Sex-specific differentiation of a male-specific abdominal muscle, the Muscle of Lawrence, is abnormal in hydroxyurea-treated and in fruitless male flies

One way to generate sexually dimorphic features is for homologous structures to undergo differential development in males and females. One example of this type of sex-specific differentiation is the formation of the Muscles of Lawrence (MOLs), a pair of large muscles found in the fifth abdominal segment of male Drosophila that have small homologues in the fifth abdominal segment of females and other abdominal segments of both sexes (Lawrence and Johnston, 1984). MOLs have two unusual features that govern their development. First, MOL fibers uniquely require male-specific innervation for their formation. This was demonstrated initially by elegant transplantation experiments (Lawrence and Johnston, 1986), and by nerve ablation experiments showing that denervation of the fifth abdominal nerve prior to metamorphosis prevents the formation of the MOL in adult males (Currie and Bate, 1995). Second, sexual differentiation of the MOL is independent of the activity of the doublesex gene, which constitutes the major output pathway for sexual development of somatic tissue (Taylor, 1992; for recent reviews: Baker, 1989; Belote, 1992; Burtis and Wolfner, 1992; McKeown and Madigan, 1992; Burtis, 1993; Taylor et al., 1994; Hall, 1994).

Adult muscle development has been well studied in Drosophila from the initial events during embryogenesis through to terminal differentiation during metamorphosis (for review see Bate, 1993). The myoblasts that found the dorsal, lateral and ventral muscles of the adult are set aside in the embryo and then divide during larval, pupal and early adult stages (Bate et al., 1991; Broadie and Bate, 1991). Midway through the pupal stage, the myoblasts migrate along the expanding adult epidermis and fuse to create multinucleate myotubes (Bate et al., 1991; Currie and Bate, 1991). These myotubes elongate, form their attachments to the epidermis and mature into muscle cells with concomitant high levels of expression of muscle-specific proteins, such as β3 tubulin (Currie and Bate, 1991).

It is unknown at what stage in adult muscle development sex-specific differences are expressed. Previously, it has been shown that the MOL first becomes recognizable by its size at around 48 hours of pupal development (Gailey et al., 1991). In this study, we show that MOL fibers in males have more muscle nuclei than do their neighboring medial fibers or their homologues in females. The number of muscle precursors in the larva and the total number of muscle nuclei that contributed to the dorsal musculature were the same in male and female hemitergites suggesting that these differences were due to sex-specific differences in the distribution of myoblasts into myotubes.

We examined the relationship between the number of nuclei in MOL fibers and expression of MOL phenotypes by ablating myoblasts with hydroxyurea (HU) in wild-type male larvae. MOL fibers, recognizable by their attachment points, formed with fewer than half the average number of nuclei present in wild-type MOL fibers. We also evaluated the effect of reducing the number of muscle nuclei on the early male-specific expression of β-galactosidase in males with a 79B actin promoter-lacZ reporter gene construct, another characteristic of MOL fibers (Courchesne-Smith and Tobin, 1989). Many recognizable MOL fibers in HU-affected hemisegments did not express β-galactosidase. Thus, the reduction in the number of nuclei within MOL fibers had less effect on earlier differentiative steps, such as forming appropriate attachment points, than on later steps, such as the male-specific expression of a reporter gene.

Mutations in fruitless (fru) lead to the abnormal development of MOLs (Gailey et al., 1991) and to the disruption of several male-specific reproductive behaviors in Drosophila (Gill, 1963; Hall, 1978; Gailey and Hall, 1989; Wheeler et al., 1989; Gailey et al., 1991). In most fru mutants, normal-sized MOL muscles failed to form, but the severity of the muscle defect depended on the individual genotype examined (Gailey et al., 1991). In the most severe fru mutant genotypes, the only muscle fibers in the MOL-position resembled neighboring nonsex-specific fibers in length and appearance. Because there were as many muscle fibers in mutant segments as in control segments, Gailey et al. (1991) proposed that MOL fibers were not deleted, but rather developed as small-sized muscles. Males mutant for other fru genotypes, such as fru¹ homozygotes or fru¹ over either one of two deficiencies, Df(3R)P14 or Df(3R)Cha^M5, had no MOL-like muscle fibers but did have slightly longer and thicker muscles suggesting that these mutations were hypomorphic compared with the deficiency transheterozygotes. In the weakest genotype, fru² (originally called P[(w⁺) ARO-1]), homozygous males often had muscles as large as MOLs in wild-type males.

Since fru mutations are the only non-homeotic mutations known to affect MOL differentiation, we examined these mutants to define the role of fru⁺ in MOL myogenesis. In this study, we found that MOL-position fibers in fru mutant males have fewer muscle nuclei compared with wild-type MOL fibers. From our current understanding of the development of abdominal muscles in adult Drosophila, there are two ways that the fru⁺ product might be expected to affect the number of nuclei present in adult muscles: (1) decreased proliferation of myoblasts in the larva and early pupa or (2) reduced fusion of myoblasts into myotubes. By comparing the number of myoblasts, identifiable as twist-positive cells, in control and several fru mutant genotypes, we showed that the fru⁺ product does not affect the initial distribution or subsequent proliferation of myoblasts in larval stages. We propose that fru⁺ controls the male-specific recruitment of myoblasts necessary to create full-sized MOL-fibers. However, based on the HU-ablations, the reduction in the number of nuclei in MOL-position fibers in fru mutants by itself is not sufficient to explain their small size suggesting that fru⁺ has additional roles in MOL fiber development.

Canton-S flies were used for the analysis of normal adult musculature and myoblasts during development. The descriptions of dorsal muscle groups in wild-type and HU-treated animals (see below) was made easier by examining animals that contained a 79B actin promoter-lacZ gene construct (P[79B actin-lacZ, ry⁺]; Courchesne-Smith and Tobin, 1989).

Most of the fru mutant stocks used in this study have been described by Hall (1978), Gailey and Hall (1989) and Gailey et al. (1991). Crosses between two deficiency stocks, Df(3R)P14/ TM6B, Tb Hu e ca and Df(3R)Cha^M5/TM6B, Tb Hu e ca, produced fru mutant flies with a severe phenotype (see Lindsley and Zimm, 1992, for descriptions of the markers). The original fru¹ is a small inversion; the distal breakpoint at 91B is uncovered when fru¹ is transheterozygous with Df(3R)Cha^M5 (Gailey and Hall, 1989) and is associated with the MOL mutant defect (Gailey et al., 1991). The fru² mutant stock, previously known as P[(w⁺)ARO-1] (Taylor et al. 1994; Hall, 1994) was derived from a recently outcrossed line (Gailey et al., 1991). Two new P-element induced fru mutations, fru³ and fru⁴, were isolated in a screen for male steriles (Castrillon et al., 1993).

All fly stocks were raised on a diet of cornmeal, dextrose, agar and yeast which included propionic acid as a mold inhibitor. Fly cultures were kept in a humidified incubator at 25±1 °C.

Dissections of adult abdomens were performed as described by Gailey et al. (1991). The abdomens were fixed for 1 hour in 4% paraformaldehyde in 0.1 M phosphate buffer and then incubated overnight in primary antibody (see below) in 0.1 M phosphate-buffered saline with 0.1% Triton X-100 and 2% normal goat serum (PBS-TX-NGS) with 0.1% sodium azide at room temperature. Following treatment with a biotinylated secondary antibody and ABC reagent (Vectakit, Vector Labs), diaminobenzidine (DAB; Sigma) was used for the final signal detection (Metcalf, 1985).

Two different antibodies were used on adult abdomens. For the nuclear counts, a monoclonal antibody (mAb) 8C5 (1:100; a gift from Dr Seymour Benzer) was used. To visualize MOL or MOL-position fibers in P[79B actin-lacZ, ry⁺] flies, an anti-β-galactosidase antibody (1:10,000; Cappel Laboratories) was used.

Because of the number of muscle nuclei and close apposition of the fibers in the MOL, these fibers were drawn with a camera lucida attachment on a compound microscope with 100× oil immersion optics to ensure accuracy. Statistical analysis was performed using a one-way ANOVA with the Statgraphics program (Statistical Graphics Corporation).

Larval body walls from wandering stage third instar larvae were processed for immunohistochemisty (see above) using a polyclonal anti-twist antibody (1:1000, a gift from Dr Bruce Paterson).

Hydroxyurea (HU) was fed to larvae to kill myoblasts. HU has been shown to kill dividing cells when present at sufficiently high concentrations during S-phase of the cell cycle, but not to harm cells that are post-mitotic or in a different phase of the cell cycle (Furst and Mahowald, 1985; Truman and Booker, 1986; Broadie and Bate, 1991). Canton-S larvae were collected from timed egg collections as newly molted second or third instar animals within 1 hour of their molt. At the appropriate times, staged larvae were placed on diet prepared with 5 mg/ml HU (see Broadie and Bate, 1991). After the HU treatment, larva were moved to normal food, in uncrowded conditions, and allowed to develop to the pharate adult stage (stage 15, Bainbridge and Bownes, 1981), then, abdomens were dissected, fixed and stained for nuclei with mAb8C5 according to the procedures listed above. Canton-S larvae were fed either during 12-20 hours of the second instar (late second instar group) or from 1-8 hours in the third instar (early third instar group) (see also, Broadie and Bate, 1991).

Because we were interested in cases where relatively few adult muscle nuclei remained, HU-affected fifth abdominal hemitergites were identified by counting the number of nuclei in all muscles (see above). We classified each hemitergite into one of four groups by comparing the number of nuclei in an experimental animal with the number of nuclei in untreated Canton-S males: group 1, no ablation where more than 183 nuclei were present, which is less that one standard deviation below the mean of wild-type males; group 2, partial ablations where from 156 to 182 nuclei were present, which is between one and two standard deviations from the mean of wild-type males; group 3, partial ablations where fewer than 155 nuclei were present, which is more than two standard deviations below the mean of wild-type males; group 4, complete ablation where no dorsal adult muscles were present (Tables 1 and 3).

To independently identify MOL fibers, late second instar (n=19), early third instar (n=39) and 0-24 hour third instar (n=19) P[79B actin-lacZ, ry⁺] larvae were fed HU as described for Canton-S males and then labelled with anti-β-galactosidase antibodies (as above).

The simple pattern of dorsal longitudinal muscle fibers in each abdominal segment is interrupted in the fifth abdominal segment of male flies by a pair of large multi-fiber MOL muscles (Crossley, 1978; Miller, 1950; Lawrence and Johnston, 1984; Gailey et al., 1991; Taylor 1992; Currie and Bate, 1995). MOL fibers and their homologues in abdominal segments in males and females were identified by staining for β-galactosidase in the P[79B actin-lacZ, ry⁺] line (Courchesne-Smith and Tobin, 1989). In pharate or newly eclosed males, only MOL fibers expressed β-galactosidase (Fig. 1A; Currie and Bate, 1995). On the day following eclosion, four to six fibers, the segmental homologues of MOL fibers in female and male abdomens, also expressed β-galactosidase (data not shown; D. A. Currie, personal communication; T.J.S. Merritt, personal communication). Two to four longitudinal muscle fibers arrayed next to the dorsal midline never stained for β-galactosidase in either males (Fig. 1A) or females (data not shown). Thus, the muscles found between the dorsal midline and the lateral insertion of the alary muscle are divided into two groups, MOL or MOL-homologue fibers and medial fibers.

Consistent with their large size, more nuclei were present in MOL fibers than in medial fibers (see also Currie, 1991). MOL fibers had around 24 nuclei per fiber irrespective of whether fibers occupied positions 3, 4, 5, or 6 from the dorsal midline (Figs 1B, 2A). Only 13% of wild-type MOL fibers contained fewer than 20 nuclei (Fig. 1B, 2A). By comparison, medial fibers had on average 15 nuclei per fiber independent of their position. Only 4% of the medial fibers examined had more than 20 nuclei per fiber. As shown in Fig. 2A, there was nearly a two-fold range in the number of nuclei present in individual medial or MOL fibers.

In order to demonstrate that the number of nuclei in MOL fibers was consistently greater than the number in medial fibers, we calculated a difference value for each animal. The average number of nuclei in medial fibers (positions 1+2) was subtracted from the average number of nuclei in MOL fibers (positions 5+6) for each segment (Table 1); a positive number denotes more nuclei in MOL fibers whereas a negative number denotes fewer nuclei in MOL muscle fibers compared with the number in medial fibers. For wild-type males, the difference values were strongly positive, with a mean of nearly 10 additional nuclei in a MOL versus a medial fiber. Thus, the MOL group has nearly 20 to 50 more muscle nuclei than the medial group of fibers.

In wild-type females, all fibers between the dorsal midline and the alary insertion had the same average number of nuclei (Fig. 2B). Because medial and MOL-homologue fibers occupy comparable positions in females as medial and MOL fibers in males, we calculated an average difference value between medial fibers (positions 1+2) and MOL-homologue fibers (positions 5+6). There was no difference between the number of nuclei in medial fibers compared with the number in MOL-homologue fibers (0.9±1.0; mean ± s.e.m.).

Two sex-specific differences were found in the distribution of muscle nuclei among muscle fibers in the fifth abdominal tergite. Medial muscle fibers in females averaged 4-5 more nuclei than medial muscle fibers in males. Conversely, MOL-homologue fibers in females had 4-5 fewer nuclei than MOL fibers in males. The sex-specific differences in nuclei number found between the two muscle groups reflect the unequal partitioning of an equivalent pool of myoblasts into medial and MOL fibers in males. Sex-specific differences were not found in the total pool of muscle nuclei contributing to the dorsal musculature (Table 2). Nor were there differences in the fraction of this pool (56% in males and 58% in females) used to create the medial and MOL or MOL-homologue muscles in males and females (Table 2).

To test directly whether the length of MOL fibers depended solely on the number of muscle nuclei, we reduced the number of myoblasts available to create the adult musculature by feeding hydroxyurea to late second and early third instar larvae (see Materials and Methods for details). To ensure that all fifth abdominal HU-affected hemitergites were detected, nuclei counts were made in abdomens labelled with a nuclei-specific antibody. To interpret the data, all of the cases were divided into four groups ranging from no ablation to complete ablation based on the total number of muscle nuclei present (Table 3).

Ablation groups in which nuclei number was reduced by less than 25% (groups 1 and 2) showed normal-sized MOL fibers (data not shown). In about 10% of group 2 hemitergites, MOLs were separated into two distinct sets of two or three fibers and in a few cases, only two MOL fibers were present rather that the normal three to five. Neither of these two conditions were seen in any untreated males. The entire adult musculature was ablated in group 4, which occurred in 14% of the animals fed in the late second instar (Table 3; see also Broadie and Bate, 1991).

Hemitergites with group 3 partial ablations did show effects on MOL and medial fiber development. In these cases, there were severe reductions in the total number of nuclei, ranging from 75% to 25% of the mean number of nuclei in wild-type males (Table 3). At least one recognizable MOL fiber was produced in 14 of the 15 group 1 hemitergites. These MOL fibers had normal or nearly normal posterior attachment points, but frequently failed to reach their usual anterior attachment points. These fibers were shorter than wild-type MOLs, but were distinctly longer than medial fibers. The nuclear distribution across the dorsal hemisegment in four examples of group 3 partial ablations are shown in Fig. 3A-D. In cases where only one-quarter to one-half of the wild-type number of nuclei remained, medial and MOL fibers contained the fewest nuclei; medial fibers were present with as few as one or two nuclei and MOL-type muscles formed with as few as six to eight nuclei, about a quarter of the average number of nuclei in wild-type MOLs (Figs 3A-C, 4A). When the number of nuclei in a group 3 hemisegment was greater than one-half the normal number, more nuclei were found in both muscle groups (Fig. 3D). The striking feature of these cases was that MOL fibers always had more nuclei compared with their medial fiber neighbors; the average difference value for hemitergites with group 3 ablations was 6.2 ± 0.8 (mean ± s.e.m.), nearly two-thirds the value obtained for wild-type MOLs (Table 1 and Fig. 4A).

In two group 3 hemitergites, no MOL fibers were detected. In the first case, only six fibers were present with a total of 12 nuclei; two extremely small fibers, which could not be identified, were found medial to the alary attachment and four fibers were located more laterally. In the other case, a complete nuclear count was not possible, but four of a total of eight fibers in the hemisegment were located between the dorsal midline and the alary insertion, about half the expected number of fibers in this region. Each fiber had 10-11 nuclei which should have been sufficient to have produced a visible MOL-type fiber. These two examples may represent cases with regional ablations of one or more myoblast nests as opposed to reductions of some but not all myoblasts within a nest (Broadie and Bate, 1991).

Because MOL fibers formed male-like attachments even with fewer nuclei than normal, we evaluated whether another male-specific feature of MOL development, the early onset of β-galactosidase expression in P[79B actin-lacZ] males, was affected in ablated hemisegments. Unexpectedly, several recognizable MOL fibers failed to express β-galactosidase. In the early third instar treatment group, 82% (n=39 hemitergites) had normal-sized MOL fibers that expressed β-galactosidase. The remaining six hemitergites did not have any β-galactosidase-positive fibers, although in five cases, one or two fibers were recognizable as MOLs from their deeper posterior attachment points. The one remaining hemitergite had 3 fibers between the dorsal midline and the alary attachment but lacked any MOL fiber. In all cases, the contralateral MOL muscle was β-galactosidase positive. Where MOL fibers were unstained, it was not possible to accurately count the number of nuclei, but the fibers were thin indicating that they did not contain many nuclei. In P[79B actin-lacZ, ry⁺] males fed as late second instars (n=19 hemisegments), 11% of the hemisegments had complete ablations of the adult musculature. All remaining cases from the second instar and all of the cases from the 24-hour third instar treatment group had MOL fibers that expressed β-galactosidase. Among these hemitergites were two exceptional cases. In one case, a single thin MOL fiber was present, which expressed β-galactosidase and had 17 nuclei (data not shown). In the second case, two, closely apposed MOL fibers were visible; the MOL fiber with 14 nuclei stained strongly for β-galactosidase, but the neighboring MOL fiber with only 9 nuclei stained only extremely weakly (Fig. 4B,C). Thus, only a subset of MOL fibers from HU-affected hemitergites, recognizable by their attachment points, expressed the full range of detectable MOL-specific phenotypes.

Because there is gradation in MOL muscle phenotype in various fru mutants (Gailey et al., 1991), we expected that the relationship between the number of nuclei and muscle size in fru mutants would conform to one of two possibilities. MOL-position fibers in fru mutants might have the same number of nuclei as wild-type MOLs, but be unable to produce a normal-sized MOL. Alternatively, MOL-position fibers in fru mutants might have fewer muscle nuclei and thus be limited to making smaller MOLs. To distinguish between these alternatives, we examined several fru mutant genotypes encompassing the full range of MOL phenotypes. The existence of MOL-position fibers in fru mutants was confirmed in P[79B actin-lacZ, ry⁺] fru¹ recombinant males which had four to six β-galactosidase-positive fibers more laterally than two to four medial fibers (data not shown).

In all of the fru mutant genotypes examined, we found that there were fewer muscle nuclei in MOL-position fibers than in MOL fibers from wild-type males. In the most severely affected mutants, such as Df(3R)P14/Df(3R)Cha^M5 males, MOL-position muscles in the dorsal region were the same size as medial muscles (Fig. 5; Gailey et al., 1991). In the scatter-plot of nuclei counts from Df(3R)P14/Df(3R)Cha^M5 males (Fig. 6A), medial fibers in positions 1, 2, and 3 had on average 15 nuclei per fiber, similar to the average number of nuclei found in wild-type medial group fibers (Fig. 2A). Muscles in positions 4, 5 and 6 had slightly fewer nuclei per fiber (Fig. 6A). Only 3% of the MOL-position fibers had more than 20 nuclei per fiber. We computed the mean difference value between fibers at positions 5 and 6, corresponding to MOL-position fibers, and those at positions 1 and 2, corresponding to medial fibers. The average difference value was slightly negative confirming that in these mutants MOL-position fibers had fewer nuclei than medial fibers (Table 1).

In fru mutants, which produced no full-sized MOL fibers and generally have only small sized MOL-position fibers (Gailey et al., 1991), the average number of nuclei for fibers in the various medial to lateral positions was similar to that found in the double deficiency fru mutants. In fru¹ homozygotes, fru¹/Df(3R)P14 and fru¹/Df(3R)Cha^M5 transheterozygotes, MOL-position fibers (4, 5, and 6) had nearly the same number of nuclei as medial fibers (1, 2, and 3; Fig. 6B). The average difference values calculated for these mutant genotypes showed that medial and MOL-position fibers had roughly the same number of nuclei (Table 1). In two other fru mutant genotypes, fru³ and fru⁴, where no MOL or intermediate-sized fibers formed (Castrillon et al 1993; data not shown), the average number of muscle nuclei per fiber was greater than for other fru mutants (Fig. 6B), but even in these animals, there were fewer nuclei in MOL-position fibers than medial fibers (Fig. 6B, Table 1). The large negative difference value for fru³ animals was due primarily to the greater number of nuclei in medial fibers not to an unusually severe reduction in the number of nuclei in MOL-position fibers (Fig. 6B).

In the weakest genotype, fru² homozygous males, a normalsized MOL often forms (Gailey et al., 1991). On average, the number of nuclei in large-sized MOL fibers was slightly larger than small-sized MOL fibers in the same positions (Fig. 6B). When the difference value was calculated for fru² MOL or intermediate-sized fibers, there were slightly more nuclei than in medial muscles; but these fibers still had significantly fewer nuclei than wild-type MOL muscles (P≤0.05; Table 1). Indeed, the mean difference value for fru² MOL fibers compared with small MOL-position fibers in fru² or in other fru genotypes was not significant (Table 1).

Since there were fewer nuclei in MOL-position fibers in fru mutants, we determined whether there were fewer muscle nuclei altogether in the hemitergite. Several mutant genotypes, fru¹ homozygotes, fru¹/Df(3R)Cha^M5 and fru¹/Df(3R)P14 transheterozygotes, had fewer muscle nuclei than did Canton-S males, but fru² homozygotes and Df(3R)P14/ Df(3R)Cha^M5 transheterozygotes were not different from wild-type males (Table 2). Between 56-61% of the total muscle nuclei were distributed in the medial two groups of muscles in fru mutants similar to the proportion contributed to these muscle in wild-type males, indicating that MOL-position muscles in fru mutants did not exist in a specially disadvantaged region within the hemisegment.

Adult dorsal longitudinal muscles derive from dorsal groups of myoblasts, detected as twist-positive cells, located near the larval dorsal oblique muscles 1, 2 and 3 (Bate et al., 1991; Currie and Bate 1991; Broadie and Bate, 1991). Fewer adult muscle nuclei, such as found in many fru mutant males, might result from a reduced number or abnormal distribution of these founder populations of larval myoblasts or to later events, such as myoblast fusion.

twist-positive cell counts in wild-type and fru mutant larvae show that both genotypes have the same number of myoblasts. In wild-type wandering third instar larvae, 18 twist-positive cells were found distributed among the three dorsal groups in the fifth abdominal hemisegment of both males and females (Table 4; see also Broadie and Bate, 1991). Occasionally, myoblasts were not clearly collected into three nests but were distributed along the intersegmental nerve or merged into two groups. Male Df(3R)P14/Df(3R)Cha^M5 larvae had the same number and distribution of twist-positive cells as wild-type male and female larvae (Table 4, Fig. 7). The lateral and ventral myoblast nests, which contribute to the lateral and ventral adult muscles, were also present in these and the other fru mutant larvae examined (data not shown) indicating that the original embryonic patterning of the adult musculature was unaffected by mutations in fruitless.

Homozygous fru¹ and fru² mutant larvae have one-third to one-half as many twist-positive cells as wild-type animals (Table 4). These differences were significant when compared to wild-type or double-deficiency males. In fru¹ and fru² males, the cluster of twist-positive cells associated with larval muscle DA3, the most lateral group, was commonly the smallest cluster, often having only two cells. When fru¹/Df(3R)P14 or fru¹/Df(3R)Cha^M5 males were examined, the number of myoblasts was not significantly different from wild-type males (Table 4), making it likely that the reduction in the number of myoblasts is a background effect due to homozygosis of the fru¹ chromosome, not to a direct fru mutant effect.

In this study we have shown that male-specific MOL fibers consistently contain ten more muscle nuclei than neighboring medial fibers, commensurate with their larger size. This differential is maintained over a two-fold range in the number of nuclei present within individual medial and MOL fibers. In contrast, female MOL-homologue and medial muscles are the same size and contain equivalent numbers of muscle nuclei. No sex-specific differences were found in the number of larval myoblasts or in the combined number of nuclei present within the adult medial and MOL or MOL-homologue muscles. Sex-specific recruitment of nuclei into MOL fibers appears to result from the unequal allocation of a common pool of myoblasts during the formation of medial and MOL muscles rather than a special set of male-specific myoblasts.

A sex-specific increase in nuclei in MOL fibers might depend on one of three different developmental processes. First, simply lengthening the time period for myoblast fusion to create MOL myotubes would lead to an increase in the number of nuclei found in MOL compared to medial fibers. In embryos, muscle size is proportional to the number of myoblasts that have fused, which in turn depends on the length of the fusion period (Bate, 1990). Second, male MOL fibers might actively recruit myoblasts that were destined to contribute to medial group muscles or other nearby muscles. Third, some combination of developmental processes such as malespecific muscle nuclei divisions in MOL myotubes and apoptosis of previously fused medial myotube nuclei could account for the numerical disparity between MOL and medial muscles. The period of myoblast division in the adult overlaps with the period of fusion but no evidence of male-specific divisions in MOL myotubes has been detected in pupae that have been exposed to 5-bromodeoxyuridine to label nuclei in S-phase (D. A. Currie, personal communication; B. J. Taylor unpublished observations).

We favor the second alternative. In HU-affected hemitergites with severe reductions in the number of myoblasts, the differential in the number of nuclei per MOL fiber compared to the number of nuclei per medial fiber is nearly as great as that in normal fibers. Under these conditions where, effectively, no myoblast surplus exists, the difference between MOL and medial fibers would be expected to diminish or disappear if due only to a passive lengthening of the fusion period. The severe reduction of nuclei in medial fibers, compared to longitudinal muscles lateral to the MOL, (see Fig. 3D) supports an active process in MOL fibers operating at the expense of medial fibers, within a general lateral to medial progression in the formation of abdominal muscle fibers.

We had anticipated that there would be a strict lower limit to the number of nuclei needed to produce either a medial or a MOL fiber. From the range of nuclei number in wild-type animals, medial fibers appeared to need at least seven nuclei, and small MOL fibers at least 15. From examination of HU-affected hemitergites, we were astonished to discover that medial muscles with one and two nuclei were able to stretch to their normal attachment sites even though these fibers were exceedingly thin. Easily recognizable MOL fibers, with deeper posterior attachments than medial muscles, formed with as few as six nuclei, about one-quarter the average number of nuclei in wild-type muscles. There were too few cases to determine whether MOL fibers developing with fewer than six nuclei would be indistinguishable from medial muscles. At the other end of the scale, the formation of full-sized MOL muscles required closer to twenty nuclei in both wild-type or HU-treated males.

From the HU ablations with control and P[79B actin-lacZ, ry⁺] males, it is apparent that the number of MOL fibers and the expression of β-galactosidase from the 79B actin promoter were more sensitive to the number of nuclei than was the outgrowth of muscle fibers toward their attachment points. In none of the severely affected hemisegments were more than two MOL-type fibers found compared with three to five fibers present in wild-type animals. Additionally, recognizable MOL fibers did not always express β-galactosidase. Even fibers of equivalent size had extremely divergent reporter expression that correlated with the number of nuclei present (see Fig. 4B,C). We interpret the lack of β-galactosidase staining in five other MOL fibers in HU-treated P[79B actin-lacZ, ry⁺] males as due to a lack of sufficient muscle nuclei to promote muscle maturation.

An alternative explanation for the loss of β-galactosidase expression in HU-treated males is the death of the MOL-specific motorneurons, which would prevent the induction of the MOL (Lawrence and Johnston, 1986; Currie and Bate, 1995). A direct effect of HU on motorneurons seems improbable since abdominal motorneurons for skeletal muscles in holometabolous insects are generated in the embryo and so should be resistant to the HU treatment (e.g. Taylor and Truman, 1974; Giebultowicz and Truman, 1984; Thorn and Truman, 1989; Truman and Booker, 1986; Broadie and Bate, 1991; B. J. Taylor, unpublished observations). Indeed, in our experiments, even smaller MOL fibers were recognizable as MOLs by their attachment points suggesting that they had been innervated by MOL-specific motorneurons.

Thus, the steps in muscle development that appear to be the most dependent on the number of nuclei are involved in terminal differentiation. Attachment to the epidermis by developing abdominal muscles is an event that occurs about midway during the pupal stage (Currie and Bate, 1991; Gailey et al., 1991) followed by terminal differentiation. The inability of MOL fibers with fewer nuclei to respond appropriately perhaps reflects a failure to reach this fully differentiated state. Further examination of the MOL fibers in HU-affected hemitergites would be needed to demonstrate whether general muscle cell differentiation, such as the accumulation of glutamate receptors at the muscle endplate or the expression of other muscle-specific proteins, was also reduced, or whether only the sex-specific differences associated with MOL development were affected in fibers having a fraction of the normal complement of nuclei. Our results indicate that the formation of the proper attachment points and the expression of a late stage muscle-specific reporter gene are independent phenomena that have different requirements for the number of muscle nuclei.

Considering the small size of MOL-position fibers in severe fru mutants, one of our initial expectations was that fewer muscle nuclei would be present in MOL-position fibers concomitant with their reduced size. Indeed, MOL-position fibers in nearly all fru mutants had either the same or slightly fewer muscle nuclei than did neighboring medial fibers (see Table 1), a significant difference compared to the relationship between MOL and medial fibers in wild-type males. These results suggest that one role of the fru⁺ product is to mediate the male-specific recruitment of nuclei into MOL fibers. The fru mutant effect appears to be specific to MOL muscle development and not a generalized inability to form dorsal abdominal muscles.

We propose that the first developmental stage where fru⁺ acts to generate a normal MOL muscle occurs during the fusion of myoblasts into MOL myotubes. Since the number of twist-positive cells in fru mutant larvae was the same as in wild-type larvae, there appears to be no fru mutant effect at early stages of myogenesis. Subsequent divisions by these myoblasts in the pupal stage generated a sufficient pool of myoblasts to have generated a normal-sized MOL in at least two of the strong fru⁻ genotypes when compared to the HU-ablation treatments (see Tables 2 and 3). In addition, the medial and MOL-position fibers in fru mutant males used the same fraction of the total muscle nuclei pool as did medial and MOL fibers in wild-type males. The remaining step in muscle development which would affect the final number of nuclei in muscles is fusion, leading us to the conclusion that in fru mutant males, MOL-position fibers were apparently unable to actively recruit additional myoblasts.

The reduction in the number of nuclei in MOL-position fibers in fru mutant males is not sufficient to account for the small size of these fibers. In one of the five fru mutant genotypes tested, fru¹/Df(3R)P14, the total number of nuclei fell within the range found for the HU-treatment group 2 partial deletions. Two other fru mutant genotypes, fru¹/fru¹ and fru¹/Df(3R)Cha^M5, had total nuclei counts similar to the HU-treatment group 3 partial deletions. Normal sized or reduced MOL fibers were present in virtually all of the HU-treated hemitergites whereas MOL-position fibers were no larger than medial fibers in virtually all of the severe fru mutant genotypes. In the genotype, fru², where normal sized MOL fibers form, the number of either total muscle nuclei or the number of nuclei in MOL-position fibers is not different from other fru genotypes where MOLs do not form. On the basis of the distribution of nuclei into MOL and medial fibers in hemisegments where the number of total nuclei had been reduced to that of various fru mutants, we would have expected that MOL fibers would have been able to sequester essentially normal numbers of nuclei and that medial fibers in these mutants would have had even fewer nuclei than found in the medial fibers of wild-type males.

Our findings support a dual role for fru⁺ activity in the development of the MOL. The first role is in the male-specific recruitment of additional myoblasts to the MOL myotubes and in an additional, independent step, fru⁺ product leads to the elongation of MOL fibers to their appropriate attachment points. Due to the unusual dependence of MOL development on innervation by male-specific motorneurons, fru⁺ may act either in the MOL-specific motorneuron, in the generation of the male-specific inductive activity, or in the myoblasts to mediate the response to the male-specific inductive signals.

We are grateful to Drs Seymour Benzer and Bruce Paterson for the generous gift of antibodies. Dr Sara Tobin kindly provided the P[79B actin-lacZ, ry⁺] line. The fru stocks were gifts from Drs Steve Wasserman, Diego Castrillon, Jeffrey Hall and Don Gailey. We would also like to thank Carol Rivin, Barbara Wakimoto, Doug Currie and Thomas Merritt for their careful reading of the manuscript. We want to thank Douglas Currie and Thomas Merritt for permission to cite their unpublished work. We also appreciate the many fruitful discussions with Dr Doug Currie, Dr Carol Rivin, Dr Jeffrey Hall, Dr Bruce Baker, Adriana Villella and Dr Mike McKeown during the course of these experiments. This research was supported by National Science Foundation grant IBN-9210785 to B. J. T.