Systematic transcriptomics reveals a biphasic mode of sarcomere morphogenesis in flight muscles regulated by Spalt

Muscles organise pseudo-crystalline arrays of actin, myosin and titin filaments to build force-producing sarcomeres. To study how sarcomeres are built, we performed transcriptome sequencing of developing Drosophila flight muscles and identified 40 distinct expression profile clusters. Strikingly, two clusters are strongly enriched for sarcomeric components. Temporal gene expression together with detailed morphological analysis enabled us to define two distinct phases of sarcomere development, which both require the transcriptional regulator Spalt major. During the sarcomere formation phase, 1.8 μm long immature sarcomeres assemble myofibrils that spontaneously contract. During the sarcomere maturation phase, these sarcomeres grow to their final 3.2 μm length and 1.5 μm diameter and acquire stretch-sensitivity. Interestingly, the final number of myofibrils per flight muscle fiber is determined at the onset of the first phase. Together, this defines a biphasic mode of sarcomere and myofibril morphogenesis – a new concept that may also apply to vertebrate muscle or heart development.


Introduction
Sarcomeres are the stereotyped force producing mini-machines present in all striated muscles of bilaterians. They are built of three filament types arrayed in a pseudocrystalline order: actin filaments are cross-linked with their plus ends at the sarcomeric Zdisc and face with their minus ends towards the sarcomere center. In the center, symmetric bipolar muscle myosin filaments, anchored at the M-line, can interact with the actin filaments. Myosin movement towards the actin plus ends thus produces force during sarcomere shortening. Both filament types are permanently linked by a third filament type, the connecting filaments, formed of titin molecules (Gautel and Djinovic-Carugo, 2016;Lange et al., 2006). A remarkable feature of sarcomeres is their stereotyped size, ranging from 3.0 to 3.4 µm in relaxed human skeletal muscle fibers (Ehler and Gautel, 2008;Llewellyn et al., 2008;Regev et al., 2011). Even more remarkable, the length of each bipolar myosin filament is 1.6 µm in all mature sarcomeres of vertebrate muscles, requiring about 300 myosin hexamers to assemble per filament (Gokhin and Fowler, 2013;Tskhovrebova and Trinick, 2003).
Human muscle fibers can be several centimetres in length and both ends of each fiber need to be stably connected to tendons to achieve body movements. As sarcomeres are only a few micrometres in length, many hundreds need to assemble into long linear myofibrils that span from one muscle end to the other and thus enable force transmission from the sarcomeric series to the skeleton (Lemke and Schnorrer, 2017a). Thus far, we have a very limited understanding of how sarcomeres initially assemble into long immature myofibrils during muscle development to exactly match the length of the mature muscle fiber (Sparrow and Schöck, 2009). In particular, we would like to ! 4! understand how such sarcomeres mature to the very precise stereotyped machines present in mature muscle fibers. Across evolution, both the pseudo-crystalline regularity of sarcomeres as well as their molecular components are well conserved (Ehler and Gautel, 2008;Vigoreaux, 2006). Thus, Drosophila is a valid model to investigate the biogenesis of sarcomeres as well as their maturation. In particular, the large indirect flight muscles (IFMs) that span the entire fly thorax are an ideal model system to investigate mechanisms of myofibrillogenesis. They contain thousands of myofibrils consisting of 3.2 µm long sarcomeres (Schönbauer et al., 2011;Spletter et al., 2015).
Like all Drosophila adult muscles, IFMs are formed during pupal development from a pool of undifferentiated myoblasts called adult muscle precursors (AMPs) (Bate et al., 1991). From 8 h after puparium formation (APF), these AMPs either fuse with themselves (for the dorso-ventral flight muscles, DVMs) or with remodelled larval template muscles (for the dorso-longitudinal flight muscles, DLMs) to form myotubes (Dutta et al., 2004;Fernandes et al., 1991). These myotubes develop dynamic leading edges at both ends and initiate attachment to their respective tendon cells at 12 to 16 h APF . These attachments mature and mechanical tension is built up in the myotubes, followed by the formation of the first immature periodic myofibrils at 30 h APF when the muscle fibers are about 150 µm in length .
These immature myofibrils contain the earliest sarcomeres, which are about 1.8 µm in length . During the remaining 3 days of pupal development, the muscle fibers grow to about 1 mm to fill the entire thorax and sarcomere length increases to the final length of about 3.2 µm in adult flies (Orfanos et al., 2015).

! 5!
After myoblasts have fused to myotubes, the flight muscle specific selector gene spalt major (spalt, salm) is turned on in the developing flight muscle myotubes. Spalt major is responsible for the correct fate determination and development of the flight muscles, which includes the fibrillar flight muscle morphology and the stretch-activated muscle contraction mode (Schönbauer et al., 2011;Syme and Josephson, 2002). It does so by controlling the expression of more than 700 flight muscle specific genes or gene isoforms during development (Spletter and Schnorrer, 2014;Spletter et al., 2015).
However, how the interplay between all these isoforms instructs the formation of highly regular, pseudo-crystalline sarcomeres in the flight muscle is not understood.
Here, we studied the dynamics of flight muscle development in detail. We performed a systematic mRNA-Seq time-course of isolated muscle tissue at 8 time points from the myoblast stage until the mature adult muscle stage. Bioinformatic analysis of expression dynamics identified two gene clusters that are strongly enriched for sarcomeric genes. The temporal dynamics of these clusters enabled us to define two distinct phases of sarcomere morphogenesis: a first sarcomere formation phase in which immature myofibrils with short and thin sarcomeres are built, and a second sarcomere maturation phase during which the immature sarcomeres grow to their final length and diameter and functionally mature. Interestingly, all of the myofibrils assemble simultaneously, with the final number of myofibrils being determined at the beginning of the sarcomere formation phase. During the sarcomere maturation phase myofibril number remains constant, suggesting that every mature sarcomere needs to undergo this biphasic development. Both phases require the activity of the flight muscle selector gene spalt major, demonstrating that muscle fiber type-specific transcription is continuously ! 6! required during all phases of sarcomere formation and maturation. Together, these findings suggest that a precise transcriptional control is required to first assemble and then mature sarcomeres to their pseudo-crystalline regularity.

A time-course of indirect flight muscle development
To better understand muscle morphogenesis in general and myofibrillogenesis in particular, we focused on the Drosophila indirect flight muscles (IFMs). We hypothesised that major morphological transitions during IFM development may be induced by transcriptional changes, thus we aimed to generate a detailed developmental mRNA-Seq dataset from IFMs. IFMs are built from AMPs that adhere to the hinge region of the wing disc epithelium and are labelled with Him-GAL4 driven GFP ( Figure   1A) (Soler and Taylor, 2009). At 16 h APF, many of these myoblasts have fused to larval template muscles to build the dorsal-longitudinal flight muscle (DLM) myotubes, which initiate attachment to their tendons. At this stage, the DLM myotubes of fibers 3 and 4 have a length of about 300 µm ( Figure 1B). Fusion ceases at about 24 h APF ( Figure 1C) and attachment matures until 32 h APF, coinciding with the strong recruitment of βPS-Integrin and the spectraplakin homolog Shortstop (Shot) to the attachment sites. At this stage the myofibers have built up mechanical tension and compacted to a length of about 150 µm, coinciding with the appearance of long Shot-positive tendon extensions that anchor the muscles within the thorax. This important developmental transition is highlighted by the assembly of immature myofibrils visualised by strong F-actin staining throughout the entire muscle fiber ( Figure 1D) .
After 32 h APF, the myofibers undergo another developmental transition and begin to grow dramatically. They elongate 3-fold to reach a length of about 480 µm by 48 h APF ( Figure 1E) and about 590 µm by 56 h APF ( Figure 1F). Concomitantly, the tendon extensions shrink with the myofibers being directly connected to the basal side of ! 8! the tendon cell epithelium by 72 h APF ( Figure 1G). At the end of pupal development (90 h APF), wavy muscle fibers with a length of about 780 µm containing mature myofibrils ( Figure 1H) are present within the thorax.

Expression dynamics during indirect flight muscle development
To quantify transcriptional dynamics across the entire developmental time course, we focused on the major developmental transitions and isolated mRNA from dissociated myoblasts of dissected or mass-isolated third instar wing discs and from hand-dissected IFMs at 16 h, 24 h, 30 h, 48 h, 72 h and 90 h APF pupae, and adult flies 1 day after eclosion ( Figure 1I). We performed mRNA-Seq using at least two biological replicates for each time point (see Materials and Methods). To identify genes with similar temporal expression profiles, we used Mfuzz (Kumar and E Futschik, 2007) to cluster standard normalized read counts from all genes expressed above background (12,495 of 13,322 genes). This allowed us to confidently identify 40 distinct genome-wide clusters (Figure 2-S1), each of which contains a unique gene set ranging from 155 to 703 members (Supplementary Table 1). These clusters represent various temporal expression dynamics, with high expression at early (myoblast proliferation and fusion), mid (myotube attachment and myofibril assembly) or late (myofiber growth) myogenesis stages or a combination thereof ( Figure 1I, Figure 2-S1). These distinct patterns suggest a precise temporal transcriptional regulation corresponding to observed morphological transition points.
To verify the mRNA-Seq and cluster analysis, we selected a number of 'indicator' genes with available antibodies or GFP fusion proteins whose expression correlates with ! 9! important developmental transitions. Twist (Twi) is a myoblast nuclear marker at larval stages and its expression needs to be down-regulated after myoblast fusion in pupae (Anant et al., 1998). We find twi mRNA in Mfuzz cluster 27, with high expression in myoblasts until 16 h APF and a significant down-regulation from 24 h APF, which we were able to verify with antibody stainings (Figure 2A-C). The flight muscle fate selector gene spalt major (salm) (Schönbauer et al., 2011) and its target, the IFM splicing regulator arrest (aret, bruno) (Spletter et al., 2015) are members of cluster 26 and 14, respectively. Expression of both clusters is up-regulated after myoblast fusion at 16 h APF, which we were able to verify with antibody stainings ( Figure 2D-F, Figure 2-S2A-C). For the initiation of muscle attachment, we selected Kon-tiki (Kon) (Schnorrer et al., 2007;, member of cluster 15, which is transiently up-regulated after myoblast fusion before it is down-regulated again after 30 h APF. Consistently, we found Kon-GFP present at muscle attachment sites at 30 h APF but not at 72 h APF ( Figure 2G-I). A similar expression peak shifted to slightly later time points is found in cluster 34, which contains β-tubulin 60D (βTub60D) (Leiss et al., 1988). Consistently, we find β-Tub60D-GFP (Sarov et al., 2016) expression in IFMs at 30 h and 48 h but not 72 h APF ( Figure 2J-L). After attachment is initiated, the attachments need to mature and be maintained. As expected, we found the essential attachment components βPS-Integrin (mys) and Talin (rhea) in clusters that are up-regulated after myoblast fusion until adulthood (clusters 7 and 25, respectively). This is consistent with continuous high protein expression of βPS-Integrin-GFP and Talin-GFP at muscle attachment sites ( Figure 2M-O, Figure 2-S2D-F). Taken together, these semi-quantitative protein ! 10! localisation data nicely validate the temporal mRNA dynamics found in the mRNA-Seq data, confirming our methodology.

Sarcomeric and mitochondrial gene induction after 30 h APF
Hierarchical clustering of the core expression profiles from the 40 identified Mfuzz clusters defines 8 temporally ordered groups (Figure 3) that show progressive expression dynamics as muscle development proceeds. GO-Elite analysis (Zambon et al., 2012)  At late stages of flight muscle development, mitochondrial density strongly increases (Clark et al., 2006). Using GO-Elite, we found a strong enrichment for mitochondrial related pathways in four late up-regulated clusters, namely 3, 28, 39 as ! 11! well as the sarcomere cluster 22 (Figure 3). By comparing the clusters to systematic functional data acquired at all stages of Drosophila muscle development (Schnorrer et al., 2010), we find enrichments in clusters throughout the time course. Interestingly, genes highly expressed in flight muscle compared to other muscle types, identified as 'salmcore genes' (Spletter et al., 2015), are also enriched in the late clusters, including the mitochondrial enriched clusters 3, 28, 39 and the sarcomere enriched cluster 22 ( Figure   3). These data highlight the changes in biological process enrichments that parallel expression dynamics, with a very particular change happening during later stages of muscle development after 30 h APF. This corresponds to a time period after immature myofibrils have been assembled, which thus far remained largely unexplored.
To examine the temporal expression dynamics in more detail, we performed a After 48 h the myofibril diameter grows nearly 3-fold from 0.46 µm to 1.43 µm in adult flies ( Figure 5E-G), while fiber cross-sectional area grows nearly 4-fold from 1,759 µm 2 to 6,970 µm 2 . Strikingly, during the entire time period from 30 h APF to adults, the total number of myofibrils per muscle fiber remains constant (about 2,000 per muscle fiber).
Taken together, these quantitative data lead us to propose a biphasic model of sarcomere morphogenesis: 1) During the sarcomere formation phase, which lasts from about 30 h until shortly after 48 h APF, short and thin immature sarcomeres are assembled into immature myofibrils. 2) During the sarcomere maturation phase, starting after 48 h APF, the existing short sarcomeres grow in length and thickness to reach the mature pseudocrystalline pattern. No new myofibrils are built during the sarcomere maturation phase ( Figure 1I).
We gained additional evidence to support this biphasic assembly model on both the molecular and functional levels. First, the two phases of sarcomere assembly complement the switch in gene expression we observe from 30 h to 72 h APF. Using which then likely acquire stretch-sensitivity as the immature myofibrils grow and mature during the sarcomere maturation phase, and thus cease contracting.

Salm induces sarcomeric gene expression during sarcomere maturation
How does biphasic transcription of the various sarcomeric components instruct the biphasic mode of myofibrillogenesis? To address this important question, we performed mRNA-Seq and compared wild type to spalt-major knock-down (salmIR) flight muscles (Supplementary Table 4). We see down-regulation of mRNAs coding for sarcomeric components at 24 h and 30 h APF, and in particular at 72 h APF during the phase of sarcomere maturation ( Figure 6A, Figure 6-S1A-C). The genes down-regulated in salmIR IFM are enriched for GO terms associated with sarcomere assembly and flight behaviour ! 15! and mitochondrial genes, as well as the mitochondrial Mfuzz clusters 3, 28, 39 and the sarcomeric Mfuzz cluster 22 ( Figure 6-S1D-F). Indeed, members of cluster 22, which is strongly enriched for sarcomeric and mitochondrial genes, are less strongly induced from 30 h to 72 h APF in salmIR muscle compared to wild type ( Figure 6B, C), suggesting that salm is indeed required for the strong induction of sarcomeric protein expression after 30 h APF.
Salm is expressed shortly after myoblast fusion and constitutive knock-down of salm with Mef2-GAL4 results in a major shift of muscle fiber fate (Schönbauer et al., 2011), which may indirectly influence transcription after 30 h APF. Hence, we aimed to reduce Salm levels only later in development, to directly address its role in the second sarcomere maturation phase. To this end, we knocked-down salm with the flight muscle specific driver Act88F-GAL4, which is expressed from about 18 h APF and requires salm activity for its expression (Bryantsev et al., 2012;Spletter et al., 2015). This strategy enabled us to reduce Salm protein levels at 24 h APF resulting in undetectable Salm levels at 72 h APF ( Figure 6-S2). To test if Salm instructs the transcriptional boost of sarcomeric components after 30 h APF, we performed quantitative imaging using unfixed living flight muscles expressing GFP fusion proteins under endogenous control. We used green fluorescent beads to normalise the GFP intensity between different samples. While overall sarcomere morphology is not strongly affected in Act88F>>salmIR muscles, we found that the levels of Strn-Mlck, Fln and Mhc are strongly reduced at 72 h as compared to wild-type controls ( Figure 6D-G). This suggests that salm is indeed required for the expression boost of a number of sarcomeric proteins during the sarcomere maturation phase after 48 h APF.

16!
To investigate the consequences of late salm knock-down, we quantified the myofibril and sarcomere morphology throughout the second phase of sarcomere maturation. The myofibrils display a fibrillar morphology, confirming that the early function of Salm to determine IFM fate was unaffected by our late knock-down. At 72 h APF and more prominently at 90 h APF and in adults, Act88F>>salmIR myofibrils showed actin accumulations at broadened Z-discs ( Figure 7A-H), which are often a landmark of nemaline myopathies (Sevdali et al., 2013;Wallgren-Pettersson et al., 2011).
The myofibril width was not significantly different in these myofibrils ( Figure 7I). However, the sarcomeres displayed a strong defect in sarcomere length growth after 48 h APF in Act88F>>salmIR muscles ( Figure 7J, Supplemental Table 3), with sarcomeres only obtaining a length of 2.84 µm in adult flies, demonstrating that Salm activity is required for normal sarcomere maturation and growth.

Salm function contributes to gain of stretch-activation during sarcomere maturation
Given the defects in sarcomere length and sarcomere gene expression in Act88F>>salmIR muscles, we explored the function of these abnormal muscle fibers. As To directly test the function of a sarcomeric component during the sarcomere maturation phase, we investigated the role of the prominently induced Salm target Strn-Mlck, which is largely incorporated during the sarcomere maturation process ( Figure 5S1E). In Strn-Mlck mutants, sarcomere and myofibril morphology, including myofibril width, is initially normal. However, at 80 h APF the sarcomeres overgrow, consistently reaching lengths of more than 3.5 µm and resulting in slightly longer muscle fibers at 80 h APF (Figure 8). After overgrowing, sarcomeres appear to hyper-contract resulting in short, thick sarcomeres in 1-day-old adults ( Figure 8E (Spletter et al., 2015).
Together, these data demonstrate that sarcomere maturation must be precisely controlled at the transcriptional level to enable the precise growth of sarcomeres to their final mature size. This ensures the lifelong function of the contractile apparatus of muscle fibers.

Discussion
In this study, we generated a systematic developmental transcriptomics resource from Drosophila flight muscle. The resource quantifies the transcriptional dynamics across all the major stages of muscle development over five days, starting with stem cell-like myoblasts and attaching myotubes to fully differentiated, stretch-activatable muscle fibers. In this study, we have specifically focused on the transcriptional regulation of sarcomere and myofibril morphogenesis; however, the data we provide cover all other expected dynamics, such as mitochondrial biogenesis, T-tubule morphogenesis, neuromuscular junction formation, tracheal invagination, etc. Thus, our data should be a versatile resource for the muscle community.

A transcriptional switch correlating with two phases of sarcomere morphogenesis
Earlier work has shown that the flight muscle myotubes first attach to tendon cells and then build-up mechanical tension. This tension triggers the simultaneous assembly of immature myofibrils, converting the myotube to an early myofiber . This suggested a tension-driven self-organisation mechanism of myofibrillogenesis (Lemke and Schnorrer, 2017a). Here we discovered that myofibrillogenesis is not only regulated mechanically, but to a large extent also transcriptionally. At 30 h of pupal development, a large number of genes coding for sarcomeric proteins, including Mhc, Act88F and Unc-89/Obscurin, become up-regulated to enable the first phase of sarcomerogenesis -the assembly of short, immature sarcomeres within thin, immature myofibrils ( Figure 9). In this first phase until about At its end, the final number of sarcomeres are present within a defined number of myofibrils. These sarcomeres are contractile, but remain short and thin.
During the second sarcomere maturation phase, the existing immature sarcomeres grow in length and particularly in diameter to reach the pseudo-crystalline regularity of mature myofibrils within about two days of development. This sarcomere maturation phase is initiated by a strong transcriptional burst of the sarcomeric genes. Proteins already present in immature myofibrils like Mhc, Act88F and Unc-89/Obscurin are expressed to even higher levels, and new, often flight-muscle specific proteins like Mf, Fln and the titin-related isoform Strn-Mlck, are expressed to high levels and incorporated into the maturing sarcomeres, facilitating their dramatic growth ( Figure 9). Importantly, these matured sarcomeres no longer contract spontaneously, likely because they acquired the stretch-activated mechanism of contraction that is well described for mature Drosophila flight muscles (Bullard and Pastore, 2011;Josephson, 2006).
Our biphasic sarcomere morphogenesis model is strongly supported by the observation that the number of sarcomeres per myofibril does not increase after the sarcomere formation phase (ending shortly after 48 h APF). Furthermore, the number of myofibrils remains largely constant during the entire sarcomere morphogenesis period, suggesting that in flight muscles no new myofibrils are added after the initial assembly of immature myofibrils at 32 h APF. Together, this suggests that every sarcomere present in adult flight muscles undergoes this biphasic development.
The model is further supported by previous electron microscopy (EM) studies.
We and others found that immature myofibrils have a width of about 0.5 µm , which corresponds to about 4 thick filaments across each myofibril at the EM level at 42 h APF (at 22ºC) (Reedy and Beall, 1993). This 'core' myofibril structure built during the sarcomere formation phase is expanded dramatically after 48 h APF, reaching a mature width of 1.5 µm, corresponding to 35 thick filaments across each myofibril at the EM-level (Reedy and Beall, 1993). In total, each adult myofibril contains around 800 thick filaments (Gajewski and Schulz, 2010). The 'core' myofibril structure was also revealed by the preferential recruitment of over-expressed actin isoforms (Roper et al., 2005) and more importantly, by selective incorporation of a particular Mhc isoform that is only expressed at mid-stages of flight muscle development (Orfanos and Sparrow, 2013). This Mhc isoform expression switch corresponds to the global switch in sarcomeric gene expression between both sarcomere morphogenesis phases that we defined here. It also fits with recent observations that the formin family member Fhos is important for thin filament recruitment and growth in myofibril diameter after 48 h APF (Shwartz et al., 2016).

A role of active sarcomere contractions
Mature indirect flight muscles employ a stretch-activated mechanism of muscle contraction, thus Ca 2+ is not sufficient to trigger muscle contractions without additional mechanical stretch (Bullard and Pastore, 2011;Josephson, 2006). This is different to cross-striated body muscles of flies or mammals that contract synchronously with Ca 2+ influx. Hence, it is intriguing that immature flight muscle myofibrils do in fact contract ! 21! spontaneously, with the contraction frequencies and intensities increasing until 48 h APF.
It was recently proposed in Drosophila cross-striated abdominal muscles and in the developing cross-striated zebrafish muscles that spontaneous contractions are important for the proper formation of the cross-striated pattern (Mazelet et al., 2016;Weitkunat et al., 2017). A similar role for contractions was found in C2C12 cells by stimulating the contractions optogenetically (Asano et al., 2015). This shows that spontaneous contractions are a necessary general feature for the assembly of cross-striated muscle fibers across species.
However, flight muscles are not cross-striated in the classical sense, but have a fibrillar organisation in which each myofibril remains isolated and is not aligned with its neighbouring myofibrils ( Figure 5) (Josephson, 2006;Schönbauer et al., 2011). We can only speculate about the mechanism that prevents alignment of the myofibrils in the flight muscles, but it is likely related to their stretch-activated contraction mechanism.
This mechanism prevents spontaneous twitching due to increased Ca 2+ levels, because it additionally requires mechanical activation that can only occur during flight in the adult.
Thus, flight muscle sarcomeres not only grow and mature during the second phase of sarcomere maturation, but likely also gain their stretch-activatability.

Continuous maintenance of muscle type-specific fate
We identified a switch in gene expression between the sarcomere formation and sarcomere maturation phases. Such large scale transcriptome changes have also been observed during mouse (Brinegar et al., 2017), chicken (Zheng et al., 2009) and pig (Zhao et al., 2015) skeletal muscle development or during regeneration after injury in fish ! 22! (Montfort et al., 2016) and mouse muscles (Warren et al., 2007), indicating that muscle maturation generally correlates with large scale transcriptional changes.
It is well established that general myogenic transcription factors, in particular Mef2, are continuously required in muscles for their normal differentiation (Sandmann et al., 2006;Soler et al., 2012). Mef2 regulates a suit of sarcomeric proteins in flies, fish and mouse muscle important for correct sarcomere assembly and maturation (Hinits and Hughes, 2007;Kelly et al., 2002;Potthoff et al., 2007;Stronach et al., 1999). In Drosophila, Mef2 collaborates with tissue-specific factors, such as CF2, to induce and fine-tune expression of structural genes (Gajewski and Schulz, 2010;García-Zaragoza et al., 2008;Tanaka et al., 2008). General transcriptional regulators, such as E2F, further contribute to high levels of muscle gene expression observed during myofibrillogenesis, in part through regulation of Mef2 itself (Zappia and Frolov, 2016). However, it is less clear if muscle type-specific identity genes are continuously required to execute muscle type-specific fate. Spalt major (Salm) is expressed after myoblast fusion in flight muscle myotubes and is required for all flight muscle type-specific gene expression: in its absence the fibrillar flight muscle is converted to tubular cross-striated muscle (Schönbauer et al., 2011;Spletter et al., 2015). Here we demonstrate that Salm is continuously required for correct sarcomere morphogenesis, as late salm knock-down leads to defects in sarcomere growth during the sarcomere maturation phase, followed by severe muscle atrophy in adults.

Two phases of sarcomere morphogenesis -a general mechanism?
! 23!
Here we defined a biphasic mode of sarcomere morphogenesis in Drosophila flight muscles. Is this a general concept for sarcomere morphogenesis? Reviewing the literature, one finds that in other Drosophila muscle types which display a tubular crossstriated myofibril organisation, such as the fly abdominal muscles, the striated sarcomeres also first assemble and then grow in length (Pérez-Moreno et al., 2014;Weitkunat et al., 2017), suggesting a conserved mechanism. In developing zebrafish skeletal muscles, young myofibers present in younger somites show a short sarcomere length of about 1.2 µm, which increases to about 2.3 µm when somites and muscle fibers mature (Sanger et al., 2017;2009). Interestingly, sarcomere length as well as thick filament length increase simultaneously during fish muscle maturation, indicating that as in flights muscles the length of all sarcomeres in one large muscle fiber is homogenous at a given time (Sanger et al., 2009). Similar results were obtained in mouse cardiomyocytes measuring myosin filament length at young (2 somite) and older (13 somite) stages (Du et al., 2008) and even in human cardiomyocytes, in which myofibrils increase nearly 3fold in width and become notably more organized and contractile from 52 to 127 days of gestation (Racca et al., 2016). These observations strongly suggest that our biphasic sarcomere morphogenesis model is indeed also applicable to vertebrate skeletal and possibly heart muscles. As the progression from phase one to phase two requires a transcriptional switch, it will be a future challenge to identify the possible feedback mechanism that indicates the successful end of phase one or a possible re-entry into the sarcomere formation phase during muscle regeneration or exercise induced muscle fiber growth.

Fly Strains
Fly stocks were maintained using standard culture conditions. Characterization of normal IFM sarcomere and fiber growth was performed in w 1118 grown at 27°C. salm RNAi was performed with previously characterized GD3029 (referred to as salmIR) and KK181052 (Schönbauer et al., 2011) from VDRC (http://stockcenter.vdrc.at) at 25°C using Act88F-GAL4 to induce knock-down after 24 h APF. Act88F-GAL4 x w 1118 served as control.
The Strn-Mlck-MiMIC insertion MI02893 into IFM-specific IsoR (Bloomington stock 37038) and TRiP hairpin JF02170 were obtained from Bloomington. The salm-EGFP line was used to sort wing discs (Marty et al., 2014). Tagged genomic fosmid reporter fly lines include strn4 (Strn-Mlck-GFP, Isoform R) (Spletter et al., 2015), fTRG500 (Mhc- The rhea-YPet line used to label muscle ends for live imaging of twitch events was generated by CRISPR-mediated gene editing at the endogenous locus (S.B.L & F.S., details will be published elsewhere). The kon-GFP line was generated by inserting GFP into the kon locus after its transmembrane domain using the genomic fosmid FlyFos021621, which was integrated using Φ-C31 into VK00033 (I. Ferreira and F.S., details will be published elsewhere).

Flight tests
Flight tests were performed as previously described (Schnorrer et al., 2010). Act88F-GAL4 crosses were kept at 25°C, as higher temperatures negatively impacted flight ability, because of the very high GAL4 expression levels in this strain. Adult males were collected on CO 2 and recovered at least 24 h at 25°C before testing. Flies were introduced into the top of a 1 m long cylinder divided into 5 zones. Those that landed in the top two zones were considered 'normal fliers', those in the next two zones 'weak fliers' and those that fell to the bottom of the cylinder 'flightless'.

Immuno-staining
Wing-discs were dissected from 3 rd instar wandering larvae in 1x PBS and fixed in 4% PFA in PBS-T. Discs were stained as described below for anti-GFP. Adult and pupal flight muscles were dissected and stained as previously described . Briefly, early pupae (16 h -60 h APF) were freed from the pupal case,

Microscopy and image analysis
Images were acquired with a Zeiss LSM 780 confocal microscope equipped with an α For determining myofibril diameter, samples were imaged using a 3x optical zoom (50 nm pixel size). At least 20 cross-section images from different fibers for >10 flies were acquired for each time point. The number of fibrils per section and fibril diameter were determined with the tool 'analyze myofibrils crosswise' from MyofibrilJ.
In this tool, an initial estimate of the diameter is obtained by finding the first minimum in the radial average profile of the autocorrelation (Goodman, 1968) of the image. This estimate is used to calibrate the optimal crop area around all the cross-sections in the image, their position previously detected by finding the local intensity peaks. All of the detected cross sections are then combined to obtain a noise-free average representation of the fibril section. Finally, the diameter is calculated by examining the radial profile of the average and measuring the full width where the intensity is 26% of the maximum range.
For determining sarcomere length and myofibril width, for each experiment between 10 and 25 images were acquired from more than 10 individual flies. From each image, 9 non-overlapping regions of interest were selected, which were rotated to orient fibrils horizontally, when necessary. The tool 'analyze myofibrils lengthwise' from MyofibrilJ reports the sarcomere length (indicated as repeat) and myofibril width Live imaging of developmental spontaneous contractions was performed on a Leica SP5 confocal microscope. Prior to imaging, a window was cut in the pupal case, and pupae were mounted in slotted slides as previously described (Lemke and Schnorrer, 2017b;. At the specified developmental time point, IFMs were recorded every 0.65 seconds for 5 min. General movement within the thorax was distinguished from IFM-specific contraction, and each sample was scored for the number of single or double contractions observed per 5 minute time window. Data were recorded in Excel and ANOVA was performed in GraphPad Prism to determine significant differences. Movies were assembled in Fiji (Image J), cropped and edited for length to highlight a selected twitch event.
Quantitative imaging of fosmid reporter intensity was performed at 90 h APF in live IFM by normalizing to fluorescent beads (ThermoFisher (Molecular Probes), InSpeck™ Green Kit I-7219). IFMs were dissected from 5 flies, mounted with fluorescent microspheres (0.3% or 1% relative intensity, depending on the reporter intensity) in the supplied mounting medium and immediately imaged (within 20 minutes). Intensity measurements were obtained at 40x for at least 10 flies in regions where both IFM and at least 3 beads were visible. Control Act88F-GAL4 x w 1118 and RNAi Act88F-GAL4;; fosmid-GFP x salmIR (fosmids used include Strn-Mlck-GFP,  were imaged in the same imaging session.
Relative fluorescence fiber to beads was calculated for each image in Fiji by averaging ! 30! intensity for 3 fiber ROIs and 3 bead ROIs. Data were recorded in Excel and Student's ttest for significance and plotting were performed in GraphPad Prism.

mRNA-Seq
We previously published mRNA-Seq analysis of dissected IFMs from Mef2-GAL4, UAS-GFP-Gma x w 1118 at 30 h APF, 72 h APF and 1d adult, and Mef2-GAL4, UAS-GFP-Gma x salmIR in 1d adult (Spletter et al., 2015). We expanded this analysis in the present study to include myoblasts from 3 rd instar larval wing discs (see below) and dissected

Wing disc sorting and myoblast isolation
To perform mRNA-Seq on fusion competent myoblasts that will form the IFMs, we first dissected wing discs from wandering 3 rd instar larvae and manually cut the hinge away from the wing pouch. mRNA was isolated in TriPure reagent and sequenced as described above. We estimate this sample (Myo1) is ~50% myoblast, as the myoblasts form a nearly uniform layer over the underlying epithelial monolayer. To obtain a purer myoblast sample, we performed large-scale imaginal disc sorting followed by dissociation. We used particle sorting to isolate imaginal discs from Him-GAL4, UAS-BBM (UAS-palmCherry); salm-EGFP flies based on the green fluorescent signal. 10-12 mL of larvae in PBS were disrupted using a GentleMACS mixer (Miltenyl Biotec) and discs were collected through a mesh sieve (#0278 in, 25 opening, 710 µm). Fat was removed by centrifugation for 10 min. at 1000 rpm at 4ºC, discs were rinsed in PBS and then re-suspended in HBSS. Discs were further purified on a Ficoll gradient (25%:16%).
Discs were then sorted on a Large Particle Flow Cytometer (BioSorter (FOCA1000), Union Biometrica, Inc.), obtaining 600-1000 discs per sample. Discs were spun for 5 min at 600 rcf in a Teflon Eppendorf tube and then re-suspended in the dissociation mixture (200 µl of 10x Trypsin, 200 µl HBSS, 50 µl collagenase (10 mg/mL), 50 µl dispase (10 mg/mL)). The tube was incubated for 10 min. at RT and then transferred to a thermal shaker for 30 min. at 25°C at 650 rpm. Myoblasts were filtered through a 35 µm tube-cap filter and spun at 600 rcf for 5 min. to pellet the cells. Cells were resuspended in HBSS ! 32! for evaluation or frozen in TriPure reagent for RNA extraction. We obtained samples with ~90% purity based on counting the number of red fluorescent cells / non-fluorescent + green fluorescent cells in 3 slide regions. mRNA was isolated in TriPure reagent and sequenced as described above, generating the Myo2 and Myo3 samples.

Analysis of RNA-Seq data
FASTA files were de-multiplexed and base called using Illumina software. Reads were trimmed using the FASTX-toolkit. Sequences were mapped using STAR (Dobin et al., 2013) to the Drosophila genome (BDGP6.80 from ENSEMBL). Mapped reads were sorted and indexed using SAMtools (Li et al., 2009), and then bam files were converted to bigwig files. Libraries were normalized based on library size and read-counts uploaded to the UCSC Browser for visualization.
Genome-wide soft clustering was performed in R with Mfuzz (Futschik and Carlisle, 2005), using the DESeq2 normalized count values. We filtered the dataset to ! 33! include all genes expressed at one time point or more, defining expression as >100 counts after normalization. We then set all count values <100 to 0, to remove noise below the expression threshold. DESeq2 normalized data was standardized in Mfuzz to have a mean value of zero and a standard deviation of one, to remove the influence of expression magnitude and focus on the expression dynamics. We tested "k" ranging from 10-256.
We then performed consecutive rounds of clustering to obtain 3 independent replicates with similar numbers of iterations, ultimately selecting a final k=40 clusters with iterations equal to 975, 1064 and 1118. We calculated a "stability score" for each cluster by calculating how many genes are found in the same cluster in each run (Supplementary Table 1). Figures are from the 1064 iterations dataset. Mfuzz cluster core expression profiles were calculated as the average standard-normal expression of all genes with a membership value greater than or equal to 0.8, and then core profiles were clustered in R using Euclidean distance and complete linkage.
Enrichment analysis was performed with GO-Elite (Zambon et al., 2012) using available Gene Ontology terms for Drosophila. We additionally defined user provided gene lists for transcription factors, RNA binding proteins, microtubule associated proteins, sarcomeric proteins, genes with an RNAi phenotype in muscle (Schnorrer et al., 2010), mitochondrial genes ( http://mitoXplorer.biochem.mpg.de) and salm core fibrillar genes (Spletter et al., 2015). Full results and gene lists are available in Supplementary   Table 2. These user-supplied lists allowed us to define more complete gene sets relevant to a particular process or with a specific localization than available in existing GO terms.
Analysis was performed with 5000 iterations to generate reliable significance values.        Myofibrils and sarcomeres at these time points were stained for phalloidin (F-actin, red) and Kettin (Z-disc, green). Scale bar represents 5 µm.           After 48 h APF, all sarcomeres strongly grow in width and length by the incorporation of new structural proteins. This enables the flight muscle to gain stretch-activation. salm is required before the first phase to specify the fibrillar muscle fate and during the second ! 53! phase to boost the expression of sarcomeric proteins. Muscles are shown in red, tendons in blue. Structural proteins are illustrated as cartoons and are not drawn to scale.

Supplementary Table 1. mRNA-Seq raw data
The file includes multiple tabs containing the raw or input counts data from bioinformatics analysis, as well as a key to all original data provided in the supplementary tables. This table includes mRNA-Seq counts data, DESeq2 normalized counts data and standard normal counts data used for Mfuzz clustering for wild-type and salmIR IFM time points. The averaged core expression profiles for each Mfuzz cluster are also listed.

Supplementary Table 2. GO-Elite analysis data.
This table includes multiple tabs containing the GO-Elite analysis of enrichments in Mfuzz clusters as well as genes up-or down-regulated from 30 h -72 h APF and between wild-type and salmIR IFM. It also contains a complete list of all genes included in the 'User Defined' gene sets.

Supplementary Table 3. Summary of sarcomere and myofibril quantifications
This table includes a numerical summary of quantification values reported graphically in

GO and Gene Set Enrichments
RNAi phenotype in muscle* Salm core fibrillar genes*