Noncanonical roles for Tropomyosin during myogenesis

Nascent myotubes are faced with two major obstacles before they can form functional, contractile myofibers. First, the myotube must elongate over several cell diameters to identify and contact the appropriate tendon cells (Schnorrer and Dickson, 2004). Second, the myotube must express a wide range of muscle structural proteins and assemble those proteins into contractile sarcomeres (Rui et al., 2010). Although the mechanisms that regulate myotube precursor specification and muscle structural gene transcription have been characterized in detail (Buckingham and Rigby, 2014; Ciglar and Furlong, 2009), the molecules that coordinate myotube elongation are poorly understood.

The genetic tools in Drosophila have identified conserved cellular and molecular processes that direct striated muscle development. Drosophila somatic muscle is analogous to vertebrate skeletal muscle, and somatic muscle development initiates with the specification of myoblasts known as founder cells, that then fuse with neighboring fusion-competent cells to form nascent, multinucleate myotubes (Chen and Olson, 2004; de Joussineau et al., 2012). The nascent myotubes then elongate, activate muscle structural gene expression, attach to tendons, and assemble sarcomeres to form a functional myofiber (Rui et al., 2010; Schejter and Baylies, 2010).

Tropomyosin is a sarcomeric protein that binds thin filament actin to regulate contractions. Outside of the sarcomere, Tropomyosin also directs cytoskeletal dynamics in migratory cells. Membrane protrusions at the leading edge of migrating cells are driven by actin- polymerizing proteins in the lamellipodia, and the physical force for cell movement is derived from lamellar expansion (Ponti et al., 2004). Tropomyosins are essential actin-stabilizing proteins in the lamella that act in concert with the actin-polymerizing proteins at the leading edge, including the Wiskott–Aldrich syndrome proteins (WASP) and the actin-related proteins (ARP)2/3 complex, to drive both normal and metastatic cell migration (Bugyi and Carlier, 2010; Gross, 2014). With respect to myogenesis, the WASP and ARP2/3 protein complexes are essential for myoblast fusion (Baas et al., 2012; Berger et al., 2008; Richardson et al., 2007), and WASP proteins promote myoblast migration (Kawamura et al., 2004). Although the mechanisms that direct myotube elongation are thought to mimic those of migratory cells (Bongiovanni et al., 2012), the role of Tropomyosin during myotube elongation has not been characterized.

We previously identified the RNA-binding protein Hoi polloi (Hoip) in a screen for regulators of Drosophila myotube elongation (Johnson et al., 2013). hoip mutant embryos are largely paralytic due to defects in myotube elongation, as well as defects in myoblast fusion and sarcomeric gene expression. The Drosophila genome encodes two Tropomyosin isoforms, Tm1 and Tm2, and both isoforms were dramatically reduced in hoip embryos. Here, we show Tm2 co-localized with F-actin during myotube elongation, and Hoip-responsive sequences in Tm2 mRNA are required for Tm2 protein expression in somatic muscle. Tm2 overexpression rescued hoip myogenic defects by restoring F-actin during myotube elongation, by promoting myoblast fusion, and by enhancing the expression of additional sarcomeric proteins. Tm2 null embryos showed reduced sarcomeric protein expression, and a gain-of-function Tm2 allele disrupted myotube elongation and myoblast fusion. Tropomyosin is therefore an essential regulator of myogenesis prior to sarcomere assembly.

hoip¹ mutant embryos showed two major myogenic phenotypes. A subset of somatic myotubes [including lateral longitudinal (LL1), dorsal oblique (DO3-5), lateral transverse (LT1-4), and lateral oblique (LO1)] failed to elongate (Fig. 1A; see supplementary material Fig. S1A for a diagram of somatic muscle), and sarcomeric RNAs were dramatically downregulated in striated muscles (Johnson et al., 2013). Myoblast fusion was also affected in hoip¹ embryos. A mechanism that could explain the hoip mutant phenotype is that Hoip promotes the expression of cytoskeletal regulatory proteins that direct myogenesis. To identify potential Hoip targets, we re-examined our hoip RNA-seq data for misregulated transcripts with gene ontology (GO) terms associated with cytoskeletal regulation. Twenty-five misregulated transcripts were associated with these GO terms, including Tm1 and Tm2. Tropomyosin protein expression initiated in the somatic mesoderm during founder cell specification and is robustly expressed in elongating myotubes (Fig 1B,C). By contrast, Mhc expression did not initiate until myotube elongation was largely complete (Fig. 1D). We characterized a third sarcomeric protein, Z-band alternatively spliced PDZ-motif protein 66 (Zasp66), and found Zasp66 expression did not initiate until St16 (supplementary material Fig. S1B-D). Thus, the Tropomyosin expression pattern is temporally and spatially consistent with a role for Tropomyosin in myotube elongation, and Tropomyosin expression is temporally distinct from other sarcomeric proteins.

If Tropomyosin is required for myotube elongation, we reasoned that Tropomyosin would co-localize with F-actin at the myotube leading edge. Embryos harboring an endogenous Tm2 protein trap (hereafter Tm2^GFP; Buszczak et al., 2007) were labeled for GFP and F-actin. Tm2^GFP co-localized with F-actin at the myotube leading edge throughout elongation and during muscle attachment (Fig. 1E-G). Tm2 subcellular localization further suggested that Tropomyosin is required for myotube elongation, and that Tm2 interacts with F-actin prior to sarcomere assembly.

hoip expression is not ubiquitous, and within the embryonic mesoderm hoip expression is restricted to the striated muscle lineages and the fat body (Johnson et al., 2013). The human Hoip ortholog NHP2L1 performs multiple cellular functions that include pre-mRNA splicing and ribosomal RNA processing (Schultz et al., 2006). In fact, the crystal structure of NHP2L1 bound to the spliceosomal RNA U4 has been solved (Liu et al., 2007). NHP2L1 also shows sequence similarities to the archaeal ribosomal protein L7Ae (Kuhn et al., 2002), suggesting that Hoip could be a ribosomal component. Our previous studies showed Hoip localizes to both the nucleus and the cytoplasm of elongating myotubes (Johnson et al., 2013). Hoip could thus be required for pre-mRNA splicing, mRNA nuclear export, mRNA stability and localization, or mRNA translation.

To distinguish among these possibilities, we designed an in vivo splicing assay in which a series of Tm2 genomic constructs (Tm2GFP#1-3) were cloned upstream of a C-terminal GFP tag in a UAS transgenic vector (Fig. 2A; supplementary material Fig. S2A). For each construct, GFP expression requires correct splicing of the encoded transcript. We considered the possibility that excessive overexpression of the Tm2 constructs would be sufficient to detect GFP expression in hoip¹ embryos. To increase the sensitivity of the assay, we used random transposition to generate both low-level and high-level expressing lines for a comparative analysis.

The Tm2 genomic constructs were expressed in developing somatic muscles with RP298.gal4. As a control for RP298.gal4 activity in hoip¹ embryos, we assayed somatic muscle GFP expression from UAS.τGFP. RP298.gal4.  As UAS.τGFP. RP298.gal4 is also active in the salivary gland, salivary gland fluorescence was used to normalize somatic muscle transgene expression. Wild-type (WT) and hoip¹ muscles produced comparable levels of τGFP in somatic muscles (Fig. 2B,C; hoip¹ τGFP fluorescence=117.9% of WT, n=42). However, hoip¹ embryos that expressed the low-level Tm2-GFP constructs produced only a fraction of WT GFP fluorescence in somatic muscles (Fig 2D-J,P; Tm2-GFP#1: 18.0%, n=36; Tm2-GFP#2: 24.0%, n=48; Tm2-GFP#3: 24.7%, n=48). By contrast, hoip¹ embryos that expressed the high-level Tm2-GFP constructs in somatic muscles showed GFP fluorescence that was comparable to WT embryos (Fig. 2L-N; supplementary material Fig. S2EB-G; Tm2-GFP#1: 89.3%, n=66; Tm2-GFP#2: 167.6%, n=48; Tm2-GFP#3: 137.9%, n=36). In fact, some hoip muscles showed higher Tm2-GFP fluorescence than WT muscles. We attribute this to the fact that hoip muscles are smaller than WT muscles, and the GFP signal is concentrated over a smaller area. Our splicing assays showed Hoip regulates Tm2 protein expression during myogenesis, but surprisingly suggest that Hoip is not necessary for pre-mRNA splicing.

To test the possibility that Hoip facilitates Tm2 protein expression after pre-mRNA splicing, we generated stable insertions of a C-terminal GFP-tagged Tm2-cDNA under UAS control. Similar to the intron-containing constructs, hoip¹ embryos that expressed a low-level Tm2-cDNA.GFP showed reduced Tm2-GFP expression compared with WT embryos (Fig. 2K,P; 52.1%, n=48), and hoip¹ embryos that expressed a high-level Tm2-cDNA.GFP construct showed WT levels of Tm2-GFP expression (Fig. 2O,P; 125.0%, n=42). Thus, Tm2-GFP expression from a cDNA also required Hoip, which further argues that Hoip regulates Tm2 expression through a splicing-independent mechanism.

To confirm these observations, we generated Flag-tagged Tm2 constructs for expression in Drosophila S2 cells (Fig. 2Q). S2 cells co-transfected with Hoip and either the Tm2 genomic constructs or the Tm2 cDNA construct produced more Tm2 protein than cells co-transfected with the Tm2 constructs and a control vector (Fig. 2Q). Importantly, cells transfected with Hoip did not show enhanced expression of β-tubulin, which argues that Hoip is not a global regulator of protein expression. We could not detect Tm2-Flag protein expression from Tm2#3, which suggests that alternative exon 5a is exclusively selected in S2 cells (data not shown). The Tm2#2 genomic construct contains three exons and two introns, and cells transfected with Hoip did not show enhanced splicing of Tm2#2 transcripts compared with control transfected cells by quantitative real-time PCR (qPCR; Fig. 2R). These in vitro experiments confirmed that Hoip regulates Tm2 protein expression through a splicing-independent mechanism.

To understand if Hoip is required for Tm2 mRNA nuclear export, we assayed Tm2 mRNA localization by in situ hybridization. WT embryos showed robust Tm2 expression in all somatic muscles, whereas hoip¹ embryos expressed Tm2 at low levels in the dorsal and ventral muscle groups, but not in the lateral muscles (Fig. 3A,B). hoip¹ embryos that expressed the low-level Tm2-GFP#3 construct showed significant Tm2 mRNA in the somatic muscle cytoplasm, but Tm2-GFP protein was largely undetectable (Fig. 3C). However, hoip¹ embryos that expressed the high-level Tm2-GFP#3 showed robust Tm2-GFP protein expression in somatic muscles (Fig. 3D). Tm2 mRNAs are therefore exported from the nucleus in hoip¹ embryos, but the transcript is not robustly translated when expressed at low levels.

RNA-binding proteins typically act outside of the coding region to regulate mRNA translation and stability. However, our Tm2 transgenic constructs contained exogenous 5′ and 3′ untranslated regions (UTRs). We suspected that Hoip must act within the coding region to regulate Tm2 protein expression. To identify Hoip- responsive sequences, we co-transfected S2 cells with Hoip and a series of Tm2 coding region fragments. These assays showed that the 5′ 260 bp of the Tm2 mRNA are Hoip responsive (supplementary material Fig. S3). We deleted these sequences from the Tm2 cDNA (Tm2-ΔcDNA), and found S2 cells transfected with Tm2-ΔcDNA alone did not produce as much Tm2 protein as cells transfected with full-length Tm2-cDNA (Fig. 3E). In addition, cells co-transfected with Hoip and Tm2-ΔcDNA did not express robust Tm2Δ protein (Fig. 3E). To confirm these observations in vivo, we expressed Tm2-ΔcDNA.GFP with RP298.gal4. Although Tm2Δ.GFP protein was clearly visible in the salivary gland, Tm2Δ.GFP protein was largely undetectable in somatic muscle (Fig. 3F-H). In addition, Tm2-ΔcDNA.GFP transcripts were present in the cytoplasm of somatic muscles at levels comparable to full-length Tm2-cDNA.GFP transcripts (Fig. 3I,J). Hoip therefore acts within the Tm2 coding region to direct Tm2 protein expression in somatic muscles.

Control hoip¹ embryos that expressed τGFP in the somatic musculature showed a dramatic reduction in the number of completely elongated LL1 (16.1%, n=57) and LT1-3 (3.9%, n=147) muscles compared with WT embryos (Fig. 4A,B,O). hoip¹ embryos that expressed low-level Tm2-GFP constructs in the somatic musculature showed myotube elongation defects similar to hoip¹ embryos that expressed τGFP (Fig. 4C-H; Tm2-GFP#1: LL1=16.1%, LT1-3=15.3%; Tm2-GFP#2: LL1=26.7%, LT1-3=28.2%; Tm2-GFP#3: LL1=11.1%, LT1-3=6.8%). Remarkably, hoip¹ embryos that expressed high-level Tm2-GFP constructs in the somatic musculature showed a significant recovery in the number of elongated myotubes compared with control hoip¹ embryos (Fig. 4I-N; Tm2-GFP#1: LL1=50.4%, n=71, LT1-3=39.1%, n=213; Tm2-GFP#2: LL1=59.8%, n=67, LT1-3=55.4%, n=204; Tm2-GFP#3: LL1=69.5%, n=53, LT1-3=14.6%, n=159).

To extend this observation, we assayed Tm2-GFP expression and myotube elongation in embryos with two copies of the low-level Tm2-GFP insertions. hoip¹ embryos with two copies of the low-level expressing Tm2-GFP genomic constructs produced more Tm2-GFP than hoip¹ embryos with just a single copy (supplementary material Fig. S4A-I; Tm2-GFP#1: 497.6%, n=48; Tm2-GFP#2: 703.8%, n=48; Tm2-GFP#3: 599.4%, n=66; Tm2-cDNA.GFP: 161.3%, n=66; percent relative to single copy). In addition, hoip¹ embryos with two copies of each Tm2-GFP showed a significant recovery in the number of elongated myotubes compared with hoip¹ embryos with just a single copy (supplementary material Fig. S4I; Tm2-GFP#1: LL1=336.5%, n=62, LT1-3=290.0%, n=186; Tm2-GFP#2: LL1=304.3%, n=66, LT1-3=155.9%, n=171; Tm2-GFP#3: LL1=598.0%, n=74, LT1-3=169.3%, n=222; Tm2-cDNA.GFP: LL1=251.7%, n=73, LT1-3=326.3%, n=219). Tm2 therefore promotes myotube elongation in hoip¹ embryos.

Tropomyosin localizes to F-actin in the lamella of migratory cells, and the contractile F-actin/myosin lamellar network provides the physical force to drive cell movement (Ponti et al., 2004; Ridley et al., 2003). As Tm2 co-localized with F-actin during myotube elongation (Fig. 1E,F), we predicted that Tm2 regulates F-actin at the myotube leading edge. Compared with WT embryos (Fig. 5A-C,P; St12 DO5=9.5; St13 DO5=14.9; St15 DO5=9.9; arbitrary units, n≥12 myotubes per stage), hoip¹ embryos showed a significant decrease in leading edge F-actin (Fig. 5D-F,P; St12 DO5=3.9; St13 DO5=3.5; St15 DO5=4.5; n≥18). However, hoip¹ embryos that expressed the high-level Tm2-GFP#3 assembled significantly more leading edge F-actin than hoip¹ embryos (Fig. 5G-I,P; St12 DO5=5.0; St13 DO5=7.2; St15 DO5=9.0; n≥21). Thus, Tm2 promotes F-actin assembly at the myotube leading edge.

Similar to vertebrates, Drosophila embryos express muscle-specific actin isoforms. Actin 57B (Act57B) is a component of somatic muscle thin filaments, and Act57B expression initiates as early as St11 (Kelly et al., 2002). Compared with WT embryos, Act57B mRNA levels were dramatically reduced in hoip¹ embryos (Fig. 5J,K; supplementary material Fig. S5A,B), and hoip¹ embryos that expressed Tm2-GFP#3 showed improved Act57B expression in somatic muscles (Fig. 5L; supplementary material Fig. S5C). Importantly, the probe used to detect Act57B binds to the highly divergent 3′UTR that distinguishes it from other actin isoforms (Kelly et al., 2002). To understand how Hoip and Tm2 regulate Act57B, we used the minimal reporter gene Act57B.-593/+2.nlacZ to assay transcriptional activity (Kelly et al., 2002). Compared with WT embryos (Fig. 5M,Q; supplementary material Fig. S5D; St13=6.7; St16=8.4; arbitrary units, n≥150 nuclei per stage), Act57B.-593/+2 reporter expression was reduced in hoip¹ embryos (Fig. 5N,P; supplementary material Fig. S5E; St13=2.7; St16=5.7; n≥150). hoip¹ embryos that expressed Tm2-GFP#3 showed a slight, yet significant, increase in reporter gene expression compared with hoip¹ embryos (Fig. 5O,P; supplementary material Fig. S5F; St13=4.6; St16=6.9; n≥150).

F-actin performs multiple functions during myogenesis beyond myotube elongation. For example, F-actin stress fibers are thought to provide the template for sarcomere assembly (Friedrich et al., 2012). As Tm2 restored Act57B transcription and leading edge F-actin in hoip¹ embryos, we suspected that Tm2 might also regulate F-actin during sarcomere assembly. By St17, WT somatic muscles have assembled thin filament F-actin. However, F-actin was absent from the somatic muscles of hoip¹ embryos and hoip¹ embryos that expressed the low-level Tm2-GFP#3 (supplementary material Fig. S5H,N). Strikingly, somatic muscles of hoip¹ embryos that expressed the high-level Tm2-GFP#3 assembled thin filament F-actin (supplementary material Fig. S5I). Tm2-GFP also localized to sarcomeres in WT St17 embryos (supplementary material Fig. S5J,L). Consistent with our F-actin results, Tm2-GFP from the high-level expressing line localized to sarcomeres in hoip¹ embryos whereas Tm2-GFP from the low-level expressing line did not (supplementary material Fig. S5K,M). It is possible that the low-level expressing line did not produce enough Tm2-GFP for us to detect sarcomeric localization in hoip¹ embryos. However, these embryos also lacked thin filaments so the minimal Tm2-GFP produced likely lacked a substrate for localization. In either case, high levels of Tm2 restored sarcomere assembly in hoip¹ embryos.

Knockdown of individual sarcomeric proteins can disrupt sarcomere assembly (Rui et al., 2010). We had previously shown that sarcomeric RNAs were downregulated in hoip¹ embryos at St12-13, and that Mhc protein expression was greatly reduced in St16 hoip¹ embryos (Johnson et al., 2013). As Tm2 restored sarcomere assembly in hoip¹ embryos, we expected that Tm2 would also promote the expression of sarcomeric RNAs. Compared to WT embryos, the average ratio of sarcomeric mRNA levels decreased in St17 hoip¹ embryos [Act57B, 0.44; α-actinin (Actn), 0.49 Myosin heavy chain (Mhc), 0.38; Myosin light chain 1 (Mlc1), 0.52; Mlc2, 0.16; Troponin C at 47D (TpnC47D), 0.42 and Tm2, 0.61; Fig. 5Q]. However, the average ratio of sarcomeric mRNA levels in hoip¹ embryos that expressed Tm2-GFP#3 was comparable to WT embryos (Act57B, 1.2; Actn, 1.5; Mhc, 1.7; Mlc1, 2.4; Mlc2, 1.9; TpnC47D, 1.9; Tm2, 3.4; Fig. 5Q). Consistent with these qPCR results, Mhc protein expression, Mhc mRNA expression, and TpnC47D mRNA expression were also restored in hoip¹ somatic muscles that expressed Tm2-GFP#3 (Fig. 6A-I).

In addition to myotube elongation and sarcomere assembly, F-actin plays a key role in myoblast fusion (Schejter and Baylies, 2010). Our Act57B reporter gene experiments showed hoip¹ embryos that expressed Tm2-GFP#3 contained more β-galactosidase (β-gal)-positive nuclei than hoip¹ embryos (Fig. 5N,O). This result suggested that Tm2 promotes myoblast fusion in hoip¹ embryos. The reporter gene RP298.nlacZ is expressed in muscle founders, and subsequently in nascent myotubes, and can be used to measure myoblast fusion. hoip¹ embryos that expressed Tm2-GFP#3 showed more β-gal-positive myonuclei than hoip¹ embryos (Fig. 6J). Importantly, myonuclei number in the DO2 muscle was reduced in hoip¹ embryos compared with control embryos, even though the DO2 muscle often elongates (Fig. 6K,L; WT=9.5 DO2 nuclei, hoip¹=2.5, n≥28). hoip¹ embryos that expressed Tm2-GFP#3 showed an increased number of DO2 myonuclei compared with hoip¹ embryos (Fig. 6M; 4.6 nuclei/DO2, n=30). Tm2 therefore promotes myoblast fusion in hoip¹ embryos.

One mechanism that could explain Tm2-mediated gene expression is that Tm2 regulates muscle size, and muscle size in turn dictates sarcomeric gene expression. To test this possibility we normalized Mhc and Tropomyosin protein expression to somatic muscle size, which we refer to as the expression index. The expression index for Mhc and Tropomyosin was significantly reduced in hoip¹ embryos compared with controls. hoip¹ embryos that expressed Tm2-GFP#3 showed a significantly greater Mhc and Tropomyosin expression index than hoip¹ embryos (Fig. 6J). Therefore, the difference in Mhc expression between hoip¹ and hoip¹ Tm2-GFP#3 rescued embryos is not simply the result of an increase in muscle size.

Our rescue studies showed that Tm2 promotes myoblast fusion, myotube elongation, and sarcomeric gene expression in hoip¹ embryos. To further characterize the function of Tm2 during myogenesis, we generated a Tm2 null mutation (Tm2^Δ8-261). The Drosophila genome encodes only two Tropomyosin isoforms, Tm1 and Tm2, and both isoforms are deleted by Df(3R)BSC741. Using qPCR, we confirmed St17 Tm2^Δ8-261 homozygous embryos and St17 Tm2^Δ8-261/Df(3R)BSC741 transheterozygous embryos did not express Tm2 (supplementary material Fig. S6A). In addition, these embryos showed reduced Act57B expression compared with control embryos (supplementary material Fig. S6A; FC=0.29 and 0.31), and Act57B expression could be partially restored in Tm2^Δ8-261 embryos that expressed Tm2 under the control of RP298.gal4 (supplementary material Fig. S6A; FC=0.65). The remaining sarcomeric RNAs were expressed at near WT levels in Tm2 mutant embryos. Surprisingly, myotube elongation was largely unaffected in Tm2^Δ8-261/Df(3R)BSC741 embryos, but a subset of transheterozygous embryos showed defects in LT1-3 or LL1 morphogenesis in at least one segment (25%, n=12; Fig. 7A,B). Tm2^Δ8-261/Df(3R)BSC741 embryos also showed a significant reduction in the Mhc expression index compared with controls but somatic muscle size was unaffected (Fig. 7I). Tm2^Δ8-261 embryos that expressed Tm2 showed improved LT1-3 and LL1 morphogenesis (14% of embryos with elongation defects, n=14), and improved Mhc expression compared with transheterozygous embryos (Fig. 7C,I).

One explanation for the lack of a strong elongation phenotype in Tm2 embryos is that Tm2 is maternally contributed. Using qPCR and immunohistochemistry, we confirmed there is a maternal contribution of both Tm1 and Tm2 mRNAs (supplementary material Fig. S6B-F). However, females with Tm2^Δ8-261 or Df(3R)BSC741 homozygous mutant germ lines did not lay eggs. Previous studies showed Tm1 is required for border cell migration in the ovary (Kim et al., 2011), which suggested Tropomyosin performs essential functions in the germ line. Our results further argue that Tm1 and Tm2 are essential for oogenesis.

To further investigate a role for Tm2 during myotube elongation, we generated a dominant Tm2 allele. The human genome encodes four Tropomyosin isoforms (TPM1-4), and congenital myopathies have been associated with dominant Tropomyosin mutations (Marttila et al., 2014; Olson et al., 2001). Tropomyosin is a coiled-coil protein comprised of seven pseudo-repeat domains, and each domain contains an α-zone that interacts with F-actin when muscle is relaxed (Marttila et al., 2014). Several dominant Tropomyosin mutations occur in α-zone residues that are conserved between humans and Drosophila, including E54K (supplementary material Fig. S2). We hypothesized that the E54K mutation would disrupt Tropomyosin function in a dominant fashion, and expressed Tm2^E54K-GFP during myotube elongation with RP298.gal4. Otherwise, WT embryos that expressed Tm2^E54K-GFP in developing somatic muscles showed a significant reduction in the percent of completely elongated myotubes (Fig. 7D-F,J; line#1: LL1=75.0%, LT1-3=80.6%; line#2: LL1=76.4%, LT1-3=81.3%) and in the number of DO2 nuclei (Fig. 7G,H) compared with embryos that expressed WT Tm2.GFP. These studies further demonstrate that Tm2 is required for myotube elongation and myoblast fusion.

Our study has revealed a novel function for Tm2 during myogenesis. Functional rescue experiments showed Tm2 is epistatic to hoip and that Hoip regulates Tm2 protein expression via the coding region. Although Tropomyosins are known regulators of actin dynamics in migratory cells, we found that Tm2 regulates F-actin levels during myotube elongation and sarcomere assembly. In addition, Tm2 promoted myoblast fusion in hoip¹ embryos and regulated Mhc protein expression. To our knowledge, this is the first study to show that Tropomyosins regulate muscle development prior to sarcomere assembly. Lastly, myotubes that expressed the gain-of-function allele Tm2^E54K showed both fusion and elongation defects. This finding might have important implications in understanding Tropomyosin-related myopathies.

Tm2 protein expression is Hoip dependent, and sequences in the 5′ end of the Tm2 coding region are required for Tm2 expression in somatic muscles (Fig. 3E-J). Hoip orthologs direct spliceosome assembly (Schultz et al., 2006), and we had previously shown that an Mhc cDNA restored Mhc protein expression in hoip embryos. These findings suggested that Hoip regulates splicing, so we were surprised to discover that Hoip is not required for Tm2 pre-mRNA splicing or Tm2 mRNA nuclear export. However, our experiments with Mhc and Tm2 cDNAs did produce similar results. In the case of Mhc, we detected Mhc protein expression in hoip embryos homozygous for the Mhc transgene. In the case of Tm2, a high-level expressing insertion or two copies of a low-level insertion produced near-WT levels of Tm2. GFP in hoip embryos (Fig. 2P; supplementary material Fig. S4). These data argue that Hoip promotes robust protein expression from endogenous mRNAs, but this requirement can be overcome in hoip embryos when the transcripts are overexpressed.

Despite these similarities, Tm2 performs a distinct functional role to enhance sarcomeric protein expression. Tm2 rescued Mhc protein expression in hoip embryos (Fig. 6C), but Mhc did not rescue Tropomyosin expression (Johnson et al., 2013). In fact, several sarcomeric mRNAs were enriched in hoip embryos that expressed high levels of Tm2.GFP (Fig. 5Q). Thus, negative regulation of sarcomeric mRNAs is not simply offset by overexpressing any sarcomeric mRNA. Tm2 mutant embryos also showed reduced Mhc protein expression (Fig. 7B), but Mhc mRNA levels were unaffected (supplementary material Fig. S6A). Our results demonstrate that Tm2 performs a specific regulatory function, which is distinct at least from Mhc, to promote sarcomeric protein expression. Although Tm2.GFP ameliorated multiple myogenic defects in hoip embryos, Tm2.GFP was not sufficient to completely restore myogenesis. This incomplete rescue could be due to a number of factors including the spatial and temporal onset of Tm2.GFP expression or that Hoip has at least one additional target mRNA required for myogenesis.

Importantly, Tm2 regulates Mhc expression independent of muscle size (Figs 6J and 7I). Muscle size is thought to be dependent in part on myoblast fusion (Schejter and Baylies, 2010), and we do see a correlation between myonuclei number and muscle size in St16 embryos (Fig. 6J). Tm2.GFP promoted myoblast fusion in hoip embryos (Fig. 6K-L), and Tm2^E54K reduced myoblast fusion when expressed in nascent myotubes (Fig. 7G,H). It is possible that Tm2 regulates sarcomeric protein expression by promoting myoblast fusion, which would increase the number of sarcomeric loci, and in turn transcription of sarcomeric genes. However, the Mhc expression index is significantly different between embryos with reduced muscle size and myonuclei number (Fig. 6J). Although we cannot absolutely rule out the possibility that sarcomeric gene transcription contributes to the phenotypes we have reported, it is clear that muscle size, myonuclei number, and gene transcription alone do not account for Tm2-mediated regulation of sarcomeric protein expression.

In migratory cells, adhesions with the extracellular matrix form and disassemble at the junction of the lamellipodia and the lamella (Ponti et al., 2004). Actomyosin contractions in the lamella use these adhesion sites to move the cell forward. The lamella is characterized by discrete foci of actin polymerization and depolymerization (Ponti et al., 2004). Migrating epithelial cells that overexpress TPM1 lacked a lamellipodia, but showed rapid migration. Mechanistically, TPM1 overexpression concentrated F-actin and myosin II to the lamella and produced more cellular adhesions (Gupton et al., 2005). A similar cellular mechanism appears to direct myotube elongation. Like TPM1 (Gupton et al., 2005), Tm2 localized to punctate foci near the leading edge (Fig. 1E-G). These foci might reflect sites of Tm2-mediated actomyosin contractions. As both TPM1 and Tm2 promote F-actin assembly at the leading edge membrane, the cellular mechanism of myotube elongation likely parallels that of migratory cells. The actomyosin network also directs myoblast fusion (Kim et al., 2015). Here, actomyosin tension in the founder cell promotes pore formation at the fusogenic synapse. It will be interesting to see if Tm2 directs the assembly or even the function of the actomyosin network to promote myoblast fusion.

Although we established that Hoip regulates Tm2 through a largely post-transcriptional mechanism, Act57B.-593/+2 reporter gene expression was reduced in hoip¹ embryos (Fig. 5P), and restored in hoip¹ embryos that expressed Tm2.GFP. Accordingly, leading edge F-actin and thin filament F-actin were downregulated in hoip¹ embryos, and rescued by Tm2.GFP (Fig. 5P; supplementary material Fig. S5). The only known molecular function of Tropomyosins is to bind F-actin, so transcriptional regulation of Act57B is most likely indirect. F-actin polymerization initiates a feedback loop that drives actin transcription in some contexts (Mokalled et al., 2010), so the changes in Act57B transcription we observed are likely a downstream effect of reduced F-actin levels.

Both Tm1 and Tm2 are expressed in developing somatic muscles (BDGP insitu homepage – http://insitu.fruitfly.org/cgi-bin/ex/insitu.pl), and we confirmed these are the only Tropomyosin-encoding genes in the Drosophila genome by BLAST analysis. Tm1 mRNA was enriched in Tm2 mutant embryos at St11 and St12 (supplementary material Fig. S6B; FC=3.1, 2.1), but not at St17 (FC=0.75). These data argue that Tm1 compensates for Tm2 in Tm2 null embryos at the onset of zygotic transcription. Similarly, Tm1 mRNA was enriched in hoip embryos (Fig. 5Q; FC=5.0), and reduced in Tm2.GFP rescued hoip embryos (FC=2.5). However, elevated Tm1 levels alone were not sufficient to induce Tropomyosin protein expression in hoip embryos, whereas high-level Tm2.GFP mRNA produced significant levels of Tm2.GFP protein (Figs 2L-O and 5Q). Hoip is therefore required for the expression of both Tropomyosin protein isoforms, but different mechanisms appear to direct Tm1 and Tm2 protein expression in somatic muscle.

The modest Tm2 zygotic phenotype we characterized is consistent with other maternally contributed genes that regulate myogenesis. For example, rho and Rok single-mutant embryos develop normal somatic muscles whereas rho, Rok double-mutant embryos show myoblast fusion defects (Kim et al., 2015). Although the maternal contribution of Tm1 and Tm2 produced only a fraction of WT transcript levels (supplementary material Fig. S6B), Df(3R)BSC741 embryos showed robust Tropomyosin expression during myotube elongation and this expression persisted through St16 (supplementary material Fig. S6C-F). By contrast, a majority of somatic muscles failed to express Tropomyosin protein in hoip mutant embryos (Fig. 6B). Thus, maternally contributed Tm1/2 can be translated in the presence of Hoip to direct largely normal myogenesis.

Dominant mutations in TPM1, TPM2, and TPM3 are associated with congenital myopathies (Marttila et al., 2014) and cardiomyopathies (Karibe et al., 2001; Olson et al., 2001). The TPM1 E54K allele was identified in patients with dilated cardiomyopathy and in vitro thin filament reconstitution experiments show the E54K protein overinhibits actomyosin interactions, which decreases force generation during systolic contraction (Bai et al., 2012). Our Tm2 E54K allele disrupted myogenesis (Fig. 7D-F), which argues that Tm2-mediated actomyosin contractions are indeed required for myotube elongation. A second TPM1 allele, E40K, also interrupts actomysosin contractions (Bai et al., 2012), and TPM2 E41K is associated with congenital myopathies (Marttila et al., 2014). This raises the possibility that myotube elongation is affected in patients with TPM2-associated congenital myopathies. It will be of particular interest to characterize additional Tropomyosin alleles in vivo to better understand the disease mechanisms that underlie these myopathies.

The stocks used in this study were hoip¹ (Johnson et al., 2013), P{Gal4-kirre^rP298} and P{lacZ-kirre^rP298} (Nose et al., 1998), P{PTT-GC}Tm2^ZCL2456 and P{PTT-GA}Zasp66Z^CL0663 – referred to as Tm2^GFP and Zasp66^GFP protein traps (Buszczak et al., 2007), P{Act57B-lacZ.-593+2} (Kelly et al., 2002), P{UAS-mCD8.ChRFP}, P{UAS-mCD8.eGFP}, P{UAS-τGFP}, Df(3R)BSC741, P{neoFRT}82B, and P{ovoD1-18}3R (Bloomington Stock Center). Cyo, P{Gal4-Twi}, P{2X-UAS.eGFP}; Cyo, P{wg.lacZ}; and TM3, P{Gal4-Twi}, P{2X-UAS.eGFP} balancers were used to identify homozygous embryos.

Tm2 genomic transgenes were constructed by cloning genomic DNA or cDNA (RE15528) PCR products into pEntr (Life Technologies), and then recombined into a destination vector (TWG and AWG) as described (https://emb.carnegiescience.edu/drosophila-gateway-vector-collection). Transgenic insertions were generated by standard methods (Rainbow Transgenic Flies). Tm2-ΔcDNA deletion constructs were generated by PCR sewing as described (Johnson et al., 2013). Entry clones were fully sequenced.

The Tm2^Δ8-261 allele was generated as described (Gratz et al., 2013), and homozygous embryos were sequenced to confirm the Tm2^Δ8-261 mutation. Germ line clones were made by standard methods.

Antibodies used include anti-MHC (Kiehart and Feghali, 1986), anti-Tropomyosin (Abcam, MAC141), anti-β-Galactosidase (Promega, Z378A) and anti-GFP (Torrey Pines Laboratories; TP401). HRP-conjugated secondary antibodies and the TSA system (Molecular Probes; T20922, T20913) were used to detect primary antibodies. Antibody staining and in situ hybridization were performed as described (Johnson et al., 2013). RE15528 and LP10264 were used as templates for the Tm2 and TpnC47D in situ probes, respectively. Mhc and Act57B probes were generated from the templates described (Johnson et al., 2013; Kelly et al., 2002). Texas Red-conjugated phalloidin (Molecular Probes) was used to detect F-actin in fixed, hand devitellinated embryos.

Images were generated with an LSM700 confocal microscope (Zeiss). Control and mutant embryos were prepared and imaged in parallel where possible. Confocal imaging parameters were maintained between genotypes throughout the study. Normalized expression was calculated as mean fluorescence intensity over an entire DO2 muscle relative to salivary gland fluorescence (Tm2-GFP) or visceral muscle fluorescence (anti-Tm, anti-Mhc). Myotube leading edge F-actin levels were determined by normalizing mean fluorescence at the leading edge relative to internal fluorescence. Time-lapse microscopy was used to identify the position of the nascent myotubes prior to F-actin measurements (supplementary material Fig. S7). Act57B.-593/+2.nlacZ fluorescence was measured in single nuclei and normalized to background fluorescence. Muscle size was determined by outlining individual DO2 muscles to obtain an area. All measurements were performed with Zen2011 software (Zeiss).

Schneider 2 (S2) cells were grown in Schneider's media supplemented with 10% FBS and pen/strep. Cells were passaged biweekly and split the day prior to transfection. On the day of the transfection, cells were seeded to a density of 1.0×10⁶ cells/ml in Schneider's media with FBS. 0.8 ml of cells were transferred to one well of a 12-well plate. Transfections were performed with Effectene according to the manufacturer's instructions. 48 h post-transfection, whole cell extracts were made by pelleting cells and resuspending in lysis buffer (20 mM Tris pH 7.5, 150 mM NaCl, 1% Triton X-100, and protease inhibitors). Cells were lysed on ice for 10 min, followed by 10 min centrifugation. Western blots were performed as described (Mokalled et al., 2010) and imaged using the ChemiDoc XRS+ system (BioRad). A minimum of 3 blots was performed from independent transfections for each experiment shown.

Gateway technology was used to generate tagged Hoip constructs using the hoip cDNA clone (RE51843) as a PCR template.

Staged embryos were dechorionated and hand sorted to isolate homozygous mutants. RNA was extracted from embryos or S2 cells using TRizol, and cDNA was generated using Superscript III (Life Technologies). qPCR was performed with SYBR Select Master Mix using an ABI Prism 7000 (Life Technologies). Forward and reverse primers were designed to exons separated by at least one intron, except for splicing assay primers. Here, forward primers were designed to an exon and reverse primers were designed to the downstream intron (to detect unspliced transcripts) or to span the downstream splice donor/acceptor sites (to detect spliced transcripts). qPCR reactions were run in triplicate and normalized to RpL32 or GAPDH. See supplementary material Table S1 for primer sequences.

We thank Richard Cripps for reagents, Mayssa Mokalled for insights and discussions throughout this study and for critical reading of the manuscript, and Brenna Clay for assistance with embryology.