INTRODUCTION

The differentiation of body wall musclulature from the meso-dermal germ layer is a highly conserved process in the animalkingdom. The development and differentiation of this germlayer are differentially regulated genetic processes.

A prerequisite for the formation of the mesoderm is thedetermination of the dorsoventral axis. For Drosophila, theexistence of a cascade of maternally transcribed genes thatdrive the expression of the ventral morphogen dorsal has beenshown. The dorsal gene product regulates the expression of thezygotic genes twist and snail which are essential for mesodermformation (for review see St Johnston and Nüsslein-Volhard,1992; Steward and Govind, 1993). In contrast to the earlyevents of Drosophila mesoderm formation, much less is knownabout the differentiation of this germ layer. One of the earliestsigns of mesoderm differentiation is the separation of thevisceral and somatic mesoderm, which develop the gut mus-culature and the dorsal vessel, gonadal mesoderm, pharyngealand body wall musculature, respectively. The larval body wallmusculature of Drosophila consists of about 30 muscles perhemisegment (Crossley, 1978; Campos-Ortega and Harten-stein, 1985).

One morphological characteristic of the differentiation of

the body wall musculature is the occurrence of mononucleatedmesodermal cells that previously have undergone a determi-nation as muscle founders (Bate, 1990). From these foundercells, muscle precursors, visible as bi- and trinucleated cells,develop in a specific pattern at the end of germband extension.During germband retraction, muscle development proceeds bythe fusion of these precursor cells with fusion-competentmononucleated myoblasts to multinucleated myotubes. Thesemyotubes are formed of 4-25 myoblasts (Bate, 1990).

In vertebrate cell culture systems, this fusion process can beinduced by transfection of genes from the MyoD family, whichconsists of the four well-characterized genes MyoD, myogenin,Myf 5 and MRF 4 and code for transcriptional activators (forreview see Olson, 1990). MyoD was first identified and clonedon the basis of its ability to convert non-muscle cells to amyogenic cell fate, and its specific expression in myoblasts(Davis et al., 1987; Weintraub et al., 1989). All members ofthis gene family encode transcription factors with a basic HLHdomain which is necessary and sufficient for determining themyogenic potential in vitro (Olson, 1990). It has recently beenshown that the inactivation of the myogenin gene in micecauses muscle deficiencies and neonatal death (Hasty et al.,1993; Nabeshima et al., 1993). Besides the muscle abnormal-ities, these mutant mice reveal skeletal defects.

2611Development 121, 2611-2620 (1995)Printed in Great Britain © The Company of Biologists Limited 1995

The development and differentiation of the body wall mus-culature in Drosophila are accompanied by changes in geneexpression and cellular architecture. We isolated aDrosophila gene, termed rolling stone (rost), which, whenmutated, specifically blocks the fusion of mononucleatedcells to myotubes in the body wall musculature. β3 tubulin,which is an early marker for the onset of mesoderm differ-entiation, is still expressed in these cells. Gastrulation andmesoderm formation, as well as the development of theepidermis and of the central and peripheral nervoussystems, appear quite normal in homozygous rolling stoneembryos. Embryonic development stops shortly beforehatching in a P-element-induced mutant, as well as in 16EMS-induced alleles. In mutant embryos, other mesoder-mal derivatives such as the visceral mesoderm and thedorsal vessel, develop fairly normally and defects are

restricted to the body wall musculature. Myoblasts remainas single mononucleated cells, which express musclemyosin, showing that the developmental program of geneexpression proceeds. These myoblasts occur at positionscorresponding to the locations of dorsal, ventral andpleural muscles, showing that the gene rolling stone isinvolved in cell fusion, a process that is independent of cellmigration in these mutants. This genetic analysis has set thestage for a molecular analysis to clarify where the rollingstone action is manifested in the fusion process and thusgives insight into the complex regulating network control-ling the differentiation of the body wall musculature.

Key words: Drosophila, musculature, fusion defects in mutants,nautilus, myoblast, rolling stone gene

SUMMARY

Fusion from myoblasts to myotubes is dependent on the rolling stone gene

(rost) of Drosophila

Achim Paululat†, Susanne Burchard† and Renate Renkawitz-Pohl*

Molekulargenetik, FB Biologie, Philipps-Universität, 35032 Marburg, FRG

*Author for correspondence†The first two authors contributed equally

2612

The disruption of the MyoD or Myf-5 gene in mice has noeffect on the development of the somatic musculature, pre-sumably because of their functional redundancy in musclemorphogenesis (Braun et al., 1992; Rudnicki et al., 1992).Indeed, examination of mice carrying mutations in both ofthese genes revealed a complete absence of skeletal muscles.This suggests that either Myf-5 or MyoD is required for thedetermination or propagation of skeletal myoblasts (Rudnickiet al., 1993).

For Drosophila, the MyoD-related gene nautilus has beenisolated using the conserved basic HLH domain as a probe(Michelson et al., 1990, Paterson et al., 1991). Nautilus isexpressed in bi- and trinucleated mesodermal cells in everysegment that can be followed into mature muscles and mightbe muscle precursor cells (Abmayr et al., 1992). Two othergenes, the homeobox-related gene S59 (Dohrmann et al., 1990)and apterous (Bourgouin et al., 1992), a member of the LIMdomain family, are specifically expressed during early fusionprocesses in a subset of muscle precursors. These genes mostlikely are involved in regulating cell fate of muscle precursorsand events that determine the precise segmental pattern of thebody wall musculature.

Besides transcriptional activators, we might expect that cell-cell adhesion molecules are involved in the development ofmyoblast to mature muscles, as has been made likely by theisolation of the vertebrate muscle-specific M-cadherin gene(Donalies et al., 1991). Expression of this member of thecadherin gene family has been shown to correlate with skeletalmuscle development (Donalies et al., 1991, Moore and Walsh,1993). For vertebrates, it has been proposed that neuralcadherins also play a role in myoblast interactions (Knudsenet al., 1990a,b). Furthermore, it is likely that interactionbetween the nervous system and the somatic musculature isessential for proper development of somatic muscles. It ispossible that some genes fulfil functions both in the develop-ment of the nervous system and during muscle development;for example, neurogenic mutants of Drosophila, like Notch andDelta, also show distortions in muscle differentiation (Corbinet al., 1991). In Drosophila, there are several other genesknown to be important for the development of the somatic andvisceral musculature, most of which encode for DNA-bindingproteins (Azpiazu and Frasch, 1993; Dohrmann et al., 1990;Barad et al., 1988; Bodmer et al., 1990, Bodmer, 1993; Lai etal., 1993; Bourgouin et al., 1992).

To search for morphoregulatory molecules involved inDrosophila myogenesis, we chose a genetic approach to avoida preselection of the kind of molecule involved. We reasonedthat defects in the musculature will confer late embryoniclethality as the larvae would not be able to hatch. β3 tubulinwas used to study of the differentiation of the mesoderm. It isexpressed from the extended germband stage until shortlybefore hatching (Leiss et al., 1988). The β3 tubulin antibodystains mesodermal derivatives like somatic and visceral mus-culature and the dorsal vessel.

With the β3 tubulin antibody, we screened embryonic lethalsfor muscle defects (see Burchard et al., 1995) from a P-elementscreen of the second chromosome performed in parallel to thethird chromosome (Cooley et al., 1988).

We found one mutant with a reduced set of somatic muscles(not enough muscles (nem), Burchard et al., 1995) and a secondmutant, termed rolling stone (rost). Embryos homozygous for

rost mutations are characterized by a high number of unfusedmyoblasts and only a few myotubes. Here, we describe the rostmutant phenotype in muscle differentiation and its relation tothe differentiation of the epidermis and the nervous system.The genetic and cytological analysis suggest that rost functionis specific for myoblast fusions.

MATERIALS AND METHODS

Drosophila stocksP-element-induced embryonic lethals were obtained from T. Orr-Weaver and A. Spradling. To screen for mutants with defects inmuscle development, we stained 125 P-element mutant lines from thesecond and third chromosome, preselected for dying late in embryonicdevelopment, with the mesoderm-specific β3 tubulin antibody. Threeof the P-element insertions seemed to have a muscle-specific defect.Furthermore, we stained mutants with chromosomal deficienciescovering 40% of the genome. However, most deletions revealed suchsevere distortions that the specificity of muscle phenotypes remainedto be determined. In contrast, the P-element-induced mutant, rostP20,specifically revealed distortions in fusions from myoblasts tomyotubes and was chosen for a detailed cytological and geneticanalysis.

Genetics and EMS-mutagenesisThe P-element in the rostP20 mutant localises on the left arm of thesecond chromosome, determined by polytene in situ hybridization(data not shown). The rostP20 allele (l(2)neo14) was marked with bwand sp, and brought together by recombination with the previouslymarked chromosome of nemP8 (l(2)neo113, Burchard et al., 1995).Isogenic males of the phenotype cn bw sp were treated with ethylmethanesulfonate (Lewis and Bacher, 1968) and crossed to virginfemales of the genotype b pr cn vgD bw sp/CyO. 10000 of the resultingcn bw sp/CyO males were tested for lethality of their offsprings bycrossing with females of the genotype nemP8rostP20/CyO. Offspringcontaining a lethal mutation over the test chromosome were checkedagainst nemP8/CyO and nemP20/CyO individually to distinguishbetween nem and rost alleles. A total of 16 EMS-induced rost alleleswere obtained.

Reversion of the P-element insertionThe excision of the P-element of the mutant rostP20 should reverse itshomozygous lethality to vitality, if indeed the P-element caused themutation. This reversion experiment was performed by microinjec-tion of the transposase-producing helper plasmid p25.7wc (Karess andRubin, 1984) in the germ line of embryos with the genotyperostP20/CyO. The resulting flies were backcrossed to males or femalesof the line rostP20/CyO which contain the P-element insertion.Crossing of flies in which the P-element was excised by transposaseresults in homozygous viability in three independent lines recogniz-able by wild-type wings. Embryos stained with an antibody againstβ3 tubulin do not show any unfused myoblasts.

Antibody staining of embryosEggs laid by flies of the appropriate genetic constitution werecollected on agar-apple juice plates. In order to obtain an age distri-bution that allowed visualization of different stages of muscle devel-opment, eggs were collected over a 24 hour period. Eggs weredechorionated, permeabilized and fixed essentially as described byLeiss et al. (1988). After washing and blocking in BBT (0.15% crys-talline BSA, 10 mM Tris-HCl, pH 7.5, 50 mM NaCl, 40 mM MgCl2,20 mM glucose, 50 mM sucrose, 0.1% Tween 20), the eggs wereincubated overnight with a dilution of the appropriate antibody. Forthe analysis of the mesoderm formation, we used the twist antibody(Thisse et al., 1988; Leptin und Grunewald, 1990). For staining of

A. Paululat, S. Burchard and R. Renkawitz-Pohl

2613Myoblast fusion and rost gene expression in Drosophila

mesodermal derivatives, the anti-β3 tubulin antibody and myosinantibody were used (Leiss et al., 1989; Kiehart und Feghali, 1986).The central and peripheral nervous system was stained with the mono-clonal antibody mAb22C10 (Zipursky et al., 1984). The interseg-mental and segmental neurons were visualized with the fasII antibody(mAb1D4, Grenningloh et al., 1991). The bound antibody wasdetected with a biotinylated secondary antibody and stained with theVectastain ABC Elite-kit (VectorLabs) using diaminobenzidine asdetection agent. Double stainings were performed as described byLawrence et al. (1987). The stained embryos were embedded in Eponand photos taken under Nomarski optics with a Zeiss Axiophot micro-scope (Kodak, Ektachrome 25).

RNA in situ hybridizationIn order to visualize muscle precursor cells in rost mutants, whole-mount RNA hybridizations were performed with digoxigenin-labellednautilus cDNA probes according to Tautz and Pfeiffle (1989). Toidentify homozygous mutant embryos at early developmental stagesthat had not yet shown severe morphogenetic alterations, we stainedstrains carrying the rost mutation balanced over CyO7.1. This balancerchromosome contained a hindgut/anal-pad β-gal fusion construct(Affolter et al., 1993). Embryos were hybridized simultaneously witha lacZ and a nautilus cDNA probe. Only those embryos that show noexpression of the β-gal fusion construct are homozygous for the rostmutation.

RESULTS

Mesoderm formation is normal in rost mutantsMesoderm formation depends on a cascade of maternallyactive genes as well as on two zygotically active genes twistand snail (Thisse et al., 1987; Bouley et al., 1987). It is possibleto follow mesodermal cells during gastrulation using the twistantigen (Leptin and Grunewald, 1990). Thus, an antibody rec-ognizing the twist antigen was used to characterize mesodermformation in muscle mutants. Generally, we found no majordifference in the pattern of twist expression in comparison tothe wild type for the mutation described here (rolling stone,see below) nor for nem (not enough muscles, Burchardt et al.,1995) (data not shown). However, we cannot exclude thatsingle cells, which may be essential for the initial formation ofspecific muscles, are missing (see below and Discussion).

The gene rolling stone is essential for myoblastfusion during embryonic muscle differentiationAs a marker to follow mesoderm differentation, P-element-induced embryonic lethals were stained with the β3 tubulinantibody. The β3 tubulin is specifically expressed duringembryonic mesoderm formation (Leiss et al., 1988), as well asduring development of the adult musculature during metamor-phosis (Bate et al., 1991). The β3 tubulin isotype is firstdetectable after mesoderm formation in the extended germbandstage. It persists in splanchnopleura and somatopleura until themuscle differentation program has been completed (Leiss et al.,1988). Thus, the β3 tubulin isotype is a well-suited marker tofollow the differentiation of the major mesodermal derivatives,e.g. somatic muscles, visceral muscles and dorsal vessel. In thewild-type, this β3 tubulin antigen is present in every singlemuscle of the body wall musculature (Fig. 2A). The somaticmusculature is organized in ventral, pleural and dorsal groups.One of the analyzed P-element-induced mutations revealed avery specific distortion in myoblast fusion so that only a few

myotubes were observed (Fig. 1). We named the correspond-ing wild-type gene rolling stone (rost) and the P-element-induced allele rostP20. In homozygous rostP20 mutants,embryogenesis proceeds until shortly before hatching, whendorsal closure was visible already (Fig. 1D). In these embryos,the dorsal vessel was formed quite normally (Fig. 1D). Alsothe visceral musculature developed properly (Fig. 1B).However, major distortions were visible in the somaticmesoderm. In particular, free unfused myoblasts in themajority of mutant embryos at late stages of development char-acterize this mutant (Fig. 1C, E-G). Reflecting the variabilityof the P-element-induced phenotype, some embryos wereobserved showing far less unfused myoblasts and also astrongly reduced muscle pattern (Fig. 1H). The first distortionswere detectable during germband retraction. The embryoshown in Fig. 1A, demonstrates this earliest visible fusiondefect in the rostP20 mutant. Single myoblasts (white arrows)in the somatic mesoderm, shortly after separation from thevisceral mesoderm become visible. This observation wasproved by double stainings of rostP20/CyO7.1 embryos withantibodies against β3 tubulin and β-gal, which allow mutantand wild-type embryos to be distinguished by the expressionof the reporter gene lacZ in the wild-type embryos. Whenembryogenesis proceeded, most myoblasts remained unfused(Fig. 1E-G). Nevertheless, the myoblasts were often arrangedin specific groups in locations where in the wild-type ventral,pleural and dorsal groups of muscles were formed (Fig. 1G,E).Clearly, all muscles of the ventral (e.g. Fig. 1G), lateral (e.g.Fig. 1E,F) and dorsal group (e.g. Fig. 1D) can be affected. Wefind the same results in the analysis of the EMS-induced alleles(see below).

EMS-induced rost alleles fall into onecomplementation groupThe original mutant was induced by P-element mutagenesis.P-elements often integrate in upstream regions of a gene, thusthe P-element does not necessarily destroy the function of thegene completely. Furthermore the P-element may causemutations in two neighbouring genes. To clarify whether therostP20 phenotype shows the complete phenotype of the gene,we performed an EMS mutagensis. We obtained 16 EMS-induced alleles to rostP20 (see Material and methods).

To decide if it indeed was a single gene, we performed acomplementation analysis between all EMS-induced alleles, aswell as with the rostP20 allele. The result showed that allmutations were alleles of a single gene. We found no interal-lelic complementation. The alleles reflect the rost phenotypein that the fusion process from myoblasts to myotubes isstopped, causing late embryonic lethality. Because of their rolein this morphogenetic process, we sorted the rost allelesaccording to the degree of myotube formation (Table 1). As anexample for the strongest class, the rost15 allele is shown incomparison to the wild type (Fig. 2A for the wild type, Fig.2B-D for the rost15 allele). The dorsal vessel (Fig. 2D) and thevisceral musculature developed normally in all alleles. Thismakes it very likely that the rost gene is essential for fusion ofmyoblasts to myotubes specifically and is not involved in thedifferentiation of the visceral musculature and the dorsalvessel. Fig. 2B and C demonstrates the variability of the rostphenotype in the EMS-induced alleles similar to the variabiltyof the rostP20 mutant.

2614 A. Paululat, S. Burchard and R. Renkawitz-Pohl

Fig. 1. Staining with β3 tubulin antibody shows fusion distortions of the somatic musculature in homozygous rostP20 embryos. (A) A stage 13homozygous rostP20 embryo. The somatic musculature (sm) shows single unfused myoblasts (white arrows). (B) The visceral musculature (vm)of a homozygous stage 12 rostP20 embryo develops normally. (C) A mutant stage 14 embryo shows many unfused myoblasts (mb) in the dorsalmusculature. (D) A dorsolateral view of a stage 15 embryo. Dorsal vessel (dv) is normal while characteristic unfused myoblasts (mb) arevisible in the region of the dorsal musculature. (E) A lateral view of a stage 16 embryo illustrates the pleural musculature, which shows fewmyotubes (mt) and a great number of single unfused myoblasts (mb). The myoblasts are arranged at the location where in the wild-type pleuralmuscles are formed. (F) Higher magnification of E showing the ordered location of myoblasts (mb) preceding the formation of pleural muscles.(G) A stage 17 embryo shows unfused myoblasts (mb) in the pleural and ventral musculature. (H) This rare phenotype at stage 17 ischaracterized by a reduced number of muscles and only some unfused myoblasts are stained with the β3 tubulin antibody.

2615Myoblast fusion and rost gene expression in Drosophila

The six alleles (see Table 1) stronger than rostP20 indicatethat this phenotype very likely represents the null situation.Furthermore, we proved numerous chromosomal deletions butnone of them uncovered the rost gene. Cloning and sequenceanalysis of the strongest alleles will finally clarify the nullphenotype.

Expression of the nautilus gene is independent ofthe rost geneOne key step in the process of muscle formation is the seg-regation of muscle precursors at the stage of germband retrac-tion. Nautilus, the Drosophila homologue of the vertebratemyogenic transcription factor MyoD is expressed in a subsetof muscle precursors (Michelson et al., 1990; Paterson et al.,1991). To test whether the rost gene is essential in this earlymyogenic cells, we looked for the presence of muscleprecursor cells in rost mutants by analysing the expressionpattern of nautilus in rost mutants (Fig. 3), assuming that therost phenotype might be a result of the absence of muscleprecursors. During germband retraction, when nautilusexpression begins, the rost phenotype is not recognizable byβ3 tubulin staining, but homozygous rost/rost embryos canbe identified by lack of the balancer marker (see Material andMethods). The nautilus expression in rostP20 homozygousmutants (Fig. 3A,C,E,G) is shown in comparison to wild-typeembryos (Fig. 3B,D,F,H). In a lateral view no difference isvisible at stage 10, when nautilus is first expressed in singlecells per hemisegment (Fig. 3A,B). Shortly later, more cells

express nautilus (Fig. 3C,D). No differences are observed indorsal view. In the wild type (Fig. 3F) and in rostP20 mutants(Fig. 3E), precursors of the ventral muscles are clearlyvisible. In stage 13/14, wild-type and mutant embryos revealan identical nautilus expression pattern, as far as is detectablewith this method.

Fig. 2. Muscle phenotype of the EMS-induced allele rost15. (A) β3tubulin expression in a stage 17 wild-type embryo is shown forcomparison. vem, ventral muscles; pet 1-3, pleural external muscles1-3; pit, pleural internal muscle. (B) This homozygous rost15 embryorepresents a phenotype with a lot of myotubes and no significantproportion of unfused myotubes. The muscle pattern, however, is notcomplete and some of the muscles appear to be too short (whitearrows) and disorganized. Identifiable myotubes of the ventral part ofthe musculature (vm) and the pleural muscles pet 1-3 are marked.(C) This homozygous rost15 embryo shows a lot of unfusedmyoblasts (mb) in the pleural musculature and some myotubes in theventral musculature (vem). Often the pleural internal muscle (pit) hasdeveloped properly. (D) Dorsolateral view of a mutant stage 16embryo reveals an intact dorsal vessel (dv) and a severe reduction ofthe dorsal musculature (dm).

Table 1. Classification of rost alleles according to thedegree of myotube formation

rost alleles in which the rost alleles with rost alleles in which majority of myoblasts fused and unfused the majority of remain unfused myoblasts myoblasts are fused

rost1 rost5 rost4

rost13 rost2 rost6

rost15 rost12

rost18 rost16

rost21 rost17

rost23 rost20

Two of the EMS alleles cannot be interpreted as a second mutation had beeninduced stopping embryonic development at the extended germband stage.

The rostP20 allele shows a phenotype pending between the strongest andmedium class. The variation within this mutant stock presumably is due to theposition of the P-element relative to the transcription unit.

2616 A. Paululat, S. Burchard and R. Renkawitz-Pohl

Fig. 3. Nautilus is expressed in muscle precursors in homozygous rostP20 embryos. Wild-type embryos were stained with a nautilus cDNAprobe. rostP20/CyO7.1 (see Material and methods) embryos were stained simultaneously with a nautilus cDNA and a lacZ probe to distinguishbetween homozygous rostP20 embryos and embryos containing the balancer chromosome. (A,C,E,G) Homozygous rostP20 and (B,D,F,H) wild-type embryos. Embryos for B and D were taken from a staining of wild-type embryos, all others are from a rostP20/CyO7.1 offspring. The lacZreporter gene expression in the hindgut/anal-pad-precursor is shown in F and H (arrowhead). B and D show a lateral view of wild-type embryos(stage 10 and 11) selected by the double staining. Nautilus is expressed in the lateral muscle precursors (A) and shortly later in the medialprecursors as well (C). Compared to wild-type embryos (B,D) at similiar stages and orientation there is no obvious difference in the expressionpattern. E shows a mutant embryo (stage 10) with a normal level of nautilus expression in all precursors (arrows). A wild-type embryo of thesame stage is shown in F. Later on (stage 13) there are no differences in the nautilus expression pattern between homozygous rostP20 (G) andwild-type (H) embryos.

2617Myoblast fusion and rost gene expression in Drosophila

Myosin heavy chain is present in unfused myoblastsof rost mutantsThe analysis of several rost alleles revealed that the majorityof myoblasts remain as mononucleated cells. These mesoder-mal cells turn off twist expression and start the muscle differ-entiation program as is evident by the expression of the char-acteristic β3 tubulin isotype. We addressed the question ofwhether these cells, during differentiation, express the geneticrepertoire of muscle-specific proteins, like myosin, despitetheir failure to build myotubes. With an antibody recognizingmyosin heavy chain (Kiehart and Feghali, 1986) that stains thecomplete set of fused muscles, we analyzed several EMS-induced alleles as well as the original P-element-mediatedmutant. As a representative example, we show the results forthe rost5 allele, which exhibits a medium mutant phenotypethat is very similar to the original P-element mutant. Althoughin wild-type embryos single myoblasts did not express myosin,unfused myoblasts in the rost allele showed expression of thisprotein (Fig. 4). This myosin expression in unfused myoblastsshows that these cells complete their developmental programindependently of the myotube formation.

Ectodermal derivatives, epidermis, and central andperipheral nervous system develop normally in rostmutantsThe rost mutation causes failure of myoblast fusions. Thisdefect may be due to the missing rost gene product inmyoblasts themselves. However, it cannot be ruled out that dis-tortions in the epidermis, for example in the muscle attachmentsites, or in the development of the nervous system, caused themuscle phenotype. No defect was detected in cuticular prepa-rations of rost mutants (data not shown). For the analysis ofthe nervous system, we performed double staining experi-ments. The peripheral and central nervous system were stainedwith the monoclonal antibody mAb22C10 (Zipursky et al.,1984) and the mesoderm with the β3 tubulin antibody (Leisset al., 1988). For the wild-type, staining of the nervous systemis shown in Fig. 5A and C in dark brown and muscles areshown in light brown. In homozygous rostP20 mutants, manyunfused β3-positive myoblasts (light brown) were observed. Inthese mutant embryos, the morphology of the central andperipheral nervous system was quite normal (Fig. 5B,D),though we cannot exclude that single cells were missing.

In addition, we were interested in the pattern of motoneu-rons in the rost mutant. Previously, it was shown that inner-vation is not an essential requirement for the early stages ofmyogenesis (Bate, 1990; Broadie and Bate, 1993). The initialmyoblast fusions begin with the onset of germband retraction(Bate, 1990), yet pioneering of the motor nerves does not startbefore completion of germband retraction. Nevertheless, themotor nerves are closely associated with the developingmyotubes for many hours prior to the establishment of themature muscle pattern (Johansen et al., 1993; Broadie andBate, 1993) so that this association might be essential for thecompletion of fusions of the myoblasts.

To examine the pattern of motoneurons in the rost mutant,we performed double stainings of motoneurons and muscles inhomozygous rost embryos. The results for the EMS-inducedallele rost5 are shown in Fig. 6. With the fas II antibody (a giftfrom G. Helt and C. S. Goodman, Grenningloh et al., 1991) we

visualized all intersegmental (ISN) and segmental (SN)neurons (Grenningloh et al., 1991; Seeger et al., 1993; VanVactor et al., 1993;) in parallel to the staining of mesodermalcells with the anti-β3 antibody (Fig. 6, black: fas II, brown: β3tubulin). In mid embryonic stages (stage 14), we found anobviously normal pattern (Fig. 6A) and guiding of motoneu-rons in the absence of multinucleated myotubes, which can bewell observed at a higher magnification (Fig. 6B). In laterstages of development (stage 16), the pattern of motoneuronsin rost mutants was strongly changed (not shown), which ispresumably the result of the missing muscles.

DISCUSSION

The somatic musculature develops from mesodermal stemcells that are determined as myoblasts. During muscle devel-opment, myoblasts fuse with each other to form multinucleatedmyofibers (see Introduction). Muscle fibers insert intoapodemes and it has been suggested that this connectioninduces a regulatory cascade leading to β1 tubulin expression(Buttgereit, 1993). Previous analysis of deletion mutants of theX-chromosome revealed that deletion of a number of locidisturbs the muscle pattern (Drysdale et al., 1993); further-more, a P-element mutation has been detected, leading to astrongly reduced number of myotubes (Burchard et al., 1995).Here,we analyze the gene rolling stone (rost) which is essentialfor the fusion of myoblasts to myotubes. In mutant embryos,only a few myotubes are formed. The majority of myoblasts,however, do not fuse or only fuse partially, as is visualized byβ3 tubulin staining. However, the program of gene expressionproceeds, as is evident from the decrease of twist expression inthe wild-type, activation of β3 tubulin and myosin heavy chainexpression, the last being a typical gene product of maturemuscle fibers in wild-type embryos (Kiehart and Feghali,1986). In rost mutants, however, myosin is detected also in theunfused myoblasts. This shows that the fusion process is dis-pensable for the expression of this protein. The rost mutantphenotype may be a consequence of defects in the mesodermitself or of signals derived from ectodermal derivatives such asthe nervous system or the apodemes (muscle attachment sites).In homozygous rost mutants, mesoderm formation is quitenormal. However, we cannot exclude that some mesodermalcells are missing. In addition, the morphology of the CNS andPNS, including the motoneurons, seems normal in early stagesof development. In later stages, the motoneuron pattern showsstrong morphological abberations in a rost mutant allele. In theabsence of most of the somatic muscles, this modificationcould be a secondary effect driven by the dramatic defects inthe mesoderm. We conclude that the correct outgrowth andguidance of motoneurons is independent of the fusion ofmyoblasts to myotubes during muscle formation. Broadie andBate (1993) could not detect significant differences betweenmuscle development in the presence or absence of motorneu-ron innervation, showing that innervation is not a prerequisitefor muscle formation. Due to the limited resolution of antibodystainings, we cannot exclude disorganization of a small popu-lation of nerve cells. Our investigation of the mesodermformation and of the morphology of the nervous system allowsus to state that the gross organization is not disturbed. This isin contrast to the previously characterized gene nem, which is

2618

essential for the formation of the complete set of somaticmuscles, as well as for the development of the pentascolopidialcells of the peripheral nervous system (Burchard et al., 1995).In rost mutants, as well as in nem mutants, the visceralmesoderm, the dorsal vessel and the pharynx musculaturedevelop properly. This parallels the early separation of regula-tory pathways in these tissues for the β3 tubulin gene (Gaschet al., 1989, Hinz et al., 1992).

A similar phenotype has been observed as a consequence ofoverexpression of a mutated GTPase Drac1 in the mesoderm

(Luo et al., 1994). The role of this protein in the muscle-forming process in wild-type development remains to bedetermined. In some deletion mutants described by Drysdaleet al. (1993), unfused myoblasts were also observed. Interest-ingly, one of the EMS-induced nem alleles (nem22) also showsfusion distortions (Burchard et al., 1995). In this case,myoblasts still aggregate, which is in contrast to rost mutants.

Other mutants are known to be essential for the developmentof the visceral musculature but have only moderate effects onthe somatic musculature (Bodmer et al., 1993; Azpiazu andFrasch, 1993). In tinman mutants, the muscle pattern issomewhat disorganized and the muscle number is usuallylower than normal. In mutants for bagpipe, a gene required forthe specification of the visceral mesoderm, the somatic mus-culature appears normal (Azpiazu and Frasch, 1993).

There are two classes of cell types that are essential for theformation of the somatic musculature. First, there is a class offounder cells that precedes muscle formation. Second, there isa class of fusion competent cells that fuse with the founder cellsto form muscle precursors and muscles (Bate, 1990). Apossible function of the rost gene could be to define the foundercells or to convert premature founder cells to fusion-competentfounder cells. Failure of rost expression may have the conse-quence that the founder cells cannot be specified and the fusionprocess is disturbed.

So far a few genes that are expressed in muscle precursorcells have been isolated and their gene products localized insitu, for example nautilus, the MyoD homologue of Drosophila

A. Paululat, S. Burchard and R. Renkawitz-Pohl

Fig. 4. Lateral view of a homozygous stage 16 rost5 mutant.Expression of muscle myosin in unfused myoblasts and myotubes isobvious.

Fig. 5. The central and peripheral nervous system in homozygous rostP20 embryos. The embryos were stained with the β3 tubulin antibody(Leiss et al., 1988) to detect the musculature and with mAb22C10 (Zipursky et al., 1984) to stain the central and peripheral nervous system. (A) A lateral view of a wild-type embryo showing the pleural musculature (light-brown staining) and dorsal hair sensilla (dh) and the lateralchordotonal organs (lch5) of the peripheral nervous system (dark-brown staining). (B) Lateral view of a homozygous rostP20 mutant reveals nodefects in the dorsal hair sensilla (dh) and the lateral chordotonal organs (Ich 5). (C) The ventral cord (vc) of a wild-type central nervous systemand the ventral musculature (vem) are visible. (D) The ventral cord (vc) of a mutant embryo shows no obvious defects. Unfused myoblasts andsome myotubes of the ventral musculature (vem) are stained with the β3 tubulin antibody.

2619Myoblast fusion and rost gene expression in Drosophila

(see Introduction). We started to analyse the presence ofmuscle precursors in rost mutants by the detection of nautilusexpression with in situ hybridizations.

We found an expression of nautilus in the rolling stonemutant from the extended germband stage onwards. In com-parison to the wild type, all nautilus-expressing precursorsare visible in the rost mutant. This result indicates that thefusion defect in the rost mutant is not caused by the absenceof muscle precursors. The early muscle differentiation, char-acterized by the formation of muscle precursors is notdisturbed in rost embryos as far as nautilus expressionindicates. The high number of β3 tubulin- and myosin-expressing cells, which are not fusion competent, indicatesthat the cell specificity of myogenic cells is determined. Inconclusion we found that the precursors and the cells to fusewith them are present in rost embryos. Furthermore theexpression of nautilus in the rost mutant implicates thatnautilus is independent of rost function.

Many morphogenetic events like cell adhesion, cellmigration, cell fusion, shaping, attachment to apodemes, inner-vation and expression of muscle-specific proteins are essentialfor muscle development. Furthermore, individual muscles gaintheir identity under control of homeotic gene products (Hooper,

1986). The homeobox-containing gene S59 may also fall intothis class, as this gene is active in specific muscle precursors(Dohrmann et al., 1990). Besides transcription factors, proteinsmediating cell-cell adhesion certainly play a role in these fusionprocesses. Because myogenesis involves the fusion of mononu-cleated myoblasts to multinucleated myotubes, the recognitionand adhesion steps are believed to be important for this highlyspecific process. A number of adhesion molecules such as N-CAM (e.g. Knudsen et al., 1990a,b) and integrins seems to beimportant in myoblast fusion. Molecules like integrins areessential for muscle attachment to apodemes as is evident in themutant lethal (1) myospheroid, which eliminates the PSβsubunit (Newman and Wright, 1981; MacKrell et al., 1988;Leptin et al., 1989). Recently, a cadherin isotype was identified(Donalies et al., 1991; Moore and Walsh, 1993) that plays a rolein myogenesis.

We have started the cloning and molecular analysis of therost gene, which in combination with the molecular analysisof the EMS-induced alleles should elucidate the function ofthis gene. Determination of the spatiotemporal pattern ofexpression will indicate whether the rost gene encodes a func-tionally important molecule of the muscle cells themselves orwhether the rost gene is transcribed in another tissue but influ-ences muscle development.

We thank Ruth Hyland for excellent technical assistence, DetlevButtgereit for stimulating discussions and Heike Sauer for hersecretary work. We acknowledge D. Kiehard for providing the anti-myosin antibody, L. Zipursky for providing mAb22C10, G. Helt andC. Goodman for the fas II antibody. We thank also Marcos A.Gonzales Gaitan for help with whole-mount in situ hybrization, AlanMichelson for providing the nautilus cDNA probe and MarkusAffolter and Uwe Walldorf for the CyO7.1 balancer strain. We arevery grateful to Günther Korge for help with the interpretation ofpolytene chromosome in situ hybridizations. This work was supportedby grants from the Deutsche Forschungsgemeinschaft (Re628/7-1/7-2) and the Fonds der Chemischen Industrie to R. R.-P.

REFERENCES

Abmayr, S. M., Michelson, A. M., Corbin, V., Young, M. W. and Maniatis,T. (1992). nautilus, a Drosophila member of the myogenic regulatory genefamily. In Neuromucular Development and Disease (ed. A. M. Kelly and H.M. Blau). New York: Raven Press, Ltd.

Affolter, M., Walldorf, U., Kloter, U., Schier, A. F. and Gehring, W. J.(1993). Regional repression of a Drosophila POU box gene in the endoderminvolves inductive interactions between germ layers. Development 117,1199-1210.

Azpiazu, N. and Frasch, M. (1993). Tinman and bagpipe: two homeo boxgenes that determine cell fate in the dorsal mesoderm of Drosophila. GenesDev. 7, 1325-1340.

Barad, M., Jack, T., Chadwick, R. and McGinnis, W. (1988). A novel,tissue-specific, Drosophila homeobox gene. EMBO J. 7, 2151-2161.

Bate, M. (1990). The embryonic development of larval muscles in Drosophila.Development 110, 791-804.

Bate, M., Rushton, E. and Currie, D. A. (1991). Cells with persistent twistexpression are the embryonic precursors of adult muscles in Drosophila.Development 113, 79-89.

Bodmer, R., Jan, L. Y. and Jan, Y. N. (1990). A new homeobox-containinggene, msh-2, is transiently expressed early during mesoderm formation inDrosophila. Development 110, 661-669.

Bodmer, R. (1993). The gene tinman is required for specification of the heartand visceral muscles in Drosophila. Development 118, 719-729.

Bouley, J. L., Dennefeld, C. and Alberga, A. (1987). The Drosophiladevelopmental gene snail encodes a protein with nucleic acid bindingfingers. Nature 330, 395-398.

Fig. 6. Motoneuron development in homozygous rost5 mutants.Double staining of motoneurons (shown in black) and musculature(shown in brown) in homozygous rost embryos visualized with thefas II antibody (Grenningloh et al., 1991) and β3 tubulin antibody(Leiss et al., 1988). Intersegmental (ISN) and segmental neurons(SN) are marked. (A) This homozygous rost5 embryo at stage 14 ofembryogenesis shows a normal segmentally arranged motoneuronpattern. (B) Higher magnification of the embryo shown in A clearlyreveals independance of motoneuron outgrowth and muscle fusions.

2620

Bourgouin, C., Lundgren, S. E. and Thomas, J. B. (1992). apterous is aDrosophila LIM domain gene required for the development of a subset ofembryonic muscles. Neuron 9, 549-561.

Braun, T., Rudnicki, M. A., Arnold, H. H. and Jaenisch, R. (1992). Targetedinactivation of the muscle regulatory gene Myf-5 results in abnormal ribdevelopment and perinatal death. Cell 71, 369-382.

Broadie, K. and Bate, M. (1993). Muscle development is independent ofinnervation during Drosophila embryogenesis. Development 119, 533-543

Burchard, S., Paululat, A., Hinz, U. and Renkawitz-Pohl, R. (1995). Themutant not enough muscles (nem) reveals reduction of the Drosophilaembryonic muscle pattern. J. Cell Sci. 108, 1443-1454.

Buttgereit, D. (1993). Redundant enhancer elements guide β1 tubulinexpression in apodemes during Drosophila embryogenesis. J. Cell Sci. 105,721-727.

Campos-Ortega, J. A. and Hartenstein, V. (1985). The EmbryonicDevelopment of Drosophila melanogaster. Berlin and Heidelberg: SpringerVerlag.

Cohen, B., McGuffin, E., Pfeifle, C., Segal, D. and Cohen, S. M. (1992).apterous, a gene required for imaginal disc development in Drosophilaencodes a member of the LIM family of developmental regulatory proteins.Genes Dev. 6, 715-729.

Cooley, L., Berg, C. and Spradling, A. (1988). Controlling P-elementinsertional mutagenesis. Trends Genet. 4, 254-258.

Corbin, V., Michelson, A. M., Abmayr, S. M., Neel, V., Alcamo, E.,Maniatis, T. and Young, M. W. (1991). A role for the Drosophilaneurogenic genes in mesoderm differentiation. Cell 67, 311-323.

Crossley, A. C. (1978). The morphology and development of the Drosophilamuscular system. In The Genetics and Development of Drosophila Vol. 2b(eds Ashburner, M. and Wright, T.), pp 499-560. New York: Academicpress.

Davis, R. L., Weintraub, H. and Lassar, A. B. (1987). Expression of a singletransfected cDNA converts fibroblasts to myoblasts. Cell 51, 987-1000.

Dohrmann, C., Azpiazu, N. and Frasch, M. (1990). A new Drosophilahomeo box gene is expressed in mesodermal precursor cells of distinctmuscles during embryogenesis. Genes Dev. 4, 2098-2111.

Donalies, M., Cramer, M., Ringwald, M. and Starzinski-Powitz, A. (1991).Expression of M-cadherin, a member of the cadherin multigene family,correlates differentiation of skeletal muscle cells. Proc. Natl. Acad. Sci. USA88, 8024-8028.

Drysdale, R., Rushton, E. and Bate, M. (1993). Genes required for embryonicmuscle development in Drosophila melanogaster. Roux’s Arch. Dev. Biol.202, 279-295.

Gasch, A., Hinz, U. and Renkawitz-Pohl, R. (1989). Intron and upstreamsequences regulate expression of the Drosophila β3 tubulin gene in thevisceral and somatic musculature, respectively. Proc. Natl. Acad. Sci. USA86, 3215-3218.

Grenningloh, G., Rehm, E. J. and Goodman, C. S. (1991). Genetic analysisof growth cone guidance in Drosophila: fasciclin II functions as a neuronalrecognition molecule. Cell 67, 45-57.

Hasty, P. Bradley, A., Morris, J. H., Edmondson, D. G., Venuti, J. M.,Olson, E. and Klein, W. H. (1993). Muscle deficiency and neonatal death inmice with a target mutation in the myogenin gene. Nature 364, 501-506.

Hinz, U., Wolk, A. and Renkawitz-Pohl, R. (1992). Ultrabithorax is aregulator of β3 tubulin expression in the Drosophila visceral mesoderm.Development 116, 543-554.

Hooper, J. E. (1986). Homeotic gene function in the muscles of Drosophilalarvae. EMBO J. 5, 2321-2329.

Johansen, J., Halpern, M. E. and Keshishian, H. (1993). Axonal guidanceand the development of muscle fiber-specific innervation in Drosophilaembryos. J. Neurosci. 9, 4318-4332.

Karess, R. E. and Rubin, G. M. (1984). Analysis of P transposable elementfunctions in Drosophila. Cell 38, 135-146.

Kiehart, D. P. and Feghali, R. (1986). Cytoplasmic myosin from Drosophilamelanogaster. J. Cell Biol. 103, 1517-1525.

Knudsen, K. D., Myers, L. and McElwee, S. A. (1990a). A role for the Ca2+-dependent adhesion molecule, N-cadherin, in myoblast interaction duringmyogenesis. Exp. Cell. Res. 188, 175-184.

Knudsen, K. A., McElwee, S. A. and Myers, L. (1990b). A role for the neuraladhesion molecule, N-CAM, in myoblast interaction during myogenesis.Dev. Biol. 138, 159-168.

Lai, Z.-C., Rushton, E., Bate, M. and Rubin, G. M. (1993). Loss of functionof the Drosophila zfh-1 gene results in abnormal development of mesodermalderived tissues. Proc. Natl. Acad. Sci. USA 90, 4122-4126.

Lawrence, P. A., Johnston, P., MacDonald, P. and Struhl, G. (1987).

Borders of parasegments in Drosophila embryos are delimited by the fushitarazu and even-skipped genes. Nature 328, 440-442.

Leiss, D., Hinz, U., Gasch, A., Mertz, R. and Renkawitz-Pohl, R. (1988). β3tubulin expression characterizes the differentiating mesodermal germ layerduring Drosophila embryogenesis. Development 104, 525-531.

Leptin, M., Bogaert, T., Lehmann, R. and Wilcox, M. (1989). The functionof PS integrins during Drosophila embryogenesis. Cell 56, 401-408.

Leptin, M. and Grunewald, B. (1990). Cell shape changes during gastrulationin Drosophila. Development 110, 73-84.

Lewis, E. B. and Bacher, F. (1968). Methods of feeding ethyl methanesulfonate (EMS) to Drosophila males. Drosophila Inf. Service 43, 193.

Luo, L., Liao, Y. J., Jan, L. Y. and Jan, Y. N. (1994). Distinct morphogeneticfunctions of similar small GTPases: Drosophila Drac1 is involved in axonaloutgrowth and myoblast fusion. Genes Dev. 8, 1787-1802.

MacKrell, A. J., Blumberg, B., Haynes, S. R. and Fessler, J. H. (1988). Thelethal myospheroid gene of Drosophila encodes a membrane proteinhomologous to vertebrate integrine β subunits. Proc. Natl. Acad. Sci. USA85, 2633-2637.

Michelson, A. M., Abmayr, S. M., Bate, M., Martinez-Arias, A. andManiatis, T. (1990). Expression of a MyoD family member prefiguresmuscle pattern in Drosophila embryos. Genes Dev. 4, 2086-2097.

Moore, R. and Walsh, F. S. (1993). The cell adhesion molecule M-cadherin isspecifically expressed in developing and regenerating, but not denervatedskeletal muscle. Development 117, 1409-1420.

Nabeshima, Y., Hanaoka, K., Hayasaka, M., Esumi, E., Li, S., Nonaka, I.and Nabeshima, Y. (1993) Myogenin gene disruption results in perinatallethality because of severe muscle defects. Nature 364, 532-535.

Newman, S. M. and Wright, T. R. (1981). Histological and ultrastructuralanalysis of developmental defects produced by the mutantlethal(1)myospheroid in Drosophila melanogaster. Dev. Biol. 86, 393-402.

Olson, E. N. (1990). MyoD family: a paradigm for development? Genes Dev. 4,1454-1461.

Paterson, B. M., Walldorf, U., Eldridge, J., Dübendorfer, A., Frasch, M.and Gehring, W. J. (1991). The Drosophila homologue of vertebratemyogenic-determination genes encodes a transiently expressed nuclearprotein marking primary myogenic cells. Proc. Natl. Acad. Sci. USA 88,3782-3786.

Rudnicki, M. A., Braun, T., Hinuma, S. and Jaenisch, R. (1992).Inactivation of MyoD in mice leads to up-regulation of the myogenic HLHgene Myf-5 and results in apparently normal muscle development. Cell 71,383-390.

Rudnicki, M. A., Schnegelsberg, P. N. J., Stead, R. H., Braun, T., Arnold,H.-H. and Jaenisch, R. (1993). MyoD or Myf-5 is required for the formationof sceletal muscle. Cell 75, 1351-1359.

Seeger, M., Tear, G., Ferres-Marco, D. and Goodman, C. S. (1993).Mutations affecting growth cone guidance in Drosophila: Genes necessaryfor guidance toward or away from the midline. Neuron 10, 409-426.

St. Johnston, R. D. and Nüsslein-Volhard, C. (1992). The origin of patternand polarity in the Drosophila embryo. Cell 68, 1-20.

Steward, R. and Govind, S. (1993). Dorsal-ventral polarity in the Drosophilaembryo. Current Opinion in Genetics and Development 3, 556-561.

Tautz, D. and Pfeiffle, C. (1989). A non-radioactive in situ hybridizationmethod for the localization of specific RNAs in Drosophila embryos revealstranslational control of the segmentation gene hunchback. Chromosoma 98,81-85.

Thisse, B., El Messal, M. and Perrin-Schmitt, F. (1987). The twist gene:isolation of a zygotic gene necessary for the establishment of dorso-ventralpattern. Nucl. Acids Res. 15, 3439-3453.

Thisse, B., Stoetzel, C., Gorostiza-Thisse, C. and Perrin-Schmitt, F. (1988).Sequence of the twist gene and nuclear localization in endomesodermal cellsof early Drosophila embryos. EMBO J. 7, 2175-2183.

Van Vactor, D., Sink, H., Fambrough, D., Tsoo, R. and Goodman, C. S.(1993). Genes that control neuromuscular specificity in Drosophila. Cell 73,1137-1153.

Weintraub, H., Tapscott, S. J., Davis, R. L., Thayer, M. J., Adam, M. A.,Lassar, A. B. and Miller, A. D. (1989). Activation of muscle-specific genesin pigment, nerve, fat, liver and fibroblast cell lines by forced expression ofMyoD. Proc. Natl. Acad. Sci. USA 86, 5434-5438.

Zipursky, S. L., Venkatesh, T. R., Teplow, D. B. and Benzer, S. (1984).Neuronal development in the Drosophila retina: monoclonal antibodies asmolecular probes. Cell 36, 15-26.

(Accepted 9 May 1995)

A. Paululat, S. Burchard and R. Renkawitz-Pohl