An Org-1–Tup transcriptional cascade reveals different types of alary muscles connecting internal organs in Drosophila

The musculature of each animal species is composed of a stereotypical array of morphologically distinct muscles, which allows the precision and efficiency of body movements. Elucidating the mechanisms involved in conferring a specific morphology and function to each muscle remains a major challenge in myogenesis research. The somatic musculature of the Drosophila larva has for 20 years been a classical model with which to study this question. Each hemisegment of the larva shows a stereotypical arrangement of ∼30 distinct body wall muscles, which attach to the exoskeleton at distinctive positions (Bate, 1993). Each body wall muscle is seeded by a founder cell (FC), which undergoes multiple rounds of fusion with fusion-competent myoblasts (FCMs) (Rushton et al., 1995). FCs are born from the asymmetric division of progenitor cells (PCs) specified at precise positions and times within the somatic mesoderm. Detailed characterization of a few somatic muscle lineages has established that muscle identity (specific shape, size, orientation) reflects the expression of specific muscle identity transcription factors (iTFs) in each FC that act in a combinatorial manner (Baylies et al., 1998; Frasch, 1999; de Joussineau et al., 2012; Enriquez et al., 2012).

In addition to body wall muscles, one alary muscle (AM) in each abdominal hemisegment connects the dorsal vessel, which is the functional equivalent of the heart, to the lateral epidermis. AMs were identified some years ago in Drosophila (Rizki, 1978), but have largely escaped attention. The name alary refers to a wing-like (aliform) shape. Sparse descriptions of AMs in adults of different insect species led to proposals of their role in maintaining the heart in its dorsal position or/and controlling ostia opening and heart beating (Dulcis and Levine, 2003; Glenn et al., 2010). A detailed ultrastructural study showed that AMs adhere to the pericardial cells via flexible extracellular matrix (ECM) fibers (Lehmacher et al., 2012). In the Drosophila embryo, seven pairs of AMs attach to the dorsal vessel in a segmental arrangement; four connect to the posterior heart proper in abdominal segments A4-A7 and three connect to the aorta in segments A1-A3. The absence of AMs in thoracic segments correlates with the anterior aorta ending at the T2/T3 boundary, and bending ventrally to form a funnel-shaped outflow tract that is anchored by the heart-anchoring cells and cardiac outflow muscles (Zmojdzian et al., 2008). We recently found that both Org-1 (Optomotor-blind-related-gene-1), which is the ortholog of the vertebrate T-box factor Tbx1, and Tailup (Tup), the ortholog of the LIM-homeodomain factor Islet1, are expressed in AMs, and we devised reporter genes that allow the embryonic development of these muscles to be followed (Schaub et al., 2012; Tao et al., 2007; Boukhatmi et al., 2012).

We report here on the transcriptional control of AM development and their attachment to the epidermis and heart tube. We show that AMs run internally and affixed to the main, dorsal branch of trachea and externally to the lateral branches. We also confirmed transient, sequential contacts between several AMs and the tip cells of the anterior Malpighian tubules (MTs), as recently reported by Weavers and Skaer (2013). Our lineage analyses further revealed the existence of alary-related muscles in the thorax, which connect the epidermis of thoracic segments T1 and T2 to precise positions of the midgut. We designate these previously unrecognized types of muscle as ‘thoracic alary-related muscles’ (TARMs). Thus, AMs and TARMs together connect the circulatory system and visceral organs to the exoskeleton of the larva, and buckle the trachea, suggesting that they play both anatomical and physiological roles during larval development.

AMs, as well as TARMs, are either missing or malformed in org-1 and tup mutant embryos. Epistasis experiments and mutational analysis of tup cis-regulatory modules (CRMs) show that tup expression is under direct control of Tup itself in body wall muscles and under the control of Org-1 in AMs. Vertebrate Islet1 and Tbx1 are major players in the genetic program that controls the fate of pharyngeal mesoderm progenitors, which give rise to several aspects of the heart and head musculature (Laugwitz et al., 2005; Moretti et al., 2006; Sambasivan et al., 2009; reviewed by Vincent and Buckingham, 2010). Future studies will be required to establish which ancestral regulatory interactions have been redirected to foster the emergence of insect AMs/TARMs and hypothetical functional equivalents in other phyla.

Org-1, the Drosophila ortholog of vertebrate Tbx1, is expressed in the three larval body wall muscles, namely LO1, VT1 and SBM, as well as in AM in abdominal segments, in addition to the visceral musculature (Schaub and Frasch, 2013). Tup is also expressed in AMs, in addition to the heart and the dorsal DA1, DA2, DO1 and DO2 muscles (Tao et al., 2007; Boukhatmi et al., 2012), dorsal ectoderm, amnioserosa and precursor cells of the pharynx (Thor and Thomas, 1997; Ashe et al., 2000). Expression of the org-1-HN39-lacZ (org-1-lacZ) reporter recapitulates org-1 expression in the somatic musculature (Schaub et al., 2012), while both IsletH-GFP and tupF4-GFP, two overlapping reporter constructs, reproduce tup expression in AMs (Tao et al., 2007; Boukhatmi et al., 2012).

Staining embryos for IsletH-GFP and either Org-1 or org-1-lacZ [by β-galactosidase (β-gal)] expression confirmed that AMs are the only abdominal muscles expressing both Org-1 and Tup (Fig. 1A,B). It also revealed the existence of two multinucleated fibers straddling the thoracic T1-T3 and the T2-abdominal 1 (A1) segments, respectively (Fig. 1A-C). In addition to expressing both Org-1 and Tup, these transverse, multinucleated muscles, which do not attach to the heart, display a thin, very elongated shape similar to that of abdominal AMs (Fig. 1C). Henceforth, we refer to them as ‘thoracic alary-related muscles’ (TARMs). The discovery of TARMs prompted us to compare in detail the development and morphology of AMs and TARMs in Drosophila embryos.

The morphological description of abdominal AMs has focused, so far, on their dorsal connection to pericardial cells that surround the cardiomyocytes in the abdominal segments (Rizki, 1978). Ultrastructural and immunocytochemistry studies performed in larvae and adults showed that AMs dorsally form delta-like extensions connecting to the ECM that surrounds pericardial cells and contains Pericardin (Prc) (Lehmacher et al., 2012). Three-dimensional reconstruction of embryos stained for org-1-lacZ and either Prc or β3-tubulin (β3-Tub), which stains all muscles, shows that the AMs run immediately anterior to abdominal segmental boundaries, with dorsal projections extending on either side of the boundary (Fig. 2A,F; supplementary material Fig. S1). The anterior dorsal extension of the AM in A1 connects to the posterior aspect of the lymph gland, which is the prospective posterior signaling center (Crozatier et al., 2004). Nomarski examination of embryos stained for tup-moeGFP reporter expression (see below) shows that the dorsal extensions of two adjacent AMs connect with one another (supplementary material Fig. S1).

To further describe AM morphology visualized with org-1-lacZ, we used additional markers. Staining for stripe (sr)>GFP, a marker of tendon cells, which mediate attachment of muscles to the exoskeleton (Frommer et al., 1996), showed that the ventral AM attachment site is positioned slightly dorsal to the SBM and ventral to the DO3 and LL1 epidermal attachment sites (Fig. 2A). Triple staining for β3-Tub, tupAME-moeGFP (see description below) and βPS integrin showed βPS accumulation both at the AM epidermal attachment site and at dorsal projections in contact with either the PSC or pericardial cells (supplementary material Fig. S2). Three-dimensional reconstruction of embryos stained for sr>GFP and Prc showed that AMs form an arc of a circle inside the body of the embryo (Fig. 2B,B′,F). Triple staining for β3-Tub, org-1-lacZ and 2A12 to visualize the tracheal network (Tiklova et al., 2013) indicated that AMs run internally to the dorsal, main tracheal branch, and externally to the ventral branch, suggesting an architectural role in pressing the dorsal branch against the body wall (Fig. 2C-C‴,F).

Finally, staining for Cut, a marker of MT cells, showed that the distal tip cell of anterior MTs makes specific, integrin-mediated transient contacts with specific AMs during the MT lengthening and bending process (supplementary material Fig. S2) (Weavers and Skaer, 2013). It first contacts the AM in A5 at the A5/A6 segmental boundary, then contacts the AM in A4 at the A3/A4 boundary, to finally anchor the MT to the AM in A3 at the A3/A4 boundary (Fig. 2D-D‴,F; data not shown). The dynamic sequence of MT tip cell/AM interactions has recently been described independently by Weavers and Skaer (2013). We further observed that the anterior MT tip cell is org-1-lacZ positive (Fig. 2E). Closer examination showed that the cell immediately following the tip cell is also slightly org-1-lacZ positive, suggesting that it is the tip cell sibling and, by inference, that org-1-lacZ (Org-1) is expressed in the tip cell progenitor. Weavers and Skaer (2013) showed that adhesion between the MT tip cell and AMs is required for the looped architecture and normal dorsal positioning of anterior MTs; when either the tip cell is not specified or AMs are laser ablated, the anterior MTs are misshapen and extend too far into the anterior of the embryo. Since AMs are missing or severely malformed in org-1 mutant embryos (Schaub et al., 2012), a similar MT phenotype might have been expected. However, we observed a more severe phenotype – that is, a posterior shifting of the MT kink and tip cell position (supplementary material Fig. S3). This more severe phenotype suggests that, apart from being required for AM development, Org-1 could play a specific role at the MT tip. org-1 is also required for Dpp expression in the visceral mesoderm, which serves as a cue for anterior migration of the kinked portion of the MTs (Schaub and Frasch, 2013; Weavers and Skaer, 2013). Hence, the observed MT phenotype is explicable by a cumulative defect of disrupted anterior guidance, absence of the attachment substratum of the tip cells, and the failure of the tip cells to attach to alternative targets.

In summary, our analysis of AM morphology in fully developed embryos shows that AMs attach to the dorsal vessel, run internally to the dorsal branch of tracheae and anchor ventrally to epidermal tendon cells. In addition, the AM in A3 anchors the tip cell of the anterior MT at the end of embryogenesis. Altogether, these topological connections suggest a role of AMs in maintaining the proper anatomical organization of different internal organs of the larva (Fig. 2F).

Unlike AMs, which are vertical and run along the abdominal segmental borders, TARMs display oblique orientations and straddle several adjacent segments. In order to describe their morphology in detail, we used the same markers as for AMs plus markers of the visceral muscles. Stage 16 embryos stained for org-1-lacZ, sr>GFP and β3-Tub showed that the TARMs connect to tendon cells along the posterior boundary of segments T1 and T2. We designate them as TARM^T1 and TARM^T2. At this stage, no TARM is observed in T3. Like AMs in abdominal segments, TARM^T2 attaches close to the SBM/DO3/DO4 attachment sites, and TARM^T1 close to the DO3/DO4/LA1/LA2 attachment sites, with no SBM being formed in T1 (Fig. 3A,A′). 2A12 staining showed that, similar to AMs, TARMs run internally to the dorsal branch of the trachea in the thoracic region. Closer examination showed that this branch, which supplies oxygen to the pharyngeal muscles, loops around TARM^T1, with a 90° rotation (Fig. 3B,B′,G). Three-dimensional reconstructions showed that the posterior attachment sites of TARM^T1 and TARM^T2 are located deep inside the embryo. Triple staining for org-1-lacZ, Fasciclin III (FasIII) and Mesh (Fig. 3C,F), and double staining for HLH54F-RFP, org-1-RFP and MHC-tauGFP (Fig. 3D,E), to visualize visceral muscles and the midgut epithelium as well as the outer epithelial layer of the proventriculus (Izumi et al., 2012), showed that TARM^T2 connects to the gut at the position of the first constriction, before any morphological sign of its invagination (Fig. 3C). At this stage, TARMs^T1 make head-on contacts with the primordia of the gastric caeca (Fig. 3C′,G). These connections are integrin positive (supplementary material Fig. S2C) and are maintained at stage 16/17, when gut constrictions have resolved into loops, via βPS-integrin-dependent attachments (Fig. 3D,E; and data not shown). Rotation of the gut in connection with loop formation (Okumura et al., 2010) results in stretching of the TARMs and bilateral asymmetry of their left and right anchoring sites (Fig. 3F,F′; supplementary material Fig. S2). Through these experiments, we noticed the existence of a third pair of myofibers expressing Tup and Org-1, which connect to the proventriculus (Fig. 3F′,G; supplementary material Fig. S4). As we have identified neither its anterior attachment site nor its PC, this myofiber, named TARM*, was not considered further in this work.

In conclusion, TARMs bridge different midgut regions to the exoskeleton, while intertwining with cephalic tracheal branches. Based on anatomy, we hypothesize that AMs and TARMs could contribute to maintaining organs in their proper positions during larval foraging movements.

Org-1 is expressed in the AM progenitor at late embryonic stage 11 (Schaub et al., 2012). Staining embryos at that stage for Tup and rp298(duf)-lacZ, a marker of muscle PCs and FCs (Nose et al., 1998), identified the four PCs at the origin of the DA1, DO2, DO1 and DA2 dorsal muscles, plus another cell located several cell diameters posterior to the DA2/AMP PC (Boukhatmi et al., 2012). Double staining for Org-1 and IsletH-GFP shows that this cell is the AM progenitor and is segmentally repeated along the posterior/anterior segmental border (Fig. 4A; supplementary material Fig. S5). Double staining for org-1-lacZ and Nau showed that the AM PCs do not express Nau, in contrast to the dorsal body wall muscle PCs/FCs (Fig. 4B,B′).

The AM PC divides at stage 12, with one daughter cell remaining Org-1 positive (Fig. 4C-C″). Double staining for IsletH-GFP and Zfh1, a general AMP marker (Sellin et al., 2009; Figeac et al., 2010), previously established that the DA2 PC gives rise to a mixed DA2/AMP lineage (Boukhatmi et al., 2012). Similarly, one AM PC daughter cell is Zfh1 positive (Fig. 4D,D′). Triple staining for Org-1, Zfh1 and tupF4-GFP confirmed that asymmetric division of the AM PC gives rise to one daughter that remains Org-1 positive and a sibling that is Zfh1 positive (Fig. 4E-E″). Muscle PC into FC divisions do not occur in Regulator of cyclin A1 (Rca1) mutant embryos, and Rca1 mutations were obtained in an EMS screen utilizing org-1-RFP as a marker for abnormal development of AMs [and LO1/VT1/SBM muscles (Hollfelder et al., 2014)]. Triple staining of Rca1 embryos showed that, in the absence of division, the AM PC remains Org-1 positive and Zfh1 negative (Fig. 4F-F″), indicating that the AM fate is its default fate. Zfh1 staining does not persist in the AM sibling, unlike in the DA2 sibling (data not shown), and the ultimate fate of this cell remains unclear.

To follow the differentiation of the AM FC into muscle fiber, we performed time-lapse imaging of live embryos, as well as Org-1/tup-GFP staining (supplementary material Fig. S5). For this analysis, we used an AM-specific enhancer, tupAME-GFP (see below). Both approaches showed that AM FCs extend filopodial extensions along the dorsoventral axis and start fusing with myoblasts to form syncytial fibers that elongate until they connect dorsally to the heart and ventrally to the lateral epidermis (supplementary material Fig. S5 and Movie 1). In the abdominal segments A1-A7, each AM PC/FC gives rise to a mature AM. However, although one Org-1⁺/Tup⁺cell is specified in each thoracic segment (Fig. 4A), only two TARMs, namely TARM^T1 and TARM^T2, develop in the thorax (Fig. 1B, Fig. 3G), raising the question of the fate of the T3 progenitor. To address this question, we used org-1-lacZ in combination with the reporter line tupADME-lacZ, which allows TARMS to be distinguished (β-gal⁺, GFP⁺) from other tissues expressing only one reporter. These experiments showed that TARM^T1 and TARM^T2 grow and extend posteriorly between stage 13 and 14, whereas TARM^T3 stays globular and disappears, probably by apoptosis, at stage 14 (supplementary material Fig. S6).

In conclusion, lineage analyses and imaging of co-expressed Org-1/Tup during muscle development shows that AMs and TARMs constitute homologous muscle lineages in abdominal A1-A7 and T1-T2 segments, respectively, and that this lineage is abortive in T3.

Double staining of tup^ex4; org-1-lacZ embryos (de Navascues and Modolell, 2010) for β3-Tub and β-gal showed that AMs form in the absence of Tup activity but are severely malformed, whereas TARMs are essentially missing (Fig. 5A-B′). The patterns of expression and mutant phenotypes of tup and org-1 raised the question of whether tup was acting in parallel or downstream of org-1. We first stained stage 11-12 tup^ex4 embryos for Org-1 and observed that Org-1 expression in AM and TARM progenitors was unaffected (not shown). We then analyzed Tup expression in AM progenitors in org-1^OJ487 embryos using, as an internal reference, DA2/AMP expression. Specific loss of Tup expression in a significant fraction of AM progenitors indicated that Org1 regulates tup expression (Fig. 5C,D).

Two separate tup CRMs, namely IsletH/tupF4 and tupDME (dorsal muscle enhancer), drive expression in both dorsal muscles and AMs (see Fig. 6A). tupF4 activity is detected at late stage 11 in the DA2 and AM PCs, whereas tupDME starts to be detectable at stage 12 and remains active in the AMs and dorsal muscles up to stage 16 (Boukhatmi et al., 2012) (Fig. 5E,G). We observed complete loss of tupDME-lacZ expression in AM PCs and FCs in org-1 mutant embryos (Fig. 5F). Staining stage 16 embryos confirmed that, in the absence of org-1, both Tup (not shown) and tupDME-lacZ expression is completely lost in AMs, while remaining normal in body wall muscles (Fig. 5G,H).

To further assess whether Org-1 regulation of tup is AM specific, we expressed Org-1 in the entire mesoderm [2xPE>org-1 (Schaub et al., 2012)] and examined Tup expression using both antibodies and the AM reporter line, tupADME^ΔTup (see below and Fig. 6A). We observed a thickening of AMs and TARMs, with more Tup-positive nuclei per fiber than in wild type and, in some of segments, the formation of one extra muscle with an AM-like morphology (Fig. 5I,J). Although the thickening phenotype might reflect a differentiation phenotype, duplications of AMs suggested that pan-mesodermal expression of Org-1 is able to convert the AM FC sibling to an AM fate and that the ability of Org-1 to activate Tup is restricted to the AM and TARM lineages. To then determine the contribution of Tup activation to the org-1 gain-of-function phenotype, we expressed Tup in the entire mesoderm (2xPE>UAS-tup). In contrast to Org-1, Tup overexpression does not provoke AM duplications (not shown), reinforcing the conclusion that Tup cannot substitute for Org-1 function in engaging the AM differentiation program.

In conclusion, mutant analyses show that both org-1 and tup functions are required for AM and TARM development, with org-1 acting upstream of tup specifically in these muscles and being instructive for the AM fate.

The Org-1-dependent activity of tupDME in AMs prompted us to study this enhancer in more detail. Based on the position of a meso-ChIP CRM (Sandmann et al., 2007; Zinzen et al., 2009) and sequence conservation between different Drosophila species, we first trimmed tupDME to a 1.3 kb DNA fragment, which we call tupADME (alary and dorsal muscle enhancer) (Fig. 6A). Bioinformatics indicated the presence of three potential Org-1 binding sites (Muller and Herrmann, 1997; Schaub et al., 2012), which we designated Org-1 I, II and III, as well as one Tup binding site (Fig. 6; supplementary material Fig. S7). Mutation of the predicted Tup binding site (tupADME^ΔTup) resulted in a decrease in enhancer activity in dorsal muscles but not in AMs (Fig. 6B,C). Therefore, Tup is required for maintaining its own expression via direct positive autoregulation in dorsal body wall muscles but not in AMs.

ChIP analysis with chromatin prepared from stage 11-12 embryos (Schaub et al., 2012) indicated in vivo Org-1 binding to Org-1 sites I and II (Fig. 6D). This suggested that Org-1 I and II could mediate direct tup regulation by Org-1. To test this hypothesis, we made reporter constructs in which shorter versions of tupADME were fused to moeGFP (Fig. 6A). A subfragment containing Org-1 I and II, called tupAME (alary muscle enhancer), is active in AMs and TARMs but not in dorsal muscles (Fig. 6E,F, Fig. 7). Mutation of both Org-1 I and II completely abolished tupAME activity (tupAME^mut; Fig. 6G,H), demonstrating that Org-1 regulation of tup expression in AMs/TARMs is direct. Pan-mesodermal expression of Org-1 led to Tup and tupAME expression in two adjacent cells per segment (mean=1.91±0.44, n=90), instead of one in wild-type embryos (0.93±0.25, n=45), at stage 12, correlating with AM duplications and supporting our proposal of a conversion of the AM FC sibling to an AM fate (Fig. 6I,J; Fig. 5J). This confirmed that the ability of Org-1 to activate tup expression is specific to the AM PC or/and daughter cells. A shorter version of tupADME termed tupDME.s, which removes Org-1 I and II but preserves Org-1 III (Fig. 6A), loses expression in AMs (Fig. 7C), confirming that the Org-1 III site, which is not actively bound by Org-1 in vivo, is not functional.

The presence of two sites bound by Org-1 in vivo raised the question of whether these are redundant or, alternatively, could be specific for either AMs or TARMs. To address this question, we constructed two additional reporters termed tupAME.I and tupADME.II fused to moeGFP (Fig. 6A). tupAME.I is only active in AMs and TARMs (and in ectodermal cells connecting the frontal sac and pharynx, as with its parental construct), whereas tupADME.II is also active in the heart and dorsal muscles, consistent with the presence of the Tup autoregulatory site in this fragment. Both constructs are able to drive GFP expression in both AMs and TARMs, although much more weakly than tupAME, showing that both Org-1 I and II are functional and are both required for robust expression (Fig. 7D-F).

Together, these results allow us to draw a schematic model of tup regulation in AMs in comparison to dorsal body wall muscles, which highlights the key role of lineage-specific autoregulation during patterning of the Drosophila larval dorsal musculature (Fig. 7G).

The anatomical organization of the Drosophila larval organs is established during embryogenesis. We show here that abdominal AMs and their thoracic counterpart, the TARMs, establish physical connections between the exoskeleton and different internal organs, and that TARMs thus represent a novel type of muscle. The control of AM and TARM development by Org-1 and Tup – as orthologs of Tbx1 and Islet1, two key transcription factors of cardiopharyngeal mesoderm development in vertebrates – suggests that these transcription factors were part of an ancestral regulatory kernel that was redeployed several times during evolution to control different embryonic mesoderm derivatives.

Seven pairs of embryonic abdominal AMs connect to the heart and the aorta. Interestingly, pan-mesodermal expression of either Ubx or AbdA, two abdominal Hox proteins, resulted in the formation of three supernumerary pairs of AMs arranged in a segmental pattern in the three thoracic segments (LaBeau et al., 2009). This suggested the existence of thoracic PCs that fail to give rise to muscles in the absence of proper Hox input. Such a scenario accounted for the lack of DA3 muscle in the T1 segment (Enriquez et al., 2010). Alternatively, these thoracic PCs could give rise to previously unrecorded muscles. We discovered that both explanations apply. Thoracic Org-1⁺/Tup⁺ PCs give rise to a novel type of muscles, which we call TARMs. Only TARM^T1 and TARM^T2 form during normal development, whereas TARM^T3 differentiation is abortive. We also noted that an extra pair of muscles expressing both org and tup reporter genes in late embryos attaches to the proventriculus. The PCs at the origin of this extra pair of muscles, which we named TARM*, remain to be identified.

In summary, TARMs and AMs are generated by homologous lineages and form thin, elongated muscles of characteristic morphology and attachment sites that integrate segment-specific information.

Descriptions of AMs in adults of different insects led to proposals of their role in either maintaining the position of the heart, regulating the hemolymph flow and/or controlling heart beat. The requirement of transient contacts between some AMs and the distal tip cell of MTs revealed yet another role of AMs, namely for proper bending and positioning of MTs (Weavers and Skaer, 2013). We show here that, through linking the lateral epidermis to the dorsal vessel, AMs press the dorsal, main branch of the trachea towards the body wall. TARMs also establish intimate topological relations with cephalic branches of the trachea in their trajectory from thoracic lateral epidermis/exoskeleton to specific regions of the midgut. Rotation of the gut at the end of embryogenesis (Okumura et al., 2010) results in stretching and bilateral asymmetry of the left and right TARMs. This suggests that TARMs have elastic properties, as already indicated by their transient deformation upon interaction with the MT tip cell (Weavers and Skaer, 2013).

Although the function of AMs and TARMs during larval development and growth remains to be fully assessed, the morphological and elastic properties of AMs and TARMs suggest that they could be involved in controlling the position of the heart, trachea and gut during larval foraging movements. Whereas Org-1 and Tup are only co-expressed in the AM/TARM lineage, each is expressed in several other mesodermal or non-mesodermal tissues, precluding the use of mutants to selectively study AM and TARM function. Whether AMs and TARMs represent a new type of muscle with both spring-like and contractile properties is one of the questions to be addressed henceforth.

The morphological properties specific to each Drosophila body wall muscle are determined by the expression of specific combinations of iTFs in each FC and derived muscle (de Joussineau et al., 2012). Lineage specificity involves positive and negative cross-regulations between different iTFs, as well as iTF autoregulation. The maintenance of Org-1 and Tup expression in body wall muscles depends upon autoregulation (Boukhatmi et al., 2012; Schaub et al., 2012). We show here that tup autoregulation in dorsal muscles is direct, but does not operate in AMs; here, tup is directly regulated by Org-1. Org-1 directly regulates the expression of two other iTFs, namely slouch and ladybird, in other muscle lineages (Schaub et al., 2012). The AM/TARM Org-1>Tup hierarchy further underlines the intricate, combinatorial nature of transcriptional regulatory networks specifying Drosophila muscle identity. We have now identified an Org-1-dependent tup enhancer called tupAME, which is only active in AMs/TARMs and ectodermal cells connecting the frontal sac and pharynx. This CRM should provide a means to specifically target expression and modify AMs/TARMs in order to assess their function in larvae.

Nkx2.5, Tbx1 and Islet1 are major actors in the vertebrate genetic program controlling early heart and pharyngeal muscle development from common progenitors lying in the second heart field (SHF) in chordates (Cai et al., 2003; Prall et al., 2007; Sambasivan et al., 2009; Kelly, 2012; Watanabe et al., 2012). NK4/Nkx2.5 was recently shown to antagonize Tbx1 and repress EBF/COE function to promote cardiac versus pharyngeal muscle fate in the ascidian SHF, with Islet1 being expressed in both derivatives (Wang et al., 2013; Razy-Krajka et al., 2014). In the Drosophila embryo, the orthologs of Nkx2.5 and Islet1 are required for the formation of all (in the case of Tin) or some (in the case of Tup) of the dorsal mesodermal derivatives (heart, dorsal body wall muscles, visceral muscles, lymph gland) (Azpiazu and Frasch, 1993; Bodmer, 1993; Zaffran et al., 2006; Mann et al., 2009). Moreover, Tup represses the COE ortholog Collier (Col; Knot – FlyBase) in dorsal muscles (Boukhatmi et al., 2012), and Org-1/Tbx1 is expressed in some body wall muscles, the AMs and the visceral mesoderm, although not in myocardial progenitors. Together, these findings suggest that the Nkx2.5/Tin, Tbx1/Org-1, Islet1/Tup and COE/Col transcription factors are part of an ancestral regulatory kernel controlling diversification of heart and muscle lineages from a common progenitor pool. Future studies will be required to establish which ancestral interactions have been redirected to foster the emergence of insect AMs and TARMs.

To our knowledge, no striated muscle has been described so far to connect the (exo)skeleton to the gut, either in insects or in vertebrates. In mammals, lung and heart are separated from visceral organs by the diaphragm muscle. However, a muscular diaphragm is a defining characteristic of mammals that is not found in other vertebrates, and the ancestral origin of this recent innovation is currently unknown (Merrell and Kardon, 2013). Although highly speculative, it will be interesting to investigate whether the vertebrate muscle/septum and the insect AMs/TARMs could represent two specific adaptations of an ancestral demarcation between the circulatory systems, respiratory systems and visceral organs.

All Drosophila melanogaster stocks were grown on standard medium at 25°C. Strains used: tup^ex4 (de Navascues and Modolell, 2010), org-1^OJ487, org-1-HN39-lacZ, UAS-org-1, 2xPE-Gal4; how^24B-Gal4 (Schaub et al., 2012), IsletH-GFP (Thor and Thomas, 1997), tupF4-GFP (Tao et al., 2007), tupDME-lacZ (Boukhatmi et al., 2012), sr-Gal4 (obtained from T. Volk, Weizmann Institute of Science, Rehovot, Israel), CtB-Gal4 (Sudarsan et al., 2002), Rca1^S0844/Rca1^S2926 (I.R., unpublished; used in the heterozygous condition), HLH54Fb-RFP (Ismat et al., 2010), MHC-tauGFP (Chen and Olson, 2001), UAS-mcd8GFP and UAS-tup (Bloomington Stock Center). Mutants were balanced over marked chromosomes: FM7 ftz-lacZ; CyO wg-lacZ; TM3 twi-lacZ and homozygotes identified by the absence of β-gal expression. Pan-mesodermal expression of UAS-org-1 with 2xPE-GAL4;how^24B-GAL4 was carried out at 28°C.

For tup reporter lines, the genomic regions chr2L: 18897397..18900688 (tupADME) and chr2L: 18899851..18900688 (tupAME) according to FlyBase release R5.42 were amplified using yw genomic DNA as template. tupADME was cloned into the attB-inslacZ vector (Enriquez et al., 2010). tupAME was recombined into pDEST-MoeGFP. Positions of tupADME subfragments used for additional reporter lines are indicated in supplementary material Fig. S7. For creation of tupAME^mut-moeGFP, Org-1 binding sites were predicted using Target Explorer (Sosinsky et al., 2003) with a positional weight matrix generated by SELEX (Schaub et al., 2012), and mutated by site-directed mutagenesis as follows: tupAME TAACACAT → tupAME-orgImut TAAGCTTT; and tupAME GGGTGCCA → tupAME-orgIImut GGCTCGAG. Mutagenesis of the consensus Tup recognition sequence CTAATG (Portales-Casamar et al., 2010) to CTAGGT was achieved by PCR. The lacZ and moeGFP reporters were inserted at position 68A4 on the third chromosome by injection into nos-phiC31-NLS; attP2 embryos (Bischof et al., 2007; Markstein et al., 2008).

Immunostaining of whole-mount embryos was performed using standard techniques. The following primary antibodies were used: chicken anti-β-gal (1:2000; Abcam, ab9361); mouse anti-GFP (1:100; Roche, 11814460001); anti-β-gal (1:1000; Promega, Z3783, Lot 25580607); anti-TM1 (1:200; Babraham Institute, Cambridge, UK, MAC 141); mouse anti-Tup (1:100), anti-Prc (1:500), anti-Cut (1:200), anti-FasIII (1:20), anti-βPS (1:10), 22c10 (Futsch – FlyBase) (1:50) and 2A12 (Pavarotti – FlyBase) (1:5), all from Developmental Studies Hybridoma Bank; rabbit anti-GFP (1:500; Torrey Pines Biolabs, TP401, Lot 071519); anti-Nau (1:100; Bruce Paterson, Bethesda, USA); anti-Zfh1 (1:5000; Ruth Lehmann, New York, USA); anti-β3-Tub (1:5000; Renate Renkawitz-Pohl, Marburg, Germany); anti-Mesh (1:1000) (Izumi et al., 2012); and rat anti-Org-1 (1:100) (Schaub et al., 2012). Secondary antibodies were Alexa Fluor 488, 647 or 555 conjugated (1:300; Molecular Probes). Optimized confocal sections were acquired on a Leica SP5 or SPE microscope. Projections and three-dimensional reconstructions were created using ImageJ (NIH) and Volocity (PerkinElmer) software, respectively. Time-lapse imaging of embryos carrying GFP and RFP reporter genes was performed as previously described (Hollfelder et al., 2014).

ChIP experiments were performed using stage 11-15 2xPE; how^24B>>UAS-org-1 embryo collections and Org-1 antibodies as described (Schaub et al., 2012). The precipitated DNA was quantified in triplicate using FastStart Universal SYBR Green Master (Rox) (Roche Applied Science) on an Eco Real-Time PCR System (Illumina). Three independent precipitations per amplicon were analyzed.

We thank the Bloomington Stock Center and colleagues for Drosophila strains; Mikio Furuse, Alain Garces, Bruce Paterson and Renate Renkawitz-Pohl for antibodies; Michèle Crozatier and Alice Davy for helpful discussion and critical reading of the manuscript; Brice Ronsin at the Toulouse RIO Imaging platform; and Julien Favier for maintenance of fly stocks.