The one-eyed pinhead gene functions in mesoderm and endoderm formation in zebrafish and interacts with no tail

A series of inductive interactions pattern the early vertebrate embryo. By the end of gastrulation the three germ layers are defined and both mesoderm and ectoderm become patterned along the anterior-posterior and dorsoventral axis. Studies in frogs have shown that this process is initiated by maternally provided signaling molecules that are located in the vegetal, presumptive endodermal region of the early embryo (Nieuwkoop, 1969; reviewed by Slack, 1994; Smith, 1995). These signals induce both mesoderm formation and an organizing center, known as the Spemann organizer, in the dorsal mesoderm (Spemann, 1938; Hamburger, 1988). The organizer has a dual role in mesoderm formation. It develops into the axial mesoderm, consisting of prechordal plate anteriorly and notochord posteriorly, and it is a source of dorsalizing signals that pattern the dorsoventral axis of the adjacent mesoderm. This process leads to the characteristic vertebrate gastrula fate map, with axial mesoderm being the most dorsal tissue type, and paraxial mesoderm (developing into somites), intermediate mesoderm (pronephros) and lateral plate mesoderm (blood) being more lateral and ventral mesodermal derivatives (Keller, 1976; Dale and Slack, 1987; Kimmel et al., 1990).

The organizer and its derivatives are also involved in the induction and patterning of the neuroectoderm (Spemann and Mangold, 1924; Smith and Slack, 1983; reviewed by Ruiz i Altaba and Jessell, 1993; Doniach, 1995). Planar signals from the organizer and/or vertical signals from the underlying axial mesoderm are thought to neuralize the dorsal ectoderm and initiate anterior-posterior regionalization in the neuroectoderm. The notochord is subsequently involved in the induction of ventral cell types such as floor plate in the overlying neuroectoderm (van Straaten et al, 1988; van Straaten and Hekking, 1991; Placzek et al., 1990). Factors like noggin (Lamb et al., 1993), chordin (Sasai et al., 1994; Holley et al., 1995), and members of the TGFβ (Rebagliati et al., 1985; Kessler and Melton, 1994), FGF (Slack et al., 1988; Kimelman et al., 1992), wnt (Moon, 1993) or hedgehog (Echelard et al., 1993; Krauss et al., 1993; Roelink et al., 1994; Ingham, 1995) families are candidate signaling molecules that mediate the inductive events in mesodermal and neural patterning.

Several genes have been identified that are expressed in the organizer region in response to mesodermal patterning signals. Most prominently, the putative transcription factors Brachyury (Hermann et al., 1990; Smith et al., 1991; Beddington et al., 1992; Herrmann and Kispert, 1994), goosecoid (Blumberg et al., 1991; Cho et al., 1991), Pintallavis and HNF3β (Dirksen and Jamrich, 1992; Ruiz i Altaba and Jessell, 1992; Knoechel et al., 1992; Strähle et al., 1993; Ang et al., 1993; Ruiz i Altaba et al., 1993; Monaghan et al., 1993; Sasaki and Hogan, 1993), lim1 (Taira et al., 1992) and Xnot (von Dassow et al., 1993) are activated as a response to mesoderm inducers. These genes are thought to execute or control the embryonic patterning initiated by inductive signals. Mutational analysis in the mouse has indicated that lim1 (Shawlot and Behringer, 1995), otx2 (Acampora et al., 1995; Matsuo et al., 1995; Ang et al., 1996), HNF3β (Ang and Rossant, 1994; Weinstein et al., 1994) and Brachyury (Chesley, 1935; Gluecksohn-Schoenheimer, 1944; Herrmann and Kispert, 1994), but not gsc (Rivera-Perez et al., 1995; Yamada et al., 1995), are required during early development. All four mutants display serious deficits in the formation of the organizer and/or axial mesoderm. Brachyury mutants have a severe posterior truncation of the axis and lack a differentiated notochord. Head formation appears normal. HNF3β mutants do not form a distinct node (the mouse equivalent of Spemann organizer), lack a notochord and show variable head defects. Both otx2 and lim1 mutant mice lack structures anterior to hindbrain rhombomere 3, including the prechordal plate, midbrain and forebrain, and notochord development appears to be incomplete in some otx2 mutant embryos.

In zebrafish, three loci have been shown to affect the formation of axial tissues, floating head (flh), no tail (ntl), and cyclops (cyc). Mutants of the flh gene show defects in notochord formation during midgastrulation (Talbot et al., 1995). Instead of maintaining their axial identity, flh mutant cells in the midline develop into paraxial muscle (Halpern et al., 1995). Molecular analysis has identified flh as a mutation in the zebrafish homologue of the Xenopus Xnot homeobox gene (Talbot et al., 1995). The ntl locus is the zebrafish homologue of the mouse Brachyury gene (Schulte-Merker et al., 1992; Halpern et al., 1993; Schulte-Merker et al., 1994). It is first expressed in the entire marginal region, and later in notochord and tail bud. Notochord precursors in ntl mutants appear to be present but do not differentiate properly and retain a mesenchymal appearance. Furthermore, tail formation is impaired in ntl mutants. Embryos mutant for the cyc locus display deficits in the formation of ventral neuroectoderm, leading to eye fusion and reduction of floor plate (Hatta et al., 1991, 1994). Additionally, a reduced number of hatching gland cells, and lower levels of gsc expression indicate that prechordal plate formation is also weakly affected in cyc mutants (Thisse et al., 1994).

Although less well understood than mesodermal and neuroectodermal induction and patterning, endoderm formation might rely on the same or related signals that induce mesoderm (Asashima et al., 1991; Jones et al., 1993; Cornell et al., 1995; Gamer and Wright, 1995; Henry et al., 1996). In particular, exposure of animal caps to high concentrations of mesoderm inducers like activin can also activate endoderm specific markers (Rosa, 1989; Cornell et al., 1995; Gamer and Wright, 1995). Furthermore, injection of a dominant-negative activin receptor construct inhibits the expression of some endoderm markers (Gamer and Wright, 1995; Henry et al., 1996). The zygotic downstream responses to endodermal inducers are virtually unknown. Few genes like Mix1 (Rosa, 1989), Xlhbox-8 (Wright et al., 1988) or members of the HNF3 (Ang et al., 1993; Ruiz i Altaba et al., 1993; Monaghan et al., 1993; Sasaki and Hogan, 1993) family are expressed as an early response to endoderm formation. Genetic analysis has not yet identified any mutants that disrupt the development of the entire endoderm.

In genetic screens for additional genes that function during vertebrate pattern formation, we have discovered the one-eyed pinhead (oep) locus in zebrafish (Schier et al., 1996; Solnica-Krezel et al., 1996). The oep gene is essential for the development of several regions of the embryo, including prechordal plate, endoderm and ventral neuroectoderm. The analysis of mutant phenotypes suggest a role for the prechordal plate in dorsal-ventral but not anterior-posterior patterning of the brain. Double mutant analysis reveals a partially redundant requirement for oep and ntl in mesoderm formation.

Fish and embryos were maintained as described by Solnica-Krezel et al. (1994) and Westerfield (1994). oep^m134 (Schier et al, 1996; Solnica-Krezel et al., 1996) was isolated in the progeny of ENU-mutagenized fish from an AB background (Chakrabarti et al., 1983). oepz1 was identified in the progeny of gamma-ray mutagenized fish from an EK background (Ekk Will Waterlife Resources, Florida; Helde et al., 1994). Both alleles segregate at Mendelian ratios: 594/2318 embryos were oep^m134 (25.6%); 90/354 embryos were oep^ZI(25.4%); 227/883 embryos were oep^m134/oepz1 (25.7%). Both mutations are subject to modifiers in different genetic backgrounds. Mutant phenotypes are generally stronger (as judged from strong eye fusion or absence of eyes, notochord defects, absence of heart muscle) in AB and TU (Mullins et al., 1994) backgrounds, and weaker (partial eye fusion, normal notochord, formation of some hatching gland cells and heart muscle) in India (Knapik et al., 1996) or HK (Stainier et al., 1995) backgrounds. Additional oep alleles have been identified (Hammerschmidt et al., 1996; Strähle et al., personal communication; Kimmel et al., personal communication). Double mutants were constructed by crossing oep/⁺ heterozygous fish to ntl^b160/+ (Halpern et al., 1993) or flhn1/⁺ (Talbot et al., 1995) heterozygous fish. All three loci are unlinked. oep/⁺;flh/⁺ and oep/⁺;ntl/⁺ heterozygous fish were identified by test crosses and interbred to create homozygous double mutant embryos. Mutant loci segregate in a Mendelian fashion; crosses of oep/⁺;ntl/⁺ heterozygous fish yield 699 wild-type: 251 oep: 220 ntl: 86 oep ntl embryos (= 8.1: 2.9: 2.6: 1.0); crosses of oep/+;flh/+ heterozygous fish yield 848 wild-type: 299 oep: 304 flh: 102 oep flh embryos (= 8.3: 2.9: 3.0: 1.0).

RAPD PCR assays were performed as described by Postlethwait et al. (1994). In initial experiments, the RAPD marker 15AH.500 was found to be linked to oep. Since RAPD markers are strain-specific and therefore not informative in all mapping crosses, a sequence-tagged site (STS) marker for the 15AH.500 locus was generated by cloning and sequencing the RAPD fragment and synthesizing primers specific for that sequence (primer WTZ97, 5’-ACTTGCAGGAGTGGATCT-GAC, and primer WTZ106, 5’-CACAAAAACACCATCTGACC). In the oep mapping crosses, all of the animals display an allele of the same size, but a polymorphism was evident when the amplification products were cleaved with TaqI (Fig. 2A). We scored the 15AH.500 STS in 85 haploid embryos in oepm134 mapping crosses, and found no recombinants (0±2.3 cM; Fig. 2). To determine the location of 15AH.500 on the genetic map, we scored the STS in the linkage map cross of Johnson et al. (1996). These primers detect codominant alleles, differing in size by a few nucleotides, of the 15AH.500 locus. The 15AH.500 STS, and therefore oep, maps to one end of LG 10 (Fig. 2B). The observation that the 20K.875 RAPD marker, previ-ously mapped to LG 10 (Postlethwait et al., 1994; Johnson et al., 1996), is linked to oep (14 recombinants among 85 individuals) confirms this assignment. To test if oepz1 might represent a deletion of the region of the oep locus, PCR was performed using 12 different primers and 12 different primer pair combinations of the 15AH.500 locus. No amplification products cosegregating with oep were found in oep^z1 mutant animals (data not shown). We have recently identified another marker that is deleted in oep^z1 mutants, as judged from PCR analysis and Southern blot hybridization (J. Zhang, W. S. T. and A. F. S., unpublished results).

In vivo observations, in situ hybridization, and histological analysis were performed as described previously (Jowett and Lettice, 1994; Westerfield, 1994; Schier et al., 1996). Dorsalization by exposure to LiCl was performed as described by Stachel et al. (1993).

Genetically mosaic embryos were generated using cell transplantation techniques (Ho and Kane, 1990; Halpern et al., 1993). Donor embryos were injected with a mixture of lineage tracer dyes (5% rhodamine dextran (10 kDa) and 5% lysine-fixable biotin dextran (10 kDa; Molecular Probes) between the 1- and8-cell stage. At midblastula stages 5-50 labeled cells were transplanted into isochronic host embryos. To guarantee full expressivity of the oep phenotype, oep mutant donors or hosts were derived from fish heterozygous for oep in a AB/TU genetic background. In the assay for hatching gland formation, host embryos were allowed to develop until 28 hpf, analyzed using fluorescence microscopy, and fixed in 4% paraformaldehyde. Hatching gland cells were identified by their size, granular morphology and their location over the yolk anterior and ventral to the head. Biotin-dextran labeled donor cells were detected using the ABC-peroxidase kit (Vector Laboratories, Inc.). Subsequently, embryos were processed for in situ hybridization with digoxigenin-labeled hgg1 (Thisse et al., 1994) riboprobes, and alkaline phosphatase coupled antidigoxigenin antibodies were used to assay for hgg1 expression in hatching gland cells. When the peroxidase reaction is performed before in situ hybridization, the precipitate of the peroxidase reaction precludes the detection of hgg1 in biotin-dextran labeled cells. Therefore, host hatching gland cells (blue) could be distinguished unambiguously from donor hatching gland cells (brown). In the assay for gsc expression, donor and host embryos were treated with LiCl at the 256-to 512-cell stage (Stachel et al., 1993), and fixed at 70-80% epiboly in 4% paraformaldehyde. In situ hybridization using gsc riboprobes was performed as described above. Embryos were then processed to detect biotin-dextran labeled donor cells using the ABC-peroxidase kit. The precipitate of the alkaline phophatase reaction does not preclude the detection of biotin-dextran labeled cells. Therefore, donor cells expressing gsc could be identified unambiguously as brown cells with a bluish cytoplasmic halo.

We have isolated two one-eyed pinhead (oep) alleles in genetic screens for zygotic mutations affecting zebrafish embryogenesis (Fig. 1). Allele oep^m134 was discovered in a screen of ENU-induced mutations (Driever et al., 1996; Schier et al., 1996; Solnica-Krezel et al., 1996), and oep^z1 was isolated in a screen using gamma-rays (K. A. H. and D. Grunwald, unpublished results). Both alleles segregate as Mendelian recessive embryonic lethal mutations and display similar phenotypes (Fig. 1; see below). In addition, the gamma-ray induced oep^z1 allele displays severe general degeneration starting at the end of the segmentation period (Fig. 1C). Transheterozygous oep^m134oep^z1 embryos show all the characteristic oep phenotypes but do not express the general degeneration of oep^z1 (Fig. 1D,H).

Linkage of the RAPD marker 15AH.500 with the oep^m134 mutation allowed mapping of the oep locus to one end of linkage group 10 (Fig. 2). 15AH.500 DNA is absent in oep^z1 mutant animals (data not shown; see Materials and Methods). This result, together with the gamma-ray induced origin of oep^z1 and the general degeneration phenotype associated with oepz1, but not oep^m134 or oep^m134oep^z1, suggests that oep^z1 is a deletion of the oep locus and one or more linked essential genes.

The oep mutant phenotype was first characterized on days 2 and 3 of embryogenesis by morphological and histological analysis, and the study of marker gene expression. Analysis of mesodermal derivatives shows that oep is essential for the formation of hatching gland (Figs 1L, 3B) and eye muscles (Fig. 3D) as judged from the expression of hgg1 (Thisse et al., 1994) and myoD (Weinberg et al., 1996), respectively. Both structures have been proposed to derive at least in part from prechordal plate mesoderm, the anterior-most axial mesoderm (Adelmann, 1932; Jacob et al., 1984; Wachtler et al., 1984; Kimmel et al., 1990; Thisse et al., 1994; Inohaya et al., 1995; Kimmel et al., 1995). A minority of oep mutant embryos also show defects in the formation of notochord, the more posterior axial mesoderm (Fig. 1M-T). In particular, the notochord has a wavy, thinner appearance (Fig. 1T) or is absent most anteriorly in some oep mutants (Fig. 1P). Formation of somites, skeletal muscle, pronephros, and blood is not overtly affected in oep mutants. In a few cases the notochord is absent in the tail resulting in somite fusion ventrally.

To determine when oep functions, we analyzed the formation of the prechordal plate in living embryos and the expression of marker genes during gastrulation. The prechordal plate is clearly visible during midgastrulation as a knob-like structure constituting the anterior axial mesoderm (Kimmel et al., 1995). In oep mutant embryos this structure is not visible (Solnica-Krezel et al, 1996). During segmentation stages, hatching gland precursors form the polster, the anterior-most prechordal plate structure, in wild-type embryos, but are absent or strongly reduced in oep mutants (data not shown). Furthermore, expression of islet1 (Fig. 3F; Korzh et al., 1993; Inoue et al., 1994) and hgg1 (Fig. 3H; Thisse et al., 1994), markers for anterior prechordal plate, is absent or strongly reduced in oep mutant animals at the end of gastrulation. Markers that are expressed in the posterior prechordal plate region, such as hlx1 (Fig. 3J; Fjose et al., 1994) and gsc (see below) are also affected in oep mutants. Correspondingly, the expression of axial (Strähle et al., 1993) and shh (Krauss et al., 1993), two genes expressed in both the developing prechordal plate and ventral neuroectoderm is reduced anteriorly (Fig. 3L, and data not shown). These studies demonstrate that prechordal plate development is impaired in oep mutants by the end of gastrulation.

The homeobox gene gsc (Blumberg et al., 1991; Cho et al., 1991; DeRobertis et al., 1992) represents the earliest specific marker for prechordal plate development during gastrulation (Fig. 4). Gsc is activated in dorsal mesoderm as a zygotic response to maternal dorsalizing agents like activin, Vg1 or wnt family members shortly after midblastula transition (Cho et al., 1991). In zebrafish, gsc is expressed in the developing shield, the region corresponding to the organizer in fish (Stachel et al., 1993; Schulte-Merker et al., 1994; Thisse et al., 1994). At the onset of gastrulation gsc expression is restricted to cells of the anterior-most axial mesoderm, the prechordal plate. Expression of gsc is correctly initiated in both oep and wild-type animals before gastrulation (Fig. 4A,B). At the beginning of gastrulation and thereafter, however, gsc expression is absent or strongly reduced in oep mutant embryos (Figs. 4C-H). These data suggest that oep is required for the development of the prechordal plate before or at the onset of gastrulation.

The finding that gsc expression is activated but not maintained in oep mutants suggests that oep acts downstream of mesoderm inducers as a maintenance factor for gsc expression or gsc-expressing cells. One prediction of this model is that exposure of oep mutants to dorsalizing agents would initiate ectopic gsc expression which would not be maintained because of the later requirement for oep. Alternatively, exposure of dorsalizing agents might induce gsc and other genes and bypass the requirement for oep. Exposure to lithium ions, a dorsalizing agent before midblastula transition, leads to the ectopic activation of gsc in ventral and lateral cells in wild-type zebrafish (Stachel et al., 1993).We find that gsc expression is radially induced in lithium-treated oep embryos but cannot be maintained (Fig. 4I-L). This is consistent with a model in which oep acts downstream of dorsalizing signals and early factors that initiate gsc expression.

The above analyses indicate that oep is required for the formation of the prechordal plate before or at the onset of gastrulation. Where does oep function in this process? One possi-bility is that oep acts non-autonomously, perhaps being required for the formation of a signal(s) that controls pre-chordal plate development. Alternatively, it could be required cell-autonomously in prechordal plate precursors. To distinguish between these possibilities, we transplanted labeled wild-type cells into mutant hosts and vice versa, and assayed for the formation of hatching gland cells (Fig. 5A-C). Wild-type cells are capable of forming hatching gland in oep mutants (Fig. 5C), but we never observed the formation of hatching gland cells by mutant cells in wild-type hosts (see Fig. 5 legend). Fur-thermore, the wild-type cells in oep mutants are not able to induce or recruit mutant host cells to form hatching gland (Fig. 5C). These data demonstrate that oep acts strictly cellautonomously in the prechordal plate precursors of the hatching gland. We have extended these observations by transplanting cells of lithium-treated wild-type donors into lithium-treated mutant hosts and then assaying for maintenance of gsc expression at midgastrula stages. Consistent with the previous transplantation experiments, we find that wild-type cells are able to express and maintain gsc in mutant animals (Fig. 5D-F). These results demonstrate that oep acts cell-autonomously in prechordal plate progenitors.

Studies in frogs suggest that the prechordal plate gives rise not expression (Strähle et al., 1993, Krauss et al., 1993) in the pharyngeal endoderm of oep mutants is absent or strongly reduced at 51 hpf (Fig. 6B,F). Surprisingly, shh expression, which is found throughout the entire endoderm of wild-type embryos (Fig. 6A,C), is also strongly reduced in the posterior of oep mutants (Fig. 6D). This demonstrates that oep not only affects the formation of pharyngeal endoderm, but also more posterior endodermal structures, consistent with the observation that the gut is absent or strongly reduced in oep mutants (Fig. 6H).

To determine when the endoderm is abnormal, we analyzed the expression of collagen type II (Yan et al., 1995) during the segmentation period. We find that collagen type II expression in the presumptive endoderm is strongly reduced in oep mutants at the 12-somites stage (Fig. 6J). Fate mapping studies have shown that endodermal precursor cells are located around the entire margin of the early gastrula (Kimmel et al., 1990; Warga, 1996), and after involuting acquire a flat morphology before finally reaching the dorsal midline (Warga, 1996). We have found that during gastrulation, the axial gene (Strähle et al., 1993) appears to be expressed in all or a subset of these endodermal precursor cells positioned in close proximity to the yolk cell (Fig. 6M-O). This is consistent with the endodermal expression of HNF3-α and -β in other species (Ang et al., 1993; Ruiz i Altaba et al., 1993; Monaghan et al., 1993; Sasaki and Hogan, 1993). To determine if oep functions early in endoderm formation, we examined the expression of axial during gastruonly to prechordal plate mesoderm but also to pharyngeal endoderm (Keller, 1976; Shih and Keller, 1992). Since it is assumed that cells expressing gsc at the onset of gastrulation are fated to give rise to both prechordal plate mesoderm and pharyngeal endoderm (DeRobertis et al., 1992), the loss of gsc expression in oep mutants prompted us to determine if pharyngeal endoderm is also affected in oep mutants. As expected, we find that axial and shh lation. Consistent with an early requirement of oep in endoderm formation, axial expression is absent or strongly reduced in the hypoblast of oep mutants (Fig. 6P). Thus, oep is not only active in the dorsal mesoderm but is also required, directly or indirectly, in more lateral and ventral regions.

Since the endoderm has been implicated in the induction of the myocardium in frogs (Jacobson and Duncan, 1968; Sater and Jacobson, 1990; Nascone and Mercola, 1995; Schultheiss et al., 1995), we examined the heart in oep mutants. We find that oep mutants have defects in heart muscle as judged from morphological observations and the expression of ?-tropomyosin (Fig. 6L; Thisse et al., 1993). The heart is small or absent, and sometimes cardia bifida is apparent. Pharyngeal endoderm has also been implicated in the induction of cartilage formation in pharyngeal neural crest cells (Graveson and Armstrong, 1987; Seufert and Hall, 1990), and we find that cartilage formation in the jaw region of oep mutants is reduced as judged from Alcian blue staining at 80 hpf (data not shown).

The prechordal plate has been implicated in several aspects of neural patterning (reviewed by Ruiz i Altaba, 1993). Based on embryological experiments (Spemann, 1931; Mangold, 1933; Gerhart et al., 1989; Ang and Rossant, 1993; Ang et al., 1994) and the analysis of mouse mutants in lim1 (Shawlot and Behringer, 1995) or otx2 (Acampora et al., 1995; Matsuo et al., 1995; Ang et al., 1996), the prechordal plate has been proposed to be an equivalent of the head organizer, a source of signals inducing anterior neural structures including forebrain and midbrain. Additionally, embryological studies in amphibians have suggested that the prechordal plate is involved in the separation of the eye forming region (Adelmann, 1936) or the induction of the eye forming region, but not of the entire forebrain (Dixon and Kintner, 1989; Ruiz i Altaba, 1992; Papalopulu and Kintner, 1993). The absence of a prechordal plate in oep mutants has allowed us to test some predictions of these models in zebrafish. Morphological analysis indicates that oep mutants have a normal anterior-posterior patterning of the brain (Fig. 1E-H). Telencephalon, midbrain and hindbrain are distinct. Anterior-posterior patterning of the brain in oep mutants is also revealed by the analysis of expression patterns of pax-a, pax-b, and krox-20 expression in forebrain, midbrain and hindbrain subregions during segmentation (Krauss et al., 1991; Püschel et al., 1992a (Krauss et al., 1992; Püschel et al., 1992b; Oxtoby and Jowett, 1993; Fig. 7B). These data do not support models in which the prechordal plate is essential for forebrain or midbrain formation during zebrafish gastrulation.

In contrast, dorsoventral patterning in the forebrain is severely disrupted in oep mutant embryos, most clearly mani-fested as severe cyclopia (Fig. 1J). Ventral forebrain regions are not present as indicated by the loss of the hypothalamus (Fig. 1F-H) and the severe reduction or absence of shh (Fig. 7F; Krauss et al., 1993) and nk2.2 (Fig. 7H; Barth and Wilson, 1995) expression in ventral neuroectoderm. These observa-tions are in accord with classical embryological studies by Adelmann (1936), who showed that the absence of head mesoderm can lead to cyclopia, a phenotype very reminiscent of oep. These results provide further evidence in favor of the idea that the prechordal plate might be involved in the induction of ventral cell types in the forebrain.

Oep is not only required for the formation of ventral structures in the forebrain, but is also involved in the proper formation of the floor plate. As judged from morphological analysis and the expression of floor plate markers like shh (Fig. 7I-L), typeII collagen, axial or F-spondin (Klar et al., 1992; data not shown), the number of floor plate cells in oep mutants is reduced at 28 hpf. To determine when the ventral neuroectoderm phenotype becomes apparent, shh expression was studied at the end of gastrulation. Normal shh expression in the ventral neuroectoderm is not observed in oep mutant animals (Fig. 7D), but is found in the developing notochord. We conclude that oep is required for the specification of ventral neuroectoderm during gastrulation. This defect may result indirectly from the lack of proper prechordal plate formation or from a direct effect of oep on neuroectoderm.

Two zebrafish mutants, ntl and flh, affect the formation of the notochord (Halpern et al., 1993; Schulte-Merker et al., 1994; Talbot et al., 1995). To determine the embryological effects of disrupting the development of the entire axial mesoderm and to test for possible genetic interactions, we constructed oep flh and oep ntl double mutants, and studied the expression of the flh and ntl genes in oep mutants.

The homeobox gene flh, a homologue of the Xenopus Xnot gene (von Dassow et al., 1993), is first expressed in the blastula margin and then becomes restricted to the shield and notochord (Talbot et al., 1995). Consistent with the studies described above, these flh expression domains are present in oep mutants (data not shown). We conclude that oep is not required for early flh expression. During somitogenesis flh is expressed in progenitors of the hatching gland in wild-type but not oep mutant embryos (data not shown).

Oep flh double mutants do not show any new dramatic defects that are not already present in the single mutants (Fig. 8E). Defects seem mainly additive as in the case of the axial mesoderm (absence or reduction of prechordal plate (oep) and notochord (flh) at the beginning of somitogenesis), or reflect the individual mutant phenotypes as in the case of cyclopia (oep), or somite patterning (lack of muscle pioneers as in flh). The correct number of somites forms, and pronephros and blood develop normally in oep flh double mutants.

However, more detailed inspection reveals that oep flh double mutants have enhanced floor plate and adaxial muscle phenotypes. A few floor plate cells are present in oep mutants (Figs 7K,L, 9C, 10B), and the floor plate seems normal an-teriorly, but scattered posteriorly in flh mutant embryos (Figs 9E,F, 10C). In contrast, double mutants display a complete absence or very severe reduction of floor plate cells during seg-mentation (Fig. 9G,H) and at 28 hpf (Fig. 10D).

Adaxial muscle cells lie adjacent to the notochord (Thisse et al., 1993; Kimmel et al., 1995). Both oep and flh single mutants develop adaxial cells as judged from the expression of myoD (Weinberg et al., 1996) at the 11-somites stage (Fig. 11). Interestingly, expression of myoD is strongly reduced in the posterior region of oep flh double mutants (Fig. 11E,K). This result suggests that oep and flh or the structures that they primarily affect have a partially overlapping role in patterning structures adjacent to the axial mesoderm.

Ntl is first expressed in the entire marginal zone, then in the developing notochord, and finally in the tail bud (Schulte-Merker et al., 1992). Consistent with the analysis of oep described above, no loss of ntl expression was observed in oep mutants (Fig. 3H, and data not shown). We conclude that oep is not required for the regulation of ntl gene expression.

Oep ntl double mutants show a dramatic deficit in the formation of mesoderm (Fig. 8F). As compared to the single mutants where notochord (ntl), prechordal plate (oep), and a subset of adaxial cells (ntl) are affected, double mutants retain only the anterior-most somitic mesoderm as judged from the expression of ?-tropomyosin (Fig. 12D) and myoD (Fig. 11F,L). Somitic defects are already apparent during gastrulation. Expression of snail1 (Hammerschmidt and Nüsslein-Volhard, 1993; Thisse et al., 1993) in somitic progenitors is slightly reduced in ntl mutant embryos (Fig. 12G), but drastically reduced or absent in oep ntl double mutants (Fig. 12H). The formation of ventral mesodermal cell types like blood is also reduced in oep ntl double mutants as judged from the expression of gata1 (Detrich et al., 1995) during somitogenesis (Fig. 12L). Intermediate mesoderm seems to form in oep ntl double mutants as judged from the expression of pax-b (Püschel et al., 1992b; Krauss et al., 1992) and lim1 (Toyama et al., 1995) during somitogenesis (data not shown). Midline structures are also severely affected in oep ntl double mutants; shh is expressed in the notochord and floor plate of wild-type embryos at the 12-somite stage (Fig. 9A,B). In oep mutants, few floor plate cells are present, but the notochord appears normal (Figs 9C,D, 10B). In ntl mutants, a string of cells expresses shh in the trunk (Fig. 9I,J, 10E). In contrast, oep ntl double mutants have no shh expressing cells, except in the anterior trunk region (Figs 9K,L, 10F). Interestingly, these cells are not located only at the midline, but are displaced laterally as patches of shh expressing cells (Fig. 9L). Inspection of shh, axial and F-spondin expression at later stages, when these genes are specifically expressed in the floor plate, suggests that the few remaining midline cells in oep ntl double mutants correspond to floor plate cells (data not shown). These data demonstrate that the oep ntl double mutants have an enhanced early midline phenotype. Consistent with this observation, expression of axial in the presumptive notochord is often absent in oep ntl double mutant embryos at the end of gastrulation (data not shown). In summary, the drastic deficits in oep ntl double mutants indicate that oep and ntl interact and have partially redundant roles in the formation of mesoderm and trunk midline structures.

The examination of gene expression patterns in oep mutants indicates that oep is involved in the early steps of prechordal plate and endoderm formation. Both gsc expression in pre-chordal plate progenitors and axial expression in presumptive endodermal cells are absent or strongly reduced in oep mutants. Our transplantation studies show that oep acts cell-autonomously in prechordal plate precursor cells. Furthermore, oep is also cell-autonomously required in the marginal progenitors of axial-expressing hypoblast cells (AS,SN & WD, unpublished results). Studies in frogs suggest that the formation of endoderm and dorsal mesoderm is initiated by maternal signaling molecules like Vg1 or activin (Rosa, 1989; Kessler and Melton, 1994; Cornell et al., 1995; Gamer and Wright, 1995; Henry et al., 1996). The initial response to these factors seems to be normal in oep mutants as genes like gsc or Brachyury are turned on normally after midblastula transition. This early activation has been shown to be independent of zygotic gene activity (Cho et al., 1991; Smith et al., 1991), and, as expected, is independent of the zygotic oep locus. As a zygotic downstream gene, oep is required in the execution of the muscles are derived from the prechordal plate, the first invo-luting dorsal cell group (Adelmann, 1932; Keller, 1976; Jacob et al., 1984; Wachtler et al., 1984). Fate map studies suggest, but have not unequivocally demonstrated, that the prechordal plate has the same fate in zebrafish (Kimmel et al., 1990; Halpern et al., 1993; Kimmel et al., 1995). Furthermore, the anterior portion of the zebrafish prechordal plate gives rise to the pillow (polster), which later differentiates into the hatching gland (Thisse et al., 1994; Kimmel et al., 1995). The oep mutant phenotype and the correlation between the early absence of prechordal plate and the later deficit in eye muscles, hatching gland and pharyngeal endoderm support the view that the zebrafish prechordal plate has a fate similar to that estab-lished in other vertebrates.

The oep mutant phenotype is reminiscent of proposals of Adelman (1936), who suggested that the prechordal plate might be a midline signaling center. Transplanting the eye-forming region of neural plate Amblystoma embryos to the belly region, Adelman found that two separated eyes develop only when the underlying head mesoderm is included in the transplant. Transplanting isolated neural plate alone leads to the formation of a single, median eye. These results led to the suggestion that the prechordal plate might be involved in the formation of ventral brain structures, leading to the separation of the eye field into two units. The phenotype of oep mutants is consistent with this proposal. Oep mutants display extreme cyclopia and a loss of ventral forebrain structures. In this scenario, the primary role of oep is in the formation of the prechordal plate and the forebrain defects would then be a consequence of the loss of a ventralizing center underlying the programs leading to the formation of endoderm and pre-chordal plate.

Studies in amphibian and avian embryos have suggested that pharyngeal endoderm and eye forebrain. Alternatively, oep might (also) have a direct autonomous function in the formation of the ventral forebrain. The prechordal plate has also been implicated in the induction of anterior neuroectoderm (reviewed by Ruiz i Altaba, 1993; Doniach, 1995). Classical transplantation experiments have shown that presumptive head mesoderm can induce the formation of anterior neural structures in host embryos (Mangold, 1933; Spemann, 1931). In contrast, in these experiments, the presumptive posterior mesoderm induced more posterior neural structures. These and other results led to the proposal that the neural plate is patterned along the anterior-posterior axis by vertical signals from the underlying mesoderm. The head mesoderm would then correspond to the head organizer and the more posterior mesoderm to the tail organizer. The head organizer proposal has recently been revived by the finding that lim1 mutant mice lack structures anterior to rhombomere 4 in the hindbrain (Shawlot and Behringer, 1995). As lim1 is expressed in the organizer and prechordal plate, one possible interpretation of the lim1 mutant phenotype is that the loss of prechordal plate and corresponding signaling function leads to the loss of the head organizer and head. Similar proposals have been put forward to explain the otx2 mutant phenotype (Acampora et al., 1995; Matsuo et al., 1995; Ang et al., 1996). The absence of the prechordal plate in oep mutants provides a test of this model in zebrafish. We find that despite the absence of prechordal plate during gastrulation, anterior-posterior patterning of the brain is largely undisturbed, allowing the formation of structures like telencephalon and midbrain. This finding supports models of planar induction (Ruiz i Altaba, 1992, Doniach et al., 1992), in which signals from the organizer (including prechordal plate precursors) directly induce anterior-posterior patterning in the neuroectoderm, without the need for underlying head mesoderm. Explant studies in frogs have also suggested that some aspect of forebrain development can be induced in a planar fashion (Papalopulu and Kintner, 1993). Thus, head organizer genes like lim1 might exert their effects in the organizer, prior to the formation of the prechordal plate.

The finding that oep is defective in the early formation of endoderm provides the first example of such a phenotype in vertebrates. Fate map studies in zebrafish have shown that the endoderm derives from the most marginal region of the late blastula embryo, partially overlapping with mesodermal pre-cursors (Kimmel et al., 1990, 1995; Warga, 1996). During gastrulation this cell population involutes and streams from the margin towards the animal pole. Mesoderm and endoderm are not at first distinguishable as separate germ layers but form the hypoblast, an apparently single layer of cells. Our observation that axial (Strähle et al., 1993) is expressed in a sub-population of hypoblast cells located in close proximity to the yolk cell, and that this cell population is specifically affected in the endoderm mutant oep, suggests that axial-expressing cell represent some or all of the endoderm precursors during gastrulation. Thus, distinct populations of mesoderm and endoderm cells are present in the hypoblast of the gastrula.

The deficit of endoderm in oep mutants provides a tool to study the postulated roles of endoderm in the formation of mesodermal and ectodermal tissues like heart (Jacobson and Duncan, 1968; Sater and Jacobson, 1990; Nascone and Mercola, 1995; Schultheiss et al., 1995) or pharyngeal cartilage (Graveson and Armstrong, 1987; Seufert and Hall, 1990). We find that both heart muscle and pharyngeal cartilage differentiation are compromised in oep mutants. Although we cannot exclude an autonomous role of oep in these structures, oep mutant defects are consistent with embryological data that suggest an inductive role for endoderm.

The floor plate, the ventral-most cell type in the neural tube, is thought to be induced by signals from the notochord (van Straaten et al, 1988; van Straaten and Hekking, 1991; Placzek et al., 1990). The signaling molecule shh has been implicated in this process, as it can induce floor plate and is expressed in the notochord (Krauss et al., 1993; Echelard et al., 1993; Roelink et al., 1994). We find that oep mutations disrupt early floor plate development, and that this is not caused by a failure of oep mutant notochord cells to express shh. These results suggest that oep might be a downstream component or response to the shh signaling cascade, or might act in a parallel pathway. Alternatively, the requirement for oep might be less direct, and earlier deficits in oep mutant animals, e.g. the absence of prechordal plate, might lead tiimpaired floor plate development. It is conceivable that the early contact of prechordal plate cells with the overlying neuroectoderm during gastrulation contributes to floor plate induction, either by direct induction or by priming the neuroectoderm to respond to later signals from the notochord. Analysis of genetic mosaics would determine the role of oep in floor plate induction, but the significant number of floor plate cells that develop in oep mutants has precluded this analysis. It is therefore unclear if oep is directly and cell-autonomously involved in the formation of floor plate, or if the ventral deficits are due to the role of oep in other structures like phenotype of oep mutants. Whereas a variable number of floor plate cells form in oep or flh mutants, oep flh double mutants show a dramatic loss of floor plate cells. Furthermore, as evidenced by reduced myoD expression, posterior adaxial muscle cells are also affected in oep flh embryos. Both adaxial and floor plate cell precursors are located adjacent to the notochord, and are also contacted by the prechordal plate during gastrulation. It is conceivable that the loss of both prechordal plate and notochord in oep flh double mutants leads to a significant reduction of axial mesodermal signaling activity and consequently affects the induction of adjacent structures more severely that in single mutants. It is also possible that oep and/or flh play direct autonomous roles in floor plate formation and act in a partially redundant or additive manner, a function that is uncovered in double mutants.

Oep ntl double mutants show extreme defects in the formation of mesodermal cell types, revealing a partially overlapping requirement of oep and ntl in mesoderm formation. This finding offers an explanation for an enigma concerning the expression and function of ntl. The ntl gene is first expressed in the marginal zone, including all mesodermal precursor cells; however, this expression domain does not seem to be required as judged from the normal development of mesoderm like muscle, pronephros, blood or heart in ntl mutants (Halpern et al., 1993). Mutants for ntl are only defective in the proper formation of notochord, muscle pioneer cells and tail. These mutant phenotypes have been suggested to reflect the requirement and expression of no the axial mesoderm.

The additive defects in oep flh double mutants suggest that oep and flh act primarily independently. Two structures, however, are more strongly affected in double mutants than in either of the single mutants. Loss of flh seems to enhance the floor plate tail in notochord and tail bud after the onset of gastrulation (Halpern et al., 1993). Our results suggest that the early requirement for ntl in mesoderm formation might be partially masked by the activity of oep. The oep or ntl genes alone might be sufficient for normal somite and blood formation. Removing both genes simultaneously reveals their partially redundant function and leads to drastic deficits in mesoderm development.

How do oep and ntl interact ? We find that oep disrupts the development of axial expressing cells derived from the entire marginal zone of the gastrulating embryo. Ntl is expressed in this region during this time. It is tempting to speculate that both oep and ntl act in marginal cells to ensure proper mesoderm formation. Both oep and ntl might act to specify mesodermal structures. This suggestion is consistent with studies in frogs, showing that ectopic expression of Brachyury can also induce the formation of mesodermal cell types other than notochord (Cunliffe and Smith, 1992, 1994; O’Reilly et al., 1995). Alter-natively, oep and ntl might also be required for the proper migration of mesendodermal cells. The latter proposal is supported by the study of wild-type and Brachyury mutant chimeric mouse embryos (Wilson et al., 1993, 1995). Brachyury mutant cells appear defective in proper mesodermal cell movements, which results in the progressive accumulation of mesoderm cells near the primitive streak. This ultimately blocks the formation of posterior mesoderm. If a similar scenario is applicable to zebrafish, oep and ntl might lead to migratory abnormalities in the marginal zone. It is interesting to note that oep mutants display reduced convergence-extension (Solnica-Krezel et al., 1996). Gastrulation movements could be partially blocked in oep ntl double mutants and lead to the observed deficits.

The idea that ntl has the same function in zebrafish as Brachyury in mouse is supported by the high degree of sequence conservation between the two genes and the resem-blance of their expression patterns (Halpern et al., 1993; Schulte-Merker et al., 1994). Indeed, both ntl and Brachyury mutant embryos display defective notochord differentiation and posterior truncations. However, it is clear that several char-acteristics of the mouse Brachyury mutant phenotype are not present in the zebrafish ntl mutant. In particular, Brachyury mutants form no more than 8 somites in the anterior trunk region and lack the floor plate (Chesley, 1935; Gluecksohn-Schoenheimer, 1944; Grueneberg, 1958; Beddington et al., 1992; Dietrich et al., 1993; Conlon et al., 1995). In contrast, ntl mutants form all trunk somites and a floor plate. It is inter-esting to note that the trunk phenotype of oep ntl double mutants is more closely related to the Brachyury mutant phenotype. Namely, midline structures are severely affected and somites are found only in the anterior trunk region in oep ntl double mutants. We might speculate that zebrafish oep has some of the functions or features of the mouse Brachyury gene product, thereby masking a broader role of ntl.

During the formation of axial mesoderm, oep primarily functions in the formation of the prechordal plate, whereas flh appears to promote notochord formation (Halpern et al., 1995; Talbot et al., 1995). The flh mutant phenotype has been interpreted as a cell fate specification defect. Similarly, oep might act as a prechordal plate cell fate specification gene. It has to be emphasized, however, that oep might also be involved in the formation of the notochord, a function revealed in a minority of oep mutants that show notochord defects, and the midline defects in oep ntl double mutants. Therefore, we suggest that oep acts in axial mesoderm precursor cells located in the organizer region where it is mostly required to allow the proper formation of the anterior-most, first involuting axial mesoderm. The deficits in ventral neuroectodermal structures in oep mutants are reminiscent of the cyclops mutant phenotype (Hatta et al., 1991, 1994). Cyc embryos display partial eye fusion, absence of floor plate cells, and a slight reduction of the prechordal plate (Thisse et al., 1994). The extreme cyclopia and absence of prechordal plate demonstrate that the head phenotype of oep is more severe than in cyc mutants. Interestsingly, the floor plate phenotype of cyc mutants appears to be more severe: only a few floor plate cells form in the tail region of cyc embryos. These comparisons suggest that oep and cyc are involved in the same or similar developmental pathways, but to different extents. Further, oep seems to be involved in additional processes as judged from the endoderm phenotype and the severe mesoderm defects in oep ntl double mutants.

In summary, the findings presented here establish oep as an essential zygotic component downstream of several inductive interactions in the vertebrate embryo, namely the formation of axial mesoderm, endoderm and ventral neuroectoderm. As a first step towards the molecular isolation of oep, we have mapped the oep locus to linkage group 10 on the zebrafish genetic map. So far, no candidate genes map to this region, and we can exclude oep as a mutation in a number of genes that have been invoked in organizer function or development, including gsc, lim1, axial, or shh (J. Postlethwait, W. S. T., and M. Gates, personal communication). Phenotypic analysis and the study of genetic mosaics suggest that oep candidate genes should be expressed in the dorsal mesoderm and the marginal zone at the onset of gastrulation. The molecular isolation of the oep locus should offer further insights into how oep functions in the patterning of all three germ-layers, and into the nature of its interaction with ntl.

We are grateful to David Grunwald, Chuck Kimmel, and John Postlethwait for support and helpful discussions during the course of this work. Jiaojiao Zhang helped in identifying oep^z1 as a deletion. We thank U. Strähle, P. Ingham, R. Warga, and J. Postlethwait for sharing unpublished results. The following colleagues kindly provided us with cDNA clones: T. Jowett, S. Krauss, U. Strähle, P. Ingham, A. Fjose, D. Duboule, B. Riggleman, R. Weinberg, A. Barth, S. Wilson, B. and C. Thisse, Y.-L. Yan, J. Postlethwait, S. Schulte-Merker, M. Hammerschmidt, C. Nüsslein-Volhard, L. Zon, H. Okamoto, D. Kimelman, A. Klar, T. Jessell, R. Toyama, and I. Dawid. We are grateful to Gord Fishell, Rolf Karlstrom, Ariel Ruiz i Altaba, and Lila Solnica-Krezel for critical reading of the manuscript. This work was supported by NIH RO1-HD29761 and a sponsored research agreement with Bristol Myers-Squibb to W. D., NIH RO1-AI26734 to J. Postlethwait, NIH 1PO1 HD-22486 to the University of Oregon Zebrafish Program Project, NIH 1RO3 RR07832-01A1 to D. Grunwald, a Jane Coffin Childs Memorial Fund fellowship to W. S. T. and EMBO and Swiss National Fond fellowships to A. F. S.