We have studied the pattern of axon outgrowth from the identified embryonic Drosophila motorneurons, RP1, RP3, RP4 and RP5, from the onset of axonogenesis to the time of arborization over target muscles. Lucifer Yellow was intracellularly injected into each of these neurons to obtain a detailed description of the morphology of their growth cones and of the pathways that they follow. We have divided the sequence of axon growth from these neurons into five major phases. In the first phase, the growth cone of each RP axon grows medially along its contralateral homologue along the anterior commissure. Each RP axon follows a separate path across the midline in the anterior commissure. After crossing the ventral midline, the axons wrap around specific contralateral RP somata. In the second phase, each axon grows posteriorly and dorsally down the contralateral longitudinal connective, fasciculating with the other RP axons. In the third phase, the axons turn into the intersegmental nerve via the anterior nerve root, then cross over to the segmental nerve, before contacting the external surfaces of intermediate muscles 15/16. They do not fasciculate with the pioneering aCC and RP2 axons at this time. In the fourth phase, the axons advance laterally across the ventral muscle group. During this phase, each axon extends processes over a number of inappropriate muscles as well as contacting its correct, target muscle. In the final phase, the processes to inappropriate muscles are withdrawn, generating the mature pattern of motor axon projections. There is no consistent, clear difference between the RP motorneurons in the relative timing of axon outgrowth.

We would like to understand how the specific pattern of connections between motorneurons and muscles is established during the course of embryonic development. The advanced genetic techniques available in the fruitfly Drosophila melanogaster (Rubin, 1988; Thomas and Crews, 1990) make it an attractive model organism for investigating the molecular basis of this developmental phenomenon. However, there are considerable technical difficulties associated with working at the cellular level on an embryo of the small scale of Drosophila. Consequently, only a limited amount of information is available about the cellular processes underlying the establishment of specific neuromuscular connections in this insect. This is in contrast to our detailed view of motor axon growth in the much larger locust embryo (Ball et al. 1985; Whitington, 1989; Myers et al. 1990). This lack of information at the cellular level has hampered the interpretation of genetic/molecular studies of neuromuscular specificity.

In reports of previous studies of the Drosophila embryo (Thomas et al. 1984; Canal and Ferrus, 1987; Jacobs and Goodman, 1989b), the choices of pathway made by a subset of identified motorneurons within the central nervous system (CNS) have been described. However, earlier descriptions of motor axon growth in the periphery (Hartenstein, 1988; Johansen et al. 1989) have been based on immunohistochemical techniques that stain groups of axons and do not reliably reveal the somata of these motorneurons. Thus a description of the complete sequence of axon growth from individual, identified motorneurons, from the earliest stages of axon outgrowth to arrival at the target muscle, is lacking. Detailed information of this type is needed to pinpoint the nature and degree of specificity of cues that guide the axons to their targets as well as the cellular mechanisms that generate the mature pattem of connections. Furthermore, this level of resolution would greatly assist interpretation of data such as the cellular localization of gene products or the pattem of defects seen in developmental mutants.

We have recently characterized the motorneuron population in abdominal segments A3 to A7 of the Drosophila embryo and have determined the muscle targets of those motorneurons that can be reliably recognized from their soma positions (Sink and Whitington, 1991). In the present study, we have used intracellular dye injection to determine the cellular contacts made by the growth cones of a subset of these motorneurons, neurons RP1, RP3, RP4 and RP5, from the onset of axon outgrowth to the time of arborization over their muscle targets. We have pinpointed a number of neuronal and non-neuronal cells that are likely to be involved in the guidance of these motor axons at various phases of their growth. Furthermore, we have revealed that the final specific pattern of neuromuscular connections is attained in part by the initial extension of processes over both target and nontarget muscles, followed by the retraction of inappropriate processes. A similar developmental sequence has recently been described for motorneurons in the locust embryo (Whitington, 1989; Myers et al. 1990).

Wild-type Drosophila melanogaster (Oregon-R) embryos from our laboratory colony were used in this study. The embryos were collected from yeast-pasted food jars, chemically dechorionated by agitation in a 25 % commercial bleach solution, and staged according to the morphological entena of Campos-Ortega and Hartenstein (1985). They were dissected on a clean glass slide under Drosophila saline (6 5 g NaCl, 0.14 g KC1, 0.20 g NaHCO3, 0.12 g CaCl2, 0 01g NaH2PO4 in 11 of distilled water) held in a Vaseline dam. The anterior end of the egg was cut off and the embryo was squeezed from the vitelline membrane The embryo was dissected longitudinally along the dorsal midhne with a sharpened tungsten needle, the digestive system removed and the bodywall gently flattened onto the slide.

For intracellular dye injections, the RP motorneuron somata were identified in unfixed embryos using Zeiss Nomarski optics and a Leitz water immersion objective Somata were penetrated with 30–60 MQ microelectrodes filled with a 5% Lucifer Yellow (LY) solution. The LY was iontophoretically injected with a 0 2 nA DC hyperpolarizing current for 20 s

Anti-LY immunohistochemistry was used to stain LY-injected neurons with a non-fading reaction product that could be visualized in transmitted light under Nomarski optics. Injected embryos were fixed in 4 % paraformaldehyde in Millomg’s buffer, washed in phosphate-buffered saline (PBS), and incubated overnight at 4°C in anti-LY antibody (prepared in our laboratory, according to the procedure of Taghert et al 1982) diluted 1 500 in PBS/0.4% Triton X-100/0.25% bovine serum albumin (PBT). Embryos were then washed in PBS, and incubated overnight at 4°C in horseradish peroxidase-conjugated goat anti-rabbit IgG antibody (Amersham) diluted 1.250 in PBT. The embryos were then washed in PBS, incubated for 60 min in 0.5% diaminobenzidine in PBS, and reacted with 0.006% hydrogen peroxide until the injected neurons were stained dark brown. Following a final wash in PBS, the embryos were cleared and mounted in 100% glycerol.

Preparations were examined on a Zeiss photomicroscope with Nomarski optics using a Zeiss Planapo 100 × or 63 × oil immersion objective. Stained neurons were drawn with the aid of a camera-lucida and photographed with Tech Pan film.

Muscle arborization of neurons RP1, RP3, RP4 and RP5 in the stage 16 embryo

We have previously described (Sink and Whitington, 1991) the pattern of arborizations of neurons RP1, RP3, RP4 and RP5 (hereafter called the RP neurons) on abdominal muscles in stage 16 embryos, a stage near the end of embryogenesis (hatching takes place at stage 17). The axons of RP1 and RP4 arborize along the medial face of muscle 13; the axon arborization of RP3 runs anteriorly and posteriorly along the border between muscles 6 and 7; while the axon of RP5 arborizes on muscles 13, 14, 15 and 16 (Fig. 1). For the purposes of this study, we have taken these to be the mature arborization patterns for each of these neurons. The somata of each of these neurons can be reliably recognized under Nomarski optics throughout axonogenesis. They lie in stereotypic positions with respect to the scaffold of axons formed by the transverse commissures and the longitudinal connectives in each segment. RP1 and RP4 lie dorsal to this axon scaffold, RP3 at middle levels of the scaffold and RP5 ventral to the scaffold.

Fig. 1.

Camera-lucida drawings showing arborizations of RP motorneurons over target muscles in stage 16 embryos (A) RP1 contacting muscle 13, (B) RP3 contacting muscles 6 and 7, (C) RP4 contacting muscle 13, (D) RP5 contacting muscles 13, 14, 15 and 16. Scale bar=10μm.

Fig. 1.

Camera-lucida drawings showing arborizations of RP motorneurons over target muscles in stage 16 embryos (A) RP1 contacting muscle 13, (B) RP3 contacting muscles 6 and 7, (C) RP4 contacting muscle 13, (D) RP5 contacting muscles 13, 14, 15 and 16. Scale bar=10μm.

Fig. 2 is a schematic map of the region of abdominal musculature encountered by these neurons as they grow to their target muscles. Our description of the muscle pattern accords with that provided by Crossley (1978), except that we recognize a new muscle, lying just lateral and slightly external to muscle 14. This muscle appears in Crossley’s figures but is unnumbered. A recent study (Bate, 1990) used the number 14.2 for this muscle and renumbered muscle 14 as 14.1. We prefer to use number 30 for this new muscle as Bate’s nomenclature may imply that there is a special relationship between this muscle and muscle 14.

Fig. 2.

Schematic diagram of the region of musculature encountered by the RP neurons, showing the muscles in three adjacent abdominal segments at internal, intermediate and external levels. External is closest to the cuticle. The arrow runs along the ventral midline and points anteriorly.

Fig. 2.

Schematic diagram of the region of musculature encountered by the RP neurons, showing the muscles in three adjacent abdominal segments at internal, intermediate and external levels. External is closest to the cuticle. The arrow runs along the ventral midline and points anteriorly.

General comments

The following description of the patterns of axon growth from the RP neurons is based on a total of 58 fills of RP1 in 51 embryos, 79 of RP3 in 57 embryos, 67 of RP4 in 49 embryos, and 37 of RP5 in 32 embryos. These observations were spread across the developmental period under investigation and were confined to neurons in abdominal segments A3–A7. Within an individual embryo at a given stage of development, any one of these motorneurons showed a similar morphology across these segments. There was no indication of a consistent anteroposterior sequence in axon outgrowth from segmentally homologous neurons within segments A3–A7.

The developmental sequence of axon growth from motorneurons RP1, RP3, RP4 and RP5

For the purposes of this study we have divided the sequence of axon growth from the RP motorneurons into five phases. Jacobs and Goodman (1989b) have described the pattern of axon growth for RP1 and RP3 from axonogenesis up to entry of the axon into the peripheral nerve. Our observations, obtained by intracellular LY injections, agree closely with their data, drawn from reconstructions of electron micrographs.

Growth across the midline

Axonogenesis begins in stage 12 for each of the RP neurons and follows a similar pattern in each case. A lamellipodial veil approximately 5 μm long extends from the soma towards the ganglion midline. Short filopodia (1–1.5 μm) are present along the lateral edges of the veil while one to two longer filopodia (up to 5 μm) are at the leading edge (Fig. 3A).

Fig. 3.

Camera-lucida drawings of RP motorneurons during growth across the midline and into the CLC. (A) Extension of a lamelhpodium by RP1 towards the ganglion midline. (B) Fasciculation of a pair of RP1 axons in the anterior commissure. (C) Dual fill of RP1/3 showing separate pathways taken by these axons in the anterior commissure. (D) RP1 axon entering the CLC Arrow indicates a long filopodium at the leading edge of the growth cone. (E) Branches of RP1 axon wrapping around both the medial and lateral sides of the contralateral RP1 soma, ac, anterior commissure; pc, posterior commissure; conn., connectives; CLC, contralateral longitudinal connective. Scale bar=10μm.

Fig. 3.

Camera-lucida drawings of RP motorneurons during growth across the midline and into the CLC. (A) Extension of a lamelhpodium by RP1 towards the ganglion midline. (B) Fasciculation of a pair of RP1 axons in the anterior commissure. (C) Dual fill of RP1/3 showing separate pathways taken by these axons in the anterior commissure. (D) RP1 axon entering the CLC Arrow indicates a long filopodium at the leading edge of the growth cone. (E) Branches of RP1 axon wrapping around both the medial and lateral sides of the contralateral RP1 soma, ac, anterior commissure; pc, posterior commissure; conn., connectives; CLC, contralateral longitudinal connective. Scale bar=10μm.

Slightly later, the veil thins into an axon and extends medially, crossing the embryonic midline in the process. Dual fills in the same segment show that each of the RP axons fasciculates with the axon of its contralateral homologue at this time (Fig. 3B). Each RP axon pair runs in the anterior commissure, separated from the others along both the dorsoventral and the anteroposterior planes; the axons of RP1 and RP4 lie dorsal to those of RP3 and RP5. Growth of the RP axons appears to be largely synchronous at this stage; none is obviously leading the others. This finding is based upon fills of two different RP motorneurons in the same hemisegment in the same individual (Fig. 3C). Fills of homologous neurons in different segments in the same individual reveal small differences in the extent of axon outgrowth, but these differences are not consistent and there is no apparent gradient along the anteroposterior axis (data not shown).

During stage 13, the growth cone of the RP1 motorneuron wraps around the anterior and lateral surface of its contralateral homologue and at the same time the somata move medially to take up their final positions inside the rectangular grid formed by the anterior and posterior commissures and the longitudinal connectives (Fig. 3D). The axons of RP3, RP4 and RP5 show a similar behaviour. However, while RP3’s axon wraps around its contralateral homologous soma (Fig. 6A), the axon of RP5 appears to associate with the contralateral RP3 soma and not with the contralateral RP5 soma. Filopodia from RP4’s growth cone contact the contralateral RP1 soma as well the contralateral RP4 soma. In five fills of RP1 (Fig. 3E) and four of RP3, the growth cones of these neurons also extended processes along the medial and posterior surfaces of their contralateral somata. Such branches are apparently withdrawn soon thereafter as they were never seen in later stage embryos.

Fig. 6.

Photomicrographs of RP neurons at different stages of development. (A) RP3 axon entering the CLC. The axon has grown along the anterior and lateral surface of the contralateral RP3 cell (arrow). Scale bar=5μm.

(B) RP1 axon diverging from the ISN into the SN outside the ganglion. Arrowheads indicate inadvertently filled ghal cells. Scale bar=10gm. (C) RP4 axon at target muscle 13 (arrowhead) showing processes in contact with non-target muscles 14 and 15 (arrows). Scale bar=10μm.

Fig. 6.

Photomicrographs of RP neurons at different stages of development. (A) RP3 axon entering the CLC. The axon has grown along the anterior and lateral surface of the contralateral RP3 cell (arrow). Scale bar=5μm.

(B) RP1 axon diverging from the ISN into the SN outside the ganglion. Arrowheads indicate inadvertently filled ghal cells. Scale bar=10gm. (C) RP4 axon at target muscle 13 (arrowhead) showing processes in contact with non-target muscles 14 and 15 (arrows). Scale bar=10μm.

Growth down the contralateral longitudinal connective

Contact with a contralateral RP soma brings the growth cone of each of the RP motorneurons within filopodial grasp of the contralateral longitudinal connective (CLC), which has been previously pioneered by the axons of neurons pCC, MPI, vMP2 and dMP2 (Jacobs and Goodman, 1989b); the longest filopodia span the width of the CLC at this stage. During stage 14, the axon of each of the RP neurons extends posteriorly down the CLC along a route that is oblique and sometimes not straight (Fig. 4A,B). The previously separate RP axons fasciculate within the CLC (Fig. 4A) and lie dorsally within this tract. Dual fills reveal that the RP axons have extended approximately the same distance posteriorly, although in different embryos there are variable differences in the extent of growth of these axons (Fig. 4A).

Fig. 4.

Camera-lucida drawings of RP motorneurons during axon growth into the anterior root of the ISN; along the ISN, and onto oblique intermediate muscles. (A) Dual fill of RP1 and RP5 showing fasiculation of their axons within the CLC. RP5’s growth cone (arrow) is just behind RPl’s. (B) RP1 axon turning into the anterior root (AR) of the ISN Note long filopodia in the direction of the turn. (C) RP1 axon crossing from the ISN to the SN. The subdivision of the peripheral nerve into the ISN and the SN fascicles is indicated by the dots (D) Branching of RP1 axon on the external face of the intermediate, ventral muscles 15 and 16. A short dendritic branch has extended from the soma into the ipsilateral connective (arrow) at this time. AR, anterior root of the ISN; ISN, intersegmental nerve; SN, segmental nerve. Scale bar= 10 μm

Fig. 4.

Camera-lucida drawings of RP motorneurons during axon growth into the anterior root of the ISN; along the ISN, and onto oblique intermediate muscles. (A) Dual fill of RP1 and RP5 showing fasiculation of their axons within the CLC. RP5’s growth cone (arrow) is just behind RPl’s. (B) RP1 axon turning into the anterior root (AR) of the ISN Note long filopodia in the direction of the turn. (C) RP1 axon crossing from the ISN to the SN. The subdivision of the peripheral nerve into the ISN and the SN fascicles is indicated by the dots (D) Branching of RP1 axon on the external face of the intermediate, ventral muscles 15 and 16. A short dendritic branch has extended from the soma into the ipsilateral connective (arrow) at this time. AR, anterior root of the ISN; ISN, intersegmental nerve; SN, segmental nerve. Scale bar= 10 μm

Growth to the ventral muscle group

In stage 14, the RP axons diverge from the CLC, turning laterally into and growing along the dorsal surface of the anterior root of the intersegmental nerve (ISN; Bate, 1982), which has been previously pioneered by the axon of neuron aCC. Typically, several long (up to 5μm) filopodia are directed towards the anterior root of the ISN as the growth cone turns (Fig. 4B). While the growth cone is within the nerve root, filopodia extend across its whole width. All of the RP axons show the same behaviour at this stage.

The axons of motorneurons aCC and RP2 extend into the periphery along the anterior and posterior roots of the ISN, respectively, before the axons of the other RP neurons (Jacobs and Goodman, 1989b). This makes the aCC and RP2 axons candidates for guiding the RP axons by a process of selective axon fasciculation. To test this hypothesis, we performed dual fills of the RP neurons and aCC or RP2. We observed that the RP axons do not fasciculate with aCC or RP2 within the ISN.

Once outside the ganglion, the RP axons cross over from the ISN to the segmental nerve (SN; Bate, 1982) (Figs 4C, 6B). The RP axons remain fasciculated during this stage. Shortly thereafter, early in stage 15, the RP axons extend processes that contact the external surface of the ventral muscles 15 and 16 (Fig. 4D).

At about this time, each of the RP somata sends out a collateral branch that extends into the ipsilateral longitudinal connective (Fig. 4D). This branch is consistently present later in embryonic and larval development, attaining a length of up to 5μm in late embryonic stages. Dendritic branches also start to appear within the neuropile contralateral to the soma (Fig. 4D).

Growth over the ventral muscle group

Having reached muscles 15/16, the RP axons diverge from the main SN nerve into SN branch b (SNb; Johansen et al. 1989) and contact the internal surface of muscle 28. RP5 also sends a process into nerve SN branch d (SNd; Johansen et al. 1989). Lateral to muscle 28, the RP axons grow between the most internal and its adjacent, intermediate muscle sheet (Fig. 2). While growing over the ventral muscles, the RP growth cones become more expansive and send processes over the surface of the muscles that they contact (Figs4D, 5, 6C). Some of these processes may be axonal branches rather than filopodia, since they appear to be thicker than typical filopodia, are of non-uniform width and are sometimes branched. The pattern of muscles contacted by a RP neuron in any given embryo is variable (illustrated by RP3, Fig. 5). However, examination of a number of embryos suggests that each of the RP axons is able to arborize over all or most of the following muscles; the internal surface of muscles 12, 13, 14, 15, 28 and 30 and the external surface of muscles 6 and 7.

Fig. 5.

Camera-lucida drawings of RP3 in 4 different embryos at a stage when the growth cone is contacting the target muscle 13. Note the variability in the pattern of muscles contacted and in the arborization patterns on individual muscles. Scale bar=10μm.

Fig. 5.

Camera-lucida drawings of RP3 in 4 different embryos at a stage when the growth cone is contacting the target muscle 13. Note the variability in the pattern of muscles contacted and in the arborization patterns on individual muscles. Scale bar=10μm.

The pattern of initial axon arborization on a given muscle by an individual motorneuron is both variable and more widespread than at the end of embryogenesis, as shown, for example, by the arborization of RP3 over muscles 6/7 in Fig. 5. Whereas in stage 16 the arborization of this neuron is confined to the adjoining edges of muscles 6 and 7, during stage 15 it is often found spanning the complete mediolateral extent of these muscles.

These growth trajectories bring the RP axons into the vicinity of their target muscles, 13 (RP1 and RP4), 6/7 (RP3) and 13, 14, 15 and 16 (RP 5), over which they form the stereotypic arborization patterns evident at late embryonic stages (Sink and Whitington, 1991). In some cases, RP axons continue to grow beyond their target muscles and arborize over more distal muscles after having apparently contacted their target muscle (e.g. RP3 sometimes extends processes to muscles 12/13, Fig. 5, RP1, RP4 and RP5 to muscle 12,). However, they were never observed to extend distal to muscle 12.

Retraction of inappropriate axonal branches

During the period late stage 15, the RP motorneurons have arborizations over both their target muscles and the non-target muscles, which they had contacted en route to their targets. Late in stage 15 and early in stage 16, the processes in contact with inappropriate muscles are withdrawn, leaving the mature arborization patterns (Fig. 1).

This is the first study in the Drosophila embryo to have examined the cellular contacts made by the growth cones of identified motorneurons through the whole period of their growth, from the onset of axonogenesis to the time of arborization over muscle targets. Our observations of the behaviour of the RP axons provide insights into: (a) the cellular identity of putative guidance cues for these axons; (b) the degree of specificity of these cues with respect to different RP axons; and (c) the cellular mechanisms that underlie the generation of the stereotypic axonal morphologies of mature RP motorneurons. We discuss these issues below with respect to the five phases of RP motor axon growth that we define in this study.

The earliest stage of axon outgrowth involved the formation of a lamellipodium on the medial side of the RP somata. Lefcort and Bentley (1989) have presented evidence that, in the case of the pioneer neurons in the grasshopper limb bud, the orientation of the mitotic spindle during the final cell division prior to the formation of these neurons determines the orientation of axon emergence. Whether such intrinsic factors or some external cue determines the direction in which the lamellipodium is formed in the RP motorneurons awaits further investigation.

After this early phase, the RP axons appear to use a variety of external cellular structures as guidance cues. The first cellular feature with which the RP axons associate is the axon of their contralateral homologue. Our observations, together with those of Jacobs and Goodman (1989b) on RP1 and RP3, strongly suggest that this behaviour is due to specific self-self axon recognition, since the RP growth cones are within filopodial reach of all the RP axons, and grow out at the same time, yet fasciculate selectively with their homologous partners.

In the case of RP1, RP3 and RP4, this apparent self-self recognition continues as their growth cones wrap around their contralaterally homologous somata. However, the growth cone of RP5 changes its growth substratum at this point, appearing to associate with the contralateral RP3 soma, rather than with its homologue. This suggests that different markers are present on the RP cells which are selectively recognised by the different RP growth cones. Molecules that are selectively expressed on subsets of embryonic Drosophila CNS neurons have been identified (fasciclin I, Zinn et al. 1988; fasciclin III, Patel et al. 1987); however, the effects of deleting such molecules have not been examined at the single cell level.

Contact with contralateral RP somata brings the axons of the RP neurons within filopodial reach of the CLC, down which they subsequently grow, before diverging into the anterior root of the ISN and advancing to the edge of the CNS. Other axons precede the RP axons into the CLC and ISN (Jacobs and Goodman, 1989b) and are possible candidates for guidance of the RP axons in these phases of their growth. However, dual fills show that two of these pioneers, aCC and RP2, are not followed by the RP axons. The RP axons may be guided by a trail of glial cells located on the dorsal side of the CNS, which has been suggested as the cue guiding the posteriorly directed growth of the pioneering aCC axon (Jacobs and Goodman, 1989b). The trajectory of the RP axons along the CLC is frequently not straight, and may be the result of a preferential adhesion to the adjoining edges of these glial cells. Studies in both vertebrates (Rakic, 1971; Alvarez-Buylla and Nottebohm, 1988) and invertebrates (Bastiani and Goodman, 1986; Carr and Taghert, 1988; Jacobs and Goodman, 1989a) show that glial cells can delineate pathways for both cell and axon migration. Molecules specifically assisting neuronglial cell adhesion (Ng-CAMs) have been identified in vertebrates (Grumet et al. 1984; Grumet and Edelman,1988) In Drosophila, a Ng-CAM, neuroglian, has been identified (Bieber et al. 1989) but its functional role is unclear.

The RP axons encounter a choice point at the edge of the ganglion, leaving the ISN and crossing over to the SN, before exiting the segmental nerve onto the external surfaces of ventral muscles 15 and 16. In other cases, insect motor axons are guided into the periphery by sensory axons (Hartenstein, 1988) or pioneer motor axons (Whitington, 1989). Our observations, together with the results of the anti-HRP study by Johansen et al. (1989), suggest that the RP axons are not preceded onto the ventral muscle group by either sensory axons or other motor axons. Therefore, the RP axons apparently do not exit the segmental nerve by tracking other axons.

Glial cells known to be present along the proximal region of the segmental nerve (Fredieu and Mahowald,1989) may act as guideposts, providing the cue for the RP axons to leave the nerve. The CAM neuroglian is expressed on both glia and dorsal neurons in the Drosophila embryo (Bieber et al. 1989), and could serve a role in guiding the RP axons from the nerve.

Alternatively, the RP axons may be exiting the segmental nerve in response to a chemoattractant released by the ventral muscles. Insect motor axons still exit the ganglion, albeit by different nerves, in the absence of the target muscles (Whitington et al. 1982; Whitington and Seifert, 1982; Nässel et al. 1986). Similarly, we have found that the RP3 axon still exits the segmental nerve in the absence of target muscles (Sink and Whitington, in preparation). ‘This dismisses a role for a diffusible, target-derived neurotropic factor in guiding axons from the ganglion, but it is still possible that the RP axons are guided from the nerve by such a cue from more proximal muscles.

Once in contact with the ventral muscle group, the RP growth cones became more expansive, and extend distally over the surfaces of these muscles. The pathways followed by the RP axons across the ventral muscles are not delineated by any obvious physical structures. It is likely, therefore, that the pathways are based on preferential adhesion by the RP growth cones for substrata expressed on the surface of the muscles. Preferential growth cone adhesion has been identified as a mode of pathfinding both in vitro (Letourneau, 1975) and in vivo (Berlot and Goodman, 1984; Caudy and Bentley, 1986). The extension of processes by each individual RP motorneuron over many of the muscles in the ventral group indicates the broad distribution of adhesive substrata. The same molecular markers may therefore be used to guide the different RP motoraxons to their targets. This strategy of pathway delineation reduces the number of molecular markers that are needed to obtain a stereotypic pattern of axonal arborization over closely situated muscles. Similarly the fasciculation of RP axons during growth down the CLC and into the ISN may reflect the usage of a common set of guidance cues by these neurons.

In summary, our results indicate that the RP axons use a variety of guidance cues as they grow towards their muscle targets, including guidepost cells, selectively labelled fascicles and adhesive substrata. We have recently embarked on a series of ablation experiments to test axon utilisation of individual guidance cues. These experiments will form the basis for a search for the molecules on these cues that mediate axon guidance. Several cell adhesion molecules (CAMs) have already been identified in Drosophila (fascichn III, Patel et al. 1987; neuroglian, Bieber et al. 1989 and the gene product of l(2)gl, Klämbt et al. 1989) and implicated in axon fasciculation. In mutants for these genes, the CNS tracts appear relatively normal. One interpretation of this result is that axon fasciculation at the tract level involves the use of more than one CAM. Alternatively, these mutants may possess subtle defects in axon behaviour that might only be revealed by examination at the level of identified axons, with a thorough background knowledge of normal axon behaviour such as is provided in the present study.

Our observations indicate that two cellular mechanisms underlie the generation of the stereotypic axonal morphologies of mature RP motorneurons, namely (a) refiable axon steering along appropriate pathways and (b) more diffuse initial axon branching followed by retraction of inappropriate processes.

After crossing the ventral midline, processes of the RP growth cones were rarely found on the posterior and medial sides of the contralateral RP somata. This may result from a directed, preferential growth of processes from the outset along the anterior and lateral sides of the RP somata, due to an asymmetrical distribution of molecular markers on the surfaces of these cells. Alternatively, the growth cone may initially wrap around the entire cell, but those processes on the anterior and lateral sides of the cell may contact the CLC before those on the posterior and medial sides, due to their physical proximity to that pathway. As a result, the former branches may outcompete the latter ones for a limited maintenance factor and be retained at their expense. If retraction took place quickly, branches on the medial and posterior sides would be seen less frequently than those on the lateral and anterior sides.

When growing down the CLC and into the ISN, the RP motor axons appear to take a well-defined growth trajectory from the outset, with little evidence of branching outside their mature, appropriate pathways.

Once in contact with the ventral muscle group, the RP growth cones extend processes over non-target muscles. The inappropriate processes are retracted during stage 15–16 and are rarely observed in older stage 16 embryos. Studies of motor axon outgrowth in the embryos of locusts (Myers et al. 1990) and leeches (Kuwada, 1984; Baptista and Macagno, 1988) have shown that inappropriate axon branching consistently occurs during axon outgrowth by specific embryonic motorneurons. This is contrary to the commonly held view that motor axon outgrowth is highly specific and error-free (Ball et al. 1985; Tosney and Landmesser, 1985; Eisen et al. 1986; Myers et al. 1986).

Inappropriate branches in the locust embryo rarely contact a non-target muscle (Myers et al. 1990), whereas inappropriate processes in the Drosophila embryo extend over non-target muscles. We have yet to determine if these processes form synapses with the non-target muscles. Functional connections with nontarget muscles have been observed for target-deprived locust motorneurons (Whitington, 1985), demonstrating that insect motorneurons do possess the capacity to innervate muscles other than their normal targets.

Retraction of inappropriate processes may be the result of contact with the target muscle and the subsequent stabilisation of the appropriate branches at the expense of the inappropriate processes. Baptista and Macagno (1988) found that inappropriate branches in the leech are retained in the absence of normal muscle targets. This suggests that a signal is derived from the target muscle that stabilises the contacting branch, and causes the retraction of inappropriate branches. Alternatively, competition may occur between appropriate and inappropriate axon arborisations for innervation sites on the muscles. The results of our muscle ablation experiments argue for the first of these alternatives (Sink and Whitington, in preparation).

In several cases, RP motor axons grew beyond their target muscle(s) to contact more distal muscles, even after they had apparently contacted their target muscle. This observation provides evidence against the hypothesis that contact with the target muscle immediately leads to cessation of axon growth, a conclusion that is consistent with a recent study of axon growth by the S interneuron in the leech embryo (McGlade-McCulloch and Muller, 1989). However, it does not exclude the hypothesis that target muscle contact may stop axon growth following a delay.

In relation to the establishment of axon arborization patterns within the target muscle, we have observed that the initial arborizations of RP motor axons are both variable and more widespread than at late stages of embryogenesis. This finding appears to conflict with the claim by Johansen et al. (1989) that Drosophila motorneurons establish their stereotyped arborizations without significant process pruning, a conclusion based upon a comparison of arborizations in stage 16 embryos with those in stage 17 embryos and larvae (Fig. 11, Johansen et al. 1989). However, our observations show that the RP motor axons have virtually attained their mature morphologies by stage 16; more widespread branching patterns are only seen during stage 15.

This research was supported by a grant from the Australian Research Council (PMW) and an Australian Postgraduate Research Award (HS) We thank Friedrich Bonhoeffer for use of facilities at the Max-Planck-Institut für Entwicklungsbiologie, Tübingen, Federal Republic of Germany, during part of this study.

Alvarez-Buylla
,
A
and
Nottebohm
,
F
(
1988
)
Migration of young neurons in the adult avain brain
Nature
335
,
353
354
Ball
,
E E
,
Ho
,
R. K.
and
Goodman
,
C S
(
1985
).
Development of neuromuscular specificity in the grasshopper embryo Guidance of motoneuron growth cones by muscle pioneers
J Neurosa
.
5
,
1808
1819
Baptista
,
C A
and
Macagno
,
E R
(
1988
).
Modulation of the pattern of axonal projections of a leech motor neuron by ablation or transplantation of its target
Neuron
1
,
949
962
.
Bastiani
,
M J
and
Goodman
,
C S
(
1986
)
Guidance of neuronal growth cones in the grasshopper embryo III Recognition of specific glial pathways
J Neurosa
6
,
3542
3551
Bate
,
M
(
1982
)
Pioneer neurons and axonal pathways
N R P Bull
20
,
838
843
.
Bate
,
M
(
1990
).
The embryonic development of larval muscles in Drosophila
.
Development
110
,
791
804
.
Bentley
,
D
and
Caddy
,
M
(
1983
).
Pioneer axons lose directed growth after selective killing of guidepost cells
Nature
304
,
62
65
.
Berlot
,
J
and
Goodman
,
C. S
(
1984
)
Guidance of peripheral pioneer neurons in the grasshopper Adhesive hierarchy of epithelial and neuronal surfaces
Science
223
,
493
496
.
Bieber
,
A. J
,
Snow
,
P M
,
Hortsch
,
M.
,
Patel
,
N. H
,
Jacobs
,
J R.
,
Traquina
,
Z R
,
Schilling
,
J
and
Goodman
,
C. S.
(
1989
).
Drosophila neuroghan a member of the immunoglobulin superfamily with extensive homology to the vertebrate neural cell adhesion molecule LI
Cell
59
,
447
460
.
Campos-Ortega
,
J A.
and
Hartenstein
,
V
(
1985
)
The Embryonic Development of Drosophila melanogaster Berlin Springer-Verlag
Canal
,
I
and
Ferrus
,
A
(
1986
).
The pattern of early neuronal differentiation
in
Drosophila melanogaster J Neurogenet
3
,
293
319
Carr
,
J. N
and
Taghert
,
P H.
(
1988a
)
Formation of the transverse nerve in moth embryos I. A scaffold of non-neuronal cells prefigures the nerve
Devi Biol
130
,
487
199
Caddy
,
M.
and
Bentley
,
D
(
1986
)
Pioneer growth cone steering along a senes of neuronal and non-neuronal cues of different affinities
J Neurosa
6
,
1781
1795
Crossley
,
A C
(
1978
)
The morphology and development of the Drosophila muscular system
. In
The Genetics and Biology of Drosophila
, Vol
2b
, (ed
M
Ashbumer
and
T.R.F.
Wnght
), pp
499
-
560
, Academic Press, New York.
Eisen
,
J S
,
Myers
,
P Z.
and
Westerfield
,
M
(
1986
).
Pathway selection by growth cones of identified motoneurones in live zebrafish embryos
Nature
320
,
269
271
.
Fredied
,
J R
and
Mahowald
,
A. P
(
1989
).
Glial interactions with neurons dunng Drosophila embryogenesis
Development
106
,
739
748
.
Gao
,
W-Q
and
Macagno
,
E R
(
1988
)
Axon extension and retraction by leech neurons, severing early projections to [peripheral targets prevents normal retraction of other projections
Neuron
1
,
2FB
-
2T1
Grümet
,
M.
and
Edelman
,
G. M.
(
1988
)
Neuron-gha cell adhesion molecule interacts with neurons and astrogha via different binding mechanisms
J Cell Biol
106
,
487
503
Grümet
,
M
,
Hoitman
,
S
and
Edelman
,
G M
(
1984
)
Two antigenically related neuronal CAMs of different specificities mediate neuron-neuron and neuron-gha adhesion Proc natn
.
Acad. Sa USA 81,267-271
Hartenstein
,
V
(
1988
)
Development of Drosophila larval sensory organs spatiotemporal pattern of sensory neurones, peripheral axon pathways and sensilla differentiation
Development
102
,
869
886
Jacobs
,
J. R.
and
Goodman
,
C S.
(
1989a
)
Embryonic development of axon pathways in the Drosophila CNS 1. Role of glia
J Neurosa
9
,
2401
2411
Jacobs
,
J R
and
Goodman
,
C S
(
1989b
)
Embryonic development of axon pathways in the Drosophila CNS U. Behavior of pioneer growth cones
J Neurosa
.
9
,
2412
2422
Johansen
,
J.
,
Halpern
,
M E
and
Keshishian
,
H
(
1989
).
Axonal guidance and muscle fiber-specific innervation in Drosophila embryos
.
J Neurosa
9
,
4318
4332
Klambt
,
C
,
Müller
,
S
,
Lützelschwab
,
R
,
Rossa
,
R
,
Totzke
,
F
and
Schmidt
,
O
(
1989
)
The Drosophila melanogaster l(2)gl gene encodes a protein homologous to the cadhenn celladhesion molecule family
Devi Biol
133
,
425
436
Kuwada
,
J. W
(
1984
)
Normal and abnormal development of an identified leech motor neuron
J Embryol exp Morph
79
,
125
137
Lefcort
,
F
and
Bentley
,
D
(
1989
)
Organization of cytoskeletal elements and organelles preceding growth cone emergence from an identified neuron tn situ
.
J Cell Biol
108
,
1737
1749
Letohrnead
,
P C.
(
1975
)
Cell-to-substratum adhesion and guidance of axonal elongation
Devi Biol
44
,
92
101
Mcglade-Mccdlloch
,
E.
and
Müller
,
K. J
(
1989
)
Developing axons continue to grow at their tip after synapsing with their appropriate target
Neuron
2
,
1063
1068
Myers
,
C M
,
Whttington
,
P M
and
Ball
,
E. E.
(
1990
)
Embryonic development of the innervation of the locust extensor tibiae muscle by identified neurons formation and elimination of inappropriate axon branches
Devi Biol
137
,
194
206
Myers
,
P Z.
,
Eisen
,
J S
and
Westerfield
,
M
(
1986
).
Development and axonal outgrowth of identified motomeurons in the zebrafish
J Neurosa
6
,
2278
2289
Nassel
,
D R
,
Helgee
,
A
and
Sivasubramanian
,
P
(
1986
)
Development of axon paths of motomeurons after removal of target muscles in a holometabolous insect
Devi Brain Res
26
,
211
21
Patel
,
N H
,
Snow
,
P M
and
Goodman
,
C S
(
1987
).
Characterization and cloning of Fascichn III a glycoprotein expressed on a subset of neurons and axon pathways
in
Drosophila Cell
48
,
975
988
Rakic
,
P
(
1971
).
Neuron-glia relationships dunng granule cell migration in developing cerebellar cortex A golgi and electron microscopic study
in
Macacus rhesus J comp Neurol
141
,
283
312
.
Rubin
,
G. M
(
1988
)
Drosophila melanogaster as an expienmental organism
Science
240
,
1453
1459
Sink
,
H
and
Whitington
,
P M
(
1991
)
Location and connectivity of abdominal motomeurons in the embryo and larva of Drosophila melanogaster
J Neurobiol (in press)
Taghert
,
P H
,
Bastiani
,
M. J
,
Ho
,
R K
and
Goodman
,
C S
(
1982
)
Guidance of pioneer growth cones Filopodial contacts and coupling revealed with an antibody to Lucifer Yellow
Devi Biol
94
,
391
399
Thomas
,
J B
,
Bastiani
,
M J
,
Bate
,
M
and
Goodman
,
C S
(
1984
)
From grasshopper to Drosophila a common plan for neuronal development
Nature
310
,
203
207
Thomas
,
J. B
and
Crews
,
S T
(
1990
)
Molecular genetics of neuronal development in the Drosophda embryo
FASEB J
4
,
2476
2482
Tosney
,
K W.
and
Landmesser
,
L. T
(
1985
)
Specificity of early motoneuron outgrowth in the chick embryo
J. Neurosa
5
,
2336
2344
Whitington
,
P M
(
1985
).
Functional connections with foreign muscles made by a target-deprived insect motomeuron
Devi Biol
107
,
537
540
.
Whitington
,
P M
(
1989
)
The early development of motor axon pathways in the locust embryo The establishment of the segmental nerves in the thoracic ganglia
Development
105
,
715
721
Whitington
,
P M
,
Bate
,
M
,
Seifert
,
E
,
Ridge
,
K
and
Goodman
,
C S
(
1982
)
Survival and differentiation of identified embryonic neurons in the absence of their target muscles
.
Science
215
,
973
975
Whitington
,
P M
and
Seifert
,
E
(
1982
)
Axon growth from limb motomeurons in the locust embryo: the effect of target hmb removal on the path taken out of the central nervous system
Devi Biol
93
,
206
215
Zinn
,
K
,
Mcallister
,
L
and
Goodman
,
C S.
(
1988
)
Sequence analysis and neuronal expression of fascichn 1 in grasshopper and
Drosophila Cell
53
,
577
587