Specifying the path of the intersegmental nerve of the Drosophila embryo: a role for Delta and Notch

How a neuron finds its synaptic targets is one of the fundamental issues in neural development (Ramon y Cajal, 1984). There is a wealth of data about the cell biology of axon guidance (reviewed in Dodd and Jessell, 1988; Bixby and Harris, 1991; Hynes and Lander, 1992), and the goal of much recent work in this area has been to identify molecules whose distribution, structure and activity all suggest that they are likely to be determinants of a choice in axon guidance (Bastiani et al., 1987; Cunningham et al., 1987; Tessier-Lavigne et al., 1988; Furley et al., 1990; Stahl et al., 1990; McIntire et al., 1992). One approach to uncovering a guiding molecule and its receptor is to look for mutations of Drosophila that give rise to specific aberrations in nerve trajectories. Such mutations may affect nuclear proteins required to specify the identity or tran-scriptional program of a neuron (Ghysen et al., 1983; Doe et al., 1988; Vaessin et al., 1991), or the cell surface and effector molecules that directly control axon extension (Gertler et al., 1989; Hedgecock et al., 1990; Grenningloh et al., 1991). However, there is no guidance decision for which the entire regulatory heirarchy is known: guiding molecule, receptor, and their regulators and effectors. We have therefore analyzed the development of a single nerve in the peripheral nervous system (PNS) of the fly embryo to identify the molecules that specify its trajectory.

In the Drosophila embryo, each side of each bilaterally symmetric thoracic and abdominal segment includes an axon bundle, termed the intersegmental nerve (ISN), that courses along a fixed path near the anterior boundary of that segment (Campos-Ortega and Hartenstein, 1985; Ghysen et al., 1986; Fig. 2A,B). The ISN contains both sensory axons that grow ventrally, towards the central nervous system, and motor axons that grow dorsally. While it is not known what genes or structures determine the trajectory of the ISN, it has been noted that this nerve grows, in turn, on three different substrata. In the most ventral part of the embryo, the ISN passes between layers of myoblasts (J. Johansen et al., 1989) whereas, in the most dorsal region, the motor axons of the nerve grow internal to the dorsal cluster of sensory neuron cell bodies. In between, in the lateral part of the embryo, the ISN grows in close apposition to the transverse branch of the embryonic trachea (Fig. 2A,B; Hartenstein, 1988), which develops prior to the initiation of peripheral axon extension. In fact, in this middle third of the pathway, the pioneer growth cone of the ISN migrates along the tracheal wall, leading to speculation that the trachea might define part of the ISN tra-jectory (Hartenstein, 1988).

The mechanisms that determine the shape of the trachea are not themselves understood, though it is known that a Drosophila homolog of the fibroblast growth factor receptor (D-FGFR) is expressed in the cells of the trachea and is specifically required for the morphogenesis of this tissue (Glazer and Shilo, 1991). If the trachea does in fact provide a guidance signal to the ISN, one might expect the trachea to produce a molecule for which there is a receptor on peripheral axons. One candidate for such a ligand-receptor pair is the proteins encoded by the genes Delta and Notch. In addition to their expression at other times and places in the developing fly, Delta is transcribed in the embryonic trachea (Vaessin et al., 1987), whereas Notch protein is expressed on the axons of the ISN (K. Johansen et al., 1989; Fehon et al., 1991).

Delta and Notch are large, transmembrane proteins that bear a number of EGF-like (epidermal growth factor-like) repeats in their extracellular domains (Wharton et al., 1985; Vaessin et al., 1987; Kidd et al., 1988; Kopczynski et al., 1988). These proteins are required for many of the cell-cell interactions that regulate the development of the embryonic and adult nervous system of the fly (Lehmann et al., 1983; Cagan and Ready, 1989; Hartenstein and Posakony, 1990), as well as being important for the development of several non-neuronal structures (Lindsley and Grell, 1968; Corbin et al., 1991; Ruohola et al., 1991). For example, in embryos mutant for either Delta or Notch, essentially all ectodermal cells in the neurogenic region become neurons (Lehmann et al., 1983; Campos-Ortega and Jan, 1991) and none become epidermal cells, due to a failure of cell-cell com-munication (Doe and Goodman, 1985; Technau and Campos-Ortega, 1987). It is generally believed that the Delta and Notch proteins interact directly with one another to effect those processes that require both genes, based on the following observations. First, lesions in these two genes give rise to very similar mutant phenotypes in a number of developmental contexts (Lindsley and Grell, 1968). Second, these loci interact genetically: the effect of a mutation in either gene is strongly influenced by the gene dosage of the other (Lindsley and Grell, 1968; Vaessin et al., 1985; de la Concha et al., 1988), and the phenotypes of some Notch mutations are suppressed by particular Delta mutations, in an allele-specific manner (Brand and Campos-Ortega, 1990). Third, in cultured cells, Notch protein can act as a receptor for Delta (Fehon et al., 1990; Rebay et al., 1991). This last result is consistent with data from genetic mosaics, which suggest that, in various situations, Delta generates the developmental signal for which Notch is part of the receiving mechanism (Hoppe and Greenspan, 1986, 1990; de Celis et al., 1991; Heitzler and Simpson, 1991). Most of the available data on the roles of Delta and Notch concern their activities in cell fate determination, but one report has described axonal defects in the adult peripheral nervous system in flies bearing particular mutant alleles of Notch (Palka et al., 1990). Furthermore, the Notch mutant pheno-type has been interpreted as reflecting a primary role for the protein in cell adhesion (Hoppe and Greenspan, 1986), a cellular property that is believed to be a major determi-nant of axon guidance.

The experiments reported here suggest a mechanism for specifying the trajectory of the intersegmental nerve. In par-ticular, we employ temperature shifts of temperature-sensi-tive mutant alleles of the genes Delta and Notch to show that, separate from their well-studied roles in the determi-nation of cell fate, these genes are also required for the growth of ISN axons through the lateral region of the embryonic periphery. Since the ISN grows on the surface of the trachea through this part of the embryo and Delta is expressed in the trachea while its receptor, Notch protein, is expressed on the associated peripheral axons, we specu-late that the trachea uses the interaction of Delta and Notch to direct the growth of ISN axons. We further test the idea that the trachea is important for peripheral axonogenesis by using a mutation of the D-FGFR to perturb tracheal devel-opment and show that the resulting tracheal defects are reflected in corresponding axonal defects.

We observe that Delta or Notch mutations, which block normal development of the intersegmental nerve, also pre-vent formation of the longitudinal axon tracks between the segmental ganglia in the CNS. We propose, therefore, that the guidance system that we have defined for the lateral section of the ISN also specifies the trajectory of one ele-ment of the axon scaffold of the CNS.

Drosophila melanogaster stocks were maintained on standard media at 25°C (or 18°C for temperature-sensitive mutants). In experiments involving temperature shifts, embryos were collected on grape-juice agar at 18°C and shifted to 32°C as specified for the particular experiment. For experiments at temperatures other than 25°C, times are given here as the equivalent developmental time at 25°C, not as the actual elapsed time (Ashburner, 1989). We assumed the rate of development to be decreased 50% at 18°C, and increased 10% at 32°C, relative to the rate at 25°C. The fol-lowing mutant stocks were used: fz GF^3b (Df3L 70B-70D6)/TM6B Tb e; fz GS^1a (Df3L 70C6-15 -70E4-6)/TM3 Sb; N^ts1/N^ts1; N^ts1/N^ts1, SM1Cy Dp(1;2)w⁺51b7N⁺/+; stDl^6B37(ts)e/TM3 Sb Ser; st Dl^via1e/st Dl^via1e. The wild-type stock was either Oregon R or y w iso 2,3 OreR (Bier et al., 1989). Temperature-shift exper-iments with Delta were performed using the trans-heterozygous combination st Dl^6B37(ts)e/st Dl^via1e. The FGFR gene was deleted in the trans-heterozygous combination fz GF3b/fz GS1a; fz stocks were obtained from the Bloomington Stock Center. All stocks not otherwise specified are described in Lindsley and Grell (1968).

Immunocytochemistry and in situ hybridization were performed by standard methods (Bodmer and Jan, 1987; Bier et al., 1989; Tautz and Pfeifle, 1989) except samples for confocal microscopy were blocked with 5% normal donkey serum (in 0.1 M sodium phosphate, pH 7.2, 0.3% Triton X-100 with 2% bovine serum albumin) for 15 minutes at room temperature immediately prior to incubation with secondary antibody. Neurons were labelled either with anti-HRP (Jan and Jan, 1982), obtained from Cappel and affinity-purified, or with mAb 22C10 (Zipursky et al., 1984). Anti-tracheal antibody was mAb 68G5D3 (Sandra Barbel, Larry Ackerman, LYJ and YNJ, unpublished results). Peroxidase-cou-pled goat secondary antibodies were from BioRad, fluorescein- and Texas Red-coupled donkey secondary antibodies were from Jackson Immunologicals. Peroxidase-developed samples were cleared in toluene and mounted in Permount (Fisher) for Nomarski microscopy using a Nikon Optiphot; fluorescent samples were mounted in 90% glycerol with 2% n-propyl gallate and 20 mM Tris pH 7.6, and examined on a Zeiss Axioplan equipped with BioRad scanning confocal microscopy hardware and software. For in situ hybridization, Delta cDNA 3.2 (Vaessin et al., 1987) was labelled with digoxigenin (Boehringer Mannheim) and used to probe wild-type or mutant embryos. For different experiments, either DNA probes were prepared by ‘random-oligo’ labelling, or antisense RNA probes were prepared by transcription in vitro. Staging of embryos for all experiments was as described by Campos-Ortega and Hartenstein (1985).

It has been observed that the axons of the intersegmental nerve (ISN) bear Notch protein (Fehon et al., 1991). We thus undertook a series of temperature-shift experiments with temperature-sensitive mutant alleles of Notch, and also of Delta, to uncover any specific function these proteins might have in nerve guidance. Embryos were grown for varying lengths of time at the permissive temperature (18°C) in order to provide the Delta and Notch activity required for neurogenesis, and were then raised to restric-tive temperature (32°C) and maintained at that temperature until embryonic stage 15/16 (after the completion of periph-eral axonogenesis). As background, we now briefly sum-marize the main events in neurogenesis and peripheral axonogenesis (Campos-Ortega and Hartenstein, 1985; Hartenstein, 1988; J. Johansen et al., 1989; Sink and Whit-ington, 1991). Neurogenesis begins at about stage 9 (4 hours of development) and the pioneers of the ISN are born by stage 13 (9 hours), though the full complement of motoneurons is not readily identifiable until stage 15. The motor pioneer of the ISN (aCC) exits the CNS during stage 13, at roughly the same time that the dorsal sensory pio-neer of this nerve (dh1) initiates its ventrally directed axon outgrowth (about 9.5 hours). The two pioneer axons con-tact opposite ends of the transverse branch of the trachea late in stage 13. The axons meet along the trachea and fas-ciculate during stage 14. (Note that the transverse branch of the trachea has formed by the end of stage 12.)

Appropriate temperature-shifts of Notch^ts or Delta^ts mutants caused defects in the growth and guidance of the ISN (Fig. 1A-D). In representative segments, axons were found either to grow inappropriately, for example, extend-ing into adjacent segments (Fig. 1B,D, arrows), or to cease extending altogether (Fig. 1A,C, open arrowheads). Tem-perature shifts done in parallel with wild-type embryos, or with Notch^ts embryos containing a wild-type copy of the Notch gene on another chromosome (Fig. 1E), did not pro-duce such guidance errors. Unexpectedly, these defects were localized to roughly the lateral one-third of each seg-ment: outside the lateral region of the embryo, the ISN did not appear to be adversely affected in these temperature-shift experiments (Fig. 1A-D, triangles). In our experi-ments, initiating the shift from permissive to restrictive tem-perature between 3 and 6 hours of development (equivalent time at 25°C; see Materials and Methods) reproducibly caused aberrations in axon elongation while minimizing neurogenic defects; shifting the temperature between 6 and 9 hours typically failed to produce a mutant axonal pheno-type. The ISN grows through the lateral region of the embryo between about 10 and 11 hours of development. The delay between the initiation of the temperature shift and the onset of phenotypic aberrations presumably reflects perdurance of the mutant protein, and is consistent with the delay seen in analogous experiments examining Notch^ts effects in larvae and pupae (Cagan and Ready, 1989; Hartenstein and Posakony, 1990). In those experiments, it was necessary for animals to be at restrictive temperature for at least 4-8 hours (depending on the tissue) before any mutant phenotype was observed. Similarly, it has been observed in genetic mosaic experiments that Notch activity remains at significant levels for several hours following the generation of a mutant clone in embryos (Hoppe and Greenspan, 1990). The latency of Delta^ts effects in ovaries appears to be at least as long as is the case for Notch^ts (Ruo-hola et al., 1991).

We noted that, in the lateral part of the embryo, that is, the region in which we observed axonal defects arising from Notch^ts and Delta^ts, the axons of the ISN are known to extend along the surface of the trachea (Hartenstein, 1988; diagrammed in Fig. 2A). We therefore tested whether the defects that we observed might be indirect consequences of tracheal aberrations caused by the Delta or Notch muta-tions. Notch^ts and Delta^ts embryos were subjected to the same temperature-shift protocols described above and then labelled both with an antibody directed against the trachea and one directed against neurons. For comparison, the asso-ciation of the ISN with the trachea in a wild-type embryo is shown in Fig. 2B. In this micrograph, the nervous system is in green and the trachea in red; their apposition in the lateral region is highlighted with a bracket. The anti-tra-cheal antibody labelled the lumen of the trachea, thus the apparent distance between the green- and red-stained struc-tures is the thickness of the tracheal cells themselves. In the mutants (Fig. 2C-E), we readily found examples of seg-ments in which tracheal development was normal (open arrowheads), but displaying defects in the growth and guid-ance of the ISN like those described in the previous sec-tion (arrows), suggesting that Notch and Delta are directly required for ISN guidance.

The finding that Notch and Delta are required for growth of the ISN along the trachea recalled the report that Delta RNA is expressed in the trachea (Vaessin et al., 1987), which we have verified (E. G. and Y. N. J., unpublished observations). Taken together, these observations suggested the idea that Delta protein on the trachea might provide a path for the elongation of ISN axons. We tested this model further by disrupting tracheal development and examining the effect of such disruptions on axonogenesis. Specifically, we took advantage of the observation that mutation of a Drosophila homolog of the FGF receptor (D-FGFR) pre-vents tracheal morphogenesis but not the birth or differen-tiation of the tracheal cells (Glazer and Shilo, 1991; Klambt et al., 1992). In the embryonic periphery, the D-FGFR is expressed only in the trachea and is never detected in neu-rons. A subset of ventral midline cells also express the D-FGFR, but mutation of this gene has only very subtle effects on the CNS (Klambt et al., 1992). Moreover, the pioneer motor axon of the ISN (aCC) does not cross the midline. Thus, any peripheral neuronal defect in the mutants must arise as an indirect consequence of tracheal defects.

Mutant embryos lacking the D-FGFR suffered severe defects in tracheal morphology and had corresponding defects in the associated nerves, as detailed below. The most apparent axonal aberrations were observed in segments that lacked trachea altogether, or in which the trachea was dis-placed dorsally beyond the filopodial reach of ISN motor axons. In these cases, the ISN either stopped growing or fol-lowed the trachea in an adjacent segment (Fig. 3A,A′,B,B′), aberrent nerves indicated by arrows; trachea denoted by open arrowheads; segment lacking trachea highlighted with asterisk). We never observed a segment in which the ISN grew through the lateral part of the embryo in the complete absence of tracheal tissue. In segments where residual tra-cheal material developed, the ISN appeared to grow along this tissue, however irregular the morphology of that tra-cheal remnant might be. We never found an instance in which the ISN stalled after successfully contacting the trachea. Note that Delta RNA was expressed normally in the tracheal tissue that developed in mutant embryos (data not shown). In all cases, ISN development appeared normal in the ventral region of the embryo (triangles).

Our experiments employed overlapping chromosomal deficiencies that completely remove the receptor gene (see Materials and Methods). Thus, the receptor itself is not directly required for association of the ISN with the trachea. These deletions are also believed to remove several other genes (Glazer and Shilo, 1991), however, and thus we cannot rule out the possibility that one of these other genes might contribute to the mutant axonal phenotype. Nonethe-less, the perfect correlation of tracheal and axonal defects in these tracheal mutants implies that ISN development depends, at least in part, upon the trachea. Also consistent with this idea was the observation that in a small subset of cases, temperature shifts of Notch^ts embryos blocked the development of the transverse branch of the trachea. In these rare cases, segments that lacked the transverse trachea due to the absence of Notch function did not support growth of the lateral section of the ISN (data not shown). This result provides independent confirmation of the idea that a critical role is played by the trachea in facilitating ISN development.

In the course of our analysis of the role of Notch and Delta in peripheral axonogenesis, we observed that these two mutations produced very similar mutant phenotypes in the axon scaffold of the CNS (Fig. 4B,C). Specifically, both mutations prevented formation of the connectives (longitu-dinal tracks) between the segmental ganglia (arrows), while having far more modest effects upon the commissures (cross-tracks) and connectives within the ganglia (Fig. 4, open arrowheads). This phenotype is distinct from a vari-ety of other classes of mutant phenotypes that have been described for the CNS (Mayer and Nusslein-Volhard, 1988; Rothberg et al., 1988; Thomas et al., 1988; Smouse et al., 1988). We therefore suggest that Delta and Notch are part of a regulatory system that determines a variety of indi-vidual choices in nerve guidance, including the trajectory of this element of the CNS and of the lateral third of the ISN.

How is an axon guided to its synaptic targets? For one nerve, the intersegmental nerve (ISN) of the Drosophila embryo, the following model emerges from the experiments presented here (Fig. 5). Expression of Delta protein on the embryonic trachea defines a path for axons of the ISN and these axons use the Notch protein on their growth cones to recognize and follow that path. These conclusions follow, in part, from our observation that appropriate temperature shifts of Notch^ts or Delta^ts mutants prevent proper growth of the ISN through the lateral part of the embryo, and that this nerve growth is also blocked if tracheal morphogene-sis is disrupted by mutation of the D-FGFR. The pattern of expression of these genes is also consistent with the pro-posed model, as Delta is expressed in the trachea (Vaessin et al., 1987), while its receptor, Notch, is on the axons of the intersegmental nerve (K. Johansen et al., 1989; Fehon et al., 1991). The regulatory system described here for guid-ance of the lateral part of the ISN evidently also specifies the trajectory of a section of each longitudinal nerve track in the CNS: mutations in Delta or Notch specifically dis-rupt the growth of CNS axons between the segmental gan-glia.

One caveat to our interpretation of the Notch^ts and Delta^ts temperature-shift experiments is that we cannot formally rule out the possibility that some of the observed defects in axon guidance actually reflect transformations of neu-ronal identity, with consequent confusion of axonal rout-ing. This is particularly a concern in the case of Notch, which is expressed in neurons throughout embryonic devel-opment. However, careful analysis of the embryonic PNS phenotypes of neurogenic mutations has provided no evi-dence for such changes in identity (Goriely et al., 1991). Furthermore, in the adult PNS, supernumerary neurons pro-duced as a consequence of a Notch mutation are wild type in the pattern of their projections (Simpson and Carteret, 1990). Analogous detailed analyses of cell fate in the CNS of neurogenic mutants have not been reported, potentially complicating the interpretation of the mutant CNS pheno-types. For Delta, we consider it unlikely that we were observing axonal defects stemming from errors in neuro-genesis, since late temperature shifts that did not produce the Delta neurogenic phenotype (neural hyperplasia) nonetheless resulted in defects in axon guidance (see, for example, Figs 1C, 2E) and those extra neurons that were produced appeared to be morphologically normal (Fig. 1C,D). Moreover, the level of Delta RNA in neurons is decaying rapidly at the stages of embryogenesis when the ISN is extending (Vaessin et al., 1987; Kopczynski and Muskavitch, 1989; E. G. and Y. N. J., data not shown).

Might other molecules contribute to guidance of the ISN by the trachea? While we have shown that Delta is neces-sary to define this nerve path, we have not tested whether it is sufficient. One candidate for a co-regulator is the gene Serrate, which encodes a protein that is similar in structure to Delta and is expressed in parts of the trachea (Fleming et al., 1990; Thomas et al., 1991). As is the case for Delta, Serrate interacts genetically with Notch and cells express-ing Serrate protein bind to Notch-expressing cells (Rebay et al. 1991). The embryonic Serrate mutant phenotype (Fleming et al., 1990) is rather different from the Delta and Notch phenotypes described here, but additional experi-ments will be required to determine whether Serrate plays a role in ISN guidance. Two lines of evidence argue that additional guidance information is indeed apt to reside in or near the trachea. First, expression of Delta RNA appears fairly homogeneous throughout the transverse part of the trachea (though we emphasize that we have not been able to examine the distribution of Delta protein, only of Delta RNA). Nonetheless, motor axons grow only in the dorsal direction and sensory axons grow only ventrally. We have not identified an explicit marker of directionality. Further-more, axons of the ISN remain associated with the trachea over only a short distance, about 30 microns for the case of the motor axons, yet Delta RNA is present throughout this branch of the trachea. We have not identified the signal that separates axons away from the trachea at the appro-priate point. Finally, Notch is expressed in the trachea, albeit at lower levels than it is in the nervous system. Our results do not exclude the possibility that this expression, or expression of Delta or Notch in tissues that we have not examined, may contribute to ISN guidance.

Notch protein is expressed throughout the ISN, yet Notch and Delta are required for the formation of only a short section of its trajectory. This observation suggests that the trachea-based Delta- and Notch-dependent guidance system may define an independent domain of guidance informa-tion, separable from the regulatory systems that guide the ISN through more dorsal and more ventral parts of the embryo. If this idea is correct, it should be possible to iden-tify other, genetically separable, guidance systems for other sections of the ISN. The notion that different steps in the path of a single axon may be guided by separate regulatory mechanisms is consistent with the specificity of mutant phe-notypes observed in the CNS in our experiments, as well as with results from other systems (Bastiani et al., 1987; Furley et al., 1990; Hedgecock et al., 1990; Grenningloh et al., 1991; McIntire et al., 1992).

The role that we propose for Delta and Notch in specifying nerve trajectory raises the question of how, mecha-nistically, these proteins control axon growth. In particular, it is striking that a pair of proteins that are central to the determination of the fates of many cells can produce such a different effect when active in axons. The Delta-Notch interaction could, in principle, guide the ISN by a purely adhesive mechanism (Hammerback et al., 1988), but we prefer the idea that Notch activates a specific intracellular effector system in axons (Gertler et al., 1989; Schuh et al., 1989; Silver et al., 1990). For example, it seems plausible that the signal transduction cascade that alters cell pheno-type in response to Notch action elsewhere in development, that is, the ‘neurogenic gene cassette’ (Ruohola et al., 1991), includes some of the molecules responsible for effecting Notch-dependent growth of the ISN along the trachea. By this view, the next critical question is to iden-tify those molecules that impart to the Delta-Notch regula-tory system the particular ability to promote directed axon extension in the current developmental context.

We are grateful to many members of our laboratory, especially Harald Vaessin and Michael Brand, for helpful discussions at all stages of these experiments. We particularly acknowledge Kirsten Bremer, Michael Brand, Hannele Ruohola and Harald Vaessin for much advice about the Delta and Notch experiments, Larry Ack-erman for assistance with confocal microscopy and in preparing the figures, and Sandy Barbel for performing the in situ hybridiza-tions. We thank Ralph Greenspan, Sandy Johnson, Tom Jongens, Elaine Ostrander, Marc Tessier-Lavigne, Andrew Tomlinson and Robin Wharton for reading drafts of the manuscript. E. G. is an Associate, and L. Y. J. and Y. N. J. are Investigators, of the Howard Hughes Medical Institute.