Cell polarity triggered by cell-cell adhesion via E-cadherin

Polarization of the cell-motility apparatus is a crucial first step for migration of cell populations in contexts such as morphogenesis and wound healing. During migratory polarization, many structures, including the centrosome, Golgi complex, lamellipodia and microtubules, organize towards the leading edge of the cell, whereas other structures, including the nucleus, stress fibers and mature focal adhesions, localize towards the cell rear (Ridley et al., 2003; Lauffenburger and Horwitz, 1996; Wittmann and Waterman-Storer, 2001). Scrape wounding of a monolayer of cells triggers reorientation of this migration machinery (Gotlieb et al., 1981; Kupfer et al., 1982; Etienne-Manneville and Hall, 2001; Gomes et al., 2005) and thus has been widely used to study cell polarization. In this model, Cdc42 is required for polarization in several cell types (Nobes and Hall, 1999; Palazzo et al., 2001; Watanabe et al., 2004; Tzima et al., 2003; Shen et al., 2008), and it has largely been assumed that adhesion to the extracellular matrix (ECM) in protrusions into the scrape wound is the source of the polarizing signal (Etienne-Manneville and Hall, 2001). Interestingly, fibroblast wound-edge cells polarize before significant cell protrusion takes place (Gomes et al., 2005) and intercellular communication appears to be required (Schlessinger et al., 2007), suggesting that cell-ECM adhesion is not the first or only polarizing signal. Asymmetries in cell-cell contact formed by monolayer wounding also might contribute to polarization of the cell-migration machinery.

In a variety of settings, cell-cell contact induces changes in the structural organization of cells. Cell-cell interaction at the immune synapse polarizes the nucleus and Golgi complex of the T cell to direct the transport of immune effectors (Kupfer et al., 1983). Epithelial cell-cell contact that is mediated by epithelial (E)-cadherin is required for proper establishment of apical-basal polarity (Wang et al., 1990; Nejsum and Nelson, 2007). Furthermore, E-cadherin engagement leads to significant rearrangements of actin and microtubule cytoskeletons, proximal (Chu et al., 2004; Drees et al., 2005; Yamada and Nelson, 2007; Scott et al., 2006; Kovacs et al., 2002) and distal (Ehrlich et al., 2002; Yamada and Nelson, 2007; Adams et al., 1998; Vaezi et al., 2002) to the cell-cell contact. These lines of evidence suggest that cell-cell adhesion can induce cells to segregate organelles and domains, and rearrange the cytoskeleton.

Here, we sought to investigate whether cell-cell contact might also guide polarization of the cell-migration machinery. Using a micropatterning approach, we found that E-cadherin-mediated cell-cell adhesion polarizes and directs the cell-migration machinery away from neighboring cells, and occurs prior to cell migration. Such polarity requires the actin cytoskeleton and Cdc42 activity. Polarity that is induced by cell-cell contact probably contributes to coordinated cell movements such as those occurring during morphogenesis and wound healing.

We observed that, in the absence of serum, normal rat kidney epithelial cells (NRK-52E) failed to polarize and migrate in sparse culture, yet did so robustly at the edge of an experimental scrape wound (Fig. 1A). The dynamics of this process at the wound edges were such that cells were randomly oriented immediately after wounding, but by 4 hours were oriented such that their centrosomes `faced' the wound in both the presence or absence of serum (Fig. 1B-D; supplementary material Fig. S1). Such polarization was similar to that observed in other cell types (Kupfer et al., 1982; Nobes and Hall, 1999; Etienne-Manneville and Hall, 2001; Palazzo et al., 2001), and was accompanied by significant protrusive activity of the wounded edge.

To eliminate such protrusions as a potential cause of reorientation, we plated cells onto patterned substrates to form monolayers with defined edges. Here, edge cells polarized similarly to the scrape wound (Fig. 1E-G). To explore preliminarily whether monolayer geometry might guide polarity, we examined multicellular patterns for which predicted forces vary along the edge (Fig. 1H) (Nelson et al., 2005). We observed similar high levels of orientation regardless of monolayer shape or size, suggesting that geometry did not contribute to edge-cell polarity.

We reasoned that asymmetric cell-cell contacts experienced by cells at the edges of scrape wounds and patterns might underlie their orientation. To test this idea, we plated cells into agarose microwells (Nelson et al., 2007), such that two cells land within a microwell, spread against the substrate and form a cell-cell contact between them (Fig. 1I,L). In this setting, protrusion and adhesion beyond the pattern are prevented. Soon (2 hours) after plating, few cells had made contact, but those that had were randomly oriented (Fig. 1J,K). However, 6 hours after plating, 80% of cells were oriented, similar to the level observed in scrape-wounded and patterned monolayers, both in the presence or absence of serum (supplementary material Fig. S2). Orientation was reminiscent of what occurs following mitosis, in which daughter cells polarize and respread away from the cytokinetic furrow (Robbins and Gonatas, 1964). However, cells that were inhibited from proliferating with mitomycin C still polarized (supplementary material Fig. S3), indicating that mitosis did not drive orientation here.

Measurements revealed that 12 hours after plating, the nucleus was located near the cell-cell contact, whereas the centrosome was oriented to the distal side of the nucleus (Fig. 1M,N), accounting for polarity. The disruption of microtubules with nocodozole (10 μM) led to mislocalization of the centrosome (supplementary material Fig. S4), which is consistent with the results from studies that demonstrated the importance of microtubules in centrosome centration (Vallee and Stehman, 2005; Gomes et al., 2005). Together, these data suggest that cell-cell contact is important in directing cell polarization.

Although nuclear and centrosomal positions are associated with directional ruffling and migration (Ridley et al., 2003; Gotlieb et al., 1981), such ruffling is prevented in agarose microwells. To examine ruffling activity when the nucleus and centrosome were polarized, we patterned cells on a flat surface that enabled membrane protrusion (Fig. 2A; supplementary material Movie 1). Kymographs from time-lapse recordings suggested that membrane ruffling was present distal, but not proximal, to the cell-cell contact. To report the extent and location of ruffling, we summed the absolute difference images between all successive frames of time-lapse recordings. The resulting `activity map' revealed higher activity distal compared with proximal to the cell-cell contact (Fig. 2B), indicating that membrane ruffling was also polarized.

Together with the organelle analysis, these data suggested that cells were `primed' for directed movement away from cell-cell contacts. To test this directly, we used electrodesorption (Jiang et al., 2003) to release cell patterns, allowing migration. Oriented cells that were released from patterns migrated away from each other in a directed fashion (Fig. 2C; supplementary material Movie 2). Collectively, these data implicate a role for cell-cell contact in initiating polarized cell structure, ruffling and migration.

Because cell-cell contact correlated with polarization of the migration apparatus, we turned to identifying the adhesion receptor(s) that are responsible. Removing Ca²⁺ from the media after cells had polarized abolished orientation (Fig. 3A), suggesting that Ca²⁺-dependent cell-cell adhesion was required for polarity. E-cadherin is a crucial receptor in organizing epithelial cell-cell adhesions and associated cytoskeletal structures (Scott et al., 2006; Yamada and Nelson, 2007; Adams et al., 1998), and its homophilic binding is Ca²⁺-dependent. To directly modulate E-cadherin binding, we infected cells with adenovirus harboring either GFP (Ad-GFP) or a truncated, dominant-negative mutant of E-cadherin (Ad-EΔ) prior to plating in patterns. Ad-GFP- but not Ad-EΔ-infected cells displayed typical E-cadherin localization to the cell-cell contact, and polarized in both the presence or absence of serum (Fig. 3B,C; supplementary material Fig. S5), indicating that E-cadherin engagement is required for the establishment of polarity. To determine whether E-cadherin engagement was required to also maintain polarity, we waited until cells polarized in the patterns (6 hours), and then infected them with either Ad-EΔ or Ad-GFP. Oriented cells that were infected with Ad-EΔ subsequently randomized their orientation (supplementary material Fig. S6), demonstrating that E-cadherin is required for both the establishment and maintenance of polarity. Antagonizing E-cadherin also contributed to loss of polarized membrane ruffling on patterned surfaces (Fig. 3D,E; supplementary material Movies 3 and 4). These findings suggest that E-cadherin engagement at the cell-cell contact is required to trigger polarization of the cell-migration machinery on patterned substrates.

The contribution of E-cadherin engagement to polarization and migration induced by scrape wounding remained unclear. Cells lacking E-cadherin engagement (supplementary material Fig. S7) were still able to develop lamellipodial extensions and migrate directionally to close an experimental scrape wound at a rate that was comparable to their control counterparts (Fig. 3F). Surprisingly, even in the presence of this directional extension, new adhesion and motility, Ad-EΔ-infected cells exhibited random centrosomal-nuclear polarity during wound closure (Fig. 3G). These data suggested that cells might principally use cell-cell contact to guide polarization of the nuclear-centrosomal axis.

The actin cytoskeleton is required for scrape-wound-induced reorientation (Gomes et al., 2005). To test whether the actin cytoskeleton was involved also in cell-cell-contact-induced polarity, we disrupted actin filaments with cytochalasin D (1 μM) (Fig. 4A). Although this manipulation did not prevent the accumulation of β-catenin and E-cadherin at the cell-cell contact (Fig. 4B), it substantially disrupted polarity in both the presence or absence of serum (Fig. 4C; supplementary material Fig. S8). In particular, shifts in both nuclear and centrosomal position accounted for randomized orientation in cells that had a compromised actin cytoskeleton (Fig. 4D).

The Rho family of GTPases – RhoA, Rac1 and Cdc42 – are central mediators of actin-cytoskeleton reorganization (Ridley and Hall, 1992; Ridley et al., 1992; Nobes and Hall, 1999), and have been linked to cadherin-mediated cell-cell adhesion (Braga and Yap, 2005). Inhibiting RhoA signaling via adenoviral-based delivery of the dominant-negative mutant Rho^N19, the ROCK inhibitor Y27632 (50 μM) or the myosin-II inhibitor blebbistatin (25-100 μM) all attenuated actomyosin-based stress fibers (data not shown), but had no impact on cell-cell contact-induced polarity (Fig. 4E; supplementary material Fig. S9). Likewise, antagonizing Rac1 signaling by treating cells with NSC23766 (50 μM) abrogated lamellipodial formation (data not shown), but had no impact on polarity (Fig. 4F). Taken together, these results demonstrate that RhoA and Rac1 signaling are dispensable for polarity induced by cell-cell contact.

Cdc42 is known to be essential for polarity in numerous settings (Etienne-Manneville, 2004) and can be activated by E-cadherin-mediated cell-cell adhesion (Kim et al., 2000; Noren et al., 2001). To test whether Cdc42 is involved also in polarity that is induced by cell-cell contact, we antagonized Cdc42 signaling via short interfering RNA (siRNA). We observed significant inhibition of Cdc42 expression and activity, but no attenuation of the cell-cell contact, relative to scrambled siRNA (Fig. 4G,H). Cells with antagonized Cdc42 signaling in microwells failed to orient in the presence or absence of serum (Fig. 4I; supplementary material Fig. S10). Similar to disruption of the actin cytoskeleton, shifts in both nuclear and centrosomal position accounted for this loss in polarity (Fig. 4J). Taken together, these data suggest that the polarizing signal that is triggered by cell-cell adhesion involves E-cadherin, the actin cytoskeleton and Cdc42.

It has been thought that the primary orientation signal at the edges of scrape-wounded monolayers arises from new cell-matrix adhesions as cells extend into the woundbed (Etienne-Manneville and Hall, 2001). Here, we show that asymmetric, E-cadherin-mediated cell-cell contact at wound edges is required to orient cells, even on patterned substrates that prevent cell-matrix protrusions beyond the pattern. These findings are consistent with prior reports in which scrape-induced polarization occurs prior to significant protrusion formation (Gomes et al., 2005). Interestingly, our studies show that cells with compromised cell-cell contacts have disrupted polarization axes, but nonetheless extend lamellipodia and close wounds with similar efficiency as their control counterparts. These data suggest that cell-cell contact is required for polarization of the nuclear-centrosomal axis, but additional signals such as new cell-ECM adhesion are required for fully directed migration. This decoupling of polarization of various intracellular structures highlights the complexity by which cells probe asymmetries in their extracellular environment to produce coordinated responses, and motivates a more detailed examination of the separate effects of cell-cell and cell-matrix cues.

Previous studies have identified a role for Cdc42 in mediating cell-ECM-induced polarization of actin (Etienne-Manneville and Hall, 2001; Etienne-Manneville, 2004; Thery et al., 2006). Our data indicate that cell-cell contact also requires Cdc42 activity and the actin cytoskeleton to polarize cells. Interestingly, whereas some components of the polarizing machinery are shared (actin, Cdc42), others are not; serum and myosin II are required in some scrape-induced polarity models (Palazzo et al., 2001; Gomes et al., 2005) but not here in cell-cell contact-induced polarity. It is therefore not unreasonable to suggest that motile cells might integrate directional signals from multiple sources, including soluble factors, cell-ECM adhesion and cell-cell adhesion, in order to direct migration. Although additional studies will be required to elucidate the interaction of these and other signals in mediating polarity, it is clear that cells have developed mechanisms to respond to asymmetries in cell-cell adhesions, such as those occurring during morphogenesis and wound healing.

NRK-52E cells (ATCC) were cultured in 10% fetal bovine serum (Gibco) in DMEM (Gibco) unless otherwise indicated. Cells were G0-synchronized as previously described (Liu et al., 2006). Reagents included: 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI; Sigma), anti-pericentrin (Covance), anti-α-tubulin (Sigma), anti-β-catenin (Santa Cruz Biotechnology), anti-ZO-1 (Zymed), anti-Cdc42 (BD), anti-E-cadherin (BD), phalloidin–Alexa-Fluor-488 (Invitrogen), mitomycin C (Sigma), blebbistatin (EMD), Y27632 (Tocris), NSC23766 (Calbiochem), scrambled (5′-UAACGACGCGACGACGUAAtt-3′) and Cdc42 (5′-GACUACGACCGUUAAGUUAtt; CAACUGUUUUUGACAACUAtt-3′) siRNA (Ambion), OptiMEM (Gibco), and Lipofectamine 2000 (Invitrogen). Adenoviral vectors harboring mutant E-cadherin or Rho bicistronic to GFP were prepared as previously described (Liu et al., 2006; Nelson and Chen, 2003).

Surface and microwell patterns were generated via microcontact printing and agarose patterning as previously described (Tan et al., 2004; Nelson et al., 2007). Dynamically adhesive substrates were generated using self-assembled monolayers of HS(CH₂)₁₅CH₃ (Sigma) and HS(CH₂)₁₁(OCH₂CH₂)₃OH (EG₃; ProChimia) on gold (Jiang et al., 2003). A potentiostat (Gamry Instruments) was used to desorb EG₃ and thereby permit cell migration.

Cells were fixed in ice-cold 1:1 methanol:acetone for 10 minutes at –20°C (for E-cadherin), or pre-warmed 2% paraformaldehyde in microtubule-stabilizing buffer [1 mM EGTA, 1 mM MgSO₄, 4% (w/v) poly(ethylene glycol) 8000 and 1% (v/v) Triton X-100 in 0.1 M PIPES, pH 6.75] for 10 minutes at 37°C (for α-tubulin, β-catenin, pericentrin, ZO-1 and phalloidin), and labeled with antibodies in 10% goat serum.

GTP-loaded Cdc42 was measured using a kit (Upstate) as previously described (Liu et al., 2006). Total and GTP-loaded Cdc42 were normalized to α-tubulin or GAPDH and quantified relative to control.