Abstract
During the first day of embryogenesis in the zebrafish, a precise and relatively simple network of neurons develops, pioneering axonal pathways and apparently functioning to mediate reflexive motor responses to touch stimuli. We have begun to use zygotic lethal mutations to analyze the assembly of this ‘primary’ embryonic nervous system. Here we focus on spinal primary motoneurons, their inputs from hindbrain Mauthner neurons, and their outputs to segmental body wall muscle. The mutation nic-1 blocks synaptic transmission between nerve and muscle, yet embryonic primary motoneurons appear normal, suggesting that functional interactions with their targets are not involved in regulating their development. The mutation spt-1 directly disrupts development of this muscle, and the mutation cyc-1 appears to directly block specification of the floor plate. Both spt-1 and cyc-1 affect aspects of primary neuronal development, and they probably do so indirectly. The nonautonomous actions of these mutations are local and they produce variable neuronal phenotypes. The observations can be interpreted to mean that some cellular interactions that specify the neurons and their axonal paths occur at close range and involve multiple, possibly combinatorial, transmitterindependent pathways.
Introduction
During development, nerve cells characterized by reproducible and often cell-specific features in morphology, chemistry, and connectivity arise at predictable times and locations within the nervous systems of vertebrates and invertebrates alike (reviews: Kandel, 1976; Bastiani et al. 1985; Kimmel and Westerfield, 1990; Eisen, 1991). Stereotyped development of each of many classes of neurons clearly plays a major role in generating the structural diversity of nervous systems, and consequently, we believe, their functional complexity. This lesson has been learned by observing normal development of a variety of types of nerve cells in a variety of different organisms. Furthermore, we know from experimental studies in which the neighbor relationships among developing cells are perturbed that cell-cell interactions are critical for establishing normal patterning. We can now begin to set out, again for each of a variety of different systems, schemes showing which cells must interact, and when they must interact, in order that different elements of the pattern be properly generated.
Our goal is to understand at a genetic level the mechanisms of these cell interactions. As illustrated so beautifully by recent studies of Caenorhabditis elegans and Drosophila, we can learn the genetic controls that underlie cell behaviors during development by studying simple systems, where critical observations and manipulations of the developing cells themselves are possible, and where mutations can be obtained to block gene functions. We have found, and show here, that the same approach can be applied to neuronal development in a vertebrate, the zebrafish Brachydanio rerio. We study the first neurons that arise during embryogenesis, and consider how their development is changed by mutations that appear to act not in these cells directly, but in cells in their environments.
Zygotic lethal mutations that act during embryogenesis
We have begun to use procedures developed earlier by Streisinger and his group at the University of Oregon to obtain the mutations (Fig. 1). Most of these mutations have been introduced by gamma-ray mutagenesis at the blastula stage. At this time, the future germ line of the fish appears to be represented by only a few primordial cells (Walker and Streisinger, 1983); if a mutation is introduced into one of these cells, the cell will generate a clone of mutant descendants that will make up a substantial fraction (e.g. 20%) of the germ line. The large size of the mutant clone facilitates both the initial identification and the later recovery of the mutation, since fewer fish need to be examined than if the clone were smaller. Furthermore, the presence in the blastula of several potential target cells for mutagenesis in each fish increases the mutation rate over what we would observe by mutating gametes directly. The procedures are quite efficient; currently we are identifying new mutations, by the embryonic phenotypes they produce, in about one-third of the mutagenized fish that we examine.
Gamma rays efficiently generate recessive lethal mutations in zebrafish. In our current paradigm, mutant eggs (shaded) are obtained from female fish reared from embryos that were gamma-irradiated at blastula stage. Fertilization of the eggs with sperm genetically inactivated by exposure to ultraviolet (u.v.) light initiates development of haploid embryos, and the mutations are identified in these haploids if they produce a distinctive embryonic phenotype. Even if no lethal mutation is present, the haploids die shortly after the time of hatching, so for recovery of the mutation other eggs are fertilized with noninactivated wild-type sperm and diploid embryos are reared. After the diploids reach adulthood, those bearing the mutation in heterozygous form are identified by examining their progeny, either as haploids or from intercrosses.
Gamma rays efficiently generate recessive lethal mutations in zebrafish. In our current paradigm, mutant eggs (shaded) are obtained from female fish reared from embryos that were gamma-irradiated at blastula stage. Fertilization of the eggs with sperm genetically inactivated by exposure to ultraviolet (u.v.) light initiates development of haploid embryos, and the mutations are identified in these haploids if they produce a distinctive embryonic phenotype. Even if no lethal mutation is present, the haploids die shortly after the time of hatching, so for recovery of the mutation other eggs are fertilized with noninactivated wild-type sperm and diploid embryos are reared. After the diploids reach adulthood, those bearing the mutation in heterozygous form are identified by examining their progeny, either as haploids or from intercrosses.
Since most mutations produce recessive phenotypes, finding them can be difficult in diploid animals where the recessive mutations are hidden by the dominant wild-type alleles. Streisinger et al. (1981) circumvented this problem by examining embryos produced parthenogenetically from the mutagenized fish, either as haploids (Fig. 1) or homozygous diploids. New recessive mutations are identified in F1 screens of the progeny of the mutagenized fish. If haploids are used, essentially the whole genome is available for examination.
Our strategy is to be able to study lethal mutations, for they have been extremely useful in analysis of Drosophila embryogenesis (Nüsslein-Volhard and Wieschaus, 1980), and are also proving to be so in C. elegans (J. Rothman, personal communication). We have found that, as is the case for these invertebrates, zebrafish embryos bearing zygotic lethal mutations that perturb the development of specific cells or their arrangements usually survive embryogenesis and only begin to die as young larvae, well after the time of hatching at three days. This is true for mutations like spt-1 and cyc-1, discussed below, which produce severe phenotypes and are expressed as early as the gastrula stage. We take advantage of the fact that embryogenesis can proceed in the presence of a lethal mutation by screening the embryos not for the presence of the lethal phenotype itself, which could be present for a variety of uninteresting reasons, but for associated, embryonically expressed, specific cellular changes. We select those mutations that produce changes we wish to study, and then, after we recover the mutations, we further select ones that we will characterize in some detail by looking for early expression of the phenotype. We believe that this strategy enriches for mutations in genes that are most important for patterning of the embryo.
Stereotyped development of primary neurons
During the first day of embryogenesis, a precisely organized and relatively small and simple network of neurons develop. These neurons pioneer central and peripheral axonal pathways, and some of them interact functionally, apparently mediating the first reflexive motor responses to touch stimuli (review, Kimmel and Westerfield, 1990). We and others (Wilson et al. 1990; Chitnis and Kuwada, 1990; Kuwada et al. 1990a,b’, Wilson and Easter, 1991a,6) have used this early ‘primary’ nervous system as a target system for developmental analysis, essentially ignoring the huge numbers of neurons that begin to become postmitotic only a few hours later.
The genesis of the primary neurons, as postmitotic cells, occurs during or shortly after gastrulation, about 10–1211 after fertilization, and long before the neural tube itself is formed (Kimmel and Westerfield, 1990). A few hours later, just after the trunk somites have appeared, axons begin to emerge from cells present at specific locations all along the neuraxis. Characteristically, in each location, pathfinding occurs by the growth cone of a single pioneering neuron (Eisen et al. 1986). The routes these pioneers take are stereotyped, and within a few hours in many of the locations they are joined by follower axons growing along the leading ones and generating small bundles. This same pattern of early development has been described, for example, for diencephalic commissural neurons (Wilson and Easter, 1991a, 19916), the central axons of sensory neurons in the trigeminal ganglion (Metcalfe et al. 1990), the peripheral axons of lateral line sensory neurons (Metcalfe, 1985), the long central axons descending (Mendelson, 1986) and ascending (Kuwada et al. 1990b) between the hindbrain and spinal cord, and for the spinal motor axons innervating body wall muscle (Eisen et al. 1986; Myers et al. 1986). By the end of the first embryonic day, the axonal bundles, still very small, collectively form a simple scaffold (Wilson et al. 1990) of bilateral longitudinal tracts and commissures in the CNS, and nerve roots communicating with the periphery (Fig. 2).
A simple scaffold of axonal pathways is present in the one-day embryo. Major pathways are projected onto a left side view of the CNS. Sensory nerves and tracts (see Metcalfe, 1989; Metcalfe et al. 1990) are omitted. Abbreviations; DLF, dorsolateral fascicle; DVDT, dorsoventral diencephalic tract; FB, forebrain; HB, hindbrain; IC, intersegmental commissure; MB, midbrain; MLF, medial longitudinal fascicle; TAC, tract of the anterior commissure ;TPC, tract of the posterior commissure; TPOC, tract of the postoptic commissure; SC, spinal cord; SN, segmental nerve; SOT, supraoptic tract; VLF, ventrolateral fascicle; VLT, ventral longitudinal tract; VTC, ventral tegmental commissure. Compiled from Hanneman et al. (1988), Kuwada et al. (1990b), Myers et al. (1986), Trevarrow et al. (1990) and Wilson et al. (1990). Scale bar, 100μm.
A simple scaffold of axonal pathways is present in the one-day embryo. Major pathways are projected onto a left side view of the CNS. Sensory nerves and tracts (see Metcalfe, 1989; Metcalfe et al. 1990) are omitted. Abbreviations; DLF, dorsolateral fascicle; DVDT, dorsoventral diencephalic tract; FB, forebrain; HB, hindbrain; IC, intersegmental commissure; MB, midbrain; MLF, medial longitudinal fascicle; TAC, tract of the anterior commissure ;TPC, tract of the posterior commissure; TPOC, tract of the postoptic commissure; SC, spinal cord; SN, segmental nerve; SOT, supraoptic tract; VLF, ventrolateral fascicle; VLT, ventral longitudinal tract; VTC, ventral tegmental commissure. Compiled from Hanneman et al. (1988), Kuwada et al. (1990b), Myers et al. (1986), Trevarrow et al. (1990) and Wilson et al. (1990). Scale bar, 100μm.
Advantageous for study are the primary motoneurons in the trunk spinal cord and primary interneurons in the hindbrain that project to the spinal cord, so-called reticulospinal neurons. Both classes are segmentally organized, and the cells are all recognizable as individuals; a single type being present on each side of the midline of a single CNS segment (i.e. one cell in each Aem/segment), which permits extremely refined analysis. Thus, the Mauthner neuron, a well-known reticulospinal neuron present in the center of fourth hindbrain hemisegments, invariably is the first neuron in its segment to develop. For example, the Mauthner neuron’s time of origin and time of axon initiation both occur about an hour earlier than an adjacent, more medial, reticulospinal neuron in the same segment, MiMl (Fig. 3). The Mauthner axon crosses the midline to project contralaterally into the spinal cord. The MiMl axon joins that of the Mauthner within the same bundle, but its pathway is ipsilateral (Metcalfe et al. 1986).
The MiM1 and Mauthner (M) reticulospinal interneurons occupy neighboring positions in the fourth hindbrain segment. Transverse section, with the brain midline at the center. The brain contains one of each cell on each side of the midline but here only one member of each of the pairs is labeled, with horseradish peroxidase, according to whether the projection is ipsilateral (MiM1) or contralateral (M; see Fig. 5). On both sides the MiM1 neuron lies just medial to the Mauthner cell. Drawn from Mendelson (1986). Scale bar, 25 μm.
The MiM1 and Mauthner (M) reticulospinal interneurons occupy neighboring positions in the fourth hindbrain segment. Transverse section, with the brain midline at the center. The brain contains one of each cell on each side of the midline but here only one member of each of the pairs is labeled, with horseradish peroxidase, according to whether the projection is ipsilateral (MiM1) or contralateral (M; see Fig. 5). On both sides the MiM1 neuron lies just medial to the Mauthner cell. Drawn from Mendelson (1986). Scale bar, 25 μm.
Within each trunk spinal hemisegment, three, sometimes four, identified primary motoneurons are present (Fig. 4) that innervate the overlying muscle hemisegment in a nonoverlapping invariant pattern. The motor axons of these neurons also develop in an invariant caudal-to-rostral sequence; the first axon to emerge is that of the most caudal cell, CaP, and it innervates ventral myotome (see also Eisen, this volume).
Primary motoneurons are arranged segmentally in the spinal cord. Left side view. A cluster of motoneurons in each spinal hemisegment gives rise to a single ventral root to the overlying muscle segment. Each cluster invariably contains three identified motoneurons (one; the CaP motoneuron is indicated by an arrow), and sometimes a fourth one is also present (Eisen et al. 1990; and see also Eisen, this volume). From Eisen and Pike (1991) with permission from Cell Press. Scale bar, 25 μm.
Primary motoneurons are arranged segmentally in the spinal cord. Left side view. A cluster of motoneurons in each spinal hemisegment gives rise to a single ventral root to the overlying muscle segment. Each cluster invariably contains three identified motoneurons (one; the CaP motoneuron is indicated by an arrow), and sometimes a fourth one is also present (Eisen et al. 1990; and see also Eisen, this volume). From Eisen and Pike (1991) with permission from Cell Press. Scale bar, 25 μm.
During the early part of the second embryonic day, light touches to the head or body consistently begin to elicit muscular contractions (Grunwald et al. 1988). The functional circuitry for these reflexive behaviors appears to involve the neurons we have just been considering (Fig. 5). Trigeminal sensory neurons innervate head skin (Metcalfe et al. 1990) and form the first sensory input to the Mauthner neuron (Kimmel et al. 1990), which projects in turn to the primary motoneurons (Myers, 1985). A touch to the head elicits a contralateral body flexure, which would be expected if the crossing Mauthner axon mediated the reflex. Parallel pathways, perhaps involving segmental homologues of the Mauthner neuron in the hindbrain (Metcalfe et al. 1986), might also be present from very early stages of development.
A circuit of three kinds of primary neurons mediates mechanosensory reflexive movements. Sensory neurons (SN) including trigeminal neurons of the head project ipsilaterally to hindbrain reticulospinal interneurons including the Mauthner (M) neuron. The Mauthner axon crosses the midline and descends to segmentally repeating spinal motoneurons (MN) which innervate body wall muscle (mus). All of the connections are excitatory, and all of the neurons are actually present bilaterally. Contractions contralateral to the side stimulated, as mediated by this pathway, would bend the embryo away from the stimulus.
A circuit of three kinds of primary neurons mediates mechanosensory reflexive movements. Sensory neurons (SN) including trigeminal neurons of the head project ipsilaterally to hindbrain reticulospinal interneurons including the Mauthner (M) neuron. The Mauthner axon crosses the midline and descends to segmentally repeating spinal motoneurons (MN) which innervate body wall muscle (mus). All of the connections are excitatory, and all of the neurons are actually present bilaterally. Contractions contralateral to the side stimulated, as mediated by this pathway, would bend the embryo away from the stimulus.
Even before the tactile reflexes appear, myotomes contract spontaneously, beginning in each segment at the time the CaP growth cone enters the periphery (Myers et al. 1986; Hanneman and Westerfield, 1989). The contractions are probably elicited by spontaneous release of acetylcholine from the motoneurons (Grunwald et al. 1988). One can observe both the elicited and spontaneous movements of these early embryos very easily (Fig. 6), allowing screens for mutations that block development of nerve or muscle, or the functional interactions between them.
The nic-1 mutation blocks spontaneous body contractions. Double exposure of a one-day old wild-type (right) and paralyzed mutant (left) embryo, both about 1.5 mm long. The wild type, but not the mutant, moved during the second exposure. Failure of such movement provides for a simple screening procedure that yields mutants in which defects occur in muscles, or neurons, or as in this case, the synapses between them.
The nic-1 mutation blocks spontaneous body contractions. Double exposure of a one-day old wild-type (right) and paralyzed mutant (left) embryo, both about 1.5 mm long. The wild type, but not the mutant, moved during the second exposure. Failure of such movement provides for a simple screening procedure that yields mutants in which defects occur in muscles, or neurons, or as in this case, the synapses between them.
nic-1: primary motoneurons develop in the absence of function of the muscle acetylcholine receptor
Screens for paralysis of the embryo have yielded mutations whose effects appear restricted to muscle cells, allowing us to address whether activity is involved in early nerve cell development, as it appears to be for some later aspects (Purvis and Lichtman, 1985). In the mutants, the ability of muscle to contract is blocked either because the muscle is defective mechanically, as in fub-1 (Felsenfeld et al. 1990, 1991) or because synaptic transmission fails, as in nic-l (Westerfield et al. 1990). In both cases, primary motoneurons appear to innervate their target muscles normally. These findings argue that development and patterning of the early neuromuscular synapses does not require muscle cell function.
The more meaningful analysis is of nic-l, since its action appears to be at the nerve-muscle junction itself. As determined by intracellular recording, synaptic potentials are absent in mutant muscle cells, even when an acetylcholine agonist, carbachol, is applied to the embryo (Westerfield et al. 1990). Furthermore, the nic-1 muscle cells do not bind detectable alpha-bungarotoxin (Fig. 7), suggesting that acetylcholine receptors are missing from their surfaces, or that they fail to form clustered patches. Yet labeling studies of the primary motoneurons reveal that they are normally arranged in nic-l mutants, that their axons grow to the muscles normally, and that they make normal arbors in the muscle, e.g. as shown in Fig. 8 for the CaP motoneuron. The axons of the primary motoneurons innervating other regions of the myotome, MiP and RoP, also appear normal, as revealed by intracellular dye-labeling (J. S. E., unpublished observations).
Muscle fibers in nic-1 mutants fail to bind detectable alpha-bungarotoxin. Left side views of a trunk muscle segment, viewed with Nomarski optics, are on the left, and fluorescent labeling with rhodamine-bungarotoxin in the same field is shown on the right. Only the wild-type muscle (upper) is labeled. From Westerfield et al. (1990) with permission from Cell Press. Scale bar, 40 μm.
Muscle fibers in nic-1 mutants fail to bind detectable alpha-bungarotoxin. Left side views of a trunk muscle segment, viewed with Nomarski optics, are on the left, and fluorescent labeling with rhodamine-bungarotoxin in the same field is shown on the right. Only the wild-type muscle (upper) is labeled. From Westerfield et al. (1990) with permission from Cell Press. Scale bar, 40 μm.
CaP motor axonal arbors are similar in wild-type and nic-l mutant embryos. Left side views. The axons were orthogradely filled in 3-day old embryos with Dil. The arrow indicates where the main trunk of the motor axon crosses the horizontal myoseptum of the muscle segment, in both cases branches only occur in the ventral part of the segment (below the arrow). From Westerfield et al. (1990) with permission from Cell Press. Scale bar, 40 μm.
CaP motor axonal arbors are similar in wild-type and nic-l mutant embryos. Left side views. The axons were orthogradely filled in 3-day old embryos with Dil. The arrow indicates where the main trunk of the motor axon crosses the horizontal myoseptum of the muscle segment, in both cases branches only occur in the ventral part of the segment (below the arrow). From Westerfield et al. (1990) with permission from Cell Press. Scale bar, 40 μm.
Motor end plates that appear normal under the electron microscope are present (Westerfield et al. 1990). Based on these observations, it may be that synaptic receptordependent interactions between the developing motoneurons and their targets is unnecessary for all aspects of early nerve cell (and muscle cell) development.
spt-1: mesoderm may critically influence primary motoneuronal patterning
However, analysis of another mutation, spt-1, or ‘spadetail’, strongly suggests that other kinds of interactions with muscle are absolutely essential fc motoneuronal patterning. This mutation deforms the embryo (Fig. 9); specifically in the region of the bod trunk, where somite-derived segmental body was muscle is reduced and disorganized. Developments analysis showed the region where trunk somites should form is deficient in mesoderm and that there is corresponding excess of mesoderm in the tail bud. The phenotype appears early, during gastrulation. Together these two observations suggested that the mutation act to disrupt the early morphogenesis of trunk muscle precursors (Kimmel et al. 1989). Cell transplantation between spt-1 and wild-type embryos confirmed this finding, and showed, moreover, that the mutation’ action on mesoderm is autonomous (Ho and Kane 1990). In the experiment summarized in Fig. 10, wild-type and mutant cells were labeled with different colored fluorescent dyes and co-transplanted before gastrulation into an unlabeled wild-type host. According to their genotypes, the cells moved apart during gastrulation. The wild-type cells migrated to the trunk and differentiated later as muscle, and the mutant cells migrated to the tail and formed mesenchyme. The same segregation of wild-type and mutant cells occurs if they are placed into a mutant host. However, the cells don’t segregate if they are transplanted into regions of the fate map that make other mesoderm, such as notochord, and, importantly for the argument here, ectodermal spinal cord precursors do not segregate. Thus the autonomous action of the mutation is apparently specifically on somite-forming mesoderm (Ho and Kane, 1990).
The spt-1 (‘spadetail’) mutation deforms the embryo. Wild-type (left; about 3 mm long) and mutant (right) at day 3. The defects are largely limited to the bod; trunk, except for excess cells at the tip of the tail. Notice the pronounced axial disturbances. Reproduced with permission from Cell Press.
The spt-1 (‘spadetail’) mutation deforms the embryo. Wild-type (left; about 3 mm long) and mutant (right) at day 3. The defects are largely limited to the bod; trunk, except for excess cells at the tip of the tail. Notice the pronounced axial disturbances. Reproduced with permission from Cell Press.
In spite of this, primary motoneurons in the spt-1 trunk are severely perturbed (Eisen and Pike, 1991). Identifiable motoneurons are reduced in number, and are dislocated, not segmentally arranged as in the wild type. Their axons are much shorter than normal, as if arrested in their outgrowth, and are frequently oriented incorrectly such that one cannot distinguish the different types of primary motoneurons normally present (Fig. 11; compare with Fig. 4). Remarkably, neuron-like cells, which by immunolabeling appear to be primary motoneurons, are sometimes located just outside of the spinal cord, specifically in mutants and not wild types, suggesting they migrated there in a completely abnormal manner.
The spt-1 mutation autonomously disorients convergence movements of trunk somite precursor cells during gastrulation. Wild-type (red) and mutant (green) cells were co-transplanted to the lateral marginal zone, from which trunk somites normally arise, of a wild-type host near the beginning of gastrulation. The labeled populations (A) move apart during gastrulation (B). Later (C), in the one-day embryo (1.5 mm long) the wild-type cells have moved correctly to the trunk where they differentiate as body wall muscle, and the mutant cells have moved incorrectly to the tail where they form mesenchyme. From Ho and Kane (1990) with permission from Nature.
The spt-1 mutation autonomously disorients convergence movements of trunk somite precursor cells during gastrulation. Wild-type (red) and mutant (green) cells were co-transplanted to the lateral marginal zone, from which trunk somites normally arise, of a wild-type host near the beginning of gastrulation. The labeled populations (A) move apart during gastrulation (B). Later (C), in the one-day embryo (1.5 mm long) the wild-type cells have moved correctly to the trunk where they differentiate as body wall muscle, and the mutant cells have moved incorrectly to the tail where they form mesenchyme. From Ho and Kane (1990) with permission from Nature.
Motoneuronal development is perturbed in the spt-1 trunk, zn-1 labeled section taken at day 1 and oriented as in Fig. 4, which shows the corresponding view of a wild-type embryo at the same stage. Neurons that may correspond to primary motoneurons, as evidenced by their labeling characteristics with zn-1, are disorganized and malformed. From Eisen and Pike (1991) with permission from Cell Press. Scale bar, 25μm.
Motoneuronal development is perturbed in the spt-1 trunk, zn-1 labeled section taken at day 1 and oriented as in Fig. 4, which shows the corresponding view of a wild-type embryo at the same stage. Neurons that may correspond to primary motoneurons, as evidenced by their labeling characteristics with zn-1, are disorganized and malformed. From Eisen and Pike (1991) with permission from Cell Press. Scale bar, 25μm.
The severity of the changes of the motoneurons is variable, and locally, within segment-like regions in the trunk, seems to match the disturbances to the muscle (Eisen and Pike, 1991). This finding suggests that the mutation is affecting the neurons indirectly, through its disturbance of the mesoderm. Striking support for this suggestion was obtained by transplanting single motoneurons, before they grew axons, between mutant and wild-type embryos (Fig. 12): Wild-type motoneurons transplanted singly into mutant hosts are made to look mutant, and mutant motoneurons transplanted into wild-type hosts are made to look wild type. Thus, the action of spt-1 on the motoneurons is nonautonomous at the level of single cells; presumably the mesodermal disturbances are responsible for the defects.
Disruption of primary motor axonal development in the spt-1 trunk is nonautonomous. Dye-labeled CaP motoneurons developing in live embryos are shown, in situ on the left (A: wild type and B: mutant), and after transplantation on the right (C,D). Postmitotic cells were transplanted before axogenesis. (C) A mutant cell transplanted to a wild-type host later appears wild type in morphology. (D) A wild-type motoneuron, transplanted to a mutant host later exhibits the mutant phenotype. From Eisen and Pike (1991) with permission from Cell Press. Scale bar, 25 μm.
Disruption of primary motor axonal development in the spt-1 trunk is nonautonomous. Dye-labeled CaP motoneurons developing in live embryos are shown, in situ on the left (A: wild type and B: mutant), and after transplantation on the right (C,D). Postmitotic cells were transplanted before axogenesis. (C) A mutant cell transplanted to a wild-type host later appears wild type in morphology. (D) A wild-type motoneuron, transplanted to a mutant host later exhibits the mutant phenotype. From Eisen and Pike (1991) with permission from Cell Press. Scale bar, 25 μm.
We do not know what specific mesodermal changes are responsible for the motoneuronal phenotype. One possibility is that a special class of cells, we term the muscle pioneers, are normally themselves responsible, at least in part, for motoneuronal patterning. These cells differentiate earlier than all other muscle cells (Felsenfeld et al. 1991) at the location where the horizontal septum forms in each myotome, and just where the growth cones of the primary motoneurons normally pause during their outgrowth (Eisen et al. 1986). Furthermore, muscle pioneers are missing in the trunk myotomes of spt-1 mutants (Hatta et al. 1991a), and the location of the apparent arrest of motor axons in spt-1 approximately corresponds to where they should have been. These changes thus specifically link the muscle pioneers to the disorder of motoneurons in the mutant. Recently however, another mutation, ntl-1, has been characterized, in which the muscle pioneers are also missing (M. Halpern, unpublished data) but in which, unlike spt-1, motor axons do not arrest at the site of the horizontal myoseptum. These findings suggest that the muscle pioneers must not uniquely provide a signal to the motor growth cones that is essential for them to pass this ‘choice point’ in their development. On the other hand, the muscle pioneers could well be a part of the signalling system.
All of the neuronal defects in spt-1 that have been observed are restricted to the body trunk, where the mesodermal defects are localized. Primary motoneurons in the tail, where myotomes are normal in morphology, are normal. Additionally, in the head, neurons that are presynaptic to the spinal motoneurons appear normal. For example, the Mauthner cell, which must be clearly deprived of a substantial part of its synaptic output, appears completely normal in structure in spt-1 mutants (Fig. 13A,B). It is possible that early development of the Mauthner neuron is insensitive to changes in its target. However, it seems equally possible that targets are required, but that enough of them are present (some in the spt-1 trunk and the full complement of them in the tail) to provide sufficient stimulus to the developing central neuron. The issue requires a study that more effectively removes the targets of the Mauthner neuron.
The Mauthner axon navigates the same way in wild types (A) and in spt-1 mutants (B), but errors are frequently encountered among cyc-1 mutants (C). Dorsal views of whole mount preparations labeled with the 3A10 antibody at 1 day. Unpublished studies of K. H. Scale bar, 50 μm.
cyc-1-. the floor plate may locally guide CNS axons
Analysis of another mutation, cyc-1 or ‘cyclops’, seems to reveal the presence of patterning interactions occurring between primary interneurons like the Mauthner cell and an interesting class of cells that are not synaptic targets of the interneurons in question; rather they form the floor plate, a nonneuronal structure of the CNS (Hatta et al. 1990; Kuwada and Hatta, 1990). As is the case for spt-1, the cyc-1 mutation produces an early embryonic phenotype that progressively becomes very severe and eventually involves several cell types (Hatta et al. 1991b).
A prominent phenotype of the mutation is cyclopia; in mutants the normal pair of lateral eyes, much smaller than normal, are variably fused together into a single median one lying ventral to the forebrain (Fig. 14). The forebrain itself is vastly depleted of ventral cells, and in fact cyc-1 deletes ventral cells along the whole length of the neuraxis. However, in the caudal brain and spinal cord the phenotype is much more subtle. In these regions only the floor plate appears to missing, and our studies suggest that the direct action of the mutation is to block specifically the development of cells that normally would be located precisely in the ventral midline of the CNS. Possibly all of the other changes, including cyclopia, are secondary consequences of this primary defect. Among these secondary consequences are errors in pathfinding by specific neurons whose axons normally course in the immediate neighborhood of the floor plate (Fig. 13C).
The cyc-1 (‘cyclops’) mutation produces a severe phenotype in the head, including cyclopia and reduced brain size. Wild-type (A) and mutant (B) embryos at day 3. From Hatta et al. (1991b) with permission from Nature.
In the wild-type embryo, the floor plate is extremely simple in structure; it is a one-cell-wide row located in the spinal cord between the central canal and the notochord (Fig. 15A). The close relationship between the notochord and the floor plate appears to be crucial for its development: Transplantation experiments done classically with amphibian embryos (Hamburger, 1988) and very recently with avian embryos (Placzek et al. 1990b) suggest that the notochord specifically induces the floor plate to form. Yet the floor plate does not form in cyc-1 mutants, even though the notochord and other developing mesoderm looks normal.
cyc-1 deletes the floor plate (FP), a single cell row beneath the central canal (CC) in the wild-type spinal cord. Axons of the medial longitudinal fasicles (MLF), which normally are present in bundles on both sides of the floor plate, are scattered in the ventral midline of mutants, just where the floor plate normally lies. Other structures appear normal, including the notochord (NOTO) and the primary motoneurons CaP (C), MiP (M), and RoP (R).
cyc-1 deletes the floor plate (FP), a single cell row beneath the central canal (CC) in the wild-type spinal cord. Axons of the medial longitudinal fasicles (MLF), which normally are present in bundles on both sides of the floor plate, are scattered in the ventral midline of mutants, just where the floor plate normally lies. Other structures appear normal, including the notochord (NOTO) and the primary motoneurons CaP (C), MiP (M), and RoP (R).
Mosaic analysis using cell transplantation provides insight into this matter (Hatta et al. 1991b). Wild-type ectodermal precursor cells, transplanted to mutant hosts at early gastrula stage can differentiate as floor plate, if they develop at the appropriate location (Fig. 16A). Furthermore, mutant cells in their near vicinity can be incorporated into this floor plate (arrow in A), revealing directly that the mutation does not prevent floor plate cell differentiation. Rather, it appears to block initial floor plate specification (see Kimmel et al. 1991), and to do so autonomously, since transplantation of mesodermal precursor cells of the notochord cannot rescue the defect (Fig. 16B). These findings suggest that if in zebrafish, as in other vertebrates, inductive signalling is involved in floor plate specification, cyc-1 blocks the ectodermal response to the mesodermal signal, not its production by the chorda mesoderm.
cyc-1 autonomously blocks floor plate specification. Left-side views of the tail of mutant hosts bearing transplants of wild-type cells (fluorescently labeled) that developed as notochord (NOT) and floor plate (FP in A) or notochord alone (B). When wild-type cells make floor plate, adjacent mutant cells (unlabeled) can also do so (arrow in A), revealing intact homeogenetic floor plate induction in the mutant. From Hatta et al. (1991b) with permission from Nature. Scale bar, 25 μm.
cyc-1 autonomously blocks floor plate specification. Left-side views of the tail of mutant hosts bearing transplants of wild-type cells (fluorescently labeled) that developed as notochord (NOT) and floor plate (FP in A) or notochord alone (B). When wild-type cells make floor plate, adjacent mutant cells (unlabeled) can also do so (arrow in A), revealing intact homeogenetic floor plate induction in the mutant. From Hatta et al. (1991b) with permission from Nature. Scale bar, 25 μm.
We interpret the observed recruitment of mutant cells into the mosaic floor plate as being due to a second kind of normally occurring embryonic induction, so-called homeogenetic induction (‘like begets like’). This induction must fail in mutants (as contrasted with mosaic embryos) because, in this case and owing to the direct action of the mutation, cells that normally produce the signal are missing. The failure thus provides an example of a nonautonomous change due to the mutation.
The same kind of transplantation experiments reveal that the action of cyc-1 on axonal pathfinding is also nonautonomous (Hatta et al. 1990; K. H., unpublished data). In wild-type embryos the bilateral pair of axon bundles that include the Mauthner axons do not touch the floor plate directly, but are removed from it by a single undistinguished cell. In the mutant spinal cord the axons do not form a pair of compact bundles (as in Fig. 15A), but are diffusely scattered in a single, unpaired region in the ventral midline (Fig. 15B). In the regions of the spinal cord of mosaics where the floor plate is present (due to the inclusion of wild-type cells) the axons, derived from genetically mutant neurons, form normal looking bundles on either side of the floor plate, and, as in the wild type, are separated from it by a single cell.
The floor plate may thus normally be involved in signalling the axons to take an orderly longitudinal course beside it through the spinal cord. Its influence might be direct, e.g. through a diffusible signalling molecule. Alternatively, its action could be indirect, mediated through the neighboring cells that are themselves contacted by the axons. The mosaic analyses do not distinguish between these alternatives, but they show that the neighboring cells need not be genetically wild type for signalling to occur. Furthermore, ablation experiments done in wild-type embryos provide suggestive evidence for a direct action of the floor plate on axonal pathfinding, for if stretches of it are specifically ablated shortly before the axons appear, a phenocopy of the axonal disorder cyc-1 is produced, occurring just where the floor plate is missing (K. H., unpublished data).
A role of the floor plate in axonal navigation in zebrafish was suggested previously from descriptive studies of the outgrowth of specific axons in the spinal cord (Kuwada et al. 1990a). At least with respect to its local action, its influence on axons also seems like that proposed for the corresponding structure in the spinal cord of birds and mammals (Plazcek et al. 1990a). However, based upon transplantation studies in avian embryos, Jessell et al. (1989) have proposed that the floor plate plays an essential role in establishing dorsoventral polarity of the neural tube, and is required for the specification of ventral neurons including motoneurons (see also Yamada et al. 1991). Only few motoneurons are present in Xenopus embryos lacking a floor plate and notochord (Clarke et al. 1991). Accordingly, one would predict that motoneurons might be reduced in number or missing in cyc-1 embryos, but in fact both primary and secondary motoneurons are present as usual and their axons extend into the myotomes. It could be that the extent of influence of the floor plate on cells around it varies among different vertebrates, being less in zebrafish than in tetrapods. Interestingly, our mosaic analyses suggest that the floor plate has expanded organizational influence in the zebrafish midbrain as compared with the spinal cord, a difference along the neuraxis which correlates with the severity of the cyc-1 mutant phenotype (K. H., unpublished data). Although the floor plate looks morphologically simple, in fact we know from labeling studies that it is heterogeneous along its length: At the junction of the midbrain and hindbrain the floor plate cells express an Engrailed-like homeodomain antigen that is absent elsewhere (Hatta et al. 1991a), and in a short segment of the trunk spinal cord the cells express a Hox-related homeoprotein antigen that is absent elsewhere (Molven et al. 1990).
Conclusions
Although in its very early days, mutational analysis in zebrafish would seem to be a productive avenue for study of interactions involved in specifying the fates of early embryonic cells, including cells of the nervous system (Fig. 17). Mutations can introduce extremely delicate lesions into the embryo, and thus provide for a very refined level of analysis of the changes the lesions produce. Furthermore, we believe that one can gain insight about patterning of nerve cells not only from the phenotypic changes associated with a mutation, but also from what is left unchanged. Thus, we argue that it is significant that motoneurons are present in cyc-1 mutants when the floor plate is missing, and that their early development occurs normally in nic-1 mutants, where nerve-muscle synaptic function is blocked. However, these findings are negative ones, and conclusions need to be most carefully drawn. For example, suppose we have been fooled about the nature of the primary lesion in cyc-1, and that at early developmental times a floor plate is present, but that it is cryptic. This structure, so far invisible to us, might function to induce the development of motoneurons, thus accounting for the difference we supposed might be present between fishes and birds.
Mutational analyses uncover interactions during development of cells mediating an embryonic reflexive tactle-response behavior. The circuit (see Fig. 5) is from touch-sensitive trigeminal sensory neurons (SN) through the Mauthner neuron (M) and spinal motoneurons (MN) to body wall muscles, cyc-1 appears to block floor plate (FP) specification by the notochord (NOT), and because the FP is missing, M-axonal pathfinding is disturbed, spt-1 blocks morphogenesis of trunk somites, prevents muscle pioneers (MP) from forming, and reveals somite-dependent features of MN development, including their arrangement and axonal development, nic-1 blocks formation of acetylcholine receptor clusters on muscle cells, ntl-1 (a new mutation currently being studied by M. Halpern and R. K. Ho) blocks development of both NOT and MP, In both nic-1 and ntl-1 mutants MN appear normal, suggesting that muscle synaptic receptors, NOT, and MP are each nonessential for MN development.
Mutational analyses uncover interactions during development of cells mediating an embryonic reflexive tactle-response behavior. The circuit (see Fig. 5) is from touch-sensitive trigeminal sensory neurons (SN) through the Mauthner neuron (M) and spinal motoneurons (MN) to body wall muscles, cyc-1 appears to block floor plate (FP) specification by the notochord (NOT), and because the FP is missing, M-axonal pathfinding is disturbed, spt-1 blocks morphogenesis of trunk somites, prevents muscle pioneers (MP) from forming, and reveals somite-dependent features of MN development, including their arrangement and axonal development, nic-1 blocks formation of acetylcholine receptor clusters on muscle cells, ntl-1 (a new mutation currently being studied by M. Halpern and R. K. Ho) blocks development of both NOT and MP, In both nic-1 and ntl-1 mutants MN appear normal, suggesting that muscle synaptic receptors, NOT, and MP are each nonessential for MN development.
As compared with other sorts of lesioning methods, glass needles and laser microbeams for example, mutations enjoy a special position, for they directly perturb the genetic machinery that is driving the developmental process. It is not yet possible, with zebrafish, to use a mutation to initiate molecular studies of the mutated gene itself, but certainly this approach holds great promise for the future.
The interactions revealed by both spt-1 and cyc-1 appear to occur in the local vicinities of the cells in question. In spt-1 the appearance of motoneurons is correlated with the severity of changes in their target muscle, not in muscle in even immediately adjacent segments. The homeogenetic induction revealed by cyc-1 appears locally restricted, and the axonal disturbances produced in cyc-1 are restricted to the neighborhood of the floor plate; axons that course only at some distance to it are normal. Long-range signalling events may well accompany vertebrate neural development, but so far such signals have not been revealed by our experiments.
Finally, it seems likely that we are studying patterning interactions that occur by multiple, perhaps combinatorial, signalling events. Mosaic analysis in cyc-1 suggests that there are two pathways for floor plate induction, signalling from the notochord and homeogenetic signalling within the ectoderm, cyc-1 embryos make motoneurons, even though the floor plate might normally be involved in motoneuronal specification. The ventral axonal pathways that normally neighbor the floor plate are present in cyc-1 mutants, even though they are disorganized, and some instances of axonal navigation that normally occur at the floor plate occur reliably in mutants, e.g. the normal decussation of the Mauthner axon is invariably present (Fig. 13C). Thus, other cues than the floor plate must be used for pathfinding by this axon. In spt-1 .we see that motor axonal outgrowth can at least begin, even though the cells they first normally grow to, the muscle pioneers, are missing. Understanding whether or not the muscle pioneers play any role in motor axonal guidance will involve analysis of another mutation, ntl-1, that also deletes them. The ability to study several mutations that act on the same cell type has greatly facilitated developmental analysis in flies and worms, and we imagine will also be useful in continuing studies of the fish.
ACKNOWLEDGEMENTS
We thank Mamie Halpern and Robert Ho for sharing unpublished observations, and Nigel Holder for comments on the manuscript. Original work was supported by NIH grants HD22486 and NS17963.