A dual requirement for neurogenic genes in Drosophila myogenesis

During embryogenesis Drosophila constructs a complex pattern of larval muscles beneath the newly forming epidermis. There are 30 muscles per hemisegment from A2 to A7, (comparable patterns in other segments) and each muscle has a highly specific set of properties: size, shape, orientation, attachment and innervation. Muscle formation requires that neighbouring cells fuse to form multinucleate muscle precursors, and that adjacent precursors segregate to form separate elements of the final pattern. Thus in some way, individual elements of the pattern are specified and segregate as groups of cells to form particular muscles. One model of this process suggests that for each muscle, a cell called a founder cell is specified in the mesoderm (Bate1990). A characteristic programme of differentiation is initiated in such a cell (or cells) and neighbouring cells are entrained to this programme as they fuse with the founder to form a syncytial muscle precursor. Indeed, it has been shown for one subset of the muscles, that myogenesis is associated with the appearance, at specific locations in the mesoderm, of cells that express the homeobox-containing gene, S59, and these cells have been described as founders (Dohrmann et al., 1990). Neighbouring cells fuse with these founders to form the precursors of a small number of S59- expressing muscles in each segment. More recently, a number of other putative transcription factors (such as apterous [Bourgouin et al., 1992] and vestigial [vg; Williams et al., 1991]) have been shown to be similarly expressed in subsets of the larval muscles and their apparent founder cells in the embryonic mesoderm (for review see Bate, 1993).

Virtually nothing is known of the actual mechanisms that underlie the events of myogenesis in the fly, but there is at least a formal similarity between these events and the process of neurogenesis in insect embryos (Bate, 1990 and see Campos-Ortega, 1993 for review). During embryonic neurogenesis, neuroblasts, the precursor cells of the central nervous system, are defined by their position in an epithelium and segregate from surrounding cells which will later differentiate to form the ventral epidermis. Each neuroblast now divides repeatedly to produce a specific family of cells, which is unique to its position in the neuroepithelium. During myogenesis, founder cells are defined in the developing mesoderm and then fuse with neighbouring cells to form a syncytium with highly specific properties, which segregates from its neighbours. In both cases specification and segregation are an integral part of pattern formation. Although the mechanisms by which neuroblasts are specified are not yet clear, it has been exhaustively demonstrated that the segregation of neuroblasts from epidermoblasts in the ectoderm of the embryonic fly depends on the function of the neurogenic group of genes that includes Notch (N), Delta (DI), big brain (bib), neuralised (neu), mastermind (mam) and the Enhancer of split (E(splf) complex. Loss of function in any of these genes disrupts the segregation process and leads to an overproduct ion of ncuroblasls at the expense of surrounding epithelial cells (for review see Campos Ortega. 1993). The products of the neurogenic genes are expressed in the embryonic mesoderm (Kidd et al., 1986; Knust et al., 1987; Johansen et al., 1989: Kopczynski and Muskavilch. 1989; Fehon et al., 1991; Haenlin el al. 1990; Rao et al., 1990; Smoller et al., 1990; Bclllerct al.. 1991; Kooh et al. 1993) and it has been shown that their functions are required for the normal development of a variety of mesodermal derivatives (Corbin ct al.. 1991; Ruohola ct al.. 1991; Hartenstcin ct al.. 1992). although the precise nature of this requirement remains unclear. Most interestingly, mesodermal domains of expression of nautilus (Michelson et al., 1990). the fly homologue of vertebrate Myol) (Weintraub et al., 1991), are clearly enlarged in embryos mutant for any of the neurogenic genes (Corbin et al., 1991). We have used a variety of methods to analyse the requirement for these genes in Drosophila myogenesis. 11ère we show that mutations in the neurogenic genes lead to an expansion of the domains of S59 and vg expression in the mesoderm and produce severe abnormalities in muscle patterning and muscle differentiation. These mutant phenotypes appear to relied two distinct requirements for neurogenic gene functions during myogenesis: the first is for the proper segregation of founder cells, the second for the proper differentiation of syncytial muscle libres. Whether these relied an autonomous requirement for the function of any or all of the neurogenic genes in the mesoderm itself is. as yet. unclear.

Fggs were collected from wild-type (Oregon R) and mutant (for details, see below) strains of Hies on agar/apple juice plates (Wieschaus and Niisslein-Volhard. 1986) and kept al 25°C. Embryogenesis lasts 21 hours at 25°C and embryonic stages arc given according to the scheme of Campos-Ortega and Harlcnstcin (1985).

The following strains of flies were used in this work: Dl^FX3 (Lehmann et al., 1983); Df(1)svr, N^55e11 (Brand and Campos Ortega. 1988). provided by J.A. Campos Ortega; bib^ID05;; mam^1J113:neu^IF65 (Lehmann et al., 1983); E(spl)^BX22 (Preiss et al., 1988): K(spl)^8D06(Df3RE(spl)l; Jürgens et al., 1984) provided by D. A. Hartley. We are very grateful for the provision of these Flies.

For immunocytochemistry, eggs were collected, dcchorionated, lixed and devitcllcnised according to the methods of Wieschaus and Niisslein Volhard (1986) and stained with antibodies using standard protocols and the Vectastain ABC Elite Kit from Veclalabs. The anti-vestigial antibody was generously provided by Sean Carroll.

Polarising and Nomarski microscopy of late embryonic muscles anil cuticles follows the protocol given by Drysdale et al. (1993). In summary, late embryos are dcchorionated and placed singly in saline on a slide beneath a coverslip. The saline layer is then sucked from beneath the coverslip with a tissue, so that the coverslip Hattens the embryo, extruding the internal organs and leaving a Battened cuticle with attached muscles. This preparation is then viewed on (he stage of a Nikon polarising microscope (to view (he muscle birefringence between crossed polarised beams) before transfer to a Zeiss Axiophol microscope (to sec the cuticle pattern with Nomarski optics). Photographs were taken on Kodak Technical Pan in both cases.

Embryos were dissected anil stained with toluidine blue according to the protocol given by Truman and Bate (1988).

Muscle nomenclature is according to the scheme given in Bate (1993).

In wild-type embryos, S59 is expressed in developing muscles (Fig. 1). in the embryonic CNS and in a small sector of the midgut. Expression begins about 6–7 hours after egg laying, in a single cell in the ventral mesoderm of each hemisegment. This cell divides to produce a pair of S59-expressing progeny. Al the end of stage I I a second cluster of S59-expressing cells appears in the ventral mesoderm. 4 cells in each abdominal hemisegment. but larger clusters in the thorax. As the germ band retracts, a pair of more dorsal mesodermal cells begins to express S59 in each abdominal hemisegment (Fig. 2). These cells are not present in the thorax, and there arc special patterns posterior to A7. so that, in all. 7 pairs of more dorsal S59-cxprcssing cells segregate out on each side as the germ band retracts. These early patterns of mesodermal expression arc a prelude to muscle formation and the recruitment of additional S59-cxpressing nuclei as myoblasts fuse with the S59-cxprcssing muscle founders during and after germ band retraction. The final pattern of S59-cxpressing muscles consists of 3 muscles in each abdominal hemisegment (muscles VT1. VA2 [ventral] and muscle DT1 [dorsolateral]; nomenclature from Bate. 1993), with special patterns in the thorax and telson (Fig. 1). CNS expression begins shortly after the onset of expression in the mesoderm and evolves into a segmenlally repeated pattern of presumed ganglion cells after germ band retraction. Cells in the midgut begin to express .S’59 after the fusion of the anterior and posterior midgut primordia in 12-to 13-hour embryos. A detailed description of these expression patterns is given by Dohrmann el al. (1990).

In embryos homozygous or hemi/.ygous for mutations in any of the neurogenic genes we have looked al (Dl, N, E(spl), bib. mam, nen) we find a characteristic derangement of the normal pattern of mesodermal S59 expression in the extended germ hand (Fig. 2). At stage 11 we find that (he expansion of S59 expression in the CNS expected in a neurogenic mutant embryo is accompanied by the formation of enlarged clusters of S59-exprcssing cells in the ventral mesoderm. These clusters arc located in the expected positions for wild type expression, but they arc enlarged to a degree characteristic for the mutation involved and apparently closely correlated with the severity of the neurogenic phenotype (Fig. 2). In extreme cases, adjacent clusters coalesce, so that the exact number of cells in any cluster is difficult to resolve. As the germ band retracts, the more dorsal patches of mesodermal S59 expression appear in the mutant embryos. As in the wild type these dorsal patches are restricted to the abdominal segments where the precursor of muscle DTI normally forms. In mutant embryos the patches are enlarged and. once again, the degree of derangement apparent in these more dorsal cells is characteristic of the mutation concerned (Fig. 3). Thus. as the germ band retracts, each of the mutant embryos shows a complex pattern of expanded S59-cxprcssing cells dorsal to the developing CNS which itself shows a hyperplasie pattern of S59 expression.

Unexpectedly, al later stages, ventral S59 expression (corresponding to muscles VT1 and VA2 in the wiki type) disappears from the mesoderm of all the mutant embryos except for a few scattered cells and patches of persistent expression adjacent to the more ventral epidermis, which remains in embryos with a weaker neurogenic phenotype. The dorsal expression remains in more or less expanded clusters beneath the epidermis which occurs outside the domain of the expanded nervous system. That these clusters are indeed equivalent to the dorsal clusters of wild type embryos is shown by their segment specificity: they are only present in the abdomen where they form 7 distinct groups in each of the mutants, mirroring the segment specificity of muscle DTI in the wild type (Fig. 3). Once again, the degree of aberration in these persistent clusters is closely correlated with the strength of the neurogenic phenotype. In the weaker alleles, dorsal cells migrate into the characteristic strip-like arrangement of differentiating muscle, as found in wild type embryos (Fig. 3). These muscle-like configurations only appear in late embryos (more than 13 hours AEL). but it is important to note that, in such cases, the muscles have formed, or appeared to form from a larger than normal cluster of S59-cxpressing cells in the retracting germ band (Fig. 2). Thus in the weaker mutants, there is an apparent correlation between the expansion of the nervous system and the degree of disruption to late phases of muscle differentiation. In such mutants the ventral mesoderm (close to the expanded nervous system) is deranged, whereas dorsal muscles (associated with apparently undisturbed epidermis) may form relatively normally. Dorsally and vertically however, in such mutant embryos, the clusters of S59-cxpressing cells are expanded as the germ band shortens.

We used a second marker, vg. to examine the effects of loss of neurogenic gene functions on a different subset of the forming muscles, vg encodes a novel nuclear protein (Williams et al., 1991). which, like S59. is expressed in a subset of the larval muscles, and also in the developing CNS (Fig. 1), vg too is expressed in small numbers of mesodermal cells prior to germ band retraction (Fig. 4) and these appear to act as muscle founder cells, recruiting neighbouring cells by fusion to form the syncytial precursors of the vg-expressing muscle subset. The sequence and pattern of mesodermal vg expression has yet to be described in detail (Bale et al., unpublished data), but it is clear that vg-expressing ventral longitudinal muscles are prefigured by a regular arrangement of small numbers of vg-expressing cells sitting just internal to the CNS in the late stage 11 embryo (Fig. 4). In neurogenic mutant embryos these cells arc replaced by enlarged clusters of vg-expressing cells al equivalent positions overlying the enlarged C’NS (Fig. 4). In mutant embryos with a strong neurogenic phenotype, ventral vg expression is lost as development proceeds. In weaker alleles, such as mam¹¹¹¹’, where there is still considerable epidermis formed vcntrally, vg expression persists ventrally, and vg-expressing cells are clearly incorporated into longitudinally arranged muscle libres (big. 5). ‘I’hus the normal segregation of the vg-expressing muscle founder cells, like their S59-cxpressing neighbours, requires neurogenic gene functions and continued rg expression and the formation of t’g-expressing muscle libres may require the presence of epidermis.

Late wild type embryos (just prior to hatching) show striking patterns of birefringence when Battened on a slide and viewed with a polarising microscope (Drysdale el al 11993]; sec also Materials and methods and Fig. 6). The birefringence is largely caused by the highly oriented structure of the contractile proteins in the muscle myofibrils, together with a smaller contribution from crystals of uric acid in the Malpighian tubules, which are randomly extruded from the embryo as il Hattens. Muscle birefringence can be used lo assay both muscle differentiation ami muscle patterning, and we have used it lo look al the later development of muscles in neurogenic mutant embryos (Fig. 7). In each case we selected embryos that were al least 24 hours old. but whose internal tissues had not yet begun to degenerate, flattened them on slides and compared the phenotypes seen with polarised light (muscles, Malpighian tubules) with those seen with Nomarski optics in bright-iicld illumination (cuticle, tracheae). These preparations revealed a striking parallel between the degree of muscle differentiation and the extent of the differentiated cuticle formed in each of the mutant embryos. All the mutants have birefringence indicating that they succeed in differentiating myobrils with contractile proteins (Fig. 7). Weaker alleles (bib^ID05, mam^IJ113, E(spl)^BX22) also result in fused muscles in patterned arrangements, particularly in the more dorsal regions of the embryo, whereas stronger alleles result in few if any fused muscles, although the unfused. biréfringent libres may, nonetheless, have a patterned arrangement. In extreme cases (Dl^FX3, E(spl)^8DO6) the birefringence reveals a rather random pattern of line fibres apparently radiating out from the small fragments of dorsal cuticle, which arc still present in these embryos (Fig. 7).

To explore the possibility that the mesodermal phenotypes we observe in neurogenic mutants are simply a secondary consequence of the enlargement of the central and peripheral nervous system and the concomitant reduction of the epidermis, we used die experimental approach adopted by Corbin et al. (1991) and looked at the phenotype of embryos that were doubly mutant for N and for a deficiency which removes the Achactc-Scutc complex (AS-C). Embryos that arc homozygous for deficiencies in AS-C fail to develop the proper complement of neuroblasts, and in severe cases, most of the CNS is missing (see Campos Ortega. 1993 for review). In addition, the PNS is reduced and, in the case of Df( 1 )svr (which removes the entire AS-C. together with the neighbouring genes elav and vnd. which arc also required for CNS development | Garcia Bellido and Santamaria. 1978: Garcia-Bellido. 1979; Jimenez and Campos-Ortega. 1979. 1987. 1990: While. 1980; Campos et al., 1985]), all PNS neurons arc missing, except for the chordotonal organs. In the double mutants there is a substantial rescue of the epidermal phenotype in lateral and dorsal regions although vcntrally the epidermis, and the CNS arc largely missing (Brand and Campos-Ortega. 1988; our observations: see Fig. 7). Following the experiments of Corbin cl al. (1991) we reasoned that if the mesodermal phenotype, or some aspects of it, were secondary to the loss or derangement of epidermal cells inherent in the neurogenic phenotype, then these mesodermal effects, like the epidermal phenotype, should be ameliorated in the double mutant.

We find (M. B. and E. R.. unpublished observations) that S59-expression is deranged in the ventral mesoderm of DJ( I ).wr embryos, but dorsally. expression is normal, and muscle DTI develops in its appropriate position. In the double mutants, we therefore concentrated our attention on the dorsal expression pattern. In single mutant N^55e11 embryos, we find expanded clusters of S59-expressing cells as the germ band retracts, and these persist in the shortened embryos, in a more or less scattered arrangement, with no sign of aggregation to form a normal muscle (Fig. 8). In the double mutants we also find expanded domains of S59- expressing cells dorsally, even though the enlarged anlage of the CNS is no longer present, and the extent of the epidermis appears to be far greater than in N^55e11 (Figs 7, 8). These expanded clusters persist as the germ band retracts and form large, compact, conspicuous clusters in the shortened embryo. Interestingly, unlike the case of N^55e11, there is also some persistent ventral expression in the double mutant at late stages, and this appears to correlate closely with the expanded lateral epidermis present in these embryos. Clearly, the amelioration of the nervous system phenotype in the double mutants does not reduce the expansion of the dorsal S59-expressing clusters. On the other hand, the late loss of S59-expression from ventral mesodermal cells characteristic of N^55e11 mutants is partially rescued in the double mutant Df( 1 )svr;N^55e11 and this rescue appears to be related to the increased territory covered by epidermal cells in these embryos.

The single mutant, N^55e11, has a characteristic phenotype in polarised light (Fig. 7K,L), which fits into the series described above, in which increasing severity of the neurogenic phenotype correlates with apparent loss of myoblast fusion and muscle patterning. A shield of cuticle forms dorsally and, radiating out from this, there is a fan of biréfringent fibres, with a few fibres crossing the shield itself. By contrast, in the double mutant, Df( 1 )svr;N^55e11 there is a substantial rescue of both the cuticular and the muscle phenotypes (Fig. 71,J). The cuticle has expanded both in the anteroposterior and in the dorsoventral axes, although there is still no sign of ventral denticle belts. The muscle phenotype shows obvious signs of the strap-like structures typical of fused muscle fibres, and many of the fibres are oriented and attached in a clearly patterned fashion. It is only at the ventral margins of the cuticular shield that this organisation breaks down and here there are signs of disorganised and unfused muscle fibres. Thus the phenotype of the single and double mutant embryos seen in polarised light reveals a strong correlation between the strength of effects on the epidermis and nervous system and the degree of disorganisation in the late myogenic phenotype associated with muscle differentiation.

Although birefringence gives a good indication of the degree of myoblast fusion during myogenesis, it is not possible to say unequivocally from the birefringence alone, whether fusion has or has not occurred. To assess the degree of myoblast fusion in different neurogenic mutants we dissected wild type and mutant (recognisable by their expanded nervous systems) embryos 12-13 hours AEL (stage 16) and stained them with toluidine blue. In wild type embryos, preparations of this kind can be used to follow the sequence of muscle differentiation from the onset of myoblast fusion (Bate, 1990). Using this method we were able to find fused cells in the somatic mesoderm of all the neurogenic mutants, with the possible exception of E(spl)^8DO6. As with the phenotype of the birefringence, there is a clear correlation between the extent of the neurogenic phenotype, that is the expansion of the CNS and PNS, and the loss of epidermis, and the degree to which myoblast fusion occurs. Thus, in the weaker mutants there are well formed syncytial muscles in an orderly arrangement, inserted on the dorsal and lateral epidermis, as predicted from the observations using the polarising microscope, whereas in the strongest mutants, there is little apparent order in the mesoderm and cell fusion occurs rarely, to produce occasional bi- or trinucleate syncytia. In E(spl)^8DO6, we did not even find rare fusions, but this finding should be treated with caution - the embryos are difficult to dissect, and the number of fusion events that we see in other strong neurogenic mutants (Dl^FX3, N^55e11; neu^lF65) is small. We do not exclude the possibility of rare fusion occurring in E(spl)^8DO6. In those embryos such as mam and bib with large numbers of syncytial muscle fibres, relatively large well-formed muscles attach to the most dorsal regions of the epidermis. Ventrally, there is a tendency for the normal pattern of muscles to be replaced by large numbers of thinner, syncytial, spindle-shaped fibres arranged in parallel.

All the mutants, again with the possible exception of E(spl)^8DO6, show a striking phenotype in the relatively disorganised mesoderm where syncytial muscles fail to form. Concentrated in particular on the margins of the expanded nervous system abutting the edges of the epidermis, there are conspicuous clusters of prominent mesodermal cells, which in some cases appear to form rosettes about a central focus (Fig. 9). These cells are unfused and may represent a default state adopted when fusion fails to occur among muscle forming cells. This view is strengthened by observations in weaker mutants and in the double mutant Df(I)svr;N^55e11, where the unfused cells in these marginal clusters are often associated with cells which have fused to form bi- or trinucleate syncytia, but have not differentiated into well formed muscles (Fig. 9). Dorsally in the double mutant, the almost complete absence of fusion seen in N^55e11 is rescued and syncytial muscles develop and insert on the expanded epidermis, again confirming our observations with the polarising microscope. Summarising, we find that the degree of myoblast fusion is closely related to the extent of the epidermal territory in a neurogenic mutant. Where cells fail to fuse, conspicuous clusters of mesodermal cells are formed, and if epidermal territories are expanded, cells in these clusters may be recruited to fusion.

The muscles of the Drosophila larva form a complex pattern of individual units each of which has a characteristic and distinctive set of properties. Muscles differ in, and can be recognised by, their position, insertion points, size, orientation and innervation. These properties are manifested by groups of myoblasts which, during and shortly after germ band retraction aggregate and fuse to form syncytial precursors for each of the differentiated muscles. Thus, the muscle pattern is formed by a mechanism which segregates mesodermal cells into separate and distinct muscle forming groups. There arc two alternative ways in which the properties of the syncytial muscle precursors might he defined. Since cell fusion can lead to the entrainment of nuclei in a shared cytoplasm to common patterns of gene expression, the properties of the syncytium could be defined by the prior segregation of a set of founder cells, each of which would seed the formation of a different precursor by fusing with neighbouring cells. Alternatively, cells could be assigned in groups, the members of each group fusing to form a particular syncytial precursor. The founder cell hypothesis is consistent with the developmental history of the S59-expressing muscle subset. Expression begins in a small number of mesodermal cells with which surrounding myoblasts fuse to form the precursors of the three S59-exprcssing muscles in each segment. As they fuse, these myoblasts are recruited to S59-expression. Although S59 is the type example of such a “founder cell gene” there is an increasing catalogue of genes with similar expression patterns that define other subsets of the larval musculature and which are expressed in small numbers of putative founder cells prior to germ band retraction (for review, see Bate. 1993).

In this paper we have shown that the normally limited domains of S59-expression are expanded in the mesoderm of embryos mutant for the neurogenic genes that we have studied. A second marker, vg, shows that this effect is not restricted to those myoblasts that express S59. but probably represents a general requirement for neurogenic gene functions during the segregation of all muscle precursors. In addition, in the mutant embryos, we find that later phases in the sequence of myogenesis arc deranged: there arc defects in fusion and there is a premature loss of S59-expression from regions of the mesoderm which initially express it. The relationship between the two phenotypes (expanded domains of S59-expression and disrupted myogenesis) and the neurogenic phenotype is different. S59-cxprcssing clusters are expanded dorsally and ventrally, so that the expansion of the domains of S59 expression may be independent of both the expansion of the nervous system and the consequent loss of epidermis, both of which are most prominent ventrally. However, the disruption of muscle differentiation seems to be closely related to the enlargement of the nervous system. The degree of disruption of myogenesis closely follows the extent to which the epidermis is reduced and the nervous system expanded and in the weaker mutants there is an obvious ventral to dorsal decline in the disruption to the muscles. A priori there is no simple reason for such a gradation in the phenotype unless the disruption to myogenesis is directly linked to the derangement of the ventral neurogenic region. In addition, when the nervous system phenotype is partially rescued (Df(I)svr;N^55e11) there is a dramatic improvement in myogenesis. By contrast, the double mutant embryos still have expanded domains of S59 expression. Thus, the double mutants emphasise the relative independence of the expansion of S59-expressing clusters from the nervous system phenotype, as against the close relationship between the degree of expansion of the nervous system and the extent of disruption to myogenesis. One explanation for this finding would be that there is a direct requirement for neurogenic gene functions in the mesoderm for the proper segregation of S59-expressing cells, whereas the later effects on myogenesis are a secondary consequence of the disruptive effects of the neurogenic phenotype on the proper patterning and differentiation of the ectoderm.

All the neurogenic genes are expressed in the mesoderm during embryogenesis (Kidd et al., 1986; Knust et al., 1987; Johansen et al., 1989; Kopczynski and Muskavitch, 1989, Fehon et al., 1990; Haenlin et al., 1990; Rao et al., 1990; Smoller et al., 1990; Bettier et al., 1991; Kooh et al., 1993), but without further experiments it is impossible to say whether any or all of the effects we describe here are a direct consequence of loss of functions in the mesoderm itself. It could be, for example, that loss of neurogenic genes causes a general derangement of the ectoderm, which indirectly causes a disruption to the normal patterning of mesodermal cells immediately adjacent to it. Experiments in other insects have shown that the ectoderm is essential for normal mesodermal development. In particular Bock (Seidel et al., 1940) found that the differentiated cell type formed by any given sector of the mesoderm depends on the region of the ectoderm that it comes to underlie after migrating inwards at gastrulation. These experiments strongly suggest that there is an inductive role for the ectoderm in mesodermal patterning in insects. However, the partitioning of the mesoderm that Bock studied, i.e. its subdivision into different sectors (heart, fat body, visceral, somatic and gonadal mesoderm) is successfully achieved in many if not all the mutants that we looked at (Hartenstein et al., 1992) and it is not clear that the inductive effects that Bock inferred would be directly relevant for a process like the selection of mesodermal cells for S59 expression. A simpler interpretation of the results we report here is that, during the development of the wild type embryo, domains of prospective gene expression are established in the somatic mesoderm from which the muscles will be formed. Within these domains, individual myoblasts are selected for expression by a. mechanism which at the same time suppresses expression in other cells of the domain. In the neurogenic mutants this mechanism fails, so that large clusters of S59 (or vg)-expressing cells are formed. The simplest model for this process would locate it in the mesoderm itself, but it would also be possible to formulate a mechanism whereby a normally restricted S59-activating signal passing from the ectoderm to the mesoderm would become more widespread, so leading to ectopic expression in the mutant embryos. On the basis of the results we report here, we cannot distinguish between these two alternatives, and it will be necessary to make an analysis of mosaic embryos in order to find the primary focus for the effect of neurogenic mutations on S59 and vg expression.

The close connection between the strength of neurogenic phenotypes and the degree to which late myogenesis is disrupted suggests that there may be a role for ectodermal derivatives in the normal development of muscle in the fly. Poulson (1945) described the failure of muscle development in embryos deficient for Notch. On the basis of results of embryo culture experiments, Cross and Sang (1978) suggested that there might be a direct requirement for Notch in muscle development, although they also pointed out that they could not exclude a secondary effect of the deranged nervous system in their cultures. Our results show that the Notch phenotype in late myogenesis fits into a consistent series, which includes the other neurogenic loci and matches the degree of muscle disruption to the strength of the neurogenic phenotype. This is in agreement with the observations of Lehmann et al. (1983) who described mesodermal defects associated with the “strong” neurogenic phenotype and suggested that these could be “explained by taking neural and epidermal defects into account”. In addition, in Df(l)svr;N^55e11 embryos where the neurogenic phenotype is partially rescued by a complete loss of the AS-C, there is a clear rescue of disrupted muscle differentiation. This result strongly suggests a direct link between hyperplasia of the nervous system and defects in late phases of myogenesis, including myoblast fusion. We speculate that the epidermis may be essential for muscle assembly and that the effects of the hyperplastic nervous system on myogenesis depend on the associated loss of epidermis. Where epidermis is present, even in strong mutants, some myoblast fusions may occur, and in the less extreme phenotypes, well-formed muscles develop in dorsal regions away from the enlarged nervous system. In contrast, the formation of muscles is always deranged ventrally, and it is at the ventral margins of the residual epidermis that we find clusters of mesodermal cells in a novel arrangement, which may represent some intermediate state of differentiation when fusion is prevented. We do not know whether any of these clusters correspond to expanded domains of S59 or vg expression, although it seems likely that included in this population are the ventral mesodermal cells that lose expression of S59 or vg during later embryogenesis in the stronger mutants. Possibly the epidermis, or some product of epidermal cells, is a required substratum for fusion to occur (other mutations that disrupt the epidermis also lead to a reduction of myoblast fusion; M. Frasch, unpublished observations). We could envision, for example, an initial step in which a cell adheres to and flattens on the epidermis before neighbouring cells can fuse with it. Clusters of cells seen in toluidine blue stained embryos might then represent mesodermal cells attempting to adhere to the margin of the residual epidermis, or cells with no competent partner with which to fuse. It is interesting in this context that Volk et al. (1990) report that, in serum free medium, cultured Drosophila myoblasts fail to adhere to and flatten on coverslips if laminin is not present. Adhesion and flattening are a prelude to the formation of multinucleate myotubes in their cultures. Prior to and during germ band retraction, laminin mRNAs are expressed in the mesoderm and the epidermis, but not at all in neurons (Montell and Goodman, 1989). Thus laminin might be substantially reduced in neurogenic mutant embryos, although we have as yet no direct observations to substantiate this idea. It is also interesting that, in Drosophila, syncytial muscle precursors always begin to form in close contact either with the CNS or the epidermis. It might be argued that, since the earliest signs of fusion occur not next to the epidermis, but in close proximity to the developing CNS (Bate, 1990), the epidermis itself cannot be essential for myogenesis. However, at the stage at which these first fusions occur, just before germ band retraction, the nervous system is still an integral part of the ventral embryonic epithelium, which, in a wild type embryo, includes both the precursors of the ventral epidermis and the CNS. Thus, the precursors of the epidermis are present both in the region of the forming CNS and outside it, and in a neurogenic mutant, where epidermal precursors would have been transformed into the neural pathway, they would be missing in the region of the expanded CNS. Although we suggest that there is a general phenotype for late myogenesis associated with loss of epidermis in neurogenic mutants, we do not exclude the possibility of a direct involvement for individual neurogenic genes in muscle differentiation which is masked by the derangements we describe here. In this respect, it may be significant that neu^IF65 embryos seem to have a worse muscle phenotype than we would predict from the extent of the cuticle they produce.

During the normal sequence of muscle development, the onset of expression of genes like S59 and vg in a limited set of mesodermal cells is followed by expression in additional cells as they are recruited to forming syncytia by fusion. In neurogenic mutant embryos, larger than normal clusters of mesodermal cells initiate the expression of S59 and vg in a way which suggests that the early limitation of expression seen in the wild type requires the functions encoded by the neurogenic genes. The additionally expressing cells in the mutant embryos are likely to include those that would have been recruited to expression by fusion in the wild type. However, it is not clear what the consequences of precocious gene expression in these cells may be for the process of myogenesis itself. One model of myogenesis would suggest that each muscle is seeded by the definition of a single founder cell which recruits neighbouring cells by fusion. If early expression of S59 or vg is in some way diagnostic for such founder cells, then in a neurogenic mutant embryo additional founder cells would be present in the expanded clusters of S59- or vg-expressing cells. Thus we might expect that if muscles formed in such embryos, they would be present as multiple copies, each muscle being founded by a separate cell. In weaker mutant embryos, such as mam, where vg-expressing cells are incorporated from expanded ventral clusters into ventral muscles, these muscles are indeed present as multiple copies. However, the issue is complicated by the fact that, more dorsally in such embryos, substantial muscles may be formed, apparently as single copies. As this example shows, although we find evidence of a dual requirement during muscle development for the functions encoded by the neurogenic genes, where and how these genes operate during myogenesis and the relationship between the two requirements we have demonstrated are, as yet, unresolved.

We are grateful to Alfonso Martinez Arias and Mary Baylies for many thought provoking discussions and much useful advice. This work is supported by a grant from the Wellcome Trust to M.B.