Zygotic loss of function of either E-cadherin or N-cadherin has no consequences for mesoderm morphogenesis

Drosophila E-cadherin (Tepass and Hartenstein, 1994) is expressed at low levels throughout the early blastoderm embryo from maternally provided mRNA. The zygotic transcription of the gene is activated at high levels in the cellularising blastoderm stage embryo, but remains inactive in the future mesoderm through repression by the transcription factor Snail (Oda et al., 1998). Thus, the ectoderm expresses high levels of E-cadherin, whereas the mesoderm continues to express the lower maternal level that remains detectable until after the mesoderm has invaginated. N-cadherin (encoded by CadN) is expressed in the early mesoderm under the control of Twist, and is later expressed in neuronal tissues (Iwai et al., 1997). The protein encoded by the neighbouring CadN2 locus has a partially redundant function in photoreceptor morphogenesis (Prakash et al., 2005).

Although the consequences of the loss of function of E-cadherin and N-cadherin in Drosophila have been described previously (Dottermusch-Heidel et al., 2012; Iwai et al., 1997; Oda et al., 1993; Prakash et al., 2005; Tepass and Hartenstein, 1994), the precise roles of these proteins in the earliest stages of mesoderm morphogenesis have not been studied in detail. We therefore re-analysed mutant embryos during gastrulation for subtle morphological defects. To delete all zygotic N-cadherin transcribed from both the CadN and CadN2 loci (Hill et al., 2001), we used Df(2L)E48, a deficiency that uncovers a small region containing both genes [see Materials and Methods; Fig. 1B (Walsh and Brown, 1998)] (Nick Brown, personal communication). These experiments confirmed that neither the (zygotic) loss of E-cadherin (shg) nor the loss of the N-cadherin (CadN, CadN2) genes resulted in defects in the formation of the ventral furrow, the segregation of the germ layers or spreading of the mesodermal cell layer (data not shown). This deficiency allowed normal development and mesoderm differentiation beyond the end of gastrulation to the fully extended germband stage (Fig. 1A,B).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Phenotypes of CadN and E-cad (shg) mutants. (A,B) Control and CadN mutant embryos stained for Twist (Twi, brown) and N-cadherin (N-cad, blue). (C–H) Stage-11 embryos of the indicated genotypes stained for Even-skipped (Eve, blue) and N-cadherin (brown) (C–F) or Fasciclin3 (Fas3, blue) and N-cadherin (brown) (G,H). Eve is expressed in 11 segmental pericardial-cell clusters (PC) in the dorsal mesoderm. Fas3 is expressed in the visceral mesoderm (VM). Control embryos are from a Df(2L)E48/CyO collection of embryos [homozygous for either the balancer or Df(2L)E48/CyO].

Changing the complementary cadherin gene expression during early embryogenesis

Embryos that are homozygous for either of the mutations discussed above lack zygotic expression of either E-cadherin or N-cadherin, but in each case the other cadherin is still present. It is plausible that differential expression of the remaining cadherin might be sufficient to keep the cell populations apart during mesoderm spreading, in the same manner as qualitative or quantitative differences in cadherin expression lead to the sorting of cell populations in vitro (Miyatani et al., 1989; Nagafuchi et al., 1987; Ninomiya et al., 2012; Nose et al., 1990; Steinberg and Takeichi, 1994) and in vivo (Godt and Tepass, 1998; González-Reyes and St Johnston, 1998; Hayashi and Carthew, 2004). We therefore analysed mesodermal development in embryos lacking both zygotic E-cadherin and N-cadherin. These mutant embryos contain only the maternally supplied E-cadherin, which is ubiquitously expressed at low levels throughout the whole embryo and is uniformly expressed in ectoderm and mesoderm (Oda et al., 1998). Such embryos develop normally during gastrulation and form a ventral furrow (data not shown). In addition, later stages of mesoderm development, including mesoderm spreading, were not significantly affected. Staining with antibodies against Even-skipped (Eve) or Fasciclin 3 (Fas3), which highlights either clusters of heart and muscle precursor cells in the dorsal mesoderm (PCs; Fig. 1C–F) or the developing visceral mesoderm (VM; Fig. 1G,H) (Tepass and Hartenstein, 1994), respectively, showed that zygotic cadherins are not necessary for early mesoderm development or subsequent differentiation. In particular, the complementary expression of the cadherins was not required for an early morphogenetic function during gastrulation. The results also suggested that adhesion provided by maternal E-cadherin is sufficient to sustain mesoderm morphogenesis.

Effects of increased levels of E-cadherin on invagination and spreading

It is possible that the maintenance of reduced levels of E-cadherin is crucial for mesoderm morphogenesis, and hence that there is a need to counter the increase in E-cadherin levels that is normally seen in the ectoderm during gastrulation. This assumption fits with the importance of Snail-mediated transcriptional repression for mesoderm morphogenesis (Seher et al., 2007). To test this hypothesis, we overexpressed E-cadherin using the full-length E-cadherin trangenes UAS cadh5 and UAS cadh5/9, and either a maternal-Gal4 or a twist-Gal4 driver both in wild-type embryos and in embryos lacking CadN. We initially investigated whether high levels of E-cadherin per se are detrimental for mesoderm development (even in the presence of N-cadherin), whereas the second set of experiments tested whether N-cadherin might be necessary to counteract the potentially negative effects of high E-cadherin expression in the mesoderm.

In blastoderm stage embryos, E-cadherin localises to the developing adherens junctions at a sub-apical position in the blastoderm epithelium. With the onset of cell shape changes in the mesoderm, the adherens junctions disassemble and re-localise at the most apical point of the cell–cell contact (Kölsch et al., 2007). This post-transcriptional modulation of E-cadherin is also indirectly controlled by Snail (Kölsch et al., 2007), at least in part through repression of the Bearded gene (Chanet and Schweisguth, 2012), a feature also seen during sea urchin gastrulation (Wu and McClay, 2007). However, E-cadherin is not restricted to this apical point, but is present throughout the lateral cell membranes, in decreasing concentrations towards the basal end of the cell (Mathew et al., 2011). Armadillo does not mirror this distribution, but is highly concentrated only at the most apical position (Mathew et al., 2011) (Fig. 2D).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Mis-expression of E-cadherin in the embryonic mesoderm. (A–C) Cross sections through the ventral furrow in stage-6 control (A) and E-cadherin overexpressing (B,C) embryos stained for E-cadherin (E-cad, red) and Neurotactin (Nrt, green, to visualise cell membranes), and with DAPI (blue). Overexpressed E-cadherin is seen both at its natural subcellular location at the adherens junctions and ectopically, in the cytoplasm near the nucleus (arrowhead in B,C). (D) Single-pixel fluorescence intensity measurements of E-cadherin and Armadillo along the lateral membranes (5 µm from apical to basal) in control (twi-G4, grey) and E-cadherin-overexpressing embryos (twi-G4>>cadh5, black). (E,F) Detail showing the mesodermal cell layer apposed to the ectoderm in stage-10 control embryos (E) and in embryos overexpressing E-cadherin in the mesoderm (F), stained for N-cadherin (green) and E-cadherin (red) to reveal their subcellular localisation. (G) The fluorescence intensity of E-cadherin and N-cadherin on the mesodermal membrane traced from the end closest to the ectoderm (exterior) to the end at the yolk surface (interior). Scale bars: 10 µm. Error bars represent s.e.m.

In embryos expressing E-cadherin under the control of twist-Gal4, the overexpressed E-cadherin becomes detectable at the time of ventral furrow formation (Fig. 2A–C). In addition to a raised level along the lateral membranes, an atypical accumulation of E-cadherin is seen in the perinuclear region (Fig. 2B′,C′, arrowhead), probably representing protein that is trapped in the secretory pathway (Pacquelet and Rørth, 2005), suggesting that there is a limit on the amount of protein that can be transported to the plasma membrane. Higher levels of membrane-associated E-cadherin are also seen by the time the mesoderm has invaginated. At this point, when a significant difference in E-cadherin expression between the ectoderm and the mesoderm becomes evident in wild-type embryos because of the transcriptional upregulation of E-cadherin in the ectoderm, the mesoderm retains the early low level of expression. A strong signal is seen at the apical adherens junction and along the lateral cell membranes. The ectopic expression is also detectable in the cells neighbouring the ventral furrow that do not invaginate (i.e. the mesectoderm and the adjacent ventral ectoderm), because the early Twist expression domain extends into this area. Like maternally provided E-cadherin in the wild-type embryo (Clark et al., 2011), the ectopically expressed E-cadherin (Fig. 2F′) persists after the invaginated mesoderm has dispersed into individual cells. In addition, unlike endogenous E-cadherin, which is localised mainly at the interface between mesodermal and ectodermal cells and at the mesodermal cell membrane that faces the yolk (Fig. 2E′), ectopic E-cadherin is localised along the entire mesodermal cell membrane once the mesodermal cell layer has become established (Fig. 2F′).

In wild-type mesodermal cells, some N-cadherin localises to the plasma membrane, but it is more abundant in a punctate pattern in the cytoplasm (Fig. 2E). When N-cadherin is overexpressed throughout the blastoderm, it is also mostly seen in the cytoplasm in ectodermal cells and does not colocalise with E-cadherin (supplementary material Fig. S1A). In embryos expressing high levels of E-cadherin in the mesoderm, the membrane localisation of N-cadherin seemed to be further reduced (Fig. 2G), whereas the punctate cytoplasmic pattern was unaffected or slightly enhanced (supplementary material Fig. S1C). Thus, the plasma membrane localisation of cadherins might be subject to tight regulation, and it seems that E-cadherin is more efficiently localised to the plasma membrane than is N-cadherin.

The absence of defects during gastrulation shows that high levels of E-cadherin in the mesoderm do not affect cell segregation between mesoderm and ectoderm, and that the transcriptional downregulation of E-cadherin is not a prerequisite for the mesoderm to lose its epithelial state or for the creation of a single-cell layer. This is unexpected, given the widespread conservation of the regulation of EMT (Nieto, 2011), and indicates that a different mechanism or a combination of mechanisms must contribute to the mesodermal EMT.

A redundant role for E-cadherin downregulation during mesodermal EMT?

The FGF signalling pathway is necessary for mesoderm spreading and migration (Michelson et al., 1998; Vincent et al., 1998; Wilson et al., 2005). In the absence of FGF, which is expressed in the ectoderm (Gryzik and Müller, 2004; Stathopoulos et al., 2004), the mesodermal cells fail to extend filopodia in the direction of the source of FGF (Clark et al., 2011). As soon as the mesoderm has invaginated, cell division occurs, with the rounding up of cells contributing to the dispersal of the mesoderm and the dissolution of adherens junctions. When both mitosis and FGF signalling are blocked (in dof, stg double mutants), E-cadherin retains its junctional localisation in the mesoderm for a longer time than in the wild type, and the epithelial structure of the invaginated tube also persists for longer. Nonetheless, the mesodermal cells eventually spread out on the ectoderm (Clark et al., 2011). Thus, neither mitosis nor FGF signalling nor transcriptional downregulation of E-cadherin seem to be necessary to allow the mesodermal EMT.

To test whether the three processes act redundantly to facilitate mesoderm dispersal, we overexpressed E-cadherin in the mesoderm of dof, stg double-mutant embryos. We observed no further exacerbation of the mesodermal defects at early stages (Fig. 3B,C). Despite the abnormally high levels of E-cadherin in such embryos, the mesodermal cell layer nevertheless flattened out on the underlying ectoderm. However, although the transition from an epithelial tube to a cell layer occurred both in dof, stg and in dof, stg, twi-Gal4>>E-cadherin embryos, the invaginated epithelial tube retained its epithelial structure and round cross-section for longer in the mutants than in the wild type (Fig. 3A–C; see also Clark et al., 2011), and the establishment of contact between ectoderm and mesoderm was also delayed (supplementary material Fig. S2C–E). The cells on the ventral side of the invaginated tube eventually established a close apposition to the ectoderm, but their shapes were abnormal. The cell mass subsequently lost its round structure and flattened against the ectoderm, spreading sideways, but, apart from the reduced number of cells, this stage was similar in appearance to wild-type embryos (Fig. 3F). The cells eventually covered most of the ectoderm in many cases, but the more dorsally located cells failed to reach the ectoderm, so that two layers of mesodermal cells persisted, and only in rare cases was a single layer established. In many embryos, the inner dorsal cell layer seemed to disintegrate (supplementary material Fig. S2H–K), suggesting that the ectoderm might also provide a trophic signal.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Overexpression of E-cadherin in a dof, stg double mutant. (A–C) Cross sections of control (twi-Gal4) and dof, stg double-mutant embryos either alone or in combination with E-cadherin overexpression in the mesoderm (twi-Gal4»cadh5, dof, stg). Embryos are stained for E-cadherin (E-cad) and Neurotactin (Nrt), as indicated, showing the mesodermal cells after completion of epithelial invagination. (D–F) In control (D), dof, stg mutant (E) and dof, stg mutant embryos overexpressing E-cadherin (F), the mesoderm flattens out on the ectoderm.

In summary, even when a high level of E-cadherin is maintained, and FGF signalling and mitosis are blocked, the mesodermal cells are able to spread out and form a new cell layer. Although indirect effects of Snail on E-cadherin localisation and dynamics are not addressed by these experiments, the results show that Snail-mediated transcriptional repression of E-cadherin is not essential for the mesodermal EMT. If it is important at all for E-cadherin to be downregulated, it is more likely that the need lies elsewhere or that downregulation is accomplished post-transcriptionally.

Effect of high levels of E-cadherin on mesoderm differentiation

We did detect abnormalities in embryos overexpressing E-cadherin at later stages. At the extended germband stage, the pattern of expression of Eve in mesodermal derivatives was altered. At this stage, Eve is expressed both in the segmental neuronal precursors and in clusters of cells in the dorsal mesoderm that contain the precursors for pericardial cells and the dorsal muscle DA1. The neuronal pattern of Eve expression was normal in embryos expressing high levels of E-cadherin in the mesoderm, but the staining for the mesodermal clusters was absent or drastically reduced (Fig. 4A,B; supplementary material Fig. S3A–D). The Eve-positive cell clusters arise in a dorsally located band of the mesoderm that receives differentiation signals from the overlying ectoderm (Lee and Frasch, 2000; Lockwood and Bodmer, 2002), and that also contains precursors for other mesodermal derivatives, such as the somatic dorsal muscles. When we stained older embryos of the same genotype for Mef2, a marker of these cells, we found that they were also reduced or absent (Fig. 4C,D). Thus, the dorsal mesoderm apparently fails to differentiate properly in embryos overexpressing E-cadherin. This could be due to either the failure of the mesodermal cell layer to spread sufficiently far to reach the region in which it can receive the appropriate signals, or its inability to receive or respond to a differentiation signal. We excluded a third possibility, namely, that cells were unresponsive because they were dying, by staining the embryos with antibodies against activated caspase3, which did not reveal abnormal cell death (data not shown).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Mesodermal differentiation in embryos mis-expressing E-cadherin. Control (A,C,E) and E-cadherin-overexpressing embryos (in the mesoderm, B,D,F) stained for Eve (stage 11; A,B) to mark pericardial-cell clusters (PC), stained for Mef2 (stage 16; C,D) to mark somatic, visceral and cardiac muscle nuclei or assayed for bagpipe (bap) expression (E,F) to mark the visceral mesoderm (E,F).

To investigate the first hypothesis, we focused on the region in which the Eve-positive cells should arise, to see whether the mesodermal cell layer was present or absent in these regions. Fig. 5 shows an embryo in which only one Eve-positive pericardial cell cluster is detectable (Fig. 5A). We analysed optical sections at a level just below the ectoderm. The sections reveal the mesoderm underneath the ectoderm from the ventral (top) to the dorsal (bottom) region in each of the selected segments (Fig. 5C–E). It is evident that the mesoderm in each segment extends equally far dorsally, and, in particular, that in the two segments lacking Eve-staining, mesodermal cells are present in the region in which the Eve-positive cells should arise. Z-sections confirm the presence of two cell layers at the positions indicated in Fig. 5C–E. Thus, the mesodermal cell layer in these embryos has extended as far dorsally as it does in normal embryos, and mesoderm is present in the region where it can receive ectodermal differentiation signals. The defect must therefore be an inability to receive or interpret the signals from the ectoderm. An additional morphological defect in older embryos (supplementary material Fig. S3E–G) might be caused by mesectodermal cells with abnormally high levels of E-cadherin failing to internalise.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Mesodermal spreading in embryos overexpressing E-cadherin in the mesoderm. Embryos expressing Cadh5/9 in the mesoderm using the twist-Gal4 driver stained for Even-skipped (Eve, green) and Armadillo (red), and stained with DAPI (blue). (A) Overview of a whole stage-11 embryo. (B,B′) Higher magnification of three hemisegments in the boxed region in A. The regions selected for higher magnification analysis are outlined in B. (C–E) Single optical planes from the imaged stack of the region shown in B. The focal plane lies within the ectoderm at the top of the image and in the mesoderm at the bottom of the image. The white lines mark the positions at which the yz- and xz-sections in C′–E′ and C″–E″ were taken. (C′–E′) The white dashed line highlights the mesoderm. Scale bars: 10 µm.

High levels of E- but not N-cadherin perturb Wingless-dependent mesoderm differentiation

Overexpression of E-cadherin in the ectoderm has been shown to interfere with the ability of the cells to respond to the Wingless signal (Sanson et al., 1996). The excess of E-cadherin is thought to bind to and reduce the pool of β-catenin (Armadillo) that is available to transmit the Wingless signal to the nucleus. We tested whether high levels of E-cadherin in the mesoderm affected the distribution of Armadillo or other proteins associated with E-cadherin (Fig. 6A,B,E,F; supplementary material Fig. S4). In cells overexpressing E-cadherin, we found that the level of Armadillo at the plasma membrane was strongly increased (Fig. 6D). The Armadillo-interacting protein α-catenin also showed increased levels (supplementary material Fig. S4A). Another E-cadherin-associated protein, p120-catenin, which is barely detectable or absent in wild-type mesodermal cells, showed no increase after overexpression of E-cadherin (supplementary material Fig. S4B). This suggests that overexpression of E-cadherin interferes with the normal participation of Armadillo in Wingless signalling, and with the differentiation of Wingless-dependent mesodermal cell types (Baylies et al., 1995; Wu et al., 1995). To further test the notion that E-cadherin interferes with mesoderm differentiation because of its effect on Armadillo signalling, rather than its function in adhesion, we expressed a form of E-cadherin that can mediate adhesion, but that does not have an Armadillo-binding site in its intracellular domain. This construct, E-cadΔβ/αCat, can mediate all E-cadherin and β-catenin functions at adherens junctions, including cell sorting, epithelial integrity and cell migration (Pacquelet and Rørth, 2005), suggesting that the sole function of the β-catenin-binding domain at cell junctions is to facilitate the interaction with α-catenin. In contrast to full-length E-cadherin, overexpression of E-cadΔβ/αCat in the mesoderm had no effect on the differentiation of the pericardial cells, or any other aspect of mesoderm development, thus identifying the Armadillo-binding domain as a key mediator of the effects of E-cadherin on Wingless signalling (Fig. 6L).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

Distribution of β-catenin (Armadillo) in mesodermal cells overexpressing E-cadherin. Sections of late-stage-10 control (A,E) and E-cadherin-overexpressing embryos (B,F) stained for E-cadherin (E-cad, green) and Armadillo (Arm, red) as indicated. Dashed boxes in A′ and B′ indicate the areas that are enlarged in E and F, respectively. (C,D) Section of a late-stage-10 control (C) and an N-cadherin-overexpressing embryo (D) stained for N-cadherin (N-cad, grey). (C) The fluorescence intensity was adjusted in Photoshop so that the endogenous N-cad signal is visible. (G–I) Staining for Armadillo (red) in mesodermal cells overexpressing E-cadhΔB/αcat (G), N-cad (H) or E-cadhΔCyt/N-cadh (I). Scale bar: 10 µm. (J,K) Stage-11 embryos expressing UAS-Ncadh (J) or UAS-Cadh5/9 (K) under the control of nullo-Gal4, stained for Engrailed (En, green). (L) Quantitation of Eve-positive cell clusters in stage-11 embryos of the indicated genotypes. Overexpression of the full-length cadherin transgenes (cadh5, cadh 5/9) and dominant-negative Tcf produced significant reductions (P<0.001) in the number of clusters; n≥10 embryos in each case. Error bars represent s.e.m.

If the loss of Eve-positive heart precursor cells is due to the inability of the dorsal mesoderm to respond to Wingless, one might expect the cell population to be transformed to the alternative Wingless-independent mesodermal fate. The heart precursors arise in the sloppy-paired domain (which includes the wingless expression domain) of the dorsal mesoderm; the cells in the neighbouring domain differentiate into visceral mesoderm, which is characterised by expression of the transcription factor Bagpipe (Bap). The expression domain of bap increases when wingless signalling is reduced (Riechmann et al., 1997). Consistent with this, overexpressing E-cadherin in the mesoderm resulted in a substantial expansion of the bap expression domain in the dorsal mesoderm (Fig. 4E,F). Thus, high levels of E-cadherin do lead to the same fate transformation as the loss of Wingless.

If the reason for downregulation of E-cadherin by Snail in the mesoderm is to allow efficient Wingless signalling, then N-cadherin might be expected not to have negative effects on Wingless signalling. We first tested this in the ectoderm, where E-cadherin has previously been shown to be able to abrogate Wingless signalling (Sanson et al., 1996). Expression of E-cadherin using the nullo-Gal4 driver led to the stabilisation of Armadillo at cell membranes, and efficiently downregulated the expression of engrailed (Fig. 6K), a Wingless target. By contrast, expression of N-cadherin using the same driver had no effect on engrailed expression (Fig. 6J). Thus, high levels of E-cadherin but not N-cadherin interfere with Wingless signalling in the ectoderm.

Similarly, the overexpression of N-cadherin in the mesoderm had no effect on the differentiation of any of the dorsal mesodermal derivatives (Fig. 6L). Its overexpression led to increased recruitment of Armadillo to the puncta in the cytoplasm, rather than an enrichment at the plasma membranes (Fig. 6H). Thus, the mesoderm can tolerate increased levels of N-cadherin expression but not increased levels of E-cadherin. This provides the first experimental demonstration that N-cadherin differs from E-cadherin in its ability to modulate Wingless signalling, a property that might rely on the different affinities of N- and E-cadherin for Armadillo that have been demonstrated biochemically in the nervous system (Loureiro and Peifer, 1998). Further support for this conclusion comes from the finding that it is specifically the intracellular domain of E-cadherin that is responsible for the interference with Wingless signalling. Overexpression of a chimeric construct in which the extracellular domain of E-cadherin is fused to the cytoplasmic domain of N-cadherin (E-cadhΔCyt/N-cadh) had no effect on mesoderm differentiation (Fig. 6L).

The physiological effects of expressing the various cadherins could either be due to biochemical differences in their actions, or to the fact that they are expressed at different levels. A comparison of the levels of overexpression that were achieved by the various constructs, by using an antibody against the extracellular domain of E-cadherin, revealed that cadh5, cadh5/9 and E-cadhΔCyt/N-cadh were expressed at similar levels. E-cadΔβ/αCat was detectable at lower levels but was, nonetheless, sufficient to sustain the movement of border cells in ovaries (Pacquelet and Rørth, 2005). The levels of overexpressed N-cadherin also significantly exceeded the endogenous level (Fig. 6C,D), and, like E-cadherin overexpression, overexpression of N-cadherin caused Armadillo recruitment and stabilisation (Fig. 6H).

The difference in the effect of E-cadherin and N-cadherin could be due to either the difference in their subcellular localisation (junctional vs cytoplasmic), or a biochemical difference in the intracellular domain. The chimeric protein E-cadhΔCyt/N-cadh localised primarily at the plasma membrane and was able to stabilise Armadillo at the junctions (Fig. 6I). This demonstrates that the increased recruitment of Armadillo to the plasma membrane is not sufficient to cause suppression of Wingless signalling. Instead, N-cadherin might recruit a different sub-pool of Armadillo as shown previously in a different context (Loureiro and Peifer, 1998).

Curiously, in apparent contradiction to this conclusion, the effect of overexpressing E-cadherin in the mesoderm was weakened if N-cadherin levels were lowered by half or if it was removed completely (Fig. 7C–F). In other words, although N-cadherin competes less efficiently for Armadillo when overexpressed, its reduction counteracts the sequestering of Armadillo by high levels of E-cadherin in the mesoderm (Fig. 7G,H). We discuss possible explanations for this paradox below, but opposite effects of N-cadherin and E-cadherin have also been documented in the context of eye morphogenesis (Mirkovic and Mlodzik, 2006).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 7.

Overexpression of E-cadherin in CadN-deficient embryos. Stage-11 embryos of the indicated genotypes were stained for Even-skipped (Eve, blue) and N-cadherin (N-cad, brown) (A–F), and E-cadherin (E-cad, green) and Armadillo (Arm, red) (G,H) to visualise the effects on the pericardial-cell clusters and on Armadillo localisation, respectively. Scale bars: 10 µm.

In summary, the complementarity of cadherin expression in the ectoderm and mesoderm is not needed for early mesoderm morphogenesis. Instead, the transcriptional downregulation of E-cadherin by Snail might be required because of the potential for high levels of E-cadherin to interfere with cell differentiation in the mesoderm.

On the requirement of the cadherin switch

We confirm that there is no requirement for either zygotic E-cadherin or N-cadherin for the formation of the ventral furrow. These results differ from observations in other models, including chicken (Nandadasa et al., 2009), and from previous suggestions that the switch is necessary in the fly (Oda et al., 1998). In both flies and mice, the early effects of the loss of E-cadherin function [maternal E-cadherin in flies and zygotic in mice (Larue et al., 1994; Riethmacher et al., 1995; Tepass et al., 1996)] preclude an analysis of gastrulation in the complete absence of E-cadherin.

Drosophila gastrulation and the EMT

One motivation for the investigations presented here was the resemblance of Drosophila gastrulation to the EMT, and the observation of cadherin switching in other EMT settings. As classically defined, an EMT is associated with the loss of cell contacts and the consequent dissociation of cells from the sheet, a process mediated by the action of the ‘EMT inducers’, Snail, Zeb and Twist, and which includes transcriptional repression of E-cadherin as a key molecular signature. Cancer metastasis and the formation of the neural crest are thought to represent conventional EMTs by these criteria (Nieto, 2009; Nieto, 2011). Although the loss of cell contacts mediated by the downregulation of E-cadherin is one requirement for an EMT, others, including cytoskeletal reorganisation and matrix remodelling to facilitate a change in cell shape, motility and migration are also necessary. Another contribution to motility during an EMT is the cadherin switch. N-cadherin is expressed primarily by mesenchymal and neural tissues, and promotes motility (Iwai et al., 1997; Li et al., 2001; Nandadasa et al., 2009). Its upregulation is a feature of some, but not all, EMTs (Nieto, 2011), and might result in the modulation of adhesion strength that is necessary for motility.

Drosophila gastrulation converts a stable epithelium into a more mesenchymal tissue. This relies on several functions of the transcription factor Snail, evident from the fact that the loss of E-cadherin is insufficient to rescue the effects of a Snail mutant (our unpublished observations). Unlike a classical EMT, however, single cells do not delaminate from the sheet during Drosophila gastrulation (Thiery et al., 2009), and the resulting mesoderm remains as a layer. Whereas cell shape changes that are driven by apical constriction require the tethering of the actin cytoskeleton to adherens junctions (Dawes-Hoang et al., 2005), the changes that accompany spreading, dispersal and migration of the mesoderm are less well understood, but are thought to rely on the loosening of cell contacts for the acquisition of motility. In neuronal growth cones, N-cadherin interacts, through its extracellular domain, with the FGF receptor (Cavallaro and Christofori, 2004; Williams et al., 2001). The latter is also a key regulator of mesoderm migration in Drosophila. Whereas FGF signalling is downstream of the transcription factors Twist and Snail in the Drosophila mesoderm, it is upstream of Snail in vertebrate gastrulation and in the regulation of EMT (Ciruna and Rossant, 2001), and seems to function in a manner analogous to Twist in its requirement for the maintenance of snail expression. Despite these intriguing parallels, we found that the complete loss of N-cadherin has no effect on invagination.

Another possibility is that the switch is a redundant or permissive mechanism that facilitates other processes that allow the mesodermal cells to disperse and form the mesodermal cell layer. FGF receptor signalling in the mesoderm is needed for the efficient and even distribution of the mesodermal cells over the ectoderm (Bae et al., 2012; Wilson et al., 2005). FGF signalling mediates the close apposition of the basal side of the newly invaginated mesodermal epithelium against the ectoderm, and, subsequently, the filopodia-mediated motility of the cells away from the site of invagination. At the same time, the mesodermal cells undergo a wave of mitoses, which might help the dispersal of the cells (Clark et al., 2011). However, as shown previously, inhibiting mitosis and FGF signalling does not prevent the formation of a mesodermal cell layer (Clark et al., 2011). Here, we show that a cell layer is still formed even when the level of E-cadherin is increased in addition to the inhibition of mitosis and FGF signalling.

Our results suggest that gastrulation can be thought of as a collective EMT. The mesoderm collectively loses its epithelial character and polarity, and acquires motility that is directional and involves adhesion to the substrate. This might necessitate that the cells maintain some contact with one another, rather than lose all contact. Consistent with this, E-cadherin in the mesoderm is not completely lost from the mesodermal cells. Although the zygotic transcription of E-cadherin is upregulated in the ectoderm and this upregulation is repressed in the mesoderm, the maternally provided E-cadherin persists in the mesoderm (Clark et al., 2011; this study). This might also explain the lack of strong effects of cadherin overexpression on gastrulation.

An alternative possibility is that EMT in the mesoderm is accomplished by mechanisms other than the transcriptional downregulation of E-cadherin expression. Indeed, in addition to changes in the levels of E-cadherin, changes in localisation of E-cadherin also accompany mesoderm morphogenesis. The transcription factor Snail is required for an early change in the location of the adherens junctions from a subapical to an apical position in the invaginating epithelium, although a low level of protein is distributed over a broader domain in the lateral cell membranes (Kölsch et al., 2007; Mathew et al., 2011). The adherens junctions are subsequently dismantled after the invagination is completed, cells lose their apico-basal polarity, and E-cadherin is detected over the whole cell surface. As the mesodermal cell layer is established, E-cadherin becomes concentrated at the interface with the ectoderm and on the cell surface facing the yolk. Experimentally overexpressed E-cadherin shows a broadly similar behaviour, with some additional staining in the perinuclear region, presumably as a result of saturating the secretory system or the adherens junctions. These findings suggest that changes in cell adhesion mediated by mechanisms other than the transcriptional downregulation of E-cadherin might be responsible for the post-invagination behaviour of the mesoderm. They further strengthen the idea that multiple events occurring downstream of Snail contribute to the EMT.

Mechanisms that do not rely on Snail or on the transcriptional repression of E-cadherin can also contribute to EMT. For example, the EMT in the Drosophila posterior midgut is driven by Serpent-mediated transcriptional repression of the apical membrane regulator Crumbs, which in turn influences the subcellular localisation of E-cadherin (Campbell et al., 2011).

Although EMT, as classically defined, necessitates the loss of polarity and the transcriptional downregulation of E-cadherin by the EMT transcription factors (Snail, Twist and Zeb), a precise understanding of the full range of mechanisms that are involved is still missing. Analysis of gene regulatory networks in the sea urchin also suggests that multiple networks accomplish distinct features of the EMT (Wu and McClay, 2007), and Snail might mediate only a subset of these.

Cadherins and mesoderm differentiation

The most striking phenotype observed in the mesoderm upon overexpression of full-length E-cadherin is the failure of the pericardial cells to differentiate, and an expansion of visceral mesoderm precursors. The overexpression of N-cadherin had no effect, highlighting qualitative differences between the two cadherins in their ability to influence differentiation. Mesoderm differentiation relies on combinatorial signalling (Bate, 1990; Baylies et al., 1998). PC-cell specification depends on Wingless signalling (Bate and Rushton, 1993), which can be modulated by adhesion through Armadillo and can influence adhesion qualitatively and quantitatively (Jaiswal et al., 2006; Sanson et al., 1996; Ulrich et al., 2005). The overexpression of E-cadherin abrogates Wingless signalling in the Drosophila wing, through sequestration of Armadillo (Sanson et al., 1996). Whereas overexpressed full-length E-cadherin effectively recruited Armadillo to the membrane in the mesoderm and ectoderm, N-cadherin overexpression was less efficient at doing so. Instead, overexpression of N-cadherin resulted in an observable increase in the cytoplasmic pool of Armadillo.

Three observations are intriguing: (1) the inability of overexpressed N-cadherin to influence Wingless signalling, (2) the fact that the concomitant removal of N-cadherin can rescue the wingless phenotype produced by E-cadherin overexpression and (3) the insufficiency of junctional Armadillo recruitment as an explanation for the wingless phenotype. The first observation suggests that higher levels of N-cadherin are necessary to downregulate Wingless signalling or that N-cadherin is less efficient than E-cadherin at sequestering Armadillo. Both of these explanations might rely on the predominant subcellular localisation of both N-cadherin and the recruited Armadillo (cytosolic rather than membrane-associated). However, our experiments with the chimaeric cadherins reveal that the junctional localisation of overexpressed E-cadherin and the consequent junctional recruitment of Armadillo are not sufficient to account for the downregulation of Wingless signalling by E-cadherin. It is also possible that the binding affinities of Armadillo to E-cadherin and N-cadherin are different. A third possibility is that the effects of the two cadherins on Wingless signalling are driven by distinct mechanisms. N-cadherin has been reported to antagonise Wingless signalling through its interaction with the Wnt co-receptor LRP, which targets β-catenin for degradation (Haÿ et al., 2009). This mechanism requires the localisation of N-cadherin at the membrane, which in the Drosophila mesoderm is low. These results suggest that differences between E- and N-cadherin in their ability to recruit, bind and retain Armadillo – both qualitative differences (different pools or isoforms) and quantitative differences (different binding strengths) – might underlie their effects on Wingless signalling. Taken together, these possibilities also provide an explanation for why the loss of N-cadherin rescues the phenotype produced by the overexpression of E-cadherin: this might result from the reduced downregulation of Wingless signalling in the absence of N-cadherin. A distinct form of Armadillo binds to N-cadherin in the nervous system (Loureiro and Peifer, 1998). Whether the mesoderm also expresses different forms of Armadillo or other linker proteins remains unknown. In addition, the different biochemical properties of CadN and CadN2 that are suggested by differences in their length and sequence need to be examined experimentally.

The regulation of cadherins during morphogenesis

Although we have focused on the expression and localisation of E-cadherin and N-cadherin, many other post-transcriptional modes of regulation affect cell behaviours during morphogenesis. Regulation of localisation through the activity of the Rho1 GTPase accompanies apical constriction during the invagination of the ventral furrow [causing more apical localisation of adherens junctions (Dawes-Hoang et al., 2005; Kölsch et al., 2007)]. The formation of epithelial folds seems to rely solely on the progressively basal shift of the adherens junctions, driven by the polarity regulator Bazooka and Rap1 (Wang et al., 2012; Wang et al., 2013). Junction remodelling by localised endocytosis underlies cell intercalation during germband extension (Levayer et al., 2011). Differences in cell behaviour can also result from modulating the link with the cytoskeleton (Roh-Johnson et al., 2012). In addition, the effective increase in adhesion upon E-cadherin overexpression in the mesoderm might be modulated by the membrane trafficking pathways that either facilitate the degradation of E-cadherin or control its incorporation into the membrane secretory pathway (Langevin et al., 2005; Lohia et al., 2012). Indeed, the low levels of p120-catenin in the mesoderm might be part of a mechanism that ensures E-cadherin degradation. It is possible that one or more of these mechanisms also operate in the mesoderm and serve as additional controls that ensure that there is no net change in functional adhesion upon cadherin overexpression, explaining the lack of effects of overexpression of E-cadherin on early mesoderm morphogenesis. By contrast, signalling is more sensitive to small changes in the level of E-cadherin.

Conclusions

In summary, our results rule out qualitative differences in E- and N-cadherin as a mechanism for the segregation of germ layers. They also rule out a role for the transcriptional repression of E-cadherin by the transcription factor Snail in the spreading and migration of the mesoderm. Our results suggest that N-cadherin does not have a role in early mesoderm morphogenesis, and they uncover striking differences in the ability of E- and N-cadherin to influence Wingless signalling that might rely on differences in the interactions of these cadherins with Armadillo. The regulation of adhesion by other mechanisms might be responsible for allowing mesoderm cell dispersal. A detailed understanding of the related cell biology will be of great interest.