WASP and SCAR have distinct roles in activating the Arp2/3 complex during myoblast fusion

Intercellular fusion of myoblasts is a crucial process for somatic muscle formation during embryogenesis of Drosophila and vertebrates, including mammals (Taylor, 2006). Additionally, in vertebrates, muscle growth after birth requires continuous fusion of myoblasts arising from satellite cells, the stem cells of vertebrate muscles. The somatic musculature of Drosophila melanogaster is well-characterized and can serve as a model in which to study and unravel the underlying molecular mechanisms involved in myoblast fusion.

In Drosophila, myoblast fusion results from a dynamic relationship between founder cells (FCs) and fusion-competent myoblasts (FCMs) (Bate, 1990). The FCs seed the fusion process and are thus responsible for the unique identity of each muscle fiber. The larger population of FCMs migrate towards, recognize and adhere to the FCs (for reviews, see Abmayr et al., 2005; Baylies et al., 1998; Chen and Olson, 2004; Taylor, 2000). Following these cellular interaction steps, an FC then fuses with one or two FCMs, giving rise to a precursor cell, which in turn merges with additional myoblasts in a second fusion step until the appropriate muscle size is reached (Bate, 1990; Rau et al., 2001). Studies on mammalian cell cultures have revealed that there are strong parallels to this model in vertebrates at the molecular level (Horsley and Pavlath, 2004).

Recently, several essential molecules required for the fusion of FCs and FCMs have been reported. For example, during cell adhesion, membrane-spanning Immunoglobulin (Ig)-molecules are involved in establishing a ring-shaped fusion-restricted myogenic-adhesive structure (FuRMAS) (Kesper et al., 2007). The Ig-domain molecule Dumbfounded [Duf, also known as Kin of Irre (Kirre) (see FlyBase)] (Ruiz-Gomez et al., 2000; Strünkelnberg et al., 2001) forms the ring at cell-contact points in FCs, whereas the Ig-domain protein Sticks and Stones (Sns) (Bour et al., 2000) forms a corresponding structure in FCMs. Both proteins interact heterotypically to initiate the process of myoblast fusion. The progress of fusion is manifested in the expansion of the FuRMAS at the contact sites between FCs and FCMs from 1 μm to 5 μm (Kesper et al., 2007). Interestingly, the Ig-domain protein Duf was recently identified in zebrafish, suggesting a conserved molecular pathway in insects and vertebrates (Srinivas et al., 2007).

Other known components of the FuRMAS include the adaptor protein Rolling pebbles7 [Rols7, also known as Antisocial (Ants)] (Chen and Olson, 2001; Menon and Chia, 2001; Rau et al., 2001) in FCs and Blown fuse (Blow) (Doberstein et al., 1997; Schröter et al., 2006) in FCMs. Rols7 induces the second fusion step (Rau et al., 2001) and is expressed as a ring in FCs, similar to Duf, and delivers the signal from the FC membrane into the cytosol by interacting with the intracellular domain of Duf via its tetratricopeptide (TPR) repeats (Kreisköther et al., 2006). Blow, by contrast, is concentrated as a plug in the center of the ring formed by Sns (Kesper et al., 2007).

During the fusion process, the actin cytoskeleton of myoblasts is dynamically organized. Consistent with this, the center of the adhesion ring in FCs and FCMs contains filamentous actin (F-actin) (Kesper et al., 2007). Recently, we reported that the actin regulators kette (Schröter et al., 2004) and wasp (Schäfer et al., 2007) are essential for myoblast fusion. Kette is the Drosophila homolog of HEM2 (NAP1), and can be found in a complex with SCAR (homolog of the mammalian WAVE) and is a member of the Wiskott-Aldrich-syndrome protein (WASP) family (reviewed by Ibarra et al., 2005). Moreover, kette was shown to interact genetically with blow during myoblast fusion (Schröter et al., 2004). WASP family members possess a conserved VCA (verprolin homologous, cofilin homologous and acidic) domain, which is involved in the binding of G-actin and Arp2/3 (Rohatgi et al., 1999; Miki and Takenawa, 1998; Machesky et al., 1999). The wasp^3D3-035 allele lacks the CA domain of the VCA domain and thus neutralizes the function of maternal WASP (Schäfer et al., 2007). Furthermore, genetic data indicate that Kette antagonizes the function of WASP (Schäfer et al., 2007). The WASP-interacting partner Verprolin1 [Vrp1, also known as Solitary (Sltr) and Wip; the Drosophila homolog of the mammalian WIP], however, acts in a complex with WASP to ensure successful myoblast fusion (Kim et al., 2007; Massarwa et al., 2007).

The branching of F-actin is initiated by the de novo nucleation of actin, triggered through the activity of the Arp2/3 complex (reviewed by Takenawa and Miki, 2001). Alone, the Arp2/3 complex is inactive and requires a set of binding partners, including members of the WASP family, to become active (Machesky and Insall, 1998). The binding of these proteins has been proposed to lead to a conformational change in the position of the seven subunits of the Arp2/3 complex, especially in the relative position of the subunits Arp2 and Arp3 (Robinson et al., 2001; Volkmann et al., 2001).

Here, we describe for the first time an EMS-induced Arp3-null allele, Arp3^{schwächling}, having an abnormal myoblast-fusion phenotype. Ultrastructural analysis of Arp3^{schwächling} and wasp^3D3-035 revealed the formation of a fusion pore in both mutants. Whereas in Arp3^{schwächling} mutants the membrane between fusing myoblasts is removed, in wasp^3D3-035 embryos, membrane breakdown is not completed. The finding that fusion in Arp3^{schwächling} embryos is disrupted despite the formation of a fusion pore indicates that the pore does not expand in Arp3^{schwächling} mutants. Therefore, to gain a deeper insight into the F-actin regulation of myoblast fusion, we examined Arp3 wasp double-mutant embryos, which show a complete block of myoblast fusion. This phenotype suggests that WASP is not the only actin regulator controlling F-actin polymerization during fusion. Indeed, our studies on scar single mutants and scar vrp1 double mutants strongly imply that the first and the second fusion steps require a different set of F-actin nucleation factors.

The fusion mutant schwächling was detected in the same mutagenesis collection in which we previously identified kette and wasp (Schröter et al., 2004; Schäfer et al., 2007). Upon analysis of the muscle pattern of the ethyl methane sulphonate (EMS)-induced pointed allele pnt^3D2-26 with the anti-β3-tubulin antibody, we observed an arrest in myoblast fusion. To address whether the mutation in pnt is responsible for the fusion defect, we examined the embryonic muscle pattern of the loss-of-function mutant pnt^Δ88. However, pnt-null mutants did not show a myoblast-fusion phenotype (data not shown), indicating that the mutation in pnt is not responsible for the observed phenotype. Therefore, it is highly likely that the fusion phenotype is caused by a further mutation on the pnt^3D2-26-containing chromosome. Complementation tests demonstrated that the additional mutation was not allelic to known mutations in genes required for myoblast fusion, thereby revealing a new fusion-relevant component, which we have named schwächling. To characterize the schwächling mutant phenotype in detail, we separated the pnt locus from the schwächling mutation by using meiotic recombination. Homozygous schwächling mutants display a severe defect in myoblast fusion (Fig. 1B, arrow).

To map the region in which schwächling is localized, we used the deficiency kit for the third chromosome. The deficiency Df(3L)ZP1 was allelic to schwächling, and both homozygous (Fig. 1C, arrow) and transheterozygous (Fig. 1D, arrow) embryos had a myoblast-fusion defect. Df(3L)ZP1 is a large deficiency (66A17-66C5) resulting in the removal of approximately 90 open reading frames (ORFs). We also found that the deficiency Df(3L)pbl-X1 (65F3-66B10) is allelic to schwächling. To further narrow-down this region, we employed P-element insertions and performed complementation tests. The lethality-causing EP-element insertion Arp66B^EP3640 did not complement schwächling. Moreover, the EP insertion Arp66B^EP3640 was mapped to the cytological position 66B6 and was found to be inserted in the 5′UTR of the Drosophila homolog of the Arp3 gene Arp66B (FlyBase). We observed that homozygous Arp66B^EP3640 mutant embryos do not exhibit a strong myoblast-fusion defect (Fig. 1E); however, in transheterozygous Arp66B^EP3640/schwächling embryos, a disruption in myoblast fusion was clearly visible (Fig. 1F, arrow). These observations suggest that schwächling is a new allele of Arp3; we named this allele Arp3^{schwächling}. Further support for this finding also comes from RNA in situ hybridizations, which show that the Arp3 transcript is expressed from stage 10 onwards in the mesoderm (Fig. 1G,H).

To investigate the molecular basis of the EMS mutation in Arp3^{schwächling}, we determined the sequence of Arp3 in Arp3^{schwächling} homozygous-mutant embryos. Fig. 1J shows the gene structure of Arp3. In Arp3^{schwächling} embryos, the Arp3 gene contains a G→A nucleotide transition at the splice acceptor site of the first intron (Fig. 1J). This point mutation possibly leads to an aberrantly spliced transcript, retaining the first intron, that becomes translated into a truncated protein of 15 amino acids (aa) and can be thus considered as a null allele. To test this notion, RNA was isolated from wild-type, Arp3^{schwächling} heterozygous and homozygous embryos, and used as template in reverse transcriptase-PCR (RT-PCR) assays. We amplified a 509-bp fragment representing the spliced mRNA form of exons 1 and 3 (Fig. 1I, lower arrowhead). In Arp3^{schwächling} heterozygous and homozygous embryos, we additionally obtained a 674-bp product containing the first intron (Fig. 1I, upper arrowhead). However, because this product was present in low amounts, it is highly likely that the spliced form of the product was still present, probably because of the maternal contribution of Arp3 mRNA.

Further analysis revealed that the determination of FCs and FCMs in Arp3^{schwächling} mutant embryos occurs normally (data not shown). Consequently, the phenotype in Arp3^{schwächling} embryos is due to a failure of myoblasts to fuse. Myoblast fusion in vertebrates and Drosophila proceeds in two distinct steps (Bate, 1990; Beckett and Baylies, 2007; Horsley and Pavlath, 2004; Rau et al., 2001). During the first step, a bi- to tri-nucleated precursor cell forms, which undergoes additional rounds of fusion until the final size of the mature myofiber is reached. The transcription factor Even skipped (Eve) marks the 14 nuclei of the segmentally repeated dorsal muscle DA1 as well as the pericardial cells. To estimate at which stage of fusion Arp3^{schwächling} mutant embryos are arrested, we counted the nuclei of the DA1 muscle marked by anti-Eve (Fig. 2B,D). The DA1 muscles in Arp3^{schwächling} embryos were found to contain, on average, three nuclei, indicating that the first fusion step, forming a bi- to trinucleated precursor cell, had completed.

Arp3 is an actin-related protein that is one of the seven proteins forming the Arp2/3 complex, which is involved in creating new branching actin filaments. The activation of the Arp2/3 complex requires the binding of the WASP family members. In Drosophila, there are two members of this family – WASP and SCAR. Both proteins share a common C-terminal Verprolin homologous (V), Cofilin homologous (C) and acidic (A) domain (also known as the VCA domain). The V domain binds monomeric G-actin and the CA domain binds the Arp2/3 complex (Machesky and Insall, 1998; Miki and Takenawa, 1998; Rohatgi et al., 1999). Previously, we reported that the lack of the CA and A domain of WASP blocks myoblast fusion after precursor formation (Schäfer et al., 2007). Thus, in both Arp3^{schwächling} and wasp^3D3-035 mutants, myoblasts complete the first fusion step, but fail to undergo the second one.

The second fusion step takes place as several distinct steps that can be distinguished at the ultrastructural level (Doberstein et al., 1997; Schröter et al., 2004). First, electron-dense vesicles appear on adjacent myoblast membranes. This structure is known as the pre-fusion complex (Fig. 2E, arrow). The vesicles are thought to disintegrate and give rise to the so-called electron-dense plaques (Fig. 2F, arrow). At the contact area between a precursor cell and an FCM (Fig. 3A,A′, between the arrows), membranes start to vesiculate, initiating the breakdown of the membrane (Fig. 3A′, arrowheads).

In ultrathin sections of Arp3^{schwächling} embryos, we observed the formation of a fusion pore between a growing myotube and an FCM (Fig. 3C, overview; Fig.3C′, arrows). However, we did not observe any plasma membrane remnants between the cells, as seen in wild-type embryos (Fig. 3A′, next to the arrowheads). We therefore conclude that Arp3^{schwächling} embryos stop fusion after membrane breakdown, when a fusion pore has been formed. By contrast, ultrathin sections of wasp^3D3-035 embryos revealed that membrane breakdown fails to complete in these mutants (Fig. 3B; Fig. 3B′, between arrowheads). Hence, wasp is required for the creation of a fusion pore, whereas Arp3 is required to integrate FCMs into the growing myotube after fusion-pore formation.

Recently, two papers have described the identification of a WASP-interacting partner named Vrp1 as an essential component for myoblast fusion (Kim et al., 2007; Massarwa et al., 2007). To date, two vrp1 alleles, namely wip^D30 and sltr^S1946, are known to stop fusion after precursor formation. Because we observed that wasp^3D3-035 mutants arrest fusion during membrane removal, but that a very different phenotype for vrp1 has recently been reported (Kim et al., 2007), we reinvestigated the phenotype of vrp1 mutant embryos by electron microscopy. For this purpose, we used the characterized wip^D30 allele. Our analysis confirmed that wip^D30 mutants (Fig. 3D,D′) are arrested during membrane breakdown – a finding also reported by Massarwa et al. (Massarwa et al., 2007) – and that this arrested stage is the same as found in wasp^3D3-035 mutants (Fig. 3B,B′).

To gain further insight into the relevance of F-actin regulation during myoblast fusion, we generated and investigated Arp3^{schwächling} and wasp^3D3-035 double-mutant embryos. As described previously, the wasp allele wasp^3D3-035 carries a point mutation that leads to loss of the CA domain of the VCA module (Schäfer et al., 2007). wasp^3D3-035 mutants show an unfused myoblast phenotype (Fig. 4B,B′). Because the CA domain is essential for binding of the Arp2/3 complex, we predicted that the phenotype of the double mutants would resemble that of wasp^3D3-035 mutant embryos. Instead, we observed a strongly enhanced myoblast-fusion phenotype (Fig. 4C,C′). To analyze the Arp3^{schwächling} wasp^3D3-035 mutant phenotype more precisely, we examined the DA1 muscle with the nuclear marker anti-Eve. The DA1 muscle was found to only contain one nucleus per hemisegment (Fig. 2C,D), indicating that fusion is blocked completely in the double mutant. To examine the interaction of Arp3 and wasp during myoblast fusion more closely, we performed the double-mutant analysis with weaker alleles of Arp3 and wasp. This analysis revealed that, in Arp3^{schwächling} wasp³ and Arp66B^EP3640 wasp^3D3-035 double mutants, fusion is also blocked completely (data not shown). wasp and Arp3 mRNA are maternally contributed. In the case of wasp, maternal wasp is sufficient to complete myoblast fusion (Schäfer et al., 2007). We therefore expected that fusion would not be blocked completely, but that the first fusion step, but possibly not the second, would take place in wasp Arp3 double-mutant embryos. The enhanced fusion defect observed in the double mutant led us to propose that wasp is only required during the second fusion step. As a consequence, a different Arp2/3 activator besides WASP must regulate the activity of the Arp2/3 complex during the first step of myoblast fusion.

The activity of the Arp2/3 complex is regulated by nucleation-promoting factors such as WASP and SCAR. To determine whether scar is required for myoblast fusion, we examined a hypomorphic and a null allele of Drosophila scar, and generated double mutants with scar and components of the WASP complex.

First we examined the muscle pattern of homozygous scar^k13811 and scar^Δ37 mutants (Zallen et al., 2002) using the anti-β3-tubulin antibody. We observed a weak fusion defect in these embryos. The unfused myoblast phenotype in homozygous scar^Δ37 embryos was more severe than in homozygous scar^k13811 mutants (Fig. 5). Nevertheless, unfused myoblasts were visible in both scar mutants and muscles seemed to be reduced in size in scar^Δ37 mutants (Fig. 5E, arrowhead). We further noticed that dorsal closure is incomplete in both scar mutants. Similar to wasp and Arp3 mRNA, scar mRNA is maternally contributed (Zallen et al., 2002). This might explain why scar-null mutants still show considerable myoblast fusion. To investigate whether scar and wasp interact genetically during myoblast fusion, we performed a dosis experiment and generated scar wasp double mutants. We used the scar^k13811 and wasp³ alleles, which display only a very weak myoblast-fusion phenotype, for this experiment. Taking out one copy of wasp³ in a scar^k13811 mutant background intensified the appearance of unfused myoblasts (Fig. 5B, arrow). This phenotype became even more severe in scar^k13811 wasp³ double-mutant embryos (Fig. 5C, arrow). These findings support our notion that myoblast fusion in Drosophila is regulated via both members of the WASP family. WASP and its interacting partner Vrp1 are both essential for myoblast fusion. Thus, we next examined whether scar and vrp1 also interact genetically.

In addition to the known vrp1 alleles, there is a piggyBac insertion from the Harvard Drosophila stock collection available named vrp1^f06715. We found that vrp1^f06715 does not complement the deficiencies Df(2R)Exel6078 and Df(2R)Exel7010, and is allelic to wip^D30. Fig. 5F shows that some fusion takes place in vrp1^f06715 embryos; however, at the end of embryogenesis the growing myotubes are still surrounded by numerous unfused myoblasts. This phenotype is similar to the one observed in wip^D30 mutants. To assess the role of the WASP-Vrp1 complex and SCAR during myoblast fusion, we generated scar^Δ37 vrp1^f06715 double mutants. Immunostainings of these double mutants with anti-β3-tubulin revealed a dramatically increased myoblast-fusion defect (Fig. 5G) similar to that observed in Arp3 and wasp double mutants.

Taken together, our double-mutant experiments with scar wasp and vrp1 clearly show that scar and wasp interact genetically during myoblast fusion. Thus, SCAR and the WASP-Vrp1 complex act together during myoblast fusion to regulate the activity of the Arp2/3 complex.

To study whether Verprolin is distributed similarly to F-actin during myoblast fusion (Kesper et al., 2007), we generated an anti-Vpr1 antibody and analyzed the subcellular localization of Vrp1 in embryos carrying the rP298 enhancer-trap insertion. The enhancer-trap line rP298 expresses lacZ under the control of the Duf regulatory region, and thus marks all nuclei of the FCs and growing muscles (Nose et al., 1998; Ruiz-Gomez et al., 2000). Vrp1 was recently reported to localize to specific foci in FCMs (Kim et al., 2007); however, we observed that this is only the case in FCMs that have not yet formed a filopodia (Fig. 6A′, asterisk). After filopodia formation, however, Vrp1 becomes localized to the tip of the filopodia. In Fig. 6A′, the arrowhead points to a dinucleated myotube in which Vrp1 is faintly expressed. Thus, Vrp1 is concentrated like F-actin at the site of cell-cell contact in FCMs, but not at the site of the FCs. To clarify whether the localization of Vrp1 depends on cell contact, we examined the distribution of Vrp1 in an sns-null allele. Although we observed Vrp1 to be still present in sns^20-23 (Paululat et al., 1995) embryos, the protein could only be found in discrete foci (Fig. 6B′, asterisks). Accordingly, Vrp1 is restricted to the filopodia of FCMs, in which it attaches to an FC/growing myotube. We therefore assume that Vrp1 increases, together with WASP, the formation of F-actin at the tip of the filopodia. The localization of Vrp1 to the filopodia depends on cell-cell contact, as revealed by immunostainings with anti-Vrp1 on sns^20-23 single mutants and on wasp^3D3-035 Arp^{schwächling} double mutants (Fig. 6B′ asterisk; Fig.6C′, asterisk and arrow). This further shows that the correct localization of Vrp1 in the double mutant is not disturbed because myoblasts are capable of recognition and adherence. When we applied the anti-Vrp1 antibody to vrp^f06715 mutant embryos, no staining was detectable, indicating that no protein is generated in vrp1^f06715 embryos. Taken together, we found that Vrp1 becomes restricted to the filopodia formed by the FCMs after successful cell-cell contact.

To test whether the initial step of recognition and adhesion between FCs and FCMs is indeed not disturbed in Arp3^{schwächling} and wasp^3D3-035 single mutants, or in the double mutants, we investigated the distribution of Duf in those mutants. In wild-type embryos, we observed the typical ring-like structure in which Duf is distributed (Fig. 7A, arrowhead). Furthermore, FCMs could be seen contacting growing muscles (Fig. 7A, asterisk). Fig. 7B shows that, in Arp3^{schwächling} mutants, several FCMs successfully attached to a precursor cell, but did not become integrated into the precursor cell (asterisks). Also, in wasp^3D3-035 mutants, FCMs were seen to adhere to a growing myotube (Fig. 7C). These findings are not surprising, because the characterization of the Arp3^{schwächling} and wasp^3D3-035 mutant phenotypes has already demonstrated that some fusion takes place in both mutants. However, can FCMs that fail to fuse completely in Arp3^{schwächling} wasp^3D3-035 double mutants attach to FCs or is the failure of fusion in the double mutant due to a secondary effect, e.g. cell migration? Fig. 7D shows that FCMs still find their target, thereby suggesting that no secondary effect is involved (asterisks).

The genetic data presented in this study provide new insights into the regulation of branching F-actin during myoblast fusion. There are three classes of protein that initiate the polymerization of new actin filaments: the actin-related protein complex Arp2/3, the Formins and Spire. These protein classes are evolutionary conserved in most eukaryotes and promote new actin assembly by a distinct mechanism. The Arp2/3 complex is the only known protein complex that initiates new actin filaments branching off an existing filament. Upon cell-cell contact, F-actin is mostly present in the tip of the FCM and in the area of the FC/growing myotube to which the FCM is attaching (Kesper et al., 2007). Our analyses of Arp3, wasp and scar mutants indicate that branching F-actin is essential for myoblast fusion. Based on our double-mutant analyses, we postulate that the WASP-Vrp1 complex promotes branched F-actin formation positively during the second step of myoblast fusion, thereby allowing us to propose a new model for actin regulation during myoblast fusion (see below).

This work emanated from the identification of an Arp3-null allele. Fusion is severely disrupted in Arp3^{schwächling} mutants. However, stainings with anti-Duf clearly show that this is not due to a failure in cell-cell recognition and adhesion. For example, FCMs still attach to the site of the growing myotubes where Duf is expressed in Arp3^{schwächling} and wasp^3D3-035 mutant embryos. Interestingly, myoblasts also adhere successfully in Arp3^{schwächling} wasp^3D3-035 double mutants that fail to fuse completely (compare Fig. 1B and Fig. 4B with Fig. 4C). Duf serves to attract FCMs, which can migrate towards the Duf-expressing source (Ruiz-Gómez et al., 2000). Hence, the ability of FCMs to migrate is a prerequisite for myoblast fusion. The migration of cells, however, also depends on actin-cytoskeleton regulation. Our observations clearly show that the nature of the fusion arrest in Arp3^{schwächling} and wasp^3D3-035 single mutants, and Arp3^{schwächling} wasp^3D3-035 double mutants, is not due to either the inability of FCMs to migrate nor to a failure of FCs/myotubes and FCMs to recognize and adhere, but rather to a specific defect in cell-cell fusion.

Electron microscopy analyses on Arp3^{schwächling} and wasp^3D3-035 mutants further assisted us to dissect the process in which F-actin formation is required during myoblast fusion. The WASP-Vrp1 complex is involved in the formation of a fusion pore, as reported by Massarwa et al. (Massarwa et al., 2007). In line with their study, we observed that mutations in wasp^3D3-035 and wip^D30 stop fusion during membrane breakdown, but not after pre-fusion-complex formation, as reported by Kim et al. (Kim et al., 2007). We found that the pre-fusion complex, which has been described to consist of 1.4 vesicles per pre-fusion complex (Estrada et al., 2007), looks identical in both of the mutants as well as in wild-type embryos.

Fig. 8A presents a new model for myoblast fusion derived from the data presented in this study, Table 1 summarizes the molecules that are required for the fusion process. Our studies on Arp3 mutants indicate that the polymerization-based force of branching F-actin is required beyond the stage of membrane breakdown. After membranes have been removed between fusing myoblasts, the FCMs must become integrated into the FC/growing myotube. Because Arp3^{schwächling} mutants fail to fuse after membrane removal, we propose that F-actin is required to allow integration into the growing myotubes. After the Duf- and Sns-mediated recognition of FCs and FCMs, further essential proteins for myoblast fusion, e.g. Blow, become recruited to the point of fusion in FCMs. We found that Vrp1 localizes in the tip of the filopodia of FCMs. Kim et al. (Kim et al., 2007) and Massarwa et al. (Massarwa et al., 2007) have shown that the localization of WASP to the membrane depends on Vrp1 activity. Thus, the recruitment of WASP presumably leads to an increase of F-actin at the site of fusion. As a result, the formation of a fusion pore is initiated and expands until the FCM becomes finally pulled into the FC/growing myotube.

Our morphological and statistical analysis on the nuclei of the DA1 muscle suggested that no fusion takes place in Arp3 wasp double mutants. Because wasp^3D3-035 mutants should lack the ability to promote F-actin formation and stop fusion, like Arp3^{schwächling} mutants, after precursor formation, this was a surprising result. We therefore examined whether the actin regulator SCAR additionally controls F-actin branching during myoblast fusion. Double-mutant and epistasis experiments of scar and wasp revealed that this was indeed the case. The complete disruption of myoblast fusion in Arp3 wasp double-mutant embryos – despite the presence of maternal wasp and Arp3 – in conjunction with the finding that SCAR is required for myoblast fusion, led us to propose that SCAR and WASP control different steps of myoblast fusion (Fig. 8B). So far, our genetic data suggest that WASP is only required during the second fusion step. This is in line with previous data presented by Massarwa et al. (Massarwa et al., 2007). It further implies that SCAR is essential for the first fusion step. However, myoblast fusion is not disrupted completely in scar^Δ37-null mutant embryos. The maternally provided gene product of scar is probably able to compensate for the loss of zygotic scar during myoblast fusion. Nevertheless, scar germline clones with reduced levels of maternal and zygotic SCAR protein show an enhanced myoblast-fusion phenotype (Richardson et al., 2007). Because the loss of maternal scar disrupts oogenesis, it is not possible to eliminate the maternal component of scar completely (Zallen et al., 2002). Hence, further experiments are required to determine whether SCAR controls the first fusion step alone or acts in functional redundancy with an additional factor.

Myoblast fusion in scar vrp1 double mutants was blocked completely. This might indicate that SCAR regulates the first fusion step together with the WASP-interacting partner Vrp1. The Vrp1 protein is expressed from stage 10 onwards, shortly before the first fusion step is initiated. Members of the Verprolin/WIP family have been reported to influence actin polymerization in a WASP-independent manner (reviewed by Anton and Jones, 2006; Aspenström, 2005). It remains to be investigated whether this is also the case during Drosophila myoblast fusion.

In addition to regulating the first fusion step, the activity of SCAR might also be required for the second fusion step. Support for this notion comes from our previous findings that the vertebrate homolog of Kette, HEM2, which is present in a complex with SCAR, is essential for myoblast fusion (Schröter et al., 2004). Kette and WASP have antagonistic functions during myoblast fusion (Schäfer et al., 2007). Because we have previously shown that the expression of WASP in FCs failed to rescue the wasp mutant phenotype, one could assume that the activity of WASP is only required in FCMs (Schäfer et al., 2007). We therefore predict that the polymerization of branching actin filaments might be regulated in a myoblast-type-specific manner (Fig. 8B).

In summary, our observations suggests for the first time that the cellular machinery leading to the formation of F-actin is controlled by a different set of nucleation-promoting factors during the first and second fusion steps. Future studies should now focus on the mechanistical details to reveal how these factors become activated in FCs and FCMs.

The EMS mutant Arp3^{schwächling} was mapped using the deficiency kit from the Bloomington Stock Center. We used the following deficiencies and P-insertions to uncover the Arp3 gene (breakpoints are indicated): Df(3L)ZP1 (66A17-66C5), Df(3L)pbl-X1 (65F3-66B10), Df(3L)66C-G28 (66B8-66C10) and Arp66B^EP3640. We found that Df(3L)66C-G28 is also allelic to schwächling and Arp66B^EP3640, suggesting that the distal breakpoint of the deficiency (66B8) is not mapped precisely. Oregon R was used as wild-type strain. We used Dr/TM3 Deformed lacZ and If/CyO hindgut lacZ flies to rebalance our stocks with a blue balancer. Genetic experiments were carried out using the following fly strains: wasp³/TM3 (Ben-Yaacov et al., 2001), wasp^3D3-035/TM3 (Schäfer et al., 2007), scar^Δ37/CyO and scar^k13811/CyO (Zallen et al., 2002). The vrp1^f06715/CyO strain was ordered from the Drosophila Stock Collection at the Harvard Medical School. The wip^D30 allele was obtained from Eyal Schejfer (Weizman Institute, Rehovot, Israel). We used genomic recombination to generate Arp3^{schwächling} wasp^3D3-035, Arp3^{schwächling} wasp³, Arp66B^EP3640 wasp^3D3-035 and scar^Δ37 vrp1^f06715 double mutants. The rP298 enhancer-trap line (Nose et al., 1998) was used to label founder cells. All crosses were performed at 25°C.

Embryos were collected on apple-juice plates and aged at 25°C/18°C. Most embryos were fixed as described previously (Hummel et al., 1997). To visualize the muscle pattern, we used the anti-β3-tubulin antibody (Buttgereit et al., 1996; Leiss et al., 1988) at a dilution of 1:12,000 (for DAB staining) and 1:5000 (for fluorescence staining). To distinguish between balancer (lacZ)-carrying and homozygous mutant embryos, we additionally used the anti-β-galactosidase antibody (Cappel: 1:5000 dilution). The anti-Eve (3C10: 1:30 dilution) antibody was obtained from the Hybridoma Bank. Anti-Duf stainings (1:1000 dilution) were performed as described previously (Kesper et al., 2007). Cy2- and Cy3-conjugated secondary antibodies were purchased from Dianova. Fluorescent images were acquired using a Leica TCSsp2 confocal scanning microscope.

Whole-mount in situ hybridizations were performed as described previously (Tautz and Pfeifle, 1989). The full-length Arp3 cDNA LD35711, which was labeled for RNA in situ hybridizations, were obtained from the DGRC.

Polyclonal antiserum against Vrp1 was generated by injection of guinea pigs with recombinant HIS-tagged protein corresponding to residues 837-936 of Vrp1 in pETM11 (details available upon request). The resulting guinea pig antiserum (Medprobe) was IgG-purifed on a Protein A column (Pierce) prior to use at 1:1000 for immunohistochemistry.

The insertion of the EP element was confirmed by inverse PCR. We used CfoI to digest the genomic DNA isolated from Arp66B^EP3640, and Pry1 and Pry2 (BDGP) as primer pairs for inverse PCR. This revealed that the EP-element insertion is located 123 bp in front of the ORF of Arp3. To sequence the Arp3 gene in Arp3^{schwächling} mutants, genomic DNA was isolated from homozygous Arp3^{schwächling} mutant embryos and the coding region of Arp3 was amplified by PCR. Sequencing was performed by AGOWA (Berlin). For RT-PCR, total RNA was prepared using TRIzol (Invitrogen) from embryos. We used the OneStep RT-PCR Kit (Qiagen) to amplify a 509-bp spliced and a 674-bp unspliced fragment. We used the following primer pairs: forward primer located at the 5′ UTR of LD35711 (5′-CTGGGTTTATCTGCAACTG-3′) and reverse primer located within the third exon of LD35711 (3′-GAACGTCTCAAACATTATCTC-5′).

Embryos were fixed in 18% glutaraldehyde:heptan (1:1) for 20 minutes. X-Gal staining was used to detect homozygous mutant embryos that were then devitillinized manually in 0.5% formaldehyde/0.2 M HEPES. Embryos were washed three times in PBS for 10 minutes each, and then the PBS was replaced with 1% OsO₄ and 0.15% ferrisyanide in PBS for 60 minutes. Embryos were washed three times in H₂O for 10 minutes each and incubated with 4% uranylacetate for 60 minutes. Afterwards, embryos were dehydrated in an ethanol series and incubated for 3 hours in Epon (Epoy-embedding kit, Fluka)/Ethanol (1:1). Ultrathin sections were obtained using an Ultracut E microtome (Reichert-Jung) and analyzed with the Hitachi HU-12A electron microscope.

This work was supported by the Deutsche Forschungsgesellschaft (DFG) by the Graduate School 1216 awarded to R.R.-P. and to S.-F.O., as well as by the European Network of Excellence (MYORES). R.H.P. is a Swedish Cancer Foundation Research Fellow supported by the Swedish Research Council (621-2003-3399). We thank Aurelia Fuchs for carrying out the first phenotypic analysis of scar mutant embryos. We further thank Sabina Huhn and Helga Kisselbach-Heckmann for technical assistance. The electron microscopy studies were supported by EMBO ASTF 215-2006. The first ultrastructural analysis on wild-type embryos was carried out in collaboration with Andreas Brech in Harald Stenmark's laboratory at the Radium Hospital in Oslo, Norway.