Post-transcriptional regulation of myotube elongation and myogenesis by Hoi Polloi

The transcriptional regulatory networks that direct muscle precursor cell specification and the expression of muscle structural genes have been well defined. However, the possible post-transcriptional contribution to mesoderm development is only beginning to come to light (Biedermann et al., 2010; Toledano-Katchalski et al., 2007; Yarnitzky et al., 1998). The unique properties of Drosophila, including external development and an extensive array of genetic tools, have allowed the discrete cellular processes directing muscle development to be dissected in detail (Guerin and Kramer, 2009a; Schejter and Baylies, 2010; Schnorrer and Dickson, 2004).

Embryonic somatic muscle development in Drosophila is a multistep process that initiates with the specification of founder cells from a field of myogenic competent cells in the mesoderm (Carmena et al., 1995; Jagla et al., 1998). Founder cells express a unique set of muscle identity genes, encoding transcription factors, that direct differentiation into one of 30 somatic muscles (de Joussineau et al., 2012). Once specified, muscle founders begin the process of migration and elongation that can be divided into three phases (Schnorrer and Dickson, 2004). During the first phase, founder cells migrate to their correct position within the segment. The second phase begins when the founder cells initiate myoblast fusion and form polarized myotubes that elongate along a single axis. The myotubes then form extensive filopodia in the direction of initial polarity, presumably in response to guidance cues from tendon cells in the overlying epidermis (Guerin and Kramer, 2009a; Schnorrer and Dickson, 2004). The center of the myotube remains localized while the ends of the myotube elongate towards their respective muscle-attachment sites (Schnorrer and Dickson, 2004). The final phase of elongation initiates when the myotube ends reach their muscle attachment sites and filopodia no longer form. The myotube then localizes integrin-mediated adhesion complexes with the overlying tendon cells to establish strong myotendinous junctions (Schejter and Baylies, 2010).

The mechanisms that control myotube elongation during somatic muscle morphogenesis are poorly understood. Slit is the single guidance molecule known to direct both myotube elongation and target site recognition, but loss of Slit modestly affects the elongation of only a subset of myotubes (Kramer et al., 2001). Nascent myotubes must undergo extensive cytoskeletal rearrangements during elongation, and recent work has focused on the role of microtubule dynamics in this process (Folker et al., 2012; Guerin and Kramer, 2009b). Tumbleweed (Tum) is a Rac family GTPase-activating protein that becomes localized to the nuclear periphery via its association with the microtubule-associated protein Pavarotti (Pav). Loss of pav or tum disrupts microtubule polarity and polarized growth, mislocalizes the minus-end microtubule nucleator γ-tubulin and causes modest myotube elongation defects (Guerin and Kramer, 2009b). A second regulator of microtubule dynamics, Dynein heavy chain (Dhc64C), is also required for myotube elongation but its role is restricted to the final stages of elongation (Folker et al., 2012). Although microtubule dynamics plays a key role in the process, the mechanisms that initiate myotube elongation and the downstream targets of intracellular messenger proteins, such as Tum, remain largely unknown.

The cellular events that regulate myotube morphology are distinct from the molecular processes that direct terminal differentiation and structural gene expression. Embryos defective in myoblast fusion express Myosin Heavy Chain (MHC) in unfused mononucleate founder cells; this striking phenotype has been exploited in genetic screens to identify novel regulators of myoblast fusion (Chen and Olson, 2001). Components of the sarcomere, the basic unit of muscle contraction, are subject to extensive post-transcriptional regulation. For example, Drosophila MHC is encoded by a single genomic locus that can produce 480 unique protein isoforms (Zhang and Bernstein, 2001). These isoforms encode variant regions of the MHC globular head and provide diversity in contractile performance (Kronert et al., 1994). However, the RNA-binding proteins (RBPs) that regulate the production of different MHC isoforms have not been identified.

In a screen for genes that regulate somatic muscle morphology in Drosophila, we identified the putative RBP Hoi polloi (Hoip). hoip embryos show two dramatic phenotypes: myotube elongation does not initiate, even though founder cell specification and myoblast fusion initiate normally and striated muscles fail to express multiple sarcomeric proteins, including MHC and Tropomyosin (Tm). hoip expression is tissue specific and within the muscle lineage is restricted to striated muscle precursors. By RNA deep sequencing (RNA-seq), we found that known regulators of myotube elongation are expressed correctly in hoip mutant embryos, suggesting Hoip orchestrates a previously unrecognized post-transcriptional mechanism to initiate elongation. Functional rescue experiments demonstrate that Hoip directs pre-mRNA splicing during myogenesis. The human Hoip ortholog NHP2L1 can rescue the hoip phenotype in Drosophila, and morpholino (MO) knockdown experiments in zebrafish indicate that nhp2l1 is a conserved essential regulator of myogenesis. This is the first study to show a tissue-specific role for Hoip or its orthologs in vivo, to identify a robust genetic block in the second phase of myotube elongation, and to address post-transcriptional regulation of sarcomeric gene expression by a putative RBP during Drosophila embryogenesis.

All stocks were obtained from the Bloomington Stock Center unless otherwise noted. The stocks used in this study were: Df(2L)ED90, Df(2L)ED678, Df(2L)Exel6024, Df(2L)Exel7043, Df(2L)Exel7042, Df(2L)Exel8041, Df(2L)BSC216, Df(2L)BSC108, Df(3R)Exel6191, P{lacW}hoip^k07104, P{Mhc.EMB} (Wells et al., 1996), Mhc¹ (Wells et al., 1996), P{Gal4-kirre^rP298}, P{kirre^rP298.lacZ} (Nose et al., 1998) and the Baylor P-element Mapping Kit (Zhai et al., 2003). The Cyo, P{Gal4-Twi}, P{2X-UAS.eGFP}; Cyo, P{wg.lacZ}; and TM3, P{ftz.lacZ} balancers were used to identify homozygous embryos.

Isogenic, starved w¹¹¹⁸ males were fed 25 mM EMS in 1% sucrose overnight and mated en masse as shown (supplementary material Fig. S1A). Mapping of hoip¹ was performed as described previously (Zhai et al., 2003).

Antibodies used include anti-Mef2 (Lilly et al., 1995), anti-Nau (Wei et al., 2007), anti-Tin (Venkatesh et al., 2000), anti-MHC (Kiehart and Feghali, 1986), anti-Kr (Kosman et al., 1998), anti-MF20 (DSHB), anti-Tropomyosin (Abcam, MAC141), anti-GFP (Torrey Pines Laboratories), anti-βPS (DSHB), anti-Talin (DSHB), anti-22C10 (DSHB) and anti-βgal (Promega). Alexa488-, Alexa633- and HRP-conjugated secondary antibodies in conjunction with the TSA system (Molecular Probes) were used to detect primary antibodies. Antibody staining and in situ hybridization was performed as described (Johnson et al., 2007). Mef2 was directly conjugated with Zenon633 (Molecular Probes) for co-labeling with rabbit primary antibodies. The following templates were used to generate in situ probes: RE51843 (hoip; DGRC) and Mhc.Exon4-6 (a gift from S. Bernstein, SDSU, San Diego, CA, USA). Images were generated with LSM510 and LSM710 confocal microscopes. Control and mutant embryos were prepared and imaged in parallel.

Appropriately aged embryos were dechorionated (Drosophila) or devitellinated (Danio), hand sorted to isolate homozygous mutants where needed, and homogenized in TRizol (Invitrogen). RNA was then extracted as per manufacturer’s specification. cDNA was generated using Superscript III (Invitrogen) and qPCR was performed with SYBR Green (Promega) using an ABI Prism 7000. Primers were designed to be intron spanning. qPCR reactions were run in triplicate and normalized to RpL32 (Drosophila) or GAPDH (zebrafish) expression. Primer sequences are provided in supplementary material Table S6.

COS-1 cells were transfected with 1 μg of DNA according to manufacturer’s instructions (Fugene6, Roche), maintained for 48 hours, collected and lysed with NP40 lysis buffer. Western blots were performed as described previously (Mokalled et al., 2010).

UAS constructs were generated by subcloning hoip (DGRC RE51843) and NHP2L1 (DF/HCC HSCD00326196) ORFs into pUASt. HA-tagged Hoip was made by PCR amplifying the hoip ORF and cloning the PCR fragment into pEntr (Invitrogen); after sequence verification, L/R clonase (Invitrogen) was used to recombine hoip into THW. For reporter genes, genomic DNA was PCR amplified, cloned into pCRII (Invitrogen), sequence verified, subcloned into pH.Stinger.eGFP (Barolo et al., 2004). Transgenic vectors were injected by standard methods to establish stable transgenic insertions. Multiple independent lines were characterized for each construct. Primer sequences are available upon request. Site-directed mutagenesis to generate Hoip.-225GFP.ΔE was carried out as described previously (Johnson et al., 2011).

The SOLiD Total RNA-Seq Kit was used for RNA purification from 6-10 hour embryos and DNA library construction. Libraries were prepared in duplicate and sequenced on a 5500xl SOLiD Sequencer (Life Technologies) using a paired-end reading strategy. Sequencing reads were mapped to the UCSC reference genome using LifeScope (Life Technologies), then assembled and quantified using the Cufflinks algorithm (Trapnell et al., 2010). Assembled sequence reads were visualized using the Integrative Genomics Viewer (Robinson et al., 2011) and GO analysis was performed using the DAVID 6.7 bioinformatics resources (Dennis et al., 2003).

Fertilized, one-cell stage Tg(α-actin:GFP) zebrafish embryos were injected with 0.08 ng or 0.8 ng of nhp2l1b ATG-MO (GGTTCACTTCAGCTTCAGTCATCTT) or 0.8 ng of Cntrl-MO (Gene Tools). At 14 hpf, embryos were live imaged for GFP, fixed and analyzed for GFP and MF20 expression, or used for RNA isolation and qPCR analysis, by standard methods. Similar experiments were performed for nhp2l1b 5′UTR-MO (TTACTTAATAACACACGGTCCTCTC). Annealed oligonucleotides encoding the ATG-MO target sequence were cloned into 5′RV EGFP T7TS (Small et al., 2005). Linearized template was transcribed with T7 mMessage machine (Ambion) to generate injectable RNA. Control and morphant embryos were prepared and imaged in parallel.

We performed an EMS genetic screen to identify novel mutations that specifically affect somatic muscle morphology. Our screening strategy employed two reporter genes: MHC.τGFP, which expresses a membrane-localized GFP in embryonic somatic muscles (Chen and Olson, 2001); and Hand.nGFP that expresses a nuclear-localized GFP in cardioblasts (CBs) by embryonic stage 12 (St12) (Johnson et al., 2011). The Hand.nGFP reporter served as a control to distinguish mutations that affected early mesoderm patterning events and cardiac cell fate specification, allowing us to focus on mutations that specifically affected somatic muscle morphology. We screened ∼10,000 genomes and identified 96 mutations on the second chromosome that caused somatic muscle defects (supplementary material Fig. S1A). We mapped two mutations to Kon-tiki (Kon), a known regulator of target site recognition (Schnorrer et al., 2007), and a second mutation to mind bomb 2, a known regulator of muscle attachment (Carrasco-Rando and Ruiz-Gómez, 2008).

One mutation that solely affected the pattern of MHC.τGFP expression mapped to a region of chromosome 2L that contains eight genes (Fig. 1A). Genetic analysis in this region showed that the mutation failed to complement P{lacW}hoip^k07104, so we named the EMS allele hoip¹. The Hoip orthologs Snu13 in yeast and NHP2L1 (Non-Histone Protein 2 Like-1) in humans are RNA-binding proteins that bind noncoding RNAs associated with the spliceosome (Dobbyn and O’Keefe, 2004; Vidovic et al., 2000; Watkins et al., 2002). Compared with live wild-type (WT) embryos (Fig. 1B), somatic muscle morphology was aberrant in live hoip¹ homozygous (Fig. 1C), hoip¹/Df(2L)ED690 (Fig. 1D) and hoip¹/P{lacW}hoip^k07104 transheterozygous embryos (Fig. 1E). In particular, the lateral transverse (LT1-4), lateral longitudinal (LL1), lateral oblique (LO1) and dorsal oblique (DO3-5) somatic muscles showed pronounced membrane extensions towards target sites, yet remained rounded in hoip mutant embryos (Fig. 1C-E; supplementary material Table S1). This failure in myogenesis prevented hoip¹, hoip¹/Df(2L)ED690 and hoip¹/P{lacW}hoip^k07104 embryos from emerging from the chorion after embryogenesis.

The P{lacW}hoip^k07104 insertion was originally identified in a peripheral nervous system (PNS) screen (Kania et al., 1995; Prokopenko et al., 2000). P{lacW}hoip^k07104 embryos showed disorganized dorsal clusters of PNS neurons and axonal path finding defects (Kania et al., 1995). However, the P-element itself was not revertible (Kania et al., 1995) and P{lacW}hoip^k07104 embryos showed global patterning defects that we did not observe in other hoip mutant combinations (supplementary material Fig. S1B,C). We assayed PNS morphology in hoip¹ and hoip¹/P{lacW}hoip^k07104 embryos (n≥3) but could not confirm the previously reported PNS defects (supplementary material Fig. S1D-F). These data suggest a lethal mutation on the P{lacW}hoip^k07104 chromosome outside hoip disrupts a powerful regulator of embryonic patterning that could affect PNS development.

Sequencing the hoip¹ allele revealed a G37E missense mutation (Fig. 1F; supplementary material Fig. S1I) within the predicted Hoip RNA-binding domain (Schultz et al., 2006). Based on the predicted crystal structure of this domain (supplementary material Fig. S1G), the acidic amino acid substitution would be expected to eliminate Hoip RNA-binding activity (Vidovic et al., 2000). A tagged hoip protein harboring the G37E mutation was detectable by western blot in transfected COS-1 cells, demonstrating that hoip¹ is not a protein null mutation (supplementary material Fig. S1H).

To understand the stage of myogenesis regulated by Hoip, we performed time-lapse studies in embryos expressing τGFP under the control of the founder cell driver rp298.gal4. Our analysis focused on the LL, LT, LO and DO muscles, as these muscles were most often disrupted in hoip mutant embryos (supplementary material Table S1). Our analysis began in late St12 embryos that expressed rp298>τGFP in nascent myotubes (Fig. 2A). In wild-type embryos, myotubes showed obvious polarization and had elongated to 50% of segment width after 30 minutes. By 60 minutes, the myotubes had largely completed their extension and created extensive filopodia for attachment site recognition (Fig. 2A; n=2). By contrast, nascent myotubes failed to elongate in hoip¹ embryos after 30 minutes even though the myotubes showed an initial polarity (Fig. 2B). By 60 minutes, myotubes failed to extend to 50% of segment width in hoip¹ embryos and had lost their polarity (Fig. 2B; n=3). These results demonstrate that myotubes failed to elongate and reach their attachment sites in hoip¹ embryos.

To understand whether Hoip controls target site recognition, we repeated time-lapse imaging at higher magnification to document filopodia in detail (supplementary material Fig. S2; n=3). hoip¹ embryos extended filopodia exclusively in the direction of myotube polarity. This phenotype contrasts with that of Kon mutant embryos, which initiate myotube elongation but, instead of orienting filopodia solely toward muscle attachment sites, extend ectopic filopodia in all directions (Schnorrer et al., 2007). Kon is a transmembrane receptor that regulates myotube target site recognition. As hoip myotubes do not phenocopy Kon myotubes, we conclude that Hoip does not control attachment site recognition.

To understand whether tendon cells, which mediate muscle attachment, were specified in hoip¹ embryos, we assayed expression of βPS (myospheroid), an effector of muscle attachment. βPS is expressed in tendon cells and localizes to myotendinous junctions after attachment-site recognition (Martin-Bermudo and Brown, 2000). βPS was clearly detectable in the epidermis of hoip¹ embryos, but showed diffuse localization along the dorsoventral axis compared with wild-type embryos (Fig. 2C,D). As βPS is also expressed in somatic muscle, we next examined Talin expression. Talin is restricted to tendon cells and acts as a linker between βPS and the cytoskeleton. Similar to βPS, Talin is clearly expressed in the epidermis of hoip¹ embryos (Fig. 2E,F). Taken together, these results show that the rounded muscle phenotype in hoip¹ embryos is due to a block at, or prior to, myotube elongation and is not a result of tendon cell mis-specification.

To further characterize the myogenic phenotype in hoip¹ embryos, we examined founder cell specification and myoblast fusion. After specification, founder cells undergo an initial round of fusion that is complete by the end of St12 (Bate, 1990). Subsequent fusion then determines final muscle size and each muscle undergoes a unique number of fusion events. MHC serves as a classic marker for identifying myoblast fusion defects; embryos with defects in the first round of myoblast fusion robustly express MHC in single unfused founder cells (Chen and Olson, 2001). To control against possible dominant mutations on the EMS chromosome, we compared MHC expression in hoip¹ embryos with heterozygous hoip¹/Cyo.lacZ embryos. Strikingly, the somatic musculature of St16 hoip¹ embryos showed almost no MHC protein expression, whereas hoip¹/Cyo embryos showed normal MHC expression and somatic muscle morphology (Fig. 3A,B; Table 1).

Another method for identifying myoblast fusion defects is to quantify the temporal expression of muscle identity genes in the dorsal mesoderm (Chen and Olson, 2001). The identity gene nautilus (nau) is expressed in a subset of founder cells that give rise to the somatic muscles affected in hoip¹ embryos, including DA3, DO3, DO4, DO5, VA1, LO1 (Wei et al., 2007). The number of Nau+ nuclei in hoip¹ embryos was comparable with hoip¹/Cyo.lacZ embryos at St12, but significantly less at St14 (supplementary material Fig. S3A,B). Thus, founder cell specification and the first round of myoblast fusion proceed normally in hoip¹ embryos; however, the later rounds of fusion do not occur.

To confirm this result, we used the founder cell transgene rp298.nlacZ to assay founder cell specification and myoblast fusion. Similar to Nau, the number of lacZ-positive nuclei was comparable between hoip¹ and hoip¹/Cyo.lacZ embryos at St12 (supplementary material Fig. S3C,D), but significantly reduced at St14 and St16 (Fig. 3C-F; supplementary material Fig. S3E-H). However, the fusion defects in hoip¹ embryos were not restricted to the muscles that showed elongation defects. For example, DO1 and DA1 muscles elongated normally in hoip¹ embryos but showed dramatically fewer lacZ-positive nuclei than hoip¹/Cyo.lacZ embryos (Fig. 3C,D). However, LL1 and LO5 were multinucleate in hoip¹ embryos but failed to elongate (Fig. 3E,F).

One explanation for the somatic muscle defects in hoip¹ embryos was that MHC itself is required for myotube elongation. However, embryos homozygous for the null mutation MHC¹ (Wells et al., 1996) showed normal myotube elongation (supplementary material Fig. S3I,J). Sarcomere assembly occurs after myotube elongation and myofiber attachment (Rui et al., 2010) and embryos defective for the first round of myoblast fusion do express MHC (Chen and Olson, 2001). Together, these observations demonstrate that muscle morphogenesis is genetically separable from muscle structural expression and prompted us to define a secondary role for Hoip during myogenesis.

In Drosophila, MHC is the single muscle myosin and is expressed in cardiac, somatic and visceral muscle (Bernstein et al., 1983). Mef2 is also expressed in all myogenic cells and is a direct transcriptional activator of MHC (Bour et al., 1995). Mef2 was expressed at comparable levels in the somatic and cardiac mesoderm of hoip¹ and hoip¹/Cyo.lacZ embryos (Fig. 3A,B,G-L), even though MHC was largely absent from both tissues in hoip¹ embryos (Fig. 3B,H). The expression of a second transcription factor, Tinman (Tin), is restricted to and orchestrates the maturation of a subset of cardioblasts (CBs) (Reim et al., 2005). Cardiac Tin expression was also comparable between hoip¹ and hoip¹/Cyo.lacZ embryos (Fig. 3I-L). As Drosophila CBs are mononucleate, do not undergo elongation, yet fail to express MHC in hoip¹ embryos, we conclude that Hoip regulates muscle maturation (i.e. muscle structural protein expression) independently of myotube elongation.

Even though MHC expression was largely absent from CBs and somatic muscles in hoip¹ embryos, MHC expression in mature visceral muscle was comparable between hoip¹ and hoip¹/Cyo.lacZ embryos (Fig. 3M,N). These results suggest that Hoip performs tissue-specific functions during myogenesis to specifically regulate striated muscle maturation.

To confirm that Hoip regulates myogenesis after founder cell specification, we expressed Hoip with rp298.Gal4 in hoip¹ embryos and assayed MHC expression. Founder cell-specific expression of Hoip was indeed sufficient to rescue myotube elongation and MHC protein expression in somatic muscles of hoip¹ embryos (Fig. 3O; Table 1). Hoip therefore regulates myogenesis after founder cell specification in a mesoderm cell-autonomous manner.

In situ hybridization using a probe antisense to the full-length hoip transcript (Fig. 4A) showed hoip expression initiates at low levels in the mesoderm and endoderm of St9 embryos (supplementary material Fig. S4A-C). Robust expression hoip mRNA could be detected in St11 embryos and, consistent with the tissue-restricted MHC phenotype in hoip¹ embryos, is expressed in the Mef2-expressing cells of the somatic and cardiac mesoderm, but not in the Mef2-expressing cells of the visceral mesoderm (Fig. 4B). hoip mRNA is also expressed in the fat body and the endoderm at this stage, but is absent from the neuroectoderm ventral to the Mef2 expression domain. hoip mRNA continues to be expressed in the somatic musculature throughout embryogenesis (Fig. 4C; supplementary material Fig. S4D,E). We did not detect hoip in the PNS by in situ hybridization.

To confirm the in situ results, we generated a GFP reporter construct that contained 225 bp of genomic DNA upstream of the hoip-coding sequence (Hoip.-225.GFP). This reporter gene directed GFP expression in a pattern that recapitulated hoip mRNA expression in St11 and St13 embryos (Fig. 4D,E). Interestingly, the Hoip.-225 sequence contains a conserved E-box sequence (CANNTG, supplementary material Fig. S4F). Basic helix-loop-helix (bHLH) transcription factors bind E-box sequences and a hoip reporter gene with a mutated E-box (Hoip.-225ΔE.GFP) initiated GFP expression in a manner comparable with Hoip.-225.GFP (Fig. 4F) but did not maintain GFP expression in the mesoderm through St13 (Fig. 4G). These findings demonstrate that hoip is expressed in the striated muscle lineage after precursor cell specification, and that maintenance of hoip expression depends on a conserved E-box sequence that is likely a bHLH target.

We next assayed Hoip localization in the somatic musculature using an HA-tagged Hoip transgene. Surprisingly, Mef2>Hoip-HA embryos showed both nuclear and cytoplasmic localization of Hoip-HA (supplementary material Fig. S5). Hoip may thus perform multiple molecular functions during myogenesis.

The Hoip orthologs Snu13 in yeast and NHP2L1 in humans are spliceosomal RNA-binding proteins (Dobbyn and O’Keefe, 2004; Vidovic et al., 2000; Watkins et al., 2002). In eukaryotes, spliceosomes that contain small nuclear (sn) RNAs are believed to remove intronic sequences from pre-mRNAs, whereas spliceosomes that contain small nucleolar (sno) RNAs orchestrate ordered cleavages along pre-rRNAs. Snu13/NHP2L1 proteins preferentially bind to GA-rich RNA sequences in the kink-turn motif of both snRNAs and snoRNAs (Cléry et al., 2007; Nottrott et al., 1999; Schultz et al., 2006; Vidovic et al., 2000).

We took a non-biased approach to identify potential Hoip targets in the developing mesoderm. As robust hoip expression initiates at St11 and continues at high levels through St13, we performed RNA-seq in St11-13 (6-10 hour) embryos. Our analysis identified 353 transcripts that were differentially expressed and 60 transcripts that were expressed approximately at wild-type levels but inappropriately processed in hoip¹ embryos (supplementary material Tables S2, S3). This RNA-seq analysis also identified the G37E missense mutation in hoip¹ embryos, confirming the initial genomic sequencing data. In addition, 45S pre-rRNA was processed correctly in hoip¹ St11-13 embryos, suggesting that Hoip does not regulate ribosome biogenesis during these stages of development (supplementary material Fig. S6A). These in vivo results demonstrate that Hoip is not required to process all pre-mRNA or pre-rRNA transcripts during embryogenesis.

We analyzed the misregulated transcripts in hoip¹ embryos by Gene Ontology (GO) functional annotation clustering and found the most significant cluster associated with the GO term Contractile Fiber (Fig. 5A). Strikingly, transcripts within the Contractile Fiber cluster (Table 2) include Mhc and other sarcomere components, including inflated (if), Myosin light chain 2 (Mlc2), Tropomyosin 2 (Tm2), Troponin C at 47D (TpnC47D) and Troponin C at 73F (TpnC73F). We confirmed the RNA-seq data for these sarcomeric genes by quantitative PCR (qPCR) and found that each transcript was dramatically downregulated in hoip¹ embryos compared with controls (Fig. 5B). The RNA-seq data showed that the embryonic sarcomeric actins Act57B and Act87E, and Mef2, the only known robust transcriptional regulator of terminal muscle differentiation genes in Drosophila, were expressed at wild-type levels in hoip¹ embryos (supplementary material Fig. S6B, Table S2). The developmental time point of our RNA-seq coincided with the onset of muscle structural gene expression at St12. However, hoip¹ embryos fail to express MHC protein at all developmental stages (Fig. 3B). These observations suggest that Hoip is required to both initiate and maintain muscle structural gene expression during embryogenesis.

We examined 22 genes experimentally shown to regulate myotube elongation, attachment site recognition or myotendinous junction formation in our RNA-seq data (supplementary material Table S4). Surprisingly, pav expression was not changed in hoip¹ embryos, whereas tum was upregulated (fold change=2.04). However, tum overexpression does not affect myotube elongation (Guerin and Kramer, 2009b). Of the remaining 19 genes, only MSP-300 showed significant downregulation in hoip¹ embryos (fold change=0.27). Unlike hoip¹ embryos, MSP-300 mutant embryos show only a modest somatic muscle phenotype that initiates late in embryogenesis (Rosenberg-Hasson et al., 1996); however, MSP-300 larvae do show defects in nuclear positioning and microtubule organization (Elhanany-Tamir et al., 2012). The 22 genes regulating somatic muscle morphology also showed normal splicing in hoip¹ embryos, further arguing that Hoip regulates the expression of other transcripts to initiate myotube elongation.

The RNA analyses suggested that Hoip processes pre-mRNAs encoding sarcomeric proteins but does not regulate transcription or rRNA processing. To confirm these results, we performed a functional rescue experiment with the MHC^emb transgene (Fig. 5C), which uses the endogenous MHC promoter to express a MHC cDNA specifically in somatic muscle (Hess et al., 2007; Wells et al., 1996). If Hoip regulated either ribosome biogenesis or processing of an mRNA whose protein product activates MHC transcription, then MHC^emb would not be expected to rescue MHC protein expression in hoip¹ embryos. However, if Hoip acts post-transcriptionally to splice the MHC pre-mRNA, then the MHC cDNA, which is expressed from the MHC^emb transgene, would generate a functional mRNA that would be appropriately translated in hoip¹ embryos.

Indeed, the MHC^emb transgene restored MHC protein expression in somatic but not cardiac muscle of hoip¹ embryos (Fig. 5D-F,I; supplementary material Fig. S7). This experiment corroborated our MHC.τGFP expression studies in which the endogenous MHC promoter directed GFP protein expression throughout the somatic mesoderm of hoip¹ embryos (Fig. 1C-E). In addition, the MHC^emb transgene did not restore somatic muscle morphology in hoip¹ embryos, further confirming that Hoip regulates myotube elongation independent of MHC expression. We conclude Hoip is required to perform at least one splice in the MHC pre-mRNA and functions independently of ribosome biogenesis to direct somatic muscle maturation.

Two separate transgenes harboring MHC promoters (MHC.τGFP and MHC^emb) were able to direct cDNA expression in hoip¹ embryos (Fig. 1C, Fig. 5F). However, MHC mRNA was nearly undetectable in hoip¹ embryos by RNA-seq and qPCR (Fig. 5B). We assayed MHC mRNA localization by in situ hybridization and found only punctate, nuclear MHC mRNA localization in hoip¹ embryos, even though wild-type embryos showed MHC mRNA localized throughout the myofiber (Fig. 5G,H). These transgenic and in situ results clearly demonstrate that MHC is transcribed in hoip¹ embryos but that the transcript fails to translocate out of the nucleus.

The remarkable homology between Hoip and human NHP2L1 (supplementary material Fig. S1I) suggested that the function of Hoip during myogenesis is conserved across species. To test this hypothesis, we expressed human NHP2L1 in founder cells of hoip¹ embryos with rp298.gal4 and assayed MHC protein expression. Although human NHP2L1 did not rescue hoip¹ embryos as effectively as Drosophila Hoip (Table 1), we observed a significant restoration of myotube elongation and MHC expression in the somatic musculature, demonstrating that Hoip and human NHP2L1 can perform similar functions during myogenesis (Fig. 6A,B).

Zebrafish nhp2l1b is expressed in the paraxial mesoderm at 10 hour post-fertilization (hpf) and throughout the myotome at 19 hpf (Thisse and Thisse, 2001). We asked whether nhp2l1b is essential for zebrafish muscle development using two independent MOs to knockdown endogenous nhp2l1b. One MO targets the nhp2l1b translational start site with 100% identity (ATG-MO), but does not target the highly divergent nhp2l1a (supplementary material Fig. S8A-C). The other MO targets the nhp2l1b 5′UTR 57 bp upstream of the translational start site and shows no sequence similarity to nhp2l1a (supplementary material Fig. S8B). The α-actin:GFP transgenic (Tg) line harbors a skeletal muscle reporter (Higashijima et al., 1997) and Tg(α-actin:GFP) embryos injected with control-MO showed robust GFP expression in the somitic mesoderm by 14 hpf (Fig. 6C). However, both ATG-MO- and 5′UTR-MO-injected embryos showed little or no GFP expression (Fig. 6D,E). The MF20 antibody reacts with all muscle MyHC isoforms, and we observed robust somitic MF20 staining in control-MO but not ATG-MO-injected embryos at 14 hpf (Fig. 6F,G).

Tg(α-actin:GFP) expression showed a dose-dependent response to MO concentrations (Fig. 6H). Sixty-five percent (n=259) of embryos injected with 0.08 ng ATG-MO and 96% (n=233) of embryos injected with 0.8 ng ATG-MO failed to initiate α-actin:GFP expression; 43% (n=58) of embryos injected with 0.08 ng 5′UTR-MO and 72% (n=87) of embryos injected with 0.8 ng 5′UTR-MO failed to initiate α-actin:GFP expression; 11% (n=302) of embryos injected with 0.8 ng Cntrl-MO showed changes in α-actin:GFP expression. With the exception of skeletal muscle marker expression, 0.8 ng ATG-MO-treated embryos appeared normal through 16 hpf; however, MO treatment induced lethality after 16 hpf (supplementary material Fig. S8F,G). The ATG-MO also blocked eGFP translation when the nhp2l1b ATG-MO target site was placed upstream of the eGFP-coding sequence (supplementary material Fig. S8D,E). Thus, the ATG-MO targets nhp2l1b.

By qPCR, we found that several transcripts encoding sarcomeric proteins were present at reduced levels in ATG-MO 14 hpf embryos, including two slow MyHCs, two troponins and one tropomyosin. Importantly, the expression of other genes essential for mesoderm development, such as mef2 and notch1, was unaffected in ATG-MO embryos. Taken together, these results indicate that the function of Hoip is highly conserved and that nhp2l1b regulates myogenesis in vertebrates.

The results of this study reveal a specific and essential role for the putative RBP Hoip in the control of embryonic muscle development. hoip is expressed in the striated muscle lineage and regulates two distinct processes: myotube elongation and sarcomeric protein expression. Using functional rescue experiments, we have established that Hoip regulates MHC pre-mRNA splicing but not MHC transcription. The human hoip ortholog human NHP2L1 can rescue myogenesis in hoip mutant embryos, and antisense nhp2l1b knockdown blocks muscle development in zebrafish. This study is the first to identify a tissue-specific function for hoip or its orthologs in vivo and highlights the essential role of post-transcriptional gene regulation during tissue morphogenesis.

Non-coding RNAs that function as ‘core’ spliceosome components can have tissue-specific functions in vivo. For example, the mouse genome encodes multiple U2 snRNA genes and the rnu2-8 U2 snRNA is differentially expressed in the mouse nervous system with peak expression levels in the cerebellum (Jia et al., 2012). Rnu2-8 knockout mice show normal splicing of constitutive exons, but incomplete splicing of alternative exons solely within the cerebellum (Jia et al., 2012), and rnu2-8 U2 snRNA is essential for neuron survival in the cerebellum.

The Hoip orthologs Snu13/NHP2L1 proteins have been characterized as ‘core’ spliceosome components that bind the kink turn motif of U4 snRNAs, U3 snoRNAs and U14 snoRNAs (Schultz et al., 2006). However, we found Hoip expression and function is restricted to the striated muscle lineage within the Drosophila mesoderm. RNA-seq and MHC^emb rescue data clearly demonstrate that Hoip is not a global regulator of pre-mRNA splicing or ribosome biogenesis, but acts specifically on a set of RNAs that encode functionally related proteins.

The MHC^emb transgene contains a fully spliced MHC cDNA that encodes exons 2-19 (Wells et al., 1996). The 5′ end of the transgene comprises genomic DNA that initiates 450 bp upstream of the first transcriptional start site and terminates in exon 2. This 5′ sequence contains the necessary enhancer and promoter elements for transgene expression in the somatic mesoderm, as well as three alternative transcriptional start sites. The 3′ end of the transgene contains a complete exon 19 with multiple polyadenylation (poly A) sites, but it does not contain an exogenous poly A signal (i.e. SV40). As MHC^emb rescues MHC protein expression in hoip embryos, Hoip must not regulate transcriptional start site selection or 5′UTR stability. In addition, the endogenous 3′UTR is sufficient to restore MHC protein expression, indicating that Hoip does not regulate poly A site choice, polyadenylation itself or 3′UTR stability. Hoip therefore acts post-transcriptionally to control at least one splicing event in exons 2-19.

The apparent specificity with which Hoip targets sarcomeric RNAs is striking. One explanation for this specificity is that the Hoip paralog Nhp2 fulfills the spliceosome functions that Hoip does not. A more intriguing hypothesis is that Hoip facilitates ribonucleotide modifications in a subset of transcripts. For example, NHP2L1/snoRNA processomes direct 2′-O methylation of ribosomal pre-RNAs (Watkins et al., 2002) and 2′-O methylation has been reported to enhance pre-mRNA splicing in some contexts (Ge et al., 2010). Perhaps Hoip confers a similar modification to sarcomeric RNAs that is permissive for pre-mRNA splicing; in this model, unmodified pre-mRNAs would not be spliced and would degrade in the nucleus. We envision Hoip-mediated modifications to be a rate-limiting step that ensures proper stoichiometry of sarcomeric proteins.

This study also identified myotube elongation defects in hoip embryos. To our knowledge, this is the first mutation reported that blocks the initiation of myotube elongation (supplementary material Table S4). Accordingly, we failed to identify robust misregulation of genes known to regulate myotube elongation, attachment site recognition or myotendinous junction formation in hoip embryos (supplementary material Table S4). The strength of the hoip phenotype suggests that a suite of proteins is required for myotube elongation. A second possibility is that some sarcomeric proteins could be required for myotube elongation. We have shown that MHC does not regulate elongation; however, tropomyosins regulate actin dynamics outside the sarcomere that influence cell polarity and cell outgrowth in Drosophila (Li and Gao, 2003; Zimyanin et al., 2008).

One known regulator of microtubule dynamics during elongation, Tum, is a Rac family GTPase-activating protein (RacGAP). We searched the RNA-seq data for other Rac/Rho family regulators and identified two RhoGAPs and one Rho guanine nucleotide exchange factor (RhoGEF) that were misregulated in hoip embryos (supplementary material Table S5). We also identified two Rab family GTPases, the function of which in synaptic endosome trafficking has been well established (Gurkan et al., 2005). Rab family members are essential for microtubule-dependent outgrowth of tracheal and bristle cells (Nagaraj and Adler, 2012; Schottenfeld-Roames and Ghabrial, 2012) and the myotube guidance molecule Grip localizes to endosomes at the ends of extending myotubes (Swan et al., 2004). Finally, overexpressing the tendon cell regulator Stripe in the ectoderm upregulates a number of novel genes (Gilsohn and Volk, 2010a; Gilsohn and Volk, 2010b), including the transmembrane protein Tetraspanin 42Ea (Tsp42Ea). Tsp42Ea orthologs function in cell motility and signal transduction, and Tsp42Ea was downregulated in hoip embryos. It will be interesting to determine which of these genes are required for myotube elongation in vivo.

The zebrafish Hoip ortholog nhp2l1b is expressed in the paraxial mesoderm during the 10- to 14-somite stage and in the myotome during the 20- to 25-somite stage (Thisse and Thisse, 2001). Thus, the expression and myogenic function of hoip/nhp2l1b appears to be conserved in vertebrates. Despite being one of the most intensely studied developmental systems, skeletal muscle continues to reveal novel developmental mechanisms. In the future, it will be of particular interest to characterize the interactions between transcriptional and post-transcriptional mechanisms that coordinate final muscle morphology and function.

We appreciate Mike Buszczak for insights and discussions throughout this study. We thank Sandy Bernstein, Elizabeth Chen, Richard Cripps and Bruce Paterson for reagents.