collier transcription in a single Drosophila muscle lineage: the combinatorial control of muscle identity Free

Drosophila Collier (Col; also known as Knot) belongs to the COE(Col/Olf1/EBF) family of transcription factors, which contains a single member in metazoans, except for vertebrates, in which four genes have been identified(Dubois and Vincent, 2001; Liberg et al., 2002; Pang et al., 2004). col was initially characterised for its expression and function in a specific region of the embryonic head that corresponds to both a mitotic domain (MD2) and a gnathal parasegment (PS0)(Crozatier et al., 1996). col is also expressed in, and required for the formation of, a single somatic muscle, the embryonic Dorsal/Acute 3 (DA3) muscle(Crozatier and Vincent, 1999),thereby providing a unique entry site for studying the transcriptional control of muscle identity.

The embryonic musculature of Drosophila melanogaster is highly stereotyped, with a standard arrangement of around 30 somatic muscles in each trunk hemisegment. Each muscle fibre is an individual syncitium that can be distinguished by its position, shape, epidermal attachment sites and innervation (Bate, 1993; Baylies et al., 1998). Muscle fibres are seeded by founder cells (FCs), which are themselves generated from progenitor cells singled out from promuscular clusters by Notch-mediated lateral inhibition (Carmena et al.,1995; Ruiz Gomez and Bate,1997). FCs undergo multiple rounds of fusion with fusion competent myoblasts (FCMs) to form a myofibre. The current view is that `muscle identity' transcription factors (TFs) endow FCs with the capacity to execute the fusion and differentiation programme specific to each muscle fibre(Baylies and Michelson, 2001; Frasch and Leptin, 2000). The`identity TF code', at least in part, reflects the initial position of the promuscular cluster and derived progenitor cell. Pioneering work on the control of expression of the homeodomain transcription factor Even-Skipped(Eve) in dorsal muscle progenitors showed that it involved the combinatorial activity of TFs functioning downstream of Wingless (Wg), Decapentaplegic (Dpp)and receptor tyrosine kinase (RTK) signalling, [dTCF (Pan - FlyBase), Mothers against Dpp (Mad) and Pointed (Pnt), respectively]. Integration of this positional information and tissue-specific (mesodermal) information at the level of the eve promoter was responsible for activating Eve-expression in promuscular clusters (equivalence groups) from which Eve progenitors were selected by Notch (N) signalling(Carmena et al., 2002; Carmena et al., 1998; Halfon et al., 2000; Halfon et al., 2002). Large-scale analyses of gene expression in conditions of perturbation of components of Eve regulation suggested that related transcriptional codes could be responsible for different patterns of progenitor gene expression(Estrada et al., 2006; Philippakis et al., 2006; Sandmann et al., 2006). The eve enhancer reproducing Eve expression in muscle progenitors was not active, however, in recruited FCM nuclei(Halfon et al., 2000),indicating that different cis-regulatory elements (and TFs) could be required for specifying promuscular clusters and maintaining a TF identity code.

Here we used Col expression as both a determinant and read-out of DA3 muscle identity to ask how positional information that defines promuscular clusters is relayed into the FC identity and extended to fused FCM nuclei. We first identified the cis-regulatory regions controlling coltranscription at several steps during formation of the DA3 muscle and defined a DA3-muscle-specific cis-regulatory module (CRM). Detailed analysis of this CRM revealed the existence of three separate steps: Col activation in promuscular clusters, upregulation in the selected progenitor and DA3 FC and activation in the nuclei of FCM incorporated in the growing DA3 myofibre during the muscle fusion process. Comparison of the DA3 muscle CRM between several Drosophila species identified a set of conserved sequence motifs with functional significance supported by the expression patterns of reporter genes containing the D. virilis (D. vir) DNA. Conserved binding sites for the mesodermal TFs Twist (Twi), Nautilus (Nau, the Drosophila orthologue of MyoD) and Mef2(Andres et al., 1995; Huang et al., 1996; Ip et al., 1992; Kophengnavong et al., 2000)and a putative Col-binding site necessary for positive autoregulation were present in different subdomains of the DA3 muscle CRM, correlating with the separate phases of col regulation. We show that colauto-regulation is crucial for a reiterative, two-step activation of col transcription in each `naïve' FCM incorporated into the DA3 muscle. Nau, which was previously reported to be required for DA3 muscle formation (Keller et al.,1998), is also required for col transcription in the DA3 muscle, beyond the FC stage. Pan-FC expression of either Col, Nau or both proteins resulted in ectopic col transcription in different sets of muscles. Together, our results show that separate sets of cis-regulatory elements ensure col activation in the DA3/D05 promuscular cluster,progenitor and DA3 myofibre. Nau and Col act together in ensuring that all nuclei within the DA3 myofibre activate Col and express the same differentiation programme, thereby directly supporting the concept of combinatorial control of muscle identity.

The following strains were used: w¹¹¹⁸ as a wild-type(wt) reference and for P element transformation using standard procedures(Rubin and Spradling, 1982);rp298-Gal4 (Menon and Chia,2001); col¹(Crozatier et al., 1999) and nau¹⁸⁸ (Balagopalan et al., 2001) EMS-induced loss-of-function alleles; vg^83b27-R, a γ-ray induced amorphic allele;UAS-col (Vervoort et al.,1999); hs-col(Crozatier and Vincent, 1999);UAS-nau (Keller et al.,1997); UAS-lacZ (Bloomington Stock Center, Indiana, USA). UAS-mcd8::GFP (Grueber et al.,2003).

The P5cl construct was described in Crozatier and Vincent(Crozatier and Vincent, 1999). Other Pcl constructs were generated by cloning different fragments of col upstream DNA (for the restriction sites used, see Fig. S2 in the supplementary material) into pCaSpeRβ-gal or pPTGal4(Sharma et al., 2002). Mutagenesis of the putative Nau- and Col-binding sites in P2.6cl was done by PCR. The D. vir constructs were generated by restriction digestion of genomic DNA isolated from a λ phage library (J. Tamkun,University of California, Santa Cruz, CA).

Embryos were fixed and processed for antibody staining and/or in situ hybridisation as described (Crozatier et al., 1996). The nau intronic probe covers all three nau introns and the two corresponding exons. The following primary antibodies were used: rabbit anti-Col (1/400); mouse anti-Col (1/100); rabbit anti-MHC (1/500; D. Kiehart, Duke University, Durham, NC); mouse anti-β-gal (1/1000, Promega); rabbit anti-GFP (1/1000 Torrey Pines Biolabs); Secondary antibodies were Alexa Fluor 488 and Alexa Fluor 647 conjugated goat anti-rabbit, Alexa Fluor 647 conjugated goat anti-mouse(Molecular Probes 1/300); Rhodamin RedX conjugated donkey anti-mouse (Jackson Laboratory 1/300); biotinylated goat anti-mouse (Vector Laboratory, 1/1000). For double fluorescent in situ hybridisation/immunostaining, we used biotinylated col and digoxygenin-labelled nau intronic probes and the ABC kit from Vector Laboratory, followed by fluorescent tyramide staining (Alexa fluor 555 or 488 conjugated tyramide from Molecular Probes) and Fast Red. Primary antibodies against Col, GFP or MHC were used at five times the usual concentration. Monoclonal Col antibodies were generated in collaboration with Jeannine Boyes and Georges Delsol, U 563 INSERM,Toulouse Purpan.

Pairwise sequence alignments of col upstream sequences from various Drosophila species(http://flybase.bio.indiana.edu/static_pages/news/articles/2007_03/genomes_papers3.html)were done using NCBI-BLAST (bl2seq), Genome Browser (UC Santa Cruz) and Evoprinter (NINDS, NIH, Bethesda) and manually edited following eye-inspection. Search for individual binding sites for transcription factors made use of Genomatix Matinspector, Possum(http://zlab.bu.edu/~mfrith/possum/),cis-analyst(http://rana.lbl.gov/cis-analyst/cgi/viewer.php)and FlyEnhancer(http://genomeenhancer.org/fly;M. Markstein) and manual inspection based on the literature. Access to the Mef2 and Twi in vivo binding sites(Sandmann et al., 2007; Sandmann et al., 2006) was via the E. Furlong's lab site(http://furlonglab.embl.de/data/).

col belongs to the class of Drosophila regulatory genes with numerous introns, large amounts of flanking sequence and multiple expression sites (Crozatier and Vincent,1999; Nelson et al.,2004; Philippakis et al.,2006). During embryogenesis, col is expressed in the MD2/PS0 head region, the somatic DA3 muscle, precursor cells of the lymph gland, a small set of multidendritic (md) neurons of the peripheral nervous system and specific neurons of the central nervous system (CNS)(Baumgardt et al., 2007; Crozatier et al., 2004; Crozatier et al., 1999; Crozatier and Vincent, 1999; Orgogozo and Schweisguth,2004). We previously generated a lacZ reporter transgene(P{5col::lacZ}, abbreviated P5cl, Fig. 1A) containing 5 kb of col upstream DNA, which faithfully reproduced coltranscription both in the MD2/PS0 and the DA3 muscle, starting at the progenitor stage and not in promuscular cluster(s)(Crozatier and Vincent, 1999). To identify the missing cis-regulatory information, we tested a longer construct containing the entire 9 kb region separating col from CG10200, the next predicted upstream gene(http://flybase.bio.indiana.edu/; P9cl, Fig. 1A). In addition to the head and DA3 muscle, P9cl expression reproduced col expression in md neurons and a subset of neurons in the CNS. A DNA fragment located further upstream, between CG10200 and the next predicted gene CG10202, was independently shown to drive col expression in the anteroposterior organiser of the wing imaginal disc(Hersh and Carroll, 2005). However, neither this construct nor P9cl reproduced Col expression in promuscular clusters (Fig. 1D). The col transcription unit is immediately flanked at its 3′ end by another gene, BEAF32 (Fig. 1A), making rather unlikely the presence of cis-regulatory elements within this region. However, it contains ten different introns, of total length around 30 kb, the cis-regulatory content of which remains to be assessed (see Discussion).

To delineate more precisely the CRM driving col expression in the DA3 muscle, we tested a series of constructs containing 2.6, 2.3, 1.6 and 0.9 kb of DNA upstream of the col transcription start site, respectively(Fig. 1A). P2.6clretained the information necessary for col expression in MD2/PS0 and the DA3 progenitor and muscle (Fig. 1C), although we noted that P2.6cl expression in muscle progenitors was less robust than P9cl. P2.3cl was also activated in MD2/PS0 at stage 6 and the DA3 muscle. However, unlike P9cl or P2.6cl, P2.3cl was not activated in the DA3/DO5 progenitor but only at the FC stage (Fig. 1C;ectopic lacZ expression was observed in clusters of neuroectodermal cells at embryonic stage 11). This difference indicated that cis-regulatory elements required for col expression in the DA3/DO5 progenitor reside between positions -2.6 and -2.3 and act separately from those required for expression in the DA3 FC and muscle. P1.6cl was only active in MD2/PS0, whereas no expression at all could be detected with P0.9cl(data not shown). Together, expression data from this series of reporter constructs allowed the mapping of the CRM required for col-specific expression in the DA3/DO5 muscle progenitor and DA3 FC/myofibre to a DNA fragment between positions -2.6 and -1.6 upstream of the coltranscription start (Fig. 1E).

We took advantage of the recently available genome sequences of several Drosophila species to search for conserved motifs in the colupstream DNA, as it has often proven to be effective to identify functionally important cis-regulatory elements(Wasserman et al., 2000; Yuh et al., 2002). Among these species, D. virilis (D. vir) is the most distant from D. melanogaster (D. mel)(Tamura et al., 2004). We first verified that Col expression in D. vir was similar to that in D. mel embryos (Fig. 2A and see Fig. S1 in the supplementary material) and could infer from this that the regulatory information controlling coltranscription in the DA3 muscle lineage has been conserved. Sequence comparison of 9 kb of the col upstream region between D. mel, D. vir and four other Drosophila species, D. yakuba, D. ananassae, D. pseudoobscura and D. mojavensis revealed numerous stretches of high sequence conservation, of sizes up to 100 bp (see Fig. S1 in the supplementary material). Ten conserved motifs of size >20 bp, numbered 1 to 10 from 5′ to 3′, were found in the same order and at the same relative position between position -2.6 and the start of transcription in all six Drosophila species (Fig. 2B and see Fig. S2 in the supplementary material). To test the relevance of this conservation, we generated lacZ reporter constructs containing either D. vir or D. mel DNA(Fig. 2B). P.3.4cl^vir corresponds to D. mel P2.6cl, whereas P3.4-1.3cl^vir and P2.6-0.9cl are truncated versions covering motifs 1 to 10. All four reporter genes showed expression in the DA3 muscle, starting at the progenitor stage, confirming the evolutionary conservation of a DA3-muscle-specific CRM(Fig. 2A). A Gal4 driver line containing only the -2.6 to -1.6 region (P2.6-1.6cG), harbouring only motifs 1 to 7 (Fig. 2B), was also specifically expressed in the DA3 muscle(Fig. 2A). This confirmed that the DA3 muscle CRM is located between positions -2.6 and - 1.6. We noticed,however, that expression of P2.6-1.6cG was weaker and more sporadic than P2.6-0.9cl, suggesting the existence of cis-regulatory element(s) between positions -1.6 and -0.9 contributing to robust DA3 muscle expression. We then searched within the conserved motifs 1 to 10 for consensus binding sites of known TFs that could account for col activation in the DA3 muscle. This identified a binding site for the mesodermal basic helix-loop-helix (bHLH) protein Twi (Ip et al., 1992; Kophengnavong et al., 2000) (within motif 2 Fig. 2B), correlating well with the position of the muscle progenitor cis-element (Fig. 1E) and a potential EBF/Col-binding site (Travis et al., 1993) within motif 7. Further visual inspection of the sequence alignments identified other conserved TF-binding sites, including one Mef2-binding site (Andres et al.,1995) within the -1.6 to -0.9 fragment and one consensus binding site for Nau (Huang et al.,1996; Kophengnavong et al.,2000). On the one hand, the position of the Mef2 site correlated well with the requirement of the -1.6 to -0.9 fragment for robust DA3 muscle expression (Fig. 2A). On the other hand, the presence of a Nau-binding site was particularly intriguing as Nau is required for DA3 muscle formation(Keller et al., 1998). Potential binding sites for other TFs(Bergman et al., 2005; Vlieghe et al., 2006) could be found in the DA3 CRM, but we limited here our annotation to the conserved sites (see Fig. S2 in the supplementary material). The relative paucity of known TF-binding sites in the conserved sequence motifs found in the DA3 muscle CRM leaves largely open the question of the roles of these motifs in col regulation.

Heat-shock-driven, ubiquitous expression of the Col protein activated endogenous col transcription in a few muscles other than DA3, mainly the VL1 and, more sporadically, the DA2 muscle(Crozatier and Vincent, 1999). By using different col-lacZ reporter genes, we mapped the cis-regulatory element(s) responsible for this muscle-specific activation to the DA3 muscle CRM (Fig. 2B-Dand data not shown). As it is restricted to the DA3 and VL1 (and possibly DA2)muscles, we reasoned that col auto-activation was dependent upon a specific combination of TF expressed in these muscles. Of the known TFs expressed in somatic muscles, only Vestigial (Vg) and Nau are expressed in DA3 and VL1 (Bate et al., 1993; Dohrmann et al., 1990; Keller et al., 1998; Paterson et al., 1991). nau mutant embryos lack a subset of muscle fibres, with DA3 being the most severely affected (Balagopalan et al.,2001; Keller et al.,1998). By contrast, no muscle phenotype has yet been described for vg loss-of-function mutations. vg mutants show reduced wings and severe flight muscle defects but are viable and fertile, allowing the study of their maternal plus zygotic phenotype. We did not observe abnormal Col expression or abnormal DA3 muscle formation in vg mutant embryos,indicating that Vg is not required for formation of this muscle (data not shown).

In situ hybridisation with a col intronic probe that labels nascent transcripts revealed that col transcription is activated in the nuclei of those FCMs that are recruited to form the DA3 muscle(Crozatier and Vincent, 1999). To further investigate the mechanisms behind this observation, we compared the patterns of Col accumulation and col transcription during the process of DA3 muscle formation (Fig. 3A-C). We found that, throughout the FC/FCM fusion phase (stage 12-15), each DA3 muscle syncitium contains on average one or two nuclei, which stain positive for Col but do not transcribe col (see also Fig. 4). Close-up analysis of fusion events in stage 13 embryos further revealed that only nuclei containing high levels of Col protein activated col transcription(Fig. 3A). This strongly suggested that Col accumulation is a prerequisite for auto-activation in newly fused FCM nuclei. In support of this interpretation, all the DA3 muscle nuclei transcribe col after completion of the muscle fusion process(Fig. 3B), although this uniform expression phase is only transient, as col transcription declines abruptly during stage 16 to become undetectable(Fig. 3C). From these observations, we conclude that activation of col transcription occurs through a reiterative two-step mechanism, ensuring the same transcriptional programme to all nuclei of the DA3 myofibre. In a first step, nuclei from FCMs newly incorporated into the growing syncitium import some of the Col protein present in the muscle precursor (inset in Fig. 3A). In a second step, col transcription is turned on in these nuclei.

First evidence for col autoregulation during DA3 muscle formation came from the observation that col transcription is not maintained in the DA3 FC in col mutant embryos(Crozatier and Vincent, 1999). In order to investigate this phenotype in more detail, we constructed a P9col-Gal4 driver (P9cG), allowing us to express a membrane-bound form of GFP in the DA3 muscle progenitor and to specifically follow the fate of this progenitor in col mutant embryos(UAS-mcd8GFP/P9cg; Fig. 3D). In wt embryos, mCD8GFP remains expressed and is detected both intracellularly and at the plasma membrane of the DA3 myofibre. In col mutant embryos, mCD8GFP expression is lost early but stability of the protein at the plasma membrane allows the detection of the mutant DA3 fibres. This experimental set-up confirmed that fusion of FCMs with the DA3 FC is drastically impaired in col mutant embryos and that coltranscription is neither maintained in the DA3 FC nor activated in the nuclei of FCM, which sometimes fuse to form an abortive DA3 muscle precursor(Fig. 3E). These data establish that col auto-regulation and the muscle DA3 identity programme are intimately connected.

Nau activity is also required for formation of the DA3 muscle, although the described DA3 nau mutant phenotype is not as severe as for col (Keller et al.,1998). The presence of a consensus Nau-binding site in the col DA3 muscle CRM raised the possibility that one Nau function could be to regulate col transcription. To address this possibility, we first compared nau and col transcription in wt DA3 muscles,using in situ hybridisation to primary transcripts and Col immunostaining. This revealed that col and nau are transcribed together in the DA3 progenitor, FC and muscle precursor up to early stage 13(Fig. 4A-C). Subsequently, only col transcription persists in the DA3 muscle(Fig. 4D). We then looked at col transcription in embryos homozygous for the null allele nau¹⁸⁸ (Balagopalan et al., 2001; Wei et al.,2007). Based on Col and Myosin heavy chain (MHC) antibody staining(Fig. 4E,F and data not shown)the DA3 muscle was completely absent in around 5% of segments, abnormal in orientation in 45% and rather normal-looking in about 50% of segments,consistent with previous reports (Keller et al., 1998). Low amounts of Col protein were observed in nuclei of the `normal-looking' DA3 muscles (Fig. 4E,F), allowing us to look at col transcription in these nuclei. In wt embryos at stage 15, each DA3 muscle syncitium contains on average nine nuclei, which are all strongly stained with Col antibodies, and seven to eight are positive for col transcription(Fig. 4E; see also Fig. 3B). The DA3 fibres present in nau¹⁸⁸ embryos contained only seven nuclei on average, with most showing a low level of Col protein. However, at most two or three of those transcribed col(Fig. 4F). This result indicated that Nau activity is required, in addition to Col, for activation of col transcription in the FCM nuclei that are recruited by the DA3 FC. The Col protein that is detected in nau mutant embryos probably derives from earlier, Nau-independent col transcription. Supporting this conclusion, one nucleus, probably the FC nucleus, shows high levels of col transcripts in nau mutant embryos at late stage 12, when DA3 muscle precursors contain two or three nuclei(Fig. 4G,H). In summary, nau and col are expressed in the DA3 FC and both Nau and Col are required for col activation in the nuclei of newly recruited FCM,thereby ensuring that all nuclei within the DA3 myofibre acquire the same identity.

To further test the hypothesis of a combinatorial role of Nau and Col in conferring the DA3 muscle its identity, we examined the activation pattern of P2.6cl at stage 15 after either Nau alone, Col alone or Nau+Col were ectopically expressed in all muscle FCs (rp298Gal4 driver)(Menon and Chia, 2001). rp298Gal4-driven Col expression resulted in ectopic P2.6clexpression in several muscles other than DA3, including DA2, DT1 and VL2,although this expression was most robust in VL1, as seen in hs-colexperiments (Fig. 5A,B),without major phenotypic effects, at least at the level of muscle fibre morphology (data not shown). By contrast, rp298Gal4-driven Nau expression, while altering the pattern of muscle fibres, as previously documented with a heat-shock construct(Keller et al., 1997),provoked ectopic expression of P2.6cl only in a single muscle, the DA2 muscle (Fig. 5C). These data confirmed that, despite a more general role than Col in somatic myogenesis (Keller et al.,1997; Wei et al.,2007), Nau is generally unable by itself to ectopically activate col transcription. When Col and Nau were expressed together, P2.6cl was activated in the same muscles as with Col alone, but much more strongly (compare Fig. 5B with D), confirming that Nau potentiates the ability of Col to activate its own transcription. Interestingly, P2.6cl was activated by Nau+Col in a few muscles, including the SBM muscles, which did not respond to the presence of Col alone, indicating that Nau and Col may act synergistically. Still, many muscles remained refractory to this combination and did not express P2.6cl, suggesting that other competence factors are lacking or that negative regulation exerted by Notch and/or other factors may be dominant in these muscles.

The evolutionary conservation of a Nau-binding site and a potential EBF-binding site within the DA3 muscle CRM(Fig. 2B and see Fig. S2 in the supplementary material) suggested that regulation of coltranscription by Nau and Col could be direct. We independently mutated the putative Nau- and EBF-binding sites within the P2.6cl construct,giving rise to P2.6cl^nau and P2.6cl^col, respectively(Fig. 6F). P2.6cl^nau expression was either lost from the DA3 muscle or much reduced compared with P2.6cl(Fig. 6A,C), suggesting that Nau directly regulates col transcription. Unlike the case with P2.6cl, however, ectopic P2.6cl^nau expression was observed, indicating that the mutated E-box in the Nau site could also mediate binding of repressing factor(s) in absence of Nau. Col binds in vitro to the EBF consensus binding site (TTCT/CNNGGGAA)(Travis et al., 1993),consistent with sequence conservation of the COE DNA-binding domain(Dubois and Vincent, 2001)(V.D., unpublished). The closest match to the consensus EBF recognition site found within the DA3 CRM is the sequence ATGTCTGGGGAT, which is part of the conserved motif 7 (Fig. 6F and see Fig. S2 in the supplementary material). Gel-shift assays and immunoprecipitation of DNA-protein complexes formed by co-incubation of Col with synthetic oligonucleotides overlapping this predicted EBF-binding site failed to reveal strong binding in vitro (V.D., unpublished). Nevertheless,DA3-specific expression of P2.6cl^col in vivo was almost undetectable when this site was mutated(Fig. 6B,F), suggesting that it mediates col auto-regulation. To provide a different test of this in vivo function, we looked at P2.6cl^col activation in conditions of ectopic Col expression. Unlike P2.6cl(Fig. 6D), P2.6cl^col expression was activated very weakly, if at all,in the VL1 (and DA2) muscles (Fig. 6E). These results reinforce the conclusion that the predicted EBF/Col-binding site present within the conserved motif 7 is required for positive col autoregulation.

The stereotyped pattern of Drosophila body wall muscles relies upon the specification of FCs that seed the formation of individual muscles at specific positions in the somatic mesoderm(Baylies et al., 1998; Rushton et al., 1995). The current view is that the properties specific to each muscle result from the selective expression, in each FC, of distinct combinations of `muscle identity' TFs. However, experimental evidence for such a combinatorial code has remained sparse. We addressed here this question, using as a paradigm Col expression as both a determinant and read-out of DA3 muscle identity.

Functional dissection of the DA3 muscle CRM present in the colupstream region showed that col expression in the DA3 FC can be separated from its expression in the DA3/D05 progenitor and the promuscular cluster. It thus revealed the existence of three steps in the transcriptional control of muscle identity (Fig. 7). That col expression in the DA3/D05 progenitor could be uncoupled from that in promuscular clusters was in apparent contradiction with the previous conclusion from pioneering studies on Eve expression in dorsal muscle progenitors that this expression issued from Eve activation in promuscular clusters. Restriction of Eve expression to progenitors was considered a secondary step, mediated by N-signalling during progenitor selection by lateral inhibition (Carmena et al., 1998; Halfon et al.,2000). To reconcile our data and this model, we propose that the muscle DA3 CRM is only active in the DA3/D05 progenitor because it lacks some positively acting cis-elements necessary to counteract N-mediated repression of col transcription (Fig. 7A). We have indeed previously shown that coltranscription is repressed by N during the progenitor selection process(Crozatier and Vincent, 1999). We also noted that a Twi-binding site is present in the `progenitor' subdomain of the DA3 CRM (Fig. 2B and Fig. 7A). The functional importance of this site is supported by its in vivo occupancy in 4- to 6-hour-old embryos when selection of the DA3/DO5 progenitor takes place(Sandmann et al., 2007). Together, Twi in vivo binding and the col/P2.6cl/P2.3clexpression data suggest that Twi activity contributes to colexpression in the DA3/DO5 progenitor but may not be sufficient to override N repression of col transcription before progenitor selection. Additional binding sites for Twi present in the col upstream region,between positions -8.7 and -8.3, are also bound by Twi in vivo(Sandmann et al., 2007) and probably contribute to the robustness of P9cl expression in progenitor cells, but the question of which cis-regulatory elements mediate col activation in promuscular clusters remains open. From their Eve expression studies, Michelson and colleagues developed a computational framework to identify other FC-specific genes(Estrada et al., 2006; Philippakis et al., 2006). This framework, named Codefinder, integrates transcriptome data and clustering of combinations of binding sites for five different TFs (Pnt, dTCF, Mad, Twi and Tin). col/kn was selected by Codefinder owing to the presence of five clusters of binding sites, four of which are located within introns(Philippakis et al., 2006). It remains to be determined which of these could be responsible for colactivation in promuscular clusters, but it is interesting to note that another in vivo Twi-binding site in 4-6-hour-old embryos correlates with the 3′-most cluster (Sandmann et al.,2007). In addition to Twi, conserved binding sites for Nau and Mef2 are found within the DA3 CRM. The Mef2 binding site is located in a region required for robust DA3-muscle expression of a reporter gene(Fig. 2B, Fig. 7B; and see Fig. S2 in the supplementary material). A direct control of col transcription by Mef2 during the muscle fusion process is further supported by the recent finding that Mef2 binds in vivo to the col upstream region between 6 and 8 hours of embryonic development(Sandmann et al., 2006).

Detailed analysis of col auto-activation revealed a reiterative,two-step process: import of pre-existing Col protein in the FCM nuclei that incorporate into the growing DA3 myofibre precedes activation of coltranscription (Fig. 3). This process ensures that all incorporated FCM nuclei acquire the same identity. Nau is required for maintaining col transcription in the DA3 muscle precursor and this control is probably direct. The presence of a putative EBF-binding site in the DA3 muscle CRM also correlates with the Col requirement for maintaining its own transcription beyond the FC stage(Crozatier and Vincent, 1999). Thus, despite the failure of our assays to detect strong Col binding to this site in vitro, it appears to be essential for col auto-regulation in vivo. This suggests that in vivo binding is potentiated by one or more specific co-factor(s) present in the DA3 muscle. One co-factor is probably Nau, as the ability of Col to activate its own transcription in newly recruited FCM is dependent upon Nau activity(Fig. 7B). Nau is not sufficient, however, as many muscles containing both Nau and Col proteins do not activate col transcription(Fig. 5). Interestingly, mouse EBF (also known as Ebf1 and Olf1 - Mouse Genome Informatics) and E2A (Tcfe2a -Mouse Genome Informatics), a bHLH protein of the same subgroup as MyoD(Simionato et al., 2007), have been shown to act on the same target promoter and synergistically upregulate transcription of B-lymphocyte-specific genes, although no direct physical interaction between EBF and E2A could be found in vitro. This suggested that functional interaction of EBF and E2A, similar to Col and Nau, requires yet another factor (O'Riordan and Grosschedl,1999). Taking into account the restricted pattern of ectopic col activation in hs-col conditions, we hypothesised that Vg could be another component of the DA3 combinatorial identity(Bate, 1993; Frasch, 1999). However, we found that Vg is not required for DA3 muscle specification, leaving open the question of which factor may bridge Col and Nau functions.

Unlike col or P2.6cl, P2.3cl is expressed in the DA3 FC and muscle precursor but not the DA3/DO5 progenitor, showing that coltranscription in the progenitor and muscle precursor is under separate control. These two phases of col regulation are intimately linked,however, as Col is required for activating its own transcription in the nuclei of FCM recruited by the DA3 FC. This regulatory cascade may explain how pre-patterning of the somatic mesoderm and muscle identity are transcriptionally linked in the Drosophila embryo. As discussed above, the ability of Col to auto-regulate depends upon the presence of Nau,another muscle identity TF. Col and Nau act as obligatory co-factors for maintenance/activation of Col expression in all nuclei of the DA3 muscle, thus bringing to light a clear case of combinatorial coding of muscle identity.

We thank the Bloomington Stock Center, S. Menon and S. Abmayr for fly stocks, D. Kiehart for antibodies, A.M. Michelson for sharing unpublished results, M. Markstein for access to Fly Enhancer version 2, J. Boyes and G. Delsol for help with generating Col antibodies and S. Plaza and D. Cribbs for discussion. We acknowledge the help of the Toulouse RIO Imaging platform and B. Ronsin for confocal microscopy. This work was supported by CNRS,Ministère de la Recherche et de la Technologie, Université Paul Sabatier (grant to G. Delsol, Inserm U563 and A. Vincent) and Association Française contre les Myopathies. J.E. was supported by a fellowship from MRT.