Cross-repressive interactions of identity genes are essential for proper specification of cardiac and muscular fates in Drosophila

The molecular pathways involved in the early events of heart and muscle formation are thought to be conserved from invertebrate to vertebrate species (for reviews, see Bodmer and Venkatesh, 1998; Jagla et al., 1998; Yun and Wold, 1996). Considerable progress in understanding how the heart and the muscle progenitors are specified has been made using Drosophila as a model system. However, the mechanisms that underlie commitment of cells to cardiac and muscular lineages and the requirements for generating the diversity of heart and muscle cells are not fully understood. In Drosophila embryos, heart tube derives from the dorsal mesoderm. It is located close to a subset of syncytial muscle fibres and composed of a small number of cell types, and is thus well suited for studying the genetic regulation of cell commitment.

To date, different intrinsic and extrinsic factors involved in progressive cell fate determination in the dorsal mesoderm of Drosophila have been identified (Corbin et al., 1991; Gisselbrecht et al., 1996; Shishido et al., 1997; Jagla et al., 1997; Carmena et al., 1998b; Park et al., 1998, Nose et al., 1998; Gajewski et al., 1999; Su et al., 1999; Gajewski et al., 2000; Crozatier and Vincent, 1999; Fosset et al., 2000; Ward and Skeath, 2000; Lee and Frasch, 2000, Halfon et al., 2000). The activity of the homeobox gene tinman (tin) (Azpiazu and Frasch, 1993; Bodmer, 1993) maintained by ectodermal Decapentaplegic (Dpp) signals delimits dorsal mesodermal cells (Frasch, 1995) and makes them competent to form heart, dorsal muscles and visceral mesoderm. tin expression in these cells is necessary but not sufficient to specify cardiac and muscular fates, as ubiquitously distributed tin products do not appear to cause any dorsalisation (Yin and Frasch, 1998). Segmentally repeated clusters of cells competent to become the heart and muscle progenitors form at the intersection of transversely striped wingless (wg) domain and the dorsal dpp/tin domain thus suggesting that tin expression is only instructive in conjunction with Wg and Dpp signalling pathways (for a review, see Frasch, 1999). In response to combined Dpp/Tin/Wg activities the mesodermal cells underlying the epidermal Dpp/Wg overlap switch on the proneural lethal of scute (l’sc) [l(1)sc – FlyBase] gene encoding a bHLH protein (Carmena et al., 1998b). Subsequent Receptor Tyrosine Kinase (RTK) action subdivides this territory into small clusters of equivalent cells from which the individual progenitors segregate (Carmena et al., 1998b; Halfon et al., 2000). This step involves a process of lateral inhibition governed by the neurogenic genes, which include Notch and its ligand Delta. Progenitor cells are thought to produce a pair of sibling cells, either the founders of dorsal muscles or precursors of two types of heart cells, the cardioblasts and the pericardial cells. The distinct fates of founder cells derived from the same progenitor are determined by asymmetric segregation of the Numb protein and by lineage-specific expression of identity genes (Ruiz Gomez and Bate, 1997; Jagla et al., 1998). Recent studies have provided important insights into the mechanisms underlying specification of a subset of pericardial cells expressing the identity gene even skipped (eve) (Carmena et al., 1998a; Park et al., 1998). These cells arise from the Numb-devoid founder cell generated after asymmetric division of dorsal progenitor P2 (Ruiz Gomez and Bate, 1997; Carmena et al., 1998a). The absence of Numb makes the prospective eve-pericardial cells competent to respond to Notch signalling, a requirement essential for their final fate (Park et al., 1998). In addition, the intrinsic program of pericardial cell specification involves the function of the GATA factor Pannier (Pnr) (Gajewski et al., 1999) and its antagonist U-shaped (Ush) (Fosset et al., 2000).

Studies on eve-positive muscle and heart lineages have indicated that the closely located dorsal progenitors P2 and P15 are specified sequentially (Carmena et al., 1998a). As demonstrated by Knirr et al. (Knirr et al., 1999), for a slouch (slou)-expressing progenitor (P5/25) and adjacent ladybird (lb)-positive SBM progenitor (P8), the sequential segregation also takes place in the lateral somatic mesoderm. In this case, an ectopic expression of lb in P5/25 and the resulting supernumerary SBM muscles observed in slou mutant embryos clearly indicate that the repression of lb by slou is a prerequisite for the specification of the progenitor P5/25 (Knirr et al., 1999). Similar negative interactions between lb and the muscle specific homeobox (msh) gene are essential for proper fate specification of muscles derived from the msh-expressing progenitors P21/22, P23/24 and the lb-positive SBM progenitor (Jagla et al., 1999). Thus, the switches of cell fates resulting from the repression of one identity gene by an other have been previously reported in Drosophila; however, the mutual repression between identity genes was not considered as an integral part of intrinsic program governing cell fate specification.

In this report, we focus on the interactions of three homeobox-containing identity genes expressed in adjacent populations of heart and muscle precursors: the lb genes that specify a subset of heart cells (Jagla et al., 1997); the msh gene required for the specification of two dorsal muscles (Nose et al., 1998); and the eve gene, which determines two pericardial cells and a dorsal muscle (Su et al., 1999). Using misexpression experiments and confocal microscopy, we demonstrate that msh, eve and lb gene products act as mutual repressors. Taking in consideration cross-repressive interactions of vertebrate identity genes during specification of neural fates (Briscoe et al., 2000; Pierani et al., 2001; Moran-Rivard et al., 2001) we anticipate that mutual repression between identity genes represents an important part of intrinsic mechanisms that lead to the diversification of cell types in both invertebrate and vertebrate development.

The following Drosophila strains were used: numb2 (Uemura et al., 1989); msh-lacZ line rH96, the msh null alleles; msh^Δ68 and msh^Δ89; and the UAS-msh-m25-m1 line (Nose et al., 1998), which exhibits high levels of ectopic msh expression when crossed with a GAL4 driver. For the lb gain-of-function experiments, we used double transgenic Hs-lbe/Hs-lbl flies (Jagla et al., 1997) or UAS-lbe-16-1 strain (Jagla et al., 1997). The lb mutants referred to as lbdef have been previously described (Jagla et al., 1997). The UAS-eve line and the thermosensitive evets mutants (Su et al., 1999) were kindly provided by R. Bodmer.

Ectopic mesodermal expression of msh, lbe and eve was induced using the Gal4-UAS system with the mesoderm-specific effector line 24B-Gal4 (Brand and Perrimon, 1993). The influence of ectopic lb expression on the specification of msh-positive muscle precursors was analysed in heat-shocked embryos derived from the cross of Hs-lbe/Hs-lbl males and rH96 virgin females. Hs-lbe and Hs-lbl transgenes were induced at 5 hours after egg laying (AEL) by a 20 minutes heat shock in 37°C water bath. To invalidate eve function during specification of muscle and heart progenitors, the evets embryos were shifted from 18°C to 29°C at 5 hours AEL.

Embryos were stained with the following primary antibodies: monoclonal anti-Lbe 1:1 (Jagla et al., 1997); rabbit anti-Kr, 1: 1000 (provided by G. Vorbrüggen); guinea pig anti-Kr, 1:500 (provided by D. Kosman); rabbit anti-β-galactosidase, 1:5000 (Cappel Laboratories); rabbit anti-Eve, 1:2000 and guinea pig anti-Eve, 1:1000 (provided by D. Kosman); rabbit anti-Odd (1:500) (provided by J. Skeath); rabbit anti-Tin, 1:800 (provided by M. Frasch); rabbit anti-dMef2, 1:1000 (provided by H. Nguyen); rabbit anti-β3-Tubulin, 1:200 (provided by R. Renkawitz-Pohl); rabbit anti-Myosin Heavy Chain, 1:2000 (provided by D. Kiehart); and monoclonal anti-Pc, 1:3 (mAb No.3, anti-pericardin antibody, provided by T. Volk). Labelled cells were detected using the ABC-Elite-peroxidase kit (Vector Laboratories) with diaminobenzidine as a substrate, or with secondary antibodies conjugated with Cy2, Cy3 or Cy5 (Jackson). In some cases, immunostaining was amplified using the Tyramide-Biotin (indirect) or Tyramide-Cy3 (direct) systems (NEN). To visualise heart and the muscle cells simultaneously, the whole-mount embryos were analysed using an OLYMPUS FV300 confocal microscope. The whole-mount in situ hybridisation experiments were performed with a digoxigenin-labelled msh antisense-RNA probe according to Tautz and Pfeifle (Tautz and Pfeifle, 1989). The full-length msh cDNA clone kindly provided by T. Isshiki was used as a template.

Previous studies (Carmena et al., 1998b) have indicated that dorsally located heart and muscle founders derive from a precluster of l’sc-expressing cells determined by an interplay of epidermal Wg and Dpp signalling pathways. At the beginning of germband retraction, in each abdominal hemisegment one can distinguish segmental heart units composed of six cardioblasts and ten pericardial cells (Ward and Skeath, 2000), as well as closely adjacent precursors of four dorsal muscles (Fig. 1A). Interestingly, the cardioblasts and the pericardial cells within the same hemisegment differ along the anteroposterior axis as pairs of cells that could be characterised by the unique combinatorial code of transcription factors (Fig. 1A). The most anterior pair of cardioblasts expresses tin, the adjacent pair is tin/lb-positive, while the most posterior do not express tin and can be detected in a seven up-lacZ (svp-lacZ) enhancer trap line (data not shown) (Ward and Skeath, 2000) as well as with the anti-Pericardin (anti-Pc) antibody (Fig. 1A,B). Among the pericardial precursors, as the most recent data indicate (Ward and Skeath, 2000), four pairs derive from the anterior compartment and one pair originates from the posterior compartment. Three out of four anterior pericardial pairs express tin (Fig. 1A,C). One tin-positive pair co-expresses lb and another one co-expresses eve (Fig. 1A,C). The tin-only expressing pericardial cells lie behind the cardioblasts and are often difficult to distinguish (Fig. 1C).

Two remaining tin-negative pairs of pericardial cells have recently been shown to express the Odd-skipped protein (Ward and Skeath, 2000) (Fig.1A). Two out of four Odd-pericardial cells are located in the posterior compartment and co-express Svp (Fig. 1A). Precursors of dorsal muscles DA1, DO1, DA2 and DO2 are located more lateral to the pericardial cells. Like the heart cells, they express a combination of transcription factors that allows monitoring of their identity during development (sketched in Fig. 1A). Among them, the most dorsally located DA1 and DO1 can be detected using anti-Eve and the msh-lacZ enhancer trap line rH96, respectively (Fig.1D) (Nose et al., 1998). In addition, in the late stage embryos, the cardioblasts and the myoblast nuclei could be visualised using anti-Mef2, while the pericardial cells could be visualised using anti-Pc antibodies (Fig. 1B). At stage 13/14 of embryogenesis, weak Mef2 staining is also detected in pericardial cells (data not shown, see also Fig. 7A). In this study, the influence of deregulated activity of heart and muscle identity genes was monitored using the cell fate markers described above.

Our previous data have shown that lb functions to specify the identity of two pericardial cells (lb-Pc) and when misexpressed, is able to change the fates of adjacent eve-expressing pericardial cells (eve-Pc) (Jagla et al., 1997) (Fig. 2C,D). The analysis of lb gain-of-function experiments with the use of a general heart marker tin has led us to the conclusion that ubiquitous mesodermal expression of lb results in heart hyperplasia (Jagla et al., 1997). To elucidate the origin of supernumerary heart cells in 24B-GAL4/UAS-lbe and Hs-lbe/Hs-lbl embryos (see Materials and Methods), we analysed specification of the most dorsal muscle founders and formation of corresponding syncytial fibres. The DA1 founder was monitored with anti-Kr (Fig. 2A,B) and anti-Eve (Fig. 2C,D) while the specification of DO1 founder was revealed with anti-Kr (Fig. 2A,B) or with anti-lacZ in rH96 embryos (Fig. 2E,F). The resulting muscle fibres were detected using anti-Myosin Heavy Chain (MHC) (Fig. 2I). We found that the pattern of all founder-specific markers was significantly affected (Fig. 2B,D,F). The uniform mesodermal expression of lbe led to the 60% loss of DA1 founders and 40% loss of DO1 founders (Table 1; Fig. 2B). Affected Kr expression (Fig. 2B) was not linked to selective Kr repression, as the reduced number of DA1 and DO1 muscle precursors was also detected by anti-Eve (Fig. 2D) and anti-Msh (Fig. 2F) antibodies.

This specific loss of identity of most dorsally located muscle founders most probably leads to morphological defects observed in dorsal somatic musculature of late stage embryos overexpressing lb genes (Fig. 2I). As revealed by the anti-MHC antibody (Fig.2I), the gaps in somatic musculature are associated with the significantly enlarged heart area, suggesting that the supernumerary heart cells are generated by a shift in fate of muscle founders, rather than by aberrant proliferation of cardiac precursors. Interestingly, the majority of cardiac cells induced by ectopic lb expression corresponds to the tin-positive cardioblasts (Fig. 2H). On average, mesodermal ectopic expression of lb results in the generation of 7.5 tin-labelled cardioblasts per hemisegment instead of 4 (Table 1). In addition, the accumulation of supernumerary tin-positive pericardial cells is seen in about 20% of hemisegments (Fig. 2H′). These pericardial cells are distinct from eve-Pc, as ectopic lb suppresses eve-positive fates (Fig. 2D). Altogether, these observations suggest that lb, when ectopically expressed, is able to repress msh and eve, leading to the transformation of dorsally located muscle founders into the tin-positive heart cells.

One possible explanation of the exclusive non-overlapping expression of Lb, Msh and Eve proteins in small subsets of adjacent dorsally located mesodermal cells (see Fig. 1D) is that these genes are able to repress each other. To test this possibility, we induced ubiquitous mesodermal expression of msh. Previous work (Nose et al., 1998) has indicated that msh is required for the specification of two dorsal muscles and is not expressed in the heart. Indeed, as shown in Fig. 1D, msh is not expressed in the heart field. However, our data clearly demonstrate the ectopic msh leads to alterations in both dorsal muscle and heart cell populations (Fig. 3). The number of heart cells is dramatically reduced (Fig. 3). The majority of pericardial cells and about 70% of tin-expressing cardioblasts are lacking (Table 1; compare Fig. 3A with 3B). Among suppressed cardiac cell fates are those specified by the lb and eve gene activities (Fig. 3C-F; see also Table 1).

In embryos that overexpress msh, we observe less than 10% of lb-positive and about 30% of eve-positive heart cells when compared with the wild type (Table 1). The inhibitory influence of ectopic msh on a large spectrum of cardiac fates is particularly obvious once the pericardial cells and the cardioblasts are simultaneously visualised (Fig. 3G-J). Surprisingly, msh is able to affect specification of both the anterior (β3-Tub-labelled) and the posterior (Pc-labelled) cardioblasts (compare Fig. 3G with 3H). Moreover, the major part of Pc-positive pericardial cells is missing (Fig. 3H,J). As shown by double Kr/Eve staining (Fig. 3E,F), msh is also able to change the fates of dorsal muscles. Specifically, in UAS-msh embryos, the majority of eve-positive DA1 muscle precursors are substituted by supernumerary DO1 muscles. Thus, the loss of cardiac and particular muscle cells is accompanied by overproduction of other dorsal muscle lineages (Fig. 3H,J), suggesting that msh may change the identity of dorsal mesodermal cells and recruit them to the alternative myogenic pathways.

To test whether msh determines cell fates by repressing a set of identity genes, we analysed specification of heart and dorsal muscle cells in msh null mutant embryos. Loss of msh function leads to the generation of supernumerary tin- (compare Fig. 4A with 4B) and mef-positive (compare Fig. 4C with 4D) heart cells. Interestingly, the msh-induced heart hyperplasia seems to result from the generation of two distinct types of additional cardiac cells. The first category corresponds to the supernumerary cardioblasts (arrows in Fig. 4B,D), while the second type of cells (arrowheads in Fig. 4B,D) is located more laterally than the regular cardioblasts. Based on identical position of supernumerary odd-pericardial cells observed in msh mutant embryos (arrowheads in Fig. 4F), we propose that the loss of msh function leads to the specification of enlarged number of both the cardioblasts and the pericardial cells. The increased number of heart cells in msh mutants may indicate that in the wild-type embryos, msh represses the activity of genes specifying cardiac fates. In order to check this possibility, we traced the development of DO1 and DO2 muscles in embryos lacking msh function. As indicated in Fig. 4G (arrows), at least some of the presumptive msh-positive founders (labelled by the lacZ in null msh^Δ89 strain) (Nose et al., 1998) switch on lb expression. As lb is never co-expressed with msh in wild-type embryos, we conclude that msh functions to repress lb activity.

In msh mutants, lb is de-repressed in the presumptive msh-positive muscle founder cells, which leads to their recruitment into cardiogenic pathway. However, only some msh-devoid cells are able to adopt cardiac fates (see irregular heart hyperplasia in Fig. 4B,D), indicating that msh– cells may be alternatively recruited to other differentiation pathways. Indeed, in msh– background, some of the msh-positive myoblasts (most probably fusion-competent cells that form DO1 muscle) are recruited to built enlarged eve-labelled DA1 muscle precursors (Fig. 4H). These myoblasts, once incorporated into eve-positive DA1 syncytium, lose their lacZ expression progressively (Fig. 4H), indicating that eve has a capacity to repress msh. Most likely, for this reason, in msh^Δ89 mutants we detect only few lacZ-expressing cells that contribute to DA1 muscles even if these muscle fibres are significantly enlarged (Fig. 4H). This observation is consistent with statistically rare incidence of lb and lacZ co-expression in msh^Δ89 embryos and with comparatively low lacZ expression in the presumptive msh-positive cells recruited to the cardiogenic pathway (arrow in Fig. 4H). The detection of only few cells co-expressing lacZ and lb, or lacZ and eve in germband retracted msh^Δ89 embryos (Fig. 4G,H), associated with relatively important perdurance of β-gal (Halfon et al., 2000), suggests also that the repressive action of lb and eve on msh appears early in development (most probably at the progenitor stage).

Adjacent heart and muscle cells achieve their rigidly determined states as they await metamorphic cues provided by secreted EGF and FGF signals (Buff et al., 1998), and by localised activation of transcription factors, such as Lb and Msh. Our data indicate that lb prevents presumptive cardiac cells from assuming a muscle-specific identity by repressing the muscle fate inducing gene msh. Similarly, msh is able to repress lb and to promote specification of dorsal muscle instead of cardiac fates. In contrast to msh and lb, eve is expressed in both cardiac (pericardial cells) and muscular (DA1) precursors, and is involved in cell fate specification processes in these two territories (Su et al., 1999). The requirement for eve and its mode of action can be demonstrated by misexpressing the gene outside of its endogenous expression pattern. We used 24B-GAL4 line to drive eve expression ubiquitously in differentiating mesoderm and have found that a significant number of cells from the dorsal mesoderm assume altered identities. In particular, ectopic eve represses lb and leads to the loss of lb-expressing heart cells (Fig. 5A,B; Table 1). Surprisingly, not only the lb-positive pericardial cells but also the cardioblasts are missing. This effect appears similar to the loss of cardiac cells in UAS-msh embryos (Fig. 3), except that in embryos that misexpress eve, a subset of tin-labelled cardiac cells is still specified or even overproduced (see arrowheads in Fig. 5B).

As revealed by Mef2 staining (Fig. 5D), only some of the remaining heart cells correspond to the cardioblasts, indicating that eve promotes preferentially a particular type of pericardial fate (most probably eve-Pc). This assertion is also supported by the negative influence of ectopic eve on another subset of pericardial cells expressing odd (Fig. 5E,F). Concomitant with the loss of the majority of cardiac cell types, ubiquitous mesodermal expression of eve leads to the formation of supernumerary dorsal muscles (Fig. 5D, Table 1). As shown by Mef2 (Fig. 5D) and double Kr/Odd staining (Fig. 5F), the supernumerary dorsal muscles appear preferentially in segments with the reduced number of heart cells, thus suggesting that eve is also able to convert cardiac into muscular fates. The additional Kr-positive muscles (Fig. 5F) cannot however correspond to DO1 fibres, as in situ hybridisation experiments with msh probe (Fig. 5G,H) clearly show that the ectopic eve represses msh, leading to the loss of DO1 and DO2 identities. Taken together, these observations demonstrate different regulative behaviour of cells that overexpress eve and reflect a binary decision they may take to achieve their identity.

In order to provide direct demonstration that Eve and Lb behave as repressors, we have analysed the influence of loss of function of both genes on cell fate specification in the dorsal mesoderm of the Drosophila embryos. The previously described thermosensitive evets mutants (Su et al., 1999) and tin rescue construct-carrying lbdef mutants (Jagla et al., 1997) were used in this study. In evets embryos, non-functional Eve protein is still synthesised, which makes it possible to follow the eve-positive cells. As shown in Fig. 6A (compare with wild type shown in Fig. 7C) loss of eve function at 5 hours of development (see Materials and Methods) leads to de-repression of lb in the eve-positive mesodermal cells. Consequently, activated lb represses eve so that, in the germband retracted embryos, expression of mutated Eve protein becomes progressively lower. We observe supernumerary lb-positive cells (arrowheads in Fig. 6A) in segments in which eve is completely repressed. Consistent with the model of mutual repression between lb and eve, the loss of lb function (Fig. 6B) leads to the specification of supernumerary eve-expressing cells. As demonstrated by double Kr/Eve staining (arrowheads in Fig. 6B), the additional eve-positive cells (compare with the wild type; Fig. 3E) generally adopt a pericardial identity and only some of them contribute to enlarged dorsal muscles (bracket in Fig. 6B). The repressive action of eve and lb during specification of cardiac and muscular fates is also confirmed by the analysis of msh expression (Fig. 6C,D). Compared with the wild type (Fig. 5G), in eve– (Fig. 6C) and lb– embryos (Fig. 6D), the msh transcripts are detected in much broader area, thus indicating ectopic activation of the gene.

Cell fate respecification observed after ectopic expression of lb, msh and eve is restricted to a defined subset of cardiac and/or muscular cell lineages. To learn more about the mechanisms of identity gene action, we asked whether the modifications of cell fates induced by misexpression of lb concerned cells other than siblings of founders in which the lb gene is normally expressed. As previously shown (Jagla et al., 1997) (Fig. 1), the lb genes are expressed in two cardioblasts and two pericardial cells per hemisegment, and when ectopically expressed they lead to the overproduction of tin-CB concomitantly with the loss of eve-PC and dorsal muscles (see Fig. 2). One way to check whether the affected cells arise from the same progenitors as the lb-expressing cells is to compare the influence of numb on their specification. It has been previously shown (Carmena et al., 1998a) that eve-PC and eve-DA1 founders derive from the asymmetrically dividing progenitors P2 and P15. According to the recent lineage tracing (Ward and Skeath, 2000), individual muscle and heart cells arise from progenitors that undergo either the asymmetric or symmetric divisions. We demonstrate that the number of the lb-positive cardioblasts and the lb-positive pericardial cells (arrowheads in Fig. 7A,B) is not affected in numb mutant embryos (Fig. 7) suggesting that they arise from the progenitors dividing symmetrically. Thus, the lb-positive and eve-positive dorsal mesodermal cells cannot be generated from the common progenitor cells, revealing that the ectopic lb influences the fates of cells that are not lineage linked. In addition, with the use of tyramide-enhancement system (NEN) employed to strengthen the lb staining ,we show that lb is also expressed at a weaker level in a third cardioblast (arrow in Fig. 7A) immediately posterior to the highly expressing pair. In contrast to the anterior lb-cardioblasts, this cardioblast is absent in numb mutants (see Fig. 7B), indicating that it originates from the asymmetrically dividing progenitor. The assumption that the anterior cardioblasts are generated by symmetric division, while the posterior pair by the asymmetric division is fully confirmed by double Tin/Pc staining (Fig. 7E,F). Taking into consideration these data and the so far described mesodermal phenotypes of numb mutants (Ruiz Gomez and Bate, 1997; Carmena et al., 1998a; Ward and Skeath, 2000), we propose a scheme (Fig. 7G) that summarises the influence of numb on cell fates in dorsal mesoderm. Together, our results support the lineage tracing described by Ward and Skeath (Ward and Skeath, 2000) and reveal that when lb is ectopically expressed, it can alter the identity of cells arising from the adjacent lb-negative progenitors.

How an initially pluripotent cell acquires its final fate remains one of the most intriguing unresolved issues. The analyses of genes that govern heart and muscle development in Drosophila have provided important insights into cell fate specification processes (for reviews, see Baylies et al., 1998; Wilson, 1999; Frasch, 1999; Paululat et al., 1999). It has become clear that the acquisition of cell identity requires a coordinated integration of extrinsic signals such as Wg, Hh and Dpp (Carmena et al., 1998b; Xu et al., 1998; Lee and Frasch, 2000, Halfon et al., 2000), Notch-dependent cell-cell signalling (Park et al., 1998; Crozatier and Vincent, 1999), and the intrinsic information provided by the localised expression of transcription factors encoded by the so-called identity genes (Ruiz Gomez and Bate, 1997; Jagla et al., 1998; Nose et al., 1998; Knirr et al., 1999; Crozatier and Vincent, 1999; Capovilla et al., 2001). According to the commonly accepted hypothesis (Bate, 1990; Dohrmann et al., 1990), the individual muscles and heart cells are specified at the founder cell stage. The most recent data (Ruiz Gomez and Bate, 1997; Jagla et al., 1997; Su et al., 1999: Knirr et al., 1999) indicate that the diversification of mesodermal cell types is governed by combinatorial code of identity gene expression. We have focused on intrinsic program that leads to the specification of a subset of cardiac and muscular cells located in the dorsal mesoderm of Drosophila embryos. Our data reveal cell fate plasticity in this region and suggest the instructive role of identity gene expression not only at the level of the founder cell, but also at the progenitor cell level. Gain- and loss-of-function phenotypes of lb, msh and eve clearly indicate that the cross-repressive interactions of these genes are essential for the proper cell fate acquisition.

We have previously observed the ability of ectopically expressed lb to inhibit eve in founder of the DA1 muscle (Jagla et al., 1997). This effect may be due to either a specific inhibition of eve by lb or a more general regulatory mechanism of fate specification. Data presented here favour the latter possibility, showing that the gain of lb function affects expression of several identity genes and consequently influences fates of cells in which these genes are expressed. Specifically, embryos that ectopically express lb have an increased number of tin-positive heart cells with a concomitant reduction of dorsal muscles. To demonstrate that the supernumerary cardiac cells result from cell fate switches, rather than from additional proliferation, we used msh^Δ89 mutants (Nose et al., 1998) displaying heart hyperplasia similar to that observed in embryos overexpressing lb. In this particular msh mutant, the presumptive msh-positive muscle cells monitored by lacZ start to express cardiac markers. This suggests that switches from muscular to cardiac fates contribute to heart hyperplasia induced by deregulation of identity genes. Interestingly, the ectopic expression of lb and msh leads to reciprocal phenotypes, and indicates that the identity genes specifically expressed in the heart promote dorsal mesodermal cells to enter the cardiogenic pathway, while the muscle identity genes promote the myogenic pathway. However, more detailed analysis shows that ectopic lb promotes only specific cardiac fates and ectopic msh only specific muscle identities, thus indicating that the identity genes instruct dorsal mesodermal cells to adopt the specific cardiac or muscular fates, rather than make a choice between cardiac and muscular development. This property is particularly well illustrated by the phenotypes generated by the ectopic eve, which is involved in the specification of a subset of heart and dorsal muscle cells and when ectopically expressed promotes specification of supernumerary cells of both types. Moreover, deregulated heart and dorsal muscle identity genes affect preferentially fates of mesodermal cells located in dorsal but not in ventral regions (T. J. and K. J., unpublished), thus suggesting that the identity gene action is instructive only in a permissive context. This observation is in complete agreement with the model of competence domain proposed by Carmena et al. (Carmena et al., 1998b). According to this concept, the high level of Wg and Dpp signals present in the anterodorsal region (under the intersection of Wg and Dpp epidermal domains) provides a major cue that direct mesodermal cells into cardiac or dorsal muscle development. In relation to this model, our data design a new regulatory mechanism that provides a paradigm of how the intrinsic transcription factors and extrinsic signalling molecules converge to specify cell fates.

The 24B-GAL4 line (Brand and Perrimon, 1993) used in our misexpression experiments allows the pan-mesodermal expression of identity genes to be driven from the progenitor stage onwards. This implies that the cell fate modifications we see may result either from identity gene action at the progenitor or at the subsequent founder cell stage. Endogenous identity gene expression is often detected in progenitor cells (Carmena et al., 1998a; Jagla et al., 1998; Knirr et al., 1999), even if cell fate specification is assumed after progenitor division at the founder stage. This raises a possibility that this early identity gene expression provides instructive information to the progenitor cells and contributes to progressive cell fate specification. An argument in favour of such a mechanism is our observation that the cell fate modifications induced by gain of function of each of the tested identity genes are not lineage restricted. For example in the case of lb overexpression, the affected cell lineages arise from at least three distinct progenitors (P2, P15 and P_DO1) (Carmena et al., 1998a; Nose et al., 1998). As indicated by our analysis of numb mutants, none of these asymmetrically dividing progenitors gives rise to lb-positive heart cells, which arise from progenitors that divide symmetrically.

Our findings suggest cross-repressive interactions that occur between transcription factors that specify adjacent and non-overlapping populations of muscle and heart cells. Most likely, in normal development, these interactions have a functional relevance once the progenitor cells segregate, and then continue to play an important role in the next step of cell fate diversification, namely in founder cells. The gain- and loss-of-function experiments we present indicate that the identity genes may function as repressors starting from the progenitor stage onwards. However, the earliest activation of inappropriate identity gene as a result of the loss of function of repressor (in msh^Δ89 embryos) was documented in founder cells.

In the model we propose (Fig. 8), cross-repressive interactions allow the refinement of the potentially imprecise pattern of identity gene expression induced by the interplay of Wg and Dpp signalling pathways. As elegantly demonstrated by Carmena et al. (Carmena et al., 1998b) and Halfon et al. (Halfon et al., 2000), Wg and Dpp create a permissive context for the development of cardiac and dorsal muscle precursors. In such a context, the transcription factors that specify these two types of cells (e.g. lb, eve and msh) are expected to be activated in all dorsal mesodermal cells. The local restriction of identity gene expression is, however, provided by a combinatorial signalling code mediated by two receptor tyrosine kinases, the Drosophila epidermal growth factor receptor and the Heartless (Htl) fibroblast growth factor receptor (Carmena et al., 1998b). In the model proposed by Carmena et al., transient localised activity of these two mesodermal signalling pathways is thought to subdivide the large competence domain into small clusters of equivalent cells from which individual progenitors segregate. Depending on the combination of RTKs activities, the individual identity genes are activated only in a defined equivalence group and in the resulting progenitor (Carmena et al., 1998b). Our study defines an additional step to the aforementioned model. We propose that the major role of identity genes is to maintain their restricted expression in progenitors and subsequently in founder cells by repressing other identity genes competent to respond positively to Wg and Dpp signals. These cross-repressive interactions are likely to ensure constant localised identity gene expression over time, thus providing a crucial element in establishing cell identity.

It is noteworthy that the principles of the mutual repression model outlined here resemble those governing neural identity in the ventral neural tube of chicken and mouse embryos (Briscoe et al., 2000; Pierani et al., 2001; Moran-Rivard et al., 2001). In both systems, negative regulation between the transcription factors establishes sharp boundaries of the identity gene expression and controls later aspects of cell pattern. Consistent with this interspecies homology, the cross-repression mechanism may be universally used in different biological systems as a crucial component of intrinsic control of diversification.

The cross-repression mechanism also provides an effective way for establishing discrete domains of identity gene expression; however, several questions remain to be addressed:

(1) Does the repressive influence observed between identity genes result from direct interactions

(2) How do the different RTK signalling pathways subdividing the competence domain communicate with the identity gene action

(3) What is the role of asymmetric versus symmetric division in determining cell fates in the dorsal mesoderm.

One possible way of addressing these issues, and an important challenge for future studies, will be to determine the integration of regulatory cues at molecular level by functional analysis of enhancers that drive restricted expression of identity genes.

We thank A. Nose, R. Bodmer, A. Brand, M. Frasch, D. Kosman, R. Renkawitz-Pohl, D. Kiehart, H. Nguyen, J. Skeath and the Bloomington stock centre for sending fly stocks and antibodies. We are grateful to S. Tcheressiz and M. Taylor for critical reading of the manuscript. This work was supported by grants from the INSERM, the Association pour la Recherche sur le Cancer (ARC), the Association Française contre les Myopathies (AFM) and the Human Frontier Science Program (HFSP). T. J. and Y. B. were founded by the HFSP fellowships.