Mutations affecting neurogenesis and brain morphology in the zebrafish, Danio rerio

Neural development in vertebrates starts during gastrulation with the induction and patterning of the neural plate. The morphogenic transformation of the neural plate into the neural tube and the differentiation of neuronal cell types in different waves of neurogenesis follow. Finally, precise connection between neurons in the various brain regions are established. The complexity of this process is overwhelming, suggesting interactive mechanisms between cells and their neighbors or cells and their environment. The genetic control of many of these processes is not well understood in vertebrate organisms. In Drosophila, local cell-cell interactions play an important role in coordinating developmental decisions required for the generation of different cell types (for reviews, Campos-Ortega, 1993; Ghysen et al., 1993). Neurogenic genes, such as Notch, Delta and others are involved in this process in many different tissues (Hartenstein et al., 1992). The products of these genes interact through lateral inhibition, committing a cell to a neural fate, and neighboring cells become epidermis (Seydoux and Greenwald, 1989; Heitzler and Simpson, 1991). Loss-of-function mutations cause overproduction of neurons – the neurogenic phenotype (Lehmann et al., 1983) – due to the mis-specification between the neural and epidermal pathways (Heitzler and Simpson, 1991). Gain-of-function mutations lead to overproduction of epidermis at the expense of neural structures (Palka et al., 1990) and can be mimicked by extracellular truncations of Notch (Struhl et al., 1993).

Homologs of these genes have been cloned in vertebrates, such as Notch (Coffman et al., 1990; Weinmaster et al., 1991; Ellisen et al., 1991; Reaume et al., 1992,del Amo et al., 1992; Bierkamp and Campos-Ortega, 1993; Lardelli and Lendahl, 1993) Delta (Chitnis et al., 1995; Henrique et al., 1995; Bettenhausen et al., 1995) and a rat ligand for Notch (Lindsell et al., 1995). Misexpression of constitutively active forms of Notch in cultured mammalian cells suggests that it is involved in the signaling pathways for cell fate decisions (Kopan et al., 1994; Nye et al., 1994). However, in a loss-of-function mutation in the Notch1 locus no effect on neurogenesis could be detected (Swiatek et al., 1994; Conlon et al., 1995). In Xenopus, the ectopic expression of X-Delta-1 inhibits the genesis of primary neurons, and the interference of endogenous X-Delta-1 activity causes overproduction of primary neurons (Chitnis et al., 1995). Although the progress in understanding neurogenesis in vertebrates is very encouraging, it is still not clear how and to what extent these genes are involved in cell-cell interaction and cell fate specification.

In addition to cellular interactions by lateral inhibition, the development of neurons further depends on interactions with molecules in their environment. Studies of individual cell types have revealed a tremendous diversity in the molecules that regulate the development of cells in the nervous system. These include cell adhesion molecules and the extracellular matrix (for reviews, Hynes and Lander, 1992; Müller and Kypta, 1995; Venstrom and Reichardt, 1993). These molecules have been shown to influence a variety of processes such as neuronal migration, survival, axonal growth and guidance, synapse formation and glial differentiation.

The morphogenesis of the neural tube in fish differs slightly from that of other vertebrates. There are two forms of neurulation in vertebrates. Primary neurulation, which occurs mainly in amphibia, birds and mammals (Karfunkel, 1974), proceeds by invagination of the neural plate to form a neural groove, which subsequently closes to form a neural tube. In contrast, secondary neurulation, which occurs in teleosts (Schmitz et al., 1993) and in the caudal region of the trunk in many other vertebrates (Hughes and Freeman, 1974; Nakao and Ishizawa, 1984; Schoenwolf and DeLongo, 1980), proceeds by the formation of a dense nerve rod, which then forms a lumen by cavitation. Although the morphogenetic movements accompanying formation of the neural keel in the zebrafish are different from that in other vertebrates, the zebrafish utilizes a mechanism similar to primary neurulation by infolding of the neural plate (Papan and Campos-Ortega, 1994). Mitoses occur preferentially on the ventricular side (Schmitz et al., 1993), and the siblings frequently form bilateral cell clusters across the midline (Kimmel et al., 1994; Papan and Campos-Ortega, 1994). Following its formation, the anterior end of the embryonic neural tube in every vertebrate, from fish to human, partitions into three primary brain vesicles, forebrain, midbrain and hindbrain. Further differentiation involves thickening of the wall in some locations and evagination in others. By 24 hours, several prominently sculptured regions can be seen in the zebrafish brain, including telencephalon, diencephalon, dorsal midbrain (tectum), ventral midbrain (tegmentum), cerebellum and seven hindbrain rhombomeres (Kimmel, 1993). The embryonic neurocoel is larger in the brain than in the spinal cord. Within the brain vesicles it forms expansions called ventricles.

In a large scale screen for mutants affecting the development of the zebrafish embryo, we have isolated about 100 mutants with visible brain phenotypes. In complementation tests between mutants with similar phenotypes, we have identified 16 genes that are involved in general brain development and morphogenesis (Table 1). In this paper we describe the phenotypes of 13 genes involved in different aspects of neural development, such as cell fate decision, cell-cell adhesion, and formation of the brain ventricles. parachute mutant embryos have loose cells in their brain ventricles and a displacement of some neuronal cell types in the hindbrain. Mutations in a group of six genes, snakehead, natter, otter, fullbrain, viper and white snake, cause early failure of ventricle formation of the brain, severe retardation and early death. Mutants in five genes, atlantis, big head, wicked brain, scabland and eisspalte display extra or abnormal folds in the brain. We have concentrated our analysis on white tail, the only gene identified in our screen that displays a neurogenic phenotype.

Fishes were maintained and mutagenized as described by Brand et al. (1995); Haffter et al. (1996) and Mullins et al. (1994). Embryos were collected by natural spawning, raised at 28.5°C, and staged as described by Westerfield (1994) and Kimmel et al. (1995). Sometimes the embryos were treated with 0.2 mM 1-phenyl-2-thiourea (PTU), which inhibits tyrosinase (Blagoeva and Stoichev, 1979), to maintain their optical transparency (Westerfield, 1994).

Whole-mount in situ hybridization and antibody stainings were performed as described previously (Hammerschmidt and Nüsslein-Volhard, 1993). For antibody stainings, anti-Islet-1 (1:500; Korzh etal., 1993), anti-GABA (1:500; Sigma), anti-myf-5 (1:200; Santa Cruz), anti-myogenin (1:200; Santa Cruz) and anti-GFAP (1:1000; DAKO) polyclonal antibodies and zrf-1, zn-5 (1:1000; Trevarrow et al., 1990), 3A10 (1:500; Furley et al., 1990), MF20 (1:100; Bader et al., 1982), 4D9 (1:500; Patel et al., 1989), anti-acetylated tubulin (1:1000; Sigma; Piperno and Fuller, 1985) monoclonal antibodies were used. For in situ hybridization, digoxigenin-labeled RNA probes were synthesized from krox-20 (Oxtoby and Jowett, 1993), Zash-1a, Zash-1b (Allende and Weinberg, 1994), pax[zf-b] (Krauss et al., 1991), TRP-2 (Kelsh et al., personal communication) and Notch (Bierkamp and Campos-Ortega, 1993) cDNA clones. Photographs were taken on an Axiophot photomicroscope (Zeiss).

In this paper we describe mutants in 13 genes that are involved in early neurogenesis, development of the hindbrain and formation of the brain ventricles (Table 1). We will first describe the phenotype of white tail in some detail, followed by a brief description of parachute, and an overview of the phenotypes of two groups of genes that affect early ventricle formation, and brain folding, respectively.

The gene white tail is represented by a single recessive allele. No other mutants with a similar phenotype have been isolated in our screen. During segmentation stages, homozygous wit embryos show an irregularly formed neural keel, particularly in the area of the hindbrain (Fig. 1A,B). Additionally, the somite boundaries in the posterior segments are indistinct (Fig. 1C,D) and the notochord is often slightly undulated. In the pharyngula period (24-48 hours), no rhombomeric boundaries can be seen (Fig. 1E,F). Melanophore pigmentation is absent in the posterior trunk and tail (Fig. 1G,H), and no migratory neural crest cells can be seen in the lateral pathway of the posterior trunk (Raible et al., 1992; data not shown).

Despite the complex phenotype, the development of mutant embryos does not seem to be generally retarded. The general structure and the shape of the posterior trunk and tail look rather unaffected (Fig. 1C,D); muscle cell differentiation seems to be normal (data not shown); the major subdivisions of the brain are present, and the otic vesicle length (Kimmel et al., 1995) seems to be normal (data not shown).

Due to the early CNS phenotype, we assayed the expression of a number of neuronal markers in wit homozygous mutant embryos. The monoclonal antibody (mAb) 3A10 recognizes a neurofilament-specific epitope (Furley et al., 1990) that is prominent in the Mauthner cells (Hatta, 1992 and Fig. 2A), a set of primary hindbrain interneurons which have been studied in many vertebrates (for review, Faber and Korn, 1978). In wit mutant embryos, 3-4 Mauthner cells, instead of a single cell, can be detected per hemi-rhombomere 4. Additionally, other identified hindbrain interneurons labeled by 3A10 are increased in number as well (Fig. 2B). The medial longitudinal fascicle is wider than that in wild type, suggesting an increase in axons and therefore in the neurons that give rise to these axons (compare Fig. 2A and 2B).

Does this increase in neuronal number apply only for specific neurons such as the Mauthner cell or generally for many kinds of neurons? To address this question, we used other specific and well characterized neuronal markers. The Islet-1 antibody detects LIM homeobox-containing proteins that are present in early differentiating neurons (Korzh et al., 1993). As early as in the 8-somite stage we find a striking increase in the number of Islet-1-positive cells in wit embryos (Fig. 2C-F). At this stage in the zebrafish embryo the trigeminal ganglia are labeled (Fig. 2C). In addition, there are two laterally distributed lines of Rohon-Beard sensory cells, and two medially distributed lines of primary motoneurons (Fig. 2E). In wit embryos, the trigeminal ganglia get larger (Fig. 2D), smaller motoneurons along the anteroposterior axis increase in number, and the bigger Rohon-Beard cells get more numerous not only along the anteroposterior axis but also along the mediolateral axis (Fig. 2F). By the 20-somite stage, not only CNS cells, such as the Rohon-Beard sensory cells, motoneurons and epiphyseal neurons (Fig. 2G,H), but also Islet-1-positive cells in the PNS, such as trigeminal ganglia (as mentioned before), and the presumptive anterior and posterior lateral line ganglia, are increased in number (Fig. 2G,H). Acetylated tubulin mAb labeling reveals similar results: axons of the anterior commissure (AC) and the postoptic commissure (POC) form thicker bundles and neuronal somata in the spinal cord (such as Rohon-Beard cells) are increased in number (data not shown).

In addition to the identified early differentiating neurons mentioned above, several other interneurons increase in wit as well. The antibody against the neurotransmitter gamma aminobutyric acid (GABA) labels several different types of interneurons in the brain and spinal cord (Bernhardt et al., 1992; Macdonald et al., 1994). The increase of GABA labeling can be shown at the 20-somite stage (data not shown). In the pharyngula period, not only is the GABA labeling in the hindbrain increased but also the segmented organization of the hindbrain appears disrupted (data not shown). pax-b, also called pax2 (Krauss et al., 1991; Püschel et al., 1992; Mikkola et al., 1992) expression in the hindbrain and spinal cord appears increased in wit embryos while the expression domain at the midbrain hindbrain boundary appears normal in size (compare Fig. 2I,J).

Zash-1a and Zash-1b are two achaete-scute homologs, which encode transcription factors containing basic helix-loop-helix domains (Allende and Weinberg, 1994). At about the 15-somite stage, the expression of Zash-1a and Zash-1b mRNA in wit is increased in many regions compared to normal embryos (Fig. 3A-D). In contrast, by the pharyngula period, their expression is decreased, especially the segmented expression pattern in the hindbrain (compare Fig. 3E,F and 3G,H and see discussion).

The increase in cell number is not restricted to neuroectodermal regions, but is also observed in at least two mesodermal tissues. The polster, which is a part of the prechordal mesoderm (Kimmel et al., 1990; Hatta and Kimmel, 1993) is larger in wit embryos as seen by Islet-1 labeling (Fig. 2D; Inoue et al., 1994). pax-b shows an increased area of the expression domain in the pronephric duct (Fig. 2I,J).

In summary, by using several markers, we can show that many early differentiating neurons, including sensory neurons, interneurons and primary motoneurons, as well as some meso-dermal cell types, are strikingly increased in number.

The preceding data suggest that wit mutants develop an increased number of early differentiating neurons. How do these additional cells arise? One possibility is that they increase at the expense of other cell types within the neuroepithelium. We investigated other neural cell types using markers for late differentiating neurons, glia and neural crest derivatives.

During the later pharyngula period, zn-5 mAb labeling can be seen in commissural axons and dorsal interneurons in the hindbrain (Trevarrow et al., 1990, Fig. 4A), as well as in ventral axons of the secondary motoneurons in the spinal cord (Pike et al., 1992). These late differentiating neurons do not appear before 24 hours postfertilization (hpf) and their somata are smaller than those of primary neurons (Kimmel and Westerfield, 1990). In wit homozygous embryos, zn-5 mAb labeling is reduced in the hindbrain at 36 hpf (Fig. 4A,B). By 48 hpf, there is no detectable zn-5-positive labeling of secondary motoneuron axons in the spinal cord in wit homozygous embryos (data not shown), which suggests that the secondary motoneurons are strongly reduced in number compared to normal embryos. In contrast, the zn-5 labeling of the optic nerve and ganglion cell layer in the head is increased (data not shown), which shows that the reduction in zn-5 labeling is not due to general retardation.

We also examined the expression of four markers that recognize radial glial cells in the zebrafish (Trevarrow et al., 1990; Marcus and Easter, 1995). The zrf-1 labeling is not only highly reduced in the retina (Fig. 4C,D) and the hindbrain (Fig. 4E,F) but also in the diencephalon, tectum and spinal cord (data not shown). This reduction can be detected as early as the 17-somite stage. From 24 hpf to 48 hpf, the radial glial labelings, prominent in the hindbrain in wild type, cannot be detected in wit homozygous embryos (Fig. 4E,F and data not shown). It appears that not all glial cell types are affected, as the expression of glial fibrillary acidic protein (GFAP), which is expressed in several glial cell types (Eng and Shiurba, 1988), looks normal by 24 hpf (data not shown). In contrast, the other three zrf antibodies did not show obvious reduction in labeling like zrf-1; however, the segmented labeling in the hindbrain of wild-type embryos is not seen in wit embryos (data not shown). zrf-2, zrf-3 and zrf-4, in addition to radial glial cells, may also label some other neuronal cell types (Westerfield, 1994 and data not shown).

Neural crest cells give rise to a variety of different cell types, including neurons, melanophores, endocrine cells and various connective tissue types (for review, Le Douarin et al., 1993). As described above, in wit mutants neither pigmented melanophores, nor neural crest cells migrating on the lateral migration pathway are visible in the posterior trunk and tail. The zebrafish TRP-2 homolog labels melanoblasts prior to melanin synthesis (R. Kelsh, unpublished observation). While TRP-2 labeling in the head and anterior trunk of wit mutants is indistinguishable from that of wild-type siblings, in the posterior trunk/tail there are essentially no TRP-2 labeled cells in mutant embryos (compare Fig. 4G,H). Together, these results suggest that at least one neural crest derivative (melanophore) is not formed in the posterior trunk and tail.

In live wit homozygous embryos, somite boundaries in the posterior trunk and tail are less distinct than in normal embryos (Fig. 1D). In the zebrafish, somites appear in an anterior to posterior progression at a constant rate of one somite every half hour (Hanneman and Westerfield, 1989). Although homozygous wit embryos can hardly be distinguished from wild type before the 8-somite stage, we can show that there is a delay in somitogenesis before this stage. We counted the somite number, kept individual embryos separate and determined the phenotype after 24 hpf. The delay is detected from the onset of somitogenesis (data not shown). At the 2-somite stage, the recognizable somite number in wit is about 90% of that in wild type demonstrated by statistics from a reasonable number of embryos, but it decreases further to around 60% by the 17-somite stage (compare Fig. 1C,D).

The delay in somite formation also can be shown by molecular markers. By the 11-somite stage, the difference of labeled and segmented somite number between wild types and wit homozygous embryos is about 3 somites, as assayed by expression of Myogenin (Table 2), whereas the difference is about 5 somites in Myf-5 labelings (Table 2 and Fig. 5A,B). By the 18-somite stage, the wild-type embryos have about 5 and 7 more labeled and segmented somites than wit embryos as determined by Myogenin and Myf-5 labeling, respectively (Table 2). These results support the observations in live embryos. The Myf-5 pattern reveals that, in homozygous wit embryos, in addition to the delay in somitogenesis, the somites are less sharply segmented and the caudal somite remains are unsegmented (Fig. 5A,B). MF20, which is a marker for myosin heavy chain (Bader et al., 1982 and Kopan et al., 1994), is expressed at normal levels except for the unsegmentation of somites in the posterior (Fig. 5C,D), and somitic myogenesis appears to be normal. Furthermore, the organization of myofibrils is normal in wit homozygous embryos (data not shown).

In summary, we have shown that in wit mutant embryos the number of early differentiating neurons is increased. In contrast, late differentiating neurons, some glial cells and neural crest derivatives are decreased. We find an additional effect on somite formation. A similar spectrum of phenotypes, including the effect on the mesoderm, has been observed for neurogenic genes in Drosophila, such as Notch and Delta. Therefore we examined the expression of the zebrafish Notch homolog (Bierkamp and Campos-Ortega, 1993). While in the normal embryo, during somitogenesis, high levels of Notch transcripts can be detected in the presomitic mesoderm and developing somitomeres (Fig. 5E; Bierkamp and Campos-Ortega, 1993), this expression is reduced in wit homozygous embryos (Fig. 5F).

In our screen, we found 5 alleles of different strengths of parachute (Table 1). Mutant embryos can be identified around the 10-somite stage by a slightly uneven neural keel or a small bulge in the hindbrain area (Fig.6A,B). Later, in the pharyngula period, the brain ventricles are filled with rounded-up cells and the hindbrain region is thinned while displaying an uneven surface (Fig. 6D). In the hindbrain ventricles of the weak alleles, some cell clumps are located which have a filopodia-like structure (Fig. 6E). During the hatching period abnormal growth in the hindbrain is often seen, which is associated with a bump that is devoid of pigmentation (data not shown).

Hindbrain segments consist of one segment center region and one segment border region, which contain reticulospinal neurons and commissural neurons, respectively. Centers are divided from the borders by curtain-like rows of glial fibers (Trevarrow et al., 1990). In zrf-1 mAb (Trevarrow et al., 1990) labeled normal embryos, these segmented structures are visible (Fig. 7A), while they appear disorganized in pac homozygous embryos (Fig. 7B). The disorganization also holds true for neurons as indicated by labeling with antibodies against acetylated tubulin (Fig. 7C,D).

Using the 3A10 mAb, we found that the position of the Mauthner cells is altered in pac mutant embryos (Fig. 8B-D, for detail see legend). 3A10 mAb not only labels Mauthner cells in rhombomere 4 but also some other reticulospinal interneurons in adjacent rhombomeres (Fig. 8A and Hatta, 1992). Somata positions of all 3A10-positive neurons excluding the Mauthner cells are relatively unaffected, and the pathfinding of their axons is normal or slightly affected (Fig. 8). Although a frequent displacement of Mauthner cells was observed, a less stained and more medial neuron in r4 is still present (compare Fig. 8A-C), suggesting the Mauthner cell moved from its normal position in pac embryos. At least in weak alleles, the pac phenotype gets milder with age, and some individuals may survive to adulthood.

In our screen we isolated a number of mutants displaying very early effects on brain morphology. A group of genes, the ‘snakehead’ group, snakehead (Fig. 9B), natter, otter, fullbrain, viper and white snake have very similar phenotypes. In the pharyngula period, the brain region is very much reduced in size, and the brain ventricles appear collapsed. Mutant embryos also have heart failures, enlarged heart cavity and display signs of general retardation. They die in the eary larval period. Although the general phenotype of genes in this group is very similar, there are locus-specific differences, listed in Table 1. We have examined the expression patterns of krox-20 (Oxtoby and Jowett, 1993) and Engrailed (Patel et al., 1989) in snk, nat and ott, but no detectable abnormalities can be found (data not shown).

Mutations in a group of three genes, atlantis, big head and wicked brain, all cause abnormalities in brain size or folding during the pharyngula period. In atlantis additional folds are often seen (Fig. 9D), while in wicked brain the folding is irregular, and big head embryos have larger ventricles. Many mutant embryos develop a swimbladder and may grow up to normal adults which do not display an obvious phenotype. Each gene has its own characteristic features as described in Table 1. The expression patterns of krox-20 (Oxtoby and Jowett, 1993) and Engrailed (Patel et al., 1989) is normal in atl (data not shown).

Mutants in two further genes (scabland, eisspalte) display disorganized brains during the pharyngula period. Both have additional phenotypes and show retarded development (Table 1). General brain morphology is also abnormal in three genes, bashful, grumpy, and sleepy, affecting notochord differentiation (Odenthal et al., 1996). Mutations of all three genes also cause abnormal axonal pathfinding (Karlstrom et al., 1996). A number of other genes affecting specific regions of the brain, which have been identified in our screen, are described in accompanying papers (Brand et al., 1996; Heisenberg et al., 1996).

In our screen, we isolated about 100 mutants that show a major brain phenotype. Most of the mutants were found at the second day of development when the folding of the brain vesicles is still ongoing. Several of the mutants mentioned in this paper display defects which can already be detected at much earlier stages of development. The snakehead group phenotype, as well as sbd and ele show general signs of retardation. Because of their very early effects on the brain morphology, we kept the snakehead group as well as sbd and ele, in contrast to those displaying retardation signs later, because we suspected specific patterning defects to be the primary cause of the phenotypes. This expectation was not substantiated, by analysis with some molecular markers. Further analysis will reveal the cause of the ventricle collapse in mutant embryos of snakehead group. The phenotypes of the atlantis group displaying additional and/or abnormal brain folds, which are not associated with retardation or degeneration, are apparently also not caused by misspecification of specific regions of the brain. Surprisingly, in the case of atl and bid, mutant individuals can recover and grow up to apparently normal and fertile adults. It is tempting to speculate that in these mutants functions are impaired that can be taken over by other genes if necessary.

The gene parachute, for which five alleles have been isolated, has a unique phenotype with cells apparently detaching from the surface of the outfolding neural tube. This is reminiscent of a small group of genes that were classified as causing skin defects (van Eeden et al., 1996a), as well as the dominant phenotype of epiboly mutants (Kane et al., 1996). In pac mutant embryos, several neurons were found to be displaced in an irregular fashion. Our tentative interpretation is that pac is required in processes of cell contact and adhesion within the central nervous system.

In this paper, we describe the mutant phenotype of one gene, white tail, in more detail. For wit, only one allele was isolated in the screen, displaying a unique phenotype reminiscent of the neurogenic phenotype of several Drosophila mutants. As only one wit allele is known so far, it is not clear whether all aspects of the phenotype are caused by a single mutation, or by a deletion eliminating several genes. In addition, the strength of the allele is not known. Therefore the following interpretation of the phenotype should be regarded as tentative. Aside from the putative Notch/Delta signaling, the asymmetric protein localization mechanism is also possible (for review, Jan and Jan, 1994; Doe and Spana, 1995).

The wit gene is required very early in the embryo. At the two somite stage, mutant embryos can already be recognized by a delay in somite formation. Shortly thereafter, the most prominent phenotypic feature of the wit mutation is the striking increase of several neuronal cell types as revealed by molecular markers. This neuronal hyperplasia led us to investigate whether wit might be a gene with functions similar to the neurogenic genes in Drosophila. In these Drosophila mutants, neuronal hyperplasia is at the expense of other ectodermal cell types (Campos-Ortega, 1993; Hartenstein et al., 1992).

In neural development of the zebrafish, both hindbrain reticulospinal interneurons and spinal motoneurons appear during two distinct waves of neurogenesis (Kimmel and Westerfield, 1990). Early differentiating neurons are increased, while the late differentiating neurons are decreased in wit homozygous embryos. This could be due to the lack of late differentiating neural progenitors, which have been misspecified to early differentiating neural progenitors during the early wave of neurogenesis and/or can not acquire neuronal identity during the late wave of neurogenesis. It has been proposed that Zash-1a and Zash-1b have a proneural function and are also involved in patterning of the CNS (Allende and Weinberg, 1994). We find that an increase of Zash expression around 15-somite stage (Fig. 3B,D) correlates with the additional early differentiating neurons found in wit. By 24 hpf, in contrast, we observe a strong decrease of Zash (Fig. 3F,H). One interpretation of this finding is that the cells that normally express Zash later are late differentiating neurons, and that the cells in wit embryos are misspecified to give rise to early differentiating neurons, expressing Zash during the early wave of neurogenesis.

Glial cells and neuronal cells have common progenitor cells in insects (Udolph et al., 1993), the early neural plate (Hartenstein, 1989; Soula et al., 1993), neural crest (Stemple and Anderson, 1992) and in some parts of CNS, such as the retina (Turner and Cepko, 1987), tectum (Galileo et al., 1990) and spinal cord (Leber et al., 1990). Using zrf mAbs and an antibody against GFAP, we show that a subpopulation of glial cells is reduced (Fig. 4D,4F and data not shown). These results suggest that the increase of neuronal cells is at the expense of glial cells, although only subsets of glial cells are reduced. Further experiments are required to investigate at which level of glial cell specification the wit gene is required.

The wit gene gets its name from a conspicuous lack of melanophores in the posterior trunk and tail. During neurulation in vertebrate embryos, the ectoderm becomes subdivided into three embryonic tissues, the neural tube, neural crest cells, and the epidermis. It has been demonstrated that if epidermis and neural plate are put adjacent to each other, neural crest cells are generated at the junction of the two tissues in amphibians (Moury and Jacobson, 1989) and in the chick (Selleck and Bronner-Fraser, 1995). However, it has been shown that in Drosophila neurogenic mutants, the neuronal hyperplasia is at the expense of epidermal cells (Seydoux and Greenwald, 1989; Heitzler and Simpson, 1991). Further analysis of epidermal and other neural crest derivatives is required in wit. As other neural crest cells, which normally migrate along the lateral pathway, could not be observed in the posterior of wit embryos, we propose that the neural hyperplasia observed in wit embryos is in part at the expense of neural crest derivatives. The restriction of the pigment phenotype to the posterior of the embryo may be due to the incomplete lack of function of the only wit allele.

All the cell types we found reduced are neural cells born late in neurogenesis. We therefore propose that the increase in early differentiating neurons is at least in part at the expense of late differentiating neurons and/or cells with later birthdays in the nervous system. We have not investigated the fate of other ectodermal cell types yet nor do we know whether the neurogenic region in the wit embryo is generally larger than in wild type. From the normal size of several brain regions in wit embryos, (such as the midbrain-hindbrain boundary, the otic placode and the trigeminal ganglion, see Fig. 2), it rather appears that the density of early differentiating neurons is increased within a neurogenic region which is defined independently from the wit gene.

We find that the expression pattern of several myogenic markers is delayed and that the somite boundaries are less distinct in wit. Myogenesis and muscle formation, however, are normal. This suggests that somitic myogenesis is not directly linked to somitogenesis (Borman and Yorde, 1994a,b). We have isolated a number of mutations affecting somitogenesis (van Eeden et al., 1996b), and in all of them, myogenesis is also unaffected. As only one allele of wit has been isolated so far, it is possible that the somite phenotype is not related to the neurogenic phenotype of the wit gene. However, it has been demonstrated that myogenesis is normal but somitogenesis is affected in the Notch1 knock-out mutant in mice (Conlon et al., 1995). Consistent with this, we have found that expression of zebrafish Notch (Bierkamp and Campos-Ortega, 1993) is affected in homozygous embryos (compare Fig. 5E,F and data not shown). We are currently determining whether Notch and wit are genetically linked.

In many instances, the neurogenic genes are necessary for correct implementation of cell fate decisions (for review, Artavanis-Tsakonas et al., 1995). In addition, there are a number of instances where neurogenic genes are necessary for many transitions to, and maintenance of, the epithelial state (for review, Hartenstein et al., 1992). The two apparently contrasting functions may be two sides of the same coin. During neurogenesis in vertebrates, probably through lateral inhibition as in Drosophila (Heitzler and Simpson, 1991), neural progenitors can delaminate or partially delaminate from the ectodermal epithelium, whereas during somitogenesis, loose mesenchymal cells of presomitic mesoderm reorganize into a tightly packed epithelial ball. Both regulation of neurogenesis and somite formation may rely on acquiring/maintaining epithelial stability. Recent data have indicated the important correlation between mesenchymal/epithelial development and neurogenic genes (Tepass and Hartenstein, 1995; Conlon et al., 1995; Mitsiadis et al., 1995).

The function of neurogenic genes in vertebrates has been analyzed in several cases. Ecotopic expression of X-Delta-1 inhibits production of primary neurons, and interference with endogenous X-Delta-1 activity results in an increase in primary neurons (Chitnis et al., 1995), which are reminiscent of the gain-of-function and loss-of-function mutants in neurogenic genes, respectively (for review, Campos-Ortega, 1993; Fortini and Artavanis-Tsakonas, 1993). Although the disruption of the Notch1 did not lead to the expected neural phenotype, the mutant mice showed a problem in coordinating somitogenesis (Conlon et al., 1995). The mouse mutant corresponding to Suppressor of Hairless [Su(H)], which is involved in Notch signaling (Fortini and Artavanis-Tsakonas, 1994), also displayed somite defects (Oka et al., 1995). The X-Delta-1’s role in somitogenesis is still unknown, but due to its expression in presomitic mesoderm (Henrique et al., 1995) and coexpression with X-Notch-1 in similar domains (Chitnis et al., 1995), it is likely that X-Delta-1 is also involved in somitogenesis, playing a role similar to Notch1. The failure to observe a neural phenotype in Notch1 mice could be due to the early embryonic lethality, which has been demonstrated in the study of N-myc (Moens et al., 1993). Although we only found one allele for wit in this screen, it is possible that wit^ta52b is a leaky allele. We are looking for more alleles, hoping to make this point clear. Taking these results of ours and the results of others on the alterations of cell fate in the neural tube and the defects in somite segmentation, we propose that wit is the first neurogenic mutant in the zebrafish.

We would like to thank Vladimir Korzh, Stefan Krauss, Monte Westerfield, Tom Jessell, Mark Fishman, Eric Weinberg, Andreas Püschel, Trevor Jowett and Jóse Campos-Ortega for providing antibodies and cDNA clones. We thank Suresh Jesuthasan and Tanya Whitfield for many helpful suggestions on the manuscript. Y.-J. J. wants to thank Christian Müller and Ralf Rupp for their instructive discussion. Y.-J. J. is a predoctoral fellow supported by Deutscher Akademischer Austauschdienst (DAAD).

After the acceptance of this article, we found that wit^ta52b does not complement or interact with mib^m132 (Schier et al., 1996).