Troponin I is required for myofibrillogenesis and sarcomere formation in Drosophila flight muscle

In recent years significant progress has been made in understanding how different sarcomeric proteins interact during development, myofibril assembly and muscle contraction. Formation of the highly ordered repetitive muscle cytoarchitecture is a complex process and genetic defects produce different forms of myopathy (reviewed in Clark et al., 2002; Redwood et al., 1999). It is well established that muscle contraction is regulated by changes in the concentration of Ca²⁺. The troponin-tropomyosin (Tn-Tm) complex, which is formed by three different troponin polypeptides, T, I and C (TnT, TnI and TnC), together with tropomyosin (Tm), regulates acto-myosin interactions in response to neurally stimulated intracellular release of Ca²⁺ ions. In the absence of Ca²⁺, TnI inhibits the generation of acto-myosin forces during muscle contraction (reviewed in Clark et al., 2002; Geeves and Holmes, 1999; Gordon et al., 2000). Although the role of TnI during muscle contraction is well documented, its role during muscle development has not been studied in detail. Two major reasons for this are that loss of function mutations are lethal to individuals (see nemaline TnI nulls) and in vertebrate models expression of alternate isoforms compensate during early developmental stages. For example, cardiac TnI gene knockout mice develop apparently normal hearts owing to the compensatory fetal TnI isoform but eventually acute heart failure occurs in later stages (Huang et al., 1999).

In Drosophila melanogaster the IFMs develop by fusion of myoblasts following their migration from the notum region of the wing imaginal discs during early stages of pupation. One group of IFMs, the dorsal longitudinal muscles (DLM), develop by fusion of these myoblasts to the remnants of larval oblique muscles (LOM, Fig. 1A) in the thorax region that escape complete histolysis at metamorphosis and serve as templates (TEM, Fig. 1B), whereas the other IFM group, the dorso ventral muscles (DVM), develop by de novo fusion of the myoblasts. During fusion the IFMs elongate (Fig. 1C) to span completely the developing thorax, attach to the tendon cells and, following a fibre shortening to one-third of the original size (Fig. 1D), begin to undergo myofibrillogenesis, a stage that involves high-level expression of the adult-specific structural sarcomeric proteins. Initial sarcomere organization occurs during this stage and the muscles elongate again. The tendon cells retract as the muscles increase in length and size until functional myofibres are formed (Fernandes et al., 1991; Fernandes et al., 1996; Reedy and Beall, 1993) (Fig. 1E,F). As Drosophila express many IFM-specific sarcomeric protein isoforms and no isoform changes occur during muscle differentiation or during later developmental stages, the IFMs offer an effective genetic model with which to study the null genotypes. These can be used to further our understanding of the roles of many sarcomeric proteins from early myogenesis through to the formation of functional myofibrils (reviewed in Vigoreaux, 2001). Recent studies show that kettin, a constituent of the sarcomeric Z-disc, and paramyosin, which forms part of the thick filament core, are involved in myoblast fusion (Hakeda et al., 2000; Liu et al., 2003; Zhang et al., 2000). This suggests that other sarcomeric proteins could play important roles during early myogenic processes in addition to their important roles as structural and functional components of the myofibril.

The single Drosophila TnI gene produces 10 different isoforms by differential splicing of 13 exons. Of these, isoforms differing at exon 6 (6a1, 6a2, 6b1 and 6b2) are mutually exclusive. The IFMs and TDT muscles express only the exon 6b1 isoform with or without exon 3 (Barbas et al., 1993; Beall and Fyrberg, 1991). hdp³ is a TnI mutation affecting alternative splicing of the IFM-TDT-specific exon 6b1, which results in a failure to produce all exon 6b1-containing isoforms (Barbas et al., 1993); hdp³ thus acts as an IFM-TDT-specific TnI null mutant. In the IFMs of hdp³ flies the second fibre elongation does not take place (Barthmaier and Fyrberg, 1995) and in the adults only a few remnants of muscle tissue are seen (Barbas et al., 1993; Beall and Fyrberg, 1991; Prado et al., 1999). We have used hdp³ to ask questions about the role of TnI in myoblast fusion and in the earliest stages of myofibril development. Using both a headless myosin heavy chain construct and myosin heavy chain mutations we have explored further the role of acto-myosin interactions in producing the hdp³ phenotype as proposed previously (Beall and Fyrberg, 1991) and show that inhibition of muscle contraction during early development is important. In addition, although removal or reduction of acto-myosin forces during development suppresses the major defects of the hdp³ phenotype, it does not produce the almost normal myofibrillar structure of the headless myosin construct on its own. Another aspect of the hdp³ phenotype is a reduced accumulation of the thin filaments and their proteins, which we discuss in the light of a possible co-ordinated regulation of the expression of these proteins.

Flies were maintained on standard yeast-sugar-agar medium at 25°C. Canton-S was used as a wild-type control for all the experiments unless specified. The original hdp⁵ mutation was procured from J. Vigoreaux (University of Vermont, Burlington, VT). The M6 mutation was recovered during a screen for X-chromosome flightless mutants following ethyl methanesulphonate mutagenesis. An Act88F-GFP P-element transgenic strain (inserted into the second chromosome) in which green fluorescent protein expression is controlled by the IFM-specific Act88F gene promoter (Barthmaier and Fyrberg, 1995) was obtained from K. VijayRaghavan (NCBS, Bangalore, India). The genotypes of y w hdp³ with the Mhc¹⁰ IFM null allele, with or without the headless myosin heavy chain transgenic line, Y97 (Cripps et al., 1999) were confirmed by eye colour (w or w⁺) and 1D-SDS-PAGE for full-length myosin heavy chain or the truncated form. All chromosome and gene symbols are as in Flybase (http://www.flybase.org) unless specifically described. Pupae were aged according to the method previously desribed (Fernandes et al., 1991).

Total RNA was extracted from 20 mg of flies using the Qiagen RNeasy kit and transcribed to cDNA using the Stratagene ProStar First-Strand cDNA synthesis kit. The coding region of the TnI gene was PCR extended, gel cleaned and sequenced (in-house automated sequencing facility) by using primers previously published (Barbas et al., 1993). Genomic DNA was isolated from 50 mg of Drosophila using the Qiagen DNeasy tissue kit following the manufacturer's instructions and used as PCR template to amplify TnI exons 3 and 6b1 from the mutant as both the exons were not included in the cDNA clones sequenced (see Results for details). The following primers were used to amplify and sequence the Exons 3 and 6b1. Exon 3 sense 5′-TCCTAGAACTGCAACTAACA and antisense 5′-ATTCGAACACAGCACTGCAC; exon 6b1 sense 5′-TACTATTAGTCATACGTAGA and antisense 5′-TTTGTATTCAGAGTTTTGAC. Sequences from three different PCR products for each primer pair were analysed using Lasergene software (DNASTAR, USA).

Newly eclosed flies were put into 70% ethanol and deep-frozen at –80° overnight. IFMs from 30 flies of each of the genotype were dissected in 70% ethanol as quickly as possible and put into tissue lysis buffer from the Qiagen RNeasy kit. The RNA was extracted and cDNA made as described above. 5 μl of the first strand cDNA product were PCR amplified using different thin filament gene primers and the relative amounts quantified using Gene-Genius Bio imaging (SYNGENE) gel documentation system. The following primers were used for gene-specific amplifications:

• Actin sense 5′-CCACGCCATTCTGCGTCTGG and antisense 5′-GCTGCCTTTGAAGAGCTTTCGCG;

• Tropomyosin 2 (Tm2) sense 5′-CACCATGGACGCCATCAAGAAG and antisense 5′-TTGGTATCGGCATCCTCAGC;

• Troponin T (TnT) sense 5′-GAAGATCTCGCTGTCGTTCC and antisense 5′-AATAGCAAGTTGTTAACTAC;

• TnI sense TTGTGAAGGCCAGAAATGGG and antisense 5′-GACTTCATTTCTGATCAAATCCAT;

• Troponin C (TnC41F6) sense 5′-GGATCCATATCGTATAGTAATAGTAA and antisense 5′-CATTGTCTGCCATATCTGAGC;

• Myosin (control) sense GTCCCAGGTGTCTCAGCTGT and antisense 5′-GGGTGACAGACGCTGCTTGGT.

Fly thoraces were prepared for polarised light microscopy (Nongthomba and Ramachandra, 1999) and for transmission or scanning electron microscopy (TEM and SEM) as described (Kronert et al., 1995).

After removing the heads and abdomens whole thoraces were put into cold 70% alcohol and dehydrated overnight. Thoraces were bisected using fine scissors; DLM from 30 flies were dissected carefully and transferred into 200 μl of York Modified Glycerol solution (YMG) (Peckham et al., 1990), and processed for 1D-SDS-PAGE as described (Kronert et al., 1995). Western blotting was carried out as described (Nongthomba et al., 2001) using the following anti-sera: Drosophila anti-TnI (1:1000; raised in rabbit; a gift from A. Ferrus, Cajal Institute, Madrid, Spain); Drosophila anti-TnT (1:1000; raised in rabbit; a kind gift from M. Cervera, Universidad Autónoma de Madrid, Spain); Drosophila anti-actin (1:1000; raised in rabbit); Calliphora anti-Tm (MAC 141 1:500; raised in rat; generously provided by B. Bullard, EMBL, Heidelberg, Germany). Whole thorax protein extracts were used for western blots with anti-TnI to show the presence and relative amounts of other TnI isoforms. For all hdp³ DLM protein gels the muscles were dissected from hdp³; Mhc¹² flies; the myosin null mutation inhibits the hypercontraction and degeneration of the muscles allowing their ready dissection without loss of the thin filament proteins.

After removal of the pupal cases, pupae of desired ages were dissected on a Sylgard plate (Dow Corning, USA) filled with phosphate buffered saline (PBS). After removal of all the fat cells and most of the unwanted tissues, the dissected muscles were transferred into a 1.5 ml eppendorf and the PBS was replaced with PBS containing 0.5% Triton X-100 (PBT) for 15 minutes. After fixation in 4% paraformaldehyde (PF) in PBS at room temperature for 1 hour, followed by four washes with PBT (4×15 minutes) and blocking with 1% BSA (bovine serum albumin) in PBS for 1 hour, the samples were incubated with primary antibody mixtures, and diluted with blocking solution (Vector Laboratories) overnight at 4°C. After washing in PBT, tissues were incubated in appropriately diluted secondary antibodies for 2 hours at room temperature. In some of the experiments, 2 μM rhodamine-phalloidin (Sigma) was added 20 minutes before the secondary antibody incubation was completed to reveal the F-actin filaments. Tissues were washed in PBS (4×15 minutes) and the IFMs were dissected from the attached cuticle and mounted on a slide using Vectashield (Vector Laboratories). The following antibodies were used: anti-Dmef2 (1:1000; raised in mouse; generously provided by S. Roy, Institute of Molecular and Cell Biology, Singapore); anti-muscle myosin II (1:1000, raised in rabbit; kindly supplied by S. Bernstein, San Diego State University, CA); anti-mouse Texas Red (1:500; Vector Laboratories), anti-Rabbit FITC (1:100; Sigma).

Tissues were observed with a Zeiss confocal microscope (Carl Zeiss, Germany) with appropriate filter sets for fluorescein and rhodamine. The images were taken using AIM and LSM5 image browser (CARL ZEISS, Germany). To determine the relative amount of fluorescence emission intensity of the Act88F-GFP, 40 serial 1 μm Z-axis sections of 4 μm² in area from three different areas of 12 individuals each for control and mutant genotypes were taken. The fluorescence averages (for each individual) and graphical comparisons were carried out using the LSM5 image browser. Images were processed using Adobe PhotoShop.

The Mhc⁷ mutant, which prevents expression of the IFM-specific myosin heavy chain isoform, removes thick filaments from the IFM myofibrils; conversely, in IFMs the actin null mutant Act88F^KM88 prevents thin filament formation (Beall et al., 1989). The observation that thin filaments form in Mhc⁷ IFMs (no thick filaments) and thick filaments in Act88F^KM88 IFMs (no thin filaments) argues that the two filament systems can largely assemble independently of each other. When both thick and thin filaments are removed by making both null mutants homozygous in the same fly only sarcoplasm with mitochondria is present (Beall et al., 1989). Expression of the Act88F gene begins as the myoblasts start fusing to form myotubes and the adult-specific Mhc gene isoforms appear immediately after myotube formation (Fernandes et al., 1991). It is not clear whether fibre morphogenesis depends upon the presence of either thick or thin filaments or both. Whereas the DLM of Mhc⁷ flies develop (Fig. 2A) a fibre morphology similar to wildtype (not shown), those of the Act88F^KM88 flies appear `wiggly' with the fibres bulging out in some areas and constricted in many areas (Fig. 2B). Occasionally the fibres seem to be partly split, perhaps reflecting an incomplete splitting of the DLM templates, and they appear not to fit within the thorax. All these effects must be a result of freely assembled thick filaments inside the fibres as removal of both actin and myosin restores a normal fibre morphology (Fig. 2C). These results indicate that IFM fibres (single syncytial muscle cells) can develop a normal morphology in the absence of sarcomeric actin and myosin; additionally, thick and thin filaments are clearly not required for fibre morphogenesis. Thus, many of the myogenic processes including myoblast fusion, myotube formation and the cellular processes affecting IFM morphology (Fig. 1) can occur in the absence of sarcomeric actin and myosin or both. The DVM, which do not require larval templates, also develop normally in the absence of sarcomeric actin and myosin (data not shown).

The wings-upA (wupA) gene is the single TnI-encoding gene in the Drosophila genome. Different muscle-specific isoforms are produced by alternative exon splicing (Barbas et al., 1993). Mutations affecting the IFM isoform have been recovered by their flightless or `wings-up' phenotype and a number of the original alleles were known as held-up. In hdp³ the 3′ AG splice site at the intron preceding alternate exon 6b1 is changed to AA, resulting in a failure to produce all exon 6b1-containing isoforms which are specific to the IFMs and TDT muscles (Barbas et al., 1993). As a result, in hdp³ flies the IFMs and TDT muscles never develop completely and myofibrils are not observed, although occasionally thick and thin filaments are present in the sarcoplasm of newly eclosed flies (Beall and Fyrberg, 1991; Nongthomba et al., 2003). The hdp⁵ mutant showed a muscle phenotype similar to that of hdp³ and reduced TnI message levels comparable with hdp³ (Beall and Fyrberg, 1991). Our sequencing of hdp⁵ genomic DNA revealed the same splice site junction mutation – AG changed to AA – indicating that hdp⁵ is synonymous with hdp³. A flightless and wings-up mutant, M6, was recovered from a screen of 3500 ethyl methanesulfonate mutagenised X-chromosomes, and contains a mutation at the same site. The presence of different silent mutational changes in all three mutants confirms that they arose independently, even from the original mutagenesis that produced the hdp³ and hdp⁵ alleles (Deak et al., 1982). This suggests that the 3′ AG splice site of the intron preceding exon 6b1 is a mutagenic hot spot. These wupA alleles (hdp^3,5 and M6) were all recovered by selection for the flightless and wings-up phenotypes. As such it was a requirement that they affect the IFMs. They all affect a specific splicing that results in non-expression of IFM-TDT TnI isoforms, causing them to be extreme mutations that affect only the IFM and TDT muscle groups; they are thus viable. The apparent mutational hotspot of the exon 6b1 3′ AG splice site may be due to the limited sites that can mutate to produce a strong IFM-specific phenotype.

Since TnI gene transcripts are detectable in myoblasts (A. Ferrus, personal communication), we followed the flight muscle development starting from myoblasts present in the third instar imaginal discs. By using an Act88F-GFP construct to visualise the early development of the IFMs, Barthmaier and Fyrberg described how the hdp³ IFMs partially elongate after myoblast fusion but then shortly afterwards start degrading so that only small muscle remnants are seen in later pupae (Barthmaier and Fyrberg, 1995). More detail of the developmental defects is not known. Staining of myoblasts for Dmef2 showed that the number of myoblasts present in the wing discs of the hdp³ mutant was comparable with wild-type controls (data not shown). We found no evidence of defective myoblast migration from the notum region of the wing disc. The hdp³ myoblasts [Fig. 3A, 22 hours after puparium formation (APF), shown in red; Dmef2 immunostaining] fuse to the DLM larval templates expressing GFP (Fig. 3A; yellow colour shows DLM containing fused myoblasts) confirming the results of Barthmaier and Fyrberg (Barthmaier and Fyrberg, 1995). Myotube formation then took place normally followed by the shortening of the fibres (Fig. 1). However, the elongation of fibres (Fig. 1E) that normally follows this shortening (Fig. 1D) was never completed in the mutant. Dmef2 immunostaining at a slightly later stage (40 hours APF) shows that the hdp³ myotubes remained quite short (Fig. 3C) compared with wildtype (Fig. 3B), but that the nuclei within the myotube are arranged in lines in both genotypes. This staining further confirms that the hdp³ myoblasts fused to the muscle templates in a manner similar to that of wild-type myoblasts. At 42 hours APF, when muscle fibres normally extend completely to the cuticle (Fig. 3D), the mutant fibres started degrading, a process that is first seen at the fibre ends (Fig. 3E). Within a few hours, the fibres were seen clumped together and pulled to the sides (Fig. 3F) and, with time, complete degradation occurred so that no fibres could be seen in newly eclosed adults (not shown).

All the myofibrils of developing IFMs are laid down inside sleeves of microtubules beginning at around 40-42 hours APF (Reedy and Beall, 1993). This corresponds to the time when the fibres are pulled from their attachment sites in hdp³ mutant. At this time, immunostaining with anti-myosin antibodies and rhodamine-phalloidin labelling of F-actin shows the wild-type individual myofibrils to consist of thick and thin filaments assembled into clearly demarcated repetitive I-A-I bands (Fig. 4A). By contrast, the hdp³ mutant myofibrils are clumped as filamentous bundles with no clear sarcomere-like structures, although thick and thin filaments are visible but scattered without any regular structure (Fig. 4B). The 40-42 hour APF wild-type myofibrils exhibit uniform length sarcomeres demarcated by straight, electron-dense Z-discs (Fig. 5A). By 46-48 hours APF, well-defined sarcomeres are present within each myofibril (Fig. 5B). However, in the hdp³ mutant IFMs no myofibrillar or sarcomeric structures are present. Z-discs are usually absent, but stress-fibre-like structures (SFLS), which normally precede sarcomere assembly (Reedy and Beall, 1993), and are larger and have bulbous Z-disc like electron dense bodies (Fig. 5C), are seen occasionally. These structures do not persist in the mutant fibres. By 46-48 hours APF, when the fibres are completely bundled up close to one of the two attachment sites, thick and thin filaments are visible but without any sarcomeric structure or Z-discs (Fig. 5D).

Previously, we have defined muscle hypercontraction as an aberrant process whereby a muscle develops normally but pulls apart owing to unregulated interactions of the thick and thin filaments (Naimi et al., 2001; Nongthomba et al., 2003). Although hdp³/+ heterozygote adults show a completely normal wing posture, they are dominant flightless and exhibit abnormal muscle structure (Barbas et al., 1991; Deak et al., 1982; Prado et al., 1999). Pupal dissections of hdp³/+ show that the muscles develop normally but that hypercontraction starts at late pupa stages (70-75 hours APF) similar to hdp² homozygotes (Nongthomba et al., 2003). When observed under polarised light, one-third of the hdp³/+ flies (n=77) showed IFMs with a hypercontraction muscle phenotype in which the muscles had been pulled to one of the attachment sites starting from the middle of the fibres (Fig. 6B). No IFM hypercontraction was seen in wild-type controls (Fig. 6A). Ultrastructurally, the outer areas of hdp³/+ myofibrils show loosely packed thick-thin filaments (Fig. 6D), a characteristic feature of the hypercontracting alleles, while more central core regions show the regular uniform lattice characteristic of wild-type myofibrils (Fig. 6C). Z discs and M lines, which are prominent in normal sarcomeres (Fig. 6E), are `streamed' in hdp³/+ (Fig. 6F).

We previously showed that the structure of IFMs from all hypercontracting alleles, in which muscle damage occurs after the muscles are formed, can be restored by reducing the force produced within the muscles (Nongthomba et al., 2003). The experiments involved genetically removing the wild-type myosin using mutations of the myosin heavy chain gene Mhc, which expressed no myosin in the IFMs, and replacing it with a transgenic construct which expresses a headless myosin that can assemble into thick filaments (Cripps et al., 1999) but cannot interact with the thin filament to produce force. The IFM muscle defects in hdp³/+ flies are completely restored to a normal structural condition (not shown) when combined in a genotype with two copies of the Mhc¹⁰ null allele and the headless myosin construct insert Y97. Earlier experiments (Beall and Fyrberg, 1991) had shown that the effects of homozygous hdp³ on the amount and length of fibres could be suppressed by also making the flies homozygous for an Mhc null allele. In this case no thick filaments form. So, is the homozygous hdp³ phenotype caused by unregulated interactions between thick and thin filaments similar to that suggested for hdp³/+, but taking place in early developmental stage owing to complete absence of TnI, or does TnI absence affect some earlier aspect of myogenesis? We planned to answer this question by decreasing the amount of functional myosin heads available to assemble into sarcomeres using an IFM myosin null mutation, Mhc¹⁰, but at the same time maintaining the assembly of thick filaments using a transgenic headless myosin line Y97. In hdp³/Y (two Mhc⁺ copies), where there are no visible muscles, IFM or TDT, in the adult thorax (Fig. 7A), decreasing the concentration of functional myosin head (7B) hdp³/Y; Mhc¹⁰/+ (one Mhc⁺ copy) and (7C) hdp³/Y; Mhc¹⁰/+; Y97/Y97 (one Mhc⁺ copy, two `headless' copy) – increased the amount of muscle fibres present (Fig. 7B,C). However, in these fibres no myofibrillar structures were visible until all the functional myosin heads had been removed, hdp<sup>3</sup>/Y; Mhc<sup>10</sup>/Mhc<sup>10</sup>; Y97/Y97 (no Mhc<sup>+</sup> copy, two headless copies) (Fig. 7D). This result suggests that even very small amounts of functional myosin are enough to tear the fibres. This is supported by the fact that similar experiments with non-functional myosin mutants having mutations in either actin or ATP-binding sites partially suppress the effects of hdp<sup>3</sup> on fibre morphology (Nongthomba et al., 2003). The ultrastructure of one of these, Mhc<sup>P401S</sup>, in combination with hdp<sup>3</sup> and the Mhc<sup>7</sup> myosin null shows an absence of myofibrils (Fig. 7E). Some thick filaments are seen, but there are no clear areas of inter-digitated thick-thin filaments. As reported earlier (Beall and Fyrberg, 1991; Fyrberg et al., 1990), condensed I-Z-I discs (serially repeated Z-discs connected by short stretches of thin filaments, which we refer to as `tiger-tails'), and Z-disc like structures (isolated Z-discs with attached thin filaments) are present in hdp³, hdp⁴ and hdp⁵ IFMs lacking myosin. All these structures show similarity to Z-discs by their electron-dense staining and by their orthogonal association with thin filaments. They show close similarity to human nemaline rod structures characteristic of thin filament myopathies (Sparrow et al., 2003). The results suggest that even weak interactions between myosin heads and thin filaments are enough to generate local force or cause conformational change to pull the filaments apart (discussed below). In the absence of all full-length myosin but with the headless myosin construct, hdp³; Mhc¹⁰; Y97, IFM thick filaments, some thin filaments, isolated Z-discs and tiger-tails are readily visible, but sarcomeric structures fail to assemble (Fig. 7F). If the function of the TnI is just to block the myosin heads interacting with thin filaments as they assemble, then the hdp³; Mhc¹⁰; Y97 genotype was expected to produce an ultrastructural phenotype similar to Mhc¹⁰; Y97 in which sarcomeric structures assemble (Cripps et al., 1999). Since we did not observe sarcomeres in hdp³; Mhc¹⁰; Y97 IFMs the primary defect of hdp³ must have other causes.

The large reduction in thin filaments within the hdp³ electron micrographs suggests that the absence of TnI might affect either thin filament assembly/stability or the expression of other thin filament genes. We used RT-PCR to examine the mRNA expression pattern of all the thin filament proteins in the hdp³; Mhc¹² genotype (the presence of Mhc¹² null mutation prevents fibre hypercontraction by removing the thick filaments). No complete TnI isoform with exon 6b1 and exon 3 was ever recovered, but remaining exons downstream could be detected readily in products extended using a primer within exon 6b1, which suggests that the splice defect preceding 6b1 does not affect the splicing of exons 3′ of this mutation. However, message levels were considerably reduced indicating the unstable nature of the transcript. Similarly, more than a fourfold decrease in the expression of TnC and TnT mRNA were seen (Fig. 8A). Tropomyosin and actin mRNA showed an approximately twofold decrease, whereas there was no change in the message level of the Mhc control (Fig. 8A). Evidence for a possible downregulation of other thin filament genes also comes from the activity of the Act88F promoter. Fluorescent expression of Act88F-GFP in the IFMs of hdp³; Act88F-GFP was reduced (36.66±12.98 arbitrary fluorescence units) significantly (Student's t=4.67, df=22, P<0.001) compared with wild-type Act88F-GFP controls (77.17±14.20). At least partly as a result of the decreases in mRNA levels, reductions in the amount of all the thin filament proteins are seen (Fig. 8B). As reported previously (Barbas et al., 1993), the 40 kDa TnI adult isoform is completely missing. However, the smaller isoform (TnI-E^*), which is normally expressed in thoracic muscles other than IFM or TDT, is present in the mutant whole thorax protein extract at levels comparable with those in the wild type (C-S). Immunoblotting of TnI isoforms in samples from wild-type IFMs dissected during muscle development (Fig. 8C) shows for the first time that the small IFM-TDT-specific isoform (includes exon 6b1 but no exon 3) (Barbas et al., 1993) is the sole TnI isoform produced early in IFM development, but it is replaced by the larger IFM-TDT isoform (exon 3 included) towards the end of the pupal period and is totally absent in IFMs from adults. Its disappearance may be due to dilution by high levels of expression of the larger isoform but even on over-development of the immunoblot we failed to detect the small isoform in the adult samples, and so this probably indicates an isoform replacement process. In the hdp³ IFM neither of these two isoforms was found, confirming directly that hdp³ is an IFM-specific null mutation. We have not confirmed the absence of TnI in the mutant TDT muscle. Surprisingly, arthrin, an IFM-specific ubiquitinated actin isoform (Ball et al., 1987) is also absent, indicating a possible correlation between the ubiquitination of actin and the troponin complex.

Muscle development occurs in two distinct stages. The first involves acquisition of muscle cell fate and fusion to form the syncytial myotubes. The second is the differentiation of the muscles including the intracellular assembly of the sarcomeres which are necessary for muscle contraction (reviewed in Baylies and Michelson, 2001; Buckingham, 2001; Naya and Olson, 1999; Roy and VijayRaghavan, 1999).

The development of normal fibre shape and attachment (Fig. 2) in the absence of expression of sarcomeric actin or myosin, or both, shows for the first time that fibre differentiation involves two independent processes: fibre morphogenesis and myofibrillogenesis. The mechanisms underlying fibre morphogenesis have not been identified but two major cytoskeletal networks that probably drive it are the non-muscle acto-myosin and dynein/kinesin-microtubule systems. Cytoplasmic actin and non-muscle myosin II are localised at muscle fibre membranes and the growing tips of the developing IFM myotubes extending towards their attachment sites (U.N. and J.S., unpublished). In developing IFM, expression of tubulins begins during the myoblast fusion stage (Fernandes et al., 1996), and microtubules form rings before the thick-thin filament assembly of the nascent myofibrillar lattice within them (Reedy and Beall, 1993). These microtubular structures may be important for fibre morphogenesis in the absence of the sarcomeric thick and thin filaments. In a vertebrate cell culture system it has been shown that the muscle-specific RING-finger protein (MURF), which is associated with microtubules, is required for myoblast differentiation, myotube formation and muscle morphogenesis (Pizon et al., 2002; Spencer et al., 2000).

We have shown that fibre morphogenesis and myofibrillogenesis are separable, independent processes, but muscle contraction inevitably affects fibre shape. It is therefore not inconsistent that hypercontracting mutations of sarcomeric proteins can seriously affect and destroy muscle fibre shape and integrity (Nongthomba et al., 2003).

Although the location and function of most sarcomeric proteins in the mature myofibril are known, a number are expressed in post-mitotic myoblasts and myotubes (Costa et al., 2002; Liu et al., 2003; Ojima et al., 1999; Zhang et al., 2000). Genetic studies have shown that D-titin is required for myoblast fusion (Zhang et al., 2000) and paramyosin is important in the same process (Liu et al., 2003). The roles of these proteins in fusion are not yet clear. The completion of fibre morphogenesis in the hdp³ mutant would seem to suggest that TnI is not required for these early myogenic processes. However, the hdp³ mutant is a `null' in the sense that it affects the splicing and thereby translation of the IFM-specific isoform. The viability and otherwise normal behaviour of hdp³ flies argues that in all other muscles the wupA gene expresses normal or sufficient quantities of TnI.

Most heterozygotes for contractile protein null mutants exhibit flight impairment and abnormal IFM muscle structure. Most of these are believed to result from stoichiometric imbalances between sarcomeric proteins as they assemble (Beall et al., 1989; Vigoreaux, 2001). It is well established that in the regulation of striated muscle contraction TnI acts as the inhibitory regulatory element for actin-myosin interactions in the absence of the Ca²⁺ binding to TnC (reviewed in Gordon et al., 2000). In hdp³ homozygotes the absence of TnI seems to lead to an extreme muscle phenotype (Fig. 7A) in which the IFM do not develop because of an absence of inhibition of muscle contraction (see below). However, hdp³/+ heterozygotes show an IFM hypercontraction phenotype (Nongthomba et al., 2003), in which the IFM develop normally but subsequently during late pupal stages undergo destructive contractions. These observations suggest that during early IFM development the TnI molecules, expressed from the wild-type wupA gene, assembled into troponin complexes are sufficient to inhibit muscle contraction, but that once the muscles can be activated at around 72-75 hours APF (Nongthomba et al., 2003) the proportion of TnI-containing troponin complexes along the thin filaments are insufficient to re-establish the `relaxed' muscle state.

In hdp³ homozygotes there is none of the IFM-specific TnI isoform (Fig. 8C). Myosin heads will be able to interact directly with the myosin-binding surface of nascent F-actin, and cause contraction of the fibre at a time (Fig. 3E) when it is normally extending and thereby prevent assembly of developing sarcomeres. Evidence that this process is driven by unregulated acto-myosin interactions comes from the following observations: (1) that myosin mutants that affect the actin-binding sites or hamper the ATP hydrolysis partially suppress the hdp³ fibre phenotype (Nongthomba et al., 2003); (2) that removing all the myosin (Beall and Fyrberg, 1991) or (3) replacing the myosin with a headless isoform (Fig. 7D) completely suppress the hdp³ fibre phenotype. Ultrastructurally, at the earliest stages of myofibrillogenesis hdp³ IFM show SFLS-like structures, which normally precede sarcomere assembly (Reedy and Beall, 1993) but in the mutant they are thicker, rounded Z-bodies (Fig. 5C). The lack of Z-discs at later stages (Fig. 5D) supports the proposal that unregulated acto-myosin interactions in hdp³ myofibrils prevent assembly of sarcomeres or destroy them as they form.

If unregulated acto-myosin interactions in developing hdp³ myofibrils were the root cause of the muscle phenotype then the ultrastructural appearance of hdp³; Mhc¹⁰; Y97 myofibrils should be no different from those of Mhc¹⁰; Y97 flies. This is not the case. The latter genotype forms respectable IFM sarcomeres and myofibrils, although not with the completely regular structure and integrity of wildtype (Cripps et al., 1999). This result suggests that the hdp³ phenotype is caused by more than aberrant regulation of thick-thin filament interactions. We have shown that the hdp³ mutation affects the expression levels of mRNA and proteins of other thin filament components but is without effect on expression of the myosin heavy chain mRNA. Previously, similar reductions in the accumulation of associated thin filament proteins have been reported for other IFM thin filament null mutants: TnT (Fyrberg et al., 1990), Tm2 (Karlik and Fyrberg, 1985) and Actin88F (Mogami et al., 1982; Mogami and Hotta, 1981). The reduction in expression of this group of thin filament genes in hdp³ flies may explain the effects of the mutant on the muscle phenotype that is not suppressible by reducing force production during fibre development.

Secondary effects of one thin filament mutant on the expression of other thin filament genes, and the regulatory interactions that this implies, are not restricted to Drosophila IFM. In zebra fish, a mutation of the TnT2 gene (silent heart, sih) has been found to severely reduce not only TnT expression but also TnI3 and Tm-α (Sehnert et al., 2002), and, in a myoblast cell culture system, a β-actin gene mutation causes reduced protein and mRNA levels of Tm2 and Tm3 (Schevzov et al., 1993). Furthermore, relative mRNA output between contractile gene families in humans has been found at different stages of development, and independently of isoform switching, which suggests the existence of some form of communication between these genes (Wade et al., 1990). These and recent microarray expression profiling of different developmental stages of Drosophila, where many muscle genes were found to be expressed at the same time (Arbeitman et. al., 2002; White et. al., 1999), suggest the presence of common regulatory interactions between contractile protein gene families. Further microarray profiling of different IFM developmental stages of wild type and of null mutants of the thick-thin filament protein genes will be required to elucidate these pathways.

The mechanism by which the mutant transcripts or proteins affect those of the normal thin filament genes is unclear. The various mutants include nonsense and splicing defects. Transcript splicing and mRNA export are mediated by the exon-junction complex (EJC), which contains several proteins involved in nonsense-mediated decay (NMD) (Reed and Hurt, 2002; Reed, 2003; Vasudevan and Peltz, 2003). It is likely that after the nuclear translation, a pioneer round of translation occurs (Buhler et al., 2002) and the mutant splice defect isoform is eliminated. Therefore the TnI isoform with the 6b1 exon is never formed, which explains why this isoform was never recovered in previous (Barbas et al., 1993) and present studies. PCR recovery of low levels of TnI cDNA with the later exons from hdp³ IFM may represent partially processed mRNA in the nucleus. Whether this pathway is somehow linked to reduced expression of other related thin filament genes remains a conjecture.

Elucidation of how these genes are coordinately regulated seems likely to be important in understanding human mutations that cause muscle disease. Mutations in human TnI are associated with hypertrophic cardiomyopathy (Kimura et al., 1997) and distal arthrogryposis (Sung et al., 2003). In mice a cardiac troponin-I knockout leads to lethality (Huang et al., 1999). In this study we have shown that the Drosophila IFM and the wupA gene can provide a model system to explore the function of TnI in normal muscle development and disease. The hdp³ mutation would seem a useful tool for starting an investigation of the regulation of other thin filament protein genes during myofibrillogenesis in vivo.

Thanks to S. Roy, A. Ferrus, B. Bullard, M. Cervera, D. Kiehart, J. Vigoreaux, K. VijayRaghavan and S. Bernstein for generously providing the antibodies and flies, and to Debbie Girdlestone for comments on the manuscript. This work was supported by BBSRC and BHF grants to J.C.S.