Genes involved in forebrain development in the zebrafish, Danio rerio

The process of neural induction in vertebrates has been postulated to be subdivided into two steps (for a review see Saxen, 1989): first there is an induction of anterior fates within the whole neural plate, called ‘activation’. This is followed by a ‘transformation’ of the neural plate to give posterior character; a gradient emanating from the posterior of the embryo regionalizes the neural axis into brain and spinal cord. This morphogenetic gradient is established by the mesendoderm; early involuting mesendoderm induces anterior neural fates and late involuting mesendoderm induces posterior ones (Eyal-Giladi, 1954).

Several candidate molecules involved in neural induction and subsequent patterning of the neural plate have been identified. Noggin, a secreted polypeptide (Smith and Harland, 1992; Lamb et al., 1993), follistatin, a secreted antagonist of activin (Hemmati-Brivanlou et al., 1994) and chordin, the vertebrate homologue of the Drosophila gene short gastrulation (sog) (Sasai et al., 1994, 1995; Francois and Bier, 1995) can induce neuroectoderm of anterior character, thereby mimicking the ‘activation’ step during neural induction. Retinoic acid as well as basic fibroblast growth factor may be involved in generating the morphogenetic gradient leading to the ‘transformation’ of the neural plate, since both have been shown to induce the development of neural tissue with posterior character (Durston et al., 1989; Kengaku and Okamoto, 1995). However, the role of these factors in vivo and the mechanisms by which these signals are transmitted to the ectoderm remain unclear, as results of a complete ‘knock-out’ of the genes have not been published (except for follistatin, where neural induction appears to be unaffected; Matzuk et al., 1995).

A number of experiments in amphibians suggest that signals needed for induction and patterning of the neural plate can be transmitted in different ways (Spemann, 1938; Kintner and Melton, 1987). There is evidence for both vertical transmission of signals from the underlying mesendoderm to the ectoderm and horizontal signaling within the plane of the ectoderm: exogastrulated embryos and explants of dorsal tissue, which mimic exogastrulae (‘Keller sandwiches’), have been employed to analyze neural-specific gene expression within the ectoderm in the absence of vertical signals (Keller and Danilchik, 1988; Ruiz i Altaba, 1990). Exogastrulated embryos form mesendoderm, which does not involute, therefore neural inducing signals from the mesendoderm must be exclusively transmitted horizontally through the plane of the ectoderm (Holfreter, 1933; Hamburger, 1988). In exogastrulae and ‘Keller sandwiches’, the expression of neural-specific genes along the anterior-posterior axis of the ectoderm is induced, suggesting that vertical signals from the underlying axial mesendoderm are not required (Ruiz i Altaba, 1992; Doniach et al., 1992; Papalopulu and Kintner, 1993). However, differentiation of specific cell types characteristic of the forebrain and ventral neural tube is not observed. These cell types are therefore likely to depend on both vertical and planar signals (Ruiz i Altaba, 1992, 1994; Papalopulu and Kintner, 1993).

The underlying axial head mesendoderm, the prechordal plate, is known to be essential for induction and patterning of the forebrain. Whereas the expression of genes specific for the forebrain anlage has been shown to be independent of vertical signals from the prechordal plate, the differentiation of forebrain derivatives like eyes, as well as the capability of the presumptive forebrain to induce midbrain (Nieuwkoop and Albers, 1990), depends on both vertical and planar signals (Ruiz i Altaba, 1992, 1994; Papalopulu and Kintner, 1993). Moreover analysis of the zebrafish cyclops (cyc) and one-eyed-pinhead (oep) mutants shows that a reduction of the prechordal plate is associated to a deletion of ventral forebrain tissue, suggesting that vertical signaling is involved in the formation of the dorsal-ventral forebrain axis (Hatta et al., 1994; Thisse et al., 1994; Hammerschmidt et al., 1996; Warga et al., unpublished data).

Subdivision of the forebrain along its major axes, similar to that shown for the hindbrain and spinal cord (Keynes and Stern, 1984), is still a matter of debate (for review see Fraser, 1993; Puelles and Rubenstein, 1993). Boundaries of expression domains of various markers found in the forebrain have been shown in many cases to coincide with morphological boundaries seen at different stages of forebrain development (Puelles et al., 1987; Figdor and Stern, 1993; Puelles and Rubenstein, 1993, Macdonald et al., 1994; Barth and Wilson, 1995). The diencephalon for example has been shown to be composed of four neuromeres along its anterior-posterior axis which are anatomically, molecularly and functionally definable (Figdor and Stern, 1993). However, many of the other proposed neuromere subdivisions, in particular within the telencephalon, must still be examined in more detail with respect to functional properties such as lineage restriction.

In this study we describe the isolation and initial characterization of zebrafish mutants exhibiting specific defects in the formation of the forebrain: patterning of the forebrain along the anterior-posterior axis is affected in mbl, positioning of the eyes is aberrant in slb, and differentiation of the roof plate separating the telencephalic primordia is defective in kas. The analysis of these mutant phenotypes may be helpful in elucidating important steps in forebrain patterning and differentiation.

Maintenance and breeding of fish has been described by Mullins et al. (1994). Identification and isolation of mutants together with the complementation analysis is described in an accompanying paper by Haffter et al. (1996).

Antibody and in situ stainings were performed as described previously (Hammerschmidt and Nüsslein-Volhard, 1993). For antibody stainings, anti-islet (Korzh et al., 1993; 1:500), anti-pax6 (also called pax[zf-a]) (Macdonald et al., 1994; 1:400), anti-pax2 (also called pax[zf-b]) (Mikkola et al., 1992; Püschel et al., 1992b; 1:100), antintl (Schulte-Merker et al., 1992; 1:1000), and anti-fkd2 (Warga et al., unpublished data; 1:2000) polyclonal antibodies, as well as anti-acetylated tubulin (Sigma; 1:1000) and anti-fret43 (Larison and Bremiller, 1990) monoclonal antibodies were used. For in situ hybridization digoxigenin-labeled RNA probes were synthesized from the full-length shh (Krauss et al., 1993), zotx-2 (Li et al., 1994), zash-1b (Allende and Weinberg, 1994), myoD (Weinberg et al., 1996), pax2 (Krauss et al., 1991a,b), and fkd3 (Odenthal et al., unpublished data) cDNA clones. For double stainings (in situ and antibody labeling), embryos first underwent complete in situ hybridization, followed by complete antibody staining. Sections of embryos were prepared after whole-mount stainings. Embryos were dehydrated (methanol, 10 minutes), permeabilized (acetone, 10 minutes), incubated in araldite/acetone (1:1; 12 hours) and embedded in araldite. The araldite-blocks were polymerized (12 hours, 80°C) and sectioned on a vibrotome in 10 μm intervals. Sections were mounted in araldite. Photographs were taken using an Axiophot photomicroscope (Zeiss).

Transplantations were done as described by Ho and Kane (1990). Cells from wild-type donor embryos labeled with rhodamine-dextran at the one-cell stage were transplanted into the animal pole blastoderm of 4-hour host embryos obtained by crossing mbl heterozygotes. The results were analyzed at 30 hours. A quarter of the hosts were expected to be mutant. Only partial rescue of the eye and forebrain phenotype could be identified since a complete rescue was expected to result in a phenotype indistinguishable from the wild-type siblings.

Screening for various classes of mutants was done by visual inspection of live embryos on the second, third and sixth day of development. 131 mutants were isolated and kept on the basis of morphologically recognizable defects of the nervous system. 18 of these mutants show regionally restricted defects within the nervous system (Brand et al., 1996a,b; Jiang et al., 1996; and this paper) whereas the others exhibit more general defects of neuronal tissue (Jiang et al., 1996; Furutani-Seiki, 1996). We also isolated mutants that showed defects of the axial mesendoderm accompanied by a general reduction of the ventral central nervous system (CNS) such as alleles of the cyclops (cyc) mutation (Hatta et al., 1991; Thisse et al., 1994; Warga et al., unpublished data). These mutants will be described elsewhere (Brand et al., 1996b; Hammerschmidt et al., 1996; Odenthal et al., 1996).

The zebrafish embryonic forebrain (prosencephalon) forms at the anterior end of the neural keel and is subdivided into a dorsal-anterior telencephalon and a ventral-posterior diencephalon (Ross et al., 1992). Structures formed or induced by the forebrain such as the retinae, olfactory placodes and the epiphysis are clearly visible in 24-hour embryos (Kimmel et al., 1995) and were used as morphological landmarks to discriminate forebrain mutants.

Four mutants defining three genes which exhibit a clear neuroectodermal defect restricted to the forebrain will be described in this paper: one allele of masterblind (mbl), two alleles of silberblick (slb) and one allele of knollnase (kas) (Table1).

The single mbl allele found (tm13) exhibits a recessive lethal phenotype which does not resemble any other phenotype seen in the screen. The strength of the phenotype is slightly variable within a single egg-lay and depends on the genetic background. The phenotype of live mbl embryos is characterized by the absence of optic vesicles seen already at 14 hours (Fig. 1A,D). Eventually the eyes (Fig. 1B,E) and olfactory placodes do not develop (Fig. 1C,F). The degree of variation ranges between the complete lack of eyes in severe cases to the formation of very small eyes (about 1/10 of wild type size) in milder cases. Phenotypic alterations are restricted to the anterior region of the CNS in mbl mutants; no other specific defects could be detected in mbl embryos as compared to wild type siblings. General necrosis of the mutant embryo is observed by 120 hours.

Primary neurons and their axons form a simple and well-described arrangement within the zebrafish embryonic CNS (Wilson et al., 1990; Ross et al., 1992). To determine which structures are altered in the CNS of mbl embryos, stainings using antibodies that recognize these elements were performed. Labeling of 24-hour mutant embryos for acetylated tubulin (Chitnis and Kuwada, 1990) revealed that the telencephalic clusters of primary neurons and olfactory placodes are missing (Fig. 2A,B,D,E). Evaluation of the axonal scaffold at 24 hours showed that the anterior commissure connecting the bilateral telencephalic neuronal clusters and the postoptic commissure connecting the bilateral nuclei of the tract of the postoptic commissure (nTPOC) do not form in the mutant (Fig. 2B,E). To test if the absence of a postoptic commissure is due to mbl embryos lacking the neurons which normally pioneer this commissure, we labeled embryos for acetylated tubulin and islet proteins (using an anti-pan-islet antiserum recognizing members of the LIM homeodomain protein family; Korzh et al., 1993). Anti-tubulin labeling showed that many differentiated neurons are present in the nTPOC of 24-hour mbl embryos (Fig. 2A,D). However, anti-islet labeling revealed major differences in the neurons between wild-type and mutant embryos in that none of the nTPOC cells express islet proteins in mbl embryos at 24 hours (Fig. 2C,F). This result is consistent with the possibility that an islet-proteins-expressing subpopulation of the nTPOC cells, normally responsible for pioneering the postoptic commissure, is absent in mbl. The alterations in the organization of the CNS in mbl embryos are schematically illustrated in Fig. 2G.

Structures associated with the ventral diencephalon, such as the anterior pituitary, appear to be reduced in size, as seen by a reduction in the number of pax6-positive cells within the anterior pituitary of mbl embryos compared to wild type at 30 hours (Fig. 3A,D) (Krauss et al., 1991b,c; Püschel et al., 1992a; Macdonald et al., 1994). In contrast, the epiphysis, a dorsal diencephalic structure, is strongly expanded in mutant embryos, as seen by staining neurons within the epiphysis for islet proteins (Fig. 3B,E) and for fret43 (Larison and Bremiller, 1990) at 28 hours (Fig. 3C,F).

We examined whether the alterations in the architecture of the forebrain in mbl embryos are preceded by changes in the anlage of these structures at earlier stages of development. Therefore we analyzed the expression of genes in specific sub-domains of the embryonic forebrain by in situ hybridization. The diencephalic expression domain of zotx-2 (a zebrafish relative of Drosophila orthodenticle, Li et al., 1994) is expanded anteriorly in mutant embryos at 14 hours (Fig. 4A,D). Correspondingly the telencephalic expression domain of zash-1b (a zebrafish relative of Drosophila genes of the achaete-scute complex, Allende and Weinberg, 1994) is absent as shown at 12 hours (Fig. 4B,E). In wild-type embryos, sonic hedgehog (shh) (a zebrafish homologue of Drosophila hedgehog, Krauss et al., 1993) is expressed within the forebrain in the posterior diencephalon and in the presumptive anlage of the hypothalamus at 14 hours (Krauss et al., 1993; Barth and Wilson, 1995). In mbl mutants it forms a continuous band of strong expression extending to the anterior end of the neural keel (Fig. 4C,F). This is consistent either with an anterior extension of the posterior diencephalic expression domain or a posterior extension of the anterior expression field. In summary, these results suggest that the anlage of the telencephalon is already strongly reduced in mutant embryos at early neural keel stages whereas the anlage of the posterior diencephalon is extended anteriorly.

Since the trigeminal ganglion in mbl embryos appears to be abnormally shaped at 24 hours (Fig. 2B,E), we examined the anlage of the trigeminal placodes at earlier stages. By 12 hours, cells within the trigeminal placode are islet-positive in the wild type (Korzh et al., 1993). In mbl embryos there are ectopic clusters of islet-positive cells anterior to the trigeminal placodes. Moreover, the trigeminal placodes that are positioned normally appear slightly enlarged in mutant embryos (Fig. 5B,D). Thus the trigeminal placodes are expanded anteriorly at the expense of the lens and olfactory placodes which are not formed in mbl (Fig. 1B,E and data not shown).

The forebrain defects in mbl embryos may be caused by the lack of signals from the prechordal plate to the neuroectoderm. We therefore examined the expression of zebrafish goosecoid (gsc) (Stachel et al., 1993; Schulte-Merker et al., 1994a) and fkd2 (a zebrafish forkhead domain transcription factor, Odenthal et al., unpublished data; Warga et al., unpublished data) which predominantly stain mesendodermal cells within the prechordal plate during gastrulation. Expression of gsc and fkd2 within the prechordal plate before the tailbud stage (10 hours) appears normal (data not shown). At 10 hours and during the early somite stages, gsc expression overlying the anterior-lateral edge of the prechordal plate is clearly reduced in mbl mutants (Fig. 5A,C). However, this reduction of gsc expression may reflect a neuroectodermal rather than a mesendodermal defect since gsc has been reported to be primarily expressed in neuroectodermal cells in this region (Thisse et al., 1994). We therefore have no clear evidence for the prechordal plate being affected in mbl.

To determine if mbl acts in a cell-autonomous manner, cell transplantation experiments were performed. We transplanted 10-20 rhodamine-dextran-labeled wild-type cells into mutant embryos at late cleavage stages prior to the onset of gastrulation and asked if these cells could bring about the formation of eyes in mbl embryos. Only egg-lays in which none of the untransplanted mutant embryos showed any sign of eyes were included in this study. In 7 out of 33 cases, eye-like structures (exhibiting characteristics of eyes such as production of melanin, expression of pax6; data not shown) composed of 10-30 labeled cells were found in mutant embryos (Table 2 and Fig. 6). Mutant cells were not seen to be recruited into these structures, indicating that mbl acts in a cell-autonomous manner within the neuroectoderm. In all transplantations leading to the formation of eye-like structures in mbl, other labeled wild-type cells were observed in various regions of the forebrain (Table2). Therefore it cannot be excluded that mbl is also needed in a non-cell-autonomous way within other regions of the forebrain to promote the formation of eyes.

Although two cases of transplantation of wild-type cells into the head mesenchyme of mbl embryos were observed (Table 2) these experiments were not performed in sufficient detail to gain insight into whether mbl function is also needed outside the CNS in a cell-non-autonomous manner to induce the formation of eyes.

Pax6, a member of the murine paired-box-containing gene family (Walther and Gruss, 1991), and its homologues have been shown to be essential for the development of eyes in such divergent species as mice and flies (Hill et al., 1991; Quiring et al., 1994). Since mbl is the only mutant in our screen which lacks eyes, a phenotype reminiscent of mice homozygous for mutations in the pax6 gene (Hill et al., 1991), we tested whether mbl is linked to the zebrafish pax6 homologue. A recombination analysis using a restriction fragment length polymorphism (RFLP; as previously described by Schulte-Merker et al., 1994b) showed that mbl is not linked to pax6 (data not shown).

slb is a zygotic mutation with two alleles (tx226, tz216) of approximately the same phenotypic strength. slb displays a recessive phenotype which is variable within a single egg-lay and depends on the genetic background. The phenotype is fully penetrant at early stages, but only partially penetrant at later stages of development. Adult viability was not tested.

During the early somite stages, slb embryos are shorter than wild type. The polster, a structure of the anterior prechordal plate thought to contribute to the hatching gland (Ballard, 1973, Kimmel et al., 1990), appears smaller (Fig. 7A,D). At 24 hours, in about half of the mutant embryos, the retinae are not properly separated anteriorly (Fig. 7B,E) whereas other parts of the embryo appear normal. Mutants with a strong phenotype show a nearly complete fusion of the retinae, mutants exhibiting a weak phenotype have only slightly closer retinae. At 120 hours the jaw is deformed (Fig. 7C,F).

In order to elucidate the eye-phenotype of slb embryos, we stained embryos for pax2 (Krauss et al., 1991a,b), which labels the presumptive optic stalk region, and pax6 (Macdonald et al., 1994), which labels the developing retinae. At 20 hours the pax2 expression domain is broader across the midline, suggesting that the optic stalks may be thicker (Fig. 8A,C), whereas the retinae are still separated from each other (Fig. 8B,D).

A gastrulation phenotype in slb mutants is transiently visible between 10 and 12 hours: the anlage of the notochord in the posterior axial mesendoderm is broadened and shortened, as visualized by the expression of no tail (ntl) in notochord precursor cells at 10 hours (Fig. 9) (Schulte-Merker et al., 1992). In contrast, formation of the paraxial mesoderm appears to be unaffected in slb mutants, as assayed by the expression of myoD, a marker for paraxial/presomitic mesoderm (Weinberg et al., 1996) which is unperturbed in mutants at 10 hours (Fig. 9B,C). The prechordal plate in the anterior axial mesendoderm appears to be abnormally shaped, as shown by staining mesendodermal cells within the prechordal plate for fkd2. Clustering of fkd2-positive cells at the anterior edge of the prechordal plate, where the polster forms (Warga et al., unpublished data), is not seen in mutant embryos at 10 hours (Fig. 10A,C). In parallel to the situation in the hypoblast, the presumptive neural plate in slb embryos appears to be broader during gastrulation, particularly in the head region as shown by the expression of fkd3 (a zebrafish forkhead domain transcription factor, Odenthal et al., unpublished data) and pax2 (Mikkola et al., 1992; Püschel et al., 1992b) in the dien-cephalon anlage at 10 hours (Fig. 10A-D). The expression domain of shh in the anterior-most part of the neural keel overlying the prechordal plate is shortened in slb mutants (Fig. 10B,D). At later stages of development (16 hours), however, slb embryos show normal expression of shh (data not shown).

kas is a zygotic lethal mutation. A single allele (ty122) was found with a recessive phenotype which is distinct from all other phenotypes seen in the screen. In kas embryos formation of the telencephalic ventricle which underlies the dorsal midline roof plate tissue and separates the telencephalic primordia is defective, as shown in a 24-hour embryo in Fig. 11A,D. Horizontal sections through the telencephalon of 24-hour embryos stained for acetylated tubulin show that the arrangement of cells at the presumptive midline is severely dis-organized in kas mutants (Fig. 11B,E). The malformation of the roof plate is accompanied by a defasciculation of the anterior commissure connecting the telencephalic clusters as shown by staining axons for acetylated tubulin at 24 hours (Fig. 11C,F). In contrast, commissures crossing the ventral midline such as the postoptic commissure appear normal in kas (data not shown). Thus in kas mutants the dorsal midline within the anterior forebrain does not differentiate properly.

The mutants analyzed in this study were isolated in the course of a large scale mutagenesis screen for mutants affecting embryonic development in zebrafish (Haffter et al., 1996). Only mutants with easily visible morphological changes in the architecture of the forebrain could have been detected. Since the forebrain in teleosts is a rather inconspicuous structure in comparison to other parts of the embryo (for instance the notochord) we may have missed a significant portion of forebrain mutants exhibiting more subtle defects. Furthermore, there may be mutants with still undiscovered defects in the forebrain, which were identified and subsequently classified on the basis of different phenotypic criteria like motility (Granato et al., 1996) and changes in the retinotectal projection (Trowe et al., 1996; Karlstrom et al., 1996).

Both mbl and kas were identified by single alleles and are therefore clearly underrepresented in comparison to other genes in this screen (Haffter et al., 1996). In the case of kas, the phenotype appears so subtle that weaker alleles might have been missed. In the case of mbl, a mutant with an identical phenotype was found and subsequently lost, which might have been allelic to mbl considering the singularity of this phenotype.

In mbl mutants the telencephalon, the anterior pituitary and the optic vesicles are absent or reduced in size. The epiphysis, a dorsal diencephalic structure, is expanded anteriorly. We propose that in mbl embryos the part of the forebrain which is reduced or absent is actually transformed into the enlarged forebrain structures. Both the expression pattern of various markers within the forebrain at early neural keel stages and the morphological alterations seen in later stages of development are consistent with this hypothesis.

The situation in the forebrain of mbl mutants is reminiscent of homeotic transformations of body segments described in Drosophila and subsequently in vertebrates (for review see Lawrence and Morata, 1994; Krumlauf, 1994): combinatorial expression of homeobox genes is known to specify segmental identities. Altering this combinatorial code by either shifting or turning off the expression of these genes leads to a trans-formation of segmental identities, so that a whole segment takes on a new identity. In recent studies the expression patterns of various homeobox genes within the vertebrate forebrain have been mapped and correlated to the proposed neuromere subdivisions (Simeone et al., 1992). Since the mbl phenotype presumably represents a transformation within the forebrain, these homeobox genes are likely candidates for involvement in the generation of the mbl phenotype.

The nearly complete loss of a restricted subset of forebrain structures such as the telencephalon and the eyes in mbl mutants indicates that there may be a common pathway for the determination of these structures. This may indicate that they share segmental identity. How could such a ‘segment’ fit into the proposed neuromere subdivisions of the forebrain? The forebrain is thought to be subdivided into neuromeres, which are defined by morphological landmarks and the expression domains of various genes respecting the proposed neuromere boundaries (Figdor and Stern, 1993; Puelles and Rubenstein, 1993; Macdonald et al., 1994). In mbl mutants the reduced structures are located anteriorly to the epiphysis, whereas the epiphysis itself is enlarged. Just anterior to the epiphysis lies the middiencephalic boundary (MDB), presumably equivalent to one of the proposed neuromere boundaries separating the ventral from the dorsal thalamus, the zona limitans interthala-mica described in other species (Figdor and Stern, 1993; Puelles and Rubenstein, 1993; Barth and Wilson, 1995). The MDB has been reported to overlap with the anterior diencephalic expression boundary of both shh and zotx-2 (Krauss et al., 1993; Li et al., 1994; Barth and Wilson, 1995), which is shifted to the anterior end of the neural keel in mbl mutants. This observation, together with the finding that the morpho-logical alterations in mbl embryos are restricted to an area within the forebrain anterior to the MDB, supports the view that the MDB defines a segmental border between neuromeres. The outcome of the transplantation experiments indicates that mbl acts in a cell-autonomous manner within the neuroectoderm. It therefore appears likely that mbl is needed by a population of cells at the anterior end of the neural keel to differentiate into the structures missing in the mutant. What makes this population of cells distinct from other cells located in more posterior parts of the neural keel? It could be either the presence of signals originating from the underlying axial mesendoderm, the prechordal plate, or planar signaling within the neural plate. The observation that a nearly complete loss of the prechordal plate in the zebrafish one-eyed-pinhead (oep) mutant does not substantially affect anterior-posterior patterning of the forebrain (Hammerschmidt et al., 1996) indicates that the prechordal plate is not the source of signals specifically patterning the anterior-posterior forebrain axis. Planar signaling within the neural plate is therefore likely to be involved in determining the identity of the cells affected in the mbl mutant. If they loose their ability to respond to these signals, as may be the case in mbl, they adopt more posterior fates, resulting in a transformation of anterior into more posterior forebrain structures.

slb mutants display a very restricted defect within the CNS: an anterior fusion of the eyes. The zebrafish cyclops (cyc) mutant, which has been isolated and analyzed previously (Hatta et al., 1991), shows a similar but stronger eye phenotype. The ventral midline tissue within the neural keel in cyc embryos is reduced (Hatta et al., 1991, 1994; Patel et al., 1994; Macdonald et al., 1994), accompanied by a reduction in the underlying axial mesendoderm (Thisse et al., 1994; Warga et al., unpublished data). In contrast, there is no general reduction of midline tissue detectable in slb mutants (data not shown). Also, the type of eye fusion in slb differs from the cyc phenotype, in that the eye primordia contain pax2-expressing cells demarcating the area of the forming optic stalk. Pax2-positive cells are reduced or absent and so little or no optic stalk tissue is detectable in cyc mutants (Hatta et al., 1991, 1994; Ekker et al., 1995; Macdonald et al., 1995). The slb eye phenotype therefore seems to be due to an aberrant differentiation of midline tissue restricted to the anterior forebrain rather than the result of a massive deletion of the ventral CNS tissue as may be the case in the cyc mutant.

The partial fusion of the eyes in slb embryos may be due to the shortened expression domain of shh along the anterior-posterior axis during early somite stages: it has been shown that overexpression of shh induces pax2 and represses pax6 expression which is followed by a hypertrophy of optic stalk-like structures and a reduction of the retinae (Macdonald et al., 1995; Ekker et al., 1995). Anteriorly shortened shh expression, as seen in slb, may therefore lead to an ectopic induction of pax2 and/or ectopic repression of pax6 expression resulting in deformed optic stalks and retinae being placed closer together.

From the tailbud stage until the early somite stages (10-14 hours) the elongation of the body axis in slb mutants is delayed, as indicated by a short and broad notochord and an abnormally shaped prechordal plate. However, at the 10 somite stage (14 hours) the differentiation of the notochord and prechordal plate appears normal. The primary defect in slb may be a reduction in medial-lateral intercalation of cells in the axial mesendoderm. Medial-lateral intercalation is thought to mediate convergence and extension movements at the dorsal side of the gastrula (Keller and Tibbets, 1989; Warga and Kimmel, 1990). Alternatively, migration of the prechordal plate towards the animal pole may be impaired, which in turn may slow mediallateral intercalation and thereby extension of the axis.

How may this gastrulation phenotype be linked to the neuroectodermal phenotype? It has been speculated for a long time that the initially singular eye field within the neural plate is split into two bilateral fields by means of vertical signals originating from the axial head mesendoderm, the prechordal plate, to the neuroectoderm (Adelmann, 1930; Brun, 1981). Thus the partial eye fusion in slb may be causally related to the deformation of the prechordal plate. We propose that the delay in the differentiation of the anterior prechordal plate into the polster (Ballard, 1973; Kimmel et al., 1990) causes the neuroectodermal phenotype in slb. If this is true, the polster should be involved in the separation of the eye fields within the neuroectoderm.

The analysis of the zebrafish cyc mutant has shown that a reduction of the prechordal plate and chordal mesendoderm is associated with both a severe reduction of the ventral CNS and a fusion of the eyes (Thisse et al., 1994; Warga et al., unpublished data). In contrast, in slb mutants the ventral CNS appears relatively unaffected and the mesoendodermal defect is most prominent in the polster region. It is therefore tempting to speculate that posterior prechordal plate and chordal mesendoderm are needed for general induction of ventral CNS whereas the polster is specifically required for positioning of the eye fields within the neural plate. This hypothesis is supported by the initial analysis of the schmalspur (sur) mutant where the posterior prechordal plate but not the polster is reduced, accompanied by a reduction of ventral CNS tissue but not a fusion of the eyes (Brand et al., 1996b; Warga et al., unpublished data).

An alternative scenario which may explain the eye fusion in slb mutants is suggested by a recent publication presenting a fate map of the zebrafish nervous system (Woo and Fraser, 1995). It has been proposed that the initial singular eye field is physically separated into two bilateral fields by diencephalic progenitor cells moving towards to the animal pole during gastrulation. An impaired extension of the neural plate in slb as reflected by the shortened shh and broadened pax2 and fkd3 expression fields may slow down the migration of diencephalic progenitors, leading secondarily to an incomplete separation of the eye field.

The analysis of the kas mutant phenotype shows that the dorsal midline (roof plate) separating the bilateral telencephalic clusters in the anterior forebrain is malformed in the mutant. As a consequence, axons connecting the telencephalic clusters are not bundled into one anterior commissure anymore, instead they cross the midline between the telencephalic clusters at multiple locations. It seems that kas specifically affects the formation of the telencephalic ventricle, which forms the roof plate in the anterior-most part of the neural keel. It will be interesting to see if anterior-lateral parts of the neural plate, which are thought to give rise to anterior-dorsal structures of the neural keel (Kimmel et al., 1994; Papan and Campos-Ortega, 1994) are already affected in this mutant.

We isolated a class of mutants with specific defects in the development of the forebrain. The analysis of these phenotypes provides new insight into the mechanisms by which the forebrain is patterned along its anterior-posterior and dorsalventral axis: specification of the anterior forebrain in mbl, positioning of the eyes in slb and formation of the roof plate in kas. Taking advantage of the zebrafish as an experimental system, further analysis of these mutants may help in understanding the fundamental processes underlying induction and differentiation of the vertebrate forebrain.

We isolated several mutants which show smaller eyes at various stages of embryonic development. A preliminary characterization of the mutant phenotypes did not reveal specific defects in the process of eye differentiation. We divided these mutants into subgroups based on the association of the eye phenotype with other phenotypes like brain necrosis and pigmentation (Table 3A). We did not find mutants with smaller eyes in which the rest of the embryo develops normally. This preliminary synopsis of mutants showing alterations in the size of the eyes will be followed by a more detailed analysis of the various eye phenotypes, including mutants which were isolated on the basis of other phenotypic criteria like retinotectal projection and motility (Trowe et al., 1996; Karlstrom et al., 1996; Granato et al., 1996). An example of a mutant with a strongly reduced eye size associated with a reduction in the total length of the body (microps) at the sixth day of development is shown in Fig. 12B,D.

Mutants that show abnormal lens differentiation were subdivided into seven complementation groups (Table 3B): in bumper (3), rosine (1) and korinthe (1) the lenses begin to degenerate at the fourth day of development, in sunrise (1) the lenses protrude from the eyes and the cornea is closely apposed to the lens and in helderziend (1) the colour of the lenses appears clear instead of greenish at the sixth day of development. The remaining complementation groups, dreumes (1), leprechaun (2), earache (1) and ukkie (1) display changes in the size of the pupil combined with an abnormal differentiation of the ears and fins (Whitfield et al., 1996; van Eeden et al., 1996). We have no information as yet about which processes of lens differentiation are affected in these mutants. An example of a mutant with degenerating lenses at the sixth day of development is shown in Fig. 12A,G.

We thank E. Weinberg for the kind gift of myoD, zotx-2 and zash-1b cDNA, I. Mikkola and S. Krauss for providing pax2/6 antibodies and shh cDNA, and V. Korzh for providing the pan-islet antibody. We are grateful to S. Wilson for help with the initial characterization of the mbl phenotype and many valuable comments on the manuscript. We would also like to thank Robert Geisler, Suresh Jesuthasan, Rolf Karlstrom, Stefan Schulte-Merker and Siegfried Roth for critical reading of the manuscript.