αE-catenin regulates cell-cell adhesion and membrane blebbing during zebrafish epiboly

The adherens junction (AJ) comprises protein complexes involved in cell-cell adhesion and collective cell migration in embryonic development (Halbleib and Nelson, 2006). Cell-cell adhesion at the AJ is mediated by classical cadherins, such as E-cadherin, via trans interactions between their extracellular domains on opposing cells (Stemmler, 2008). The cytoplasmic domain of classical cadherins binds β-catenin and p120-catenin (Shapiro and Weis, 2009), and β-catenin in turn binds the actin-binding protein α-catenin that connects the AJ with the actin cytoskeleton (Benjamin and Nelson, 2008). α-Catenin might have additional roles in cell polarity, signal transduction and as a tension transducer (Gladden et al., 2010; Yonemura et al., 2010; Schlegelmilch et al., 2011).

Early models proposed that vertebrate α-catenin provides a static bridge between the AJ and the actin cytoskeleton (Rimm et al., 1995). However, recent studies have caused a re-evaluation of α-catenin regulation and function. In vitro biochemical studies revealed that αE-catenin exists as either a monomer that binds strongly to β-catenin but weakly to F-actin, or a homodimer that binds strongly to F-actin but not to β-catenin, and the homodimer exist as a cytoplasmic pool (Drees et al., 2005; Yamada et al., 2005; Benjamin et al., 2010). These studies raised the possibility that αE-catenin regulates F-actin independently of the E-cadherin cell-cell adhesion complex. In fact, selective depletion of the αE-catenin cytoplasmic pool increased lamellipodial activity and the rate of cell migration without affecting cell-cell adhesion (Benjamin et al., 2010; Borghi et al., 2010). Thus, αE-catenin appears to play roles in maintaining cell-cell adhesion and suppressing membrane dynamics associated with cell migration.

Gastrulation is a crucial stage of animal development that depends on morphogenetic processes (including epiboly) that involve cell migration, cell sorting and tissue remodeling while maintaining cell-cell adhesion (Solnica-Krezel, 2005). Epiboly is defined as the spreading and thinning of one tissue over another (Lepage and Bruce, 2010). In Danio rerio (zebrafish) gastrulation, two tissues undergo epiboly to spread over and enclose the yolk cell: the external-most cell layer, which is termed the enveloping layer (EVL), and the deep cells, which become the embryo proper (Lepage and Bruce, 2010). EVL epiboly may be driven by the changing morphology of the yolk syncytial layer (YSL) (Koppen et al., 2006), a cytosolic, yolk-free region of the yolk cell found immediately under the deep cells that also spreads during epiboly (Carvalho and Heisenberg, 2010). The leading edge of the EVL and the YSL are connected by tight junctions, and depletion of the tight junction protein claudin E impairs EVL epiboly (Siddiqui et al., 2010). The YSL might provide a pulling force on the EVL margin through an actin ring just below the EVL margin (Koppen et al., 2006). A major contributor to deep cell epiboly is radial intercalation, in which deeper cells intercalate between and displace more superficial cells causing the presumptive ectoderm to become thinner and spread over the yolk (Kane et al., 2005; Lepage and Bruce, 2010).

E-cadherin (Cdh1 – Zebrafish Information Network) plays a crucial role during zebrafish gastrulation. Loss of E-cadherin expression due to mutation or depletion by morpholino arrests deep cell epiboly at mid-gastrulation (Babb and Marrs, 2004; Kane et al., 2005; Shimizu et al., 2005). The delay in epiboly in E-cadherin mutants and morphants appears to be due to defective radial intercalation. Mutant or morphant deep cells fail to stably integrate into the superficial layer and drop back to the lower layer, thereby slowing or blocking epiboly (Kane et al., 2005; Slanchev et al., 2009). By contrast, E-cadherin is not required for EVL epiboly (Shimizu et al., 2005; Koppen et al., 2006; Lin et al., 2009).

Few studies have examined the role of αE-catenin in cellular dynamics during early vertebrate development. In Xenopus, depletion of αE-catenin causes a delay in gastrulation (Kofron et al., 1997), and genetic deletion of αE-catenin (Ctnna1 – Mouse Genome Informatics) in mice is lethal in the pre-implantation stage (Torres et al., 1997). Here, we analyzed the role of αE-catenin at the single cell level during coordinated cell movements in vertebrate epiboly, and show that αE-catenin plays roles in cell-cell adhesion, membrane dynamics and cell migration.

Zebrafish wild-type embryos were raised at 28.5°C and staged as described (Kimmel et al., 1995). Wild-type strains TL and WIK were used.

Primary antibodies used were: α-catenin (for western blotting, BD; for immunostaining, Sigma), pan-cadherin, β-catenin and tubulin (Sigma); pERM and pMLC (Cell Signaling); HA-tag (Covance); zebrafish E-cadherin (Babb and Marrs, 2004). Rhodamine Phalloidin was obtained from Invitrogen, blebbistatin from Sigma. Antisense morpholinos (Gene-Tools) used were: ctnna morpholino 1 (targeting 5′UTR), CAAAATGGAGGGATGAGACTTTTAC; ctnna mismatch morpholino 1, CAAATTCGACGGATGACAGTTTTAC; ctnna morpholino 2, TAATGCTCGTCATGTTCCAAATTGC; cdh1 morpholino, ATCCCACAGTTGTTACACAAGCCAT (Babb and Marrs, 2004); ezr morpholino, CGCGAACATTTACTGGTTTAGGCAT (Link et al., 2006). Capped mRNA was synthesized using mMessage-Machine (Ambion).

Zebrafish ctnna c-DNA was obtained from Open Biosystems (Image ID 2639279) and subcloned into pCS2⁺. The HA tag was introduced through the 3′ primer sequence. Mouse gfp-mctnna (Benjamin et al., 2010) was subcloned in pCS2⁺.

Capped mRNA and antisense morpholino were injected at the 1- to 2-cell stage at the following doses: membrane localized GFP (Wallingford et al., 2000), 100 pg; ctnnaha, 200 pg; gfp-mctnna, 200 pg; ctnna morpholino1, 4 ng (high dose) and 0.8 ng (low dose); ctnna mismatch morpholino1, 5 ng; ctnna morpholino2, 10 ng; cdh1 morpholino, 8 ng; ezr morpholino, 8 ng (high dose) and 1 ng (low dose).

Embryos were fixed in 4% paraformaldehyde in PBS. Whole-mount immunohistochemistry was performed as described previously (Koppen et al., 2006).

Live embryos were mounted in 2% methylcellulose in 0.3× Danieau’s buffer at 28.5°C (Saude et al., 2000). Embryos at segmentation stages were imaged with an Axiovert microscope (Carl Zeiss). Differential interference contrast (DIC) and confocal images were taken using an LSM 510 (Carl Zeiss).

Image analysis was performed using ImageJ software (http://rsbweb.nih.gov/ij/). Time-lapse images were recorded for 20 minutes at y1 minute intervals, or for 5 minutes at 4 second intervals. The imaging focal plane was set at 6-8 μm from the basal membrane of EVL at the animal pole. The 20 minute time-lapse recordings used embryos at 50-60% epiboly. The 5 minute time-lapse recordings used embryos at 50% epiboly. For the blebbistatin experiments, embryos were incubated in 50 μM blebbistatin or DMSO for 30 minutes.

Radial intercalation and reverse radial intercalation rate represent the percentage of cell undergoing radial intercalation (RI) or reverse radial intercalation (revRI) of the total population of cells in the field of view at the end of the recording period (20 minutes).

Staging by ‘percent epiboly’ was calculated by measuring the distance from animal pole to the EVL or deep cell margins, and dividing by the animal-vegetal axis length. Three independent experiments were performed; n=25 embryos.

Length-to-width ratio of the deep cells at the margin was measured as previously described (Koppen et al., 2006). Three independent experiments were performed; n=40 cells (eight embryos).

The fluorescence intensity (FI) determination for phosphorylated myosin light chain (pMLC) staining was performed as previously described (Kitt and Nelson, 2011). Three independent experiments were performed; n=300 cell contacts (15 embryos). The control was normalized to one.

Wild-type embryos were injected with membrane-localized GFP mRNA with or without ctnna morpholino 1. Labeled cells were transplanted at the blastula stage into control or morphant hosts.

Embryos were collected at dome stage and fixed with 4% paraformaldehyde overnight at 4°C. Whole-mount ISH was performed as described (Thisse and Thisse, 1998). Probes used were: Keratin4 (Thisse et al., 2001), no tail (Schulte-Merker et al., 1992) and pax2.1 (Krauss et al., 1992).

Dome stage embryos were dispersed using trypsin/EDTA. The cells were plate at a concentration of 5000 cells/ml and incubated overnight at 28.5°C overnight with or without EGTA (5 mM).

Embryo homogenates were lysed with 4×lysis buffer (62.5 mM Tris-HCl pH 6.8, 10% glycerol, 2% SDS, 0.5% β-mercaptoethanol, 0.01% Bromophenol Blue) and heated at 100°C for 5 minutes. Extracts were resolved in a 6% polyacrylamide gel and transferred to a PVDF membrane using the Criterion System (Bio-Rad).

Statistical analyses were performed using an unpaired t-test with unequal variance using Excel (Microsoft) and Prism (Graph Pad).

αE-catenin (Ctnna – Zebrafish Information Network) is ubiquitously expressed during gastrulation and early zebrafish development (Dong et al., 2009; Slanchev et al., 2009). This was confirmed by immunofluorescence microscopy of whole-mount embryos and western blotting of embryo extracts with antibodies specific for αE-catenin (supplementary material Fig. S1A-D; see also Fig. 1D and Fig. 2D).

To investigate the role of αE-catenin in early zebrafish development, we blocked ctnna mRNA translation by injecting two different translation blocking morpholinos (mo): ctnna mo1 targeted the 5′UTR of ctnna mRNA, and ctnna mo2 targeted a sequence that overlaps with 5′UTR and the open reading frame (ORF). Embryos injected with 4 ng of ctnna mo1, but not uninjected control embryos, had defects in morphology at the end of gastrulation (95.2±4.8%; Fig. 1A-C,I). Embryos injected with 10 ng of ctnna mo2 also displayed the same morphological defects as ctnna mo1, but with weaker penetrance (79.9±4.4%; Fig. 1I). Some ctnna morphants did not close the blastopore over the yolk (Fig. 1C, arrow).

The specificity of the epibolic defect caused by ctnna mo1 was shown by three types of experiments. First, injection of a mismatched mo1 (mm-mo1; 5ng) did not cause epibolic defects (1.5±1.1%; Fig. 1I). Second, αE-catenin was depleted by ctnna mo1 as shown by western blotting of whole embryo extracts at 90% epiboly (Fig. 1D). Third, the defects caused by the ctnna mo1 were rescued by co-injecting the ctnna mo1 with a synthetic mRNA for αE-catenin tagged with HA (ctnnaha), which did not bind the mo1; only 35.9±7% of rescued mo1 embryos had defects in morphology at the end of gastrulation (Fig. 1I). Note that αE-catenin-HA was expressed and localized at the plasma membrane and cytoplasm of deep cells (Fig. 2G) and EVL (A.S., unpublished results). Injection of up to 200 pg of ctnnaha mRNA alone did not cause any apparent developmental defects (Fig. 1I).

To understand whether the delay in epiboly in the ctnna morphant was due to a morphological or developmental defect, we performed in situ hybridization (ISH) for no tail and pax2.1 (pax2a – Zebrafish Information Network) in embryos fixed at 80% epiboly (supplementary material Fig. S1E,F). The ISH for no tail revealed a pattern that was typical for a delay in epiboly (Babb and Marrs, 2004). By contrast, pax2.1 expression was absent from the ctnna morphant at a stage at which it was already present in the control; note that pax2.1 expression appeared after 90% epiboly in the ctnna morphant (not shown). The delay in pax2.1 expression indicates that depletion of αE-catenin caused a developmental delay.

At later stages of embryogenesis (mid-segmentation), 57.9±2.9% of embryos injected with ctnna mo1 had a strong developmental delay, and those that reached the mid-segmentation stage had an abnormal morphology (Fig. 1E-H,J). Furthermore, 38±2.9% of the embryos died by mid-segmentation. Rescue of ctnna mo1 reduced the percentage of delayed embryos to 19.4±3.4% and the percentage of dead embryos to 7±3.4% (Fig. 1J). All of the ctnna morphants displayed defects in notochord and somite morphology (Fig. 1G,H). Injection of the ctnna mm-mo1 did not cause any defects at this stage (2±1.5% dead embryos, Fig. 1J), whereas injection of ctnna mo2 caused complete lethality (A.S., unpublished results). These defects in later stages of embryogenesis were not pursued further. Instead, we focused on defects in epiboly to study αE-catenin functions in cell-cell adhesion, membrane dynamics and cell migration. For the rest of this study, we did not use the ctnna mo2 owing to the weak penetrance in the epibolic defect; from this point, we refer to ctnna mo1 and mm-mo1 as ctnna mo and mm-mo, respectively.

To investigate the delay in gastrulation caused by αE-catenin depletion, we measured the extent of epiboly at mid-gastrulation. Uninjected embryos or embryos injected with ctnna mo, ctnna mo + ctnnaha mRNA (rescue), or ctnnaha mRNA were fixed at 80% epiboly, as defined by the control uninjected cohort. Embryos were stained with a β-catenin antibody to mark the plasma membrane for analysis of embryos by confocal microscopy (Fig. 2A-C); percentage epiboly was defined as the distance between the animal pole and the EVL or deep cell margins (Fig. 2A-C, arrows). Compared with the control, ctnna morphants were delayed in the migration of both the EVL margin (80.8±2.02% epiboly; control, 88.9±0.8% epiboly) and deep cell margin (71.4±2.16% epiboly; control, 83.7±1.7% epiboly) (Fig. 2A,B,H). Rescue embryos were not delayed (Fig. 2C,H), nor were embryos injected with ctnnaha mRNA (Fig. 2H).

We analyzed the levels of αE-catenin at 50% epiboly by immunostaining whole-mount embryos in order to confirm that it had been depleted in the deep cells. Ctnna morphant deep cells had significantly reduced levels of αE-catenin at the plasma membrane and in the cytoplasm compared with control (Fig. 2D,E). Note that localization of αE-catenin at the plasma membrane and cell-cell contacts was dependent on expression of E-cadherin (Fig. 2F). αE-catenin-HA was expressed and localized at the plasma membrane (Fig. 2G).

Next, we examined the morphology of cells at the EVL margin at 80% epiboly using F-actin staining. Control EVL cells had an elongated shape, which is typical of this stage (Fig. 3A). Elongation of EVL cells at the margin coincided with movement of the EVL margin toward the vegetal pole (Koppen et al., 2006). By contrast, ctnna morphant cells did not have this elongated morphology (Fig. 3B). This difference in morphology was shown quantitatively by measuring the length-width ratio (LWR) of EVL margin cells in control (LWR 2.25±0.22) and ctnna morphants (LWR 1.36±0.08) (Fig. 3F). The morphology of the cells at the EVL margin was rescued by co-injection of ctnna mo and ctnnaha mRNA (LWR 2.05±0.3) (Fig. 3C,F). Injection of ctnnaha mRNA alone did not cause defects in cell morphology at the EVL margin (LWR 2.22±0.18; Fig. 3F).

As αE-catenin forms a complex with E-cadherin, we compared the morphology of ctnna morphants with cdh1 morphants (8 ng cdh1 mo). We found that morphological defects in cells in the EVL margin were specific to ctnna morphants (LWR 1.36±0.08), and not to cdh1 morphants (LWR 2.32±0.14) (Fig. 3D,F), as previously shown (Koppen et al., 2006; Lin et al., 2009). Moreover, co-injection of cdh1 mo and ctnna mo (8 ng and 4 ng, respectively) resulted in a similar morphology to the embryos injected with ctnna mo only (LWR 1.33±0.01) (Fig. 3E,F).

It has been reported that defects in EVL differentiation can induce defects in epiboly (Fukazawa et al., 2010). In order to rule out the possibility that the defect in EVL morphology and the delay in epiboly were caused by a defect in EVL differentiation, we performed ISH for keratin4 (krt4) mRNA at the dome stage (supplementary material Fig. S2). The ctnna, cdh1 and ctnna-cdh1 morphant embryos displayed normal krt4 expression compared with the control, indicating that the EVL had differentiated normally in these embryos.

These results indicate that the failure of EVL margin cells to acquire an elongated morphology in ctnna morphants correlates with the delay in the EVL epiboly (Fig. 2B). This phenotype was not caused by E-cadherin depletion and was not rescued by co-depletion of E-cadherin and αE-catenin. Thus, αE-catenin appears to be necessary for cells at the EVL margin to acquire an elongated morphology, but E-cadherin is dispensable.

To investigate the nature of the deep cell margin delay (Fig. 2B,H), we used time-lapse microscopy to examine the behavior of individual deep cells (DCs) at the animal pole at 50-60% epiboly in embryos injected with mRNA for membrane-bound GFP (mbGFP). Time-lapse microscopy of control DCs undergoing radial intercalation (RI) showed that DCs appeared from a deeper layer, intercalated between other cells and were retained in the outer epiblast layer (Fig. 4A; supplementary material Movie 1). DCs in ctnna morphants (4 ng mo) also appeared in the superficial epiblast layer, similar to control DCs (Fig. 4B). Strikingly, however, some of the DCs in ctnna morphants moved back into the deeper layers (Fig. 4B; supplementary material Movie 2) in a process termed reverse radial intercalation (revRI) that was never observed in control embryos (Fig. 4A; supplementary material Movie 1).

We also analyzed the behavior of DCs in cdh1 morphants; depletion of E-cadherin was confirmed by immunostaining using an E-cadherin-specific antibody or a pan-cadherin antibody (supplementary material Fig. S3A-D), indicating that E-cadherin is the predominant, if not the only, classical cadherin expressed in the most external layer of DCs at the animal pole at 50% epiboly. In cdh1 morphants, DCs intercalated into the superficial epiblast layer (Fig. 4C; supplementary material Movie 3). Similar to ctnna morphants, however, some of the cdh1 morphant DCs underwent revRI (Fig. 4C; supplementary material Movie 3). These observations confirm results of previous studies of cdh1 mutants and morphants (Kane et al., 2005; Slanchev et al., 2009).

We measured the rates of RI and revRI, defined as the percentage of cells that intercalated into or left the external epiblast layer, respectively, during 20-minute time-lapse movies for nine embryos (Fig. 4D). The RI rate was similar in the control (19.17±1.12), cdh1 morphant (23.19±2.3) and ctnna morphant (23.2±1) (Fig. 4D). We did not observe any revRI events in control embryos (rate 0), but the revRI rate was 10.7±1.5 in cdh1 morphants and 9.3±1 in ctnna morphants (Fig. 4D). The revRI defect in the ctnna morphant was partially rescued by injection of a synthetic mRNA encoding a GFP-tagged version of mouse αE-catenin (gfp-mctnna) (Fig. 4D; for expression and correct cell membrane localization of GFP-αE-catenin see supplementary material Movie 9). Injection of the mismatched ctnna mm-mo did not alter the rate of RI (19.24±2.22; Fig. 4D) and, similar to the control embryos, did not cause any revRI events (rate 0). Finally, the revRI caused by αE-catenin depletion was cell autonomous (7/42 cells for 5 min time lapses; supplementary material Movie 4). These results indicate, as shown previously for the cdh1 mutant/morphant (Kane et al., 2005; Slanchev et al., 2009), that the delay in DC margin epiboly was due to cells that were unable to remain in the most external epiblast layer and moved back into the deeper layers.

We explored the possibility that the revRI phenotype was caused by defects in cell-cell adhesion by performing an ex vivo cell-cell aggregation assay. Embryos were dissociated using trypsin and the cells were allowed to re-aggregate overnight. Cells from control embryos formed large cell aggregates (supplementary material Fig. S4A). Incubation of wild-type cells with 5 mM EGTA, which chelates Ca²⁺ and inhibits Ca²⁺-dependent cadherin-mediated cell-cell adhesion, completely inhibited cell aggregation (supplementary material Fig. S4B). Depletion of E-cadherin or αE-catenin also inhibited cell-cell adhesion (supplementary material Fig. S4C,D). This result supports the hypothesis that the revRI phenotype of DCs depleted of αE-catenin is due to reduced cell-cell adhesion.

Time-lapse movies of ctnna morphant DCs at 50% epiboly revealed dynamic plasma membrane activity that was not observed in controls (supplementary material Movies 5, 6). To investigate the nature of this membrane activity, we took time-lapse movies at a faster interval rate (4 second intervals for 5 minutes) in the most external DCs layer at the animal pole. Time-lapse images revealed that ctnna morphant DCs had plasma membrane activity that resembled blebbing (Fig. 5B, arrows indicate membrane blebs; supplementary material Movie 8) (Charras et al., 2006). These cells displayed up to eight bleb-like protrusions during the time-lapse recording, with an average number of 2.44±0.17 bleb-like protrusions per cell per 5 minutes (Fig. 5D,E). By contrast, >80% of both control and cdh1 morphant cells had no bleb-like activity, and the remainder had fewer than three bleb-like activities over the same period (Fig. 5A,C,D; supplementary material Movies 7, 10); overall, the average bleb-like activity of both control and cdh1 morphant cells was ∼0.2 per cell (Fig. 5E). Co-injection of gfp-mctnna mRNA with ctnna mo reduced the number of bleb-like protrusions to an average similar to that observed in control embryos (0.49±0.11) (Fig. 5D,E; supplementary material Movie 9), whereas injection of gfp-mctnna mRNA did not induce blebbing (Fig. 5D,E). Injection of the mismatched ctnna mm-mo did not trigger blebbing of DCs, similar to control cells (0.12±0.032 blebs per cell; Fig. 5D,E).

To address how αE-catenin regulates plasma membrane blebbing, we tested whether αE-catenin expression was required in the blebbing DCs and/or in the surrounding DCs. Transplantation experiments were performed in which DCs from ctnna morphants were transferred into control embryos and vice versa. Control DCs transplanted into ctnna morphant embryos did not display protracted plasma membrane blebbing (23/23 cells in ten embryos) (Fig. 6A; supplementary material Movie 11). By contrast, ctnna morphant DCs transplanted into a control cell background displayed protracted plasma membrane blebbing (22/42 cells in ten embryos Fig. 6B; supplementary material Movie 12) similar to that of DCs in the ctnna morphant from which they were transplanted. Thus, plasma membrane blebbing of ctnna morphant DCs was cell autonomous, and increased plasma membrane blebbing in surrounding ctnna morphant DCs did not provoke blebbing in control DCs.

Myosin II activity is required for plasma membrane blebbing (Blaser et al., 2006), and we tested whether myosin II was also involved in blebbing induced in ctnna morphants. Embryos were incubated with vector (DMSO) or blebbistatin, a well-characterized inhibitor of myosin II (Kovacs et al., 2004); note that embryos displaying multinucleated cells were not analyzed further owing to defects in cytokinesis and cell division. Blebbistatin inhibited plasma membrane blebbing in ctnna morphants, indicating the involvement of myosin II in this process (Fig. 6C-F, arrows indicate blebbing; supplementary material Movies 13-16).

Unlike ctnna morphant DCs, cdh1 morphant DCs did not exhibit increased plasma membrane blebbing. Therefore, we investigated further the relationship between αE-catenin and E-cadherin in regulating plasma membrane blebbing. One possibility that could explain the difference in phenotype is that myosin II is activated in ctnna morphant DCs but not in cdh1 morphant DCs. To explore this possibility, we performed immunostaining of whole-mount embryos at 50% epiboly using an antibody that recognized the phosphorylated form of myosin light chain (pMLC); phosphorylation of MLC is correlated with myosin II activation (Amano et al., 1996). We took confocal images of the external epiblast layer at the animal pole (Fig. 7A-C). Quantification of the fluorescence intensity (FI) of pMLC staining at cell-cell contacts revealed that pMLC is enriched equally at cell-cell contacts in both cdh1 (FI: 1.49±0.22) and ctnna morphants (FI: 1.44±0.14) compared with the control FI value normalized to 1 (Fig. 7D). These results indicate that the difference in blebbing phenotype between the two morphants is not dependent on differences in myosin II activity.

Another possibility is that E-cadherin-mediated cell-cell adhesion triggers or initiates plasma membrane blebbing by contacting αE-catenin-depleted cells. To address this possibility, we performed another transplantation experiment in which ctnna morphant DCs were transferred into a cdh1 morphant host. The ctnna morphant DCs did not show plasma membrane blebbing when they were surrounded by cdh1 morphant DCs (34/34 cells, nine embryos; Fig. 8A; supplementary material Movie 17), in contrast to ctnna morphants DCs transferred into control embryos (Fig. 6A; supplementary material Movie 12). We also performed a co-depletion of E-cadherin and αE-catenin. Significantly, these morphants had very little plasma membrane blebbing (0.016±0.07 blebs for cell; Fig. 8B-D; supplementary material Movie 18), similar to cdh1 morphants and controls embryos (Fig. 5; supplementary material Movies 7, 10), but unlike the ctnna morphants (2.44±0.17 blebs for cell; Fig. 5; supplementary material Movie 8). Taken together, these results indicate that plasma membrane blebbing in the ctnna morphant DCs is dependent on E-cadherin-mediated cell-cell contacts.

Next, we investigated the possibility that E-cadherin could function with other proteins involved in plasma membrane blebbing, such as ezrin (Ezr). Ezrin is an ERM-domain protein that regulates cell cortex stability and cell adhesion and plays important roles in blebbing dynamics (Charras and Paluch, 2008; Arpin et al., 2011); depletion of ezrin increases blebbing in zebrafish mesoderm cells (Diz-Munoz et al., 2010). We first confirmed that ezrin depletion increased blebbing of epiblast DCs. Indeed, DCs in ezr morphants (8 ng mo, high dose) had extensive plasma membrane blebbing (Fig. 9A,B; supplementary material Movie 19), similar to ctnna morphant DCs (Fig. 5B; supplementary material Movie 8).

We next investigated whether ezrin and αE-catenin synergize to suppress plasma membrane blebbing, by co-injecting embryos with low concentrations of both ezr and ctnna mo. Note that DCs in embryos injected with a low concentration of either ctnna mo (0.8 ng, low dose) or ezr mo (1 ng, low dose) did not have increased plasma membrane blebbing (Fig. 9A,B; supplementary material Movies 20, 21). However, co-injection of these low concentrations of ezr and ctnna mo resulted in increased plasma membrane blebbing (Fig. 9A,B; supplementary material Movie 22). By contrast, co-injection of a high concentration of cdh1 mo (8 ng) and ezr mo (8 ng) caused a reduction in membrane blebbing compared with embryos injected with a high concentration of ezr mo alone (Fig. 9A,B; supplementary material Movies 19, 23). Taken together, these results indicate that blebbing in both the ctnna and ezr morphant DCs is dependent on E-cadherin-mediated cell-cell adhesion. As depletion of both ezrin and αE-catenin had an additive effect, it is likely that these proteins function in the same pathway to suppress blebbing.

ERM proteins are auto-inhibited, and this inhibition is relieved by phosphorylation, which allows binding to partners and localization at the plasma membrane (Arpin et al., 2011). We used an anti-phospho-ERM domain antibody to investigate activation and localization of ERM-domain proteins in DCs at 50% epiboly in the background of αE-catenin or E-cadherin depletion (Fig. 9C-E). Control DCs had weak phospho-ERM domain staining at the plasma membrane (Fig. 9C). By contrast, DCs in ctnna or cdh1 morphants displayed strong staining at the plasma membrane, indicating activation and plasma membrane recruitment of ERM domain proteins in these mutant backgrounds (Fig. 9D,E). Note that DCs in an ezr morphant did not display any change in the staining intensity or subcellular localization of αE-catenin or E-cadherin (supplementary material Fig. S5); however, we cannot exclude the possibility that the lack of changes in E-cadherin and αE-catenin localization in ezrin morphant is due to the activity of moesin (Link et al., 2006). These results indicate that αE-catenin and E-cadherin localize to the plasma membrane independently of ezrin, and that they might be involved in regulating cycles of ERM protein activation and inactivation required for plasma membrane.

Recent biochemical studies and analysis in mammalian tissue culture cells have indicated important roles for the actin-binding protein αE-catenin in both E-cadherin-mediated cell-cell adhesion and an E-cadherin-independent role in controlling membrane dynamics and cell migration (Benjamin and Nelson, 2008; Borghi et al., 2010). E-cadherin mediated cell-cell adhesion and cell migration are regulated by the actin cytoskeleton (Harris and Tepass, 2010), and must be coordinated during complex cell movements in gastrulation (Hammerschmidt and Wedlich, 2008). Here, we investigated the role of αE-catenin in early zebrafish development during coordinated cell movements in epiboly.

The ctnna morphant displayed an abnormal morphology (Fig. 1C) and a strong delay in epiboly (Fig. 1B) compared with controls. As shown in Figs 2 and 3, depletion of αE-catenin affected epiboly of both the EVL and deep cell layers. The combination of defects observed in those tissues might contribute to the overall delay in epiboly and abnormal morphology shown in Fig. 1B,C. The observation that both the EVL and deep cell layers were affected and that there was a delay in onset of pax2.1 expression (supplementary material Fig. S1E,F) indicates that, in addition to these morphological defects, there might be a developmental delay in the ctnna morphant. An intriguing possibility is that the delay of pax2.1 expression could be caused by the defect in cell adhesion described in Fig. 4 and supplementary material Fig. S4. Additional experiments are needed to understand the possible role of αE-catenin in the YSL.

The delay in the EVL cell margin correlated with the failure of cells at the margin to elongate properly (Fig. 3). There is a correlation between the elongation of these cells and the advancing EVL margin (Koppen et al., 2006). In contrast to the ctnna morphants, the cdh1 morphants did not displayed a cell elongation defect (Fig. 3), as reported by others (Shimizu et al., 2005; Koppen et al., 2006; Lin et al., 2009). Furthermore, depletion of E-cadherin did not rescue the EVL morphology defects in the background of the ctnna morphant. These results indicate that whereas E-cadherin is dispensable for the EVL epiboly, αE-catenin is necessary for this process. Although we cannot exclude the possibility that the function of E-cadherin is replaced by another cell-cell contact complex, such as tight junctions (Siddiqui et al., 2010), we suggest that this phenotype might reflect the role of αE-catenin in regulating the actin cytoskeleton independently of E-cadherin. This function of αE-catenin might take place in the EVL and/or in the YSL.

The delay in epiboly of the deep cell margin in the ctnna morphant was caused by defects related to cellular processes during RI (Figs 2, 4; supplementary material Movies 1-4). Ctnna morphant deep cells were capable of RI but were unable to stabilize cell-cell contacts with their neighbors in the external layer, and subsequently moved back into the deeper layer in a process that we term reverse radial intercalation (revRI) (Fig. 4; supplementary material Movies 2-4). Significantly, cdh1 morphants had a similar phenotype to ctnna morphants (Fig. 3; supplementary material Movie 3), as shown previously for cdh1 mutants (Kane et al., 2005; Slanchev et al., 2009). Using ex vivo cell aggregation assays, we also showed that depletion of αE-catenin impaired cell-cell adhesion to a similar extent as E-cadherin depletion or Ca²⁺ chelation (supplementary material Fig. S4). These results indicate that ctnna and cdh1 morphant deep cells share the same defect in cell-cell adhesion and, thus, αE-catenin appears to be important in E-cadherin-mediated adhesion between deep cells and the overlying EVL (Shimizu et al., 2005; Slanchev et al., 2009). Taken together, these data highlight the importance of αE-catenin in E-cadherin-dependent cell-cell adhesion in early zebrafish development.

However, our results also identified differences in cell behavior in ctnna and cdh1 morphants (Fig. 5; supplementary material Movies 5-10). Ctnna morphant deep cells exhibited extensive plasma membrane blebbing, which was not observed in the cdh1 morphant or controls. Significantly, plasma membrane blebbing in Ctnna morphants was cell autonomous (Fig. 6; supplementary material Movies 11, 12).

Recent studies have highlighted the importance of plasma membrane blebbing in cell migration in certain cell types (Cunningham et al., 1992; Blaser et al., 2006; Fackler and Grosse,2008). Blebbing is a type of plasma membrane protrusion driven by detachment of the plasma membrane from the cell cortex and cytoplasmic hydrostatic pressure that causes the plasma membrane to transiently protrude outwards (Charras and Paluch, 2008; Tinevez et al., 2009). Plasma membrane blebbing, unlike lamellipodial formation, does not involve actin polymerization in this initial protruding phase. Once the membrane bleb is formed, ERM proteins, such as ezrin or moesin, are recruited to the plasma membrane to reconnect the membrane to the actin cytoskeleton in the cell cortex (Charras et al., 2006; Diz-Munoz et al., 2010). Finally, activation of Rho signaling and myosin II activity, and F-actin assembly cause retraction of the membrane bleb (Charras et al., 2006; Charras and Paluch, 2008; Kardash et al., 2010).

We investigated mechanisms that might trigger plasma membrane blebbing in ctnna morphant DCs. Immunostaining for pMLC showed that there was an enrichment of activated myosin II in both morphant DCs, compared with controls (Fig. 7). This result indicates that the lack of membrane blebbing in the cdh1 morphant was not due to reduced activation of actomyosin contractility.

We also observed that depletion of E-cadherin or αE-catenin caused increased levels of phospho-ERM domain proteins at the plasma membrane (Fig. 9C-E). For the ctnna morphant, this might be in response to increased blebbing because ERM proteins are recruited to the site of blebbing in order to retract the bleb (Charras et al., 2006). However, this could not be the case in cdh1 morphant cells that lack blebbing. Alternatively, recruitment of ERM protein to the plasma membrane might be independent of blebbing. ERM domain proteins regulate adherens junction assembly and positioning (Van Furden et al., 2004; Pilot et al., 2006; Arpin et al., 2011), and increased pERM staining in cdh1, and perhaps ctnna morphant cells, might be a response to reduced levels of functional adherens junctions.

To address the relationship between αE-catenin and E-cadherin in blebbing, we showed that the blebbing phenotype of ctnna morphant cells could be rescued by either transplanting ctnna-depleted cells into a cdh1 morphant background, or co-depletion of αE-catenin and E-cadherin in the same cells (Fig. 8; supplementary material Movies 17, 18). Moreover, this pathway involves ERM domain proteins as suboptimal co-depletion of ezrin and αE-catenin caused blebbing, and co-depletion of ezrin and E-cadherin rescued the blebbing phenotype induced by ezrin depletion alone. Taken together with recent studies (Kardash et al., 2010; Speirs et al., 2010), our results demonstrate that plasma membrane blebbing induced by either αE-catenin or ezrin depletion is dependent on E-cadherin-mediated cell-cell engagement/adhesion.

We suggest that as deep cells in the most external layer became more adhesive (Kane et al., 2005), increased E-cadherin levels decrease the membrane-to-cortex attachment that normally opposes membrane blebbing. We posit that αE-catenin, like ezrin, counter this by stabilizing membrane-to-cortex attachment and suppressing membrane blebbing. Normally, this allows deep cells to complete radial intercalation and establish cell-cell adhesion with each other and the overlying EVL. This balance between functions of E-cadherin and αE-catenin-ezrin is disrupted by depleting αE-catenin or ezrin, which results in weaker membrane-to-cortex attachment and protracted membrane blebbing (Figs 5, 9), and is counter-balanced by co-depletion of E-cadherin (Figs 8, 9). This is the first insight into the diverse roles of αE-catenin during early stages of vertebrate development in vivo, not only in cell-cell adhesion as shown previously in Xenopus and mice (Kofron et al., 1997; Torres et al., 1997), but also in regulating membrane dynamics and cell migration as suggested from studies in tissue culture cells (Benjamin et al., 2010; Borghi et al., 2010).

We thank Will Talbot and Lilianna Solnica-Krezel for training and advice. We thank Dr James Marrs for the zebrafish E-cadherin antibody.