The zebrafish space cadet gene controls axonal pathfinding of neurons that modulate fast turning movements

Vertebrate motor behaviors such as respiration, saccadic eye movements and locomotion are controlled by dedicated neural circuits, composed of a defined complement of neuronal cell types with precise and stereotypic neural connections. The function of these neural circuits in the mature nervous system depends on earlier events, when embryonic neurons establish precise and selective connections with their synaptic partners. This process of assembling neural circuits includes specification of neuronal cell types, outgrowth and pathfinding of axons, synapse formation, and refinement of these connections (for a review, see Albright et al., 2000). A key point in understanding neural circuits concerns the role that individual neural cell types within a neural circuit play to control a defined motor behavior.

Genetic studies mainly in the invertebrate Caenorhabditis elegans have provided several examples in which specific neuronal cell types or their synaptic connections contribute to a neural circuit that underlies a defined behavior (e.g. Bargmann et al., 1993; McIntire et al., 1993). For example, mutations in the gene unc-4 change the synaptic VA motoneuron input from AVA interneurons, which are responsible for reverse motions, to inputs from AVB interneurons, which mediate forward movements (Miller et al., 1992; White et al., 1992). This has demonstrated a functional role for AVA interneurons in backward movements of C. elegans. In vertebrates, studies that aim to correlate individual neuronal cell types with a defined behavior present a more difficult task, in part because of the enormous complexity of the nervous system. Traditionally, electrophysiology and lesion experiments have been used to examine the role of individual cell types for defined behaviors, such as locomotor behavior (for reviews see: Bate, 1999; Grillner and Wallen, 1999). In the mouse, a variety of mutants with locomotor defects have been studied, including reeler (Alter et al., 1968; D’Arcangelo et al., 1995; Hirotsune et al., 1995), lurcher (Le Marec et al., 1997; Zuo et al., 1997) and staggerer (Hamilton et al., 1996; Lalonde, 1987). Although the genes mutated in these mice have been identified, it still remains a formidable task to understand the functional contribution of the affected neural cell types for the particular motor behavior they control.

The zebrafish is emerging as an excellent genetic model system in which to study the assembly and the function of neural circuits that underlie locomotor behavior (Bate, 1999; Fetcho and Liu, 1998). As in other developing vertebrates, zebrafish embryos and larvae display a well-defined set of developmentally regulated, stereotypical locomotor behaviors (Eaton et al., 1977; Kimmel et al., 1974; Saint-Amant and Drapeau, 1998). Most of the neural circuits controlling embryonic and larval motor behaviors are thought to reside in the spinal cord and in the hindbrain (Saint-Amant and Drapeau, 2000). A large number of the neuronal cell types in the larval hindbrain and spinal cord can reliably be identified by their stereotyped position, morphology and axonal trajectories (Bernhardt et al., 1990; Kimmel et al., 1982; Metcalfe et al., 1986). In recent years, several powerful methods to study neural circuits in the zebrafish have been adapted or developed, including electrophysiological recordings from identified neurons (Ribera and Nusslein-Volhard, 1998; Hatta and Korn, 1998), optical monitoring of neural activity in intact embryos (Fetcho and O’Malley, 1995) and laser-ablation of individual neurons (Liu and Fetcho, 1999). These methods, combined with the large collection of locomotor defective mutants (Granato et al., 1996), provide a unique opportunity to study the genetic, molecular and cellular basis of neural circuits that underlie locomotion in vertebrates.

We have sought genes required for assembly and function of neural circuits that control defined locomotor behaviors. To identify mutants in which the absence of a distinct neuronal population or of its axonal trajectories correlates with specific motor defects, we examined a collection of mutants with presumed defects in the neural circuitry that underlies locomotion (Granato et al., 1996). We have now used high-speed video microscopy to show that mutant space cadet larvae fail to activate high-speed turning movements properly, such as the escape response. Using a trans-synaptic labeling approach, we demonstrate that space cadet larvae lack axonal connections between hindbrain spiral fiber neurons and the postsynaptic Mauthner neuron, a central circuit component that controls fast turning movements. Moreover, severing spiral fiber axons in wild-type larvae causes space cadet-like locomotor defects, providing functional evidence that spiral fiber neurons play a vital role in the circuitry controlling fast turning movements. Finally, analysis of retinal ganglion cell axons provides compelling evidence that the space cadet gene functions in pathfinding of spiral fiber axons, integrating these neurons into the Mauthner cell circuits. These studies provide a rare example in vertebrates in which a defined motor behavior is directly correlated with a small population of identified hindbrain neurons.

All experiments were performed using both space cadet alleles, i.e. te226a and ty85d, in a Tübingen strain background (Granato et al., 1996). Both alleles are embryonic lethal and display similar phenotypes. The retinotectal phenotype was previously unknown. To prevent melanin formation from obscuring antibody stainings, embryos were kept at 28.5°C in E3 medium (5 mM NaCl; 0.17 mM KCl; 0.33 mM CaCl₂; 0.33 mM MgSO₄; 10⁻⁵% Methylene Blue; pH 7-8) containing 0.2 mM PTU (1-phenyl-2-thiourea, Sigma).

Escape responses were elicited by a small tap to the head with a polished glass probe. The trials were recorded with a high-speed camera that captures images digitally at 1000 frames/s (EG&G Reticon, Sunnyvale, CA), and the data were analyzed as described previously (Liu and Fetcho, 1999). Turn duration and the time between turns were determined by counting frames (1 frame=1 msecond).

120 hours post fertilization (hpf), old larvae were fixed and pressure injected with DiI (Molecular Probes) dissolved in dimethylformamide (0.25%). Images were recorded using a LSM Zeiss confocal microscope.

Reticulospinal neurons were labeled retrogradely (Moens et al., 1996) with 3×10³M_r tetramethylrhodamine-dextran (Molecular Probes). Analysis of labeled reticulospinal neurons was carried out using an LSM confocal microscope (Zeiss).

Antibody staining was performed as previously described (Zeller and Granato, 1999) with few modifications. The following primary antibodies were used: 3A10 (1:50, Hatta, 1992, kindly provided by Dr T. Jessell) or zn-5 (1:500, which recognizes DM-GRASP (Fashena and Westerfield, 1999); Antibody Facility, University of Oregon). Stained embryos and larvae were viewed using Nomarski optics on a Zeiss Axioplan microscope. Images were acquired using a digital camera (Progress 3012, Kontron), saved on a Macintosh computer and processed with Adobe PhotoShop 4.0.1 software. For confocal microscopy, Alexa 488 or 594-conjugated secondary antibodies were used (Molecular Probes).

120 hpf larvae were anesthetized in 0.04% tricaine (3-aminobenzoic acid ethyl ester, Sigma) and positioned with the dorsal side upwards into rectangular wells made with 1.2% agarose in 0.3× phosphate-buffered saline (PBS) on a glass slide using a plastic mold. Using a sharpened tungsten needle (tip diameter ∼1 μm, Fine Science Tools), different brain regions were separated along or across the midline. The anterior and the posterior extent of the otic vesicle, as well as the position of the two otoliths within the otic vesicle were used as landmarks. Operated larvae were transferred to Ringer’s solution and their swimming patterns were analyzed 2 to 16 hours later. The larvae were fixed in 4% paraformaldehyde and stained with 3A10 to assess the precise extent of the cuts.

A solution of 25% calcium green dextran (3×10³M_r CGD, Molecular Probes, in 10% Hank’s) was pressure injected into the ventral spinal cord of 4 dpf larvae to label the Mauthner soma. Larvae with CGD-labeled Mauthner cells were identified 24 hours later using a fluorescence dissecting scope, and neurobiotin (Vector Laboratories, 10% in 10% Hank’s) was pressure injected into one labeled Mauthner soma. Successful targeting of the Mauthner soma resulted in a CGD-mediated increase in fluorescence. After 12 to 16 hours, neurobiotin-labeled cells were visualized using either the ABC Vectastain kit combined with nickel intensified DAB (Vector Laboratories), or with the TSA fluorescein system (NEN). Control injections of neurobiotin adjacent to the Mauthner soma did not label neurons away from the injection site.

Mutant space cadet larvae develop characteristic motor defects around 96 hours post fertilization (hpf), when mature locomotor behaviors start. During the first 96 hours of development, zebrafish embryos and larvae exhibit increasingly complex locomotor behaviors, beginning with spontaneous, alternating contractions (17-21 hpf), then touch response-induced contractions (21-28 hpf) and touch-induced swimming (28 hpf-96 hpf; Eaton and Farley, 1973; Eaton et al., 1977; Granato et al., 1996; Kimmel et al., 1974; Saint-Amant and Drapeau, 1998). At 96 hpf, wild-type larvae become very active and display two patterns of more mature motor behavior, similar to those observed in adult fish. First, tactile or vibrational stimuli elicit a fast escape response – a maneuver that occurs when the fish are confronted with a sudden aversive stimulus, such as a predator (Eaton et al., 1977). When viewed from above, the larva bends its body and tail into a C-like profile, resulting in a fast displacement of the head away from the source of the stimulus (Fig. 1A1-A4). The C-like bend is followed by a counter turn – a less powerful contraction of the body and tail musculature on the opposite side (Fig. 1A5,A6). The larva then rapidly accelerates away from its initial position (Fig. 1A7-A12). The second behavior, spontaneous swimming, does not involve a C-like bend of the body, but occurs through a series of fast, bilateral tail flexures (Fig. 1A7-A12).

In space cadet larvae, all of the above behaviors appeared on time, indistinguishable from their wild-type siblings. However, we observed abnormal motor behaviors during stimulus-induced escape responses and during spontaneous swimming (Granato et al., 1996). In response to tactile or vibrational stimuli, space cadet larvae often performed multiple C-like bends towards the same side (Fig. 1B). The initial turn and incomplete counter turn (Fig. 1B1-B6) are followed by a second large turn towards the same side (Fig. 1B7,B8). As a result, space cadet larvae rotate circumferentially, rather than performing a stereotypic 180° escape response. Similarly, in place of fast, alternating tail flexures that generate the normal swimming motion, we observed aberrant motor episodes characterized by repeated tail flexures to the same side (Fig. 1C). Both, aberrant escape responses and swimming can first be detected at 72 hpf in about 20% of the mutant larvae, and by 96 hpf, 100% of the mutant larvae exhibit these defects. Thus, among larvae derived from two heterozygous space cadet fish, 25% displayed aberrant escape responses, suggesting that the swimming phenotype is 100% penetrant. It is also worth noting that individual mutant larvae performed multiple C-like bends to the left- and to the right-hand side, indicating that there is no obvious preference towards which side aberrant flexures occur.

To analyze the behavioral defects in more detail, we used a high-speed camera (1,000 frames/second) and recorded locomotor patterns of individual space cadet larvae and their wild-type siblings. We focused our analysis on stimulus induced escape responses, because their kinematics have been described in great detail (Eaton et al., 1977; Liu and Fetcho, 1999). For both space cadet and wild-type larvae, we compared the angular velocity and duration of the initial turn (Fig. 1A1-A4,B1-B4), and the angular velocity of the counter turn (Fig. 1A5,A6,B5,B6). In cases where there were successive turns towards the same side (Fig. 1B7,B8), we examined the angular velocity of the second turn and the interval from the peak of the initial turn to initiation of the second turn. We concentrated on the angular velocities because these parameters are most indicative of the high-performance turns exhibited in normal escape behaviors (Liu and Fetcho, 1999). The results are summarized in Table 1. In wild-type siblings, all trials (n=60 from six animals) yielded a stereotypic escape response with kinematic parameters comparable with values previously determined for wild-type larvae (Liu and Fetcho, 1999). In contrast, space cadet mutant larvae exhibited three types of responses (63 trials from nine animals, Table 1). In 70% of the trials, space cadet larvae responded with a single turn escape response, with similar kinematic parameters to those exhibited by their wild-type cohorts. Although the angular velocity was slightly slower than in wild-type larvae (19.57±0.7°/mseconds in space cadet larvae versus 21.89±0.4°/mseconds in wild-type), these values are within the range of normal escape responses. In 30% of the trials, space cadet larvae responded with double turns – two successive large turns in the same direction. In 78% of these double turns, performance was markedly reduced, as evidenced by a decrease in the angular velocities of the initial turn, counter turn and the second turn. In contrast, in 22% of the double turn episodes, both turns were high-speed with kinematic parameters similar to those of wild-type escape responses (19.78±0.5°/mseconds for the first turn and 23.43±0.4°/mseconds for the second turn). Thus, the ability of space cadet larvae to perform wild-type-like escape responses indicates that mutations in the space cadet gene may not affect the execution of the escape response but may regulate the occurrence of high-speed turning movements.

The observation that space cadet larvae produce aberrant turning movements suggested to us a defect in the neural circuits controlling the escape response. In teleost fish and amphibians, the escape response is controlled by the well-studied network of the hindbrain Mauthner cell (reviewed by Faber et al., 1989; Zottoli and Faber, 2000). We therefore used antibody staining and axonal tracing techniques to examine if brain trajectories, including those of the Mauthner cell circuits, were misguided or absent in space cadet mutants. Immunohistochemical analysis using several antibodies (3A10, zn5 and anti-acetylated tubulin) revealed that major axonal trajectories in the CNS of space cadet larvae were unaffected (Fig. 2A,B and data not shown), while a small subset of hindbrain commissures was strongly affected (Fig. 2C,D).

Using the 3A10 antibody, we detected three defects in the larval space cadet hindbrain. First, a ladder-like array of commissures in the caudal hindbrain was affected. In wild-type larvae this array consists of seven commissures, the identity of which is unclear (Fig. 2C). In space cadet larvae, the first six commissures were present but in a small fraction of mutants appeared less well organized (asterisks in Fig. 2D), while the caudal most commissure was absent in 35% of the mutants examined (n=46). Second, in 100% of space cadet mutant larvae the Mauthner axon cap was affected. In wild-type larvae the 3A10 antibody stains the Mauthner soma, the axon and a structure surrounding the initial axonal segment, called the Mauthner axon cap (arrowheads in Fig. 2C,E). The structure and function of the Mauthner axon cap has been studied extensively in adult goldfish (for a review, see Zottoli and Faber, 2000) and to some extent also in zebrafish (Eaton et al., 1977; Hatta and Korn, 1998; Kimmel et al., 1981). In 120 hpf wild-type zebrafish larvae, the axon cap is composed of glial cells surrounding a complicated neuropil that contains the terminal endings of several neurons, including spiral fiber neurons (Eaton et al., 1977; Kimmel et al., 1981). In space cadet mutant larvae, the 3A10 antibody failed to stain the Mauthner axon cap (arrowheads in Fig. 2D,F). Since this antibody recognizes an axon-associated antigen (Serafini et al., 1996), the lack of immunoreactivity in space cadet axon caps suggests that terminal endings of spiral fiber neurons and/or other neurons are reduced or absent. The third defect we found was that in all space cadet mutants analyzed (n=150), two prominent commissures in rhombomere 3 were strongly reduced or absent (arrows in Fig. 2F). The 3A10 antibody staining of these commissures is restricted to the axons and therefore did not reveal the number and location of their somata. In summary, our analysis reveals that in all space cadet mutants a very small and specific set of hindbrain commissures is affected, including two prominent commissures in rhombomere 3 (Fig. 2F).

To investigate if mutations in the space cadet gene affect axonal trajectories throughout the nervous system or more specifically a defined subset of hindbrain commissural axons, we examined various axonal trajectories throughout the nervous system. We focused on the hindbrain, where an exceptionally detailed map of the reticulospinal neurons is available (Metcalfe et al., 1986). We labeled reticulospinal neurons retrogradely with rhodamine dextran, and at 120 hpf examined the trajectories of 12 classes of reticulospinal neurons, of which five extend axons contralaterally (see Table 2). Analysis of reticulospinal neurons in space cadet larvae (n=41) did not reveal any significant differences in cell number or axonal trajectories, when compared with wild-type larvae (see Table 2 and Fig. 2G,H). Similarly, 3A10-positive commissural axons of presumptive CoPA interneurons in the spinal cord (n=30; data not shown) and 3A10 positive trajectories in the midbrain of space cadet mutants (n=200, Fig. 2J) were indistinguishable from those in wild-type larvae (n=50, Fig. 2I). We conclude that the space cadet gene, rather than being an essential component for the development of many commissural trajectories, exhibits a function that is specific for a selective set of commissural axons, including two rhombomere 3 commissures.

We reasoned that if the swimming phenotype was related to the absence of the rhombomere 3 commissures, then in wild-type larvae they should develop between 72 and 96 hpf, when space cadet swimming deficits become apparent. Alternatively, if the absence of these commissures was unrelated to the swimming defect, then the onset of the swimming phenotype might differ from the time period when these commissures develop. To distinguish between these two possibilities, we examined the development of the two commissures in wild-type and space cadet mutant larvae. In wild-type larvae the two commissures became detectable around 72 hpf (Fig. 3A), and over the next 48 hours the thickness of both tracts increased dramatically (Fig. 3C,E). In contrast, in space cadet mutants these two commissures were strongly reduced or absent at 72 hpf, suggesting that space cadet plays an essential role during the time period these commissural axons first establish a path towards their synaptic targets (Fig. 3B,D,F). Thus, formation of these rhombomere 3 commissures coincides with the onset of the space cadet swimming phenotype, indicating that reduction or absence of these commissures may account for the space cadet swimming phenotype.

To determine if the lack of the two rhombomere 3 commissures in space cadet larvae can account for their aberrant swimming, we performed a series of surgical lesion experiments. Commissural tracts in different regions of wild-type larval brains were severed using a sharpened tungsten needle. The lesions were performed at 96 hpf, and the swimming patterns of the operated larvae were observed over the next 12 hours. The results of these experiments are summarized in Table 3. Separating the midbrain along the midline resulted mostly in larvae with wild-type-like escape and swimming movements (83%, Fig. 4A, Table 3). In contrast, separating the hindbrain along its entire length resulted in a series of rapid, unilateral tail flexures, characteristic of the space cadet motor behavior (84%, Fig. 4B, Table 3). Moreover, when we restricted the lesion to a region encompassing only rhombomere 3 (red line in Fig. 4C), we mostly observed a space cadet-like swimming phenotype, showing that the absence of these two commissures in space cadet larvae can account for their aberrant swimming patterns (57%, Table 3).

We had observed that the two rhombomere 3 commissures were continuous with axonal tracts extending caudally towards the Mauthner cell. To test if these longitudinal tracts included the axons critical for alternating swimming movements, we transected the nervous system between rhombomere 3 and the Mauthner cell (green line Fig. 4C). In the majority of the larvae (61%) we observed a space cadet-like swimming phenotype, consistent with the idea that the axons crossing the midline in rhombomere 3 extend towards the Mauthner cell. To confirm that we had severed all commissures in the rhombomere 3 region, but only in this region, we stained the hindbrain of the operated larvae using the 3A10 antibody. In larvae that displayed space cadet-like swimming after severing commissures only in rhombomere 3, Mauthner cell axons crossing the midline in the adjacent rhombomere were unaffected (compare Fig. 4D with 4E). In summary, our lesion experiments demonstrate that the circuits controlling alternating tail movements include two rhombomere 3 commissures, and that their absence can account for the space cadet swimming phenotype.

We next examined the identity of the neurons whose axons cross the midline in the two rhombomere 3 commissures. Two observations suggested that these axons might synapse on the Mauthner neuron within the region of the axon cap. First, in wild-type larvae, development of the two commissures in rhombomere 3 temporally correlated with development of 3A10 reactivity in the Mauthner axon cap (Fig. 3). Second, in space cadet mutants both the two rhombomere 3 commissures and 3A10 reactivity in the axon cap region were reduced or abolished (Fig. 2). We therefore asked which neurons are located rostral to the Mauthner neuron and extend commissural axons synapsing on the Mauthner neuron within the region of the axon cap. Studies in the adult goldfish have identified three different classes of neurons that form synapses within this region: commissural PHP neurons, collateral PHP neurons, and spiral fiber neurons (reviewed by Faber et al., 1989; Zottoli and Faber, 2000). However, only spiral fiber somata are located at a distance (300-700 μM) rostrally to the Mauthner cell (Scott et al., 1994). In adult goldfish, the soma of spiral fiber neurons are located rostral to the axon cap, from where they extend thin axons crossing the midline in a commissure at the level of the cell bodies. After crossing, goldfish spiral fiber axons project first caudally and, at the level of the Mauthner cell, laterally into the axon cap (Scott et al., 1994). Thus, based on their soma position and their commissural axonal pathways reported in adult goldfish, spiral fiber neurons appeared to us as a good candidate for the neuronal cell type affected in the hindbrain of space cadet mutants.

To determine if spiral fiber neurons are affected in space cadet mutants, we sought to visualize these neurons and their axonal trajectories. In zebrafish larvae, only the characteristic spiraling endings of spiral fiber axons have been identified (Eaton et al., 1977; Kimmel et al., 1981). In the absence of markers that label spiral fiber axons and somata, we took advantage of observations made in adult fish, namely that these neurons communicate with the Mauthner cell through electrical synapses containing gap junctions (Nakajima and Kohno, 1978). Mauthner cells were labeled retrogradely by injecting a fluorescent indicator, calcium green dextran (CGD), into the ventral spinal cord (Fig. 5A). Labeling the Mauthner neurons with CGD allowed us to target selectively the Mauthner soma for injection of neurobiotin, a small tracer able to pass through gap junctions (Fig. 5B; for details see Material and Methods). In wild-type larvae, neurobiotin injection in one of the two Mauthner cells labeled distinct sets of ipsilateral and contralateral neurons (Fig. 5C). The identity of most neurons located at the same level and caudally to the Mauthner cells is unclear, but these neurons might include those known to form gap junctions on the ventral Mauthner cell dendrite (Nakajima and Kohno, 1978). In 42% of the injected wild-type larvae (n=111), we observed two groups of labeled cell bodies, located rostrally and on the contralateral side of the injected Mauthner neuron (Fig. 5C,E). Each of the two groups contained 10-14 labeled cell bodies, extending thin axons that cross the midline at the level of their cell bodies, thus forming two distinct commissures in rhombomere 3 (arrows in Fig. 5C,E). After crossing the midline, the two commissures fuse and project caudally to the Mauthner cell. Thus, based on their soma positions, axonal trajectories and because they are dye coupled with Mauthner neurons, we conclude that the two groups of neurobiotin labeled, rostral commissural neurons correspond to spiral fiber neurons.

To confirm that the axons of spiral fiber neurons constitute the two 3A10-positive commissures absent in space cadet mutants, we first labeled presumptive wild-type spiral fiber neurons trans-synaptically with neurobiotin and then stained with 3A10 antibody. Confocal microscopy revealed that neurobiotin-labeled spiral fiber axons cross the midline within the region of the two 3A10-positive rhombomere 3 commissures, demonstrating that these two commissures contain the axons of presumptive spiral fiber neurons (arrows in Fig. 5D). Finally, we used the trans-synaptic labeling strategy to examine spiral fiber neurons in space cadet larvae (n=96). Both at the same level and caudal to the Mauthner cells, we observed trans-synaptically labeled neurons with a distribution similar to wild-type larvae (Fig. 5F). By contrast, none of the 96 space cadet larvae examined contained labeled cell bodies that were consistent with the position and axonal trajectories of spiral fiber neurons (Fig. 5H). Thus, in space cadet mutants, spiral fiber neurons fail to form gap junction synapses with the Mauthner neuron, consistent with our results that in space cadet larvae, 3A10-positive spiral fiber axons are strongly reduced or absent.

Our analysis shows that in space cadet mutants, two rhombomere 3 commissures, which contain spiral fiber axons, fail to develop (Fig. 3). This axonal defect suggested to us that the space cadet gene might play a role in axonal pathfinding. However, the 3A10 antibody staining of spiral fiber commissures was restricted to the axons and therefore did not reveal the number and location of their somata (Fig. 2). In the absence of additional spiral fiber specific molecular markers, we cannot exclude the possibility that the axonal defects arise as a consequence of an ‘earlier’ defect, e.g. a defect of spiral fiber neuron survival or specification. Our initial analysis of the space cadet nervous system had revealed axonal defects in two classes of neurons: hindbrain spiral fiber neurons described above and retinal ganglion cell (RGC) neurons. Consequently, we examined the development of RGC neurons to determine if the space cadet gene plays a role primarily in neural survival, neural specification, axonal outgrowth or axonal pathfinding.

DiI labeling of space cadet RGC axons revealed striking pathfinding defects. In wild-type larvae, RGC axons exit the eye at about 34 hpf (Stuermer, 1988). Axons from both eyes cross each other at the ventral midline of the diencephalon to form the optic chiasm before projecting to the contralateral optic tectum. Upon reaching the contralateral tectal lobe, retinal axons project topographically to form a precise map of the visual world within the brain. At 120 hpf, retinotectal trajectories can be readily visualized in fixed larvae by DiI labeling. In all wild-type larvae examined (n=350), all RGC axons crossed the midline and projected to the contralateral tectum (Fig. 6A). In space cadet larvae (n=456), retinal ganglion axons displayed three pathfinding errors which varied in their penetrance. In space cadet larvae, RGC axons either failed to exit the eye (0-20%, 10% average; data not shown), exited the eye but stalled at or near the midline (0-20%, 10% average; Fig. 6C), or projected aberrantly to the ipsilateral tectum (20-64%, 40% average; Fig. 6B). Thus, in a significant fraction of larvae, RGC axons projected bilaterally, similar to the phenotype observed in a small group of zebrafish mutants with retinotectal pathfinding defects (Class I; Karlstrom et al., 1996). To determine if the RGC axonal phenotypes were caused by lack of RGC specification or by progressive RGC degeneration, we examined the well-defined histology of the retina as well as the differentiation of RGCs in space cadet mutants (Dowling, 1987; Larison and Bremiller, 1990; Trevarrow et al., 1990). Histological sections of space cadet larvae (Fig. 6E) revealed that the organization of the different retinal layers as well as the cellular morphology within the RGC layer was indistinguishable from those observed in the wild-type retinae (6D). Moreover, the expression of DM-GRASP protein in space cadet retinal ganglion cells (Fig. 6F) was indistinguishable from the pattern and levels observed in wild-type retinae (Fig. 6G). In summary, our analysis indicates that mutations in the space cadet gene do not overtly affect RGC specification or survival, but suggests a role for space cadet in axonal pathfinding.

Our results suggest that defects in spiral fiber neurons account for the space cadet locomotor phenotype and that spiral fiber neurons play a vital role in controlling fast turning movements. These conclusions are supported by several lines of evidence, including high-speed kinematic analysis, space cadet spiral fiber defects and surgical lesion experiments. In addition, we demonstrate that retinal ganglion cell axons display striking pathfinding errors, providing compelling evidence for a role of the space cadet gene in spiral fiber axon pathfinding. Thus, our work provides direct functional evidence for the role of spiral fiber neurons, and identifies space cadet as a pathfinding gene required to integrate spiral fiber neurons into the circuits that underlie fast turning movements.

In space cadet mutants, aberrant turning movements correlate with axonal defects of spiral fiber neurons. Spiral fiber neurons were first described in adult catfish in 1915 by Barthelmez (Barthelmez, 1915), and later in adult goldfish by Bodian, as fibers forming ‘irregular spirals around the axon neck’ (Bodian, 1937). Histological characteristics of spiral fiber neurons include (1) somata located rostrally to the Mauthner cell, (2) thin, commissural axons, (3) projections into the region of the axon cap where they spiral around the Mauthner axon, and (4) gap junctions with the Mauthner axon (Scott et al., 1994). Our trans-synaptic labeling approach visualizes several populations of neurons that are synaptically connected via gap junctions with the Mauthner neuron. Two groups of these synaptically coupled neurons display histological characteristics consistent with them being spiral fiber neurons (Fig. 6).

Our analysis of the space cadet phenotype reveals a role for spiral fiber neurons in controlling fast turning movements. Three lines of evidence suggest that the space cadet swimming phenotype is caused by defective spiral fiber neurons. First, we find that in 100% of space cadet mutants two 3A10-positive rhombomere 3 commissures are absent (Fig. 2F), or project ipsilaterally rather than crossing the midline (K. L. and M. G., unpublished). Second, severing these two commissures in wild-type larvae evokes unilateral tail flips, similar to those observed in space cadet mutants. Third, the two 3A10-positive rhombomere 3 commissures contain axons of spiral fiber neurons, and in space cadet mutants these axons fail to make synaptic contacts with their target, the Mauthner neuron. Although we examined in detail many identified cell types in the hindbrain (Fig. 2, Table 2), it is possible that the 3A10-positive commissures in rhombomere 3 contain axons of a second, unidentified class of neurons, and that defects in these neurons elicit the space cadet phenotype. In summary, our results demonstrate that in space cadet larvae spiral fiber axons are affected, and that the temporal onset and spatial localization of this defect correlates very strongly with aberrant swimming patterns, suggesting that indeed defective spiral fiber trajectories are responsible for the space cadet swimming phenotype.

Studies in adult goldfish have implicated spiral fiber neurons as part of the presynaptic neural circuits controlling the excitability of the Mauthner cell (Scott et al., 1994). The Mauthner cell in teleost fish and amphibians is part of a ‘brainstem escape network’, which acts as a sensory integration system receiving input from many sensory systems to generate high-speed turning movements in response to aversive stimuli (Fig. 7A; Eaton et al., 1991). Afferent sensory projections to the Mauthner cell arise from several systems, including the vestibular system, through VIII nerve axons (Faber et al., 1989). The excitability of the Mauthner cell is thought to be controlled by a presynaptic inhibitory network including commissural and collateral PHP neurons (Fig. 7A; reviewed by Faber et al., 1989; Zottoli and Faber, 2000). Both groups of neurons form synapses on the Mauthner cell within a highly specialized region, the axon cap, where they produce electrical and chemical inhibition. A third class of neurons known to develop synapses within the region of the axon cap are the spiral fiber neurons (Fig. 7A; Nakajima and Kohno, 1978). Initially hypothesized to inhibit the Mauthner cell electrically, more recent, electrophysiological studies in adult goldfish suggest that spiral fiber neurons mediate directly excitation of the Mauthner neuron (Scott et al., 1994). In addition, spiral fibers have been proposed to activate PHP neurons, thereby providing indirectly inhibition to the Mauthner cell (Fig. 7A; Scott et al., 1994). Therefore, spiral fiber neurons have been identified as prime candidates in determining sensitivity of the Mauthner neuron to activation (Scott et al., 1994).

Based on the axonal and locomotor defects observed in space cadet mutants, it is tempting to speculate that spiral fiber neurons act in two ways to modulate the excitability of their postsynaptic targets. First, as suggested by work in goldfish, spiral fiber neurons play a vital role in determining whether the Mauthner neuron fires (Scott et al., 1994). In space cadet mutants, spiral fiber axons fail to connect to the Mauthner cell, and the absence of spiral fiber-mediated activation of the Mauthner neuron (at the axon cap) may reduce its overall excitability (Fig. 7B). Consistent with this notion, in 25% of the behavioral trials, space cadet larvae displayed slow double turns, as evidenced by a decrease in the angular velocities of the turns and counter turns (Table 1). Such slow turning movements are observed when the Mauthner cell and its two functional homologues are ablated (Liu and Fetcho, 1999), suggesting that during the slow turns observed in space cadet larvae, the circuits that mediate high speed turning are not activated.

Our analysis provides evidence for a second function of spiral fiber neurons, which is to ensure that high-speed turning movements do not occur spontaneously or at low stimulus levels, but only in response to strong stimuli, such as the appearance of predators. In wild-type fish, sensory inputs, e.g. from the vestibular system through the VIIIth nerve, activate the Mauthner cell and, similar to the situation in goldfish, may also activate PHP neurons, which in turn inhibit the Mauthner cell (Fig. 7A; Faber and Korn, 1978). The consequence of this arrangement is that at weak stimulus strengths, PHP-mediated inhibition dominates, but at higher sensory input strength, this inhibition is overcome (Faber and Korn, 1978). Because of their axonal defects, space cadet spiral fiber axons might fail to synapse and activate PHP neurons (Fig. 7B), thereby decreasing the overall levels of PHP-mediated inhibition impinging onto the Mauthner cell. The reduced levels of inhibition may allow weak sensory stimuli to activate the Mauthner cell (Fig. 7B), thereby causing the spurious turning movements we observe in space cadet mutants. Thus, we propose that spiral fiber neurons play an essential role modulating the excitability of postsynaptic targets, i.e. the Mauthner cell.

Our analysis does not allow us to distinguish whether spiral fiber neurons control turning movements exclusively through the Mauthner cell, or they influence additional hindbrain neurons. Laser ablation experiments have demonstrated that high-speed escape responses are controlled by at least three paired reticulospinal neurons, the Mauthner cells and two additional pairs of hindbrain neurons, MiD2cm and MiD3cm (Liu and Fetcho, 1999). It is therefore conceivable that spiral fiber neurons influence multiple hindbrain neurons, and that the space cadet swimming phenotype might involve more neurons than the Mauthner cell (Fig. 7). Consistent with this view is the decrease in angular velocity (15.3°/mseconds slow double turns, Table 1), which also occurs after killing Mauthner, MiD2cm and MiD3cm (13-14°/mseconds), but not Mauthner alone (23°/mseconds; Fig. 6, Liu and Fetcho, 1999). Furthermore, we find that in wild-type larvae, some 3A10-positive spiral fiber axons extend caudally from the Mauthner cell, providing further evidence that spiral fiber neurons influence additional synaptic targets (K. L. and M. G., unpublished).

The CNS defects observed in mutant larvae suggests a role for the space cadet gene in axonal pathfinding. In space cadet mutants spiral fiber axons are absent (Fig. 2), or project ipsilaterally rather than crossing the midline (K. L. and M. G., unpublished). Similarly, retinal ganglion cell axons fail to exit the eye, stall around the midline, or select an inappropriate path at the midline (data not shown; Fig. 6B,C), demonstrating that mutations in the space cadet gene affect pathfinding of a small set of commissural neurons. In recent years several components of a sophisticated system guiding commissural axons towards and across the CNS midline have been identified (reviewed in: Terman and Kolodkin, 1999; Van Vactor and Flanagan, 1999). Mutations in any of these midline guidance genes, such as unc-6/netrin, unc-40/DCC, commissureless, robo and slit, affect axonal trajectories of many if not most commissural neurons (Hedgecock et al., 1990; Rothberg et al., 1990; Seeger et al., 1993). In contrast, our analyses using three different antibodies (zn-5, anti-acetylated tubulin and 3A10) reveal that most commissural trajectories in the space cadet forebrain, midbrain, hindbrain and spinal cord appear unaffected (Fig. 2 and data not shown). Thus, different from the phenotypes observed in midline gene mutants, our analysis of the space cadet phenotype reveals axonal defects in only a small set of commissural neurons, including retinal ganglion cells and spiral fiber neurons. Several explanations can account for a selective axonal phenotype. For example, space cadet activity might be necessary for pathfinding of many commissural trajectories, but the two available space cadet mutations represent only partial loss-of-function alleles. Alternatively, other genes with similar function may partially substitute for the absence of space cadet activity. Although we cannot exclude that the space cadet gene product still retains some of its activity, or that other genes partially compensate for space cadet, we favor the idea that space cadet function is dedicated to pathfinding of ‘late’ developing commissures.

The commissural neuronal cell types affected in space cadet mutants share two features. First, axonogenesis of RGC and spiral fiber neurons occurs late in development. Most commissural trajectories, including those of commissural spinal interneurons, commissural reticulospinal neurons and those of commissural neurons projecting in major axonal tracts of the brain develop between 17 and 28 hpf (Bernhardt et al., 1990; Chitnis and Kuwada, 1990; Wilson et al., 1990). In contrast, retinal ganglion cell axons leave the eye only around 32-34 hpf (Stuermer, 1988), while 3A10-positive spiral fiber neurons begin axonogenesis around 72 hpf (Fig. 3). Second, RGC and spiral fiber growth cones, rather than extending on existing trajectories, pioneer a new path. In the zebrafish, RGC growth cones establish their own commissural path adjacent to, but separate from, the postoptic commissure (Wilson et al., 1990). As determined by 3A10 immunoreactivity, presumptive spiral fiber axons do not follow any existing hindbrain commissural tracts, but appear to establish new paths across the midline (Fig. 3). Thus, space cadet might not be part of the well-studied midline guidance system, but might play a role for pathfinding of a small set of commissural neurons extending their axons late. Consistent with this notion, we have mapped space cadet to a small genomic region, in which no midline guidance genes have been reported (K. L. and M. G., unpublished). In summary, our data provide compelling evidence that the space cadet gene plays an essential role for RGC and spiral fiber growth cones to pioneer late commissural pathways.

Our studies provide direct evidence that spiral fiber neurons, first described more than 80 years ago, play a vital role in the circuits controlling high-speed turning movements. To our knowledge, the analysis of the space cadet mutant phenotype is one of the few examples in vertebrates that provides compelling evidence for the functional role of a defined neuronal cell type for a well-studied behavior. Future studies are required to elucidate the precise mechanisms by which spiral fiber neurons contribute to the circuits underlying high-speed turning movements. Similarly, studies to uncover the molecular identity of the space cadet gene are underway to reveal the mechanism by which space cadet activity guides retinal ganglion cell and spiral fiber axons.

We thank the following colleagues: Dr Tom Jessell for 3A10 antiserum; Steve DiNardo, Gerald Downes, Mary Mullins, Peter Sterling, Roland Dosch, Julie Waterbury, Jing Zhang and Jörg Zeller for critical comments; Steve Zottoli for helpful discussions; and Helena Robinson-Bey for care of the fish. This work was supported by grants from the Belgian American Educational Foundation to K. L. and from the March of Dimes Birth Defects Foundation (5FY97-0722) to M. G.