Activating and repressing signals in head development: the role of Xotx1 and Xotx2

At the beginning of the century, pioneering embryological experiments performed on amphibians by Spemann and collaborators showed that transplantation of a gastrula dorsal blastopore lip to a region fated to form ventral mesoderm, resulted in the formation of a secondary axis (Spemann, 1938). The dorsal blastopore lip was called ‘organizer’ because of its ability of recruiting neighbouring host cells to form axial structures of the induced secondary axis. Moreover, the observation that early dorsal lips are able to induce secondary heads while late dorsal lip transplantations give rise to ectopic posterior structures, led to a distinction between a ‘head’ and a ‘trunk- tail’ organizer (Spemann, 1938; Hamburger, 1988). The Spemann’s organizer itself is induced by signals released at early blastula stage from a group of dorsovegetal cells known as the Nieuwkoop centre (Nieuwkoop, 1973). Candidate Nieuwkoop centre signals, such as the secreted factors Vg1, noggin and members of the Wnt family, were also isolated during the last few years (reviewed in Kessler and Melton, 1994; Slack, 1994).

Once induced, the organizer is involved in mesoderm patterning, where dorsalizing factors produced by the organizer itself, like secreted factors noggin (Smith and Harland, 1992), chordin (Sasai et al., 1994) and follistatin (Sasai et al., 1995) and the homeoprotein goosecoid (Niehrs et al., 1994), are antagonized by ventralizing signals represented by members of the BMP family (Graff et al., 1994; Fainsod et al., 1994; Piccolo et al., 1996; Zimmerman et al., 1996). The organizer region is also responsible for neural induction. According to a current model for neural induction, proposed by Nieuwkoop (1952), the neuroectoderm is initially specified as anterior (‘activation’) and only later posterior neural structures are derived from anterior neuroectoderm (‘transformation’) under the effects of posteriorizing signals. Neuralizing factors able to induce a condition similar to the ‘activation’ state are noggin (Lamb et al., 1993), follistatin (Hemmati-Brivanlou et al., 1994) and chordin (Sasai et al., 1995), while putative ‘transforming’ factors are retinoic acid (Durston et al., 1989) and FGF (Cox and Hemmati-Brivanlou, 1995; Lamb and Harland 1995). Regions that will give rise to the head show a migratory behaviour that is different from the one shown by regions that will form trunk and tail structures (Keller et al., 1992). In fact, head-forming regions do not undergo those cell intercalation movements that are typical of trunk- and tail-forming regions and are responsible for the elongation of these structures.

Recently, a new family of homeobox genes related to the Drosophila orthodenticle gene has been described (reviewed in Finkelstein and Boncinelli, 1994; Boncinelli and Mallamaci, 1995). As for the Drosophila gene, expression pattern of this gene family is strictly related to the development of anterior head structures. Among the family members, the most conserved genes in Vertebrates are Otx1 and Otx2 (Simeone et al., 1993; Bally-Cuif et al., 1994; Li et al., 1994; Pannese et al., 1995; Kablar et al., 1996). Mouse knock-out experiments have recently confirmed the hypothesis that these genes maybe involved in anterior brain patterning since Otx2^−/− mice lack forebrain and midbrain regions (Acampora et al., 1995; Ang et al., 1995; Matsuo et al., 1995), while Otx1^−/− mice show several brain abnormalities (Acampora et al., 1996). Nonetheless, the early expression pattern of these genes, which has been studied in detail in Xenopus, suggests an earlier role in development (Pannese et al., 1995; Blitz and Cho, 1995; Kablar et al., 1996). A common feature of the early expression domains of Xotx1 and Xotx2 is that they correspond to presumptive head regions that have been shown not to undergo convergent extension movements mediated by cell intercalation.

We previously studied Xotx2 and found that its overexpression resulted in embryos where severe posterior deficiencies are associated with the presence of additional cement glands and ectopic neural tissue. In this study, we show that embryos microinjected with Xotx1 display posterior defects similar to those observed in Xotx2-injected embryos but do not show any ectopic structure. These effects may be due to Xotx1 inhibitory action on convergent extension movements and tail organizer activity. Moreover, Xotx1 and Xotx2 expression is activated by Siamois (Lemaire et al., 1995), noggin (Smith and Harland, 1992) and Xwnt-8 (Smith and Harland, 1991), factors able to mimic head organizer induction in injected embryos, and repressed by a posteriorizing agent like retinoic acid. These data, taken together with Xotx1 and Xotx2 expression patterns demarcating anterior head regions from the earliest stages of development, suggest that Xotx genes are involved in head- organizing activity. We propose that, in order to allow proper head development, the head-organizing centre releases not only positive signals inducing anterior structures, but also negative signals capable of repressing the formation of posterior structures.

Eggs were obtained from female Xenopus laevis by injecting them with 800 IU of gonadotropin. Fertilization and embryo culture were performed as described by Newport and Kirschner (1982). Staging was according to Nieuwkoop and Faber (1967). Exogastrulae preparation and histological examination were performed according to Pannese et al. (1995).

Whole-mount in situ hybridization was performed on staged albino embryos as described by Harland (1991). The antisense or control sense-strand RNA probes were generated from linearized plasmids containing full-length cDNA inserts of Xotx1, Xotx2, pintallavis, Xbra, Xnot2, goosecoid, en-2 and XAG-1.

The following antisense RNA probes were synthesized using T7 or SP6 polymerase: Xotx1 (240 bp EcoRI-PvuII fragment), Xotx2 (170 bp BamHI-EcoRV fragment), rpS8 (90 bp DdeI fragment). Antisense RNA probes were hybridized to total RNA at appropriate temperature. RNase digestion and electrophoresis were carried out as described (Simeone et al., 1993).

Capped synthetic RNAs were generated by in vitro transcription of full coding sequences of Xotx1, Xotx2, human OTX1, Xwnt-8, noggin and Siamois or control construct ΔXotx1 (generated from a deleted Xotx1 clone and coding for a mutated protein lacking the homeodomain). All RNAs were resuspended in 88 mM NaCl, 5 mM Tris Ph 7.5 and injected in the animal pole of 1-, 2- and 4-cell-stage embryos, in the case of Xotx1 and Xotx2, or in the vegetal region of 2-cell-stage albino embryos, in the case of noggin, Xwnt-8 and Siamois.

Host embyos (stage 10⁺⁾ were transferred into 1× MBS (Gurdon, 1976) saline and dechorionated; the ventral blastocoele cavity was accessed with a cut made with a hair knife and the graft implanted using forceps and a hair loop. After 1 hour, the embryos were transferred in 0.1× MBS saline and cultured until stage 40. Dorsal marginal zone explants were isolated from stage 10 injected embryos and cultured flat under a coverslip in 1× MBS saline until control embryos reached stage 23.

We microinjected synthetic Xotx1 mRNA into the animal hemisphere of fertilized eggs and single blastomeres of 2- and 4-cell-stage embryos. Embryos injected with as little as 150 pg of Xotx1 mRNA display posterior deficiencies that range from strongly reduced trunk and tail structures to less severe phenotypes showing normal trunk and shortened tail (Fig. 1; Table 1). These effects are dose-dependent as increasing the amount of injected RNA results both in a higher percentage of affected embryos and in the predominance of the most severe phenotype. The site of injection is also determinant; embryos that received Xotx1 RNA in a dorsal blastomere at 4- cell stage exhibit posterior defects at a significantly higher frequency than embryos microinjected ventrally. Control microinjections using RNA synthesized from a frame shift Xotx1 construct (ΔXotx1), predicted to generate a mutated protein lacking the homeodomain, did not yield any specific effect. Moreover, we did not observed any of the above described phenotypes after microinjection of comparable amount of mRNA encoding the homeoproteins gsc (Cho et al., 1991a) and Pintallavis (Ruiz i Altaba and Jessel, 1992) (not shown). On the contrary, overexpression of human OTX1 produced embryos with reduced trunk and tail (Table 1). His- tological examination of Xotx1-injected embryos reveals the presence of somites, sometimes reduced in number and size, and notochord, which appeared larger and shorter than in control embryos, indicating that mesoderm induction was not affected in the injected embryos.

The phenotypes shown by Xotx1-injected embryos closely resemble those obtained after microinjection of Xotx2 mRNA. A major difference between the two is the presence of ectopic cement glands in Xotx2-injected embryos. To determine whether small clusters of ectopic cement gland cells, not readily detectable at the histological analysis, were induced by Xotx1 overexpression, we hybridized injected embryos with XAG-1, an early cement-gland-specific marker (Sive et al., 1989). Fig. 2B shows that even this more sensitive assay failed to show any ectopic expression of cement gland cells in Xotx1- injected embryos.

We also investigated possible synergisms between Xotx1 and Xotx2 by co-injection of equimolar amount of both RNAs. Surprisingly, the morphology of co-injected embryos remarkably resembled that of embryos injected with Xotx1 alone, showing posterior defects but no secondary cement gland. When these embryos were hybridized with XAG-1 (Fig. 2), it became apparent that Xotx1 strongly inhibits and, in some cases, completely abolishes, Xotx2 ability to induce ectopic cement glands (Fig. 2D). This inhibitory effect is only observed when equimolar amounts of the two RNAs are injected, as changing the ratio Xotx2:Xotx1 to 1:0.5 completely rescue the Xotx2 phenotype (Table 1). In any case no cooperative effect was noticed.

Xotx1-injected embryos appear indistinguishable from controls until early-mid gastrula stages (stage 10.25-10.5). Soon after these stages, blastopore closure is suspended in most injected embryos or, in less extreme cases, delayed. Suspension of blastopore closure results in embryos with reduced trunk and tail, whereas delay of the same movements, often ending with a not perfectly closed blastopore, produces the short tail phenotype. To directly follow Xotx1-injected cells during gastrulation, we co-injected a fluorescent dextran lineage label and Xotx1 mRNA into both dorsal blastomeres at the 4-cell stage (not shown). These lineage tracing experiments showed that posterior mesoderm did not involute and extend anteriorly, but instead migrated laterally around the blastopore. Nevertheless, the distribution of the labelling in anterior regions, corresponding to anterior mesoderm and presumptive anterior neuroectoderm, is similar in both Xotx1 and control embryos injected with fluorescent tracer alone, in agreement with the observation that Xotx1 overexpression does not affect head development.

To characterize further the effects of Xotx1 overexpression, we assayed the expression of genes that mark distinct cell populations during gastrulation in embryos injected into one blastomere at 2-cell stage in which the uninjected side serves as an internal control. As probes for whole-mount in situ hybridization, we used goosecoid, a marker of anterior mesoderm, Xbra, a pan-mesodermal marker whose expression at late gastrula become restricted to the notochord (Smith et al., 1991), Pintallavis, expressed both in the notochord and in the floor plate (Ruiz i Altaba and Jessel, 1992), and Xnot2, a tail bud marker (Gont et al., 1993). The injected sides of the embryos showed a remarkably reduced (or completely absent) expression of Xbra, Xnot2 and Pintallavis (not shown) at late gastrula stage (Fig. 3), whereas no difference in the expression of goosecoid was detected (data not shown). The expression of the same markers does not appear to be affected before stage 11.5; in the case of Xbra, this supports the observation that mesoderm induction occurs normally in these embryos. It is interesting to note that, in normal embryos, the expression of these down- regulated genes is strictly related to trunk and tail regions undergoing cell intercalation, whereas goosecoid, which is not affected by Xotx1 injection, is expressed by deep migratory cells that move anteriorly driven by crawling movements (Niehrs et al., 1993).

Convergent extension movements can be studied in exogastrulae where they are readily detectable as they take place without involution of the marginal zone (Keller, 1991). We analyzed exogastrulation in Xotx1-injected and control embryos by microinjecting embryos at 4-cell-stage into both dorsal blastomeres with RNA encoding Xotx1 or ΔXotx1. When embryos reached stage 7.5, they were forced to exogastrulate by culturing them upside- down without vitelline membrane in a hypertonic salt solution. Under these conditions, embryos injected with ΔXotx1 RNA behaved like uninjected controls, producing normally elongated exogastrulae at high frequency (Fig. 4A; Table 2). On the contrary, most of the exogastrulae obtained from Xotx1 overexpressing embryos are shorter than controls and, in the marginal zone, appear to be poorly elongated (Fig. 4B; Table 2). In most extreme cases (Fig. 4B, top row left embryo), the embryo exogastrulates thus separating ectoderm from mesendoderm, but then completely fails to elongate, resulting in an almost radially symmetrical exogastrula.

To study convergent extension using an independent approach, we also analyzed morphogenetic movements in dorsal marginal zone (DMZ) explants. Embryos were microinjected at 4-cell stage into dorsal blastomeres with Xotx1 or ΔXotx1 RNA and DMZ explants were isolated at stage 10. When control embryos reached stage 23, elongation of Xotx1- injected (Fig. 4D) explants was strongly inhibited compared with ΔXotx1-injected explants (Fig. 4C). We conclude that Xotx1 overexpression inhibits convergent extension movements that are essential for trunk and tail formation.

Tail formation is always impaired in embryos overexpressingXotx1 or Xotx2, even in less severe phenotypes showing normal head and trunk development. The expression of Xnot2, an early tailbud marker, is repressed in Xotx1-injected embryos. This led us to think that these embryos might lack tail organizer activity. The presence of such an activity is typically tested transplanting the tail organizer region into a receiving embryo and evaluating its ability to induce an ectopic tail. We therefore transplanted putative tail organizer regions from late gastrulae into the blastocoele of an early host gastrula: the donor gastrulae had developed from embryos injected with Xotx1, Xotx2 or ΔXotx1 RNA into both dorsal blastomeres at 4-cell stage. Based on recent fate mapping data (Tucker and Slack, 1995), the expression of Xnot2 and our preliminary transplantation experiments, we identified the region with the highest tail organizer activity in a normal stage 12.5 embryo, as an approximately 600 μm by 600 μm square including the dorsal blastopore lip. When this region was dissected from either ΔXotx1-injected stage 12.5 embryos, or uninjected controls, and transplanted into host embryos, an ectopic tail was induced in the majority of cases (Fig. 5E; Table 3, area B). These secondary tails appear to be quite normal by external examination, showing rows of melanocytes and fin structures; on histological analysis, they display a neural tube with a central canal, notochord and myotomes throughout the length of the tail (Fig. 6A,A′). When a corresponding region from a Xotx1-injected neurula of the same age was transplanted, in most cases, ectopic structures different from tails (Fig. 5C; Table 3: ‘bulging structures’) were induced while only in a minority of cases were ectopic reduced tails present (Fig. 5D; Table 3). These ectopic structures appear like swollen vesicles and do not show typical tail features, such as rows of melanocytes (although few dispersed melanocytes are present), or a fin, at an external inspection. The histological analysis reveals the absence of notochord and neural tube while most of the structures are composed of mesenchyme and not well-organized muscle cells (Fig. 6B,B′). Moreover, even the ectopic short tail-like structures, induced at lower frequency, show the lack of well-defined axial structures when assayed histologically although, in this case, patches of notochord and neural cells are present in the proximal portion of the ectopic structure.

Several dorsal cells that received Xotx1 RNA migrate laterally instead of involuting around the blastopore lip. To test if these cells have any tail organizer activity, we also trans- planted dorsolateral regions, including the lateral blastopore lip, from Xotx1-injected neurulae into receiving embryos (Table 3, area D). The induced structures are indistinguishable from those obtained after transplantation of the dorsal region of the same embryos, with a prevalence of ‘bulging’ structures.

Interestingly, when the same experiments were performed transplanting corresponding regions taken from Xotx2-injected embryos, ectopic anterior structures were often induced (Fig. 5A,B). The extent of induced anterior structures ranged from an ectopic cement gland to a complete secondary head. As shown in Table 3, transplantation of putative tail organizer (area B) from Xotx2-injected embryos results mainly in the induction of ectopic cement gland associated with eyes or complete heads, while transplantation of the dorsal-lateral region (area D) induces at higher frequency ectopic cement glands. ‘Bulging’ structures were also induced but at much lower frequency than observed for Xotx1 transplantations.

As a control, we also transplanted anterior neuroectoderm together with underlying anterior mesoderm taken from a subset of Xotx1- or ΔXotx1-injected embryos that were used for tail organizer dissections. Anterior structures were induced by anterior neuroectoderm and mesoderm explants from both types of embryos (Table 3). Finally we checked if head organizer activity was also affected in Xotx1-injected embryos. This appeared not to be the case as the transplantation of a region corresponding to the head organizer taken from Xotx1-injected early gastrulae efficiently produced secondary heads comparable to those obtained after trans- plantation of head organizer from control embryos (Table 3). In agreement with the observations of Slack and Isaacs (1994), we noticed that head structures induced by anterior implants (Table 3, areas A and C) localized in the anterior third of the ventral portion of the host embryo, while ectopic tails developed in the middle-posterior thirds. Notably ‘bulging’ structures always localized in the posteriormost region close to the proctodeum, thus revealing the intrinsic posterior character of these structures. On the contrary, despite their posterior origin, anterior structures and secondary heads induced by Xotx2 putative tail organizer transplantations always localized in the ventral-anterior region of the host embryo.

To investigate how Xotx1 and Xotx2 respond to caudalizing signals or to signals inducing the head organizer, RNAs from embryos treated with retinoic acid (RA), which acts as a posteriorizing agent (Durston et al., 1989), or microinjected with noggin (Smith and Harland, 1992) or Xwnt-8 (Smith and Harland, 1991) secreted factors able to induce the head organizer, were used for RNase protection analysis. Both Xotx1 and Xotx2 are repressed by RA although with a slightly different kinetics (Fig. 7). In fact, Xotx2 expression is already remarkably reduced at stage 10.25 after RA treatment, while Xotx1 expression at this stage is less severely affected. By stage 14, the expression of the two genes is completely turned off. On the contrary, microinjection of either noggin or Xwnt-8 stimulates the expression of the two Xotx genes already at stage 10.5. In particular, at stage 10.5, while Xwnt-8 leads to a similar increase (8-fold) of both Xotx1 and Xotx2 expression, noggin overexpression appear to stimulate Xotx1 more than Xotx2 (12- fold and 6-fold respectively).

We also analyzed the effects of overexpression of Siamois, a head organizer inducer, which probably acts as a mediator of Pre-MBT Wnt-signalling pathway (Lemaire et al., 1995), on Xotx1 and Xotx2 expression. As shown in Fig. 8, microinjection of Siamois in a vegetal position activates ectopic expression of both Xotx1 and Xotx2 already at stage 10.5.

Xotx1 and Xotx2 display several similarities and some differences in structure, expression pattern and overexpression phenotype. The two predicted proteins share the same general structure and homeodomains with proteins that belong to the bicoid class. The expression patterns of Xotx1 and Xotx2 are similar but not perfectly superimposable with Xotx1 expression domain, mainly restricted to midbrain regions, always contained within that of Xotx2, which includes midbrain and forebrain (Pannese et al., 1995; Kablar et al., 1996). When overexpressed in early embryos, Xotx1 and Xotx2 produce similar effects in impairing trunk and tail development but Xotx2 appears to be able to induce ectopic cement gland and neural tissue (Pannese et al., 1995, this work). When we co- injected Xotx1 and Xotx2 RNAs, we observed that Xotx1 exerts an inhibitory effect on the ability of Xotx2 to induce ectopic cement glands. Thus Xotx1 might function in repressing Xotx2 cement-gland-inducing activity in posterior archencephalon and anterior deuterencephalon regions where they are both expressed. According to this hypothesis only the fraction of Xotx2 that is expressed in the anteriormost neural plate, where Xotx1 is not present, would be able to induce cement gland. Given that it has been recently proposed that the natural cement-gland-inducing tissue is just the anteriormost neural plate (Drysdale and Elinson, 1993), and that cement gland induction by noggin and GSK3β (Itoh et al., 1995?? is associated with Xotx2 transcription activation, it is possible that Xotx2 may play a key role in this inductive event.

In order to better characterize the negative effects of Xotx genes on the development of posterior structures, we focused on Xotx1 because overexpression of this gene produces exclusively these effects. We believe that Xotx1 phenotype is specific because: (i) it is obtained even after injection of relatively low doses of RNA (150 pg); (ii) the severity of the effects is dose and site of injection dependent; (iii) injection of comparable amounts of other homeobox genes produces different effects, while these same effects are displayed by overexpression of other genes of the Otx family like Xotx2 and human OTX1 and (iv) genes expressed in convergent-extending cells are specifi- cally repressed by Xotx1 injection.

Lineage tracing experiments showed that posterior cells that received Xotx1, instead of undergoing normal convergent extension movements, migrate laterally around the blastopore. This leads to a delay of blastopore closure, more pronounced than the slight delay (about 0.5 stage) described as an aspecific effect of microinjection (Karnovsky and Klymkowsky, 1995), and to a disruption of trunk and tail development. Studies of morphogenetic movements in exogastrulae and DMZ explants confirmed that Xotx1 inhibits convergent extension. Whole- mount in situ hybridizations performed with markers that label specific cell populations during gastrulation also showed that Xotx1 overexpression specifically represses late gastrula stage expression of Xbra, Pintallavis and Xnot2, genes expressed in trunk and tail cells that undergo mediolateral intercalation movements. Notably, the development of anterior regions is not affected by Xotx1 microinjection as shown by lineage tracing and gsc expression that appeared to be normal as is the case for en-2 (data not shown).

Interestingly, it has been recently reported that expression of a deleted form of a XB/U cadherin (Kuhl et al., 1996) results in a dominant negative effect producing embryos that closely resemble the Xotx1 phenotype. The only difference is that, while in dominant negative XB/U cadherin embryos microcephaly is often observed, this defect is only rarely displayed by Xotx1-injected embryos. The similarity of the two phenotypes suggests that Xotx1 might inhibit convergent extension acting on cell adhesion molecules; this activity could be either direct or mediated by the repression of other genes (like Xbra, Pintallavis and Xnot2).

Impairment of tail development in both Xotx1- and Xotx2- injected embryos, as well as the down regulation of Xnot2 in Xotx1-injected embryos, suggested that the tail organizer activity could be repressed in embryos overexpressing Xotx genes. This appeared to be the case since transplantation of putative tail organizer regions from Xotx1-injected embryos, resulted in the induction of ‘bulging’ structures clearly different from secondary tails generated in control experiments. Intriguingly, when putative tail organizer were trans- planted from Xotx2-injected embryos anterior structures and secondary heads were induced. Thus Xotx2 seems to be able to reprogram a tail organizer into a head organizer. A possible interpretation of these data is that the inhibition of convergent extension is sufficient only to disrupt tail organizer activity, while in order to reconstitute a head organizer a specific an- teriorizing activity is also required. Xotx2 at variance with Xotx1, appear to possess this activity as also shown by its ability to induce ectopic cement glands in microinjected embryos.

Xotx genes appear to be good candidates for the role of genes involved in head-organizing activity. Since an activity shared by Xotx1 and Xotx2 is the repression of posterior structures formation, we propose that the head organizer releases not only a positive signal promoting the formation of anterior structures but also a negative signal that represses, in presumptive head regions, the development of posterior structures. There are examples of genes involved in head-organizing activity whose overexpression produces one or both effects. Genes that, when microinjected, generate both ectopic anterior structures and reduction of trunk-tail structures are: Xotx2, noggin, follistatin and Xwnt-8. However, there are also genes that show either only the positive signal, like gsc, Xlim (Taira et al., 1995) and Siamois (Lemaire et al., 1995), or only the negative signal, as is the case for Xotx1. Although negative signals may be more difficult to reveal, a confirmation of their existence should also come from loss-of-function mutants whenever the knocked-out function is not taken over by other genes. One example could be represented by Lim1^−/− mutant (Shawlot and Behringer, 1995). As expected for a putative head organizer gene, embryos homozygous for the null Lim mutation lack anterior head structures; unexpectedly, in some cases, duplication of neural axis posterior to rhombomere 3 was also observed. An explanation for this observation could be that Lim1 mediates both activating and repressing head organizer signals. In this case, the lack of this gene would remove not only the positive signal, leading to the absence of anterior brain, but also the negative signal, resulting in duplication of more posterior structures. On the contrary, the tail organizer might generate both positive and negative signals as well. In this case, positive signals inducing duplication of posterior structures are probably mediated by Xnr3 (Smith et al., 1995), Xnot2 and XlHbox6 (Cho et al., 1991b), while negative signals leading to anterior fate suppression could act through FGF (Isaacs et al., 1994), RA, Xhox3 (Ruiz i Altaba and Melton, 1989), Pintallavis and BMP4 (Graff et al, 1994; Fainsod et al.,1994). The existence of positive and negative signals released by the anterior pole is well extablished in Drosophila. In fact, bicoid acts as an anterior morphogen not only by stimulating the expression of anterior genes like orthodenticle, empty spiracles and button head, but also through repression of posterior genes. The latter activity is accomplished by bicoid either through hunchback activation or through a direct repression of caudal translation (Rivera-Pomar et al., 1996; Dubnau and Struhl, 1996). At the posterior end of Drosophila embryos, nanos is active which, in turn, represses hunchback translation, thus allowing formation of posterior structures. Similar mechanisms might also operate in Vertebrates.

We thank Eddy De Robertis for the Goosecoid and Xnot2 probes, Hazel Sive for the XAG-1 probe, Jim Smith for the Brachyury probe, Richard M. Harland for the noggin and Xwnt-8 plasmids, Patrick Lemaire for the Siamois plasmid and Ariel Ruiz i Altaba for the Pintallavis probe. We are grateful to Giuseppina Barsacchi, Robert Vignali and Jonathan Slack for helpful discussions and critical reading of the manuscript. This work was supported by grants from EC BIOTECH and BIOMED Programmes, the Telethon Italia Program and the Italian Association for Cancer Research (AIRC). M. A. was recipient of an AIRC fellowship.