p120-catenin controls contractility along the vertical axis of epithelial lateral membranes

ABSTRACT

In vertebrate epithelia, p120-catenin (hereafter referred to as p120; also known as CTNND1) mediates E-cadherin stability and suppression of RhoA. Genetic ablation of p120 in various epithelial tissues typically causes striking alterations in tissue function and morphology. Although these effects could very well involve p120's activity towards Rho, ascertaining the impact of this relationship has been complicated by the fact that p120 is also required for cell–cell adhesion. Here, we have molecularly uncoupled p120's cadherin-stabilizing and RhoA-suppressing activites. Unexpectedly, removing p120's Rho-suppressing activity dramatically disrupted the integrity of the apical surface, irrespective of E-cadherin stability. The physical defect was tracked to excessive actomyosin contractility along the vertical axis of lateral membranes. Thus, we suggest that p120's distinct activities towards E-cadherin and Rho are molecularly and functionally coupled and this, in turn, enables the maintenance of cell shape in the larger context of an epithelial monolayer. Importantly, local suppression of contractility by cadherin-bound p120 appears to go beyond regulating cell shape, as loss of this activity also leads to major defects in epithelial lumenogenesis.

Highlighted Article: The distinct activities of p120 towards E-cadherin and Rho are molecularly and functionally coupled, and this in turn enables the maintenance of cell shape in the larger context of an epithelial monolayer.

INTRODUCTION

A universal principle of organization for polarized epithelial cells is the physical and functional segregation of membranes into three distinctive domains. Basal membranes impart anchorage, lateral membranes organize adhesive contacts and apical membranes establish a free surface for exchange of materials (Mostov et al., 2003). However, epithelial tissues vary widely in size and shape to accommodate diverse epithelial functions (Gumbiner, 1996, 2005). Although determinants of epithelial cell fate have been well described, the molecular mechanisms controlling cell height and shape are poorly understood. In Drosophila melanogaster, recent evidence suggests that, in addition to cell fate specification, molecular gradients of the morphogen Decapentaplegic (Dpp) exercise important position-specific effects on epithelial cell architecture (Gibson, 2005; Shen and Dahmann, 2005). Although mechanistic details are still unclear, the Dpp pathway appears to modulate epithelial cell height, in part, by controlling compartmentalization of Rho1 (the Drosophila homolog of RhoA) activity along the length of the lateral cell membrane (Gibson, 2005; Shen and Dahmann, 2005; Widmann and Dahmann, 2009). Whether (and how) Rho activity affects cell height in vertebrate epithelial systems is currently unknown.

A potentially important discrepancy between Drosophila and vertebrate systems is the relative function of p120-catenin (hereafter referred to as p120; also known as CTNND1), which binds directly to the cytoplasmic juxtamembrane domain of E-cadherin in both systems. In C. elegans and Drosophila, p120 is considered dispensable, as genetically null adults are clearly viable (Fox et al., 2005; Myster et al., 2003; Pacquelet et al., 2003; Pettitt et al., 2003), albeit sensitive to stress (Stefanatos et al., 2013). In contrast, p120 gene ablation in vertebrates is embryonic lethal. Notably, conditional p120 KO phenotypes in mice are often striking but highly variable from one epithelial cell type to another. In many human cancers, p120 is mislocalized and/or downregulated to varying degrees, suggesting potential roles in tumor progression or metastasis (Davis and Reynolds, 2006; Kurley et al., 2012; Smalley-Freed et al., 2010). Interestingly, in contrast to its Drosophila and C. elegans counterpart, vertebrate p120 is essential for cadherin stability. Removal of p120 in most epithelial cell types causes rapid internalization of the cadherin complex in vitro and in vivo (Davis, 2003; Davis and Reynolds, 2006; Kurley et al., 2012; Marciano et al., 2011; Smalley-Freed et al., 2010; Xiao, 2003). In Drosophila, the E-cadherin-containing adherens junction is often restricted to an apical compartment delimited by the septate junction. Compartmentalized suppression of Rho occurs along the lateral domain by a cadherin-independent mechanism and and promotes an increase in cell height (Widmann and Dahmann, 2009). Vertebrate E-cadherin (along with p120-, α- and β-catenins), by contrast, is typically localized along the entire lateral membrane (Wu et al., 2014). Notably, vertebrate p120 is well established as an inhibitor of Rho (Fang, 2004; Noren et al., 2000; Ponik et al., 2013; Reynolds et al., 2000; Schackmann et al., 2011; Wildenberg et al., 2006; Zebda et al., 2013). In the cytoplasm, inhibition of Rho by p120 occurs by a Rho GDP-dissociation inhibitor (Rho GDI)-like mechanism and is mutually exclusive with binding to E-cadherin (Anastasiadis et al. 2000; Kourtidis et al., 2015). Cadherin-bound p120, by contrast, can interact with a spectrum of Rho mediators [e.g. Rho guanine-nucleotide-exchange factors (GEFs), Rho GTPase-activating proteins (GAPs), ROCK proteins and Shroom3) depending on parameters such as cell and/or tissue type, and subcellular localization (Lang et al., 2014; Noren et al., 2000; Ponik et al., 2013; Smith et al., 2011; Wildenberg et al., 2006). For example, in many polarized columnar epithelia, p120 interacts apically with specific RhoGEFs that modulate apical constriction (Lang et al., 2014) and basolaterally with RhoGAP family members (Klompstra et al., 2015; Zebda et al., 2013; Kourtidis et al., 2015). These observations suggest that p120 acts as a coordinating hub for mediators of local Rho activity and raise the possibility that p120 in vertebrates could participate in regulating lateral cell height through local suppression of Rho.

Vertebrate p120 function has been extensively studied in conditional knockout (KO) mice (Davis and Reynolds, 2006; Kurley et al., 2012; Marciano et al., 2011; Perez-Moreno et al., 2006; Smalley-Freed et al., 2010). Phenotypes vary widely depending on the organ and tend to involve striking changes in tissue morphology. Although these phenotypes could potentially be linked to Rho (Ponik et al., 2013), distinguishing Rho-mediated effects from those caused by cadherin destabilization have been inherently difficult because the effects on Rho mediated by p120 are epistatic to, and dependent on, its cadherin-stabilizing activity. Nonetheless, cadherin stability cannot by itself account for the wide spectrum of p120-KO phenotypes observed in vitro and in vivo (Davis and Reynolds, 2006; Dohn et al., 2009; Kurley et al., 2012; Perez-Moreno et al., 2006, 2008; Ponik et al., 2013). Additionally, we and others have found that physiologically relevant results are often masked or blocked altogether when the cells are cultured on hard surfaces (Baker and Chen, 2012; Brugge, 2012; Dohn et al., 2009; Paszek et al., 2005; Töyli et al., 2010). Moreover, epithelial cells that are columnar in vivo adopt completely different shapes when cultured by conventional means on plastic. MDCK cells, for example remodel into very flat disc-shaped cells featuring wide basal footprints and lateral domains that make strong cell–cell contacts but that are otherwise almost non-existent. We have therefore transitioned to two-dimensional (2D) cultures on thick collagen pads (which enable cuboidal to columnar morphology) and/or three-dimensional (3D) cell cultures in collagen. Here, using a vertebrate epithelial cell model (i.e. MDCK II cells), we separate the cadherin-stabilizing and RhoA-suppressing functions of p120 under conditions that, for the first time, permit selective assessment of phenotypes caused by the impact of p120 on Rho. Unexpectedly, selectively removing the Rho-suppressing p120 activity dramatically disrupted the integrity of the apical surface, irrespective of E-cadherin stability. The physical defect stems from excessive actomyosin contractility along the vertical axis of lateral membranes, causing dramatic basal dislocation of the tight junction and expansion of the apical domain, leaving cell polarity intact. Moreover, the impact of this excess contractility goes beyond regulation of cell shape, as the effect is accompanied by major defects in epithelial lumenogenesis. Importantly, this defect is completely reversed by inhibition of ROCK proteins or myosin, irrespective of E-cadherin stability. Thus, although most p120 ablation phenotypes can be attributed to adhesion defects, the phenotypes described here are rescued by suppression of Rho but not E-cadherin.

RESULTS

p120 ablation disrupts the apical surface of MDCK cell monolayers leaving cell polarity intact

In many epithelial cell types, p120 ablation leads to complete loss of cell–cell adhesion (e.g. MCF10A and A431 cells) (Kurley et al., 2012; Xiao, 2003), making it difficult to distinguish between direct consequences of p120 loss and collateral fallout associated with loss of all contact-dependent signaling. Moreover, p120 activity has important effects that manifest only in the context of adhesion-intact cell monolayers (e.g. lumen formation and collective migration) and are thus masked by loss of cell–cell contacts. MDCK cells circumvent many such issues because intercellular adhesion can be maintained by E-cadherin-independent junctions upon knockdown of p120, despite the near complete loss of adherens junctions. Notably, tight junctions and desmosomes are unaffected (Dohn et al., 2009).

When cultured on plastic, the morphologies of wild-type (WT) and p120-knockdown (KD) MDCK cells were essentially identical (data not shown). When plated on collagen, however, the cells polarized, and developed sufficient height to qualify as cuboidal or columnar cell monolayers, even when subconfluent. In this scenario, p120 KD induced dramatic changes in cell morphology. By contrast, overexpression of p120 (isoform 1A or 3A) by at least twofold had no overall impact on cell shape (Fig. S1C-E). By using transmission electron microscopy (TEM), we observed large gaps between neighboring cells only in p120-KD cells (Fig. S1F). Although the tight junction was retained, the apical surface at cell–cell contacts was substantially distorted (Fig. S1F, white arrow). To further characterize this effect, the cells were immunostained for ezrin (an apical marker) and the tight junction marker cingulin. Normally, ezrin staining is confined to a thin zone (i.e. the apical section) across the top of the epithelial monolayer and highlights the perfectly flat apical surface of WT MDCK monolayers. Note that there was little or no detectable ezrin staining in other confocal planes (e.g. middle or basal sections) (Fig. 1A,B). However, in the p120-KD cells, confocal cross sections showed that ezrin staining was clearly present well into the middle section of the cell (Fig. 1A, arrows). Fig. 1C shows confocal cross sections of the same experiment; note the aberrant presence of ezrin staining in the intercellular space across the middle section (Fig. 1C, arrow). The lower panels of Fig. 1C show that p120 staining was confined to the lateral membranes in WT cells and was substantially downregulated in p120-KD cells. 3D reconstructions graphically illustrate the topography of ezrin-stained WT and p120 KO surfaces viewed from the top (left panel) and bottom (right panel) (Fig. 1D). Whereas the WT surface was relatively flat (see Movie 1), the KD surface was deeply invaginated, as illustrated dramatically by the bottom-up view (Fig. 1D, arrow, lower right panel, see also Movie 2).

Fig. 1.

p120 KD induces apical membrane expansion leaving apico-basal polarity intact. (A) p120 KD causes striking intercellular invaginations (lower panel, arrows) of the otherwise normally flat apical membrane (upper panel). x-z-stacks are shown. (B) Schematic illustrating the apical membrane expansion phenotype. The dotted line shows the apical and middle confocal sections where images were taken in C. (C) Membranous ezrin staining present in the middle section indicates deeply invaginated apical membranes in p120-KD cells. (D) 3D reconstruction of WT and p120 KD apical membranes (ezrin staining). Top-down (left) and bottom-up (right) views are shown. The bottom-up view in the p120-KD cells illustrates extensive and continuous apical expansion extending well into the lateral domain (white arrowheads). Also see Movies 1 and 2. (E) Rescue of flat apical morphology by addback of p120. A mosaic area containing p120-KD cells (right) and the same cells rescued by expression of human p120 (hp120, left) is shown. The white dashed box outlines the inset magnified on the right. The cell in the crosshairs shows junctional staining of p120 on three sides. The fourth side (white arrow) lacks a p120-positive neighbor and exhibits expansion of the apical membrane on that side only. (F) Co-staining with ezrin and the tight junction marker cingulin, showing that the boundary between the apical and lateral compartments are retained in both WT and p120-KD cells. (G) A maximum intensity projection of cingulin staining confirms that tight junctions are intact and continuous in both WT and p120-KD cells. (H) Treatment with 20 μM myristoylated PKCζ-PS (aPKC inhibitor) for 48 h fails to rescue the apical membrane defect in p120-KD cells. (I) Immunoblotting (WB) shows that p120 KD does not affect the protein stability of ezrin, irrespective of the culture condition used. Tub, tubulin. Scale bars: 10 µm.

Interestingly, the dome-shaped apical surfaces are reminiscent of apical expansion phenotypes induced by overexpression of members of the apical polarity complex (i.e. Par3 and Par6 proteins) (Chalmers et al., 2005) or downregulation of members of the lateral polarity complex (i.e. Lgl, Scrib or Dlg proteins) (Yamanaka, 2006). Overexpression of gp135 (also known as CNTN1) or downregulation of KIBRA (also known as WWC1 in mammals) can also induce apical expansion (Nielsen et al., 2000; Yoshihama et al., 2011). For the polarity proteins, apical expansion is driven by mis-targeting of apical membrane proteins to lateral membranes, irrespective of the placement of tight junctions (Tanentzapf and Tepass, 2002). gp135 accumulation, by contrast, physically expands the apical membrane, in part through recruitment of NHERF (also known as SLC9A3R1) and ezrin. Apical membrane also expands in 3D MDCK cysts upon KIBRA KD, owing to hyperactivation of aPKC. Notably, in the latter cases (gp135 and KIBRA KD), the mechanism was shown to not involve mis-targeting to apical or basolateral membranes. Instead, the tight junction is retained at the boundary between apical and lateral membranes but displaced basally by the expanding apical surface (Nielsen et al., 2000; Yoshihama et al., 2011). To determine whether either of these mechanisms was responsible for the p120-KD phenotype, we first examined the effect of p120 KD on the placement of tight junctions relative to the apical surface. Importantly, cingulin staining showed that tight junctions localized at the very tip of the surface invaginations, illustrating that the boundary between the apical and lateral domains remains intact (Fig. 1F). Maximal intensity projections of cingulin staining to the x-y plane showed that the p120-KD tight junctions remained circumferentially continuous, although being dislocalized basally (Fig. 1G). Because this result was morphologically consistent with the gp135–NHERF–ezrin and KIBRA-KD mechanism, we measured ezrin protein levels in the WT and p120-KD cells and found no differences (Fig. 1I). Additionally, inhibition of atypical protein kinase C (aPKC) activity with the myristoylated PKCζ-PS inhibitor failed to rescue the apical membrane defect in p120-KD cells (Fig. 1H). Thus, although morphologically similar to the apical expansion phenotypes associated with polarity genes, it appears that neither of the mechanisms accounts for the p120 ablation phenotypes here.

p120–E-cadherin interaction is essential for maintaining a flat apical membrane

In p120 rescue experiments, analysis of mosaic p120 rescue suggested that suppression of the apical membrane invagination is dependent on the p120–E-cadherin-mediated cell–cell contacts. For example, the cell in the crosshairs of Fig. 1E (right panel) makes E-cadherin-based adhesions on three sides (Fig. 1E, arrowhead). The fourth side (Fig. 1E, arrowheads), by contrast, does not, and exhibits apical defects (i.e. invagination) on that side only. Thus, although the phenotype could in theory be linked to a cytoplasmic or nuclear p120 function, the data strongly implies a mechanism involving the cadherin-bound fraction of p120. Moreover, in exploratory studies, we deleted each of the Arm repeats in p120 individually and conducted p120 KD and add-back experiments to identify the domains in p120 required to suppress apical expansion. Interestingly, suppression was selectively mediated by repeats 1–6, exactly the same repeats that mediate E-cadherin binding to p120 (Ireton, 2002) (Fig. S1A,B).

To further discriminate roles played by cadherin-bound and cytoplasmic p120 fractions, we used a single amino acid p120 mutant described recently (p120K401M) (Ishiyama et al., 2010) to selectively uncouple its interaction with E-cadherin (Figs 2C, 3A,B). p120-1A and p120-3A isoforms containing the K401M mutation localized exclusively to the cytoplasm (Fig. 2A) and failed to co-immunoprecipitate with, or stabilize, either E- or N-cadherin (Fig. 2B; Fig. S2A,B). Importantly, these mutants retained the interaction with Kaiso (also known as ZBTB33 in mammals), whose p120 interaction domain is known to overlap with that of E-cadherin (Fig. 2B). p120 KD and add-back experiments with these p120K401M mutants failed to rescue the apical invagination phenotype (Fig. 2D), indicating that cytoplasmic p120 is inactive.

Fig. 2.

The E-cadherin-bound fraction of p120 is essential for suppression of apical expansion. (A) The E-cadherin-uncoupling point mutation K401M was introduced into p120 isoforms 1A and 3A. In contrast to WT p120, expression of the mutant p120 constructs in p120-KD cells reveals exclusively cytoplasmic staining. (B) Biochemical validation of selective uncoupling from cadherin binding of p120K401M mutants. p120 immunoprecipitation (IP) confirms that K401M p120 mutants no longer bind E-cadherin or N-cadherin, but retain the interaction with Kaiso. (C) Schematic illustrating E-cadherin mutations that block E-cadherin endocytosis (LLAA) and uncouple p120 binding (E762A). The K401M mutation in p120 is also shown (lower panel). (D) K401M mutants fail to rescue apical expansion of p120-KD cells. z-stacks are shown to illustrate the behavior of the apical membrane. White arrowheads mark the tight junction, as evidenced by cingulin staining. Scale bars: 10 µm.

Fig. 3.

The p120–E-cadherin interaction is required for suppression of the apical expansion phenotype. (A) Schematic of p120 constructs used. Note the CAXX motif is fused to K401M mutant p120, not WT p120. (B) Schematic illustrating the cellular localization of E-cadherin and p120 mutants used in following experiments. (C) Both WT and p120-KD cells were stably transfected with CAAX-box tagged K401M (KM) isoform 1A and 3A constructs. As expected, p120-KM-CAAX mutants did not stabilize E-cadherin at cell junctions. Notably, localization of the CAAX box constructs is essentially identical to that of endogenous p120. (D) Lateral membrane association of p120 is not sufficient to suppress apical expansion in p120-KD cells. Notably, despite precisely colocalization with endogenous p120, the E-cadherin-uncoupled CAXX-box p120 constructs do not rescue the apical expansion defect and show no sign of dominant active activity in WT cells. (E) p120-KD cells were stably transfected with an endocytosis-deficient human E-cadherin mutant (LAEA) containing the E762A p120-uncoupling mutation. In contrast to endogenous E-cadherin in the same cells, this LAEA mutant is retained at the cell surface in the absence of p120. Note that two distinct E-cadherin monoclonal antibodies are used here; the rr-1 monoclonal antibody specifically recognizes endogenous canine (e.g. MDCK) E-cadherin, whereas the BD (610181) pan-E-cadherin monoclonal antibody binds both canine and human E-cadherin. For reasons that are unknown, rr-1 recognizes only E-cadherin on the cell surface; the endocytosed pool is not recognized (e.g. the bright Golgi staining of endogenous E-cadherin staining in p120-KD cells is seen only with the BD E-cadherin monoclonal antibody). (F) Forced surface retention of E-cadherin in the absence of p120 did not rescue the apical expansion defect (bottom panel). (G) Quantification of the apical expansion phenotype by measuring the LAI. Note that p120 KD+LAEA has a similar LAI to WT+DA-RhoA, and both are significantly larger than WT. *P<0.01; ***P<0.0001; ns, not significant (unpaired two-tailed t-test). Scale bars: 10 µm.

To determine whether direct interaction between p120 and E-cadherin is essential for the steady state suppression of apical invagination, we first asked whether simply targeting p120 to lateral membranes (irrespective of E-cadherin binding) was sufficient to support normal p120 activity. The p120K401M mutants were fused at the C-terminus to a CAAX-box motif (Fig. 3A,B), known to relocate cytoplasmic proteins to the plasma membrane (Seabra, 1998). Interestingly, the localization of these p120K401M-CAAX mutants was almost indistinguishable from that of endogenous p120, including exclusion from the apical surface (Fig. 3C,D). However, although p120K401M-CAAX was abundantly expressed on lateral membranes, the unstable internalized pool of E-cadherin in the p120-KD cells was clearly not rescued by these mutants (Fig. 3C). When plated on collagen, p120-KD cells expressing the p120K401M-CAAX mutants showed no sign of rescue of the apical invagination phenotype (Fig. 3D). Thus, although the localization of p120 to lateral membranes is essential for suppression of apical invagination, it is clearly not sufficient.

Next, we asked whether E-cadherin was sufficient to suppress apical invagination, irrespective of p120 binding. Because E-cadherin is rapidly degraded if p120 is not bound, we first generated an endocytosis-resistant E-cadherin mutant (from human E-cadherin) by changing the classic di-leucine endocytosis motif to alanine residues (Miyashita and Ozawa, 2007; Nanes et al., 2012). To further rule out any contribution from bound p120, we also introduced the p120-uncoupling point mutation (i.e. E-cad E762A) (Ishiyama et al., 2010), resulting in the construct termed Ecad-LAEA (see schematic, Figs 2C, 3B). When expressed in p120-KD MDCK cells, Ecad-LAEA localized normally to lateral membranes (Fig. 3E, third column), whereas endogenous E-cadherin remained entirely cytoplasmic (Fig. 3E, first column). Because the E-cadherin monoclonal antibody RR1 used in the first column is canine specific, it detects only the endogenous MDCK E-cadherin. Notably, the E-cadherin monoclonal antibody used in column-3 (BD) recognizes both canine and human E-cadherin, permitting visualization of Ecad-LAEA. p120 staining in the Ecad-LAEA cell line, by contrast, was comparable to that in the p120-KD cells and was almost undetectable (Fig. 3E). Ecad-LEAE is thus stably retained on lateral membranes despite almost undetectable p120 binding. Importantly, Ecad-LEAE rescued the junctional localization of α-catenin, indicating that p120 interaction is not required for the recruitment/coupling of this important co-mediator of E-cadherin function (Fig. S3A,B), and cell–cell adhesion in the cadherin-deficient A431D cells (Fig. S3C,D). To determine whether Ecad-LAEA could, in fact, suppress apical expansion without assistance from p120, the cells were again plated on collagen and examined, by performing ezrin staining, for apical membrane invagination. Unexpectedly, although Ecad-LAEA retention was in fact unaffected by p120 loss, it was nonetheless unable to reverse the apical invagination phenotype (Fig. 3F). We further quantified the phenotype by measuring the length of apical invagination (LAI) as a proxy for the extent of apical expansion. As shown in Fig. 3G, p120 KD induced a 3.2-fold increase of LAI (from 1.154 to 3.717 µm). Although Ecad-LAEA rescue did reduce the magnitude of the effect slightly (i.e. 2.7-fold versus 3.2-fold), these results show that stable surface retention of E-cadherin by itself is unable to appreciably rescue the p120 ablation phenotype. Moreover, as described later, directly activating RhoA in WT cells led to a comparable level of increase in LAI to that seen upon Ecad-LAEA rescue (Figs 3G, 4D). Thus, instead of directly modulating RhoA activity, restoring the E-cadherin–actin linkage by Ecad-LAEA could potentially slightly limit the scale of apical membrane deformation by activated junctional contractility.

Fig. 4.

Activation of the RhoA–ROCK–myosin pathway underpins both the apical expansion and the lumen formation defects caused by p120 KD. (A) Inhibition of either ROCK or myosin rescues WT apical membrane organization in p120-KD cells. WT and KD cells were cultured overnight on collagen gels and then treated with either DMSO, Y27632 (10 µM) or Blebbistatin (20 µM) for another 48 h. (B) Quantification of the LAI in WT and p120-KD cells upon treatment with DMSO, PKCζ-PS inhibitor, Y27632 or Blebbistatin. Both Y27632 and Blebbistatin brought LAI down to basal levels, whereas DMSO or PKCζ-PS inhibitor had no effect. ***P<0.0001; ns, not significant (unpaired two-tailed t-test). (C) Inhibition of ROCK does not cause E-cadherin relocalization to cell junctions in p120-KD cells. (D) DA-RhoA expression by itself in WT cells effectively recapitulates the apical expansion defect. Note that overexpression of DN-RhoA has no effect on apical membrane organization. (E) Expression of DA-RhoA induces apical expansion without affecting p120 localization. (F) E-cadherin localization is not affected by the overexpression of DN-RhoA or DA-RhoA. (G) Immunoblotting (WB) confirms expression of DN-RhoA or DA-RhoA, and shows that the stability of cadherin complexes are not affected. (H) Inhibition of ROCK blocks the apical expansion induced by DA-RhoA. Scale bars: 10 µm.

Apical invagination following loss of p120 is a function of excess contractility along lateral membranes

Another established consequence of p120 ablation is activation of the RhoA–ROCK–myosin pathway (Dohn et al., 2009; Perez-Moreno et al., 2006; Wildenberg et al., 2006), suggesting that the apical invagination phenotype might result from an imbalance in actomyosin activity. To test whether the presence of collagen alters the ability of p120 to inhibit this pathway, we blotted for downstream effectors of ROCK proteins in WT, p120-KD and p120 rescue cells cultured on collagen. As shown in Fig. S4G, p120 KD induced a notable increase of phosphorylation of cofilin-1, and p120 rescue (Rescue, p120 KD+p120 1A) strongly suppressed this activity. Although the antibodies against phosphorylated myosin light chain (MLC, also known as MYL2) we tried did not work well on western blots, we did notice a dramatic increase in the protein level of MLC, and an upward band shift corresponding to the active phosphorylated form of MLC (Fig. S4G, MLC blot long exposure). These data all point to increased ROCK and myosin activity in p120-KD cells compared to WT or p120 rescue cells. To further test this hypothesis, we blocked the RhoA–ROCK–myosin pathway by inhibition of either ROCK proteins (with Y27632) or myosin II (with Blebbistatin). Indeed, even in the near complete absence of adherens junctions (Fig. 4C), both inhibitors completely reversed the apical defect associated with p120 KD (Fig. 4A). Using the same quantification method described earlier, both Y27632 and Blebbistatin were found to reduce the LAI of p120-KD cells so that it was the same level as WT (Fig. 4B). By contrast, treatment with either DMSO or an aPKC inhibitor (PKCζ pseudo substrate) had no effect on the LAI (Fig. 4B). The defect, therefore, is not due to cadherin loss per se but is instead caused by excessive activation of Rho. It is worth noting that two Rho-uncoupled p120 mutants have been described previously, one involving deletion of an N-terminal region (ΔNTR), the other an intermediate sequence (ΔIns) of six amino acids located between ARM5 and ARM6 (Yanagisawa et al., 2008). However, this RhoGDI-like activity is reported when p120 is overexpressed and is apparently restricted to unbound p120, as it is not detected in E-cadherin-associated fractions (Anastasiadis and Reynolds, 2001; Reynolds et al., 2000; Yanagisawa et al., 2008). Furthermore, these mutants (p120-ΔNTR and p120-ΔIns) completely rescued the apical defect in p120-KD cells (Fig. S4A,B), indicating that the suppression of Rho by cadherin-bound p120 is in fact independent of the RhoGDI-like mechanism.

To better understand the impact of constitutively elevated Rho activity upon p120 loss, we next examined consequences of directly activating (DA-RhoA) or suppressing (DN-RhoA) RhoA, respectively, through expression of previously characterized Myc-tagged dominant active (Myc–RhoAG14V) and negative (Myc–RhoAT19N) RhoA constructs (Hall, 1998). When introduced separately into WT MDCK cells, DA-RhoA effectively recapitulated the apical invagination defect (Fig. 4D). Of particular interest is that the junctional presence of p120 was strongly maintained (Fig. 4E), as was E-cadherin (Fig. 4F), and yet the apical invagination defect was readily apparent. In contrast, cell morphology was unaffected by DN-RhoA (Fig. 4D). Other variables, including the levels of members of E-cadherin complexes, were unaffected by either construct (Fig. 4G). Furthermore, blocking ROCK activity effectively resolved the apical invagination defect induced by expression of DA-RhoA (Fig. 4H). Thus, the apical defect is strongly associated with excessive contractility and fully dependent on ROCK.

To characterize the localization of active RhoA (the RhoA-GTP form) upon p120 KD, a recently developed RhoA sensor, GFP–AHPH, was transiently transfected into both WT and p120 KD MDCK cells cultured on collagen gels. Cells expressing a low level of GFP were imaged because high level expression causes the sensor to mislocalize diffusely to the cytoplasm. Interestingly, in collagen-plated WT MDCK cells, RhoA-GTP was exclusively detected at the apical or (less frequently) basal domains. It was virtually never observed with p120 along the lateral membrane (Fig. 5A, left panel). However, in p120-KD cells, RhoA-GTP relocalized to cell–cell contacts (Fig. 5A, right panel) and notably, was no longer observed at the apical membrane. Confocal analysis placed the signal just under the apical invaginations (Fig. 5B, right panel). This data further supports the notion that p120 locally suppresses RhoA activity. We then examined the effect of p120 KD on myosin II, a downstream effector of RhoA. Indeed, the major myosin isoform (NMMIIA, also known as MYH9) in MDCK cells was normally excluded from lateral cell junctions (Fig. 5A, arrow) (Yamada and Nelson, 2007; Yu et al., 2008). However, p120 KD induced a striking accumulation of NMMIIA at lateral membranes (Fig. 5C, arrow, and 5D). Quantification of this phenomenon on a junction-by-junction basis was facilitated by the fact that NMMIIA staining was largely present or absent. NMMIIA was, for example, robustly present on the 22.2±4.3% of the p120 KD cell–cell junctions (i.e. 102/486), but essentially never observed at WT junctions (i.e. 3/803; Fig. 5E). Notably, the myosin bundles did not recapitulate the circumferential-belt-like localization observed in some epithelial systems (Ebrahim et al., 2013; Smutny et al., 2010). Generation of cell contractility is mediated by conformational changes in the head domain of NMMIIA (Hall et al., 1982). Importantly, the recruitment of NMMIIA was almost invariably accompanied by apical membrane expansion (i.e. 98%, see Fig. 5F), as evidenced by immunostaining for ezrin (Fig. 5G). Strikingly, when present, NMMIIA was found at the very tip of the ezrin-demarcated apical invagination (Fig. 5G, lower-right panel), which was shown to terminate abruptly at the tight junction (Fig. 1F). The data suggest an explanation for why the apical surface is so dramatically affected by loss of a lateral membrane protein. Apparently, the contractile force generated in the absence of p120 is a function of locally activated RhoA and the subsequent accumulation of myosin at the tight junction, which then transduces the force directly to the apical membrane, accounting for its invagination. We propose that this vertical suppression of contractility is a core function of p120, and ultimately essential for the characteristic rectangular morphology of individual epithelial cells and their collective ability to assemble a perfectly flat apical surface (Fig. 5H).

Fig. 5.

Aberrant RhoA-GTP and NMMIIA accumulation invariably marks the basal end of the apical invagination. (A,B) GFP–AHPH was transiently transfected into either WT or p120-KD cells on collagen. Cells were then fixed and stained for p120 (A) or ezrin (B). Only cells expressing low levels of GFP–AHPH were imaged, and 100 observations were made for both WT and p120-KD cells. Note that RhoA-GTP normally localizes to the apical membrane and is excluded from the lateral membrane in WT cells. Upon p120 KD, RhoA-GTP became concentrated at the basal end of apical invagination (arrows). (C,D) NMMIIA is excluded from cell junctions in WT MDCK cells but is recruited to cell junctions upon p120 KD (arrows). Representative colonies are shown. A fluorescence line-scan analysis of NMMIIA and p120 staining, for the line shown in C, is shown side by side. (E) Quantification (mean±s.e.m., n=603 for WT, n=486 for p120-KD) of junctional NMMIIA accumulation on a junction-by-junction basis reveals robust its robust presence at 22.2±4.3% of p120 KD cell–cell contacts. **P<0.01; ns, not significant (unpaired two-tailed t-test). (F) Correlational analysis of apical expansion and junctional recruitment of NMMIIA in p120-KD cells. The first two columns show the percentage of ezrin invaginations (Invag) with (+) and without (−) junctional staining of NMMIIA. The second set of columns show the percentage of the junctions containing junctional NMMIIA with (+) and without (−) ezrin invaginations. 98% of the NMMIIA-containing junctions are accompanied by ezrin-stained apical invaginations. Ezin Invag was quantified on a junction-by-junction basis. A junction is defined as the interface between two cells and each cell normally has four to five such interfaces. Ezrin invag is defined by the presence of strong strand-like staining at these interfaces. Quantified from 146 junctions that are Ezrin invag+ and 88 junctions that are NMMIIA+. (G) When present, NMMIIA bundles (arrow) invariably localizes to the tip of the invaginating apical membrane (arrowhead). Representative images and magnified insets of x-y and x-z confocal stacks (white dashed boxes) are shown side by side. (H) Working model illustrating proposed role of p120 in maintenance of the epithelial phenotype and underlying mechanisms. Scale bars: 10 µm.

Suppression of junctional contractility is essential for lumen formation

p120 KO often induces rather dramatic defects of epithelial morphogenesis in various organs. Among these, a common irregularity is an impairment in generating internal lumens. Although this has been postulated to simply be an adhesion defect, here, we aimed to use our established methodology to examine whether the activity of p120 towards RhoA plays a role in epithelial lumenogenesis. To mimic this morphogenetic process in vitro, we turned to two complementary assays that enable assessment of lumen formation under spontaneous or inducible conditions. The so-called ‘dome assay’ takes advantage of the fact that cells attaching to the plate in the context of collagen spontaneously form a two-layered colony separated by multiple lumens (Fig. 6A–C). Interestingly, in the absence of p120, the bilayer formed as in the WT scenario but no lumen was generated (Fig. 6A,B). Note that the lumens were clearly outlined by dense circular actin (Fig. 6A) and appeared as transparent bubble-like structures under bright field illumination (Fig. 6B, black arrowheads). In contrast, the structures observed in p120-deficient bilayers were gaps, not lumens. Notably, under bright-field microscopy, the transparent bubble-like structures, which are indicative of sealed lumens, are completely absent from the p120-KD bilayers (Fig. 6B, compare upper and lower panels).

Fig. 6.

p120 is required for lumen formation through suppression of contractility. (A,B) Confocal immunofluorescence (A) and bright field (B) imaging showing the presence of lumens (white arrows and black arrowheads, respectively) in WT but not p120-KD cultures. Images are of day 9 MDCK ‘dome’ structures (shown schematically in C). White dashed boxes designate insets shown at higher magnification to the right. Bright field imaging discriminates lumens (transparent bubble-like structures, black arrowheads) from intercellular spaces, the latter being visible by immunofluorescence (A), but not bright field (B). (C) Schematic illustration of MDCK dome formation (left) or inducible lumen formation (right) upon collagen overlay. Before collagen overlay, MDCK cells cultured on collagen form characteristic epithelial apical-basal polarity with apical membrane facing the medium. After collagen overlay, apical membranes are redistributed to cell junctions by transcytosis to initiate lumen formation, causing the monolayer to reorganize into a bilayer. (D) p120 KD blocks the formation of lumens induced by collagen overlay. The effect is rescued by p120 addback (addition of human p120; +hp120). Tubular and circular lumens are visualized by ezrin staining (arrows). Lumens are present in WT and p120 addback, but not p120-KD cells. (E) x-y confocal and z-stacks of WT, p120 KD and addback collagen overlay cultures. z-stacks confirm in 3D the presence of sealed lumens when p120 is present (arrows). In contrast, actin staining in the absence of p120 is diffuse, indicating lack of lumens. The schematic shows the distinction between lumens, which are sealed, and gaps, which are irregular spaces between cells. (F) Inhibition of ROCK or myosin rescues lumen formation in p120-KD cells. MDCK WT and p-120-KD cells were cultured on collagen were treated with either DMSO, Y27632 (10 µM) or Blebbistatin (20 µM) for 24 h, overlaid with collagen, and cultured for another 3 days (with daily replacement of drug). A low-magnification image, taken at 20×, is shown on the left and two magnified insets are shown on the right. Note that lumens are recognized as the circular intensified actin staining (arrow) surrounded by the nucleus. Scale bars: 10 µm (A); 20 µm (B,D–F).

The second assay, the so-called ‘collagen overlay assay’, was designed to be inducible, and more importantly, to form lumens independent of cell proliferation (Hall et al., 1982). Briefly, MDCK cells were seeded at very high density on a layer of collagen and allowed to arrest growth as a confluent monolayer (Fig. 6C). The cells were then overlaid with a second layer of collagen, invoking an intrinsic epithelial differentiation program that drives de novo lumen formation and regenerates the free apical surface (Fig. 6C). Normally, this process involves relocation of apical proteins to intercellular junctions where the nascent lumen is formed (Hall et al., 1982), as illustrated by ezrin-stained circles (Fig. 6D,E). However, in the absence of p120, ezrin localized randomly across the entire cell membrane and lumen formation did not occur. Importantly, lumen formation was rescued by restoring p120 expression (Fig. 6D,E). Thus, p120 is essential for lumen formation, apparently independently of its role in cell proliferation.

To distinguish potential Rho-mediated effects from those caused by cadherin destabilization, we used the same p120-CAAX and Ecad-LAEA mutants described above and tested their ability to rescue lumen formation in p120-KD cells. Indeed, neither p120 (Fig. S4E) nor E-cadherin (Fig. S4F) alone at the cell–cell junction was able to rescue this lumen formation defect (Fig. S4, arrows showing where lumens form). Additionally, p120 mutants lacking the NTR or Ins regions faithfully rescued lumen formation (Fig. S4C,D, arrows), consistent with a RhoGDI-independent mechanism. By contrast, ROCK and myosin inhibitors efficiently reversed the effects of p120 KD, this time in the context of lumen formation. Using the same collagen overlay assay, we found that, upon addition of either Y27632 or Blebbistatin, p120-KD cells were able to target apical membrane to specified foci, restoring the ability to generate lumina (Fig. 6F, arrows; the pictures were intentionally taken at low magnification to show restoration of the pattern of lumen formation in p120-KD cells in the presence of drug treatment). Although the effects of Blebbistatin and Y27632 were not identical, both clearly rescued lumen formation, as exemplified by the distinct apical membrane foci surrounded by the nucleus (Fig. 6F, inset, arrows). Taken together, these observations indicate a pivotal role for cadherin-bound p120 in controlling junctional contractility during epithelial morphogenesis.

DISCUSSION

Generation of an epithelial monolayer from individual cells is a coordinated process involving cell–cell adhesion and acquisition of shape (Fig. 7B) (e.g. squamous, cuboidal or columnar). It is initiated by lateral cell–cell contacts, which then expand radially to form lateral membranes. The phenomenon is well characterized in 2D MDCK cell cultures and involves a zippering process along the x- and y-axis (Yamada and Nelson, 2007). Notably, our model is similar in concept, except that expansion also occurs in the z-axis to accommodate the vertical dimension induced by plating the cells on collagen. Normally, this process ends in the formation of a perfectly flat apical membrane (Fig. 7A, third and fourth panels). In the absence of p120, the process apparently fails, resulting in a phenotype essentially identical to that exhibited by the WT cells at the earliest stage of epithelial maturation illustrated experimentally in Fig. 7A (1 h time point).

Fig. 7.

The activities of p120 towards E-cadherin and Rho are molecularly and functionally coupled to enable the maintenance of cell shape in the larger context of an epithelial monolayer. (A) Apical invagination is an inherent early feature and gradually retreats during epithelial monolayer formation. WT MDCK cells were trpsinized and plated at a confluent density on 2D collagen gels. Cells were fixed at different time points and stained for ezrin and p120. Confocal images were 3D reconstructed to generate the tilted view. The z-axis stacks of each channel and the merged view are shown side by side for each time point. Scale bar: 10 µm. (B) Conceptually, epithelial maturation involves cell–cell adhesion (step I) and acquisition of shape (step II). The cadherin-stabilizing activity of p120 is well established and essential for adhesion (step I). By contrast, by binding to E-cadherin, p120 locally suppress the RhoA–ROCK–myosin pathway to establish a low-tension zone along the lateral membrane, which enables epithelial maturation into a geometrically organized monolayer (step II). Finally, this local suppression of contractility by p120 is further required for various crucial epithelial functions including cyst formation and lumenogenesis.

Whereas the cadherin-stabilizing activity of p120 is well-established and essential for adhesion (Fig. 7B, step I), the role of the Rho-suppressing activity of p120 has been elusive. Here, we separate the cadherin-stabilizing and RhoA-suppressing functions of p120 by analyzing a stabilized E-cadherin mutant that is retained on the cell surface irrespective of p120 binding. Surprisingly, the RhoA-suppressing activity of p120 is crucial for suppressing contractility along the vertical axis of lateral epithelial membranes. Moreover, this function is essential for maintenance of individual cell shape in the overall context of collective epithelial architecture (Fig. 7B, step II). Although establishment of epithelial cell shape is generally attributed to tension-generating mechanisms, such as apical constriction, here, we demonstrate that, along the lateral cell membranes, it is in fact suppression of contractility that is crucial.

Importantly, the impact on cell shape (following p120 KD) was not rescued by forced E-cadherin stability (through the LAEA mutation, see Fig. 3F), suggesting that the invagination phenotype is not primarily a function of cadherin stability. Although the stabilizing mechanism is well established, it is possible that cadherin signaling and/or other activities are nonetheless compromised. However, the fact that the LAEA E-cadherin mutant interacts normally with β-catenin, and restores α- and β-catenin to endogenous levels, indicates that the cytoskeletal linkage is largely intact. Moreover, in cadherin-deficient cell lines, such as A431D, the behavior of the mutant is almost indistinguishable from that of WT E-cadherin. Furthermore, inhibition of the Rho–ROCK–myosin pathway completely rescued the apical membrane defect, despite the fact that the entire E-cadherin complex was absent (Fig. 4B). Additionally, ectopic expression of constitutively active Rho had no effect on E-cadherin stability and yet effectively recapitulated the apical defect associated with p120 ablation (Fig. 4D,E). Remarkably, in WT cells, activated RhoA was detected almost exclusively at the apical membrane, whereas the signal shifted entirely to the lateral membrane (along with recruitment of myosin) upon depletion of p120. Thus, cadherin-bound p120 apparently maintains a low tension zone along the lateral membrane through suppression of RhoA. Notably, the mutually exclusive presence of E-cadherin and myosin II at cell–cell contacts is a relatively common phenomenon. For example, convergent extension in Drosophila is dependent on the segregation of E-cadherin and myosin II to the dorsal-ventral and anterior-posterior edges, respectively (Simões et al., 2010). Similarly, E-cadherin drives compaction in the early mouse embryo by redirecting myosin away from cell–cell contacts (Maître et al., 2015).

Interestingly, although constitutive membrane targeting is frequently sufficient to activate receptor-associated cofactors, the CAAX-box-mediated targeting of cadherin-uncoupled p120 to basolateral membranes did not rescue p120 ablation. The experiment was remarkable in that the localization of the E-cadherin-uncoupled CAAX-p120 protein and that of WT E-cadherin-bound p120 was essentially indistinguishable, and yet only the E-cadherin-bound p120 was active. The crystal structure of the p120/E-cadherin complex shows the JMD core of E-cadherin embedded in a groove along one side of the Arm repeats, leaving most of the p120 surface still exposed and available for interaction with other proteins (Ishiyama et al., 2010). Thus, one possibility is that suppression of Rho by p120 is enabled by (and perhaps dependent on) interaction with E-cadherin. Plausible mechanisms include an E-cadherin-triggered conformational change in p120, or alternatively, de novo generation of a new ‘combinatorial’ binding site consisting of polypeptides from both p120 and E-cadherin. A third possibility is that E-cadherin simply holds p120 in an ‘active’ orientation.

Notably, MDCK cells do express other cadherins (e.g. N-cadherin) (Stewart et al., 2000), which along with E-cadherin share a common cellular pool of p120 (Carnahan et al., 2010). Apparently, for that reason, E-cadherin KD alone in these cells does not noticeably alter p120 localization (Fig. S2C,D), and removing just E-cadherin has no effect on either apical organization or lumen formation (Fig. S2E,F). Thus, E-cadherin is not the only classical cadherin that can engage p120 to suppress Rho activity. For example, although E- and N-cadherin have clearly evolved disparate roles, as exemplified by their alternative usage in epithelial-to-mesenchymal transition (EMT), they (and probably other classical cadherins) are apparently redundant with respect to regulation of junctional tension.

An emerging paradigm in Drosophila development is the specification of cell height by compartmentalization of Rho activity along lateral epithelial membranes (Gibson, 2005; Shen and Dahmann, 2005; Widmann and Dahmann, 2009). For example, Dpp morphogen gradients specify the timing and amount of RhoGAP transcription during wing development. RhoGAP then accumulates along lateral membranes to suppress Rho. Removal of Dpp interrupts the pathway, causing unscheduled elevation of Rho-mediated contractility and shortening of the lateral membranes. Although acute p120 ablation is not directly comparable to the elegant spatiotemporal sculpting orchestrated by developmental programs, the end result is consistent with the Drosophila paradigm and indicative of the essential role of p120 in suppressing lateral contractility. Interestingly, although the RhoGAP involved has yet to be identified in our MDCK model, a very recent study in C. elegans has identified a previously uncharacterized linker, PAC-1-interacting coiled-coil protein-1 (PICC-1), that bridges C. elegans p120 (JAC-1) to PAC-1, a RhoGAP with specificity toward Cdc42 and Rho (Anderson et al., 2008; Klompstra et al., 2015). Remarkably, PICC-1 turns out to be the worm homolog of vertebrate CCDC85B (also known as DIPA), a direct p120-binding partner identified recently in our laboratory (Markham et al., 2014). This newly identified role for p120 seems rather important and unlikely to be confined to C. elegans, as virtually all of the C. elegans players in the story can be matched to highly conserved vertebrate homologs. Moreover, several clues support the notion that the scenario just described is quite likely to be conserved in the mammalian embryo. Thus, it appears that recruitment of various RhoGTPase modulators (such as the RhoGAP PAC-1) to control local GTPase activities might in fact be the paradigm for which p120 was originally intended in ancient metazoa.

Finally, suppression of contractility by p120 apparently goes well beyond the control of cell height and shape. Indeed, lumen formation is also dependent on appropriate regulation of tension at this level. Exactly how contractility controls these events is not yet clear, in part because NMMIIA activity impacts upon almost all of the cellular processes that influence epithelial morphogenesis (Vicente-Manzanares et al., 2009). It is important, however, that virtually all of the phenotypes induced by p120 ablation are effectively reversed by specific inhibition of ROCK. Although the cadherin-stabilizing function of p120 is clearly essential, the extent of rescue by ROCK inhibition across multiple phenotypes reinforces the notion that p120 is also a key regulator of cellular tension. Interestingly, E-cadherin is increasingly recognized as a mechanosensor of intercellular forces (Engl et al., 2014; le Duc et al., 2010; Smutny et al., 2010; Taguchi et al., 2011). The fact that p120 stabilizes E-cadherin, on one hand, and regulates contractility, on the other, places p120 at the intersection between sensing and transducing mechanical forces at sites of cell–cell adhesion.

MATERIALS AND METHODS

Antibodies and reagents

The primary antibodies against p120 (monoclonal antibody pp120, 0.5 μg/ml, BD; polyclonal antibody F1aSH, 1:500; and monoclonal antibody 15D2, 2 μg/ml) were generated as described previously (Reynolds et al., 1994); 15D2 was used for immunoprecipitation, pp120 was used for western blotting, and pp120 and F1aSh were used for immunofluorescence. Other antibodies were against: E-cadherin (1:1000, BD), E-cadherin (rr-1, 1:500), N-cadherin (1:500, BD), β-catenin (1:1000, Sigma-Aldrich), α-catenin (1:500, Sigma, C2081), ezrin (1:1000, BD), xingulin (1:500, a gift from Sandra Citi, Department of Cell Biology, University of Geneva, Switzerland), tubulin (1:1000, VAPR), cleaved caspase-3 (1:500, Cell Signaling), Kaiso (1:500, VAPR), Flag (1:1000, Sigma-Aldrich), Myc (9B11,1:500, Cell Signaling), and NMMIIA (1:500, Covance). The nucleus was stained with Hoechst 33258 (Sigma-Aldrich) (1:1000), Actin stained with phalloidin conjugated to Alexa Fluor 594 or 488 (1:200, Invitrogen). Secondary antibodies for western blotting were anti-mouse-IgG conjugated to Alexa Fluor 680 (Invitrogen) and anti-rabbit-IgG conjugated to IRdye 800 (Rockland Immunochemicals, Boyertown, PA). Secondary antibodies used for immunofluorescence analysis included anti-mouse-IgG, anti-mouse-IgG2a, anti-mouse-IgG1, anti-mouse-IgG2b, and anti-rabbit-IgG conjugated to Alexa Fluor 488 or 594 (Invitrogen) Reagents used include ROCK inhibitor Y27632 (EMD, Millipore), Blebbistatin (EMD, Millipore), myristoylated PKCζ pseudosubstrate (ENZO) and DMSO (Fisher).

Plasmids

pRetroSuper retroviral vectors expressing short hairpin RNA (shRNA) directed against canine p120 was generated as previously described. LZRS-Neo-3XFlag-Gateway vector was used for exogenous expression of p120 (full-length and mutants). LZRS-Neo-MS was used for exogenous expression of E-cadherin (full-length and mutants), DN-RhoA–Myc and DA-RhoA–Myc. Point mutations were generated using site-directed ligase-independent mutagenesis (SLIM). GFP–AHPH was a kind gift from Alpha Yap (Division of Cell Biology and Molecular Medicine, Institute for Molecular Bioscience, The University of Queensland, Brisbane, Australia).

Virus production and transduction

Retrovirus was generated by transfecting the Phoenix 293 cells using the calcium phosphate method. Retrovirus constructs used were all based on LZRS-neo as described previously (Ireton, 2002). Virus was harvested at 48 h post-transfection by passing the supernatant through a 0.45-μm filter. Target cells were infected by incubation with retrovirus-containing medium containing 4 μg/ml Polybrene overnight, which was then replaced with normal culture medium. At 48 h post-infection, cells were selected using G418 (1 mg/ml) for 7 days. Lentiviral particles were generated by transfecting 293T cells with the petrosuper shRNA plasmid of interest, psPAX2 packaging plasmid and pMD2.G envelope plasmid, using the calcium phosphate method. Lentivirus was harvested at 48 h post-transfection, and target cells were infected as described above. At ∼48 h post-infection, infected cells were selected using puromycin (0.5 mg/ml) for 2 days.

Cell culture

MDCK II cells were cultured in Dulbecco's modified Eagle's medium (DMEM; Life Technologies) supplemented with 10% fetal bovine serum (FBS; Hyclone, Thermo Scientific) and 1% penicillin–streptomycin (Life Technologies/Invitrogen). Phoenix 293 and 293T cells were cultured in DMEM supplemented with 10% heat-inactivated FBS and 1% penicillin-streptomycin. Collagen solution (1 ml) was made on ice by adding the following solutions sequentially: 75 μl double-distilled H₂O, 100 μl 10× DMEM (D2429, Sigma), 100 μl 200 mM HEPES, 50 μl 74 mg/ml NaHCO₃, 670 μl collagen I (354236, BD) and 1 drop of 40 mg/ml NaOH). For 2D collagen culture, 85 μl collagen solution was added into the chamber and allowed solidify in the incubator for 20 min, then a 10-μl resuspension of cells was mixed with 200 μl DMEM and added on top of the collagen gel. This was then cultured for 3 days and then fixed and stained.

Collagen overlay assay

MDCK II cells were resuspended at 1.5×10⁶ cells/ml as described above. Then, 55 μl collagen solution was added into each chamber and allowed to solidify. A 50-μl cell resuspension mixed with 200 μl DMEM was then added on top of the collagen gel. After 24 h, the medium was carefully removed from the chamber, and 85 μl collagen solution was added and allowed to solidify for 20 min. Then, 200 μl DMEM was added on top of the collagen gel, and the cells cultured for another 48 h before fixation and staining. Transparent lumen and tubules should be visible for WT MDCK II cells under bright-field microscope.

Immunofluorescence, immunoblotting and immunoprecipitation

Lysate preparation, western blot and immunoprecipitation procedures, immunostaining on 2D coverslips are as described previously (Dohn et al., 2009). For immunostaining of cells on collagen, the entire collagen gel was transferred from the chamber slide to a 24-well plate and fixed using 4% paraformaldehyde in PBS+ (500 ml PBS, 1 mM CaCl₂ and 0.5 mM MgCl₂) for 30 min. This was then washed with PBS+ and permeabilized with 0.025% saponin in PBS+ for 1 h, before being washed again with PBS+ and incubated with quench solution (75 mM NH₄Cl and 20 mM glycine in PBS+) for 1 h. After another wash, the gel was incubated with blocking buffer (1% BSA, 1% goat serum, 0.025% Saponin in PBS+) for 1 h. Diluted primary antibody with blocking buffer was added and the gel was incubated at 4°C overnight, followed by washing for 4 h and incubation with secondary antibody at 4°C overnight. After a 4-h wash, the nucleus was stained with Hoechst 33258 (Sigma-Aldrich) for 30 min. Collagen gels were then transferred onto the slides, mounted with prolong gold and stored at 4°C before viewing. Images were collected using Zeiss LSM 510 confocal microscope at 40× or 63× magnification. For z-stacks, 0.45-μm sections were taken. All images and movies were then processed with ImageJ with the 3D view plugin or Volocity 6.3 demo.

Statistics

Statistical analyses were preformed using Prism (GraphPad) with two-tailed Student's t-tests or Mann–Whitney tests.