Snail1a and Snail1b cooperate in the anterior migration of the axial mesendoderm in the zebrafish embryo

The Snail family of zinc-finger transcription factors fulfils important roles in vertebrate development. Indeed, these genes are involved in a variety of processes that require cell movement and shape changes within the organism. Most significantly, they are required to induce the epithelial to mesenchymal transition (EMT) through which epithelial cells loose their cell-cell adhesion contacts, becoming migratory and invasive. EMT is fundamental for the formation of many tissues during vertebrate embryonic development, implicating Snail family members in these processes(Nieto, 2002). Moreover, as EMT also occurs during pathological processes, Snail genes have been implicated in tumour progression and renal fibrosis(Nieto, 2002; Thiery, 2002; Boutet et al., 2006).

In amniotes, EMT is fundamental for the formation of tissues as important as the mesoderm and the neural crest. However, no large-scale EMT has been observed during gastrulation in anamniote embryos where a complex interplay of different morphogenetic movements gives rise to a mass sheet-like migration and the cells maintain their contacts while moving. These morphogenetic movements can be described as: (1) epiboly towards the vegetal pole over the yolk cell; (2) internalisation of the mesendoderm at the margin; and (3)dorsal convergence and extension along the anteroposterior axis (C&E)(Keller et al., 2000; Solnica-Krezel, 2006). In contrast to the Xenopus gastrula, C&E movements in zebrafish seem to be independent of both epiboly and mesendoderm internalisation(Myers et al., 2002), and convergence can even be separated from extension(Myers et al., 2002; Glickman et al., 2003). Indeed, the interdependence of these two processes in Xenopus may in part be due to the strength of the adhesive forces that maintain cells together during involution and convergent-extension. However, during internalisation in fish, cells can be found in groups or even as individual cells (Carmany-Rampey and Schier,2001; Montero et al.,2005).

Although Snail genes have been implicated in the formation of the neural crest and the mesoderm in different vertebrate species (reviewed by Nieto, 2002), functional analyses in zebrafish have only been carried out on snail1a. Four Snail genes have been described in teleosts: snail1, snail2, snail3and slug (Thisse et al.,1993; Hammerschmidt and Nüsslein-Volhard, 1993; Thisse et al., 1995; Locascio et al., 2002; Manzanares et al., 2004). Phylogenetic analyses have shown that snail1 and snail2 are duplicates of the Snail1 gene that arose after gene duplication in the teleost lineage (Postlethwait et al.,1998; Manzanares et al.,2001). snail1 and snail2 have been renamed snail1a and 1b, respectively. Similarly, slug has been renamed snail2, and the most recently described family member is snail3 (Barrallo-Gimeno and Nieto,2005).

The patterns of snail1a and 1b expression in the neural crest suggest that as in other species, Snail genes are fundamental for the development of this tissue (Thisse et al.,1993; Hammerschmidt and Nüsslein-Volhard, 1993; Thisse et al., 1995). However,in the absence of massive EMT at gastrulation these genes were considered to be less important in convergence and extension(Locascio and Nieto, 2001). Nevertheless, snail1a has since been implicated in the anterior movement of the axial mesendoderm in the fish embryo(Yamashita et al., 2004) where elegant time-lapse studies demonstrated that individual cells delaminate from the embryonic shield (Montero et al.,2005). Here we show that snail1b is important for the movements that take place during extension, with snail1a and 1b fulfilling non-redundant functions in the movements that direct the migration of the axial mesendoderm.

Zebrafish were raised and kept at 28°C under standard conditions, and they were staged according to Kimmel et al.(Kimmel et al., 1995). AB and Tup-Lof wild-type strains were used in all experiments.

All morpholino antisense oligonucleotides were obtained from Gene Tools LLC and used as described by Nasevicius and Ekker(Nasevicius and Ekker, 2000). The antisense morpholinos (MOs) blocking snail1b translation hybridise to bases -28 to -4 (MO1) and -2 to +23 (MO2) with respect to the start codon, and the snail1a MOs are as previously described(Yamashita et al., 2004). Likewise, the cdh1 MOs were as described by Babb and Marrs(Babb and Marrs, 2004), and the standard control morpholino was used for control injections (Gene Tools, LLC). All oligonucleotides were injected into the yolk of one- or two-cell embryos(3-20 ng/embryo) using a pressure injector (PicoSpritzer).

Rescue and overexpression analyses were performed with snail1bmRNA that lacked the 5′UTR region to which MO1 binds, and that were synthesised and capped using the mMessage mMachine Kit (Ambion). Rescue was achieved by co-injecting 3 ng of morpholino plus 250 pg mRNA per embryo, and 150 pg of mRNA per embryo was injected for overexpression studies. Phenotypic analyses involved at least 50 embryos for each specific marker studied under each experimental condition.

In situ hybridisation of whole-mount zebrafish embryos was performed as described by Nieto et al. (Nieto et al.,1996), with minor modifications, and fluorescent whole-mount in situ hybridisation was carried out as described by Denkers et al.(Denkers et al., 2004). In these experiments, the following probes were used: snail1a(previously snail1) and snail1b [previously snail2(Thisse et al., 1993; Thisse et al., 1995)]; cdh1 [cadherin 1; also known as E-cadherin(Babb and Marrs, 2004)]; hgg1 (Vogel and Gerster,1997) [also known as ctsl1b (cathepsin L) -ZFIN]; spt (Griffin et al.,1998) [also known as tbx16 (T-box gene 16) -ZFIN]; myoD (Weinberg et al.,1996); papc (Yamamoto et al., 1998) [also known as pcdh8 (protocadherin 8) - ZFIN]; dlx3 (Akimenko et al., 1994); and ntl [no tail(Schulte-Merker et al.,1992)]. The zebrafish anti-Cdh1 antibody(Babb and Marrs, 2004) was used at a dilution of 1/500. Embryos were embedded in gelatin to obtain vibratome sections (50 μm).

Donor embryos were injected at the one- to two-cell stage with fluorescein-dextran (10,000 kDa; Molecular Probes), alone or together with the appropriate dose of the snail1b morpholino. At the shield stage,10-20 cells were transplanted into the shield of an unlabelled sibling host. Alternatively, donor embryos were also injected with mRNA encoding a membrane-tagged RFP. After an incubation of 1 hour and at approximately 70%epiboly, grafted embryos were mounted dorsal side up in 0.8% low-melting point agarose in Danieau's solution, and images were obtained on a Leica SP2 confocal microscope using a 40× water immersion objective. Embryos were allowed to develop to the tail-bud stage, when they were photographed on a Leica FLIII dissecting microscope. The distances that three individual cells moved in each experiment (wt→wt, n=7, mo→wt, n=5)were assessed using the Manual Tracking plug-in for ImageJ 1.37v software.

For time-lapse, multiple focal plane (4D) microscopy of zebrafish embryos,the embryos were co-injected with snail1b morpholino and 100 ng/μl of membrane-tagged-GFP mRNA [with myristoylation and palmitoylation sequences from the Lyn tyrosine kinase (Teruel et al., 1999), kindly provided by Reinhard Köster]. The embryos were then embedded at the appropriate stage in low gelling 1% agarose in Danieau buffer. Imaging was performed on a Leica SP2 confocal system mounted on an inverted microscope using a 10× objective, and z-stacks were obtained every 5 minutes and processed for maximum intensity projection using ImageJ software.

To analyse the function of snail1 genes during fish development,we have adopted an approach that involves the use of morpholino antisense oligonucleotides. We initially decided to study snail1b because of its prominent expression in the premigratory neural crest(Thisse et al., 1995). To disrupt snail1b activity, an oligonucleotide against the 5′UTR region (MO1) and another that overlapped the start codon of the snail1b cDNA (MO2; Fig. 1A) were designed. The vast majority of embryos that were injected with 1-1.5 ng of the morpholino survived, although at higher doses (3 ng) only 80% of the embryos reached the tail bud stage and only 60% survived for 24 hours after the injection. Unexpectedly, we found that injecting snail1b MOs induced an early phenotype at gastrulation stages and the shortening of the body axis at later stages(Fig. 1B-E,H,I). The phenotype we observed was somewhat surprising given the pattern of snail1bexpression in the mesoderm, where snail1a rather than snail1b is more prominently expressed(Fig. 1K-R). Indeed, the main sites of snail1a expression in the mesoderm at gastrulation stages are the margin, the adaxial cells (Fig. 1K,L) and the paraxial mesoderm(Fig. 1N) (see Thisse et al., 1993). By contrast, the main mesodermal sites for snail1b are the converging cephalic mesoderm (Fig. 1Q),the adaxial cells and a population of the axial mesendodermal cells(Fig. 1O,R)(Thisse et al., 1995).

In all cases, the phenotypes were highly penetrant with more than 90% of the living embryos developing a similar phenotype at each concentration of morpholino. No defects were observed in uninjected embryos, in embryos injected with buffer alone (Danieau), or in those embryos injected with a negative control MO.

The injection of either morpholino alone (MO1 or MO2) induced similar defects and with a similar dose response. Since MO1 was used in all our control experiments (see also Fig. S1 in the supplementary material), it was used in most of the experiments shown at a dose of 3 ng per embryo unless otherwise indicated. To confirm the specificity of the phenotype, we co-injected MO1 along with the coding region of snail1b mRNA. Significantly, both the phenotypic and lethality effects observed after MO1 injection could be rescued in up to 80% of embryos (n>100; Fig. 1F,G,J), indicating that the shortening of the body axis was induced specifically by the snail1b MOs. In order to assess whether we were dealing with morphogenetic defects or with a failure to specify the mesoderm, we assessed the expression of axial [no tail(Schulte-Merker et al., 1992)]and paraxial [paraxial protocaherin(Yamamoto et al., 1998)]mesoderm markers at gastrulation stages. Both these markers were detected in the embryos, indicating that fate specification was not significantly affected. In fact, somites still formed in the axial mesoderm, although they had an abnormal shape, and evidence of the compression was detected by myoD expression at segmentation stages(Fig. 2). Together, these data indicated that snail1b MOs induced a severe impairment in anteroposterior extension without affecting mesodermal fate.

Since the anterior migration of the axial mesendoderm is an important component of the extension movements and snail1b is expressed in this territory, we analysed the alterations in the snail1b MO-injected embryos using the prechordal plate markers, hatching gland gene 1[hgg1 (Vogel and Gerster,1997)] and spadetail [spt(Griffin et al., 1998)]. Injection of snail1b MO expanded the expression domains of both these markers, extending spt expression more than hgg1(Fig. 3A-C,F-H). By contrast,ectopic overexpression of snail1b mRNA induced the anterior compression of the hgg1-domain and altered the distribution of the spt-positive cells but not the number of expressing cells(Fig. 3K-M and data not shown). These two complementary phenotypes are compatible with snail1b being involved in the regulation of cell-cell adhesion during anterior migration. Snail is a strong repressor of cdh1 (E-cadherin)(Cano et al., 2000; Batlle et al., 2000) and as expected, injection of the snail1b MO enlarged the domain of cdh1 expression in the axial mesendoderm(Fig. 3J), as well as the hgg1 expression domain. Furthermore, less cdh1 was detected throughout the embryos than overexpressed snail1b(Fig. 3N,O). Snail1b did indeed appear to be acting as a repressor of cadherin in these embryos, as the morphant phenotype could be ameliorated by reducing cadherin function with the injection of cdh1 MOs (Fig. 3P-S). For these rescue experiments, embryos were injected with the usual dose of snail1b MO and a low dose of cdh1 MO (500 pg) to avoid the severe phenotype of strong cdh1 knockdown(Babb and Marrs, 2004), and thus, generating a partial loss of Cdh1 function. The prechordal plate adopts a normal shape in embryos injected with both snail1b and cdh1 MOs (3 ng and 500 pg, respectively), and its shape in cadherin morphants is similar to that observed in embryos with Snail1b gain of function(compare Fig. 3V with O). Cells still disaggregate in the posterior region of double morphants as observed in the cdh1 MO-injected embryos (asterisks in Fig. 3R,S, and not shown). It therefore appears that increased cell adhesion most probably accounts for the defects in extension movements in snail1b morphants.

Adhesion appeared to be the main defect underlying the aberrant movement of the prechordal plate when snail1b function is compromised. However, we were puzzled by the fact that snail1b is not expressed in the prechordal plate but rather, in a population of axial mesendoderm that seems to lie just posterior to it (Thisse et al.,1995) (Fig. 1R). Moreover, as a transcription factor, snail1b should act in a cell-autonomous manner. To better define the territory of snail1bexpression, we contrasted the distribution of hgg1 and confirmed that these two markers identified independent cell populations that partially intermingle at their interface as they migrate during gastrulation(Fig. 4A-C). By the tailbud stage, cells that express hgg1 and snail1b were completely separated such that a sharp boundary had formed(Fig. 4D). We also compared the distribution of the posterior axial mesodermal marker no tail(ntl) with the snail1b-expressing population and we were unable to identify double-labelled cells. As such, the posterior limit of snail1b expression abuts with the anterior limit of ntlexpression (see Fig. S2 in the supplementary material).

Our data indicate that the expression of snail1b in the axial mesendoderm is restricted to a segment that lies posterior to the hgg1-expressing cells and anterior to those expressing ntl. Accordingly, these cells occupy a gap between the hgg1 and ntl expression domains that can be clearly observed at the tailbud stages. This expression data indicated that snail1b is excluded from the precursor cells of the hatching gland where it seems to induce a cell non-autonomous phenotype, as hgg1-positive cells lag behind in snail1b morphant embryos (Fig. 3H). Conversely, the hgg1-expressing cells accumulate significant levels of Cdh1 (Fig. 4E-G), further evidence that they lack snail1b. To analyse whether snail1b MOs induce an additional cell autonomous effect, we grafted 10-20 wild-type or snail1b morphant shield cells to the shield of wild-type hosts that were left to develop to the tailbud stage. The wild-type grafted cells were found in the prechordal plate, whereas the snail1b morphant cells remained at much more posterior locations(Fig. 5B). Indeed, when we analysed the behaviour of the individual cells, the abundant lamellipodia that projected anteriorly in wild-type grafted cells were not observed in the snail1b morphant cells, which often migrated in compact groups(Fig. 5C,D and Movies 1 and 2 in the supplementary material) and more slowly(Fig. 5E). Thus, as might be expected for a transcription factor, Snail1b also fulfils a cell autonomous function favouring cell migration.

Snail1a has been implicated in anterior movement of the axial mesendoderm in the zebrafish embryo, since the prechordal plate in morphant embryos was located at a more posterior position(Yamashita et al., 2004). However, snail1a is not expressed in the prechordal plate(Thisse et al., 1993), again suggesting that, like snail1b, this transcription factor seems to act in a cell non-autonomous manner to direct the migration of the prechordal plate cells. Accordingly, we carefully analysed snail1a expression sites in the gastrulating fish, identifying a previously neglected domain of extraembryonic snail1a expression in the yolk syncytial layer YSL located underneath and anterior to the prechordal plate cells at 70% epiboly(compare Fig. 6A with B, in territories anterior to the dotted line encircling the hgg1-positive region, and Fig. 6D). This expression domain can be better assessed in sections(Fig. 6D). Higher levels of snail1a expression surrounding the hgg1-expressing cells are already evident at 60% epiboly, where the previously described high levels of expression in the margin are also observed(Fig. 6E,F).

When the expression patterns of the two snail1 genes were compared with that of hgg1 (Fig. 6), their distribution was mutually exclusive, as they are with respect to cdh1 (Figs 4 and 6). Indeed, as shown above, hgg1-expressing cells also express cdh1(Fig. 4E), and both gene transcripts are excluded from snail1a or snail1b expressing cells (Fig. 4 and Fig. 6G,H). Thus, a dynamic and complementary pattern of snail1 and Cdh1 expression divides the embryo into territories with different intercellular adhesion properties(summarised in Fig. 6I with data from Figs 2, 4 and 6).

Since snail1a and 1b are expressed in independent populations and they generate a distinct phenotype when disturbed individually, we wondered whether they cooperate to promote extension movements during fish gastrulation. We co-injected snail1a and snail1b MOs and found that the defects in the axial mesendoderm were compounded (Fig. 7D; n>20). Hence, at the end of gastrulation the double morphants show the defects produced in both single morphants(Fig. 7B-D). The anterior limit to axial mesendoderm migration in these morphant embryos was more posterior than in control embryos [as in snail1a morphants in Yamashita et al.(Yamashita et al., 2004) and Fig. 7B,J; n>20],and they developed an abnormally elongated prechordal plate, similar to the snail1b morphants (Fig. 3 and Fig. 7C,K; n>100). Significantly, the cells expressing snail1b in the snail1a morphants migrate normally and often continue migrating on top of the hgg1-expressing cells(Fig. 7F).

The cells that express snail1a located anterior to the prechordal plate are unaffected when the function of snail1b is compromised,explaining how the anterior limit of the prechordal plate is established correctly (Fig. 7C). However,the polster is formed before epiboly has been completed (see Fig. S3 in the supplementary material) and when the morphant embryo reaches the tailbud stage, the prechordal plate has already adopted an abnormal shape(Fig. 2H and Fig. 7C). Since the snail1b MO used does not induce mRNA degradation, we can follow the cells in which Snail1b function has been knockdown since they still express its mRNA (Fig. 7C,G). In these embryos, time-lapse studies of the movements of the cells containing snail1b transcripts (the translation of which is impaired by the morpholinos) indicated that their anterior movement was much slower than in control embryos (not shown). When we observed the movements of the prechordal plate cells, we found that the anterior cells move forward normally while those at its posterior edge appeared to struggle to advance and lagged behind(for an example, see Movies 3 and 4 in the supplementary material). At the tail bud stages the lagging hgg1 cells were still intermingled with cells that contained snail1b transcripts(Fig. 7G), explaining the aberrant elongated shape of the prechordal plate(Fig. 2H and Fig. 7C,K). We did not find any double-labelled cells, indicating that snail1b knockdown is not accompanied by a change in cell fate towards the hgg1-expressing prechordal plate cells. Nevertheless, the enlargement of the area occupied by the prechordal plate (hgg1-expressing cells) in embryos injected with snail1b MOs could be due to increased proliferation. This does not seem to be the case because we were unable to detect a significant change in the total number of hgg1-positive cells in wild-type, snail1a,snail1b and snail1 double morphants (n=3 for each condition) when their nuclei were counted(Fig. 7M-P). In addition, even in the most compact areas, the number of nuclei per field in the hgg1-positive territory in snail1b or in snail1aplus 1b morphants was around 50% of that observed in wt or snail1a morphants. (Fig. 7M-P). These data indicate that the same number of cells occupied a much larger area, explaining the elongated aberrant shape of the polster region observed in these embryos (Fig. 3 and Fig. 7C,D).

In summary, our data reveal the importance of the cooperation between the two populations of snail1-expressing cells and that of prechordal plate expressing hgg1 and cdh1 in the formation of the prechordal plate.

In this work, we describe the cooperation of snail1 genes in the anterior migration of the axial mesendoderm. The snail1a and 1b genes were generated by the duplication of the vertebrate Snail1 gene in the teleost lineage(Manzanares et al., 2001), and their preservation may reflect the appearance of new functions associated to either gene or their subfunctionalisation. According to the duplication-degeneration-complementation model (DDC)(Lynch and Force, 2000), their subfunctionalisation would imply the complementary loss of expression between the two duplicates. Our functional analysis in the axial mesendoderm highlights the translation of this complementarity into functional cooperation.

The mutually exclusive expression of snail1 and Cdh1 generates an interesting framework of transient and dynamic adhesion that somehow reconciles different models of the regulation of cell adhesion during the directed migration of prechordal mesoderm cells (see Solnica-Krezel, 2006). At the beginning of gastrulation, the internalisation of the first axial mesendodermal progenitors occurs through the delamination of individual cells(Montero et al., 2005). This process is reminiscent of the EMT that occurs during mesodermal delamination in amniotes as they do not express cadherin. In agreement with this view, in the shield, these individual cells express goosecoid and snail1a (Thisse et al.,1993). Similarly, the first axial mesodermal cells to ingress during gastrulation in Xenopus embryos delaminate as individual cells that transiently express Snail1, as discussed previously (see Morales and Nieto, 2004). Upon delamination, cells very quickly re-express Cdh1 and migrate as a cohesive anterior group. Cdh1 re-expression can be seen as a rapid mesenchymal to epithelium transition (MET, the reverse of EMT), which occurs concomitant with the loss of snail1a and appears to be necessary for the proper migration of the mesendodermal cells(Montero et al., 2005). As such, injection of cdh1 MO(Montero et al., 2005) impairs anterior migration, as occurs in different cdh1 hypomorphic mutants(Kane et al., 2005; Shimizu et al., 2005). Mesendodermal progenitors involute normally in these embryos(Montero et al., 2005), as expected considering that wild-type involuting cells lack Cdh1 because of the expression of snail1a. Thus, cadherin is required for mesendodermal extension in cells that lack snail1a and 1b, in the prechordal plate and in the posterior regions but it is excluded from the delaminating cells, which undergo a very transient EMT, and in those located posterior to the hgg1-positive domain.

The migration of the prechordal plate cells is impaired in snail1aand snail1b morphants due to the upregulation of Cdh1 in the corresponding populations. Hence, the prechordal plate forms at an anomalous posterior position in embryos deficient in snail1a function(Yamashita et al., 2004). We found similar phenotypes when 20 ng of snail1a MO was injected, an unusually high dose when compared to that used for the snail1b MO(1.5-3ng). Nevertheless, the expression of hgg1 showed that the prechordal plate maintained its identity, again indicating that Snail1a is not a fate determinant but rather an inducer of movement(Barrallo-Gimeno and Nieto,2005) and in agreement with the fact that the hgg1 domain does not express any of the snail1 genes.

Interestingly, although it impairs the movements of the posterior prechordal plate cells, the knockdown of snail1b does not prevent the correct migration of those located in its anterior domain since the pioneers seem to reach the appropriate anterior position. Again, this behaviour reinforces the crucial role of snail1a in driving the anterior migration of the prechordal plate(Yamashita et al., 2004). The correct movement of the prechordal plate cells depends on the expression and motility of the snail1a cell population located underneath and in front of them. In addition, the intermingling of the more posterior hgg1-expressing prechordal plate cells and those expressing snail1b at gastrulation stages in the wild-type embryo may explain why the polster shows an abnormally elongated shape in embryos where snail1b function is compromised. The increased cell to cell adhesion among the axial mesendodermal cells that have aberrantly lost snail1b expression disturbs the motility of the prechordal plate snail1-negative and hgg1-positive cells that are forced to lag behind, and that sometimes even detach from their pioneers. This effect would only be observed when a significant proportion of axial mesendodermal cells have lost Snail1b function, as occurs in snail1b morphants. As such, small grafts of snail1b morphant cells into wild-type hosts only show a cell-autonomous effect, being unable to perturb the behaviour of surrounding normal axial mesendodermal cells. This suggests that a sort of `community effect', i.e. certain number of abnormal cells, is required.

Our data indicate that the cell-autonomous function of snail1b is necessary for the proper migration of the axial mesendoderm as observed by the defects in migration of snail1b morphant cells transplanted into WT embryos. In addition, the influence of snail1a and 1b loss of function on snail1-negative, hgg1-positive prechordal plate cells can be summarised as follows: the posterior cells push and the YSL is likely to provide an active substrate for prechordal plate migration(D'Amico and Cooper, 2001; Rohde and Heisenberg, 2007). The latter explains the abnormal posterior position of the prechordal plate in snail1a morphants (Yamashita et al., 2004) (this work). In summary, the precursors of the polster require the activity of Snail1a, which exerts a cell-non-autonomous effect on them, to reach their correct position, and they also require Snail1b function to acquire a normal shape. Selective adhesion allows hgg1-expressing prechordal plate cells to compact and separate from the more posterior cells while migrating. Indeed, a clear morphological boundary between the polster and the more posterior cells is evident in gastrulating fish embryos. Hence,for the axial mesendoderm to migrate properly, cells located adjacent to the hgg1-expression domain must migrate actively. Although this does not exclude the need for prechordal plate cells to actively migrate(Heisenberg and Tada, 2002),it highlights the importance of a force driven by Snail1a in the YSL and of the influence of actively migratory snail1b-expressing cells located posteriorly.

We have unveiled an important function for snail1 genes in the migration of the axial mesendoderm. The products of both genes repress cadherin expression and by downregulating adhesion, they facilitate the migration. The combination of the cell autonomous action of these two transcription factors in two independent populations influences the behaviour of the snail-free prechordal plate region destined to become the hatching gland. In addition, these results extend and confirm that Snail genes regulate cell adhesion and cell movements in different tissues, and that their function as adhesion regulators may or may not be associated with the induction of a full EMT depending on the cell context (i.e. neural crest cells or delaminating mesendodermal precursors versus mesendodermal precursors located posterior to the hgg1-positive cells). As such, the EMT is much more than just cell adhesion or cadherin downregulation, highlighting the need to analyse the different Snail targets and partners during the induction of cell movements in the developing embryo.

We thank Steve Wilson for his generosity and very helpful suggestions. The initial steps of this work were carried out in his lab thanks to an EMBO short-term fellowship to M.J.B. This work was supported by grants of the Spanish Ministry of Education and Science (DGICYT-BMC2002-0383;MEC-BFU2004-02665 and MEC-BFU2005-05772) and the regional Government of Valencia (GV04B-292) to M.A.N.; GV05-028 to A.B.G. and Fondecyt 1031003, ICM P02-050, ICGEB CRP/CHI03-03 to M.L.A. M.J.B. and A.B.-G. hold Ramon y Cajal Contracts (Spanish Ministry of Education and Science) and H.A. is contracted under the I3P Programme (European Social Fund/CSIC).