Anterior neurectoderm is progressively induced during gastrulation: the role of the Xenopus homeobox gene orthodenticle

How is the anteroposterior axis of an embryo specified? This question has been an area of intense interest since the early days of experimental embryology. In 1924, Spemann and Mangold demonstrated that a region of tissue adjacent to the dorsal lip of the blastopore of an amphibian gastrula, when transplanted into the ventral side of a host embryo, induced a secondary embryonic axis (Spemann and Mangold, 1924). The donor tissue, since referred to as Spemann’s organizer, contributed mostly to the notochord with host tissues constituting the bulk of somitic mesoderm and neural tissue. These results and subsequent experiments demonstrated that the mesoderm provides signals directing neural induction and anteriorposterior positional specification during gastrulation (Hamburger, 1988).

Classic embryological studies also provided a framework for understanding the mechanisms governing anterior-posterior specification of the neural plate. Early models suggested that at least two signals were required for patterning of the anteriorposterior (A-P) axis; one signal instructed the development of anterior neural structures and the other instructed posterior neural structures (Waddington, 1940; Yamada, 1950; Nieuwkoop et al., 1952; Toivonen and Saxen, 1955). These two signals were hypothesized to exist as opposing gradients with maximum concentrations at the future anterior and posterior ends of the embryo (Toivonen and Sáxen, 1955). Subsequently, Nieuwkoop and coworkers provided experimental evidence supporting a two-step activation-transformation model (Nieuwkoop et al., 1952). According to this model, naive ectoderm initially receives an activating signal from the anterior mesoderm that instructs it to adopt an anterior neural fate (Sala, 1955). This process is then followed by a transformation step that posteriorizes the ectoderm providing A-P positional values (Nieuwkoop et al., 1952; Toivonen and Saxen, 1955).

The earlier work of Mangold suggested A-P positional information is transmitted by vertical induction from underlying mesoderm (Mangold, 1933). Later, Eyal-Giladi further suggested that the anterior neural-inducing signal emanates from the anterior mesoderm when this mesoderm first makes vertical contact with the overlying neurectoderm. This ectoderm would then subsequently receive caudalizing signals as it comes into contact with more posterior mesoderm passing beneath it during the subsequent movements of gastrulation (Fig. 1, Eyal-Giladi, 1954). Recently, however, the vertical signalling aspect of these models has been challenged by the discovery of planar signalling (Kintner and Melton; 1987, Ruiz i Altaba, 1992; Doniach et al., 1992; Keller et al., 1992). At the beginning of gastrulation, prior to involution, the organizer mesoderm and prospective neurectoderm exist as a continuous sheet of tissue in the embryo (see illustration in Fig. 8A). When these regions are excised from the embryo and allowed to develop in culture such that no vertical contact between mesoderm and ectoderm is permitted, convergence-extension and A-P specification of the ectoderm is still observed (Keller et al., 1992; Doniach et al., 1992). Since the mesodermal component of the explants is required for the processes of convergence-extension and A-P specification of the ectoderm, it has been suggested that diffusible signals can pass in a planar fashion from the mesoderm to the ectoderm in vivo.

Currently the molecules involved in planar and/or vertical signalling are unknown. However, two candidates have been identified which have neural-inducing activity in vitro and are expressed in prechordal and notochordal mesoderm in vivo during gastrulation. One of these, follistatin, is an inhibitor of activin signalling and is thought to block activin-mediated repression of neural induction (Hemmati-Brivanlou et al., 1994). This allows for the specification of neural ectoderm instead of epidermis, the ground state of the ectoderm. The second molecule, noggin, induces general neural markers as well as an anterior neural marker, otxA (which is likely to be an alternative genomic copy of Xotx2, described in this report), suggesting that noggin plays a role in patterning anterior neurectoderm (Lamb et al., 1993).

In addition to the involvement of extracellular factors in neural patterning, homeodomain proteins are important components in the specification of A-P information during the development of the metazoan body plan. Homeodomain proteins, such as bicoid in Drosophila and goosecoid in vertebrates, are involved in axial specification (Nusslein-Volhard, 1991; Cho et al., 1991b). The Drosophila homeodomain protein orthodenticle (otd), and other members of a class of homeodomain proteins, which include bicoid and goosecoid, is distinguished from other classes of homeodomain proteins by its unique DNAbinding specificity (Hanes and Brent, 1989; Treisman, 1989; Blumberg et al., 1991; Simeone et al., 1993). Drosophila otd is expressed as an anterior band at the cellular blastoderm stage (Finkelstein et al., 1990; Finkelstein and Perrimon, 1990). Mutation of otd results in a loss of antennal segment-derived sensory organs (Wieschaus et al., 1984). Two mouse homologs of otd (otx1 and otx2) have been identified that are expressed in nested fashion in the early mouse brain (Simeone et al., 1992). We sought to identify Xenopus homologs of mouse orthodenticle to serve as anterior neural markers in our studies of the role of prechordal mesoderm in the specification of anterior neurectoderm. While our work was being conducted, the mouse otx2 gene was shown to be expressed throughout the embryonic egg cylinder before gastrulation and localized to the anterior primitive streak during gastrulation (Simeone et al., 1993).

In this study, we describe the isolation and expression of the Xenopus Xotx2 gene, a homolog of mouse otx2. Xotx2 is initially expressed in Spemann’s organizer and is subsequently induced in the overlying presumptive posterior neurectoderm. As gastrulation proceeds, Xotx2 expression is induced in progressively more anterior ectoderm in register with underlying Xotx2-expressing mesoderm, in a pattern reminiscent of the Eyal-Giladi model for neural positional specification (Eyal-Giladi, 1954). Later, Xotx2 is expressed in the midbrain, forebrain, and prospective cement gland anlage from midgastrula to early tailbud stages and expression persists in the brain until later stages of embryogenesis. The cement gland resides anterior to the forebrain and is generally regarded as the most anterior ectodermal structure specified by neural-inducing signals (Sive et al., 1989; Drysdale and Elinson, 1993). We demonstrate that ectopic expression of Xotx2 in the skin induces the differentiation of ectopic cement glands as well as inducing a cement gland marker (XAG1) in isolated animal cap ectoderm.

A mixed gastrulae cDNA library (Cho et al., 1991b) was screened at low stringency with a random-primed probe derived from a 340-bp fragment of the Drosophila orthodenticle cDNA containing the home-odomain (Finkelstein et al., 1990). The conditions of hybridization were 42% formamide, 6× SSC, 50 mM sodium phosphate, pH 6.7, 5× Denhardt’s, 0.2% SDS, 0.1 mM EDTA, 5% dextran sulfate, and 100 μg torula RNA/ml at 37°C. Filters were washed in 0.5× SSC at 37°C and exposed to Kodak X-ray film. The cDNA inserts from candidate clones were in vivo excised according to the manufacturer’s instructions (Stratagene). Nine positive cDNA clones were isolated from a total of 500,000 plaques.

Total embryonic RNA was isolated from staged Xenopus embryos according to Sargent et al. (1986). 20 μg of total RNA was loaded into each lane of a formaldehyde gel and ethidium bromide fluorescence confirmed that all lanes were loaded approximately equally (data not shown). RNA was transferred to nitrocellulose filters according to standard procedures (Sambrook et al., 1989). Filters were hybridized at 65°C in 0.5 M sodium phosphate, pH 7.2, 7% SDS, 1 mM EDTA, and 5% dextran sulfate with a random-primed 1.5-kb XhoI fragment of Xotx2 that lacks the homeodomain.

The method of Harland (1991) was used. The Xotx2 probe consisted of digoxigenin-labelled antisense transcripts derived from the 1.5-kbp XhoI fragment of the Xotx2 cDNA which had been subcloned into pBluescript-KS. The Krox20 probe was generated by digestion with EcoRI and transcription with T7 RNA polymerase. Embryos were sectioned according to the procedure of Norenburg and Barrett (1987).

Embryos were fertilized and microinjected with synthetic mRNAs as described previously (Cho et al., 1991b). Mesoderm-ectoderm conjugates were prepared from a pair of early gastrula stage albino embryos, one member of which was injected at the 2-4 cell stage with lysinated-rhodamine dextran (relative molecular mass of 10×10³. Molecular Probes Inc., Oregon). Involuted organizer mesoderm from an injected embryo was removed at stage 10.25 and conjugated with animal cap ectoderm derived from an uninjected stage 10.25-10.5 embryo. Conjugates were allowed to develop in 0.3× Barth until sibling embryos reached stages 14-16 (Nieuwkoop and Faber, 1967), and were then fixed for whole-mount in situ hybridization.

Keller sandwiches were prepared essentially as described by Keller and Danilchik (1988), Keller (1991) and Doniach et al. (1992). Early gastrula embryos were lightly stained with Nile Blue and manually dechorionated prior to staging according to Nieuwkoop and Faber (1967), and Keller (1991, and personal communication) as follows. Stage 10−, bottle cell constriction just barely becomes visible on the dorsal side; stage 10.0, bottle cells appear as a straight line and constriction has advanced as is apparent from the intensity of concentration of Nile Blue and the width of the constriction, however the lip has not yet begun to pucker inward significantly; stage 10+, the bottle cells occupy greater width, begin to curve ventrally, and the lip puckers inward substantially; stage 10.25, the bottle cells have progressed laterally to occupy an approximate 180° arc and the dorsal lip occupies an approximate 75° arc; stage 10.5, the bottle cells have progressed to the ventral midline and the dorsal lip occupies an approximate 180° arc. To construct Keller sandwiches, two explants are dissected from two identically staged embryos and sandwiched together with their inner surfaces apposed after removing highly migratory head mesoderm from the inner surface of the explants. Surgeries were performed using eye brow hair knives and hair loops and sandwich explants were allowed to develop in Sater’s modified Danilchik’s solution (Keller et al., 1992) until sibling embryos reached stage 17-19 (Nieuwkoop and Faber, 1967). Sandwiches were then fixed for whole-mount in situ hybridization.

Total cellular RNA isolated from animal cap ectoderm was reverse transcribed at 37°C for 60 minutes using random hexamers in a 50 μl reaction volume as described by Makino et al. (1990). cDNA was then PCR amplified in a 10 μl reaction volume containing 300 ng of primer. The conditions for the thermal profile were as follows: 94°C, 1 minute; 55°C, 1 minute; 72°C, 1 minute, for 24-30 cycles. A final extension step of 72°C for 10 minutes was included. Amplification of all fragments was within the linear range of the PCR reaction after 30 cycles except histone H4 which was saturated after 24 cycles of PCR. All PCR products were analyzed on a 5% urea-polyacrylamide gel. The oligos used for PCR amplification were as follows: N-CAM F:5′AGATGCAGTCATTATTTGTGATGTC-3′, R:5′-CTGGATGTCC-TTATAGTTGATCTC-3′ (Collett and Steele, 1993); Xotx2 F:5′-GGAGGCCAAAACAAAGTG-3′, :5′-TCATGGGGTAGGTCCTCT-3′; XAG1 F:5′-GAGTTGCTTCTCTGGCA-3′, R:5′-CTGACTGTC-CGATCAGAC-3′ (H.Sive, personal communication); histone H4 F:5′-CGGGATAACATTCAGGGTATCACT-3′, R:5′-ATCCATG-GCGGTAACTGTCTTCCT-3′ (kindly provided by B. Blumberg).

A cDNA library constructed from mixed gastrula stages of Xenopus laevis was screened at low stringency with a probe containing the homeodomain region of the Drosophila orthodenticle (otd) gene (Finkelstein et al., 1990). The longest cDNA isolated was approximately 2.1 kb and analysis of the DNA sequence (GenBank accession numbers U19813 and U19814) revealed an 870-bp open reading frame (ORF) flanked by 207 bp of 5′ and approximately 1 kb of 3′ untranslated sequence, respectively. The homeodomains of all the orthodenticle homologs are highly conserved (Simeone et al., 1993), with the Xenopus homolog differing at only 3 of 60 amino acid residues from the homeodomain sequence of the Drosophila orthodenticle protein. Furthermore, the homeodomain of the Xenopus protein differs by a single amino acid substitution from the mouse and human otx2 proteins and at two amino acid residues from the mouse and human otx1 proteins. The orthodenticle homeodomains, including this Xenopus homolog, contain a lysine at position 9 of the third helix, putting them in the same homeodomain class as bicoid and goosecoid. This lysine residue has been demonstrated to affect DNA-binding specificity of homeodomain proteins (Hanes and Brent, 1989; Treisman et al., 1989). Mouse otx2 and goosecoid have both been demonstrated to bind bicoid-like DNA binding sites (Blumberg et al., 1991; Simeone et al., 1993).

Throughout the remainder of the protein, the conservation in amino acid sequence and the position of corresponding gaps (data not shown, see Simeone et al., 1993) strongly suggest that this Xenopus gene is a homolog of the mouse, human and zebrafish otx2 genes (the Xenopus protein is 92% identical to the otx2 proteins) and not that of otx1 (at shared residues the Xenopus protein is approximately 70% identical to the otx1 proteins). Comparison of the Xotx2 amino acid sequence with the available sequence from Xenopus otxA (Lamb et al., 1993) demonstrates that these two genes encode related but not identical molecules. These differences occur in the sequence encoding the 5′ untranslated regions of these mRNA’s as well as in the amino acid sequence. An amino acid sequence that would unambiguously identify otxA as an otx2 homolog is presently not available, but it seems likely that Xotx2 and otxA represent divergent genomic copies of different otx2 genes present in the pseudotetraploid genome of Xenopus laevis (Graf and Kobel, 1991). Since the sequence of the gene described in this study is virtually identical to the otx2 genes of mouse, human and zebrafish, we refer to our Xenopus gene as Xotx2.

We examined the temporal expression of Xotx2 RNA by northern analysis using a 1.5-kb probe consisting of sequences downstream of the homeodomain and the entirety of the 3′ untranslated region (Fig. 2). A single band of approximately 2.1 kb was detected corresponding to the size of the Xotx2 cDNA, suggesting that this cDNA is approximately full-length. Xotx2 RNA was not detected at early stages of development up to and including blastula stages, but is easily detected during gastrula stages and persists at high levels at least until swimming tadpole stages.

The distribution of Xotx2 RNA in the developing Xenopus embryo was examined using whole-mount in situ hybridization (Harland, 1991). Xotx2 transcripts are first detected in the early gastrula embryo. These results differ slightly from the reported expression pattern of otxA in that otxA is expressed in the late blastula throughout the marginal zone and quickly becomes restricted to the superficial layer on the dorsal side (Lamb et al., 1993). We cannot exclude the possibility that Xotx2 also behaves in this fashion immediately prior to onset of bottle cell constriction. Xotx2 hybridization in the gastrula is localized to cells of Spemann’s organizer in a region consisting of an approximately 45° arc from the mid-sagittal line (Fig. 3A), in a pattern indistinguishable from that of Xenopus goosecoid (Cho et al., 1991b; Inoue et al., unpublished data). As gastrulation proceeds further, the abundance of Xotx2 RNA in the ectoderm increases, moves anteriorly, and by stage 11 expands mediolaterally (data not shown). At later stages of gastrulation, as mesoderm and ectoderm undergo continued gastrulation movements, the Xotx2 hybridization signal moves both further anteriorly and concomitantly fans out to form a large cap over the anterior end (Fig. 3B).

In order to determine the identity of cells expressing Xotx2 during gastrulation, whole-mount in situ hybridized embryos were sectioned and further analyzed. Near the beginning of gastrulation, Xotx2-expressing cells of the organizer are found in the deep cell layers (Fig. 3C) comprising the presumptive head mesoderm (Keller, 1975, 1976). During gastrulation these cells are the first to invaginate and constitute the leading edge of migrating mesoderm, which travels along the inner surface of the blastocoel roof. At a slightly later stage (approximately stage 10.25) Xotx2 expression is detectable, albeit weakly but reproducibly, in ectoderm overlying the anterior-most portion of the mesodermal signal (Fig. 3D). In addition, mesodermal Xotx2 RNA can be detected more posteriorly in cells which constitute the presumptive anterior notochord (Figs 3B and 8).

This pattern of expression is maintained at the anterior end of the embryo throughout the remainder of gastrulation (Fig. 3B,E). In the late gastrula embryo (stage 12) the domain of Xotx2 expression overlaps a region of columnar cells (Fig. 3E, see arrow) which constitute the future brain (Nieuwkoop and Faber, 1967; Keller, 1980). Xotx2 expression is also found in a region anterior to the columnar cells (see arrow in Fig. 4B) prior to morphological delineation of the anterior neural plate boundary, which subsequently bisects the Xotx2 expression domain into anterior neural plate and the prospective cement gland (Figs 3E, 4A,B; for comparison see also Jamrich and Sato, 1989).

In open neural plate stages, the expression of Xotx2 is found in ectoderm and mesodermal cells of the prechordal plate (Fig. 4A and B). As in the gastrula, the ectodermal expression continues along the anterior-posterior axis approximately in register with the mesodermal domain of expression. Prechordal mesodermal Xotx2 expression occupies a narrow strip in the dorsal midline approximately 150 μm in width (data not shown). Ectodermal Xotx2 expression overlaps the anterior border of the neural plate and persists weakly in the cement gland anlage until early tailbud stages (Fig. 4A-D).

During subsequent development of the neural tube, Xotx2 is detected in the anterior brain and developing eye anlage (Fig. 4C). This Xotx2 domain eventually develops into the forebrain and midbrain (Fig. 4D,E), with the posterior border of Xotx2 expression demarcating the midbrain-hindbrain junction as revealed by staining with Krox20 (Fig. 4F) and engrailed-2 (data not shown) probes. Krox20 is a marker for hindbrain rhombomeres 3 and 5 whereas engrailed-2 overlaps the midbrain-hindbrain junction (Papalopulu et al., 1991; Hemmati-Brivanlou et al., 1991). Later, expression of Xotx2 decreases in the presumptive diencephalon prior to tailbud stage 28 and Xotx2 marks the presumptive telencephalon and mesencephalon (Fig. 4E,F).

A role for ectodermal Xotx2 expression was examined by microinjection of Xotx2 RNA into the animal pole region. These experiments demonstrated that up to 76% (Table 1, Exp.1) of injected embryos developed ectopic cement glands in the skin (Fig. 5A-C). Cement glands were first visible at neurula stages due to concentration of pigment granules during cement gland differentiation. Larger ectopic cement glands secreted mucus (data not shown) and sectioning revealed the presence of columnarized cells containing pigment granules (compare Fig. 5, panels D,E and F) demonstrating bona fide cement gland induction. No other obvious ectopic structures could be identified in the vicinity of ectopically induced cement glands. In contrast, microinjection of transcripts encoding other homeodomain proteins such as goosecoid (data not shown, and Cho et al., 1991b), XlHbox1 (Wright et al., 1989), XlHbox6 (Cho et al., 1991), and non-homeodomain transcripts such as β-globin (our unpublished observation) and lacZ (Inoue et al., unpublished data) into animal pole blastomeres did not induce ectopic cement glands.

To assess the ability of Xotx2 to induce cement glands in ectoderm in the absence of inductive contributions from other non-ectodermal structures in the embryo, we injected Xotx2 RNA into the animal hemisphere of 4-cell embryos and dissected animal caps at the blastula stage. These ectodermal explants were cultured until sibling embryos had reached the late neurula stage. Total RNA was isolated and subjected to RTPCR analysis for relative levels of RNA encoding the cement gland marker XAG1. XAG1 RNA is expressed in vivo at low levels during gastrulation and is dramatically up-regulated in the ectoderm upon cement gland induction (Sive et al., 1989). In uninjected control animal caps XAG1 was detected at low levels (Fig. 6A, lane 2). However, in animal caps expressing Xotx2 we found a large increase in XAG1 RNA levels (Fig. 6A, lane 3), demonstrating that Xotx2 can induce a marker of cement gland differentiation in isolated animal cap ectoderm.

The cement gland is generally considered to be the most anterior structure patterned in response to neural inductive signals (Sive et al., 1989; Drysdale and Elinson, 1993). Since Xotx2 induces ectopic cement glands in vivo as well as a cement gland marker gene in isolated animal cap ectoderm, and is expressed in neural tissue throughout early embryogenesis, we examined the possibility that Xotx2 may also play a role in the induction of neural tissue. Induction of N-CAM was detected in Xotx2-injected animal caps (Fig. 6B, lane 4) although the level of induction is lower than that of noggininjected animal caps (Fig. 6B, lane 3). From these experiments, we conclude that ectopic expression of Xotx2 appears to induce N-CAM, albeit less efficiently than the induction of N-CAM by noggin. Although Xotx2 may play a role in neural induction, it is not apparent whether the induction of N-CAM observed has any in vivo relevance. Therefore, a more extensive examination using other pan-neural and position-specific markers will be necessary to provide a thorough understanding of the role of Xotx2 in this process.

Since Xotx2 is expressed in the forebrain and midbrain, and appeared to induce N-CAM in isolated animal caps, we attempted to obtain non-cement gland neural phenotypes by microinjection of Xotx2 into blastomeres at a variety of developmental stages. Injection into blastomeres contributing to prospective neurectoderm at various stages typically resulted in a high frequency of embryos exhibiting spina bifida (Table 1, and data not shown), as well as embryos with truncated (or short) tail structures (Fig. 5H) which were often associated with incompletely closed blastopores (Fig. 5G and data not shown). The short-tail phenotype does not appear to reflect a posterior-toanterior transformation since diminution of the tail is not accompanied by enlargement of the head or the remainder of the trunk. Therefore, we conclude that these phenotypic analyses have failed to demonstrate an in vivo function for Xotx2 in neural specification.

Since ectodermal expression of Xotx2 followed the migration of the underlying anterior mesoderm during gastrulation, we tested the hypothesis that vertical contact by anterior mesoderm was required for the ectodermal expression pattern observed. The lineage tracer lysinated rhodamine-dextran (LRD) was injected into the marginal zone of all blastomeres of twoand four-cell stage embryos (Fig. 7A). At early gastrula stage (stage 10.25), involuting anterior mesoderm was dissected from these embryos and wrapped in unlabelled animal cap ectoderm isolated from sibling embryos at early gastrula stages 10.25-10.5. When these mesoderm-ectoderm conjugates reached early neurula stages (stage 14-16), they were fixed and subjected to whole-mount in situ hybridization using an Xotx2 probe. No hybridization signal was detected in animal cap ectoderm controls that lacked mesoderm (Fig. 7C), demonstrating that Xotx2 expression in the ectoderm is not cell autonomous. Conjugates strongly expressed Xotx2 and most of this signal was present in the unlabelled ectoderm (Fig. 7B,C). In addition, cement glands were also induced by underlying mesoderm (data not shown). We conclude that ectodermal Xotx2 expression and cement gland differentiation can be induced by underlying anterior mesoderm via vertical signalling.

Since vertical signalling appeared to play an important role in establishing the in vivo pattern of neurectodermal Xotx2 expression, we sought to examine Xotx2 expression in the absence of vertical interactions. This is also important as it remained possible that ectodermal Xotx2 expression may also be induced by signals travelling through the plane of tissue shared by the mesoderm and ectoderm at the beginning of gastrulation (Keller and Danilchik, 1988; Doniach et al., 1992). We examined these issues using Keller sandwich explants. Since it has recently been demonstrated that anterior mesoderm can provide vertical signals to overlying ectoderm in the stage 10+ gastrula (A. Posnansky and R. Keller, personal communication), we constructed a series of Keller sandwiches from embryos isolated at stages 10−, 10.0, 10+, 10.25, 10.5 (Nieuwkoop and Faber, 1967; Keller, 1991; and Keller, personal communication) as described in the Materials and Methods. In the stage 10− embryo (Fig. 8A, left), the head and notochordal mesoderm reside as a continuous strip of cells (Keller, personal communication) and have not begun involution. In the stage 10.0 embryo, head mesoderm has begun involution but has not yet contacted ectoderm fate mapping to the posterior neural plate (Keller et al., 1992; Keller, personal communication). During the subsequent stages of gastrulation, head mesoderm and notochordal mesoderm migrate further anteriorly and by approximately stage 11 -11.5 the head mesoderm comes to underlie ectoderm of the prospective forebrain (Keller et al., 1992). Sandwich explants were assessed for their pattern of Xotx2 expression by whole-mount in situ hybridization at equivalent late neurula stages (stages 17-19).

The expression of Xotx2 could be detected in the mesoderm and the ectoderm at the anterior extremes of the explants (100%, n=26, and Fig. 8). The mesodermal staining corresponds to the anterior mesoderm comprising the presumptive prechordal mesoderm and anterior notochord, while the ectodermal stain corresponds to the anterior ectodermal domain of Xotx2 expression (Fig. 8). Interestingly, Keller sandwiches constructed at different stages of development all had similar patterns of ectodermal Xotx2 expression (Fig. 8A and B, and data not shown) at late neurula stage, despite the differences in the strength of induction of the hindbrain-specific marker Krox20 (compare Fig. 8A and B), a marker sensitive to vertical inductive influences (A. Posnansky and R. Keller, personal communication). Therefore, we concluded that both vertical and planar signals may contribute to the ectodermal expression of Xotx2 in the developing Xenopus embryo.

The spatial and temporal expression of Xotx2 in the anterior mesoderm appears virtually identical to the expression of the homeobox gene goosecoid (Figure 3A, Cho et al., 1991b; Inoue et al., unpublished data). In addition, all treatments that affect goosecoid gene expression (LiCl, UV, retinoic acid, activin, and XWnt8) were shown to affect Xotx2 expression in a similar manner (data not shown, Cho et al., 1991b; Inoue et al., unpublished data; Christian and Moon, 1993). Since bicoid regulates the expression of orthodenticle in the cellular blastoderm stage Drosophila embryo, it is tempting to speculate that this model for orthodenticle gene regulation might be extended to vertebrates. Therefore, synthetic goosecoid mRNA was coinjected with a lineage tracer (LRD) into a single C-tier blastomere at the 32-cell stage and then embryos were examined for Xotx2 expression by whole-mount in situ hybridization at the early gastrula stage. Xotx2 RNA colocalized with the lineage tracer in all embryos examined (6 of 6 ventrally injected embryos, Figure 3F). We conclude that Xotx2 may be a target gene regulated by goosecoid.

Analysis of the spatial expression of Xotx2 during gastrulation has demonstrated that Xotx2 expression is initially absent in the ectoderm. But, at a slightly later stage, Xotx2 is expressed weakly in ectoderm overlying the head mesoderm. This ectodermal signal increases in strength thereafter. Since isolated animal cap ectoderm does not express Xotx2 in the absence of inducing signals emanating from the mesoderm, the underlying mesoderm appears to be important for the proper induction of expression of Xotx2 in the ectoderm. Neurectodermal Xotx2 expression is in approximate register with the underlying Xotx2-expressing anterior mesoderm throughout the remainder of gastrulation and neurulation. Although in register expression between the mesoderm and ectoderm has previously been described for other homeodomain genes such as XlHbox1 (Xenopus HoxC6) and HoxB1 (in mouse), these genes are expressed later in development, after the morphogenetic movements of gastrulation have been completed (De Robertis et al., 1989; Frohman et al., 1990). Xotx2 is the earliest expressed gene indicating that position-specific determination occurs across germ layers during gastrulation.

Morphological features defining position in the ectoderm are absent during the majority of gastrulation and no other molecular markers exist to define the position of prospective forebrain, midbrain, hindbrain, and spinal cord during these stages of development. Therefore a precise analysis relating the position of Xotx2 expression and the boundaries of these prospective structures in sections of whole-mount in situ preparations is difficult. However, recent fate-mapping experiments by Keller and coworkers have demonstrated that anterior mesoderm initially contacts posterior neurectoderm of the prospective spinal cord during late stage 10+ (A. Poznansky and R. Keller, personal communication) and does not progress anteriorly to underlie and contact the prospective forebrain ectoderm until stages 11-11.5 (Keller et al., 1992). Since Xotx2 is already expressed in the ectoderm at stage 10.25, well before anterior mesoderm underlies and contacts prospective anterior neurectoderm, the Xotx2-expressing ectodermal cells must represent prospective posterior neurectoderm. The observations that Xotx2 is induced in posterior ectoderm shortly after invagination of anterior mesoderm, remains in register with underlying anterior mesoderm during gastrulation, and is ultimately expressed in the anterior neural plate at the end of gastrulation suggest that a ‘wave’ of transient gene expression moves through the sheet of ectoderm from posterior to anterior.

Sive and coworkers have demonstrated that a wave of inductive activity moves through the ectoderm during gastrulation resulting in the later development of the cement gland (Sive et al., 1989). Our results suggesting a role for Xotx2 in cement gland induction are consistent with this view and further suggest that the pattern of ectodermal Xotx2 expression is an early manifestation of the inductive wave described by Sive and coworkers. Expression of the cement gland markers (Sive et al., 1989) and Xotx2 RNA during gastrulation (our present finding) correlates well with the model for neural induction proposed by Eyal-Giladi (1954; see Fig. 1). This model, a modification of the activation-transformation model of Nieuwkoop et al. (1952), proposes that ectoderm is first activated to an anterior neural state by underlying mesoderm early in gastrulation. Following activation of the overlying ectoderm, the ectoderm undergoes a transformation in response to signals produced by subsequent contact with more posterior mesoderm slightly later in gastrulation. The observed pattern of expression of the anterior brain marker Xotx2 during gastrulation (and the inductive wave described by Sive and coworkers) may reflect a response to the early activation event proposed by Nieuwkoop (1952) and Eyal-Giladi (1954).

Domains of ectodermal and mesodermal expression of Xotx2 are in register throughout gastrulation, suggesting that the underlying mesoderm mediates activation of Xotx2 expression in the ectoderm through vertical signalling as suggested by the model of Eyal-Giladi (1954). Vertical induction of Xotx2 in the ectoderm was examined by constructing mesoderm-ectoderm conjugates in which lineage-labelled anterior mesoderm from an early gastrula was wrapped in unlabelled naive animal cap ectoderm, also from an early gastrula. Xotx2 expression was strongly induced in the ectoderm of these conjugates, demonstrating that vertical signalling can induce ectodermallyexpressed Xotx2 during gastrulation.

Induction of numerous neural-specific markers has been demonstrated in the absence of underlying mesoderm in Keller sandwich explants (Doniach et al., 1992; Papalopulu and Kintner, 1993). Under these conditions the mesoderm and ectoderm are in planar contact as attempts by the mesoderm to underlie the ectoderm are mechanically prohibited. One of these markers, a Xenopus distal-less homolog expressed in the forebrain, is expressed in Keller sandwiches, indicating that planar signals are capable of patterning expression of anterior markers (Papalopulu and Kintner, 1993). We addressed whether Xotx2 may also be induced by planar signalling for three reasons. Firstly, it is not yet known whether planar signals alone can induce all anterior marker genes. Secondly, the in-register expression pattern of Xotx2 during gastrulation suggested that Xotx2 induction may be mediated by vertical signals in vivo. Thirdly, it has recently been demonstrated that vertical signalling occurs as early as stage 10+ (A. Posnansky and R. Keller, personal communication), suggesting that vertical signalling may have influenced the results of previous studies that concluded a role for planar induction in the A-P patterning of neurectoderm in Keller sandwiches. Therefore, we constructed Keller sandwiches from gastrulae in which mesoderm had not yet involuted as well as later stage gastrulae with varying extents of mesodermal involution. However, all of the Keller sandwiches expressed Xotx2 strongly in the ectoderm even in the absence of any involuting mesoderm, demonstrating that planar signals may also contribute to the expression of Xotx2 in the anterior ectoderm.

Although molecules responsible for planar signalling are largely unknown, noggin has been suggested to play a role in patterning anterior neurectoderm (Lamb et al., 1993). The fact that Xotx2 is induced by noggin in isolated animal cap ectoderm, and that Xotx2 is expressed in the ectoderm of Keller sandwiches, suggest that noggin may act as a planar signal to induce Xotx2. noggin is capable of inducing various anterior neural marker genes as well as cement glands, the most anterior ectodermal structures specified by neural inducing signals (Lamb et al., 1993). Since the cement gland anlage lies at the anterior boundary between the skin and the neural plate and the domain of Xotx2 expression overlaps this boundary, it is noteworthy that ectopic expression of Xotx2 in ectoderm induces cement gland differentiation in the skin and induces the cement gland marker XAG1 in isolated animal cap ectoderm. These results suggest that differentiation of the cement gland at the anterior neural plate border may require overlapping fields of information (Drysdale and Elinson, 1993) specifying both skin and expression of Xotx2.

Development of the head, and the rules governing its formation appears to be quite distinct from those of trunk. While the formation of trunk structures is dependent on a combination of HOX gene expression (HOX codes; Hunt and Krumlauf, 1992), head patterning appears to be regulated by other classes of non-clustered homeobox genes such as orthodenticle (Finkelstein and Perrimon, 1990; Cohen and Jurgens, 1991). Conservation of orthodenticle homeobox sequences among disparate species such as Drosophila, Xenopus, and mouse and their common localized expression patterns in anterior head regions suggest that the underlying molecular mechanisms of head specification may also have been conserved throughout evolution.

The expression patterns of the Xenopus Xotx2 and mouse otx2 genes are quite similar. Both genes are expressed in anterior neurectoderm by the end of gastrulation and later expressed in presumptive foreand mid-brain. Despite these similarities between Xenopus and mouse otx2 expression patterns, some differences do exist. For example, in mouse, otx2 transcripts are expressed abundantly in the epiblast of prestreak embryos and otx2 expression persists in the entire embryonic ectoderm after the onset of gastrulation. However in Xenopus, Xotx2 transcripts in ectoderm, prior to gastrulation, have not been detected by whole-mount in situ hybridization. A difference is also found in their mesodermal expression patterns. Although the expression of Xotx2 is strongly detected in the organizer mesoderm of gastrulating Xenopus embryos, the expression of mouse otx2 in anterior primitive streak (node), the organizer equivalent in the mouse, was not reported (Simeone et al., 1993). Whether the apparent lack of mouse otx2 expression in mesoderm is due to insufficiently sensitive methods to detect the transcripts or represents a situation unique to the mouse system has yet to be determined. Despite some dissimilarities in the expression of Xenopus and mouse otx2, we conclude that Xotx2 is most closely related to mouse otx2 due to similarities in their amino acid sequences and overall patterns of expression. Since Xotx2 is the earliest anterior neural gene currently available in Xenopus and this gene appears to be well conserved among vertebrates, further characterization of the function and regulation of Xotx2 is likely to provide insights into the mechanisms of vertebrate neural induction.

We are grateful to Dr David Finkelstein for providing the cDNA encoding Drosophila orthodenticle. We thank Drs N. Papalopulu, and R. Harland for providing us with the Xenopus Krox20 and engrailed2 probes and noggin cDNA, respectively. We thank Dr H. Sive for advice on PCR primers for XAG1 and Dr B. Blumberg for primers to histone H4. We thank Drs C. Hashimoto, K. Inoue, A. Poznansky, D. Gardner and T. Doniach for helpful discussions on whole-mount in situ hybridization. We thank Drs C. Kintner, T. Doniach, A. Posnansky, R. Keller for technical advice on preparation of Keller sandwiches and for sharing data prior to publication and Richard Harland for sharing his otxA results prior to publication. We gratefully acknowledge the assistance of Sam Kim with sequencing portions of the Xotx2 cDNA. We also acknowledge M. Artinger, D. Bittner and M. Selleck for critical readings of the manuscript. This work was supported by NIH grant HD29507-02, Basil O’Connor Starter Research Award No.5-FY93-0795, and grant JFRA-431 from the American Cancer Society to K. W. Y. C. I. L. B. was supported by a NIH training grant.

While this manuscript was in review, further analyses of the mouse and zebrafish otx2 expression patterns were published (Ang, S.-L., Conlon, R. A., Jin, O. and Rossant, J. (1994). Development120, 2979-2989; Li, Y., Allende, M. L., Finkelstein, R. and Weinberg, E. S. (1994). Mech. Dev. 48, 229-244.) In agreement with our results in the frog.