Mutations in two genes affect the formation of the boundary between midbrain and hindbrain (MHB): no isthmus (noi) and acerebellar (ace). noi mutant embryos lack the MHB constriction, the cerebellum and optic tectum, as well as the pronephric duct. Analysis of noi mutant embryos with neuron-specific antibodies shows that the MHB region and the dorsal and ventral midbrain are absent or abnormal, but that the rostral hindbrain is unaffected with the exception of the cerebellum. Using markers that are expressed during its formation (eng, wnt1 and pax-b), we find that the MHB region is already misspecified in noi mutant embryos during late gastrulation. The tectum is initially present and later degenerates. The defect in ace mutant embryos is more restricted: MHB and cerebellum are absent, but a tectum is formed. Molecular organisation of the tectum and tegmentum is disturbed, however, since eng, wnt1 and pax-b marker gene expression is not maintained. We propose that noi and ace are required for development of the MHB region and of the adjacent mid- and hindbrain, which are thought to be patterned by the MHB region.
Presence of pax-b RNA, and absence of pax-b protein, together with the observation of genetic linkage and the occurrence of a point mutation, show that noi mutations are located in the pax-b gene. pax-b is a vertebrate orthologue of the Drosophila gene paired, which is involved in a pathway of cellular interactions at the posterior compartment boundary in Drosophila. Our results confirm and extend a previous report, and show that at least one member of this conserved signalling pathway is required for formation of the boundary between midbrain and hindbrain in the zebrafish.
One of the hallmarks of the central nervous system (CNS) is the enormous number of different cell types that cooperate in its function. The great majority of these cell types derive from a common embryonic primordium, the neural plate, but the processes leading to diversification of the neural plate are only poorly understood. The neural plate forms from dorsal ectoderm as a consequence of an inductive influence from the underlying mesoderm (Spemann, 1938). Qualitative differences between the type of inducing mesoderm, as well as interactions between cells of the developing neural plate, are thought to result in the first subdivisions of the neural plate in a process termed regionalization (Nieuwkoop, 1989). As a result of regionalization, gross anatomical subdivisions become recognizable in the embryonic vertebrate CNS, such as the fore-, mid- and hindbrain, and the spinal chord. At least the fore- and hindbrain are thought to be further subdivided into segmentally arranged neuromeres, called prosomeres and rhombomeres, respectively (Puelles et al., 1987; Lumsden and Keynes, 1989; Figdor and Stern, 1993; Rubenstein et al., 1994, and references therein), which are then further sub-divided into individual areas producing neurons, which contain specific and diverse cell types. The process of regionalization is therefore one of the first steps on the way to generating functional diversity in the developing vertebrate brain. Consistent with this view, position within the developing neural plate, rather than lineage history, has been shown to be an important determinant of the individual neuronal cell type in the case of zebrafish primary motoneurons (Eisen, 1994). It was therefore of interest to isolate and study mutations that affect major subdivisions of the brain, because they might define the mechanisms involved in regionalization of the neural plate.
The midbrain derives from the mesencephalic neural plate and includes as major derivatives the optic tectum and the ventral tegmentum. Additional neuromeric subdivisions within the midbrain have been tentatively suggested (Rubenstein and Puelles, 1994). Transplantation and inversion experiments in the chick embryo show that the midbrain develops in intimate association with the primordia of the fore- and hindbrain (Alvarado-Mallart, 1993; Wassef et al., 1993; Nakamura et al., 1994; Marin and Puelles, 1994). Rostral hindbrain tissue, when transplanted into caudal forebrain territory, leads to the expression of midbrain-hindbrain markers, not only in the transplanted tissue, but also in the neighbouring forebrain tissues. When such transplants are allowed to develop, the induced cells show a midbrain-like character (Gardner and Barald, 1991; Martinez et al., 1991; Bally-Cuif et al., 1992). Similar inductive interactions between fore- and hindbrain primordia are suggested by rotation experiments within the developing midbrain primordium (Marin and Puelles, 1994). If only the primordium of the midbrain is rotated, the transplanted piece heals in without any apparent reversal of polarity. If, however, a piece of rostral hindbrain tissue (the MHB primordium) is rotated along with the midbrain primordium, a duplicated midbrain with an inverted polarity results. Similarly, prospective MHB tissue grafted into hindbrain rhombomeres can induce alar rhombomeric cells to form cerebellar tissue (Martinez et al., 1995). These experiments identify the MHB region as an important organizing center with a role in midbrain induction and patterning (Marin and Puelles, 1994; Rubenstein and Puelles, 1994).
Midbrain development can be visualized with the aid of several molecular markers, such as wnt1, eng or pax-b. The corresponding genes are orthologues of the wingless, engrailed and paired loci in Drosophila that are involved in defining, through cell interactions, the posterior compartment boundary in the larval segments (reviewed by Ingham, 1991; Martinez-Arias, 1993). Mutational analysis of these genes in the mouse indicates that their vertebrate orthologues also function during development of the mid- and hindbrain (Thomas and Capecchi, 1990; McMahon et al., 1992; Wurst et al., 1994; Millen et al., 1994; Urbanek et al., 1994). In zebrafish, the function of the paired box-containing pax-b gene was studied by injecting an antibody to the pax-b protein into developing embryos (Krauss et al., 1992b). In the antibody-injected embryos, morphological malformations and a reduction in the levels of wnt1, eng2 and pax-b RNA were observed at the midbrain-hindbrain boundary, leading to the suggestion that pax-b is involved in formation of this structure (Krauss et al., 1992b), consistent with misexpression experiments of pax-b RNA (Kelly and Moon, 1995). We describe the isolation and initial characterization of mutations in two genes that affect the formation of the midbrain-hindbrain boundary. Our results suggest that these genes are required for successive steps in development of the isthmocerebellar primordium, and that one of these genes is the pax-b gene.
MATERIALS AND METHODS
Maintenance of fish, embryo collection and staging
Fish were raised and kept under standard laboratory conditions at about 27°C (Westerfield, 1994; Brand et al., 1995). Mutants were isolated as described in the accompanying paper (Haffter et al., 1996). Mutant carriers were identified by random intercrosses, and identified carriers for the mutation were then outcrossed to wild-type fish to maintain the stock. To obtain embryos showing the mutant phenotype, two heterozygous carriers for a mutation were crossed to one another. Typically, the eggs were spawned synchronously at dawn of the next morning, and embryos were collected, sorted, observed and fixed at different times of development at 28.5°C. In addition, morphological features were used to determine the age of the embryos, as described by Kimmel et al. (1995). In some cases, 0.2 mM phenylthiourea (PTU) was added to prevent melanization. For photography, live embryos were mounted in methylcellulose as described by Westerfield (1994).
Acridine orange staining
To detect degenerating cells, live embryos were dechorionated and placed for 1 hour into PBS, pH 7.1 with 2 μg/ml of acridine orange (Sigma). After two brief washes in PBS, the embryos were mounted in methylcellulose and viewed with fluourescence microscopy, using the FITC filter set.
Immunocytochemistry and antibodies
Whole-mount detection with antibodies was as described by Schulte-Merker et al. (1992). The following antibodies and concentrations were used: anti-acetylated tubulin (Sigma), 1:1000; mAb3a10 (Furley et al., 1990), 1:3; Pan-Isl (Korzh et al., 1993), 1:500; Zn12, Zn5 (Trevarrow et al., 1990), 1:1000 each; mAb 4D9 (Patel et al., 1989), 1:3. Secondary antibodies from a Vectastain elite kit were used at a dilution of 1:300 for detection.
Whole-mount in situ hybridisation
Digoxigenin-labelled RNA probes were prepared using an RNA labelling and detection kit from Boehringer. Hybridisation and detection with an anti-digoxigenin antibody coupled to alkaline phosphatase (Boehringer) was as described by Schulte-Merker et al. (1992), with modifications by C. Houart. RT-PCR isolation and sequencing of mutant noi alleles will be described elsewhere (K. Lun and M. Brand, unpublished).
In our screen, we recovered seven recessive embryonic lethal mutations that affect formation of the boundary between mid- and hindbrain (MHB). Based on complementation tests, these mutations define two genes: no isthmus (noi) with six alleles of varying strength, and acerebellar (ace) with one allele (Table 1). In the following phenotypic analysis of noi, we always used one of the strong alleles, unless stated otherwise. Two other non-complementing mutations found by Schier et al. (1996) define the spiel-ohne-grenzen (spg) gene, which complemented both noi and ace mutations.
Development of noi embryos
Wild-type embryos during the pharyngula period [24-48 hours of development (h)] show a conspicuous constriction of brain tissue at the boundary between the mid- and hindbrain (Figs 1A, 7E). Embryos that are homozygous mutant for noi (Fig. 1B) or ace (Fig. 1D) lack this constriction completely, along with the cerebellum that derives from its posterior part. In addition, embryos that are homozygous for strong noi mutations lack the tectum, a dorsal portion of the midbrain (Fig. 1B, asterisk). For two weaker alleles of noi, a partially formed tectum was observed (Fig. 1C) that can still be recognized by ingrowing retinal axons (Trowe et al., 1996). Outside of the brain, noi mutant embryos during the late pharyngula stage lack a pronephric duct, which is normally visible as a tubular epithelium above the yolk in wild-type embryos (Fig. 1E,F, arrows). Also, whereas circulation is well established through the common cardinal vein of 36 h wild-type embryos (Kimmel et al., 1995), circulation is feeble in noi mutant embryos, where most blood cells accumulate in a dent on the surface of the ventral yolk sac (Fig. 1G,H), and the otic vesicle is often slightly reduced in size. The reason for the defective circulation is not known. Presumably because of this defect, embryos that are homozygous for a strong allele of noi show extensive edema on the third day of development, followed by severe degeneration and retardation in many tissues on the fourth day of development. Embryos homozygous for weak alleles survive to day 6 or 7, but fail to develop a swim bladder and eventually die.
Tectal cells undergo cell death in noi embryos
Living homozygous noi mutant embryos could be distinguished from their wild-type siblings from the late segmention period (about 22 h) of development onwards, based on a greater turbidity of the cells of the developing tectum. Staining of such embryos with acridine orange (AO), a dye that specifically detects apoptotic cell death in Drosophila (Abrams et al., 1993), revealed that a large block of predominantly dorsal cells die in the midbrain of noi mutant embryos for three different strong alleles (n=18), but not in their wild-type sibling embryos (n=13; Fig. 2). We observed no staining above wild-type levels at earlier stages (6-somite and 14-somite stages; n=30, wild type and mutants) or in other tissues, such as the pronephric duct. Much smaller numbers of AO-positive cells were seen in the tectum of embryos homozygous for weak alleles of noi (not shown). We conclude that a tectum initially forms in noi mutant embryos, but that it subsequently degenerates. We observed no increased cell death in homozygous mutant ace embryos (not shown).
Overall organisation of the CNS in noi
We looked at the overall neuromeric organisation and the formation of early axonal fascicles in the developing noi mutant brain, using an antibody to acetylated tubulin (Fig. 3; Chitnis and Kuwada, 1990; Ross et al., 1992; Wilson et al., 1990).
Despite the absence of major portions of the brain, other areas are developed remarkably normally in homozygous mutant noi embryos at the pharygula stage. The establishment of the major comissures and longitudinal connectives in the fore- and hindbrain is not affected, like the medial longitudinal fascicle and its nucleus, which is thought to straddle the forebrain-midbrain boundary (Macdonald et al., 1994; Fig. 3A,B). Specific hindbrain neurons like the Mauthner cell and several of the reticulospinal neurons (Kimmel et al., 1985; Metcalfe et al., 1986; Hanneman et al., 1988) also seem to form normally, as revealed by staining with the 3A10 antibody (Fig. 3C,D). On the other hand, the distance between the forebrain-midbrain boundary (marked by the position of the tract of the posterior commissure) and the first hindbrain commissure (arrowheads in Fig. 3A,B) appears to be severely shortened in mutant embryos. Also, formation of the tract of the ventral tegmental commissure is severely reduced in mutant embryos (Fig. 3A,B, arrow in A). The nuclei of early differentiating cranial neurons of the mid- and hindbrain can be visualized with an antibody recognizing several Isl proteins (Pan-Isl; Korzh et al., 1993). In the tegmentum of 32 h wild-type embryos stained with this antibody, two bilateral clusters of neurons are seen, which probably are the nuclei of the oculomotor and trochlear nerves. In noi mutant embryos, only a small remnant of these clusters is observed, and the distance to the hindbrain nuclei is severely shortened (Fig. 3E,F; arrowheads); in addition, the trochlear nerve is absent from Zn5-stained mutant embryos (not shown). Together, these results point to a strong reduction of the occulomotor and trochlear nuclei in noi mutant embryos. As was seen in acetylated tubulin- and 3A10-stained embryos, organization of the fore- and hindbrain is largely normal in mutant noi embryos at this stage. This analysis indicates that CNS defects in noi mutant embryos are largely restricted to much, but not all, of the midbrain, the cerebellum and the MHB region.
Expression of markers of the MHB is affected in noi
Our analysis of the brain structure of noi mutant embryos suggested that absence of the dorsal derivatives of the midbrain (tectum) and rostral hindbrain (cerebellum) could be due to a specific deletion of tissue from the MHB region. We therefore examined the expression of markers for both dorsal and ventral portions of the midbrain-hindbrain boundary in wild type and noi mutant embryos during the pharyngula stage. In wild-type embryos stained with the 4D9 antibody, which recognizes all three zebrafish eng proteins (Patel et al., 1989; Ekker et al., 1992b), expression is seen in both dorsal and ventral portions of the MHB, and the neighbouring caudal midbrain and rostral hindbrain (Fig. 4). In 30 h embryos carrying a strong allele of noi, the MHB constriction is not formed, and consequently no eng staining is found in this area. In addition, expression of eng protein is also not detectable in the neigh-bouring caudal tegmentum and rostral hindbrain, which are not overtly missing in noi mutant embryos (Fig. 4A,B). Expression of eng proteins is normal in other domains, such as the muscle pioneers or the clusters of hindbrain neurons (not shown). Expression of other markers, such as wnt1 (Molven et al., 1991) and zash1A (Allende and Weinberg, 1994) is also affected at the MHB in mutant embryos (Fig. 4). In wild-type embryos at the 20-somite stage, wnt1 expression is detected in an anterior cluster underneath the epiphysis, along the dorsal edge of the tectum, in the MHB, and in segmental clusters in the hindbrain (Fig. 4E). In noi mutant embryos, wnt1 expression at the MHB is abolished, whereas other expression domains are not affected (Fig. 4F). The distance between the anterior cluster and the first rhombomere is decreased in noi mutant embryos, suggesting that the cells of the MHB area are absent at this stage. Nevertheless, a tectum is still present at this stage in noi mutant embryos, and its posterior edge appears to be directly joined to the first rhombomere (Fig. 4F). zash1A expression in wild-type embryos at the pharyngula stage occurs in two ventral clusters in the tegmentum and ventral rhombomere 1 or 2, leaving a gap at the MHB (Fig 4G, between arrowheads; Allende and Weinberg, 1994). In noi mutant embryos, zash1A expression extends across this gap (Fig. 4H, arrowhead), again indicating that the intervening tissue may be deleted or respecified. No defects are observed in the expression of markers at the forebrain-midbrain boundary (pax-a) or the middien-cephalon (shh, fkd3; not shown). We conclude that expression of eng, wnt1 and zash1A is specifically affected at the midbrain-hindbrain boundary in noi mutant embryos.
Development of the MHB primordium is affected in noi
In order to determine if the primordium of the MHB is already affected in noi mutant embryos, we looked for expression of eng proteins in noi embryos during somitogenesis that were doubly stained with krx20, a marker for rhombomeres 3 and 5 (Oxtoby and Jowett, 1993; Fig. 4C,D). In wild-type embryos at the 8-somite stage, eng expression is detected in a broad band across the developing MHB region of the neural keel, located half way between the rhombomere 3 stripe of krx20 and the optic vesicle. No expression of eng is detected in noi mutant embryos, whereas expression of krx20 is unaffected in the same embryos (Fig. 4D). The distance between the rhombomere 3 stripe and the posterior optic vesicle seems unchanged in DAPI-counter-stained specimens, indicating that the MHB primordium is not missing in the mutant embryos at this stage. We also do not detect any eng staining at the end of gastrulation, shortly after the onset of eng staining in wild-type embryos (n=16 of 67 embryos; not shown). In contrast, pax-b RNA expression, as another marker of the MHB primordium, is normal in late gastrula and 6-somite stage mutant noi embryos, but is fading to near invisibility at the 14-somite stage (Fig. 5G,H). We conclude that the MHB primordium is initially present, but defective in marker gene expression in noi mutant embryos.
pax-b protein, but not RNA, is eliminated in noi
pax-b is expressed during development of the MHB, pronephros, optic stalk and certain hindbrain and spinal chord neurons (Fig. 5A,C; Krauss et al., 1991a; Mikkola et al., 1992). In assaying the expression of pax-b protein as another marker of the MHB region, we noticed its entire absence in the mutant embryos (n=22 of 88; Fig. 5B,D). A survey of the different noi alleles using whole mount staining with a polyclonal anti-pax-b antibody showed that pax-b protein staining is eliminated in mutant embryos for three of the four strong alleles, but present for the two weak alleles (Table 2). Using whole mount in situ hybridisation, we determined that pax-b RNA is still present in 28 h homozygous mutant embryos for all alleles in all tissues, excepting the MHB (Fig. 5E,F). pax-b RNA expression is seen in the pronephric duct, optic stalk and the otic placode of mutant embryos, though the staining area in these tissues often appears slightly smaller than in the wild-type siblings (Fig. 5E,F). The optic nerve forms normally in Zn5-stained mutant embryos, and expression of dlx3, a marker for development of the inner ear (Ekker et al., 1992a) occurs normally in the ear up to 30 h (not shown). In addition, the overall level of pax-b RNA is reduced in 30 h embryos homozygous mutant for all of the strong alleles (Fig. 5F, Table 2).
The noi mutation is linked to the pax-b gene
The preceding data suggest that noi mutations affect the pax-b gene itself. To further examine this possibility, we looked for genetic linkage between the noi phenotype and a restriction fragment length polymorphism (RFLP), which we found in the pax-b gene, using the scheme depicted in Fig. 6A. The allele noith44 was induced in the background of an inbred wild-type strain from Tübingen (Tü). Heterozygous carrier males for this allele were crossed to a female from a wild-type AB strain. In Southern blots of genomic DNA probed with a cDNA for the pax-b gene (Krauss et al., 1991a), these two strains yield a RFLP (Fig. 6B) that can be used to test for linkage between the noi mutant phenotype and the pax-b gene. If pax-b and noi are not linked, the pax-b polymorphism should assort randomly, i.e. both mutant embryos and their siblings should show both RFLP bands with equal intensity. If noi and pax-b are linked, then sorted mutant noi embryos should always carry only the Tü allele, which is what we observed (Fig. 6B). We then sequenced two independently amplified RT-PCR fragments containing the coding region of pax-b from homozygous mutant embryos for one of the strong alleles, noith44a (Fig. 6C). We found several ‘silent’ single base changes relative to the published wild-type sequence (Krauss et al., 1991a), but only one change leading to altered amino acid sequence, resulting in a stop codon in the middle of the coding region (Fig. 6D; K. Lun and M. Brand, unpublished data). The predicted shorter mutant protein (Fig. 6D) is presumably inactive or unstable. We conclude that noi mutations are located in the pax-b gene.
Phenotype of living ace embryos
Homozygous mutant acerebellar (ace) embryos can be distinguished from their wild-type siblings at the 5-somite stage, based on a slightly thicker neural keel in the area of the developing midbrain (Fig. 7A,B). Like noi mutant embryos, homozygous ace mutant embryos lack the MHB constriction at the pharyngula stage (Fig. 7C-F); they also show a similar defect in circulation (not shown), which is probably the causeof their death around day 7 of development. In contrast to noi mutant embryos, where the tectum is missing, homozygous ace embryos have a tectum that appears to be bigger than that of wild-type embryos (Fig. 7C-F, and below), and they have a normal pronephric duct (not shown). In addition, ace embryos develop a smaller otocyst that usually has only one otolith (Fig. 7E-H). On day 5 of development, overall size and the formation of semicircular canals are affected in the otocysts of ace mutant embryos (Fig. 7G,H; Whitfield et al., 1996).
Unlike in noi mutant embryos, we did not observe an increase in the amount of dying cells in acridine orange-stained ace mutant embryos at 24 h, compared to their wild-type siblings (not shown). Stainings with anti-acetylated tubulin and ZN12 antibodies showed an overall normal organisation of axonal tracts, and no decreased distance between the first hindbrain fascicle and more anterior comissures was observed with these antibodies (not shown). Likewise, the arrangement of Isl-stained cranial nuclei at 24 hours of development was normal (not shown); later stages were not examined. These findings indicate that the defect in ace mutant embryos is more restricted than in noi mutant embryos.
Expression of markers of the MHB region in ace
The absence of the MHB constriction, along with the presence of an enlarged tectum, led us to study expression of markers of the anterior portion of the constriction in ace mutant embryos. Does the constriction become incorporated into the tectum and/or tegmentum that is seen in ace mutant embryos, or is this part deleted? Does the defect in the MHB region also affect both dorsal and ventral parts of the junction between mid- and hindbrain, as seems to be the case in noi mutant embryos, or does only the cerebellum fail to form? We studied expression of eng, pax-b, wnt1 and zash1A in homozygous mutant ace embryos during the pharyngula period (Fig. 8). Expression of eng proteins, as seen by 4D9 antibody staining, is not detected in the dorsal or ventral MHB region of 28 h
ace mutant embryos (Fig. 8A,B), but is present in other parts of the body (not shown). Expression of pax-b RNA and protein, and of wnt1 RNA mark the anterior wall of the constriction, which is contiguous with the tectum. In ace mutant embryos, expression of both markers is absent from the anterior MHB, but is not affected in other places (Fig. 8C-F). Instead, the domain of wnt1 expression appears to extend further along the dorsal edge of the ace mutant tecta than in the wild-type tecta (Fig. 8E,F; the posterior end of this domain is marked by an arrow). We also observe altered expression of eng and zash1A in ace mutant tecta: in wild-type embryos, eng protein staining extends from the posterior midbrain border with gradually dimishing intensity into the tectum and tegmentum (Fig. 8A); no expression is observed in the tectum or tegmentum of ace mutant embryos at this stage (Fig. 8B). zash1A is weakly expressed in the ventricular zone of the tecta of wild-type embryos at the pharyngula stage (Allende and Weinberg, 1994; Fig. 8G, arrowhead). Concomitant with the loss of eng expression, we see a strongly increased expression of zash1A in the tecta of ace mutant embryos (Fig. 8H, arrowhead). No defect is seen in eng or pax-b expression in embryos at the end of gastrulation, but the MHB stripe of pax-b is thinner in mutant embryos at the 11-somite stage (not shown). In summary, the cerebellum is clearly absent in living ace mutant embryos. Our analysis of marker expression shows that the MHB region in general, and in particular the anterior part of the MHB constriction, are progressively lost or respecified in ace mutant embryos up to the pharyngula stage of development.
We have described mutations in two genes, no isthmus (noi) and acerebellar (ace), that are required for proper formation of the boundary between mid- and hindbrain. noi is required for a very early step in development of the MHB primordium during late gastrulation, and for development of the tectum and cerebellum. In addition, noi is also needed for maintenance of the pronephric duct. Our analysis of pax-b protein and RNA expression, and the genetic linkage between the noi mutant phenotype and the pax-b gene, strongly argue that the paired box gene pax-b is mutated in noi mutants. This is confirmed by the observation of a stop codon in the middle of the open reading frame in one of the strong noi alleles, which predicts a truncated protein of about half the size (Fig. 6; K. Lun and M. Brand, unpublished data). acerebellar (ace) seems to be required for a later and more restricted aspect of MHB development, possibly to prevent expansion of tectal identity into the MHB region, and for proper ear development. We do not yet know which gene is mutated in the ace mutant.
Detection of zebrafish mutants affecting formation of the MHB
At least three zebrafish genes (noi and ace, this paper; spg, Schier et al., 1996) can be mutated to a condition in which the affected embryos lack a MHB. Inactivation of the mouse genes wnt1, Eng1 or pax-5 (but not pax-2; M. Torres and P. Gruss, personal communication), leads to a similar phenotype (Thomas and Capecchi, 1990; McMahon et al., 1992; Millen et al., 1994; Urbanek et al., 1994), and mutations in a fourth gene, Eng-2, lead to more subtle defects in development of the cerebellum only (Millen et al., 1994) that we would not have detected with our morphological screening procedure. Such mutations may, however, have been kept if they affected other structural or functional aspects of brain development (see other papers in this issue).
Extent of the defect in noi and ace embryos
In living noi mutant embryos, development of the MHB constriction, the tectum and the cerebellum are overtly affected. Analysis of brain structure, absence of eng and wnt1 expression, and altered expression of zash1A in the mutant noi embryos, show in addition that part of the tegmentum and rostral hindbrain are affected as well. The situation is similar for ace mutant embryos, though the defects appear to be more confined to the MHB region itself. Overall, the observed defects in noi mutant embryos correlate well with the broad, early phase of pax-b expression throughout the midbrain and MHB primordia during gastrulation, suggesting that this could be the critically required expression phase, rather than the late expression that is restricted to the MHB. A requirement for the broader, early expression phase was also postulated for the mouse wnt1 gene (Rowitch and McMahon, 1995). Alternatively, the progressive loss of midbrain tissue in noi mutant embryos may be due to loss of the organizing ability postulated for the MHB region (Marin and Puelles, 1994; see below).
Determination of the MHB region
Establishment of pax-b mRNA (but not protein) expression during late gastrulation stages occurs normally in noi mutant embryos, whereas eng expression is abolished. Therefore, positional information leading to formation of the MHB primordium must be generated independently of the pax-b gene.
pax-b most likely functions during early determination or differentiation of the MHB primordium. In contrast, early expression of pax-b and eng are not affected in ace mutants, indicating that ace affects a later or more restricted aspect of MHB development.
Are noi and ace involved in regionalization of the neural plate?
The defects of noi and ace mutant embryos in the midbrain and MHB region beg the question of how this affects development of the adjacent brain regions. We have not examined this question in detail yet. Loss of the wnt1 and pax-b expression at the MHB, along with extension of the dorsal mesencephalic wnt1 domain and the increased size of the tectum in ace mutant embryos, argue that the MHB region is transformed to a more anterior fate. In noi mutant embryos, the aberrant extension of the tegmental domain of Zash1A posteriorly (Fig. 4G,H) indicates that a transformation of the MHB region to a more anterior fate may have occurred. The altered marker expression we observe in noi and ace could therefore reflect disturbed interactions in regionalization of the neural plate. Since pax-b expression is found in the neural ectoderm, but not the underlying mesoderm, these regionalization processes must also depend on interactions in the ectoderm. Similar transformations within the forebrain anlage are observed for embryos mutant for the masterblind mutation found in our screen, which lack eyes and the telencephalon, and show an expanded posterior diencephalon. Transplantation chimaeras argue that masterblind is autonomously required in the ectoderm, again suggesting that interactions influencing regionalization of the forebrain occur in the ectoderm (Heisenberg et al., 1996).
Functional requirement of the pax-b gene
Previous experiments have already indicated a requirement for pax-b function. Injection of an antibody to pax-b protein lead to malformation of the MHB, and to reduced expression of molecular markers of this region (Krauss et al., 1992b), but it could not be tested whether the observed defects were solely due to inactivation of pax-b protein. Our results confirm and extend these studies. pax-b is expressed during development of the midbrain and the MHB, optic stalk, otic placode, certain hindbrain and spinal chord neurons and the pronephros (Krauss et al., 1991a; Mikkola et al., 1992). The requirement for pax-b function is most obvious for the MHB region, a large portion of the midbrain, and for the cerebellum and the pronephric duct. The common functional requirement for pax-b in development of the more primitive pronephros in fish, and for Pax-2 in metanephric development of the mouse (Torres et al., 1995), points to an interesting conservation in the genetic mechanisms involved in the formation of these structures. Recent experiments also suggest a function for pax-b in partioning the optic vesicle into optic stalk and neural retina, in response to midline-derived signals (Macdonald et al., 1995; Ekker et al., 1995). Consistently, a mutation in the human Pax2 gene leads, as a dominant trait, to optic nerve colobomas and kidney dysfunction (Sanyanusin et al., 1995). Our findings suggest that pax-b has only a minor role, if any, in development of the eyes and inner ear as a whole, since the eyes and the optic nerve appear to form normally in the mutant embryos, and the otic vesicle is only visibly affected at a time when the embryo is already rather sick from defective circulation. It remains to be determined, however, if more subtle aspects of optic stalk and inner ear development depend on the pax-b gene.
In the zebrafish, pax-b is currently the only member of the pax gene family known to be expressed at the MHB. Although only pax-b was isolated in two independent studies (Krauss et al., 1991a; Püschel et al., 1992), further pax genes expressed at the MHB may yet exist. In mice, Pax-2, -5 and -8 are expressed in overlapping domains at the MHB (Asano and Gruss, 1992; Stoykova and Gruss, 1994). Pax-8 mutants have not been described yet. Inactivation of Pax-5 leads to defective development of the posterior midbrain and cerebellum, but the defects seem to be considerably milder than in noi mutant embryos (Urbanek et al., 1994). Homozygous mutant Pax-2 mice lack kidneys, ureters and genital tracts (Torres et al., 1995), but have a normal midbrain-hindbrain boundary (M. Torres and P. Gruss, personal communication). Molecularly, pax-b is clearly related to both Pax-2 and Pax-5: whereas Pax-5 is only expressed at the MHB and not in the developing optic stalk, ear and kidney, pax-b and Pax-2 are expressed in all of these tissues (Asano and Gruss, 1992; Krauss et al., 1991a). On the other hand, the paired domain of pax-b is almost identical to the one of Pax-5 (1 in 120 amino acids changed; Krauss et al., 1991a), consistent with the requirement for pax-5 in MHB development. We therefore propose that the strong requirement for a single pax gene in the zebrafish may reflect the ancestral situation before the separation of mammals and teleosts in evolution. Alternatively, pax genes may have been lost in teleosts that were retained in the mammalian lineage. In either case, loss of pax-b function in the zebrafish is predicted to be equivalent to the knockout phenotype of all three mouse pax genes expressed at the MHB.
Evolutionary conservation of the pax-b pathway
The first vertebrate members of the pax gene family were isolated using a ‘paired-box’-containing probe from the Drosophila paired locus (reviewed by Noll, 1993; Chalepakis et al., 1993). The Drosophila paired gene cooperates with wingless, hedgehog and engrailed in a signaling pathway controlling the formation of the posterior compartment boundary (reviewed by Bejsovec and Martinez Arias, 1991; Ingham, 1991; Martinez Arias, 1993). Several vertebrate orthologues of members of the Drosophila signaling cascade are now known to be expressed at the MHB in the zebrafish (Krauss et al., 1991b; Ekker et al., 1992b; Püschel et al., 1992; Krauss et al., 1992a), and targeted inactivation of the mouse orthologues wnt-1, Eng-1, Eng-2 and of pax-5 have demonstrated a functional requirement for these genes in midbrain and/or cerebellum development (Thomas and Capecchi, 1990; McMahon et al., 1992; Wurst et al., 1994; Millen et al., 1994; Urbanek et al., 1994). The previous results of the antibody injection experiments by Krauss et al. (1992b) and our results show that at least one of the zebrafish orthologues of the Drosophila signaling pathway, pax-b, also functions in MHB development in the zebrafish; ace, and perhaps the spg gene (Schier et al., 1996) as well, might encode some of the missing members. Our results suggest that noi mutations in the pax-b gene affect the MHB primordium at a very early stage of its development, and that one of the functions of the pax-b transcription factor may be to regulate expression of the eng genes. Similarly, normal eng and pax-b expression in ace mutant embryos up to the 8-somite stage shows that ace is required for a later step in development of the MHB than noi. The expression of and phenotypic requirement for pax-b RNA and protein thus precedes the time when the requirement for ace function becomes apparent, although eventually, the phenotype of noi and ace come to resemble each other at the pharyngula stage. The ace gene might therefore well be one of the downstream targets of the pax-b protein, and could also be a member of the conserved signaling cascade.
noi and ace are required for development of an important organizing center in the brain
In Drosophila, the signaling events leading to the establishment of the compartment boundary take place between neighbouring cells. By inference, a similar signaling process might be necessary to establish a proper boundary between mid- and hindbrain during vertebrate neural plate development; this hypothesis can be tested in the zebrafish by performing a test for non-autonomy in transplantation experiments using the mutants we have described. In chicken, rotation experiments of the midbrain including or excluding the MHB region (Marin and Puelles, 1994), and transplantations into naive forebrain territories (Gardner and Barald, 1991; Itasaki et al., 1991; Martinez et al., 1991; Bally-Cuif et al., 1992) have shown that the MHB region can orchestrate development of the midbrain. Our observation of cell death of tectal cells in noi mutant embryos, and of aberrant marker gene expression in ace mutant embryos, could reflect the absence of such a signal. The MHB region and the tectum seem to be affected in mutant noi or ace embryos in different ways. In mutant noi embryos, absence of eng or wnt1 expression from the MHB is detected long before apoptotic cells are seen in the tectum. Likewise, in ace mutant embryos, absence of the MHB is observed when a tectum is clearly present, although abnormal in marker gene expression. Absence of noi and ace might affect the MHB and the tectum in different, but direct ways: e.g. noi mutations may cause a respecification in the case of the cells of the MHB, and death in the case of the tectum. Alternatively, defective development of the MHB region may lead to secondary defects in the tectum, resulting in its loss in noi mutant embryos, and in altered marker gene expression in ace mutant embryos. Such a scenario would be consistent with the proposed organizing abilities of the MHB region that have been demonstrated in chick embryos. The conserved signaling pathway that pax-b, and perhaps ace and spg, are involved in could be the substrate matter of this patterning influence.
We would like to thank our colleagues in the zebrafish community for generously sharing antibodies and probes, in particular Terje Johannsen, Vladimir Korzh, Stefan Krauss and Ingvild Mikkola, as well as Christine Dreyer, Nigel Holder, Tom Jessel, Trevor Jowett, Anders Molven, Eric Weinberg and Monte Westerfield. M.B would like to thank his colleagues for numerous discussions, and Francisco Pelegri, Suresh Jesuthasan and Luis Puelles for comments on the manuscipt. Thanks also to Peter Andermann and Eric Weinberg, who helped in the analysis of Zash expression, and especially to Corinne Houart, for her lovely in situ protocol and many discussions. Silke Hein helped greatly in final stages of this work. M.B. was supported by a Helmholtz stipend of the BMFT.