Distinct genetic programs guide Drosophila circular and longitudinal visceral myoblast fusion

Drosophila stocks

The following Drosophila stocks were used in this study: Df(3 L)BK9/TM3,Sb,Dfd-lacZ (Deficiency for rols7, BDSC), rols ^XX117 /TM3,Sb,Dfd-lacZ 43], blow ¹ /CyO and blow ² /CyO 44], kette ^J4–48 /TM3,Sb,Dfd-lacZ and kette ^G1–37 /TM3,Sb,Dfd-lacZ 45], mbc ^C1 /TM3,Sb and mbc ^D11.2 /TM3,Ser 25], lmd ²⁰² /TM3,Sb,Dfd-lacZ (Holz and Renkawitz-Pohl, unpublished), rp298-lacZ 46], sns-mCherry-NLS 47], bap-lacZ 12], HLH54F-lacZ 21], HLH54F-GFP 24], twist promoter-GFP-actin 35], rolsIn1-lacZ (see below), the protein trap lines P{PTT-un1}sls ^ZCL2144 (sls::GFP) [48] and P{unk}trol ^GFP311 (trol::GFP) [49], wip ^D30 /CyO 50], wip ^f06715 /CyO,hg-lacZ, wip ^f06715 scar ^Δ37 /CyO,hg-lacZ, arp3 ^schwächling wasp ^3D3–035 /TM3,Sb,Dfd-lacZ 36], and arp3 ^schwächling /TM3,Sb,Dfd-lacZ 36]. The scar ^Δ37 allele was obtained from the Bloomington Drosophila Stock Center. To distinguish between homozygous mutants and heterozygous flies, balancer chromosomes carrying lacZ or GFP insertion markers (Bloomington stocks 6662 and 6663) were used. As the wild-type reference, we used w ¹¹¹⁸ or balanced sibling embryos.

Immunohistochemical analyses of Drosophila embryos

Embryos were fixed and immunohistochemically analyzed as described previously [32]. Guts were prepared from embryos and larvae fixed with 4% formaldehyde and not treated with methanol and were stained the same way as whole mount embryos. The following antibodies were used at the noted dilutions: mouse anti-FasIII [51] 1:50 (Developmental Studies Hybridoma Bank), rabbit anti-Duf/Kirre [28] 1:1000, rabbit anti-β3Tub [52] 1:3000, rabbit anti-β-gal 1:5000 (Biotrend), rabbit anti-Blow [10] 1:200, rabbit anti-DMef2 [53] 1:500, rabbit anti-GFP 1:500 (abcam), rabbit anti-Rols7 (directed against the first 300 amino acids of Rols7) 1:500, rat anti-RFP (Clontech) 1:1000, and rat anti-Tm (abcam) 1:1000. Embryos stained with anti-Rols7, anti-Duf/Kirre, and anti-Blow were heat fixed, and the reaction was enhanced using an Individual Indirect Tyramide Reagent Pack (Perkin Elmer). Primary antibodies were detected using either biotinylated secondary antibodies at a dilution of 1:250 and the Vectastain Elite ABC Kit (Vector Laboratories) or fluorescent-labeled antibodies at a dilution of 1:200 (Dianova). DNA was stained with Hoechst reagent (5 μg/ml; Sigma-Aldrich), and F-actin was stained with TRITC-coupled phalloidin (10 μg/ml; Sigma-Aldrich). For all stainings, specimens were embedded in Epon or Fluoromount-G™ (Southern Biotech) and observed under a Zeiss Axiophot 2 microscope, Zeiss AxioObserver.Z1 inverse microscope, or Leica TCS Sp2 confocal microscope.

Fluorescence in situ hybridization (FISH)

A DIG-labeled rols probe directed at the transcript encoding the C-terminus of Rols7 was synthesized by in vitro transcription of rols cDNA LD1 [43] using a DIG-RNA labeling kit (Roche Diagnostics) and following the manufacturer’s instructions. Formaldehyde-fixed embryos were hybridized in situ with the DIG-labeled rols probe essentially as described in Lécuyer et al. [54]. DIG-labeled rols probes bound to embryos were detected with biotinylated anti-DIG antibody (1:2000; Roche Diagnostics) and the TSA™ Fluorescein System (Perkin Elmer). Fluorescence in situ hybridizations were analyzed by confocal microscopy at various depths within the samples (stacks), which ensured that many ends of the spindle-like FCs were visible in a single picture.

Generation of transgenic flies carrying promoter constructs

rols7 and rols6 reporters were constructed by ligating different parts of the presumptive control regions of the genes to the Escherichia coli lacZ gene and integrating these constructs into pChap to establish transgenic Drosophila lines as reported previously [55]. In addition, the nucleotides sequence encoding the first intron of the rols gene was cloned into the vector pChabHsp43-lacZ [56], yielding rolsIn1-lacZ, which was injected into w ¹¹¹⁸ flies to establish transgenic lines following standard procedures.

Longitudinal and circular visceral muscles differ from somatic muscles in several aspects

The gut musculature of Drosophila larvae is multinuclear and striated [6–8] and comprised of a dense network of circular and longitudinal muscles that can be visualized by scanning electron microscopy [10]. At the light microscopy level, the network can be visualized with the protein-trap allele sls::GFP 48, 49, 57–59], the ECM protein Trol/Perlecan tagged internally with GFP (trol::GFP) 49, 60], and TRITC-coupled phalloidin to visualize F-actin (Additional file 1: Figure S1A–C). Trol/Perlecan is an ECM component; the ECM around the trunk visceral mesoderm (TVM) is required for longitudinal FCs to migrate along the TVM [61, 62].

The morphological data are schematically summarized in Figure 1A–C. At late stage 11, the trunk mesoderm contains on each side of the embryo a row of FCs for the circular visceral muscles and two to three rows of FCMs, which have been proposed to be a common pool for circular visceral muscles and longitudinal visceral muscles. The FCs for longitudinal visceral muscles are determined in the caudal region of the embryo (Figure 1A). At early stage 12 (Figure 1B), the circular visceral muscles are binuclear and arose by one-to-one fusion of a circular FC from the TVM with a visceral FCM [8, 11]. In rp298-lacZ 46], the FC-derived nucleus in the small syncytia remained β-Gal positive, while the nucleus derived from the visceral FCM was β-Gal negative (Figure 1C and Additional file 1: Figure S1C; [11]), as observed earlier in embryos at stage 12. The binucleated circular visceral muscles stretch out in the dorsoventral direction. The FCs of the longitudinal visceral muscles migrate over the stretching circular visceral muscles and reach the anterior part of the trunk mesoderm in early stage 12. At this time, the longitudinal visceral muscles are still mononucleated (Figure 1D) and are surrounded by FCMs (Figure 1B). What happens between this stage and late embryonic stages, with their typical network of circular and longitudinal visceral muscles, is still unknown (Figure 1C).

The longitudinal visceral FCs form multinucleated nascent myotubes during migration and stretch to thin long myotubes at the end of embryogenesis

Since far less is known about myoblast fusion during development of longitudinal visceral muscles than of circular visceral muscles, we focused on longitudinal muscles. We used flies carrying the reporter construct HLH54F-lacZ, in which longitudinal FCs are marked by β-Gal expression [21]. These cells migrate from the circular visceral muscles anteriorly along the TVM from stage 11 until stage 13 [20–22, 24]. When we focused on the time and efficiency of myoblast fusion in these migrating cells, we observed that they were mononucleated when they arrived at the TVM in early stage 12 of development (Figure 1D). At this time point, they formed protrusions, and in some cases, we observed closely adjacent HLH54F-lacZ- positive cells (Figure 1D insert, arrowheads). The longitudinal FCs were arranged dorsally and ventrally of the TVM when they migrated (Figure 1E). At late stage 12, we detected binucleated and trinucleated nascent myotubes, in part connected by thin cytoplasmic bridges (Figure 1F, arrows). By the time the circular muscles had stretched dorsally, the longitudinal FCs were arranged perpendicularly to the circular muscles and already contained several nuclei (Figure 1G and H, arrowheads). At the end of embryogenesis, when the gut was constricted, the longitudinal gut muscles covered the whole midgut evenly. At this stage, β-Gal expression in myotubes appeared in very thin areas, not much wider than the nuclei, with even thinner protrusions between them (Figure 1I).

During longitudinal visceral muscle myogenesis, Duf/Kirre, Rols7, and Blow localize at distinct foci at the sites of fusion

We further analyzed the fusion process of longitudinal FCs and visceral FCMs by using cell-type-specific duf- and sns-reporter constructs. The adhesion molecule Duf/Kirre is expressed in all somatic and visceral FCs, while Sns is expressed in all somatic and visceral FCMs [8, 11]. We observed that in embryos carrying rp298-lacZ, the nuclei of the longitudinal FCs were β-Gal positive at stages when fusion presumably occurred (Figure 2A–A’), concordant with earlier data that showed rp298-lacZ- positive nuclei in migrating longitudinal FCs as well as in the mature longitudinal visceral muscles [8, 11].

To follow the visceral FCMs, we established flies carrying both sns-NLSmCherry 47] and rp298-lacZ. At stages when the longitudinal FCs migrated and were still mononucleated, sns-NLSmCherry-positive FCMs from the somatic mesoderm and the TVM were in proximity to the rp298-positive FCs (Figure 2A”, arrowheads).

To follow the fate of sns-mCherry-expressing FCMs, we established a fly strain carrying sns-NLSmCherry (nuclear signal) and HLH54F-lacZ (cytoplasmic signal). HLH54F-lacZ also allowed us to follow the nascent longitudinal muscles, which increase in size during development (Figure 2B–D). These nascent myotubes were still surrounded by numerous sns-NLSmCherry-expressing myoblasts (Figure 3D). The sns-NLSmCherry signal often appeared to be in the nuclei of the nascent longitudinal myoblasts, which might indicate fusion between FCs and FCMs (Figure 2D, arrowheads); however, due to the numerous sns-NLSmCherry- positive cells in proximity to these nascent myotubes, this was difficult to evaluate. Nevertheless, we hypothesized that the longitudinal FCs fuse with sns-NLS mCherry-positive visceral FCMs.

Consequently, we analyzed the localization of fusion-relevant proteins during longitudinal myogenesis in mid and late stage 13, when fusion takes place to create the longitudinal visceral muscles (see Additional file 2: Figure S2 for circular visceral myogenesis). First, we analyzed whether and where Duf/Kirre and Rols7 are present in the longitudinal visceral FCs (Figure 2E–E”, F–F”). Duf/Kirre was expressed in the longitudinal FCs while they migrated over the circular visceral muscles, in agreement with the expression of rp298-lacZ (compare Figure 2E’ to Figure 2A’). Duf/Kirre often localized with a striking subcellular distribution at both ends of the spindle-like FCs (Figure 2E’, arrow). At a higher magnification, it became evident that Duf/Kirre was limited to those sites of the FC that were in contact with a visceral FCM (Figure 2E”, arrowhead). Next, we used an antibody directed against the first 300 amino acids of Rols7 and detected the protein in the visceral mesoderm (Figure 2F). In part, Rols7 was found in foci at the ends of the spindle-like longitudinal FCs (Figure 2F’). At higher magnification, it was evident that these foci were also at the contact sites with an FCM (Figure 2F”).

This accumulation of Duf/Kirre and Rols7 is comparable to the FuRMAS structure observed during fusion in the somatic mesoderm [32, 42]. The FuRMASs are characterized by the ring-like distribution of Duf/Kirre, Rst/IrreC, Rols, and Sns in FCs and by an F-actin-rich core in the FCMs (for recent reviews, see [1, 29]). To analyze whether F-actin foci appear in FCMs during longitudinal visceral muscle myogenesis, we investigated actin-GFP expression under the control of the twist promoter. Indeed, we observed F-actin foci in those FCMs (Figure 2G”, arrowhead) that contact the longitudinal FCs (Figure 2G’, G” arrow).

Blow is a regulator of WASp-mediated Arp2/3-dependent F-actin polymerization during somatic myogenesis and accumulates as dense foci in FCMs [10, 31, 32, 34]. We found Blow in FCMs that contact the longitudinal visceral FCs (Figure 2H’, H”). At these stages, the spindle-like FCs were about 15 μm in length, while the diameter of the Blow and actin foci was mainly 0.5 to 1.0 μm.

Taken together, these findings suggest that FuRMAS-like structures exist in longitudinal FCs and FCMs. However, Duf/Kirre and Rols were not observed in ring-like structures, in contrast to Duf/Kirre and Rols in somatic FCs, but rather appeared as foci. This difference might be due to the much smaller size of the observed structures compared to those in somatic myoblasts.

The rols7 transcript localizes in the longitudinal FCs before fusion

The rols gene is regulated by two promoters, which leads to rols7 and rols6 transcripts with specific 5’ exons [43]. We investigated which isoform of Rols is required in longitudinal FCs. We analyzed the promoter regions responsible for transcription of rols7 and rols6, focusing on longitudinal myogenesis.

Indeed, the rols7 promoter contained distinct regulatory regions for transcription in the somatic mesoderm and visceral mesoderm (see Figure 3A for a summarizing scheme). An intron between exons 1 and 2 of rols7 controlled transcription in the circular visceral mesoderm and during development of longitudinal muscles (Additional file 3: Figure S3, rolsIN1-lacZ reporter). To clarify the situation in longitudinal visceral myogenesis in more detail, we used these rolsIn1-lacZ transgenic embryos stained with fluorescent antibodies against β-Gal and the cell surface glycoprotein Fasciclin III (FasIII, [51]) to label the membranes of the trunk mesoderm. In embryos expressing rolsIn1-lacZ, the longitudinal FCs clearly expressed β-Gal when they migrated along the TVM in mid-embryogenesis (Figure 3B, B’, arrowheads). Also later, when the circular muscles stretched dorsally, β-Gal was expressed in the longitudinal FCs that were already binucleated (Figure 3C). Thus, the rolsIN1-lacZ reporter allowed us to follow the fusion stages during longitudinal visceral myogenesis.

Rols7 was detected at distinct foci at both ends of the spindle-like FCs in the developing longitudinal visceral muscles (Figure 2F–F”). To investigate whether this particular localization of Rols7 is regulated at the level of rols7 mRNA localization in the longitudinal FCs at this stage, we hybridized bHLH54F-lacZ embryos in situ with fluorescent probes aimed against rols7 mRNA and stained with antibodies against β-Gal to visualize FCs (Figure 3D–D”). In the longitudinal FCs, rols7 transcripts were mostly concentrated in speckles, and often towards the tips of the spindle-shaped cells, which indicated a targeted distribution of rols7 mRNA during fusion (Figure 3D’ and D”, arrows).

Rols7 is required for fusion but not for orientation or migration of longitudinal FCs

Since rols7 mRNA partially localized to the polar, presumptive sites of fusion in the longitudinal FCs, and since the Rols7 protein localized at distinct foci during longitudinal visceral myogenesis, we then asked whether rols7 is required for fusion in both circular and longitudinal visceral myogenesis. We analyzed rols-deficient embryos expressing different reporter constructs that mark visceral FCs or FCMs. First we used bap-lacZ transgenic lines to distinguish between unfused visceral FCMs and unfused somatic FCMs. The transcription factor Bagpipe (Bap) is expressed in all visceral myoblasts of the trunk mesoderm. After fusion of circular FCs with neighboring FCMs to form binucleated syncytia, only a small number of remaining unfused FCMs can be detected directly beneath and above the stretching circular myotubes (Figure Three I–L in [11]). Klapper et al. 11] suggested that these remaining FCMs fuse with the migrating longitudinal FCs. Thus, in wild-type embryos, bap-lacZ mainly marks the circular visceral muscles directly after fusion (Figure 4A) and β-Gal was later expressed in the visceral muscles of the midgut (Figure 4E). In rols7 mutant embryos, unfused β-Gal- positive visceral myoblasts were located in the interstitium between the somatic mesoderm and visceral mesoderm (Figure 4F).

To examine the origin of these unfused visceral myoblast cells, we analyzed rols mutant embryos at earlier stages of visceral muscle formation. rols mutants exhibited more unfused FCMs, which indicated a visceral fusion defect (compare Figure 4A and B), in agreement with expression of Rols7 in circular (Additional file 2: Figure S2B’) and longitudinal visceral FCs (Figure 2F–F”). However, rols mutant circular muscles stretched normally in the dorsal direction (Figure 4B and G) and the overall circular muscle morphology visible with bap-lacZ and other markers, such as FasIII, appeared to be mostly regular with only sporadic small gaps (Figure 4G, arrowhead). We conclude that these gaps are the result of minor failures in circular myoblast fusion and that the majority of unfused myoblasts (Figure 4F) result from failure in longitudinal visceral myotube formation.

This conclusion was supported by the midgut morphology of rols mutant embryos in late stages of embryogenesis. Using anti-β3-Tubulin to visualize midgut muscle morphology, we observed a chambering defect between the 1^st and 2^nd midgut chamber in rols mutants during late embryogenesis (Figure 4F, arrowheads). This defect resembles the midgut phenotype of known circular visceral fusion mutants [11, 25], although in the rols mutants, the defect is most likely due to the malformation of the longitudinal visceral muscles.

Since the circular muscles were only slightly affected in rols mutants, we focused on the development and fusion process of the longitudinal gut muscles by analyzing the differentiation of longitudinal FCs in rols mutant embryos carrying the reporter construct HLH45F-lacZ 21]. We found that the longitudinal FCs migrated correctly along the circular muscles during mid-embryogenesis and were arranged dorsally and ventrally to them (compare rols mutant in Figure 4G with wild-type in Figure 1E). However, shortly before constrictions formed, the cells did not align entirely perpendicular to the circular muscles; instead, they formed protrusions in other directions and were mainly mononucleated (Figure 4C, arrowheads) at a stage when binucleated and trinucleated syncytia were detectable in the wild-type (Figure 1F). At later stages, we observed gaps between the normally evenly distributed cells (compare rols mutant in Figure 4H–J to the wild-type in Figure 1I). We sometimes detected mononucleated cells with protrusions stretching in all directions, but rarely detected stretched and binucleated cells in the anterior part of the gut. Interestingly, the posterior part of the gut was still surrounded by dense stripes of longitudinal muscles in late embryogenesis (Figure 4H and J). When we looked at the cells at a higher magnification, we observed mainly binucleated cells with protrusions orientated in the correct anterior–posterior direction (Figure 4D).

In summary, although the longitudinal FCs migrated correctly, fusion of these cells was disturbed in rols mutant embryos. As a consequence, only binucleated syncytia were detected at the end of embryogenesis, which could also indicate a delay in their fusion. In the anterior part of the gut, the phenotype was more severe, and the longitudinal muscles mainly comprised mononucleated cells. Notably, longitudinal muscle fusion proceeded in rols mutant embryos, but in analogy to somatic muscle fusion, the fusion process may be limited to the first fusion step that gives rise to binucleated longitudinal syncytia.

Longitudinal muscle development requires Lame duck

The requirement for Rols7 in longitudinal muscle fusion could indicate a broader similarity between myoblast fusion in somatic and visceral longitudinal myogenesis. Therefore, we analyzed longitudinal visceral muscle development and midgut morphology in the background of mutations that disturb somatic muscle fusion at different fusion-relevant steps.

Lame duck (Lmd), a homolog of the Gli family of transcription factors in Drosophila, is an essential regulator during FCM specification in somatic myogenesis, and thus has an indirect effect on fusion by regulating the expression of sns 63]. We previously identified lmd ^E202 as a new allele of lame duck by screening for genes relevant for myogenesis ([19, 34]; Holz and Renkawitz-Pohl, unpublished data). An analysis of longitudinal visceral muscle development in homozygous lmd ^E202 mutant embryos with the HLH54F-LacZ reporter and β3-Tubulin antibodies (Figure 5A–C) revealed only mononucleated lacZ-positive cells, which indicated that Lame-duck-dependent specified FCMs are also required for longitudinal visceral muscle formation.

In contrast, longitudinal visceral muscle migration and later spreading as well as protrusion formation appeared unaffected in mutant embryos at stages 13 and 14/15 (Figure 5A and B), although morphological defects could be detected within the underlying circular muscle strands. At the end of embryogenesis (Figure 5C), midgut chambering and constriction formation remained incomplete in lmd ^E202 embryos, reflecting a visceral phenotype as already observed in sns mutant embryos [25].

Longitudinal muscle development requires the F-actin-regulating proteins Myoblast City, Blow, and Kette

Since we found FuRMAS-like structures with actin foci during fusion of longitudinal FCs with FCMs (Figure 2E–H), we analyzed regulators of F-actin polymerization required for somatic myoblast fusion. Mbc is a guanine nucleotide exchange factor (GEF) for the Rac1-GTPase, which is involved in the activation of the Scar/WAVE complex [64]. The loss of Mbc function leads to a block of somatic myoblast fusion, and Mbc accumulates together with Rac1 in actin foci [47]. Furthermore, Mbc is necessary and sufficient in FCMs for myoblast fusion [47]. However, longitudinal visceral muscle migration appeared to be unaffected in transheterozygous HLH54F-lacZ; mbc ^C11 /mbc ^D11.2 embryos (Figure 5D). In contrast to the situation in the wild-type, in the mutant, the longitudinal FCs along the remaining visceral mesoderm were not only still mononucleated at stage 13/14, but also were roundish and had shorter protrusions (Figure 5D), in agreement with data from dye injections into mbc mutant muscles [8]. In late embryonic development, the first midgut constriction was missing in mbc mutant embryos (Figure 5F). Filamentous protrusions of the longitudinal FCs were then clearly visible at the posterior half of the midgut, while only a few longitudinal FCs with shorter, randomly orientated protrusions were visible in the anterior midgut regions.

Mutations in numerous other members of the Scar/Wave and WASp complexes involved in the Arp2/3-dependent F-actin polymerization machinery induce characteristic fusion defects in the somatic musculature. However, initial studies have only revealed a minor influence of these factors, e.g., Blow and Kette (also referred to as Hem-2 or Nap1), on circular visceral muscle development [10, 38]. Therefore, we asked whether these genes are also dispensable for myoblast fusion to form the longitudinal visceral muscles.

While Blow was expressed in the FCMs during fusion and longitudinal myogenesis (Figure 2H–H”), kette was broadly transcribed in the embryo until stage 14, and is also maternally contributed [45]. We analyzed whether gut constrictions were correctly formed in kette and blow mutants. Although the gut chambers were constricted, their proportions differed from that of the wild-type (not shown). We then asked whether the longitudinal muscles develop correctly in blow and kette mutants. In HLH54F-lacZ; blow ² mutants (Figure 5G and I) and in transheterozygous HLH54F-lacZ; blow ¹ /blow ² embryos (Figure 5H), longitudinal FCs migrated along the circular mesoderm, although not only dorsally and ventrally, but also along the whole TVM. Moreover, these cells were mononucleated at this time and were rounder than those in wild-type embryos. At later stages, protrusions of the longitudinal FCs stretched not only in anterior–posterior directions, but also in dorsal–ventral directions (Figure 5I).

When we stained HLH54F-lacZ; blow ² mutant embryos with an antibody against DMef2, which marks the nuclei of all muscle cells, we observed that cells in the anterior midgut region were mononucleated before the gut became constricted (Figure 5M, arrow). After constriction formation, cell protrusions were much shorter than in the wild-type, and gaps appeared between the neighboring cells (Figure 5I). Although we observed that the phenotype was less severe in the posterior midgut regions, HLH54F-lacZ- positive cells remained mainly mononucleated, and only occasionally were binucleated cells found (Figure 5H).

Longitudinal FC migration along the TVM was also observed in transheterozygous HLH54F-lacZ; kette ^J4–48 /kette ^G1–37 embryos, although some HLH54F-lacZ- positive cell were localized aberrantly (Figure 5J, arrows). During later stages, longitudinal FCs formed protrusions that were mainly oriented in the correct anterior–posterior direction (Figure 5K, double arrows). At this stage, some HLH54F-lacZ-positive myoblasts appeared to be binucleated. However, double staining with anti-DMef2 revealed mainly mononucleated and a few binucleated HLH54F-lacZ- positive myoblasts in kette ^J4–48 mutant embryos at this time (Figure 5N, arrow). At the end of embryogenesis, although the longitudinal FCs had formed thin cell protrusions and aligned along the midgut, overall fewer elongated longitudinal FCs appeared to be present, and some areas along the circular muscles were not covered with longitudinal muscles (Figure 5L).

Scar/Wave is mainly essential for longitudinal visceral fusion

Mbc and Kette are both involved in Scar/Wave-dependent activation of the Arp2/3 complex during myoblast fusion [10, 31, 47]. To determine whether also scar is involved in longitudinal FC and visceral FCM fusion, we analyzed midgut constriction formation in scar ^Δ37 mutant embryos. Gut constriction formation and gut morphology showed no defects in comparison to wild-type embryos (Figure 6A and B). scar is also maternally transcribed, and maternal scar compensates for the loss of zygotic scar in somatic myoblast fusion. Furthermore, scar cooperates with WASp-dependent Arp2/3 regulation during somatic myoblast fusion, and in this system fusion is only blocked completely in scar ^Δ37 wip ^f06715 double mutant embryos [30, 36]. Our findings that blow mutants also displayed longitudinal visceral fusion defects further imply an involvement of WASp, as recent studies have shown that Blow competes with WASp for Wip-binding [31]. However, wip single mutants, Arp3 ^schwächling single mutants, and Arp3 ^schwächling wasp double mutants did not show severe defects in longitudinal visceral muscle formation (Additional file 4: Figure S4B, C and I–I”), which might be due to maternal contribution in the case of Arp3 and wasp. However, at stage 13, some longitudinal FCs showed aberrant cell migration (Additional file 4: Figure S4H, arrow). Nevertheless, at the end of embryogenesis, gut morphology and constrictions almost like those in the wild-type appeared, and at least trinucleated muscles were observed (Additional file 4: Figure S4I–I”). This finding may indicate that WASp-dependent Arp2/3 activation is involved in the migration of longitudinal FCs.

To analyze the fusion defects of scar ^Δ37 wip ^f06715 double mutants more closely, we examined gut constriction formation (Figure 6C) and introduced the HLH54F-GFP marker for longitudinal FCs into scar ^Δ37 wip ^f06715 double mutants (Figure 6D, F). We observed migrating, mostly mononuclear FCs in stage 13 embryos (Figure 6D, arrow), some of which showed aberrant migration positions (Figure 6D, arrowhead). The fusion defects were severe; we detected mainly mononucleated, binucleated, and occasionally trinucleated HLH54F-GFP-positive cells (Figure 6F, arrowheads). Next we asked whether removing a copy of wip influences the scar phenotype (Figure 6G, H). The FCs at stage 13 were mainly mononucleated during their migration (Figure 6G); we observed fewer FCs at aberrant positions than in embryos with two mutant copies of wip (Figure 6D). At stage 16, longitudinal muscles were mononucleated and binucleated (Figure 6H, arrowheads). These muscles showed long extensions and a relatively parallel arrangement in the posterior part of the embryo (Figure 6H). Next we analyzed wip homozygous mutants with one mutant copy of scar ^Δ37 (Figure 6I, J). In the wip homozygous embryo, many FCs were at aberrant positions (Figure 6I, arrowhead) and showed aberrant protrusions (Figure 6I, arrow). Muscles at stage 15/16 were mainly mononucleated or binucleated. We concluded that Scar/Wave is an important regulator for longitudinal fusion. The analysis of scar wip double mutants further revealed that Wip enhances the defects in longitudinal myogenesis. Because wip single mutants did not show any defects, it needs to be clarified whether Wip influences indeed fusion or other processes during longitudinal visceral myogenesis, e.g., migration or stretching.

In conclusion, fusion of the longitudinal FCs with FCMs was severely disturbed in rols, mbc, blow, kette, and scar mutant embryos. Our results also revealed significant differences in protrusion formation between these mutants. While cell protrusions frequently formed, but without longitudinal syncytia formation, in lmd mutants and in the posterior midgut regions of mbc mutant embryos, blow and mbc mutant embryos had shorter protrusions and abnormal orientation of longitudinal FCs at the anterior midgut. Importantly, the analyses of scar mutants indicated that the Arp2/3 activator Scar/Wave but not WASp is mainly essential for the fusion of longitudinal myoblasts and suggested that blow may act in this context independently of WASp/Wip-dependent Arp2/3-based actin polymerization.

FuRMAS-like cell-adhesion structures with F-actin foci form during longitudinal visceral myoblast fusion

At the site of fusion, Duf/Kirre and Rols7 were localized in FCs, and foci of Blow and F-actin-GFP were localized in FCMs. These findings correspond to the FuRMAS structure including the characteristic F-actin-foci observed during somatic myoblast fusion. However, we found distinct differences in the need for and/or function of Scar/Wave and WASp complexes in the two types of myoblast fusion.

Longitudinal visceral fusion appears to be mainly dependent on Kette and Scar/Wave

The analyses of single and double mutants that possess defects in Arp2/3-induced actin polymerization demonstrated that Scar-dependent Arp2/3 activation is essential for longitudinal myoblast fusion. It has been proposed that the activation of Scar/Wave depends on the small Rac-GTPase [64]. The myoblast-fusion-relevant GEF for Rac is Mbc [47]. Rac involvement is likely required in longitudinal fusion, as longitudinal myoblasts did not fuse in mbc mutants. Furthermore, Kette was essential for fusion, in agreement with the function of Kette as part of the Scar/Wave complex. However, the scar phenotype was enhanced when also wip was deleted. Although, wip single mutants did not display longitudinal fusion defects, Wip and its interaction partner WASp might be required for the migration of longitudinal FCs or stretching of longitudinal muscles.

In contrast to mbc, kette and scar, we detected only minor defects in the Arp2/3 subunit mutant arp3 ^Schwächling . We hypothesized that highly maternally contributed arp3 mRNA might be sufficient to allow longitudinal fusion. This is in contrast to somatic myoblast fusion [35, 36] and axonal pathfinding during CNS development [65]. We further found that gut constrictions were not indicative for loss of longitudinal visceral fusion (Figure 6A–C).

Blow might act independently of WASp/Wip during longitudinal visceral fusion

During somatic myoblast fusion, Blow is needed to stabilize the WASp/Wip complex, which in turn activates the Arp2/3 complex [31]. Therefore, Blow may act independently of the WASp/Wip complex in the formation of the longitudinal muscles. Furthermore, Blow genetically interacts with Kette during somatic myoblast fusion [34], which indicates a WASp/Wip-independent function of Blow.

Different relative positions of FCs and FCMs characterize fusion of circular visceral, longitudinal visceral, and somatic muscles

Differences in the relative position of FCs and FCMs have to be kept in mind when comparing the three different myoblast fusion events in the embryo (Figure 7). The circular muscles originate from a row of FCs fusing with the adjacent row of FCMs, giving rise to a binucleated cell. Thus, fusion takes place in an epithelial-like situation (Figure 7A; only the row of FCMs adjacent to the FCs is shown). However, in longitudinal visceral and somatic myogeneses, FCs and FCMs are not arranged like an epithelium, but rather FCs attract FCMs to the site of fusion [26, 27]. Importantly, longitudinal fusion takes place during migration of FCs, which means that defects in migration may mimic a fusion phenotype.

The complexity of known regulators of myoblast fusion increases from circular to longitudinal to somatic myoblast fusion

Just as in the somatic mesoderm, in the visceral mesoderm, fusion of myoblasts depends on the presence of the immunoglobulin super family receptors Duf/Kirre and/or Rst/IrreC in the FCs, and Sns in the FCMs (Figure 7). In both duf,rst double mutants and sns mutants, the FCMs fail to adhere to the circular FCs and fusion is completely blocked. Therefore, the adhesion molecules appear to mediate the heterotypic adhesion between these two cell types in all embryonic myoblast fusions [8, 11, 66, 67]. Mutants of the GEF-encoding gene mbc displayed mononuclear circular and longitudinal visceral myoblasts as well as somatic myoblasts; thus, also Mbc appears to be essential in all myoblast fusion events.

The first differences in the myoblast fusion events became apparent when we analyzed the adaptor Rols, which is specifically expressed and binds to Duf/Kirre in the FCs. Rols was not essential for the formation of binucleated circular visceral muscles (Figure 7A) but was essential for visceral longitudinal and somatic fusion (Figure 7B, C). We found that Duf/Kirre and Rols localize to the spindle-like FCs during longitudinal visceral muscle myogenesis (Figure 7B), which suggested an interaction as in somatic myogenesis (Figure 7B and C). In rols mutants syncytial longitudinal muscle development was disturbed, and we found mostly binucleated mini-muscles. The first fusion events of the longitudinal FCs might function without Rols7, analogous to the first phase of fusion in the somatic mesoderm, which is independent of Rols7 function [40, 41, 43]. In longitudinal visceral fusion, the adaptor protein Rols7 might bind to Duf/Kirre to allow efficient fusion via a positive feedback loop, as is the case in the somatic mesoderm [68].

Actin foci are typical at the sites of longitudinal visceral and somatic fusion. The Scar/Wave and WASp/Wip complexes bind and activate the Arp2/3 complex and thereby enable actin polymerization during somatic myoblast fusion [30, 33, 36, 37, 69]. Our study and other studies showed that actin regulator molecules that perform an essential function during somatic fusion (Figure 7C) play a different role in morphogenesis of the gut muscles, even though they are expressed in the visceral mesoderm. In blow and kette mutants, stretching and outgrowth of the circular visceral muscles is disturbed, although fusion itself is not affected [10]. This is in contrast to the situation during longitudinal visceral fusion, where Blow, Kette, and Scar/Wave were required (Figure 7A, B). In the case of kette and scar mutants, it should be considered that a pool of both transcripts is supplied maternally ([45, 65]) and that this pool might allow the observed limited number of fusions in homozygous kette and scar mutants.

Both circular and longitudinal visceral myoblast fusion seemed to be independent of the actin regulators WASp and the WASp-interacting protein Wip (our results) and do not exhibit any obvious defects in gut muscle development [38]. However, WASp and Wip influence the migration of longitudinal FCs.