Toll, a muscle cell surface molecule, locally inhibits synaptic initiation of the RP3 motoneuron growth cone in Drosophila

Neuronal growth cones in a developing brain select their targets with very high cellular resolution (Goodman, 1996; for a review). At the level of an individual synapse, the process of target recognition is manifested as a motile growth cone contacting an appropriate target cell and progressively converting itself into a more stable presynaptic terminal, fully equipped with neurotransmitter releasing machinery (Haydon and Drapeau, 1995 for a review). Recent studies have begun to characterize a variety of molecules that are provided locally and likely promote synaptogenesis of exploring growth cones. These include Neurexins (neurexin family) and N-cadherins and E-cadherins (cadherin family) at mammalian central synapses and Fasciclin III (immunoglobulin superfamily) at Drosophila neuromuscular synapses (Chiba et al., 1995; Fannon and Colman, 1996; Ulrich et al., 1995). Other cell- associated molecules, including GPI-linked T-cadherin (cadherin family) in the hamster motor system and Semaphorin I / Fasciclin IV (semaphorin family) in the grasshopper sensory system, are proposed to inhibit normal growth cone elongation (Fredette et al., 1996; Kolodkin et al., 1992). It is likely that combinations of these and other locally provided molecules ensure the high degree of cell-cell recognition required for establishing functional neuronal networks. How these local growth cone guidance molecules work in vivo is beginning to be tested at the single-cell level.

Toll is an important multifunctional developmental gene in Drosophila. It encodes a transmembrane molecule with extracellular leucine-rich repeats (LRR) and a cytoplasmic interleukin 1 receptor (IL-1R) domain (Hashimoto et al., 1988). In the early embryo, maternally supplied Toll acts as a cell membrane receptor during dorsoventral axis determination and has served as a model for studying transmembrane gene regulation (Morisato and Anderson, 1995). In contrast, later embryonic development relies on zygotic Toll expression, which is found in the gut, epidermis, nervous system and musculature; the roles of zygotic Toll, however, have received attention only recently (Halfon et al., 1995; Nose et al., 1992; Wharton and Crews, 1993). Here we focus on the role of muscular Toll expression in regulating synaptic initiation of an identified motoneuron, RP3.

The timing of Toll expression in the ventral musculature of a Drosophila embryo is dynamic. It roughly coincides with axon pathfinding and synaptogenesis of the motoneurons that innervate the ventral muscle cells (Halfon et al., 1995; Keshishian et al., 1993; Nose et al., 1992; this study). For example, the growth cone of the RP3 motoneuron extends past proximal Toll-positive muscle cells before reaching its more distal targets, muscle cells 6 and 7. However, it initiates synaptogenesis on its target muscle cells only after Toll expression on them is down-regulated. The RP3 growth cone likely responds positively to Fasciclin III, an immunoglobulin-like cell adhesion molecule expressed on the target muscle cells, but still manages to avoid targeting errors in embryos lacking Fasciclin III (Chiba et al., 1995). Therefore it appears that multiple mechanisms can instruct the RP3 growth cone to promote synaptogenesis at appropriate sites and/or discourage innervation at inappropriate sites. While no neurons express Toll, the specific expression pattern of Toll in the ventral musculature suggests that a role of Toll may be to inhibit synaptogenesis of specific motoneurons (e.g. RP3) on inappropriate muscle cells.

The molecular structure of Toll suggests that this cell surface molecule may mediate growth cone-muscle interaction. A 125×10³M_r (1097 amino acids) single-pass membrane glycoprotein with several distinct domains, Toll could potentially mediate various intercellular and cytoplasmic interactions (Hashimoto et al., 1988, 1991; Keith and Gay, 1990; Mitcham et al., 1996; Sims et al., 1989; Yamagata et al., 1994). There is good evidence that the external portion of Toll contains multiple N-glycosylation sites as well as intramolecular disulfide bonds (Hashimoto et al., 1991). Most notably, Toll has extracellular regions of 24-amino acid tandem repeats (a total of 15) with high leucine concentration (i.e. leucine-rich repeats, or LRR). This LRR motif is shared by a number of other Drosophila cell surface molecules, e.g. Connectin, Chaoptin, 18-wheeler, Kekkon and Toll-like, as well as mammalian neural receptors, e.g. NLRR-1 and GARP (Chiang and Beachy, 1994; Eldon et al., 1994; Keith and Gay, 1990; Musacchio and Perrimon, 1996; Nose et al., 1992; Ollendorff et al., 1994; Reinke et al., 1988; Taguchi et al., 1996). Structural analyses indicate that the LRR motifs could mediate protein-lipid as well as homophilic and heterophilic protein-protein interactions (Kobe and Deisenhofer, 1995; Krantz et al., 1991). Nose et al. (1994) have demonstrated that the structurally related Connectin protein, when ectopically expressed in some of the ventral muscle cells, can function as a repulsive signal to motoneuron growth cones. All these provide circumstantial evidence suggesting that Toll, being expressed at specific muscle cell surfaces, may act locally to inhibit synaptogenesis of specific motoneuron growth cones on inappropriate muscle cells.

In this paper, we have combined genetic manipulation and single-cell visualization, demonstrating that the timing and cell specificity of muscular Toll expression can affect synaptogenesis of RP3 and other motoneuron growth cones. Our results support the idea that a variety of locally provided cues control synaptogenesis and, in particular, that cell-bound molecules that inhibit synaptogenesis of specific growth cones contribute to development of neuronal networks.

We have analyzed Toll null mutant embryos as transheterozygotes between two null alleles kni^IIDDf(3R)Tl^9QRXsr e^sca / TM1 p^popa^9c and opa^9cDf(3R)ro^XB3ru h th st cu Tb ca / TM1 p^pkni^IID. As described by Halfon et al. (1995) all embryos except for the trans-heterozygotes reach early embryonic lethality. Thus, at the late embryonic stages (stages 14- 16), all survivors from this cross are Toll null, a result independently confirmed by immunocytochemistry (data not shown). As in the study by Halfon et al. (1995), who examined the embryos of the same genotype, our analyses on the motoneuron growth cones excluded the data from the hemisegments which did not have a normal complement of muscle cells. Toll misexpressor flies have been generated anew (see below) and analyzed as homozygotes. Transgenic p(RK20′-lacZ)^ET1C-10 (constructed by Andrew Tomlinson, and provided to us by John Thomas) was crossed into Toll null mutants to visualize RP motoneuron cell bodies. Canton S strain was used as a wild-type control.

The use of Toll null mutants has potential complications due to subtly aberrant development of neurons near the CNS midline (e.g. RP motoneurons). Since no neuron expresses Toll, any such anomaly is likely due to yet unknown effects of losing the normal Toll expression in the midline glia (Halfon et al., 1995). We found that overall development of RP motoneurons remains largely intact. First, Fasciclin III immunocytochemistry, which reveals the cell bodies of RP1, RP3 and RP4, showed that a ‘Fasciclin III-positive RP3’ is identified, at the focal level of the longitudinal connectives and segmental commissures, in 87% of the cases in wild-type embryos (n=74) and in an only slightly lower 81% of Toll null mutants (n=76) (see Halfon et al., 1995 for more pronounced but similar results). Second, use of another molecular marker gene RK20 (Stefan Thor and John B. Thomas, personal communication), which is expressed heavily in RP3 and RP5, revealed that a ‘RK20-positive RP3’ is present at very similar rates in both wild type (93%, n=54) and Toll null mutants (89%, n=66). In addition, relying solely on Nomarski optical analysis of unstained fillet preparations, we could still identify a ‘morphological RP3’ in its expected position and of an expected size (i.e. standard criteria with unstained preparations) in 79% of the Toll mutants (n=46), comparable to wild-type embryos (80%, n=52) (Fig. 1A,C). Finally, when any of these cells identified as RP3 were injected with intracellular dye, a single axon was seen to take the normal ‘RP3 pathway’, i.e. crossing the midline through the anterior commissure, veering off posterolaterally and reaching the lateral edge before exiting the CNS (Fig. 1B,D; see Chiba et al., 1993, 1995; Halpern et al., 1991; Sink and Whitington, 1991 for the ‘RP3 pathway’). Therefore, the cells that we identify as RP3 in both wild-type and Toll null mutant embryos have similar morphological, molecular and pathfinding properties to wild type. These observations suggest that, at least within the CNS, RP3 development remains virtually normal even with the Toll mutation and, further, that this well-studied motoneuron is a useful probe for studying the effects of losing Toll during motoneuron growth cone-muscle interactions.

A transgene consisting of a muscle-specific promoter from the Muscle myosin heavy chain gene (Wassenberg et al., 1987) and Toll cDNA (Toll⁺, Hashimoto et al., 1988) was cloned into pCaSpeR4 (Rubin and Spradling, 1982) for P element-mediated genomic transformation. Seven independent homozygous viable lines were obtained. We have analyzed two of them, w ; P(w⁺, Mhc′-Toll⁺)⁹⁸ (insertion on the second chromosome) and w ; P(w⁺, Mhc′-Toll⁺)²⁸ (insertion on the third chromosome), which express Toll in all muscle cells at very similar levels and onsets. The first sign of Mhc′-driven Toll expression appears in ventral and lateral portions of the musculature during stage 14 (approximately hour 12 of embryogenesis), about 2-4 hours before the onset of neuromuscular synaptogenesis (Fig. 2C). At this point, levels of ectopic Toll expression are slightly below those of endogenous Toll expression in the ventral muscle cells. During stages 15 though early 16 (hours 13-15), the levels of misexpression gradually increase as well as spread out through the entire musculature. By late stage 16 (hours 16-18), the levels of misexpressed Toll surpass those of peak endogenous expression in the ventralmost muscle cells. As with endogenous Toll, ectopic Toll accumulates at the muscle clefts (Fig. 2D). The Mhc′-driven Toll expression remains detectable at lower levels in all muscle cells throughout the subsequent larval stages (data not shown). Unlike ectopically expressed Fasciclin III under the same Mhc promoter (Chiba et al., 1995), misexpressed Toll does not appear to overtly increase cell-cell adhesivity between Toll-misexpressing muscle cells. When dissected and pulled towards the dorsal midline, larval muscles 6 and 7 remain closely apposed in 26% of the cases (n=100) compared to 15% (n=83) in wild type. Despite the ectopic Toll expression, other membrane-bound putative growth cone guidance molecules such as Fasciclin III and Connectin retain normal expression patterns (data not shown). It is therefore likely that motoneuron growth cones experience normal molecular environments in all respects save that they now face ectopically placed Toll on all muscle cells during the onset of synaptogenesis. We have used both Mhc′-Toll⁹⁸ and Mhc′-Toll²⁸ lines for analysis with Fasciclin II immunocytochemsitry and Mhc′-Toll⁹⁸ for dye injection.

Lucifer yellow (Molecular Probe), a fluorescent dye, was injected into individual RP3 motoneu- rons and growth cones were visualized as in previous studies (Chiba et al., 1993, 1995). We chose RP3 primarily because the collective visualization of all motoneuron growth cones with Fasciclin II immunocytochemistry showed likely RP3 targeting errors at its normal site of innervation, i.e. muscle cells 6 and 7. In addition, RP3 is one of the easiest cells on which to perform dye injection. Late stage 16 (hours 16-18) embryos were selected from overnight egg collections based on morphothan logical criteria, i.e. they had already past the ‘four-lobed gut’ stage and initiated hindgut convolution. Identification of RP3 cell bodies within the unstained CNS was based on unique position, shape and size (see Fig. 1A,C). Images of dye-filled RP3 growth cones were captured digitally using a cooled CCD camera (Hamamatsu C5985). To minimize ambiguity, we have rejected all those cases where more than one neuronal cell body was labeled in a given segment. Muscle cell injection was done using the same protocol.

Primary antibodies against the following molecules were used (Chiba et al., 1993 for immunocytochemistry procedures): Toll (mAb Toll_cd, Hashimoto et al., 1991), Fasciclin II (mAb 1D4, Grenningloh et al., 1991), Fasciclin III (mAb 7G10, Patel et al., 1987), Connectin (Meadows et al., 1994), Lucifer yellow (Molecular Probe), and β- galactosidase (Promega). Larvae were fillet-dissected prior to immunocytochemistry, while embryonic preparations were dissected afterwards. Synaptic boutons were counted at the 6/7 muscle clefts of abdominal (A2-A7) segments in third instar larval specimens stained with Fasciclin II antibodies.

Toll is dynamically expressed in the ventral half (proximal to the CNS) of the Drosophila embryonic musculature. This part of the musculature includes targets of the SNb and SNd motoneuron groups (Fig. 3B,C). Fig. 3C summarizes three stages of muscular Toll expression relative to the known outgrowth sequence of the SNb growth cones and specifically the RP3 motoneuron growth cone, a well-studied SNb growth cone (see Halfon et al., 1995; Keshishian et al., 1993; Nose et al., 1992 for the temporal expression control of Toll; also see Broadie et al., 1993; Halpern et al., 1991; Sink and Whiting- ton, 1991 for the RP3 growth cone development). At embryonic stage 14 (approximately 12 hours into embryogenesis), five ventral (proximal) muscle cells (7, 15, 16, 17 and 28) express the Toll gene at high levels. This is before SNb growth cones exit the CNS. Toll protein preferentially accumulates at muscle-muscle contact sites or ‘clefts’, e.g. the apposition between muscle cells 6 and 7, and between 15 and 16 (Fig. 2A). During stage 15 (hours 13-14) the SNb growth cones begin to contact the ventral-most oblique muscle cells 15, 16 and 17. The 6/7 cleft, the normal synaptic site of RP3, loses Toll during this period. During stage 16 (hours 15-18), the SNb growth cones extend further and innervate their targets. The remaining Toll-positive ventral muscle cells gradually lose Toll until the protein remains weakly detectable only at the 15/16 cleft (Fig. 2B). Later-arriving growth cones of SNd, which include the growth cone of motoneuron RP5, innervate this cleft just as its Toll expression becomes undetectable. Consequently, the RP3 growth cone briefly contacts Toll-positive muscle cells before reaching its more distal targets and down-regulation of Toll expression at the muscle 6/7 cleft precedes the onset of RP3 synaptogenesis at that site by approximately 1 hour. Thus, spatiotemporal control of Toll expression in the embryonic musculature raises the possibility that one of Toll’s roles may be to inhibit inappropriate synaptogenesis by the RP3 growth cone as it grows towards proper targets.

In the following experiments, as we examine the effects of genetic manipulation of muscular Toll expression, we pay special attention to those growth cones that normally either innervate or grow past Toll-expressing muscle cells, i.e. the growth cones of SNb and SNd nerve branches and more specifically the RP3 growth cone, which targets muscle cells 6 and 7.

To decide whether Toll expression in the musculature may influence motoneuron growth cones, we have examined nerve ending patterns in embryos lacking zygotic Toll expression. Maternal Toll, which is still supplied, becomes undetectable by early gastrulation, i.e. by stage 6 (hours 3-4) (Hashimoto et al., 1991). Halfon et al. (1995) have reported slightly abnormal neural as well as muscular development in this mutant, presumably due indirectly to losing Toll expression in the midline glia and epidermis. However, their and our own analyses show that the overall development of both the CNS and PNS remains intact at both the gross level and the level of individual RP neurons (see Materials and methods). We hypothesized that, if the normal function of muscular Toll expression is to prevent inappropriate synaptogenesis of specific motoneuron growth cones, lack of Toll would permit abnormal growth cone activities and possibly ectopic synapse formation.

To visualize the motoneuron growth cones collectively, we used Fasciclin II immunocytochemistry. Consistent with previous observation (Halfon et al., 1995; Nose et al., 1994), the five major motor branches, ISN, SNa, SNb, SNc, and SNd, all show largely normal development (Fig. 4C for SNa). However, we found that SNb and SNd motoneurons in Toll null mutant embryos often fail to form distinct terminals on their normal targets (Table 1). For example, the SNb sub-branch innervating at the muscle 6/7 cleft (normally contains the growth cones of RP3 and 6/7b motoneurons; Keshishian et al., 1993) is present in only 61% of the cases and is often reduced in size (e.g. Fig. 4D). Similarly, other subbranches innervating muscle cells 12 and 13 are sometimes missing. On the contrary, extra-thick endings are occasionally seen at muscle 13 and more often at the muscle 15/16 cleft (Fig. 4D, white circle). At this level of analysis, we can not tell whether specific motoneurons, e.g. RP3, reach their normal target regions but fail to innervate or misroute their axons into other nerve branches (see below). The results, however, are consistent with Toll playing a role in motoneuron- muscle interactions.

To further study possible roles of Toll in motoneuron growth cone targeting at the level of single growth cones, we next specifically visualized the RP3 growth cone in Toll null mutants by using intracellular dye injection. This technique provides the resolution necessary to specifically follow RP3 targeting in mutant embryos. It is important to note that, in both wild type and Toll null mutants, the RP3 growth cone extends contralaterally and continues posterolaterally before exiting the CNS; once outside the CNS, it extends normally until it reaches the most ventral (proximal) muscle cells 15, 16 and 17 (see Materials and methods). Thus, in Toll null mutants, the initial axon outgrowth of RP3 within the CNS and ventral musculature is consistent with the wild-type situation. Subsequently, however, a noticeable deviation is detected.

In wild-type embryos, RP3 continues to elongate distally and innervates muscle cells 6 and 7 by extending its growth cone approximately 4 μm internally through the 6/7 cleft (Fig. 5A,B; also see Fig. 6A). In contrast, the majority of RP3 growth cones in Toll null mutants (90%) appear to either stall under the 6/7 cleft, failing to innervate any muscle cell at all, or misinnervate neighboring ventral muscle cells (15 and/or 16) which would normally have Toll on their surfaces during the same period (Fig. 5C,D; Table 2). Such a misinnervation pattern is never seen in wild-type embryos. These results demonstrate that Toll is involved in normal motoneuronmuscle interactions, possibly inhibiting synaptic initiation of the RP3 growth cone.

To further test the idea that muscular Toll expression influences motoneuron growth cones and possibly inhibits synaptic initiation of specific motoneurons, we have next created an experimental situation where Toll is misexpressed under the control of a muscle-specific promoter from the Muscle myosin heavy chain (Mhc) gene (see Materials and methods). This genetic manipulation is reciprocal to the Toll null situation in the musculature, with Toll now being expressed on all ventral muscle cells by stage 14 (hour 12) (Fig. 2C). Mhc-driven Toll expression on these muscle cells begins 1-2 hours before RP3 and other motoneuron growth cones contact them. The levels of Toll misexpression gradually rise towards late stage 16 (hours 16-18) (Fig. 2D). For the 6/7 cleft, this means that down-regulation of Toll at this site is delayed for at least 1 hour, i.e. until after the RP3 growth cone contacts. Therefore, this experiment tests the importance of precisely controlling not only the cell specificity, but also the timing of Toll expression in the embryonic musculature. If Toll is normally responsible for preventing inappropriate synaptogenesis of specific motoneurons, then misexpression of Toll on the ventral muscles might lead to a lack of normal innervation at these sites.

Fasciclin II immunocytochemistry in these Toll misexpressing embryos revealed that all major motor nerve branches, i.e. ISN, SNa, SNb, SNc and SNd, develop normally (Fig. 4E for SNa). Thus, elimination of the normal dorsoventral (distal-to- proximal) gradient of Toll expression at the musculature does not lead to any gross effects on motoneuron-muscle interaction.

However, a closer examination on the SNb growth cones has revealed that subtle errors occur among some of them. In particular, the SNb sub-branch innervating the muscle 6/7 cleft is either absent or much reduced (Fig. 4F, black arrow; Table 1). Similarly, other SNb sub-branches and SNd branch at 15/16 cleft are also sometimes missing (Table 1). It is not possible with this method to determine each growth cone’s exact position, i.e. whether it has mistargeted alternative muscle cells as in the case with Toll null mutants, or merely stopped short of synaptogenesis at the correct targets (see below). Overall, motoneuron growth cones in Toll misexpressing embryos appear less elaborated than in the wild type (compare in Fig. 4 and Table 1). These results suggest that simply altering the timing of Toll expression in normally Toll-positive muscle cells may result in local inhibition of synaptogenesis.

To visualize the effects of Toll misexpression on the RP3 growth cone in particular, we again used intracellular dye injection. We hypothesized that, if Toll expression in ventral muscle cells affects the RP3 growth cone in an inhibitory manner, misinnervation by the growth cone would not occur as in the previous Toll null mutants. Instead, we would expect that, while other cues directing RP3 to its target region likely remain intact (see ‘Toll Misexpression in Embryonic Musculature’ in Materials and methods), Toll misexpression would act locally to inhibit synaptogenesis at the 6/7 cleft.

We found indeed that the RP3 growth cone, unlike in the Toll null mutants, has no problem extending past the proximal musculature (e.g. muscle cells 15, 16 and 17) and reaching the point just beneath (external to) the 6/7 cleft. After this point, however, the RP3 growth cone almost always fails to enter the cleft and to initiate synaptogenesis on the internal side (Fig. 5E,F; Table 1). Confocal microscopy with double labeling for the motoneuron growth cones and muscle cells 6 and 7 has further confirmed that the RP3 growth cone stalls beneath (external to) the 6/7 cleft, 3-4 μm before the normal termination site (Fig. 6). It is not possible with our analysis to determine whether the growth cone actively withdraws or passively stalls upon contacting Toll-expression muscle cells. We conclude that, when expressed by target muscle cells at the wrong time, Toll does not prevent the RP3 growth cone from reaching the correct target region but instead inhibits initiation of synaptogenesis at that site.

To determine if Toll misexpression in muscles leads to permanent deficits in motoneuron innervation, we examined motoneuron endings in first and third instar larval, 40-45 and 70-80 hours after the onset of embryogenesis, respectively. As before, we used Fasciclin II immunocytochemistry to visualize innervation patterns (Fig. 7). We especially focused on the 6/7 cleft because we have noted poor synapse development by the RP3 growth cone at this site in embryos. Mhc′- driven Toll misexpression in muscles persists at low levels throughout larval development (data not shown). Nevertheless, most muscles become innervated similarly to wild-type larvae (Table 3). While the 6/7 cleft in wild type is always innervated by late embryonic stage 16 (hours 16-18) (e.g. Fig. 4B), the innervation rate in the Toll misexpressors, as measured by Fasciclin II immunocytochemistry, increases steadily from 50% at stage 16 (e.g. Fig. 4F) to 89% at the first larval stage (e.g. Fig. 7C), and finally to 100% at the third larval stage (e.g. Fig. 7D; Table 3). Other ventral muscles (12, 13, 15 and 16), which often lack normal innervation during embryogenesis, become innervated in a manner similar to wild type by the late first larval stage (Table 3). Although this visualization method did not allow us to trace individual motoneu- ron axons specifically, we noted that both the site of innervation and size of boutons on a given muscle surface were similar to those in wild type (Fig. 7). There is, however, a quantitative difference. The average number of synaptic boutons at the 6/7 cleft in Toll misexpressing third stage larvae is 31.3±1.1 (mean ± s.e.m., n=76), 31% less than in the wild type (45.0±2.1, n=79). These data suggest that the prolonged presence of Toll on ventral muscle cells delays synaptogenesis of RP3 and other SNb growth cones at least until early larval stages, in some cases causing further long-lasting deficits in terminal elaboration. Overall, these results are con- sistent with the idea that Toll is a part of a large array of often redundant molecular cues that growth cones utilize during neuromuscular recognition (Tessier-Lavigne and Goodman, 1996 for a review).

This study shows that the transmembrane molecule Toll plays a role during synaptic initiation of Drosophila motoneurons. Based on its molecular structure and expression patterns in the musculature, we hypothesized that Toll may contribute to the target recognition and/or synaptic initiation of specific motoneuron growth cones. We tested this through reciprocal loss-of-function and gain-of-function genetic manipulations followed by collective and individual visualization of motoneuron growth cones. We found that overall organization of the motor axon pathways remains virtually intact despite these genetic manipulations. However, on a finer level, lack of Toll results in ectopic innervation by a specific motoneuron (i.e. RP3), while misexpression of Toll in the musculature inhibits synaptic initiation of the same motoneuron. Our study supports the idea that cell surface molecules of the leucine-rich repeat (LRR) gene family may play a role during synaptogenesis, and also hints at the importance of precise temporal, as well as spatial, control of growth cone guidance molecules.

Our data show that muscular Toll can influence synaptic initiation of motoneuron growth cones locally. For example, synaptogenesis by the RP3 growth cone on its normal target muscle cells, 6 and 7, can fail when either Toll’s normal down- regulation at these muscle cells does not occur or its persistent expression at other, more proximal non-target muscle cells is absent. The data with Fasciclin II immunocytochemistry in the Toll null mutants suggest that, while some additional motoneurons whose growth cones grow through SNb₁₂ and SNb₁₃ nerve branches are also likely to be affected by contact with Toll- expressing muscle cells, certain other growth cones are insensitive to such contact. For example, under normal conditions the SNd growth cones (e.g. RP5) are able to initiate synaptogenesis at the cleft between muscle cells 15 and 16 before the cleft loses Toll completely. In addition, when Toll is ectopically expressed in all muscle cells, synaptogenesis proceeds apparently normally for the growth cones of the SNa nerve branch (Fig. 4E). Based on our results, we propose that contacting Toll-expressing muscle cells delays synaptogenesis of a specific set of motoneuron growth cones, including RP3, regardless of whether or not these growth cones are destined to synapse with those selfsame muscle cells.

The molecular mechanisms by which muscle-provided Toll elicits such a growth cone response are unclear. The apparent similarity between the RP3 growth cone response to Toll misexpression at the 6/7 cleft and the deletion of a motoneuron-provided ‘synapse initiation molecule’, late bloomer (lbl), is intriguing (Kopczynski et al., 1996). In both cases, the growth cone succeeds in reaching the 6/7 cleft but delays synapse initiation until later times. One possible scenario is that muscle-provided Toll physically interferes with the initiation of a neuromuscular synaptogenesis complex, of which lbl is a part. Excessive Toll may interfere in this process by promoting muscle-muscle adhesion, presumably through heterophilic interactions with unidentified ligands present on these muscle cells (Keith and Gay, 1990, for evidence of heterophilic inter- actions). Such increased adhesion may block growth cones from extending through intermuscular space, accessing guidance molecules present on muscle cell surfaces, and/or activating the ‘synapse initiation molecules’ (e.g. lbl) (Fig. 8A). An alternative scenario is that growth cones bear yet unidentified Toll receptors, which may or may not interact with lbl. When activated, these Toll receptors may inhibit growth cone differentiation (Fig. 8B). Immunological visualization of RP3 and other growth cones in Toll-misexpressing embryos seems consistent with the latter ‘instruction’ hypothesis. Mature innervation was sometimes missing even in SNb₁₃ and SNb₁₂ nerve branches, which do not differentiate at ‘clefts’. In Toll null embryos, many growth cones appear to slide away from correct targets, while some seem to branch abnormally (Table 1). This can be taken as further supporting evidence for Toll normally providing signals to prevent growth cones from excessive synaptogenesis. An in vitro study using neuronally derived cell lines suggests that Toll can bind to as yet unidentified molecules probably present in Drosophila neuronal tissues (Keith and Gay, 1990). These Toll-binding molecules may function as growth cone receptors, which act to suppress the initiation of synaptogenesis. Determining the primary factor by which Toll may inhibit synaptogenesis is difficult at the moment (Fig. 8).

In summary, membrane molecules with pronounced LRR motifs have been proposed to play roles during the develop- ment of both vertebrate and invertebrate nervous systems (Eldon et al., 1994; Musacchio and Perrimon, 1996; Nose et al., 1992, 1994; Taguchi et al., 1996; Yamagata et al., 1994). Our study demonstrates that synaptic target recognition could involve local inhibition of synaptic initiation mediated by Toll, a protein with LRR motifs.

Expression of molecules involved in growth cone guidance often occurs under highly specific controls. In vivo studies have demonstrated that such control of cell specificity and dosage is essential to proper development of neuronal networks (e.g. Lin and Goodman, 1994). Our experiment suggests that as little as a 1 hour delay in the down-regulation of Toll in RP3’s target muscle cells (6 and 7) can have significant consequences on its synaptic initiation. This prompts further testing of the idea that precise timing control of growth cone guidance molecules is at least as important as spatial and dosage control.

Extrinsic cues that guide neuronal growth cones vary widely in their structures and functions. They include both diffusible molecules exerting an influence from relatively long distances, e.g. Neurotrophins, neurotransmitters, Netrins, Semaphorins and cell-bound molecules providing more local signals, e.g. NCAM, L1, cadherins, RAGs, Connectin, Fasciclin III (Chiba and Keshishian 1996; Culotti and Kolodkin, 1996; Davenport et al., 1996; Friedman and O’Leary, 1996; Garrity and Zipursky, 1995; Goodman, 1996; Phillips and Armanini, 1996; Tessier-Lavigne and Goodman, 1996 for reviews).

To how many of these extrinsic guidance molecules is a given growth cone capable of responding? Is it that each growth cone has a limited number of receptors for external signals and remains blind to the majority of potential cues? Or is it the case that growth cones share a wide array of receptors and, given opportunities, are capable of responding to numerous cues? If the former were the case, a crucial factor producing differential growth cone responses would be the genetic control of receptor expression unique to each growth cone, whereas if the latter were true, spatiotemporal control of extrinsic cue distribution would be the primary determining factor.

This and several other recent studies have utilized single cell dye injection and begun to describe the behavior of the RP3 growth cone under various experimental conditions. It appears that this growth cone is able to specifically recognize muscle cells 6 and 7 as its normal targets (Chiba et al., 1993). However, it can also respond to multiple molecular cues in distinct manners. For example, the RP3 growth cone is attracted to muscle cells which either endogenously or ectopically express Fasciclin III, an immunoglobulin-like transmembrane molecule (Chiba et al., 1995). In contrast, the same growth cone is inhibited from synapsing on its normal targets when the targets ectopically express Semaphorin II, a diffusible molecule with high homology to vertebrate Collapsin (Matthes et al., 1995). Similarly, in this study, we have shown that synaptogenesis of the RP3 growth cone is inhibited on muscle cells that express Toll. Such reciprocation in cell-bound molecule action seems a part of the molecular environment normally experienced by the developing RP3 growth cone in the periphery. Interestingly, another cell surface molecule of the LRR gene family, Connectin, which is normally present in regions of musculature never accessible to the RP3 growth cone, seem able to steer away this growth cone when placed ectopically on the muscle cells contacted by it (Nose et al., 1994). Whether or not growth cones in general are responsive to normally inaccessible molecules awaits future examination. However, in light of RP3’s ability to respond even to cues it would not normally experience, it is intriguing to view individual growth cones as ‘opportunists’, ready to utilize whatever extrinsic cues are available. Such a view implies that precise control over the temporal and spatial expression patterns of potential growth cone guidance cues is a crucial factor governing brain development.

We thank Carl Hashimoto (Yale University) for Toll antibodies and cDNA, Haig Keshishian and Marc Halfon (Yale University) and Akinao Nose (NIBB, Japan) for Toll null mutant flies, Andrew Tomlinson (Columbia University), John Thomas and Stefan Thor (Salk Institute) for RK20′-lacZ transgenic flies, Corey Goodman (UC Berkeley) for Fasciclin II antibodies, Rob White (University of Cambridge) for Connectin antibodies, Mike Gorcyzyca and Vivian Budnik (University of Massachusetts) for advice on the first instar larval dissection, and Chris Doe for comments on this manuscript. This work was supported by grants from NSF (IBN-95-14531), NIH (NS35049) and the Markey Foundation to A. C.