Muscle-derived Myoglianin regulates Drosophila melanogaster imaginal disc growth

Organ growth and size are finely tuned by intrinsic and extrinsic signaling molecules. In Drosophila, the BMP family member Dpp is produced in a limited set of imaginal disc cells and functions as a classic morphogen to regulate pattern and growth by diffusing throughout imaginal discs. However, the role of TGFβ/Activin-like ligands in disc growth control remains ill-defined. Here we demonstrate that Myoglianin (Myo), an Activin family member, and a close homolog of mammalian Myostatin, is a muscle-derived extrinsic factor that uses canonical dSmad2 mediated signaling to regulate wing size. We propose that Myo is a myokine that helps mediate an allometric relationship between muscles and their associated appendages.


INTRODUCTION
How organ growth is regulated to achieve proper scaling with final body size has been a long standing question in developmental biology. Organ growth is a dynamic process that is finely tuned by highly conserved signaling molecules that are produced and act within the tissue of origin (intrinsic) or on a distant tissue (extrinsic) (Boulan et al., 2015;Hariharan, 2015;Johnston and Gallant, 2002;Mirth and Shingleton, 2012). These growth signals regulate final organ size through three distinct processes: cell division (proliferation), mass accumulation (cell growth) and cell death (apoptosis). The Drosophila wing imaginal disc, the larval precursor to adult wings and thorax, has been widely used as a model to study the molecular mechanisms that control organ growth and size control (Martin et al., 2009). Genetic analysis has identified numerous wing disc intrinsic factors, such as Wnt, Hedgehog, and Bone Morphogenetic Protein (BMP) homologs, that are required for proper patterning and proliferation (Hariharan, 2015). In contrast, examples of extrinsic factors that regulate imaginal disc growth are limited. The Insulin-like signaling pathway is one clear exception. Insulin-like peptides are produced in the brain and fatbody and are well known to promote both growth and proliferation in many tissues including imaginal disc cells (Boulan et al., 2015;Droujinine and Perrimon, 2013;Grewal, 2012;Nijhout, 2003;Shingleton et al., 2007;Sopko and Perrimon, 2013;Teleman, 2009). Insulin-like factors act in a physiological capacity to coordinate overall organismal growth with nutrition. Identification of other extrinsic growth signals could provide new insights into mechanisms governing coordinated organ scaling.
The largest and most diverse group of signaling molecules is the Transforming Growth Factor β (TGF-β) superfamily of growth and differentiation factors. The superfamily is subdivided into the BMP and the Activin/TGF--like subfamilies based on sequence and signaling relationships. In higher vertebrates, these growth factors play a myriad of roles during development from cell fate specification and stem cell maintenance to control of cellular growth and proliferation (Morikawa et al., 2016). Both branches of TGF-β are conserved in Drosophila, where limited functional redundancy and a powerful genetic toolkit provide an attractive system for study (Peterson and O'Connor, 2014;Upadhyay et al., 2017). The Drosophila BMP ligand Decapentaplegic (Dpp) functions as a classic intrinsic morphogen, since it is secreted from a limited number of imaginal disc source cells and then diffuses throughout the tissue, to regulate proliferation and patterning based on its concentration gradient (Campbell and Tomlinson, 1999;Lecuit et al., 1996;Restrepo et al., 2014). Conversely, the Activin-like branch also plays a significant role in disc growth, however it is much less characterized (Brummel et al., 1999;Hevia and de Celis, 2013;Hevia et al., 2017).
In Drosophila Activin-like signaling is initiated by the three ligands, Myoglianin (Myo), Activinβ (Actβ), and Dawdle (Daw). These ligands bind to and activate a hetero-tetrameric receptor complex on the cell membrane consisting of two type-I receptor subunits encoded by the baboon (babo) locus and two type-II receptor subunits encoded by either the punt or wishful-thinking (wit) genes or perhaps a combination of the two (Upadhyay et al., 2017). Activated Babo phosphorylates cytosolic dSmad2 which then accumulates in the nucleus and, along with the co-Smad Medea, regulates expression of target genes (Brummel et al., 1999;Hevia et al., 2017). Notably, there are three splice isoforms of babo with different ligand binding domains, which presumably provide ligand binding specificity (Jensen et al., 2009). Babo null mutants die at the larval-pupal transition and have significantly smaller wing imaginal discs, and escapers produced by hypomorphic alleles also have smaller adult wings (Brummel et al., 1999).
Conversely, expression of constitutively activated Babo in wing imaginal discs leads to larger adult wings that contain more cells, suggesting that Activin-like signaling has a role in regulating proliferation (Brummel et al., 1999;Hevia and de Celis, 2013). Additionally, Babo can signal non-canonically through Mad (the BMP Smad) to stimulate wing disc overgrowth, but only in the background of a protein null allele (Peterson and O'Connor, 2013).
For both the dSmad2 canonical and the Mad non-canonical signaling pathways, it remains unclear which Babo isoform and which Activin-like ligands, or combination thereof, regulates disc growth, nor is it clear which tissue(s) provide the ligands that promote disc growth. Expression analysis by in-situ hybridization or RNAseq data suggests that all Activin-like ligands are expressed in the imaginal wing disc, albeit at low levels, perhaps indicating collective contribution to stimulating tissue growth (Graveley et al., 2011;Hevia and de Celis, 2013). However it is important to recognize that mRNA expression does not necessarily correlate with secretion of functional ligand, especially since TGF-−type ligands undergo a series of post-translational modifications and ligand processing to generate active molecules, and these processes may vary within different tissues (Akiyama et al., 2012).
Here we report our analysis of Activin-like ligands and Baboon isoforms for their roles in regulating imaginal disc growth. Through a series of genetic experiments, we demonstrate that among the three Drosophila Activin-like ligands, only Myo is required for proper growth of imaginal discs. We also identify the receptors Babo-A and Punt along with the co-factor Plum as being necessary for Myo regulation of wing imaginal disc growth via phosphorylation of dSmad2. Surprisingly, we identify the tissue source of Myo that controls disc growth to be the larval muscle.

Myoglianin is required for proper disc growth
All Drosophila Activin-like ligands (Myo, Act, and Daw) have tissue specific function yet signal via the sole Type-I receptor Babo. Hence, the Babo mutant phenotype is likely a composite of the three individual ligand mutant phenotypes. The imaginal discs of Babo mutants are approximately 30% smaller than wild-type but do not exhibit any obvious defects in tissue patterning (Brummel et al., 1999).
We set out to identify which Activin-like ligand(s) are required for proper imaginal disc growth. Null myo, actβ, and daw mutants are larval or pharate lethal (Awasaki et al., 2011;Ghosh and O'Connor, 2014;Zhu et al., 2008). Therefore, we analyzed the size of imaginal wing discs from late 3rd instar larvae of null mutant animals. For this initial analysis, we used a simple yeast/sugar diet (5Y5S), because daw mutants are sensitive to standard acidic cornmeal food (Ghosh and O'Connor, 2014). Interestingly, only 6 myo mutants yield small discs, phenocopying babo mutants. Myo wing discs are approximately 40% smaller than yw controls, but retain their normal shape and pattern (Fig 1 B,E). Actβ mutant discs had no significant size defects, however daw mutant discs were slightly smaller, which may be attributable to the reduction of insulin secretion in the daw mutants (Fig 1 C-E) (Ghosh and O'Connor, 2014). Previous studies suggested that all Activin-like ligands are expressed in the wing disc (Hevia and de Celis, 2013).
Despite this observation, our result clearly suggests that only myo is functionally required for growth of the discs, and that other ligands may have other functions apart from disc growth. We next asked whether other imaginal tissues such as the haltere and leg discs are also affected by the loss of myo. Indeed, both these imaginal discs are also approximately 40-50% smaller in myo mutants and maintain their proportional size (Fig 1F-I). This suggests that Myo signaling regulates tissue growth in all three of these imaginal discs to similar extents.
To determine why myo mutant discs are small, we measured cell size, cell growth, proliferation, and apoptosis. First, to assess cell size, we used phalloidin staining to mark the apical actin belt, and measured the apical surface area of disc epithelial cells and found that wing disc cells of myo mutants are 22% smaller than controls (Fig J-L). Next, we examined proliferation by staining for phospho-Histone 3(ser10) to mark mitotic cells, and found no significant difference between myo discs and controls at mid L3 stage (Fig 1, M -O). Third, using the apoptotic marker cleaved-caspase 3, we found no increase in apoptosis in myo mutants at the early L3 stage (Fig S1 A-B).
Another possible explanation for why myo mutant discs do not grow properly is TOR pathway regulation. To address this, we stained growing early L3 imaginal discs for phospho-S6, which is a target of pS6k and a readout of TOR pathway activation (Romero-Pozuelo et al., 2017). We find no decrease in pS6 staining in myo mutant discs, indicating that TOR signaling is not perturbed in these cells . To summarize, myo mutant discs are smaller in part due to smaller cells. However, the 22% decrease in cell size is insufficient to completely account for the 40-50% decrease in total tissue size. Because our proliferation and apoptosis assays are snapshot experiments of fixed tissues, we conclude that the portion of final disc size that cannot be accounted for by smaller cell size represents either a small proliferation reduction or a slight increase in apoptosis rate that cannot be detected at individual time points but over the course of the entire developmental period leads to an additional ~20% decrease in disc size.
Myo is homologous to vertebrate Mstn, which negatively regulates skeletal muscle size by inhibiting proliferation of muscle stem cells (Morikawa et al., 2016) and regulation of protein homeostasis (Lee et al., 2011;Rodriguez et al., 2014;Trendelenburg et al., 2009). Recent studies in Drosophila using RNAi tissue specific knockdown suggested that Myo also functions as a negative regulator of muscle size similarly to vertebrate Mstn (Augustin et al., 2017). We re-examined this issue using myo null alleles.
Surprisingly, we find that larval muscles #6 and #7 (Segment A2) at the late L3 stage contain the same number of nuclei, indicating no defects in myoblast fusion during embryogenesis, yet they are 13% smaller than heterozygous controls (Fig1 P-S). We conclude that with respect to muscle size control, myo is not functionally conserved with vertebrate Mstn.

Myo signals canonically via dSmad2
Canonical TGF-/Activin pathway activation in Drosophila is mediated by C-terminal phosphorylation of dSmad2 by the Type-I receptor Baboon (Brummel et al., 1999;Peterson and O'Connor, 2014;Upadhyay et al., 2017). To determine if Myo acts through the canonical signaling pathway, we examined phosphorylated dSmad2 (p-dSmad2) levels in wing disc extracts of various mutants. Wing disc extracts of late L3 larvae from yw, myo heterozygotes and null animals were Western blotted and probed with anti phospho-dSmad2 (p-dSmad2) antibody. We observe a dose dependent reduction in p-dSmad2 levels upon loss of myo, i.e. in heterozygous myo-/+ the p-dSmad2 signal is reduced compared to wild-type, whereas in homozygous null mutants there is no signal under our detection conditions. These findings indicate that Myo provides the majority of canonical signal to the wing disc. (Fig1 T-U).

Myo signaling in imaginal discs requires a specific isoform of Baboon
Baboon is the sole Activin branch Type I receptor in Drosophila, but encodes three splice isoforms (Awasaki et al., 2011;Jensen et al., 2009). The three splice isoforms are a result of mutually exclusive alternative splicing of the fourth exon (4a, 4b, 4c) giving rise to three receptors (Babo-A, Babo-B, and Babo-C) (Brummel et al., 1999;Jensen et al., 2009) that differ only in their extracellular, ligand binding domains. It has been suggested that different tissues express different isoforms, perhaps in order to tune tissue-specific responses to various ligands (Awasaki et al., 2011;Jensen et al., 2009). Ectopic Myo expression in the larval brain is lethal, however viability can be partially restored by simultaneously knocking down Babo-A but not Babo-B or Babo-C (Awasaki et al., 2011). These findings suggest that Myo signals specifically via Babo-A. Since only myo is required for proper growth of imaginal discs (Fig1 B), we reasoned that Babo-A should be the major Babo isoform expressed in the discs, a conjecture confirmed by qPCR analysis of dissected late L3 wing discs ( Fig.2-A). We verified that our primers for Babo-B and Babo-C are functional by testing other tissues such as carcass, brain, and fat body where we find different profiles for each isoform. (Fig. S2).
To confirm that the Babo-A isoform is indeed required for wing growth, we employed the GAL4/UAS system to conduct tissue and isoform-specific RNAi knockdown experiments. Using esg (entire disc), ci (anterior), and hh (posterior) GAL4 lines, we knocked down specific Babo isoforms and measured overall tissue size and cell size/density. In support of the qPCR data, we find that knocking down Babo-A alone with esg-GAL4 results in 40% smaller adult wings, whereas knocking down Babo-B or Babo-C resulted in no size difference ( Fig.2 B-E, J). Ci-GAL4 driving babo-a RNAi reduces the anterior compartment by 50% (Fig.2 F -I, L). In other experiments, we verified that the babo-b and c isoform specific RNAis are functional and produce phenotypes, thereby indicating that lack of wing phenotype by knockdown with these isoforms is not due to poor efficiency of 9 these RNAis.
We next assessed cell size in these wings by measuring trichome density in adult wings. RNAi for babo-a using esg-GAL4 or ci-GAL4 results in a fractional decrease (-6%) or increase (12%) in cell density respectively (Fig.2 K, M) indicating that the esg driver results in larger cells whereas the ci driver results in smaller cells. Assuming no difference in the number of wing disc precursors after embryogenesis, we infer that the esg driver leads to smaller wing size (

Babo-A, Punt, and Plum mediate non-canonical signaling in wing imaginal discs
In addition to canonical signaling, Babo can also signal non-canonically by cross-talk with downstream components of the BMP/Dpp pathway (Peterson and O'Connor, 2013). Non-canonical Babo activity is only detectable in a protein null mutant of dSmad2 or a strong RNAi knockdown of dSmad2 (Peterson and O'Connor, 2013;Peterson et al., 2012;Sander et al., 2010). In the absence of its preferred target dSmad2, Babo promiscuously activates Mad, the BMP ortholog of dSmad2, which then represses brk expression throughout the disc leading to overgrowth of the tissue along the anterior-posterior axis (Peterson and O'Connor, 2013). We hypothesized that the non-canonical signaling is mediated by Babo-A and not via other isoforms ectopically signaling in the absence of dSmad2. Babo activation can be visualized in dSmad2 loss-of-function clones using the brk (B14-LacZ) reporter (Müller et al., 2003;Peterson and O'Connor, 2013). We used this assay to generate dSmad2 RNAi clones and confirmed the ectopic brk-LacZ silencing (Fig. 3A). Next, we performed genetic epistasis analysis by simultaneously knocking down dSmad2 along with each of the babo isoforms in clones and then assayed for brk-LacZ rescue. Knocking down dSmad2 and babo-a rescued brk reporter expression ( Fig. 3 A  TGFβ signaling requires ligands binding to hetero-tetrameric type-I and type-II receptor complexes. There are two type-II receptors in the Drosophila genome: punt and wishful thinking (wit), both of which can signal with Type-I receptors from either branch of the TGFβ superfamily (Upadhyay et al., 2017). Punt is more commonly utilized during development, and facilitates Dpp signaling in the wing imaginal disc in combination with the BMP type I receptors Tkv and Sax (Brummel et al., 1994;Fuentes-Medel et al., 2012;Letsou et al., 1995;Parker et al., 2006). In cell culture models however, Punt can also signal in combination with Babo and Activin-like ligands (Brummel et al., 1999;Jensen et al., 2009;Letsou et al., 1995). Conversely, Wit functions primarily at the neuromuscular junction with the BMP ligand Gbb (Marqués et al., 2002) although it has been shown to bind Myo in combination with Babo in mammalian cell culture (Lee-Hoeflich et al., 2005). Since both punt and wit are expressed in the wing imaginal disc, it is unclear whether one or both type II receptors are required for Myo-Babo signaling (Childs et al., 1993;Marqués et al., 2002). To address this issue we utilized the brk silencing assay in RNAi clones as described above. We find that while RNAi for punt suppress the dSmad2 induced brk silencing (Fig. 3 E vs G), RNAi for wit does not suppress brk silencing in the same context ( Fig. 3 E vs F). Taken together, these experiments demonstrate that Babo-A requires Punt, but not Wit, for proper control of brk expression in this non-canonical wing disc signaling assay and leads us to conclude that Myo and Babo-A signal with Punt to control normal wing disc growth.
In addition to Type-I and Type-II receptors, various additional co-receptors have been found to interact with TGF-β ligands and receptors to facilitate signaling (Heldin and Moustakas, 2016).
Drosophila plum encodes an immunoglobulin superfamily protein, and is required for Myo stimulated pruning of mushroom body neurons (Yu et al., 2013). Therefore, we examined whether plum also facilitates Babo activity in the context of wing imaginal disc growth. Indeed, we find that plum-RNAi in the wing disc rescues the dSmad2 mutant overgrowth phenotype, confirming that Plum is required for Myo/Babo signaling in the wing disc similar to its requirement in mushroom body neurons (Fig. S4). This observation together with the previous report (Yu et al., 2013) suggest that Plum may be widely utilized as a co-receptor for Myo/Babo signaling in Drosophila.

Muscle-derived myoglianin signals to the wing disc
Growth factors such as Wg and Dpp are tissue intrinsic ligands, produced by and acting on the wing disc cells. We sought to determine the source of Myo that signals to the imaginal discs. The name myoglianin refers to the two tissues (muscles and glia) in which myo expression was initially observed (Lo and Frasch, 1999). More recent RNA in situ hybridization results suggest that myo is expressed in the wing imaginal disc at low levels (Hevia and de Celis, 2013). To precisely determine the source of Myo that acts on the wing disc, we first examined the expression pattern of myo in late L3 larvae using myo-GAL4 reporter to drive UAS-CD8-GFP (Awasaki et al., 2011). We detect GFP expression in both the glial cells of the brain and the ventral ganglia, as well as the larval body wall muscles (Fig. 4 A-B), but we do not observe any GFP signal in any imaginal discs (Fig. 4 A). Thus, we conclude that either myo is not expressed in the wing disc or that our myo-GAL4 construct does not include all of the endogenous enhancers.
To determine the functional source of Myo that controls imaginal disc growth, we conducted loss-of-function studies in the dSmad2 null mutant background, and looked for rescue of the wing disc overgrowth phenotype. First, using null mutant alleles, we confirmed that loss of myo in dSmad2 mutant background suppress the dSmad2 overgrowth, resulting in small discs similar in size to myo mutants ( Fig.   4 C-E). Next, we knocked down myo using tissue specific GAL4 drivers. To confirm efficacy of the RNAi constructs, we first used the da-GAL4 driver to ubiquitously knock down myo in the dSmad2 background and confirmed that it suppresses wing disc overgrowth (Fig. 4 F vs G). Next, knocking down myo using either esg-GAL4 (wing disc) or repo-GAL4 (glia) failed to suppress the dSmad2 wing disc overgrowth ( Fig. 4 F vs H-I). This result suggests that the wing discs are receiving a strong Myo signal even when knocked down in the imaginal discs or glia. In contrast, RNAi for myo in body wall muscles, using either mhc-GAL4 or mef2-GAL4 drivers, completely suppresses the dSmad2 disc overgrowth phenotype phenocopying the dSmad2;myo double mutant . Therefore, we conclude that muscles are the major source of Myo that stimulates imaginal disc growth.

Overexpression of Myo in muscle or fat body rescues myo mutant disc growth
While the suppression of disc overgrowth in a dSmad2 mutant background strongly implicates muscle as the source of ligand, we also examined if Myo expression in muscle, or other non-disc tissues, could rescue the size of myo mutant discs. First ubiquitous overexpression of myo in wild type larva results in wings that larger than normal (Fig. S5). Next, we examined whether overexpression of myo in either glia, muscles, or fat body is able to rescue size of myo mutant discs. Overexpression in glia (repo- depending on the source. Collectively, these results indicate that muscle-derived Myo acts as an endocrine-like ligand via the hemolymph to activate a specific receptor complex on disc cells, consisting of Babo-A, Punt and Plum, that signals to dSmad2 to promote imaginal disc growth.

DISCUSSION
Understanding the physiologic roles of inter-organ signaling during development is an emerging and complex field. In particular, how and which inter-organ signals regulate organ/tissue growth remains an open question. In the present study, we demonstrate that the Drosophila Activin-like ligand Myoglianin is a myokine that acts systemically to fine tune imaginal disc growth. We suggest Myo tuning of disc growth may play an important role in maintaining proper allometric scaling of appendage size with the size of the muscles that control their movements.

Only Myo regulates imaginal disc size
Although Babo/dSmad2 signaling has been previously implicated in imaginal disc growth control (Brummel et al., 1999;Hevia and de Celis, 2013;Peterson and O'Connor, 2013), the ligand(s) responsible and their production sites(s) have not been identified. Previous in situ hybridization and RNAi knockdown experiments suggested that all three Activin-like ligands (Myoglianin, Activin, and Dawdle) contribute to control of wing size (Hevia and de Celis, 2013). However, we find no expression in imaginal discs, with the exception of Act which is expressed in differentiating photoreceptors of the eye imaginal disc (Moss-Taylor et al., 2019;Zhu et al., 2008). More importantly, using genetic null mutants, we show that only loss of myo affects imaginal disc size. The discrepancy in phenotypes between tissue-specific knockdown results and the genetic nulls is often noted and not fully understood (Di Cara and King-Jones, 2016; Gibbens et al., 2011). In addition to simple off-target effects within the wing disc itself, one possible explanation is that many GAL4 drivers are expressed in tissues other than those reported, potentially resulting in deleterious effects for the animal that indirectly affect imaginal disc size. Another possibility is that in Actβ and daw genetic null backgrounds a non-autonomous compensatory signal is generated by another tissue and this signal is not activated in the case of tissue-specific knockdown (Di Cara and King-Jones, 2016). We think both of these explanations are unlikely in this instance since we demonstrate that only the Babo-A receptor isoform is expressed and required in discs and we previously showed that Daw only signals through the Babo-C isoform (Jensen et al., 2009). Therefore, it is unclear why knockdown of daw in the wing disc would result in a small wing phenotype as previously reported (Hevia and de Celis, 2013). We conclude that the small wing phenotypes caused by RNAi knockdown of Act or daw are likely the result of off-target effects and that Myo is the only Activin-type ligand that regulates imaginal disc growth.

Plum and Punt are required for efficient Babo-A signaling to dSmad2 in the wing disc
The signaling ability of TGF- ligands is modulated by the specific combinations of receptors and co-receptors to which they bind (Heldin and Moustakas, 2016). In Drosophila, the receptor requirements for effective signaling through dSmad2 likely vary for each ligand and tissue (Jensen et al., 2009). In this report, we find that Myo signaling in the wing disc requires Punt as the type II receptor, Babo-A as the type I receptor and Plum as a likely co-receptor. Furthermore, we establish that Myo is the exclusive Activin-like ligand signaling to the discs since loss of Myo eliminated detectable phosphorylation of dSmad2 in the wing imaginal disc. Since Babo-a is the only isoform expressed in wing discs we also conclude that Myo is able to signal through this isoform in the absence of other isoforms. Whether Myo can also signal through Babo-a or c is not yet clear, but in the context of mushroom body remodeling Babo-A also appears to be the major receptor isoform required (Awasaki et al., 2011;Yu et al., 2013). Plum is also required for mushroom body remodeling, suggesting that Plum and Babo-a are both required for efficient Myo signaling. However, it is noteworthy that Plum null mutants are viable (Yu et al., 2013) while Myo null mutants are not. This observation suggests that Plum is not required for all Myo signaling throughout the animal.
The requirement of Punt as a type II receptor for production of an efficient signaling complex with Myo may be context dependent. In the mushroom body, indirect genetic evidence suggests that the two Type II receptors function redundantly (Zheng et al., 2006). Although both punt and wit are expressed in imaginal discs (Childs et al., 1993;Marqués et al., 2002), only loss of punt produces a phenotype in the brk reporter assay. The requirement of both Punt and Plum for Myo signaling may also explain why a previous attempt to study Myo signaling in a heterotypic cell culture model failed (Lee-Hoeflich et al., 2005). In that study, Myo was found to form a complex with Wit and Babo-a in COS-1 cells but no phosphorylation of dSmad2 was reported. One explanation is that effective signaling by Myo requires Punt, Plum, and Babo-A, and none of these is found in mammalian cells. Collectively, our results provide in vivo functional evidence for a Myo signaling network that regulates growth.

Proliferation and cell size
Final tissue size is determined by several factors including cell size, proliferation, death rates, and duration of the growth period. While we did observe cell size changes upon manipulation of Myo signaling, the direction of change depended on the genotype. In myo mutants, estimation of cell size via apical surface area indicates that the cells are ~20% smaller than wild-type. Although this measurement does not indicate the actual volume of the cells, it gives an indication of cell density in the epithelial sheet of the wing pouch, which is analogous to counting cells in the adult wing. RNAi knock down of babo-a in the entire disc produced smaller adult wings with larger (less dense) cells. This result differs from the myo mutant, but is similar to the reported adult wing phenotypes of babo mutants (Hevia and de Celis, 2013) and larval disc phenotypes of dSmad2 mutants (Peterson and O'Connor, 2013). When babo-a is knocked down in one compartment, that compartment is reduced in size with smaller cells. We conclude that tissue size reduction is the consistent phenotype upon loss of Myo signaling, but cell size changes depend on the specific type of manipulation.
While cell size effects may be context dependent, it is notable that neither reduction in size of imaginal discs nor adult wing surface area can be explained solely by a cell size defect. Therefore, we believe that a subtle proliferation or apoptotic defect also contributes to the phenotype. Consistent with the possibility of a proliferation defect is the observation that the large disc phenotype exhibited by dSmad2 protein null mutants is clearly dependent on Myo, and we have previously shown that this is the result of enhanced proliferation (Peterson and O'Connor, 2013). Likewise, earlier studies showed that expression of activated Babo in wing discs also leads to larger wings with slightly smaller cells which is most easily explained by an enhanced proliferation rate (Brummel et al., 1999). It is worth noting that this proposed enhanced proliferation rate will be very difficult to detect since cell division is random with regard to space and time during development (Wartlick et al., 2011). Thus the final ~40% reduction in adult wing size translates to disc cells dividing on average 1 less time throughout the entire time course of larval development. Therefore, without prolonged live imaging, this small reduction in proliferation rate will not be detectable using assays that provide only static snap shots of cell division.
One attempt to shed light on the transcriptional output of TGF-β signaling responsible for wing disc size employed microarray mRNA profiling of wild-type versus dSmad2 gain-and loss-of-function wing discs (Hevia et al., 2017). However this study did not reveal a clear effect on any class of genes including cell cycle components, and it was concluded that the size defect is the result of small expression changes of many genes Consistent with this view are dSmad2 Chromatin Immunoprecipitation experiments in Kc cells which revealed that dSmad2 is associated with many genomic sites and thus may regulate a myriad of genes (Van Bortle et al., 2015).

Is myoglianin function evolutionarily conserved?
Insect myoglianin is a clear homolog of vertebrate Myostatin (Mstn/GDF8) (Hinck et al., 2016), a TGF- family member notable for its role in regulating skeletal muscle mass. Mstn loss-of-function mutants lead to enlarged skeletal muscles (Grobet et al., 1997;McPherron and Lee, 1997). Mstn is thought to affect muscle size through autocrine signaling that limits muscle stem cell proliferation, as well as perturbing protein homeostasis via the Insulin/mTOR signaling pathways (Morikawa et al., 2016;Rodriguez et al., 2014). Similarly, Gdf11, a Mstn paralog, also regulates size and proliferation of muscles and adipocytes, and may promote healthy aging (Egerman et al., 2015;McPherron et al., 2009;Sinha et al., 2014). Mstn and Gdf11 differ in where they are expressed and function. Mstn is highly expressed in muscles during development while Gdf11 is weakly expressed in many tissues. Both molecules are found to circulate in the blood as latent complexes in which their N-terminal prodomains remain associated with the ligand domain. Activation requires additional proteolysis of the N terminal fragment by Tolloid-like metalloproteases to release the mature ligand for binding to its receptors (Uhlén et al., 2015;Wolfman et al., 2003). Interestingly in Drosophila, the Myoglianin N-terminal domain was also found to be processed by Tolloid-like factors, but whether this is a prerequisite for signaling has not yet been established (Serpe and O'Connor, 2006 ). In terms of functional conservation in muscle size control, however, the role of Drosophila Myo remains ambiguous. Analysis of the null mutant in the present study indicates that it has little effect on muscle size. However a recent study employing a muscle-specific RNAi knockdown of myoglianin (myo), reported finding larger muscles similar to the vertebrate observation (Augustin et al., 2017). This discrepancy between the tissue-specific RNAi knockdown and the null phenotypes is something that is found relatively frequently in both Drosophila and also in vertebrates and may represent activation of compensatory pathways in the null phenotype that are not triggered by tissue specific arthropods, it appears that the Myo/Activin pathway is required to promote muscle growth, while in vertebrates they inhibit muscle growth. Whether arthropod Activin-like ligands promote muscle growth by regulating Insulin/Tor signaling in the opposite direction compared to vertebrates remains to be determined.

Intrinsic (intra-organ) vs extrinsic (inter-organ) mode of Myo signaling and the significance of muscle to disc inter-organ communication
The most intriguing finding of this study is that muscle-derived Myo acts non-autonomously to regulate imaginal disc growth. This is in stark contrast to the two BMP ligands, Dpp and Gbb, which are produced by disc cells and act autonomously within the disc itself to regulate both growth and pattern (Haerry et al., 1998;Minami et al., 1999). The fact that a TGF- ligand can act in an endocrine-like manner is not particularly novel since many vertebrate members of the TGF- family, including Myostatin, the closest homolog to Drosophila Myoglianin, are found in the blood (Sinha et al., 2014).
Even the disc intrinsic molecule Dpp has been recently shown to be secreted into the hemolymph where it circulates and signals to the prothoracic gland to regulate a larval nutritional checkpoint (Setiawan et al., 2018). Several additional reports indicate that ligands from the Drosophila Activin-like subfamily also circulate in the hemolymph and function as inter-organ signals. For example, muscle derived Actβ and Myo signal to the fat-body to regulate mitochondrial function and ribosomal biogenesis, respectively (Demontis et al., 2014;Song et al., 2017). In addition, Dawdle produced from many tissue sources can signal to the Insulin producing cells and the mid-gut to stimulate Insulin secretion and repress expression of sugar metabolizing genes, respectively (Chng et al., 2014;Ghosh and O'Connor, 2014). Thus many TGF- type factors act as both paracrine and endocrine signals depending on the tissue and process involved.
The phenotype of the myo mutant animal supports our claim that endogenous Myo contributes to imaginal disc growth. The ectopic expression assay produced various wing disc sizes when Myo was expressed in different tissues, indicating that the growth response likely depends on the level of Myo being produced in the distal tissue. Loss of glial derived Myo is not sufficient to suppress overgrowth of dSmad2 mutant discs, but overexpression of Myo in glia did partially rescue size of myo null wing discs, likely because the repo-Gal4 driven overexpression produces more ligand than endogenous glia.
Likewise, expression from a large tissue such as muscle or fatbody likely produces more Myo than glia leading to normal disc growth or even overgrowth. It is also possible that Myo signaling activity is modified depending on the tissue source. Like other TGF− family members, Myo requires cleavage by a furin protease at its maturation site to separate the C-terminal ligand from the prodomain (Lee-Hoeflich et al., 2005;Lo and Frasch, 1999), and Myo may also require a second cleavage by a Tolloid protease family member to achieve full dissociation of the prodomain from the ligand to ensure complete activation (Serpe and O'Connor, 2006). Either of these cleavage reactions, or any other step impacting the bioavailability of active Myo ligand, may vary with tissue or may be modulated by environmental conditions.
What is the rationale for larval muscle regulating imaginal discs size? A possible reason is that for proper appendage function, the muscle and the structure (leg, wing, and haltere) that it controls should be appropriately matched to ensure optimal organismal fitness for the environmental niche the adult occupies. For example, a large muscle powering a small wing might result in diminished fine motor control. Conversely, a small muscle may not be able to power a large wing to support flight. However, the multi-staged nature of muscle and appendage development complicates this picture. Larval muscles are histolysed during metamorphosis and do not contribute to the adult muscle. However, remnants of larval muscles in the thoracic segment are preserved as fibers that act as scaffolds upon which the larval myoblasts infiltrate and fuse to become the adult indirect flight muscles (Dutta et al., 2004). Thus, at least for the indirect flight muscles, the size of the larval muscle scaffold might contribute to the building of a bigger adult muscle. Another possibility invokes a signal relay system. Wg and Serrate/Notch signaling from the wing disc epithelial cells control myoblast proliferation during larval development (Gunage et al., 2014). Thus it may be that Myo signaling from the larval muscles stimulates proliferation of the disc epithelial layer which in turn enhances Wnt and Serrate/Notch signaling to myoblasts to increase their 20 number thereby coordinating the adult appendage size with muscle size. A final scenario is that, since muscle is a major metabolic and endocrine organ, Myo production may be regulated by the general metabolic state of the larva. If healthy, high levels of Myo, in concert with other growth signals such as insulin, leads to a bigger fly with large wings, and if the metabolic state is poor then lower Myo levels leads to diminished proliferation and a smaller cell size resulting in a smaller fly with small wings.
Regardless of the precise mechanism, the ability of the muscle to control appendage size has interesting implications in terms of evolutionary plasticity. The proportionality of insect wing size to body size can vary over a large range (Shingleton et al., 2008), but the mechanism responsible for determining this particular allometric relationship for a given species is not understood. We have recently demonstrated that in Drosophila, motoneuron derived Act, another TGF- superfamily member, can dramatically affect muscle/body size (Moss-Taylor et al., 2019). Therefore, it is tempting to speculate that evolutionary forces might modulate the activity of these two genes to produce an appropriate body-wing allometry that is optimal for that species' ecological niche.        (red squares) leads to widening with a ratio of 0.6; RNAi for babo (all isoforms, purple triangles) or babo-a (green circles) alone rescues the widening back to wild type ratios. These discs are smaller than 25 wild type due to loss of canonical Baboon activity ( Fig.1 and Fig.2). There is no effect to widening by knocking down either babo-b or babo-c (blue and yellow circles vs red squares).

27
Larval muscle and wing disc epithelium actin-belt were stained with rhodamine-phalloidin. To measure wing disc size, tissues were stained with DAPI.
Tissues for fluorescence microscopy were mounted with 80% Glycerol in PBS (0.1% TritonX-100). Wide-field images taken with a 20X objective on a Zeiss Axiovert microscope. CARV Confocal images were captured using a 20X objective on a Zeiss Axiovert. We also used the Zeiss LSM710 for confocal and "tile-scanning" imaging.
For adult wings, animals were dissected in 95% ethanol, and mounted with (50:50) Canadian balsam and wintergreen oil solution. Images were taken with a 4X objective (for whole wing) and 40X objective (for trichome density).
Images were processed in FIJI (imageJ). Wing disc size were calculated with the DAPI staining then using the threshold and measure functions. For larval muscle size and adult wing size, outlines were drawn using the polygon tool.

Western Blot:
Mid-late third instar larvae were dissected in cold PBS, and wing imaginal discs were collected and washed on ice. Total protein was extracted with RIPA buffer (Cell Signaling) with cOmplete protease inhibitor (Roche) and PhosSTOP phosphatase inhibitor (Roche). Lysates were mixed with reducing sample buffer, boiled, and 10 µl of sample were run on 4-12% graded precast gels (Invitrogen). Resolved protein samples were transferred to PVDF membrane. Membrane blocking and probing were performed using standard protocols for ECL. Antibodies used for western blot: phospho-Smad2 (Cell Signaling #3108), Tubulin (Sigma #T9026), goat anti-mouse HRP, and goat anti-rabbit HRP.

qPCR:
28 Late-third instar larvae were dissected in cold PBS, and tissues were pooled and washed in PBS on ice. Tissues were homogenized in Trizol (Invitrogen) and Total RNA purified using RNeasy Mini Kit (Qiagen). cDNA libraries were made using SuperScript-III (Invitrogen). qPCR was performed using SYBR green reagent (Roche) on a LightCycler 480. Rpl23 was used as a housekeeping gene for normalization. We confirmed that primer sets for all three isoforms amplify equivalently using Babo-A, B and C control cDNA templates. The following qPCR primers were used for babo isoforms:

Statistics:
Statistical analysis were performed using Prism 6 (Graphpad) software. All experiments were repeated at least twice. Sample size was determined using published methods in the field. Data collection and analysis were not blind. Quantitative data are plotted as mean with standard deviations. Significance were determined using 2-tailed t-test with Welch's correction or an ANOVA with multiple comparisons. P values < 0.05 = *, < 0.01 = **, < 0.001 = ***.