Tup/Islet1 integrates time and position to specify muscle identity in Drosophila

Understanding how a cell acquires its identity is a fundamental problem in developmental biology. The somatic musculature of the Drosophila embryo is a classical model with which to study the regulatory processes that generate cellular diversity in response to positional information provided by embryonic body axes. Each hemi-segment shows a stereotypical arrangement of 30 somatic muscles connected to the exoskeleton, one alary muscle that attaches to the dorsal vessel, which is the equivalent of the heart, and six muscle stem-like cells [known as adult muscle precursors (AMPs)] (Bate and Rushton, 1993; Bate et al., 1991; Figeac et al., 2010; LaBeau et al., 2009). Each skeletal muscle is a syncytium seeded by a founder cell (FC), which undergoes multiple rounds of fusion with fusion-competent myoblasts. FCs are born from the asymmetric division of progenitor cells (PCs), themselves selected by Notch (N)-mediated lateral inhibition from equivalence groups of myoblasts, termed promuscular clusters at distinct positions within the somatic mesoderm (Carmena et al., 1995; Ruiz-Gómez and Bate, 1997). Characterization of selected muscle lineages has shown that muscle shape, size and orientation reflect the expression of specific ‘identity’ transcription factors (iTFs) in each PC and FC (Bate and Rushton, 1993; Baylies et al., 1998; Frasch, 1999; Tixier et al., 2010). Maintaining the expression of a subset of iTFs in one FC, but not its sibling (Ruiz-Gómez et al., 1997; Rushton et al., 1995), and propagation of this iTF code into all nuclei to the growing myofibre (Crozatier and Vincent, 1999; Dubois et al., 2007; Knirr et al., 1999) determine the final identity of each muscle. There are four dorsal muscles (DA1, DA2, DO1, DO2) and six dorsolateral (DL) muscles (DA3, DO3, DO4, DO5, DT1 and LL1) (Bate and Rushton, 1993; Nose et al., 1998). Detailed studies of the iTF Even-skipped (Eve) expression in the DA1 and DO2 lineages showed that eve cis-regulation integrates positional information provided by ectodermal Decapentaplegic (Dpp) and Wingless (Wg) signalling, tissue-specific information provided by the mesodermal transcription factors Twist (Twi) and Tinman (Tin), and Receptor Tyrosine Kinase (RTK) signals (Carmena et al., 1998; Duan et al., 2007; Knirr et al., 1999). Although these pioneering studies have provided a useful framework, there remain many gaps in our understanding of how each PC acquires a specific identity according to its position.

From a genetic screen designed to identifying new muscle identity genes (L.D., unpublished), we found that removal of the LIM homeodomain TF Tailup/Islet1 (Tup) causes a DA3 duplication. Islet1 was first characterized in vertebrates, as a regulator of insulin expression in the pancreas (Karlsson et al., 1990) and in motoneuron specification (Pfaff et al., 1996). Recently, it has been shown to be necessary for second heart field (SHF)-derived cells to populate the heart (Laugwitz et al., 2005; Moretti et al., 2006). Drosophila tup was initially characterized for its role in neuronal identity and axon pathfinding (Thor and Thomas, 1997), and was later shown to be required for development of the lymph gland and the specification of cardiac cells (Mann et al., 2009; Tao et al., 2007).

We report here that Tup is expressed in the four dorsal muscles and the alary muscles, and that the dorsal musculature is severely disorganized in tup-null embryos. We then focus our studies on the DA2 to DA3 muscle transformation. The DA2 PC is selected from myoblasts transiently expressing both Col and Tin. Tup activation in this PC during this time window is required to repress col transcription and impose the DA2 identity. Tup is then required for maintaining its own expression in dorsal muscles beyond the PC stage, suggesting a handover relay mechanism between two, early and late, cis-regulatory modules. Our findings provide a new dynamic view of transcriptional regulation that patterns the Drosophila dorsal musculature and novel parallels with specification of pharyngeal muscles in chordates.

Drosophila mutant and transgenic strains were: tup^ex4 (de Navascués and Modolell, 2010), col¹ (Crozatier et al., 1999), rp298-lacZ (Nose et al., 1998), Mhc-GFP (Eric Olson and Elisabeth Chen, Dallas, USA), twist-Gal4 (Baylies and Bate, 1996), IsletH-GFP, UAS-tup (Thor and Thomas, 1997), N^55e11 (Salzberg et al., 1994), tin^EC40 (Bodmer, 1993) and Vg-lacZ (Richard Cripps, Albuquerque, USA). Mutant strains were balanced over marked chromosomes: Cyo wg-lacZ; Cyo dfd-EYFP; TM3 twist-lacZ. Df(2L)Exel7072 and Df(2L)Exel7073 were obtained from the Bloomington Stock Center.

The 4_0.9col-moeGFP construct fuses a col genomic DNA fragment that contains 4 kb of upstream sequence and 560 bp of transcribed sequence extending into the Col protein region, in frame with a D-moesin GFP fusion (Polesello et al., 2002). It was inserted into the AttP-pS3AG transgenesis vector (Williams et al., 2008). tup DME-lacZ was constructed by placing a PCR-amplified tup genomic fragment, FlyBase positions 2L: 18897397..18900688 (release: r5.42), in the attB-inslacZ vector (Enriquez et al., 2010). 4_0.9col.moeGFP and DME-lacZ were inserted at position 68A4 on the third chromosome by injection into nosC31NLS;attP2 embryos (Bischof et al., 2007; Markstein et al., 2008).

Antibody staining and in situ hybridization with intronic probes were as described previously (Dubois et al., 2007). Primary antibodies were: guinea pig anti-Col (1:5000; Adrian Moore, Saitama-Ken, Japan), anti-L’sc (Stephen Crews, Chapel Hill, USA), anti-SrB (Talila Volk, Reohovot, Israel), mouse anti-Col (1:100) (Dubois et al., 2007), anti-Tup (1:100; Developmental Studies Hybridoma Bank), anti-GFP (1:100; Torrey Pines Biolabs), anti-β-galactosidase (1:1000; Promega), rabbit anti-Tin (1:750; Manfred Frasch, Erlangen, Germany), anti-Nau (1:100; Bruce Paterson, Bethesda, USA), anti-Kr (Ralf Pflanz, Goettingen, Germany), anti-Eve (1:1000) (Kosman et al., 1998), anti-b3-tubulin (1:5000; Renate Renkawitz-Pohl, Marburg, Germany), anti-Zfh1 (1:5000; Alain Garces, Montpellier, France), anti-Vg (Andrew Simmonds, Edmonton, Canada) and rat anti-Runt (1:300) (Kosman et al., 1998). Secondary antibodies were: Alexa Fluor 488- and 555-conjugated antibodies (1:300; Molecular Probes), and biotinylated goat anti-mouse (1:1000; Vector Laboratories). The tup intronic probe spans 2 kb of the first intron. Confocal sections were acquired on a Leica SP5 microscope at 40× magnification and 1024/1024 pixel resolution. Images were assembled using Photoshop software.

tup CRM predictions were based on the positions of Tin, Twi and Mef2 in vivo binding sites (Sandmann et al., 2007; Zinzen et al., 2009). These data were accessed via the Eileen Furlong’s lab webpage (http://furlonglab.embl.de/data/browse_chip_mod/).

Our knowledge of the regulatory program that controls the exquisite pattern of Drosophila larval muscles is still fragmentary. To identify new players, we undertook a genetic screen, starting from a collection of small overlapping deficiencies covering the second chromosome and using Col expression as a read-out. Two overlapping deficiencies, Df(2L)Exel7072 and Df(2L)Exel7073, revealed a duplication of the Col-expressing DA3 muscle in most segments (81%, n=249; 82%, n=230, respectively; Fig. 1A-D). The genomic region deleted in both deficiencies includes two genes, short spindle 3 (ssp3) and tup (Fig. 1A). tup^ex4 null embryos (de Navascués and Modolell, 2010) show a DA3 duplication in 78% of segments (n=102) (Fig. 1E-E′) and a similar phenotype is observed in hemizygous tup^ex4/Df(2L)Exel7072 embryos (not shown). Lack of tup is therefore responsible for the DA3 duplication phenotype observed in Df(2L)-Exel7072 and Df(2L)Exel7073 embryos. To better visualize the DA3 muscle, we expressed a moe-GFP fusion protein under control of a DA3-specific cis-regulatory module (CRM) (4_0.9colmoeGFP). Double phalloidin and GFP staining confirmed the transformation of one dorsal muscle into a DA3-like muscle in tup^ex4 embryos (Fig. 1F,G). Phalloidin staining also revealed that all dorsal muscles are affected to some extent in tup^ex4 embryos, with individual muscles often not recognisable from segment to segment, showing that tup activity is crucial for patterning the dorsal musculature.

Each muscle can be uniquely identified by its position and shape at stage 15. Staining for Tup showed its expression in the four dorsal muscles (Fig. 1H,H′), complementary to DA3 Col expression (Fig. 1H,J,K), in addition to alary muscles (Fig. 1I) (Tao et al., 2007). As muscle identity is conferred by activation of specific combinations of iTFs in each PC, we looked at Tup expression at the PC stage. As general marker of PCs/FCs, we used rp298lacZ (Beckett and Baylies, 2007; Enriquez et al., 2012; Nose et al., 1998). Staining of late stage 11 rp298lacZ embryos for lacZ and either Tup or Col, showed that Tup expression is expressed in the four dorsal-most PCs (Fig. 2A). The dorsal FCs have been positioned relative to each other and identified, based on the expression of specific iTFs, except for DA2 (Beckett and Baylies, 2007). We tentatively identified the DA2 PC as a cell expressing Nautilus (Nau), but not Col, and located immediately dorsal to the DA3/DO5 PC (Enriquez et al., 2010; Enriquez et al., 2012). Double staining for Tup and either Eve or Kr confirmed that this cell expresses Tup, and is Eve and Kr negative (Fig. 2D,E). Thus, Tup is expressed in the four dorsal PCs and contributes to their specific iTF code. Kr expression is reduced in the DO1 muscle in tup hypomorphic mutants (Mann et al., 2009). Accordingly, Kr expression is completely lost in tup^ex4 embryos, confirming that it crucially depends upon Tup (data not shown). On the contrary, Eve and Runt remain expressed in the DA1 and DO2 muscles, respectively, although an abnormal clustering of Runt-expressing nuclei correlates well with the abnormal morphology of the DO2 muscle in tup mutant embryos (supplementary material Fig. S1). Together, these data indicate that Tup acts both upstream of some ITFs and in combination with others in specifying dorsal muscle identities.

Only two dorsal muscle lineages, DA1 and DO2, were previously established, based on Eve expression. Both are mixed muscle/Eve pericardial cell (EPC) precursor lineages (Carmena et al., 1998; Speicher et al., 2008). To establish the DA2 lineage, we used a 7 kb tup cis-regulatory region, IsletH, which was initially characterized for its ability to drive GFP expression in a subset of post-mitotic motoneurons and interneurons (Thor and Thomas, 1997). We found that IsletH is active at stage 11 in the DA2 PC, as identified by its position, immediately dorsal to the DA3/DO5 PC (Fig. 2G). Following IsletH-GFP expression showed that the DA2 PC divides at stage 12. Double staining for GFP and Zfh1, a general AMP marker (Figeac et al., 2010; Sellin et al., 2009), established that this division is asymmetric, with one daughter cell becoming an AMP (Fig. 2H). The DA2 PC thus gives rise to a mixed somatic muscle/AMP lineage. Staining of stage 16 embryos for Zfh1 and GFP showed that the DA2 sibling AMP is the dorsal-most DLAMP (Fig. 2I). This AMP still expresses Zfh1 in tup mutant embryos, suggesting that tup is not required for its specification (not shown).

To determine which muscle ectopically activates Col expression and adopts a DA3 morphology upon loss of tup, we double stained wild-type and tup^ex4 embryos for Col and Vestigial (Vg). Vg marks the DA1, DA2, DA3 and LL1 muscles (Bate et al., 1993). In wild-type embryos, only the DA3 muscle expresses Col (Fig. 3A,A′). In tup^ex4 embryos, the DA2 muscle also expresses Col, correlating with its abnormal morphology, showing that it is transformed into DA3 in absence of tup activity (Fig. 3B,B′).

The DA2>DA3 transformation observed in tup mutants is also observed upon Col ectopic expression in the entire mesoderm (Twi>Col) (Enriquez et al., 2012). This phenotype similarity suggested that Tup could antagonize Col activity. To test this possibility, we overexpressed Tup in the entire mesoderm (Twi>Tup embryos). Staining for Col and Vg showed a loss of Col expression in the DA3 muscle (Fig. 3C) and high frequency of DA3>DA2 transformations (88% of segments, n=77), similar to a col loss-of-function phenotype (Fig. 3D). We conclude that Tup is able to repress Col expression in the DA3 muscle, correlating with the DA3>DA2 identity shift (Fig. 3C′,D′). The opposite DA2>DA3 and DA3>DA2 transformations observed in tup mutants or col gain-of-function and col mutants or tup gain-of-function, respectively, show that Col and Tup distinguish between the DA2 and DA3 muscle identities. Of note, however, Col is activated in the DA2>DA3 muscle in tup^ex4 embryos, whereas the DA3>DA2 transformation in col embryos occurs without Tup activation (Fig. 3B′,D′). This difference suggested that Tup is not instructive for the DA2 fate. To test this idea further, we looked at the muscle pattern of col¹; tup^ex4 double mutant embryos. β3-Tubulin staining showed a cumulative phenotype; that is, dorsal muscle defects and a lack of DA3 muscle (supplementary material Fig. S2). Staining col¹; tup^ex4 embryos for Vg confirmed that, in absence of tup and col activity, the DA2 and DA3 muscles display a DA2 morphology (Fig. 3E,E′). It indicates that DA2 is default morphology and suggests that DA2 iTFs other than Tup remain to be identified.

To describe more precisely the DA2>DA3 morphological transformation in tup mutants, we compared the epidermal attachment sites of the DA2 and DA3 muscles in wild-type and in tup^ex4 embryos. Staining Vg-lacZ embryos for lacZ and Stripe B (StrB), a marker of tendon cells (Volk and VijayRaghavan, 1994; Volohonsky et al., 2007), indicated that the wild-type DA2 and DA3 attachment sites are not overlapping (Fig. 3F). StrB staining of wild-type and tup^ex4; 4_0.9colmoeGFP embryos showed an overlap between the wild-type DA3 and tup^ex4 DA2>DA3 insertion sites (Fig. 3G,H). Thus, in absence of tup, the DA2 muscle targets the DA3 tendon cells. Some partial DA2>DA3 transformations were also observed in tup mutant embryos, with triangular shaped muscles (Fig. 3H). This phenotype underlines the existence of two successive steps in the attachment of elongating muscles to specific tendon cells, a transient attachment, followed by a selection of the final attachment sites, with the second step being tightly regulated by the iTF code (Enriquez et al., 2012). Whether iTFs other than Col and Tup contribute to the robustness of the selection of DA2 versus DA3 attachment sites remains an unresolved issue.

Temporal aspects of Col expression in the DL PCs showed that they are specified sequentially. The DA3/DO5 PC is specified first, followed by the DO3/DT1 PC (only in the A1-A7 segments; not further considered here) and, finally, the LL1/DO4 PC (Enriquez et al., 2012). We previously noticed that the Nau-positive cell now identified as the DA2 PC (Fig. 2), singles out from the Col-positive cluster earlier than the DA3/DO5 PC and transiently expresses low amounts of Col (Enriquez et al., 2012). This observation suggested that the repression of Col by Tup was exerted in the DA2 PC, raising the question of the relative kinetics of tup and col transcription in this cell. To address this question, we used in situ hybridization with intronic probes to detect tup and col nascent transcripts. Double in situ hybridization revealed that tup is strongly transcribed in only one cell among the cluster of myoblasts that transcribe col at stage 10 (Fig. 4A,A′). Examination of single confocal sections showed that this unique tup- and col-positive cell has singled out and adopted a more external position than the rest of the cluster (Fig. 4B-D). By stage 11, however, whereas tup remains transcribed, col transcription is not detected anymore in this cell, unlike in the DA3/DO5 FCs and LL1/DO4 PC, which is singling out at this stage (Fig. 4E-H). We conclude that col and tup are co-transcribed in the DA2 PC, prior to col extinction. To determine whether tup maintenance and col extinction are linked, we compared col transcription in the DA2 PC in wild-type and tup^ex4 embryos, using nau transcription as an internal reference. In wild-type embryos, col and nau transcripts are detected in the DA3/DO5 PC, identified by the accumulation of Col protein, while only nau transcripts are detected in the DA2 PC (Fig. 4I,L). In tup^ex4 mutant embryos, nau and col transcripts, and Col protein are detected in the DA2 PC (71% of segments; n=60) in addition to the DA3/DO5 PC (Fig. 4J,L), showing that Tup represses col transcription in the DA2 PC. Promuscular col transcription at stage 10 is not affected in twi>tup embryos (supplementary material Fig. S3), confirming that Tup repression of col is exerted at the PC stage. The lower level of Col protein in the DA2, compared with DA3 PC, in tup mutant embryos (Fig. 4J), suggests, however, another level of col regulation.

Eve and S59 expression in selected lineages showed that PC-specific expression of some iTFs is established in two steps: first, activation in an equivalence group of competent cells, followed by restriction to PC(s), which involves N-mediated lateral inhibition (Buff et al., 1998; Carmena et al., 2002; Knirr and Frasch, 2001). col expression in the DL PCs follows this scheme (Fig. 4A-F) (Crozatier and Vincent, 1999; Enriquez et al., 2010). To determine whether tup expression in the DA2 lineage is first activated in an equivalence group, we looked at tup transcription in N mutant (N^55e11) embryos where lateral inhibition is defective. In absence of N, several myoblasts co-transcribe col and tup (Fig. 4M,M′), revealing the existence of a group of cells from which the DA2 PC is selected. Only the dorsal-most cells of the col cluster transcribe tup (Fig. 4M,N), providing an explanation of why col repression by Tup only occurs in the DA2 PC. In turn, it raised the question of which mechanisms control the different registers of tup and col activation.

Tin is a major regulator of cell fate in the dorsal mesoderm (Bodmer, 1993; Liu et al., 2009), which shows a dynamic expression pattern (Azpiazu and Frasch, 1993; Bodmer, 1993; Johnson et al., 2011; Xu et al., 1998; Yin et al., 1997). Tin is first expressed in the entire trunk mesoderm, followed by upregulation in the dorsal mesoderm, in response to Dpp signalling from the dorsal ectoderm (stage 10, Fig. 5A). It is during this second phase that Tin has been proposed to specify different dorsal mesoderm fates. Later, Tin expression is restricted to cardiac progenitors (late stage 11, Fig. 5D), before directing heart cell diversification (Fig. 5E,F). Of specific interest here, is the relationship between Tin, Col and Tup expression during the second phase of Tin expression. At stage 10, a small group of myoblasts express both Col, Tin and the proneural protein Lethal of scute (L’sc) (Fig. 5A,B), suggesting that it corresponds to the equivalence group of tup/col transcribing myoblasts revealed in N mutant embryos (Fig. 4N). Accordingly, residual Tin is detected in the DA2 PC (Fig. 5C). The overlap between Tin and Col expression is transient, however, because at late stage 11, while the Col-expressing cluster has expanded ventrally, Tin expression has retracted dorsally (Johnson et al., 2011; Reim and Frasch, 2005), such that Col and Tin are now expressed in complementary domains (Fig. 5D). As a consequence, the DA3/DO5 (and DT1/DO3) progenitors, which are born more ventrally, and later than the DA2PC, are selected from cells that do not anymore express Tin (Fig. 5E).

Tin role in patterning the dorsal muscles has not been studied in detail, except for its requirement for Eve expression in the DO2 and DA1 progenitors (Knirr and Frasch, 2001). In silico search and chip-on-chip experiments suggested that Tin could directly regulate the expression of several iTFs, including col (Enriquez et al., 2010; Liu et al., 2009; Philippakis et al., 2006). Promuscular Col expression is not lost in tin^EC40-null embryos, however, but extends dorsally, to reach the dorsal edge of the embryo proper and abut the amnioserosa, which is expanded in these mutants (Bodmer, 1993) (Fig. 6A,B). Thus, during its first phase of expression, Tin sets the dorsal limit of Col expression, via a mechanism that remains unknown. At stage 15, Col is ectopically expressed in the DA2 muscle in 56% of tin^EC40 segments (n=146), correlating with a DA3-like morphology (Fig. 6C,D). This mutant phenotype, reminiscent of tup mutants, prompted us to compare tup and col transcription between wild-type and tin^EC40 embryos. At stage 11, while col is transcribed in several PCs in tin mutant embryos, tup transcription is strongly reduced in all dorsal cells and lost in the DA2 progenitor (Fig. 6F), indicating that Tin acts upstream of tup in specifying DA2 identity. Consistent with this loss of tup transcription, col remains transcribed in the DA2 muscle throughout differentiation (Fig. 6H). However, it is not expressed in more dorsal muscles where tup expression is also lost.

In summary, the staggered patterns of Tin and Col activation in stage 10 embryos define three distinct types of PCs along the DV axis: the DA1, DO1 and DO2 PCs, which express Tin and Tup; the DA2 PC, which expresses Tin, Tup and Col; the three DL PCs, which express Col but not Tin. Both spatial and temporal aspects of Tin, Tup and Col regulation are involved in distinguishing the DA2/AMP and DA3/D05 identities (Fig. 8).

The absence of tup transcription and IsletH expression in the DA2 PC in tin^EC40 embryos (Fig. 7B,C) suggested that activation of tup transcription by Tin is mediated by IsletH, consistent with in vivo binding of Tin (Sandmann et al., 2007) (Fig. 7A). IsletH activity in the DA2 muscle is only transient, however (Fig. 7H), indicating the need for (an)other cis-regulatory region(s) responsible for maintaining tup transcription throughout muscle development. To identify other CRMs active in the mesoderm, we profited from data concerning the position of ChIP-validated binding sites of Tin, Twist and Mef2, which define so-called meso-ChiP CRMs (Liu et al., 2009; Sandmann et al., 2007; Zinzen et al., 2009). One such CRM maps to IsletH and overlaps tup-F4, a 1.5 kb IsletH subfragment previously shown to drive GFP in the heart, lymph gland and alary muscles (Tao et al., 2007); tup-F4 and IsletH give identical GFP patterns in dorsal muscles (data not shown). A second Meso-Chip CRM located upstream of IsletH binds Twi and Mef-2, but not Tin (Fig. 7A). An overlapping DNA region was retrieved from an in silico search for non-coding sequences selectively enriched for combinations of Twi-, Tin- and Ets-binding sites (Philippakis et al., 2006; Warner et al., 2008) (A. Aboukhabil and M. Bulyk, personal communication). Taking into account ChiP-ChIP data, and sequence conservation during evolution, we identified a 3 kb fragment, designated below as DME (dorsal muscle enhancer), that drives lacZ expression in heart cells, the dorsal and the alary muscles (Fig. 7A,H; supplementary material Fig. S4). DME starts to be active in the DA2 FC at stage 12 (Fig. 7E,F) and remains active in the dorsal and alary muscles up to stage 16, at a time when IsletH-driven expression is no longer detected (Fig. 7G,H). The sequential activity of IsletH and DME in the DA2 lineage was reminiscent of the sequential activity of an early CRM (CRM276) and a late CRM (4_0.9col CRM) in the control of mesodermal col transcription (Enriquez et al., 2010). It raised the possibility of a similar relay process, whereby Tup protein expressed in the DA2 PC under control of IsletH, would subsequently auto-regulate its own expression via DME. To test this possibility, we examined IsletH and DME-lacZ expression in tup^ex4 embryos and found that, while IsletH remains active (Fig. 7D), DME-lacZ expression is completely lost from the DA2 PC (Fig. 7I,J). Staining stage 16 embryos confirmed that DME activity in the dorsal muscles is dependent upon Tup (Fig. 7K,L). In situ hybridization confirmed that tup transcription in dorsal muscles is lost in the absence of tup (not shown). We conclude that Tup-positive autoregulation, mediated directly or indirectly by DME, is required for tup expression in dorsal muscles throughout differentiation (Fig. 7M).

Drosophila Tup/Islet plays key roles in specification of neurons (Thor and Thomas, 1997) and cardiac and lymph gland progenitor cells (Mann et al., 2009; Tao et al., 2007). We show here that it is required for somatic muscle development.

The pattern of rp298lacZ expression and the three-dimensional arrangement of FCs distinguished four groups: dorsal, dorsolateral, lateral and ventral (Beckett and Baylies, 2007; Nose et al., 1998). Whether this topology reflects specific genetic programs has remained unclear. Tup and Col are expressed in the four dorsal and three DL PCs, respectively, supporting the notion of DV regionalization of the somatic mesoderm. This notion was evoked by regional Pox meso (Poxm) expression in most ventral and lateral FCs (Duan et al., 2007). As other known iTFs are only expressed in subsets of dorsal PCs/FCs (Tixier et al., 2010), it raised the possibility that Tup could reside at the top of dorsal iTF cascades. Our data show that this is not the case, as Tup, although required for Kr expression in the DO1 PC (Mann et al., 2009) (data not shown) and for Col repression in the DA2 PC, is not required for expression of Eve, Runt and Vg in the DA1, DO2 and DA2 and DA1 lineages, respectively (Fig. 2; supplementary material Fig. S1). Likewise, Col is required for expression of some iTFs but not others in DL PCs (Enriquez et al., 2012). Together, the patterns of Col, Eve, Kr, Poxm, Runt, Tup and Vg expression in wild-type and tup or col mutant conditions underline the intertwined, combinatorial nature of transcriptional regulatory networks specifying muscle identity. The DA2 PC gives rise to the DA2 muscle/DL AMP mixed lineage. Each abdominal hemisegment features six AMPs at stereotypical positions (Figeac et al., 2010). Other AMPs originate from mixed lineages, e.g. the ventral VA3/AMP and lateral SBM/AMP lineages. The VA3/AMP and SBM/AMP PCs express Poxm and S59, and Lb, respectively (Jagla et al., 1998; Ruiz-Gómez and Bate, 1997). Tup expression in the DA2/AMP lineage confirms that different AMPs express different iTFs at the time of specification. Whether, as for somatic muscles, the iTF code confers specific properties to each AMP remains an unresolved issue.

How PCs born at similar positions in the somatic mesoderm come to express different combinations of iTFs and acquire distinct identities has remained elusive. For example, what distinguishes the fate of the two Eve-expressing PCs, which are sequentially born from the same dorsal cluster, is unknown (Carmena et al., 2002; Carmena et al., 1998). One other example is the expression of S59 and Lb, each in one of two abutting ventrolateral PCs: the LO1/VT1 and SBM PCs. Activation of both Lb and Slo expression in the two PCs is controlled by the same upstream regulator, Org-1 (Schaub et al., 2012). Subsequent reciprocal secondary cross-repression results in exclusive S59 or Lb expression (Junion et al., 2007; Keller et al., 1997), but the nature of the presumed positional bias responsible for the oriented resolution of this cross-repression has not been explored. In the case of the adjacent DA2 and DA3 PCs, we show here that Tup activation by Tin in the DA2 PC is instrumental in distinguishing between DA2 and DA3 identities. The DA2 PC is selected from a small group of cells at the intersection between Tin and Col expression domains. Thus, the relative registers of tin and col expression along the DV axis provide precise positional information. Another key is timing. The overlap between Tin and Col expression is only transient, such that only the earlier-born Col-expressing PC expresses Tin. This provides a unique temporal window for Tup activation and Col repression (Fig. 8). The transient overlap is due to the dorsal restriction of Tin expression to cardial cells during stage 11 (Yin et al., 1997). This dynamic process is controlled by JAK-STAT signalling activity in the mesoderm, which is itself modulated by Tin activity (Johnson et al., 2011; Liu et al., 2009). The key function of Tup in the DA2 PC, which is to distinguish between two muscle identities, illustrates how cell identity can be specified with single-cell precision when temporal dynamics are combined with positional information.

Some iTFs are expressed during all steps of myogenesis, from promuscular stage to muscle attachment (Baylies et al., 1998; Frasch, 1999). Schematically, two major phases of expression can be distinguished: (1) PC specification when multiple iTFs are expressed in different PCs and extensive cross-regulation occurs, leading to FC-specific iTF patterns (Enriquez et al., 2012; Junion et al., 2007; Knirr et al., 1999; Ruiz-Gómez and Bate, 1997) (Fig. 2; supplementary material Fig. S1); and (2) muscle differentiation when the FC pattern is maintained and propagated into the syncytial fibre (Rushton et al., 1995) via transcriptional activation of the iTF code in newly fused FCMs (Crozatier and Vincent, 1999; Dubois et al., 2007; Knirr et al., 1999). Analysis of col regulation in the DA3 lineage showed that these two phases rely on two separate, early (CRM276) and late (4_0.9) CRMs, the activity of the late CRM requiring Col provided under the control of the early CRM. We termed this auto-regulatory loop a CRM handover mechanism (Dubois et al., 2007; Enriquez et al., 2010). Here, we have provided evidence that tup transcriptional regulation in the DA2 muscle follows the same rule. On the one hand, tup activation by Tin is mediated by an early CRM, IsletH; on the other, tup expression is maintained in differentiating muscles via a late CRM, DME, the activity of which depends upon Tup (Fig. 7). We propose that this handover relay mechanism could be a widespread mode of iTF regulation, as it efficiently links early steps of muscle specification in response to positional information with final muscle identity.

Tup and Tin are key components of the transcriptional regulatory cascade that controls early cardiogenesis, with Tin acting to activate Tup, the expression of which then persists after Tin has ceased to be expressed. We have now shown that a similar cascade operates in the somatic muscle mesoderm. Tup and Tin expression in both the heart and dorsal somatic muscles recalls Nkx2.5 (Tin ortholog) and Islet1 expression in the pharyngeal mesoderm, which contributes to some head muscles and part of the vertebrate heart (Kelly et al., 2001; Lescroart et al., 2010). Nkx2.5 is required for deployment of the SHF (Moretti et al., 2006; Prall et al., 2007) and Islet1 marks SHF progenitors that contribute both to the right ventricle and the arterial pole of the forming heart and a subset of skeletal pharyngeal muscles (Cai et al., 2003; Nathan et al., 2008; Tzahor and Evans, 2011). Similarly, in the simple chordate Ciona intestinalis, Nk4 (Tin/NKx2.5) marks the cardio-pharyngeal mesoderm at the origin of the heart and atrian siphon muscles (ASMs) that are evocative of vertebrate pharyngeal muscles. Islet-expressing cells also contribute to ASMs (Stolfi et al., 2010; Tolkin and Christiaen, 2012), suggesting an evolutionarily conserved link between cardiac and pharyngeal muscle development. Interestingly, the ascidian Col/EBF ortholog Ci-COE, is expressed in ASM precursors and is a crucial determinant of the ASM fate (Stolfi et al., 2010), reminiscent of Xenopus XCoe2 expression and requirement in pharyngeal arches for aspects of jaw muscle development (Dubois et al., 1998; Green and Vetter, 2011). It is now well established that distinct genetic networks govern skeletal myogenesis in the vertebrate head and trunk (Bothe and Dietrich, 2006; Sambasivan et al., 2011). The repertoire of TFs differentially deployed in the head mesoderm includes Tbx1, the Drosophila ortholog of which, Org-1, has recently been shown to act as a muscle iTF (Schaub et al., 2012). Tin/Nkx2.5, Tup/Isl1, Org-1/Tbx1 and Col/EBF may thus be part of a repertoire of transcription factors co-opted and diversified to regulate muscle patterning in Drosophila trunk and head muscle patterning in chordates.

We thank the Bloomington Stock Center and colleagues for Drosophila strains and antibodies; Martha Bulyk for sharing unpublished PhylCRM predictions; and Laetitia Bataillé, Caroline Monod, David Cribbs and Stéphane Zaffran for critical reading of the manuscript. We acknowledge the help of Brice Ronsin, Toulouse RIO Imaging platform, and Frédéric Luce and Julien Favier for maintenance of fly stocks.