Cloning and characterization of three Xenopus slug promoters reveal direct regulation by Lef/beta-catenin signaling.

In amphibians and birds, one of the first steps of neural crest cell (NCC) determination is expression of the transcription factor Slug. This marker has been used to demonstrate that BMP and Wnt molecules play a major role in NCC induction. However, it is unknown whether Slug expression is directly or indirectly regulated by these signals. We report here the cloning and characterization of three Xenopus Slug promoters: that of the Xenopus tropicalis slug gene and those of two Xenopus laevis Slug pseudoalleles. Although the three genes encode proteins with almost identical amino acid sequences and are expressed with similar spatiotemporal patterns, their 5'-flanking regions are quite different. A striking difference is a deletion in the X. tropicalis gene located precisely at the transcription initiation site that results in the X. tropicalis promoter being inefficient in X. laevis. Additionally, we identified two regions common to the three promoters that are necessary and sufficient to drive specific expression in NCCs. Interestingly, one of the common regulatory regions presents a functional Lef/beta-catenin-binding site necessary for specific expression. As the Lef.beta-catenin complex is a downstream effector of Wnt signaling, these results suggest that Xenopus Slug is a direct target of NCC determination signals.


From UMR 144, CNRS/Institut Curie, 26 rue d'Ulm, 75248 Paris Cedex 05, France
In amphibians and birds, one of the first steps of neural crest cell (NCC) determination is expression of the transcription factor Slug. This marker has been used to demonstrate that BMP and Wnt molecules play a major role in NCC induction. However, it is unknown whether Slug expression is directly or indirectly regulated by these signals. We report here the cloning and characterization of three Xenopus Slug promoters: that of the Xenopus tropicalis slug gene and those of two Xenopus laevis Slug pseudoalleles. Although the three genes encode proteins with almost identical amino acid sequences and are expressed with similar spatiotemporal patterns, their 5-flanking regions are quite different. A striking difference is a deletion in the X. tropicalis gene located precisely at the transcription initiation site that results in the X. tropicalis promoter being inefficient in X. laevis. Additionally, we identified two regions common to the three promoters that are necessary and sufficient to drive specific expression in NCCs. Interestingly, one of the common regulatory regions presents a functional Lef/␤-catenin-binding site necessary for specific expression. As the Lef⅐␤-catenin complex is a downstream effector of Wnt signaling, these results suggest that Xenopus Slug is a direct target of NCC determination signals.
In vertebrates, neural crest cells (NCCs) 1 form a unique population of cells. NCCs are determined during neurulation at the border of the neural plate, delaminate from the neural tube during and after its closure, and migrate extensively before colonizing various tissues in which they differentiate into various neuronal and non-neuronal cell types (1,2). In several species, the related zinc finger transcription factors Slug and Snail are specific early markers of this population: Slug in chick (3), Snail in mouse (4,5) and zebrafish (6), and both genes in Xenopus (7,8).
In Xenopus laevis embryos, Slug is expressed from the be-ginning of neurulation in the presumptive neural crest domain, a strip of cells at the border between the neural plate and presumptive epidermis (7). Many studies have used this marker to investigate the molecular basis of NCC induction.
Interactions between the neural and non-neural ectoderm seem to be crucial for this induction (9,10). As it is decisive for neural induction, the role of BMP4 inhibition in NCC determination has been extensively studied. By overexpressing a BMP4 antagonist, noggin or chordin, this inhibition has been shown to be necessary but insufficient for XSlug expression, which occurs only in synergy with another lateralizing signal (10 -12). Both fibroblast growth factor and Wnt pathways have been suggested as sources of this additional signal (7,11,13).
Recent studies have suggested that Wnt is a direct signal required for NCC determination, whereas fibroblast growth factor-dependent XSlug induction would be a consequence of an indirect effect mediated by Wnt (10). These studies have contributed to our understanding of the mechanisms of XSlug induction, but most are difficult to interpret definitively because it is difficult to determine whether XSlug is a direct or indirect target of the injected molecules. We cloned the slug promoter to overcome this problem and to analyze precisely the pathways leading to XSlug induction. In this study, we isolated one promoter from Xenopus tropicalis and two different promoters from X. laevis. The sequences of the three promoters seemed very different at first glance; but by detailed comparison, we were able to distinguish two conserved regions that appeared to be necessary and sufficient for specific expression in NCCs. This promoter, used with a simple expression system, is a very powerful tool for targeting NCCs. It can thus be used to study the role of several putative candidate molecules in the determination, migration, and differentiation of these cells. We also report here for the first time an extensive comparison between X. tropicalis and X. laevis promoters, indicating the possibility of differences in transcription mechanisms in these two species, which are assumed to be very closely related. Finally, the two regulatory regions are potential targets for several transcription factors and would be very useful for determining effective XSlug inducers. We report here the presence in one of these regions of a functional Lef/␤catenin-binding site that is necessary for expression in NCCs. This suggests that the Wnt pathway is directly involved in Xenopus Slug regulation and that Slug is one of the direct targets of NCC determination signals. nucleotides derived from the X. laevis cDNA, these primers allowed us to amplify two complete X. laevis Slug genes. PCR was carried out with the high fidelity Bio-X-Act polymerase (Bioline, London, United Kingdom), and various clones were sequenced with an ABI310 Genetic Analyzer (Applied Biosystems, Foster City, CA). 5Ј-and 3Ј-ends of mRNA were precisely located with the GeneRacer kit (Stratagene, La Jolla, CA). In silico sequence analyses were performed at the Infobiogen facilities (Evry, France).
Embryos and Injections-Wild-type X. laevis embryos were used for all injection experiments. Each construct was inserted in the Sleeping Beauty plasmid between the two inverted terminal repeats (14). These inverted terminal repeats enable genomic integration if transposase is co-injected, which was not the case here. Since this plasmid is remarkably stable and nontoxic, it can also be used alone for ectopic expression in Xenopus embryos. In this study, each construct was injected, in a volume of 15 nl, into one blastomere of a two-cell stage embryo. To compare the effects of different constructs, all the injections in a series were performed at the same molarity. If not otherwise stated below, 2 ϫ 10 Ϫ17 mol of plasmid was injected in 15 nl; for example, this corresponded to 100 pg of the ␣3000-GFP construct. Injections were performed in 0.33ϫ Marc's modified ringers and 4% Ficoll (15). Embryos were then transferred to 0.33ϫ Marc's modified ringers and cultured at 18°C until the desired stage of development. Stages were determined as described by Nieuwkoop and Faber (16).
RT-PCR Experiments-Radioactive semiquantitative RT-PCR, with histone H4 used as an internal control, was performed with a pool of 10 embryos as previously described (17). To detect the GFP transcript, we used the following specific primers: U, 5Ј-GATCCACAGCCACCAT-GAGTA; and D, 5Ј-AAAGCATTGAACACCATAAGT. To detect Slug transcripts, the downstream primer 5Ј-TCGGAAAGTTTGGTTTG-GAGTC was combined with 5Ј-CGGGCGAGGACGAAACCAT for slug ␣, with 5Ј-AGGCCGGGGAGGAAACATA for slug ␤, and with 5Ј-CTGT-GCTACCCCAGCCTGACAT for the amplification of both Slug genes. PCR products were subjected to electrophoresis on 8% native acrylamide gels and analyzed with a PhosphorImager (Amersham Pharmacia Biotech).
In Situ Hybridization-Whole-mount in situ hybridization experiments were performed as described by Harland (18), but with minor modifications (7). For each X. laevis Slug gene, the 5Ј-UTR was used as a short specific probe.

RESULTS
Cloning of the X. tropicalis slug Gene-X. laevis is a pseudotetraploid; and during evolution, some of its genes have become silent. So, to ensure that we did not clone an inactive Slug promoter, we cloned the promoter of the slug gene from X. tropicalis, a very close diploid relative of X. laevis. A 500-bp X. laevis Slug probe corresponding roughly to human slug exon 2 was used to screen an X. tropicalis genomic library at high stringency. We obtained three overlapping positive clones that enabled us to sequence the whole slug gene and 4 kb of the 5Ј-flanking region. The predicted amino acid sequence of the protein encoded by the X. tropicalis gene is very similar to that of the published X. laevis Slug sequence (7), with 99% identity overall and all five zinc fingers totally identical. The organiza-tion of the gene is identical to that of the human gene (19), consisting of two short introns (of 700 bp and 1 kb, respectively) located at the same relative positions in the coding region in both species.
The 5Ј-and 3Ј-ends of the X. tropicalis Slug cDNA were amplified from late neurula total RNA with the GeneRacer kit. This enabled us to obtain the complete X. tropicalis Slug cDNA sequence. This cDNA was 1.8 kb in size, slightly shorter than the mouse Slug mRNA (20).
A transcription initiation site (ϩ1) was identified 142 nucleotides upstream from the start codon, and a good putative CAP site (21) was present nearby: 80% maximum score and centered on ϩ7. Moreover, the upstream region contained a putative TATA box (70% score and centered on Ϫ28), a putative CCAAT box, and two GC boxes. These results suggest that the cloned flanking region is a good candidate for the X. tropicalis slug promoter.
The X. tropicalis slug Promoter Is Inefficient in X. laevis Embryos-To test the activity of the X. tropicalis slug promoter in vivo, deletion constructs were produced with the GFP gene as a reporter gene. These constructs were cloned in the Sleeping Beauty plasmid (14). Injected into X. laevis at the two-cell stage, this plasmid remains present at high levels throughout early development, at least until late tail-bud stages, and enables expression of ectopic proteins under the control of specific promoters. 2 This technique has the great advantage of being simple and harmless for the embryos. Thus, a large number of injected embryos can be obtained in a single experiment, making it possible to obtain statistically significant results with many different constructs.
We assayed various fragments of the X. tropicalis slug promoter, from 200 bp to 4 kb in length, in which the GFP coding region was inserted just downstream from the CAP site. The GFP coding region inserted downstream from the 5Ј-UTR of the X. tropicalis Slug cDNA was used as a negative control (Fig.  1A). Several series of injections were performed with each of these constructs, but no fluorescence was ever detected. We tested for GFP mRNA by semiquantitative RT-PCR of stage 19 NF injected embryos. The GFP transcript was detected in various quantities, depending on the construct injected (Fig. 1A). The constructs p200-GFP and p3900-GFP (including 200 and 3900 bp of the X. tropicalis slug promoter, respectively) gave relatively high levels of expression. These results show that the X. tropicalis slug promoter was able to drive GFP expression in X. laevis embryos. Nevertheless, the efficiency of this promoter could not be compared with that of the endogenous promoter in this assay.
We investigated promoter efficiency by making new constructs with the two most efficient regions of the X. tropicalis slug promoter. These fragments were cloned upstream from a modified X. laevis Slug cDNA containing an insertion (Fig. 1B). This enabled us to detect in injected embryos the products of both the endogenous and exogenous promoters with the same set of primers and thus to perform comparative RT-PCR. The results show that the exogenous and endogenous products were produced in similar amounts (Fig. 1B). As expected, the relative quantity of exogenous Slug decreased over time because the number of plasmid copies/cell decreased during the course of development. 2 However, with this method, each cell should be considered to contain ϳ20 copies of the plasmid at stage 19 NF. Therefore, if the two promoters were similar in efficiency, the exogenous product would be produced in substantial excess over the endogenous product.
Our results indicate that the region of the X. tropicalis slug promoter isolated was inefficient in X. laevis embryos. There may be two reasons for this: either this region did not contain all the regulatory elements, or the regulatory process differed between X. tropicalis and X. laevis.
Cloning of Two X. laevis Slug Pseudoalleles-A PCR approach was used to clone the 5Ј-flanking region of the X. laevis slug gene. Comparison of the sequences of Slug promoters from X. tropicalis, humans (19), and mice 3 revealed no similarity except a common 20-bp region, located at Ϫ3.2 kb in X. tropicalis, Ϫ1.5 kb in humans, and Ϫ1.2 kb in mice. We designed two primers in this common region and used them with two primers corresponding to the Slug coding region to perform nested PCR on X. laevis genomic DNA.
We obtained two types of amplicon containing 3 and 1.7 kb upstream from the ATG codon, respectively. Several subfragments were amplified, and a large number of clones were sequenced to obtain the complete sequences of these two genes. The genetic organization of these two clones was identical ( Fig.  2A), and their predicted coding sequences were very similar (Table I), with 96% identity at the nucleotide level, compared with 67% identity between XSlug and the related zinc finger protein Xenopus Snail (22). This high level of similarity indicated that we had cloned two Slug pseudoalleles. The first (with 3 kb of 5Ј-flanking region) appeared to correspond to the published X. laevis slug gene and is referred to hereafter as slug ␣. In addition to a few differences in the coding region, the second clone is slightly divergent in the 5Ј-and 3Ј-UTRs (Table I). This clone is referred to hereafter as slug ␤.
Both X. laevis Slug Genes Are Expressed in NCCs-Before studying the regulatory capacities of the 5Ј-flanking regions of the two X. laevis Slug clones, we needed to determine the precise expression pattern of each gene. As previous in situ hybridization studies (7, 23) used part of the coding region as a probe, the domains expressed may be those of Slug ␣, Slug ␤, or both Slug mRNAs.
To discriminate between the two pseudoalleles, we first used an RT-PCR approach. Amplification was performed with primers derived from a divergent region of the 5Ј-UTR, which were therefore specific for a particular Slug gene. Semiquantitative PCR using embryos at various stages of development showed that both pseudoalleles were expressed at very similar levels and in exactly the same temporal pattern (Fig. 3A). Moreover, in situ hybridization with neurula embryos and short specific probes derived from the 5Ј-UTR showed that the two X. laevis Slug genes were expressed in the same location, at the border of the neural plate, as expected (Fig. 3B).
The 5Ј-Flanking Regions of Both X. laevis Slug Genes Drive Specific Expression in NCCs-We assessed the efficiency of the X. laevis promoters obtained using constructs consisting of the GFP coding region under the control of the full-length slug ␣ or slug ␤ flanking region (␣3000-GFP and ␤1700-GFP, respectively). Both promoters drove strong GFP expression in NCCs at the edge of the neural plate throughout the neurulation process and later both in premigratory and migrating cells (Fig. 4). Expression was detected at stage 14 NF at the border of the neural plate. The highest level of expression was observed at stage 19 NF, with the strong expression in the cephalic NCCs and a lower level of expression in the rest of the neural crest territory. At stage 25 NF, GFP was strongly expressed in the cephalic premigratory NCCs. The migrating cells colonizing the branchial arches gave a positive, but weaker GFP signal. The cells from the mandibular pathway displayed the strongest expression among cells colonizing the branchial arches. This expression pattern closely resembled that of endogenous Slug according to in situ hybridization data (23). This suggests that the isolated flanking regions of both slug ␣ and slug ␤ contain the key regulatory elements of the promoter. 3 P. Savagner, unpublished data. The bars indicate the means Ϯ S.D. of three independent experiments. Negative control lanes of the gel correspond to samples where no reverse transcriptase was added. The p200-GFP and p3900-GFP constructs gave the strongest expression. B, to compare the efficiency of the X. tropicalis slug promoter with that of the endogenous promoter, a modified X. laevis Slug cDNA was inserted downstream from the p200 and p3900 promoters (upper panel). With a single set of primers (U and D), both endogenous (endo) and exogenous (exo) Slug cDNAs were amplified (middle panel). This made it possible to compare mRNA levels (lower panel). The relative amount of exogenous product was much lower than expected with the Sleeping Beauty expression system, indicating that the X. tropicalis slug promoter is inefficient in X. laevis embryos. a.u., arbitrary units; u, uninjected; St, stage.
Comparison of the Three Xenopus Slug Promoters-The three Xenopus Slug genes presented similar expression patterns, but their 5Ј-flanking regions were different. More detailed analysis revealed two conserved regions ( Fig. 2A and Table I). Region A was ϳ250 bp long and was located just upstream from the CAP site. Region B was ϳ300 bp long, and its position was variable: around Ϫ3 kb in X. tropicalis slug, around Ϫ2.7 kb in X. laevis slug ␣, and around Ϫ1.5 kb in slug ␤. The sequences between these two regions were divergent. X. tropicalis presented a 315-bp fragment of GT repeats, located around Ϫ800 bp. These repeats were absent from the X. laevis Slug promoters, but slug ␣ and slug ␤ displayed 11 and 4 repeats, respectively, of a 79-nucleotide motif. Although this motif has been identified in regulatory as well as transcribed regions of the X. laevis genome (24 -26), it has never been shown to have a regulatory function.
Additionally, comparison of the sequences of the genes from X. tropicalis and X. laevis revealed a striking difference at the transcription initiation site. Although the surrounding regions are almost identical in the three promoters, both X. laevis genes have a 9-bp insertion just downstream from the identified CAP site of the X. tropicalis promoter (Fig. 2B). This insertion creates a second putative CAP site, with a score even higher than that for the first site. This may account for the inefficiency of the X. tropicalis promoter in X. laevis. Alternatively, divergent regions, such as the repeats present in X. laevis but not in X. tropicalis, may be involved in Slug regulation.
The Extra CAP Site in X. laevis Is Necessary for Efficient Transcription-We investigated whether the extra CAP site is necessary for efficient transcription in X. laevis by producing a construct derived from ␤1700-GFP. Seven nucleotides were deleted at the 3Ј-end of the promoter, eliminating the second putative CAP site (construct ␤1700⌬CAP-GFP) (Fig. 5A). This construct did not drive detectable GFP expression when injected into X. laevis embryos (data not shown). GFP mRNA was detected in these embryos by RT-PCR, but at a level of about one-fifth that found in embryos injected with ␤1700-GFP (Fig.  5A). Thus, the second CAP site present in the X. laevis genes is . Asterisks indicate differences between the two species. We predicted CAP sites and determined score as described by Bucher (21). In the X. laevis promoter, a 9-bp insertion creates an extra CAP site, with a better predicted score than the initial CAP site and that of X. tropicalis. necessary for efficient transcription in this species. The ␤1700⌬CAP-GFP construct was very similar to the X. tropicalis constructs used in terms of the CAP site (Fig. 5B). Interestingly, the level of GFP mRNA detected with this ␤1700⌬CAP-GFP construct was similar to that obtained with the most efficient X. tropicalis constructs (compare Figs. 1A and 5A, the same unit is used in both cases). This suggests that the inefficiency of the X. tropicalis promoter in X. laevis was mainly due to the absence of the second CAP site in this promoter.
We also investigated the putative role of the first intron in Slug regulation. We injected a construct containing the ␤1700 promoter, the first exon, the first intron, and the start of the second exon in frame with the GFP cDNA (construct ␤1700int1-GFP) (Fig. 5A). Similar results were obtained with this construct and with ␤1700-GFP, suggesting that the first intron of the slug gene contains no major regulatory elements.

Combination of the Two Common Regions Is Necessary and
Sufficient for Targeted Expression-We investigated the regulatory properties of the various regions of Slug promoters using several constructs containing different portions of the slug ␣ promoter (Fig. 6). The amount of GFP and its distribution in injected embryos were analyzed both by fluorescence monitoring throughout development and by semiquantitative RT-PCR of stage 19 NF embryos. Two different amounts of plasmid were injected (referred to as 1ϫ and 8ϫ) and gave different and complementary results. In both cases, no fluorescence was detected with the negative control, which consisted of the 70 bp just upstream from the transcription start site containing the TATA box. This ␣70TATA-GFP construct gave the basal level for RT-PCR experiments. The ␣200A ⌬ -GFP construct, with three-fourths of region A, drove strong expression both in NCCs and in presumptive epidermis. This reveals the presence in this region of a positive regulatory element, activated both in NCCs and in presumptive epidermis. With the ␣270A-GFP construct, which contained all of region A, no GFP was detected with 1ϫ injections, and mRNA levels were similar to the basal

FIG. 3. The two X. laevis Slug genes have identical expression patterns.
A, RT-PCR experiments were performed on X. laevis embryos at various stages of development. A specific primer was designed for each gene based on the 5Ј-UTR. The graph shows quantification using histone H4 as an internal control. The two Slug mRNAs displayed similar temporal expression patterns and were produced at similar levels. B, the distribution of Slug ␣ and Slug ␤ mRNA was determined in stage 14 NF embryos by in situ hybridization using 5Ј-UTR sequences as short specific probes. The two genes were expressed at the same site in the presumptive cephalic NCCs (arrowheads) as previously described for Slug mRNA. a.u., arbitrary units.

FIG. 4. Both X. laevis Slug promoters drive specific expression in NCCs.
Constructs containing the GFP coding region under the control of the slug ␣ or slug ␤ promoter were injected into one blastomere of a two-cell stage embryo, and GFP levels were monitored throughout early development. The dashed lines indicate the midline of the embryo. e, eye vesicle; c, cement gland. With both constructs, a pattern similar to that for endogenous Slug was obtained. Note that at stage (St.) 25 NF, migrating cells colonizing the branchial arches were still GFPpositive, but the signal was weaker because the acquisition time required to detect it was three times longer (inset in the upper right panel). levels obtained with ␣70TATA-GFP. This indicates the existence of a negative regulatory element in the 5Ј-fourth of region A. With 8ϫ injections, this ␣270A-GFP construct gave a similar nonspecific expression pattern compared with ␣200A ⌬ -GFP, but with lower mRNA levels. The ␣1000-GFP and ␣2200-GFP constructs (with 1 and 2.2 kb of promoter, respectively, and containing region A, but lacking region B) had similar effects compared with ␣270A-GFP for both 1ϫ and 8ϫ injections. This suggests that the intermediate divergent regions of the promoters are not very important for transcription regulation. If the region B was present, as in the full-length ␣3000 promoter, the specific Slug pattern was observed in 1ϫ injected embryos, as described above. This suggested the presence of another positive regulatory element in region B. No GFP expression was ever observed with constructs containing region B and the TATA box, but lacking the rest of region A (see, for example, ␣500B-GFP and ␣2300B-GFP), even for 8ϫ injections. These results suggest that both conserved regions A and B were necessary for specific expression in NCCs. In addition, the intermediate region did not seem to possess regulatory elements. As the relative positions of regions A and B differ in the three promoters, we wondered whether the intermediate region might play an essential role as a spacer. We tested this hypothesis by injecting the ␣700BA-GFP construct, which contains a very short intermediate region. This construct gave a similar expression pattern compared with the full-length promoter, indicating that the two conserved regions alone are sufficient to drive specific expression in NCCs.
Region B Contains a Functional Lef/␤-Catenin-binding Site Necessary for Its Positive Regulatory Action-In silico analysis FIG. 5. The second CAP site is necessary for X. laevis Slug promoter function, whereas the first intron is not. A, the second CAP site was deleted from the slug ␤ promoter (␤1700-⌬CAP construct) (left panel). This construct was injected into embryos, and GFP expression was quantified by RT-PCR analysis and compared with that driven by the full-length promoter (right panel). No fluorescence was detected with this construct in living embryos, and the mRNA level at stage 19 NF was about one-fifth that obtained with the full-length promoter. A construct containing the full-length promoter and the first intron gave results similar to those obtained with the promoter alone. Int1, intron 1; n, number of embryos analyzed. B, shown is the comparison of the sequences surrounding the transcription initiation site in the ␤1700-⌬CAP construct and in the X. laevis and X. tropicalis Slug promoters. The transcription start sites are indicated by an uppercase G. Note that the X. laevis construct lacking the second CAP site was very similar to the X. tropicalis promoter in this region. a.u., arbitrary units.

FIG. 6. Combination of the two conserved regions is necessary and sufficient to drive specific expression in NCCs. Various constructs (left panel)
were injected into one blastomere of a two-cell stage embryo at two different concentrations. In 1ϫ injections, 2 ϫ 10 Ϫ17 mol of plasmid was injected in 15 nl. This corresponded to 100 pg of the ␣3000-GFP construct. In 8ϫ injections, eight times as many plasmids were injected in the same volume. GFP expression was analyzed both by fluorescence monitoring in living embryos and by semiquantitative RT-PCR with stage 19 NF total RNA. n, number of embryos analyzed; fluo, localization and intensity of the observed fluorescence (major phenotype); nc, neural crest; ep, presumptive epidermis. The bars indicate the amount of GFP mRNA relative to histone H4 mRNA. The images show the GFP expression pattern obtained with the ␣70TATA-GFP (TATA), of Xenopus Slug promoter sequences predicted a number of putative binding sites for transcription factors. These sites included a consensus binding site for Lef-1 in the middle of region B (centered on Ϫ2839 in the slug ␣ promoter). The Lef/Tcf transcription factors form complexes with ␤-catenin, acting as downstream targets of the Wnt signaling pathway, which has been shown to be directly involved in NCC determination (10). We investigated the possible role of this site in Slug regulation by performing electrophoretic mobility shift assays to study the in vitro interaction of Lef with the slug promoter. Purified Lef-1 protein specifically bound a 22-bp fragment from region B containing the putative site (Fig. 7A). This binding was inhibited by a nonradioactive competitor containing the Lef-1 consensus binding site (27). Furthermore, the addition of ␤-catenin led to the detection of a ternary DNA⅐Lef⅐␤-catenin complex that was supershifted upon addition of an anti-␤catenin antibody (Fig. 7B). Finally, the same shift and supershift complexes were detected if a probe containing the Lef-1 consensus binding site was used in the assay (data not shown).
We investigated the function of this binding site in vivo with a new construct consisting of the X. laevis slug ␣ promoter from which 8 bp had been deleted in the core of the Lef-binding site (␣3000⌬Lef-GFP). No fluorescence was detected in X. laevis embryos injected with this construct, and RT-PCR analysis showed that GFP mRNA levels were similar to the basal levels obtained with the ␣70TATA-GFP construct (Fig. 7C). Thus, this region seems to be necessary for the positive regulation driven by region B, which suggests that this regulation involves the Lef/␤-catenin signaling pathway.

The XSlug Promoter Is an Efficient Tool for Targeting
NCCs-We report in this study the cloning of XSlug 5Ј-flanking regions, which contain the regulatory elements required for specific expression in NCCs. This promoter is a powerful tool for investigating the molecular mechanisms controlling the behavior and fate of these cells. Only a few studies have investigated these mechanisms, mainly through difficult grafting experiments (28 -31). In such experiments, donor tissues injected with an exogenous mRNA must be transplanted to the precise position of the presumptive neural crest. These manip-ulations are delicate and time-consuming and result in only a small number of analyzable embryos. In addition, the position of the graft may differ slightly between embryos, rendering data interpretation difficult. Overall, such experiments require a considerable amount of work if statistically significant results are to be obtained.
In contrast, the XSlug promoter, combined with an efficient vector like Sleeping Beauty, facilitates the targeting of exogenous proteins to NCCs. Given the pattern of expression driven by this promoter, the role of various molecules can be investigated in at least three important phases of NCC evolution: 1) the quiescent state, corresponding to the entire neurulation process in Xenopus; 2) the epithelium-mesenchyme transition and delamination from the neural tube; and 3) migration along various specific pathways.
Xenopus Slug Gene Evolution-We report here the complete sequencing and comparison of the two X. laevis pseudoalleles and the X. tropicalis homolog of the same gene. This kind of comparison has been reported only once before, for the short ␣-globin gene (32). Such comparisons are very informative, both about the isolated gene itself and about the relative evolution of these two species.
As far as Slug is concerned, this comparison identified regions that may be important in the action of the protein and for its regulation. As expected, the three predicted amino acid sequences of the Slug proteins show a high level of conservation, with Ͼ97% of residues identical. The five zinc fingers are totally identical in the three genes, confirming their crucial importance. In contrast, the 29-amino acid Slug-specific box described by Sefton et al. (33) displays two nonequivalent substitutions in these very closely related species, calling into question the role of this region in Slug function. At the nucleotide level, comparison of the divergent 5Ј-flanking sequences led to the identification of two conserved regions that seem to contain the key regulatory elements of the promoter. These regions are highly conserved, displaying ϳ90% identity; and domains thought to be of particular importance, such as the TATA box and the Lef-binding site, are totally identical in the two species.
These findings indicate that there has been strong selection FIG. 7. Region B contains a functional Lef/␤-catenin site necessary for positive regulation. A, shown are the results from a mobility shift assay with a fragment of region B including the predicted Lef/␤-catenin-binding site as a probe. Purified Lef-1 bound to this fragment (arrow). This binding was inhibited in a dose-dependent manner by adding various amounts (2ϫ to 50ϫ) of a DNA fragment including the published Lef-1 consensus binding site. No competition effect was detected with a mutated Lef-1 site (mut. 10ϫ lane). The arrowhead indicates free probe. comp., competitor. B, if Lef-1 was incubated with a GST/␤-catenin fusion protein before adding the probe, a DNA⅐Lef-1⅐␤-catenin complex was detected by electrophoretic mobility shift assay (middle arrow). This complex was not very stable under the conditions used, but its existence was confirmed by adding an anti-␤-catenin antibody (Ab; upper arrow). The ␤ lane is a negative control in which the probe was incubated with GST/␤-catenin alone. The GST lane is a negative control in which Lef-1 was incubated with GST alone before adding the probe. The arrowhead indicates free probe. C, GFP expression in stage 19 NF embryos injected with the indicated constructs was assessed by RT-PCR. The bars indicate the means Ϯ S.D. of three independent experiments. The ␣3000⌬Lef construct, which contained a promoter lacking the Lef-binding site, gave levels of GFP much lower than those obtained with the full-length ␣3000 promoter, but similar to those obtained with the negative ␣70TATA control. a.u., arbitrary units. pressure and emphasize the importance of a functional Slug protein produced at precisely the right time and place for NCC determination. The related gene snail is also precociously expressed in NCCs, suggesting that Slug and Snail are both important in neural crest formation in Xenopus, whereas only one of these two factors is maintained in this area in other vertebrate species (33).
The analysis of divergent regions provides some information about the similarity and relative evolution of the X. laevis and X. tropicalis genomes. These two species are thought to be closely related, but very little is known about similarities between them in terms of nucleic acid sequence. One study compared rDNA and concluded that there was a sister group relationship between X. laevis and X. tropicalis, which seem to be more closely related to each other than to other morphologically similar pipid frogs (34). One possible mechanism of genomic evolution is the insertion of DNA fragments into nonessential regions. Two examples are found in the divergent regions of Xenopus Slug genes. There is a 79-nucleotide motif in the intermediate region of the X. laevis Slug promoters, but not in the equivalent region of the X. tropicalis promoter. Eleven repeats of this motif are present in slug ␣, whereas only four are present in slug ␤. This suggests that an ancestral motif was inserted after separation of the two species but before X. laevis genomic duplication. In the first intron, 16 repeats of a 24nucleotide motif are present in slug ␣ only, accounting for the difference in the length of this intron between this gene and the other two genes. In this case, the ancestral motif must have been inserted after genomic duplication. Further analysis of the divergent regions should make it possible to date more precisely this genomic duplication in the evolution of Xenopus frogs.
The most striking difference observed in our comparison of Slug genes concerns the transcription initiation site. A 9-bp insertion in X. laevis has created an extra CAP site that seems to be the functional site in this species, as its removal renders the promoter inefficient in vivo. Consistent with this result, the X. tropicalis promoter was unable to drive physiological expression when injected into X. laevis embryos, although it possesses all the other important regulatory regions. This demonstrates that closely related species may present different mechanisms, even in a regulatory process that is quite similar overall between the two species. Thus, care should be taken when using X. tropicalis material (especially DNA material) in X. laevis or vice versa.
Mechanism of XSlug Activation-The three Xenopus Slug promoters cloned displayed two conserved regions, which are critical for the specific regulation of this gene: the combination of these two regions is necessary and sufficient to drive expression in NCCs. These conserved sequences are not present in the 5Ј-flanking regions of the human (19) and mouse 3 Slug genes, but this is consistent with the observed differences in the Slug expression pattern between Xenopus and mouse (33). The only sequence conserved between these four species is a 20-bp motif at the 5Ј-end of region B in Xenopus, suggesting that this motif may be involved in regulation.
The first conserved region in Xenopus Slug promoters (region A) is located just upstream from the CAP site and contains basal transcription regulation elements and at least two specific responsive elements. In the middle of this region, between the TATA box and Ϫ200, there is a positive element that can be activated in all non-neural ectodermal cells, but not in neural tissue, as shown by injection of the ␣200A ⌬ -GFP construct. This specificity suggests interaction with factors specific to the presumptive epidermis. The 5Ј-fourth of region A contains a negative element, which is also efficient in non-neural ectoder-mal cells. Indeed, injected at physiological levels, the ␣270A-GFP construct gave no expression. Injected at higher doses, this construct promoted expression at the same location as ␣200A ⌬ -GFP. This indicates that its negative regulation is probably mediated by one or several nuclear factors: in overexpression experiments, some of the injected plasmid would titer these factors, and "free" plasmid would be responsible for the observed expression. These factors may be specific to nonneural ectodermal cells or ubiquitous.
The second conserved region (region B) contains a positive regulatory element that, in combination with those of region A, is responsible for the specific activation of XSlug genes in NCCs. The distance between these two regions is not critical, as it differs greatly between the three genes, and the elimination of almost all of the intermediate region does not affect expression. Region B contains a functional Lef/␤-catenin-binding site that seems to be required for Slug activation in NCCs. This is consistent with the finding that Wnt signaling is required for the activation of this marker (10) and indicates that Slug is a direct target of this pathway. As Wnt is thought to be one of the initial signals in NCC determination, Slug induction would be one of the first steps of this determination. Nevertheless, slug is not a master gene in NCC formation, as its overexpression in ectodermal explants is insufficient to elicit neural crest gene expression (10). Thus, other specific genes, such as snail, may be regulated by the same determination signals.
Mechanism of Xenopus NCC Determination-Our study confirms that the Wnt signaling pathway plays a direct role in Xenopus NCC induction. It also indicates that other signals are necessary, as the Wnt-responsive element of the XSlug promoter is insufficient to promote expression, as shown by injection of the ␣2300B-GFP construct. This is consistent with previous experiments showing that Wnt alone does not induce NCC markers in ectodermal explants (10). Based on previous data on Xenopus NCC induction, BMP4 is an attractive candidate for the second signaling molecule. Interestingly, some promoters have been shown to be directly regulated by interaction between the Wnt and BMP pathways. For example, in Drosophila melanogaster, Wingless and Decapentaplegic (Wnt and BMP orthologs, respectively) act in synergy to stimulate transcription of the Ultrabithorax gene (35). Similar cooperation between these two signaling pathways was recently demonstrated in Xenopus during formation of the Spemann's organizer (36,37). However, the mechanism is likely to be different in NCCs, as Slug expression requires activation of the Wnt pathway, but inhibition of the BMP4 pathway (10). It is not clear whether total or partial inhibition is required, and there are therefore two possible hypotheses: 1) NCC induction results from combination of the Wnt signal with low levels of BMP4, or 2) the total absence of BMP4 is required for NCC induction in response to the Wnt signal. Further studies of the Slug promoter, particularly of the regulatory elements in region A, should help us to understand the sophisticated mechanism underlying NCC induction.