rolling pebbles (rols) is required in Drosophila muscle precursors for recruitment of myoblasts for fusion

In Drosophila, the establishment of muscle founder cells prefigures the larval pattern of body wall muscles (Bate and Martinez Arias, 1993). These cells are descendants of progenitor cells, which are determined by the Notch pathway via lateral inhibition (Carmena et al., 1995) and specified by signaling pathways such as the EGF receptor pathway (Buff et al., 1998). Each progenitor divides asymmetrically and gives rise to two founder cells, each of which expresses a distinct set of transcription factors and is determined to build a specific myofiber (Ruiz-Gomez, 1998; Frasch, 1999; Paululat et al., 1999a).

Founder cells first form bi- or trinucleated cells called precursor cells by fusion to the second class of mesodermal cells, the fusion-competent myoblasts (FCMs) and then enlarge by further fusion to form mature myotubes (Bate, 1990). Recently, Ruiz-Gomez et al. (Ruiz-Gomez et al., 2000) analyzed dumbfounded (duf), which encodes a member of the immunoglobulin family and presumably serves as an attractant in the founder cells. duf was the first gene shown to be essential in the founder cell to mediate already the first fusion to muscle precursor cells. These precursor cells then recruit additional FCMs from a common pool in the surrounding mesoderm to form myofibers. In Drosophila, myotube formation is finished within a few hours.

At the ultrastructural level, distinct steps in the fusion event are evident (Doberstein et al., 1997). First, cells recognize each other and align. At the area of fusion where the cell membranes will later be degraded, vesicles accumulate between the individual precursors and the FCMs. This prefusion complex dissolves and is replaced by electron-dense plaques of unknown composition; afterwards membrane breakdown occurs and the cells fuse. The electron microscopic studies of Doberstein et al. (Doberstein et al., 1997) were carried out with stage 13 and stage 14 embryos at a time when the majority of precursor cells have been established but abundant fusion to mature muscles still takes place.

Genetic approaches uncovered several genes essential for myoblast fusion (Paululat et al., 1999b; Frasch and Leptin, 2000; Taylor, 2000). The ultrastructural analysis of Doberstein et al. (Doberstein et al., 1997) revealed that mutations in individual fusion relevant genes interrupt myoblast fusion at distinct steps with regard to cell adhesion, prefusion complex and plaque formation. The gene myoblast city (mbc) is expressed in, besides other tissues, all myoblasts and is here required during an early step to mediate proper recognition and alignment between the founder cells and the FCMs (Rushton et al., 1995). Mbc shows homologies to the human DOCK180 protein and interacts with the Drosophila homolog of CRK, an adaptor protein (Erickson et al., 1997; Doberstein et al., 1997; Galletta et al., 1999). However, as it is present only in the cytoplasm, its precise function is still unclear. The same holds true also for Blown Fuse (Blow), a protein that is uniformly distributed in the cytoplasm of all myoblasts and essential to proceed from prefusion complex formation to plaque formation, all the more as it shows no homologies to proteins identified so far (Doberstein et al., 1997). In sticks and stones (sns) mutants, which correspond to the 43-49 fusion mutant originally found on the rost^P20 chromosome (Paululat et al., 1999b; Bour et al., 2000), myoblasts accumulate after plaque formation (Doberstein et al., 1997). SNS is expressed in FCMs, but not in founder cells, and encodes a 162 kDa nephrin-like protein that crosses the membrane and contains Ig repeats in the extracellular domain (Bour et al., 2000). Overexpression of a dominant negative form of the GTPase Drac1 leads to incomplete membrane breakdown (Luo et al., 1994). Interaction between Mbc and Drac1 may be required to rearrange the cytoskeleton during myoblast fusion as Mbc has been isolated independently as a suppressor of Drac1 (Nolan et al., 1998). Both proteins are involved in a number of developmental processes and are clearly essential, but not specific, for myoblast fusion.

Despite considerable insight into several steps of fusion, only a limited number of components and their interactions are known at the molecular level. An important question is whether the establishment of muscle precursor cells by initial fusion and further growth to muscles by recruiting further fusion competent cells is just a timewise difference, or whether distinct genes are required to proceed from the precursor to the mature muscle. We present data that suggest rolling pebbles (rols) might be a key component specifically required in the muscle precursor cell to recruit surrounding myoblasts for fusion. Alternatively the loss of Rols might reduce the efficiency of fusion so that only mini-muscles are formed.

The rols relevant deficiency line Df(3L)BK9 (which corresponds to Umea stock #54650) as well as the enhancer trap lines P1027 and P1729 (Spradling et al., 1995) were obtained from the Bloomington and Umea Stock Centers. P1729 corresponds to stock numbers 11729 and P{PZ}l(3)08232, and carries a P-element at 68F01-03; P1027 corresponds to stock number 11027 and P{lArB}A490.2M3, the P-element was placed in 68F. The alleles rols^AD328 and rols^XX117 were found in an EMS mutagenesis screen for the 3rd chromosome carried out by Skeath and Doe in which the FGF-receptor mutant heartless was found (Gisselbrecht et al., 1996).

rols^AD328, rols^XX117, rols^P1027 and rols^P1729 were all allelic, as determined by failure to complement each other and the deficiency Df(3L)BK9. rols alleles were kept over a marked balancer TM3 Sb Dfd-lacZ to distinguish homozygous mutants and heterozygous embryos at early stages. In the rP298 enhancer trap line, the P-element is localized on the X-chromosome (Nose et al., 1998) and the enhancer trap pattern corresponds to dumbfounded (Ruiz-Gomez et al., 2000). By appropriate crosses we established a fly stock carrying the rols deficiency on the 3rd chromosome over the marked balancer TM3 Sb Dfd-lacZ and the rP298 insertion on the X chromosome.

The P-element in rols^P1729 and in rols^P1027 was remobilized by using the transposase source of the line ry⁵⁰⁶P[ry⁺Δ2-3]/ry⁵⁰⁶P[ry⁺Δ2-3]. The loss of the P-element (visible by loss of rosy⁺ marker) was detected by crossing single jumpstarter males with females of the balancer line 120 (Bloomington Stock Center). Isogenic individual rosy lines were analyzed concerning lethality, phenotype and allelic situation to the deficiency Df(3L)BK9. As we noticed a second lethal hit on the P1027 chromosome (outside the breakpoints of the deficiency), we crossed this strain five times (Davies et al., 1996) against the ry⁵⁰⁶ allele and named the corresponding strain with the P-element in the 68F locus after rebalancing rols^P1027ber. Although no second hit was observed in rols^P1729, the P1729 line was crossed in the same way (and named rols^P1729ber).

Immunostaining was performed as described previously (Paululat et al., 1995). We used antibodies against β3 tubulin (Leiss et al., 1988; Buttgereit et al., 1996) and against Mef2 (Bour et al., 1995) to stain mesodermal derivatives and developing muscles. A polyclonal anti β-galactosidase antibody (Biotrend, dilution of 1:5000) was used to visualize muscle precursors of the enhancer trap line rP298, the expression of the marked balancer, and the enhancer trap pattern in the rolling pebbles alleles rols^P1027 and rols^P1729. Anti-Eve (Frasch et al., 1987) was used to selectively stain dorsal acute muscle 1 (DA1) and pericardial cells. The Vectastain ABC Elite-kit (Vector Laboratories) was used as detection system. The embryos were embedded in glycerol or dehydrated and embedded in Epon and photos were taken under Normarski optics with a Zeiss Axiophot microscope, digitalized and processed in Adobe Photoshop.

Whole mount in situ hybridization was carried out essentially as described by Tautz and Pfeifle (Tautz and Pfeifle, 1989). We synthesized antisense DIG-labeled probes by single strand PCR using primers specific for the rols6 transcript (st-Primer: 5′ ACTGCCAGATCGACGACCGGTTG) or the rols7 transcript (lt-primer: 5′ TCCAGCGAGGAGTTCATAGAT), respectively. In addition, for sns we used antisense probes from a 6 kb sns cDNA clone in pBSKII kindly provided by Susan Abmayr (Bour et al., 2000).

The electron microscopic analysis was carried out as described by Doberstein et al. (Doberstein et al., 1997) using the rols^AD328 allele.

Plasmid rescue was carried out using standard molecular biology methods. Genomic DNA of rols^P1729 flies was digested with XbaI, ligated overnight and transformed into high efficiency ultra competent cells (XL2-Blue, Stratagene). Thus, we obtained a 5 kb genomic fragment flanking the P-element insertion. Accordingly, we obtained a 4 kb genomic fragment by digesting genomic DNA of rols^P1027 flies using SalI.

A Canton-S genomic library (Stratagene, made from 0-12 hours embryos) was screened using the 5 kb plasmid rescue fragment as an initial probe. Several overlapping clones were isolated, subsequently subcloned into the Bluescript plasmid vector (Stratagene) and used for additional screenings. A 5′ extension of 3 kb was obtained by the analysis of the P1 phage DS00075 (Kimmerly et al., 1996) so that the different phage clones finally spanned a genomic region of 60 kb.

We isolated three cDNAs from the embryonic LD-cDNA library from the BDGP EST Project (made by Ling Hong) by using a genomic 3 kb probe from the most 5′ region of the rols transcription unit. Two of these three cDNAs were identical (6.3 kb), whereas the third one, 5.8 kb length, differed in the 5′ region.

To determine exon/intron boundaries we compared the sequences of the rols cDNAs to genomic sequences. For this purpose, we took advantage of the already published genomic sequences of Drosophila (Adams et al., 2000) and used the independent sequenced genomic fragments from our library screening. To search for protein domains, we applied the SMART-Tool of the EMBL (Schultz et al., 1998; Schultz et al., 2000).

Sequence data for rols7 and rols6 have been deposited with the EMBL/GenBank Data Libraries under Accession Numbers AF386647 (rols7) and AF386648 (rols6).

Primers were designed to amplify a common part of the rols mRNA leading to a rols PCR fragment of 580 bp from the mRNA and a genomic fragment of 830 bp (primer LDI-14, CCACCATCGCTATGCGAGC; and LDI-20, CTTCAGGTTGTGATTGCTCACG). A primer pair specific for rols7 leads to a 705 bp fragment (primer LDA, 5′ GCCACATTGGATTATCAGTGA; primer LDR, 5′ CATGATGTCATTCCGATTGAAG flanking the rols7 specific intron). A rols6 specific combination leads to a 708 bp fragment (primer LDIIDO, 5′ TCGGCAAATCATCGCCGATAC; and primer LDIISP, 5′ACTGCCAGATCGACGACCGGTTG, flanking the small rols6 specific intron). A primer pair (B1E1D, CCAAGCTGGTCAGTGC; and B1E2U, GAGCCTCGTTGTCGATG) was used to amplify a 601 bp fragment of the β1-tubulin mRNA. All primers were chosen so that they flanked an intron in the pre-mRNA, to distinguish amplification from genomic DNA. As the β1-tubulin fragment (601 bp) is very similar in size to the rols PCR-product, β1-tubulin primers were omitted in the rols PCR reactions, but included as internal control in the rols6 and rols7 reactions. Per RT-PCR reaction with approx. 20 ng poly (A⁺) RNA (isolated with the Qiagen One-step kit) amplifications were carried out at the appropriate annealing temperature with 30 to 35 cycles.

As donor for transplantation, we used the offspring of UAS-lacZ/UAS-lacZ; rols^Df(3L)BK9 /TM3 Sb Dfd-lacZ. The rols allele Df(3L)BK9 was marked with UAS-lacZ on the 2nd chromosome (P1776 from the Bloomington Stock Center, this Drosophila stock was additionally marked with Ubx/Sb on the 3rd chromosome) and balanced over TM3 Sb Dfd-lacZ. The strain daughterless GAL4 (daGAL4) with ubiquitous GAL4 expression (Wodarz et al., 1995) (supplied by R. Klapper) served as a recipient (Klapper, 2000). All transplantation experiments were carried out during blastoderm stage (Holz et al., 1996) and beginning gastrulation. Mosaics were generated by removing approximately 25 cells from the mesodermal anlage of donor embryos and transplanting three to five cells homotopically in up to six host embryos. Transplantations were carried out between 40-60% EL (egg length). To select hetero- and homozygous donor embryos, single developed donors were dissected and stained for β-galactosidase activity achieved by the TM3 Sb Dfd-lacZ- balancer. All host embryos with cells from an individual donor were raised together in culture and dissected as third instar larvae. Spreading of the musculature was achieved by briefly dipping the larvae into a 60°C water bath (Hooper, 1986). Histochemical demonstration of β-galactosidase expression was carried out as described by Meise and Janning (Meise and Janning, 1993). The detection of β-galactosidase expression was only possible in syncytial tissues where UAS-lacZ and daughterless Gal4 are co-expressed. This method of syncytium detection was established by Klapper et al. (Klapper et al., 2001a).

We screened a collection of deletion mutants for defects in muscle formation to identify novel genes required for myoblast fusion (Bloomington and Umea stock collection, about 70% of the collection for 3rd chromosome). The mesoderm specific β3-tubulin isotype is expressed before myoblast fusion (Buttgereit et al., 1996). Thus, we used a β3-tubulin specific antibody to visualize mesodermal cells in mutants. This strategy has been successfully applied for P-element-induced mutations described previously (Burchard et al., 1995; Paululat et al., 1995). Most of the deletion mutants show severe distortions (e.g. in segmentation). However, embryos homozygous for a small deletion on the left arm of the third chromosome, Df(3L)BK9 exhibited a clear muscle phenotype. This deletion maps to the cytological positions 68E2-3; 69A1. Mesoderm differentiation is severely disturbed compared with wild-type embryos (compare Fig. 1C-H with Fig. 1A,B); most reliably, we detect clusters of unfused myoblasts, which fail to form myofibers at late stages of embryogenesis (Fig. 1D,F). We named the transcription unit responsible for the fusion defects rols, owing to the appearance of the scattered distribution of the unfused myoblasts.

The relative size of the deletion suggests that many genes are removed. To identify the gene responsible for the mutant phenotype, we analyzed lethal P-element inserts located within the deleted region for defects in muscle formation. We identified two P-element insertions, P1027 and P1729 from the Bloomington collection, which lead to embryonic lethality and show the same muscle fusion phenotype as the deficiency. The original P1027 chromosome contains two mutations leading to lethality in the homozygous situation. These mutations were separated by meiotic recombination and the strain with a single P-element insertion at 68F was named rols^P1027ber. Embryos homozygous for rols^P1027ber and as well as for rols^P1729ber (a cleansed rols^P1729 chromosome, see Material and Methods) show strong fusion defects (Fig. 1H). As rols^P1027 fails to complement rols^P1729, we concluded that these two P-element inserts were allelic. Additionally, remobilization of both P-elements in the cleansed P-element insertion lines restored viability, strong evidence that these P-elements disrupt the function of the rols gene (data not shown).

Furthermore, we identified two EMS-induced alleles from the collection of Skeath and Doe (Gisselbrecht et al., 1996), which failed to complement the deficiency as well as the P-alleles. As they exhibit an identical fusion phenotype, we named them rols^AD328 and rols^XX117. Looking at the muscle morphology, fusion was severely disturbed as in all other rols alleles (Fig. 1G). Before dorsal closure, we often observed many unfused myoblasts per segment. In embryos at stage 16/17, only a small number of irregularly shaped myofibers were present, leading to a very rudimentary muscle pattern and a varying proportion of unfused myoblasts (see Fig. 1C-H). The disappearance of many myoblasts might be explained in part by cell death of unfused myoblasts which are cleared away by macrophages, as previously observed in another muscle fusion mutant (Rushton et al., 1995).

Embryos homozygous for the mutant rols alleles develop until shortly before hatching as dorsal closure is evident (shown for Df(3L)BK9, for example) (Fig. 1E). Many unfused myoblasts were present and we found some muscle-like fibers with only a few nuclei (Fig. 1C-H). We propose that these muscle-like fibers represent muscle precursor cells that stretched and tried to contact the epidermis, as originally observed for founder cells in mbc mutant embryos (Rushton et al., 1995). The persisting myoblasts often adhere to the myofiber-like cells (see Fig. 1D,F). Furthermore, we often observe that these myoblasts extend filopodia, which are directed towards muscle-like fibers. However, the extent of the phenotype is variable mainly after stage 16; the number of unfused myoblasts, and the appearance and number of myofibers also differ significantly between embryos of the same stage. Cardioblast development is not obviously disturbed, as the typical repetitive pattern of four β3 tubulin stained cardioblasts and two unstained cardioblasts is evident (Fig. 1E,F,H). Analysis of gut morphogenesis often reveals incomplete formation in at least a quarter of the mutants when compared with the wild type (data not shown), which might be evidence for defects in the visceral muscles of the midgut. Recently, it was shown by Klapper et al. (Klapper et al., 2001b) and San Martin et al. (San Martin et al., 2001) that the visceral musculature also consists of small syncytia. Indeed, also the visceral mesoderm contains rP298 expressing cells, presumably founder cells, and sns expressing cells, presumably fusion competent cells. San Martin et al. (San Martin et al., 2001) and Klapper et al. (Klapper et al., 2001b) have shown that in duf mutants and in sns mutants, no fusion can be detected in the visceral mesoderm, implying that the founder cell hypothesis also holds true for the visceral musculature of the midgut.

Both P-elements in P1027 and P1729 contain a lacZ gene coupled to a basal promotor, which can be driven by genomic enhancer elements. The enhancer trap pattern in rols^P1027 is mainly restricted to the endoderm (see Fig. 6A), while β-galactosidase is expressed in the muscle attachment sites (apodemes) in the epidermis of rols^P1729 embryos (see Fig. 6B). Therefore we analyzed whether these tissues show morphologically detectable distortions in rols mutants. Apodeme integrity was checked using an antibody against Alien, which marks all muscle attachment sites (Goubeaud et al., 1996; Dressel et al., 1999). No major aberrations could be detected; therefore we conclude that at least apodeme formation is normal in rols^P1729 mutant embryos (data not shown). As many genes have a function in myogenesis and neurogenesis, we checked in addition for the gross morphology of the nervous system by staining with Mab22C10 (Zipursky et al., 1984). Again, no major specific aberrations in the nervous system were detectable (data not shown).

Myogenesis can be disturbed by disrupting muscle progenitor determination leading to a loss of founder cells, by interfering with muscle precursor formation or the recruitment and fusion of fusion-competent myoblasts. Therefore, we addressed the question of whether muscle founder cells and/or muscle precursor cells are correctly determined in rols mutants. We analyzed the formation and development of muscle founders and muscle precursors in the rols deficiency Df(3L)BK9 using the enhancer trap line rP298. In a wild-type background, rP298 expresses β-galactosidase in the nuclei of all muscle founder cells at the extended germ band stage and in the nuclei of all muscle precursor cells at later stages of embryonic development (Fig. 2A) (Nose et al., 1998). We detect a similar pattern of β-galactosidase expression in embryos homozygous for the Df(3L)BK9 deficiency. This result indicates that founder/precursor cell formation occurs normally in the absence of rols function (Fig. 2B). As we assume that the deficiency represents the null situation for rols, this shows that at least the majority of muscle founder/precursor cells are specified correctly in the absence of rols function.

As already mentioned, two steps may be distinguished during the formation of the myofibers at the morphological level (Bate, 1990). First, the founder cells recruit some myoblasts to form precursors with two to four nuclei. Second, these precursor cells attract further FCMs to form the multinucleated myotubes with up to 24 nuclei. When we analyze later embryonic stages according to the rP298 pattern, we observe from stage 15 onwards a reduced number of rP298-positive nuclei (compare Fig. 2C with 2D). To determine more precisely at which step myotube formation is blocked in rols mutants at least for one individual muscle, we analyzed the even skipped pattern, which marks dorsal muscle 1 and the pericardial cells (Fig. 2); as it was shown that Eve is expressed in fusion mutants (Paululat et al., 1999a). In embryos homozygous for the rols^P1729 allele, Eve expression in pericardial cells appears normal; however, the DA1 muscle expresses Eve in two to five nuclei, as opposed to up to 14 nuclei in wild-type embryos. As the P-element-induced mutation might not be a null mutation for rols, we analyzed Eve expression in the deficiency Df(3L)BK9. In this rols null mutant, development for dorsal muscle 1 proceeds beyond the founder cell stage, as we again observe two to three nuclei (Fig. 2H). This is in contrast to the situation in mbc mutants, where the Eve-positive cells are clearly mononucleated (Carmena et al., 1998). These data suggest that in the absence of rols, muscle founder cells fuse with a restricted number of myoblasts to become muscle precursor cells, but that muscle development becomes stalled at this stage.

For the steps that follow muscle precursor formation, light microscopy is not sufficient to resolve at which step myogenesis is interrupted during myofiber formation. We therefore analyzed the rols phenotype of the strong EMS allele rols^AD328 at the electron microscopic level. At early stage 13, myoblasts are closely apposed to each other (Fig. 3A) and morphologically indistinguishable from wild-type myoblasts. Syncytial cells with three nuclei are present, which presumably represent muscle precursor cells. However, by stage 14, precursor cells and FCMs have lost their close alignment and have scattered throughout the mesoderm (Fig. 3B). In wild-type embryos, a prefusion complex forms between precursor cells and FCMs after cell adhesion (Doberstein et al., 1997). In rols^AD328 mutants, the prefusion complex and later steps of fusion are absent. Whether this failure of cell alignment and further progression in myogenesis is due to defects in the precursor cells, in the FCMs or both remains to be clarified. To address this and other issues, we cloned the rols gene.

The rols allele rols^P1729 provided an entry point into the genomic region and facilitated the identification (for details see Materials and Methods) of a 5 kb fragment of genomic DNA that flanks the P-insert. We used the 5 kb genomic fragment to isolate several genomic clones from phage libraries. We used the phage clones to screen a cDNA library and we sequenced all positive cDNA clones. These data gave access to the corresponding sequence of the Drosophila genome in the database (Adams et al., 2000). From the analysis of the isolated cDNAs, we identified two putative rols transcripts (5.8 and 6.3 kb) that differ at the 5′ end. These transcripts are named rols7 (long transcript) and rols6 (short transcript) (Fig. 4A). The encoded proteins are named Rols7 and Rols6, accordingly. Sequence comparison between the cDNAs and the genomic region revealed a complex intron-exon structure as well as common and selective exons (Fig. 4A). The six most 3′ localized exons are common to both transcripts and include a single polyadenylation signal, as well as the majority of the open reading frame (Fig. 4A). rols6 contains two unique 5′ exons with the translation initiation codon localized in the second exon. The longer transcript contains two specific exons as well as the translation initiation codon in the second exon unique for rols7. At the 5′ site, rols is flanked by another gene previously identified as coding for Semaphorin 5c (Khare et al., 2000). The positions of the P-element integrations in rols^P1027 and rols^P1729 were determined by cloning and analyzing flanking sequences. This places the P-element in P1729 in intron sequences shared by the rols7 and the rols6 transcription units (Fig. 4A). In P1027, the P-element is inserted approximately 70 bp upstream of the transcription initiation site of rols6, which has been determined by primer extension analysis (data not shown). This P-element is localized in the second intron of rols7. Both P-elements are inserted into the same gene, and excision of both P-elements can revert the rols phenotype. This provides strong evidence that this gene is rols.

Sequence analysis of rols6 and rols7 predicts two distinct large proteins to be encoded by rols (Fig. 5). Conceptual translation indicates that Rols7 encodes a 1900 amino acid protein, whereas rols6 reveals an open reading frame of 1670 amino acids. Rols6 harbors 79 specific amino acids at the N terminus with a high level of acetic amino acids, while Rols7 is characterized by 309 specific amino acids at the N terminus (Fig. 5).

In order to gain insight into the possible function of Rols we analyzed the predicted domains of the protein using the SMART-tool of the EMBL (Schultz et al., 1998; Schultz et al., 2000). Two striking structural features are depicted in the common part of Rols6 and Rols7 (Fig. 5). Both proteins contain a RING-finger motif that has been shown to participate in protein-protein interactions (Freemont, 2000). In some cases, specialized RING-fingers are essential parts of proteins involved in the ubiquitination pathway having E3-ligase activity (Freemont, 2000). Furthermore, nine ankyrin repeats are predicted which represent a classical structure for protein-protein interactions. Close to the C terminus, an additional protein-protein interaction motif is found, a so-called TPR motif eventually overlaid by a coiled-coil region (Blatch and Lassle, 1999). Apart from these structural domains, the database research revealed no homologous proteins in other organisms. The predicted structure of Rols implies multiple protein-protein interactions during myoblast fusion.

We determined the temporal expression profiles for rols6 and rols7 by RT-PCR using primers specific for each transcript. Poly (A)⁺ RNA from embryos collected at 4 hours intervals was isolated and purified. Using this approach, we first detected the characteristic fragment for rols7 between 4 and 8 hours of development. rols7 reaches a maximum level of expression at 8-12 hours and then declines to low levels in 12- to 16- and 16- to 24-hour-old embryos (Fig. 4B). We first detect the rols6 transcript between 4-8 hours of development. rols6 reaches maximum levels between 12-16 hours of development and remains at high levels up to 24 hours (Fig. 4B). Therefore, the expression of rols7 coincides with the time course of the process of fusion between founder/precursor cells and FCMs, while maximum levels of rols6 occur at later stages.

The characterization of the P-element-induced alleles rols^P1027 and rols^P1729 provided strong evidence that the identified gene is responsible for the rols phenotype. Furthermore rols mRNA is present during the time of myoblast fusion in embryonic mRNAs. However, the RT-PCR experiments do not resolve the tissue(s) in which the particular transcripts are expressed. Therefore we used in situ hybridization to analyze the cellular transcript distribution during embryogenesis with specific probes for rols7 and rols6 (see Materials and Methods). rols7 expression is first observed at the extended germ band stage in a few mesodermal cells per segment (Fig. 6C) and in the head region where the precursors of the pharyngeal muscles arise (Fig. 6D). Transient expression of rols7 is observed in the visceral mesoderm during germband retraction (Fig. 6D) and then vanishes. The functional significance of rols7 expression in the visceral mesoderm remains to be clarified.

During germband retraction, myogenesis proceeds and several cells express rols7 in every segment. The precursors formed at stage 13 accumulate rols7 mainly around one nucleus (Fig. 6E,F). The mRNA is of very low abundance so that colocalization experiments with other markers were not of sufficient quality. Therefore, we compared the expression pattern of rols7 to that of Mef2 which is expressed throughout embryogenesis (Nguyen et al., 1994; Lilly et al., 1995; Bour et al., 1995; Taylor, 1995) and with rP298. In these experiments we used rP298 to identify muscle founder cells. Only a few myoblasts are rP298 positive (Fig. 6H) when compared to Mef2 in all nuclei of myoblasts (Fig. 6G). By contrast, the rols7 expression (Fig. 6E) is very similar to the expression pattern of rP298 (compare Fig. 6E with 6H) in the somatic mesoderm, indicating that rols7 is expressed in progenitor/founder cells up to the muscle precursor cells.

The pharyngeal musculature is also syncytial and we show that the development of these muscles is also dependent of rols (Fig. 1H). Previously, we have shown that sticks and stones mutants show fusion defects in the development of the pharyngeal musculature (Paululat et al., 1995) (EMS alleles previously named rost) (Paululat at al., 1999; Bour et al., 2000). Our expression analysis indeed revealed two distinct cell populations in the clypeolabrum at late stage 12 (Fig. 6K-N) that later invaginate to form the pharyngeal muscles. rols7 and rP298 show overlapping expression patterns in the ventrally localized mesodermal cells of the clypeolabrum (Fig. 6K,M), while sns transcripts, which mark FCMs, are localized in the dorsally adjacent mesodermal cells (Fig. 6N). As the rols7 pattern overlaps with the rP298 pattern (reflecting duf expression), we suggest that also the pharyngeal musculature consists of founder cells and fusion competent cells.

rols6 exhibits a distinct transcript profile. At the extended germ band stage, rols6 is expressed strongly in the invaginating endoderm (Fig. 6I). At stage 14 and later, rols6 is expressed in ectodermal cells of the head region (Fig. 6L) and the developing malpighian tubules are stained weakly (Fig. 6J). Furthermore, a stripe-like expression pattern is observed in the ectoderm at stage 16 (data not shown). The expression pattern of rols6 is thus very similar to the enhancer trap pattern observed in rols^P1729 in the apodemes and in rols^P1027 in the endoderm (Fig. 6A,B). This distinct distribution shows that rols7 is transcribed during development of the Drosophila pharyngeal and body wall musculature. As only rols7 is detected in myoblasts during fusion and ceases beyond stage 14, we propose that this is the relevant transcript. Comparison with the rP298 staining indicates that rols7 is expressed starting with progenitors/founders up to muscle precursor cells during myoblast fusion.

The expression profile of rols7 suggests that it may act at the level of muscle founder or precursor cells. The analysis of the mutant phenotype reveals the appearance of elongated mini-muscles containing more than one nucleus in several cases, which was shown in detail for dorsal muscle 1. In order to test the hypothesis that Rols acts on the precursor level, we examined the fusion competence of rols mutant cells in a wild-type background with a cell transplantation strategy adapted from Klapper et al. (Klapper et al., 2001a). The wild-type host embryos contain a daGAL4 construct, which is expressed in all cells, while the transplanted cells of the rols mutant contain an UAS-lacZ gene. Thus rols mutant cells in a wild-type background will express only the reporter β-galactosidase after successful cell fusion with the wild-type host cells. One might expect that successful fusion is dependent on the transplanted cell type. As we transplant ventral mesodermal cells at the cellular blastoderm, they might develop either to founders or to fusion-competent cells the last being the far more abundant cell type. If Rols is indeed required in muscle precursor cells – as suggested by its expression and mutant phenotype – but not in FCMs, wild-type precursor cells of the host embryo should be able to recruit FCMs from rols mutants to form myofibers that express β-galactosidase. However, transplanted donor cells that develop into precursors should not be able to recruit further host myoblasts for fusion.

We chose the deficiency Df(3L)BK9 as a null allele for rols. As we transplant mutant mesodermal cells into wild-type embryos, the loss of rols6 in the endoderm and semaphorin 5c in the ectoderm as well as the loss of other genes localized in the deficiency should not influence this assay. After transplantation the recipient embryos were allowed to develop until 3rd instar larval stage and the muscle pattern was examined with respect to myotube development.

From a total of 191 transplantations, 119 embryos (62%) reached the third larval instar. Of these larvae, 53 showed a clone derived from the transplanted cell descendants, which had fused to host cells. In 11 cases, the clones derived from homozygous rols donors and were found in the musculature (Fig. 7). We compared these clones at the morphological level with 42 clones derived from control donors (hetero- or homozygous blue balancer donors) with regard to size, shape and correct attachment. The 11 clones derived from homozygous rols embryos were found in the ventral, lateral and dorsal body wall muscles of third instar larvae, and they demonstrate that at least a population of rols mutant cells is able to fuse with host cells to multinucleate myotubes (Fig. 7A-C). These data show that rols mutant cells can participate in fusion. As these large muscle clones are abundant, we suggest that in these clones rols mutant FCMs fuse with wild-type founders. Moreover we found the clones at the same positions in the thoracic and the abdominal segments and approximately in the same size (Fig. 7A) as the control clones. In the case of transplanted rols mutant cells, we found in five larvae small muscle-like structures referred as mini-muscles (Fig. 7B-D) or compact, not elongated muscle-like cells (Fig. 7B, arrowhead). These mini-muscles represent only a small part of the observed clones and contain two to five nuclei. We consider these mini-muscles as precursor cells formed by a rols mutant founder and fusion competent cells of the donor. In one case, we observe two of these mini-muscles close to each other. We interpret the appearance of duplicated mini-muscles as evidence for two precursor cells derived from asymmetric cell division of a progenitor. On the basis of these observations, we conclude that the mini-muscles can be hybrid precursor cells derived from a rols mutant founder (with UAS-lacZ) and one to four cells of the wild-type host (carrying daGAL4). Therefore we suggest that Rols acts in muscle precursors to recruit further FCMs. With rols-deficient cells, we found that nearly 50% of the host embryos contain mini-muscles. We detected in four cases of the 43 control transplantations (in about 10% of the embryos) with hetero- or homozygous balancer embryos similar defects and assumed that this is probably a result of a dosis effect or may be due to incompatibility of homozygous balancer cells in the mosaic clones.

In summary, the ultrastructural data, the immunostaining of founder and precursor cells as well as the transplantation assays clearly show that rols is required for proceeding from the precursor level towards multinuclear myotubes. This correlates convincingly with the immunohistological localization of the Rols protein at the cell borders as shown by Menon and Chia (Menon and Chia, 2001).

The rolling pebbles gene encodes two distinct transcripts with independent transcription initiation sites: rols7 and rols6. rols7 starts to be expressed in muscle progenitor cells but is detected still in muscle precursors until stage 14. The encoded proteins Rols7 and Rols6 share several putative protein-protein interaction domains as a RING-finger motif, ankyrin and a TPR repeat. Rols7 is distinguishable by 309 unique N-terminal amino acids, Rols6 by 79 unique N-terminal amino acids. We focus on the possible role of Rols7 during myoblast fusion.

An important question is at which cellular level rols is essential. The early enhancer trap pattern of rP298, which reflects the expression of the duf gene, is normal in rols null mutants, suggesting that muscle progenitors and founders are specified correctly. However, after stage 14, the number of rP298-positive cells decreases significantly. Electron microscopic analysis of rols^AD328 shows cells with three nuclei, presumably muscle precursor cells. Further phenotypical analysis of rols mutants revealed small myofiber-like structures with only a few nuclei also at later stages of myogenesis. These muscle precursors stretch and contact the epidermis, as it has been observed in other fusion mutants (Paululat et al., 1999b). From the presence of muscle precursors in rols mutants we deduce that rols is not essential for their formation. It is possible that another protein compensates for rols function in these early fusion steps leading to precursor formation. Further on, muscle precursors attract additional myoblasts for fusion in the wild type. In the rols deficiency allele these precursor cells form the small myofiber-like cells and seem to attract some fusion-competent cells, as we observed filopodia of fusion-competent cells directed towards these myofiber-like cells; however, they do not fuse to them.

In in situ hybridization studies, we detect rols7 RNA only in muscle progenitor descendants, implying that Rols7 is essential in the precursor cells but not in the surrounding myoblasts. However, the rols7 transcript is rare and we would not have detected a tenth of the mRNA level in FCMs. We therefore established an in vivo system to determine the functional competence of muscle precursor cells and fusion-competent cells by transplanting rols-deficient cells into wild-type host embryos. The majority of the transplanted cells formed multinucleated myofibers with normal shape and attachment to the epidermis. This clearly showed that transplanted fusion-competent cells from rols-deficient embryos (these are far more abundant than muscle founder cells) behave like wild-type cells in the establishment of multinucleated myofibers, and Rols is not required in myoblasts for fusion to wild-type precursor cells. By contrast, we observed mini-muscles with two to five nuclei, which originate from fusion between transplanted and host cells. Based on these results, we concluded that the muscle progenitor cells are correctly determined in rols mutants and that they divide to two founder cells, which is supported by our observation that occasionally two precursor cells are located close to each other in the host embryo. These precursor cells represent heterokaryons of rols^– and rols⁺ nuclei; nevertheless, they behave like Rols-defective precursors. Thus, the recruited rols⁺ nuclei obviously do not complement for the rols^– nucleus. As a consequence, we observe mini-muscles instead of normal sized myotubes. It is well known that the muscle progenitors and founders are determined by intrinsic components like transcription factors and extrinsic factors as signaling molecules from the overlaying ectoderm (Baker and Schubiger, 1995; Baylies et al., 1998). This integration of determining factors is probably missing in rols⁺ nuclei of the mini muscles, and we suggest that this is the reason that they are unable to activate rols and thus to complement the rols-deficient founder cell nucleus. Determined founder cells recruit one to three myoblasts for muscle precursor cell formation. After this step, development to myofibers is interrupted in rols mutant cells in the wild-type background. These data are confirmed by our observation of cells with up to three Eve-positive nuclei characteristic for dorsal muscle 1 in the analyzed rols deficiency and our ultrastructural analysis, showing cells with three nuclei in rols mutants.

We conclude from these observations that the first phase of myoblast fusion leading to muscle precursor cells with two to four nuclei, is unaffected in rols mutants. Another indication of a two step fusion process might be the observation that Eve-positive precursors for dorsal muscle 1 are formed in a strong sns allele (Paululat et al., 1999a). For the second phase of fusion leading to multinucleated myofibers, Rols7 is essential in muscle precursor cells.

Comparison with the onset of the rols7 mRNA detection already at progenitor/founder stage to the analysis of the mutant phenotype and the transplantation assays shows biological activity of Rols in the precursor cells with two to four nuclei. This correlates well with a change in protein distribution from uniform cytoplasmic to membrane associated during fusion. This reorganization is dependent on Duf (Menon and Chia, 2001). Our data obtained with the rP298 enhancer trap line suggest that the early activity of Duf is independent of Rols, while the maintenance of expression is dependent on the Rols-mediated fusion steps.

As described in detail in the introduction myofiber formation is prefigured by the specification of progenitor cells out of an equivalence group. After asymmetric division of a progenitor cell two founder cells are determined that lead to two distinct myofibers, or to one myofiber and one founder cell set aside to form an adult muscle during metamorphosis. Subsequently myoblast fusion starts. On the morphological level precursor cells with two to three nuclei can be distinguished (Bate, 1990), then fusion proceeds by recruiting additional myoblasts resulting in a mature myofiber. It has recently been shown that duf is essential for this first step of fusion (Ruiz-Gómez et al., 2000) as well as mbc (Carmena et al., 1998), while other fusion mutants (e.g. blow) allow precursors to form but stop at subsequent steps of fusion (Paululat et al., 1999b). Our analysis enables us to integrate rols into this cascade. We show that precursor cells are often correctly determined in rols mutants but that they are not able to recruit additional myoblasts for fusion. We started to analyze this feature in detail by the analysis of individual muscles first by Eve expression in dorsal muscle 1 in rols-deficient embryos. For this muscle, we can clearly show that founder cells start to fuse to mini-muscles of two to four nuclei, presumably precursor cells. Thus rols should be positioned downstream of duf, where even the muscle precursors are not formed. However, duf may have additional functions later in the fusion process. The transplantation assays furthermore demonstrate that rols function is restricted to the muscle precursor cells and not needed in FCMs. The phenotypic analysis at the EM level revealed that rols mutants neither form a prefusion complex nor electron-dense plaques, which shows that rols is required earlier than blow and sns. In agreement with these data, Menon and Chia (Menon and Chia, 2001) have shown that the Rols protein distribution is not affected in sns and blow, but is clearly dependent on duf.

Both Sns and Duf belong to the immunoglobulin superfamily and have been inferred to mediate cell adhesion, Sns in FCMs and Duf in muscle progenitors and founders. Duf furthermore attracts myoblasts to founders. Whether duf is also involved in the cell fusion process itself is not known, as the mutations available so far are null mutants. Rols clearly belongs to a new class of molecules that are essential for fusion. In stage 13 embryos, cell adhesion between muscle precursor cells and fusion-competent cells initially looks like wild-type but cells appear to lose contact. After cell adhesion, we would expect that a signaling mechanism acts between both cell types. Rols may mediate directly or indirectly a signal after recognition and adhesion between precursor cells and fusion-competent cells, leading to later steps of fusion. In the absence of these later steps, adhesion may not be maintained. Although the prefusion complex and the subsequent steps are symmetrically exhibited by both cell types, it might be expected that a specific signal from the precursor cell to the fusion-competent cells is essential to initiate such a process (Fig. 8). Ankyrin repeats are especially predestinated to mediate further steps in a signal transduction cascade on the precursor site. The RING-finger structure is a further candidate for protein-protein interaction in the precursor cell. Proteins with such a domain have been found in many organisms, and some of them are involved in the ubiquitin-mediated protein-degradation pathway as they function as E3-ubiquitin ligases (Zheng et al., 2000). However, comparison of the RING-finger sequence present in Rols to a Drosophila Cbl homolog reveals no agreement of otherwise very conserved amino acids between human, C. elegans and D. melanogaster Cbl homologs. Therefore, a role for Rols in an ubiquitination pathway driven protein degradation is quite unlikely, but cannot be completely excluded. Taken together, the isolation and characterization of rols, the very specific phenotype of mutants analyzed so far and its extremely interesting distribution during the fusion process points to a specialized function exclusively on the side of the muscle precursor cell. So the search for possible interacting partners besides Drosophila Titin (Menon and Chia, 2001) will probably lead to further insights into the molecular mechanisms of the fusion process.

We thank W. Janning, A. Nose, the Drosophila Stock Centers in Bloomington and Umea for fly stocks, our colleagues M. Frasch, D. Kiehart and H. T. Nguyen for antibodies, and S. Abmayr for sns cDNA. We thank very much Sree Devi Menon and Bill Chia for communication and exchange of manuscripts prior to publication. We are indebted to R. Klapper and W. Janning for communication of methods before publication and for supply of the daGAL4 strain. We thank Heike Sauer for excellent secretarial assistance, Peter Seum, Miriam Findt and Dominik Helmecke for technical assistance, and Thomas Kremser for technical advice. This research was supported by the Deutsche Forschungsgemeinschaft to R. R.-P. (Re 628/10-3 and 10-4, SFB 397), the Kerckhoff-Stiftung and the Fonds der Chemischen Industrie. A. M. M. is an Associate investigator of the Howard Hughes Medical Institute. This research was also supported by NSF grant (IBN-0077727) to J. B. S.