Crossveinless and the TGFβ pathway regulate fiber number in the Drosophila adult jump muscle

Higher animals are characterized by complex arrangements of skeletal muscles, which are used for posture and locomotion. Understanding the origin of these muscles is of crucial importance to human medicine, as a number of debilitating diseases strongly impact skeletal muscle function and maintenance(Nishino and Ozawa, 2002). It is also apparent that several muscle diseases affect some muscle groups to a greater extent than other groups; thus, understanding how individual skeletal muscles arise is a central challenge in the field.

Although there is still much to learn regarding muscle patterning in mammalian muscle specification, some significant insight has been afforded by invertebrate systems. In grasshopper embryos, Ho et al.(Ho et al., 1983) were the first to identify, among the myoblast pool, individual larger cells that appeared to seed the formation of individual muscle fibers. These `muscle pioneers' were subsequently identified in Drosophila(Bate, 1990), where it was shown that single skeletal muscle fibers arise from the fusion of a muscle pioneer or `founder' cell with a small number of fusion-competent myoblasts.

Each embryonic founder cell is also largely responsible for the acquisition of fiber-specific phenotypes, such as patterns of gene expression, innervation and muscle attachment locations. This was concluded based upon myoblast fusion mutants, where unfused founder cells still attempt to make appropriate orientations and connections (Rushton et al., 1995) (reviewed by Baylies and Michelson, 2001). Furthermore, specific muscle phenotypes arise from individual patterns of regulatory gene expression within founder cells (Crozatier and Vincent,1999; Knirr et al.,1999; Clark et al.,2006). Thus, understanding the genetic pathways that contribute to founder cell specification will impact our understanding of muscle specification. Along these lines, signaling pathways including the Wingless pathway (Cox and Baylies,2005) and the epidermal growth factor pathway(Buff et al., 1998) contribute to founder cell selection in the embryo. Nevertheless, there are still several parts of this specification process that have yet to be uncovered.

During Drosophila metamorphosis, most of the larval skeletal muscles degenerate and are replaced by new muscles arising from imaginal myoblasts (Crossley, 1978; Currie and Bate, 1991; Fernandes et al., 1991). These adult myoblasts are specified during embryogenesis and many become associated with the imaginal discs (Poodry and Schneiderman, 1970; Bate et al., 1991). Subsequently, the adult myoblasts migrate from the discs to the future locations of the muscles, and myoblasts from each disc give rise to a variety of physiologically distinct muscles(Lawrence, 1982). However,mechanisms that control the specification of many of these muscles have yet to be fully elucidated.

Does the founder cell model of muscle development also hold true for adult myogenesis in Drosophila? This question is important given the complex organization of the adult skeletal musculature: individual fibers can span several segments and multiple fibers are often arranged into larger muscles that are more characteristic of those found in mammals. In the adult thorax, there are two major types of muscles (reviewed by Bernstein et al., 1993): the fibrillar indirect flight muscles (IFMs) are adapted to contract at high frequency to provide the power for flight. The IFMs comprise six pairs of medial fibers termed the dorsal longitudinal muscles (DLMs), and three pairs of lateral muscles termed the dorsoventral muscles (DVMs). In addition to the fibrillar flight muscles, physiologically distinct tubular muscles are located laterally and ventrally in the thorax. Tubular muscles function in walking,jumping and angling the wings. Tubular muscles each are composed of several individual fibers grouped together. Most prominent among the tubular muscles is the tergal depressor of the trochanter (TDT, or `jump') muscle, a large fiber found in all Diptera studied which attaches the dorsal notum to the second pair of legs. This muscle is essential for the escape response of the fly (Nachtigall and Wilson,1967).

Several groups have demonstrated that the formation of some adult skeletal muscle fibers are associated with cells showing the characteristics of founders. These include the abdominal muscles, where each individual muscle is pre-figured by a cell expressing the canonical founder cell marker, dumbfounded/kirre, usually detected as an enhancer trap termed duf-lacZ, or rp298(Ruiz-Gomez et al., 2000). Precursors of the adult indirect flight muscles also express duf-lacZ(Dutta et al., 2004), and the ablation of these founder cells significantly disturbs the formation of the DVMs (Atreya and Fernandes,2008). Interestingly, whereas founders have been observed to prefigure adult muscle development, relatively little is known of the mechanisms responsible for their specification at this stage. In fact, the process of singling out founder myoblasts, which in the embryo requires in part lateral inhibition via the Notch pathway(Carmena et al., 1995; Carmena et al., 1998), appears to occur in a Notch-independent manner in the adult(Dutta et al., 2004). Thus,understanding specification of adult muscles should provide further new insight into muscle specification mechanisms.

In this work, we have analyzed the basis of a mutation that affects the morphology of the jump muscle of the adult thorax. This mutation, which causes a reduction in TDT fiber number and defects in the morphology of the muscle,arises from mutation of the crossveinless (cv) gene, the established function of which is to modulate TGFβ signaling in the specification of wing crossveins. We demonstrate that cv functions in muscle development as part of the TGFβ pathway, which is activated autonomously in the adult myoblasts in order to control the number of founder cells specified for the TDT. By manipulating the TGFβ pathway, the TDT,which normally comprises 20-30 muscle fibers, can be modified to consist of as little as five fibers or as many as 50 fibers. Overall, these studies define an important function for the TGFβ pathway in adult muscle specification that might also be used in the formation of the more complex muscles found in higher animals.

Drosophila were grown on Carpenter's medium(Carpenter, 1950) at 25°C unless indicated. Stocks carrying cv¹, cv⁴³,dpp¹⁰⁶³⁸, X chromosome deficiencies, the second chromosome deficiency covering gbb, and Gal4 driver lines (unless noted) were obtained from the Bloomington Drosophila Stock Center. UAS-Dad (Tsuneizuni et al.,1997) and UAS-tkv^*(Adachi-Yamada et al., 1999)were from Stuart Newfeld (Arizona State University, AZ, USA); UAS-cvwas from Larry Marsh (UC Irvine, CA, USA); 1151-Gal4(Anant et al., 1998) was from L. S. Shashidhara (CCMB, Hyderabad, India); and duf-lacZ was from Upendra Nongthomba (India Institute of Science, Bangalore, India). The wild-type strain Simms-L was locally captured in Albuquerque (NM, USA) and determined to be conspecific based upon visual examination and crossing with established laboratory strains.

Samples were prepared for paraffin sectioning as described by Lyons et al.(Lyons et al., 1990) and modified by Cripps et al. (Cripps et al.,1998). Sections were cut at 8-12 μm, and stained with Hematoxylin and Eosin (Sigma) for evaluation of TDT structure. Stained slides were dehydrated through 100% ethanol, soaked in xylene, and mounted in Cytoseal-XYL (VWR Scientific Products). TDT fibers were counted from both sides of the thorax and treated as independent samples. Averages for each genotype were calculated and comparisons were performed using Student's t-test at Graphpad.com.

Cryosections were prepared by embedding adult flies in OCT medium followed by freezing. Sections were cut at 15 μm at -18°C and air dried. Next,samples were fixed for 5 minutes at room temperature with 1.9% v/v formaldehyde in 1×PBS, washed and used for antibody staining as described in the following section.

For pupal dissections, newly pupariated animals were marked, and aged for the appropriate time until harvesting and dissection. All pupal samples were dissected in a Sylgard-coated petri dish (Dow Corning) and pinned open. After dissection, samples were fixed for 30 minutes on ice with 5% formaldehyde in 1×PBS, washed in PBTx [1×PBS, 0.2% v/v Triton-X100, 0.2% w/v Blocking Agent (Roche)], and then subjected to blocking and antibody incubations (see below).

For documentation of adult wings, wings were removed from adult flies and stored in 70% (v/v) ethanol overnight, then transferred twice to 100% ethanol. Wings were next soaked in 100% xylene, and mounted using Cytoseal-XYL (VWR Scientific Products) for photography.

Fixed and washed samples were subjected to immunostaining essentially as described by Patel (Patel,1994) and modified by Molina and Cripps(Molina and Cripps, 2001). Primary antibodies used were: anti-βPS-integrin 1:10(Brower et al., 1984)(University of Iowa Developmental Studies Hybridoma Bank, IA); anti-Z(210)1:100 (Vigoreaux et al., 1991)(kindly supplied by Jim Vigoreaux, University of Vermont, VT, USA); anti-MEF2 1:2000 (Lilly et al., 1995)(kindly supplied by Bruce Paterson, NIH); rabbit anti-β-galactosidase 1:1000 (AbCam); and mouse anti-β-galactosidase (Promega). For immunofluorescence, secondary antibodies were Alexa conjugated (Molecular Probes), and mixed with Alexa-488 phalloidin at 1:500 (Molecular Probes) and DAPI used at 2 μg/ml (Sigma). For immunohistochemistry, secondary antibodies and detection were carried out using the Vectastain Elite staining kit and diaminobenzidine (DAB) substrate according to the manufacturer's recommendations.

In situ hybridization was carried out using digoxigenin-labeled probes(Roche) according to the method of O'Neill and Bier(O'Neill and Bier, 1994). Probes for cv were generated from the plasmid pOT2/SD27025(Drosophila Genomics Resource Center, Indiana University, IN, USA): sense probes were synthesized from plasmid cut with XhoI using T7 RNA polymerase; antisense probes were generated from plasmid cut with EcoRI using SP6 RNA polymerase.

Images were collected using an Olympus BX-51 stereomicroscope using either DIC or fluorescence optics. Digitally captured images were assembled into figures using Adobe Photoshop.

Jump tests were performed essentially as described by Cripps et al.(Cripps et al., 1994). Briefly, flies lacking wings and aged 2-3 days after eclosion were induced to jump from a platform elevated 10 cm above a piece of white paper. The landing location was marked and the lateral distance from the edge of the platform to the landing point was measured in mm. Average distances were calculated for more than 20 individuals for each genotype and compared using Student's t-test.

The jump muscle of the Drosophila thorax comprises large and small cells, organized into a rosette. There are generally 26-28 large cells and four small cells (O'Donnell et al.,1989; Peckham et al.,1990). The cells are visualized in paraffin sections cut horizontal to the axis of the muscle (Fig. 1A), and also are outlined in anti-β-PS integrin stains of horizontal cryosections (Fig. 1B). Large and small cells can additionally be distinguished based upon the presence (large cells) or relative absence (small cells) of the Z(210) Z-disc-associated protein (Fig. 1C) (Vigoreaux et al.,1991).

Control animals in some of our publications have shown fewer fibers than the canonical 26-28 (Baker et al.,2005), prompting us to evaluate TDT fiber number in a range of strains. The number of small cells only occasionally deviated from four (a range of three to five); however, there was significant variation in the number of large cells (summarized in Table 1). Some strains showed the published 26-28 large fibers(Fig. 1A), other lines showed as few as 18 large fibers (Fig. 1B). These data include a locally caught strain named Simms-L,which showed an average of 22 large cells. Nevertheless, in all but one of these strains, the shape and arrangement of the fibers was maintained. Given the importance of this muscle to the escape response of the fly, we sought to investigate the genetic basis of this variation in more detail.

One strain tested carried the X-chromosome linked visible mutations yellow, crossveinless, vermillion, forked, carnation (y cv¹ v f car). For this strain, the number of TDT large fibers was severely reduced, to an average of 13(Table 1; Fig. 1D,E). Although both large and small cells were detectable (Fig. 1F), the overall morphology of the muscle was usually abnormal,with the cells failing to form a ring of fibers. We also observed an overall increase in the size of the individual fibers. We attribute the latter observation to a common size of myoblast pool being divided into fewer distinct fibers in the mutants. This would cause more myoblasts to contribute to each fiber and might result in an increase in fiber volume. The model that fiber size corresponds to the number of contributing myoblasts has already been proposed for the indirect flight muscles(Farrell et al., 1996). Given the large reduction in fiber number in the y cv¹ v f carstrain, we defined the cause of this phenotype.

Using standard genetic crosses and analyzing TDT fiber number in individuals, the reduction in fiber number in the X-chromosome stock arose from a recessive mutation that mapped to the marked X chromosome (data not shown).

To localize the mutation, we crossed y cv¹ v f car with a wild-type strain to generate y cv¹ v f car/+ + + + +females, which were then backcrossed to wild-type males. Male recombinant offspring from this cross were collected, genotyped using the visual markers,and individually assessed for TDT fiber number. The results are shown in Table 2. In all cases the TDT fiber phenotype segregated according to the allele of cv that was present: cv⁺ recombinants showed a normal fiber number,whereas cv¹ recombinants showed defective TDTs. These results placed the TDT mutation in the proximity of cv.

To localize the mutation more precisely, we analyzed the phenotype of females heterozygous for the cv¹ allele on one X chromosome and a deficiency on the other X. Combination of cv¹ with Df (1) N73 (deficient for 5C2 to 5D5-6)showed a normal TDT phenotype (Fig. 2A), whereas cv¹ heterozygous with Df (1)C149 (deficient for 5A8-9 to 5C5-6) uncovered the TDT phenotype(Fig. 2B). Similarly, we found that duplication of this segment of the X chromosome could rescue the TDT phenotype: in cv¹/Dp(1;Y) dx⁺1 males (where a wild-type X chromosome segment comprising 5A8-9 to 6D8 is translocated to the Y), the TDT phenotype was normal (Fig. 2C). These data placed the TDT mutation in the genomic region 5A8-9 to 5C2 of the X chromosome.

Given the close linkage of the TDT mutation to the cv¹allele, and given that cv resides at 5A13 of the X chromosome, we next tested whether mutation of cv might be responsible for the TDT phenotype by analyzing cv⁴³, an allele which is null for cv gene function, and which has a genetic background distinct to that of cv¹ (Vilmos et al.,2005). cv⁴³ homozygotes also showed a severe TDT phenotype (Fig. 2D), and cv¹/cv⁴³ female heterozygotes had defective TDT structure (Fig. 2E). These studies demonstrated that mutation of cv, in addition to affecting the crossveins of the adult wing (Bridges,1920), affects the normal development of the adult jump muscle.

We also determined whether genetic rescue of the cv mutation would rescue the jump muscle phenotype. We used the Gal4-UAS system(Brand and Perrimon, 1993) to determine whether Gal4-driven expression of cv was sufficient to rescue the mutant phenotype. We also followed the crossveinless phenotype in the wing to control for our manipulations; wild type and mutant are indicated in Fig. 3A,B.

When we generated a homozygous stock of the genotype w¹¹¹⁸cv⁴³; UAS-cv, this stock showed some rescue of the crossvein phenotype in the wings (Fig. 3C, left panel). Upon sectioning these adults, the TDT phenotype was also rescued (Fig. 3C,right panel). We attribute this finding to the UAS-cv transgene used being slightly leaky, such that in the homozygous condition it generates sufficient Cv protein to rescue the two phenotypes. By contrast, when we studied w¹¹¹⁸ cv⁴³; UAS-cv/+ (i.e. mutants carrying just one copy of the UAS-cv transgene), adults showed the mutant crossvein and TDT phenotypes (Fig. 3D), indicating that two copies of the UAS construct were required for rescue.

Rescue of the crossvein and TDT phenotypes confirmed that loss of cv function was responsible for the TDT phenotype. We extended the rescue experiment by crossing w¹¹¹⁸ cv⁴³;UAS-cv homozygous females to males carrying either the ectodermal driver patched-Gal4 (Wilder and Perrimon, 1995), or the mesodermal driver 24B-Gal4(Brand and Perrimon, 1993). In male offspring for both of these cases (offspring of the genotype w¹¹¹⁸ cv⁴³/Y; UAS-cv/ptc-Gal4 or w¹¹¹⁸ cv⁴³/Y; UAS-cv/+; 24B-Gal4/+),significant rescue was observed for both the crossvein and the TDT(Fig. 3E,F).

The observation that either a mesodermal or an ectodermal driver could rescue the wing and muscle phenotypes of cv⁴³ can be explained by the demonstration that Cv is a secreted protein(Shimmi et al., 2005; Vilmos et al., 2005), and might diffuse from a source to the target tissue. It is interesting to note that 24B-Gal4 can direct sufficient Cv synthesis that the wing vein phenotypes can be rescued. In this instance, it is possible either that 24B-Gal4 is expressed at some levels outside of the mesoderm, or that the production of Cv in this background sufficiently stabilizes the ligand to which it binds in order to allow signaling over large distances.

To determine whether TDT defects are a common phenotype among mutants showing wing crossvein alterations, we studied the fiber number in mutants for crossveinless-2 (cv-2) and detached (det). In both of these cases, no obvious fiber defect was observed(Table 1), indicating that cv must play either a unique or a more crucial function in muscle development than other members of this mutant class.

We next determined whether reductions in the number of TDT fibers affected TDT function. This muscle is solely responsible for the jump response in flies(Nachtigall and Wilson, 1967;Elliot et al., 2007), thus we carried out jump tests of wild-type, mutant and rescued flies. For w¹¹¹⁸ controls, the jumping distance was 48 ±2 mm, whereas w¹¹¹⁸ cv⁴³ males jumped 38±3 mm. This difference was significant using the Student's t-test (P=0.02). When we analyzed the jumping ability of w¹¹¹⁸ cv⁴³; UAS-cv rescued males flies (the same genotype used in Fig. 3C),the jumping distance was rescued to 50±3 mm, essentially indistinguishable from wild type. These experiments revealed that the TDT defects in cv mutants result in subtle, albeit significant defects in jump muscle performance. Clearly, reduction in TDT fiber number causes severe abnormalities in both muscle morphology and muscle performance.

Recently, cv was shown to be a component of the TGFβsignaling pathway, facilitating Dpp signaling in the formation of crossveins(Shimmi et al., 2005; Vilmos et al., 2005). To determine whether the cv muscle phenotype arises from disruptions in TGFβ signaling, we tested the effects upon TDT development of activation and repression of the TGFβ pathway.

To achieve this, we firstly crossed the adult myoblast driver 1151-Gal4 to a line carrying UAS-tkv^*, an autonomous constitutive activator of the TGFβ pathway. We analyzed TDT structure in adult offspring (termed 1151>UAS-tkv^*)using both paraffin sections and immunostained cryosections. Compared with wild type (Fig. 4A-C), the mutant offspring showed a striking increase in the number of TDT fibers,although each individual fiber was somewhat smaller than in wild-type muscles(Fig. 4D,E). We consistently observed over 50 fibers per TDT, approximately twice the normal number. By immunostaining we observed that most of the fibers showed accumulation of Z(210), a characteristic of the large TDT cells(Fig. 4F). Thus, fiber number,but not fiber fate, was altered in these mutants.

We next analyzed the effects of repressing the TGFβ pathway upon TDT formation, by creating flies of the genotype 1151>Dad, where Dad encodes an intracellular repressor of the TGFβ pathway. Here, there was a severe reduction in the total number of TDT fibers to approximately six (Fig. 4G-I).

The opposing effects upon TDT fiber number of activating or repressing the TGFβ pathway provided strong evidence for the involvement of this pathway in TDT development. Moreover, as we used a Gal4 driver line active in the adult myoblasts, we also conclude that the TGFβ pathway is activated in the myoblasts themselves, rather than in an adjacent cell population.

To define the cellular basis for the TDT defects, we studied the early development and specification of the TDT in duf-lacZ at late larval and early pupal time points. By correlating this expression with markers of the muscle lineage, we defined crucial events in TDT formation, which could be compared with cv mutants. First, the TDT is pre-figured by the specification of founder cells at the proximal region of the T2 leg imaginal disc. The founders were somewhat difficult to discern during the late larval stage (Fig. 5A), either because they were just being specified, or because the duf-lacZ reporter was only just becoming activated. In the early pupal stage [2 hours after puparium formation (APF)] 12-15 duf-lacZ-expressing founder cells were observed in this region of the imaginal disc(Fig. 5B), and by 8 hours APF had increased to an average number of 21.5±1.3 founder cells(Fig. 5C).

As pupal development proceeded, this group of founder cells migrated laterally and dorsally, among a large number of MEF2-positive myoblasts. By 16 hours APF, the TDT founders had increased in number to 25.4±1.5,similar to the final number of large plus small fibers in the mature TDT of duf-lacZ adults (Fig. 5D; Table 1). Later than 16 hours APF there was no significant increase in the number of founder cells for the TDT, and instead the formation of linear fibers began to take place.

These observations first suggested that the fibers of the TDT each develop according to the founder cell model. Second, the specification of TDT founders takes place over several hours at the end of the larval stage and early during pupal development. These findings are consistent with the observations of Rivlin et al. (Rivlin et al.,2000) who elegantly followed the appearance and migration of`muscle pioneers' for the TDT, which are probably coincident with the founder cells. In addition, similar data are presented by Atreya and Fernandes(Atreya and Fernandes, 2008),although their study did not focus specifically upon TDT development.

We studied next TDT development in cv mutants by creating a duf-lacZ cv⁴³ recombinant chromosome. We followed founder cell number in duf-lacZ controls and duf-lacZ cv⁴³ mutants during formation of the muscles (16 hours APF images are shown in Fig. 6A,C),and we studied the appearance of muscle fibers at 24 hour APF(Fig. 6B,D). Owing to the relatively large size of the muscles at this stage, the images presented do not show all of the founders cells or muscle fibers forming the muscles under study. Representative focal planes are shown for each sample.

Interestingly, we observed that the initial specification of TDT founders was normal in the cv mutants. By 8 hours APF, the number of founder cells in the mutants was 20.2±0.8, which was not significantly different from wild type at this timepoint. By 16 hours APF, however, there were slightly fewer TDT founder cells observed in the cv⁴³mutants, averaging 18.3±0.5 founder cells(Fig. 6C). This number was significantly less than wild-type at 16 hours APF (P<0.01,Student's t-test), and identified this timepoint as the first at which defects became apparent in the cv mutants. This value of∼18 founder cells is slightly larger than the number of large plus small fibers in the cv mutants, and it is possible that there is additional loss of founder cells prior to the initiation of fiber formation.

At 24 hours APF, the formation of individual fibers could be easily discerned in both wild-type (Fig. 6B) and cv⁴³ mutants(Fig. 6D). However, the number of fibers was clearly reduced in the mutants compared with controls, and the mutant fibers were also disorganized. Taken together, these studies indicated that in the cv mutants the process of founder cell specification was stalled, and that the reduction in founder cell numbers was the direct cause of the fiber number defect.

We next determined whether founder cell number and the formation of initial TDT fibers was altered when we interfered with TGFβ signaling via expression in the myoblasts of activated tkv or of Dad. Consistent with our observations of fiber number in the adults, we saw striking effects upon founder cell and fiber number in young mutant pupae. For expression of activated tkv, the number of founder cells was clearly increased at 16 hours APF. This is apparent in comparing Fig. 6E with 6A. The increase in founders was also reflected by an increase in the total number of fibers that were initially specified (Fig. 6F).

When we expressed Dad in the adult myoblasts, the opposite effect was observed: there were significantly fewer founder cells (only two or three can be discerned in Fig. 6G),although we also noted that there was an overall reduction in the total number of myoblasts, as determined by anti-MEF2 staining. At 24 hours APF, TDT fibers were difficult to identify as they were so few in number, and also showed severe hypoplasty (Fig. 6H).

Taken together, the studies shown in Fig. 6 define for the first time a crucial role for TGFβ signaling in the specification of TDT founder cells. These results will be further evaluated in the Discussion.

We next sought to identify the TGFβ ligand responsible for controlling TDT fiber number. Ligand encoded by the dpp gene is expressed at high levels in the imaginal discs and during pupal development(Masucci et al., 1990) and expression of lacZ from a dpp enhancer trap lies close to the TDT founder cells throughout early pupal development (data not shown). Furthermore, ligand encoded by glass bottom boat (gbb) is also expressed broadly in the imaginal discs(Khalsa et al., 1998). Thus,we focused upon these genes as potential contributors to TDT founder cell specification.

To determine whether Dpp or Gbb might be important for TDT founder specification, we tested whether double-heterozygotes for a ligand-encoding gene and for cv mutant alleles showed significantly fewer TDT fibers than did single heterozygotes for either gene mutation. The results of this analysis are presented in Fig. 7A,B. For Dpp, we observed, first, that the number of fibers in the dpp¹⁰⁶³⁸/+ heterozygotes was reduced slightly compared with that of cv⁴³/+, suggesting that haploinsufficiency for dpp might be an important factor in muscle specification. More importantly, the combination of both cv and dpp mutant alleles resulted in a fiber count further reduced relative to either single heterozygotes alone, and which was significant based upon Student's t-test (Fig. 7A). These studies identified Dpp as at least one of the TGFβ ligands that impacts TDT fiber number during pupal muscle development.

For Gbb, we also investigated whether haploinsufficiency would exacerbate the TDT phenotype in a cv⁴³/+ background. In this instance, we observed a more striking effect upon fiber number in the double heterozygotes (Fig. 7B). The double mutants displayed a highly significant reduction in fiber number when compared with controls. These data suggest that both Dpp and Gbb ligands might play important roles in specification of TDT muscle fibers.

To complement these data, we studied the pupal expression pattern of cv by in situ hybridization. We generated cv sense and antisense probes, and validated their functionality in embryos. As reported by Vilmos et al. (Vilmos et al.,2005), we observed specific expression of cv surrounding the embryonic tracheal pits (Fig. 7C), whereas no signal was obtained from a sense control probe(Fig. 7D).

In 16 hour APF pupal samples, we also observed expression of cv in the forming cuticle, close to where the TDT myoblasts had migrated(Fig. 7E). Although the signal was relatively weak, it was reproducibly greater than in sense probe control preparations that were stained in parallel(Fig. 7F). We are currently generating promoter-lacZ gene fusions of cv, in order to expand upon these data. These findings will be presented elsewhere.

Taken together, identification of the most likely TGFβ ligands that contribute to TDT founder cell specification, and presentation of the pupal cv expression pattern at 16 hours APF further strengthen our model for the specification of TDT founder cells via the TGFβ pathway.

The specification and action of founder cells is crucial to normal muscle development in Drosophila, where founder cells impart unique phenotypes for each muscle. Here, we have shown that the founder cell model also holds true for the adult tubular muscles such as the TDT, or jump,muscle. More importantly, we demonstrate that the specification of TDT founder cells arises from activation of the TGFβ pathway, a pathway that has not been previously implicated in controlling founder cell number at the adult stage of development. Our data demonstrate that it is most probably the activation of the TGFβ pathway within myoblasts that impacts founder cell specification, as in our overexpression experiments we used a Gal4driver which is active in the adult myoblasts.

The TGFβ pathway has roles in a range of developmental processes, and the addition of adult muscle development here extends a detailed list(reviewed by Kollias and McDermott,2008). In mammals, TGFβ can affect muscle development in several ways, including inhibition of differentiation (e.g. Yanagisawa et al., 2001) and inhibition of muscle regeneration in vivo(Cohn et al., 2007). Interestingly, the TGFβ molecule myostatin acts in mammalian muscle development to modulate the number of muscle fibers: in myostatin mutants in a variety of mammalian models, there is a profound increase in both muscle fiber number and muscle mass (reviewed by Kollias and McDermott, 2008). Despite these similarities, there is not yet sufficient evidence to suggest that Cv-mediated modulation of the Dpp pathway plays a similar role in Drosophila to that played by myostatin in mammals. This is at least partly because the effects of cv mutants that we observe here are restricted to a small subset of the adult musculature. A Drosophilagene named Myoglianin and encoding a molecule with strong sequence similarity to myostatin has been described, although no mutant alleles have been characterized (Lo and Frasch,1999).

Our genetic interaction data suggest roles for both Dpp and Gbb in TDT fiber specification. Each of these ligands function in wing vein development(e.g. Khalsa et al., 1998; Ralston and Blair, 2005);thus, a combinatorial role for them in founder cell specification would not be unprecedented. We also note that the Drosophila genome encodes a number of additional TGFβ-related molecules(O'Connor et al., 2006), and such molecules, in addition to Dpp and Gbb, might contribute to TDT founder cell specification.

Although manipulation of the TGFβ pathway showed clear effects upon the numbers of TDT founder cells, we also note that inhibition of the pathway,via UAS-Dad, caused a decrease in the number of total myoblasts as visualized by MEF2 staining. This observation suggests that, in addition to founder cell specification, the TGFβ pathway in adult myoblasts impacts either myoblast proliferation or myoblast survival. This observation is consistent with our finding that, in cv mutants, the number of founder cells reduces slightly as pupal development proceeds.

Specification of TDT founder cell number appears to be subject to significant variability in Drosophila. This observation contrasts sharply with many of the other muscles of the animal, which show relatively invariant fiber numbers. These include the skeletal body wall muscles of the larva (Bate, 1993), and the indirect flight muscles of the adult(Cripps and Olson, 1998; Farrell et al., 1996). Perhaps the variability in TDT fiber number is a reflection of the multi-step TGFβ pathway responsible for its specification, where variation in one or a few of the signaling components required for founder cell specification will ultimately impact the number of founder cells specified. Similar to the studies presented here, a genetic approach should be able to identify additional genes whose products function in this founder specification pathway and are responsible for the strain-specific differences in TDT fiber number that we have characterized. The identified genes might encode novel new members of the TGFβ signaling pathway active in myoblasts, or might identify mild alleles of known pathway members.

Variability in the organization of skeletal musculature is also documented in higher animals. In humans, the palmaris longus muscle of the forearm is commonly used as a source for tendon transplants. However, this muscle is absent, either unilaterally or bilaterally, in up to 15% of humans, depending strongly upon the ethnicity of the population being tested(Thompson et al., 1921; Thompson et al., 2001; Sebastin et al., 2005). This variability also has a strong heritable component(Thompson et al., 1921), and there is also significant variation in the size of this muscle(Sebastin et al., 2005). Similarly, the plantaris major also shows variable loss in humans(Vanderhooft, 1996). Our studies, which define how muscle fibers in the adult can be specified, point to potential mechanisms that might contribute to this variability in higher animals.