The E-cadherin/AmotL2 complex organizes actin filaments required for epithelial hexagonal packing and blastocyst hatching

The E-cadherin/AmotL2 complex organizes actin filaments required for epithelial hexagonal packing and blastocyst hatching
Sebastian Hildebrand1,2, Sara Hultin2, Aravindh Subramani2, Sophie Petropoulos1, Yuanyuan Zhang 2, Xiaofang Cao3, John Mpindi4,5, Olli Kalloniemi4,5, Staffan Johansson3, Arindam Majumdar2,6, Fredrik Lanner1 & Lars Holmgren2 Epithelial cells connect via cell-cell junctions to form sheets of cells with separate cellular compartments. These cellular connections are essential for the generation of cellular forms and shapes consistent with organ function. Tissue modulation is dependent on the fine-tuning of mechanical forces that are transmitted in part through the actin connection to E-cadherin as well as other components in the adherens junctions. In this report we show that p100 amotL2 forms a complex with E-cadherin that associates with radial actin filaments connecting cells over multiple layers. Genetic inactivation or depletion of amotL2 in epithelial cells in vitro or zebrafish and mouse in vivo, resulted in the loss of contractile actin filaments and perturbed epithelial packing geometry. We further showed that AMOTL2 mRNA and protein was expressed in the trophectoderm of human and mouse blastocysts. Genetic inactivation of amotL2 did not affect cellular differentiation but blocked hatching of the blastocysts from the zona pellucida. These results were mimicked by treatment with the myosin II inhibitor blebbistatin. We propose that the tension generated by the E-cadherin/AmotL2/actin filaments plays a crucial role in developmental processes such as epithelial geometrical packing as well as generation of forces required for blastocyst hatching.

A central question during development is how single cells form functional multi-cellular organ structures. The high reproducibility indicates intricate synchronization of cellular processes such as migration, proliferation and cell shape changes. Much attention has been focused on how growth factors form biochemical gradients that govern some of these processes1–3. However, less is known regarding how mechanical signals or forces modulate cell shape and control cellular expansion4, 5.Cells perceive and respond to exogenous mechanical forces via different points of contact in the outer mem- brane. Forces exerted on the extra-cellular matrix are detected by epithelial cells via integrins in focal adhesions which transfer tension from the extracellular matrix to the cytoskeleton6. Low rigidity in the extra-cellular matrix transfers less extracellular force and thereby promotes the formation of organ-like epithelial structures in vitro whereas increased force or stiffness in the matrix causes loss of tissue architecture associated with tumor progres- sion and promotes cell proliferation7–10.Recent evidence has also shown that actomyosin contractility is transmitted via the adherens junctions. External forces applied to cadherins have indicated a mechanical coupling between the cytoplasmic domain of cadherin and the actin cytoskeleton11. Cellular interactions and the force-mediated morphological changes are also important for the processes involved in organ development. One example is apical contraction where theapical side of cells contracts to a wedge-like shape required for sheets of cells to fold or bend to form invaginationse.g. during Drosophila germ band extension, vertebrate gastrulation or neural tube formation12–14. An important issue is how force is transmitted from E-cadherin to the cytoskeleton. Classical cadherins are normally associ- ated to p120, β and α–catenins, which are essential for the connection to actin filaments. Recent evidence sug- gests that α-catenin may undergo force-dependent conformational changes that regulate binding of the minimal cadherin-catenin complex to an actin filament under force.

Force-induced conformational changes also allows binding of effector proteins such as vinculin dependent on junctional maturity and myosin II activity15, 16.The angiomotin scaffold protein family is comprised of angiomotin (amot), angiomotin like 1 (amotL1) and angiomotin like 2 (amotL2). Each protein exists in two different isoforms, whereat the two amotL2 isoforms are called p100 amotL2 and p60 amotL2. All three amot family members have been studied extensively in endothelial cells, demonstrating their importance in cell migration, polarization, proliferation and tight junction stability17–20. Furthermore, the amot family of proteins has been shown to be vital for maintaining polarity, regulating cell growth and motility, and facilitating tight junction stability21–24. Amot has been reported to bind F-actin, thereby controlling cell shape in endothelial cells25 and facilitating actin cytoskeleton remodeling in epithelial cells26. p100 amotL2 has been shown to localize to the cellular junctions of epithelial tissue cells with so far undescribed functional impact27. We have previously shown that amotL2 is essential for normal vascular development, specifically during vasculogen- esis where amotl2 associates to VE-cadherin to mediate actomyosin-dependent mechanical force required for aortic expansion28. Finally amot and amotL2 have further been shown to control lineage specification of the first cell type of the mammalian embryo, the trophectoderm29, 30 which also is the first epithelial tissue to form.In this report, we have analyzed the functional role of amotL2 in epithelial cell-cell junctions in several cul- tured epithelial cells lines in vitro as well as in zebrafish skin epithelium and mammalian trophectoderm in vivo. We show that amotL2 is a component of the E-cadherin complex that is essential for the formation of radial actin filaments. Functionally, depletion of amotL2 and subsequent loss of radial actin fibers resulted in stalled hatching of mouse and human embryos from the zona pellucida.

Results
Analysis of AMOTL2 mRNA levels in organ tissues revealed a ubiquitous expression in all organs except lymphoid, blood and bone marrow cells (Supplemental Fig. 1a). Furthermore, amotL2 expression in 755 human cell-lines in vitro indicated that amotL2 is primarily expressed in epithelial cells (Supplemental Fig. 1b).To analyze potential role of amotL2 in formation and maintenance of cell-cell contact, we depleted amotL2 protein levels using shRNA carrying lentiviruses targeting approach as previously described31, 32. Three epithelial cell lines were utilized: Madin-Darby Canine Kidney (MDCK) cells, which are tumorigenic kidney epithelium cells derived from dog, Caucasian colon (Caco-2) cells, a human epithelial colorectal adenocinoma cell line and an immortalized human keratinocyte cell line (HaCaT) derived from human skin. The knock-down efficiency was analyzed by immunofluorescent staining and by western blot (Fig. 1a,b and Supplemental Fig. 2a and b). The effect of amotL2 depletion on the junctional localization of tight junction protein ZO-1 and the adherens junc- tion protein E-cadherin was assessed by immunofluorescence staining (Fig. 1c). As shown in the profile plots in Fig. 1d, both proteins still localized to cell-cell junctions. Amot proteins bind to the Par3 and Crb3 apical polarity protein complexes21, 33, 34. We therefore assessed whether removal of amotL2 protein would perturb apical-basal polarity. MDCK cells were stained antibodies against the apical protein podocalyxin (Fig. 1e). As shown in Fig. 1f, podocalyxin could be detected in the apical membranes of both Ctrl and amotL2 shRNA MDCK cells.p100 amotL2 is required for actin organization.

The amotL2 depleted cells exhibited a markedly altered cellular morphology when cultured at 50% confluency compared to that of the control cells. At 50% con- fluency MDCK cells grow in loosely packed colonies. The changes in cellular surface area and junction length were quantified and a ~6-fold increase in cell area and a ~4-fold increase in junction length was detected in amotL2 shRNA cells (Fig. 2a–f). The cellular surface area phenotype could be rescued by re-expression of p100 amotL2 (Supplemental Fig. 2c) or was partially restored in densely packed cell cultures (Fig. 2a,c and e).Next we investigated whether changes in cytoskeletal components could explain the altered cell shape. However, the distribution of F-actin using phalloidin staining was dramatically altered. Typically, control cells exhibited junctional actin as well as fibers that were apparently connecting to filaments of the neighboring cells via membrane junctions (Fig. 2g and h). In contrast, amotL2 shRNA cells typically lacked the radial actin fibers that connected perpendicular to the cell membrane (Fig. 2g,h and Supplemental Fig. 2). The localization of the tubulin or keratin networks were not visibly affected (Supplemental Fig. 3).The packing of cell layers into hexagonal shapes was originally observed by Lewis in 1928 and is a geometrical form that is universally conserved, extending from plants, sea urchins to human epithe- lia35, 36. ZO1 staining of control or amotL2 shRNA MDCK showed clear differences in surface areas and geometrical shape (Fig. 3a). The number of sides bordering neighboring cells was quantified and their relative frequencies were summarized in the bar diagrams in Fig. 3b and c. Caco-2 and MDCK cells con- sisted to the great majority of cells with either pentagonal or hexagonal shape (Fig. 3b). Following amotL2 knock-down with shRNA the cell shape was drastically changed with most cells bordering to four neigh- boring cells (Fig. 3c). Interestingly, densely packing cells did not affect (in contrast to the cell area shown in Fig. 2) the geometrical shape of the epithelial cells (Fig. 2d and e). With the rationale that amotL2 knock-down cells showed deficiency in the microfilament network, we hypothesized that loss of acto- myosin contractility could account for the change in cell shape.

As such, we treated control cells with the myosin II inhibitor, blebbistatin, before quantifying cell geometry. Myosin II inhibition via blebbistatin treatment has been shown to facilitate actin fiber disassembly37. Interestingly, a 2-hour treatment with bleb- bistatin drastically affected epithelial cell shape, mimicking the effect of amotL2 depletion (Fig. 3a and f ).Taken together, we show that amotL2 is required for organizing the actomyosin network to control cell area, size and shape.AmotL2 is part of the E-cadherin junctional protein complex. We have previously shown that p100 amotL2 associates to the VE-cadherin complex in endothelial cells28. The corresponding cadherin in epithelial cells is E-cadherin which has been extensively studied as loss of protein function or expression hasbeen implicated in tumor progression and invasion38, 39. Next we assessed whether amotL2 was associated to the E-cadherin junctional complex in epithelial cells. For this purpose, we performed co-immunoprecipitation analysis and could show that amotL2 was directly or indirectly bound to E-cadherin as well as α and β-catenin (Fig. 4a). In addition, we could further demonstrate that amotL2 is associated with the junctional scaffold protein MAGI1 and actin (Fig. 4a). We and others have shown that the amot family of proteins associates to the scaffold junctional protein MAGI1 via a WW protein interaction motif27, 28, 40. It has further been reported that MAGI1 associates to VE-cadherin via binding to β-catenin41. These findings raised the possibility that MAGI1 acts as a direct link between actin-amotL2 and E-cadherin. To identify the amotL2 domains responsible for the binding to E-cadherin, we performed co-immunoprecipitation analysis using deletion mutants covering the N-terminal protein interaction motifs as well as the C-terminal PDZ-binding motif (Fig. 4b). Using this strategy, we identi- fied a domain of 87 a.a. which was responsible for the interaction between amotL2 and E-cadherin (Fig. 4c).

This domain contains two potential WW-protein interaction sites (Fig. 4b), which were mutated by substituting the tyrosine to an alanine. However, none of these sites appeared to be essential for the association to E-cadherin as analyzed by co-immunoprecipitation (data not shown). The association of MAGI1 and actin was mapped to the 101–220 a.a. N-terminal domain of p100 amotL2 which was distinct from the E-cadherin interaction site. By mutating the tyrosine to alanine we could show that MAGI1 binding was lost in the LPTA mutant (Fig. 4d). In addition, the association of actin to amotL2 was dependent on the LPTY motif (Fig. 4d). All Angiomotin protein family members also contain a C-terminal PDZ-binding motif (Fig. 4b). This motif binds directly to the PATJ/MUPP1 or the Par3 polarity proteins21, 33, 42. However, neither the LPTY/PPQY motifs nor the C-terminalPDZ-binding motif was required for the interaction with E-cadherin (Fig. 4d). In conclusion, these data show that E-cadherin MAGI1/actin associate to amotL2 via separate domains.AmotL2 sensitizes epithelial layers to external mechanical force. Our findings indicated a role of amotL2 in the coupling of actin fibers to membrane anchors and thereby potentially connecting endogenous cellular tensile forces to neighboring cells. This also opened up the question whether this protein complex could integrate external mechanical forces and transmit them at a supra-cellular level. In order to address this issue, we used a cellular stretch assay (Fig. 5a). In this system HaCaT (human keratinocytes) cells were grown to conflu- ency on fibronectin-coated elastic silicone membranes. Mechanical force was loaded on the membrane using a uniaxial cyclic stretch apparatus. After 2 hours of 30% cyclic stretch, approximately 50% of the control cells exhib- ited disruptions at the cellular junctions resulting in a discontinuous epithelial sheet (Fig. 5b and c). The actin filaments in the remaining cells reorganized perpendicular to the applied force.

In contrast, the epithelial layer of the amotL2 shRNA cells remained intact (Fig. 5b and c). We used blebbistatin to investigate whether inhibition of actomyosin contractility could prevent the cell junction disruption in a similar fashion. A 15 min pre-treatment with blebbistatin completely prevented the mechanical stress induced junctional rupture in the stretch assay. These data indicated that amotL2 influences cell sheet topology by promoting actin-mediated cellular stiffness,allowing cellular adaptation to extracellular force. AmotL2 shRNA or blebbistatin treatment promotes a more fluid epithelium that is more resilient to mechanical stress due to higher elasticity.AmotL2 controls epithelial geometry in Zebrafish skin epithelium. The complex development of epithelial structures can be studied in zebrafish, which allows for an easy visualization of these processes at a cellular level in vivo. In zebrafish, amotl2 is duplicated with paralogues on chromosomes 6 (amotl2a) and 2 (amotl2b)28.We used an anti-sense morpholino (MO) approach to target the translation initiation sites of both amotl2 par- alogues. Bright-field images of control and amotl2 MO-treated embryos at 48 hours post-fertilization are shown in Fig. 6a and b. The amotl2 morphants suffered from pericardial edema due to circulatory defects as previously described (Fig. 6b)28. We could show amotl2 expression in zebrafish skin cells with whole mount immunofluores- cent stainings by using the previously established amotL2 antibody (Fig. 6c)28. AmotL2 was confined to cellular junctions and cytoplasmic filamentous structures. The knock-down efficiency of the MO in skin cells was assessed by whole mount immunofluorescent staining and by qPCR as previously published (Fig. 6c)28. The effect on thecytoskeletal organization was analyzed by phalloidin staining. Similar to the epithelial amotL2 shRNA cells, junc- tional actin was clearly detectable whereas the non-junctional actin filaments were almost completely removed (Fig. 6d).

E-cadherin staining was strictly localized to cell-cell junctions in control MO skin cells but more dif- fusely spread in the membrane and cytosol of skin cells in the amotL2 morphants (Fig. 6e), as also earlier detected in the amotL2 shRNA epithelial cells (Fig. 1f). Junctional stainings using antibodies against ZO1 showed that the cellular surface area was almost doubled (Fig. 6f and h). Analysis of the skin epithelium also revealed a decrease in cells with hexagonal shape and an increase of cells with fewer sides towards their neighboring cells (Fig. 6g). Co-injection of the amotl2 MO with human p100 AMOTL2 mRNA could rescue the actin architecture in the skincells (Supplemental Fig. 4) and restore geometry and cellular size (Fig. 6g and h). We concluded that amotL2 is essential for the packing of skin epithelial cells in vivo.AmotL2 is expressed in the human and mouse trophectoderm. The first specified cell-type to be established during mammalian development is the trophectoderm (TE) that belongs to the class of epithelium with tight and adherens junctions and exhibit cell polarization (Fig. 7a). The trophectoderm is specified through a positional mechanism where the outer cells sense that they are in an outer position. This positional sensing pro- cess has been shown to involve components of the Hippo signaling pathway together with amot and amotL229, 30 to drive expression of the trophectoderm transcription factor Cdx243. Ultimately, correct formation of the tro- phectoderm is crucial for normal embryo development as it mediates the implantation of the blastocyst into the uterus and gives later rise to the placenta44.

Since amot together with amotL2 has been implicated in the segrega- tion of the inner cell mass (ICM) from the trophectoderm29, 30 we reasoned that it would be of relevance to investi- gate the expression pattern of amotL2 and its potential role in regulating cell shape and lineage specification of the epithelial trophectoderm. Mining of recently published single-cell RNA sequencing data sets showed significantlyhigher expression of amotL2 in the trophectoderm of both human and mouse blastocysts compared to the inner cell mass (Fig. 7b and c)45, 46. Antibody staining confirmed this finding at a protein level and showed subcellular localization towards the lateral cell-cell junctions in the trophectoderm (Fig. 7d–g).AmotL2 is essential for radial actin fibers and cell shape of the blastocyst trophectoderm. Phalloidin stainings of human and mouse blastocysts confirmed the presence of both, junctional actin aligning along the cell junctions and radial actin fibers oriented perpendicular to the cellular junctions, spanning through- out the cytoplasm (Fig. 7h–k).To investigate whether radial actin fiber maintenance in trophectoderm cells also depends on amotL2 expres- sion, as observed in epithelial cell lines in vitro and in zebrafish in vivo, we silenced the amotL2 gene in mouse blastocysts. For this purpose we used a floxed mouse strain28 and injected Cre mRNA in the fertilized zygotes to excise the amotL2 gene. As a control, GFP mRNA was injected into the embryos. Injected embryos were culture in vitro and analyzed at early and late blastocyst stage. Successful excision of the floxed amotL2 was confirmed through antibody staining (Supplemental Fig. 5a).As amot-family members have been implicated in trophectoderm versus inner cell mass specification29, 30 we first explored this issue. We stained control and amotL2−/− mouse blastocysts for the trophectoderm specific transcription factor CDX2, which was correctly expressed only in outer trophectoderm cells of both blastocyst cohorts (Supplemental Fig. 5b).

We further analyzed the number of cells and the ratio of trophectoderm versus inner cell mass cells in amotL2+/+ and amotL2−/− blastocysts, which was equal in both populations (Fig. 8d, Supplemental Fig. 5c). Turning focus towards the blastocyst actin cytoskeleton we could detect a distinct absence of radial actin fibers in Cre injected mice whereas the junctional actin remained intact (Fig. 8a and b). In addition to the loss of radial actin fibers, Cre mRNA injection resulted in approximately 60% increased cell area size of trophectoderm cells contributing to an increased blastocyst volume (Fig. 8b, e and f).To further support that the observed changes in cell area size and volume were dependent on remodeling of radial actin fibers, we incubated mouse blastocysts with blebbistatin (100 µM blebbistatin for 6 hours). Thistreatment phenocopied Cre mRNA injected blastocysts resulting also in increased cell area size and volume with- out affecting cell numbers (Fig. 8c,g–i and Supplemental Fig. 5c).To further investigate if the loss of amotL2 and actin fibers would also impact the hexagonal cell shape of tro- phectoderm cells, we evaluated 3D images of control and amotL2 excised blastocysts at early and late blastocyst stage. The most abundant cell shape of control blastocyst trophectoderm cells was hexagonal (Fig. 9a,d and e). In contrast, amotL2 excised blastocysts were mainly assembled out of trophectoderm cells comprising a pentagonal cell shape (Fig. 9b,d and e). Again, similar results were observed following pharmacological disruption of actin fibers using blebbistatin treatment (Fig. 9c,f and g).

AmotL2 is required for blastocyst hatching. The mammalian embryo is enveloped by the zona pel- lucida, a glycoprotein extracellular shell, that prevents premature implantation into the uterine wall47, 48. When the embryo is ready to implant it must therefore first hatch through the zona pellucida before attaching to the endometrium. This process is mediated by combined enzymatic digestion of the matrix together with expansion and contraction cycles that drive the blastocyst out of the shell49, 50. As we have detected a loss of radial actin fibers in the mouse blastocyst after amotL2 excision, we explored whether these radial actin fibers are required for the process of hatching as described earlier51, 52. Indeed, during in vitro culture, 90% of all control blastocysts hadhatched whereas only 20% of the amotL2 excised embryos completed the same process but were trapped in an intermediate hatching state (Fig. 10a,b and c).Again, pharmacological disruption of actin filaments during the same period using blebbistatin reproduced the hatching phenotype (Fig. 10d and e). Finally, similar results were observed in blebbistatin treated human embryos where 75% of the control blastocysts hatched whereas 75% of the blebbistatin treated blastocysts remained in the hatching state (Fig. 10f and g) suggesting a conserved function also in humans.

Discussion
The maintenance of epithelial cellular shapes depends on the ability of cells to relay and sense mechanical forces. Our report provides mechanistic insight on how these forces may be organized by the formation of cell-cell con- tacts. We show that amotL2 links E-cadherin to cytoskeletal actin and thereby affects the cytoskeletal organiza- tion, cell geometry and cell topology of epithelial cells in vitro and in vivo.We provide evidence that amotL2 is essential for the development of specific actin fibers that connect in a perpendicular fashion to the outer membrane. These radial actin fibers are anchored in the membrane by cou- pling to the E-cadherin/AmotL2/MAGI1 ternary protein complex. Therefore, the amotL2 complex differs from the EPLIN/E-cadherin complex that connects junctional actin to the membrane53, 54. In contrast, reduction of amotL2 levels in epithelia in vitro and in vivo results in the loss of radial actin fibers. Our observations suggest that amotL2 organizes actin and thereby physical properties of epithelial sheets. This was supported by the in vitro stretch experiments, which pointed to a significant change in physical properties, most likely due to the loss of radial actomyosin filaments. The importance has previously been highlighted in model systems of monolayer epithelial cells without matrix attachment, where reducing actin depolymerized with Latrunculin B decreased cell sheet stiffness by approximately 50%55.

The organization of epithelial cells into hexagonal shapes is highly evolutionary conserved35, 36. This appears to be the optimum method of packaging cells in regards to transducing force and minimizing energy expendi- tures. The integration of cells in hexagonal patterns is also of functional importance. Hexagonal packing in the vertebrate lens minimizes light scattering by plasma membranes, controls the orientation of cell division as well as planar cell polarity and the organization of cilia. Previous studies have indicated the importance of cadherin-mediated cell-cell contacts and actomyosin contractility in controlling cell geometry56–58. Recent work has also provided evidence that tensile stress in the cellular cortex control hexagonal packing. Relaxation of the contractile force in the junctional actin resulted in lengthening of the adherens junctions and altered cell geome- try59. Our observations suggest that amotL2 expression is critical for maintaining cellular geometry by supporting contractile forces required for the shaping of cuboidal epithelial cells and secondly by controlling hexagonal packing of cells in planar epithelium. Although it is not entirely clear how amotL2 affects the epithelial geometry, it is conceivable that either the loss of contractile actin filaments running perpendicular to the membrane is the cause of the loss of geometry or that contraction of junctional actin is also affected. The latter is supported by the observed lengthening of junction in amotL2 depleted cells.

The function of amot family members has previously been explored in the mouse blastocyst linking amot and amotL2 to Hippo signaling-dependent segregation of trophectoderm and inner cell mass29, 30. In these studies the inner cell mass cells of amot and amotL2 double-targeted embryos failed to fully mature the inner cell mass cells with ectopic expression of the trophectoderm marker gene Cdx2. This phenotype was not evident in our study where only amotL2 was targeted indicating overlapping and redundant function with amot in controlling troph- ectoderm versus inner cell mass maturation.
Since the change in cell shape has not been described in amot null embryos, our data suggest a specific func- tion of amotL2 in linking E-cadherin to radial actin fibers. This observation was further supported by the distinct subcellular localization of the two family members, where amot is localized to the apical domain of trophecto- derm cells30, 51 whereas we detect amotL2 expression at the lateral side of trophectoderm cells.It is becoming increasingly clear that the mechanisms that control mouse and human preimplantation devel- opment is very poorly conserved, which can explain why mouse studies have not been all that helpful to improve human infertility treatments60. As the function of amotL2 appears to be universal between different types of epithelial cells from zebrafish to humans it is encouraging to see similar expression patterns of amotL2 as well as hatching phenotype in mouse and human embryos. This is further stressed by our detected shared hatching phenotype were both mouse- and human- blastocysts are trapped in an intermediate hatching state. This result suggests that the hatching process is controlled by orchestrate protease thinning of the zona pellucida together with proper biophysical characteristics of epithelial trophectoderm cells which we now show is dependent on amotL2 controlled radial actin fibers. We anticipate that amotL2 controlled epithelial properties will be important for many other epithelial structures throughout development and normal versus pathological physiology.

The following primary antibodies were used: actin (abcam, ab 3280); for human and dog cells: AmotL2 (polyclonal antibodies reactive to human AmotL2 C-terminal peptide, NH2- CLDSVATSRVQDLSDMVEILI -COOH); for zebrafish: AmotL2 (polyclonal antibody reactive to zebrafish amotL2 C-terminal peptide NH2- CQKAPSAVDLFKGVDDVSAE- COOH), custom made by Innovagen (Lund, Sweden); for mouse blastocysts: AmotL2 (GeneTex, GTX 120712; for human blastocysts: AmotL2 (Aviva, OAAB05455); α-catenin (BD, 610193); β-catenin (BD, 610154); CDX2 (BioSite, MU392A-UC); for human and dog cells: E-cadherin (BD, 610181); for mouse blastocysts: E-cadherin (Sigma, U3254); Ezrin (BD, 610602); GFP (Life technologies, A10262); MAGI1 (Sigma, WH0009223M3); Tubulin (Sigma, T5168); YAP1 (Santa Cruz, sc-101199); ZO-1 (Invitrogene, 339100). The following secondary antibodies were used: alexa fluor 405 anti- mouse (GE, A12380); alexa fluor 488 anti-mouse (GE, A11001); alexa fluor 488 anti-rabbit (GE, A11008); alexa fluor 488 anti-chicken (GE, A11039); alexa fluor 488 anti-rat (GE, A21208); alexa fluor 647 anti-mouse (GE, A31571); ECL anti-mouse IgG horseradish peroxidase (GE, NA931V); ECL anti-rabbit IgG horseradish peroxi- dase (GE, NA934V). To visualize the cellular actin fibers the F-actin probe Texas Red-X Phalloidin (Life technol- ogies, T7471) was used. Nuclei were visualized using DAPI (Sigma, F6057) in epithelial cells and zebrafish skin and with Hoechst Blebbistatin 33342 (Life technologies, H3570) in mouse- and human-blastocysts.