Notch activates sonic hedgehog and both are involved in the specification of dorsal midline cell-fates in Xenopus

The floor plate is an epithelial structure located on the ventral midline of the vertebrate neural tube and constitutes an important source of signals involved in the induction of motor neurons and axonal pathfinding(Tanabe and Jessell, 1996; Colamarino and Tessier-Lavigne,1995; Stoeckli and Landmesser,1998). Two cell populations have been described as contributing to this structure: one in the midline, the medial floor plate (MFP), which expresses sonic hedgehog (shh) and hnf3β, and the other one, the lateral floor plate (LFP), which expresses hnf3β and flanks the medial cells (Odenthal et al.,2000). Studies on the avian embryo showed that the floor plate does not develop after removal of the caudal notochord but appears ectopically after grafting notochordal tissue onto the lateral or dorsal regions of the neural tube. Therefore, it was concluded that the floor plate is induced by the notochord (Placzek et al.,2000). This scenario, reinforced by work with other vertebrate models, considers Shh as the inducer(Tanabe and Jessell, 1996) and this is supported by the finding that mice lacking shh function fail to develop the floor plate (Chiang et al.,1996). It was proposed from over-expression experiments in frog embryos that secreted Shh from the notochord promotes Gli1 expression in the midline of the developing neural plate. Gli1, in turn, activates the winged-helix transcription factor hnf3β, or the related gene pintallavis (plvs) in Xenopus, which then activates shh in floor plate precursors(Lee et al., 1997). Finally,Shh induces motor neurons in the neighbouring ventral neural tube(Echelard et al., 1993; Krauss et al., 1993; Roelink et al., 1994; Roelink et al., 1995; Marti et al., 1995; Ericson et al., 1996; Lee et al., 1997).

Careful experiments revisited those studies performed on birds and found that the reason why floor plate did not develop after the removal of the notochord was that the floor plate precursors were removed also, since Hensen's node (equivalent to the amphibian's Spemann's organiser) generates both midline structures (Catala et al.,1996; Teillet et al.,1998; Le Douarin and Halpern,2000). This agrees with several studies in vertebrates including chicken, mouse, fish and amphibians, which show that precursors of the floor plate, notochord and dorsal endoderm originate within the organiser(Spemann and Mangold, 1924; Selleck and Stern, 1991; Catala et al., 1995; Catala et al., 1996; Wilson and Beddington, 1996; Shih and Fraser, 1995; Melby et al., 1996; Amacher et al., 2002; Latimer et al., 2002). Therefore, if the notochord and the floor plate share the same embryonic origin, it will be necessary to reconcile this scenario with the role and the hierarchy of the molecules described above. This is complicated by the findings that in zebrafish embryos, only the development of the LFP seems to be Shh dependent, while the medial cells appear to be dependent on Nodal activity (Rebagliati et al.,1998; Sampath et al.,1998; Schauerte et al.,1998; Odenthal et al.,2000).

Notch signalling is best known from its central role in lateral inhibition during neurogenesis. Within a proneural cluster, the future neuron is the source of the membrane-bound ligand Delta, which interacts with the receptor Notch on the surface of the neighbouring cells. The receptor is cleaved to render the intracellular domain (Notch^ICD) that enters the nucleus and, in association with CSL intracellular transducers such as Su(H),activates the transcription of target genes that lead to the suppression of the neuronal fate in the cells surrounding the neuronal precursor(Schroeter et al., 1998; Bray, 1998; Wolfe and Haas,2001; Kopan and Goate,2002).

Different models provided evidence that Presenilins facilitate the Notch signalling pathway. Most authors favour the idea that their γ-secretase activity, which mediates the proteolytic cleavage of the amyloid precursor protein (APP), is also involved in the final proteolytic step that cleaves the Notch receptor to produce Notch^ICD(De Strooper et al., 1999; Struhl and Greenwald, 1999, Struhl and Greenwald, 2001; Chan and Jan., 1999; Taniguchi et al., 2002). It was also shown that a γ-secretase-independent mechanism may play a partial role in Notch signal transduction(Berezovska et al., 2000; Berechid et al., 2002) and other research suggested that Presenilin is required before the generation of Notch^ICD (Ye et al.,1999; Ray et al.,1999).

We have previously shown that both shh and presenilinrepress primary neurogenesis in Xenopus laevis embryos(Franco et al., 1999; Paganelli et al., 2001). Since presenilin stimulates shh expression in the floor plate, we suggested that the effects of presenilin could be exerted through shh. We also proposed that in the primary neurogenesis cascade, shh acts very upstream of the lateral inhibition step mediated by Notch, modulating the activity of prepattern genes. Bearing in mind the relationship between Notch and Presenilin, and because lateral inhibition does not account for the repression of primary neuron development after presenilin overexpression, we wondered whether Notch signalling could modulate shh expression.

To answer this question, we either activated or prevented Notch signalling in Xenopus embryos and found that Notch stimulates shh and plvs expression in the floor plate and represses the notochordal markers chordin (chd) and brachyury (bra). These changes are accompanied by an expansion of the floor plate and a reduction of the notochord size. We propose that Notch may execute a binary decision, favouring floor plate development at the expense of the notochord, and this preferentially occurs before mid gastrula. We also show that shhdown-regulates chd and suggest that shh itself may be involved in reinforcing the binary decision executed by Notch. We present evidence that Presenilin can also modulate this switch in a Notch-dependent way.

Albino Xenopus laevis embryos were obtained using standard methods(Ruiz i Altaba, 1993), staged according to Nieuwkoop and Faber(Nieuwkoop and Faber, 1994)and fixed with MEMFA (Harland,1991).

Synthetic capped mRNAs for microinjection were obtained as described previously (Franco et al.,1999). To direct the material to the future dorsal midline populations, injections were delivered into the animal hemisphere of one blastomere at the 2-cell stage at approximately 30-40° from the equator and close to the cleavage plane, following the observations of Vodicka and Gerhart (Vodicka and Gerhart,1995). The Xotch antisense oligodeoxynucleotide (Xotch Mo) used was a 25-mer morpholino oligo (Gene Tools, LLC) with the base composition 5′-GCACAGCCAGCCCTATCCGATCCAT-3′. The X-ps-α antisense morpholino oligonucleotide (X-ps-α Mo), the standard control morpholino oligo (Control Mo) and the X-ps-α construction used for in-vitro transcription were the same as those employed by Paganelli et al.(Paganelli et al., 2001). The templates for mRNA synthesis has been described previously: X-shh(Franco et al., 1999); notch^ICD_,hGR/ICD22 and X-su(H)^DBM (Wettstein et al., 1997). Nuclear translocation of hGR/ICD22 was induced with dexamethasone (Sigma, D4902).

For RNA interference of shh function, a deletion was made in the construct used for making X-shh antisense probes described by Franco et al. (Franco et al., 1999)by digestion with Eco0109 I and BstEII, followed by fill-in and ligation. The resulting construct contains a 935 bp insert encoding the N-terminal region of X-Shh, from bp 51 to bp 986 of the cDNA sequence(Ekker et al., 1995). Sense and antisense RNA were obtained after linearisation with EcoRI or KpnI and transcription with T7 or T3 mRNA polymerase, respectively. 1μg of each template was used for in vitro transcription, which was carried out with the Megascript kit (Ambion). Synthetic RNAs were purified with the Qiagen RNeasy mini kit. Equimolar amounts of sense and antisense RNAs were annealed in 1 mM MgSO₄, 30 mM NaCl at 70°C for 15 minutes and then at 37°C for 45 minutes. The quality of double strand RNA (ds-RNA) was tested by native agarose gel electrophoresis in TBE in the presence of ethidium bromide. Gel mobility was shifted according to ds-RNA of the expected length.

The amounts of synthetic mRNAs, ds-RNA and morpholinos injected are indicated in the table and figures. All injections included 0.5 ng of nuc-lacZ mRNA as tracer. For comparison, equal amounts of nuc-lacZ mRNA were delivered in each set of experiments.

Semi-quantitative RT-PCR analysis was performed essentially as described by Paganelli et al. (Paganelli et al.,2001). The number of cycles and the template input for PCR were determined empirically in each case, within the linear range of amplification. The forward (F) and reverse (R) primer sequences, the product sizes, and the number of cycles were as follows: ef1α: F 5′-CAGATTGGTCCTGGATATGC-3′, R 5′-ACTGCCTTGATGACTCCTAG-3′, 268 bp., 26 cycles for stage 12, 25 cycles for stage 15; X-shh: F 5′-ATGCTGGTTGCGACTC-3′; R 5′-CCCGCCAGACTTGG-3′, 581 bp., 36 cycles for stage 12, 32 cycles for stage 15.

X-gal staining, preparation of digoxigenin-labeled antisense RNA probes and whole-mount in situ hybridisation were performed as described previously(Franco et al., 1999), except that the proteinase K step was omitted in in situ hybridisation when embryos would be further processed for immunohistochemistry.

50 μm sections were cut using an Oxford Vibratome and mounted onto gelatine coated slides as described by Hollemann et al.(Hollemann et al., 1996),except that 4% formaldehyde was used instead of glutaraldehyde during the embedding, in order to reduce background fluorescence. For immunolocalization of the C-myc epitope, slides were washed three times in PBS, for 10 minutes each, incubated for 1 hour at room temperature in blocking buffer containing 5% nonfat milk (Molico, Nestlé) in PBS, then overnight at 4°C with mouse 9E10 anti-Myc monoclonal antibody (Santa Cruz) diluted 1/500 in blocking buffer, in a wet chamber. The following day, slides were washed three times at room temperature with PBS containing 0.1% Tween 20 (10 minutes each), twice with PBS (10 minutes each), and incubated for 2 hours at room temperature in a wet chamber in the dark with anti-mouse immunoglobulins-FITC (Dako F0232)diluted 1/200 in blocking buffer. After washing with TBS (pH 7.5) containing 0.1% Tween 20, slides were mounted in PBS:glycerol (1:2).

To investigate whether Notch signalling can modulate shhexpression, we first examined the effects of bilateral injections of notch^ICD mRNA, which encodes a constitutively active form of the receptor independent of ligand binding, and analysed the expression of shh by semiquantitative RT-PCR. Active Notch significantly increased the levels of shh transcripts, and this was evident both at late gastrula and at neurula stages (Fig. 1A). Next, we wanted to know whether changes in the spatial distribution of shh transcripts could account for this up-regulation. For this purpose, we examined the effects of unilateral injections of notch^ICD by in situ hybridisation.

Early neurulae showed that notch^ICD enhances shh expression on the injected side (83% of injected embryos, n=24; Fig. 1C). This was evident as an increase in the density of shh-positive cells and/or a lateral expansion of the domain. As with presenilininjections, we did not observe ectopic expression of shh in other regions of the embryo.

To corroborate whether endogenous Notch activity was indeed involved in this modulation, we prevented Notch signalling by injecting X-su(H)^DBM mRNA, which encodes a dominant-negative variant of the Notch transducer X-Su(H). Whole embryos at the neural plate stage showed that shh transcripts were down-regulated (82%, n=90; Fig. 1D).

Since shh mRNA is present both in the prospective floor plate and notochord at this stage, we wanted to know which structures were affected. For this purpose, we analysed transverse sections of more advanced neurulae, when the notochord and floor plate are more clearly distinguished from each other. Control embryos show shh transcripts in the floor plate and a dorsal-ventral gradient in the notochord, with highest levels close to the floor plate (Fig. 1B′). Embryos injected with notch^ICD revealed an expansion of the shh domain corresponding to the floor plate and a concomitant reduction of the notochord size on the injected side(Fig. 1C′,C″). Conversely, X-su(H)^DBM-injected embryos showed a decrease in shh transcripts in the floor plate domain, while the notochord size was simultaneously augmented on the injected side(Fig. 1D′,D″).

The Notch-induced up-regulation of shh in the floor plate domain may happen by a direct or indirect regulation and/or by favouring the development of the floor plate. Therefore, we studied the consequences of activating the Notch pathway on the expression of another gene that was described to participate in floor plate development. Plvs mRNA precedes and overlaps shh expression and is first detected in the dorsal marginal zone of late blastulae. During gastrulation it is expressed at the dorsal midline by cells that undergo convergent-extension movements. At the early neurula stage, transcripts are distributed throughout the dorsal midline in the three germ layers, i.e. the prospective floor plate, the notochord and the dorsal endodermal cells lining the archenteron(Ruiz i Altaba and Jessell,1992) (Fig. 1E,E′).

Whole embryos at the neural plate stage show that plvs expression also increases on the injected side when the Notch pathway is stimulated (75%, n=16; Fig. 1F) and transverse sections confirm that this occurs in the floor plate domain, which appears expanded (Fig. 1F′,F″). Conversely, X-su(H)^DBM-injected embryos showed a decrease in plvs expression (95%, n=19; Fig. 1H). Thus, Notch signalling positively regulates the expression of two genes that were shown to be able of promoting floor plate development.

To further study the effect of Notch signalling on notochord development,we analysed whether activating or blocking the pathway could change the expression of the mRNA encoding the secreted polypeptide Chd. In early neurulae, chd transcripts are normally present in the notochord and the prechordal mesoderm (Sasai et al.,1994) (Fig. 2A,D). At this stage we found that notch^ICD strongly reduced chd expression on the injected side (66%, n=35; Fig. 2B). Transverse sections of slightly more advanced embryos again showed that the notochord size, now visualised by the expression of chd, was reduced on the injected side, while the overlying layer containing the prospective floor plate was concomitantly thickened (Fig. 2E). In contrast, when Notch signalling was interfered with by injecting X-su(H)^DBM or an antisense morpholino oligonucleotide against Xenopus notch-1 (Xotch Mo), chd expression was increased, appearing as a more dense and superficial staining in whole embryos (63%, n=49 for X-su(H)^DBM; 75%, n=16 for Xotch Mo; Fig. 2C,I). Transverse sections revealed that this effect was the result of an increase of chd-positive cells, which adopted more dorsal positions, as if they were occupying the normal place of floor plate cells (arrow, Fig. 2F).

Therefore, Notch signalling inhibits chd expression and this may occur through a direct or indirect regulation and/or by inhibiting the development of the dorsal axial mesoderm. Thus, we studied the effects of activating the Notch pathway on the expression of bra, which is necessary for notochord development(Chesley, 1935; Halpern et al., 1993; Conlon et al., 1996; Smith, 1997).

In Xenopus, transcription of bra begins at mid-blastula transition, but strongest levels are achieved when gastrulation starts. At this stage, transcripts are distributed throughout the entire marginal zone(presumptive mesodermal cells). During gastrulation, mesodermal cells that migrate anteriorly turn-off bra expression with the exception of the notochord. During neurula stages, transcripts persist only in the notochord and in a circumblastoporal ring (Smith et al., 1991) (Fig. 2J,L).

Neurula stage embryos showed that notch^ICD strongly reduced the notochordal expression of bra (86%, n=42; Fig. 2K,M). We conclude that activation of the Notch pathway down-regulates the expression of two notochordal markers, the transcription factor Bra and the secreted polypeptide Chd, and concomitantly decreases the notochord size; at the same time, Notch up-regulates two molecules expressed by the floor plate, the transcription factor Plvs and the secreted factor Shh, and increases the floor plate size. All this evidence suggests that a binary choice determines between floor plate and notochordal fates and when Notch signalling is active, favours floor plate development at the expense of the notochord.

To determine the time of highest competence for the proposed binary decision in response to Notch signalling, we performed time-course experiments by injecting hGR/ICD22 mRNA, a construction encoding the ligand-binding domain of the human glucocorticoid receptor fused to the amino termini of Notch^ICD. Thus, nuclear translocation of Notch^ICD can be induced upon dexamethasone administration(Wettstein et al., 1997). Control embryos were injected with the construction but left untreated. Two windows of induction with 10 μM dexamethasone were assayed: one that included the first half of gastrulation (from stage 9+ to stage 11) and the other included the second half (from stage 11 to stage 13). In situ hybridisation revealed that in response to Notch signalling, nearly three-fold more embryos were able to up-regulate shh before stage 11 than after this stage. Meanwhile, chd down-regulation occurred in nearly two-fold more embryos before stage 11 than after this stage(Table 1). These results suggest that the Notch-mediated binary decision that chooses floor plate fate in preference to the notochord for the trunk region mainly takes place around early gastrulation. To further test this hypothesis, we analysed whether the effect of activating or blocking Notch signalling on chd expression could also be detected at early gastrula. We chose chd instead of shh because of its stronger and broader expression in the early organiser (see below), so that unambiguous differences in transcripts levels between the injected and non-injected side could be distinguished. We found that notch^ICD decreased chd transcripts in the organiser on the injected side (90%, n=21; Fig. 3B; see also Fig. 5B,C), while X-su(H)^DBM produced the opposite effect (63%, n=49; Fig. 3C).

Thus, the down-regulation of chd by Notch signalling is already initiated at early gastrula, when the competence of chd and shh for responding to active Notch is highest.

It was reported that Xenopus notch-1 (Xotch) transcripts are present in domains of primary neurogenesis from late gastrula stages(Chitnis et al., 1995). RNAse protection assays revealed maternal and zygotic Xotch mRNA and its presence in the three germ layers, especially in the dorsal mesoderm, when gastrulation starts (Coffman et al.,1990). A more precise study of the early distribution would help to understand whether Xotch transcripts are present at the right time and place to elicit the effects suggested above. Therefore, we performed in situ hybridisation of early and late gastrula/early neurula stages and attempted to correlate Xotch mRNA distribution with the expression patterns of chd and shh.

At early gastrula, Xotch transcripts are present both in the involuting and non-involuting dorsal marginal zone(Fig. 4A) and extend towards the animal pole but are absent from the ventral marginal zone and the vegetal yolk mass (not shown). Sagittal sections reveal Xotch transcripts in the epithelial and subepithelial layers of the dorsal blastopore lip(Fig. 4D), where chdis strongly expressed (Fig. 4B,E). Shh transcripts are first detected at this time(Fig. 4C), but are delayed in relation to chd, which began to be expressed in the dorsal marginal zone shortly before gastrulation started(Sasai et al., 1994). It is noticeable that shh expression is more confined than chd and Xotch: transcripts are mainly found in several cells in the subepithelial layer of the organiser, with some faint distribution in the epithelial layer (Fig. 4F). Neither Xotch nor shh mRNA are detected in the deep zone of the organiser, which mostly contains the involuted precursors of the prechordal mesoderm and expresses chd. Therefore, when gastrulation starts, a population of cells containing notochord precursors expresses the three markers analysed.

When gastrulation ends, apart from the known expression in the presomitic mesoderm and neural ectoderm, Xotch transcripts are detected in floor plate precursors, with higher levels in the posterior region, and are absent from the notochord (Fig. 4G,J). Around this time, chd mRNA is restricted to the developing notochord(Fig. 4H,K) and shhexpression is evident in the three dorsal midline layers, i.e. the prospective floor plate, the notochord and the lining of the archenteron(Fig. 4I,L). Across these populations, a dorsal to ventral gradient of shh expression is evident, which later correlates with the higher levels of expression found in the floor plate and in the dorsal region of the notochord(Fig. 1B′).

Thus, at early gastrula, Xotch transcripts are present in a population of cells in the organiser which is known to contain notochord precursors and co-express chd and shh, giving support to the idea that Notch signalling may be involved in the regulation of these genes and in cell-fate decisions during the specification of the dorsal midline structures. Later on, among the dorsal midline layers, Xotchtranscripts are only present in the developing floor plate, suggesting that notochord cells in the trunk can no longer respond to Notch ligands to divert towards floor plate fate.

Early gastrulae injected with notch^ICD showed that the patches of X-gal staining coincided with the territories of chdrepression but in some embryos, repression of chd was also seen towards more lateral regions at a distance from the X-gal patch(Fig. 5B,C). Because X-gal staining could be underestimating the cells harbouring the injected notch^ICD mRNA, we made use of the c-Myc epitope fused to the Notch^ICD fragment encoded by the synthetic mRNA that we injected in our experiments. Thus, we were able to identify more precisely those cells that inherited and translated the notch^ICDmRNA by immunofluorescence for the Myc-tag and compared their localisation with the expression of chd.

Sagittal sections of early gastrulae showed down-regulation of chdexpression in patches of Myc-tag-positive cells (compare Fig. 5E,F with 5D). In addition, chd down-regulation was also observed in Myc-tag-negative cells that were surrounded by Myc-tag-positive cells (compare Fig. 5H,I with 5G). Transverse sections of early neurulae also revealed Myc-tag-positive cells displaying a strong down-regulation of chd intermingled with Myc-tag-negative cells that did express chd, although in much lower levels than cells on the non-injected side (Fig. 5J-M). Therefore, since chd expression was repressed in the same cells that inherited the notch^ICD mRNA and also in neighbouring cells, our results suggest that active Notch has the ability to down-regulate chd expression in both cell-autonomous and non-cell-autonomous ways. Transverse sections of more advanced neurulae showed that Myc-tag-positive cells were chd negative and rarely contributed to the notochord; instead, they were located in the overlying layer containing the floor plate (Fig. 5N,O; see also Fig. 2E), which appeared thickened, while the notochord was concomitantly reduced in size and chd-positive cells were decreased on the injected side as shown before, indicating that cells that inherited notch^ICD mRNA diverted their notochord fate and became incorporated into the floor plate. This developmental series suggests that cells within the organiser are able to repress chd expression in response to active Notch and later segregate to the prospective floor plate.

Bearing in mind the preceding results and the proposed role of Shh as inducer of floor plate development, we wanted to analyse whether an enhancement of Shh signalling could contribute to reduce the number of chd-positive cells. Gastrulae unilaterally injected with X-shh mRNA showed a decrease in chd expression in the early organiser (83%, n=36; Fig. 3D), in a manner reminiscent to that obtained after Notch activation (Fig. 3B). Next, we wanted to know what happened if we interfered with shh function. Double-stranded RNA has been successfully used as a potent and specific reagent for silencing different genes in Xenopus embryos(Oelgeschläger et al.,2000; Nakano et al.,2000; Lau et al.,2001; Zhou et al.,2002). When we specifically degraded the endogenous transcripts with X-shh double-stranded RNA (X-shh-ds) (compare Fig. 3G,G′ with F,F′), we observed a significant increase of chd-positive cells on the injected side (44%, n=43; Fig. 3E), resembling the effect of blocking Notch signalling with X-su(H)^DBM(Fig. 3C). Because the X-gal staining was not reduced in X-shh-ds-injected embryos when compared with those injected with nuc-lacZ mRNA alone (compare Fig. 3E with 3A), we conclude that the phenotype that we observe is not due to non-specific mRNA degradation or to non-specific interference of protein translation. To verify the specificity of the effects produced by X-shh-ds, we analysed the expression of N-tubulin and gli3, which were shown to be down-regulated by X-shh overexpression(Franco et al., 1999). We also examined the distance between the optic vesicles in tadpoles, since targeted disruption of the mouse shh gene leads to cyclopia(Chiang et al., 1996). As expected, X-shh-ds increased the number of primary neurons (55%, n=38; Fig. 3H),expanded the gli3 domain (55%, n=38; Fig. 3I) and produced several grades of cyclopia (80%, n=35; Fig. 3J). Moreover, another hint that the action of X-shh-ds RNA is specific is that it does not deplete other unrelated endogenous transcripts, for example N-tubulinor even chd (notice in Fig. 3H that the X-gal staining is extensively distributed on the injected side of the embryo, indicating that X-shh-ds RNA, although present in the domains of N-tubulin expression, does not promote the degradation of N-tubulin mRNA; for chd see Fig. 3E). Thus, Shh signalling specifically down-regulates chd and this effect is already evident when the proposed switch controlling notochordal and floor plate fates modulated by Notch mainly takes place. This correlation suggests that Shh signalling may be in part mediating the effect of Notch in the specification of the different cell populations that configure the dorsal midline.

Because Presenilins have been implicated in Notch signalling, we wanted to know whether Notch mediates the up-regulation of shh promoted by presenilin. Therefore, we unilaterally co-injected Xenopusembryos with presenilin-α (X-ps-α) and X-su(H)^DBM mRNAs and analysed the expression of shh at neurula stages. When injected alone, X-ps-αexpands shh expression on the injected side (84%, n=19; Fig. 6B), as we have previously described (Paganelli et al.,2001). Transverse sections show a lateral expansion of the floor plate domain and a concomitant reduction of the notochord domain on the injected side (Fig. 6E),resembling the consequences of Notch activation. When Notch signalling was prevented by co-injection of X-su(H)^DBM (i.e. down-stream of the processing step of the receptor where Presenilin is thought to intervene), X-ps-α was unable to up-regulate shh. Moreover, the effect of blocking the Notch pathway prevailed, and we observed a reduction of shh in the floor plate domain (82%, n=34; Fig. 6C,F).

If Presenilin affects shh expression and the development of the floor plate and the notochord by facilitating Notch signalling, then, chd expression should be impaired by presenilin in the same way as after Notch activation. A set of experiments was performed to test this hypothesis. Over-expression of X-ps-α resulted in a down-regulation of chd (41%, n=39; Fig. 6H,K). Conversely,blocking X-ps-α translation with a specific antisense morpholino oligonucleotide had the opposite effect, and in these X-Ps-α-depleted embryos chd expression appeared as a more dense and superficial staining than in embryos injected with a control morpholino(80%, n=5; compare Fig. 6N,P with M,O), similar to the effect observed after blocking Notch signalling with X-su(H)^DBM. When Notch signalling was prevented by co-injection of X-su(H)^DBM, X-ps-α was unable to down-regulate chd, and the effect of blocking Notch signalling prevailed again: chd was up-regulated and transverse sections showed a higher density of chd-positive cells on the injected side (57%, n=23; Fig. 6I,L).

Together, these results suggest that Notch signalling requires Presenilin activity to modulate the binary switch that decides between notochord and floor plate fates.

Although substantial evidence that Presenilins facilitate Notch signalling has been collected, we found that Notch-mediated lateral inhibition was probably not responsible for the repression of primary neurogenesis after presenilin overexpression in Xenopus(Paganelli et al., 2001). However, since presenilin was able to up-regulate shhexpression, which also represses primary neurogenesis(Franco et al., 1999), we considered the possibility that Notch could be participating earlier in the cascade, modulating shh expression. We confirm this hypothesis and also present evidence that the Notch pathway is involved in cell fate decisions during the specification of the dorsal midline.

Our data show that Notch signalling promotes the expression of shhand plvs, two markers of floor plate specification, and together expands the floor plate. At the same time it represses the expression of both notochordal markers examined, chd and bra, while concomitantly reduces the notochord size.

Grafting and ablation experiments in birds have shown that the notochord and the floor plate derive from the Hensen's node during primary neurulation,and from the cordoneural hinge (the remains of Hensen's node) during secondary neurulation [(Catala et al.,1995; Catala et al.,1996; Teillet et al.,1998) for revision see Le Douarin and Halpern(Le Douarin and Halpern,2000)]. Lineage tracing revealed that the Xenopus late organiser, which is the equivalent of the avian cordoneural hinge, also originates both structures during tail formation(Gont et al., 1993) and the pioneer work of Spemann and Mangold in amphibians clearly demonstrated that the implanted dorsal lip differentiates into notochord and floor plate in the trunk [see figure 25 in Spemann and Mangold(Spemann and Mangold, 1924)]. When the cellular origin of the Xenopus organiser was traced in the 32-cell embryo, the progeny of the B1 blastomere gave rise to 70% of the organiser, and descendants were found both in the notochord and the floor plate [see figure 6 in Vodicka and Gerhart(Vodicka and Gerhart, 1995)]. Fate maps of the embryonic shield, the teleost equivalent of the Spemann's organiser, also established that this region contributes to both structures(Shih et al., 1995; Melby et al.,1996; Amacher et al.,2002; Latimer et al.,2002). Therefore, if the floor plate and the notochord derive from a common cell population, our results indicate that Notch may be executing a binary cell-fate decision: when active, it promotes floor plate specification at the expense of the notochord. This is consistent with other findings in zebrafish embryos: missense mutants of the Notch ligand delta A (dlA)develop excess notochord and reduced numbers of floor plate and hypochord cells, while overexpression of dlA leads to the opposite effect(Appel et al., 1999).

For the trunk region, the proposed binary decision appears to take place mainly around the beginning of gastrulation, since the effect of activating or blocking Notch signalling on chd expression is already evident at early gastrula and the competence of chd and shh to respond to active Notch is highest from stage 9+ to stage 11. Therefore, at least two parallel mechanisms seem to contribute to stop this Notch- mediated cell-fate switch in the dorsal midline precursors while gastrulation proceeds:first, Xotch transcripts ultimately disappear from the developing notochord, suggesting that these cells become refractory to divert to floor plate in response to Notch ligands; second, the competence of the binary switch for responding to active Notch decreases throughout gastrulation. Interestingly, X-delta-1 transcripts are present in the dorsal blastopore lip at stage 10.5 (Ma et al.,1996) and then disappear from the involuted cells(Wittenberger et al., 1999). Functional experiments should elucidate whether X-Delta-1 is the ligand that operates the proposed binary switch, but its down-regulation in the involuted cells may underlie another key component in the mechanisms that contribute to stop Notch signalling in the dorsal midline.

Besides the previously reported up-regulation of shh, we show here that presenilin also down-regulates chd. Because both activities were prevented when the transduction of the Notch pathway was impeded, we conclude that Presenilin is required by Notch signalling during the binary switch that specifies dorsal midline fates. X-ps-αexpression appears to be ubiquitous, but transcripts are present at the right time to modulate Notch signalling during dorsal midline development(Tsujimura et al., 1997).

Evidence from several sources suggest that Notch activation may trigger a cascade linking shh and plvs, which ultimately results in favouring floor plate development. First, notch^ICDinjection increases both transcripts in floor plate precursors (this paper). Second, their expression domains overlap, but plvs precedes shh, beginning at late blastula(Ruiz i Altaba and Jessell,1992) (this work). After neural tube closure plvs mRNA is replaced by transcripts from the closely related gene hnf3β, and it was proposed that the combined expression of both transcription factors in Xenopus is equivalent to that of hnf3β in rats and mice(Ruiz i Altaba et al., 1993a). Third, functional correlation from overexpression experiments in frog embryos associates plvs/hnf3β and shh with the specification of floor plate fate: hnf3β promotes the ectopic expression of floor plate markers including shh and hnf3β itself, in the neural tube; plvs promotes the ectopic expression of F-spondin, a marker of differentiated floor plate, in the dorsal neural tube, and shh induces the ectopic expression of F-spondin,hnf3β and shh itself(Ruiz i Altaba et al., 1993b; Ruiz i Altaba et al., 1995; Roelink et al., 1994). It could be argued that in these studies, the ectopic expression of floor plate markers within the neural ectoderm was obtained at tadpole stages and could not be detected earlier, when floor plate is normally specified. Furthermore,in Xenopus embryos the regulatory relationship between plvsand shh was not directly tested and the ability of hnf3β/plvs and shh to promote floor plate development was not analysed in the context of their normal functional domain. Therefore, functional experiments in this context should be carried out to test the hierarchy of the relationship between shh and plvsupon Notch activation.

Evidence from other vertebrate models also indicate that shh and hnf3β are functionally linked. Mice lacking shhactivity did not develop a distinct floor plate despite the presence of a differentiated notochord during early stages, and hnf3βexpression in the ventral neural tube was never initiated(Chiang et al., 1996). However, mice homozygous for targeted mutations of the hnf3βgene failed to develop the notochord and the floor plate and lacked shh expression, but the severe impairment in the development of the node and its derivatives did not allow us to determine whether hnf3β is directly required for shh expression(Ang and Rossant, 1994; Weinstein et al., 1994). However, HNF3-binding sites have been found in the promoter and other regulatory regions of the mouse and zebrafish shh gene, including intronic enhancers that direct the expression in floor plate and notochord,and it was suggested that shh expression in both structures is regulated by HNF3-dependent and independent mechanisms(Chang et al., 1997; Müller et al., 1999; Epstein et al., 1999). However, the enhancer that directs hnf3β expression in floor plate cells in mice contains a Gli binding site, which was proposed to respond to Shh signalling (Sasaki et al.,1997). Thus, evidence collected from mice suggest that Shh protein from the notochord induces HNF3β during specification of the floor plate and HNF3β in turn activates shh expression in floor plate cells. Analysis in zebrafish support the idea that shh is a target of HNF3β, nevertheless, it was proposed that, unlike floor plate development in the mouse, zebrafish embryos employ two distinct mechanisms for floor plate specification, one dependent on Nodal activity, which induces MFP, and the other dependent on Shh, which induces LFP(Schauerte et al., 1998; Odenthal et al., 2000).

In Xenopus, chd is able to promote notochord development in mesodermalised animal caps and in u.v.-ventralised embryos, both in cell-autonomous and non cell-autonomous ways(Sasai et al., 1994). Chd is a potent antagonist of ventralising BMPs(Sasai et al., 1995; Piccolo et al., 1996), and it was suggested that notochord formation requires co-repression of both BMP and Wnt signalling (Yasuo and Lemaire,2001). Inactivation of Chd protein by the metalloprotease Xolloid leads to strong ventralised phenotypes up to neurula stages, and later the notochord is frequently absent (Piccolo et al., 1997). Lack of Chd activity in zebrafish embryos disrupts posterior notochord development, and Ntl (Bra) protein is absent from the posterior notochord (Hammerschmidt et al.,1996; Schulte-Merker et al.,1997). Thus, Chd has been mainly regarded as an inhibitor of ventralising signals during mesodermal patterning and its requirement for notochord development may reflect this fact.

Experiments conducted in Xenopus embryos showed that brais one of the direct targets of mesoderm-inducing factors and promotes development of posterior mesoderm in ectodermal explants(Smith et al., 1991). The dorsal-ventral character of this mesoderm depends on braconcentration, the highest dose being able to promote somitic muscle but never notochord formation (Cunliffe and Smith,1992; Cunliffe and Smith,1994), and hence, it is unable to induce chd in this kind of explants (Taira et al.,1997). However, when co-expressed with either plvs or the BMP antagonist noggin (which is also present in dorsal mesoderm), bra can promote notochord development(Cunliffe and Smith, 1994; O'Reilly et al., 1995). Besides, lack of bra function in mouse and zebrafish or changing its behaviour from transcriptional activator to repressor in Xenopusresults in the absence of posterior mesoderm and failure of notochord differentiation (Chesley,1935; Halpern et al.,1993; Conlon et al.,1996). It appears therefore that the transcription factor Bra is necessary but not sufficient in the pathway that leads to notochord differentiation, and other molecules such as Plvs and BMP antagonists may cooperate in this process. In this context, it will be interesting to test whether the combined action of bra and the BMP antagonist chd is sufficient to promote notochord formation. In conclusion, our results show that Notch signalling down-regulates the expression of two molecules required for notochord development.

Notably, it was demonstrated that the floor plate, revealed by shhexpression, is widened in zebrafish ntl mutant embryos(Halpern et al., 1997),suggesting that Bra activity antagonises floor plate development while promoting notochord formation. Moreover, in line with our findings, it was recently described that Notch represses ntl cell-autonomously(Latimer et al., 2002). Therefore, bra may constitute a key target in the binary switch that decides between notochord and floor plate fates under the control of Notch signalling. Further experimentation will be needed to elucidate whether a hierarchical relationship links chd and bra in this switch.

The enhancement of shh and plvs transcripts that we observe could be due to a Notch-dependent up-regulation of both genes or it may be a consequence of favouring floor plate development. Analogously, chd and bra down-regulation may be the result of their repression by active Notch or a consequence of disfavouring notochord development. However, as discussed above, shh and plvs have been implicated in the specification of floor plate fate, and bra and chd are necessary for proper notochord formation. Therefore, more than being regarded as mere markers, changes in the expression of these genes appear to be inherent in the fate decisions that are taking place. Thus, it seems more likely that Notch signalling triggers floor plate specification at the expense of the notochord through the opposite regulation of plvs/shh and chd/bra. Although further experimentation should elucidate if shh is required for Notch effects, and whether the activation of shh is direct or indirect, we have shown that shh down-regulates chd, and this can already be seen at early gastrula, suggesting that an enhancement of Shh signalling may be necessary for limiting the number of notochordal precursors. This mechanism may underlie some aspects of the role of shh in promoting floor plate development.

Classically, Notch activation has been considered as a mechanism by which a cell remains in a progenitor state to be available for subsequent waves of differentiation or as a way to repress the specification of certain cell types in favour of others. However, recent evidence suggests that Notch may have an instructive role in specifying glial fate(Gaiano et al., 2000; Morrison et al., 2000; Furukawa et al., 2000; Scheer et al., 2001) (for a review, see Gaiano and Fishell,2002; Lundkvist and Lendahl,2001). In this scenario, there could be at least two explanations for the role of Notch signalling in the development of the dorsal midline. (1)A permissive role for floor plate development, which implies that within the population of dorsal midline precursors, Notch activation may repress the notochordal fate and allow the development of floor plate identity through some default mechanism. In this context, it is intriguing that some markers of floor plate specification are also expressed by the notochord (e.g. shh,plvs) whereas notochordal markers seem to be exclusively present in the notochord (e.g. bra, chd), and depleting embryos of Bra activity, as in the zebrafish mutant ntl(Halpern et al., 1997),favours floor plate development at the expense of the notochord. Active Notch may thus deplete the dorsal midline precursors of molecules required for the specification of notochord, allowing floor plate to develop. (2) An instructive role, implying that, apart from the repression of genes required for notochord development, Notch signalling may actively promote floor plate specification by increasing the expression of genes that specify floor plate fate (e.g. shh, plvs).

In conclusion, we propose that the early organiser contains a population of cells with the potential to develop either as floor plate or notochord. Activation of Notch (which is present in the dorsal blastopore lip) in response to a ligand, which may be X-Delta-1 but this remains to be elucidated, may switch-off the genetic program for notochord specification in a subset of cells (evident by repression of bra and the cell-autonomous down-regulation of chd) and switch-on the program for floor plate development (instructive hypothesis), including the enhancement of plvs and shh expression, or allow this program to proceed(permissive hypothesis). In turn, secreted Shh could refine the segregation of both populations by limiting the number of notochordal (chd-positive)cells by a non cell-autonomous mechanism. This is consistent with the developmental profile of chd and shh expression. Their spatial patterns partially overlap in the early gastrula organiser but chd precedes shh and displays a broader domain. Later, when gastrulation finishes, this spatial relationship is reversed: while shh is expressed throughout the dorsal midline cell populations, chd expression is excluded from the prospective floor plate and the dorsal lining of the archenteron, and thus covers a subset of the shhexpression domain. This dynamic profile and the results of our functional experiments may thus underlie a negative feed back of shh over chd (Fig. 7). Whether the cell-autonomous down-regulation of chd elicited by Notch is due to the activation of a transcriptional repressor or is mediated by secreted Shh acting on the same cell, where it is activated by Notch (i.e. in an autocrine way), remains to be elucidated.

Finally, we have observed that at the early neural plate stage shhand plvs expression completely overlap in the floor plate domain(data not shown). If shh expression is restricted to the MFP and hnf3β/plvs is expressed in both floor plate populations in Xenopus, as has been described for zebrafish, mouse and rat embryos (Odenthal et al.,2000), our observation would suggest that LFP has not been specified yet. Hence, at this stage, these cells co-expressing shhand plvs, which we have shown to be susceptible to Notch signalling,may correspond to the MFP. This raises the possibility that the MFP derives from the same organiser population as the notochord, and thus Notch may either permit or instruct some cells to adopt MFP fate, while repressing the notochordal fate. Indeed, recent findings from the avian embryo demonstrate that the MFP derives from Hensen's node while the LFP is formed by the neuralised ectoderm (Charrier et al.,2002). Whether Shh signalling from the MFP further induces LFP on the neural ectoderm in Xenopus, as suggested for zebrafish and avian embryos (Odenthal et al.,2000; Charrier et al.,2002), remains to be tested.

We wish to acknowledge the following colleagues for providing us with the constructs for making synthetic mRNA: Chris Kintner for X-su(H)^DBM and hGR/ICD22; Thomas Hollemann for notch^ICD; Stephen Ekker for the full-length X-shh; and Richard Harland for nuc-lacZ. We are also grateful to the following researchers for providing us with the constructs for making in situ hybridisation probes: Igor Dawid for N-tubulin,Abraham Fainsod for X-brachyury, Thomas Hollemann for Xotch,Tomas Pieler for gli3, Ariel Ruiz i Altaba for pintallavisand Eddy De Robertis for chordin. We thank to Marianela Candolfi and Daniel Pisera for their collaboration. A.E.C. is indebted to Dr T. Pieler, who supported our work consistently and without prejudice. S.L.L., A.R.P. and A.E.C. are from CONICET. O.O. and M.V.R.S. were supported by fellowships Ramón Carrillo-Arturo Oñativia 2000 and 2001, respectively, and P.G.F. by a fellowship from ANPCyT. This paper was supported by grants to A.E.C. from CONICET (PIP 0848/98), ANPCyT (BID802/OC-AR PICT 3410) and Beca Ramón Carrillo-Arturo Oñativia 2000-2001.