|Home | About | Journals | Submit | Contact Us | Français|
Muscle formation and repair depends critically on the fusion of myoblasts. Despite the importance of this process, little is known about the cellular and molecular mechanisms regulating fusion. Forward genetic screens in Drosophila melanogaster have uncovered genes that, when mutated, prevent myoblast fusion. Analyses of these gene products have indicated that the actin cytoskeleton and its regulation play a central role in the fusion process. In this review, we discuss recent advances in the field, including new imaging approaches to analyze fusion as well as a description of novel genes required for fusion. In particular, we highlight what has been learned about the requirement of a specific actin structure at the site of fusion. We also place these findings from Drosophila within the context of myoblast fusion in vertebrates.
The body wall musculature of the developing Drosophila embryo consists of 30 muscles in each abdominal hemisegment (1–3) (Figure 1). Each muscle is formed by the fusion of two cell types: a founder cell (FC) that seeds the formation of a specific muscle and fusion-competent myoblasts (FCMs) that provide mass. Each FC has a unique identity, characterized by the expression of transcription factors such as Slouch (also known as S59), Even-skipped, Krüppel, Vestigial, Apterous and Nautilus (4–6). The specific combination of transcriptional regulators expressed in each FC determines the unique morphology of the final muscle and controls the number of fusion events, muscle shape and orientation, innervation and sites of attachment to the overlaying epidermis. In contrast, FCMs are thought to be a naïve population of myoblasts. Prior to fusion, FCMs express the transcriptional regulator Lame duck (Lmd) (7–9). Upon fusion to an FC, the FCM nucleus turns off lmd transcription and begins to express the combination of transcriptional regulators of the FC (6,10). Depending on the particular muscle, fusion will occur between 2 and 25 times (2).
Recently, the period in which all embryonic myoblast fusions happen has been determined. Myoblast fusion occurs over a 5.5-h period, during stages 12–15 [7.5–13 h after egg lay (AEL)], although not all FCs begin fusion at the same time (2,10). It was thought that smaller muscles, which arise from fewer fusion events, stop fusing sooner than larger muscles. However, recent analysis of individual muscles has shown that fusion for smaller muscles is completed at stage 15, similar to larger muscles (11). Moreover, this analysis has indicated that there are two temporal steps of fusion. During the first 3 h of fusion (stage 12–13, 7.5–10.5 h AEL), 9–27% of fusion events occur, depending on the muscle analyzed, while the remaining 73–91% of fusion events occur in the final 2.5 h of the process (stage 14–15, 10.5–13 h AEL) (11).
Genetic studies have uncovered a number of gene products required for fusion (12–14) (Table 1). These gene products fall into several classes based on their predicted functions and localization. Four transmembrane proteins containing immunoglobulin domains function in the recognition and adhesion of myoblasts. Dumbfounded/Kirre (Duf) and Roughest/Irre (Rst) function in the FC, while Sticks and Stones (Sns) and Hibris (Hbs) function in the FCM (15–19). Duf and Rst bind directly to Sns, and these proteins are the primary mediators of myoblast adhesion. Duf and Rst perform redundant functions, and removal of both genes leads to a complete fusion defect (15). Removal of sns also causes a complete fusion defect (17). Removal of hbs causes a mild fusion phenotype, and it has been proposed that Hbs regulates Sns during particular stages of fusion (18,19).
Downstream of adhesion in the FC is Rolling Pebbles/Antisocial (Rols), a multi-domain adaptor protein (20–22). Rols is thought to regulate adhesion and fusion by promoting Duf recycling to the membrane following each fusion cycle (23). Additionally, Rols is thought to physically link Duf to components and regulators of the cytoskeleton, namely Titin and Myoblast City (Mbc), the Drosophila Dock180 homologue (20,22). Removal of rols, titin, mbc or elmo, an mbc interacting partner, leads to a severe fusion defect (20–22,24–28).
Genetic analysis has identified an essential role of the actin cytoskeleton and its regulators in fusion. Mbc and its homologue Dock180 are guanine nucleotide exchange factors (GEF) that play an established role in regulating the small GTPase Rac (29–31). An additional protein, Loner/Schizo, is also a GEF that acts on the GTPase ADP ribosylation factor (ARF)6 and appears to be required for proper Rac localization during fusion (32). Recently, analysis of a null allele of ARF6 revealed no effect on myoblast fusion, suggesting that Arf6 either does not play a role during fusion or may not be the only target of Loner during myoblast fusion (33). Rac, among its many proposed functions, affects the cytoskeleton through regulation of a multi-protein complex consisting of Kette/Nap1, Sra-1/CYFIP/Pir121 and HSPC300 bound to SCAR/WAVE (34,35). This complex appears important in stabilizing and localizing SCAR/WAVE during fusion (36–39). SCAR/WAVE in turn is a potent activator of Arp2/3, an actin-related protein that promotes the polymerization and branching of actin (40–42). Removal of loner, Rac, kette or SCAR leads to a strong fusion defect (32,39,43–45). Mutation of Arp3 leads to a partial fusion defect, although reagents are not available for a complete loss-of-function analysis (39). The WASP protein presents an alternative mechanism of activating Arp2/3 (40,42). Mutations in WASP and its regulator Wasp-interacting protein/Solitary (D-WIP/Sltr) lead to strong fusion defects, further highlighting the critical nature of the Arp2/3 pathway to the fusion process (46–48).
Two additional proteins have less clear roles in the fusion process. Blow is a PH domain-containing protein, suggesting that lipid binding and/or membrane localization are important to its function (12). blow also genetically interacts with kette, suggesting that Blow may also be involved in regulation of the cytoskeleton (45). The Singles Bar (Sing) protein contains a MARVEL domain, which is thought to be important for membrane apposition events such as tight junction formation and vesicular transport (49). Removal of blow or sing leads to a strong fusion defect (49,50).
Several approaches have been taken to better understand the cell biological processes required for fusion, including confocal imaging approaches focused on the actin cytoskeleton and transmission electron microscopy (TEM). Actin-based cytoskeletal rearrangements have been defined using improved methodologies in both fixed and live preparations (39). This approach has allowed identification of an F-actin focus (also known as a fusion-restricted myogenic-adhesive structure or FURMAS) at the site of myoblast adhesion that occurs just prior to fusion (39,51) (Figure 2). Many of the proteins implicated in myoblast fusion localize to the actin focus, suggesting that this is the site of their activity during fusion. Crucially, live imaging of the actin cytoskeleton using a green fluorescent protein::actin protein has revealed that actin foci mark the actual site of subsequent myoblast fusion, something that had not been previously identified in any system (39). All fusion events are preceded by the formation and dissolution of this dynamic structure, and data indicate that the cell membranes are intact when an actin focus is present. Drosophila myoblast fusion is a relatively rapid process, with the average fusion event being approximately 12 min from appearance of the focus to its dissolution and incorporation of a new nucleus into the myotube (39).
Actin foci have been analyzed in the fusion mutants, leading to the organization of the mutants into three distinct classes (39) (Figure 2). In class one, no or fewer actin foci are detected: sns mutants and duf/rst-deficient embryos have no actin foci, while rols mutants have a drastically reduced number of actin foci. In class two mutants, normal-sized foci are detected, but with increased number, presumably because of the block in fusion. The loner mutation falls into this class. The third class of mutations is typified by more and larger foci. Members of this class include mbc, blow, kette, Rac, Scar and sing mutants (39) (B. E. R. and M. K. B., unpublished observations). The actin focus phenotype of D-WIP/sltr mutants remains controversial. Focus size appears identical to wild type, placing D-WIP/sltr into the second class of fusion mutants. Kim et al. (46) report that sltr-deficient FCMs do not form actin foci, as judged by overlap with the FC-specific protein Rols7 (46). However, in our hands, using the same technique, actin foci are detected in both FCs and FCMs in sltr mutants (B. E. R. and M. K. B., unpublished observations). Clear labeling of cell membranes will be necessary to resolve this issue. Finally, the actin focus phenotypes of Wsp and elmo mutants remain to be tested, while the actin focus phenotype of Arp2/3 mutants awaits stronger loss-of-function reagents.
TEM has also been used to analyze myoblast fusion at the ultrastructural level. Through this approach, Doberstein et al. identified several different stages of fusion and ordered them into a sequential process based on the prevalence of a structure at a particular developmental stage (50). This model proposed that following adhesion between an FC/myotube and an FCM, a set of electron-dense vesicles form and align across the membranes of each cell, forming a ‘prefusion complex’. Next, an electron-dense plaque forms, presumably from the contents of the prefusion complex. Finally, the cells align with apposed plasma membranes at which point fusion pores open and the cells undergo cytoplasmic mixing and ultimately fusion to become one cell.
Despite the insights to the fusion process provided by ultrastructural studies, questions remain about the proposed sequential order in which the steps of fusion take place. Fusion is not a synchronized process with all myoblasts starting and stopping the fusion cycle at the same time. Hence, many fusion events at different steps in the fusion cycle are captured by a particular section prepared for TEM, and relative prevalence of subcellular structures does not provide direct evidence of a transition between these steps. Further development of methodologies to confirm the order of the subcellular events during fusion is necessary.
TEM has been used to analyze several fusion mutants. mbc mutants almost completely lack prefusion complexes, suggesting that Mbc is required for the prior step of myoblast recognition/adhesion (50). blow mutants contain normal numbers of prefusion complexes but lack electron-dense plaques, consistent with a role for Blow in their formation (50). sing mutants contain elevated numbers of prefusion complexes compared with wild type, suggesting that Sing is necessary for fusion to progress past this step (49). Larger electron-dense plaques are found in kette mutants compared with wild type, indicating that Kette may be responsible for their breakdown (45). The TEM phenotype of D-WIP/sltr mutants remains controversial. Kim et al. (46) have reported that D-WIP/sltr mutants appear to have a defect in vesicle targeting as paired vesicle complexes are seen not only at the site of FC/FCM adhesion but also between FCMs in these mutants (46). However, Massarwa et al. (47) found that D-WIP/sltr mutants progress to the final stage of fusion, with fusion pores formed between adherent cells (47). Further studies are required to resolve these conflicting observations.
Given these data sets, it is tempting to equate the actin foci with the electron-dense material (prefusion complex and plaques) seen by TEM. However, there is a strong argument against this conclusion. Enlarged actin foci have been observed in a number of fusion mutants, including kette, mbc and blow (39). Whereas all show enlarged foci, TEM analysis reveals distinct phenotypes; mbc mutants are deficient in recognition/adhesion, blow mutants lack electron-dense plaques and kette mutants show a block at plaque formation (45,50). Therefore, it is possible that all steps revealed by TEM occur at the time in which an actin focus is present. Recent TEM work suggests that actin is important for the targeting of vesicles to plasma membrane at the site of myoblast adhesion (46). How the diffuse actin observed in these studies relates to the dynamic but concentrated accumulation of F-actin into foci at the plasma membrane as seen by live and fixed imaging remains in question. Hence, the connections between what has been detected at the light microscope level and that of the TEM remain to be clarified.
The prevalent model from recent studies of Drosophila myoblast fusion proposes that fusion takes place in two distinct steps (21,45). This two-step model states that initial myoblast–myoblast fusion requires certain gene products and subcellular events to produce a bi- or trinucleate myotube (known as a precursor cell). This model also states that succeeding myotube–myoblast fusion events require a different set of gene products and subcellular events to undergo additional rounds of fusion until a muscle of the correct size is formed. This two-step model is based on the initial identification of two types of fusion mutants: those in which no fusion occurs (interpreted as a block in initial myoblast–myoblast fusion) and those in which some fusion occurs (interpreted as a block in later myotube–myoblast fusion).
An in-depth re-examination of the fusion profile of individual muscles in fusion mutants provides evidence that is inconsistent with a requirement for distinct gene products in each of the two steps of myoblast fusion (11). This analysis shows that all fusion mutants are capable of occasional fusion events. These fusion events can occur at any stage of myoblast fusion and not just at the stage of initial myotube/precursor cell formation. Conversely, in fusion mutants in which two to three fusion events per FC frequently occur, there are also FCs that do not fuse at all, demonstrating that these gene products can be required for both initial or later fusion events depending on the FC (11). To explain these observations, we propose that limited fusion events that occur in fusion mutants are because of residual gene product activity. For example, mbc, Rac, kette and loner all have RNA expression in the early embryo that could be maternally loaded and therefore allow some fusion, even in a zygotic null mutant embryo (27,32,43,52). This hypothesis could also explain the differences in final nuclei number observed between various fusion mutants as well as the intermediate fusion blocks often observed with rescue experiments (11,20,21,45).
Two additional observations provide additional insight to the timing of fusion events. First, live imaging studies have indicated that the average time for an individual fusion event to occur does not change between early and late fusion (39). Second, examination of the three-dimensional arrangements of myoblasts during fusion indicated that the frequency of fusion events increases later during the fusion process, and this increased frequency of fusion correlates with the mobilization, migration and recruitment of FCMs from inner layers of somatic mesoderm (11). Taken together, these data have led us to propose a new model to describe the process of myoblast fusion at the tissue level. As in the past model, we find that there are two steps of fusion. In the first step, which corresponds to early fusion (stage 12–13, 7.5–10.5 h AEL), rare and limited fusion events occur between FCs and adjacent, external FCMs. During the second step, which corresponds to late fusion (stage 14–15, 10.5–13 h AEL), frequent fusion events occur as FCMs move externally to fuse with growing myotubes. In contrast to the two-step model outlined above, our model proposes that all gene products and subcellular behaviors (e.g. actin foci, prefusion complexes and fusion plaques) previously identified are required for both steps of the fusion process. We propose that the transition between the first step of early and infrequent myoblast–myoblast fusion events and the second step of later and frequent myotube–myoblast fusion events is instead because of a limiting factor such as Duf and/or FCM migration. Levels of Duf at the membrane are known to be tightly regulated, and migration of FCMs does correlate with later, more frequent fusion events (11,23).
The basic cellular events that occur during myoblast fusion in both Drosophila and mammals, namely recognition, adhesion, alignment and breakdown of myoblast membranes, appear identical. However, in contrast to the Drosophila embryo in which each muscle is composed of a single, multinucleate myofiber, muscles in mammals are composed of bundles of many myofibers. Myofibers form by the fusion of mononucleated myoblasts to create nascent multinucleated myofibers. These myofibers further mature, as seen by increases in size that require additional myoblast fusion. Myofiber size can change dramatically depending on context: for example, myofiber size can decrease because of disuse or increase because of exercise (53,54).
The fusion events of myoblasts that generate these myofibers occur in two well-characterized phases. In the first phase, de novo myoblast–myoblast fusion events generate a syncytial myofiber with several nuclei. The second phase of fusion occurs by the fusion of mono-nucleated myoblasts with these multinucleate structures (myoblast–myotube fusion events). Both phases of myoblast fusion are essential for the development of the musculature, regeneration of skeletal muscle that has become damaged as a result of injury or disease and in hyperplasia. The second phase is also implicated in growth after atrophy and maintenance (53,54). A role for this second phase in muscle hypertrophy, however, is now being debated (55–57). What contrasts these two phases from what is seen in the two steps in Drosophila fusion appears at the mechanistic level: specific gene products and processes have been identified in mammals that regulate each of these phases during muscle development and regeneration (see below).
In vitro studies, as well as analysis of primary myoblasts derived from mouse knockout models, have identified several gene products that are crucial for mammalian myoblast fusion (54). Proteins that affect myoblast–myoblast fusion include components of calcium signaling pathways, membrane proteins, cell surface metalloproteases, ion channels, secreted extracellular signals and factors that regulate myoblast motility (Table 2). Disruption of these gene products impairs or blocks myoblast–myoblast fusion. However, despite the importance of these molecules to the fusion process, the exact mechanism(s) whereby these factors regulate this process, either directly or indirectly, remain unclear.
Specific gene products have been identified that regulate the second phase, or myoblast–myotube fusion, events. The NFATC2 transcription factor regulates fusion through the regulated secretion of interleukin-4, which recruits free myoblasts for fusion to damaged or growing myofibers (58,59). Importantly, muscles still form in NFATC2−/− mice, and isolated cultures of primary NFATC2−/− myoblasts fuse into small thin myotubes, indicating that initial fusion events that form syncytial structures occur. These myofibers, however, are significantly reduced in size and have lower myonuclear numbers, indicating a critical role for NFATC2 in later myoblast–myotube fusion events (58).
Myoferlin, a member of the ferlin family and homologue of human dysferlin, is highly expressed in developing skeletal muscle. Like dysferlin, myoferlin is membrane associated and binds phospholipids in a calcium-dependent manner (60,61). Similarly to NFATC2, mutations in myoferlin lead to defects in myoblast–myotube fusion events. While muscle fibers do form in myoferlin-null mice, these fibers have reduced cross-sectional area and are much smaller than wild-type littermates. Regeneration of muscle in myoferlin-null mice also is significantly impaired following injury. Together, these results indicate a role of myoferlin in the fusion of myoblasts with existing myofibers during development and regeneration (61). Enrichment of myoferlin is observed at the apposed membranes between fusing structures. The localization of myoferlin described in this study might provide a possible means of identifying the actual site of fusion between mammalian myoblasts. As was shown in Drosophila, identifying the site of fusion would provide new insight to the roles played by all proteins implicated in fusion process (Table 2).
Genetic and cell biological studies in Drosophila have highlighted an essential role of the cytoskeleton for the cellular events required during myoblast fusion (39,62). Cytoskeletal remodeling is important for the cell shape changes required for recruitment and adhesion of myoblasts and for cellular events leading to membrane merger between myoblasts. Although cytoskeletal remodeling during mammalian myoblast fusion has not been extensively described, there is emerging evidence of its importance throughout the fusion process. Several recent studies have determined that regulation of myoblast migration, which is regulated by actin cytoskeleton remodeling (63), is critical for myoblast–myoblast and/or myoblast–myotube fusion events (64–67). These myoblast migrations are thought to enhance or promote myoblast–myoblast and myoblast–myotube contacts, which are critical for myoblast differentiation and fusion to proceed (64–70).
Additional roles for cytoskeletal remodeling in the subsequent steps leading to membrane merger between mammalian myoblasts are also emerging. For example, examination of C2C12 myoblasts after the switch to differentiation media that contains the actin polymerization inhibitor cytochalasin D reveals a block in myoblast fusion (71) (Figure 3). Likewise, analysis of two GEFs that have been linked to cytoskeletal rearrangements in other systems, including Drosophila fusion, has provided evidence for a role of these proteins in mammalian myoblast fusion. Specifically, targeted knockdown of Dock180, the mammalian homologue of Drosophila mbc, and Brag2, the mammalian homologue of Drosophila loner, leads to several defects in C2C12 myotube differentiation, including a block in myoblast fusion (72). Moreover, these fusion defects were retained in vivo: injected knockdown cells failed to form myofibers when injected in to the tibialis anterior of the CB17/severe combined immunodeficiency mice. Further examination of these knockdown C2C12 myotubes also showed distinct effects of Dock180 and Brag2 on myotube morphology. Whether these defects in cell morphology also contribute to the fusion defects is not known.
Another cytoskeletal system, the microtubules, has also been implicated in the fusion process. Treatment of myoblasts with agents that cause microtubule depolymerization, such as nocodazole, leads to impairment of myotube formation (73). In a recent set of experiments, depletion of EB3, a microtubule tip-binding protein through short hairpin RNA interference (shRNAi), led to altered myoblast morphology in differentiating cells, concomitant with a reduction in fusion of shRNA-expressing myoblasts (74). In these shRNAi cells, microtubule capture and stabilization at the cellular cortex of membrane protrusions were disrupted. However, it remains to be determined whether the impaired myoblast fusion is because of a direct role of microtubules in fusion process per se or a secondary effect resulting from disrupted cortical organization of microtubules.
While further research is required to determine if the roles for processes such as cytoskeletal remodeling are conserved between Drosophila and mammals during myoblast fusion, recent studies have indicated that some gene products required for fusion are conserved between Drosophila and zebrafish. For example, Kirrel, the zebrafish orthologue of Drosophila Duf, is required for zebrafish fast muscle fusion (75). Similarly, loss of Rac function in zebrafish embryos also disrupted fast muscle myoblast fusion and muscle formation (75). An additional study determined that knockdown of Dock1 and Dock5, orthologues of Drosophila Mbc, results in the failure of fast-twitch myoblasts to fuse (76). Importantly, this study also found that zebrafish orthologues of Crk, a protein that biochemically interacts with Mbc in Drosophila, also interacts with Dock1 and Dock5, strengthening the conclusion that these proteins are functionally conserved between Drosophila and zebrafish (76). As discussed previously, recent work has also found that Dock180 (mbc) and Brag2 (loner) have a conserved role in regulating myoblast–myotube fusion events in mammals (72). Additionally, Brag2 and Dock180 also regulate the fusion of macrophages, suggesting that fusion machinery is also conserved across multiple cell types (72).
While the Dock180-regulated pathway appears to be conserved among insects, fish and mammals, it remains to be seen if there is functional conservation of other genes critical for Drosophila myogenesis in mammals. For example, mice deficient for Neph1, the mammalian homologue of Kirrel and Duf, do not display defects in skeletal muscle formation but rather have impaired kidney function (77). This suggests that the function of Duf/Kirrel/Neph1 during embryonic myogenesis may not be functionally conserved from Drosophila and zebrafish to mammals. However, a role for Neph1 may be masked by other Kirrel protein family members in mammals with potentially overlapping functions during muscle development (77). Furthermore, Neph1 function may not be essential for initial, myoblast–myoblast fusion events but still be critical for later myoblast–myotube events, which are critical for muscle growth and regeneration, similar to NFATC2 (58) and myoferlin (61). The identification and characterization of the fusion site between mammalian myoblasts as well as further study of mammalian orthologues of Drosophila and zebrafish fusion proteins remain important avenues for future research in this field. Note in proof: another allele of Arp3 has been recently identified and has been shown to affect myoblast fusion in Drosophila (84).
We thank members of the Baylies Lab and R. Krauss for discussions and critical reading of the manuscript. This work was supported by Sloan-Kettering Institute, National Institutes of Health grants (GM 586989/GM 78318) to M. B. and a Muscular Dystrophy Association Research Development Grant to S. N. (MDA4153).