Mouse myoblasts can fuse and form a normal sarcomere in the absence of β1 integrin expression

Skeletal myogenesis starts when the dermomyotome forms from somitic mesoderm. Soon after, mononucleated primary myoblasts start to differentiate and to migrate to their peripheral targets where they become postmitotic, contact each other and fuse into elongated syncytia called primary myotubes or myofibers. This first fusion event is followed by the migration of secondary myoblasts which align along the primary myotubes, fuse and form the secondary myofibers. Finally, primary as well as secondary myotubes form a typical cytoarchitecture by organizing sarcomeres into stacks called myofibrils, become innervated, and differentiate into distinct fast and slow contracting muscle groups.

A large number of experiments suggest that cell adhesion molecules including integrins, cadherins and members of the immunoglobulin superfamily play crucial roles during myogenesis (reviewed by Gullberg and Ekblom, 1995; McDonald et al., 1995a). Integrins are heterodimers composed of an α and a β subunit that bind extracellular matrix (ECM) and cell counter receptors (Hynes, 1992; Haas and Plow, 1994). So far 16 different α subunits and 8 different β subunits are known. The β1 subunit can associate with at least 10 different α subunits forming the largest subfamily of integrins. The cytoplasmic domain of β1 integrin can interact with cytoskeletal proteins such as talin and α-actinin and with signal transducing proteins such as focal adhesion kinase (FAK; Schaller et al., 1995) and integrin-linked kinase (ILK; Hannigan et al., 1996).

Skeletal muscle cells express a large variety of integrins which contain either the β1 or the αv subunit. The β1 subunit is expressed throughout skeletal myogenesis: it is present in myoblasts and in myotubes where it concentrates in myotendinous junctions (MTJ) and costameres (Z-discs of the outer-most myofibrils) (Bozyczko et al., 1989). During myogenesis, β1 integrin appears in distinct cytoplasmic domain splice variants: whereas myoblasts express the commonly found β1A form, adult fibers synthesize the musclespecific β1D variant (van der Flier et al., 1995; Zhidkova et al., 1995; Belkin et al., 1996). The expression of the αv subunit which can associate with 5 β subunits (β1, β3, β5, β6, β8) is high during the embryonic and fetal period of muscle development and is low in the mature adult muscle tissue (Hirsch et al., 1994).

Skeletal muscle cells express eight different α subunits which associate with β1 integrin, and most of them show a developmentally regulated expression pattern. In the chick α1 (Duband et al., 1992), α3 (McDonald et al., 1995b), α5 (Enomoto et al., 1993), α6 (Bronner-Fraser et al., 1992) and αv (McDonald et al., 1995b) are present in the embryo and down-regulated in the adult. α4β1 integrin is made during secondary myogenesis on primary myotubes where it colocalizes with its ligand VCAM-1 which is expressed by secondary myoblasts (Rosen et al., 1992). α7 integrin is abundantly expressed at all stages of muscle development (Bao et al., 1993) and appears in several alternatively spliced forms (Collo et al., 1993; Martin et al., 1996; Velling et al., 1996). The adult mature skeletal muscle expresses small amounts of α9β1 (Palmer et al., 1993).

Most of these integrin receptors are distinctly localized on the membrane of the myofiber. In developing myotubes, for example, the α5 subunit is localized in adhesion plaque-like structures along the myotube (Lakonishok et al., 1992). Conversely, the αv subunit is enriched at the costamere (McDonald et al., 1995b) and, in the mouse, at the MTJ (Hirsch et al., 1994). In the chick embryo, α3 localizes transiently at the MTJ (McDonald et al., 1995b) where it is then replaced by α7 (Bao et al., 1993).

The restricted expression pattern of β1 integrins together with results from anitbody perturbation experiments suggested a role in several steps of myogenesis such as myoblast migration (Jaffredo et al., 1988), differentiation and fusion into myotubes (Menko and Boettiger, 1987; Rosen et al., 1992; McDonald et al., 1995b) and sarcomere formation. For example, α4β1 and its ligand VCAM-I have been implicated to have a role in secondary myotube formation. These findings are contradicted by the observation that myoblasts derived from α4 integrin-deficient ES cells form secondary myotubes both in vivo and in vitro (Yang et al., 1996). Studies on quail primary myoblasts indicate that α5β1 and α6β1 modulate myogenesis by balancing proliferation and differentiation (Sastry et al., 1996). Overexpression of each of these two receptors shows a distinct effect: whereas the expression of α5 correlates with active proliferation, α6 suppresses cell divisions and induces differentiation. Finally, several lines of evidence suggest that β1 integrins are involved in the organisation of sarcomeres. First, in the Drosophila mutant lacking the β1 integrin homologue βPS (Bogaert et al., 1987; Leptin et al., 1987), muscles detach from the insertion sites after the first contractions and give rise to a spheroidal morphology of the developing embryo. In culture, βPS-deficient myotubes show a rounded morphology and severely malformed sarcomeres in which Z-bands fail to form (Volk et al., 1990). Second, anti-β1 integrin antibodies delay the organisation of α-actinin along Z-bands of quail myotubes (McDonald et al., 1995b). Third, β1-null cardiac muscle cells have a shorter life span in vivo which is associated with abnormal sarcomeres (Fässler et al., 1996).

The ablation of the β1 integrin gene in mice leads to periimplantation lethality (Fässler and Meyer, 1995; Stephens et al., 1995). Despite this early lethal phenotype β1-null ES cells can contribute to the formation of many but not all organs in β1-null chimeric mice. The chimerism, for example, is high in skeletal muscle (Fässler and Meyer, 1995) and skin (Bagutti et al., 1996), but low in most other tissues and absent in hematopoietic organs (Hirsch et al., 1996). In this paper we report that β1-null myoblasts contribute to trunk as well as limb muscles in vivo. Furthermore, we show that β1-null myoblasts derived from chimeric embryos as well as newborn mice can fuse and form sarcomeres which appear to be organized like control sarcomeres. These results were confirmed by in vitro differentiation of β1-null ES cells into embryoid bodies. Although myoblast fusion was significantly slower than in wild-type controls, embryoid bodies lacking β1 integrin contained fully differentiated myotubes. These data suggest that several events during skeletal myogenesis such as myoblast migration, fusion and sarcomere formation occur without β1 integrin.

The following ES cells were used: 2 wild-type (+/+) ES cell lines D3 (Doetschman et al., 1985) and R1 (Nagy et al., 1993); G119 which is heterozygous (+/ −) for the β1 integrin gene mutation and D3-derived; G101 which is wild type (+/+) for β1 integrin gene, mock transfected and D3-derived; G201 which is β1-null (−/ −) and D3-derived (Fässler et al., 1995) and G110 which is β1-null (−/ −) and R1-derived (Fässler and Meyer, 1995). The cell clones G119, G201 and G110 have a fusion cDNA of β-galactosidase and neomycin (geo) inserted in frame with the ATG of the β1 integrin gene (Fässler et al., 1995). The cell clone G101 contains a randomly integrated β-galactosidase gene which is ubiquitously expressed (E. Hirsch and R. Fässler, unpublished).

ES cells were cultured either in the presence (for the generation of chimeras) or in the absence (for the generation of embryoid bodies) of a fibroblast feeder layer in DMEM supplemented with 20% heat-inactivated fetal calf serum (FCS; GibcoBRL, Gaithersburg, MD), 0.1 mM β-mercaptoethanol (Sigma Chemical Co, St Louis, MO), 1× nonessential amino acids (Gibco-BRL) and 1,000 U/ml recombinant leukemia inhibiting factor (Gibco-BRL).

The following primary antibodies were used: rabbit anti-β1 integrin (Bottger et al., 1989); rabbit anti-αv (Hirsch et al., 1994;); rabbit anti-myoD, anti-myf5, anti-myogenin (all gifts from Thomas Braun, University of Braunschweig, Germany); rabbit anti-proliferating cell nuclear antigen PCNA (a gift from Bjorn Olsen, Harvard Medical School, USA); rabbit anti-α7B (Velling et al., 1996); sheep anti-α/β dystroglycan (a gift from Kevin Campbell); mouse monoclonal antibody (mAb) against desmin (Dako, Germany); mouse mAb against sarcomeric α-actinin (Sigma A-7811); mouse mAb against titin (gift of Dieter Fürst, University of Potsdam, Germany); mouse mAb against skeletal myosin (fast; Sigma; M4276); mouse mAb against myosin heavy chain (slow; Medac Diagnostika, Germany; clone WB-MHCs); mouse mAb against myosin heavy chain (fast; Medac Diagnostika; clone WB-MHCf); mouse mAb against vinculin (Sigma; clone no.vin-11-5).

The following secondary antibodies were used: goat anti-rabbit FITC (Jackson Immunoresearch Laboratories, USA); goat anti-mouse CY3 (Jackson Immunoresearch Laboratories).

For generating chimeric mice, blastocysts were isolated at day 3.5 post coitum (p.c.) from C57BL/6 mice, injected with 15 ES cells and transfered into the uterus of pseudopregnant recipient (C57BL/6xDBA)F1 females (2.5 days p.c.). Tissue sections were prepared from four 6-week-old male chimeric mice which were transcardially perfused with phosphate buffered saline (PBS) followed by 4% paraformaldehyde, 0.02% NaN3 in PBS. After a short postfixation and equilibrium in 30% sucrose the tissues were embedded in OCT compound (Miles), frozen in dry ice-pentane and stored at -80°C. Tissues were cut into 10-20 μm sections on a Leitz cryostat and collected on gelatin-coated slides. LacZ staining and counterstaining with eosin followed published protocols (Fässler and Meyer, 1995). Finally, sections were covered with Canada Balsam, coverslipped and examined with a Zeiss AXIOPHOT microscope.

Hindlimbs and forelimbs were dissected from chimeric mice at E16 or after birth. The remaining body was assayed for lacZ expression (Fässler and Meyer, 1995) to determine the extent of chimerism. Muscles were isolated, minced and enzymatically dissociated with a mixture of collagenase B (0.1%; Boehringer Mannheim) and dispase (grade II, 2.4 U/ml; Boehringer Mannheim). The released cells were cultured in a 1:1 mixture of DMEM and F-10 media supplemented with 20% FCS and penicillin/streptomycin. Cultures were treated for 10 days with 800 μg/ml of G418 (GibcoBRL) to kill all wild-type cells derived from the host blastocysts.

For differentiation of mononucleated cells into myotubes, cultures were plated at subconfluence on 4-well Lab-Tek chambers (Nunc, Germany) and incubated up to 10 days in DMEM supplemented with penicillin/streptomycin and 2% horse serum.

For in vitro differentiation, ES cells were trypsinized and resuspended at 3×10⁴ cells/ml in differentiation medium. Drops of 20 μl (corresponding to 600 cells) were put onto the lid of a bacterial Petri dish filled with PBS. After inversion of the lid, cells were incubated in hanging drops at 37°C, 5% CO2 for 2 days. This culture method leads to ES cell aggregates which differentiate and form so called embryoid bodies (EBs). Afterwards, EBs were transferred to fresh bacteriological dishes and incubated in differentiation medium for 8 more days resulting in a total incubation period of 10 days in suspension. Finally, EBs were plated on gelatin-coated glass coverslips where they were allowed to adhere and incubated as indicated in the result section.

To determine culture conditions for skeletal muscle differentiation in vitro, 10 EBs were plated on a gelatin-coated well of a 12-well plate. Five different differentiation media were tested: DMEM supplemented with penicillin-streptomycin, 15% FCS; DMEM supplemented with penicillin-streptomycin, 1× non essential amino acids, 0.1 mM β-mercaptoethanol, 5% horse serum (Boehringer Mannheim, Germany), 10% FCS (5:10 medium); DMEM supplemented with penicillin-streptomycin, 1% DMSO and 15% FCS; DMEM supplemented with penicillin-streptomycin, 15% horse serum; DMEM supplemented with penicillin-streptomycin, 1× nonessential amino acids, 0.1 mM β-mercaptoethanol, 15% dextran charcoal-treated FCS. Each medium was tested on an entire 12-well plate. After plating, differentiation was scored daily for 30 days by microscopic inspection of myotubes. The percentage of differentiation was calculated as the number of EBs with a group or a field of myotubes per 100 EBs.

All subsequent experiments were performed by plating 8-10 EBs on glass coverslips coated with 0.1% gelatin followed by culture in 5:10 medium. EBs were kept in culture for a maximum period of 30 days. Medium was changed every day. An incubation of X number of days after plating is indicated as 10+X d.

The fusion index (a measuring unit for the ability of cells to fuse into multinucleated cells) was calculated as the ratio of nuclei in myotubes (spindle like cells with more than 3 nuclei) to the total number of nuclei in a field of myosin heavy chain positive cells.

For immunofluorescence analysis cells were fixed for 5 minutes in 4% paraformaldehyde in PBS, permeabilized with 0.1% Triton X-100 in PBS for 10 minutes and incubated with primary antibodies for 1 hour at room temperature. After washing with PBS specimens were incubated with secondary antibody for 1 hour at room temperature.

Cell death was analyzed following the protocol supplied by the manufacturer (Boehringer Mannheim, Germany). Briefly, embryoid bodies were fixed in 4% paraformaldehyde in PBS, postfixed for 2 minutes in methanol at −20°C, treated for 2 minutes in 0.1% Triton X-100 in 0.1% sodium citrate, and incubated with terminal deoxynucleotidyl transferase which catalyzes the addition of fluorescein-conjugated dUTP to free 3’-hydroxyl groups in apoptotic cells. For postive control, cells were treated for 2 hours with 9 μM herbimycinA before fixation.

Mutant and control myoblasts isolated from mouse embryos were seeded either on cell culture Petri dishes (Falcon, Germany) or on permanox plastic Lab-Tek chambers (Nunc) and differentiated as described above. Myotubes were fixed in a 100 mM Hepes/Pipes buffer (pH 7.35) containing 1.75% paraformaldehyde, 2% glutaraldehyde and 15% picric acid for 1 hour at room temperature. Afterwards myotubes were treated with 100 mM Hepes/Pipes buffer containing 1% tannic acid for 30 minutes at room temperature and finally osmified with 0.5% OsO₄. Prior to embedding in Epon resin (Agar Scientific, Stansted, UK) myotubes were dehydrated in a graded series of ethanols. Ultrathin sections (30-60 nm) were mounted on Formvar-coated copper grids, stained with 0.2% uranyl acetate and lead citrate and examined with a Zeiss EM 902A electron microscope.

In our previous study we reported that β1-null myoblasts can migrate to the forelimb and contribute to the formation of the triceps muscle. At least two different myogenic subpopulations develop in the dermomyotome which form the musculature of limb and trunk, respectively. The facial muscle is derived from prechordal mesoderm. To test whether all subpopulations can form and reach their targets in the absence of β1 integrin, forelimb muscles (M. triceps, M. biceps), hindlimb muscles (M. quadriceps, M. soleus, M. gastrocnemius), trunk muscles (M. psoas, intercostal muscle) and facial muscles were assayed for lacZ activity in four 6-week-old β1-null chimeric mice. Only a small number of β1-null cells allow normal development of chimeric mice (Fässler and Meyer, 1995). Therefore, we had to inject 138 blastocysts to obtain four β1-null chimeric animals with a coat color contribution derived from the mutant ES cells which ranged between 10 and 25%. Two of these chimeric mice showed lacZ staining in all muscle samples (Fig. 1B shows lacZ staining in M. triceps). One chimeric mouse had lacZ-positve myotubes in forelimb and body wall musculature and none in hindlimb and facial muscles. The fourth chimeric mouse contained a few lacZ-positve areas in the forelimb muscles, none in the hindlimb muscles and areas with strong lacZ staining in the M. psoas (Fig. 1A). The intensity of the lacZ staining varied strongly between myotubes (Fig. 1A,B) in all animals tested. The variable degree of chimerism was also confirmed by comparing the amount of wild-type and ES cell specific isoforms of glucose phosphate isomerase (GPI) in muscle tissues of β1-null chimeric mice. Similar to an earlier report from our laboratory the contribution of β1-null cells to muscle ranged between 2 and 28% and as a consequence of this low chimerism all lacZ-positive myotubes expressed β1 integrin (see Fässler and Meyer, 1995, and data not shown). Muscles from all β1-null chimeric mice analysed so far showed no histological alterations specific for a muscular dystrophy.

These data indicate that β1-null myoblasts can form and migrate to all peripheral targets where they are able to fuse with wild-type myoblasts to form chimeric myotubes which express β1 integrin.

To test whether β1-null myoblasts are able to differentiate and fuse with each other, primary cultures of β1-deficient myoblasts were established from the limb muscles of β1-null chimeric 16-day-old embryos and newborns, respectively. In seven experiments, muscle tissue was dissected from 8-10 chimeric embryos or newborns, respectively. Afterwards, embryonic myoblasts or satellite cells were dissociated from the fibers and cultured in the presence of high concentration of G418. Wild-type cells derived from the host embryo did not express the neomycin resistance gene and hence were sensitive to the G418 selection. β1-null cells were derived from the targeted ES cells and, therefore, expressed the neomycin resistance gene. Due to the high G418 concentration most of the wild-type cells died within the first 4 days. Immunofluorescence assays indicated that after a culture period of 7 days all surviving cells were β1 integrin-deficient. Myoblasts were, however, routinely selected for 10 days before fusion was induced. Control cultures were obtained either by culturing the cells without G418 (giving rise to a mixed population of wild-type and mutant cells) or by isolating neomycin-resistant myoblasts from chimeras generated by injecting β1 integrin heterozygous ES cells or mock transfected ES cells, respectively.

For the present study only normal β1-null chimeric embryos were analysed. The contribution of β1-null cells in chimeric embryos usually does not exceed 25% (see also Fässler and Meyer, 1995). Therefore, a large number ofblastocysts was injected and transferred into foster mice: 425 blastocysts with β1-null ES cells, 47 blastocysts with β1 heterozygous ES cells, 92 blastocysts with either wild-type or mock transfected ES cells. To increase the yield of β1-null cells, muscles from 8-10 chimeras were pooled for each experiment. The selected β1-null myoblasts were expanded and could be propagated for more than 5 passages, indicating that β1-null myoblasts can proliferate in the absence of β1integrin.

Usually, myoblasts are enriched by panning contaminating fibroblasts on a fibronectin-coated plastic dish. Since the ablation of β1 integrin alters adhesion to fibronectin, subconfluent monolayers of a mixed population of G418-resistant myoblasts and fibroblasts were cultured in differentiation medium for 10 days during which myotube formation was examined daily by phase contrast microscopy. The first elongated myofibers appeared after approximately 45 days and fusion of myoblasts was completed after 10 days both in β1-null and control cultures, indicating that terminal differentiation proceeds apparently normally in the absence of β1 integrins. Myofibers were analyzed by immunofluorescence using antibodies specifically reacting with β1 integrin (Fig. 2C,D), desmin (Fig. 2A,B), myosin light chain, myosin heavy chain, α-actinin and titin. Whereas both control and β1-null myotubes expressed desmin (Fig. 2A,B), α-actinin (see Fig. 7A,B) titin, and various myosin chains (not shown) at similar levels and with similar distribution, β1 integrin was only made by wild-type myotubes (Fig. 2C,D). Staining of nuclei with bisbenzimide revealed that normal and β1-null myofibers were multinucleated (Fig. 2A,B) containing a minimum of 4 nuclei. The average number of nuclei did not vary between normal and β1-null myotubes.

These results were obtained with embryonic myoblasts as well as newborn satellite cells and clearly demonstrate that a pure population of β1-null myoblasts can efficiently fuse to form multinucleated myotubes.

To confirm our results with a second independent experiment, β1-null ES cells were differentiated in embryoid bodies. To obtain comparable differentiation conditions for normal and β1-null ES cells, 5 different media were tested for their ability to induce myogenic differentiation. Normal and β1-null embryoid bodies were cultured 10 days in suspension and then 30 days on gelatinzed glass coverslips. The percentage of embryoid bodies that contained groups of myotubes was scored daily for each culture condition. Myogenic differentiation in normal as well as β1-null embryoid bodies was strongly influenced by the different culture conditions (Fig. 3). Whereas culture medium containing 15% horse serum favored differentiation of wild-type myotubes, many cells in β1-null embryoid bodies started to detach from the well. The remaining cells rarely differentiated into myotubes. The presence of DMSO in the FCS as well as DCC-stripped FCS increased the number of myotubes as compared to regular or untreated FCS both in normal and β1-null embryoid bodies. Culture medium containing 5% horse serum and 10% FCS allowed efficient and similar myogenic differentiation in normal as well as β1-null embryoid bodies. Since the latter conditions biased none of the ES cell lines all further experiments were performed in the presence of 5% horse serum and 10% FCS.

Microscopic inspection of embryoid bodies indicated that myotube formation occurred but was significantly delayed in β1-null embryoid bodies. Immunostaining for fast myosin heavy chain confirmed this observation. Fig. 4 shows a representative area in normal and β1-null embryoid bodies cultured for 10 days in suspension and then for 5, 15 or 25 days, respectively, on gelatinized glass coverslips. At 10+5 days half of the myosin heavy chain-positive cells in normal embryoid bodies have fused and have formed big nests of myotubes (Fig. 4A and Fig. 5). At 10+15 days fusion of normal myoblasts was completed (Fig. 4C and Fig. 5). In contrast, myotube formation was delayed in β1-null embryoid bodies: it occurred rarely at 10+5 days (Fig. 4B and Fig. 5), increased to 30% at 10+15 days and to 50% at 10+25 days (Fig. 4D,E and Fig. 5).

A recent study demonstrated that cells expressing myosin heavy chain are postmitotic and committed to terminally differentiate into myotubes (Andres and Walsh, 1996). To test whether the population of non-fused, mononucleated myosin heavy chain-positive cells in β1-null embryoid bodies are abnormally committed and still proliferating, embryoid bodies were stained with an antibody recognizing DNA polymerase processing factor PCNA which is only present in dividing cells. In all β1-null embryoid bodies tested, none of the cells which expressed myosin heavy chain stained for PCNA (Fig. 6A,B) clearly indicating that these cells were postmitotic. To exclude the possibility that this cell population is undergoing apoptosis, β 1-null embryoid bodies were double labeled for myosin heavy chain and the presence of nicked DNA. None of the round, myosin heavy chain-positive cells was apoptotic under normal conditions (Fig. 6C). After addition of 9 μM herbimycinA to β1-null embryoid bodies, however, apoptotic cells were readily detectable (Fig. 6D, see green nuclear staining).

These data indicate that β1-null myoblasts behave like normally committed myoblasts: they are postmitotic and, in the presence of serum, protected from apoptosis (Wang and Walsh, 1996).

To test whether muscle-specific genes are activated during differentiation of β1-null myogenic cells, myotubes were stained for the presence of muscle-specific transcription factors and components of sarcomeres. Nuclei of both normal and β1-null myofibers contained similar amounts of MyoD (Fig. 7C,D), Myf-5 and myogenin (not shown). Immunostaining of primary cultures isolated from normal and β1-null chimeric embryos showed that the expression of MyoD appeared 2 days after switching to the differentiation medium both in normal and β1-null myotubes. These results indicate that the absence of β1 integrin does not alter the expression of early markers of muscle differentiation.

In a next experiment mutant myofibers derived from chimeric mice and embryoid bodies were tested for the expression of sarcomeric components typically expressed and assembled by terminally differentiated skeletal muscle fibers. Immunostaining revealed a highly organized distribution of α-actinin (Fig. 7A,B), titin (not shown) and several myosin heavy chain isoforms (Fig. 8) in normal and β1-null myofibers. In particular, myosin appeared in repeated units of myofibrils which correctly organized in register generating the striated pattern typical of muscle cells (Fig. 8A-D). This typical distribution of sarcomeric proteins was evident in myofibers derived from β1-null chimeric embryos and newborn mice (Fig. 7A,B) as well as β 1-null embryoid bodies (Fig. 8). Normal and mutant myoblasts derived from both chimeric mice and embryoid bodies expressed the slow myosin heavy chain. Moreover, many fibers of both genotypes also expressed a fast twitch myosin heavy chain which is a distinctive marker of type II myofibers (Fig. 8C,D). Since type II myofibers are supposed to originate from the second generation of myoblasts (Kelly and Rubinstein, 1980), the presence of the fast type II specific myosin isoform suggests that both primary and secondary myogenesis can occur in the absence of β1 integrins.

To test for subtle changes in the architecture of sarcomeres, myotubes derived from normal and β 1-null chimeric embryos were analysed at the ultrastructural level using electronmicroscopy. The mesh of actin/myosin fibrils was organized in stacks and Z bands correctly appeared at the border of each contracting unit both in normal (Fig. 9A) and β1-null myotubes (Fig. 9B). Most filament bundles showed a regular deposition and were arranged in parallel orientation. The distance between two Z bands was constant and approximately 1.8 μm in length in normal and β1-null myotubes (Fig. 9A,B). Whereas in all specimens the M-line was scarcely developed, A bands containg thick myosin filaments and I bands consisting of thin actin filaments were detectable on both sides of Z bands.

These data together with the results from the immunostaining show that sarcomeres assemble and form apparently normally in the absence of β1 integrin expression.

Integrin subunit α7 can associate with the β1 subunit to form a functional integrin which is highly expressed on myotubes. To exclude the possibility that in the absence of β1 integrin the α7 subunit associates with an unknown or unusal β subunit, we performed double immunostaining of α7 (Fig. 10A,B) and myosin heavy chain (Fig. 10C,D) on control and β1-null myotubes. Antibodies showed the expected expression pattern of α7 integrin on normal mytubes (Fig. 10A) and demonstrated its absence on β1-null myofibers (Fig. 10B).

Skeletal muscle expresses two subfamilies of integrins, one containing the β1 subunit and the other containing the αv subunit. Previous reports demonstrated that αv integrins are distributed along Z-bands (McDonald et al., 1995b) and the MTJ region (Hirsch et al., 1994) in embryonic myofibers of vertebrates. On the other hand, β1 integrins are found all over the sarcolemma (Bozyczko et al., 1989). To test whether the absence of β1 integrins alters the expression of αv integrins, normal and β1-null myotubes were doublestained for myosin heavy chain and αv integrin, respectively (Fig. 10E-H). Like in wild-type myotubes, the localisation of the αv integrin subunit was high at the tip of the β1-null myofibers and did not signficantly extend into the interior of the tube (Fig. 10F,H). In addition, the staining intensity was similar on both normal and β1-null myotubes.

The dystrophin/dystroglycan complex is another important adhesion molecule expressed on myotubes. Like integrins, this complex links the ECM with the actin cytoskeleton (Henry and Campbell, 1996). Immunostaining of normal and β1-null myotubes revealed that the distribution of dystroglycan was not altered on β1-null myotubes (Fig. 11A,B).

Several membrane-associated cytoskeletal components of the costameres interact directly or indirectly with the cytoplasmic domain of β1 integrin. To test whether costameres are formed, mature β1-null myotubes were immunostained for vinculin (Fig. 11C,D) and talin (not shown). The expression pattern of both vinculin and talin was similar in normal and β1-null myotubes.

Altogether these data demonstrate that the absence of β1 integrin on myotubes does not lead to an apparent alteration of αv integrin, talin and vinculin expression. Dystroglycan is expressed on normal and on β1-null myotubes and could, at least partially, compensate for the lack of β1 integrin.

In the present study we examined the role of β1 integrin during skeletal myogenesis of the mouse using genetically altered mice and ES cells. Several previous reports have shown that members of the β1 integrin subfamily control several events during muscle development including migration of myoblasts, terminal differentiation and myoblast fusion. Studies from Drosophila also suggest that β1 integrin may regulate sarcomere assembly. Despite these evidences we found that muscle development proceeds without obvious defects in the absence of β1 integrin when β1-null myoblasts were isolated from β1-null chimeric embryos and mice and fused. Moreover, formation of normally developed myotubes can also occur in β1-null embryoid bodies although it is significantly delayed.

In a previous report we have already shown that limb muscle of β1-null chimeric mice contains myotubes which are positve for lacZ expression (Fässler and Meyer, 1995). Since lacZ-positive myotubes also express β1 integrin these results indicated that β1-null cells differentiate in the dermomyotome, migrate to the limb region and fuse with normal myoblasts to form a chimeric myotube. There is clear evidence, however, that muscle in limbs and trunk originates from different sublineages in the dermomyotome: two different sublineages in the dorso-medial part contribute to the myotome which produce the epaxial muscle of the deep back and the hypaxial muscle of the body wall, respectively, and a third sublineage located in the lateral part generates the musculature of the limbs (reviewed by Yun and Wold, 1996; Rawls and Olson, 1997). Therefore, it was important to extend our previous investigation and analyse muscle from various trunk as well as limb regions of β1-null chimeric mice. Extensive analysis of several β1-null chimeric mice showed that β1-null myoblasts contribute to muscle tissue derived from all three myotomal compartments. Furthermore, the extent of contribution to the musculature was similar in limbs and trunk indicating that none of the sublineages is favored in the absence of β1 integrins. The presence of lacZ-positive myotubes is a clear indication that β1-null myoblasts are able to migrate from the dermomyotome to peripheral targets, contrasting earlier reports showing that antibodies against β1 integrin inhibit myoblast migration (Jaffredo et al., 1988). In line with these myoblast results are observations showing that several migratory cell types including neural crest cells and neuroblasts can reach their targets without expressing β1 integrin (Fässler and Meyer, 1995). Also for these cells, antibody perturbation assays or expression of antisense mRNAs have suggested that β1 integrin plays a crucial role for their ability to migrate in vivo (Thiery et al., 1985; Galileo et al., 1992). Blocking experiments with antibodies against distinct integrin subunits have extended these observations and identified α4β1 integrin for mediating migration of neural crest cells (Sheppard et al., 1994) and α7β 1 integrin for mediating migration of myoblasts (Echtermeyer et al., 1996; Yao et al., 1996). These findings, however, could not be confirmed in knockout experiments: mice lacking α4 develop neural crest-derived tissues (Yang et al., 1996) and mice lacking α7 have a normal musculature at birth (Mayer et al., 1997) indicating that migration is not affected by the absence of these integrins. One possible explanation for the discrepancies could be that migration of β1-null myoblasts is not a cell-autonomous process but achieved by the interaction of mutant and normal cells. It is also possible that only certain cells in the dermomyotome, a sublineage of dermomyotomal cells, can migrate without β1 integrin expression. Finally, other integrins or adhesion molecules may compensate for the lack of β1 integrin resulting in a cell-autonomous ability of migration. One way to distinguish between these alternatives is to generate mice which are β1-null in all myoblasts but are otherwise normal. This can be achieved by introducing a conditional null mutation into the β1 integrin gene.

After reaching peripheral target sites myogenic cells stop to proliferate, become postmitotic and fuse into multinucleated myotubes. When β 1 integrins of primary chicken myoblasts are blocked in vitro with the monoclonal antibody CSAT they remain in a proliferative state, do not express markers of terminal differentiation and fail to fuse (Menko and Boettiger, 1987). These results suggested that β1 integrins control the cell cycle of myoblasts and are crucial for entering the GO phase. A direct role of β1 integrin in myoblast fusion was suggested from the expression pattern of α4β1 and its counter-receptor VCAM-1 on the surface of primary myotubes and secondary myoblasts, respectively (Rosen et al., 1992). The requirement of this interaction for myogenesis was demonstrated by inhibiting myotube formation in cultures of the mouse myogenic cell line C2C12 with antibodies against α4β1 or VCAM-1 (Rosen et al., 1992). Nevertheless, we demonstrate that these events can occur in the absence of β1 integrin expression. This could be shown in two experimental systems. First, we isolated myoblasts from β1-null chimeric embryos and newborn mice, respectively. To isolate β1-null myoblasts from wild-type cells, cultures were treated for several days with high concentrations of G418 which kills normal cells but allows mutant cells to survive. Afterwards the remaining myoblasts were fused and analysed. β 1-null myoblasts from all stages of development fused as efficiently as their normal counterparts. Furthermore, the expression of muscle-specific transcription factors and of sarcomeric proteins occurred at the same time in β1-null and normal myotubes. These data extend our observation from muscle tissue in β1-null chimeric mice and demonstrate that fusion and terminal differentiation of β1-null myoblasts can also occur in the complete absence of wildtype myoblasts. Second, we have differentiated normal and β1-null ES cells in embryoid bodies. Also here, we were able to obtain fully differentiated β1-null myotubes which were indistinguishable from their control counterparts. Again, several explanations can be considered for the difference between our results and previous reports: first, the absence of β1 integrins is compensated by other integrins or adhesion molecules in totipotent cells but not after cell lineage decisions have occurred. Immuno staining of normal and β1-null myotubes showed that the αv integrin subunit remained mainly confined to the tips of β1-null myofibers and did not significantly spread throughout the sarcolemma compensating for the lack of β1 integrin at sites where it is normally located (Bozyczko et al., 1989). This finding does not exclude the possibility that αv is still slightly upregulated and/or redistributed which would escape detection with the assays used in this study. More acurate comparisons (western blotting, immunoprecipitation or RNA analyses) of αv integrin expression are difficult to perform with embryoid bodies which are composed of many different cell types expressing different amounts of αv integrin. Second, functional redundancy could allow formation of β1-null myotubes. A candidate receptor is dystroglycan which, like several β1 integrins, binds laminin and is expressed throughout the myofiber (Henry and Campbell, 1996). Immunostaining revealed a similar distribution of dystroglycan on normal and β1-null myotubes. Third, technical problems of antibody perturbation experiments may have led to wrong interpretations of former studies. For example, antibodies cluster β1 integrins and may activate selective intracellular signaling pathways (Miyamoto et al., 1995). Furthermore, integrin-specific inhibitors may induce trans-dominant effects that block the function of other integrins in the same cell (Diaz-Gonzalez et al., 1996). Therefore, activities of blocking anti-β1 integrin antibodies may trigger cellular responses which go beyond the sole blockage of receptor occupancy and inhibition of cell adhesion.

The in vitro experiments with β1-null embryoid bodies clearly indicate that some β1 integrins modulate myotube formation under certain conditions. Although fully differentiated myotubes were present in β1-null embryoid bodies their formation was significantly delayed and much less efficient when compared to normal embryoid bodies. Similar differences with our β1-null ES cells were obtained by Rohwedel et al. (unpublished). At present we cannot offer experimental explanations for the different results obtained with β1-null chimeric mice and β1-null embryoid bodies. One explanation for these differences is that β1-null myoblasts from mice differentiate and migrate in their normal environment where they are exposed to appropriate growth factors and substrates at the right time and at the correct concentration. This is clearly not the case in embryoid bodies where the cells differentiate in the presence of bovine and horse serum. To evaluate the influence of culture medium on myotube formation we have tested several different culture conditions and found that the amount and type of serum greatly influences the formation of myotubes in β1-null but not so much in normal embryoid bodies. An alternative explanation comes from a recent report indicating that a balanced expression of α5β1 and α6β1 is important for a regulated myoblast proliferation and differentiation (Sastry et al., 1996). Whereas ectopic expression of α5β1 keeps myoblasts in the proliferative state, overexpression of α6β1 inhibits proliferation but not differentiation. Moreover, decreasing α6β1 by antisense mRNA expression leads to an increased proliferation rate and to a reduction of myoblast fusion. Whereas myoblasts from either β1-null chimeric embryos or mice did not show an obvious alteration in proliferation and/or differentiation, many myoblasts derived in β1-null embryoid bodies were incapable of fusing even after an extended culture period. It is conceivable that cells expressing neither α5 nor α6 containing integrins remain in an unbalanced state. The lack of α5β1 could have blocked their proliferation (these myocytes are post-mitotic and apoptosis protected) but the simultaneous absence of α6β1 could have delayed their differentiation and fusion into myotubes. Although these explanations may account for the delayed differentiation in embryoid bodies, they do not operate during differentiation of β1-null myoblasts derived from β1-null chimeric mice where maybe other adhesive systems are operating and replacing them.

Additional evidence that myoblasts can form and fuse into myotubes without β1 integrin comes from genetic studies of invertebrate mutants. Deletion of the β1 integrin homolog (βPS) in Drosophila enables differentiation of myoblasts and formation of myotubes which, however, have altered sarcomeres lacking Z-bands (Volk et al., 1990). Similar defects were observed in muscle of C. elegans lacking βPAT-3, the β1 integrin homolog of worm (Hresko et al., 1994; Williams and Waterston, 1994; Gettner et al., 1995). Whereas mouse β1-null myoblasts differentiate and fuse, we did not find evidence that the expression of β1 integrin is necessary for the assembly of sarcomeres. Moreover, we found that β1-null myotubes could normally distribute costameric components such as vinculin and talin, suggesting that the organisation of the cytoarchitecture in mouse myotubes is achieved in a β1 integrin-independent manner. Similar observations have been reported for focal adhesion sites in Drosophila myotubes where these intracellular scaffold structures form first and then drive βPS to its final localization (Martin-Bermudo and Brown, 1996). Once located to the areas of mechanical stress, β1 integrin may organize specialized attachments necessary to stabilize the connection between the contractile apparatus and the surrounding basement membrane. Such a model would imply that the major role of β1 integrin in skeletal muscle cells is not the formation but the long-term maintenance of the myofibrillar structure. Although our data provide only indirect evidence for this hypothesis, several studies are in agreement with such a view. Antibody perturbation experiments with quail myotubes show that the normal distribution of α-actinin in Z bands is lost when myotubes loose their anchorage (McDonald et al., 1995a,b). We have recently reported that β1 integrins might have a crucial function in providing strength for the attachment of sarcomeres in cardiac muscle cells. β1-null cardiomyocytes differentiate in embryoid bodies and in β1-null chimeric mice (Fässler et al., 1996). The differentiation of β1-null cardiomyocyte is delayed and associated with an agedependent disorganization of sarcomeres. Additional evidence for a role of a member of the β1 integrin subfamily in maintaining sarcomeres comes from the analysis of α7-null mice. The absence of α7β1 does not impair skeletal myogenesis but leads to the degeneration of MTJ and sarcomeres and to the development of a muscular dystrophy (Mayer et al., 1997).

We thank Stefan Benkert for expert technical assistance, Drs Thomas Braun, Dieter Fürst, Bjorn Olsen and Kevin Campbell for antibody gifts and Drs Rick Horwitz, Cord Brakebusch, Uwe Rauch and Bernhard Bader for critically reading the manuscript. This work was supported by the Swedish National Reserach Foundation (no. 12091-300 to R.F.), the Swedish Medical Research Council (no. 7147 to S.J.) and the King Gustaf V’s 80-års fond.