A developmental pathway controlling outgrowth of the Xenopus tail bud

The cells of the Xenopus tail bud give rise to the neural tube, somites and fin of the posterior part of the tadpole tail (Tucker and Slack, 1995a). As the tail grows, the structures derived from the tail bud become continuous with, and are indistinguishable from, those of the trunk-derived mid-axis. Despite some detailed mapping of the tail bud anatomy on the basis of gene expression domains (Gont et al., 1993; Beck and Slack, 1998; Gawantka et al., 1998), to date nothing is known about the molecular mechanisms involved in tail bud outgrowth and differentiation.

We have previously shown that tail bud determination is dependent on the interactions between three regions that are brought into apposition at the completion of gastrulation (Tucker and Slack, 1995b). These are the main neural plate (N), the posterior 100 μm of the neural plate (M), which is actually mesodermal in both fate and commitment, and the underlying dorsal mesoderm (C). A tail bud will only form when the junction of the N and M regions directly overlies C. We believe that the action of C is exerted over the period of the early neurula, stages 13-16, as it is over this stage range that the capacity to form double tails following duplication of the NM junction by neural plate rotation becomes well established (Tucker, 1995). However, outgrowth of the tail bud does not start until about stage 27, well after the neural tube has closed and the primary body axis of the embryo is complete. We have previously shown that there are two phases of gene expression in the Xenopus tail bud (Beck and Slack, 1998). An early group of genes is expressed from gastrulation onwards and continues to be expressed in the tail bud until the end of tail outgrowth. These genes include X-Notch-1 and X-delta-1. The second phase of gene expression occurs just before tail bud outgrowth, at around stage 26, and includes lunatic fringe (lfng), Xwnt3a and Xhox3. Our results showed that the early and late gene domains together defined seven distinct regions of the tail bud, and this has now been increased to ten regions following the study of additional genes (Gawantka et al., 1998). The particular feature of interest for the present study is the small region of overlap between X-Notch-1 and X-delta-1, expressed together in the posterior wall, and lunatic fringe (lfng), expressed in the dorsal roof of the neural tube. The region of overlap corresponds to the junction between N and M territories in the posterior of the embryo, which defines the direction of tail outgrowth. At around the same time, Xhox3 (a Xenopus homologue of Drosophila evenskipped), is expressed in a subset of cells of the overlap region corresponding to the caudal tip, or most distal cells, of the future tail bud. We reasoned that, by analogy with other systems undergoing outgrowth such as the Drosophila wing disc and the chick limb (Fleming et al., 1997; Laufer et al., 1997; Panin et al., 1997; Rodruigez-Esteban et al., 1997), Notch signalling may be restricted by lfng to the overlap region, and this could be the molecular basis of the N-M interaction in the posterior wall that leads to outgrowth of the tail bud.

In the present study we have investigated the mechanisms underlying tail bud outgrowth in Xenopus using a simple grafting assay to study the effects of ectopic expression and inhibition of late tail bud genes. We show that ectopic tail-like projections can be formed by posterior grafts expressing a constitutively active form of X-Notch-1, Notch ICD. The resulting buds express Xhox3 as well as many other tail bud markers. Furthermore, grafts expressing Xhox3 form identical ectopic tails, suggesting that Xhox3 is a downstream target of Notch signalling. Evidence that Xhox3 is actually a critical step in bud outgrowth is provided by the use of an antimorphic form of Xhox3, which prevents ectopic tail formation by Notch ICD, and also prevents tail development in intact embryos.

These results provide evidence that the N-M interaction initiates a new molecular pathway controlling tail bud outgrowth, involving activation of Notch signalling at the future tail tip and consequent activation of Xhox3. We also show that expression of Xwnt3a is required in addition to Notch signalling and that tail outgrowth may restricted to the posterior as a result of this additional requirement. We conclude that extension of the caudal neural tube and tail outgrowth in Xenopus occurs by mechanisms which show some similarity to those involved in appendage outgrowth in other organisms. Given the similar expression patterns for Notch, Wnt3a and evenskipped-related genes in mouse (Takada et al., 1994; Gofflot et al., 1997) and zebrafish (Joly et al., 1993; Westin and Lardelli, 1997), this may provide a general mechanism for formation of the tail in all vertebrates.

Xenopus laevis embryo collection was as described in Godsave et al. (1988). Embryos were cultured in 1× NAM salts (110 mM NaCl, 2 mM KCl, 1 mM Ca(NO₃)₂.4H₂O, 1 mM MgSO₄.7H₂O, 0.1 mM Na₂EDTA) with 1 mM NaHCO₃, 5 mM Hepes, pH 7.5, and 2.5 μg/ml gentamycin sulphate (Sigma) initially and then transferred to NAM/10 before gastrulation (0.1× NAM salts with 5 mM Hepes, pH 7.5, and 2.5 μg/ml gentamycin sulphate). Embryos were staged according to Nieuwkoop and Faber (1967).

Full-length Xhox3 cDNA was truncated just 3′ to the homeobox (Pro 233) and fused in-frame to the engrailed repressor domain (encoding amino acids (aas) 1-298) in CS2+ vector to create Xhox3EnR, or to the VP16 activator domain (aas 410-490), also in CS2+, to create Xhox3VP16 (Fig. 1). An inducible version of the antimorphic construct Xhox3VP16 was subsequently made by in-frame fusion of the glucocorticoid receptor ligand-binding domain (aas 512-777; Kolm and Sive, 1995) 3′ to the VP16 moiety. The sizes of the fusion constructs were tested by in vitro translation using a rabbit reticulocyte lysate system (Promega). Synthesised proteins were labelled with [³⁵S]methionine and analysed by SDS-PAGE and autoradiography.

Capped synthetic mRNA was made using an SP6 MEGAscript in vitro transcription kit (Ambion) with the addition of 5 mM cap analogue (Ambion) and reduction of GTP to 0.5 mM in the reaction. Notch ICD cDNA in CS2+ was cut with NotI and transcribed with SP6 (Coffman et al., 1993). eFGF and Xwnt3a cDNAs in pSP64T were cut with BamHI and transcribed with SP6 (Isaacs et al., 1995; Wolda et al., 1993). Lfng and Xhox3 cDNAs were subcloned into CS2+, cut with NotI and transcribed with SP6 (Wu et al., 1996; Ruiz-i-Altaba and Melton, 1989a). Xhox3VP16 and Xhox3VP16GR were cut with NotI and transcribed with SP6, and Xhox3EnR was cut with SacII, filled using Klenow DNA polymerase and transcribed with SP6.

Embryos were injected with mRNA into all four blastomeres at the 4-cell stage and cultured to stage 9. Animal caps were removed in NAM/2 (0.5× NAM salts with 1 mM NaHCO₃ and 5 mM Hepes, pH 7.5) and cut into 600×100 μm strips for grafting (Fig. 2A). Host embryos at stage 14 were placed in NAM/2 containing 10 μg/ml type IX trypsin (Sigma). The host neural plates were slit with a tungsten needle perpendicular to the anterior-posterior axis and an opening of 600×100 μm was provided for the graft. As the slit widened automatically, no tissue needed to be removed from the host, therefore the host tail bud region is unaffected. Both graft and host were then transferred to NAM/2 containing 20 μg/ml soybean trypsin inhibitor (Sigma), and hosts were placed in agar wells. The grafts were positioned so that the open face of the graft contacted the underlying notochord and presomitic mesoderm of the host, then held in place using small glass bridges and left to heal for 30 minutes. The bridges were then removed and the host embryos cultured individually in NAM/2 overnight at 18°C, then transferred to NAM/10 and cultured for 24 hours at 18°C before fixation.

Ectopic projections formed by the grafts were scored for tail characteristics using the index of Tucker and Slack (1995a). One point is awarded each for the presence of notochord, neural tube and paired somites, two points for a complete fin, one point for tapering and one point for elongation. A total of seven points is given for a normal tail.

Many of the grafts were labelled with nuclear β-galactosidase by incorporating a total of 50 pg lacZ mRNA into the injections (Pownall et al., 1996). X-gal staining could then be used to trace the final position of the graft. Embryos were fixed in MEMFA (4% formaldehyde, 100 mM MOPS, 1 mM MgSO₄.7H₂O, 1 mM EGTA, pH 7.4) for 30 minutes, washed twice for 10 minutes in PBSA, then once in LacZ buffer (10 mM K₃Fe(CN)₆, 10 mM K₄Fe(CN)₆, 2.2 mM citric acid, 156 mM NaHPO₄) for 10 minutes, before staining in LacZ buffer containing 2.7 mg/ml X-gal (Boehringer Mannheim) at 37°C for 15 minutes. Embryos were then washed twice for 10 minutes with PBS containing 0.1% Tween 20 (Sigma) and re-fixed for 30 minutes in MEMFA. These embryos could subsequently be used for in situ hybridisations.

For FDA (fluorescein dextran amine) labelling, host embryos were injected bilaterally at the 2-cell stage with a total of 9.6 nl of dialysed FDA at 50 mg/ml as in Tucker and Slack (1995a).

Whole-mount in situ hybridisations used the method of Harland (1991), with modifications as detailed in Pownall et al. (1996). Digoxygenin-labelled antisense probes for Xbra, Xhox3, Xcad3, lfng, X-shh and X-delta-1 were made as described in Beck and Slack (1998). Probe for Xenopus FGF-8 was made as described in Christen and Slack (1997) and probe for Xwnt3a was made by cutting XP-2 with ClaI and transcribing with T7 as in Wolda et al. (1993).

Histology of embryos and animal caps was as detailed in Godsave et al. (1988). Samples were block-stained overnight in Borax Carmine in 35% ethanol, washed for 4 hours in 70% ethanol, 1% HCl, dehydrated and paraffin-embedded. 8 μm serial sections were cut, rehydrated and counterstained with Picroblueblack, dehydrated and mounted in Depex (BDH). Sections were photographed using a colour CCD camera (Hamamatsu) as in Chalmers and Slack (1998). Images of FDA-labelled sections were captured using a cooled monochrome CCD camera (Hamamatsu).

RNA was made from groups of 30 animal caps by homogenising in 250 μl of 0.1 M NaCl, 50 mM Tris, pH 8.0, 5 mM EDTA and 0.5% SDS, extracting with phenol-chloroform and precipitating with ethanol. RNAse protection assays were performed on 5 μg of RNA (5-10 caps) using the method of Isaacs et al. (1992). Probes were cut and transcribed as follows: Xhox3 was cut with DdeI and transcribed with T7 (230 bp protected; Saha and Grainger, 1992). NCAM was cut with EcoRI and transcribed with SP6 (200 bp protected; Balak et al., 1987). The ubiquitously expressed control probe ODC was made by cutting with BglII and transcribed with T7 polymerase (91-bp protected; Isaacs et al., 1992).

We have previously shown that tail bud outgrowth occurs at the position where there is an overlap in the expression of lfng with X-Notch-1 and X-delta-1 (Beck and Slack, 1998). By analogy with the situation in the Drosophila wing disc, this suggested to us that fringe might be needed for the activation of Notch by Delta and that activation of Notch might be a necessary condition for outgrowth.

We therefore grafted small strips of animal cap tissue expressing a constitutively active form of Notch, Notch ICD (Coffman et al., 1993), into the posterior neural plate of stage-13 host embryos (Fig. 2A). Of 143 cases, 81% of the grafts formed tail-like projections (Fig. 2D, see also Table 1). By contrast, grafts of pieces of animal cap tissue taken from uninjected embryos or embryos injected with β-galactosidase mRNA formed ectopic projections in only 2% of cases (n=94, Fig. 2B, Table 1). Other mRNAs, including eFGF and lfng (Table 1) also gave negative results, showing that the Notch effect is specific. β-galactosidase labelling of the grafts showed that Notch ICD ectopic tails are formed from the part of the graft which comes to lie more posteriorly (Fig. 2E). By contrast, control grafts were simply incorporated into the host neural tube (Fig. 2C). Notch ICD ectopic tails score 5/7 points on the tail index of Tucker and Slack (1995a) as they contain neural tube, fin and loose mesenchyme, and are elongated and somewhat tapered. However the notochord and somites found in a normal tail are always missing (Fig. 2F,G). Graft-derived tails form in the host midline and the ectopic neural tube branches directly from that of the host. Participation of host tissue in the ectopic tails was examined by making Notch ICD grafts to FDA-labelled hosts. This showed that the ectopic tail neural tube and mesenchyme are primarily graft-derived while the epidermis is mainly host-derived (Fig. 2H,I).

Grafts were positioned between 100 and 200 μm anterior of the blastopore slit at stage 13 to avoid physically splitting the M region (the posteriormost neural plate fated to become posterior tail somites; see Tucker and Slack, 1995b). Grafts placed more posteriorly were often seen to generate a split tail, where the host forms two tails with contributions from the graft. As these arise equally from Notch ICD or control animal caps we believe they arise simply by mechanical splitting of the host tail bud forming region.

We have previously shown that Xhox3 is expressed in the distal tip of the tail bud, occupying a region of the posterior wall centered on the overlap of lfng and X-delta-1 expression (Beck and Slack, 1998). Xhox3 is therefore a potential downstream target of Notch activation. Ectopic tails formed by grafts of Notch ICD-expressing tissue also switch on expression of Xhox3 at the distal tip, resembling the expression in normal tails (Fig. 3B). Despite the lack of notochord and somites in the ectopic tails, the Notch ICD tail buds resemble real tail buds in the expression and localisation of the tail bud markers: Xbra, Xwnt3a, X-delta-1, lfng, Xshh, FGF-8 and Xcad3 (Fig. 3). Recent identification of a functional Suppressor of hairless binding site in the ascidian Brachyury promotor suggests that Xbra may also be a direct target of Notch signalling (Corbo et al., 1998). While Xbra is always expressed in ectopic tails, eFGF, which has been shown to activate Xbra during gastrulation (Isaacs et al., 1994), it fails to form an ectopic tail in the grafting assay (Table 1).

The late onset of Xhox3 expression in the distal tip of the tail bud suggests that it may be a result of Notch signalling (Beck and Slack, 1998), and might therefore be an important component of the outgrowth mechanism. We tested this by grafting pieces of animal cap tissue expressing Xhox3 into the posterior neural plate of stage 13 hosts. These grafts generated ectopic tails at high frequency (Fig. 4A, 62%, n=42, Table 1), with identical histology to the Notch ICD tails (not shown), suggesting that expression of Xhox3 can substitute for Notch signalling to enable tail outgrowth.

To further investigate Xhox3 function, two domain-swapped constructs were made. The wild-type Xhox3 was truncated just C-terminal to the homeobox and fused to the activation domain of VP16 to create XhoxVP16 and Xhox3EnR (Fig. 1). The replaced region of Xhox3 maps to the repressor domains in the homologous mammalian Evx and Drosophila evenskipped proteins (Biggin and Tijan, 1989; Johnson and Krasnow, 1992; Han and Manley, 1993; Tenharmsel et al., 1993; Briata et al., 1995, 1997).

The effects of the fusion proteins on tail outgrowth were tested using the same grafting procedure. As expected, Xhox3EnR grafts into the posterior neural plate generated ectopic tails even more effectively than Xhox3 itself (85%, n=20, Table 1), confirming that Xhox3, like other evenskipped family members, functions as a repressor (Fig. 4B). In contrast, Xhox3VP16 grafts did not produce any ectopic tails (n=12, Fig. 4C, Table 1), and Xhox3VP16 also prevented ectopic tail formation when co-injected with Notch ICD (n=12, Fig. 4D, Table 1). These results show that Xhox3 functions downstream of Notch signalling in tail outgrowth. Overexpression of Xhox3 in embryos has previously been shown to reduce anterior structures (Ruiz-i-Altaba and Melton, 1989b). We observe similar effects after injection of more than 100 pg/embryo of either wild-type Xhox3 or Xhox3EnR mRNA (data not shown). Conversely, expression of Xhox3VP16 results in truncations of the posterior, most obvious as a reduction in the number of tail somites from 42 to an average of 31 at stage 40 (Fig. 4F). Loss of eye pigmentation and a lack of neural crest-derived pigment cells are also seen in these embryos. Posterior defects were previously observed by Ruiz-i-Altaba and colleagues, who injected anti-Xhox3 antibodies into embryos, although eye and pigment defects were not noted (Ruiz-i-Altaba et al., 1991). Although these results are consistent with a role for endogenous Xhox3 in tail outgrowth, the tail is not the only region affected, suggesting the additional existence of an earlier role for Xhox3.

In order to confirm the specificity of Xhox3VP16, rescue experiments were conducted by co-injecting different levels of wild-type Xhox3 into whole embryos. Co-expression of very low levels of Xhox3 (Fig. 4G) can completely rescue the phenotype of Xhox3VP16-injected embryos, resulting in normal development of the tadpoles (Fig. 4H). Co-injection of higher doses of Xhox3 mRNA results in progressive loss of anterior structures, as might be expected when it is in excess (not shown). These results make it very likely that Xhox3VP16 acts as a specific activator of target genes that are normally repressed by Xhox3, and can therefore specifically antagonise its action in vivo.

The endogenous Xhox3 expression in the future tail bud becomes visible at about stage 26. However, the phenotype of embryos expressing Xhox3VP16 is suggestive of an additional earlier developmental role. We therefore constructed an inducible form of Xhox3VP16, by the addition of the glucocorticoid receptor DNA binding domain (GR) at the C terminus (Fig. 1). Transcription factors containing the GR are nonfunctional, since after translation they are sequestered in a complex by heat shock proteins (Hollenberg et al., 1993). When the synthetic steroid hormone dexamethasone is added, the transcription factor is released from the complex and can enter the nucleus and act on its target genes.

We have used this inducible system to study the effect of interfering with Xhox3 function during development of the tail. Embryos were injected with 400 pg of Xhox3VP16GR mRNA and then treated with 10 μM dexamethasone at different stages. Addition of dexamethasone to uninjected embryos at stage 14 did not affect development (not shown). Embryos injected with Xhox3VP16GR but not treated with dexamethasone also developed normally (Fig. 4I), as did those treated at stage 35 (Fig. 4J). Addition of dexamethasone at stage 26 resulted in reduced tail development (Fig. 4K), whereas addition at stage 20 resulted in embryos with vestigal tails (Fig. 4L). Eye development is delayed in more severe cases. Earlier treatment with dexamethasone at stage 14 resulted in severe head and tail defects and treatment from the time of injection was lethal (not shown). Taken together with the ability of Xhox3VP16 to block tail formation by Notch ICD, these results show that Xhox3 is indeed required for tail outgrowth, and that it performs this function by acting downstream of Notch signalling.

Generation of an ectopic tail by animal caps filled with Notch ICD only works if the grafts are placed near the late dorsal blastopore lip, but still outside the M region. This was found by placing the grafts at different anteroposterior levels along the neural plate. Posterior grafts were as described previously, anterior grafts were placed just posterior to the head neural folds and trunk grafts were placed midway along the anterior-posterior axis (Fig. 5). The grafts were co-injected with β-galactosidase mRNA to enable visualisation of the final position. Control grafts expressing this marker did not form ectopic tails regardless of position (Fig. 5A-C, Table 2). Notch ICD grafts were only able to form ectopic tails in the posterior position (Fig. 5D-F, Table 2). Together with our observation that it is tissue at the posterior end of the Notch ICD grafts which undergoes outgrowth, this suggests a requirement for a second signal, defining competence for tail outgrowth, which is tightly localised to the posterior neural plate.

We have previously reported that Xwnt3a is expressed in the extreme posterior midline just prior to tail bud outgrowth (Beck and Slack, 1998). Posterior grafts of animal cap tissue expressing very high levels of Xwnt3a only generated ectopic tails in 19% of cases (n=27, Table 1), and never did so if grafted further towards the anterior (n=16, Table 2). As Xwnt3a had only this rather limited ability to provoke tail outgrowth on its own, we considered that it might be a candidate for the factor limiting competence for Notch-induced tail outgrowth to the posterior of the embryo.

To test this, pieces of animal cap expressing both Notch ICD and Xwnt3a were grafted to the three positions along the dorsal midline, and proved to be capable of forming ectopic tails at any point along the anterior posterior axis of the host (Fig. 5G-I, Table 2). In contrast to the Notch ICD tails, which always form from the posterior region of the graft, the cores of these tails were formed by the entire graft, leaving the host’s neural tube unlabelled (Fig. 5G). As with the Notch ICD tails, the core is graft-derived and the epidermis is derived from the host (not shown). Xwnt3a is therefore a strong candidate for the posteriorly localised competence factor for tail outgrowth.

In order to evaluate the activity in isolation of Notch ICD and Xwnt3a, whole animal caps were injected as before and simply cultured to stage 38 for analysis. Uninjected caps or caps injected with β-galactosidase mRNA (50 pg) or Xwnt3a alone (250 pg) develop into atypical epidermis (Fig. 6A,C,F). Caps injected with Notch ICD alone (500 pg) look like atypical epidermis externally (Fig. 6B), but contain smaller, more densely staining cells of unknown identity (Fig. 6G). Animal caps containing both Notch ICD and Xwnt3a are strikingly different; elongation takes place during control neurula stages, and around stage 27 the caps swell and become surrounded by fin, acquiring a tail-like appearance (Fig. 6D). When these were sectioned they were found to contain multiple structures resembling neural tubes (compare Fig. 6I to the transverse section of normal tail at the same stage, Fig. 6E). In about half of the caps the neural tube-like tissue forms a core running along the centre of the long axis of the cap and is surrounded by fin-like mesenchyme and an epidermis (Fig. 6H). In the remaining cases multiple neural tubes are surrounded by condensed neural tissue and mesenchyme (Fig. 6I).

To confirm that the tubes formed by the Notch ICD/Xwnt3a caps were really neural in character, we carried out in situ hybridization for the pan-neural marker NCAM (Fig. 6J-O), and RNAse protection for NCAM and Xhox3. The protections showed that uninjected animal caps or caps injected with either Xwnt3a or Notch ICD alone contained no detectable NCAM or Xhox3 message. In contrast, RNA from caps expressing both contained high levels of NCAM and Xhox3 (Fig. 7).

The results presented here provide evidence for the existence of a molecular pathway controlling tail bud outgrowth in Xenopus (Fig. 8). Previous work has shown that three regions of the dorsal-posterior of the late gastrula must come into alignment at the end of gastrulation in order that a tail bud shall be formed. These regions are known as N (neural plate), M (neural plate posterior to N with a mesodermal fate) and C (caudal dorsal mesoderm). To form a tail bud, N and M must overlie and contact the C region (Tucker and Slack, 1995b; Fig. 8A). We believe that the C region exerts its effect in the early neurula period (Tucker, 1995). Later in development, the junction of N and M forms the tip of the tail bud as it begins to extend.

We have found that X-Notch-1 and X-delta-1 are expressed in the M region from gastrulation onwards. After fusion of the neural folds over the chordoneural hinge these genes are expressed in the lateral blastopore lips, which become the posterior wall, while the N region becomes the dorsal roof. At around stage 26 the N region begins to express lfng (Fig. 8B) and Xwnt3a is expressed in the posterior dorsal part of N. As outgrowth of the tail bud begins, expression of lfng slightly overlaps that of X-Notch-1 in the distal tip of the tail bud. Boundaries between fng+ and fng− cells have been implicated in many mechanisms involving signalling through the Notch protein, but it is the role of these boundaries in appendage outgrowth that has been best studied. In the Drosophila wing disc, fng modulates the interaction of Notch with its ligands to establish a local Notch signal along the boundary of dorsal and ventral compartments, which promotes outgrowth of the wing (Panin et al., 1997). Based on observation of lfng, X-Notch-1 and X-delta-1 expression, the Xenopus tail bud fulfils the basic requirements for the establishment of localised Notch activation in the posterior wall of the tail bud at the time outgrowth begins (Beck and Slack, 1998; Fig. 8D). In this report, we have shown that a small graft expressing the constitutively active cytoplasmic domain of Xenopus Notch and placed in the posterior neural plate can induce ectopic tail outgrowth. The resulting tails contain fin and neural tube but lack the mesodermal components, suggesting that caudal notochord and somites are not formed under the control of Notch signalling. It is likely that these structures are formed from the tail bud chordoneural hinge, as is the case in amniotes (Le Douarin et al., 1998).

The evenskipped homologue Xhox3 satisfies three criteria for being an essential component of the outgrowth mechanism: it is expressed in the forming tail tip; it will provoke outgrowth of a new tail bud in our grafting assay; and a specific antagonist will block formation of the tail in intact embryos. As expected, the domain swap experiments show that Xhox3 functions as a transcriptional repressor, like other eve-related proteins. We have shown that Xhox3 is activated by Notch signalling and that Xhox3VP16 will block the formation of tails provoked by Notch ICD. These experiments show that Notch exerts its activity by induction of Xhox3, and that Xhox3 acts downstream of Notch. Like Notch, evenskipped-type homeobox gene expression is also implicated in outgrowth formation in other organisms. Evx-2 is expresssed in the distal limb bud mesenchyme of mice and the fin of zebrafish and loss of Evx-2 function in mice leads to limb defects (Sordino et al., 1996; Herault et al., 1996), whereas Evx-1 is expressed in the tail bud in both mice and zebrafish as well as the mouse limb bud (Dush and Martin, 1992; Joly et al., 1993; Gofflot et al., 1998). Mice lacking Evx-1 function fail to gastrulate, indicating an early requirement for this gene in development (Spyropoulos and Capecchi, 1994). Our results suggest that localised Xhox3 expression in these areas of outgrowth may also be a result of Notch signalling.

We have shown that Notch ICD can only provoke tail bud outgrowth if the grafts are placed near the posterior end of the neural plate, suggesting a requirement for a posteriorly localised competence factor. Xwnt3a expression is tightly localised in this region from stage 26 and is therefore a good candidate for this factor. Animal cap grafts co-expressing Xwnt3a with Notch ICD can form ectopic tails if grafted anywhere along the anterior-posterior axis, demonstrating that Xwnt3a can act as the posterior competence factor in tail morphogenesis. It is therefore probable that it also fulfils this role in normal development. In support of this, mouse Wnt3a expression is also localised to the tail bud (Takada et al., 1994) and the apical ectodermal ridge of the limb bud (Kengaku et al., 1998). Interactions between Notch and wnt signalling have been observed in other systems and may be antagonistic (Axelrod et al., 1996; Uyttendaele et al., 1996) or synergistic (Neumann and Cohen, 1996). We have shown here that active Notch requires Xwnt3a to turn on Xhox3. However, Xhox3 may also require continued wnt signalling to promote outgrowth of the tail and caudal neural tube (Table 2).

Although the trunk- and tail bud-derived components of the tail merge seamlessly, one difference between the the mode of formation of the axial body structures in the trunk and tail of the embryo concerns the neural tube. In the anterior and trunk, the neural tube of the Xenopus embryo, like that of other vertebrates, is formed by rolling up of the neural plate, which in turn is formed during gastrulation from the ectoderm overlying the dorsal mesoderm. In contrast, the tail bud-derived neural tube is formed directly by cavitation in avian and mammalian embryos, a process termed secondary neurulation (Nievelstein et al., 1993; Catala et al., 1995; Le Dourain et al., 1998). However, amphibians are often said to be excluded from this since the condensation and cavitation occur simultaneously and the tail neural tube appears to arise as an extension of the trunk neural tube (e.g. Gilbert, 1997). The results presented here show that secondary neurulation can occur in amphibians, since direct formation of neural tubes in caps expressing Xwnt3a and Notch ICD occurs in the absence of any existing neural tube or dorsal mesoderm. These animal caps may therefore represent a simple model of the secondary neurulation process in vertebrates. In support of this, loss of wnt3a, lfng or Notch-1 function in mice has been shown to lead to posterior defects (Takada et al., 1994; Conlon et al., 1995; Evrard et al., 1998; Zhang and Gridley, 1998). This model system may be of particular interest if it turns out that some defects of the posterior spinal cord in higher vertebrates are due to faults in the Notch signalling pathway.

Thanks to Chris Kintner, Dan Kessler, Les Dale and Marko Horb for their kind donations of plasmids used in this study. We thank Harv Isaacs, Betsy Pownall and Rod Scott for their helpful comments on the manuscript. This work was supported by the Wellcome Trust, Programme Grant No. 43192.