Yorkie and JNK revert syncytial muscles into myoblasts during Org-1–dependent lineage reprogramming

The somatic musculature of the fruit fly Drosophila melanogaster as well as that of vertebrates is composed of terminally differentiated striated muscle fibers that are built by long syncytial cells. They arise during embryonic development by fusion of mononucleate muscle progenitor cells and acquire individual muscle identities that define their particular features, e.g., their shape, innervation, and attachment sites (Dobi et al., 2015). During metamorphosis, the building of the adult musculature involves a radical reconstruction of the multinucleated, striated body wall musculature. The vast majority of larval body wall muscles become histolysed and disappear, and almost all adult muscles are then newly built from stem cell–like adult muscle precursor cells (Gunage et al., 2017). However, some specific larval muscles escape histolysis and are remodeled into adult muscles, as is the case for the formation of the indirect flight muscles (Dutta and VijayRaghavan, 2006).

Recently, we discovered a new mode of larval-to-adult muscle remodeling, which applies for the development of syncytial muscles associated with the adult heart, the so-called ventral heart–associated longitudinal muscles (ventral longitudinal musculature; VLM), from a different type of larval syncytial muscles, called alary muscles (AMs; Schaub et al., 2015). The formation of the VLMs resembles a process of direct lineage reprogramming, which includes dedifferentiation and fragmentation of syncytial AMs (Fig. 1, A and B) into mononucleate myoblasts, called AM-derived cells (AMDCs; Fig. 1, A and C). The AMDCs are reprogrammed into muscle progenitors that subsequently undergo fusion with myoblasts from other sources again and redifferentiate into the VLMs (Fig. 1, A and D; and Video 1). Important components of the genetic program that regulate the development of VLMs from AMs are the T-box factor Org-1 (Drosophila Tbx1) and its direct target, the LIM homeodomain factor Tup (Drosophila Islet1), which are instrumental in initiating the dedifferentiation process. The spatial restriction of the dedifferentiation process is mediated by the action of Hox genes, in particular Ubx, which is restricted to the anterior three pairs of AMs (LaBeau et al., 2009). Likewise, the temporal regulation is mediated by stage-specific hormonal inputs by the ligand-bound nuclear receptor of the steroid molting hormone ecdysone (Schaub et al., 2015). However, the events that exert dedifferentiation and fragmentation of AMs as well as redifferentiation of the AMDCs downstream of Org-1 and Tup are as yet unknown.

Here we report that the AM lineage-specific activity of Yorkie (Yki), the transcriptional effector of the Hippo pathway and coactivator of the transcription factor Scalloped (Sd), plays a key role in initiating AM dedifferentiation and fragmentation downstream of Org-1/Tup. We demonstrate that active Drosophila JNK (dJNK) signaling is indispensable for AM fragmentation by contributing to both the activation of Yki and the activation of Drosophila Jun (dJun), thus leading to the formation of the Yki/Sd and dJun/dFos (AP-1) effector complexes. The molecular programs that are initiated by the synergistic activities of these transcriptional activators induce AM dedifferentiation and lead to the activation of Myc and piwi, which are crucial for different aspects of AM lineage reprogramming. Together, our results show that derepression of Yki downstream of lineage-specific transcription factors, in concert with active dJNK signaling, mediates lineage plasticity and initiates a cell fate switch during a naturally occurring process of muscle lineage reprogramming.

In a recently performed AM-specific in vivo RNAi screening, we identified scalloped (sd) and yorkie (yki) as crucial genes for VLM formation. org-1-GAL4–mediated knockdown of sd or yki within the AMs by either cell type specifically induced RNAi (Fig. S1, B and C) or CRISPR (Fig. 1, E and F; and Table S1) abolishes VLM formation, and leads to the presence of org-1-RFP–positive muscles that may represent remnants of the larval AMs. Yes-associated protein (YAP), the mammalian homologue of Yki, can function as coactivator for several transcription factors including members of the TEA domain family (TEAD)/transcriptional enhancer family (TEF). In Drosophila, Yki forms a complex with the single TEAD/TEF family member Sd that can activate transcriptional targets after nuclear translocation (Goulev et al., 2008; Wu et al., 2008; Zhao et al., 2008). Visualization of a GFP-tagged version of the endogenous protein clearly demonstrates that Sd is located in the nuclei of the syncytial org-1-RFP–positive AM cells (Fig. 1, G and G′). Although we were not able to detect nuclear GFP-tagged Yki protein expressed under endogenous control (Yki::GFP) in intact AMs, distinct nuclear accumulation was apparent during fragmentation of the AMs into org-1-RFP–positive AMDCs (Fig. 1, H and H′). Live imaging confirmed that the phenotypes in yki knockdown backgrounds result from the inability of the AMs to dedifferentiate into AMDCs (Video 2). Thus, we propose that Yki and Sd are crucial for the initiation of AM transdifferentiation.

The evolutionarily conserved Hippo pathway has been identified as one of the key inputs regulating Yki/YAP activity, particularly in the control of organ size via calibrating cell proliferation and apoptosis. Its core is represented by a kinase cascade consisting of Hippo (Hpo; Harvey et al., 2003; Jia et al., 2003; Pantalacci et al., 2003; Udan et al., 2003) and Warts (Wts) that phosphorylates Yki, thereby inhibiting its nuclear translocation and its function as a transcriptional coactivator of Sd (Huang et al., 2005; Goulev et al., 2008; Wu et al., 2008; Zhao et al., 2008). To dissect a potential role of Hpo signaling during AM lineage reprogramming, we overexpressed various Yki and Hpo constructs (Fig. 2 A) with org-1-GAL4. Nonphosphorylatable Yki (YkiS^168A) did not cause any effect in this assay (Fig. 2, A and C). Strikingly, however, forced expression of constitutively active Hpo (UAS-Hpo^ΔC, an N-terminal fragment possessing the Ste20 kinase domain but not the regulatory domains) completely abrogates AM transdifferentiation (Fig. 2, B and D). Since this effect is not observed upon expression of the same deletion construct carrying an inactive Ste20 kinase domain (UAS-Hpo^ΔC.K71R; Fig. 2 E), of full-length Hpo (UAS-Hpo; Fig. 2 F), or of its regulatory C-terminal part (UAS-Hpo^ΔN; Fig. 2 G) in the AMs, this strongly suggests that activated Hpo signaling blocks the induction of lineage reprogramming in the AMs. To prove that Hpo pathway–mediated suppression of Yki function provokes the observed phenotypes in Hpo^ΔC overexpression backgrounds, we coexpressed it together with phosphorylation-resistant, constitutively active Yki (UAS-yki^S168A). This caused a significant rescue of VLM formation (P ≤ 0.0001; Fig. 2, B and H), thus strengthening our hypothesis that inactivation of the Hpo kinase cascade may lead to the activation of the Yki/Sd transcriptional activator complex that is critical for inducing AM reprogramming. Strikingly, by overexpressing Hpo-insensitive Yki^S168A in a genetic background in which we induced RNAi against org-1 or its direct target tup (Schaub et al., 2015), we observed that constitutively active Yki achieves a significant rescue of VLM formation (org-1 P ≤ 0.01, tup P ≤ 0.0001; Fig. 2, B and I–L). Taken together, these results are fully consistent with a mechanism in which Org-1–dependent inactivation of the Hippo pathway and concomitant activation of Yki/Sd target gene programs are needed to trigger AM dedifferentiation into AMDCs.

In a first effort to explore possible links between Org-1 and Hippo signaling during lineage reprogramming, we tested additional known inputs into this pathway. In recent work, it was proposed that atypical PKC (aPKC) signaling can regulate the nuclear outputs of the Hippo cascade either positively, namely by dislocating and down-regulating Hippo kinase activity (Grzeschik et al., 2010), or in the case of mouse intestinal cells, negatively, namely by directly phosphorylating YAP via the aPKC member PKCζ (Llado et al., 2015). To test if aPKC activity plays a role in Hpo regulation during AM transdifferentiation, we expressed a membrane-tethered, dominant-negative form of aPKC (UAS-aPKC^CAAX-DN) and a constitutively active aPKC deletion construct (UAS-aPKC^ΔN) via org-1-GAL4. Down-regulation of aPKC activity does not have any impact on VLM formation and thus does not appear to cause increased Hpo activity in this context (Fig. 3 B). By contrast, the presence of the constitutively active form of aPKC abrogates VLM formation and AM lineage reprogramming, which could be due to increased Hpo and/or decreased Yki activity (Fig. 3 C). In line with this possibility, the loss of VLM formation in the aPKC gain-of-function genetic background can be rescued significantly by the coexpression of phosphorylation-resistant Yki (P ≤ 0.0001; Fig. 3, D and G). These results support a model in which aPKC signaling is involved in the upstream regulation of Yki activity and needs to be inactive during AM transdifferentiation.

Since it has been shown that there is a functional link between the activation of JNK signaling and down-regulation of Hpo pathway signaling (Sun and Irvine, 2013), we next asked if Basket (Bsk, dJNK) activity is required for AM lineage reprogramming. To address this question, we overexpressed a dominant-negative form of dJNK (UAS-bsk^DN) in the AMs with org-1-GAL4, which resulted in the abolishment of VLM formation (Fig. 3 E). Interestingly, coexpression of Bsk^DN and phosphorylation-resistant Yki^S168A leads to a significant rescue of VLM differentiation (P ≤ 0.0001; Fig. 3, F and G), which indicates that active dJNK normally acts in promoting nuclear Yki, which possibly also in this context involves down-regulating the Hippo pathway kinases. However, the VLM morphology in the rescued phenotypes is severely disturbed, pointing toward additional inputs from dJNK signaling into AM lineage reprogramming. As a key component of the JNK pathway, dJNK phosphorylates the transcription factor Jun-related antigen (Jra, dJun) that, together with Kayak (Kay, dFos), subsequently forms a heterodimeric protein complex, AP-1, which represents the transcriptional effector of JNK signaling (Kockel et al., 2001). By characterizing the expression pattern of an in vivo AP-1 sensor construct (TRE-GFP) during metamorphosis, we demonstrate TRE-GFP expression in org-1-RFP-positive AMDCs (Fig. 3, H and H′), thus implying that AP-1 targets are bound and activated during AM dedifferentiation. Importantly, down-regulation of either dJun or dFos in the AMs by RNAi (Fig. S1, D and E) or by inducible CRISPR (Fig. 3, I and J; and Table S1) abolishes VLM formation and interferes with AM fragmentation, thus pointing to a crucial function of AP-1 during AM transdifferentiation.

Target genes of Yki have been mainly identified upon Yki hyperactivation or ectopic expression in imaginal discs. These experiments identified Myc as a transcriptional target of Yki, and it was shown that both work together in regulating proliferation and growth in epithelial imaginal cells (Neto-Silva et al., 2010; Ziosi et al., 2010). Down-regulation of Myc via inducible RNAi or CRISPR during AM transdifferentiation strongly interferes with VLM formation and AM reprogramming (Fig. S1 F, Fig. 4 B, and Table S1). Accordingly, live imaging demonstrated that upon RNAi-mediated Myc knockdown, AM dedifferentiation and fragmentation is severely impaired (Video 3), pointing toward a crucial role for Myc in the initiation of AM reprogramming. The analysis of the expression pattern of a GFP-tagged version of Myc under endogenous control (Myc::GFP) during AM lineage reprogramming revealed that it can be detected, like Yki::GFP, in the nuclei of the AMDCs shortly after the initiation of syncytial AM fragmentation (Fig. 4, C and C′). Notably, the induction of Myc::GFP expression can be suppressed by forced expression of constitutively active Hpo (UAS-Hpo^ΔC) with org-1-GAL4 in the AMs (Fig. 4, D and D′; and Table S1), implicating Myc as a Yki target during AM lineage reprogramming.

Another target gene that has been identified upon Yki overexpression in Drosophila imaginal discs is piwi (Zhang et al., 2015). Down-regulation of piwi during AM transdifferentiation via inducible RNAi provokes loss of VLM formation (Fig. 4 E and Table S1). Live imaging revealed that upon piwi knockdown, the anterior, org-1-RFP–positive AMs do dedifferentiate and fragment into AMDCs, but AM transdifferentiation is arrested at this stage (Video 4). An endogenously controlled, GFP-tagged version of Piwi (Piwi::GFP) is not yet detectable in syncytial AMs and AMDCs during AM fragmentation but is expressed at high cytoplasmic levels during reprogramming of the AMDCs into the VLM progenitor cells (Fig. 4, F–G′). Like Myc::GFP, piwi::GFP activation is suppressed by forced expression of constitutively active Hpo (UAS-Hpo^ΔC) with org-1-GAL4 in the AMs (Fig. 4, H and H′; and Table S1). These data suggest an important role of Piwi during the reprogramming of the AMDCs into the progenitor cells of the VLMs and identifies piwi as another Yki target during AM lineage reprogramming.

Taken together, these observations implicate Myc and piwi as important examples of targets that may be activated by the synergistic activities of Yki/Sd and AP-1, which in turn are required for mediating downstream effects of the Hippo and JNK cascades during this naturally occurring reprogramming process (Fig. 5).

Although syncytial muscles are often considered a paragon of terminally differentiated cells, there are rare examples where skeletal muscle cells are induced to dedifferentiate and fragment into muscle progenitor cells, either by forced treatment or during a natural process as for the anterior Drosophila AMs (Odelberg et al., 2000; McGann et al., 2001; Jung and Williams, 2011; Sandoval-Guzmán et al., 2014; Schaub et al., 2015; Frasch, 2016). Our results herein reveal that effectors of the JNK and Hpo signaling pathways drive the lineage-specific dedifferentiation and fragmentation of syncytial AM cells into mononucleate myoblasts. With regard to muscles, JNK function in Drosophila has been associated with myonuclear positioning and visceral muscle-intrinsic left-right asymmetry (Taniguchi et al., 2007; Schulman et al., 2014), whereas Yki activity was described to be required for muscle growth during late pupal stages (Kuleesha et al., 2016). Both signaling pathways have also been implicated in dedifferentiation processes: active JNK in germline cells (Herrera and Bach, 2018), and Hippo (specifically, inactive Wts) in retinal cells (Nicolay et al., 2010). We describe a novel, synergistic function of these two pathways in a specific type of muscle and connect them to lineage plasticity. Our data imply that the synergistic action of these two pathways in AM dedifferentiation is exerted predominantly in parallel at the level of their nuclear effectors, namely nuclear Yki/Sd, which requires inactivity of Hpo/Wts, and AP-1, which requires active JNK. Additionally, our data indicate that there is cross-regulation in which JNK also contributes to Yki activation, as has been described in studies on regeneration and size regulation in wing discs. As proposed for the wing disc, this input may involve inactivation of Wts upon JNK activation (Sun and Irvine, 2013; Willsey et al., 2016). The synergistic activities of the nuclear effectors of the Hippo and JNK cascades could operate directly at the level of joint target genes. Our genetic data and expression analysis implicate Myc and piwi as important examples of such targets, which in turn are required for mediating downstream effects of the Hippo and JNK cascades during this naturally occurring reprogramming process. In future studies, it should be examined whether these putative target genes contain combinatorial binding sites for Sd and AP-1.

Given that forced YAP expression can induce reprogramming of adult mammalian cells into cells with lineage progenitor characteristics (Panciera et al., 2016); that the consensus motif for AP-1 and TEAD transcription factors are both significantly enriched in YAP-binding regions in cancer cell lines (Zanconato et al., 2015); and that the induction of pluripotent stem cells requires the active JNK signaling pathway, c-Myc, and involves up-regulation of mili (piwil2, a mouse homologue of Drosophila piwi; Takahashi and Yamanaka, 2006; Mikkelsen et al., 2008; Neganova et al., 2016), we propose that the key components that drive AM muscle lineage reprogramming have broader, evolutionarily conserved roles in mediating lineage plasticity. Specifically with regard to muscle tissues, it is worth noting in this context that in activated mouse satellite cells, YAP expression surges, activated YAP protein promotes their proliferation and inhibits their differentiation, and conditional YAP knockout in these muscle stem cells curtails muscle regeneration (Judson et al., 2012; Sun et al., 2017). Likewise, in the mouse heart, conditional depletion of the activity of the Hippo pathway kinases by various means, or expression of activated YAP, promotes heart regeneration beyond the early postnatal stages it is normally limited to. These effects involve dedifferentiation, reversal to a fetal gene expression program, and proliferation of cardiomyocytes (Liu and Martin, 2019). Although the role of Hippo inactivation and Yki/YAP activation in tissue growth has been amply documented, these and other findings reinforce the notion that, depending on the context, this pathway can also exert cellular dedifferentiation. Of note, in the tissue context examined herein, the dedifferentiated AMDCs do not proliferate; instead, upon being reprogrammed into founder cells of the VLMs, they directly undergo myoblast fusion with fusion-competent myoblasts that have multiplied and migrated from lateral sources (Schaub et al., 2015). Nevertheless, the intriguing parallels in the roles of inactive Hippo/active Yki/YAP in promoting Drosophila AM and mouse cardiomyocyte dedifferentiation, as well as in maintaining the dedifferentiated state of mouse muscle stem cells, warrant the search for additional similarities, including the potential involvement of regulatory inputs and outputs described in our current study, in muscle and heart regeneration in the mouse and other vertebrate models.

Like Hippo inactivation, Myc function is mainly known for promoting proliferation and growth. However, besides these roles it has been connected to the induction of proapoptotic genes in a cell-autonomous as well as a non–cell-autonomous manner (McMahon, 2014). Conversely, it can also exhibit antiapoptotic and proautophagic effects in cancer cells (Conacci-Sorrell et al., 2014). We infer that Myc activation during AM lineage reprogramming is instrumental in initiating the mechanistic apparatus that executes the dedifferentiation and fragmentation process. This may include the induction of the nonapoptotic activity of caspases as well as of autophagic genes involved in the disintegration of AM muscle components and syncytial fragmentation. We show that cytoplasmic Piwi function is crucially required in the generated AMDCs to proceed with transdifferentiation. Cytoplasmic expression of PIWI proteins is associated with lineage-restricted somatic stem cells from Hydra to mammals, and there seems to be a strong correlation between PIWI expression and plasticity (van Wolfswinkel, 2014). Accordingly, we envision that Piwi function in the AMDCs may be involved in the suppression of AM lineage-specific programs. This could provide increased plasticity as a prerequisite for the reprogramming of AMDCs into VLM progenitor cells. Rather than being involved in the suppression of transposable elements, it is more likely that Piwi functions through the degradation of cellular mRNAs during this process, which is also compatible with its observed cytoplasmic localization in the VLM progenitors. Precedence for this type of function exists, for example, in the Piwi-mediated degradation of a large set of uniformly distributed maternal transcripts around the onset of zygotic transcription in the early Drosophila embryo; the degradation of mRNAs (and long noncoding RNAs) during consecutive steps of spermatogenesis in the mouse via different PIWI family members so that development can transition to the next steps of meiosis and spermiogenesis; and the degradation of cellular mRNAs via Piwi in somatic niche cells to advance their development during Drosophila oogenesis (Rouget et al., 2010; Rojas-Ríos and Simonelig, 2018). Intriguingly, one important target of Piwi during this latter process is dFos mRNA, which needs to be destabilized for enabling the somatic niche cells to normally differentiate into follicle cells (Klein et al., 2016). It is thus conceivable that in AMDCs, Piwi exerts a negative feedback and destabilizes dFos, among other AM or dedifferentiation-specific mRNAs, to allow the transition toward the VLM progenitor fate.

Taking these observations together, we suggest a mechanism in which Org-1/Tup–driven lineage-restricted activation of Yki as well as JNK in the AMs triggers a mechanism that executes cellular reprogramming and whose key components have evolutionarily conserved roles (Fig. 5). Further dissection of this transdifferentiation process occurring naturally in vivo will uncover additional components linking these regulatory pathways and mediating muscle cell plasticity, which may also shed more light on general mechanisms of lineage plasticity and cellular reprogramming.

Standard Drosophila genetics was performed at 25°C. The following stocks were used: TRE-GFP (Chatterjee and Bohmann, 2012), UAS-hpo, UAS-hpo.delta.N, UAS-hpo.delta.C, UAS-hpo.deltaC.K71R (Jia et al., 2003), UAS-aPKC.CAAX.DN (Sotillos et al., 2004), hand-GFP (Sellin et al., 2006), org-1-GAL4 (line S18), org-1-RFP (Schaub et al., 2015). UAS-dsRNA-yki (34067), UAS-dsRNA-sd (55404), UAS-dsRNA-Jra (31595), UAS-dsRNA-kay (33379), UAS-dsRNA-Myc (51454), UAS-dsRNA-piwi-1 (33724), UAS-dsRNA-piwi-2 (34866), UAS-dsRNA-org-1 (62953), UAS-dsRNA-tup (51763), UAS-yki.S168A (28816), UAS-bsk.DN (6409), UAS-aPKC.CA (51673), UAS-Cas9.P2 (58986), sd::GFP (50827), and Myc::GFP (81274) were obtained from the Bloomington Drosophila Stock Center. yki::GFP (318237) and piwi::GFP (313319) were obtained from the Vienna Drosophila Resource Center.

We used the system described in Port and Bullock (2016) to create UAS inducible and t::gRNA array expressing transgenes. We amplified PCR fragments with the respective primers containing the gRNA sequences (Table 1) and 1 ng pCFD6 (Addgene, 73915) as template using Q5 Hot Start High-Fidelity 2X Master Mix (New England Biolabs, M0494L). The PCR fragments were gel purified using the QIAquick Gel Extraction Kit (Qiagen, 28704), and the respective constructs were assembled with BbsI-HF (New England Biolabs, 3539S) digested pCFD6 backbone using the NEBuilder HiFi DNA Assembly Cloning Kit (New England Biolabs, R5520S), following the instructions provided by the manufacturers.

Sequences of gRNAs (Table 1) and primers (Table 2) used for the assembly of the respective pCFD6 construct gRNAs were obtained from https://www.flyrnai.org/crispr/. Transgenes of the respective pCFD6 constructs were generated by a commercial embryo injection service (BestGene, Chino Hills, CA) using the phiC31/attP/attB system (Bischof et al., 2007) at the attP40 (UAS-t::gRNA-yki^4x, UAS-t::gRNA-Jra^4x), attP2 (UAS-t::gRNA-kay^4x, UAS-t::gRNA-Myc^4x), and VK00027 (UAS-t::gRNA-sd^4x) landing sites. Potential transformants were selected by mini-white expression.

Dissections of pupal stages until P4 (for pupal staging, see Bainbridge and Bownes [1981]) were performed after prefixation with 3.7% formaldehyde for 30 min with microsurgery scissors in toto. To dissect pharate adult stages, the pupal cases were removed before prefixation. For antibody stainings, the dissected animals were fixed in 3.7% formaldehyde for 10 min, washed three times for 30 min in PBT (PBS + 0.1% Tween 20), blocked in 10% BSA (Serva, 11930.04) for 1 h, and incubated with the primary antibodies, goat polyclonal anti-GFP (GeneTex, GTX26673, 1:1,000) and rabbit polyclonal anti-RFP (Rockland, 600-401-379S, 1:500) for 2 d at 4°C. To visualize filamentous actin, Atto 647N–conjugated phalloidin (1:1,000, Sigma-Aldrich, 65906) was added to the antibody solution. After incubation with the primary antibodies, the dissections were washed three times for 30 min in PBT and incubated with the respective secondary antibodies, Alexa Fluor 488–conjugated donkey anti-goat (Abcam, ab150129) and Alexa Fluor 555–conjugated donkey anti-rabbit (Abcam, ab150074), in 1:200 dilution overnight at 4°C. Finally, the stained animals were washed three times for 30 min in PBT and embedded into Vectashield containing DAPI (Vectorlabs, H-1200).

Confocal Z-stacks of fixed specimens were acquired with a Leica SP5 II (20×/0.7 HC PL APO Glycerol, 63×/1.3 HC PL APO Glycerol). Maximum projections of the Z-stacks were performed with Leica Application Suite X (Leica Microsystems).

Stage P3 pupae of the respective genotypes were aligned on a strip of double-faced adhesive tape connected to a slide and covered with a drop of Halocarbon oil 700 (Halocarbon Products Corp., Sigma-Aldrich, H8898) and a coverslip. Time-lapse series were acquired on a Leica SP5 II confocal system using a HC PL APO 10×/0.4 Air. Acquisition was done over a time course of 15–20 h with the following settings: pinhole 1.4 AU; 1.5× optical zoom, scan speed 200 Hz, line averaging 3, and Z-stack ∼20 sections with a step size of 3 µm and time intervals of 10 min per stack. Videos were generated using Leica Application Suite X and further processed with Photoshop CS6 (Adobe Systems).

For quantification of the VLM phenotype frequency, living adult pharate stages of the respective genotypes were aligned on a microscope slide. The org-1-HN18-RFP lineage marker in the living animals was visualized with the aid of a fluorescence-equipped microscope (Nikon, Eclipse 80i), and phenotypic classification was performed by morphological analysis of the RFP-positive cells. Pharate adult stages with differentiated VLM were defined by the presence of RFP-positive elongated muscle fibers that were ventrally attached to the adult heart and resembled the typical elongated VLM morphology. Pharate adult stages with no differentiated VLM were defined by the presence of either RFP-positive AMs (not being dedifferentiated) or spherical RFP-positive (muscle) cells. These phenotypic classes were scored relative to each other, and frequencies were calculated (Table S1). Statistical analysis of the samples was performed with the χ² test function of Excel (Microsoft). Statistical significance was defined by P ≤ 0.05.

Fig. S1 shows that induction of RNAi against the transcriptional effectors of Hpo and JNK, as well as the Yki target Myc, abolishes VLM differentiation in pharate adult stages. Table S1 shows the respective observed phenotypic frequencies of the genetic experiments. Video 1 shows the alary muscle lineage reprogramming process in a wild-type pupa. Video 2 demonstrates inhibition of alary muscle fragmentation in a yki compromised pupa. Video 3 shows inhibition of alary muscle fragmentation in a Myc compromised pupa. Video 4 demonstrates that loss of piwi function provokes arrest of transdifferentiation after alary muscle fragmentation..

We are grateful to Johannes März for helping us with dissections. We thank the Drosophila community for supplying us with fly stocks, in particular Dirk Bohmann (University of Rochester Medical Center, Rochester, NY), Sonsoles Campuzano (CBM Severo Ochoa, Madrid, Spain), Fillip Port (DKFZ, Heidelberg, Germany) and Jin Jiang (UT Southwestern Medical Center, Dallas, TX), as well as the Vienna Drosophila Resource Center and the Bloomington Drosophila Stock Center.

This work was funded by grants from the Deutsche Forschungsgemeinschaft to C. Schaub (SCHA 2091/1-1) and M. Frasch (FR 696/6-1) and was supported by the Optical Imaging Centre Erlangen.

The authors declare no competing financial interests.

Author contributions: Conceptualization: C. Schaub and M. Frasch; Investigation: C. Schaub and M. Rose; Writing – Original Draft: C. Schaub; Writing – Review & Editing: C. Schaub and M. Frasch; Funding Acquisition: C. Schaub and M. Frasch; Supervision: C. Schaub.