derailed is required for muscle attachment site selection in Drosophila

The somatic musculature of the Drosophila embryo and larva consists of a highly stereotyped arrangement of muscles. Each muscle is a single large, multinucleate fiber that can be uniquely identified by virtue of its size, morphology, epidermal attachment sites and eventual motorneuronal innervation. Many of the cellular events underlying muscle patterning are understood (for review, see Bate, 1993). The first visible signs of somatic muscle development are the fusion of mesodermal cells midway through embryogenesis at specific locations within the embryo (Bate, 1990). There is good evidence that individual muscle founder cells assume specific identities at appropriate locations and become competent to fuse with surrounding mesodermal cells, to which they impart their unique identities (Rushton et al., 1995). Such fusion events give rise to identifiable, multinucleate muscle precursors whose subsequent differentiation can be followed throughout development (Bate, 1990).

A number of observations suggest that interactions between muscles and the ectoderm play key roles in the patterning of somatic muscles. After gastrulation, myoblasts migrate laterally from the midline to assume positions in close contact with the ectoderm. As myoblast fusion proceeds, muscle precursors extend growth-cone-like processes which physically contact the underlying epidermis while navigating along specific routes toward their attachment sites. Following this extension, the establishment of appropriate epidermal attachment sites occurs, giving rise to each muscle’s characteristic orientation and length (Bate, 1990; Bier et al., 1990). Additional evidence for the importance of muscle-ectodermal interactions includes studies in which altering the developing ectoderm leads to disruptions in muscle patterning. For example, manipulation of the epidermis in the beetle, Tenebrio molitor, causes reproducible alterations in the pattern of somatic muscles (Williams and Caveney, 1980a,b). Likewise, mutations in Drosophila that disrupt normal epidermal development cause defects in muscle patterns (Bier et al., 1990; Volk and VijayRaghavan, 1994).

Although some of the molecular mechanisms underlying early mesodermal development are beginning to be uncovered (e.g., Corbin et al., 1991; Gonzalez-Crespo and Levine, 1993; Bladt et al., 1995), very little is known about how individual somatic muscles acquire their unique morphologies. In Drosophila, the stripe (sr), Toll (Tl) and myospheroid (mys) genes have been shown to be involved more generally in muscle attachment. During embryonic and pupal development, those epidermal cells destined to become sites for somatic muscle attachment express sr (Volk and VijayRaghavan, 1994; Lee et al., 1995a; Fernandes et al., 1996), which encodes a protein similar to members of the early growth response family of transcription factors (Lee et al., 1995a). sr appears to be required for proper muscle attachment, as loss-of-function mutations in the gene cause disruptions of muscle attachment and patterning for both larval (Volk and VijayRaghavan, 1994) and adult (Costello and Wyman, 1986; de la Pompa et al., 1989) somatic muscles. The cell adhesion molecule Toll is widely expressed within the epidermis and by a small subset of muscles (Nose et al., 1992). Loss of zygotic Toll expression results in a variety of somatic muscle abnormalities, including muscle loss, duplication and defective attachment (Halfon et al., 1995). Since these abnormalities are widespread and not exclusive to the Toll-expressing muscles, proper muscle patterning may depend on Toll function within the epidermis. Finally, the integrity, but not the initial recognition and establishment, of muscle attachments requires the integrin β subunit encoded by the mys gene (MacKrell et al., 1988; Leptin et al., 1989).

We previously isolated the derailed (drl) gene in a screen for genes controlling axon guidance (Callahan et al., 1995). drl encodes a receptor tyrosine kinase (RTK) family member expressed on a subset of developing embryonic neurons. In drl mutant embryos, the drl-expressing neurons fail to extend along their appropriate pathways, suggesting that the Drl RTK is an essential component in the recognition between growing neurons and their pathways. Here we show that the Drl RTK is also expressed by a small subset of embryonic muscle fibers and neighboring epidermal cells during muscle growth and attachment events. In drl mutants, drl muscles attach at abnormal locations within the epidermis. These muscle attachment defects are not the result of gross alterations of the epidermis nor loss of epidermal attachment cell precursors. In contrast, our results suggest that analogous to its role in axon pathway selection within the nervous system, Drl participates in a mechanism required for muscle attachment site recognition.

All fly strains were grown on standard cornmeal medium at 18 or 25°C. The enhancer trap line P1618 was generated by T. Volk. The D82 GAL4 enhancer trap line was generated by D. Lin and C. S. Goodman. The generation and identification of the drl^P3.765 mutation was previously described (Callahan and Thomas, 1994; Callahan et al., 1995). Additional fly stocks and chromosomes are described in Lindsley and Zimm (1992). Null alleles of drl were obtained by imprecise excision of the drl^P3.765 P element using P[ry⁺Δ2-3](99B) as a source of transposase (Robertson et al., 1988). w¹¹¹⁸; drl^P3.765 homozygous females were mated to CyO; y⁺P[ry⁺Δ2-3](99B) Sb/TM6 males. From this cross, groups of w¹¹¹⁸; drl^P3.765 /CyO; y⁺P[ry⁺Δ2-3](99B) Sb/+ males were crossed to w¹¹¹⁸; CyO/T(2;3)ES females in 550 bottles. Approximately 150,000 flies were screened for lack of w⁺ in the subsequent generation and, from these, 122 independent excision events were isolated and balanced over CyO, ActinlacZ (Bourgouin et al., 1992). DNA from each of the 122 P element excision lines was digested with XhoI and BglII, blotted, and probed with a 3.1 kb XhoI fragment that encompasses the P element insert (Callahan et al., 1995). Two excision lines, drl^R300 and drl^R343, were found to completely remove the drl coding sequence and are homozygous viable. drl^R300 and drl^R343 were tested for complementation with l(2)37Da¹, l(2) 37Cf and fs(2) TW1 (Wright et al., 1981; Wright, 1987; Gay and Contamine, 1993); both complemented the lethality or female sterility of these loci.

Standard methods were used to construct all plasmids (Sambrook et al., 1989). pUAS-tau-lacZ was constructed by ligating a 5.4 kb EcoRI fragment of pBStau-lacZ with pUAST (Lin et al., 1994) digested with EcoRI. pUAS-drl was constructed by ligating a 3.1 kb EcoRI fragment of a drl cDNA (Callahan et al., 1995) with pUAST digested with EcoRI. P[UAS-tau-lacZ] and P[UAS-drl] were introduced into the fly germ line by standard P element transformation methods (Rubin and Spradling, 1982). For both constructs, multiple independent transformants were obtained. Lines of interest were then made homozygous if viable and fertile, or balanced over FM7c, CyO or TM3. A second chromosome P[UAS-drl] insertion, UAS-drl19, was used in the GAL4 transactivation studies. P[ME4-lacZ] was generated and generously supplied by J. Botas. pME4-drl was constructed by cloning a 1.4 kb XhoI fragment of the apterous gene present in P[ME4-lacZ], into the XhoI site upstream of the hsp70 promoter of CaSpeR2/17 (Nose et al., 1994); subsequently, a 3.1 kb EcoRI fragment of a drl cDNA (Callahan et al., 1995) was blunt-ended and cloned into a blunted XbaI site downstream of the hsp70 promoter. P[ME4-drl] was introduced into the germ line and balanced as above. Multiple independent transformants were obtained. A second chromosome insertion, ME4-drl1, was used in the studies described.

Whole-mount in situ hybridizations were performed as previously described (Tautz and Pfeile, 1989) with modifications (Jiang et al., 1991). Digoxigenin-labeled sense and antisense riboprobes of drl were prepared from the 3.1 kb drl cDNA cloned in Bluescript (Stratagene) using protocols from Boehringer Mannheim. After staining, embryos were mounted directly in 90% glycerol or dehydrated in an ethanol series, briefly rinsed in xylenes and mounted in Permount.

All collections for dissections were carried out at 25°C. Embryo dissections were performed as described (Thomas et al., 1984; Callahan and Thomas, 1994). For detection of β-gal and Tau-β-gal, embryos were incubated overnight at 4°C with a rabbit anti-β-gal polyclonal antibody (Cappel) diluted 1:10,000 in PBTN (phosphate-buffered saline containing 0.1% Triton X-100 with 1% bovine serum albumin and 4% normal goat serum). Embryos were washed and incubated for 2 hours at room temperature with either a FITC-conjugated goat antirabbit antibody (Cappel) diluted 1:200 in PBTN or a biotinylated goat anti-rabbit antibody (Vector) diluted 1:200 in PBTN. Preparations processed with biotinylated antibodies were washed again and incubated for 1 hour with either fluorescein streptavidin diluted 1:200 (Vector labs) or an avidin/biotin-HRP complex (Vectastain ABC Elite Kit, Vector labs) followed by a final wash. HRP staining was visualized by a standard DAB reaction. Each wash was performed for 30 minutes in PBT (phosphate-buffered saline containing 0.1% Triton X100 with 1% bovine serum albumin).

For detection of Drl, rat polyclonal anti-Drl antibodies directed against Drl’s extracellular domain (Callahan et al., 1995) were used at a dilution of 1:500. Incubation and washing were followed by a biotinylated goat anti-rat antibody (Vector labs) diluted 1:200 in PBTN, another wash, a 1 hour incubation in either fluorescein streptavidin diluted 1:200 (Vector labs) or an avidin/biotin-HRP complex (Vectastain ABC Elite Kit, Vector labs) followed by a final wash. Before analysis, embryos were either mounted in 100% glycerol or dehydrated in an ethanol series, cleared with methyl salicylate and mounted in Canada balsam. Somatic muscle patterns were examined by incubating dissected embryos in RITC-conjugated phalloidin diluted 1:1000 (Molecular Probes) in room temperature PBTN for 30 minutes, followed by washing and mounting in a solution of 10% polyvinyl alcohol (Air Products and Chemicals) containing 2.5% DABCO, 1,4-diazabicyclo[2.2.2]octane (Sigma).

Fluorescent-labeled preparations were imaged using a BioRad MRC1000UV confocal microscope coupled to a Zeiss Axiovert 135M microscope. Images were collected using the T1/E2 block combination; each wavelength was collected using a separate and specific excitation filter and imaged on separate photomultiplier tubes using COMOS software (Biorad). Bright-field and fluorescent digital images were processed using Photoshop (Adobe Systems Inc.).

In addition to its expression within the CNS, drl is expressed in both the mesoderm and epidermis during embryogenesis. Using antibodies directed against its extracellular domain (Callahan et al., 1995), Drl protein can be detected in a subset of somatic muscles as they grow and form attachments to the epidermis. Each hemisegment in abdominal segments A2 through A7 contains 30 muscles with predictable orientations and attachments to the epidermis (Fig. 1). Beginning at hour 10 of development, low levels of Drl are detected in the precursors for muscles 21–23 as they begin to elongate dorsoventrally along the epidermis (Fig. 2A). Expression increases as these muscles continue to grow toward their attachment sites, and is still present by hour 13 as assayed by in situ hybridization (Fig. 2B) or antibody staining (Fig. 2C). During the period of attachment site selection, Drl appears enriched near the tips of muscles 21–23, especially at the locations where they contact their ventral attachment sites (Fig. 2C). By hour 15, after attachment events are completed, drl transcripts and protein are no longer detectable.

Drl protein is first detected at approximately hour 6 as stripes 3–4 cells wide in each segment. During segmental groove formation, Drl is restricted to anterior cells of each segment near the segmental grooves. At hour 9.5, expression begins to expand posteriorly from the grooves at lateral positions, resulting in broad patches of Drl-expressing epidermal cells. These lateral Drl patches overlie the differentiating muscle precursors 21–23 (Fig. 3A). By hour 11, the lateral Drl patches become more restricted, forming two smaller clusters of Drlexpressing epidermal cells located near the dorsal and ventral attachment sites for muscles 21–23 (Fig. 3B). This pattern subsequently becomes more refined, such that by hour 12.5, each cluster has become restricted to approximately 15 cells which abut and partially overlap the epidermal attachment cell clusters for muscles 21–23, as revealed by double-labeling for Drl and sr expression (see below) (Fig. 3C). As in muscles 2123, Drl ceases to be expressed in the epidermis by hour 15. Despite this pattern of drl epidermal expression, drl null mutants do not show any defects in segmental groove formation or in differentiation of the epidermis (see below).

To better understand the relationship between muscles 21-23 and their attachment sites within the epidermis, we examined muscle development in embryos carrying a single copy of P1618, a nuclear-targeted lacZ P element within the sr gene (kindly provided by T. Volk). Epidermal cells destined to become muscle attachment cells express sr before muscle attachment events (Volk and VijayRaghavan, 1994; Fernandes et al., 1996), and thus staining for β-gal expression in P1618/+ embryos allows the identification of attachment cells during myogenesis.

At hour 10, P1618/+ embryos begin to express β-gal at positions where muscles will eventually insert. This includes sites at segmental boundaries where longitudinal muscles will attach, ventral rows of cells parallel to the segmental grooves where ventral muscles will attach, and in small clusters of cells located laterally between segment boundaries where the 21–23 group will attach. In hour 11.5 embryos, muscles 21–23 have extended dorsoventrally and can be seen forming both their dorsal and ventral attachment sites within the lateral clusters of sr-lacZ-expressing epidermal cells (Fig. 4A). By hour 14, the sr-lacZ clusters have become restricted to 6–8 cells, and the ventral attachments for muscles 21–23 lie adjacent to one another within the cluster (Fig. 4B).

Normally, muscles 21–23 extend similar distances ventrally and attach adjacent to the dorsal border of muscle 12, always inserting within the sr-lacZ epidermal cluster (Figs 1, 4A,B). In drl mutants, muscles 21–23 are present in their normal locations and elongate dorsoventrally, but have ventral attachment site defects. We examined muscle morphology in two drl null alleles, drl^R300 and drl^R343, both of which remove all drl coding sequences (Callahan et al., 1995). Both alleles show the same muscle phenotype. In 20% of hemisegments of drl mutants, one or more muscles of the 21–23 group pass over their normal ventral attachment sites and appear to attach far more ventrally beyond muscle 13 (Fig. 4C,D; Table 1). This dramatic phenotype we term the ‘bypass’ phenotype. An additional 10% of hemisegments have more subtle attachment defects where at least one muscle fails to attach within the ventral sr-lacZ cluster, but does not extend beyond muscle 13 (Fig. 4D).

Importantly, both the numbers and locations of the sr-lacZ-expressing epidermal cells in drl mutants are indistinguishable from drl⁺ embryos. This indicates that the muscle phenotype seen in drl mutants is not the result of a loss of epidermal attachment cell clusters, but instead results from the inability of the muscles and their ventral tendon cells to coordinate functional attachments.

To verify that loss of drl function causes the muscle attachment defects in drl mutants, we attempted to rescue the muscle phenotype by targeted expression of a drl cDNA using the GAL4 transactivation system (Fischer et al., 1988; Brand and Perrimon, 1993). One of the UAS-drl insertions that we generated, UAS-drlⁱ, due to position effect, expresses drl widely in the embryo independently of GAL4 (data not shown). drl^R343 mutant embryos carrying two copies of UAS-drlⁱ show nearly complete rescue of the drl muscle ‘bypass’ phenotype (Table 1), clearly demonstrating that the muscle phenotype of drl mutants is due to loss of drl function.

To address the question of where drl function is required, we expressed it in muscles using the GAL4 enhancer trap line D82 (generously provided by D. Lin and C. S. Goodman). D82 embryos express GAL4 in developing somatic muscles as assayed by the ability to transactivate UAS-tau-lacZ and UAS-drl. Embryos from a cross between D82 and UAS-tau-lacZ flies (see Materials and Methods) express Tau-β-gal specifically in muscles with no detectable expression in the epidermis. Tau-β-gal expression commences at approximately hour 10, during the period when muscles 21–23 are growing and selecting attachment sites within the epidermis (Fig. 5A) and continues through the period when attachments have been established (Fig. 5B). In these D82/UAS-tau-lacZ embryos, we observed variability from segment to segment in the expression of the Tau-β-gal reporter. For example, within the anterior hemisegment shown in Fig. 5B, muscle 22 fails to express detectable levels of the marker, while other muscles express high levels. Thus, as has been previously reported in studies using GAL4-mediated expression (Lin and Goodman, 1994), there appears to be variability in GAL4 transactivation in these embryos.

To assay the rescuing ability of D82 GAL4-mediated expression of drl, we crossed drl^R343,D82 to drl^R343,UAS-drl flies. The drl^R343,D82/drl^R343,UAS-drl progeny from this cross express Drl in a manner similar to that described above for Tau-β-gal (Fig. 5C). drl^R343,D82/drl^R343,UAS-drl embryos show significant but partial rescue of the ‘bypass’ phenotype (Table 1). In contrast, drl^R343 embryos carrying either D82 or UAS-drl alone show muscle defects at frequencies similar to drl^R343 homozygotes (Table 1). Rescue cannot be due to a cryptic interaction between GAL4 and UAS sequences because drl^R343 embryos carrying both D82 and UAS-tau-lacZ do not show any rescue of the phenotype.

To overcome the variability of GAL4 transactivation, we created a direct fusion of the drl cDNA to ME4, the muscle enhancer of the apterous gene (generously provided by J. Botas). When used to drive either lacZ or a drl cDNA, ME4 directs high levels of expression in muscles 21–24 and low levels in muscles 18, 27 and 29 during the period of muscle growth and attachment site selection (Fig. 5D). Slightly later in development, ME4 also drives expression in muscle 8 and in a small subset of tracheal cells (Fig. 5E). As with D82, we could not detect any ME4-driven expression in the epidermis. However, in contrast to D82-mediated expression, ME4mediated expression in muscles 21–24 is highly reproducible, with high levels of expression in every segment examined. Mutant embryos homozygous for drl^R343 and carrying two copies of ME4-drl show complete rescue of the drl muscle bypass phenotype (Table 1; Fig. 5F). Thus, Drl expression in muscles alone appears to be sufficient for rescue, strongly suggesting that during attachment site selection Drl functions in the muscles.

We have shown that the Drl RTK is expressed by a small subset of somatic muscles and neighboring epidermal cells during their attachment to the epidermis and that, in embryos homozygous for null mutations of drl, these muscles often fail to form ventral attachments at correct locations. Loss of drl function is responsible for these defects since the muscle phenotype can be rescued by expression of a drl cDNA.

We addressed the possibility that the ventral attachment site defects seen in drl mutants might be the result of a corresponding loss of the ventral epidermal tendon cells. This would not be surprising as drl is co-expressed in at least some sr-lacZexpressing cells during attachment events and previous studies have shown that other RTK family members play roles in cell autonomous fate decisions (Hafen et al., 1987; Sprenger et al., 1989; Aroian et al., 1990). Yet, the fact that we found no differences in sr-lacZ expression in drl mutants indicates that muscle misattachment is not due to the loss of attachment cells or their precursors. Indeed, mutant muscles appear to have access to, but bypass, one or more sr-lacZ-expressing cells as they project ventrally towards inappropriate attachment sites. Our results are more consistent with a role for drl in the muscle-epidermal interactions that underlie proper attachment site selection. One possibility is that the Drl RTK participates directly in recognition events between muscles 21–23 and their ventral attachment cells. Drl itself may mediate specific contacts between muscles 21–23 and their appropriate attachment site targets. Alternatively, Drl may be more indirectly involved in muscle-epidermal recognition events. For instance, Drl function may be required in muscles 21–23 before attachment site recognition events, serving to regulate or modify other gene products that are more directly involved in recognition processes. Such a possible modulatory role for Drl in muscle-epidermal recognition would be similar to the proposed role for Drl in neuronal pathway recognition events (Callahan et al., 1995). A further possibility is that Drl is not required for muscle-epidermal recognition events, but instead is required for the subsequent cytoskeletal changes that must accompany successful attachment. In this scenario, drl mutant muscles would fail to attach at their appropriate sites because they are unable to physically anchor at any location. This possibility seems less likely considering that those muscles displaying a mutant phenotype do appear eventually to attach to the epidermis, albeit at incorrect ventral locations.

The fact that muscle defects are not 100% penetrant in drl null mutants illustrates that there are additional genes involved in muscle 21–23 attachment site selection. At present, we do not know what relationship these additional genes have to drl. One possibility is that the products of these genes are somehow modified by activation of Drl, yet are not completely dependent on Drl for their function. Alternatively, the fidelity of muscle 21–23 attachment site selection may arise from the combinatorial functioning of several distinct recognition processes of which Drl controls only one. Interestingly, misexpression of Drl in muscles that normally do not express Drl does not cause mistargeting to the 21–23 attachment sites. This may be due to a lack of competence of inappropriate muscles to respond to the Drl signal, or may reflect a restricted localization of the Drl ligand. Misexpression of a constitutively active form of Drl may help to resolve this point.

Our finding that Drl is expressed both in muscles and at epidermal locations along which the muscles grow and attach suggests that there may be both an epidermal and a mesodermal component of Drl function. For example, Drl could participate in a homotypic interaction mediating some form of recognition during muscle growth over the epidermis. Our results argue against this possibility, since we were able to rescue the muscle attachment phenotype by targeted expression of Drl in muscles. The fact that we obtained only partial rescue with D82 is likely due to variability of Drl expression in these embryos, since complete rescue could be achieved with more reproducible expression from the ME4 enhancer. These results strongly suggest that Drl function in muscles 21–23 is sufficient for proper attachment site selection, although we cannot rule out the possibility that undetectable levels of Drl in the epidermis of ME4-drl and D82/UAS-drl embryos might be sufficient to rescue the mutant phenotype.

Regardless of drl’s exact role in muscle attachment site selection, these studies afford insights into the mechanisms underlying the precision of muscle organization. First, our results suggest that similar to axonal pathfinding in the developing nervous system, specific genes are required for discrete steps during target recognition by muscles. Second, these studies of Drl further demonstrate in vivo that RTK family members are essential for specific aspects of muscle differentiation. Other examples include the mouse c-met RTK, which is involved in myoblast migration into the limb (Bladt et al., 1995) and the ErbB4 and ErbB2 receptors, which are required for cardiac muscle differentiation (Gassmann et al., 1995; Lee et al., 1995b). Since RYK, the vertebrate homolog of Drl, is also expressed in muscles (Hovens et al., 1992), it will be of interest to determine its possible role in mesodermal development. Finally, the fact that in drl mutants, the neurons that normally express Drl fail to recognize their appropriate target pathways, raises the intriguing possibility that both neurons and muscles use similar mechanisms to recognize their paths or targets, and that drl plays an analogous role in both developing systems.

We thank S. Lundgren for technical assistance; T. Volk for the enhancer trap line P1618; J. Botas for providing ME4; D. Peterson and F. Gage for assistance and advice with confocal imaging; D. Lin and C. S. Goodman for providing pUAST and the D82 GAL4 enhancer trap line. We are also thank G. Lemke and S. Thor for stimulating discussion. This work was supported by grants from the NIH, a March of Dimes Basil O’Connor Scholar Research Award and a Pew Scholars Award from the Pew Memorial Trusts to J. B. T., and from NIH/NIGMS Training Grant PHSGM07198 to C. A. C., who is a student in the Medical Scientist Training Program at UC San Diego.