The development of adult muscles in Drosophila: ablation of identified muscle precursor cells

Each segment in Drosophila has a stereotyped pattern of epidermis, nervous system and somatic muscles. A great deal of effort has gone into establishing our current understanding of the mechanisms that pattern the epidermis (reviewed in Ingham, 1988), but much less is known about the development of the internal tissues. In particular, the somatic muscles derive from the mesoderm (Miller 1950; Crossley, 1978; Lawrence and Johnston, 1986a) and may be patterned by mechanisms significantly different from those described in ectodermal tissues. Though some genes responsible for the patterning of the ectoderm are expressed and appear to function in the mesoderm, notably elements of the bithorax complex (Akam, 1983; Hooper, 1986), it is not known whether the expression of these genes is autonomously controlled or induced by the closely adjacent ectoderm. Nor is it clear what role, if any, these genes play in the establishment of the precise muscle pattern within a segment once segment identity is established. What mechanisms are responsible for patterning the musculature in the Drosophila larva and later in the adult fly? More specifically, do the mature muscle patterns derive from genetically defined cell lineages as in the epidermis or is patterning of the mesoderm directed from ectodermal structures or is some other, unknown mechanism involved?

The twist gene is expressed in the presumptive embryonic mesoderm and has been implicated in mesodermal differentiation (Thisse et al. 1988). Expression of twist declines in the embryonic myoblasts during fusion to form the embryonic muscles. However, a subset of cells continues to express twist in the late embryo. In the larva, these cells divide, giving rise to progeny that express twist throughout larval development (Bate et al. 1991). Direct observation in the developing pupal abdomen using immunocytochemistry shows that these cells continue to divide in the early pupa and eventually fuse to form adult muscles (Currie and Bate, 1991). Thus, it appears that the twistexpressing cells are the progenitors of defined mesodermal cell lineages that give rise to adult segmental muscle patterns. The primary aim of this study was to address this hypothesis..

The ablation of defined cells, or groups of cells, in order to reveal their prospective fates was one of the earliest experimental techniques in developmental biology. In insects, classical studies used cell ablation techniques to define prospective germline cells (Hegner, 1910) and more recently to define prospective fates of a variety of somatic cells (Bownes, 1975; Lohs-Schardin et al. 1979). In the present study, we have used a chemical technique to ablate the twist-expressing lineages, or subsets of these lineages, in larval stages and assay the impact on the muscle pattern in the adult. Our results indicate that the twist-expressing lineages form unique myoblast primordia for specified subsets of muscle fibres within a segment of the adult abdomen and within the adult legs. The fate of the twistexpressing primordia appears to be established at least by early larval stages and is unique to a particular myoblast lineage within a segment. Thus, a restriction in the fate of myoblast pools in early development defines elements of the final adult muscle pattern. It is not clear whether this fate restriction is controlled autonomously within the mesoderm or is directed by ectodermal derivatives such as the epidermis and the nervous system.

Wild-type (Oregon-R) Drosophila melanogaster were used for analyses of normal development. Analyses of adult muscle patterns and, subsequently, muscle ablation patterns (see below) were facilitated by use of an otherwise wild-type strain containing a myosin heavy chain (MHC) β-galactosidase construct, a generous gift of Drs Sandy Bernstein and Norbert Hess. In this construct, the β-galactosidase coding sequence was linked to the MHC promoter region permitting direct observation of the adult somatic musculature by staining for enzyme activity. Only female flies within 24 h of eclosion were used in these studies.

All larvae were reared on an artificial diet (Ashburner, 1989) at 25±1°C. On this regime; 1^st instar was 0-24 h, 2^nd instar 24-48h, 3^rd instar 48-96h, and wandering lasted approximately 10 h. To obtain precisely staged larvae, animals were collected within a 30 min window at hatching.

Three antibodies were used routinely in this study to monitor neuromuscular development in the larvae and the adult; (1) anti-/twist (twi) (Thisse et al. 1987), (2) anti-β-galactosidase (β- gal) (Doe and Goodman, 1985), (3) and anti-horseradish peroxidase (HRP) (Jan and Jan, 1982).

(1) anti-iwwt: We used an antibody against the twist gene product, a generous gift of Dr Fabienne Perrin-Schmitt. The protocol for anti-twist staining was as previously described (Bate et al. 1991).

(2) anti-β-galactosidase: We used the β-galactosidase gene product linked to the MHC promoter as a highly satisfactory adult muscle marker. β-gal expression was monitored either by using a monoclonal anti-β -gal antibody, a gift of Dr Chris Doe (Doe and Goodman, 1985), or using β-gal activity staining as follows:

Adult female flies were selected soon after eclosion, etherized and pinned to a sylgard Petri dish in PBS. An incision was made through the abdomen body wall (dorsal midline, ventral midline or laterally), the abdomen pinned flat and the internal organs removed. This abdominal preparation was fixed for Ih in 4% paraformaldehyde-PBS. in this, and all subsequent immuno-staining in the adult fly, it was essential to quench endogenous peroxidase activity by incubating fixed preparations in 0.5 % H2O2 for 15 min. After quenching, the specimens were washed as above, blocked with 10% horse serum, and incubated in a 1:100 dilution of anti-β-gal for Ih at room temperature (RT). The specimens were washed 5x in PBS-TX and incubated in 1:200 dilution of anti-mouse IgG antibody (Vectastain) for 1 h at RT. Specimens were washed as above and incubated in ABC solution (Vectastain) for 30 min at RT. The specimens were reacted with DAB, cleared and mounted for observation.

The β-gal activity staining was accomplished as follows: Adult abdomens were dissected as described above and the specimens were fixed for 20 min in 4 % paraformaldehyde in PBS. Abdomens were placed in a small volume of the reaction buffer (3.1 mM K₃Fe, 3.1 mM K₄Fe, 100mM NaPi, 150mM NaCl, 100 mM MgCl₂ and 10 mM Xgal (J. Fernandes, personal communication)) and incubated at 37 °C for 30min-lh to stain. Stained specimens were prepared for double labelling with an antibody or cleared and mounted for observation.

(3) anti-HRP: Anti-horseradish peroxidase recognizes a neuron-specific cell surface antigen in Drosophila (Jan and Jan, 1982). In this study, we use anti-HRP (Cappell) as a marker of both sensory nerves and motor nerves in the muscle ablation experiments described below. Adult abdomens were dissected and fixed as described above and stained as described previously (Currie and Bate, 1991).

Birthdating cells with 5-bromodeoxyuridine (BUdR) We used BUdR to study DNA replication in cells during larval life. The protocols we followed are described in Truman and Bate (1988).

Hydroxyurea (HU) is a DNA-synthesis inhibitor that blocks the activity of nucleotide reductase, whose catalytic activity is to convert ribonucleotides to deoxyribonucleotides (Timson, 1975). In vitro, HU treatment killed Drosophila embryonic neuroblasts in the S-phase of the cell cycle, but had no discernible effect on neuroblasts in other stages of the cycle or on previously born neuroblast progeny which continued to thrive and differentiate (Furst and Mahowald, 1985). In vivo, HU irreversibly blocked the mitotic activity of larval Manduca neuroblasts, which eventually die and degenerate, but had no discernible effect on the earlier progeny of these cells or on other cells in the larva which were not actively dividing during application of the drug (Truman and Booker, 1986). In Manduca larvae, hydroxyurea had few non-specific effects; treated animals continued to develop normally, pupated and metamorphosed as morphologically normal adults.

In the present study, we used hydroxyurea to ablate identified cells in the Drosophila larva, assessing the impact of drug treatments with specific cell markers both in the larva and the adult. HU (Sigma) was administered to feeding larvae through the diet. The drug was dissolved in a small volume of distilled water and mixed thoroughly with a measured volume of melted diet. Precisely staged larvae (see above) were allowed to feed at low densities (<20 larvae/tube) on this diet for a period of time after which they were returned to normal diet. In the first instance, the lethal dosage of this drug therapy was determined with 6 and 12 h treatments, after which the dosage level was decreased until the majority of the treated animals eclosed as morphologically normal adult flies. This experimentally determined dosage varied considerably between larval instars (younger instars being generally more susceptible to the drug) and even in stereotyped patterns within a given instar (periods immediately prior and following a moult being particularly sensitive). As a compromise, we chose the experimental concentration of 5 mg HU ml^-1 diet, with the additional limitation that animals within 3 h of a moult were not treated. When the drug was fed to larvae for 12 h bouts at this concentration, morphologically normal adult flies could be obtained from all experimental treatments. In all experiments, only externally morphologically normal adults were selected for dissection and analysis.

Based on the incorporation rates of diet-supplied [³H]thy-midine and 5-bromodeoxyuridine in mitotically active cells (White and Kankel, 1978; Truman and Bate, 1988), we estimate that it takes about an hour for HU to penetrate dividing cells. However, it may take considerably longer for the drug to reach effectively lethal concentrations. After the end of a treatment, we estimate that it takes at least 3 h for the level of HU in the larvae to drop to non-lethal levels. This estimate is based on two lines of evidence; (1) HU treatment during a moult is lethal. This sensitivity precedes the moult by approx. 3h when the tolerance level abruptly increases. (2) Uniquely identified cells accurately birthdated using our BUdR staining method can be ablated with HU at the experimental concentration if treatment occurs within 3h prior to the birthdate. Treatment at more than 3 h prior to the cell birthdate was never observed to cause cell ablation (data not shown). As a consequence of these estimates, all treatment times were adjusted so that the reported time of treatment corresponds to the time when HU is at potentially lethal concentrations within the larval tissues. Note that the HU level could never be increased to a concentration where every cell undergoing mitosis during the treatment time would be ablated. We suggest that such treatment would cause extensive death of the polyploid larval cells which are also synthesizing DNA during treatment times but are usually relatively unaffected by lower levels of the drug. Death of these cells would be expected to lead to immediate larval death. At the experimental concentration used, assuming equal access of the drug to all cells, we estimate that a given mitotic cell has a 25-30% chance of ablation.

Experimentally HU-treated larvae were divided into two groups. One of these groups was killed as wandering third instars, dissected and stained with anti-twist antibody to assess the impact of treatment on the twist-expressing cell groups. The remaining animals were allowed to pupate and eclose as adults. Young adults were killed, dissected and their nerves and muscles examined in detail to assess effects of the drug treatment. Adult nerves were examined with anti-HRP staining as described above. Adult muscles were examined with a variety of methods. (1) Abdomen muscle patterns were examined under polarized light using a Nikon polarizing microscope, with toluidine blue staining, and with f- galactosidase antibody or activity staining using the MHC-β-gal construct as described above. (2) Legs were removed from treated animals, mounted in glycerol and examined under polarized light. (3) Muscle nuclei were counted by staining adult abdomen preparations with toluidine blue or by counting nuclei labelled with the background of other staining procedures.

The adult abdominal muscles of Drosophila consist of groups of syncytial fibres, which insert on epidermal apodemes (Miller, 1950; Crossley, 1978). Abdominal hemisegments A2-A6 have an identical musculature consisting of 3 sets of muscle fibres, which can be distinguished by position and orientation (Fig. 1). (1) A group of ventral longitudinal fibres (mean=6/hemiseg-ment), which lie just lateral to the ventral midline in each hemisegment and insert at the segment boundaries. (2) A group of lateral pleural fibres (mean= 20/hemisegment), which are normal to the longitudinal axis and connect the ventral and dorsal epidermis. A small spiracular muscle is also present laterally in each hemisegment. (3) A group of dorsal longitudinal fibres (mean=20/hemisegment), which insert at the segment. boundaries. In addition, a group of remodelled larval muscles -distinguishable by their robust appearance, larval pattern of innervation and polyploid nuclei -persist in the young adult. These persistent larval muscles are associated with all three sets of adult muscles (Fig. 1). As HU treatment would not be predicted to alter the pattern of these persistent larval muscles, they served as a reliable framework for the normal muscle pattern and as important internal controls for our ablation experiments (see below). They are quickly degraded in the adult and are generally absent within 24 h of eclosion.

HU treatment in larval stages causes specific ablations of elements of the adult abdominal muscle pattern Treatment with hydroxyurea (HU) during specific developmental stages in the larva has a clear, consistent effect on the muscle pattern in the adult abdomen (Table 1). Treatment with HU in the first instar (0-24 h) has no discernible effect on the adult musculature relative to the control. In sharp contrast, treatment with HU in the second instar larva causes widespread ablations of elements of the adult abdomen muscle pattern. In Table 1, we present a summary of these findings.

Deletions in the adult abdominal muscles can be divided into those affecting each of the three components of the pattern: ventral, lateral and dorsal muscle fibres. (1) In the ventral muscle group, only all-or-none ablation of muscle fibres in a hemisegment was observed (Table 1; Fig. 2A,B). Partial ablations (that is a significant decrease in the number of fibres within a hemisegment) did not differ significantly from the control. No significant muscle ablation was observed if HU treatment occurred in the first instar (0-24 h after hatching), a high and approximately equal rate of ablation was observed in larvae treated in the early and late second instar (24-36/36-48 h), and the frequency of ablation decreased progressively for later treatments (Table 1). (2) In the lateral fibres, both partial ablation of muscle fibres and complete ablation of fibres in a hemisegment occurred (Table 1; Fig. 2C,D). In general, the frequency of partial ablations was significantly higher than the frequency of total ablations (Table 1). The treatment time for successful ablations was similar to that observed in the ventral muscles: no significant ablation in the first instar (0-24 h), high ablation rates from treatment in the second instar (24-48 h), and a rapidly decreasing number of ablations when treatment was later in development. (3) In the dorsal fibres, a similar ablation pattern to that seen for the lateral fibres was observed. Again, we see both partial and total ablation of muscle fibres in a hemisegment: partial ablations occurred at a higher frequency than total ablations, and significant rates of ablation were limited to the second and early third instars (Table 1; Fig. 2E,F).

The partial ablations observed in the lateral and dorsal fibre groups were not graded, but occurred in a quantal manner. As an example, the distribution of dorsal fibre numbers after a HU treatment (36-48 h) is shown diagrammatically in Fig. 3. In this figure, there are four local maxima in the numbers of muscle fibres: the highest (peak: 22 fibres) corresponds to the normal, wild-type fibre number, the middle two (peaks: 14 and 8 fibres, respectively) correspond to partial ablations as shown in Table 1, and the lowest peak (0 fibres) corresponds to complete ablations (Table 1). When we take into account the normal range in dorsal fibre numbers in an adult hemisegment (20.4±2.1; mean±s.D.), we observed that each peak is the approximate mean of a quantal decrease in the number of muscle fibres within a normal range. This quantal ablation revealed that the apparently continuous set of dorsal muscle fibres in each hemisegment is actually composed of three separate ‘fibre groups’, each containing an approximately equal number of fibres (6-8 fibres/fibre group; Fig. 3). Furthermore, we observed that each of these fibre groups could only be ablated in an all-or-none fashion, like the ablations observed in the ventral fibre group. Qualitatively, a similar situation can be observed in the lateral fibres, however here we observed only two ‘fibre groups’ each containing 8-10 fibres (data not shown). Again, each of the lateral fibre groups could only be ablated in an all-or-none fashion. From this quantal analysis, we conclude that there are six ‘fibre groups’ in each abdominal hemisegment (one ventral, two lateral, and three dorsal) and that each group can be ablated individually with HU treatment in the larva.

It should be noted in passing that there appears to be no absolute restriction on the distribution of muscle fibres within these fibre groups in a partially ablated region. The ablation of a single dorsal fibre group, for example, could either lead to a distinctive quantal ‘hole’ in the dorsal fibre array, or muscle fibres from the remaining two groups could spread equally through the dorsal region resulting in a reduction in the density of fibres but no obvious ‘hole’, or a range of gradations between these extremes could occur (Fig. 2D). We assume, therefore, that there are no strong limitations on epidermal insertion sites differentiating muscle fibre groups within a region.

HU treatments later than the early third instar (60 h) rarely resulted in ablations of elements of the abdominal muscle pattern (Table 1). Instead, HU treatments in the late larva would result in a decrease in muscle fibre size (data not shown). As with fibre group ablations, the reduction in muscle fibre size was observed to be quantal; the extent of the area of fibre reduction correlated closely with the ‘fibre groups’ defined by the all-or-none ablations above, and all fibres in a fibre group would be reduced in size to a similar extent. Reduction in fibre size is closely correlated with a reduction in the number of syncytial nuclei. In reduced fibres, all fibres within a group contained approximately equal numbers of nuclei per fibre. We conclude that HU treatment in the larva reduces the size of a primordium which is shared equally by all fibres within a fibre group.

We have shown that small numbers of metamerically-repeated cells which express the gene twist in the late embryo are the primordia of the adult muscles and that these cells divide in the larva and fuse during metamorphosis to form the adult abdominal musculature (Bate et al. 1991; Currie and Bate, 1991). Clearly, the late embryonic twist-expressing cells (1 ventral, 2 lateral and 3 dorsal cells per abdominal hemisegment) are strong candidates for the muscle primordia being affected in the HU ablation studies described above. To test this hypothesis, we have compared the impact of HU treatment on the larval ítvúz-expressing cells with the effect of HU treatments on the mature abdominal muscles.

We have shown that the increase of twist -expressing cells in the larval abdomen occurs by replication rather than recruitment (Bate et al. 1991). Our immediate concern here was to establish the precise birthdates of these cells for use in our ablation studies. Fortunately, iwzw-expressing cells in the larval abdomen are independently identifiable by their position, association with peripheral nerves, and the small size of their nuclei relative to the extremely large polyploid larval nuclei in the surrounding tissues (Bate et al. 1991). Thus, these cells could be identified using a nuclear stain only. Using 5-bromodeoxyuridine (BUdR) incorporation to monitor replication events, we determined that all twistexpressing cells in the wandering larva were mitotic during larval life; the number of identified cells incorporating BUdR was equal to the number of twistexpressing cells in each group at the end of larval life. Using pulsed applications of BUdR (see Materials and methods), we timed the birthdates of the twistexpressing cells in the larva (Table 3). We found that each of the ventral, lateral, and dorsal zwzsz-expressing cells incorporated BUdR during the first 6h of the second instar (24-30 h), and had divided to produce two cells by the middle of the instar (36 h; Table 3). Consistent with the increase in the number of cells expressing twist described above (Table 2), the replication of these cell groups progressed at different rates during the remainder of larval development; (1) The ventral cells divided a second time between 36-48 h, a third time during the first day of third instar (48-72 h; Table 3), and were involved in a fourth division cycle at wandering (96 h). (2) The lateral mw-expressing cells underwent a second division later than the ventral cells (42-54 h), the third division was similarly delayed (60-72 h; Table 3), but the fourth division was completed by 96 h, resulting in a higher number of twistexpressing cells than in the ventral group at wandering (Table 2). (3) The dorsal itvzsi-expressing cells underwent the slowest replication cycle; the second cycle was delayed to the third instar (48-60 h; Table 3) and the third cycle just completed by wandering (66-90 h).

In summary, the abdominal twist -expressing cell groups, each originally a single cell primordium, undergo 3 to 4 replication cycles during larval life to give rise to small clones of twist -expressing cells (8-16 cells) in the wandering third instar larva (Tables 2 and 3). The divisions in these primordia all begin in the first half of the second instar (24-36 h), but thereafter progress at different rates. Divisions are earlier ventrally, later laterally, and most delayed dorsally (Table 3). At wandering (96 h), the lateral primordia have the largest number of cells, the ventral primordium is smaller and the dorsal primordia have the fewest cells.

Precise determination of cell birthdates in the larval abdomen allowed us to compare directly the impact of HU treatment on the twist -expressing cell groups with the ablations generated in the adult muscle pattern. In the first instance, animals were treated with HU for a limited period of time (12 h) and then allowed to mature to wandering third instars (96 h) when the twistexpressing cell groups were examined. A summary of the impact of drug treatment on the twist -expressing primordia at this stage is presented in Table 4.

In general, ablation of the twist -expressing primordia was in excellent agreement with our expectations based on the birthdates of these cells (Table 3 and 4). Exposure to HU before the observed birthdates of the twist -expressing cells (first instar; 0-24 h) did not result in any discernible ablations relative to the control (Table 4). In sharp contrast, treatment with HU during the initial replication cycle (24-36 h) resulted in significant ablations of the twist-expressing primordia in all three groups in an all-or-none fashion (Table 4; Fig. 5). Treatment with HU at progressively later times resulted in a progressive decrease in the number of complete ablations in all three groups (Table 4). Instead, the percentage of partial ablations, (significant decreases in the numbers of cells per primordium), initially increased with HU exposure in the late second and early third instar (Table 4). This was the expected result given that each group contained an increasing number of cells during development (Table 2), thus decreasing the likelihood that all the cells within a particular group would be successfully ablated. Treatment with HU later in the third instar (60-96 h) did not result in significant ablations in the zw/sz-expressing cell groups. We attribute this to the fact that cell ablations later in a lineage would have relatively less effect on the final larval primordia and that, while cell death was almost certainly occurring, the intrinsic variability in the size of primordia at wandering would mask small changes in cell number (Table 2).

In summary, ablation of the twist -expressing cells was consistent with our expectations based on their birthdates and the number of cells within each group (compare Tables 2, 3 and 4). Note that complete ablations occurred at a significant level only if HU was provided during the first 1-2 replication cycles, when the small number of cells made these groups vulnerable (Table 4; Fig. 5). In this connection, it should also be noted that single cell primordia are highly susceptible to ablation via cell death. Even in the control groups, missing primordia were observed with some frequency (Table 4). HU treatment in older primordia successfully ablated a number of cells within the primordia but rarely resulted in complete ablation of a group.

A strong correlation was observed between the ablation of twist -expressing lineages in the larva and the ablation of muscle ‘fibre groups’ in the adult abdomen (Fig. 5; compare Tables 1 and 4). Treatment with HU in the first instar (0-24 h) did not cause ablation of the quiescent twist -expressing cells nor did it result in aberrations in the mature abdominal muscle pattern. In contrast, HU treatments in the second instar tended to ablate the newly replicating twist -expressing lineages and also resulted in the ablation of fibre groups in the mature muscle pattern (Fig. 5). Furthermore, ablations occurred independently in the six twist -expressing larval lineages, in an all-or-none fashion, and, similarly, in the six muscle fibre groups of the adult abdomen (Tables 1, 4; Figs 2, 5). Note particularly that the ablation success was very similar spatially and temporally between the two groups (Fig. 5). Finally, treatment with HU in the late third instar larva rarely resulted in the complete ablation of a rwzsr-expressing lineage and, similarly, rarely resulted in the ablation of an adult muscle fibre group.

Treatment with HU in the late larva often resulted in significant partial ablations of twist -expressing primordia and also resulted in a gross reduction in the size of muscle fibres within a fibre group in the adult abdomen. In both cases, the treatment resulted in a reduction in cell number, i.e. the number of twist -expressing cells within a larval primordium and the number of syncytial nuclei within the muscle fibres of a fibre group (Table 5). We observed that a reduction in the number of twist -expressing cells in a larval primordium was often proportionally greater than the reduction in the number of muscle nuclei in an adult fibre group (Table 5). This raises the possibility that extra divisions of twist -expressing cells may be occurring to compensate for ablated cells. Alternatively, it may be that cells within a nv/sr-expressing primordium divide at different rates and so particular ablations within this group have a variable impact on the final complement of nuclei. In either case, it is clear that a variable number of nuclei can contribute to a given fibre (Table 5) and that fibre number rather than the number of nuclei per fibre is conserved. Finally, it is clear that an adult muscle fibre group is defined as a single primordium at the time of HU treatment in the larva and that subpopulations within a primordium are not uniquely fated to form subpopulations of fibres within a muscle fibre group.

Because they have epidermal insertions, we would expect the pattern of the muscles to be affected by changes in the adult epidermis. We argue that the epidermis is not being affected by our HU treatments for the following reasons: (1) Adult histoblast nests (primordia of the adult abdominal epidermis) do not start replicating until the late third instar (Guerra et al. 1972; K.S.B. and M.B., unpublished observations). Hence, it is very unlikely that HU treatment in the second instar, when we observed the strongest effects on muscle patterning, would perturb these cells. (2) All preparations were routinely screened for defects in the epidermis based on pigmentation, bristle number and bristle pattern. Only a very small percentage (<3 %) of segments with ablated muscles showed any discernible defect in the epidermis. We conclude that the muscle ablations that we have described cannot be attributed to an epidermal defect.

When animals were treated with HU very late in larval development (90+ h) we did, occasionally, succeed in ablating or perturbing cells in the histoblast nest, which, in turn, resulted in adult segments either missing adult epidermis, missing patches of adult epidermis, or showing a greatly reduced number or aberrant pattern of epidermal bristles (data not shown). All these aberrations affected the underlying muscle pattern: absence of adult epidermis resulted in a lack of recognizable muscle; missing patches of adult epidermis correlated with a reduced number of muscle fibres in the affected segment and aberrant bristle patterns were reflected in aberrant muscle patterning (data not shown). This last epidermal defect rarely resulted in significant muscle ablations but rather resulted in the correct number of muscle fibres making incorrect insertions in the epidermis so that muscles would be ‘clumped’, or would cross each other to make incorrect insertions (data not shown).

There is good evidence that motorneurons may play a role in establishing the adult musculature in the Drosophila abdomen (Lawrence and Johnston, 1986b) and other work (reviewed in Nuesch, 1985) indicates that successful innervation is necessary for muscle maintenance. It is well known in holometabolous insects, specifically Drosophila (Truman and Bate, 1988), that the majority of the adult nervous system is born postembryonically and is a potential target in these experiments. Indeed, our observations (Broadie and Bate, in prep.) indicate that HU treatment can ablate postembryonic neuroblasts or subsets of their lineages. However complementary studies that we have made show that adult motorneurons are born embry-onically and not in the larva, and we conclude that ablation of the motor network cannot be responsible for the muscle ablations we observe.

The adult abdomen is an easily dissected epithelium which has a simple superficial array of 50 muscle fibres that can be easily divided into 3 subsets for ready analysis (Fig. 1). In addition, the larval abdomen has a relatively simple array of twist-expressing primordia that are spatially distinct and composed of a small number of cells. In contrast, the adult thorax dissection is more difficult, the thoracic muscle pattern much more complex and the larval thorax has a much greater number of twist-expressing cells distributed in an apparently complicated array both within and outside the imaginal discs. Nevertheless, we were interested in establishing a connection between the phenomena observed in the abdomen and in muscles derived from precursors in the imaginal discs. As a compromise, we chose to study the relatively simple muscle pattern in the adult legs (Fig. 6A) which can be readily analysed with a simple procedure (see Materials and methods).

The adult leg is composed of five easily distinguishable segments, the coxa, trochanter, femur, tibia and, most distally, the tarsus. Only the first 4 have muscles and of these we focus on the tibia, femur and trochanter, which have invariant muscle patterns in each of the thoracic segments.

Miller (1950) defined the leg musculature by name and number and, as far as this is consistent with our observations, we follow those conventions here (Fig. 6A,B). Starting proximally, Miller defined a single muscle in the trochanter described as a reductor muscle (no. 38). We would add that this muscle appears to be divided bilaterally into two similar muscle masses each composed of 8-12 fibres which span the longitudinal length of the trochanter from the coxal to the femoral joint (Fig. 6A). Miller defined three muscle masses in the femur: a tibia levator muscle (no. 39) and two tibia depressor muscles (no. 40, no. 41). The levator is a large, robust longitudinal muscle (approx. 16-20 fibres) which lies anteriorly in the femur, inserting at the trochanteral joint and on the integument proximal to the femorotibial joint (Fig. 6A). The first depressor (no. 40) is a similar muscle which occupies the posterior femur region. The remaining depressor (no. 41) is an oblique muscle composed of 8 fibres which lies just proximal to the femorotibial joint and inserts in the joint and the anterior femur integument. This muscle underlies the tibia levator (no. 39) and is difficult to observe normally. Finally, the tibia contains four morphologically distinct muscles (Fig. 6A). Miller (1950) identifies two of these, the tarsus levator (no. 42) and tarsus depressor (no. 43), but does not mention two others. The depressor (no. 43) is the most prominent muscle in the tibia. It is composed of 18-20 paired V-shaped fibres which attach to a prominent medial tendon and the tibial integument in an oblique orientation (Fig. 6A). The levator is a smaller muscle composed of 6-10 fibres which lies in the anterior, distal tibia. These fibres have an oblique orientation and insert on the anterior tibial integument and a small tendon attached to the tarsal joint (Fig. 6A). The third tibial muscle (named no. 43a to fit in Miller’s classification) is composed of a few (4-8) robust fibres which run obliquely in the proximal tibia to insert just distal to the femorotibial joint and the posterior tibial integument (Fig. 6A). Finally, the fourth tibial muscle (named no. 43b) is located at the extreme distal end immediately adjacent to the tibiotar-sus joint. It is a small muscle composed of 4 transverse fibres which are normally difficult to identify in polarized light. Because of this difficulty, we do not consider this muscle further.

Thus, we examine the muscles of three segments of the adult leg: the reductor muscle of the trochanter, the 2 depressors and 1 levator of the femur, and the 2 levators and 1 depressor of the tibia. This simple pattern of 7 muscles in 3 segments lends itself to a straightforward analysis of changes induced by HU treatment in the larvae.

By treating larvae with HU during specific periods, we were able to ablate subsets of muscles in the thoracic legs in a fashion similar to the muscle ablations in the abdomen (Tableó; Fig. 6). As in the abdomen, successful ablations were only achieved with HU treatment in the second and early third instar (24-60 h) (Table 6). In general, it was more difficult to generate muscle ablations in the leg than in the abdomen, reflecting either a decreased accessibility of disc cells to the drug treatment, or a biological phenomenon, such as a larger number of precursors per muscle. A detailed schedule of leg muscle ablations is given in Table 6.

Our most substantial finding was that subsets of the adult leg muscle pattern in each segment could be ablated without apparent effects on the remaining muscles in the segment or in adjacent segments (Fig. 6). We observed approximately 10 muscle fibre subsets in the trochanter, femur and tibia which could be ablated independently; 2 in the trochanter, 5 in the femur, and 3 in the tibia (Table 6; Fig. 6). Notice that 10 fibre groups were defined in this way by HU ablation, but that only 7 morphologically distinct muscles can be observed. As in the abdomen, muscle fibre groups and morphologically-defined muscles were not necessarily co-extensive. The 10 ablation groups can be divided into two classes. (1) Ablation of a complete muscle as defined by morphological criteria. Thus, all 3 muscles in the tibia and the distal depressor muscle (no. 41) in the femur were all ablated in an all-or-none fashion (Fig. 6) and, as a consequence, each appears to be defined in the larva as a single group. (2) The ablation group defined a subset of fibres within a morphologically defined muscle mass. Thus, the levator (no. 39) and depressor (no. 40) of the femur, and the reductor (no. 38) of the trochanter can all be partially ablated and appear to be composed of more that one defined fibre group each (Fig. 6). As with partial ablations in abdominal muscle groups (Fig. 3), partial ablations in leg muscle groups were not graded, but appear to occur in a stepwise fashion. All partial ablations in the leg were observed to result in an approximate 50% reduction in fibre numbers and, as a consequence, each muscle appeared to be composed of two defined muscle fibre groups. Finally, as in the abdomen, partial ablation of fibres within a muscle was not always observed as a distinct hole in the muscle but could result in a reduced number of fibres occupying the same area as the original muscle mass. As in the abdomen, we conclude that fibres are not rigorously restricted in the epidermal insertion sites that they can occupy.

Poodry and Schneiderman (1970) identified a class of so-called adepithelial cells dispersed between the folds of the disc epithelium, but within the basement membrane. These cells align during pupal metamorphosis and fuse to form the adult muscles (Reed et al. 1975). The adepithelial cells express twist in the embryonic imaginal discs and throughout larval development (Bate et al. 1991).

The twist-expressing adepithelial cells in the embryonic imaginal discs are both numerous and spatially complicated (Bate et al. 1991). However, we were intent on determining if the array and proliferative activity of these putative muscle precursors was consistent with the schedule of HU ablations of the adult leg musculature. We concentrated on the prothoracic leg disc as the number of twist -expressing cells was lowest in the prothorax and the embryonic prothoracic disc was distinct from other primordia (unlike the more posterior leg discs, which were intimately associated with the wing and haltere discs) allowing clear identification of associated twist -expressing cells.

A subset of cells associated with the prothoracic leg disc (mean=14) expresses twist in the stage 16 embryo. This array remains unchanged in the late first instar larva, begins to increase early in the second instar, and the number of cells expressing twist increases gradually for the remainder of larval life (data not shown). Thus, we observed that the progression of twist -expressing adepithelial cells was strikingly similar to the progression of the twist -expressing cells described in the abdomen. As in the abdomen, BUdR incorporation data (not shown) indicates that the adepithelial twistexpressing cells in the leg discs were mitotically active in the larva and that replication commenced early in the second larval instar. Unfortunately, because of the high replication rates and the large number of cells in the leg discs, it was impossible to determine if all twistexpressing cells derive as a lineage from embryonic twist -expressing cells, as in the abdomen (Bate et al. 1991), and it was also not feasible to divide the cells into groups because of their close packing (Fig. 7A). Therefore, it has not been possible to define particular twist -expressing lineage groups in the disc by position only, as in the abdomen.

Treatment with hydroxyurea (HU) in the early larva produced widespread ablations of the iivAi-expressing cells in the leg discs assayed in the wandering third instar larva (Fig. 7). A detailed analysis of these ablations was not possible in the absence of a detailed understanding of the cellular array. Nevertheless, we can say that treatment in the first instar had little discernible impact on the array of twist -expressing cells, whereas treatment during the second and early third instars caused a large but variable rate of cell ablation (Fig. 7). Later HU treatment in the third instar had a noticeable but greatly decreased impact on the number of twist-expressing cells at wandering. Thus, the ablation twist-expressing cells was consistent with our observations of their birthdates, and was superficially similar to the schedule of muscle primordia ablations in the abdomen.

In conclusion, it seems likely that the twist-expressing adepithelial cells in the leg imaginal discs are the unique primordia of adult leg muscle fibre groups. The mitotic activity and HU ablation schedule of these cells is consistent with the schedule of HU ablations of elements of the final muscle pattern. It is possible that the embryonic twist-expressing adepithelial cells found lineages in the larva that are uniquely fated to form specific muscle fibre groups in the leg as is the case in the abdomen.

The origins of the adult muscles in Drosophila have long been debated. Crossley (1978) reviewed the hypothesis that adult thoracic muscles derive from the adepithelial cells in the imaginal discs, and that the muscles in the adult abdomen derive from a similar class of cells associated with the abdominal histoblast nests. Direct demonstration of this hypothesis and the determination of the origin of these putative cells has not been forthcoming, largely due to the lack of a specific marker for the muscle precursors. In the accompanying paper (Bate et al. 1991), we show that a group of cells in the late embryo express the gene twist. These cells are associated with the anlagen of the imaginal discs and there is also a set in each abdominal segment, which are not. however, associated with the histoblast nests. Direct observation shows that the abdominal iwBr-expressing cells and their descendants fuse to form the adult musculature (Currie and Bate, 1991) and we conclude that the late embryonic twistexpressing cells are the precursors of adult muscles.

The primary objective of this study was to show whether the late twist -expressing cells are the unique progenitors of adult muscles by ablating abdominal twist-expressing cells in the larva and analyzing the effect on the adult abdominal muscle pattern. We were able to demonstrate that ablating twist-expressing cells in the larva correlates with ablations of adult muscles. Furthermore, effective ablation times corresponded exactly with the period of early cell cycles in the twistexpressing cells, the extent of ablation in the twistexpressing cells was paralleled by the extent of adult muscle ablations, and complete ablation of the twistexpressing cell nests was coextensive with the complete ablation of specific adult muscle fibre groups. We found that partial ablations of twist -expressing lineages correlated closely with a decrease in the number of syncytial nuclei in muscle fibre groups, implying that a reduction in the number of twist-expressing cells resulted in a reduction in the number of available muscle precursors. Finally, the number of muscle precursors in the early larva calculated on the basis of a clonal analysis of the adult ventral longitudinal muscles (Lawrence and Johnston, 1982) agreed well with the number of ventral twist-expressing cells present during the same period. This extensive evidence confirms our view that the abdominal twist-expressing cells are the sole progenitors of the adult abdominal muscles.

A more surprising finding was that single cell primordia identified in the embryo are the unique precursors of specific groups of adult muscle fibres. In each embryonic abdominal hemisegment, six spatially distinct twzst-expressing cells can be observed; one ventrally, two laterally and three dorsally. Ablation studies indicate that each of these cells gives rise to a specific set of 6-10 adult abdominal muscle fibres. Furthermore, removal of any one of these primordial cells results in removal of the adult structure. That is, loss of these cells cannot be compensated for by adjacent cell groups. It appears that it is the maintenance of the primordium itself that is essential for defining a muscle fibre group. There is no evidence that the ablation of any single cell or subset of cells within a twist-expressing lineage results in ablation of the adult muscle or a subset of its fibres. Rather, there is a uniform decrease in the number of syncytial nuclei throughout the adult muscle group. It appears, therefore, that all cells within a larval primordium are equivalent and equally available to fuse with any fibre within a muscle fibre group.

Two other lines of evidence suggest that the muscle groups in the adult are founded by discrete populations of myoblast cells. Lawrence (1982) described four clonally related muscle groups in the adult thorax arising from a small number (relative to the cuticle primordia) of primordia which he suggested are present in the imaginal discs. Furthermore, he demonstrated that there are eight muscle anlagen present in the early larva which increase greatly in size during the second larval instar (Lawrence, 1982). In conclusion, Lawrence suggests that each imaginal disc contains muscle precursors which will construct a specified subset of thoracic muscles. This, as well as a very similar conclusion for muscle development in the head (Vijayraghavan and Pinto, 1985), agrees well with our analysis of muscle specification in the abdomen. Furthermore, the twist-expressing cells that we have described in the imaginal discs are good candidates for the muscle anlagen as described by Lawrence. In addition, Deak and co-workers (Deak, 1977; Deak et al. 1982) isolated a number of X-chromosomal mutations that specifically affect different muscle groups in the thorax. For example, the mutations ewg, l(l)93p and sr ablate or damage the dorsal longitudinal muscles (DLMs), whereas the closely adjacent and morphologically indistinguishable dorsal ventral muscles (DVMs) are completely wild type. This evidence, Deak maintains, suggests different origins for these muscle groups. In combination with earlier studies involving extirpations and transplantations of imaginal discs (Zalokar, 1947), Deak maintains that his genetic evidence points to distinct and separate groups of myoblasts which give rise to the DLMs and the DVMs (see also Fernandes et al. 1991). Again, this conclusion parallels our evidence for restricted muscle precursor groups in the abdomen, and the twist-expressing cells in the discs provide strong candidates for the discrete myoblast populations as defined by Deak.

In summary, based on the evidence presented in this study and the corroborative reports from other groups (Lawrence, 1982, 1985; Deak et al. 1982), we suggest that adult muscles in the Drosophila abdomen are not derived from a central mobile pool of precursor cells but rather that each muscle fibre group in the adult is represented by a discrete pool of precursors in the larva. We also suggest that each pool has a unique fate and cannot be replaced by cells in other precursor pools and that there is little mixing of muscle precursors during development. Our evidence suggests that each muscle primordium is founded by an early precursor cell in the embryo which divides to give rise to a group of cells in the larva that remain together during development and, in turn, fuse to form a small number of specific adult muscle fibres.

It seems likely that myoblast specification occurs in muscles derived from the imaginal discs in a fashion similar to that which we have described in the abdomen. It is well established that the muscles of the adult leg, at least, arise from a population of adepithelial myoblasts within the larval leg imaginal discs (Ursprung et al. 1972; Crossley, 1978; Lawrence and Johnston, 1986a). Several parallels can be drawn with our observations of abdominal muscle development, twist -expressing cells are present in the embryonic and larval discs and expression appears to be in the adepithelial cells only. These cells begin to divide early in the second larval instar, treatment with a DNA-synthesis inhibitor during this period will ablate subsets of these cells, and distinct muscle masses in the adult leg are ablated if, and only if, treatment with the DNA-synthesis inhibitor takes place during the first rounds of twist-expressing cell division. Based on this evidence, we can state that the fate of the adult leg muscles is specified at least in the second instar and removal of the fated primordia cannot be compensated for by other cell populations. When compared with our results in the abdomen, it seems highly probable that the ‘fated primordia’ in the second instar leg disc are subsets of the zwzsz-expressing adepithelial cells, possibly, as in the abdomen, single cells.

Schubiger (1968) showed that the leg disc of the mature larva is similar to other imaginal discs of Drosophila in that it possesses a ‘mosaic determination’; that is, the primordia of many adult structures can be localized in an ‘anlage plan’ of the disc. This specification is very strict, with defined primordia for the claws, groups of sensilla, and even down to a single bristle on the trochanter (Poodry and Schneiderman, 1970). Given these studies, it is not surprising that the primordia of different muscle masses are defined in the larva in a similar fashion. At a superficial level, it is more difficult to imagine twist -expressing primordia in the leg discs forming isolated pools of muscle precursors, as the cells are closely spaced and not spatially distinct as in the abdomen. However, closer observation of the disc structure belies this argument. The twist -expressing adepithelial cells are not actually adjacent but are largely separated by folds in the disc epithelium (Poodry and Schneiderman, 1970). Furthermore, during eversion of the leg disc, these twistexpressing primordia would be widely separated as the folds in the disc epithelium are extended. Thus, it is quite possible that pools of muscle precursors in the larval leg disc are spatially discrete and consequently fated to establish different adult muscle masses.

Defining the fate of a cell or cell group does not provide any information about the state of commitment of those cells, but establishing prospective fates is a prerequisite for the interpretation of any experiments concerned with early developmental decisions. It appears that identified twist -expressing cells in the early Drosophila larva are fated to give rise to specific muscle fibre groups in the adult abdomen and leg. During normal development, this fate is unique to a particular cell, and its progeny, and cannot be taken on by other cells in its absence. This fate limitation could be caused by either of two radically different mechanisms. First, each of the twist -expressing primordial cells could be determined only to give rise to a specific muscle mass. Thus, in the abdomen, each of the six twist -expressing lineage groups would be genetically programmed to give rise to the ventral muscle fibre group, the anterior lateral fibre group, a specific dorsal fibre group, etc. Garcia-Bellido (1975) has proposed such a mechanism to explain pattern formation in the Drosophila epidermis and, indeed, grafting experiments in Musca have demonstrated that the abdominal histoblasts are determined to form a defined part of the adult abdominal integument (Bhaskaran, 1973). According to this genetic determination model, activation of a ‘selector gene’ occurs once in development and remains clonally irreversible to provide the genetic basis of a stable state of determination. Furthermore, the products of these selector genes are required throughout subsequent development in order to maintain a certain developmental pathway. Thus, we suggest that twist, or a similar gene, might determine the myoblast fate of the twistexpressing cells. We could extend this argument to hypothesize that the expression of other genes subdivides the total twist -expressing cell pool into smaller pools which are determined to specific muscle fates. Alternatively, the twist -expressing cells could be genetically equivalent, their fate being defined only by their spatial location and the subsequent localization of their progeny. According to this model, initiation of a developmental pathway is determined only by location and an ability to follow that pathway. Thus, the fate of a particular twist -expressing primordium to form, for example, the ventral longitudinal fibres in the adult abdomen, derives only from the position of the primordium rather than from a particular character of the founder cell. This simpler model would also explain the results that we have obtained in this study without requiring a sophisticated level of genetic control and determination.

Evidence has been presented in this study and elsewhere (Lawrence, 1982; Lawrence and Brower, 1982; Lawrence and Johnston, 1982; Deak et al. 1982) that appears to support both of these possible mechanisms. Support for determined myoblast primordia has been most strongly presented by the genetic analysis of Deak and co-workers (Deak, 1977; Deak et al. 1982), which conclusively demonstrates a genetic distinction between two closely adjacent muscle groups in the thorax, the DLMs and DVMs, as described above. While this evidence does support a genetic difference in these muscle groups (the focus for these mutations is mesodermal), it should be noted that the genetic specificity is not necessarily in the adult myoblasts themselves. Lawrence’s clonal analysis of thoracic muscle groups (1982) also lends some support to the theory of myoblast determination. However, clonal restriction of fate does not necessarily require that the cells be genetically determined; barriers to cell movement or other isolation of the muscle precursor groups would give similar results. However, there is some evidence, from direct observation in the pupa and clonal analysis in the abdomen, that small numbers of myoblasts may be capable of migration between (Lawrence and Johnston, 1982, 1986a) and within segments (Currie and Bate, 1991). This strengthens the case for a determinative mechanism and removes a major tenet of the argument for a mechanism of spatial isolation.

Nevertheless, there is some evidence that supports a non-determinative mechanism for myoblast segregation. For example, it has been shown that myoblasts introduced into a host abdomen from the wing imaginal disc of a donor can contribute, under these experimental conditions, to a wide range of developing muscles (Lawrence and Brower, 1982). This is evidence that these myoblasts are not determined to a particular muscle fate. However, it should be added that other explanations have been presented for this observation that do not demand that myoblasts are maintained in a non-determined state. For example, Lawrence and Johnston (1984) have proposed that myoblasts may be entrained to a ‘new determined state’ by the majority of the nuclei in the muscle syncytium. Lawrence and Johnston (1986a) have used this same argument of ‘myoblast entrainment’ to suggest that strict segregation of determined precursors in the mesoderm need not be maintained as in the epidermis, so that migrations of small numbers of myoblasts between muscle groups as described above may not be significant.

In conclusion, this study indicates that specific myoblast groups in the larvae have a unique and restricted fate but fails to resolve whether this fate is defined by genetic determination or some mechanism of spatial isolation of genetically equivalent precursors. If the former hypothesis is true, we would expect differences in gene expression patterns to differentiate all six twist -expressing primordia in the embryonic/lar-val abdominal hemisegment, the 10-14 muscle primordia in the imaginal leg discs, as well, presumably, as subsets of muscle primordia in the other imaginal discs.

We are grateful to Emma Rushton, Rachel Drysdale, Helen Skaer and Amy Bjesovec for much advice and support during the course of this study and for critically reading earlier versions of this manuscript. We also thank Douglas Currie for many illuminating discussions. Stocks were kindly provided by Dr Sandy Bernstein and Dr Norbert Hess and antibodies by Dr Chris Doe and Dr Fabienne Perrin-Schmitt. This work was supported by a Fulbright Fellowship and an Oliver Gatty Studentship to K.S.B., and a Wellcome Trust grant to M.B.