Myofibrillogenesis in the developing chicken heart: assembly of Z-disk, M-line and the thick filaments

The heart is the first functional organ in the developing embryo (Lyons, 1996). Contractions of cardiomyocytes can be observed in chicken embryos already after 36 hours in ovo (at the 9 somite stage) and after 12 further hours the entire blood flow is managed by a rhythmically contracting meshwork of myofibrils (Tokuyasu and Maher, 1987a). The assembly of myofibrillar proteins into their functional unit, the sarcomere, must therefore be a rapid and well co-ordinated process. In order to explain this phenomenon numerous studies were performed on primary cultures of cardiomyocytes and several models have been proposed. Thick actin bundles, so-called stress fibre-like structures (SFLS), are a prominent feature of cultured cardiomyocytes and might act as a scaffold during sarcomere assembly (Dlugosz et al., 1984). This hypothesis was extended further with the observation that IZI-complexes, structures which contain the Z-disk component sarcomeric α-actinin as well as α-actin and titin, might be put in register on these filamentous structures to build non-striated myofibrils (NSMF) and that thick filaments, consisting mainly of myosin, might be assembled independently and only become incorporated later into the myofibrils (Wang et al., 1988;

Schultheiss et al., 1990). A different theory proposes the importance of the so-called premyofibril as a precursor structure during myofibril assembly (Rhee et al., 1994; LoRusso et al., 1997). Premyofibrils are composed of minisarcomeres, i.e. narrower spaced Z-disks which contain nonmuscle isoforms of contractile proteins such as myosin and do not contain titin. Later the distance between individual Z-disks increases to reach the distance normally observed in the sarcomere and non-muscle myosin is replaced by its muscle isoform (Rhee et al., 1994; LoRusso et al., 1997). However, all these models emerged from studies on cardiomyocytes in vitro. Since these possessed already myofibrils at the time of isolation from the tissue, it cannot be excluded that some of the observations do not reflect the situation during de novo myofibril assembly in situ. Firstly, it is possible that parts of myofibrils might be degraded and recycled during the cellular adaptation process to culture conditions and secondly the artificial two-dimensional environment might favour specific cytoskeletal structures which do not occur in situ, such as stress fibres. In order to understand the biogenesis of cardiomyocyte cytoarchitecture, it is therefore important to find out how myofibrils are assembled in situ.

So far comparatively little is known about myofibrillogenesis during normal development. Studies on embryonic mouse skeletal muscle indicated that sarcomeric proteins are expressed in a defined sequence during development with desmin, an intermediate filament protein, appearing first, subsequently followed by titin and later all the other sarcomeric components (Fürst et al., 1989). This led to the proposal that titin might act as an integrator of IZI-complexes in situ as well and that A-band assembly happens independently. However, in skeletal muscle myofibrils develop over the course of one week (Fischman, 1986), whereas a beating heart takes only a few hours to mature. Therefore different mechanisms of myofibrillogenesis might be employed during cardiac muscle development. In their pioneering studies on embryonic chicken heart Tokuyasu and Maher (1987a) reported that α-actinin and titin became organised in a periodical fashion at a time when actin and myosin were still distributed diffusely throughout the cell. Later, as development proceeded, myosin was observed in a sarcomeric arrangement while actin appeared still filamentous (Tokuyasu and Maher, 1987a). The arrangement of actin changes during heart development; initially it is present as a filamentous meshwork, then thick actin bundles are formed prior to the appearance of mature I-bands (Shiraishi et al., 1992). By studying the localisation of vinculin, N-cadherin, fibronectin and phospho-tyrosine it was shown that myofibrils from neighbouring cardiomyocytes interconnect already at early stages of myofibrillogenesis and it was suggested that these interaction sites might be important for the alignment of the myofibrils (Tokuyasu, 1989; Shiraishi et al., 1993, 1995, 1997a).

Despite all these studies, no thorough comparative analysis of the localisation of several myofibrillar proteins during heart development has been reported so far. In order to investigate whether sequential expression of sarcomeric components occurs during cardiac myofibrillogenesis and whether intermediate stages of myofibril assembly or scaffolding structures are important during myofibrillogenesis in situ, we performed confocal microscopy on immunostained whole mount preparations of embryonic chicken hearts that were isolated around the time when myofibrillar contractions start. By applying doubleand triple-labelling techniques we were able to compare the localisation of different components of the sarcomere within the same myofibril. All sarcomeric proteins investigated were already accumulated well before the onset of myofibrillogenesis and only two major stages of myofibril assembly could be resolved. In the immature heart before beating, filamentous actin is found near the cell membranes and only α-actinin and titin are organised in a distinct fashion. As soon as the first contractions can be observed, the sarcomeres appear to be fully organised with the exception of the I-bands, which attain their definitive length only at a later stage of development (E. Ehler et al., unpublished).

Embryonic chicken heart whole mount preparations

Fertilised eggs from white Leghorn hens (Hungerbühler, Flawil, Switzerland) were incubated at 37°C for about 1.5 days. Embryos were removed from the eggs, transferred into cold PBS (phosphate buffered saline) and their stage age was determined under a dissection microscope by counting the somites.

PFA-fixed heart whole mount preparations

The heart was dissected in cold PBS and fixed in 3% PFA (paraformaldehyde) in PBS with 0.002% Triton X-100 (PBT) for 1 hour. After several washes in PBT and permeabilisation with 0.1% Triton X-100 in PBS for 30 minutes, the hearts were treated with hyaluronidase (1 mg/ml; Sigma, St Louis, MO, USA) in PBT for 30 minutes at RT (room temperature) in order to remove the cardiac jelly and to ensure access for the antibodies to the inner myocardial wall (Tokuyasu and Maher, 1987a). After further washes in PBT and blocking with 5% NGS (normal pre-immune goat serum), 1% BSA (bovine serum albumin) in PBS for 30 minutes, the hearts were incubated with the primary antibody mixtures, diluted in blocking solution, overnight at 4°C. After 3⨯ 2 hours washing in PBT the secondary antibodies were applied for either 6 hours at RT or overnight at 4°C. The hearts were washed in PBT (6⨯ 1 hour) and mounted in 0.1 M Tris-HCl (pH 9.5)-glycerol (3:7) including 50 mg/ml n-propyl gallate as anti-fading reagent (Messerli et al., 1993a).

The dissected hearts were digested with hyaluronidase as above, washed with PBT and subsequently fixed in 90% methanol, 10% DMSO overnight at 4°C (Robson and Hughes, 1996). After rehydration for 2⨯ 30 minutes in PBS the blocking of unspecific binding sites and the antibody labelling were carried out as described above.

Hearts from 11-day-old chicken embryos were digested with collagenase (108 units/ml, Worthington Biochemical Corp., Freehold, NJ, USA) in ADS buffer (116 mM NaCl, 20 mM Hepes, 0.8 mM NaH₂PO₄, 1 g/l glucose, 5.4 mM KCl, 0.8 mM MgSO₄, pH 7.35) and cultured as described (Sen et al., 1988). Cells were plated into dishes coated with 10 µg/ml fibronectin in plating medium (67% Dulbecco’s MEM (Amimed AG, Basel, Switzerland), 17% Medium M199 (Amimed AG), 10% horse serum (Gibco, Life Technologies, Basel, Switzerland), 5% fetal calf serum (Gibco) and 1% penicillin/ streptomycin (Gibco)). After one day the medium was replaced by maintenance medium (78% Dulbecco’s MEM, 20% Medium M199, 1% penicillin/streptomycin, 1% horse serum and 10⁻⁴ mol/l phenylephrine (Sigma)) In order to reduce the number of contaminating fibroblasts, glutamine was left out of the culture media and cytosine arabinoside (10 µmol/l, Sigma) was added to the cultures.

For immunofluorescence 4-day-old cultures were either fixed with 4% PFA in PBS for 15 minutes at RT, then blocked in 0.1 M glycine in PBS for 5 minutes and permeabilised in 0.2% Triton/PBS for 10 minutes or alternatively fixed in methanol for 5 minutes at −20°C. After blocking with 5% NGS in 1% BSA/PBS for 20 minutes the first antibody was incubated for 1 hour at RT and after washing with PBS the secondary antibody was applied for 45 minutes. The specimens were washed in PBS and mounted as mentioned above.

The imaging system consisted of a Leica inverted microscope DM IRB/E, a Leica true confocal scanner TCS NT and a Silicon Graphics workstation. The images were recorded using a Leica PL APO 100x/1.4 oil, PL APO ⨯63/1.4 oil or a PL APO ⨯63/1.4 water immersion objective. The system was equipped with an argon/krypton mixed gas laser. Image processing was done on a Silicon Graphics workstation using ‘Imaris’ (Bitplane AG, Zurich, Switzerland), a 3-D multi-channel image processing software specialised for confocal microscopy images (Messerli et al., 1993b).

Dissected hearts from embryos of the same stage were pooled on dry ice, resuspended in a modified version of SDS-sample buffer (3.7 M urea, 134.6 mM Tris, pH 6.8, 5.4% SDS, 2.3% NP-40, 4.45% β-mercaptoethanol, 4% glycerol and 6 mg/100 ml Bromophenol Blue; Laemmli, 1970) and boiled for 1 minute. For the positive and negative controls heart and brain were isolated from HH stage 35 embryos, which were staged according to Hamburger and Hamilton (1951); the tissue was homogenised by freeze-slamming and solubilised in the modified SDS-sample buffer.

The SDS-samples were run on 7.5% or 12% polyacrylamide minigels (Bio-Rad, Glattbrugg, Switzerland) except for the titin blot where 2-12% gradient gels were prepared according to the method of Matsudaira and Burgess (1978) using the buffer system of Laemmli (1970). The proteins were blotted overnight onto nitrocellulose (Hybond-C extra, Amersham, Zurich, Switzerland; Towbin et al., 1979).

After the transferred proteins had been visualised with Ponceau Red (Serva, Heidelberg, Germany), unspecific binding regions were blocked with 5% non-fat dry milk in washing buffer (0.9% NaCl, 9 mM Tris, pH 7.4, 0.1% Tween-20) for 1 hour at RT. Primary and secondary horseradish peroxidase (HRP)-conjugated antibodies were diluted in washing buffer with 0.1% milk and incubated for 1 hour, respectively, with intermittent washing in washing buffer. After a final wash in washing buffer a chemiluminescence reaction was performed according to the manufacturer (Amersham and Pierce, Socochim, Lausanne, Switzerland, respectively) and results were visualised on Fuji Medical X-Ray films. Bands of the expected molecular mass, as judged by Kaleidoscope Prestained Standards (Bio-Rad) were present with all the antibodies used in the positive control (HH 35 heart) and absent in the negative control (HH 35 brain for sarcomeric proteins; data not shown).

The monoclonal mouse (mM) anti-myomesin (clone B4; Grove et al., 1984) and the polyclonal rabbit (pR) anti-chicken heart myosin binding protein-C (MyBP-C; Bähler et al., 1985) were both characterised in this laboratory. The mM anti-sarcomeric α-actinin (clone EA-53) and anti-desmin (clone DE-U-10) antibodies were obtained from Sigma (Buchs, Switzerland). The mM anti-α-sarcomeric actin (clone Alpha-Sr-1) antibody was from DAKO (Zug, Switzerland) and the mM anti-non-muscle myosin IIB from Chemicon (Temecula, CA, USA). The mM mouse anti-titin (clone 9D10) antibody was obtained from the Developmental Studies Hybridoma Bank (maintained by the University of Iowa, Department of Biological Sciences, Iowa City, IA52242, USA). The mM antisarcomeric myosin heavy chain (clone A4.1025; Pavlath et al., 1989) antibody was donated by Dr Simon Hughes (London, UK), the polyclonal rat (pRt) anti-myomesin (Obermann et al., 1996), the pRt anti-MyBP-C and the pR anti-sarcomeric α-actinin antibodies were a kind gift from Dr Mathias Gautel, (Heidelberg, Germany) and the mM anti-titin (clone T41; M-line epitope; Obermann et al., 1996) antibody was generously donated by Prof. Dieter Fürst (Potsdam, Germany). The anti-non-muscle myosin incubations were performed on methanol fixed specimens, for the rest PFA-fixation was used.

For the triple immunofluorescence stainings, combinations of FITC, Cy3 or TRITC and Cy5-conjugated secondary antibodies as well as phalloidin were used. Incubation of the whole mount preparations with the secondary antibody mixtures alone gave no significant signal. The secondary antibodies for immunofluorescence were purchased from Cappel (FITC anti-mouse Ig; FITC anti-rabbit Ig; Organon Teknika, Pfäffikon, Switzerland) and Jackson Immuno Research (Cy3 antimouse Ig; Cy5 anti-mouse Ig; Cy5 anti-rabbit Ig; Dianova, Hamburg, Germany). For double labelling of anti-titin 9D10 and anti-sarcomeric myosin monoclonal antibodies and anti-non-muscle and sarcomeric myosin monoclonal antibodies, respectively, subclass specific antimouse IgM-(FITC-conjugated, Sigma) and anti-IgG-(TRITC-conjugated, Cappel, Organon Teknika) antibodies were used. For visualising the titin T41 and the desmin DE-U10 antibodies a biotinylated anti-mouse antibody (DAKO, Zug, Switzerland) was used followed by incubation with FITC-Streptavidin (DAKO). FITC- and TRITC-conjugated phalloidin was obtained from Molecular Probes (Leiden, The Netherlands). HRP-conjugated anti-mouse Igs (DAKO), anti-rat Igs (DAKO), anti-rabbit Igs (Calbiochem, Luzern, Switzerland) and anti-mouse IgMs (Sigma) were used for immunoblotting.

In skeletal muscle sarcomeric proteins were reported to be expressed sequentially during embryonic development (Fürst et al., 1989). In order to check for sequential expression during embryonic heart development we analysed SDS-samples of dissected chicken hearts of different stages by immunoblots with specific antibodies. A selection of sarcomeric proteins was investigated (sarcomeric α-actin, sarcomeric α-actinin, titin, sarcomeric myosin heavy chain, myomesin, MyBP-C and desmin), all of which were already expressed in the 6 somite heart (Fig. 1A-G, lane 1), well before myofibrillar contractions start.

Next we wanted to know how the different components of the myofibril are assembled to sarcomeres during early heart development. Whole mount preparations of isolated hearts of different developmental stages (8 somites, 9 somites and 11-12 somites; Fig. 2 left, middle and right column, respectively) were stained for immunofluorescence analysis using either a combination of phalloidin and antibodies to sarcomeric α-actinin (Fig. 2, rows one and two) or a combination of antibodies to titin (N-terminal epitope 9D10) and to sarcomeric myosin heavy chain (Fig. 2, rows three and four) or by performing single-labelling with antibodies to myomesin (Fig. 2, row five). Analysis by confocal microscopy revealed all the major sarcomeric components already in the 8 somite stage heart, before beating starts (Fig. 2A,D,G,J,M). However, major differences in the localisation within the cell were observed for the individual components. Whereas F-actin, as visualised by phalloidin, is preferentially found in filamentous structures, which run along the cell membranes (Fig. 2A), both the Z-disk component sarcomeric α-actinin as well as titin (N-terminal epitopes), which makes up the elastic, third, filament system, appear in a punctuate staining pattern, already partially organised along lines (Fig. 2D,G, arrows). Sarcomeric myosin heavy chain as a marker for the A-band and myomesin as representative of the M-line are both still distributed diffusely throughout the cytoplasm in the 8 somite heart (Fig. 2J and M, respectively). At the time when beating starts, which is at the 9 somite stage, all components of the sarcomere are localised correctly in the myofibrils (Fig. 2E,H,K,N), with the exception of actin, which is still partially nonstriated (arrowheads in Fig. 2B; compare with the organisation of sarcomeric α-actinin in Fig. 2E. Mature I-bands are indicated with arrows in Fig. 2B). The staining for titin (Fig. 2H) and sarcomeric myosin heavy chain (Fig. 2K) reveals striations in the same myofibrils, with tiny doublets for the titin 9D10 epitope and broader double bands for the myosin heads, however, at the 9 somite stage these striations are not yet as distinct as in older myofibrils (Fig. 2I,L). The myomesin antibody delineates the M-bands in the first myofibrils (Fig. 2N). By the 12 somite stage most of the actin is properly segregated into I bands and the myofibrils have become thicker (Fig. 2C,F,I,L,O).

Antibodies, which react with N-terminal epitopes of titin at the Z-disk or in the I-band show that this part of titin is already organised from the earliest time points of myofibrillogenesis onwards. When double labellings with the 9D10 antibody, which recognises a titin epitope in the PEVK region in the I-band (Wang et al., 1991) together with an antibody against sarcomeric myosin heavy chain are performed, both titin and myosin are localised in their sarcomeric pattern in the 9 somite stage heart (Fig. 2H,K). However, striations of titin appear slightly earlier than myosin double bands (arrows in Fig. 2H and K, respectively). A similar pattern is observed in cultured embryonic chicken cardiomyocytes (ECC), where titin doublets can already be observed at a time when myosin is not yet organised properly (data not shown).

To investigate whether the M-line region of titin is also organised at the very early stages of heart development and whether titin functions indeed as a sarcomeric ruler for myofibril assembly we performed triple stainings with the monoclonal T41 titin antibody, which recognises an epitope in the M-line (Obermann et al., 1996; Fig. 3C,G; green in A,E), together with a polyclonal sarcomeric α-actinin antibody (Fig. 3D,H; light-blue in A,E) and phalloidin (Fig. 3B,F; red in A,E). Fig. 3A and E show an overlay of the three channels, where the sarcomeric α-actinin-staining appears pink due to the localisation in the middle of the I-bands. We could demonstrate that M-line titin is partially organised in the 8 somite heart (arrows in Fig. 3A-D), however, this organisation is less marked than the one observed with antibodies which recognise Z-disk or I-band epitopes of titin (compare with Fig. 2G) and most of the staining appears diffuse. There is also very little T41 titin staining in peripheral areas in cultured embryonic chicken cardiomyocytes (ECC), where sarcomeric α-actinin can already be detected (data not shown). In Fig. 3E-H, a triple staining of a 10 somite heart, a mature myofibril and an assembling myofibril can be seen in the same field. Alternative spacing between α-actinin and T41 titin is obvious in the mature myofibrils at the left hand side but also when the sarcomeres are less organised (arrowheads for sarcomeric α-actinin and arrows for titin T41, respectively in Fig. 3E-H). Thus it seems that although the M-line end of titin (the C terminus of the protein) can be visualised with antibodies slightly later than the N terminus at the Z-disk, titin can fulfil its function as a ruler of sarcomere assembly before the rest of the myofibrillar proteins becomes localised.

We compared the localisation of myomesin and myosin binding protein-C (MyBP-C) together with actin to resolve whether there are any temporal differences in assembly of the M-line and the A-band. Both myosin-binding proteins show a similar, completely diffuse localisation at the 8 somite stage (Fig. 4B,C).

In contrast, organisation to a sarcomeric pattern is clearly apparent in the 9 somite heart (Fig. 4D-F) and an increase in the number of assembled myofibrils can be observed in the 11 somite heart (Fig. 4G-I). Sometimes the organisation of myomesin seems to precede that one of MyBP-C (Fig. 4D,E,F, arrows directed at the same positions in the myofibril), similar to observations made in cultured ECC (data not shown). We were unable to detect thick filaments apart from the filamentous actin structures. This suggests that although myosin filaments are assembled independently as shown by electron microscopy of embryonic chicken heart (Manasek, 1968; Hiruma and Hirakow, 1985), their integration into the sarcomere can only happen together with the M-line structure composed of the antiparallel C termini of titin and of myomesin.

Studies on embryonic cardiomyocytes in vitro have underlined the potential importance of non muscle myosin IIB during myofibrillogenesis as a component of the so-called premyofibril (Rhee et al., 1994). In cultured ECC there is a clear segregation of non-muscle (arrowheads in Fig. 5K; green in I) and muscle myosin (Fig. 5J; an overlay of the three channels can be seen in 5I where muscle myosin is red), with non-muscle myosin being present together with sarcomeric α-actinin (Fig. 5L, light blue in I) in an alternating fashion in the so-called premyofibril region at the periphery of the cell (Rhee et al., 1994). To investigate the significance of non-muscle myosin for myofibrillogenesis in situ we performed triple labellings of non-muscle myosin (Fig. 5C,G, green in A,E), sarcomeric myosin (Fig. 5B,F, red in A,E) and α-actinin (Fig. 5D,H, light-blue in A,E) on whole mount preparations. The localisation of these proteins can be directly compared in overlay pictures of the three channels in Fig. 5A (8 somite heart) and E (9 somite heart). Non-muscle myosin does not seem to be as important in the developing heart, since this isoform is expressed at rather low levels and localises in a diffuse fashion even at the 8 somite stage (Fig. 5C). In contrast to the results in vitro no alternation between the α-actinin dots present in the 8 somite heart and non-muscle myosin can be detected (arrows in Fig. 5A-D, pointing at α-actinin dots). There is also very little diffuse staining of non-muscle myosin in assembled myofibrils in the 9 somite heart (Fig. 5G, arrows point at a myofibril). In addition we were unable to observe socalled mini-sarcomeres in embryonic cardiac tissue. The spacing of the α-actinin dots in the 8 somite heart (arrows in Fig. 5D) matches the distance between mature Z-disks (Fig. 5H) and is not as narrow as in the premyofibril regions in the periphery of ECC (Fig. 5L,K).

Desmin is the major intermediate filament protein in striated muscle (Lazarides and Hubbard, 1976) and is thought to be the first sarcomeric component expressed in skeletal muscle (Fürst et al., 1989). In chicken heart, expression of desmin can be already detected at the 6 somite stage (Fig. 1G), and in the 7 somite heart a clear punctuate pattern is observed with an antidesmin antibody (Fig. 6B′). These dots run sometimes on both sides of the filamentous actin along the cell borders and sometimes they seem to precede actin filament extension (Fig. 6A′,B′, arrows and arrowheads, respectively). Intermediate filaments might therefore provide a guiding scaffold for actin along the membranes. By the 8 somite stage this precise localisation of desmin dissolves and only a diffuse localisation can be observed (Fig. 6B). In the 10 somite heart desmin is still completely diffuse (Fig. 6E) and becomes only organised at Z-disks in the most mature sarcomeres in the 12 somite heart (arrows in Fig. 6H). Staining for sarcomeric α-actinin (Fig. 6C,F,I) was used as a reference for the Z-disk together with phalloidin to visualise filamentous actin (Fig. 6A′,A,D,G). Therefore, like in skeletal muscle in situ and in vitro, desmin is expressed already at the earliest stages of myofibril assembly but attains its mature localisation pattern only quite late during myofibrillogenesis (Fürst et al., 1989; van der Ven et al., 1993).

By analysing triple-stained whole mount preparations of developing chicken hearts using confocal microscopy we were able to compare the organisation of several components of the sarcomere during myofibrillogenesis in situ. Our study presents the first comparative analysis since the pioneering work of Tokuyasu and Maher (1987a,b) and highlights important differences between the process of myofibrillogenesis in cardiomyocytes cultured in vitro and in the developing heart in situ.

Myofibrils in cultured cardiomyocytes are typically striated in the centre of the cell with non-striated extensions into the cellular periphery (Messerli et al., 1993a). These extensions terminate in so-called subsarcolemmal adhesion plaques (SAP), are characterised by a concentration of the cellsubstrate contact protein vinculin and are thought to represent the region of new myofibril assembly (Lu et al., 1992). However, when we follow a myofibril in a three-dimensional representation of our confocal data sets from the developing heart no transitions between striated and non-striated stretches can be seen and there is also no evidence for SAP. The observed myofibrillar structures are always either immature i.e. possess dense body-like structures together with filamentous actin or mature with all the sarcomeric proteins in their correct striation pattern. Immature structures tend to be associated with membranes, possibly mirroring the function of SAP in cultured cardiomyocytes, whereas mature myofibrils stretch throughout the cytoplasm and can even run from one cell into another.

Another important difference between myofibrillogenesis in vitro and in situ is the absence of premyofibrils in the developing heart. The spacings between neighbouring α-actinin dots resembles mature sarcomere distances even before beating and there is no evidence for a function of non-muscle myosin IIB as a space-holder for muscle myosin isoforms in between neighbouring α-actinin dots at these early stages. Thus the premyofibril hypothesis seems not to be applicable for myofibrillogenesis in situ, an interpretation which is supported by the fact that mice which are homozygous null for non-muscle myosin IIB are able to assemble sarcomeres (Tullio et al., 1997). However, many observations which were made on cultured cardiomyocytes can be confirmed by analysis of whole mount preparations. For example dense body-like structures composed of α-actinin, the N terminus of titin and actin can also be detected from the earliest stages of sarcomere assembly in situ and seem to act as basic organising structures during myofibrillogenesis.

Based on the data presented here together with results obtained by us and others by the study of myofibrillogenesis in cultured cardiomyocytes, we are able to suggest a probable sequence of events which finally lead to a functional myofibril. At present our model only deals with observations made on chicken hearts and other species might show variations of this process (van der Loop et al., 1992).

The first step in myofibrillogenesis seems to involve a complex of sarcomeric α-actinin, the N terminus of titin and membrane-associated actin filaments. These complexes, which have been termed dense body-like structures, are already present in an organised fashion in the 8 somite heart and seem to be the first building blocks for myofibrillogenesis in situ (our results, Tokuyasu and Maher, 1987b; Fürst et al., 1989) as well as in vitro (Schultheiss et al., 1990; Lu et al., 1992). Recent models of Z-disk organisation provide an explanation how these three molecules might interact with each other in order to make up nucleating structures and to provide the first cues for a polarised organisation of both the actin filaments as well as of titin (Gregorio et al., 1998; Young et al., 1998).

Once these dense body-like structures have assembled, there must be some mechanism to put them in register. The most likely candidate for such a function is titin, which extends from Z-disk (N terminus) to M-line (C terminus) in the mature sarcomere. However, to fulfill its role as a ruler of the sarcomere, titin molecules have to unfold and the C termini of neighbouring titin molecules have to interact with each other in an anti-parallel fashion as proposed in a recent M-line model (Obermann et al., 1996). Several groups have observed that the organisation of N-terminal titin epitopes precedes the one of more C-terminal epitopes in cultured myogenic cells (Schultheiss et al., 1990; Komiyama et al., 1993; van der Loop et al., 1996) as well as during skeletal muscle development in situ (Fürst et al., 1989). Of several antibodies to titin M-line epitopes we tested, only one worked in chicken tissue. Using this antibody (clone T41) we also observed a delayed organisation of the M-line end compared to the Z-disk end. One possible explanation for this lack of signal at time-points when the titin N termini are already clearly spaced is that the M-line ends are not interacting properly yet and that the epitopes for T41 are not concentrated enough to give a signal which can be revealed in the confocal microscope. Alternatively potential epitope masking cannot be ruled out. The M-line end of titin is phosphorylated during development (Gautel et al., 1993) and the epitope of T41 lies in a region, where there is a high occurrence of KSP motifs (Gautel et al., 1993; Obermann et al., 1996). We suggest that weak interactions occur already between the C termini of individual titin molecules while they are unfolding. However, the titin molecules might be stretched out during the changes in cardiomyocyte size and shape between the 8 and 9 somite stage and only the appropriate interactions are strengthened. The increase in cell size is probably transmitted by interactions of α-actinin with the membrane at future costameric sites. Vinculin, the major component of costameres (Pardo et al., 1983), is organised in an interrupted pattern along the membranes already in the 7 somite heart (Tokuyasu, 1989) and evidence from experiments on cultured cardiomyocytes has indicated that costameric structures as well as cell spreading are important for myofibrillogenesis. Antisense oligonucleotides to vinculin inhibit myofibril assembly (Shiraishi et al., 1997b) and when cultured cardiomyocytes are prevented from spreading on the substrate by the addition of cytochalasin D to the culture medium no assembly of new myofibrils occurs (Rothen-Rutishauser et al., 1998).

When the degree of organisation of titin is compared with other myofibrillar components it is apparent that it precedes the accumulation of for example thick filaments (Wang et al., 1988; Handel et al., 1991; Komiyama et al., 1993). This suggests that there is a basic stabilising structure of the sarcomere, which is composed of α-actinin at the Z-disk, titin and an integrating molecule at the M-line. The most likely candidate for this function is myomesin, which is expressed in all kinds of striated muscle (Grove et al., 1989) and becomes localised in its definite pattern from the time when the first sarcomeres can be observed (this study, Auerbach et al., 1997). Myomesin is known to interact with both titin and myosin and might thus serve an analogous function to α-actinin in the Z-disk, namely to link contractile filaments, in this case myosin, to titin (Obermann et al., 1997). Evidence for an independent assembly of myosin filaments during myofibrillogenesis comes from several different experiments. Isolated myosin molecules have the ability to form bipolar filaments even in the test tube (Goldfine et al., 1991) and separate myosin filaments have been detected during myofibrillogenesis in electron microscope studies of intact tissue (Manasek, 1968; Hiruma and Hirakow, 1985) and also in cultured cardiomyocytes (Schultheiss et al., 1990). In the confocal microscope thick filaments can only be resolved once individual myosin filaments have been assembled and were integrated by myomesin at the M-line. Support for the existence of a basic framework in the sarcomere comes from experiments, where sarcomeric components are selectively removed. When the integrity of the thin filaments is destroyed by gelsolin or by expression of inappropriate actin isoforms, the overall appearance of the sarcomeres is maintained and the thick filaments are held in place despite the absence of one of its essential constituents (Funatsu et al., 1990; Von Arx et al., 1995). The basic framework of the sarcomere is not even affected when both thick and thin filaments are removed (Funatsu et al., 1993).

Most of the myofibrillar proteins have been targeted to their proper place in the sarcomere as soon as the first contractions are observed in the 9 somite heart. However, there is an important difference in the way how thick and thin filaments are assembled. Whereas thick filaments have their mature length of 1.6 µm from the first time they can be identified onwards, thin filaments attain their definite length only later during development by the 12 to 13 somite stage (E. Ehler et al., unpublished). By this time nascent intercalated disks are already apparent and the individual myofibrils get properly registrated in their definite orientation along the cell axis, probably by interaction with desmin and also with costameres (Tokuyasu, 1989; Shiraishi et al., 1993, 1995, 1997a). This lateral alignment of myofibrils is the last step during myofibrillogenesis and seems to be important mainly for optimal stress management, since cardiomyocytes from mice, which are homozygous null for desmin or vinculin, respectively, can assemble functional sarcomeres (Li et al., 1996; Milner et al., 1996; Xu et al., 1998).

Future experiments are necessary to test this model of sarcomere assembly for example by ablating the genes for the proposed key players of myofibrillogenesis α-actinin, titin and myomesin. Antisense oligonucleotide experiments inhibiting titin synthesis in cultured cardiomyocytes have underlined its essential role during myofibrillogenesis (Jutta Schaper, personal communication) and overexpression of an N-terminal fragment of titin can lead to a dominant negative phenotype resulting in myofibril disassembly (Turnacioglu et al., 1997), however the effects of perturbation of α-actinin or myomesin remain to be shown.

We thank Prof. Hans M. Eppenberger for his continuing support and members of our group for stimulating discussions. Evelyne Perriard supplied us with primary cultures of embryonic chicken cardiomyocytes. The donation of antibodies by Drs Dieter Fürst, Mathias Gautel and Simon Hughes is gratefully acknowledged. This research was funded by the Swiss National Science Foundation (grant numbers 31.37537/93 and 31.52417/97) and a predoctoral training grant from ETH to B.M.R.