Mutations affecting the development of the embryonic zebrafish brain

Embryonic development of the vertebrate brain involves several steps. First, the neural plate forms on the dorsal side of the embryo and is regionalized along the anterior-posterior and dorsal-ventral axes. The formation of the neural tube from the neural plate and the development of brain ventricles then contribute to the typical morphology of the embryonic brain. Finally, the differentiation of neurons and glia and the establishment of proper synaptic connections lead to functional circuitry in the nervous system.

Embryological studies have established that the formation and regionalization of the central nervous system result from a series of inductive interactions between different regions in the embryo (Doniach, 1992; Jessell and Dodd, 1992; Kintner, 1992; Ruiz i Altaba, 1993). Initially, signals from dorsal mesoderm neuralize and pattern the adjacent ectoderm along the anterior-posterior axis. Signals from the axial mesoderm, including the notochord, and from the nonneural ectoderm specify dorsal-ventral regions in the neuroectoderm. Local interactions within the neural plate lead to further regionalization. For instance, the region of the midbrain-hindbrain boundary seems to serve as a source of activity involved in patterning the adjacent midbrain (Alvarado-Mallart, 1993; Marin and Puelles, 1994; Bally-Cuif and Wassef, 1995; Joyner, 1996). Once established, some of these domains, e.g. the hindbrain rhombomeres, seem to behave as domains of celllineage restriction (Lumsden, 1990; Krumlauf et al., 1993; Keynes and Krumlauf, 1994). The embryological mechanisms involved in the subsequent formation of the neural tube and brain ventricles are less well understood. Both intrinsic and extrinsic forces seem to contribute to neurulation (Schoenwolf and Smith, 1990), whereas cerebrospinal fluid pressure appears to be an important requirement for the enlargement of the embryonic brain (Desmond and Jacobson, 1977).

In recent years several molecules have been implicated in these developmental processes. Signaling molecules like noggin (Harland, 1994), follistatin (Kessler and Melton, 1994) and chordin (Sasai et al., 1994) are candidates for neural inducers. Transcription factors like krox-20 or members of the otx, emx, pax, engrailed and Hox gene families and secreted factors like wnt-1 and sonic hedgehog have been suggested to be involved in the regionalization of the brain (Finkelstein and Boncinelli, 1994; Joyner and Guillemot, 1994; Krumlauf, 1994; Smith, 1994; Stuart et al., 1994; Ingham, 1995). Adhesion molecules like N-cadherin or NCAM have been implicated in neural tube formation (Papalopulu and Kintner, 1994).

Genetic approaches provide a critical test for the postulated role of these molecules in the formation of the brain (Rossant and Hopkins, 1992; Joyner and Guillemot, 1994). Indeed, the analysis of mice with targeted mutations in wnt-1 (McMahon and Bradley, 1990; Thomas and Capecchi, 1990), pax5 (Urbanek et al., 1994), engrailed-1 and engrailed-2 (Joyner et al., 1991; Wurst et al., 1994), krox-20 (Schneider-Maunoury et al., 1993; Swiatek and Gridley, 1993) and Hoxa-1 (Carpenter et al., 1993; Dolle et al., 1993) have demonstrated the essential role of these loci in the regionalization of the anteroposterior axis of the embryonic brain. In contrast, in the case of NCAM (Cremer et al., 1994; Ono et al., 1994) and follistatin (Matzuk et al., 1995), mutant mouse embryos do not show defects that directly support the postulated roles of these factors.

Classical genetic studies have also led to the identification of mutations affecting murine brain development (Lyon and Searle, 1989). Kreisler and swaying (an allele of wnt1) mutant embryos show defects in the regionalization of the brain (Lane, 1967; Frohman et al., 1993; McKay et al., 1994), and the reeler mutation results in the malpositioning of neurons in the brain (Rakic and Caviness, 1995). The recent molecular isolation of the genes affected in these mutants demonstrates the potential of the classical genetic approach for the identification of essential components of brain development (Thomas et al., 1991; Cordes and Barsh, 1994; D’Arcangelo et al., 1995).

The relative simplicity of the embryonic zebrafish (Danio rerio) brain, and the powerful embryological and genetic methodology applicable to its analysis, promise further insights into vertebrate brain morphogenesis (Wilson and Easter, 1992; Kimmel, 1993; Driever et al., 1994). During zebrafish embryogenesis the brain rudiment is already visible at the end of gastrulation (9 hours after fertilization) as a distinctly thickened structure. Within the next 20 hours the primary organization and morphology of the zebrafish brain is established (Kimmel et al., 1995), neural primordia become regionalized (Fjose, 1994; Woo and Fraser, 1995), the neural canal and brain ventricles form (Papan and Campos-Ortega, 1994), and neuronal differentiation and axonogenesis lead to a highly stereotyped axonal scaffold (Chitnis and Kuwada, 1990; Wilson et al., 1990; Ross et al., 1992). By 28 hours postfertilization (hpf) the zebrafish brain has the typical morphology of embryonic vertebrate brains (Fig. 1). Furthermore, the neuroectodermal fate map (Woo and Fraser, 1995), the expression patterns of genes like engrailed (Davis et al., 1991; Hatta et al., 1991a; Ekker et al., 1992; Fjose et al., 1992) and sonic hedgehog (Echelard et al., 1993; Krauss et al., 1993; Riddle et al., 1993; Roelink et al., 1994), and the position of axon tracts (Easter et al., 1994), are very similar to other vertebrates, supporting the notion of evolutionarily conserved mechanisms of morphogenesis in the vertebrate neuroectoderm.

Genetic studies in zebrafish have identified one mutation affecting early brain development, the cyclops mutation (Hatta et al., 1991b). The main features of cyclops mutant embryos are the absence of ventral neuroectodermal structures and fusion of the eyes. Genetic mosaics have indicated that the cyclops gene product is cell-autonomously required for the formation of ventrally located cells in the neuroectoderm, both in the forebrain and more posteriorly in the floor plate (Hatta et al., 1991b, 1994). These midline defects lead indirectly to aberrant axonal patterning in the medial longitudinal fascicles (Bernhardt et al., 1992; Hatta, 1992). The studies on cyclops have highlighted the potential of a combined embryological and genetic approach in zebrafish to study vertebrate brain development.

In a large-scale screen for mutations affecting zebrafish embryogenesis, we have identified more than 60 mutations affecting the morphology of the embryonic zebrafish brain. Mutant phenotypes range from defects in the regionalization of the dorsal-ventral and anterior-posterior axis of the neuroectoderm to abnormalities in the formation of neurons, fasciculation of axons and integrity of brain ventricles.

Mutations were induced by the mutagen N-ethyl-N-nitrosurea (ENU) and recovered in an F ₂ screen as described (Solnica-Krezel et al., 1994; Driever et al., 1996). 2383 embryonic and early larval lethal mutations were identified, 63 of which are described here. All embryos were maintained at 28.5°C and staged according to Kimmel et al. (1995).

Embryos were initially observed under a dissecting microscope. For further analysis and photographic documentation, embryos were dechorionated, anesthetized in 0.02% 3-aminobenzoic acid methyl ester (Sigma), embedded in 2% methylcellulose and photographed either under a dissecting microscope or using differential interference contrast (DIC/Nomarski) optics on a Zeiss Axiophot microscope (Westerfield, 1994).

Whole-mount in situ hybridization was performed with digoxigenin-labeled RNA probes (Oxtoby and Jowett, 1993). Expression patterns were documented after clearing in benzylbenzoate/benzyl alcohol (2:1) and mounting in Permount (Fischer Scientific). The following probes have been used: krox-20 (Oxtoby and Jowett, 1993); pax[zf-b] (Krauss et al., 1991a; Püschel et al., 1992b); pax6 (Krauss et al., 1991b,c; Püschel et al., 1992a); hlx1 (Fjose et al., 1994); wnt1 (Krauss et al., 1992a; Kelly et al., 1993); dlx2 (Akimenko et al., 1994); engrailed-2 (Ekker et al., 1992; Fjose et al., 1992); sonic hedgehog (Krauss et al., 1993); rtk1 (Xu et al., 1994); islet1 (Inoue et al., 1994). Antisense RNA probes were used to analyze the following structures: krox20, rhombomeres 3 and 5; pax[zf-b], midbrain-hindbrain boundary region, optic stalk, otic vesicles, pronephros, a subset of commissural interneurons; pax6, optic vesicles, diencephalon, rhombencephalon (anterior border in middle of rhombomere 1), spinal cord; hlx1 at 28-32 hpf, thin stripes adjacent to interrhombomeric boundaries, tegmentum, tectum, part of dorsal thalamus, part of ventral thalamus; wnt1, dorsal neuroectoderm, anterior portion of midbrain-hindbrain boundary region; dlx2, pharyngeal arch primordia, subregion of telencephalon and diencephalon, pectoral fin bud; engrailed-2, midbrain-hindbrain boundary region, muscle pioneers, jaw muscle precursors; sonic hedgehog, ventral neuroectoderm including hypothalamus and floor plate; rtk1, rhombomeres 1, 3, and 5; islet1, primary neurons including a subset of motorneurons, Rohon-Beard neurons, a subset of interneurons, trigeminal ganglion neurons, acoustic nerve ganglion neurons, epiphysial neurons, polster.

Immunocytochemistry with monoclonal antibodies anti-acetylated α-tubulin (Piperno and Fuller, 1985) and 3A10 (Furley et al., 1990; Hatta, 1992) was performed after fixation for 1-2 hours at room temperature in 4% paraformaldehyde as described (Solnica-Krezel and Driever, 1994).

For methacrylate sections, embryos were fixed in paraformaldehyde (4%, overnight), dehydrated in ethanol and embedded in JB-4 resin (Polyscience Inc.). 5 μm sections were cut on a Leica 2065 microtome.

Embryos were photographed on 160 ASA Ektachrome Tungsten Film. Images from slides were scanned on a Kodak Professional RFS2035 Plus Film Scanner. Figures were assembled using Adobe Photoshop 3.0 software (Adobe Corporation).

Mutations affecting the formation of pharyngeal arches, mother superior^m188, quadro^m271, little richard^m433, mont blanc^m610 (Neuhauss et al., 1996) or the size of the ear, quadro^m271, m471, m574, helter skelter^m504, golas^m618 (Malicki et al., 1996b) were also analyzed for defects in hindbrain patterning by morphological and gene expression analysis using the following markers: hlx1 and dlx2 (for mutants affecting pharyngeal arches and/or ear), rtk1 (for ear mutants), and krox20 and pax[zf-b] (for pharyngeal arch mutants). No clear abnormalities could be identified in the rhombencephalon of mutant embryos.

In order to test allelism of isolated mutations, complementation analysis among members of the phenotypically defined groups of mutations was performed. Complementation between two mutations was tested by crossing identified heterozygous parents of each mutation and screening their offspring for the mutant phenotype. A minimum of 30 embryos per complementation cross was analyzed. All mutations segregate as mendelian recessive loci. Limited complementation has been performed with mutations of similar phenotypes isolated in Tübingen. The following loci have been identified in both screens: oep, cyc, boz, snk, sly, gup, bal. spg^m216 was found to complement the midbrain-hindbrain boundary mutants noi and ace that were isolated in Tübingen.

In a systematic F₂ screen for mutations affecting zebrafish development, the morphology of the brain of living zebrafish embryos was examined at days 1, 2 and 3 of development with dissecting stereo microscopes. At these stages, the size and shape of telencephalon, diencephalon, tectum, tegmentum, midbrain-hindbrain boundary, hindbrain and brain ventricles can be scored (Fig. 1). More subtle features (e.g. sub regions of the hindbrain or particular neurons) are not identifiable at this level of analysis. Among 2383 embryonic and larval lethal mutations identified, we have isolated 63 mutations constituting 24 loci that lead to abnormal brain morphology by 28 hours postfertilization (hpf). An additional 50 mutations lead to CNS degeneration during somitogenesis and are described in an accompanying paper (Abdelilah et al., 1996).

Here we describe the genetic and phenotypic characterization of mutations affecting brain morphogenesis. Mutant embryos were analyzed using dissecting microscopes and compound microscopes with Nomarski interference contrast illumination and with molecular markers. Mutations with similar phenotypes were tested for complementation (see Table 1). The general features of identified brain mutants are described in Table 1. Mutants can be broadly classified into two groups, one affected in regionalization along the anteriorposterior or dorsal-ventral axis of the neuroectoderm, and the other affected in general morphological features of the brain. These phenotypes are described in detail below.

The midbrain-hindbrain boundary (MHB) region consists of the posteriormost midbrain and the anteriormost hindbrain region, also including the cerebellum (Fig. 1). We have identified one locus, spiel ohne grenzen (spg^m216 and spg^m308), that is required for the formation of this region. Morphological inspection (Fig. 2B) and the aberrant expression of pax[zf-b] (a member of the pax-2/5/8 family; Fig. 3J) and engrailed-2 (Fig. 2D) at the MHB indicate that a large portion of the MHB region is deleted in spg mutants at 28 hpf. Phenotypes range from the absence of the ventral portion to a complete deletion of this region. Both the adjacent prospective tectum and posterior hindbrain are present, as judged from both morphological observations, as well as the expression patterns of hlx1 (Fig. 2F) and krox20 (data not shown); however, more subtle defects are visible. Hlx1 expression in the hindbrain of spg mutants appears less distinct than in wild type (Fig. 2F), and the otic vesicles are reduced in some mutant embryos.

Gene expression, fate mapping and transplantation studies indicate that the anlage of the MHB region is established during the end of gastrulation and at the beginning of somitogenesis (Hatta et al., 1991a; Krauss et al., 1991a; Püschel et al., 1992b; Alvarado-Mallart, 1993; Oxtoby and Jowett, 1993; Marin and Puelles, 1994; Woo and Fraser, 1995). To determine when the spg defect becomes apparent, embryos were analyzed for the expression of pax[zf-b] and pax6 (expressed in forebrain and hindbrain) at the beginning, middle and end of somitogenesis (Fig. 3). Already at the beginning of somitogenesis (1-somite stage), pax[zf-b] but not pax6 shows an aberrant expression pattern in spg^m216 mutant embryos (Fig. 3B). Pax[zf-b] expression in the MHB anlage is limited in its anterior-posterior extent and reduced in the medial (future ventral) region of the neural plate. Other domains of pax[zf-b] expression are not affected. At the 10-somite stage, the pax[zf-b] expression domain is severely restricted and absent ventrally (Fig. 3D,F). Concomitantly, the expression domains of pax6 in the forebrain and hindbrain are shifted closer to each other, nearly touching ventrally. Furthermore, the rostral boundary of pax6 expression in the hindbrain is affected, suggesting that the spg phenotype extends into rhombomere 1. Consistent with reduced pax[zf-b] expression, the domains of wnt1 (Fig. 3H) and engrailed-2 (data not shown) are also severely reduced in the MHB region at mid-somitogenesis. By 26 hpf pax[zf-b] expression in the MHB region is lost in most mutant embryos (Fig. 3J). We conclude that spg is required for the development of the anlage of the MHB region as early as at the beginning of somitogenesis.

We have identified mutations in four loci that affect the dorsalventral patterning of the brain. Single alleles of the one-eyedpinhead (oep^m134; Schier et al., unpublished data; Strähle et al., unpublished data), uncle freddy (unf^m768) and bozozok (boz^m168) loci and three alleles of the previously described cyclops locus (cyc^m101, cyc^m122, cyc^m294; Hatta et al., 1991b) have been isolated. At 28 hpf all mutants show variable fusion of the eyes, and ventral neuroectodermal structures like the hypothalamus and floor plate are reduced (Fig. 4). oep^m134 embryos have one, often smaller eye (cyclopia). cyc^m294 and cyc^m122 behave like the previously identified cyc^b16 allele and show partial fusion of the two eyes (synopthalmia), whereas cyc^m101 seems to be a weaker allele, often resulting in eyes that are closer to each other antero-ventrally, but not fused. boz^m168 and unf^m768 show rather variable defects, ranging from synopthalmia to normal eyes. The reduction of ventral neuroectoderm in more severely affected unf^m768 embryos is reflected in the partial loss of sonic hedgehog expression (Fig. 5B). In addition, the axonal scaffold at the midline of unf^m768 mutant embryos is severely disrupted. The axons of the medial longitudinal fascicles are disorganized and fused at the ventral hindbrain midline (Fig. 5D). These defects are very similar to the defects described for cyclops mutant embryos (Hatta, 1992).

Mutants at the three loci sleepy (sly, 10 alleles), bashful (bal, 11 alleles) and grumpy (gup, 5 alleles) have similar notochord (Stemple et al., 1996) and brain phenotypes (Fig. 6). The entire brain is abnormally shaped and folded, and the hindbrain ventricle is enlarged. The eyes of mutant embryos are slightly smaller and tilted ventrally. Despite these gross morphological malformations, the general anterior-posterior and dorsalventral patterning appears normal (Fig. 6M-P). Detailed inspection of hlx1 (Fig. 6Q-T) and sonic hedgehog (Fig. 6I-L) expression at 29 hpf reveals subtle defects that might reflect the abnormal architecture of mutant brains. The sonic hedgehog domain seems more irregular in sly^m86, bal^m190 and gup^m189, and there is an expansion in the region of the midbrain-forebrain boundary. Hlx1 expression in the hindbrain is less distinct. Most strikingly, while the axonal scaffold is present, fewer axons and non fasciculated axons are present in mutants at these loci (Fig. 6U-X).

The ventricles of the brain start to inflate at about 17 hpf (Papan and Campos-Ortega, 1994; Kimmel et al., 1995). Parallel to the onset of circulation at about 24 hpf, ventricles enlarge further and by 28 hpf the inflated ventricles contribute to the characteristic morphology of the embryonic brain (Fig. 1). We have identified a large number of mutations that lead to reduced brain ventricles, ranging from a slight reduction to a complete absence of ventricle inflation. The weakest mutants, with a partial reduction of ventricle size, include heart mutations (Stainier et al., 1996) like silent heart (non-beating heart, Figs 7F, 8B,F), cloche (no endocardium), or bonnie and clyde (cardia bifida). All these mutants lack circulation and show a reduction in ventricle enlargement. Mutations in the loci fullbrain (Figs 7E, 8C,G), glaca (Fig. 7B), white snake (Fig. 7C), landfill (Fig. 7J), logelei (Fig. 7L), turned down (Fig. 7I) and eraserhead (Fig. 7K) result in more severely reduced ventricles (Table 1). Circulation in the head is absent in these mutants. In addition, development appears generally delayed, and tactile sensitivity is impaired. snakehead (snk, 3 alleles) mutants show the most severe form of ventricle phenotype (Figs 7D, 8D,H). The brain appears flat, unstructured and reduced in diameter. No ventricles are present at 30 hpf, only a very thin neural canal (Fig. 8D,H). In contrast to the ventricle mutants described above, the ventricle phenotype of snakehead mutant embryos is not only stronger but also readily identifiable before 24 hpf (data not shown), i.e. before the onset of circulation. Marker analysis indicates that the brain is normally patterned (Fig. 8I-L) and an axonal scaffold develops (data not shown).

zonderzen mutant embryos (Fig. 7G) develop normal circulation at the end of day 1 or on day 2 of development. Despite the initial reduction of ventricles, the brain goes on to develop normally. In all other ventricle mutants circulation is permanently affected, and the brain is smaller and starts to degenerate after two or three days of development (Fig. 8N-P).

We have identified one locus, mind bomb (mib^m132 and mib^m178), that affects neurogenesis in the entire nervous system (Fig. 9). Analysis of islet1 expression, a marker for a subset of primary neurons (Korzh et al., 1993; Inoue et al., 1994) reveals a dramatic increase in the number of cells expressing this gene in mib mutant embryos (Fig. 9D,F). The anterior-posterior and dorsal-ventral position of these dense neuronal clusters is normal, indicating that the supernumerary islet-1 positive cells do not form at ectopic sites but are localized in the regions of normal primary neuron formation (Fig. 9D,F). To determine if supernumerary cells differentiate as primary neurons or have

an aberrant fate, we analyzed the formation of Mauthner neurons in mib mutants (Fig. 9H). In wild-type embryos single Mauthner cells lie within the fourth rhombomere on either side of the midline. In contrast, multiple Mauthner cells differentiate at the same position in mib mutants, suggesting that mib is required to ensure the formation of the correct number of primary neurons during neurogenesis.

Two mutations do not fall in any of the above categories and are described here separately.

turned on^m357 shows a weak reduction of the hindbrain ventricle at 30 hpf and a slight dorsal curvature of the tail (Fig. 10B). By day 6 of development the head is tilted backwards at the level of the hindbrain and the tail is curved onto itself (Fig. 10D). Hlx1 and dlx2 expression at 29hpf and rtk1 expression at 14-somites appear normal (data not shown).

flachland^m517 mutant embryos are characterized by a flat hindbrain region at 30 hpf (Fig. 11B). This phenotype is already manifest at the 9-somite stage by a thin and flat neural rod in the hindbrain region (Fig. 11D). Krox20 and pax[zf-b] are expressed in rhombomeres 3 and 5 and in the MHB region, but the neural rod is severely level of the hindbrain (Fig. 11F).

In a large scale screen for mutations affecting zebrafish embryogenesis, we have identified mutations in 24 loci that affect the morphogenesis of the zebrafish brain. The isolated mutations provide genetic entry points into diverse processes like anterior-posterior and dorsalventral patterning, ventricle formation, neurogenesis and axonogenesis. No mutations were isolated that impair the general induction of neuroectoderm or early neurulation.

There are several explanations for the identification of a relatively small number of mutations affecting brain morphogenesis. It is important to consider firstly, that our screen has not reached saturation. 10 of the 24 loci described here are represented by a single allele. Secondly, members of large gene families often have partially overlapping or redundant expression patterns and functions. Consequently, their exact role can only be uncovered in double or triple mutants (Rudnicki et al.,1993). Thirdly, neural inducers, patterning molecules and components required for neurulation might be maternally provided or haploinsufficient. Finally, our visual screen was designed to find rather drastic defects. Mutants with more subtle phenotypes might be isolated in future screens using molecular markers or behavioral assays. In the following we will discuss the potential role of identified loci with respect to vertebrate brain morphogenesis.

Fate map studies performed in amniotes have shown that the region of the midbrain-hindbrain boundary (MHB), i.e. the posterior region of the mesencephalon and the anteriormost portion of the metencephalon, gives rise to, among other structures, the cerebellum and isthmic nuclei (Martinez and Alvarado-Mallart, 1989; Hallonet et al., 1990). This brain territory is also defined by the local expression of pax[zf-b] at the MHB, wnt1 in the anterior portion of the MHB region, and engrailed in the MHB region and fading towards the rostral mesencephalon (Wilkinson et al., 1987; Davis et al., 1991; Krauss et al., 1991a,b; Püschel et al., 1992b). Whereas it is not known how mid/hindbrain pattern is regulated in the fish, grafting and rotation experiments in the chick indicate that the MHB region is specified by the 10-to 14-somites stage and can induce ectopic engrailed expression and tectum formation in host tissue (Alvarado-Mallart, 1993; Bally-Cuif and Wassef, 1994, 1995; Marin and Puelles, 1994; Joyner, 1996).

We have isolated one locus, spiel ohne grenzen (spg), that is required for the proper formation of the MHB region as early as at the beginning of somitogenesis. spg mutant embryos consistently lack most of the MHB region, but the adjacent prospective tectum and posterior hindbrain region seem intact, although we cannot exclude more subtle defects. The early deficit suggests that spg is required for the initial establishment of the MHB region. Alternatively, spg might be one of the first factors required for the maintenance of this brain region. Absence of the spg gene product might lead to defects in growth, survival or specification of cells in the anlage of the MHB. Since the pax[zf-b]-expressing region seems to be partially deleted in spg, and since there is no obvious cell death detectable in this region at 1- and 10-somite stages (unpublished results), we suggest that a specification and/or proliferation defect is responsible for the observed phenotype in spg mutants. It is noteworthy that the injection of antibodies against the pax[zf-b] protein into zebrafish embryos causes a downregulation of pax[zf-b] transcripts and MHB defects similar to spg (Krauss et al., 1992b). It is conceivable that the spg phenotype might be in part a direct consequence of the reduced expression of pax[zf-b] in the MHB region (also see (Urbanek et al., 1994; Torres et al., 1995).

The spg phenotype is similar to the MHB region deficits observed in mouse embryos mutant for wnt-1 or engrailed-1 (McMahon and Bradley, 1990; Thomas and Capecchi, 1990; Wurst et al., 1994). Similar to spg, en-1 and wnt-1 mutant phenotypes are already apparent during early somitogenesis. Studies of engrailed expression in wnt-1 mutant embryos suggest that the MHB anlage is initially established normally (1-to 4-somites stage) but then progressively deleted (McMahon et al., 1992). Similarly, the expression of engrailed-2 in en-1 mutant animals is already perturbed by the 10-somites stage and is restricted to a dorsal patch in the midhindbrain region by E9.5 (Wurst et al., 1994). The direct comparison of wnt-1, en-1 and spg mutants is complicated by the variability of phenotypes. Strong wnt-1 phenotypes involve the deletion of both the MHB region and tectum), a defect that is stronger than the spg phenotype. Weaker wnt-1 phenotypes, however, are mainly restricted to deficits in the cerebellum, a phenotype more reminiscent of spg. As for spg, it is not clear if wnt-1 and en-1 are required for the survival, growth or specification of the MHB region. The observation that the ectopic expression of wnt1 has a strong mitogenic effect in the CNS might point towards a role of wnt-1 in cell proliferation (Dickinson et al., 1994). Cell lineage and transplantation studies should determine the primary defects in spg mutants.

Embryological studies have established that signals from the notochord induce ventral neuroectodermal structures in the overlying neural plate (Jessell and Dodd, 1992; Ruiz i Altaba and Jessell, 1993). The sonic hedgehog signaling pathway seems to be directly involved in this process, both in the more posterior neuroectoderm and in the forebrain (Smith, 1994; Ericson et al., 1995; Ingham, 1995). We have isolated four loci, including the previously identified cyclops locus (Hatta et al., 1991), that lead to ventral deficits in the neuroectoderm. Ventral deficits encompass the entire neuroectoderm, from a reduction of floor plate cells posteriorly, to deficiencies in the forebrain more anteriorly. These defects suggest that the four identified loci are good candidates for factors involved in the production, transmission or response to the sonic hedgehog signal.

Recent studies have suggested that sonic hedgehog emanating from the diencephalic midline also regulates the formation and partitioning of the optic primordium (Ekker et al., 1995; Macdonald et al., 1995). Loss of this signal might lead to cyclopia, the formation of a single or fused, often median eye. The eye fusions in the mutants described here might therefore result from the deletion of ventral forebrain structures i.e. a source of sonic hedgehog. Our observation that sonic hedgehog is absent in the anteriormost ventral brain region of unf^m768 embryos is consistent with this view. This notion is also supported by the finding that the eye phenotype of cyclops mutant animals can be indirectly rescued by the presence of wild-type cells in the ventral forebrain region of mutant embryos (Hatta et al., 1994).

It is interesting to note that at least three of the identified loci (cyclops, one-eyed-pinhead^m134, bozozok^m168) also show defects in the formation of the prechordal plate (Thisse et al., 1994; Solnica-Krezel et al., 1996; Schier et al., unpublished data; Straehle et al., unpublished data). Studies in Amblystoma have suggested that during gastrulation and neurulation the eye field region is split into two domains, due to the influence of underlying prechordal plate (Adelmann, 1936). Nervous system fate maps indicate that the zebrafish neural retina also derives from a single coherent region that later bifurcates (Woo and Fraser, 1995). It is therefore possible that eye fusions and ventral defects observed in cyclopic mutants result, in part, from prechordal plate defects.

The three loci bashful, sleepy and grumpy affect the formation of the notochord (Stemple et al., 1996) and brain. Despite a severely aberrant brain morphology and enlarged hindbrain ventricle, primary patterning appears normal in these mutants. In contrast, the axonal scaffold is disorganized. We cannot distinguish if these defects are due to abnormal specification, proliferation or survival of neurons. Our observations support the view that correct neuronal patterning is not simply a consequence of normal regional patterning in the brain, but involves additional mechanisms. This conclusion is further supported by the mind bomb mutant phenotype (see below).

Interestingly, all three loci show defects in notochord differentiation. Further analysis will show whether the neural and notochord phenotypes are directly linked. This analysis might help to investigate the later functions of the notochord (and possibly the head mesoderm) in brain morphogenesis, as opposed to its early role in dorsal-ventral patterning of the neuroectoderm.

The ventricles of the brain start to enlarge subsequently to the formation of the neural tube and become fully inflated after the onset of blood circulation. Very little is known about the mechanisms leading to ventricle enlargement. Our analysis of mutant embryos indicates that mutations that permanently block circulation in the brain (e.g. mutants with a defective cardiovascular system) show an incomplete inflation of brain ventricles. Subsequently, the brain does not enlarge and degenerates. Physiological studies suggest a possible mechanism underlying these phenotypes. In the adult animal, a complex balance between the cerebrospinal fluid and blood is responsible for the proper integrity of brain ventricles. The choroid plexus, a secretory epithelium, maintains the chemical stability of the cerebrospinal fluid by bidirectional transport and secretion between blood and central nervous system. Studies on perfused isolated sheep choroid plexus have indicated that a diminished rate of perfusion, with its accompanying diminished capillary pressure, can lead to the reduction of cerebrospinal fluid secretion and pressure (Deane and Segal, 1979). Animal models with lowered systemic arterial pressure support this finding. In these studies, the extent of blood flow through the brain was reduced, resulting in a decreased rate of cerebrospinal fluid production (Carey and Vela, 1974; Weiss and Wertman, 1978). During chick embryogenesis, a positive cerebrospinal fluid pressure seems to be an important requirement for the normal enlargement of the embryonic brain (Desmond and Jacobson, 1977). Based on these observations, it is conceivable that the lack of circulation in cardiovascular mutants in zebrafish results in a reduction or absence of cerebrospinal fluid pressure, and thus impairs the inflation of brain ventricles. The possible connection of blood flow, blood pressure and ventricle inflation also suggests that an increase of blood flow or blood pressure might lead to an enlargement of ventricles. Consistent with this notion, we have found that mutants with reduced body length (no tail: Halpern et al., 1993; floating head: Talbot et al., 1995; dopey or sneezy: Stemple et al., 1996) have enlarged brain ventricles (unpublished observations). We might speculate that these mutants, due to a shorter vascular system, have increased blood flow or blood pressure, resulting in enlarged brain ventricles. At least one locus, snakehead, is required for ventricle formation before the onset of circulation. At the current level of analysis, we cannot exclude the possibility that some of the other identified ventricle mutations may also have a more direct role in brain morphogenesis.

During early neurogenesis, primary neurons arise as clusters in the brain and at distinct medial-lateral positions of the neural plate, forming columns of primary sensory neurons, interneurons and motorneurons. We have identified one locus, mind bomb (mib), that is required for the specification of the correct number of primary neurons. The supernumerary primary neurons in mib embryos still arise at their correct position, but form denser and broader domains. The observed abnormalities in the mib nervous system are strikingly similar to the phenotype of neurogenic mutants in Drosophila and to defects induced in Xenopus embryos by the injection of an antimorphic Delta construct (Campos-Ortega, 1993; Chitnis et al., 1995). In both systems, interference with the Notch-Delta system for lateral specification leads to the overproduction of primary neurons. It is conceivable that the mib gene product is a component of a lateral specification system, where a cell committed to a primary neural fate forces its neighbors to remain uncommitted or to follow a different fate. The observation that supernumerary cells form only locally suggests further that these regions constitute equivalence groups or proneural fields. The pleiotropic phenotype of mib might indicate that similar mechanisms are used in many regions in the embryo (Greenwald and Rubin, 1992; Artavanis-Tsakonas et al., 1995; Conlon et al., 1995). In this model, mib would play a major role in the process of lateral specification and cell com-mitment. Alternatively, mib might be involved in the control of cell proliferation to prevent the overproliferation of primary neuron precursors. Blocking cell proliferation in mib mutants could provide a test of the latter scenario.

The identification of developmental genes via the systematic isolation of mutants defective in morphogenetic processes has led to fundamental insights into the mechanisms that govern the development of Drosophila and Caenorhabditis elegans (Nüsslein-Volhard and Wieschaus, 1980; Horvitz and Sternberg, 1991). These mutants have not only led to the identification of key molecules, but in some cases have revealed the mechanistic logic of morphogenetic pathways. The relatively small number of isolated mutations and the diversity ofassociated phenotypes do not yet allow for the construction of a complex genetic pathway controlling zebrafish brain morphogenesis. Rather, we have isolated essential entry points into different aspects of brain development. First, the identified loci provide a basis for the molecular isolation of important com-ponents in brain morphogenesis. Second, detailed embryological studies in wild-type and mutant embryos promise to unravel important principles of brain development. Finally, the analysis of double mutant phenotypes, and the possibility of genetic modifier screens, will lead to the identification of further key components. Thus the loci described here provide the genetic framework for the further study of brain morphogenesis in zebrafish.

We thank Colleen Boggs, Jane Belak, Lisa Vogelsang, Jeanine Downing, Heather Goldsboro, Kristen Diffenbach, Lisa Anderson, Ioannis Batjakas and Scott Lee for help during the various stages of the screen. The following colleagues kindly provided us with cDNA clones: T. Jowett, S. Krauss, P. Ingham, A. Fjose, D. Duboule, M.-A. Akimenko, M. Ekker, M. Westerfield, Q. Xu, N. Holder, G. Kelly, R. Moon and H. Okamoto. We are grateful to Kathleen Cantwell, Alex Joyner, Will Talbot and Steve Wilson for critical reading of the manuscript. This work was supported in part by NIH RO1-HD29761 and a sponsored research agreement with Bristol Myers-Squibb (to W. D.). Further support in the form of fellowships came from HFSP and the Fullbright Program (to Z. R.), the Helen Hay Whitney Foundation (to D. L. S. and D. Y. S.), the Medical Research Council of Canada (to M. H.), the Damon Runyon-Walter Winchell Cancer Research Fund (to J. M.) and EMBO and Swiss National Fond (to A. F. S.).

logelei does not complement otter (Jiang et al., 1996). The name of the locus will be otter.