Induction and patterning of the telencephalon in Xenopus laevis

The vertebrate central nervous system (CNS) is composed of a variety of discrete regions with diverse neuronal morphology and connectivity. Of outstanding interest is the study on how these different areas are generated during development, particularly within the forebrain, which contains the most complex structures in the vertebrate CNS, such as the telencephalon of mammals. Similarly to the Hox genes, which are involved in patterning the trunk CNS, several regulatory genes were proposed to define specific regions within the most rostral brain (Simeone et al., 1992; Bulfone et al.,1993; Shimamura et al.,1995; Shimamura et al.,1996; Rubenstein et al.,1998). Loss-of-function or gain-of-function experiments with these genes, either single or in combination, in fact lead to disruption of proper development within selected areas of the anterior CNS (reviewed byRubenstein et al., 1998;Wilson and Rubenstein, 2000;Boyl et al., 2001).

Particular interest has been focused on the signals that promote the spatially restricted expression of patterning genes within the developing CNS. Perhaps the best known model that has been proposed to explain neural patterning is the activation/transformation model of Nieuwkoop and co-workers(Nieuwkoop et al., 1952;Nieuwkoop and Nigtevecht,1954; Foley et al.,2000; Foley and Stern,2001; Stern,2001), who suggested that early induction and patterning of the neuroectoderm occurs in two steps. During a first step (`activation'), the dorsal ectoderm is initially induced from the adjacent and underlying mesendoderm to presumptive forebrain neuroectoderm. Subsequently, during the second step (`transformation'), some of this tissue receives caudalizing signals from the posterior dorsal mesoderm. This model has received strong molecular support from studies on Xenopus. Several factors that can work as `activators' have been identified in the secreted molecules noggin(Lamb et al., 1993; Zimmerman et al., 1996), chordin (Sasai et al.,1994; Piccolo et al.,1996), follistatin(Hemmati-Brivanlou et al.,1994), Xnr3 (Smith et al.,1995; Hansen et al.,1997) and cerberus (Bouwmeester et al., 1996; Piccolo et al.,1999). They are all expressed in the dorsal mesendoderm during gastrula/neurula developmental stages and work as extra-cellular antagonists of bone morphogenetic proteins (BMPs). Molecules with characteristics of`transformers' include retinoic acid, Wnts and FGFs, all of which can activate expression of posterior neural genes in neuroectoderm(Sasai and de Robertis, 1997;Gamse and Sive, 2001).

While the two-signal model may be sufficient to explain how the CNS is subdivided into main regions such as forebrain, midbrain, hindbrain and spinal cord, it does not explicitly account for the complex subregionalization of the forebrain itself. In principle, this could result from either a gradient of a single anterior inducing activity, or from the integration of multiple,qualitatively different, activities. Inhibition of BMP signaling appears to be a crucial step in forebrain induction, as shown by the double knockout ofchordin and noggin in the mouse(Bachiller et al., 2000). However, several lines of evidence suggest that, within the most anterior region of the neural plate, inhibition of BMP signaling needs to be integrated by other activities that counteract Wnt and Nodal signaling, thereby promoting forebrain development (Glinka et al.,1997; Piccolo et al.,1999). Some of these molecules have been identified as the Wnt-inhibitors Dkk1, Frzb1, crescent and sFRP2(Leyns et al., 1997;Wang et al., 1997;Glinka et al., 1998; Pera et al., 2000), the Nodal inhibitor Lefty1(Meno et al., 1999), or cerberus, a triple inhibitor of BMP, Wnt and Nodal(Piccolo et al., 1999), all of which are expressed in anterior mesendodermal tissues. Moreover, IGF signaling also appears to be required for head formation in Xenopus(Pera et al., 2001). Finally,patterning of the most anterior parts of the CNS may be integrated by additional signaling molecules, such as FGFs, Nodal, hedgehog proteins, Wnts and BMPs, involved in locally modifying the regional character of the forebrain neuroectoderm after its initial induction(Shimamura and Rubenstein,1997; Furuta et al.,1997; Ye et al.,1998; Barth et al.,1999; Golden et al.,1999; Gunhaga et al.,2000; Shanmulingam et al., 2000;Heisenberg et al., 2001;Rohr et al., 2001;Wilson and Rubenstein,2000).

Although all these data have started to clarify the molecular mechanisms that govern induction and patterning of the forebrain region, the fact that experiments were often performed on whole embryos did not allow in many cases the dissection of the activity of single inducing/patterning molecules, and to distinguish their direct actions on the neuroectoderm from indirect actions due to effects on mesendodermal tissues. This can be carried out easily in the frog embryo by means of dissection/recombination and misexpression methods that allow the overexpression of genes in the context of tissue conjugation experiments. In this paper, we report on some of the tissue and molecular signals at work in the induction and patterning of the anterior CNS inXenopus, with particular attention to the telencephalon. We show that dorsoventral patterning of the telencephalon is a complex process that cannot be elicited by simple inhibition of BMP signaling. Moreover, by dissection/recombination experiments, we identify the anterior dorsal endoderm(ADE) of the leading edge of the Xenopus gastrula embryo as a source of signals that can regulate dorsoventral patterning of the telencephalon, in possible cooperation with the adjacent prechordal mesendoderm. Finally, in animal cap assays, we have used different combinations of inducing and patterning molecules to show that dorsoventral telencephalic patterning can be reconstructed, at least partially, in naive ectoderm by the combined action of the ADE molecule cerberus and FGF signals.

Embryos were obtained and staged as previously described(Nieuwkoop and Faber, 1967;Newport and Kirschner, 1982). Embryos and explants were processed for whole-mount in situ hybridization as previously described (Harland,1991), except for proteinase K treatment, which was omitted, and for bleaching of pigment, performed as described by Mayor et al.(Mayor et al., 1995).Fig. 1 shows the expression patterns of the neural markers used in this study at stage 22/23 or stage 30/31.

Capped RNAs were synthesized from linearized plasmid templates as described(Krieg and Melton, 1984). Embryos were injected with 10-2000 pg mRNA/embryo at the one- and eight-cell stage as previously described (Vignali et al., 2000). The following template plasmids were used.

• cerberus: pcer-HA, pcer-S(Piccolo et al., 1999) and pcer-ΔC1 (Fetka et al.,2000).

• chd: pCS2-Chd (Sasai et al.,1994).

• ΔXFGFR-4a: ΔXFGFR-4a-pSP64T(Hongo et al., 1999).

• Nxfz8: pCS2-Nxfz8 (Deardoff et al., 1998).

• Smad7: pCS2-Smad7 (Nakayama et al., 2001).

RT-PCR was performed as described by Henry and Melton(Henry and Melton, 1998). Embryo RNA was extracted with RNA-NucleoSpin kit (Macherey and Nagel) and retro-transcribed with Superscript II (Invitrogen). PCR primers and conditions were drawn fromhttp://www.hhmi.ucla.edu/derobertis/index.html,except for cpl-1 (see Knecht and Harland, 1997), XBF-1 and nrp-1 (seeHongo et al., 1999). ForXemxl, 35 cycles were used with primers GCAGAAGCCTTTGTCAGTGG(forward) and CCTCCAGTTT-CTGCCTCTTG (reverse); for eomes, 32 cycles were used with primers GCCTACGAAACAGACTACTCCT (forward) and TAATGGAGGGAGGGGTTTCTAC (reverse).

For animal cap and dissection/recombination assays, RNAs were injected in the animal pole of one-cell stage embryos. Animal caps were dissected from stage 9 or stage 10.5 embryos in 1×MBS; after healing, caps were cultured in 0.5×MBS until early tailbud stage 22/23, or to late tailbud stage 30/31 alongside with sibling embryos.

Dissections and culturing of dorsal ectoderm from gastrula stage embryos were similarly performed.

In conjugate experiments, embryo fragments were similarly dissected,recombined and cultured. Peptide-releasing beads (SIGMA H-5263) were washed in 1×PBS and then incubated overnight at 4°C in 5 μl of 1×PBS,0.1% BSA containing either human bFGF (100 or 200 ng/μl; ICN) or mouse FGF8b (100, 200 or 400 ng/μl; R&D). Beads were implanted within pairs of animal caps dissected from either injected or uninjected embryos.

Dissections were performed in the presence of gentamycin (50 μg/ml final concentration).

We have previously shown that BMP antagonists, such as noggin(Lamb et al., 1993) and Xnr3(Hansen et al., 1997),although able to trigger anterior neural fate in Xenopus animal caps,are not sufficient to specify dorsal telencephalon(Pannese et al., 1998). Yet,the dorsal blastopore lip of the early Xenopus gastrula can efficiently activate dorsal telencephalic markers in animal cap tissue(Pannese et al., 1998). In order to identify what signals from the dorsal blastopore lip are involved in the induction of dorsal telencephalon, we first aimed to define their timing of action during development. We therefore removed the dorsal ectoderm (DE)from gastrula stage Xenopus embryos, cultured the explants up to the corresponding of early tailbud (stage 22/23) or late tailbud (stage 30/31)stage, and assayed their state of specification by in situ hybridization using several neural markers, including dorsal telencephalic markers.

In particular, we dissected, from stage 10.5 midgastrula embryos, fragments of DE of about 500 μm comprised between the animal pole and about half way between the dorsal blastopore lip and the leading edge of the involuting dorsal mesendoderm, as outlined in the scheme inFig. 2. Data from one to seven independent experiments (Table 1) — depending on the analyzed marker — indicate that this DE region is already specified to develop as anterior neural tissue. In fact, explants cultured up to stage 22/23 showed a strong expression of the pan-neural marker nrp-1 (Knecht et al., 1995), of the fore/midbrain marker Xotx2(Pannese et al., 1995), of the general telencephalic marker XBF-1 (also expressed in the nasal part of the eye) (Papalopulu and Kintner,1996), of the eye marker Xrx1(Casarosa et al., 1997) or of the ventral forebrain marker Xvax1b(Liu et al., 2001)(Fig. 2A-C,F,G;Table 1). However, only few of the explants showed a faint staining for the dorsal telencephalic markerXemx1 (Pannese et al.,1998) (Fig. 2H;Table 1). We also found that more posterior markers such as Xkrox-20(Bradley et al., 1993) andXhoxB9 (Wright et al.,1990) were not activated at all, or activated only in few explants(Fig. 2D,E;Table 1). By contrast, when the explants were cultured to stage 30/31, not only did they express Xrx1,Xvax1b and the ventral forebrain marker Xnkx2.1(Small et al., 2000), but an evident activation of Xemx1 also occurred(Fig. 2K-M,Table 1; seeFig. 6B andTable 1 for Xnkx2.1). To test for possible mesoderm contamination, a proportion of explants were assayed for expression of chd(Sasai et al., 1994) (at the equivalent of stage 12.5) or muscle actin(Mohun et al., 1984) (at stage 22/23), and found deprived of expression for either marker(Fig. 2I,J;Table 1). Therefore, although some aspects of forebrain specification have already taken place by midgastrulation, the onset of expression of dorsal telencephalic genes appears to be significantly delayed in stage 10.5 explants. However, when DE was dissected from late gastrula embryos, clear expression of Xemx1 was already detectable at stage 22/23 (data not shown). These observations suggest that further contact with the dorsal mesendoderm may be required between mid-gastrula and end of gastrulation, to ensure a proper temporal specification of the dorsal telencephalon.

Because signals produced from dorsal mesendoderm may be important for proper induction of dorsal telencephalon(Pannese et al., 1998), we decided to assay the inducing/patterning abilities of different regions of this tissue.

In order to do this reproducibly, different parts of the involuting mesendoderm were dissected at stage 10.5, cultured to stage 12.5, and assayed with Xhex, chd, gsc and Xnot-2 probes as diagnostic molecular markers. We identified four different pieces, which were used in our recombination experiments. Three of these fragments are contained within one another, and correspond to the yellow, red and green pieces in the schemes ofFig. 3. A first fragment of about 100 μm, corresponding to the anterior dorsal endoderm (ADE; in yellow in Fig. 3), strongly expressedXhex (Jones et al.,1999) (Fig. 3A;Table 1); contamination by prechordal mesoderm was excluded by absence of hybridization to a chdprobe (Fig. 3B;Table 1), that specifically labels the whole axial mesendoderm, but not the most anterior dorsal endoderm(Sasai et al., 1994). A larger fragment of about 200 μm, corresponding to the anterior half of the involuted anterior dorsal mesendoderm (ADME; in red inFig. 3), weakly expressedgsc (Cho et al.,1991), a marker of prechordal mesoderm, strongly expressedchd, but did not express Xnot-2(Gont et al., 1993), a marker of presumptive notochord (Fig. 3C,D; Table 1; and data not shown). By contrast, a still larger fragment of about 300 μm,corresponding to the anterior three-quarters of the involuted ADME (in green in Fig. 3), strongly expressedgsc while showing a weak spot of Xnot-2 staining only in a minority of explants (Fig. 3E,F; Table 1). Finally, a fourth fragment of about 120 μm, corresponding to the posterior quarter of the involuted ADME of stage 11 embryos (in brown inFig. 3), showed a weakgsc, but a strong Xnot-2, expression(Fig. 3G,H;Table 1). Thus, we conclude that the `yellow' fragments correspond to the anterior endoderm of the leading edge (ADE), while those in `red' or `green' appear to contain exclusively, or almost exclusively, prechordal mesendoderm; finally, the `brown' fragment is mainly composed of presumptive notochord tissue with little — if any— prechordal mesendoderm.

We first separately analyzed the inducing properties of the prechordal mesendoderm (`green' fragment) compared with those of the presumptive anterior notochord (`brown' fragment). Their different inducing abilities were tested by conjugating either `green' or `brown' fragments with stage 9 animal caps,followed by in situ hybridization analysis of the conjugates at the corresponding of stage 22/23. A weak anterior neural induction was detected in the conjugates with the prechordal (`green') fragment, as shown by the occurrence of either weak or no activation of nrp-1, Xotx2, Xrx1,XBF-1 and Xemx1 genes after extensive color reaction(Fig. 4A-E;Table 1). By contrast,efficient induction of neural tissue took place in the conjugates made with presumptive anterior notochord (`brown') tissue, as shown by the strong activation of nrp-1; localized weak expression of Xotx2, Xrx1,XBF-1 and Xemx1 was detected only in a minority of explants(Fig. 4F-J;Table 2). By RT-PCR assay, very weak or no activation was detected for nrp-1, N-CAM, XBF-1, Xotx2,Xrx1 and Xemx1 in conjugates with the prechordal mesendoderm(Fig. 4K). Instead nrp-1,N-CAM, XBF-1, Xotx2 and Xrx1 were readily detected in conjugates with the anterior chordomesoderm, while Xemx1 was very weakly expressed in these recombinants (Fig. 4K), possibly owing to the presence of contaminatinggsc-positive cells in the `brown' fragment(Fig. 3G). Therefore, the prechordal mesendoderm and the anterior notochord significantly differ in their neural inducing abilities, but neither tissue is able to efficiently induce dorsal telencephalic character in naive ectoderm. Differences between in situ hybridization and RT-PCR results may reflect the different potencies of the two techniques in detecting localized or average levels of expression.

However, when the prechordal mesendoderm (`green' piece ofFig. 3) was conjugated to neuralized stage 10.5 DE (upper scheme ofFig. 5), it was able to restore appropriate expression (both in timing and intensity of signal) of dorsal telencephalic genes (Xemx1) within the conjugates cultured up to stage 22/23 (Fig. 5A;Table 1). In fact, a smaller region of this `green' fragment may be sufficient for this patterning activity: when stage 10.5 DE (as shown inFig. 2) was removed from embryos together with the underlying fragment of ADME (`red' piece inFig. 3), as outlined in the lower scheme in Fig. 5, again appropriate strong expression of Xemx1 was observed at stage 22/23(Fig. 5B;Table 1).

It has been proposed that the ADE in Xenopus, and the corresponding structure known as AVE in the mouse, may play a pivotal role in forebrain development (Bouwmeester et al.,1996; Thomas and Beddington,1996). We therefore tested whether the ADE alone, without the adjoining prechordal mesendoderm, could elicit Xemx1 activation in midgastrula DE. We explanted DE fragments from stage 10.5 embryos together with the underlying ADE (`yellow' piece ofFig. 3), as in the lower scheme in Fig. 6, and cultured them up to stage 22/23. Control explants, made of DE alone (upper scheme ofFig. 6), displayed a strongXrx1 (as a positive control of neuralization, data not shown), but noXemx1 activation (Fig. 6A; Table 1); by contrast, most of the ADE-containing recombinates expressed Xemx1(Fig. 6C;Table 1). We also tested whether the ADE had any effect on ventral forebrain marker specification, and surprisingly found that while Xemx1 expression was maintained in stage 30/31 recombinates (data not shown), Xnkx2.1 expression was suppressed in these recombinates, compared with explants of DE(Fig. 6B,D;Table 1). These results suggest that ADE may be important for specification of dorsal telencephalon, but may have an inhibitory effect on ventral forebrain specification.

The organizer-secreted BMP antagonists noggin and Xnr3 are able to activateXotx2, but not Xemx1 and Xemx2 expression, inXenopus animal caps grown to stage 22/23(Pannese et al., 1998). Because DE isolated from midgastrula embryos shows Xemx1 expression only when cultured up to stage 30/31, and not to stage 22/23(Fig. 2H,M;Table 1), we asked whether any similar delayed activation of Xemx1 could take place in animal caps neuralized by BMP antagonists.

Animal caps were dissected from stage 9 embryos injected with chdmRNA. For optimal culture to later stages caps were joined in pairs to allow a better healing of the explants (see scheme inFig. 8) and cultured up to stage 22/23 or 30/31. After injection of doses of chd ranging from 10 to 600 pg, no induction of the dorsal telencephalic markers Xemx1 andeomes (Ryan et al.,1998) was ever observed at either stage, either by in situ hybridization or RT-PCR analysis (Fig. 7B,C; Fig. 9E,F;Fig. 10;Table 2 and data not shown). Because different levels of BMP antagonism have been shown to induce neural tissue of different dorsoventral character(Knecht and Harland, 1997), we analyzed the dorsoventral organization of chd-injected caps in our assays. Strong staining with the epidermal marker XK81 was detected in chd-injected caps, indicating that explants retained epidermis and possibly a dorsal boundary between neural tissue and epidermis in the conditions used (Fig. 7D;Table 2). Presence of a dorsal neural tube boundary was also addressed by checking the expression of the telencephalic dorsal neural tube boundary marker cpl-1(Knecht et al., 1995).cpl-1 is strongly expressed in caps at low doses of injectedchd, but still detectable, though at low levels, at high doses(Fig. 10); these results are consistent with earlier observations(Knecht and Harland, 1997) and show that even in conditions that promote cpl-1 strong expression,Xemx1 and eomes are never induced. To rule out the possibility that these results could be specific to Chd with respect to other BMP antagonists, we also assayed Smad7, a global antagonist of the whole TGF-β pathway (Nakayama et al.,2001), and obtained similar results(Fig. 7E-H;Table 2;Fig. 10).

Because we showed that stage 10.5 ADME had a patterning activity on stage 10.5 neuralized DE, we asked whether it could integrate the action of BMP inhibitors to activate dorsal telencephalic genes. Therefore, stage 9 animal caps were explanted from chd injected embryos, conjugated with the ADME (`red' fragment of Fig. 3)of stage 10.5 control embryos and grown to stage 30/31 (see scheme inFig. 8). Although conjugated pairs of chd injected caps did not express Xemx1 oreomes (Fig. 8G,H;Table 2), but expressed the positive control marker Xrx1 (Fig. 8I; Table 2),chd injected caps recombined with the ADME were positive both forXemx1 and for eomes expression(Fig. 8J,K;Table 2). By contrast,uninjected caps conjugated with the ADME did not show any expression forXemx1, eomes and Xrx1(Fig. 8D,E,F), confirming the poor, if any, forebrain inducing activity of the ADME. These results demonstrate that the ADME is able to complement the action of BMP antagonists to promote development of dorsal telencephalon.

Head induction has been proposed to result from the triple inhibition of BMP, Wnt and Nodal pathways (Glinka et al., 1997; Glinka et al.,1998; Piccolo et al.,1999) by several secreted proteins. Among them, cerberus has the unique feature of being a triple BMP-Nodal-Wnt-antagonist; moreover, it is expressed in the ADE, which plays a patterning role on the anterior neuroectoderm (see above). Remarkably, when cerberus mRNA was injected into animal caps, besides a strong activation of Xrx1 in the vast majority of explants, also Xemx1 and eomes expression was found in some of the animal caps (Fig. 9A-C; Table 2).

As cerberus is a triple BMP-Wnt-Nodal-inhibitor, we decided to define which of these inhibitory activities was required for induction of dorsal telencephalic genes. To achieve this, we made use of two previously described constructs, cer-S and cer-ΔC1, encoding the C-terminal(cer-S) and the N-terminal (cer-ΔC1) regions of cerberus, which have been described as a Nodal-antagonist and as a Wnt-antagonist, respectively(Piccolo et al., 1999;Fetka et al., 2000). When the anti-BMP activity of Chd was coupled to the anti-Nodal activity of cer-S, no activation of either Xemx1 or eomes was detected(Fig. 10), in agreement with the result obtained with the general TGFβ inhibitor Smad7(Fig. 7F,G;Table 2;Fig. 10). By contrast, the combination of Chd and cer-ΔC1 was clearly able to induce bothXemx1 and eomes, as detected by in situ hybridization and by RT-PCR, while no activation was detectable in chd injected caps(Fig. 9E-H;Table 2;Fig. 10). cer-ΔC1 alone was not able to induce any expression of Xemx1, eomes, NCAM andXrx1 at doses that were able to induce dorsal telencephalic genes in combination with Chd, suggesting that cer-ΔC1, at least at these doses,lacks neural inducing ability and hence does not retain significant BMP-antagonizing activity (Fig. 10). However, we found that besides the previously described Wnt-blocking activity (Fetka et al.,2000), cer-ΔC1 retains some Nodal-antagonizing activity(data not shown). We also compared the effects of cer-ΔC1 with those of Nxfz8, a potent Wnt-antagonist (Deardoff et al., 1998). In contrast to cer-ΔC1, neither a combination of Chd and Nxfz8, nor a combination of Chd, Nxfz8 and cer-S, was able to induce expression of Xemx1 or eomes(Fig. 10), at doses of Nxfz8 that efficiently induced strong axial defects in whole embryos(Deardoff at al., 1998) (data not shown). Similar results were obtained with the analogous construct ECD8(Itoh and Sokol, 1999) (data not shown). Because the only qualitative difference between the combinations Chd+cer-ΔC1 and Chd+Nxfz8+cer-S resides in the Wnt-inhibitory activities of cer-ΔC1 and Nxfz8, these results suggest that the dorsal telencephalic inducing activity of cerberus relies on its specific anti-Wnt action; however, we cannot completely rule out the possibility that the residual anti-Nodal activity of cer-ΔC1 may also be required.

In addition, we also tested induction of the ventral forebrain markerXnkx2.1 in these same caps. Xnkx2.1 was not induced by Chd,Smad7, or the combinations of Chd+cer-S and Chd+cer-ΔC1(Fig. 10), indicating that,though cerberus is able to partially promote dorsal telencephalic fates, a full patterning of the telencephalon may require the integration of different molecular pathways.

FGFs have been proposed to play important roles both in early neural induction in the frog and the chick (Hongo et al., 1999; Streit et al.,2000), and in later patterning of the anterior neural plate, and particularly the telencephalon, in the mouse(Shimamura and Rubenstein,1997; Ye et al.,1998; Shanmugalingam et al.,2000) (reviewed by Rubenstein et al., 1998; Wilson and Rubenstein, 2000).

We therefore tested whether we could induce telencephalic genes in animal caps by integrating the activities of Chd and cerberus with that of FGF. To do this we conjugated pairs of animal caps, injected with either chd orcerberus mRNA, around a bead soaked in bFGF or in FGF8 (seeFig. 11). Animal caps were dissected at stage 10.5, when they no longer respond to mesoderm inducing signals, and therefore any effect of FGFs is a direct effect on ectoderm(Lamb and Harland, 1995). Cap competence for mesoderm induction was excluded by failure of either bFGF or FGF8 to induce Xbra (Smith et al., 1991) expression, while failure to detect expression of the pan-neural marker Sox2 excluded any direct neural inducing activity by FGFs (data not shown). However, it proved to be difficult to harvest stage 10.5 caps from chd-injected embryos, probably because excess involution of dorsal mesendoderm made it impossible to dissect caps without any underlying mesendoderm. Therefore, it would not be possible to discriminate whether FGF activity, rather than signals from the underlying mesendoderm, was responsible for any effect additional to that of Chd. We therefore co-injected chd mRNA with cer-S, acerberus construct that, by inhibiting mesoderm formation(Piccolo et al., 1999), was able to prevent any excessive involution of mesendoderm, and afterwards dissected and conjugated the animal caps to FGF beads. The effects of Chd+cer-S on caps dissected at stage 10.5 were not substantially different from those on caps dissected at stage 9, at least for the markers we tested,and resulted in no activation of the ventral gene Xnkx2.1(Fig. 11A), essentially no activation of the dorsal genes eomes(Fig. 11F) and Xemx1(Fig. 11K), and in a strong activation of Xrx1 (data not shown). However, when FGF8 was added,Xnkx2.1 expression was strongly activated in animal caps(Fig. 11B,C;Table 2), and activation was also observed for eomes (Fig. 11G,H; Table 2); by contrast, slight, if any, activation of Xemx1 was observed(Fig. 11L,M;Table 2). Similar effects were also observed for bFGF (Fig. 11D,E,I,J,N,O; Table 2).

Different results were obtained when cerberus-injected caps were explanted at stage 10.5 and conjugated in pairs either without or with FGF-soaked beads. Again, Xnkx2.1 and Xnkx2.4 were not activated by the injected RNA (Fig. 11A′,M′; Table 2); however, clear activation was observed for eomes(Fig. 11E′;Table 2) and for Xemx1(Fig. 11I′;Table 2); finally, strong activation was observed in all caps for Xrx1 (data not shown). When FGF was added to cerberus-injected caps, Xnkx2.1 andXnkx2.4 were strongly activated in almost all explants(Fig. 11B′-D′,N′;Table 2) and an increase was also observed in the expression of eomes(Fig. 11F′-H′;Table 2). Instead, no significant difference was caused by FGFs on Xemx1 activation compared with cerberus alone(Fig. 11J′-L′;Table 2).

These data indicate that FGF signals can promote ventral forebrain fates and may also be important for regulation of dorsal telencephalic fates. To further investigate this, we interfered with the FGF signaling pathway by using a dominant-negative FGF receptor, ΔXFGFR-4a, which blocks the effects of FGF8 on neural tissues (Hongo et al., 1999; Hardcastle et al., 2000). We therefore injected ΔXFGFR-4a mRNA in the animal region of Xenopus early embryos and subsequently conjugated stage 9 animal caps explanted from these embryos with a full stage 10-10⁺ organizer. Control conjugates were made with uninjected animal caps and the organizer. Experimental and control conjugates were assayed for the ventral marker Xnkx2.1 and the dorsal markerXemx1 at stage 30/31. Although in control explants both genes are strongly activated (Fig. 12A,B; Table 2), in experimental conjugates, expression of both genes was substantially suppressed(Fig. 12D,E;Table 2). By contrast, there was no apparent effect on neural induction, as the expression of the pan-neural marker Sox2 (Misuzeki et al., 1998) was essentially the same in the two sets of conjugates (Fig. 12C,F;Table 2). These data therefore suggest that FGF signals are required for correct patterning of the forebrain.

Classical models suggest that neural induction and patterning result from the combined action of two different signaling steps acting on the DE: a first step (activation) that is due to a uniform forebrain-inducing signal, and a second step in which forebrain-induced tissue is posteriorized to presumptive hindbrain and spinal cord (reviewed byGamse and Sive, 2001;Foley and Stern, 2001;Stern, 2001). Our results on injected animal caps show that molecular signaling proposed to mediate the activation step (namely BMP inhibition) is not sufficient to induce dorsal and ventral telencephalic fates, suggesting that full patterning of the forebrain requires the integration of complex signaling. In fact, although some aspects of forebrain specification may be triggered by BMP inhibitors, as shown by the activation of Xrx1 (and XBF-1, data not shown)(Andreazzoli et al., 1999) inchd-injected animal caps, expression of the dorsal telencephalic markers Xemx1 or eomes, or of the ventral forebrain markerXnkx2.1 was never observed. Therefore, signals are required to integrate the action of BMP inhibitors in order to specify both dorsal telencephalic values and ventral forebrain values. Because dorsal blastopore lip of stage 10-10⁺Xenopus embryo is able to induce the dorsal telencephalic markers Xemx1 and Xemx2 in naive ectoderm (Pannese et al.,1998), additional signaling may reside in the tissues of the organizer region, namely the ADME. In fact, our recombination experiments with explanted stage 10.5 DE and different region of ADME show that, in spite of its poor neural inducing activity, the ADME can play a patterning role on neuralized DE.

In the last few years, work on several vertebrate models has unravelled a crucial role of anterior endodermal tissues in forebrain development. In particular, the mouse anterior visceral endoderm (AVE) is essential for forebrain induction and patterning, as shown by both embryological and genetic manipulations. Indeed, removal of the AVE at the earliest stages of gastrulation impairs activation of rostral CNS markers in the epiblast(Thomas and Beddington, 1996). Moreover, before their activation in the axial mesendoderm, several genes required for forebrain development, such as Lim1, Otx2, HNF3βand nodal, are expressed in the pregastrula stage AVE, where their activities are specifically required for proper forebrain formation (reviewed by Beddington and Robertson,1998). Recently, the chick hypoblast has been proposed as the embryological and functional equivalent of the mouse AVE. In fact, genetic activities characteristic of mouse AVE are also detectable in chick hypoblast at pre-streak stages; moreover, the hypoblast induces pre-forebrain markers in the epiblast before streak formation and protects the forebrain territory from caudalizing signals by directing cell movements that distance the anterior epiblast from the organizer (Foley et al.,2000). During gastrulation, both mouse AVE and chick hypoblast are displaced by the involuting foregut endoderm; also this tissue has important functions for proper forebrain formation: in chick, removal of the foregut endoderm during gastrulation results in severely compromised forebrain patterning (Withington et al.,2001). In addition, the foregut endoderm shares some of the genetic activities of the mouse AVE or chick hypoblast, such ascerberus and Hex. Knock-out of the Hex gene in the mouse and analysis of chimeric embryos showed that Hex function is specifically required in the foregut endoderm for normal forebrain development(Martinez Barbera et al.,2000). Therefore it is likely that the AVE/hypoblast and the foregut endoderm may play similar roles and that the anti-caudalizing activity of the AVE/hypoblast is taken over at later stages by the foregut endoderm and/or prechordal mesendoderm (Foley et al., 2000; Foley and Stern,2001; Stern,2001). Although in chick and mouse this activity occurs in two separate tissues (the AVE/hypoblast and the foregut endoderm), inXenopus, the anterior dorsal endoderm (ADE) that constitutes the leading edge of the involuting dorsal mesendoderm may possess the signaling properties of both amniote tissues. Like them, the ADE is the only frog tissue that expresses cerberus and Hex. Moreover, it displays cell movements reminescent of the mouse AVE(Jones et al., 1999;Foley and Stern, 2001). Finally, the ADE will contribute to the foregut, and, similarly to the foregut endoderm of chick and mouse, it may be important to confer anterior character to the overlaying ectoderm, as judged by the ability to trigger cement gland markers in gastrula ectodermal explants(Bradley et al., 1996;Jones et al., 1999).

Our data suggest a new potential role for the Xenopus ADE in the dorsoventral patterning of the forebrain, possibly in synergism with the adjacent prechordal mesendoderm. In fact, the ADE was able to activate the dorsal telencephalic marker Xemx1 in midgastrula DE explants that,although already specified to forebrain fates, would not expressXemx1 at the early tailbud stages. Moreover, expression of the ventral forebrain marker Xnkx2.1 was suppressed in stage 10.5 DE explants conjugated to the ADE. These data suggest that the ADE may be involved in inducing dorsal telencephalic fates and repressing ventral fates within the prospective forebrain region. This patterning role was further supported by the striking observation that a fragment of ADME, including the ADE together with the anteriormost prechordal mesendoderm, was able to elicitXemx1 and eomes expression in chd-injected caps,where expression of these dorsal telencephalic markers was otherwise never detected. Notably, removal of the anterior definitive endoderm in chick embryos seems to impair proper regionalization of dorsal, but not ventral,forebrain territories, although a more specific molecular marker analysis was not performed (Whitington et al., 2001).

Previous work in Xenopus has shown that planar signals spreading from the dorsal mesendoderm are sufficient to induce, in the adjacent ectoderm, neural tissue with a remarkable degree of anteroposterior patterning, including forebrain characters(Doniach at al., 1992;Papalopulu and Kintner, 1993). However, additional vertical signaling is required from the involuting mesendoderm for proper differentiation, morphogenesis and patterning of the nervous system (Dixon and Kintner,1989; Ruiz i Altaba,1992). In line with these observations, our results suggest that vertical signals from the ADE and possibly the adjacent ADME may be specifically responsible for proper dorsoventral patterning of the telencephalon during gastrulation.

A crucial question concerns the identity of molecules mediating the patterning activity of the ADE. The secreted molecule cerberus was a likely candidate to mediate part of this activity: its expression is restricted to the ADE throughout gastrulation(Bouwmeester et al., 1996), and besides providing a BMP antagonistic effect, cerberus is also endowed with anti-Wnt and anti-Nodal activities(Piccolo et al., 1999), which could account for the patterning effects of the ADE. Remarkably, we found that cerberus was not only able to trigger anterior neural induction and early forebrain markers (such as Xrx1 and XBF-1) (see Results;data not shown) in animal caps, as do other BMP inhibitors, but also to induce the expression of the dorsal telencephalic markers Xemx1 oreomes. We then attempted to define which of the three inhibitory activities of cerberus are required for the induction of these genes. When the anti-Nodal activity of cer-S (Piccolo et al., 1999) and the anti-BMP activity of Chd were combined together, they were not able to induce Xemx1 and eomes. Instead, their efficient induction was obtained by the combination of cer-ΔC1, containing the Wnt-inhibitory activity of cerberus(Fetka et al., 2000), and Chd,while cer-ΔC1 alone did not show any telencephalic or neural inducing activity, at least in the conditions we used. Taken together, with respect to the induction of dorsal telencephalic genes, these results suggest that: (1)the anti-BMP and the anti-Wnt activities of cerberus are both required; and(2) neither of them alone is sufficient, but they might be possibly sufficient in combination. However, in our hands, cer-ΔC1 seemed to retain a partial anti-Nodal activity that has not been previously described(Fetka et al., 2000); thus, at present, a requirement for the anti-Nodal activity of cerberus in the activation of dorsal telencephalic genes cannot be completely excluded. When a different Wnt-antagonist, Nxfz8 (Deardoff et al., 1998), was tried, it did not trigger Xemx1 oreomes, either in combination with Chd or with the further addition of the anti-Nodal activity of cer-S. Because in all the different combinations that we assayed, dorsal telencephalic genes were only induced when the Wnt-inhibitory action of cerberus was included, these results would suggest that dorsal telencephalic induction may require a particular specificity of Wnt inhibition. Besides cerberus, several other inhibitors of Wnt signaling are secreted from the ADE and/or the adjacent prechordal mesendoderm, such as Dkk1 (Glinka et al., 1998),Frzb1 (Leyns et al., 1997;Wang et al., 1997), crescent and Sfrp2 (Pera and De Robertis,2000). They have different anti-Wnt specificities and different biological activities (Kazanskaya et al.,2000; Pera and De Robertis,2000); some of them may cooperate with cerberus in inducing the dorsal telencephalon. The requirement of the anti-Wnt activity of cerberus for the induction of dorsal telencephalic genes in animal caps raises the question of which Wnts need to be inhibited. In Xenopus, Xwnt7B(Chang and Hemmati-Brivanlou,1998) and Xwnt8b (Cui et al., 1995) are widely expressed in the ectodermal region of the embryo during gastrula and neurula developmental stages; furthermore,Xwnt7B expression is maintained in animal caps dissected from blastula stage embryos (Chang and Hemmati-Brivanlou, 1998). Therefore, Xwnt7B andXwnt8b potentially represent two Wnt activities whose inhibition may be necessary for patterning of the telencephalon in Xenopus. This hypothesis is strongly supported by recent work in zebrafish, showing requirement of local Wnt antagonism for telencephalic gene expression within the anterior neuroectoderm, and identifying Wnt8b as a likely target for this antagonism (Houart et al.,2002).

Because FGF8, as other FGFs (Shinya et al., 2001), is expressed in the anterior neural ridge(Crossley and Martin, 1995),and seems to mediate the ability of the latter to promote expression of the telencephalic marker XBF1(Shimamura and Rubenstein,1997; Ye et al.,1998) and also later aspects of telencephalic patterning(Fukuchi-Shigomori and Grove,2001), we tested whether FGF could have a role in the regulation of dorsal and ventral telencephalic genes. We here show that FGF8 is able to potentiate eomes expression in Chd+cer-S or cerberus injected caps. Moreover, Xemx1 activation in animal caps by the head organizer was severely compromised by overexpression of the dominant-negativeΔXFGFR-4a receptor. Together, these results suggest that cerberus and FGF8 may interact in the specification of the dorsal telencephalon.

We have also found that FGF signals (FGF8 or bFGF) are able to promote strong Xnkx2.1 expression in animal caps neuralized by cerberus or by the combination of Chd+cer-S; conversely, the dominant negativeΔXFGFR-4a receptor almost completely prevents activation ofXnkx2.1 in animal caps conjugated to early organizer tissue, without preventing neural induction. These results strongly suggest that FGF signals may be essential for specification of the ventral forebrain. Similar conclusions have been recently reached by Shinya et al.(Shinya et al., 2001), who showed that inhibition of FGF signaling, particularly from FGF3 and FGF8,suppressed development of the ventral telencephalon in zebrafish embryos.

In conclusion, our work provides evidence that inductive signals leading to specification of early dorsal and ventral telencephalic territories can be reconstructed, at least in part, on naive animal caps, by specific combinations of signaling molecules. BMP inhibition, though able to possibly provide a general telencephalic fate, is not sufficient for dorsal and ventral telencephalic specification, as it does not activate the dorsal telencephalic markers Xemx1 and eomes or the ventral forebrain markerXnkx2.1. Strong Xnkx2.1 activation instead occurred when either FGF8 or bFGF were administered to neuralized caps. By contrast,activation of both Xemx1 and eomes expression was detected in animal caps injected with cerberus or the combination of Chd and N-terminal fragment of cerberus, cer-ΔC1, and eomes induction was reinforced by the further addition of FGF8 to the explants. A model that summarizes a possible interaction between the molecules and tissues we have studied for dorsoventral patterning of the telencephalon is shown inFig. 13. According to this,the anterior neural plate is induced in dorsal ectoderm by the action of BMP inhibitors, such as Chd; this initial forebrain-presumptive region may already express region-specific genes such as Xotx2, Xrx1 and XBF-1. Upon this ground, ventral forebrain fates would be induced by FGF signals,possibly secreted from the anterior neural ridge, and inhibited by the ADE. On the same ground, cerberus, possibly through its Wnt-inhibitory activity, and FGF signaling may cooperate in the activation of Xemx1 andeomes and the specification of dorsal telencephalon.

This paper is dedicated to the memory of Rosa Beddington, who shed light onto the role of the anterior endoderm in rostral brain development. Special thanks to Gaia Gestri and Irene Appolloni for help with in situ hybridization. We thank Richard Harland and Steve Wilson for comments on the manuscript, and Hazel Sive for suggestions in starting this work. We thank Tewis Bouwmeester,Jan Christian, Eddy De Robertis, Richard Harland, Peter Klein, Paul Krieg,Harumasa Okamoto, Nancy Papalopulu, Stefano Piccolo, Kenneth Ryan, Yoshiki Sasai, Hazel Sive, Jim Smith, David Wilkinson for plasmids; Donatella De Matienzo, Marzia Fabbri and Dean Pask for technical assistance; and Salvatore Di Maria and Denise O'Connor for frog care. G. L. is grateful to E. M. for support and encouragement. This work was supported by Cofinanziamento MURST-Università di Pisa and the Wellcome Trust. G. L. was supported by an EMBO long-term post-doctoral fellowship.