Drosophila paramyosin is important for myoblast fusion and essential for myofibril formation

The assembly of striated muscle myofibrils is a complex process in which numerous structural and regulatory proteins are assembled into basic contractile units, the sarcomeres (Obinata, 1993). The formation of sarcomeres requires the assembly of thin and thick filaments of appropriate length and their precise organization into higher order structures (Epstein and Fischman, 1991). The highly organized sarcomeres effectively translate the molecular movements of myosin motors into macroscopic contraction of muscle fibers.

The mechanism by which thick filaments attain precise regularity in striated muscle remains unknown. Although previous studies demonstrated that myosin possesses self-assembly ability (Huxley, 1963), myosin filaments formed in vitro lack important features of in vivo thick filaments. Accumulating evidence suggests that the assembly of myosin into thick filaments of distinct lengths, diameters, and flexural rigidities requires the presence of other proteins (Ziegler et al., 1996). In vertebrate muscles, several additional proteins are associated with thick filaments: C-protein, H-protein (mammals) or 86-kD protein (birds), M-protein, myomesin, M-creatine kinase, skelemin, adenosine monophosphate deaminase, and titin (Epstein and Fischman, 1991; Barral and Epstein, 1999). Analogues of some of these proteins exist in invertebrates. For instance, in Drosophila melanogaster, three members of the titin family have been identified: projectin (Ayme-Southgate et al., 1991), kettin (Hakeda et al., 2000; Kulke et al., 2001), and D-titin (Machado and Andrew, 2000; Zhang et al., 2000).

Some proteins are unique to thick filaments of invertebrate striated muscles. For instance, D. melanogaster possesses paramyosin (Vinós et al., 1991; Becker et al., 1992; Maroto et al., 1995), miniparamyosin (Becker et al., 1992; Maroto et al., 1995, 1996), myosin rod protein (Standiford et al., 1997), and flightin (Vigoreaux et al., 1993; Reedy et al., 2000). Caenorhabditis elegans has paramyosin (Mackenzie and Epstein, 1980; Kagawa et al., 1989) and α-, β-, and γ-filagenin (Liu et al., 1998, 2000). The diversity of thick filament components may account for the highly variable lengths and diameters of muscle thick filaments from different species.

Paramyosin and myosin are the most abundant invertebrate thick filament proteins. Paramyosin is present in all invertebrate muscles studied (Maroto et al., 1995). This protein is a rodlike molecule with high α-helical content in its long central domain. This domain is flanked by short nonhelical NH₂- and COOH-terminal regions. Two paramyosin monomers can dimerize into a coiled coil. Analysis of paramyosin and myosin heavy chain rod sequences revealed a remarkable pattern of alternating concentrations of charge associated with a 28-residue repeat (Cohen and Parry, 1998). Interactions between these segments of opposite charge are thought to play a major role in the assembly of both of these proteins into the thick filaments (McLachlan and Karn, 1982; Kagawa et al., 1989).

Drosophila paramyosin is a protein comprised of 879 amino acid residues with a molecular mass of ∼105 kD. The central 823 residues form an α helix; this is flanked by nonhelical domains of 32 NH₂-terminal and 24 COOH-terminal residues (Becker et al., 1992; Maroto et al., 1995). By using an alternative promoter and alternative RNA splicing, the paramyosin gene produces a transcript encoding miniparamyosin. Miniparamyosin shares its COOH-terminal region with paramyosin and has a unique NH₂-terminal domain of 114 amino acids (Becker et al., 1992; Maroto et al., 1995). Paramyosin is present in both embryonic and adult muscles. However, miniparamyosin is only present in adult musculature (Maroto et al., 1996).

Paramyosin is thought to facilitate thick filament assembly. Mutant analysis in C. elegans shows that thick filament length and diameter are affected by paramyosin content (Mackenzie and Epstein, 1980). Based on biochemical, genetic, and structural studies of C. elegans thick filaments, Epstein et al. (1995) proposed a thick filament structure model in C. elegans. In this model, paramyosin, together with the filagenins, forms the tubular thick filament core in which seven paramyosin subfilaments are interconnected by an internal sleeve of filagenins. Each paramyosin subfilament contains four strands of paramyosin coiled coils throughout its length. The motor protein myosin attaches on the surface of the core, yielding the functional thick filament (Epstein et al., 1995; Müller et al., 2001). Notably, no homologues of C. elegans filagenins have been identified in D. melanogaster and miniparamyosin and flightin do not exist in C. elegans. Thus, the process of thick filament assembly in different organisms and the molecular mechanism involved in the formation of different types of myofibrils in each organism remain to be clarified.

To investigate the function of D. melanogaster paramyosin, we used a genetic approach. We functionally disabled the paramyosin gene by mobilizing a P element in its promoter region. We observed that homozygous paramyosin mutants die as late embryos and that myofibril assembly is disrupted. Surprisingly, we found that paramyosin is also required for myoblast fusion. In the absence of paramyosin, myoblast fusion is sometimes blocked, resulting in the absence of some muscle fibers. We rescued the homozygous paramyosin mutant to adulthood using a paramyosin transgene, thereby proving that defects observed in myoblast fusion and myofibril assembly arise specifically from the absence of paramyosin. Antibody localization confirmed that paramyosin is present in myoblasts before fusion and is localized in discrete foci at the contact sites of fusing myoblasts. Our results demonstrate that paramyosin functions as a cytoplasmic protein in early embryonic development and is important for myoblast fusion before its assembly into thick filaments.

A P element insertion is present in the paramyosin promoter region of fly line prm^106-5 (Fig. 1). We identified this insertion line in screen for mutants with gross defects in neuromuscular function. To this end, we screened a collection of P element insertion mutants (Deak et al., 1997) for gross defects in the motility that normally occurs in late embryos in the few hours before hatching. This line lacked normal peristaltic body wall movements and appeared to have uncontracted muscles. Sequencing of an inverse PCR product showed an insertion at coordinates 8703958-8703965 of the 3L scaffold sequence, or at nucleotides 59-66 of cDNA clone GH14085, which encodes paramyosin (genome and clone data available from Berkeley Drosophila Genome Project, http://www.fruitfly.org/). The insertion is located 174 bp upstream of the translation start site. The mutation failed to complement deficiency Df(3L)hⁱ²², consistent with a mutation at cytogenetic location 66D10; 66E1–2, and with the location of the paramyosin gene at 66D12-66D14 (for cytological data, see FlyBase, http://flybase.bio.indiana.edu:82/). The insertion of a P element in this line reduces paramyosin expression to 70% of normal and homozygous mutants die at the first instar larval stage.

To make null or strongly hypomorphic alleles of the paramyosin gene, we mobilized the P element out of the locus by crossing male flies to female flies that produce P element transposase, hoping to mutate the flanking paramyosin gene by imprecise P element excision. From 140 crosses, 70 homozygous lethal lines were obtained. Most of these are embryonic lethal and others die at the first instar larval stage. 40 of the mutants can be rescued to adulthood with the wild-type paramyosin transgene pm (Mardahl-Dumesnil, 1998), indicating that they are paramyosin-specific mutants.

We used a paramyosin-specific antibody to determine paramyosin expression levels in the lines that could be rescued by the paramyosin transgene. We identified one strongly hypomorphic mutant in which the paramyosin expression level is <1% (Fig. 2 A). We refer to this line as prm¹. Homozygous prm¹ mutants die at the late embryo stage and are rescued to adulthood by the pm transgene.

To determine the genetic lesion in the paramyosin gene of prm¹, we isolated genomic DNA from homozygous prm¹ embryos and used it as a template for PCR analysis along with paramyosin-specific primers. We then cloned PCR products into a plasmid vector and sequenced them. Sequence data showed that a 4-kb fragment 5′ to the second nucleotide of the transcription start site of the paramyosin gene is removed by imprecise excision in this line. This region contains one MEF-2 site and 3 E-boxes, and is important for paramyosin expression in larval and adult stages (Arredondo et al., 2001). This result, together with the paramyosin expression and transgene rescue data, indicates that prm¹ is a strong hypomorphic or functional null allele of the paramyosin gene.

Because paramyosin and miniparamyosin are encoded at the same locus (Fig. 1), we wondered if the mutation of the paramyosin promoter region affects miniparamyosin expression. Expression of paramyosin and miniparamyosin in pm-transgene–rescued flies was analyzed by Western blotting using paramyosin and miniparamyosin antibodies. Results showed that both paramyosin and miniparamyosin expression in rescued flies are normal (Fig. 2 B). Because the transgene used in rescue does not contain the miniparamyosin-specific exon, all the miniparamyosin expressed in rescued adult flies must be encoded by the endogenous miniparamyosin locus. This indicates that miniparamyosin expression is not affected by the deletion in the paramyosin promoter region caused by imprecise P element excision.

The deletion in prm¹ uncovers a portion of the 284-bp CG13306, an open reading frame for which no cDNAs have been reported (Fig. 1). Deletion of CG13306 is not the cause of the mutant phenotypes we document because homozygous prm¹ mutants could be rescued with the pm transgene (see Results), which is truncated within the CG13306 open reading frame.

Homozygous prm¹ embryos display grossly normal morphology including normal segmentation and epidermal denticle belts (unpublished data), but they fail to hatch at late embryonic stage 17. Compared with normal embryos, manually hatched late stage 17 prm¹ homozygous embryos are flat and motionless, suggesting possible muscle defects.

We investigated the muscle development of homozygous prm¹ embryos by staining muscle fibers of late stage 16 embryos with muscle myosin heavy chain antibody (Fig. 3). At this stage, the process of body wall muscle development is complete and each of the abdominal hemisegments (A2-A7) has 30 muscles (Bate, 1993). Each of these muscles is unique in terms of its position, size, sites of attachment, and patterns of innervation. Every muscle has a full complement of nuclei, which remains unchanged until the end of larval life (Bate, 1990, 1993). In prm¹ mutant embryos, the regular muscle pattern is disrupted. At stage 16, over 95% of the mutant embryos have obviously detectable muscle losses and aberrantly shaped muscle fibers. Compared with wild-type counterparts, these aberrant fibers are shorter or thinner and have fewer nuclei (unpublished data). Muscle fiber absence in mutant embryos usually occurs in groups. It can take place at any position in the embryo, in all three major muscle subtypes: dorsal, lateral, and ventral. In embryos in which muscle development is severely disrupted, we observed an absence of several muscle groups in different segments of the same embryo. However, muscles that successfully developed are located in normal positions. We never observed cross-segment fibers or duplicated fibers. The mutant phenotype of muscle fiber losses in prm¹ is confirmed in homozygotes for the deficiency Df(3L)hⁱ²²ki, which covers the paramyosin gene (unpublished data). The muscle fiber phenotype in the deficiency line is slightly worse than in prm¹, reflecting leaky expression of paramyosin in prm¹.

The loss of muscle fibers in prm¹ embryos could arise from defects in the specification of myoblasts, from failure of founder cells and fusion-competent cells to fuse, or perhaps from defects in the attachment of muscle fibers to the epidermis. To assess the specification of founder cells, we created a fly strain in which the enhancer trap rP298-lacZ (which marks all founders of the embryonic musculature; Nose et al., 1998) is expressed in prm¹. Staining of rP298-lacZ; prm¹ embryos with β-galactosidase antibody before myoblast fusion revealed that founder cells are normally specified in prm¹ (Fig. 4, compare A with D). We examined the specification of all myoblasts by staining early embryos with DMEF2 antibody, which marks all somatic, visceral, and cardiac myoblasts (Lilly et al., 1995). The number of DMEF2 expressing cells in prm¹ mutant embryos (Fig. 4 E) was comparable to that in wild-type embryos (Fig. 4 B). These data suggest that the initial differentiation of myoblasts in prm¹ is normal. We also stained rP298-lacZ; prm¹ embryos with β-galactosidase antibody and muscle myosin antibody to visualize founder cells and newly formed muscle fibers. We observed that unfused founder cells are present at the location of missing muscle fibers in prm¹ (Fig. 4 F). Finally, the expression pattern of paramyosin suggests that it is not likely to regulate fiber attachment to the epidermis. Expression in epidermis declines dramatically during myoblast fusion and eventually disappears after fiber formation; furthermore, paramyosin is not enriched at the ends of newly formed muscle fibers (see next section). These observations, in addition to the localization of paramyosin to sites of myoblast fusion (see next section), indicate that the loss of muscle fibers in prm¹ mutant is due to defects in myoblast fusion.

The involvement of paramyosin in myoblast fusion indicates that it has functions in addition to serving as a thick filament protein in myofibrils. Based on its structural properties, paramyosin might be a component of the cytoskeleton before myofibrillogenesis. We stained wild-type Drosophila embryos of different stages with a paramyosin antibody and observed that paramyosin is localized in the cytoplasm at very early stages (Fig. 5, A and B). Before gastrulation occurs, paramyosin antibody stains the apical surface, lateral interface, and the basal opening of the syncytial blastoderm. Note that an embryo that is homozygous for a paramyosin mutation would not be expected to display a mutant phenotype at the blastoderm stage, since transcripts expressed at this stage are maternally inherited. During gastrulation, paramyosin expression levels decline, but weak staining is present in both the mesoderm and ectoderm (Fig. 5, C and D). At stage 12, paramyosin expression increases, with higher expression levels in somatic mesodermal cells (Fig. 5 E). By stage 14, when most myoblast fusions occur (Campos-Ortega and Hartenstein, 1985), paramyosin expression increases dramatically in fusing muscle fibers of somatic and pharyngeal muscles (Fig. 5 F). Later in development, paramyosin disappears in the epidermis and is only detected in mesodermal derivatives, including body wall musculature, visceral musculature, and the heart. Previous Western blotting analysis (Vinós et al., 1991) is consistent with our antibody staining data (Fig. 5). Vinós et al. (1991) reported that paramyosin is present in relatively high amounts in the cytoskeletal pellet of mature oocytes. Its levels are very low or undetectable during gastrulation, and increase progressively during middle and late embryogenesis. The presence of paramyosin in oocytes and in embryos before myofibrillogenesis strongly suggests a role as a general cytoskeleton protein.

Because paramyosin has very high affinity to muscle myosin in myofibrils and in in vitro analysis (Epstein et al., 1976; Ziegler et al., 1996), it is possible that paramyosin also interacts with myosin at early stages of development. We observed that both muscle myosin and nonmuscle myosin II are present in myoblasts, but their localization patterns are different (Fig. 6). Although both myosins are localized in the cytoplasm of unfused myoblasts or newly formed myotubes, nonmuscle myosin II is more enriched in discrete foci at the interface of fusing myoblasts in a manner similar to paramyosin. The involvement of nonmuscle myosin II in muscle fiber formation and later in myofibrillogenesis (Bloor and Kiehart, 2001) further suggests that these two proteins interact at this stage.

Because previous studies have shown that paramyosin is a thick filament protein, the prm¹ paramyosin mutation should affect thick filament assembly and myofibril organization. We studied the subcellular distribution patterns of muscle proteins in somatic muscles of mutant and rescued embryos by staining the body wall muscle fibers with muscle myosin antibody and phalloidin to label myosin thick filaments and actin thin filaments, respectively. We observed that, as in the wild-type embryo (unpublished data), rescued embryo muscle fibers have a regular cross-striated myofibril banding pattern with myosin and actin labeling. However, the banding patterns of myosin and actin in the prm¹ mutant embryonic body wall muscles is completely disrupted (Fig. 7). This indicates that a severe reduction in paramyosin disrupts myofibril protein assembly in sarcomeres.

We compared the ultrastructure of mutant and wild-type embryonic body wall muscles by transmission electron microscopy. Well-organized sarcomeres are formed in somatic muscle fibers of wild-type embryos in stage 17. In longitudinal sections, muscle fibers display parallel arrays of thick and thin myofilaments (Fig. 8, A and B). At the end of thin filaments, Z bodies align to form Z bands (Fig. 8, A and B) that mark myofibrils into regularly separated sarcomeres. In contrast, sarcomeres are severely disrupted in prm¹ mutant embryos of the same age. Z bodies are poorly organized (Fig. 8 C). The lengths of myofilaments are greatly reduced and bundles of myofilaments associated with these Z bodies are not parallel (Fig. 8 C). Similar defects were seen in other paramyosin mutants (Mardahl-Dumesnil, 1998). The observed mutant phenotype is consistent with that of a myosin heavy chain mutant, in which organization of Z bodies in embryo somatic muscles is disrupted (O'Donnell and Bernstein, 1988). The ultrastructural defects at the electron microscopic level explain the resulting loss of myofibril striation at the light microscopic level (Fig. 7).

Transverse sections of wild-type and prm¹ mutant embryo myofibrils are shown in Fig. 8 (D and E). Compared with wild-type embryos (Fig. 8 D), prm¹ mutant embryos have reduced thick filament numbers (Fig. 8 E). However, thin filament number in the same area is relatively unaffected. This leads to some areas of myofibrils completely lacking thick filaments (Fig. 8 E, arrows). Mutation of paramyosin also disrupted thick filament structure. Thick filaments from wild-type embryo body wall muscles are hollow (Fig. 8 D, arrowheads). Each thick filament is circular in cross sections, consisting of several dense particles in the periphery. Compared with their wild-type counterparts, thick filaments of prm¹ embryo body wall muscles are solid. Dense particles in the periphery collapse into the center, forming filled thick filaments (Fig. 8 E, arrowheads). The diameters of thick filaments are also not even. We conclude that the paramyosin mutation reduces both the number and integrity of thick filaments.

To verify that the failure of myoblast fusion and aberrant myofibril assembly of somatic embryo body wall muscles in prm¹ mutant embryos are indeed caused by the paramyosin mutation, we analyzed the muscle development and myofibril assembly of rescued embryos. The phenotypes observed in prm¹ mutant embryos are restored to normal in rescued organisms. Rescued embryos have a normal embryonic body wall muscle pattern (Fig. 3, A–C). Each missing muscle fiber or group of fibers is restored and no duplicated or aberrant muscles are observed. The embryo body wall muscles of rescued embryos have a regular striation pattern of sarcomeres (Fig. 7).

As mentioned at the beginning of Results, the paramyosin transgene rescues the embryonic lethality of prm¹. Homozygous prm¹ survives to adulthood in the presence of the paramyosin transgene. Indirect flight muscles from rescued adult prm¹ flies have normal myofibril structure (Fig. 9). In longitudinal sections, myofibrils have regular Z discs and M lines and sarcomere length is unchanged compared with wild type (Fig. 9, A and B). In transverse sections, thick and thin filaments are packed in a hexagonal manner and all the thick filaments have hollow centers (Fig. 9, C–F). This suggests that paramyosin expressed from the transgene is correctly assembled into thick filaments and that this further restored the assembly of myofibrils during myofibrillogenesis.

Together, the data presented here indicate that paramyosin has dual functions at different stages of Drosophila myogenesis. Initially, paramyosin functions as a cytoplasmic protein that plays an important role in the processes of myoblast fusion. After myoblast fusion, paramyosin assists the assembly of muscle myosin molecules into well-organized thick filaments, which are capable of assembling into myofibrillar arrays.

Myogenesis is a process in which cells acquire numerous characteristics. Morphological changes that accompany myogenesis include fusions of myoblasts into multinucleate myotubes and the formation of the contractile apparatus (Bour et al., 2000). Myoblast fusion consists of cell differentiation, cell–cell recognition, alignment, and membrane fusion (Doberstein et al., 1997). Myoblast fusion precedes contractile apparatus formation. Elaboration of the contractile apparatus requires precise assembly of numerous structural and regulatory proteins into sarcomeres. In developing embryos of Drosophila, no myofibrils are assembled in unfused myoblasts and these myoblasts eventually are degraded and cleared by macrophages (Rushton et al., 1995). In this paper, we showed that paramyosin, a component of thick filaments, is not only involved in thick filament formation and myofibril assembly but also is important for myoblast fusion.

The events surrounding myoblast fusion in D. melanogaster have been studied extensively (Paululat et al., 1999; Taylor, 2002). Fusion always takes place between founders and fusion-competent myoblasts. An individual founder fuses with fusion-competent myoblasts to form muscle precursors with two or three nuclei. These precursors enlarge by recruiting and fusing with additional fusion-competent myoblasts to form multinucleate myotubes.

The most apparent mesodermal defect in embryos mutant for the paramyosin gene is random loss of muscle fibers or muscle fibers in aberrant shapes (Fig. 3). However, the defect in muscle development is not caused by reduction in myoblast number in early development (Fig. 4). Unfused myoblasts are present in the locations of missing muscles before being degraded (Fig. 4 F), suggesting a role of paramyosin in the progression of cells from myoblasts to myotubes.

The structural features of paramyosin and its cytoplasmic location seem inconsistent with a role in cell adhesion, instead suggesting paramyosin functions as a cytoskeleton protein in early embryonic stages before myofibril formation. Paramyosin molecules form rodlike coiled-coil α-helical dimers that lack domains reminiscent of cell adhesion molecules. Furthermore, paramyosin is localized beneath the cell membrane of mesodermal cells and epidermal cells (Fig. 6). Paramyosin might attach the cytoskeleton to membrane adhesion sites, akin to the role of coiled-coil intermediate filaments in other organisms. This function might arise due to the lack of cytoplasmic intermediate filament proteins in Drosophila (Goldstein and Gunawardena, 2000).

As is the case for paramyosin, there are several other cytoplasmic proteins that are important for Drosophila myoblast fusion (Paululat et al., 1999; Taylor, 2002). These include the myofibrillar protein D-titin (Zhang et al., 2000), Blown fuse (Doberstein et al., 1997), and Myoblast city (Rushton et al., 1995; Doberstein et al., 1997; Erickson et al., 1997), found in both founders and fusion-competent myoblasts, as well as Rolling pebbles (Rols)/Antisocial (Ants) present only in founders (Chen and Olson, 2001; Menon and Chia, 2001). It is suggested that Rols/Ants functions as an intracellular adaptor protein that relays signals from the immunoglobulin family membrane protein Dumbfounded (Ruiz-Gómez et al., 2000) to the cytoskeleton during myoblast fusion (Chen and Olson, 2001; Menon and Chia, 2001).

The presence of the myofibrillar protein D-titin in developing myoblasts implicates cytoskeletal components in the process of myoblast fusion (Zhang et al., 2000). Defects in myoblast fusion similar to those in prm¹ occur in Drosophila D-titin mutants. Compared with wild-type muscles, embryo body wall muscles of D-titin mutants are smaller and thinner. Occasionally, a few missing muscles are observed in severe mutant alleles (Zhang et al., 2000). The motor protein nonmuscle myosin II is also important for muscle fiber formation in Drosophila embryos (Bloor and Kiehart, 2001). In zipper mutants of nonmuscle myosin II, some ventral muscles are deleted. The involvement of D-titin and possible involvement of nonmuscle myosin II in myoblast fusion support the hypothesis that contractile elements play a role in this process. Later, we discuss a model in which paramyosin serves as part of nonmuscle myosin minifilaments that interact with the actin cytoskeleton to regulate cortical cytoskeleton dynamics and promote myoblast fusion.

Accumulating evidence suggests that the assembly of muscle myosin into thick filaments requires the presence of other proteins (Barral and Epstein, 1999). Biochemical, cell biological, and genetic studies in invertebrates support this hypothesis. In D. melanogaster, paramyosin, miniparamyosin, myosin rod protein, and flightin are all present in the A band region of some muscle sarcomeres (Vigoreaux et al., 1993; Maroto et al., 1996; Standiford et al., 1997). Flightin knockout flies show increased thick filament length in indirect flight muscles, suggesting that flightin regulates thick filament assembly (Reedy et al., 2000). In C. elegans, the thick filament protein UNC-45 acts as a myosin chaperone (Barral et al., 2002) and unc-45 mutations disrupt thick filament assembly (Barral et al., 1998). C. elegans' thick filament cores contain paramyosin and three filagenins (Liu et al., 1998). Mutation of paramyosin in C. elegans greatly reduces the length of thick filaments and disrupts the distribution of myosin isoforms (Mackenzie and Epstein, 1980; Epstein et al., 1986).

Consistent with the C. elegans studies, we found that the body wall muscles of homozygous prm¹ embryos have a marked reduction in the number of observable thick filaments, resulting in areas containing only thin filaments. Two factors might contribute to the formation of thick filaments in this line. The paramyosin from leaky expression of prm¹ may interact with other unidentified thick filament core proteins, nucleating myosin molecules into these abnormal thick filaments. Alternatively, myosin may directly interact with these unidentified proteins to form abnormal thick filaments. We conclude that paramyosin is important for the production of an adequate number of morphologically normal thick filaments.

Another specific characteristic of prm¹ is that the striated pattern of embryonic body wall muscles is disrupted. Although thin filaments assemble in the absence of paramyosin, they could not organize into regular sarcomeric patterns. The localization of paramyosin in thick filaments does not support a role as a scaffold in organizing myofibrils. Thus, abnormal interaction of thin and thick filaments might be the direct cause of this phenotype. Mutations in other Drosophila muscle contractile proteins also disrupt sarcomere organization, including actin (Sparrow et al., 1991), troponin-T (Fyrberg et al., 1990), troponin-I (Beall and Fyrberg, 1991), myosin heavy chain (O'Donnell and Bernstein, 1988; Beall et al., 1989), and α-actinin (Roulier et al., 1992). This indicates that correct interaction between thin and thick filaments is required for myofibril formation.

The formation of myofibrils has been extensively studied in embryonic vertebrate cardiac and skeletal muscle cells and several models have been proposed. In cultured cardiomyocytes, Sanger and colleagues (Dabiri et al., 1997) suggested that premyofibrils, characterized by banded patterns of α-actinin–rich Z-bodies and nonmuscle myosin IIB, form at the edges of spreading cardiomyocytes and develop into mature myofibrils. During the transition from premyofibrils to myofibrils, there is an exchange of nonmuscle myosin IIB filaments for muscle myosin II filaments and a growth and fusion of Z-bodies into Z-bands. Evidence from developing chicken embryo hearts (Ehler et al., 1999) did not substantiate the presence of premyofibrils in vivo. An alternative model, based on observations of cultured cardiomyocytes, proposes that spatially separate complexes of actin filaments and Z-bands (I–Z–I brushes) and groups of myosin thick filaments assemble independently of one another; they become spliced together by titin filaments, and then inserted at the ends of fully formed myofibrils (Ojima et al., 1999). Although these models differ in detailed steps of myofibril formation, they agree that myofibrillogenesis starts with membrane association and that thick and thin filaments form independently. According to these previous studies and our evidence of paramyosin's dual roles in myoblasts and myofibrils, we propose a model of paramyosin in Drosophila muscle differentiation.

The transition of paramyosin from a cytoplasmic location to myofibrils can be divided into several steps in our model. First, before myoblast fusion, paramyosin binds to nonmuscle myosin filaments. These paramyosin-nucleated nonmuscle myosin filaments interact with filamentous actin in the cytoskeleton of myoblasts to regulate cell shape change and membrane penetration during cell fusion. Adaptor proteins, such as Rols/Ants, help keep these paramyosin nonmuscle myosin filaments in discrete foci at this stage. Second, after myoblast fusion, nonmuscle myosin is replaced by muscle myosin. Paramyosin, together with other thick filament core proteins, nucleate muscle myosin filaments into muscle thick filament precursors. Third, actin filaments and titin of I–Z–I brushes incorporate precursor thick filaments into A-bands of nascent myofibrils, as in cultured cardiomyocytes (Dabiri et al., 1997; Holtzer et al., 1997). Filaments may be first aligned out-of-register, and then transported into position. Finally, both thin and thick filaments elongate and myofibril diameter increases by peripheral addition of myofilaments.

Overall, our work revealed a role of paramyosin in thick filament formation and myofibrillogenesis. We also discovered that paramyosin has an unexpected function in myoblast fusion. Antibody localization and analysis of contractile protein mutants available in the Drosophila system will permit testing of our model of paramyosin function in muscle differentiation.

Yellow white (yw) flies were used as a wild-type control line. Fly line yw; 0106/05 [P, w⁺]/TM2, Ubx Sb (referred to as prm^106-5 hereafter) was recovered from an enhancer-trapping screen for genes essential for neuromuscular function (Results). Line prm^106-5 was obtained from a third chromosome P element insert collection (Deak et al., 1997) that had been balanced with a TM6C, Sb, Hu chromosome. In this line, a P element is inserted in the promoter region of the paramyosin gene, 174 bp upstream of the translation start site. Line rP298 is a P[lacZ, ry⁺] insertion line on the X chromosome (Nose et al., 1998). We constructed fly line yw; MKRS, Sb/TM3, y⁺ Ser. The deficiency line yw; Df(3L)hⁱ²²ki /TM3, y⁺ Ser and transposase-producing line w; Sp/CyO; Δ2–3, Sb/TM2, Ubx were obtained from the Bloomington Stock Center.

Embryos were collected for 1 h and aged to 22 h at 25°C, by which stage most wild-type embryos (including TM6C homozygotes) had hatched (Broadie and Bate, 1993). Unhatched embryos were dechorionated and observed for spontaneous peristaltic larval movement and tested for movement in response to touching by forceps. Mutant lines whose development appeared normal, but which were defective for larval movement, were rescreened to confirm the phenotype.

To isolate new alleles of the paramyosin gene by imprecise P element excision, individual male yw; prm^106-5/TM2, Ubx Sb flies were mated to virgin female w; Sp/CyO; Δ2–3, Sb/TM2, Ubx flies, which produce transposase. The white⁺ gene in the P element was used as a marker for P element mobilization. Individual male F1 yw; prm^106-5/Δ2–3, Sb flies with blotchy eyes were selected and then mated to yw; MKRS, Sb/TM3, y⁺ Ser virgin female flies. Sibling white-eyed F2 yw; prm^106-5/TM3, y⁺ Ser flies from each cross were mated to each other to make a stable stock for each mutant line. Lines in which yw; prm^106-5/prm^106-5 flies could be recovered were probably progeny of flies with a precise P element excision and were discarded. Homozygous lethal lines likely contain new mutant alleles of the paramyosin gene and were kept for further analysis.

To identify paramyosin mutants, a paramyosin transgene P[w⁺, PM] (referred to as pm in the text) (Mardahl-Dumesnil, 1998) was crossed into these mutant backgrounds. Mutants that could be rescued to adulthood were considered to be paramyosin-specific mutants. Paramyosin-specific mutants, which are embryonic lethal, were used for paramyosin expression analysis. We used mouth hook color as a marker to differentiate homozygous mutant embryos from heterozygous embryos. Homozygous mutant embryos have yellow mouth hooks because they lack the balancer chromosome, which carries a black mouth hook marker (y⁺). Paramyosin expression levels were analyzed by Western blotting using antiparamyosin antibody (Maroto et al., 1995).

Protein electrophoresis of dechorinated embryos or dissected adult upper thoraces and Western blotting with antiparamyosin and antiminiparamyosin antibodies were performed as described by Maroto et al. (1995). Antibody detection was performed using the SuperSignal system (Pierce Chemical Co.). The relative amounts of proteins were determined by scanning densitometry of bands on films (Expression 636 scanner [Epson]; NIH Image 1.61).

The insertion site of the P element in line prm^106–5 was determined by inverse PCR and cycle sequencing, essentially as described by the Berkeley Drosophila Genome Project (http://www.fruitfly.org/p_disrupt/inverse_pcr.html). To determine the extent of the deletion resulting from imprecise P element excision, genomic DNA was extracted from prm¹ embryos. 100 ng of DNA were used as template for PCR with the Roche Expand Long Template PCR system (Roche Biosciences). Primers used were: forward primer, 5′GTTTTTAGTTCTCGGTTCTTTTTGTTGGAC3′; and reverse primer, 5′AAAAAGTTGCCGATTGCCACAAAGGCCACG3′. PCR products were cloned and sequenced.

Antibody staining of whole-mount embryos and of dissected embryos for confocal microscopy was performed as described (Bate, 1990; Rushton et al., 1995). Antibodies used are: anti–muscle myosin and anti–nonmuscle myosin II antibodies (Kiehart and Feghali, 1986); antiparamyosin and antiminiparamyosin antibodies (Maroto et al., 1995); anti-DMEF2 (Lilly et al., 1995); mouse monoclonal anti–β-galactosidase antibody (Promega); Cy5-labeled goat anti–rabbit Ig (Amersham Biosciences); and goat anti–mouse Ig Alexa Fluor 488 conjugate (Molecular Probes). FITC-phalloidin was purchased from Molecular Probes.

Electron microscopy of adult indirect flight muscles and embryos was done following previously published procedures (O'Donnell and Bernstein, 1988).

We thank Dr. Magarita Cervera for providing the paramyosin and miniparamyosin antibodies, Dr. Daniel Kiehart for providing the muscle myosin and non-muscle myosin II antibodies, Dr. Bruce Paterson for DMEF2 antibody, and Dr. Susan Abmayr for Nautilus antibody. We are grateful to Dr. Steve Barlow for advice in confocal microscopy and electron microscopy, which were performed in the San Diego State University Electron Microscopy Facility. We thank Dr. Richard Cripps and William Kronert for constructing the paramyosin transgene, and Virginia Guest for technical assistance. We are grateful to anonymous reviewers for their insightful suggestions.

This research was supported by an American Heart Association, Western States Affiliate predoctoral fellowship to H. Liu, grants MCB9604546 (National Science Foundation) and AR43396 (National Institutes of Health) to S.I. Bernstein, and grant 048476 (Wellcome Trust) to C.J. O'Kane.