2003). Conversely, overexpression of CD151 enhanced carcinoma cell-cell association (Shigeta et al., 2003), whereas an anti-CD151 antibody interfered with E-cadherin localization in HaCat cells and promoted their dispersal (Chometon et al., 2006). Collectively, these data suggest that CD151 might regulate cell-cell interactions between tumor cells; however, the effect of CD151 loss of function on E-cadherin in transformed cells has not been determined.

We recently reported that near-total, RNAi-mediated silencing of CD151 in epidermal carcinoma cells resulted in impaired adhesion and migration mediated by the CD151-associated integrins, α3β1 and α6β4 (Winterwood et al., 2006). In this previous study, we examined the behavior of individual, dissociated tumor cells. Here, we extended our analysis to conditions where cell-cell contacts were maintained. Our data uncover an important role for CD151 in promoting the stability of carcinoma cell-cell junctions, and reveal complexities that would have to be considered in any potential strategy that targets CD151 to inhibit tumor cell motility. In addition, the data help to illuminate a mechanism whereby the organization of intercellular junctions can be perturbed even though E-cadherin cell surface expression is maintained.

To determine how CD151 might regulate tumor cell motility when cell-cell contacts are maintained, we compared parental A431 epidermal carcinoma cells (WT) to CD151-silenced cells (A431 sh3 cells) in gap-filling assays. Freshly confluent monolayers were inscribed with a gap using a micropipet tip, and gap closure was monitored by time-lapse microscopy (Fig. 1A; supplementary material Movie 1). Both cell types closed the gap as continuous sheets. However, in contrast to the reduced single-cell velocity that we previously observed for the CD151-silenced sh3 cells (Winterwood et al., 2006), the sh3 cell sheets closed the gap significantly faster than their wild-type counterparts. Quantification indicated that the gap closure rate for sh3 cells was over twice as fast as for wild-type cells (Fig. 1B). The enhanced closure rate of the sh3 cell sheets was not due to a difference in proliferation, because frame-by frame analysis of mitoses occurring during assay revealed no differences (Fig. 1C).

We used a spheroid assay to further investigate how CD151 might regulate collective tumor cell migration. Spheroids, created by culturing cells overnight in tubes coated with non-adhesive poly(2-hydroxyethyl methacrylate) (poly-HEMA), were plated on laminin-5, an α3β1 integrin ligand. Both wild-type and sh3 cell spheroids expanded as intact sheets, and once again the sh3 cell sheets migrated significantly farther than wild-type cell sheets (Fig. 1D). Re-expressing CD151 in sh3 cells (creating cells designated A431 sh3 Rx) reversed the enhanced migration of the sh3 cell spheroids (Fig. 1D). Quantification showed that sh3 cell spheroids expanded over twice as far as wild-type or Rx cell spheroids (Fig. 1E). Thus, despite the fact that the motility of individual cells is significantly reduced upon silencing CD151 (Winterwood et al., 2006), the collective migration of intact, CD151-silenced monolayers is significantly enhanced. These results are summarized as linear velocities in Table 1.

The data in Table 1 indicate that all three cell types migrated significantly slower as intact monolayers than as single cells. However, whereas wild-type and Rx monolayer velocities were only ∼6% of their respective single-cell velocities, the sh3 monolayer velocity was nearly 50% of the sh3 single-cell velocity. An analysis of wild-type and sh3 collective migration velocity in the gap-filling experiments yielded very similar results (data not shown). These data suggested that: (i) the maintenance of cell-cell contacts during collective migration imposes a restraint on migration velocity, and (ii) this restraint on velocity is manifested much more in wild-type and Rx monolayers than in sh3 monolayers.

To compare our results to a scenario where cell-cell contacts are explicitly disrupted, we treated wild-type spheroids with an anti-E-cadherin function-blocking antibody. In contrast to the intact sheets observed in the previous experiments, treatment of spheroids with anti-E-cadherin caused cells to emigrate as individuals, creating a large halo of single cells around the spheroids (Fig. 1F; see also supplementary material Fig. S1 for an expanded view). Although these data were consistent with the view that cell-cell contact can restrain cell motility, they suggested that wholesale disruption of E-cadherin adhesive function was unlikely to explain the phenotype of CD151-silenced monolayers. Instead, the enhanced migration of intact sh3 monolayers suggested that cell-cell contacts were maintained, but might be perturbed, in sh3 cell monolayers.

Lastly, to test whether regulation of collective migration by CD151 depends on the extracellular substrate, we analyzed spheroids on collagen I, a ligand for α2β1 integrin. The loss of CD151 enhanced collective migration on collagen I to a similar extent as on laminin-5 (Fig. 1G). These data indicate that the ability of CD151 to act as a negative regulator of collective migration in this assay might not be strongly dependent on the exogenous extracellular substrate supplied.

We hypothesized that CD151 might regulate collective migration by regulating the organization of carcinoma cell-cell junctions. Therefore, we next examined the localization of E-cadherin and associated proteins in wild-type, sh3 and Rx cells. Whereas E-cadherin strongly localized to cell-cell contact sites in wild-type cells (Fig. 2A), its junctional localization was substantially perturbed in CD151-silenced cells (Fig. 2B). Re-expression of CD151 in A431 sh3 Rx cells restored normal E-cadherin localization (Fig. 2C). The junctional localization of E-cadherin regulatory proteins β-catenin (Fig. 2D-F), α-catenin (Fig. 2G-I), plakoglobin (Fig. 2J-L) and p120^ctn (Fig. 2M-O) were also perturbed in the sh3 cells. CD151 itself localized to cell-cell contact sites (Fig. 2P,R), which was consistent with a role in regulating cell-cell contacts, and was specifically absent in the sh3 cells, as expected (Fig. 2Q). Lastly, the receptor protein tyrosine phosphatase PTPμ, whose expression might be regulated by integrin-tetraspanin complexes in some cells (Chattopadhyay et al., 2003), appeared equally expressed in all three cell types (Fig. 2S-U). These data indicated that E-cadherin-based cell-cell junctions are less well organized in CD151-silenced cells.

To begin to explore the basis of the altered cell-cell junctions in CD151-silenced cells, we examined E-cadherin association with components of the cadherin regulatory complex. In E-cadherin immunoprecipitations, the amounts of co-precipitating β-catenin, α-catenin, p120^ctn, α-actinin and plakoglobin all appeared unchanged in CD151-silenced cells (Fig. 3A-E). In addition, the amounts of E-cadherin or p120^ctn that co-precipitated with β-catenin were unchanged, as was the amount of E-cadherin that co-precipitated with p120^ctn (Fig. 3F,G). Because many E-cadherin partners are regulated by phosphorylation, we examined the profile of phosphoproteins co-precipitating with E-cadherin, focusing on a size range that encompasses α-, β- and p120 catenin, as well as plakoglobin and α-actinin. No obvious differences were observed in any of the cell types (Fig. 3H, lanes 2-4). However, treatment of cells with epidermal growth factor (EGF) confirmed that increased phosphorylation of protein bands corresponding to β-catenin and plakoglobin (Hoschuetzky et al., 1994) could be detected by our assay system (Fig. 3H, lane 1). Moreover, repeated experiments that directly assessed β-catenin phosphorylation failed to reveal any differences (data not shown).

We also observed that E-cadherin surface expression was maintained at similar levels in CD151-silenced cells as in wild-type or Rx cells (Fig. 3I), and we observed no changes in the detergent extractability of E-cadherin in CD151-silenced cells (Fig. 3J). Next, we tested for a possible association between E-cadherin and CD151 or other tetraspanins. As shown in Fig. 3K, although E-cadherin was readily detected in a β-catenin immunoprecipitate (lane 2), no E-cadherin could be detected co-precipitating with any of several different tetraspanins (lanes 3-6) or with α2- or α3-integrin subunits (lanes 7 and 8). Conversely, α3 integrin was detected in tetraspanin immunoprecipitates (Fig. 3L, lanes 1-3), but not in a β-catenin immunoprecipitate (lane 4). Lastly, we tested for cell-surface expression of PTPμ and observed no differences in wild-type, sh3 or Rx cells (Fig. 3M), consistent with the staining data in Fig. 2S-U. Collectively, these data indicate that major biochemical disruptions in the E-cadherin regulatory complex or changes in PTPμ expression are unlikely to explain the disorganized junctions in our CD151-silenced cells. In addition, E-cadherin itself might not directly interact with CD151 or other tetraspanins.

To further define how E-cadherin localization and function might be altered in CD151-silenced cells, we tested wild-type and CD151-silenced sh3 cells in a short-term adhesion assay on purified E-cadherin-Fc fusion protein. As shown in Fig. 4, both cell types adhered equally well to E-cadherin-Fc, and adhesion could be blocked substantially with anti-E-cadherin antibody and completely with the calcium chelator, EGTA. These data suggested that initial E-cadherin-dependent adhesion events might proceed normally in CD151-silenced cells. To further test this, we examined the localization of E-cadherin during a calcium switch assay. To facilitate these experiments, we introduced an E-cadherin-GFP fusion protein (EcadGFP) into A431 wild-type and sh3 cells and sorted GFP-positive cells. In cells cultured overnight, EcadGFP strongly localized to cell-cell contacts in wild-type cells, but much less so in sh3 cells (Fig. 5A,B). After 15 minutes of exposure to low calcium, both cell types rounded up and cell-cell junctions were largely disrupted (Fig. 5C,D). Strikingly, 30 minutes after calcium restoration, EcadGFP was strongly localized to the re-forming cell-cell junctions in sh3 cells, very similar to the situation in wild-type cells (Fig. 5E,F). However, after 3 hours, EcadGFP localization in sh3 cells had begun to deteriorate (Fig. 5G,H). Collectively, these data indicated that E-cadherin-based adherens junctions can form in CD151-silenced cells but they appear unstable and fail to remain organized. The data also suggested that the adhesive function of E-cadherin, per se, might not be impaired in our CD151-silenced cells.

The formation and maintenance of E-cadherin-based adherens junctions is known to depend on the activity levels of Rho family small GTPases, which regulate the organization of the actin cytoskeleton (Arthur et al., 2002; Braga, 2002; Fukata and Kaibuchi, 2001; Yap and Kovacs, 2003). Therefore, we examined F-actin localization in relation to E-cadherin using confocal microscopy. We observed copious actin stress fibers near the basal surface of the sh3 cells, whereas F-actin was more concentrated at the cell perimeter in wild-type cells (Fig. 6A′,E′). In addition, E-cadherin was fairly evenly distributed over the entire basal cell surface of sh3 cells, whereas in wild-type cells it was more concentrated at cell-cell boundaries (Fig. 6A,E, and overlaid with actin in Fig. 6A″,E″). At more lateral and apical planes in sh3 cells, some F-actin and E-cadherin was observed at some cell-cell junctions (Fig. 6B-D,B′-D′, overlaid in Fig. 6B″-D″), but overall the junctional staining of both proteins was less prominent and less well-organized than in wild-type cells (Fig. 6F-H,F′-H′, overlaid in Fig. 6F″-H″). Confocal analysis of β-catenin revealed a similar pattern: in wild-type cells, β-catenin was prominently localized at most cell-cell junctions (supplementary material Fig. S2A-C), whereas in sh3 cells, β-catenin localized to some junctions, but in general appeared more diffuse and less well-organized (supplementary material Fig. S2D-F). Re-expression of CD151 in Rx cells restored the junctional localization of β-catenin (supplementary material Fig. S2G-I).

Because of the abundant stress fibers in sh3 cells, we next compared sh3 and wild-type cells for the levels of active RhoA, which controls formation of stress fibers. As shown in Fig. 7A, in confluent wild-type cell cultures, the level of active, GTP-loaded RhoA was significantly reduced (lane 3) compared to that in rounded wild-type cells (lane 1), which served as a positive control for RhoA activation. By contrast, the level of active RhoA in confluent sh3 cell cultures (lane 4) was significantly elevated and similar to the rounded cell positive control (lane 2). Restoring CD151 expression in Rx cells reversed the elevation of levels of active RhoA observed in sh3 cells (Fig. 7B). These data suggested that elevated RhoA activity is probably responsible for the elevated formation of stress fibers in CD151-silenced cells.

Elevated RhoA activity has been associated with a destabilization of adherens junctions (Jou and Nelson, 1998; Zhong et al., 1997) and cell scattering in response to factors such as hepatocyte growth factor (Wells et al., 2005). We therefore examined the behavior of individual cells within intact wild-type and CD151-silenced monolayers by time-lapse microscopy. These experiments revealed that sh3 cell monolayers appeared much more dynamic, with individual cells moving within the monolayer to a much greater extent than in wild-type monolayers (Fig. 8A,B; supplementary material Movie 2). Quantification showed that sh3 cells moved with a threefold higher velocity than wild-type cells (Fig. 8C). Indeed, we sometimes observed cell-cell junctions that were literally pulled apart by the highly dynamic movements within the sh3 cell monolayers (Fig. 8D; supplementary material Movie 2). To capture the remodeling of junctions more quantitatively, we measured the size of individual cell-cell contacts every 5 minutes for 2 hours. Compared to wild-type cell-cell contacts, whose size remained relatively constant, the size of sh3 cell-cell contacts oscillated dramatically as cells alternately pushed together and pulled away from one another within the monolayer (Fig. 8E,F). As a result, the average rate of change of cell-cell contact size was nearly threefold higher in the sh3 cell monolayers (Fig. 8G). The elevated dynamics within the sh3 cell monolayer corresponded to a significant decrease in the lifespan of individual sh3 cell-cell contacts (Fig. 8H). Together with the previous experiments, these data suggested that a consequence of elevated RhoA activity in sh3 cells is a more dynamic monolayer in which individual cell-cell contacts undergo continuous remodeling and display a significantly shortened lifespan. To begin to test this, we monitored intra-monolayer cell motility before and after the acute addition of recombinant, cell-permeable C3 transferase, a selective Rho inhibitor. Upon C3 treatment, the motility of individual cells within the CD151-silenced monolayer slowed to wild-type levels (Fig. 8I). C3 treatment of wild-type cells did not cause any further reduction in intra-monolayer motility (Fig. 8I).

We next tested whether the reduced motility within CD151-silenced monolayers upon C3 treatment corresponded to a restored junctional organization. Untreated sh3 cells displayed actin stress fibers and poorly organized junctions at the basal cell surface (Fig. 9A-C). By contrast, C3-treated cells showed fewer stress fibers and a more normal junctional localization of F-actin and E-cadherin (Fig. 9D-F, compared to untreated wild-type cells in Fig. 9G-I). Wild-type cells treated with C3 retained normal F-actin and E-cadherin localization, although F-actin at cell-cell boundaries appeared somewhat depleted, perhaps because RhoA activity in C3-treated wild-type cells was suboptimal for maintenance of F-actin at junctions (Fig. 9J-L). At a more lateral plane, some F-actin and E-cadherin did localize to some sh3 cell junctions (Fig. 9A′-C′), but this was more diffuse and less organized than in C3-treated sh3 cells (Fig. 9D′-F′) or untreated wild-type cells (Fig. 9G′-I′). Treatment of wild-type cells with C3 again partially depleted F-actin from the junctions, but did not perturb their organization (Fig. 9J′-L′). Collectively, the data in Figs 8 and 9 show that a RhoA inhibitor can suppress the elevated motility within CD151-silenced monolayers and restore a more normal junctional organization.

Given that α3β1 integrin is a prominent CD151 partner (Kazarov et al., 2002; Sterk et al., 2002; Yauch et al., 1998) and might regulate junctional stability in non-transformed cells (Chattopadhyay et al., 2003; Wang et al., 1999), we next began to explore the extent to which α3β1 integrin might be involved in the ability of CD151 to promote junctional organization in carcinoma cells. Using a FLAG-tagged, wild-type CD151 Rx cDNA as a template, we constructed a FLAG-tagged CD151 mutant, CD151^VR, in which the `variable region' in the CD151 large extracellular loop (EC2 domain) was replaced with that of tetraspanin TM4SF2 (see the Materials and Methods). In relatively mild Brij detergent lysates (1:1 ratio of Brij 99 to Brij 96), wild-type CD151 readily co-precipitated with α3 integrin (Fig. 10A, lane 1). By contrast, virtually no CD151^VR was detected co-precipitating with α3 integrin (Fig. 10A, lane 3). Conversely, α3 integrin co-precipitated with CD151, but only a trace of α3 integrin co-precipitated with the CD151^VR mutant (Fig. 10A, lanes 4 and 6). In CD151-silenced sh3 cell lysates, no specific signal was detected in FLAG blots, and no α3 integrin was retrieved by FLAG immunoprecipitation (Fig. 10A, lanes 2 and 5). The CD151^VR mutant migrated as a broad band of ∼45 kDa versus the ∼32 kDa band observed for wild-type FLAG-tagged CD151, probably reflecting the presence of four glycosylation sites in the TM4SF2 variable region, as compared to the single site in the CD151 variable region.

We next immunoprecipitated FLAG-tagged wild-type CD151 and the CD151^VR mutant from extracts of biotin-labeled cells. Simultaneous detection of the biotin label and the FLAG epitope revealed that wild-type CD151 and CD151^VR could both be cell-surface-labeled (Fig. 10B). In addition, CD151 and CD151^VR both associated with biotin-labeled species with apparent molecular masses (∼20-22 kDa) that correspond to tetraspanins CD9 and CD81. Thus, the CD151^VR mutant is expressed on the cell surface at a comparable level to wild-type CD151, and appears to retain association with other tetraspanins.

In contrast to wild-type CD151, which was concentrated at cell-cell junctions (Fig. 10C; see also Fig. 2P,R), the CD151^VR mutant displayed a more dispersed cellular localization (Fig. 10D). Moreover, E-cadherin, which again strongly localized to junctions in cells rescued with wild-type CD151 (Fig. 10E), was severely perturbed in cells rescued with CD151^VR (Fig. 10F). Collectively, these data suggest that the ability of CD151 to associate with α3β1 integrin might be important for its ability to promote junctional organization in carcinoma cells.

We have shown that CD151-silenced carcinoma cells display enhanced collective migration, despite our earlier observation that the same CD151-silenced cells display reduced single-cell migration velocities (Winterwood et al., 2006). This apparent paradox might be resolved if the factors that limit the much faster single-cell velocity [such as efficient detachment at the lateral or trailing edge (Winterwood et al., 2006)] are not the velocity-limiting factors in slower, collective cell migration. The collective migration velocity of wild-type cells is only ∼6% of their single-cell velocity, whereas that of CD151-silenced cells is a full 50% of the single-cell velocity. Thus, for collective cell migration, a major velocity-limiting factor might be the restraint imposed by being part of the collective, a restraint that appears significantly reduced for CD151-silenced cells. We provide evidence that the basis of this effect lies with the excessive activation of the RhoA small GTPase in CD151-silenced cells, leading to elevated actin stress fiber formation, and with a far more dynamic cell monolayer in which cell-cell contacts undergo constant remodeling and have a significantly shorter mean lifespan.

The role of RhoA in the regulation of epithelial morphology is complex. Although Rho activity is required for the formation and stability of E-cadherin-dependent cell-cell junctions (Braga et al., 1997; Takaishi et al., 1997; Yamada and Nelson, 2007), as epithelial cell monolayers reach confluence or re-establish contacts after calcium switch, RhoA activity is progressively reduced to a low basal level over a time-course of hours (Arthur et al., 2002; Noren et al., 2001). Elevated RhoA activity appears not to interfere with the early steps of junction formation (Braga et al., 1997), but excessive Rho activity can be disruptive of steady-state junctional stability and organization (Jou and Nelson, 1998; Zhong et al., 1997). These studies are consistent with our calcium switch data, in which junctions initially formed in CD151-silenced cells but failed to remain organized. Cell-cell contact-dependent regulation of RhoA can involve signaling through p190RhoGAP to reduce RhoA activity (Nimnual et al., 2003; Noren et al., 2003; Wildenberg et al., 2006). Suppression of RhoA activity via p190RhoGAP can also result from integrin ligand binding (Bradley et al., 2006; DeMali et al., 2003; Peacock et al., 2007). Thus, convergent signaling to limit or localize RhoA activity might be an important form of integrin-cadherin cross-regulation. Our data are consistent with this view and suggest that tetraspanin CD151 could play an important role in this regulatory mechanism. The inability of the CD151^VR mutant to rescue E-cadherin localization suggests that CD151 might regulate junctional organization via association with α3β1 integrin and/or other integrin partners.

Kreidberg and colleagues (Chattopadhyay et al., 2003; Wang et al., 1999) observed that E-cadherin localization was perturbed in α3-integrin-null kidney epithelial cells, an obsevation similar to that described here. Reconstituting the cells with an α3 mutant with a reduced ability to associate with CD151 in harsh detergents failed to rescue the E-cadherin phenotype (Chattopadhyay et al., 2003). In addition, CD151-silenced kidney cells showed a loss of cortical actin organization (Chattopadhyay et al., 2003). However, the mechanism suggested by these studies entailed a loss of E-cadherin–α-actinin association, elevated β-catenin phosphorylation, a loss of PTPμ expression and reduced E-cadherin-dependent adhesive strength. By contrast, we observed no change in E-cadherin–α-actinin interaction, no increase in β-catenin phosphorylation, no reduction in PTPμ and no obvious reduction in the adhesive function, per se, of E-cadherin. Instead, we found that E-cadherin-dependent junctions initially form normally in our CD151-silenced cells but fail to remain organized, probably due to the highly dynamic nature of the CD151-silenced monolayer. Thus, there could be different mechanisms operating in our human carcinoma cells than in murine embryonic kidney cells.

In another study, Shigeta et al. observed that overexpression of a CD151-GFP fusion protein in A431 cells accelerated the rate at which the cells re-aggregated after dissociation, and slowed migration rate in gap-filling experiments (Shigeta et al., 2003). In addition, an anti-CD151 antibody perturbed F-actin and E-cadherin localization. Rac and Cdc42 activities were elevated in CD151-overexpressing cells, but no change in Rho activation was observed. However, Shigeta et al. performed Rac, Cdc42 and Rho pulldown assays in subconfluent cultures, in which the basal Rho activation level was likely to be high (Noren et al., 2001) and might mask any effect of CD151 overexpression on Rho activation. In preliminary experiments, we did not observe any obvious changes in Rac or Cdc42 activation in our CD151-silenced cell cultures (data not shown). Potentially, Rac and Cdc42 regulation by CD151 are more easily uncovered by CD151 overexpression in A431 cells. Overall, our data generally agree with those of Shigeta and colleagues, with the consensus being that overexpression of CD151 enhances junctions and loss of CD151 destabilizes them.

Further support for this view comes from another recent study in which CD151 was silenced in H5SC carcinoma cells, resulting in redistribution of cortical actin and enhanced stress fibers (Hasegawa et al., 2007). Lastly, our results contrast with a recently published study that demonstrated decreased gap filling in vitro by MCF-10A cells treated with a CD151 siRNA (Yang et al., 2008). Multiple factors could account for this difference, including the use of immortalized instead of tumorigenic cell types, the presence of exogenous EGF in the MCF-10A assay, and possible differences in the balance between the ability of CD151 to enhance cell-substrate adhesion strengthening (Lammerding et al., 2003) and its ability to regulate junctional stability in MCF-10A cells versus A431 cells.

Genetic evidence from mice and humans has identified important roles for CD151 in epithelial integrity, wound healing and pathological angiogenesis elicited by tumor formation (Cowin et al., 2006; Karamatic Crew et al., 2004; Sachs et al., 2006; Takeda et al., 2007). How might the Rho-suppressive activity of CD151 be involved in these processes? Elevated Rho activity upon loss of CD151 could diminish the stability of epithelial adherens junctions, contributing to the structural failure in skin and kidney epithelia of human patients and mice lacking CD151 (Karamatic Crew et al., 2004; Sachs et al., 2006). Wound healing could also be impaired by an overly dynamic epithelium in which RhoA was deregulated and adherens junctions were destabilized. This could help to explain the seeming paradox that CD151-silenced cells can fill a gap in vitro more rapidly than wild-type cells (Fig. 1), whereas wound healing in vivo is impaired in CD151-null mice (Cowin et al., 2006). A less stable, CD151-null epithelium on a provisional wound matrix might be more easily disrupted by in vivo mechanical stresses than in the relatively quiescent environment of a tissue culture monolayer. Interestingly, genetic deletion of α3 integrin in keratinocytes led to an enhanced rate of wound healing in vivo and of gap closure in vitro, and this was ascribed to a loss of anti-migratory α3-integrin-dependent adhesion on laminin-5 (Margadant et al., 2009). However, in contrast to α3-integrin-null keratinocytes, which also migrated more rapidly than wild-type keratinocytes in single-cell assays (Margadant et al., 2009), single-cell migration of our CD151-silenced carcinoma cells was slower than that of wild-type cells (Winterwood et al., 2006). Thus, an enhanced intrinsic migration rate of individual cells due to reduced α3-integrin-dependent adhesion would be unlikely to explain the increased rate of collective migration that we observed for the CD151-silenced monolayers.

Disruptions in Rho family GTPase signaling could also interfere with pathological angiogenesis. For example, excessive RhoA signaling, such as might occur upon loss of CD151, has been linked to loss of bipolarity, detachment, cell death and sprout retraction in an organotypic model of angiogenesis (Mavria et al., 2006). Such a mechanism could help to explain the reduced pathological angiogenesis observed in CD151-null mice (Takeda et al., 2007).

The above considerations indicate that the role of CD151 in metastasis is probably be complex. Several studies have identified CD151 as a potential promoter of metastasis (Ang et al., 2004; Hashida et al., 2003; Kohno et al., 2002; Testa et al., 1999; Tokuhara et al., 2001; Yang et al., 2008). However, if downregulation of CD151 activates RhoA, destabilizes adherens junctions and enhances collective migration, then the potential exists for CD151 to be a suppressor of invasion in some settings. Recently, an anti-metastatic CD151 antibody was shown to prevent tumor-cell detachment from the primary tumor site (Zijlstra et al., 2008). Activation of the ability of CD151 to stabilize cell-cell contacts might represent another mode of action by this anti-metastatic antibody, in addition to its ability to upregulate integrin-dependent adhesion at the trailing edge (Zijlstra et al., 2008).

The potentially opposing roles for CD151 in metastasis could help to explain why its major partner, α3β1 integrin, might act as either a promoter or a suppressor of metastasis, depending on the context (Giannelli et al., 2002). Thus, future studies in which the roles of α3β1 integrin and CD151 are carefully examined in specific stages and types of malignancy will be important to determine in which contexts interfering with either protein might be advantageous. It will also be important to determine whether the ability of CD151 to promote integrin-dependent cell motility and its ability to promote stable cell-cell junctions are separable functions. If so, factors regulating the balance between these opposing functions might determine whether CD151-integrin complexes act as promoters or suppressors of metastasis in specific settings.

Monoclonal antibodies used were anti-E-cadherin (BD Biosciences #610182), SHE78-7 (Zymed Laboratories, Invitrogen), DECMA-1 (Sigma), anti-β-catenin (BD Biosciences #610154), anti-α-catenin (BD Biosciences #610194), anti-p120 catenin (BD Biosciences #610134), anti-α-actinin (BD Biosciences #612577), anti-plakoglobin (BD Biosciences #610253), anti-phosphotyrosine PY20 (BD Biosciences #610000), anti-CD9, ALB6 (Chemicon International), anti-PTPμ BK2 (Santa Cruz Biotechnology #sc-33651), anti-FLAG epitope tag M2 agarose (Sigma) and anti-RhoA (Cytoskeleton #ARHO1). Monoclonal antibodies specific for CD81 (M38), CD151 (5C11 and 11B1.G4), α2 integrin (A2-IIE10) and α3 integrin (A3-X8 and A3-IIF5) were previously referenced (Winterwood et al., 2006). Also used were rabbit polyclonal antibodies recognizing E-cadherin (gift of Jack Lilien and Janne Balsamo, University of Iowa Department of Biology, Iowa City, IA), α3 integrin (Kazarov et al., 2002) and FLAG epitope tag (Sigma). Secondary reagents were Alexa 680-goat anti-rabbit (Invitrogen), IRdye 800-goat anti-mouse and IRDye 800-streptavidin (Rockland), and Cy2-goat anti-mouse and FITC-goat anti-rabbit antibodies (Jackson ImmunoResearch). Human laminin-5 was purified from SCC-25 cell-conditioned medium as described (Winterwood et al., 2006).

The production of fluorescence activated cell sorting (FACS)-sorted A431 epithelial carcinoma cells, with near-total silencing of CD151 (A431 sh3 cells), and the A431 sh3 Rx rescue cells, in which wild-type CD151 is re-expressed, was described previously (Winterwood et al., 2006). The wild-type CD151 Rx cDNA contains two silent mutations near the center of the site targeted by the sh3 CD151 shRNA. Using the CD151 Rx cDNA as a template, the CD151^VR mutant was created with two rounds of recombinant PCR to (i) add a FLAG tag at the amino terminus, (ii) convert the endogenous initiator methionine to leucine, (iii) retain the silent Rx mutations that allow re-expression in CD151-silenced cells and (iv) swap the region of the CD151 EC2 domain from serine 158 to glycine 207 with the corresponding region from human tetraspanin TM4SF2/A15. This region corresponds to the `variable region' of CD151 and contains sequences critical for CD151 interaction with α3β1 integrin (Stipp et al., 2003). The CD151^VR mutant is not bound by anti-CD151 antibodies 5C11 or 11B1.G4, and was recognized weakly or not at all by available anti-TM4SF2 antibodies. CD151^VR is readily detected by anti-FLAG antibody. We also created a FLAG-tagged version of wild-type CD151 for use in comparison with the CD151^VR mutant.

A431 cells were cultured in standard medium (DME with 10% FBS, 2 mM L-glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin). Gaps were inscribed in freshly confluent monolayers using a yellow micropipet tip, and the cultures were refed with serum-free medium [SFM; DME with 5 mg/ml cell culture grade BSA (Sigma-Aldrich) and 25 mM HEPES pH 7.2]. Images were acquired every 5 minutes for 18-24 hours, using a stage incubator and video-microscope system described in our previous study (Winterwood et al., 2006). Gap area at 1-hour intervals was measured using NIH Image 1.63 software to calculate the gap closure rate in μm²/h. To measure mitoses during the assays, 175-μm wide strips on either side of the initial gap were analyzed frame-by-frame. Mitotic events were clearly visualized as cells rounded up, divided and spread back into the monolayer. Single-cell velocities on laminin-5 were measured for wild-type, CD151-silenced sh3 cells and Rx rescue cells in 3-hour videos, as previously described (Winterwood et al., 2006). In another series of experiments, unwounded monolayers in standard growth medium were imaged as described above, and NIH ImageJ 1.63 software was used to record the XY position of cell centroids in frames 30 minutes apart, for 35 cells per experiment. Custom Java software was then used to calculate the mean velocity of cell migration within the monolayers. The change in the size of individual cell-cell contact was followed every 5 minutes for 2 hours in one experiment. The lifespan of individual cell-cell contacts was assessed by examining frames 15 minutes apart for a total of 12 hours in another experiment.

Cells cultured on coverslips were fixed in 10% formalin in HEPES-buffered saline (HBS) with 4% sucrose and 1 mM CaCl₂, rinsed twice with TBS, and blocked with 10% goat serum in PBS. For intracellular epitopes, cells were permeabilized with either 0.2% NP-40 or 0.1% saponin during the blocking step. Cells were stained for 1 hour with primary antibodies in blocking buffer, and then for 45 minutes with Cy2-goat anti-mouse antibody. For F-actin staining, Alexa 594-phalloidin (Invitrogen) diluted 1:40 in PBS was added to cells after the secondary antibody step. Cells were mounted in FluorSave (Calbiochem) or Prolong Gold (Invitrogen) and analyzed by standard fluorescence microscopy as previously described (Winterwood et al., 2006), or by laser scanning confocal microscopy on a Leica SP2 AOBS inverted microscope.

Cells were lysed in PBS with 1% detergent, protease inhibitors (2 mM PMSF, 10 μg/ml aprotinin, 5 μg/ml leupeptin and 5 μg/ml E-64) (Roche Diagnostics), and HALT phosphatase inhibitors (Pierce Biotechnology). Detergents were NP-40, Brij 96V or a 1:1 mixture of Brij 96V and Brij 99 (all from Sigma-Aldrich). In some experiments, cells were biotinylated with 0.1 mg/ml Sulfo-NHS-LC biotin (Pierce Biotechnology) in HBS (20 mM HEPES pH 7.2, 150 mM NaCl) for 1 hour at room temperature and then rinsed three times with HBS prior to lysis. Between 2×10⁶ and 6×10⁶ cells were used for each experiment. Lysates were clarified and normalized for protein concentration. Immune complexes were collected, separated by SDS-PAGE, and blotted to nitrocellulose as previously described (Winterwood et al., 2006). Blots were blocked with AquaBlock (East Coast Bio, North Berwick, ME) and probed with primary antibodies diluted in 10% AquaBlock in TBST (20 mM Tris pH 7.5, 150 mM NaCl, 0.1% Tween-20). Blots were developed with Alexa 680-goat anti-rabbit and/or IRDye 800-goat anti-mouse secondary antibodies and analyzed with a LiCor infrared fluorescence blot imager (LI-COR, Lincoln, NE). Biotinylated proteins were detected with IRDye 800-streptavidin.

E-cadherin-Fc protein (gift of W. James Nelson, Stanford University, Stanford, CA) was purified from transfected CHO cell-conditioned medium on a protein G-Sepharose column; size, purity and dimeric structure were confirmed by SDS-PAGE and Coomassie blue staining. For adhesion assays, wells of a 96-well plate were coated overnight at 4°C with 40 μg/ml of E-cadherin-Fc fusion protein diluted in HBS with 2 mM CaCl₂ (HBSC), and then blocked with 10% heat-inactivated BSA in HBSC. To prepare cells for the assay, 7.5×10⁵ of each cell type were plated at low density in a T75 flask the day before the assay. On the day of the assay, the cells were harvested with 0.25% trypsin with 0.38 g/l EDTA·4Na (Invitrogen) and then replated at 7.5×10⁵ cells per T75 flask for 4 hours to allow for recovery. Subsequently, the cells were harvested again, this time with 0.1% trypsin without EDTA in HBSC. Harvested cells were rinsed with, and resuspended in, DME with 2% FBS. This protocol was designed to yield a single-cell suspension of contact-naïve cells with E-cadherin cell-surface expression preserved; 1.8×10⁴ cells/ml were plated per well. Some samples were treated with 5 mM EGTA or with anti-E-cadherin function-blocking antibodies for 10 minutes prior to plating. After 30 minutes, non-adherent cells were removed by rinsing; adherent cells were quantified by crystal violet staining and compared to total cells plated (as measured in poly-L-lysine-coated wells) and non-specific adhesion in BSA-coated wells, as previously described (Winterwood et al., 2006).

An E-cadherin-GFP (EcadGFP) fusion construct (gift of W. James Nelson) was transfected into A431 wild-type and CD151-silenced sh3 cells. Stable transfectants were selected with G418, and EcadGFP-expressing cells were sorted by FACS. For calcium switch experiments, cells growing on coverslips were placed in low-calcium medium (CaCl₂-free DME with 2% extensively dialyzed FBS and 5 μm CaCl₂) for 15 minutes. Then, calcium was restored to 1.8 mM for 30 minutes, 1 hour or 3hours, and cells fixed and examined by immunofluorescence microscopy.

For RhoA activity assays, 5×10⁶ cells were plated in 100 mm dishes and cultured for 3 days to achieve complete confluence. Cells were then lysed with 600 μl of 25 mM HEPES pH 7.2, 150 mM NaCl, 1% NP-40, 10 mM MgCl₂, 10% glycerol and protease inhibitors, followed by centrifugation at 16,000 × g for 4 minutes at 4°C. Active RhoA was recovered for 30 minutes at 4°C with 30 μg of GST-Rhotekin bound to glutathione Sepharose and analyzed by SDS-PAGE followed by blotting for RhoA. Blots were quantified using a LiCor infrared-fluorescence blot imager. As a positive control for RhoA activation, some dishes were harvested with trypsin plus EDTA, and the cells were held in suspension for 5 minutes prior to lysis and RhoA pulldown.

To assess the role of RhoA activity on junctional stability in CD151-silenced sh3 cells, cells plated on coverslips were treated for 1 hour with 2 μg/ml cell-permeable C3 transferase (Cytoskeleton, Denver, CO). Treated and untreated cells were then fixed and stained for E-cadherin and F-actin, as above. In parallel experiments, freshly confluent monolayers were analyzed by time-lapse microscopy, as described above, for 3 hours before and 3 hours after the addition of 2 μg/ml cell-permeable C3 transferase.