Positive and negative signals from mesoderm regulate the expression of mouse Otx2 in ectoderm explants

The vertebrate central nervous system (CNS) is divided into distinct domains along the anterior-posterior (A-P) axis early in its development, as judged by experimental analysis and the expression patterns of several families of transcription factors with homology to Drosophila pattern formation genes (reviewed in Kessel and Gruss, 1990; McGinnis and Krumlauf, 1992; Puelles and Rubinstein, 1993). Tissue explant/recombi-nation and transplantation experiments in amphibian and chick embryos have demonstrated that inductive interactions between ectoderm and mesoderm tissue, involving both vertical and planar signals, are critical in establishing the A-P patterning of the CNS (reviewed in Slack and Tannahill, 1992). Such studies have served to pinpoint the tissue source and the timing of the signals that initiate A-P patterning and have provided assays for testing putative signalling molecules in these species. However, the genetic analysis of these signalling molecules and the downstream genes that specify A-P position is largely taking place in the mouse, by targeted mutagenesis. In order to interpret the phenotypes observed, it is important to establish the basic parameters of the tissue interactions involved in neural induction in the mouse embryo itself. In an earlier report, we have shown that explant-recombination assays for neural induction are feasible in gastrulating mouse embryos. These experiments demonstrated that expression of the Engrailed (En) genes, En-1 and En-2, in the future midbrain and anterior hindbrain regions required a positive inductive signal from underlying anterior mesendoderm in the headfold stage embryo (Ang and Rossant, 1993), although En expression did not begin till the early somite stage.

In the present study, we have extended this analysis to determine the tissue interactions that lead to the restricted anterior expression domain of another putative A-P patterning gene, Otx2. In Drosophila two homeobox genes, orthodenticle (otd) (Finkelstein et al., 1990) and empty spiracles (ems) (Dalton et al., 1989), are involved in the process of patterning of anterior head structures. In mammals, two otd homologs, Otx1 and Otx2, have been isolated (Simeone et al., 1992a, 1993) as well as two ems homologs, Emx1 and Emx2 (Simeone et al., 1992a,b). Expression analysis at 9.5 days post-coitum (dpc) showed that these genes are expressed in nested A-P domains in the anterior CNS (Simeone et al., 1992a), suggesting that they may be involved in anterior specification in the vertebrate embryo. The genes are also expressed in a progressive temporal order, with Otx2 showing the earliest restricted anterior expression domain. This expression domain is established by the headfold stage, the time at which we had previously shown that a positive inductive signal from the mesoderm is required for later ectodermal expression of En genes (Ang and Rossant, 1993). However, Otx2 has been reported to be expressed throughout the ectoderm prior to gastrulation (Simeone et al., 1992b), suggesting that its continued expression at the anterior end may not require a positive signal from the underlying mesoderm, but rather that a repressive signal from the posterior mesoderm may restrict expression to the anterior end. Using explant-recombination assays, we have found evidence for such a negative signal emanating from posterior mesendoderm that helps to restrict expression of Otx2 to the anterior end at the headfold stage in vivo. This repression of Otx2 expression could be mimicked to some extent by treating embryos with exogenous retinoic acid (RA). However, we could also demon-strate that a positive signal from underlying mesendoderm is required to maintain anterior expression of Otx2. These findings are consistent with two-step models for neural induction (reviewed Gilbert and Saxen, 1993) in which an initial activat-ing signal would induce anterior forebrain fate, while a further signal would be required for subsequent restriction of anterior fate and progressive posteriorization.

Using a human Otd-related cDNA clone EST01828 (Adam et al., 1992), we screened a 8.5 day embryonic mouse cDNA library (kindly provided by Dr Brigid Hogan; Farhner et al., 1987) under low strin-gency hybridization conditions (5× Denhart’s, 5× SSC, 0.1% SDS at 42°C). Three independent phage clones were plaque purified and the EcoRI inserts were subcloned into pBluescript KS. One clone, pOtd9, showed specific expression in the anterior part of the embryo by whole-mount RNA in situ hybridization and, when sequenced, contained a partial homeodomain with homology to Otd. The insert from this clone was used to screen a 129Sv genomic library in λDash2 (gift of A. Reaume and R. Zirngibl) using high stringency hybridization conditions (5× Denhart’s, 5× SSC, 0.1%SDS at 65°C). The coding region in one positive genomic clone, g9, and the insert of pOtd9 were sequenced by the dideoxy sequencing procedure.

Mouse embryos at various stages of gestation were obtained by mating random-bred CD1 (Charles River, Canada), ROSA26 (Jackson Laboratory, USA) and C101 (S. Gasca, D. Hill and J. Rossant, unpublished) animals. The embryos were staged according to the scheme described in Ang and Rossant (1993). Briefly, prestreak embryos (6.0 –6.3 dpc) contained no mesoderm, while early streak embryos (6.5 –6.7 dpc) contained mesoderm only in the posterior half of the embryo. In mid-to late-streak embryos (7.0-7.3 dpc), the mesoderm had migrated halfway or fully to its anteriormost extent. At the headfold stage (7.5 –7.7 dpc), prominent neural folds were apparent at the anterior end of the embryo. Somite stage embryos (8.0 –8.5 dpc) contained between 1 and 10 somites. Embryos from each of the different stages are also illustrated in Fig. 2.

Ectoderm explants and recombination assays were performed as described previously (Ang and Rossant, 1993). Ectoderm explants from embryos of the early to mid-streak stages included only ectoderm from the anterior half of the embryo. From late-streak and headfold embryos, anterior explants included tissue anterior to and not including the node, while posterior refers to tissue from the remaining posterior half of the embryo. In the case of somite stage embryos, the anterior neural tube from forebrain to the anterior hindbrain region was cultured. The precise ectoderm pieces used in explants and recombinants are illustrated in Fig. 1A. All explants were cultured for 2 days in Dulbecco’s modified Eagles medium plus 15% fetal calf serum and then analyzed for expression of Otx2 by whole-mount RNA in situ hybridization. In experiments using the ROSA26 and C101 mouse lines, β-galactosidase activity was detected as described (Beddington et al., 1989).

Ectoderm explants were also cultured in media either containing 10⁻⁶ M and 10⁻⁷ M RA or in media containing carrier (DMSO) solutions alone for 24 hours. After this time, the explants were fixed. Otx2 expression in the RA-treated explants were subsequently assayed by whole-mount RNA in situ hybridization.

In situ hybridizations were carried out on whole-mount and sectioned material as described (Conlon and Rossant, 1992; Conlon and Herrmann, 1993; Ang and Rossant, 1993; Guillemot and Joyner, 1993). Single-stranded RNA probes labelled with digoxigenin- or ³⁵S-labelled UTP were synthesized from linearized template DNA as directed by manufacturer (Boehringer Mannheim Biochemicals). The Otx2 cDNA containing plasmid, pOtd9, was linearized with XbaI and transcribed in vitro using T3 polymerase to obtain an antisense tran-script. In some cases, sections were obtained from embryos stained by the whole-mount procedure according to Ang et al. (1993).

Explants were fixed in 4% paraformadehyde for an hour at 4°C and then washed twice for 10 minutes each in phosphate-buffered saline plus 0.1% Tween-20 (PBT). β-galactosidase activity was then detected according to Beddington et al. (1989), except that staining was performed at room temperature and for 4 hours only. Explants were then rinsed twice in PBT for 10 minutes each and stored in 100% methanol. Explants were subsequently analyzed for Otx2 expression by whole-mount in situ hybridization as described (Conlon and Herrmann, 1993), except that the bleaching step was omitted.

Pregnant CD-1 female mice were administered all trans retinoic acid (RA) essentially as described in Conlon and Rossant (1992). A dose of 20 mg/kg of maternal body weight was delivered by gavage. Females were treated at various stages of pregnancy, when the embryos were between the prestreak and early somite stage. Embryos were harvested 4 hours after treatment. Control mice were adminis-tered the carrier solution without RA. Otx2 expression in the RA-treated and control embryos was assayed by whole-mount RNA in situ hybridization.

Using a partial human cDNA with homology to the Drosophila Otd gene (Adam et al., 1992), we screened a mouse 8.5 day cDNA library and isolated a single 1 kb cDNA clone under low stringency hybridization conditions. The sequence of this clone showed a partial homeodomain with 100% amino acid identity to that of otd (Finkelstein et al., 1990) and the recently published mouse Otx1 and Otx2 genes (Simeone et al., 1993). Because only the 3′ part of the homeodomain, which is identical in Otx1 and Otx2, was present in this clone (Fig. 1B), we could not unequivocally determine the identity of the cDNA cloned. We therefore isolated mouse genomic DNA clones, using the cDNA clone as a probe under high stringency conditions, in order to obtain additional upstream exonic sequences. Two cognate genomic clones were obtained. One clone, g9, was mapped (Fig. 1B) and the exons were sequenced. Comparison of the deduced amino acid sequence of the two homeodomain-containing exons of g9 confirmed that we had initially isolated a partial Otx2 cDNA clone (data not shown).

Otx2 was expressed at the prestreak stage embryo as deter-mined by RNA in situ hybridization of whole embryos and tissue sections (Fig. 2A and data not shown). Expression at this stage and later in early streak stages was found throughout the epiblast which gives rise to the embryo proper, but not in extraembryonic tissues (Fig. 2A). Sagittal sections through early streak embryos revealed expression in both ectoderm and delaminating mesoderm but not in endoderm at the posterior end (Fig. 3A,B). However, between early and late streak stages, Otx2 expression became progressively restricted to the anterior half of the embryo (Fig. 2A,B). From late-streak to headfold stages, expression became further restricted to the anteriormost third of the embryo (Fig. 2C). At the late-streak and headfold stage, sagittal sections of whole-mount stained embryos showed expression in all three germ layers at the anterior end (Fig. 3C and data not shown). By the late headfold stage (Fig. 2C, right), a sharp posterior boundary of expression was apparent in the neural plate. To determine the position of this boundary relative to other patterning genes, we performed whole-mount RNA in situ hybridization with Otx2 and Hoxb-1 probes simultaneously. The anteriormost limit of expression of Hoxb-1 (Fig. 4A) is thought to represent the future rhom-bomere 3/4 boundary (Wilkinson et al., 1989; Frohman et al., 1990; Murphy and Hill, 1991). The posterior boundary of Otx2 lies anterior to this position as evidenced by the gap between Otx2 and Hoxb-1 expression (Fig. 4A).

At 8.0 –8.5 dpc (1 –15 somites), Otx2 was expressed in forebrain, midbrain and optic eminence of the central nervous system (CNS) (Figs 2D, 3D). Outside the nervous system, weaker expression was also found in the notochord, head mesenchyme and foregut at the same axial levels, as well as in the ectoderm and endoderm cells of the first branchial arch (Fig. 3D and data not shown). By 9.5 dpc, expression persisted in the neurectoderm and the first branchial arch, but was no longer detectable in the other tissues (Fig. 4B and data not shown). The sharp posterior boundary of Otx2 expression in the neural tube at this stage demarcates precisely the limit between midbrain and the hindbrain regions (Fig. 4B). In addition, Otx2 was also expressed in trigeminal neural crest cells that are condensing to form the sensory component of the trigeminal ganglion (data not shown).

The initial widespread expression of Otx2 and the progressive loss of expression from the posterior end during gastrulation suggested the possibility that Otx2 expression represented a ‘ground state’ which becomes restricted to the anterior end by a posteriorizing signal arising during gastrulation. This model predicts that Otx2 would be autonomous to the ectoderm from the earliest stages. This can be tested by an ectoderm ‘specifi-cation’ experiment, where specification is defined as the behaviour of the tissue when grown in isolation from other tissues (modified from Slack, 1991). For such experiments to be meaningful, it was first necessary to show that isolated ectoderm at the various stages did not contain or give rise to contaminating mesoderm tissues in explant cultures. We have previously shown that this is true for anterior ectoderm explants from mid-streak to headfold stage embryos (Ang and Rossant, 1993). To test if anterior ectoderm explants from the early streak stage contain or generate axial mesoderm cells, we made use again of the C101 mouse line, which has a lacZ genetrap vector inserted into its genome (S. Gasca, D. Hill and J. Rossant, unpublished data). In this line, lacZ expression marks the node and axial notochord tissue. 0/10 anterior ectoderm explants from early streak C101 embryos expressed lacZ after 2 days in culture, indicating that the explants did not contain axial mesoderm contamination. However, when total ectoderm was cultured from prestreak embryos, 80% of the cultures expressed mesoderm and primitive streak markers, namely Brachury (Herrmann, 1991) and Goosecoid (Blum et al., 1992). Total ectoderm was used in this case because it is not possible to distinguish the anterior from the posterior end of the embryo at this stage. Thus, ectoderm explants from the prestreak stage could not be used in our specification studies because of the presence of mesoderm contamination in the explants. However, anterior ectoderm from later stages, which is fated to form forebrain and midbrain regions and thus to express Otx2 (Lawson and Pedersen, 1991; Tam, 1989), is free of contaminating mesoderm tissue when explanted and can be used in specification and induction experiments.

To test for specification of Otx2 expression, ectoderm from the anterior half of early streak to somite stage embryos was dissected from underlying germ layers and cultured for two days. All anterior ectoderm explants isolated from mid-streak and later stages showed large patches of Otx2 expression after 2 days culture, as assayed by whole-mount RNA in situ hybridization (Table 1; Fig. 5B,C). In contrast, only 27% of anterior ectoderm explants of early streak embryos expressed Otx2 after culture and expression in positive cultures was limited to small groups of cells (Fig. 5A). These results demonstrate that the ectoderm is only specified to express Otx2 from the mid-streak stage onwards, despite the expression throughout the ectoderm at earlier stages in the intact embryo.

Since Otx2 expression is stable in isolated ectoderm only after the early streak stage, a positive signal from the advancing mesoderm may be required to stabilize Otx2 expression. To test this, we recombined anterior ectoderm from early streak embryos with mesendoderm from headfold stage embryos. Mesendoderm of headfold stage embryos was used in order to provide enough tissue to surround the ectoderm explants com-pletely. Two types of mesendoderm were tested for their ability to maintain Otx2 expression in ectoderm explants: anterior (which normally underlies Otx2⁺ ectoderm) and posterior mesendoderm (underlying Otx2⁻ ectoderm). 86% of recombi-nants with anterior mesendoderm showed strong Otx2 expression after 2 days culture (Table 2; Fig. 5D). In contrast, only 9% of the recombinants with posterior mesendoderm expressed Otx2 after culture (Fig. 5E; Table 2), which is even lower than observed in early streak ectoderm explants alone (Table 2). Mesendoderm from late-streak stage embryos was also tested in similar types of recombination explants and was also able to maintain Otx2 expression (data not shown). Mesendoderm tissue from younger embryos at the early and mid-streak stages was not tested because this tissue is too thin and difficult to manipulate at these stages.

It was important to show that the Otx2 expression observed in the recombinants was in the ectoderm component and not the mesendoderm, since anterior mesendoderm itself expresses Otx2 at the headfold stage. Control explants of anterior mesendoderm alone from headfold stage embryos were shown to be negative for Otx2 expression after 2 days of culture (Table 2). We further confirmed that expression was confined to ectoderm in recombinants by combining early streak ectoderm from wild-type embryos with headfold mesendoderm from embryos of the ROSA26 transgenic line. The ROSA26 mouse line contains a genetrap vector inserted into a ubiqui-tously expressed endogenous gene resulting in expression of lacZ in all cells throughout embryogenesis (Friedrich and Soriano, 1991). After two days of culture, these recombinants were analyzed for both β-galactosidase activity and Otx2 expression. 8/10 explants showed a patch of Otx2 expression, which was, in all cases, confined to the lacZ-negative ectoderm cells (Fig. 5G).

These experiments clearly demonstrate that anterior mesendoderm can stabilize Otx2 expression in anterior ectoderm. However, because the ectoderm already expressed Otx2 at the time of isolation, these experiments cannot determine if anterior mesendoderm can actively induce new expression of Otx2 in naive ectoderm. To test this, anterior ectoderm from early streak stage embryos was first cultured for 2 days, during which time the majority of the explants would turn off Otx2 expression (see above and Table 2), and then recombined with anterior mesendoderm from headfold stage embryos and cultured for two more days. 80% of these recom-binants expressed Otx2 after culture, showing that anterior mesendoderm is capable of inducing Otx2 expression in Otx2 negative ectoderm (Table 2; Fig. 5F). Experiments were also performed with posterior lateral ectoderm taken directly from late streak stage embryos, which does not express Otx2 at the time of isolation. Induction of Otx2 expression was again observed in 71% of the explants (Table 2). Posterior lateral ectoderm when cultured alone did not express Otx2 after 2 days (Table 2). These results demonstrate that anterior mesendo-derm can not only stabilize but also actively induce Otx2 expression in the ectoderm.

Since Otx2 expression was found in anterior but not in the posterior halves of embryos at late streak stage onwards, we also tested whether posterior mesendoderm could provide an active repressive signal to diminish expression posteriorly. Anterior ectoderm from late streak stage embryos was recom-bined with posterior mesendoderm from headfold stage embryos. Ectoderm from the late streak stage was chosen because 100% of ectoderm explants from this stage showed a large patch of strong Otx2 expression after 2 days culture (Fig. 5B -two rightmost explants). Mesendoderm from headfold stage was used because at this stage, there was clearly no more posterior expression in the embryonic ectoderm and enough mesendoderm tissue was available to surround most of the late streak ectoderm.

Anterior ectoderm of late streak embryos was recombined with either one or two pieces of anterior or posterior mesendo-derm of headfold stage embryos. After 2 days, these recombi-nants were scored for Otx2 expression, according to a scale based on visual estimate of the area and intensity of staining, where the strongest expression (++) was equivalent to that observed in control explants of anterior late-streak ectoderm alone or recombinants with one or two pieces of anterior headfold mesendoderm (Table 3; Fig. 6E). The range of expression seen in the experimental recombinants (++, +, −) is illustrated in Fig. 6A-C. If only those recombinants showing complete repression are included, then 32% and 54% repres-sion were observed with either one or two pieces of posterior mesendoderm, respectively (Table 3; Fig. 6A-C). If the weakly expressing (+) recombinants are also included as repressed, then the percentage of repression rises to 98% and 100% for the two experiments.

To demonstrate that lack of Otx2 expression in recombinants with posterior mesendoderm was not simply due to poor growth of the ectoderm tissue when surrounded by mesendo-derm tissue, we performed another control experiment. We recombined ectoderm tissue from late streak embryos from the ROSA26 line with headfold posterior mesendoderm from wild-type embryos. After 2 days in culture, 100% (5/5) of the recombination explants contained a large piece of lacZ-positive tissue after culture indicating that the ectoderm had survived and grown significantly in culture (Fig. 6D).

These results therefore demonstrate that a repressive signal from posterior mesendoderm may be involved in the restriction of Otx2 expression to the anterior of the headfold stage embryo.

There is increasing evidence that RA can act as a posteriorizing signal in the establishment of the A-P axis in vertebrate embryos. In particular, many different groups, including our own, have demonstrated that the addition of retinoic acid to mouse embryos can result in anterior shifts in the boundaries of expression of a number of Hox genes (Morris-Kay et al., 1991; Conlon and Rossant, 1992; Kessel, 1992; Marshall et al., 1992). Since Otx2 shows an anterior restricted expression pattern, excess posterior signal would be predicted to reduce the domain of Otx2 expression. When headfold stage embryos were treated in utero with RA and harvested 4 hours later, Otx2 expression was found to recede from the established posterior boundary (Fig. 7C). Treatment at early streak stage (Fig. 7A,B) caused a premature regression of Otx2 from the posterior end. By the early somite stage, Otx2 expression was no longer responsive to RA (Fig. 7D). This timing of sensitivity is identical to that of Hoxb-1 (Conlon and Rossant, 1992 and unpublished results). Thus, RA can cause rapid changes in Otx2 expression in vivo, consistent with its possible role as a posteriorizing signal.

We also tested whether RA could repress Otx2 expression in the ectoderm directly or whether mesendoderm was necessary for the action of RA. Ectoderm explants were isolated from late streak and headfold stage embryos and were treated with either 10⁻⁶ M or 10⁻⁷ M RA for 24 hours. We found that treatment of ectoderm explants in vitro with 10⁻⁶ M, but not 10⁻⁷ M of RA also reduced the numbers of explants showing Otx2 expression after culture by 60 –70% and expression was often weak in the remaining explants (Table 4 and data not shown). Again it is difficult to assess a quantitative effect in a qualitative assay but these results suggest that RA can mimic the effects of the posteriorizing signal both in vivo and in vitro, and can act directly on the ectoderm, although this may not be the route of actions in vivo.

Expression of the mammalian otd-related genes, Otx1 and Otx2, along with the ems-related genes, Emx1 and Emx2, divides up the developing forebrain and midbrain into nested anterior-posterior domains (Simeone et al., 1992a). Otx2 is the earliest expressed gene, showing a restricted anterior domain of expression in the future CNS by late gastrulation (Simeone et al., 1993). This gene is also expressed throughout the ectoderm prior to gastrulation but becomes progressively restricted to the anterior third of the embryo by the headfold stage (Simeone et al., 1993). In addition, we detected Otx2 expression in anterior mesoderm and endoderm during early gastrulation and in anterior regions of the notochord, branchial arches and foregut at early somite stages. Expression was not entirely neural-restricted until 9.5 days of development. This suggests that Otx2 may provide general anterior positional information in the early embryo, similar to the role of otd in Drosophila (Finkelstein and Perrimon, 1990; Cohen and Jurgens, 1990).

Observations from expression analysis alone might suggest that Otx2 expression is autonomous to the early ectoderm, but that cell-cell interactions are required progressively to repress its expression in the posterior of the embryo later in develop-ment. However, explant-recombination experiments revealed evidence for involvement of both positive and negative signals from underlying mesendoderm in defining the final expression domain of Otx2. Expression was autonomous to the ectoderm from the mid-streak stage onwards. In contrast, the majority of explants from embryos of the early streak stage lost Otx2 expression in culture, demonstrating that Otx2 expression is not yet specified at this stage. A small fraction of these explants (27%), however, did maintain a limited region of Otx2 expression in culture. One possibility is that these positive explants were from the oldest embryos within the early streak stage group where the anterior ectoderm could have been in contact with the advancing mesoderm. This hypothesis suggests that the positive Otx2 expression seen in the ectoderm explants from this early stage was actually not in the most anterior ectoderm, where Otx2 expression will later stabilize, but in the more posterior ectoderm closest to the advancing mesoderm. Thus, Otx2 expression may be stabilized in embryonic ectoderm by the advancing mesoderm in a progressive manner, posterior to anterior. This is very similar to the mechanism proposed for induction and maintenance of cement gland identity in Xenopus embryos (Sive et al., 1989). Direct evidence for a role for mesendoderm in maintaining Otx2 expression in the overlying ectoderm was provided by aggregating unspecified early streak ectoderm with anterior late streak or headfold mesendoderm. Nearly all such recombinants showed patches of Otx2 expression confined to the ectoderm after 2 days of culture. Posterior headfold mesendoderm did not allow maintenance of Otx2 expression, consistent with the notion that it is the leading edge of the mesoderm that has the capacity to maintain Otx2 expression. In the intact embryo, the signal from the advancing mesoderm presumably acts to maintain pre-existing expression in overlying ectoderm. However, we were able to demonstrate that anterior mesendo-derm can also actively induce Otx2 expression. Anterior mesendoderm can induce Otx2 expression in cultured early streak ectoderm explants that had lost Otx2 expression, as well as in non-expressing posterior lateral ectoderm from late streak stage embryos. This indicates that such mesendoderm has inducing capacity although in vivo this activity may be involved in the stabilization or maintenance, rather than induction, of Otx2 expression.

The progressive restriction of Otx2 expression to the anterior end as gastrulation proceeds could result from a passive loss of the positive signal at the posterior end as the mesoderm migrates forward, or from active repression by the later forming posterior mesendoderm. We obtained evidence for an active repressive signal present in posterior, but not anterior, mesendoderm acting on Otx2 expression in late-streak ectoderm. Nearly all recombinants showed some evidence of repression when compared with levels of Otx2 expression in control explants and recombinants. Loss of expression was not a result of loss of ectoderm cells, since all such recombinants contained ectoderm cells as judged by an independent lineage marker.

The tissue interactions involved specifically in regulating Otx2 expression have not been studied in other vertebrate species yet. However, these studies in mice can be interpreted in the context of some of the general models of neural induction that have been generated in other species. Classical experimen-tal studies in amphibian embryos using morphological markers have been largely consistent with the two-step model of neural induction, involving an initial induction to anterior neural fate (activation), followed by a progressive transformation to more posterior fates (reviewed in Slack and Tannahill, 1992; Gilbert and Saxen, 1993). More recent studies using molecular markers have also been essentially consistent with such a model (Sive et al., 1989; Lamb et al., 1993). The explant-recombination experiments described here are in agreement with the two-step model of neural induction. The whole ectoderm would be initially induced to or maintained in an anterior neural fate by a signal from anterior mesoderm as it migrates forward, as evidenced by widespread expression of Otx2 at early and mid-streak stages, and then progressively transformed by a second signal from later-forming posterior mesendoderm, as evidenced by loss of Otx2 expression in posterior regions from late-streak stage onwards. Interactions between these two signals could also act to sharpen the posterior boundary of Otx2 expression in the developing neurectoderm in vivo.

It is obviously of interest to identify signalling molecules that have these activities. The identification of noggin as a potent neural inducer in Xenopus has provided a possible candidate for the initial positive signal maintaining or inducing Otx2 expression. Noggin, a novel secreted protein expressed in the dorsal lip of the blastopore and the notochord, was first identified by its dorsalizing activity (Smith and Harland, 1992; Smith et al., 1993), but it can also induce neural tissue in gastrula ectoderm explants (Lamb et al., 1993). The neural tissue induced is anterior in nature, as assessed by the expression of Otx homologs (Lamb et al., 1993). Whether mouse noggin is expressed at the right time and place to be involved in Otx2 regulation in the mouse remains to be seen. We are currently testing whether noggin protein is capable of inducing Otx2 expression in mouse ectoderm explants. It has recently been shown that follistatin, an inhibitor of activin sig-nalling, can also induce anterior neural tissue in Xenopus animal caps (Hemmati-Brivanlou et al., 1994), providing another candidate molecule for anterior neural signalling, although the expression pattern in the mouse embryo does not necessarily support such a role (Albano et al., 1994).

As for the negative regulation of Otx2 expression, we have shown that exogenous RA can mimic the effects of the putative posterior repressing signal. The anterior expression domain of Otx2 was further restricted when gastrulating embryos were treated in utero with teratogenic doses of RA. RA alone was also sufficient to repress Otx2 expression in late streak and headfold ectoderm explants, demonstrating that RA can act directly on the ectoderm and does not necessary have to act on the mesoderm first. The repressive effect was observed at 10⁻⁶ M, but not at 10⁻⁷ M. Transgenic mice expressing a RA-responsive transgene (Rossant et al. 1991) as well as direct measurements of RA-metabolizing potential (Hogan et al., 1992) have indicated a posterior focus of retinoid activity in the gastrulating embryo, supporting the possibility that RA is not just mimicking the endogenous signal but might be directly involved in vivo. The importance of RA, as well as the putative anterior neural inducer, noggin, can be tested by mutational analysis in mice and the assays that we describe here will be useful in delineating the defects in such mutants.

We thank Kendrapasad Harpal for providing tissue sections and Dr François Guillemot for performing sectioned RNA in situ hybridization experiments. This work was supported by the Medical Research Council of Canada. J. R. is a Terry Fox Research Scientist of the National Cancer Institute of Canada and an International Scholar of the Howard Hughes Medical Institute, R. C. holds an MRC Fellowship.