|Home | About | Journals | Submit | Contact Us | Français|
The body wall musculature of a Drosophila larva is composed of an intricate pattern of 30 segmentally repeated muscle fibers in each abdominal hemisegment. Each muscle fiber has unique spatial and behavioral characteristics that include its location, orientation, epidermal attachment, size and pattern of innervation. Many, if not all, of these properties are dictated by founder cells, which determine the muscle pattern and seed the fusion process. Myofibers are then derived from fusion between a specific founder cell and several fusion competent myoblasts (FCMs) fusing with as few as 3–5 FCMs in the small muscles on the most ventral side of the embryo and as many as 30 FCMs in the larger muscles on the dorsal side of the embryo. The focus of the present review is the formation of the larval muscles in the developing embryo, summarizing the major issues and players in this process. We have attempted to emphasize experimentally-validated details of the mechanism of myoblast fusion and distinguish these from the theoretically possible details that have not yet been confirmed experimentally. We also direct the interested reader to other recent reviews that discuss myoblast fusion in Drosophila, each with their own perspective on the process [1–4]. With apologies, we use gene nomenclature as specified by Flybase (http://flybase.org) but provide Table 1 with alternative names and references.
Both founder cells and FCMs arise from small clusters of cells within the mesoderm. Through Notch-mediated lateral inhibition, a single cell within each cluster is selected. In at least some cases, this cell undergoes an additional cell division to give rise to two founder cells or a founder cell and another cell type . The FCMs arise from the same clusters of mesodermal cells, deriving from those that remain after Notch-mediated founder cell selection. As in segregation of neuroblasts and epidermal cells in the ectoderm, all cells in a cluster become founder myoblasts in the absence of Notch . The specific identity of each founder cell is then established by expression of a specific combination of “muscle identity genes”, and the absence of a particular muscle identity gene results in the loss of specific muscle fibers . The combination of muscle identity genes also determines the ultimate morphology and size of the final muscle fiber, as evidenced in recent studies by the demonstration that misexpression of identity genes alters the extent of fusion and subsequent behavior of the resulting muscle fiber . It therefore seems apparent that founder cells comprise a diverse cell population in which individuals are distinct from each other even at their earliest appearance.
Genetically, FCMs comprise a pool of cells all of which require the Gli-family transcription factor Lameduck (Lmd) for their specification and/or differentiation [2, 3, 5, 6]. Like the founder cells described above, at least a subset of these cells undergo cell division (; Abmayr et al., unpublished), though the developmental consequences of this division on cell fate or differentiation remain to be addressed. In contrast to the unique identity of each founder cell, the FCMs appear to take on the identity of the founder cell with which they fuse [5, 6, 9]. While subsets of FCMs have been reported at the level of gene expression [8, 10, 11], studies have not yet shown that FCMs are entrained to fuse with specific founder cells or established that these differences in gene expression lead to functional differences.
It is generally assumed that at least a subset of the FCMs are not initially in direct contact with founder cells or developing myotubes and find them through a combination of migration and filopodia. This hypothesis is consistent with the spatial distribution of FCMs and founder cells in fixed tissue at both early and later stages of myogenesis . One might then predict that myoblast fusion will be incomplete in the absence of molecules associated with cell migration. Recent studies demonstrating that the FCMs of embryos mutant for the small GTPases rac1 and rac2, hem or scar, all genes known to be involved in migration in other systems as well as myoblast fusion (discussed below), lack polarized filopodia and Scar protein, reflecting defects in migration . However, other studies have demonstrated that the FCMs of hem mutant embryos are capable of migration . Thus, it remains to be shown whether the presence of isolated FCMs is a consequence of defects in migration or defects in stable adhesion.
Myoblast fusion initially occurs between a founder cell and an FCM, and is followed by multiple rounds of fusion between the resulting syncitia and additional FCMs. Correct recognition and adhesion between these two cell types are controlled by molecules in the immunoglobulin superfamily (IgSF). Founder cells must express Kirre, a founder cell-specific molecule, or Roughest (Rst), which is expressed on founder cells and at least a subset of FCMs [13, 14]. FCMs must express Sticks-and-Stones (Sns) , though its paralog Hibris (Hbs) can provide limited compensatory functionality [16, 17] (Table 1, Fig. 1). Drosophila S2 cells in culture have provided a valuable assay in which to examine the interaction behavior of cell adhesion molecules [17–19], and have been used to evaluate interactions between these IgSF proteins. These studies have demonstrated that neither Sns nor Hbs mediates homotypic interaction in trans , but that Kirre and Rst can direct interaction with other Kirre or Rst-expressing cells, respectively. Despite this latter interaction, however, cells expressing Kirre or Rst have a strong preference for those expressing Sns (or Hbs) , consistent with the interactions that occur in the embryo between founder cells and FCMs. Morphological data from fixed embryos supports the model that the founder cells act as attractants for the FCMs since FCMs extend filopodia -like structures that are directionally oriented toward the founder cell or developing myotube [2, 3].
Subsequent to the above events, studies have described cytoplasmic changes coincident with and/or directing myoblast fusion that appear to be linked to cell surface receptors. Ultrastructural analysis by transmission electron microscopy (TEM) revealed a series of events in which electron-dense vesicles approximately 40nm in diameter are present on the cytoplasmic sides of juxtaposed plasma membranes over an area of about 1μm2 . These vesicles were also found budding from the Golgi apparatus, indicating that they are of exocytic origin, and associated with a microtubule network . At the membrane, these “prefusion complexes” were proposed to give rise to “membrane plaques”, comprising electron-dense material that extended for 500nm along the apposed plasma membranes. It is not clear when pores form between cells, and cytoplasmic contents actually begin to mix. However, the adjoined plasma membranes ultimately appear to vesiculate along the site of contact, possibly forming sacs of membrane that enclose the previous extracellular space .
It is not clear whether the Sns and Kirre cytodomains direct recruitment of the electron-dense vesicles to points of cell-cell contact. However, structure/function analysis demonstrated that the Sns cytodomain is essential [18, 23]. Moreover, site-directed mutations in this domain revealed important phosphotyrosine and proline-rich regions with the potential to interact with SH2 and SH3-domain containing proteins . Such adapter proteins are common intermediates between membrane receptors and downstream effectors and, while neither functional studies nor specific interaction domains have confirmed its relevance, the SH2-SH3 adaptor protein Crk  has been shown to interact with Sns by immunoprecipitation using transfected S2R+ cells .
Studies similar to those involving Sns have recently been described for Kirre . These studies revealed progressive loss of binding to Rolling pebbles (Rols) and Schizo (see below) upon progressive deletion from the C-terminus. Like Sns, phosphorylation plays an important role in Kirre function but partial rescue of mutant embryos is observed in its absence. Intriguingly, 30% of the kirre/rst mutant embryos generate syncitia with up to 4 nuclei when rescued with Kirre lacking the entire intracellular domain . As mentioned above, signal transduction from Kirre in the founder cell appears to involve its interaction with the founder cell-specific Rols protein [19, 25, 26, 46]. Fusion in embryos mutant for rols is limited to single events that occur between the founder cell and FCM, and prefusion complexes fail to form . Rols translocates to the membrane upon Kirre/Sns adhesion in wild-type embryos and is aberrant in embryos lacking Kirre/Rst . The ankyrin repeats and TPR/coiled-coil domain in Rols are essential for this localization since deletion of either domain leads to severe fusion defects . It has been proposed that Rols provides a positive feedback loop through which the appropriate amount of Kirre is maintained at the founder cell surface, consistent with the observation that cell contacts are stabilized in its presence . Rols has also been suggested to mediate activation of the small GTPase Rac1 at points of cell contact via interaction with the nonconventional guanine nucleotide exchange factor (GEF), Myoblast City (Mbc) (see subsection c below). Consistent with these latter models, Rols colocalizes with actin foci at sites of adhesion .
Embryos lacking schizo also exhibit severe defects in myoblast fusion [28, 29]. Schizo, also termed Loner, has been shown to function as a GEF for the monomeric GTPase Arf51F (ARF6) and to interact with the cytodomain of Kirre  (Table 1, Fig. 1). Over-expression of dominant negative forms of Arf51F are associated with muscle defects [28, 29]. However embryos lacking arf51F maternally and zygotically develop to adulthood and have no overt mutant phenotype , suggesting that schizo functions through a different mechanism or that another gene functions redundantly with Arf51F. Schizo colocalizes with Kirre in S2 cells and can be immunoprecipitated through its interaction with Kirre [24, 29]. Notably, purified Kirre and Schizo do not interact, suggesting that their association might be indirect [24, 29]. While the schizo pathway appears to contribute to membrane localization of the monomeric GTPase Rac1 ; see subsection c below), it does not colocalize with actin at fusion sites .
Though their exact function is still unclear, two additional proteins that are important in myoblast fusion are encoded by the blown fuse (blow) locus ) and singles bar (sing) locus  (Fig. 1). Blow has a PH domain that binds to PIP3 and may recruit proteins to the appropriate cellular compartment and/or facilitate their interaction with components of signal transduction pathways. The MARVEL domain of Sing may implicate it in membrane apposition events, such as tight junction formation and vesicular transport. Genetic interaction studies have established that blow cooperates with hem . These studies have also suggested that hem acts downstream of blow since excess HEM partially rescues blow mutants. At the ultrastructural level, blow mutants have a normal number of prefusion complexes and normal electron-dense plaques . In contrast, sing mutants have twice the number of prefusion complex, suggesting that these complexes fail to resolve at the membrane.
Fusion is impaired upon mesodermal expression of constitutively active or dominant negative forms of the monomeric GTPase Rac1 , or in loss of function mutants for both rac1 and rac2 . The likely involvement of Rac1 in myoblast fusion was independently suggested by studies establishing that mbc, which is essential for myoblast fusion  encoded an ortholog of the mammalian GEF, Dock180 . Dock180, C. elegans Ced-5 and Drosophila Mbc comprise one subunit of a highly conserved bipartite GEF for Rac1, and function in concert with Ced12/Elmo . Mbc and Ced-12 appear to activate Rac1 since their overexpression in the embryonic musculature mimics that of active Rac1 and is suppressed by loss of one copy of Rac1 . Biochemical interaction studies and transgenic rescue experiments confirmed that the SH3 domain of Mbc mediates its interaction with Ced-12 and is essential for an Mbc transgene to rescue fusion in embryos, along with the Rac1 binding Docker domain and PIP3 binding DHR1 domain . In mammalian cells, the SH2-SH3 adaptor protein Crk helps to recruit Dock180 to the membrane, thereby leading to exchange of GDP for GTP and activation of Rac1 . However, Mbc does not require any interaction with Crk to function in myoblast fusion . Rather, Mbc may to be recruited to the cytodomain of Kirre in the founder cells through interaction with Rols, since the N-terminus of Mbc interacts directly with Rols in cultured S2 cells . Schizo, which also interacts with the Kirre cytodomain, promotes membrane localization of Rac1 . It remains to be determined whether Rac1 is recruited to the membrane independently or in concert with Mbc. The involvement of Schizo in localizing Mbc to KirreSns clusters at the cell surface is also appealing because Schizo is actually expressed in both founder cells and FCMs , and might therefore provide a mechanism for Mbc localization in the latter.
Despite the clear involvement of Mbc and Rac1 in myoblast fusion (Fig. 1), their exact role remains to be determined. EM studies have shown that the myoblasts of mbc mutant embryos lack prefusion complexes, suggesting a role in formation/transport of vesicles. They both colocalize with actin foci, perhaps reflecting an indirect role in recruitment of the electron-dense vesicles through their impact on the actin cytoskeleton. Rac1 has also been shown to synergize with the SCAR complex (discussed below, Fig. 1), and SCAR is absent from sites of fusion in Rac1 mutant embryos . Studies in mammals have suggested that membrane bound Rac1 may contribute to actin polymerization via recruitment/stabilization of the phosphorylated SCAR complex in conjunction with phospholipids like PIP3 . Finally, it remains a possibility that Rac1 and/or Mbc/Ced-12 could function independently of both the SCAR complex and actin polymerization.
Studies have revealed the importance of the actin nucleation promoting factors (NPFs) SCAR and WASp in myoblast fusion [12, 27, 42, 43] (Table 1, Fig. 1). In other systems, SCAR exists as part of a multiprotein complex, the core of which is formed by interaction between Abi and Nap1 (HEM/Kette in Drosophila). SCAR, in association with Brk1/HSPC300, interacts with Abi and Sra-1 interacts with Nap1/Kette/HEM. The remaining members of the complex control SCAR stability, regulate protein-protein interactions, and convey SCAR to areas of active actin assembly . SCAR promotes actin polymerization via interaction of its C-terminal domain with the Arp2/3 complex . Of the subunits of the complex, only Kette and SCAR have been studied extensively in the musculature and shown to play critical roles in Drosophila myoblast fusion [12, 27, 32] (Table. 1). Membrane localization of SCAR appears to be mediated by Kette as in mammals, , and SCAR protein is found in filopodia-like structures in FCMs in concert with Kirre near sites of fusion . At the ultrastructural level, FCMs establish contact with myotubes in hem mutant embryos  and, though the membranes do not appear to breakdown, electron-dense plaques have been observed. Interestingly, these plaques are two to three times longer than those observed in wild-type embryos, suggesting continued recruitment of electron-dense material in the absence of fusion.
As with the SCAR complex, our understanding of the WIP/WASp complex is influenced by studies in mammalian systems. WASp is auto-inhibitory, and is inactive in its native state due to interaction of the WASp C-terminal Arp2/3 binding domain with its GTPase binding domain (GBD)-binding domain. Activation occurs upon interaction of the WASp GBD domain with GTP-bound Cdc42, promoting access to the WASp C-terminus. The role of Cdc42 in activating WASp in the Drosophila musculature is unclear, since embryos zygotically null for cdc42 have no defects in myoblast fusion . Thus, mechanisms for relieving this negative autoinhibition in the embryonic musculature, including whether maternally provided cdc42 plays a role, remain to be clarified. In mammals, WIP regulates both activation of WASp by Cdc42 and its translocation to sites where its interaction with the Arp2/3 complex promotes actin polymerization . WIP is also important for WASp stability and protects it from degradation .
Membrane localization of WASp in mammals is facilitated by molecules such as PIP2, and via interaction with adaptor proteins like Crk and Nck in which the SH3-domain binds to the proline-rich region of WASp. Interestingly, the SH2-SH3 adaptor Nck (Dock in Drosophila) can recruit WASp to the plasma membrane of mammalian cells in a WIP-independent manner . In the Drosophila embryo, the WIP ortholog Verprolin (Vrp1; also termed D-WIP and Solitary) and WASp are essential for myoblast fusion [22, 43, 45]. Vrp1 lacking the C-terminal WASp binding domain fails to rescue embryos mutant for the solitary allele of Vrp1, suggesting that the Drosophila homolog also functions through WASp (Fig. 1). Vrp1 is expressed only in the FCMs [22, 42], where its localization to sites of fusion appears to require Sns [22, 42]. Indeed, biochemical assays have shown that Sns can interact with Vrp1 through Crk, leading to a model in which Sns mediates actin polymerization at sites of myoblast adhesion via a cascade involving Crk, Vrp1, WASp and Arp2/3 (Fig. 1). Vrp1 is present in the developing myotube following fusion with FCMs, where it becomes enriched near Kirre at points of cell-cell contact . Vrp1 and Kirre interact biochemicalIy in co-transfected S2 cells though, as with Sns, sequences mediating the interaction have yet to be identified. It remains to be shown whether Vrp1 and Kirre interact in myotubes, but the absence of Vrp1 in the founder cell suggests that Kirre must mediate the initial fusion event in its absence. Examination of mutant embryos suggests that Vrp1/WASp-directed activation of the Arp2/3 complex is required for pore formation and/or membrane vesiculation [12, 45], but these conclusions are controversial due to conflicting results using GFP diffusion and TEM with either conventional  or high pressure freezing .
On the basis of genetic interaction studies, both HEM/SCAR and Vrp1/WASp activate actin polymerization via association with Arp2/3 . SCAR appears to be required earlier than WASp, since fusion stops at the “membrane plaque” stage in both hem mutants and hem, wasp double mutants . GFP transfer from the myotube to the FCMs was not observed in hem, wasp double null mutants but was observed in embryos lacking wasp alone, suggesting the presence of pores in the latter . One complication of both single and double mutant phenotypes, however, is that these NPFs do not appear to function independently. Specifically, one copy of wasp in a hem null background partially rescues myoblast fusion , suggesting that HEM has a negative impact on WASp that may be independent of its interaction with SCAR. It has also been suggested that HEM may inhibit WASp early in the fusion process to allow formation of membrane plaques. In comparison with the mutant phenotypes of components of the HEM/SCAR and Vrp1/WASp complexes, pore formation has been observed in embryos mutant for their downstream target arp66B (arp3) but membrane breakdown is impaired . However, as with many other mutant phenotypes, the perdurance of maternal gene product may obscure the true loss of function phenotype.
As described above, Sns has been implicated in polymerization of F-actin by the Arp2/3 complex via interactions between Sns, Crk, Vrp1 and WASp . Consistent with this model, Sns and Kirre become organized into a ring-like structure following adhesion that has been termed the FuRMAS (fusion-restricted myogenic-adhesive structure) . This structure has F-actin at its center, and colocalizes with other proteins associated with fusion. Independent studies have revealed highly dynamic actin foci that appear and disappear at points of cell contact coincident with myoblast fusion . The origin(s) and function of these foci, and their relationship to the FuRMAS, is not clear. They exhibit significant perturbations in mutants defective for myoblast fusion, doubling in size in embryos lacking mbc, blow, scar or hem. These data are consistent with the proposed model that the associated proteins mediate actin depolymerization , but may result from continuing actin polymerization in one cell type due to defects in the other cell type that prevent fusion. Such a phenomenon has been reported to occur in embryos mutant for the solitary allele of Vrp1, in which F-actin foci are diminished in FCMs but gradually accumulate to an abnormally high level in the adjacent founder cell along the apposing membranes .
One issue that has complicated efforts to define the sequence of events, and gene hierarchy, associated with myoblast fusion is the redundancy observed in this process. Proteins with clear functional redundancy in Drosophila myoblast fusion include the small GTPases Rac1 and Rac2, which must both be deleted to observe defects in fusion . At the earliest stage of recognition between founder cells and FCMs, Kirre and Rst also appear to have completely redundant functions in the founder cell . In comparison to these examples, embryos lacking the FCM-specific sns gene exhibit severe defects in fusion , but some founder cells are capable of limited fusion in the absence of sns as a direct consequence of its paralog hbs . Finally, in addition to their intermolecular redundancy, the cell surface receptors Kirre and Sns exhibit intramolecular redundancy within their respective cytodomains ; . The potential remains for additional redundancy, with such similar proteins as Crk and Dock and the Arf proteins, just to name a few.
Recent studies also suggest the occurrence of functional redundancy between structurally dissimilar proteins involved in myoblast fusion . For example, the limited fusion that occurs in embryos individually lacking either schizo or rols does not occur in embryos lacking both genes. Similar genetic interactions have been described for vrp1, schizo double mutants, vrp1, rols double mutants, blow, schizo double mutants and blow, rols double mutants. In all of these examples, the myoblasts of embryos lacking individual genes are capable of limited fusion to form muscle precursors but embryos lacking these double mutant combinations exhibit no fusion. One issue complicating interpretation of some such results is the potential contribution of maternally provided gene products. Moreover, it remains to be determined whether these examples represent dissimilar proteins serving identical roles mechanistically or, rather, reduce the overall efficiency of fusion by impacting two different mechanistic steps. Each of these possibilities has implications for proposed models of myoblast fusion discussed below.
As a consequence of functional redundancy and perdurance of maternal mRNA or protein (discussed above), it is often difficult to interpret the relevance of genes that exhibit no loss-of-function phenotype or genetic interaction, yet encode proteins that interact biochemically with pathway components. It is also difficult to resolve the mechanistic implications of data that derive from different loss-of-function alleles or from different experimental approaches. Nevertheless, detailed ultrastructural analysis, genetic interactions, and morphological analysis in three dimensions over time have given rise to two distinct models for the process of myoblast fusion in Drosophila.
Renkawitz-Pohl and colleagues propose that the initial fusion event, namely that occurring between a founder cell and one or two FCMs to generate a muscle precursor, is mechanistically and morphologically distinct and requires a specific subset of proteins . Precursor formation is then followed by a “second step” in which the precursor undergoes additional rounds of fusion with neighboring FCMs to give rise to a mature myotube. This model is based in part on genetic loss-of-function phenotypes in which precursors can form despite the absence of a particular gene. Analysis of these mutant embryos by TEM has suggested perturbation at various stages ranging from pore formation to membrane vesiculation . In an alternative model, Baylies and colleagues argue that fusion proceeds in two “temporal phases” that are mechanistically identical, require the same set of gene products and differ only in the rate at which fusion occurs . The first wave of fusion in this model lasts for 3 h, during which time 9–27% of fusion events take place, and is followed by a second wave of 2.5 h, during which time the remaining 73–91% fusion events occur.
Consensus between these proposed models awaits future experiments that resolve the roles of maternally provided gene products, clarify any remaining functional redundancy, and identify molecules in which an earlier requirement obscures an involvement in myoblast fusion. Detailed structural analysis of fusing cells and functional analysis of proteins will also shed light on the mechanism of fusion. On one hand, the initial fusion involves two quite dissimilar cell types, an FCM and a founder cell. By comparison, subsequent fusion events involve an FCM and a multinucleate cell imbued with the components of both the FCM and founder cell. Thus, some fusion could utilize a mechanism in which, for example, the founder cell-specific protein Kirre interacts functionally with the FCM-specific protein Vrp1, as proposed . On the other hand, any benefit to developing two mechanisms for fusing myoblast membranes remains obscure. Rather, perhaps the initial recognition, migration and fusion between an FCM and a founder cell occurs slowly and inefficiently, but similar events between an FCM and a myotube occur more efficiently due to the protein composition of the latter. Clearly, the model most accurately representing the process of fusion requires further analysis, and is one of many issues yet to be addressed in the field of Drosophila myoblast fusion.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.