Motoneuronal development in the embryonic zebrafish

Proper nervous system function requires that neurons make synaptic contacts with only the appropriate target cells. How this is accomplished is a central question in developmental neuroscience. At least three steps are likely to be crucial in the process of proper synapse formation. First, neurons must be committed to be a particular type, second they must select a pathway that will lead toward a region of the body containing an appropriate target, and finally, they must be able to form synapses with the correct type and number of cells (reviewed in Westerfield and Eisen, 1988). Many types of neurons develop cell-type specific morphologies in culture even when they are plated before axogenesis (see Banker and Waxman, 1988), suggesting that they may also be committed to select specific pathways or innervate particular targets. This idea is reinforced by studies in which changes in the expression of specific genes can switch the pathway selected by an identified neuron (Doe et al. 1988a,b). En route to their targets, navigating growth cones appear to interact with cells that influence the pathways they select (Lance-Jones and Landmesser, 1981; Goodman et al. 1982; Westerfield and Eisen, 1988). Early-developing pioneer neurons whose growth cones establish the first axonal pathways have been shown to interact with nonneuronal cells (Singer et al. 1979; Ho et al. 1983; Bastiani and Goodman, 1986; Kuwada, 1986) or undifferentiated neurons (Bate, 1976; Bentley and Caudy, 1983a,6) whose absence can prevent or alter the formation of specific axonal pathways (Silver et al. 1982; Ball et al. 1985; Bastiani and Goodman, 1986; Jellies and Kristan, 1988). The growth cones of later-developing neurons navigate through territories in which numerous axonal tracts may already be present. Apparently, these growth cones select specific axonal tracts on which to navigate; these tracts may be required for accurate pathway navigation (Raper et al. 1984; Landmesser and Honig, 1986). Interactions between navigating growth cones and cells in their environments are likely to be mediated by appropriate temporal and spatial expression patterns of specific molecules, for example extracellular matrix (ECM) components (Sanes, 1984; Tomaselli et al. 1986), cellsurface glycoproteins (Bastiani et al. 1987; Dodd et al. 1988), and diffusible factors (Letourneau, 1978; Lumsden and Davies, 1986; Tessier-Lavigne et al. 1988).

To learn how neurons become committed to innervate particular targets and to carry out proper pathfinding, my colleagues and I have been studying the development of motoneurons that innervate the axial muscles of the zebrafish. This system is well-suited for studies of neuronal differentiation, since the early embryo has relatively few neurons, many of which can be recognized as unique individuals; in addition, the optical clarity and rapid development of the embryo allow these cells to be observed at all stages in situ, in the living embryo. Zebrafish axial muscles are segmentally arranged; each muscle segment is innervated by motoneurons whose somata are located in the ipsilateral spinal hemisegment. Most muscles of adult vertebrates are innervated by a single motor axon at a single endplate (see Jansen and Fladby, 1990 for a review). In contrast, adult zebrafish axial muscles receive polyneuronal, multiterminal innervation (Wes terfield et al. 1986). Each muscle fiber is innervated by motoneurons from two separate populations, a single, identified primary motoneuron that innervates all of the muscle fibers within a specific muscle territory and by from 1 to 3 secondary motoneurons that innervate subsets of muscle fibers. The primary motoneurons can be distinguished by their earlier birthdays, larger sizes, smaller numbers, and earlier stage of axonal outgrowth from the later developing secondary motoneurons (Kimmel and Westerfield, 1990). In addition, there are genetic differences between primary and secondary motoneurons (Grunwald et al. 1988), and the two types of neurons subserve different functions during swimming behavior (Liu and Westerfield, 1990).

This paper will review some of the recent work describing motoneuronal development in the zebrafish embryo (see also Kimmel et al. this volume). Our studies suggest that zebrafish primary motoneurons are committed to become a specific cell type before they are committed to innervate particular targets. Later, but before they extend growth cones, these cells become committed to develop specific axonal trajectories. Thus, each growth cone arrives directly in the muscle region appropriate for its adult function. Commitment is unlikely to involve interactions among the primary motoneurons, but interactions with primary motoneurons may be very important for the proper development of secondary motoneurons.

In larval and adult zebrafish, every muscle segment is innervated by 3 primary motoneurons, each of which can be individually identified by the position of its soma in the spinal cord and the region of the muscle segment it innervates (Westerfield et al. 1986). CaP, the Caudal Primary, has the most caudally located soma in each spinal hemisegment and innervates the ventral region of the muscle segment. RoP, the Rostral Primary, has the most rostrally located soma and innervates the middle region of the muscle segment. MiP, the Middle Primary, has a soma between CaP and RoP and innervates the dorsal region of the muscle segment (Fig. 1).

In contrast to the primary motoneurons, we currently know much less about the secondary motoneurons. Some secondary motoneurons become postmitotic around the end of the first day of development, and new secondary motoneurons continue to be added at least through the second day of development (Myers et al. 1986). The secondary motoneurons with the earliest birthdays appear to develop in a pattern that is similar to the pattern of the primary motoneurons; cells with somata near CaP innervate ventral muscle and cells with somata near MiP innervate dorsal muscle (S. H. Pike, E. Melancon, and J. S. Eisen, unpublished data). The territories innervated by secondary motoneurons in larval and adult fish are smaller than those innervated by the primary motoneurons (Myers, 1985; Westerfield et al. 1986).

In addition to having a segmentally repeating, stereotyped organization, axonal outgrowth by the primary motoneurons occurs in a stereotyped sequence in which CaP precedes MiP and MiP precedes RoP (Fig. 2a; Eisen et al. 1986; Myers et al. 1986). The CaP growth cone pioneers the first nerve pathway into the periphery in each myotome. The CaP growth cone extends directly ventrally from the spinal cord along the medial surface of the myotome, until it reaches the region of the nascent horizontal septum, which will later separate the dorsal and ventral muscle of the myotome, where it pauses for about an hour. After pausing, the CaP growth cone resumes extension along the medial surface of the ventral myotome. The MiP and RoP growth cones initially extend caudally within the spinal cord. When they encounter CaP, they reorient and follow the CaP axon to the nascent horizontal septum. Thus, the pathway between the spinal cord and horizontal septum is a common one, followed by the growth cones of all of the primary motoneurons. The MiP and RoP growth cones also pause in the region of the horizontal septum. However, after pausing, they take different pathways; MiP’ sprouts a collateral growth cone that extends along the medial surface of the dorsal myotome and withdraws its ventral growth cone, and RoP extends its growth cone laterally, through the horizontal septum. Thus, the region of the horizontal septum defines a ‘choice point’, from which the axons of the 3 primary motoneurons diverge along individual, cell-specific pathways.

In addition to CaP, MiP and RoP, about half the spinal hemisegments in the trunk of embryonic zebrafish contain a fourth primary motoneuron, named VaP, for Variable Primary, because it is variably present (Fig. 2b; Eisen et al. 1990). In spinal hemisegments in which VaP is present, its soma is adjacent to the CaP soma, and early in development the 2 cells cannot be distinguished. However, the VaP growth cone extends only to the horizontal septum choice point. VaPs typically die without extending their growth cones further or forming any axonal branches. A few VaPs survive longer and innervate muscle fibers between the regions innervated by MiP and RoP. Although these cells persist into larval development, they have not been followed in subsequent stages, so it is not known whether they are present in adults.

The precise temporal order of outgrowth by the primary motoneurons raises the possibility that the sequence in which they grow might determine the muscle regions they innervate. Thus, the first primary motoneuron to extend a growth cone would select the ventral pathway and innervate the ventral muscle territory, the second would select the dorsal pathway and innervate the dorsal muscle territory, and the third would select the lateral pathway and innervate the middle muscle territory. This hypothesis makes a specific prediction: the first primary motoneuron to extend an axon will select the CaP pathway, the second will select the MiP pathway, and the third will select the RoP pathway. Thus, in the absence of CaP, the MiP growth cone should extend into CaP territory, rather than into MiP territory, and in the absence of both CaP and MiP, the RoP growth cone should extend into CaP territory.

To test this idea, we ablated primary motoneurons by focusing a laser microbeam onto their somata prior to axogenesis (Eisen et al. 1989; Pike and Eisen, 1990; Eisen et al. 1990). We found that, even when we ablated all but one primary motoneuron in a spinal hemisegment, the remaining cell always extended its axon along the cell-specific pathway it would have chosen if the other primary motoneurons were still present (Fig. 2c). Thus, each primary motoneuron appears to have an independent ability to pioneer both the common part of the pathway and its cell-specific pathway, suggesting that interactions among the primary motoneurons based on their sequence of outgrowth are unlikely to be important in pathway choice. These observations raise the possibility that pathway choice is governed by other factors.

One alternative possibility is that even before axogenesis, the primary motoneurons may be committed to extend growth cones along specific pathways or to innervate particular muscle regions. Commitment might involve several different aspects of neuronal differentiation, for example, expression of a set of receptors that recognizes particular extrinsic cues that would provide guidance information to the navigating growth cones, or activation of an intrinsic program that causes the cell to develop a specific morphology, as has been described for regenerating neurons of the leech (Acklin and Nicholls, 1990). The common test for commitment is to transplant a cell to a new location and observe whether it behaves in a manner appropriate for that new location, in which case it was uncommitted when it was moved, or in a manner appropriate for its original location, in which case it was committed when it was moved.

To learn whether primary motoneurons are committed to extend axons along specific pathways before axogenesis, we transplanted primary motoneurons to new spinal cord locations just prior to the time at which they would elaborate growth cones (Eisen, 1991). A cell was considered to be committed if its axonal trajectory was appropriate for the cell’s original position, and independent of the cell’s new position; a cell was considered to be uncommitted if its axonal trajectory was appropriate for the cell’s new position. We found that all of the transplanted cells developed axonal trajectories that were appropriate for their original soma positions (Fig. 3). By the definition of commitment set forth above, these cells could be considered to be committed. However, these cells behaved in two different ways, only one of which was compatible with the definition of commitment. Many of the cells in this experiment did something quite surprising. After they were transplanted, their somata moved from the new positions in which they were placed back to their original positions. Since the somata returned to their original positions, the axonal trajectories of these cells were appropriate for both their new (final) soma positions and their original soma positions. Thus, these axonal trajectories were not independent of soma position and we could not evaluate whether the cells were committed. The second result was that some of the transplanted cells remained in their new positions. These cells developed novel morphologies in which their axonal trajectories were inappropriate for their new soma positions but were appropriate for their original soma positions. Since the axonal trajectories of these cells were independent of their soma positions, by our definition these cells were committed before they were transplanted.

What features might be important in commitment? To begin to address this question, we transplanted primary motoneurons to new locations at earlier developmental stages (Fig. 3; Eisen, 1991). Again, many of these cells failed to remain in their new positions and returned to their original positions, making it difficult to evaluate whether they were committed. However, the ones that remained where they were placed developed axonal trajectories appropriate for their new soma positions, suggesting that these cells were uncommitted at the time they were transplanted. These results suggest that the soma position somehow influences the cell’s choice of axonal trajectory. One possible mechanism is that primary motoneurons in particular spinal cord locations may express receptors that recognize cues that are localized on different cell-specific pathways. Preliminary results support this idea, since motoneurons transplanted from the CaP position in the spinal cord to the horizontal septum choice point characteristically extend axons along the normal CaP pathway, but not on the MiP or RoP pathways (Gatchalian and Eisen, 1990).

The development of at least some secondary motoneurons appears to recapitulate the development of the primary motoneurons. Secondary motoneurons whose somata are in the vicinity of the CaP soma are probably the first to extend axons out of the spinal cord. The growth cones of these cells appear to fasciculate with the axons of the primary motoneurons, and to follow them to the horizontal septum choice point. These growth cones then continue ventrally, following the CaP axon, and innervate muscle fibers in the ventral region of the myotome, although we do not yet know their final destinations. Secondary motoneurons with somata in the vicinity of the MiP soma also extend growth cones to the horizontal septum choice point. Later, these cells sprout dorsal collaterals that extend along the MiP axon into dorsal muscle. The spatiotemporal relationship between the axons of the primary motoneurons and the growth cones of the secondary motoneurons suggests that the growth cones of the secondary motoneurons might require the axons of the primary motoneurons for proper pathfinding.

To test this idea, we ablated CaP and followed the development of secondary motoneurons whose growth cones would normally follow the CaP axon (Pike et al. 1989). We found that their growth cones typically were delayed in the region of the horizontal septum choice point or slightly ventral to it, and they often formed a profusion of ectopic branches in this area. Although these results suggest that the CaP axon may be required for proper extension by the secondary motoneuron growth cones, we found that later at least some of the growth cones resumed extension and in many cases a normal looking ventral nerve was eventually formed. Thus, although the CaP axon appears to be important in facilitating outgrowth by the secondary motoneuron growth cones, it does not appear to be necessary for proper pathfinding. These results show that at least some of the secondary motoneuron growth cones can pioneer the nerve pathways, and they raise the possibility that these growth cones may be able to use the same cues normally followed by the CaP growth cone.

Our results suggest that normal development by motoneurons in the embryonic zebrafish involves a number of different steps. All of the surviving transplanted primary motoneurons developed motoneuronal morphologies and not, for example, interneuronal morphologies, although there are primary interneurons with similar soma positions (see Kuwada et al. 1990). Thus, these cells appear to be committed to become primary motoneurons before they are committed to innervate specific muscle territories. Later, each cell becomes committed to develop a specific axonal trajectory; this commitment appears to depend on the location of the cell in the spinal cord. These results suggest that there may be positional cues that somehow direct primary motoneurons to extend their growth cones along specific pathways. Thus, all of the primary motoneurons may initially be equivalent, or at least have an equivalent ability to respond to localized positional cues. The time course of acquisition of primary motoneuronal identity appears similar to that of some neurons in the grasshopper (Kuwada and Goodman, 1985). Both sets of cells appear to acquire their individual identities after they become postmitotic, but before axogenesis. Because the somata of at least some of the secondary motoneurons appear to have an arrangement that is similar to that of the primary motoneurons, they may be influenced by the same sorts of positional cues. This hypothesis could be tested by transplantation experiments, similar to those performed with the primary motoneurons.

Currently, we have little information about the nature of the putative positional cues that may be influencing the differentiation of the primary motoneurons. Since pathfinding is normal after ablation of neighboring primary motoneurons, it seems likely that the primary motoneurons themselves are not the sole providers of positional information. In addition, primary motoneurons appear to develop normally in the absence of neural activity (Westerfield et al. 1990; Liu and Westerfield, 1990), suggesting that acquisition and maintenance of identity is independent of retrograde signals transported from the muscle target to the soma via the growth cone. Analysis of the spt-l(b104) mutation (Eisen and Pike, 1991; see Kimmel et al. this volume) suggests that segmented mesoderm may be important in patterning the primary motoneurons, although its precise role remains unclear.

The pathway traversed by the growth cones of the developing motoneurons has three distinct regions, where the growth cones exhibit different types of behaviors. The proximal portion of the pathway is followed by the growth cones of all primary and secondary motoneurons, thus, the cues along this pathway region may be of a ‘general’ nature (Lewis et al. 1983). The second region of the pathway is the horizontal septum choice point. The growth cones of the primary motoneurons typically pause when they reach this location. It is not clear whether the growth cones of secondary motoneurons also pause at this position during normal development, however, they are delayed in this region for some time when the primary motoneurons are ablated. Thus, the choice point region seems to be a location where the growth cones wait for some event to occur. This event might be expression of a particular pathway cue, or expression by the motoneuron of a particular receptor. Pausing of the motoneuronal growth cones at the choice point is reminiscent of the extension of growth cones in the mammalian CNS (Dodd and Jessell, 1988) and grasshopper CNS and limbs (Raper et al. 1983; Caudy and Bentley, 1986b) where en route to their final destinations, growth cones appear to be guided by a series of intermediate targets. In the case of the mammalian CNS, it may be that contact with the intermediate target influences the expression of surface molecules by the navigating growth cone (Dodd et al. 1988). The third region of the pathway is the cell-specific portion. We postulate that each cell-specific pathway may be demarcated by a different set of cues. Preliminary experiments in which we have transplanted CaP to the choice point region as a bioassay to look for differences between the pathways suggest that this view may be correct (Gatchalian and Eisen, 1990). Thus, the final basis of cell-specific pathway selection may be the ability of individual primary motoneurons to discriminate between pathways that themselves express different sets of cues.

We still know little about the specifics of pathfinding by the secondary motoneurons. Our ablation studies suggest that the primary motoneurons are important, although not essential, for proper pathfinding by the secondary motoneurons. These observations raise the possiblity that the growth cones of the secondary motoneurons may be able to recognize and follow at least two sets of cues, one on the axons of the primary motoneurons, and one on the substratum; the latter set may be the same as that followed by the growth cones of the primary motoneurons. In the future it will be interesting to learn whether there are two different sets of cues, and how they might interact during normal pathfinding by the secondary motoneurons.

Thanks to Sue Pike, Ellie Melancon, and Christine Gatchalian for unpublished results, to Ellie Melancon, Christine Gatchalian, Charles Kimmel, Sue Pike, David Raible, and Monte Westerfield for helpful discussions and comments, and to Pat Edwards for typing. Supported by NS23915, BNS8553146, HD22486, the Procter and Gamble Company, and a Searle Scholars Award.