Conserved and divergent roles for members of the Snail family of transcription factors in the chick and mouse embryo

Much of the progress made in our understanding of developmental processes over recent years has been the result of the identification of genes responsible for mutant phenotypes in the Drosophila embryo. The subsequent cloning of vertebrate homologues of such genes and the analysis of their functions has in many cases established the importance of these genes in fundamental developmental events. However, the increasing use of transgenic technology has brought to light the existence of compensatory mechanisms that exist within gene families whereby the absence of a specific family member may be compensated for by another. Such redundancy within gene families, along with the overlapping domains of expression displayed in some cases by distinct members of the same gene family, has given rise to the need to define the roles of the individual members of gene families and to compare their function in distinct species.

The zinc-finger transcription factor, snail (sna), was initially identified in the Drosophila embryo (Grau et al., 1984; Boulay et al., 1987), embryos carrying mutations in sna showing defects in mesoderm formation (Alberga et al., 1991). The cloning of vertebrate homologues of sna confirmed the possible role of this gene in mesoderm specification in distinct species (Sargent and Bennet, 1990; Nieto et al., 1992; Smith et al., 1992). Further vertebrate members of the Snail family have now been isolated and their expression patterns analysed (Hammerschmidt and Nüsslein-Volhard, 1993; Nieto et al., 1994; Thisse et al., 1993, 1995; Mayor et al., 1995). To date, it appears that, based on sequence analysis, only two members of this family exist in vertebrates, Snail (Sna) and Slug (Slu). However, in zebrafish, the two genes isolated have been called Sna1 (Thisse et al., 1993; Hammerschmidt and Nüsslein-Volhard, 1993) and Sna2 (Thisse et al., 1995), although the authors mention a higher degree of similarity between Sna2 and Slu (Thisse et al., 1995).

Whilst interest in Sna was originally due to its proposed role in mesoderm formation (Sargent and Bennet, 1990; Nieto et al., 1992; Smith et al., 1992), more recently, a Sna-related chick gene has been implicated in the control of vertebrate left-right asymmetry (Isaac et al., 1997). Slu has also generated a great deal of interest since it was first identified in the chick embryo (Nieto et al., 1994). Slu is a useful marker for premigratory neural crest (Nieto et al., 1994; Mayor et al., 1995) as well as being critical for the emigration of the neural crest from the neural tube and of the early mesoderm from the primitive streak. Indeed, this property led to the suggestion that cSlu may be required to release cells from epithelial structures permitting them to migrate, a process known as the epithelial-to-mesenchyme transition (EMT, Nieto et al., 1994). Despite the importance of these two genes during chick development, the data available regarding their patterns of expression during development are limited to discrete tissues.

Recently, a mouse homologue of Slu has been identified from a murine cell line and employed to further demonstrate the role of Slu in driving EMT in tissue culture cells (Savagner et al., 1997). However, no expression studies of this gene have been carried out during development. In addition, we have isolated mSlu from embryonic mouse tissue and analysed its pattern of expression in detail. We have completed the expression studies of the two chick genes, enabling us to compare the expression pattern of the two vertebrate members of the Snail family in both chick and mouse.

As a result of this analysis, we have observed that, at various sites during early development, the expression of Slu and Sna is inverted between chick and mouse, notably in the neural crest and the mesoderm. Other sites of expression for each of the genes are conserved between the two species. When taken together, the combined expression sites for the chick and mouse Snail family members is the same in each species. Together with these expression studies, sequence comparison analysis has allowed us to identify residues that are characteristic of either Sna or Slu proteins and thus, diagnostic for the identification of subfamily members. Finally, the comparison of both sequence and expression patterns of this gene family from Drosophila to mammals has allowed us to propose how Snail proteins may have evolved. Indeed, our data may shed new light on the origin of tissues of evolutionary significance such as the neural crest.

The embryos used throughout this study were obtained from natural matings of Balb-C (Harlan) and SJL mouse strains (Jackson), and White Leghorn chickens (Granja Rodriguez-Serrano, Salamanca, Spain). Chick eggs were incubated and opened, and the embryos staged according to Hamburger and Hamilton (1951). For mouse embryos, ages were determined as days post-coitum (d.p.c.), the day on which the vaginal plug was detected being designated 0.5 d.p.c.

Degenerate primers were designed, based upon the first and last 6 amino acid sequences of the chick and Xenopus Slu proteins (Nieto et al., 1994; Mayor et al., 1995), to amplify cDNA isolated from 9.5 d.p.c. mouse embryos. A fragment that approximated in size to the full-length chick and Xenopus cDNAs was cloned into the pGEM T-vector (Promega). When sequenced, this cDNA showed a high degree of similarity to the chick sequence and a predicted 92% identity at the amino acid level.

Whole-mount in situ hybridisation was carried out in chick and mouse embryos at various stages of development as previously described (Nieto et al., 1996). Digoxigenin-labelled probes were synthesized from the complete coding sequence of mSlu or fragments of the cDNAs corresponding to cSlu, mSna and cSna as follows: mSlu, nucleotides 1-807 (complete coding sequence); cSlu, nucleotides 1-360; mSna, nucleotides 433-824; cSna, nucleotides 258-767. Following hybridisation, the embryos were embedded in fibrowax and sectioned at 15 μm.

We have cloned the coding region of the mouse Slu gene enabling us to compare the sequences and patterns of expression of the two members of the Snail family of zinc-finger transcription factors, Sna and Slu, during early stages of chick and mouse development. The mSlu sequence is identical to that cloned by Savagner et al. (1997) and encodes a protein that shows 92% identity to the predicted cSlu protein and 65% identity to cSna protein (Table 1). Comparing the amino acid sequence of the vertebrate Slu and Sna proteins, several points become immediately obvious (Fig. 1). Of the eight vertebrate Snail family members isolated to date, three of them showed very little divergence between species and corresponded to mouse, chick and Xenopus Slu proteins (mSlu and xSlu show 92% and 91% identity to cSlu, respectively). The conservation is particularly notable in the zinc-finger domain where mSlu and xSlu show 99% and 98% identity to cSlu respectively, across the last four fingers (Table 1). The remaining five proteins correspond to the Sna subfamily, which similarly show a high degree of conservation in the zinc-finger domain, although less than that shown by Slu proteins (between 81% and 95% identity when compared to cSna). However, they are much more variable in the amino terminal portion of the protein (between 40% and 55% identity when compared to cSna).

When we analysed the amino terminal portion of the protein in greater detail, we encountered information that may prove critical for the identification of distinct family members and that may provide clues as to the evolution of this protein family. We identified a stretch of 29 amino acids immediately preceding the zinc-finger domain, which is exclusive to and highly conserved in vertebrate Slu proteins (boxed in Fig. 1). This sequence may be diagnostic for vertebrate Slu proteins. Furthermore, the vertebrate Sna proteins appear to contain several amino acid stretches of variable size at conserved positions that are absent in the Slu protein sequences (grey arrowheads in Fig. 1). The identity between each of these short sequences is not sufficiently conserved as to enable us to consider these stretches of amino acids diagnostic for Sna proteins. However, we have identified a few amino acids that are distinct in Sna from Slu but that are conserved in all Sna proteins (shaded in blue in Fig. 1). Taking all this data together, it became evident that the two genes isolated from zebrafish embryos belong to the Sna subgroup.

When compared to the three members of the Snail family in Drosophila, sna (Boulay et al., 1987), escargot (esg, Whiteley et al., 1992) and scratch (scr, Roark et al., 1995), it appears that all vertebrate members of the family show a slightly greater degree of identity to the product of esg (between 50 and 59% identity) than to that of sna (between 44 and 55% identity), scr being the most distant relative of the Drosophila genes (Table 2). The region 5′ to the zinc-finger domains in each of the Drosophila proteins is notably larger than that of the vertebrate homologues as is the case for the only Snail family homologues described in sea urchin and ascidians (Illingworth et al., 1992; Corbo et al., 1997). The vertebrate proteins of the Sna and Slu subfamilies contain sequences in this 5′ region that are specific to both esg or sna. These data indicate that the vertebrate and Drosophila genes arise from a common progenitor.

We have determined the distribution of mSlu transcripts at various stages of development by in situ hybridisation and compared the distribution of both Slu and Sna transcripts in equivalent developmental stages of chick and mouse development. The whole-mount in situs comparing the sites of expression between these genes are shown in Figs 2 and 3. A more detailed analysis with respect to their expression in specific tissues is presented in the remaining figures.

The Slu gene was first described in the chick embryo where it is expressed at high levels in both the premigratory and migratory neural crest (Nieto et al., 1994). Indeed, loss-of-function experiments indicated that this gene is critical for the emigration of the neural crest from the neural tube. For this reason, we first analysed the expression of mSlu in the cephalic neural crest, and compared it with the expression of Sna in both chick and mouse embryos. In the chick, strong expression of cSlu in the premigratory and migratory crest was observed at all stages examined (stage 8-18: Figs 2C,E, 3A,C, 4A,C,K).

However, when we analysed the expression of mSlu in embryos at 8.5 and 9.5 d.p.c., transcripts were observed in migrating neural crest cells but not in the premigratory neural crest (Figs 2I,K, 3E, 4E,G,I). In contrast, the expression of Sna in the mouse neural crest appeared to be very similar to that observed for cSlu, both in premigratory (compare Fig. 2J,L with C,E and Fig. 4H with A) and migratory cephalic neural crests cells (compare Fig. 4J with C). Transcripts of mSna were expressed in a greater number of migratory crest cells than mSlu (Fig. 4I,J).

Conversely, cSna transcripts were not observed in the chick premigratory neural crest (Figs 2D,F, 4B,D) or in the migratory crest cells at early stages (Fig. 2F). Up to stage 13, cSna transcripts were completely absent in migratory hindbrain crest cells (Fig. 3B) but cSna expression was detected in a subpopulation of migratory hindbrain neural crest from stage 14 (Fig. 4D). At stage 18, the majority of both Slu- and Sna-expressing crest had arrived at their destination although expression of both genes could still be detected in some migrating cells (Fig. 3C,D). This could be clearly seen in the hindbrain-derived crest that migrates to the branchial arches. It is interesting to note that the number of crest cells that express cSlu is significantly greater than those expressing cSna (Fig. 4C,D,K,L). Thus, it appears that, in the neural crest, mSna expression is very similar to that of cSlu, and the expression of cSna and mSlu is also very similar to each other.

In the chick embryo, both Slu and Sna are expressed from early on in development. At stage 5, cSlu transcripts were restricted to the primitive streak and the ingressing mesodermal cells, whilst cSna transcripts were absent from these cells (Nieto et al., 1994; Isaac et al., 1997; Fig. 2A,B). At an equivalent stage of mouse postimplantation development, no mSlu transcripts were detected in 7.5 d.p.c. embryonic tissues (Fig. 2G). In contrast, mSna expression is particularly strong in the mesoderm as it migrates from the primitive streak (Nieto et al., 1992; Smith et al., 1992; Fig. 2H). The expression of mSna and the absence of mSlu transcripts both in the primitive streak and early mesoderm were also observed in mouse embryos at 8.5 d.p.c. (Fig. 5E,F). Similarly, in the posterior region of stage 8 chick embryos, the expression of cSlu and the absence of cSna transcripts in these tissues was readily appreciated both in whole mounts and in sections (Fig. 5A-D). As described above for the neural crest, Sna expression in the primitive streak and early mesoderm in the mouse is similar to that of Slu in the chick.

As development proceeds and the mesoderm segregates into distinct populations, Snail family members are expressed in a more complex fashion reflecting these processes. mSlu expression was detected exclusively in the lateral mesoderm (Figs 3E, 6E,G) whereas, in contrast, mSna transcripts were observed in the paraxial mesoderm (Figs 3F, 6F,H).

The pattern of mesoderm expression of these genes in the chick was somewhat more complex. In whole-mount chick embryos at stage 13, the inverted pattern of expression of both cSlu and cSna with respect to their mouse counterparts can be readily appreciated (compare Fig. 6A,B with E,F). cSlu transcripts were restricted to the paraxial mesoderm in caudal regions, whereas cSna transcripts were detected in lateral plate mesoderm (Fig. 6A,B). However, in more rostral regions but caudal to the last-formed somite, more complicated patterns of expression were observed. In the paraxial mesoderm, cSlu continued to be expressed in the absence of cSna transcripts (Fig. 6C,D). However, in the lateral mesoderm, both genes appeared to be expressed, although in complementary domains (Fig. 6C,D). The more medial lateral mesoderm cells expressed cSlu but not cSna transcripts, whereas, in the more lateral regions of the lateral plate mesoderm, cells do not express cSlu but cSna transcripts were detected. These domains of expression appear to be complementary. This applied both to the somatic and splanchnic mesoderm although the boundary between Slu- and Sna-expressing cells is positioned at different points along the mediolateral axis in each tissue.

A distinct domain of expression in the paraxial mesoderm was observed immediately prior to the last-formed somite, which is separated from the rest of the paraxial mesoderm by a non-expressing domain. This transitory expression domain could be identified in whole mounts of mouse embryos at 8.5 and 9.5 d.p.c. labelled for Sna (Figs 2J, 6F) or conversely in stage 13 chick embryos labelled for Slu (Fig. 6A).

The somites form as a result of the segmentation and epithelial transformation of the paraxial mesoderm. Each somite subsequently subdivides into distinct domains that differentiate and give rise to different tissues. Both Slu and Sna are expressed within the somites, their patterns of expression changing depending on the differentiated state of the somite. In the early somites, mSlu transcripts were detected from as early as 8.5 d.p.c. in the whole of the somite (Figs 2I,K, 7B,G). This expression pattern was similar to that observed for cSna, which was also expressed ubiquitously across the somite between stage 8 and 14 (Isaac et al., 1997; Figs 2D, 3B, 7C,F). The expression of both cSlu and mSna was confined to cells situated ventrally in the somites at these early stages of somite development, most probably cells undergoing EMT (Fig. 7A,E,D,H). By 9.5 d.p.c., mSlu was detected in sites corresponding to the rostral halves of the somites (Fig. 3E,G). This corresponded to the neural crest cells migrating across the sclerotome as was confirmed in sections of the embryos (Fig. 7K). At these stages, mSna transcripts were detected in the neural crest and in the cells of the somite proper, being excluded from the dermatome whilst being expressed in the myotome and sclerotome (Nieto et al., 1992; Smith et al., 1992; Fig. 7L). The distribution of both Sna and Slu transcripts in somites at these stages were conserved in chick embryos of an equivalent age, cSna being expressed in the myotome and sclerotome (Fig. 7J), whilst cSlu transcripts were only observed in migratory trunk neural crest cells (Fig. 7I).

We have previously described a highly dynamic pattern of expression for cSlu in the developing limb bud (Ros et al., 1997) and we therefore investigated whether this pattern of expression might be conserved in the mouse. The expression of mSlu and mSna in the limb bud primordium was observed from 9.5 d.p.c. (Fig. 3E,F). As the limb bud developed, mSlu transcripts were observed in the ventral and dorsal ectoderm of the limb bud, as well as in the forelimb mesenchyme (Fig. 8G,H). This pattern of mSlu expression in the limb bud is very similar to that seen in the chick (compare Fig. 8C,D with G,H). However, the principal difference was the presence of mSlu transcripts in the ventral ectoderm of the limb bud where cSlu transcripts were not detected at these stages. Transcripts of mSna were observed in a more extensive domain of the limb bud mesenchyme whilst they appeared to be excluded from the limb bud ectoderm (Fig. 8E,F). Transcripts of cSna were essentially expressed in a similar extensive mesenchymal domain to that observed in the mouse and were excluded from the limb bud ectoderm (Fig. 8A,B). Sna transcripts were detected before the appearance of Slu expression (see Fig. 3). A further observation was the presence of mSlu but not mSna transcripts in the apical ectodermal ridge (AER; Fig. 8H), whilst neither gene was expressed in the AER of the chick limb bud. Moreover, neither of these genes were observed to be restricted to the zone of polarising activity in the mouse as has been reported during the early stages of cSlu expression in the limb bud (Ros et al., 1997; Buxton et al., 1997).

At later stages of forelimb development, a similar distribution of mSlu transcripts was observed at 13.5 d.p.c. to that observed at HH stage 32 in the chick limb (Fig. 8J,L), Slu transcripts being restricted to the interdigital regions of the limb. A conservation in the pattern of expression was also observed for Sna in both mouse and chick limbs (Fig. 8I,K).

In our examination of the expression pattern of mouse Slu, we observed transcripts in the lens of the developing eye at 10.5 d.p.c. (Fig. 9C). This expression in the lens was conserved in chick embryos between stage 14 and stage 18 (Fig. 9A). We were unable to detect any mouse or chick Sna-expressing cells in the eye (Fig. 9B,D).

Another facet of Sna and Slu expression that seems to have been conserved between these two species is the asymmetric left-right expression of Sna described in the chick, and thought to be related to the establishment of left-right asymmetry (Isaac et al., 1997). The transient asymmetric expression of Sna in the lateral mesoderm was observed in both mouse and chick embryos, expression being notably higher in the right side of the embryo (Figs 2D, 10). We were unable to detect an asymmetric distribution of Slu transcripts in any tissue at any of the stages examined.

The Snail family of zinc-finger transcription factors has been implicated in the formation of distinct tissues during the early development of the vertebrate embryo. We have cloned the mouse homologue of Slu, and compared its sequence and expression with that of the other mouse and chick Snail family members.

Comparing the amino acid sequences of the existing vertebrate members of the Snail family, it is apparent that they are best grouped into two subfamilies corresponding to the Sna and Slu genes already described (Fig. 1). The predicted Slu proteins show a degree of identity across their whole sequence considerably higher than that between Sna proteins, suggesting a greater evolutionary divergence among the latter. This divergence is particularly evident in the 5′ region of the Sna proteins, the zinc-finger domains of all family members being highly conserved. Members of the Slu subfamily contain a highly conserved stretch of 29 amino acids 5′ to the zinc-finger domains that we consider diagnostic for Slu proteins. Indeed, this sequence is absent from the vertebrate Sna proteins and, interestingly, is also absent from the two Snail family members isolated from zebrafish (Hammerschmidt and Nüsslein-Volhard, 1993; Thisse et al., 1993, 1995). The absence of this sequence, together with the presence of several amino acid stretches at conserved positions in Sna but not in Slu homologues, may be diagnostic for Sna proteins. In summary, we have detected diagnostic hints that permit the immediate identification of members of the Sna or Slu subfamily. Thus, with respect to the existing Snail family members, we conclude that mouse, chick and Xenopus contain one Sna and one Slu gene, whereas zebrafish has two Sna genes.

A peculiarity of the Snail family of proteins identified in distinct vertebrates is the variability in the number of zinc-finger domains that they contain. Whilst all Slu proteins identified contain 5 fingers, two of the five vertebrate Sna proteins, mSna and zSna1, contain only 4 fingers. It has been suggested that there might be a certain relationship between the number of fingers and the sites of expression that could reflect a subdivision for the family members (Thisse et al., 1995). The information that we have compiled in conjunction with the expression patterns of these genes indicates that this is not the case. The absence of the first finger in mSna and zSna1 brings into question its functional significance. With respect to this, the zinc-finger protein GL1 contains 5-finger domains of a similar structure to that of the Snail family, and it has been shown that the first of these domains does not interact with DNA (Pavletich and Pabo, 1993).

The data that we present here also enables us to clarify the confusion over the identity of certain genes within the family, for example that of the chick Snail-related gene (cSnR). Despite the high degree of identity at the amino acid level to the existing vertebrate Sna genes, it remained unclear, owing to differences observed by the authors in the patterns of expression between cSnR and mSna at early stages, as to whether this gene represented the true cSna homologue. We show that, structurally speaking, cSnR belongs to the Sna subfamily and that the combined sites of expression of cSnR and cSlu are the same as those identified in other vertebrate species for the Snail family members. On this basis, we consider it correct to assume that the cSnR does indeed represent the true chick Sna homologue, cSna.

In Drosophila, three members of the Snail family have been identified: sna, esg and scr (Boulay et al., 1987; Whiteley et al., 1992; Roark et al., 1995). Whilst sna appears to be functionally more closely related to the vertebrate genes, esg shows a slightly greater sequence similarity to both vertebrate Sna and Slu. Short stretches of amino acids that are found independently in sna or esg correspond to amino acids in the 5′ region of the vertebrate genes. Our interpretation of these sequence comparisons is that the Drosophila and vertebrate genes have descended from a common progenitor. It is possible that the duplication giving rise to the two vertebrate genes might have occurred at the described major phase of gene duplication at the origin of vertebrates (Holland et al., 1994). In support of this idea, only a single Sna/Slu homologue has been identified in the sea urchin (Illingworth et al., 1992) and ascidians (Corbo et al., 1997). The predicted protein sequence of the sea urchin gene is clearly identifiable as a member of this family owing to the absolute conservation of the first 9 amino acids and the high degree of conservation within the five zinc-finger domains. The conserved diagnostic sequence in Slu is not present in the sea urchin or in the ascidian protein, suggesting that this sequence may have arisen later in the Slu subfamily. The conservation of certain amino acids specific to the Slu protein in the zinc-finger domains is evidence that both these proteins also show a similarity to Slu.

In our analysis of embryos at early developmental stages, we observed that mSlu expression was absent in tissues that express Slu in the chick. Surprisingly, in many of the sites where chick Slu was expressed, we observed Sna expression in the mouse. Moreover, at sites of cSna expression, Slu transcripts are expressed in the mouse. Thus, an inversion in the expression of Snail family members appears to have occurred between the chick and mouse in the premigratory neural crest, early mesoderm and during early somite formation. The expression of Xenopus Sna and Slu is similar to that of both genes in the mouse (Essex et al., 1993; Mayor et al., 1995), indicating that expression at early stages of development is inverted in the chick. Thus, we propose that this inversion must have occurred in the avian lineage after the divergence between birds and mammals. This is corroborated by the fact that the sum of the expression sites of both genes is conserved in vertebrates.

The inversion in expression sites between Slu and Sna seems likely to be the result of recombination events between the regulatory sequences of both genes. Thus, our data indicate the existence of modulatory elements that can independently regulate the temporal and spatial expression of these genes. In support of this hypothesis, distinct elements have been identified in the promoter of xSna that are required for its mesodermal and ectodermal expression (Mayor et al., 1993). One explanation for our observations is that a reshuffling of these elements has occurred in the avian lineage. This would be consistent with the swapping of only some sites of expression and the conservation of others. Thus, it would be of interest to analyse the promoter regions of the family members in several species, the prediction being that some of the vertebrate regulatory sequences of mSna would be regulating cSlu and vice versa.

The expression of different family members in a particular tissue in distinct species is not unusual and probably evolves from a situation where both family members are co-expressed in such a tissue following gene duplication (see below). However, to our knowledge, the swapping of expression sites between different species as observed here, has not been described for other gene families. One implication of these findings relates to the analysis of knockouts and to the results of loss- or gain-of-function experiments referring to this family. For example, according to the data regarding Slu function in the chick (Nieto et al., 1994), one might expect that the mouse knockout for Slu should show defects in the emigration of the neural crest cells from the neural tube or in the delamination of the early mesoderm. However, in the light of our results one would more expect this phenotype for a mouse deficient in Sna function, or possibly, for the double mutant mice.

As discussed by Cooke et al. (1997), gene duplication offers an opportunity for the acquisition of new roles for different members of gene families. The expression of duplicated genes can be gradually modified such that they might be inactivated in some tissues and recruited to others where new morphological features unique to the vertebrate lineage would form (Holland and García-Fernández, 1996). This is likely to have been the case for the Snail family.

Recent duplicates would originally have overlapping expression patterns at ancestral sites that could gradually be modified. Thus, the expression of one of the duplicates in a specific tissue might be lost and the acquisition of new expression sites in an independent manner for each duplicate might also occur (see below). Once the expression of one family member were lost in a tissue, it is not difficult to imagine that evolutionary pressure would maintain the expression of the other. Otherwise, the loss of function of the two Snail family members in any tissue might give rise to serious problems during development and possible lethality. According to this, the expression sites of the ancient gene might have been distributed between the two duplicates, as seems to be the case for the mesodermal expression of Snail genes. This expression in the mesoderm has been conserved from Drosophila (Alberga et al., 1991) to vertebrates (Nieto et al., 1992; Smith et al., 1992; Essex et al., 1993; Hammerschmidt and Nüsslein-Volhard, 1993; Thisse et al., 1993; Mayor et al., 1995), where different mesodermal populations express Sna or Slu. Indeed, we observe a complementarity in the mesodermal expression of the two genes within one species in the newly formed mesoderm and its derivatives (paraxial and lateral mesoderm).

The two Snail genes isolated from zebrafish (Sna1 and Sna2) are very likely the result of an additional duplication in the fish genome, which is believed to have undergone all the evolutionary changes of the vertebrates as well as some additional ones (see Cooke et al., 1997). Indeed, there are several conserved residues specific to these two zebrafish Sna proteins. Being more recent duplicates, they maintain redundant expression sites (e.g. in the early mesoderm, paraxial mesoderm, migratory neural crest and the somites), as well as differential expression sites such as in the premigratory neural crest (Hammerschmidt and Nüsslein-Volhard, 1993; Thisse et al., 1993, 1995). Indeed, the sum of the expression sites of the two zebrafish Sna genes is equivalent to the sum of the expression sites of Slu and Sna in the other vertebrates that we have analysed. Thus, it appears that the two Sna genes identified in zebrafish are capable of performing the functions that Slu and Sna perform in other vertebrates.

Regarding the recruitment of genes to new functions following duplication, each gene may acquire new expression sites independently. This seems to be the case for the Snail family, Sna being recruited to play a role in chondrogenesis (Nieto et al., 1992; Smith et al., 1992) and Slu to play a role in lens morphogenesis. As discussed by Duboule and Wilkins (1998), the increase in developmental complexity during evolution originates not only from the appearance of new genes, but also from the recruitment of old genes to accomplish additional functions, thus generating new expression sites during development rather than providing them with emergent biochemical activities. Slu has been implicated in the process of EMT during the formation of the neural crest and the early mesoderm (Nieto et al., 1994). Whilst these processes are early patterning events, it seems likely that Slu might also have been co-opted to participate in EMT in later differentiation processes such as in the formation of the endocardial cushions (our unpublished observations) or in the growth-factor-induced EMT in a rat bladder carcinoma cell line (Savagner et al., 1997).

Considering that the history of gene families could reflect the phylogeny of structures (Duboule and Wilkins, 1998), it is interesting to note that the interchanged sites of expression between chick and mouse are phylogenetically ancient. This is compatible with their being regulated by ancestral promoter sequences, whereas many of the sites where the family member expressed is conserved, are new expression sites. These sites might have been independently acquired for each gene (Sna or Slu), possibly by the acquisition of additional cis-regulatory elements, as suggested for the Hox genes (Holland, 1992).

One of the main sites of expression of the vertebrate members of the Snail family is the neural crest. The neural crest is a tissue that, along with the placodes, is believed to have been crucial in the formation of complex sensory organs giving rise to the ‘new head’ of vertebrates as proposed by Gans and Northcutt (1983) and, thus, it has classically been considered as a vertebrate character. However, it is interesting to note that the neural crest first appears at the edges of the neural folds, precisely the region where the ascidian snail homologue is expressed (Corbo et al., 1997). This raises the possibility that these cells are the evolutionary precursors of the neural crest that also appear to be present in the cephalochordata (for a review see Baker and Bronner-Fraser, 1997; Corbo et al., 1997). Thus, the study of this gene family may be of great help in the understanding of the mechanisms that generated the neural crest during evolution.

We are grateful to Michael Sargent and Pierre Savagner for sharing sequences of partial clones of the mouse Slug gene before publication and to Alberto Ferrús, Jordi García-Fernández and the anonymous referees, for helpful advice and discussions. This work was supported by grants from the Spanish Ministry of Education and Culture (DGICYT-PM95-0024), the Comunidad Autónoma de Madrid (08.1/0020/97) and the European Union (FMRX-CT96-0065) to M. A. N.; M. S. was also the recipient of a Wellcome Trust Travel fellowship and S. S. is the recipient of a predoctoral fellowship from the Spanish Ministry of Education and Culture.