Founder myoblasts and fibre number during adult myogenesis in Drosophila

In many organisms, muscles consist of syncitial fibres or myotubes formed by the fusion of embryonic myoblasts. The formation of syncitia increases the dimensions of muscle fibres, allowing them to operate as appropriately sized contractile elements in the locomotor machinery of the animal concerned. In the Drosophila embryo, the muscles of the larva form as single myotubes of different sizes that insert onto the developing epidermis. The key to understanding the specification and assembly of this complex pattern lies in the founder myoblasts that seed myotube formation. Each founder recruits and fuses with neighbouring myoblasts to form the syncitial precursor of a myotube (Bate, 1990; Bate, 1993; Abmayr et al., 1995; Rushton et al., 1995; Baylies et al., 1998; Paululat et al., 1999a). The characteristics of these myotubes are largely, if not completely, determined by transcription factors expressed in particular founder myoblasts. In adult flies and many vertebrates, additional complexity is added to the specification of muscle properties by the fact that muscles consist of many myotubes bundled together to form a contractile element. In systems such as these, the properties of myotubes must be specified, but in addition, the number of myotubes contributing to the developing muscle must be set as well. It is possible to envisage ways in which the founder-feeder model of myotube patterning could form the basis for setting the characteristics of such multi-fibre arrays.

We have examined the formation of multi-fibre muscles, i.e. muscles that function together as a contractile unit, in the adult fly(Fig. 1). In the mesothorax of the adult, the most prominent muscles are the indirect flight muscles (IFMs),whose development has been charted in some detail(Fernandes et al., 1991; Roy and VijayRaghavan, 1999). The IFMs consist of the dorsal longitudinal muscles (DLMs), an array of six large fibres, and three groups of dorsoventral muscles: DVM-I (three fibres),DVM-II (two fibres) and DVM-III (two fibres). The mesothorax contains another large muscle, namely the tergal depressor of the trochanter (TDT) or jump muscle, which consists of many fibres bundled together as a unit. The dorsal thorax also contains the direct flight muscles (DFMs), involved in changing the orientation of the wing. Each of these muscles is a multi-fibre contractile unit. The muscles in the adult abdomen are also arranged as well-defined sets of fibres, which form dorsally, laterally and ventrally in each segment (Fig. 1)(Currie and Bate, 1991).

By examining the formation of these muscles, we provide evidence that specialized founder myoblasts are a feature of adult as well as embryonic myogenesis in the fly. As adult muscles, unlike the somatic muscles of the embryo, are composed of multiple fibres, we discuss the question of how several founders are chosen for each multi-fibre contractile unit.

To follow duf expression, the enhancer trap line rp298,which has a P element-nuclear-lacZ insertion in the promoter region of duf, was used (Ruiz-Gomez et al., 2000). The GAL4-UAS system(Brand and Perrimon, 1993) was used for directed expression of genes during adult myogenesis. The 1151-GAL4 enhancer trap strain was obtained from L. S. Shashidhara(CCMB, Hyderabad, India). This GAL4 driver is expressed in all adult myoblasts associated with the imaginal discs and nerves in the larvae. No expression is seen in the larval templates (see Anant et al., 1998). UAS-dnRac1 was obtained from the Bloomington Stock Centre. UAS-activated Notch (UAS-Nintra), UAS-dnNotch (Rebay et al.,1993), UAS-MamH and UAS-MamN were obtained as gifts from S. Artavanis-Tsakonas (MGH Cancer Centre, USA). The various UAS strains were put in the background of rp298 using standard genetic techniques. For co-localization of 1151 and duf-lacZ,1151-GAL4 was crossed with rp298; UAS-nuclear localized-GFP. Most fly cultures and crosses were grown at 25°C. The progeny of the GAL4-UAS crosses (with the exception of 1151×rp298;UAS-nls-GFP) were grown at 25°C until early second instar stages, and then shifted to 29°C until the adult stages. Controls were grown in similar conditions.

We found that the rate of development at 22°C was 0.75 times that at 25°C, and that at 31.5°C was approximately 1.3 times that at 25°C(Anant et al., 1998). Based on this, 0-hour APF pupae were grown at 22°C for 21 hours and 20 minutes(corresponding to 16 hours APF at 25°C), 24 hours (≡18 hours at 25°C), 26 hours and 40 minutes (≡20 hours APF at 25°C), 29 hours and 20 minutes (≡ 22 hours at 25°C), and then raised to 31.5°C for 6 hours, 4 hours and 30 minutes, 3 hours, and 1 hour and 30 minutes,respectively.

White prepupae (0-hour APF) were collected on moist filter paper in a Petri dish and grown at 25°C for different intervals. For the GAL4-UAS crosses and their controls, white prepupae were collected and grown at 29°C. The pupal and larval tissues were prepared for immunohistochemistry as described previously (Fernandes et al.,1991). The pupal preparations were mounted in 70% glycerol for X-GAL stained preparations, or in Vectashield mounting medium (Vector Laboratories, Burlingame, CA) for fluorescently labeled preparations.

Anti-Ewg antibody raised in rabbit was used at a dilution of 1:500(DeSimone et al., 1996). Anti-β-galactosidase monoclonal antibody and 22C10 (both from The Developmental Studies Hybridoma Bank) were used at a dilution of 1:50. Anti-Twist antibody, a gift from Siegfried Roth (University of Köln), was used at a dilution of 1:500. Vestigial antibody was a gift from Sean Carroll and was used at a dilution of 1:200. Anti-MHC antibody, a gift from Dan Kiehart, was used at a dilution of 1:500. For double-antibody stained preparations of 1151-GAL4, duf-lacZ; UASdnRac1/+ pupae(Fig. 7A-E), pupae were incubated first in anti-MHC antibody and developed in the absence of nickel sulphate (giving a light brown colour), then incubated in anti-β-galactosidase and developed in the presence of nickel sulphate(giving a black colour). For preparations stained both with X-GAL and antibody(Fig. 7F-I), pupae were first incubated overnight in X-GAL, then processed for antibody staining. DIC images were taken using a Nikon Eclipse E1000 microscope. For fluorescent images,secondary antibodies conjugated to Alexa Fluor dyes (from Molecular Probes,Eugene, OR) were used: Alexa 488 for green labeling and Alexa 568 for red labeling. Fluorescent preparations were scanned using the confocal microscope(MRC-1024, BioRad Laboratories, Hercules, CA) and analyzed using Metamorph(version 4.5, Universal Imaging).

We began our investigation of myotube formation in the adult by looking at a prominent set of muscles, the DVMs (Fig. 2A,B), which form de novo from groups of aggregating myoblasts in the adult thorax (Fernandes et al.,1991). We wanted to see how the three different DV muscles are formed from these myoblasts and how each muscle attains its typical fibre number. It could be that, as in the embryo, each fibre is seeded by the formation of a single, specialized founder cell. Alternatively, myoblasts might be selected in some other way; for example, as a group that would fuse to form a single myotube. To distinguish between these two alternatives, we used a P-lacZ insertion (rp298) in the gene dumbfounded(duf; kirre - FlyBase), which results in the expression ofβ-galactosidase in the nuclei of cells that express duf. dufitself encodes a transmembrane protein that, in the embryonic mesoderm, is exclusive to founder myoblasts and serves as an attractant for fusion competent myoblasts (Ruiz-Gomez et al.,2000). duf-lacZ expression is therefore a good marker for such cells, and we examined when and where duf-lacZ is expressed as myoblasts aggregate to form the DVMs.

At 7 hours after puparium formation (APF), at each epidermal location where a DVM fibre will develop, we see a single prominent myoblast that expresses duf-lacZ. Thus, we see three, two and two duf-lacZexpressing cells, corresponding, respectively, to the three, two and two fibres of the future DVM I, DVM II and DVM III muscles(Fig. 2C-E). These nuclei continue to express duf-lacZ at high levels as each of the syncitial fibres in the DVMs form (Fig. 2F). In addition, the fused adult myoblast nuclei within the syncitia show low levels of duf-lacZ expression(Fig. 2F). By 36 hours APF, the level of β-galactosidase in all DVM nuclei begins to fall and disappears by 70 hours APF.

These findings suggest that, just as in the embryo, the formation of myotubes in the adult may be initiated by the selection of single founder myoblasts that are identifiable by their expression of duf-lacZ. Other myoblasts would be recruited to these founders and fuse with them and(again, as in the embryo) these fusing cells would themselves be induced to express duf-lacZ, albeit at a lower level. If this view of adult fibre formation is correct, then it should be generally true for all cases in the adult where fibres form de novo from groups of aggregating myoblasts. With this in mind, we looked at the regular arrays of fibres that form dorsally and laterally in the adult abdomen. There are many such fibres laid out in a well-organized pattern and they are derived from myoblasts that, unlike the DVM cells, come not from the discs but from pools of cells associated with the abdominal nerves (Currie and Bate,1991).

The formation of the syncitial muscles in the abdomen begins at about 28 hours APF (Currie and Bate,1991). We looked at a stage prior to this to see whether single duf-lacZ-expressing cells appear before fibres form. Once again, we observed a striking correspondence between forming fibres and duf-lacZ expression in the abdomen, with every fibre preceded by a single duf-lacZ-expressing nucleus at the appropriate position. At 24 hours APF, we observe an array of duf-lacZ-expressing cells in each of the dorsal hemisegments. One of the hemisegments (A4) is shown in Fig. 3A. The monoclonal antibody 22C10 also labels these cells(Fig. 3B,C). By 28 hours APF,these cells are in positions where the future muscle fibres will form(Fig. 3D). By 50 hours APF,when the formation of the syncitial fibres is largely complete, we observed one nucleus in each fibre that expresses duf-lacZ at higher levels than the rest (Fig. 3E). We observe a similar pattern of duf-lacZ expression in the developing lateral muscles. At positions where the future lateral muscles will form there are single duflacZ-expressing cells that are also 22C10 positive(Fig. 3F). These mononucleate cells develop into multinucleate fibres (see Fig. 3G,H), which each contain several duf-lacZ-expressing nuclei.

On the face of it, the development of the DLMs follows a different scheme. Here, the muscles assemble on a set of preexisting templates provided by a small set of persistent larval fibres, the three larval oblique muscles (LOMs)(shown in Fig. 4B). Subsequently, the three templates split to form the six fibres of the DLMs(Fig. 1, Fig. 4C)(Fernandes et al., 1991). Despite this novel form of myogenesis, it has been suggested that the larval fibres serve a founder-like function in organizing the development of the DLMs(Farrell et al., 1996), and it is certainly true that the adult myoblasts aggregate on the larval fibres and fuse with them to form syncitial myotubes. With this in mind, we decided to look at the expression of duf-lacZ during DLM development. We found that duf-lacZ is indeed expressed in the founder analogues, the larval templates of the adult muscles. duf-lacZ expression begins in the three larval templates at 6.5 hours APF(Fig. 4D). All of the nuclei in each of the three templates appear to express duf-lacZ. The expression of duf-lacZ continues as the templates split and form the six fibres of the DLMs (Fig. 4E-G). By 36 hours APF, the intensity of β-galactosidase staining starts to decrease, and by 70 hours APF the staining has disappeared. The duf-lacZ-expressing nuclei in the templates could be larval nuclei or the nuclei of adult myoblasts that have already fused with the templates by 6.5 hours APF. Larval nuclei are large and easily distinguishable from the smaller nuclei of the adult myoblasts(Crossley, 1978). The transcription factor Erect wing (Ewg)(DeSimone et al., 1996) is expressed in the nuclei of the larval muscles and this expression persists as the larval muscles are transformed to templates. Ewg is also expressed in adult myoblasts as they migrate over the templates to fuse with them(Roy and VijayRaghavan, 1998). These two subsets of nuclei, both expressing Ewg, can be easily distinguished on the basis of their size (as can also be seen in Fig. 4I). A double label using antibodies against β-galactosidase and Ewg reveals that duf-lacZ-expressing nuclei co-localise with Ewg-expressing nuclei,which by their size are larval in origin(Fig. 4H-J).

If duf-lacZ expressing founder myoblasts are, as in the embryo, a decisive factor in controlling the spatial pattern of adult myogenesis, then we need to understand how this pattern of expression develops and is controlled. We first looked at the emergence of duf-lacZ expression in the myoblasts of the abdomen. The abdominal muscles develop from pools of muscle-forming cells that are associated with the nerves that innervate the larval muscle field (Bate et al.,1991; Currie and Bate,1991). These cells in turn are derived from single adult muscle progenitors that arise in the embryo as the siblings of embryonic muscle founder cells (Ruiz-Gomez and Bate,1997). Unlike their embryonic founder cell siblings, the adult precursors maintain twist expression and proliferate during larval life to form pools of nerve-associated, twist-expressing myoblasts. The adult precursors do not express duf-lacZ in the embryo(Ruiz-Gomez et al., 2000). These myoblasts then migrate, aggregate and fuse to form the muscles of the adult abdomen (Fig. 1). When we looked at the expression of duflacZ in these precursor cell populations in late third instar larvae, we found that all the cells express duf-lacZ at a low level (Fig. 5A,B). This expression persists during early pupal stages (13-14 hours APF; Fig. 5C) but, by 16 hours APF, levels of β-galactosidase in the myoblasts have declined and,at 20 hours APF (Fig. 5D), only a subset of Twist-positive myoblasts remains β-galactosidase positive. By 24 hours APF, however, there is a clear upregulation of duf-lacZexpression in particular nuclei (Fig. 5E), and by 28 hours APF, these duf-lacZ-expressing cells are positioned at the sites where individual muscle fibres will now form(Fig. 5F). These are the same nuclei that are present in 22C10-stained cells, as shown previously in Fig. 3D, and that continue to express high levels of duf-lacZ within the fully formed fibres(Fig. 3E). We find a similar sequence of duf-lacZ expression in the lateral clusters of myoblasts(data not shown). Expression does not persist into the adult and begins to diminish by 70 hours APF.

We find a similar sequence of expression in the myoblasts of the wing disc that give rise to the DLMs and DVMs(Lawrence, 1982). In late third instar larvae, duf-lacZ expression can be detected at low levels in all of these cells (Fig. 6A-C). To follow this pattern of expression in the pupa, we double-labeled duf-lacZ pupae with markers for adult myoblasts, using either UAS-GFP driven by 1151-GAL4, which is expressed in all adult myoblasts (Anant et al.,1998), or antibodies to the transcription factor Vestigial (Vg),which is expressed in most of the myoblasts on the wing disc(Sudarsan et al., 2001). These stainings reveal that the generalized low level of duf-lacZexpression had disappeared by 12 hours APF and had been replaced by expression in specific cells. For example, Fig. 6D shows 1151-driven GFP expression in myoblasts of DVM II. duf-lacZ expression is seen in a single nucleus in each of the two fibres of DVM II (Fig. 6E). Fig. 6F, a merged image, shows duf-lacZ-expressing nuclei co-localizing with GFP expression. Double labeling with antibodies against Vg and β-galactosidase again shows co-expression in specific nuclei (Fig. 6G-I). In a similar fashion, duf-lacZ expression is downregulated in myoblasts that will fuse with the larval templates: Fig. 4H,I show the absence of duf-lacZ in the myoblasts that have migrated over the larval templates. We conclude that, as in the abdomen, the duf-lacZ-expressing founders of the thoracic muscles other than the DLMs are derived from the pool of adult myoblasts, all of which initially express duf-lacZ at low levels.

The decisive function of muscle founder cells in myotube formation is revealed in the embryo by their unique capacity to form muscles in mutants where myoblast fusion is compromised. Thus in embryos that are defective in the machinery of myoblast fusion, founder myoblasts differentiate to form thin mononucleate fibres at appropriate positions and with molecular characteristics similar to those of normal wild-type muscles(Rushton et al., 1995). Other myoblasts aggregate on the differentiating founders but cannot fuse with them and do not succeed in differentiating to form muscle fibres themselves. These observations show that, in the embryo, myoblasts are of two kinds: founders,which are competent to complete myogenesis in the absence of fusion, and feeders, which can only contribute to myogenesis after fusion with a founder.

To show whether there is a similar division of myoblasts during adult myogenesis, we generated a fusion-defective phenotype during adult myogenesis,by overexpressing a dominant-negative form of the protein Rac1, a member of the small GTPase superfamily involved in the process of fusion(Paululat et al., 1999b). Overexpression of the dominant-negative Rac1 (Rac1N17) in the embryonic mesoderm severely delays the fusion process and results in abnormal fusion in the later stages (Luo et al.,1994). We found that overexpression of Rac1N17 in the adult myoblast pool severely reduces myoblast fusion, the effect being most dramatic in the lateral muscles of the abdomen and, to a lesser extent, in the thoracic muscles. Nevertheless, duflacZ-expressing myoblasts are present in the correct number at the correct positions. Preparations of wild-type and 1151GAL4/UAS-Rac1N17 pupae are shown(Fig. 7A,B). In the absence of fusion, each putative founder cell begins to express myosin at the appropriate stage, elongates and differentiates into a thin myotube. Two such lateral founders (in different planes of focus) that have begun to extend processes are shown in Fig. 7C-E. Myoblasts that have failed to fuse cluster around the duf-lacZ-expressing cells and, as in the embryo, express myosin, but do not differentiate further (Fig. 7D,E). Wild-type lateral fibres with the founder nucleus in each fibre expressing duf-lacZ are shown for comparison(Fig. 7F). X-Gal staining does not detect the low duf-lacZ expression in the remaining nuclei of these fibres. But that these fibres are multinucleate is evident when observed at a higher magnification (Fig. 7G,H). In the absence of fusion, the founders eventually develop into mononucleate, myosin-expressing fibres(Fig. 7I), like the mononucleate muscles observed in the embryos of fusion mutants.

The putative founders of the DVMs in the thorax of 1151GAL4/UAS-Rac1N17 pupae are also present in a wild-type pattern and initiate fibre formation, as shown in Fig. 7K. Some fusion does occur, but to a lesser extent than normal. The DVM II fibres shown in Fig. 7K are not mononucleate but have fewer nuclei than wild-type fibres of the same stage(Fig. 7J). These fibres ultimately give rise to muscles, albeit thin, at the correct position and with the correct number of fibres (Fig. 7M). These results suggest that where myoblast fusion is prevented during adult myogenesis, a population of duf-lacZ-expressing myoblasts segregates normally, as in the embryo, and that, like the founders in the embryo, these cells uniquely have the capacity to complete differentiation to form muscles. They also demonstrate that, as in the embryo,by the onset of fibre formation adult myoblasts are of two classes:fusion-competent cells that do not express duflacZ and founders that do express duf-lacZ. It is these latter cells that have the capacity to complete myogenic differentiation even when fusion is blocked or reduced.

In the Drosophila embryo, the diversification of muscle forming mesoderm into founders and fusion-competent cells occurs through a process of lateral inhibition mediated by Notch(Corbin et al., 1991; Carmena et al., 1995; Carmena et al., 1998). As we have now shown that single duf-lacZ-expressing cells are selected and appear to act as founder myoblasts during adult myogenesis, it is important to show whether, as in the embryo, a Notch-dependent lateral inhibition pathway mediates this selection process. To test whether Notch has a function in selecting specific myoblasts for duf-lacZ expression, we used a dominant-negative and a constitutively active form of Notch. We reasoned that if lateral inhibition were involved, then overexpression of a dominant-negative form of Notch (dnNotch) in adult myoblasts would lead to an increase in the number of duf-lacZ-expressing founders, whereas overexpression of the active form (Nintra) should suppress duf-lacZexpression altogether.

In fact, the results of these experiments appeared to be contradictory:thus, expression of UAS-dnNotch caused no change in the number of DVM founders (Fig. 8C) and flies of the genotype 1151GAL4, UAS-dnNotch had the correct number of DLM and DVM fibres (data not shown). We verified this conclusion by reducing Notch function in two additional ways. We reduced function in the Notch signalling pathway in myoblasts by overexpressing truncated forms of the protein Mastermind (Mam), an essential component of the Notch signalling pathway. Mam interacts with the intracellular domain of Notch and with Suppressor of Hairless, and forms a transcriptional activation complex(Wu et al., 2000; Kitagawa et al., 2001). Two truncated versions of Mam, MamH and MamN, when overexpressed by the GAL4-UAS system behave as dominant-negative proteins and elicit Notch loss-of-function phenotypes (Helms et al.,1999). Overexpression of either UASMamN(Fig. 8D) or UAS-MamH(data not shown) in myoblasts using 1151-GAL4 had no effect on the number of DVM founders. We further examined the role of Notch by using a conditional allele, N^ts1(Shellenbarger and Mohler,1975). Because of the close proximity of the duf and Notch loci, we could not generate duf-lacZ, N^tsrecombinants and hence used 22C10 as the marker for founder cells in the abdomen. The earliest time at which myoblasts expressing high levels of duflacZ are also labeled with 22C10 is at 24 hours APF(Fig. 3A-C). We removed Notch function for different periods (2 hours, 4 hours, 6 hours and 8 hours) before this stage by raising N^ts animals to the non-permissive temperature and looked at the number of 22C10-stained cells associated with the abdominal nerves. The numbers of 22C10-expressing cells in the dorsal(Fig. 8F) or lateral segments(data not shown) of the abdomen were examined and shown to be unaffected in these experiments. We know that all three approaches - using the dominant-negative Notch or mastermind constructs, and using N^ts animals - are effective, as they all can reduce the levels of Twist expression in adult myoblasts (data not shown), a known consequence of Notch reduction in adult myoblasts(Anant et al., 1998).

Taken together, these results suggest that Notch is not required for the selection of duf-lacZ-expressing myoblasts. However, in the converse experiment, expression of Nintra in the myoblasts does suppress the formation of founders, as we would expect for a selection mechanism based on lateral inhibition (Fig. 8H). How can these apparently contradictory findings be reconciled? Earlier studies from our laboratory showed that Notch acts as a positive regulator of Twist in the myoblast population (Anant et al.,1998). Thus, Nintra expression in adult myoblasts leads to maintained expression of Twist in these cells and to a failure of muscle differentiation. If the absence of founders that we observe is a consequence of this sustained expression of Twist in the myoblasts, then we would expect that simply expressing Twist constitutively in the myoblasts would mimic the activated-Notch phenotype. Using 1151-GAL4 to drive Twist expression in the adult myoblasts, we found that at 12 hours APF there were no DVM founders (Fig. 8I), suggesting that a decline in Twist expression, which is antagonized by the action of Nintra, is a requirement for elevated duf-lacZ expression. Indeed,founders are the first cells in the myoblast pool to show declining levels of Twist expression, with the result that the duf-lacZ-expressing founders and Twist-expressing myoblasts are mutually exclusive cell populations (Fig. 8J-L).

A central issue for our understanding of myogenesis is the organization of muscle differentiation to produce a muscle pattern. In different organisms,the fusion of groups of myoblasts to form syncitial myotubes is clearly regulated in such a way that, at appropriate positions, a set number of fibres form, resulting in a well-defined pattern of muscles on which subsequent movement depends. In the Drosophila embryo, this patterning process largely depends on the controlled segregation of a specialized class of myoblasts, the founders, each of which seeds the formation of a myotube. However, each of the larval muscles that form in the embryo consists of a single myotube, and it could be argued that the founder myoblasts are a special feature of a system that generates thirty different myotubes from the mesoderm of a single hemisegment. In the adult fly, however, the situation is more akin to that found in many vertebrates and other organisms, with each muscle consisting of multiple myotubes that associate together to form a larger contractile element. The question we have addressed here is a general one for all such systems, namely how the number and location of the forming myotubes is controlled in such a multi-fibre system.

An essential feature of embryonic myogenesis in the fly is the division of muscle-forming mesoderm into myoblasts of two kinds: founders and fusion-competent cells (Baylies et al.,1998). During embryogenesis, fusion is an asymmetric process in which founders and fusion-competent cells fuse with each other, but neither class can fuse with itself. This ensures that, wherever a founder segregates,it acts as a seed for the formation of a single myotube. During adult myogenesis, groups of myoblasts aggregate to form muscles consisting of multiple myotubes: our question is therefore, how does a fixed number of myotubes arise from this aggregate? Is it by the formation of an appropriate number of founder myoblasts and, if so, how is this controlled? Or is it by some totally different process that might be of general relevance for the formation of such multi-fibre aggregates?

Our results show that in the adult fly multi-fibre arrays arise by the choice of an appropriate number of founder cells, identifiable in the adult by the elevated expression of duf-lacZ.

An earlier study of myogenesis in the adult thorax suggested on morphological grounds that cells equivalent to founder myoblasts in the Drosophila embryo were involved in the formation of at least three muscles: the tergo-trochanteral (jump) muscle, and DVMs I and II(Rivlin et al., 2000). In addition Kozopas and Nusse showed duf-lacZ expression in one kind of adult muscle, the developing direct flight muscles (DFMs), but did not examine the function of these cells nor address the mechanisms by which they are selected (Kozopas and Nusse,2002). We decided to extend these studies by looking, during adult development, at the expression of a marker for embryonic founders, duf-lacZ. By early pupal stages we were clearly able to identify selective expression of duf-lacZ in both the thoracic and abdominal myoblasts. Furthermore, in both cases the pattern of expression mirrored the pattern of emerging fibres. Thus, where two or three fibres form, as in the DVMs, we saw two or three duf-lacZ-expressing cells amongst the aggregating myoblasts; where an array of fibres form, as in the lateral and dorsal musculature of the abdomen, we saw an array of single cells expressing the marker, prior to the formation of syncitial myotubes. We also saw that these duf-lacZ-expressing cells were able to seed individual fibres in the absence of fusion, or where fusion was severely reduced. These findings emphasize not only the presence of a specialized class of myoblasts amongst the adult muscle-forming cells, but suggest that these cells, one per fibre,are the founder myoblasts. In addition, because of the known function of Duf as an attractant for myoblasts (Ruiz-Gomez et al., 2000), it seems likely that the same patterns of asymmetric gene expression that characterize founders and fusion-competent cells prior to fusion in the embryo are recapitulated in the adult muscle-forming population as syncitial myotubes are formed.

We were surprised to find that duf-lacZ is also expressed in what have previously been thought of as analogues of founders, namely the three persistent larval muscles that act as templates in each hemithorax, organizing the development of the large DLMs from the swarms of adult myoblasts that aggregate about them as metamorphosis begins. We assume that dufexpression in these cells serves to attract the adult myoblasts to the templates, with which they then fuse to form the six fibres of the adult DLMs.

At the end of the third larval instar, duf-lacZ is detected at a low level in most, if not all, adult myoblasts. This expression disappears and is replaced by selective expression at a much higher level in the cells that we identified as founder myoblasts, as we have described and discussed above. We suspect that this initially uniform expression at a low level may reflect the origins of the adult myoblasts from lineages that generate muscle founder cells in the embryo. In those cases that have been studied in detail, it has been shown that the pools of myoblasts from which adult myotubes will form,arise from a small number of adult muscle precursor (AP) cells in the embryo. In the case of the ventral abdominal muscle of the adult, for example, it can be shown that, in each hemisegment, the pool of myoblasts that will generate the several myotubes that make up this muscle are all derived from a single ventral AP cell in the embryo (Ruiz-Gomez and Bate, 1997) (reviewed by Baylies et al., 1998). This cell in turn is the sibling of the founder myoblast that seeds the formation of larval muscle VA3. This lineage is typical of the many muscle lineages that generate muscle-forming cells in the embryo: the terminal division in each lineage generates either two sibling founder myoblasts (e.g. VA1 and VA2), or a founder and an adult precursor (VA3 and VAP). While the founders manifest their muscle forming potential in the embryo, express duf-lacZ and seed myotubes, the muscle-forming potential of the AP cells is suppressed. These cells proliferate in the larva, and first differentiate during metamorphosis as adult muscle formation begins in the pupa. Thus the adult myoblasts are clonal descendants of single cells that are themselves the products of founder myoblast-generating lineages in the embryo. We suggest that in the hormonal environment of the third larval instar, aspects of the founder lineage of these clones begin to be expressed. However, uniform expression of duf-lacZ in a population of myoblasts has no apparent functional sense. Therefore, the uniform pattern of expression (perhaps reflecting the developmental history of the cells concerned) must be replaced by local upregulation in a few cells that will act as founders and downregulation in other myoblasts that will now respond to the localized Duf signal. This is the sequence that we observe in both the abdomen and the thorax, and we suggest that it is the control of this process that is decisive for the formation of the correct pattern and number of myotubes.

Our experiments indicate that a lateral inhibition mechanism such as that which leads to the segregation of muscle progenitor cells in the muscle forming mesoderm of the embryo is not responsible for the later segregation of duf-lacZ-expressing cells from among the adult myoblasts. For these experiments, we focused our attention on the de novo segregation of the single cells that appear to seed the formation of the DVM fibres. We found that if Notch signalling is removed or blocked by the expression of dominant-negative constructs in the adult myoblasts, there was no effect on the segregation of an appropriate number of cells expressing duflacZ at the sites where the DVM or abdominal fibres are formed. This is in stark contrast to the effects of reduced levels of Notch in the embryo, which lead to an overproduction of founder cells expressing duf-lacZ. However, the DVM founder myoblasts are lost when Notch is constitutively active in the adult myoblasts and this resembles effects seen in the embryo when Notch is constitutively expressed. We suspect that this loss of adult founders reflects an indirect effect of Notch, which, when activated, maintains twistexpression in the myoblasts concerned. We find that persistent expression of twist alone blocks the appearance of founder myoblasts in the adult. Our observations lead us to conclude that lateral inhibition mediated by Notch is unlikely to be the mechanism underlying the segregation of the adult founder myoblasts.

As lateral inhibition does not appear to select cells from the muscle-forming population for duf-lacZ expression, we consider two other putative sources of muscle patterning cues: the epidermal sites at which individual muscle fibres will attach and the nerve fibres that will innervate them. In the thorax, duflacZ expression in the larval templates for the DLMs is first seen at the time stripe-expressing adult epidermal cells (Lee et al., 1995; Fernandes et al., 1996) are juxtaposed adjacent to the templates (A. Ghazi, unpublished). Preliminary results indicate that reduction of the number of stripe-expressing cells results in a reduction of duf-lacZ nuclei in the LOMs, and increasing stripe expression increases duf-lacZ expression(A. Ghazi, unpublished). The large number of stripe-expressing cells that attach to each thoracic fibre make the decisive experiments (complete removal of stripe-expressing tendon cell precursors and misexpression of stripe in a large ectopic domain) difficult to perform. The role of tendon cells, if any, in founder selection or duf expression may have to wait for other approaches that shed light on the signalling pathways involved.

Innervation might also play an important role in fibre formation through the mediation of duf expression. Laser ablation experiments have shown that DLMs can be formed even if the normal larval templates have been ablated. However, if the larval templates and the innervation are both removed then the DLMs fail to form (Fernandes and Keshishian, 1998). This suggests that, where muscles form de novo,innervation is an essential ingredient for the initiation of fibre formation and may therefore play a role in the selection of duf-lacZ-expressing founder myoblasts from the adult myoblast population. This view is reinforced by the finding that, when there is no innervation, the DVMs, which normally form de novo, do not develop at all(Fernandes and Keshishian,1998). It is well known that innervation is essential for the formation of the male specific muscle (MSM) in the abdomen(Lawrence and Johnston, 1986). There is a close association between nerve fibre branches and forming muscle fibres in the abdomen, and the MSM is itself a local aggregation of such muscle fibres. Now that we have a marker that identifies the earliest stages of fibre formation, we plan further experiments to investigate the part played by innervation and attachment in selecting the cells that seed fibre formation.

The broad conclusion is that an external cue from the region where the muscle is destined to form is likely to set the number of contributing myotubes [one external cue, Wingless, is required for maintenance of identity of groups of myoblasts (Sudarsan et al.,2001), but we have not yet established what external cues act to select individual myoblasts from this pool]. In Drosophila, this process seems to be mediated by the selection of duf-expressing founder cells, each of which seeds the formation of a fibre. The number and pattern of fibres could then be set by the strength and distribution of the founder-inducing cue. We can envisage a similar process operating in vertebrate myogenesis as myoblasts aggregate and fuse to form a pattern of primary myotubes. Whether here too the patterning of fibres depends on the induction of founder or seed myoblasts at sites of muscle formation is an important question that remains to be resolved.

We thank S. Artavanis-Tsakonas and I. Hariharan for fly strains; Siegfried Roth, Sean Carroll and Dan Kiehart for antibodies; and Veronica Rodrigues and Satyajit Mayor for insightful comments. This work was supported by grants from the Department of Biotechnology and the National Centre for Biological Sciences to K.V.R., and by a Kanwal Rekhi Fellowship to D.D. M.B. is a Royal Society Research Professor.