A single morphogenetic field gives rise to two retina primordia under the influence of the prechordal plate

Formation of bilaterally symmetric and asymmetric structures have been studied in vertebrates since the beginning of experimental embryology. It is known that several strategies are used to form these structures in the embryo. Most of the structures on both sides of an animal, such as limbs, emerge independently from separate primordia on each side of the embryo. Formation of a single heart, on the other hand, results from the fusion of two originally separate primordia (Copenhaver, 1926; Jacobs and Fraser, 1994). Here we address the question how two symmetric eyes form in the anterior neural plate.

Interest in how two eyes form in normal embryogenesis has, in part, been motivated by attempts to understand developmental origins of cyclopean eyes in abnormal situations. A single eye or two fused eyes exist in some human fetuses suffering from holoprosencephaly (Cohen, 1989; Sperber et al. 1987; Kallen et al. 1992; Muenke et al., 1994), a developmental abnormality occurring in about 1 out of 16,000 live births and 1 in 250 terminated conceptuses (Muenke et al., 1994; Cohen, 1989). Cyclopean eyes in various species have been described with reasonable accuracy since the sixteenth century. Two mechanistic explanations were offered in the last century. Speer (1819) and Meckel (1826) proposed that cyclopia resulted from the fusion of two originally separate eyes, whereas Huschke (1832) proposed that there was a single vesicle which later became two optic vesicles in normal embryos and that cyclopean eyes resulted from a failure in separating the single vesicle (reviewed by Adelmann, 1936a). The debate continued into this century. Spemann believed in the existence of two separate optic anlagen in the early neural plate (Fig. 1A; Spemann, 1938; Lewis, 1907; Stockard, 1913b; Adelmann, 1936a). His view of separate eye anlagen in normal embryos led to his interpretation of cyclopean eyes as resulting from the fusion of two eye anlagen. His opinions were shared by King (1905) and Lewis (1907, 1909). In contrast, Stockard (1907a,b, 1908, 1909a,b, 1910a,b, 1913a,b, 1914), argued that there was a single eye anlage in the median region of anterior neural plate which spread laterally and separated eventually into two eyes (Fig. 1B; Stockard, 1913a,b). LePlat (1919) came to a conclusion similar to that of Stockard except that he viewed the anlage as ‘optico-ocular’ apparatus including both the eyes and the optic chiasma (Fig. 1C). A series of experiments by Adelmann determining the developmental potential of ectodermal regions in amphibian embryos revealed that both the lateral and median portions of the anterior neural plate could give rise to eyes (Adelmann, 1929a,b, 1930, 1934, 1936), suggesting the existence of an equipotential eye field (Adelmann 1929a,b; Fig. 1D). However, since early embryos are known to have regulative ability, Adelmann cautiously pointed out that it was not clear whether eye formation by the median region isolated from its original surroundings reflects its potential in normal development (1929b). In fact, prosencephalon areas anterior and posterior to the eye field could also form eyes when isolated from their normal neighbors (Corner, 1963 and 1966; Boteren-brood, 1970). These observations have been interpreted as revealing a forebrain field which includes the telencephalon, the eyes and other regions of the diencephalon (Fig. 1E; Corner, 1963, 1966; Boterenbrood, 1970; Nieuwkoop et al., 1985), making it unclear whether there is a distinct eye field or that Adelmann’s results simply showed part of properties of the entire forebrain field. The classic studies have been largely forgotten (Hamburger, 1988, p62; Nieuwkoop et al., 1985) and the hypothesis of an eye field remains controversial.

We have been working on a new family of developmental regulators, the T domain proteins. T domain is the DNA binding domain of this new family of transcription factors (Herrmann and Kispert, 1994; Kispert et al., 1994, 1995; Kispert and Hermann, 1993) and T domain proteins have been found in both vertebrates and invertebrates (Pflugfelder et al., 1992a,b; Kispert et al., 1994; Yasuo and Satoh, 1994; Bollag et al., 1994; Bulfone et al., 1995). The prototype of the T family, the mouse T locus gene and its homologs, brachyury (Xbra) in Xenopus and no tail (ntl) in zebrafish, are essential for posterior mesoderm and notochord formation (Chesley, 1935; Grüneberg, 1958; Bennet, 1975; Hermann et al., 1990; Smith et al., 1991; Cunliffe and Smith, 1992; Schulte-Merker et al., 1992, 1994; Halpern et al., 1993). Injection of a mutant form of Xbra led to formation of the retina and the cement gland, suggesting that other T domain proteins might also exist in early Xenopus embryos (Rao, 1994).

We have now identified several genes encoding new Xenopus T domain proteins. One of them, named ET, was found to be expressed in the primordia for the retina and the cement gland. Interestingly, its early expression was in a continuous band in the anterior neural plate which later resolved into two retina primordia. To confirm that this pattern reveals a retina morphogenetic field and its resolution, we isolated and characterized the expression pattern of a Xenopus homolog of the Pax-6 gene, which is known to be important for eye formation in vertebrates and invertebrates. Xenopus Pax-6 expression pattern indeed supports the location and resolution of a retina field. Fate-mapping experiments indicate that retina field resolution is achieved by suppression of retina formation in the medial region, rather than by migration of retina forming cells into the lateral regions of the retina field. Results from experiments with Xenopus explants and chick embryos indicate that a primary signal responsible for resolving the retina field originates from the midline region of the prechordal mesoderm, the prechordal plate.

The sequences of the primers for T domain protein genes are: (GGI MGI MGI ATG TTY CCI GT) coding for GRRMPFP as the upstream primer, and (TAI GCI GTI ACI GCD ATR AA) coding for FIAVTAY as the downstream primer. The conditions for polymerase chain reactions (PCR) were: 1 cycle of 94°C, 3 minutes (min.), 53°C, 1 min. and 72°C, 2 min., followed by 36 cycles of 94°C, 45 sec., 53°C, 1 min. and 72°C, 2 min. A band of about 450 bp was obtained and subcloned into pBluescript SK. About 170 individual cDNAs were isolated and more than 90 have been sequenced. 28 of them encode 7 distinct T domain containing proteins. The sequence predicted from one of the PCR fragments corresponding to the gene named ET is shown in Fig. 2.

The sequences of the primers for Pax-6 are: (GIC CIC/T TIC CIG AT/CA/T G/CIA C) as the upstream primer, and (GGI AA/GA/G TCA/G/T ATC/T TTI GCI GC) as the downstream primer. The upstream primer was designed for PLPDST in the paired domain and the downstream primer was based on AAKIDLP in the homeo domain (the underlined sequence were unique to Pax-6). Template cDNAs were made from stage 28 and 40 embryos. The PCR conditions were the same as those used for isolating genes encoding T domain proteins. A fragment of 1.2 kb was subcloned into pBluescript SK and more than 150 colonies obtained. cDNAs from 16 individual clones were analysed and 15 of them had inserts. Sequence analyses showed that 10 of them contained the same sequence (of Pax-6) and the other sequences were not similar to any sequences in the databank. A probe made from the Pax-6 PCR fragment was used to screen the stage 28 cDNA library. 10 positives clones were isolated and the longest cDNA of 2.6 kb was sequenced.

Whole-mount in situ hybridization of Xenopus embryos was performed essentially according to the method of Harland (1991). Briefly, embryos of different stages were removed from the vitelline membrane and fixed in MEMFA (0.1 M MOPS, pH 7.4, 2 mM EGTA, 1 mM MgSO₄, 3.7% formaldehyde). Digoxigenin incorporated probes were prepared with appropriate templates and hydrolyzed. Embryos were hybridized with these probes followed by sequential washes and addition of alkaline phosphatase (AP) conjugated anti-digoxigenin antibodies. AP was detected by reaction with BCIP (5-bromo-4-chloro-3-indolyl phosphate) and NBT (nitro blue tetrazolium) or BM purple as substrates.

Anterior neural plate explants were isolated from stage 12.5 or 13 embryos. The entire prechordal mesoderm was removed from the ectoderm by hair-loops and glass needles. Ectodermal explants with or without the mesoderm were then cultured in 0.5× MMR (modified Ringer’s solution) to appropriate stages. Explants were cultured to stages 18, 19, 35 and 42 and fixed in MEMFA for in situ hybridization or morphological examination. Some stage 42 explants were hybridized with a Pax-6 probe, embedded in plastic resin (JB-4 embedding kit, Polysciences, PA) and sectioned using an ultramicrotome.

This was performed essentially according to Eagleson and Harris (1990). Small chips of DiI (1, 1′-dioctadecyl-3, 3, 3′, 3′-tetram-ethylindocarbocyanine) were placed into the medial region of the retina field at stages 12 and 12.5.

For fate mapping, DiI-labeled embryos were cultured to stage 25 and fixed with MEMFA. They were viewed under the microscope both with epifluorescence and with incident white light. The fluorescent and incident light images were taken of each embryo at the same position using a video camera and stored in the computer as Adope Photoshop RGB files. Imposition of the two images were carried out by copying the red layer of the fluorescent image and pasting it onto the red layer of the regular lighting image, resulting in a composite image (Fig. 6D).

To locate the original position of the DiI-labeled cells, photocon-version of DiI followed by in situ hybridization with Pax-6 was carried out (Izpisua-Belmonte et al., 1993; Saldivar et al., 1996). Briefly, embryos were labeled at stage 12 and fixed at stage 12.5 in 4% paraformaldehyde. They were rinsed twice in PBS and once in 0.1 M Tris-HCl (pH 7.4). They were equilibrated in a solution containing 1 mg/ml of DAB (3,3′-diaminobenzidine tetrahy-drochlorate) in 0.1 M Tris-HCl for 30 minutes. The embryos were placed on the surface of 0.6% soft agar covered with fresh DAB solution and illuminated under a 20× objective lens using rhodamine optics. DiI was photoconverted to a brown precipitate in about 20 minutes. Embryos were transferred to 0.1 M Tris-HCl for 30 minutes and dehydrated in methanol. In situ hybridization with Pax-6 was then carried out in these embryos as described above. Pigmented embryos were eventually bleached by bright light in the presence of H₂O₂.

Fertile white Leghorn chicken eggs (B & E Eggs, PA) were isolated and cultured as described by Sundin and Eichele (1992). Embryos were staged according to Hamburger and Hamilton (1951). Embryos were operated on and cultured with the ventral side up on 0.3% agar plates with 50% egg white. After the operations, each embryo was covered with 50 μl of Tyrode’s solution. Culture dishes were sealed with Parafilm membrane, and incubated in a humidified chamber at 38°C. 50 μl of Tyrode’s solution was added about every 12 hours. Tungsten needles prepared by electrolytic sharpening in 2 N NaOH were used for microsurgery.

In prechordal plate removal experiments, slits were made with a needle on both sides of the prechordal plate of a stage 5 embryo. The location of the pre-chordal region removed is diagrammed Fig. 9B and the width of the removed region is about twice that of the notochord. Whole-mount in situ hybridization of stage 6 embryos with a sonic hedgehog (shh) probe was used to verify the removal of the prechordal plate. Embryos cultured to stages 11 or 13 were examined for Pax-6 expression by in situ hybridization. Plastic sectioning of in situ hybridized embryos was carried out as previously described (Li et al., 1994).

In rescue experiments, mesodermal tissues from other embryos were transplanted into the host embryos from which the prechordal plate had been removed. Both the donor and the host embryos were operated on at stage 5. Positions of donor tissues are diagrammed in Fig. 10A. In order to allow proper adhesion of the transplanted tissues, host embryos were incubated with minimal amount of the culture medium for 2 hours before the addition of another 50 μl of Tyrode’s solution. Embryos with apparently proper adhesion were followed to later stages.

Effects of mesodermal tissues on Pax-6 expression in the retina were tested by transplantations (Fig. 10A,E). A slit was made in the mesoderm underlying the left presumptive retina primordium of a stage 5 host embryo (Li et al., 1994). A piece of mesodermal tissue was isolated from a donor embryo of the same stage and placed into the slit in the host embryo. Embryos were cultured to later stages before being fixed in MEMFA.

To search for new T domain proteins, we designed degenerate primers based on conserved sequences in the T domain (Kispert et al., 1994). DNA fragments of the expected length were amplified. The amino acid sequence predicted from one of the PCR fragments was found to be similar to that of Optomotor blind (Omb; Fig. 2A), the product of a Drosophila gene essential for optic lobe development (Pflugfelder et al., 1992a,b). The gene corresponding to this PCR fragment was named ET because of its prominent expression in the eyes (see below). The predicted ET protein is also similar to that of Tbx2, a mouse T domain protein (Bollag et al., 1994). The published expression pattern of Tbx2 mRNA is not similar to that of ET (compare Fig. 6B of Bollag et al., 1994 to data shown below), making it unclear at the present whether Tbx2 is an ortholog of ET.

The pattern of ET expression was examined by whole-mount in situ hybridization (Harland, 1991). ET is prominently expressed in the primordia of the retina and the cement gland (Fig. 3E). ET expression in the retina primordia begins as in a single band across the midline in the anterior neural plate at stage 12.5 (indicated by the red arrow in Fig. 3A). This band of expression persists at stage 15 (Fig. 3B). Expression in the medial region of this band decreases gradually, so that it is weaker at stage 16 (Fig. 3C) and disappears by stage 18 (Fig. 3D). In late stages, ET expression in the retina is clearly localized in the dorsal part of the retina, but not in the lens or the ventral half of the retina (Fig. 3H,I).

ET is also expressed in the cement gland, an anterior epidermal structure whose formation is often associated with neural induction (Lamb et al., 1993; Hemmati-Brivanolou et al., 1994; Rao, 1994). In contrast to its expression in the retina, ET expression in the cement gland does not change its pattern from the earliest time of detection at stage 12.5 (indicated by the red arrowhead in Fig. 3A) until its disappearance around stage 26 (Fig. 3A-G). Expression in the pineal gland was detected from stage 20 onwards (Fig. 3F-H). ET is also expressed in cephalic ganglia and the lateral line organ of stage 26 embryos (Fig. 3H).

The pattern of ET expression in the anterior neural plate is consistent with the possibility of a retina field in the anterior neural plate which later resolves into two retina primordia. To ask whether ET expression as a band in the anterior neural plate reflects a general feature of genes expressed in retina primordia, we isolated another early marker for eye primordia, Pax-6 and examined its expression pattern.

Pax-6 is essential for eye formation in a wide range of species. In vertebrates such as the mouse, chicken, zebrafish and human, as well as in invertebrates such as Drosophila, Pax-6 is expressed in the eye primordia (Walther and Gruss, 1991; Krauss et al., 1991; Püschel et al., 1992; Li et al., 1994; Glaser et al., 1994, Quiring et al., 1994). Pax-6 mutations in mouse, human and Drosophila result in malformation or absence of the eyes (Hogan et al., 1988; Hill et al., 1991; Ton et al., 1991; Jordan et al., 1992; Glaser et al., 1994; Quiring et al., 1994). Ectopic expression of Drosophila Pax-6 (eyeless) in imaginal discs leads to formation of supernumerary eyes, leading to the suggestion that Pax-6 is a master regulator sufficient for initiating eye formation (Halder et al., 1995).

We isolated a cDNA fragment of Xenopus Pax-6 by PCR. This fragment allowed us to isolate cDNAs encoding an apparently full-length Xenopus Pax-6 protein (Fig. 4). Although all Pax proteins are similar within the paired and homeo domains, the primary sequences of Pax-6 proteins from all species are easily distinguishable from other Pax proteins (Quiring et al., 1994). The homeodomain of the Xenopus Pax-6 is identical to those of other vertebrate Pax-6 while all differences in the paired domain of Xenopus Pax-6 occur in regions where variations have been observed among the other Pax-6 sequences. The predicted full-length Xenopus Pax-6 product is 96% identical to mouse and human Pax-6 proteins (Fig. 4).

Pax-6 mRNA is first expressed at stage 12 before the completion of gastrulation and the beginning of neurulation (Fig. 5A). In the trunk region, Pax-6 is expressed in the primordium of the neural tube, beginning with two broad stripes at stage 12 (indicated by the red arrowhead in Fig. 5A). In the anterior neural plate, Pax-6 is expressed initially in a continuous band (Fig. 5A-C). Like ET, Pax-6 expression in the median region of this band is also turned off by stage 18 (Fig. 5D-G). The band of Pax-6 expression in the anterior neural plate is broader than the ET band in the anterior/posterior dimension. On each side of the midline, Pax-6 is expressed in the entire eye and a region in the forebrain (Fig. 5E,F). The precursor region for this forebrain structure is initially located in a stripe posterior to the retina field (Fig. 5B-D,G). This stripe also resolves into two bilaterally symmetric spots during neurulation (Fig. 5B-E).

The Pax-6 expression pattern thus confirmed results obtained for ET in showing the presence of a continuous band in the anterior neural plate, supporting the suggestion that the expression patterns of these genes reveal a retina field which resolves into two retina primordia.

Pax-6 is also expressed in the lens, which is distinguishable from its expression in the retina field from stage 12.5 onwards (Fig. 5B,C). As early as stage 12.5, lens primordia, as revealed by Pax-6 expression, were separate on two sides of the embryo (Fig. 5B). Thus, there is so far no direct evidence for an eye field which includes both the retina and lens primordia.

Two hypotheses can explain why retina primordia form in the lateral but not the medial regions of the retina field. One is that retina precursor cells exist in the medial region of the retina field at early stages (e.g., stages 12 and 12.5) and they migrate into the lateral regions in later stages. The alternative hypothesis is that retina formation is suppressed in the medial region of the retina field and medial cells take on fates other than that of the retina. These hypotheses can be distinguished by following the development of cells in the medial region of the retina field. The migration hypothesis predicts that some medial cells will end up in the retina, whereas the suppression hypothesis predicts that the medial cells will stay near the midline and give rise to structures other than the retina.

Previous fate mapping studies have shown that cells in the midline of the anterior neural plate of stage 15 Xenopus embryos will form the anterior pituitary gland and the supra chiasmatic nucleus (Eagleson and Harris, 1990; Eagleson et al., 1995). However, another study tracing the lineage of two-cell stage embryos suggested that retina cells from one side of the embryo could migrate to the other side (Jacobson and Hirose, 1978). Recent studies in zebrafish embryos suggest that midline cells may contribute to both retinae (Woo and Fraser, 1995). A drawback in interpreting the lineage analyses of two-cell stage Xenopus embryos is that it cannot accurately predict the fate of midline cells since the cleavage plane at the two-cell stage is variable and independent of the midline. Results from zebrafish embryos, on the other hand, may not apply to Xenopus embryos because zebrafish embryos are different from Xenopus embryos with regard to migration of early embryonic cells (Shih and Fraser, 1995). The fate of midline cells at stages 12 and 12.5 Xenopus embryos thus remains unknown.

We examined the question of cell migration by labeling midline cells in the retina field with the fluorescent dye DiI at stages 12 or 12.5 and following the fate of the labeled cells (Fig. 6A). Some of the embryos were fixed shortly after labeling in order to locate the region of DiI labeling. DiI was photoconverted to give a brown signal in these embryos. Pax-6 was used to reveal the retina field. Double labeling by DiI photoconversion and Pax-6 in situ hybridization thus allowed determination of the spatial relationship of DiI-labeled region to the retina field. DiI labeling was indeed located in the median region of the retina field (Fig. 6B).

DiI-labeled embryos were followed to stage 25 when eyes were morphologically distinguishable. Each embryo was viewed under the microscope with epifluorescence and, immediately afterwards, with incident white light. Two images from a single embryo were thus obtained through a video camera and were individually stored. Superimposition of these two images allows precise mapping of the fluorescent cells onto the whole embryo (Fig. 6D). It is clear that DiI labeled cells do not contribute to the retinae (Fig. 6D,E). These results demonstrate that cells in the midline of the retina field do not migrate into the more lateral regions to form retina precursor cells in Xenopus embryos.

To investigate a role for the prechordal mesoderm in the formation of two retinae (Adelmann, 1936b), we examined retina formation in explants with or without the prechordal mesoderm. Explants of the anterior neural plate were isolated from stage 12.5 or stage 13 Xenopus embryos. To ensure inclusion of the retina field in the explants, we fixed some explants immediately after isolation. They were examined by in situ hybridization with the Pax-6 probe and the retina field was found to be present in the explants (data not shown).

We then examined retina formation by allowing the explants to develop to stages 35 and 42 when pigmented retina is visible (Fig. 7A). When the anterior neural plate together with its underlying prechordal mesoderm was isolated, two retinae formed (Fig. 7B,D). In contrast, only one retina formed in explants from which the pre-chordal mesoderm was removed (Fig. 7C,E). Thus, the prechordal mesoderm is essential for the formation of two retinae.

The formation of a single retina in anterior neural plate explants lacking the mesoderm could have resulted from one of the following: (1) a constant requirement of a retina inducing factor from the mesoderm for the formation of two retinae; (2) fusion of two retina primordia; or (3) failure of the single retina field to resolve. Our ability to examine retina primordia at early stages made it possible to distinguish among these different models. If the first model is true, the retina field would shrink, or in the second scenario, it would become two primordia, whereas the retina field would remain as a continuous band if the third hypothesis is correct.

We therefore examined directly the retina field with regard to the role of the prechordal mesoderm. This was carried out by hybridizing the explants with Pax-6 and ET probes. In the presence of the prechordal mesoderm, the retina field became two primordia by stage 18 (Fig. 8B,E). In the absence of the prechordal mesoderm, however, a single unresolved retina field remained (Fig. 8C,F). These results indicated that formation of a single retina in explants without the prechordal mesoderm was due to the failure of the retina field to resolve, demonstrating directly a role for the prechordal mesoderm in retina field resolution.

Explant experiments with Xenopus embryos have shown that the prechordal mesoderm is required for retina field resolution. We have carried out further experiments in chick embryos for three reasons. (1) Results discussed so far were all obtained from amphibian embryos. Their general significance remains to be tested by extending these studies to other vertebrate species. (2) Results from the explant experiments need to be confirmed by examining retina development in whole embryos. (3) The Xenopus experiments indicate retina field resolution requires the prechordal mesoderm, which is a rather large region. It is important to test the possibility that a smaller region in the prechordal mesoderm is involved in retina field resolution.

We therefore studied the consequence of removing the median region of the prechordal mesoderm, i.e., the prechordal plate, on retina formation in chick embryos. Chick embryos were operated on and cultured in albumen agar plates. The prechordal plate was removed from embryos at stage 5 (Fig. 9B) and they were cultured to stage 13 and fixed. Retinae were revealed by in situ hybridization with a chicken Pax-6 probe (Li et al., 1994). In control embryos, two retinae formed (Fig. 9C,E). In approximately 27% of embryos without the prechordal plate, a single retina, continuous from one side of the embryo to the other, formed (Fig. 9D,F). That only a fraction of embryos had cyclopia phenotype was most likely due to the difficulty in complete removal of the prechordal plate. These results thus extended the conclusions from amphibian to avian embryos and further localized the requirement for the prechordal mesoderm to its median region, the prechordal plate.

The prechordal plate could provide either a non-specific physical support or a specific signal for the formation of two retinae. To distinguish between these possibilities, mesodermal tissues from different regions were tested for their ability to rescue the cyclopia phenotype in embryos in which the prechordal plate had been removed (see Fig. 10A for a diagram of the sources of the donor tissues). The prechordal plate was able to rescue the cyclopia phenotype (Table 1), whereas mesodermal tissues taken either from the lateral parts of the prechordal region or from the lateral posterior region could not substitute for the prechordal plate (Table 1). These results indicate that the prechordal plate plays a specific role in the formation of two retinae.

Our results from Xenopus explants and chick embryos suggest that the prechordal plate sends a signal to the medial region of the retina field to suppress retina formation in the medial region of the anterior neural plate. To further test this hypothesis, we carried out transplantation experiments in chick embryos to examine whether the prechordal plate could suppress the expression of Pax-6 in the retina, which should serve as a reasonable indicator for retina development because of the known role of Pax-6 in retina formation.

A slit was made in the lateral prechordal mesoderm underlying the retina primordium on the left side of a stage 5 host embryo (Li et al., 1994). A piece of the prechordal plate or mesodermal tissues from other embryonic regions was isolated from a donor embryo of the same stage, and placed into the slit in the host embryo (see Fig. 10A and E for diagrams). Embryos were then cultured to stage 13 and fixed for in situ hybridization with Pax-6.

In experiments with prechordal plate transplants, some embryos were fixed shortly after the transplantation and examined for the expression of sonic hedgehog (shh), a marker for the prechordal plate (Marti et al., 1995; Shimamura et al., 1995). Presence of a transplanted prechordal plate was confirmed by shh in situ hybridization (Fig. 10F).

In approximately 75% of embryos with transplanted prechordal plates, Pax-6-expressing areas were smaller on the side receiving the transplants (Fig. 10G, H), indicating that Pax-6 expression was suppressed. The suppression is partial in most embryos (Fig. 10G), perhaps because contact between the transplanted prechordal plate and the anterior neural plate was not as tight as that in the wild-type situation, or because the width of the region in the prechordal plate with Pax-6 suppressing activity is not large enough to cover the entire retina primordium on the left side of the chick embryo.

Neither the transplantation of a piece of lateral prechordal mesoderm, nor the making of a slit alone, affected the size of the retina on the experimental side (Fig. 10C and Table 2). Transplantation of a piece of mesoderm from the lateral posterior region of the embryo sometimes reduced the size of the retina (Fig. 10D), perhaps due to interference of retina induction after such transplantations.

The reduction of retina in embryos transplanted with the prechordal plate was more severe than that in embryos transplanted with lateral posterior mesoderm (compare Fig. 10G and D). Strong suppression like that shown in Fig. 9H was only observed after prechordal plate transplantation, but not after lateral posterior mesoderm transplantation (Table 2). Furthermore, the number of embryos with retina size reduction in those transplanted with the prechordal plate was significantly higher than that in embryos transplanted with lateral posterior mesoderm (Table 2). Taken together, these results are consistent with the suggestion that the prechordal plate can inhibit Pax-6 expression.

Our major conclusions are that there is a single retina morphogenetic field which resolves into two retina primordia, that retina field is resolved by suppression of retina formation in the median region of the field, and that the primary signal for retina field resolution comes from the prechordal plate. Mechanisms responsible for bilateral asymmetry have been investigated in recent work on the left-right axis (Brown and Wolpert, 1990; Yost, 1990, 1991, 1992, 1995; Klar, 1994; Danos and Yost, 1995; Levin et al., 1995). Our experiments have studied formation of the eyes, a model for bilaterally symmetric structures, revealing mechanisms underlying the establishment of bilateral symmetry in the forebrain.

Our model of retina primordium formation is outlined in Fig. 11. Thus, there is a single retina field in the anterior neural plate of vertebrate embryos which becomes two primordia (Fig. 11A). This model is similar to one suggested by Adelmann who based his suggestion on the findings of retina formation in the median portion as well as the lateral portions of the anterior neural plate (Adelmann, 1929a,b). Our observations of ET and Pax-6 expression patterns in the anterior neural plate provide direct support for the existence of a single retina field in the early embryo. These results strengthened the conclusion of Adelmann in showing that the median as well as the lateral regions were indeed endowed with the potential for retina formation. Availability of the molecular markers makes it possible to locate the retina field precisely and to follow the process of its resolution into two primordia. Support for equipotentiality in the retina field also came from the experiments studying the role of the prechordal mesoderm on retina field development. Our finding of the formation of a single retina primordium in the entire field in the absence of the prechordal mesoderm (Figs 7 and 8) indicates that the entire retina field is indeed capable of forming retina. Our experiments with chick embryos further extend conclusions from studies of amphibians to other vertebrate species.

Results from our fate mapping experiments rule out the possibility that the medial region of the retina field contains retina precursor cells which later migrate into the lateral retina primordia (Fig. 6). Results from these experiments in Xenopus and ectopic transplantation of the prechordal plate in chick embryos support the suggestion that retina field resolution is achieved by suppression of retina forming potential in the median region and maintenance of retina development in the lateral regions of the retina field.

Our experiments have addressed the role of the prechordal mesoderm in influencing the bilaterality of the retina primordia. It is shown that removal of the prechordal mesoderm could result in the formation of a single retina in the retina field. Direct examination of the retina field shows that it stays as a single field after the removal of the prechordal mesoderm (Fig. 8). Prechordal plate removal and transplantation experiments in chick embryos indicate that the prechordal plate plays a specific signaling role in inhibiting retina formation in the ectoderm (Figs 9, 10). Taken together, these results suggest that the prechordal plate provides an essential signal to resolve the retina field (Fig. 11B).

The phenotype of chick embryos removed of the prechordal plate is very similar to that of zebrafish mutants defective in the cyclops gene. Thus, fusion of Pax-6 expressing domains (Hatta et al., 1994; Macdonald et al., 1995) and formation of a single eye were observed in zebrafish cyclops mutants (Hatta et al., 1991). Cell transplantation experiments have shown that midline cells in the anterior neural plate as well as floor plate cells in the neural tube are defective in cyclops mutant embryos (Hatta et al., 1991, 1994; Hatta, 1992). Failure to form separate eyes in cyclops mutant embryos was shown to be secondary to defects of these midline cells in the anterior neural plate, suggesting that the midline cells in the ectoderm is important for the separation of the two eyes (Hatta et al., 1994).

Thus, results from amphibian and chick embryos indicate that the prechordal plate is crucial for retina field resolution while results from zebrafish embryos suggest a role for the midline cells in the anterior neural plate. This scenario is quite similar to the roles of the notochord and the floor plate in inducing ventral cell types and suppressing dorsal cell types in the neural tube (Jessell et al., 1988; van Straaten et al., 1988; Placzek et al., 1990, 1993; Yamada et al., 1991, 1993; Basler et al., 1993; Liem et al., 1995; Goulding et al., 1993).

Similarity of retina field resolution to dorsal/ventral patterning in the neural tube suggests that molecules responsible for neural tube patterning may also be involved in resolving the retina field. Thus, sonic hedgehog (shh), a molecule known to be important for neural tube patterning (Echelard et al., 1993; Krauss et al., 1993; Roelink et al., 1994, 1995), is a candidate signal for retina field resolution. Injection of shh mRNA into zebrafish embryos leads to overall decrease of Pax-6 expression, reduction of eyes and increased expression of Pax-2, a marker for the optic stalk (Macdonald et al. 1995; Ekker et al., 1995b). These observations, together with the finding of shh expression in the midline of the anterior neural plate (Echelard et al., 1993; Krauss et al., 1993; Roelink et al., 1994, 1995; Ekker et al., 1995a; Ericson et al., 1995), have led to the suggestion of shh as the signal in the midline of the neural plate for patterning the retina and optic stalk (Macdonald et al. 1995; Ekker et al., 1995b). Studies involving activating and inhibiting the signal transduction pathway of shh lend further support to this suggestion (Hammerschmidt et al., 1996). Observation of cyclopia in mice lacking shh proves that shh is required for the formation of two eyes. Since expression of shh is also detected in the prechordal mesoderm (Marti et al., 1995; Shimamura et al., 1995), shh may act as the prechordal plate signal for resolving the retina field. Curious findings of human fetuses with both holoprosencephaly and digit anomalies (Young and Madders, 1987; Moerman and Fryns, 1988; Shiota and Tanimura, 1988; Atkin, 1988; Cohen, 1989) could be best explained by defects in signaling pathways used both in patterning the limb bud (Riddle et al., 1993) and in determining bilaterality in the anterior neural plate.

Classic literature has called the morphogenetic field in the anterior neural plate, the eye field (Adelmann, 1936b). Our examination of Pax-6 expression in Xenopus embryos shows that formation of the lens primordia is different from the retina primordia: there are two separate lens primordia as early as stage 12.5 when the retina field is still a single field (Fig. 5B). Thus, either there is no lens field or the lens field behaves differently from the retina field. In the former case, two lenses are determined from separate primordia. In the latter scenario, it is possible that at stage 12 (Fig. 5A), the lens field could not be distinguished from the retina field by examining Pax-6 expression. If this is true, then the lens field must resolve into two lens primordia by stage 12.5 (Fig. 5B). Since the retina field remains one field through stages 14, 15, 16, separation of this hypothetical lens field is accomplished earlier than separation of the retina field. Thus, whether or not a lens field exists, the mechanism underlying retina primordium formation appears to be different from that for lens primordium formation.

The Pax-6 expression pattern suggests that the bilaterality of forebrain regions other than the retina is established in a similar manner to that of the retina. Thus, a forebrain structure expressing Pax-6 was also found to be a continuous band just posterior to the retina field and this band became two spots over the same time course as the retina field resolution (Fig. 5), suggesting a mechanistic similarity in the formation of the bilaterally symmetric retina and forebrain structures. This suggestion could explain why, in cyclopean embryos, not only eyes but also other forebrain structures remain unseparated (e.g. Adelmann, 1934; Hatta et al., 1991).

Regulative changes of the forebrain primordial region observed by Corner and Boterenbrood have led to the proposal of a forebrain field (Corner, 1963, 1966; Boterenbrood, 1970; Nieuwkoop et al., 1985). This hypothesis suggests that an area larger than the retina field is equivalent in developmental potential (Corner, 1963, 1966; Boterenbrood, 1970; Nieuwkoop et al., 1985). It is not clear how the retina forms in this field, although the retina fate is supposed to be the default in this field (Corner, 1963, 1966; Boterenbrood, 1970; Nieuwkoop et al., 1985). In contrast to our data supporting the existence of a retina field, there is so far no evidence that complements the results of Corner (1963, 1966) and Boterenbrood (1970) in directly demonstrating a forebrain field.

We are grateful to V. Hamburger for discussions, M. Nonet and R. Kopan for help with microscopy, P. Bridgman for allowing us to use his ultramicrotome, G. Philips for help with plastic sectioning, C. Stern and C. Krull for advice on DiI photoconversion, O. Sundin for the chicken Pax-6 probe, R. Riddle and C. Tabin for shh cDNA, T. Fagaly, R. Kopan, J. Sanes, D. Van Essen and J. Lichtman for comments on the manuscript.

Two papers indicating involvement of abnormality in human shh functioning as the underlying cause of a subset of holo-prosencephaly have now been published.

Belloni, E., Muenke, M., Roessler, E., Traverso, G., Siegel-Bartelt, J., Frumkin, A., Mitchell, H. F., Donis-Keller, H., Helms, C., Hing, A. V., Heng, H. H. Q., Koop, B., Martindale, D., Rommens, J. M., Tsui, L.-C. and Scherer, S. W. (1996). Identification of Sonic hedgehog as a candidate gene responsible for holoprosencephaly. Nature Genet. 14, 353-356.

Roessler, E., Belloni, E., Gaudenz, K., Jay, P., Berta, P., Scherer, S. W., Tsui, L.-C. and Muenke, M. (1996). Mutations in the human Sonic hedgehog gene cause holoprosencephaly. Nature Genet. 14, 357-360.