Talin 1 and 2 are required for myoblast fusion, sarcomere assembly and the maintenance of myotendinous junctions

Skeletal muscle development and function are dependent on β1 integrins, a family of cell surface receptors that are formed by heterodimerization of the β1 subunit with different α subunits(Hynes, 1992). In skeletal muscle, β1 integrins localize to costameres and myotendinous junctions(MTJs), where they establish a link between the cytoskeleton and the extracellular matrix (ECM) (Mayer,2003). These connections are important for transmitting mechanical forces and for maintaining skeletal muscle fibers. Accordingly, defects in integrin function lead to muscle fiber degeneration: mutations in the gene encoding the α7 integrin subunit cause congenital myopathy in humans(Hayashi et al., 1998), and genetic ablation of either the α5 or α7 integrin subunits causes muscular dystrophy in mice (Mayer et al.,1997; Taverna et al.,1998). Integrins appear to be particularly important at MTJs. Inactivation of the α7 integrin subunit leads to detachment of MTJs from the ECM (Miosge et al., 1999),whereas inactivation of integrin-linked kinase (ILK) and talin 1 lead to detachment of integrin adhesion complexes from the muscle fiber cytoskeleton(Wang et al., 2008; Conti et al., 2008). Integrins also have important functions during skeletal muscle development, as ablation of the murine β1 integrin subunit gene, which leads to loss of allαβ1 integrins, causes defects in myoblast fusion and sarcomere assembly (Schwander et al.,2003). The mechanisms by which integrins carry out their function in skeletal muscle are still incompletely understood.

Integrins assemble signaling complexes at the plasma membrane, which contain proteins that bind to the integrin cytoplasmic domains or are recruited indirectly (Geiger et al.,2001; Liu et al.,2000). Several lines of evidence suggest that talin 1 is central for integrin signaling. Talin 1 interacts with the cytoplasmic domain ofβ1 integrins (as well as several other β subunits) and with focal adhesion components such as focal adhesion kinase (FAK) and vinculin(Nayal et al., 2004). Talin 1 also binds to F-actin, establishing a link between β1 integrins and the cytoskeleton (Nayal et al.,2004). The assembly of focal adhesions is regulated by mechanical force, which controls the recruitment of vinculin into focal adhesions(Balaban et al., 2001; Choquet et al., 1997; Galbraith et al., 2002; Riveline et al., 2001). Talin 1 is required for the force-dependent recruitment of vinculin and strengthens the interactions between integrins and the cytoskeleton(Giannone et al., 2003). Binding of talin 1 to the integrin cytoplasmic domain also enhances the strength of integrin adhesion to ligands (inside-out activation)(Nieswandt et al., 2007; Petrich et al., 2007; Tadokoro et al., 2003). However, it is less clear whether inside-out activation is essential for interactions with insoluble ligands; integrins can be activated directly by binding to insoluble ligands (outside-in activation)(Du et al., 1991), and talin ablation in Drosophila causes detachment of myofibers from integrins without loss of adhesive contact with the ECM(Brown et al., 2002).

Previously, we have shown that inactivation of the talin 1 gene(Tln1) in skeletal muscle leads to a progressive myopathy, caused by mechanical failure of MTJs (Conti et al.,2008). The phenotype resembles the defect observed in mice with a mutation in the gene encoding the integrin α7 subunit (Itga7)(Mayer et al., 1997; Miosge et al., 1999),suggesting that talin 1 mediates integrin α7β1 functions at MTJs. The Tln1-deficient mice did not show the defects in myoblast fusion and sarcomere assembly that have been observed in integrin β1-deficient mice (Schwander et al., 2003). Because vertebrates contain two genes encoding two talins (talin 1 and 2)(McCann and Craig, 1997; McCann and Craig, 1999; Monkley et al., 2001), and because talin 1 and 2 have redundant functions in integrin-mediated attachment of fibroblasts (Zhang et al.,2008), we argued that talin 2 might compensate for loss of talin 1 in skeletal muscle. Talin 2 is expressed at higher levels in skeletal muscle than talin 1 (Conti et al.,2008; Monkley et al.,2001; Senetar and McCann,2005), and talin 2 expression is upregulated during myotube formation (Senetar et al.,2007). Therefore, to determine the function of talin 2 in skeletal muscle, we generated Tln2-deficient mice (referred to as Tln2-KO), and mice lacking both talin 1 and 2 in skeletal muscle(referred to as Tln1/2-dKO). We show here that ablation not only of the talin 1 gene but also of the talin 2 gene leads to defects in the maintenance of MTJs, and we provide evidence that talin 1 and 2 mediateβ1 integrin functions in myoblast fusion and sarcomere assembly.

Tln2-KO mice were generated by ablating the first coding exon of Tln2 (see Results). Mice were genotyped by PCR using the following primers: (a) 5′-CAAACTGAATGAAGGCCCAACAG-3′; (b)5′-TCTCCACTTACTCCTTGCCC-3′; (c)5′-GCCGAGGCTACATGGAGTCAGTAT-3′. Tln1/2-dKO mice and control mice were obtained by crossing Tln1^flox/flox;Tln2^+/- mice with Tln1^flox/+;Tln2^+/-;HSA-CRE^+/- mice. Tln1^flox/flox and HSA-CRE mice have been described previously (Conti et al.,2008; Leu et al.,2003).

Muscle sections were stained with hematoxylin and eosin (H&E). To determine the number of fibers with central nuclei, random areas across the muscle were photographed and nuclei quantified. The number of fields analyzed depended on the size of the muscle type (soleus, n=3; gastrocnemius n=9; tibialis anterioris, n=6). Three mice per genotype and time-point were analyzed. The mean ±s.d. was determined and a Student's t-test performed. Immunohistochemistry, analysis of cell proliferation and western blotting were carried out as described(Conti et al., 2008; Schwander et al., 2003) using the following antibodies: (1) monoclonal against vinculin (Sigma), MHCf(Sigma), ILK (Li et al.,1999), tenascin C (Sigma), MHCs (Leica), CD9 (BD Pharmingen),sarcomeric α-actinin (Sigma), BrdU (Pharmingen); (2) rabbit polyclonal against α7 integrin (kindly provided by U. Meyer, University of East Anglia, Norwich, UK), αv integrin (Chemicon), β1 integrin(Schwander et al., 2003),collagen type IV (Chemicon), laminin α2 (Chemicon) and CX43 (Abcam). Polyclonal antibodies specific for talin 1 and 2 have been described previously (Conti et al.,2008), and correspond to residues 1830-1850 (talin 1) and 940-957(talin 2). Electron microscopy and Evans blue dye (EBD) uptake assays were performed as described previously (Conti et al., 2008). Measurements of creatine kinase (CK) levels were performed by Antech Diagnostics (Irvine, CA, USA).

Primary cultures of fetal myoblasts were prepared from hindlimb muscle from E17.5 embryos as described previously(Schwander et al., 2003). Cells were plated onto coverslips coated with 0.1% gelatin (Sigma) and grown in medium consisting of 65% DMEM (Gibco), 25% Media 199 (Gibco) and 10% fetal bovine serum (Gibco). Differentiation was induced by transferring the cultures to medium consisting of 70% DMEM, 28% Media 199, 2% horse serum (Gibco) and 0.1 mg/ml insulin (Sigma). After 3 days, cells were fixed, permeabilized with 0.5% TX-100 and stained with antibodies to α-actinin and MHCf and then with a secondary anti-mouse Alexa 488 antibody. Myonuclei were stained with DAPI (Sigma). The fusion index was determined as the ratio of myonuclei in cells with three or more nuclei to the total number of nuclei. To measure adhesion, 6×10⁴ cells were plated on poly-D-lysine, collagen type IV, laminin or fibronectin. After 90 minutes, cells were washed, fixed and myoblasts immunostained with antibodies to α7 integrin orα-actinin. Nuclei were stained with DAPI. The number of cells adhering to each substrate was normalized to the number of cells adhering to poly-D-lysine. The mean ±s.d. was determined and a Student's t-test performed. Cells were photographed using an Olympus AX70 microscope and counted using Metamorph.

Myoblasts were analyzed by fluorescence-activated cell sorting (FACS) based on α7 integrin subunit expression as described previously(Blanco-Bose et al., 2001). Cells were harvested and resuspended in PBS containing 3% BSA. To detectα7 expression, cells (10⁶/sample) were incubated with phycoeritrin (PE)-conjugated antibodies to α7 (MBL). To detect expression of active β1 integrins, cells were then incubated with antibody 9EG7 (BD Biosciences), followed by a FITC-conjugated secondary antibody. Cells were analyzed in a LSR II 2 flow cytometer (BD Biosciences).

To analyze talin 2 function during skeletal muscle development we inactivated Tln2 in mice. A gene-targeting vector was generated that included, 5′ of the first Tln2 coding exon, a neomycin(PGK-Neo) cassette flanked by LoxP sites(Fig. 1A). A third LoxP site was inserted between exons 2 and 3. Chimeric mice were generated that transmitted the targeted Tln2 allele through the germline. Crossing of these mice with a CRE deleter mouse(Schwenk et al., 1995)generated three recombination events leading to mouse lines with different Tln2 alleles: (1) mice lacking the neomycin cassette (rec);(2) mice lacking exon 2 (data not shown); and (3) mice lacking exon 2 and the neomycin cassette (Tln2^-; Fig. 1A). Mice homozygous for the Tln2^- allele were identified by genotyping using the PCR primers indicated in Fig. 1A, which amplify a 569-base-pair band specific for Tln2-KO mice (Fig. 1B). Tln2-KO mice were viable and fertile and did not differ in general appearance from wild-type mice(Fig. 2A,B). To confirm that expression of talin 2 was ablated, we analyzed protein expression by western blot using antibodies specific for talin 2(Conti et al., 2008). Two bands corresponding to intact and cleaved talin 2 were detected in muscle extracts from 1-month-old wild-type but not Tln2-KO mice(Fig. 1C). We also immunostained sections of gastrocnemius muscle in 1-month-old mice. As reported previously (Conti et al.,2008), talin 2 was concentrated at MTJs in wild-type mice, but not in Tln2-KO muscle fibers (Fig. 1D). We conclude that talin 2 expression was effectively ablated in muscle from Tln2-KO mice.

We next analyzed histological sections from Tln2-KO mice for skeletal muscle defects. In several mouse models of muscular dystrophy, muscle fibers undergo cycles of degeneration and regeneration. While the nuclei of healthy muscle fibers are located close to the sarcolemma, regenerating fibers display centrally located nuclei, providing a useful readout for muscle defects (Pierson et al.,2005). At 1 month of age, skeletal muscles from Tln2-KOmice and controls were largely indistinguishable, but a slight increase in the number of centrally nucleated fibers was noticeable in the mutants(Fig. 2C,D,I). By 7 months of age, the number of centrally located nuclei was drastically increased in Tln2-KO mice. Similar observations were made in gastrocnemius, soleus and tibialis muscles (Fig. 2E-I; data not shown). As observed in other mouse models for muscular dystrophy (Straub et al.,1997), the severity of the phenotype showed differences between distinct muscles. In Tln2-KO mice, a significantly higher percentage of centrally located nuclei was observed in the soleus when compared with the gastrocnemius and tibialis muscles (gastrocnemius: wild-type=1.27±0.93%, Tln2-KO=9.55±0.31%; soleus:wild-type=0.81± 0.16%, Tln2-KO=35.03±2.26%; tibialis:wild-type=0.11±0.09%; Tln2-KO 9.27±1.34%; in all instances n=3, values are mean ±s.d.).

The higher proportion of centrally nucleated fibers in the soleus muscle suggests that deficiency of talin 2 might predominantly affect slow (type I)fibers. To determine whether this was the case, muscle fibers were co-immunostained with DAPI and with antibodies to slow and fast myosin heavy chain isoforms (Fig. 2L-N, data not shown). Centrally located nuclei were found in both type I and type II fibers. While central nuclei were found more frequently in slow fibers, once the relative proportion of fast versus slow fibers in the soleus muscle was taken into account, no significant difference was observed(Fig. 2R). Co-immunostaining for MHCf and talin 2 showed that talin 2 was expressed at the MTJs of both fast and slow fibers (Fig. 2O-Q). Finally, analysis by western blot showed that, although talin 2 is expressed at higher levels in muscle than talin 1(Conti et al., 2008; Senetar and McCann, 2005),levels of talin 1 and talin 2 did not differ between different muscles(Fig. 2S,T). We therefore conclude that deficiency of talin 2 equally affects slow and fast fiber types. The higher number of centrally nucleated fibers in soleus muscle could be explained by it being a postural muscle(Vandervoort and McComas,1983), experiencing more stress than the gastrocnemius and tibialis muscles. Interestingly, muscles in Tln1-KO mice, including the soleus, do not show centrally nucleated myofibers(Conti et al., 2008), which could reflect the fact that talin 2 levels in muscle are higher than talin 1 levels.

Centrally nucleated myofibers are frequently found in dystrophic muscle fibers that also show defects in the stability of the sarcolemma, leading to efflux of proteins from muscle fibers. For example, this phenotype is observed in patients affected by Duchenne muscular dystrophy and in the mdxmouse (Dmd - Mouse Genome Informatics), which bear mutations in the dystrophin gene (Carpenter and Karpati,1979; Schmalbruch,1975; Straub et al.,1997; Weller et al.,1990). By contrast, mutations in the Itga7 and Tln1 genes cause muscle defects with little or no evidence of membrane damage, and the former but not the latter have centrally nucleated myofibers (Conti et al., 2008; Hayashi et al., 1998; Rooney et al., 2006). To further evaluate whether talin 2-, talin 1- and integrin α7-deficient muscle fibers shared other phenotypic features, we analyzed sarcolemmal damage by measuring creatine kinase (CK) levels in the plasma of 5-month-old mice. Tln2-KO mice presented no evidence for membrane damage(Fig. 2J). We also injected EBD in the tail vein. Occasionally, EBD-positive fibers (which by histology appeared damaged; data not shown) were detected irrespective of the genotype,validating the experimental set up (`control' in Fig. 2K); however, no obvious EBD accumulation was noted in Tln2-KO mice(Fig. 2K). Finally, while immune-cell infiltration and fibrosis are observed in mice with mutations that affect the dystrophin complex (Stedman et al., 1991), these histopathological abnormalities were not present in muscle from Tln2-KO mice (data not shown).

We conclude that Tln2-KO and α7 integrin mutant mice show signs of skeletal muscle fiber degeneration that differ from the phenotype associated with mutations affecting the dystrophin complex, providing evidence that these protein complexes regulate muscle fiber maintenance in different ways.

Talin is essential for focal adhesion assembly and turnover(Franco et al., 2004; Priddle et al., 1998). However, sarcomere and costamere assembly were maintained when Tln1was ablated in skeletal muscle (Conti et al., 2008). As talin 2 is expressed in muscle at higher levels than talin 1 (Conti et al.,2008; Senetar and McCann,2005), we determined whether the muscle fiber cytoskeleton was abnormal in Tln2-KO mice. Sarcomere integrity was evaluated by electron microscopy in 3-month-old mice. In wild-type muscle, the structure of the sarcomere was well maintained, with well-defined Z- and M-bands(Fig. 3A). In Tln2-KOmuscle, necrotic material was observed within myofibers(Fig. 3B,C). Although the M-band (asterisk in Fig. 3B)was not always evident in all areas of mutant muscle fibers(Fig. 3C), the Z-line was present and the overall striation pattern was maintained. By contrast, major defects were observed at MTJs. In wild-type muscle, actin filaments reached the sarcolemma at the end of muscle fibers(Fig. 3D,G). In Tln2-KOs, actin filaments detached from the MTJs, and necrotic and membranous material localized in the gap left by retracting myofilaments(Fig. 3E,H,I). The perturbations at MTJs of Tln2-KO mice resemble the defects in mice lacking Tln1 in skeletal muscle but were considerably more severe and prominent at an earlier age (3 instead of 6 months)(Conti et al., 2008). Unlike in Tln1-KO mice, lateral detachment of the cytoskeleton from the sarcolemma was also occasionally noted in Tln2-KO mice(Fig. 3F). These data are consistent with talin 2 being the major talin isoform in skeletal muscle(Conti et al., 2008; Senetar and McCann, 2005).

To determine whether assembly of integrin complexes was compromised in Tln2-KO mice, we evaluated the distribution of integrins and their effectors. In control muscle, α7 integrin and vinculin were localized at costameres and MTJs, whereas ILK was only found at MTJs. The distribution of these proteins was normal in Tln2-KO mice, although it appeared more diffuse, possibly reflecting MTJ disorganization(Fig. 3J-O). Importantly, talin 1 expression was undetectable or very low in wild-type mice(Fig. 3P) but was readily detectable at the MTJs of Tln2-KO mice(Fig. 3Q, arrow). Talin 1 accumulation at MTJs was likely to be caused by redistribution of talin 1 protein as western blot analysis showed that total talin 1 levels in muscle remained unchanged (see Fig. S1A in the supplementary material). We conclude that talin 2 is not essential for the assembly of integrin complexes in skeletal muscle fibers, and that talin 1 is likely to compensate for a lack of talin 2. However, defects at MTJs are observed in muscle lacking either talin 1 (Conti et al., 2008) or talin 2 (this study), suggesting that the two talin isoforms cannot completely compensate for each other or that a reduction in the total amount of talin (1 and 2) protein caused the defects at MTJs (see Discussion).

We have previously shown that mice lacking β1 integrins in skeletal muscle (refereed to as Itgb1-KOs) have a considerably more severe phenotype than mice lacking either talin 1 or 2(Conti et al., 2008; Schwander et al., 2003). This,together with the observed increased localization of talin 1 at the MTJ in Tln2-KO mice (Fig. 3P,Q), prompted us to test whether the functions of talin 1 and 2 might overlap. We took advantage of our previous observation that a Tln1^flox allele can be effectively inactivated in developing skeletal muscle by an HSA-CRE transgene(Conti et al., 2008). Using these mice and the Tln2-KO mice described here, we generated Tln1/2-dKO mice, which lack both talin 1 and talin 2 in skeletal muscle (see Materials and methods). Similar to Itgb1-KO mice(Schwander et al., 2003), Tln1/2-dKO mice had a contracted posture(Fig. 4A) and died shortly after birth. Immunohistochemical analysis confirmed that expression of talin 2 was effectively ablated in muscle from E17.5 Tln1/2-dKO embryos(Fig. 4B,C). While expression of talin 1 in skeletal muscle was too low to be detected at this stage even in wild-type mice, the severity of the phenotype of the Tln1/2-dKO mice compared with single mutants suggests that Tln1 was effectively inactivated. In addition, PCR analysis of forelimb muscles confirmed recombination of the Tln1^flox/flox allele (see Fig. S1C in the supplementary material).

Histological analysis of embryos at embryonic day 17.5 (E17.5) revealed severe defects in muscle development throughout the body. Although muscles could still be detected, they had an abnormal morphology similar to the phenotype of Itgb1-KO mice(Schwander et al., 2003). In wild-type embryos, muscle fibers were well developed and possessed a uniform size (Fig. 4D,F). In Tln1/2-dKO mice, a general disorganization of the muscles was evident, with striking variations in fiber size(Fig. 4G,H,I,K).

We next determined whether skeletal muscle defects were accompanied by changes in cell proliferation or differentiation. To evaluate proliferation,we carried out BrdU labeling experiments. Proliferating cells were located in several muscle groups, including intercostals and semispinalis muscles, and were identified by co-immunostaining for BrdU and desmin (see Fig. S2A,B,E in the supplementary material; data not shown). No difference in the number of proliferating cells per unit area (mm²) muscle tissue was observed between control and Tln1/2-dKO embryos. Next, we immunostained E16.5-E17.5 embryos with antibodies to desmin (see Fig. S2A-D in the supplementary material, red), α-actinin (see Fig. S2G,H in the supplementary material) and MHCf (see Fig. S2I,J in the supplementary material). These markers were normally expressed in Tln1/2-dKOembryos, indicating that the differentiation of myoblasts was not affected in the mutants. We therefore hypothesized that, similar to mice lacking β1 integrins (Schwander et al.,2003), defects in skeletal muscle development in Tln1/2-dKO mice might be a consequence of perturbations in myoblast fusion and sarcomere assembly.

Although no alterations in sarcomere organization were noted in Tln1-KO or Tln2-KO mice(Fig. 3)(Conti et al., 2008),simultaneous inactivation of Tln1 and Tln2 caused severe defects. In control embryos, vinculin was evenly localized at the sarcolemma(Fig. 5A). By contrast,vinculin distribution was strongly reduced and patchy in the disorganized and short muscle fibers of Tln1/2-dKO mice(Fig. 5B). ECM proteins such as collagen type IV (data not shown) and laminin were deposited around Tln1/2-deficient muscle fibers but showed signs of disorganization and discontinuity (Fig. 5C,D). Tenascin C expression is associated with processes of degeneration and regeneration in dystrophic muscle (Settles et al., 1996; Taverna et al.,1998). In control muscle fibers, tenascin C was confined at the periosteum and the tendon (Fig. 5E). In Tln1/2-dKO muscle fibers, tenascin C was expressed in areas distal from the MTJs(Fig. 5F). Severe defects in sarcomere organization were noted at the ultrastructural level. In control embryos, myofilaments and the striation of sarcomeres were well defined(Fig. 5G). In Tln1/2-dKO muscle, the organization of myofilaments was disrupted,and the assembly of Z-bands appeared rudimentary(Fig. 5H,I). Amorphous filamentous material was abundant between myofilaments. We conclude that the skeletal muscle fiber cytoskeleton is severely disrupted in Tln1/2-dKO muscle.

Immunohistochemical evaluation of the expression and localization of components of integrin adhesion complexes in skeletal muscle of E17.5 embryos revealed severe defects in Tln1/2-dKO mice. In wild-type embryos,α7-, αv- and β1-integrins were clustered at MTJs(Fig. 6A,C; data not shown),whereas none of these proteins was localized at MTJs in Tln1/2-dKOmice (Fig. 6B,D; data not shown). Likewise, the localization of vinculin and ILK was compromised(Fig. 6E-H, see also Fig. 5A,B), indicating that talin 1 and 2 are essential for the assembly and clustering of integrin adhesion complexes at MTJs. Consistent with these data, MTJs were rarely detected in Tln1/2-dKO mice by ultrastructural analysis. When present, they appeared abnormal: myofilaments and the electron-dense plaque at the muscle terminus were absent (arrows in Fig. 6I,J). Notably, these defects at MTJs differ from those in muscle from Itgb1-KO mice, where MTJs develop normally (Schwander et al.,2003). A possible explanation for this difference is the presence of the integrin αv subunit at MTJs(Hirsch et al., 1994). Unlike the integrin α7 subunit, αv heterodimerizes with several βintegrin subunits in addition to β1(Hynes, 1992). The expression and localization of αv were not affected in muscle from Itgb1-KO mice, but αv was no longer present at MTJs from Tln1/2-dKO mice (Fig. 6C,D), suggesting that the αv subunit with a heterodimeric partner other than β1 is sufficient to direct the assembly of MTJs. Taken together, our data indicate that, in Tln1/2-dKO mice, the formation of integrin adhesion complexes in skeletal muscle was affected, leading to defects in the formation of the MTJs and the assembly of the muscle fiber cytoskeleton.

Myoblast fusion depends on the alignment of the membranes of myoblasts, the formation of prefusion complexes characterized by electron-dense plaques, and the subsequent breakdown of the plasma membrane. In wild-type mice, it is difficult to capture myoblasts containing the electron-dense plaques because prefusion complexes are rapidly resolved(Schwander et al., 2003). By contrast, we could readily observe in developing skeletal muscle from Tln1/2-dKO embryos unfused myoblasts containing the electron-dense plaques (Fig. 7A-C), suggesting that fusion was perturbed.

To evaluate directly whether fusion was affected, we established primary cultures of fetal myoblasts from the hindlimbs of E17.5 embryos and analyzed myotube formation in vitro. After 3 days in culture, myoblasts isolated from wild-type embryos formed numerous long myotubes(Fig. 7D,F). By contrast, Tln1/2-dKO myoblasts attached to the underlying fibroblast layer but failed to fuse (Fig. 7E,F). Myotubes formed occasionally, but they were short and assembled a rudimentary cytoskeleton: α-actinin and MHCf were recruited in a striated pattern in wild-type myotubes, but their localization was altered in double mutants(Fig. 7G-J).

The defects in sarcomere assembly and myoblast fusion in Tln1/2-dKO mice were similar to those observed in Itgb1-KOmice (Schwander et al., 2003). In other cells types, such as fibroblasts and platelets, talin mediates the assembly of integrin complexes and integrin activation(Nieswandt et al., 2007; Petrich et al., 2007; Tadokoro et al., 2003). Defects in either of these processes (or both) could affect myoblast fusion. We therefore evaluated β1 integrin expression and activation in primary cultures of fetal myoblasts from Tln1/2-dKO mice by FACS analysis using the 9EG7 antibody, which specifically recognizes an exposed epitope whenβ1 integrins are in an active conformation(Bazzoni et al., 1995). To distinguish myoblasts from contaminating fibroblasts, we used an antibody toα7-integrin, which is specifically expressed by myoblasts(Blanco-Bose et al., 2001)(Fig. 7K). α7-integrin expression levels were normal in myoblasts from double-mutant mice(Fig. 7L,N), excluding that defects in muscle fiber development were caused by a reduction of the amount of integrin expressed at the cell surface of myoblasts. To our surprise,labeling by the 9EG7 antibody of Tln1/2-dKO myoblasts was also comparable to that of controls (Fig. 7L,M; myoblasts are represented by the gated population),indicting that integrin activation was not affected. To confirm these findings further, we seeded myoblasts on collagen type IV, laminin and fibronectin. No significant difference in adhesion was noted on any of the tested substrates(Fig. 7O). We conclude that defects in development of Tln1/2-dKO muscle are not likely to be caused by an inability of β1 integrins to interact with ligands.

To investigate the mechanisms that might cause the fusion defect in Tln1/2-dKO myoblasts further, we analyzed the expression of connexin 43 (CX43; also known as Gja1) and the tetraspanin CD9, which have been shown to regulate cell fusion. CX43 is upregulated in myoblasts preceding fusion(Araya et al., 2005; Gorbe et al., 2007) and was expressed in control and mutant myoblasts(Fig. 7P,Q). Although the localization of CD9 was affected in β1 integrin mutant myoblasts(Schwander et al., 2003), it was normally expressed in myoblasts isolated from Tln1/2-dKO mice(Fig. 7R,S). It is interesting to note that CD9 and α7 integrin were detected, although not exclusively, at the contact surface between talin1/2-deficient myoblasts(Fig. 7P-S, arrows), indicating that their recruitment was not dependent on talin1/2. Talin 1/2 are therefore likely required at a subsequent step in myoblast fusion, potentially by linking integrins to the actin cytoskeleton. Unfortunately, analysis of the organization of the F-actin cytoskeleton of myoblasts using phalloidin was not informative because of the small size and rounded morphology of these cells(data not shown).

We show here that talin 1 and 2 are essential for skeletal muscle development and function. Tln2-KO mice are viable and fertile but develop a myopathy with centrally located nuclei that is associated with defects in the maintenance of MTJs. When talin 1 and 2 are inactivated simultaneously, severe defects in myoblast fusion and sarcomere assembly are observed that are not present in the single mutants. The defects in skeletal muscle development in Tln1/2-dKO mice closely resemble the phenotype of muscle lacking β1 integrins. As talin1/2-deficient myoblasts expressed functionally active β1 integrins, defects in muscle development are likely not primarily caused by lack of an ability of β1 integrins to bind to ECM ligands but by the disruption of their interaction with the cytoskeleton. Consistent with this finding, recruitment of integrin effectors is perturbed in the remaining small muscle fibers of talin1/2-deficient mice.

Previous studies in invertebrates have shown that talin is required for the attachment of skeletal muscle fibers (Brown et al., 2002; Cram et al.,2003). The findings presented here and in our previous report(Conti et al., 2008) extend these findings and show that talin has an evolutionarily conserved function in skeletal muscle attachment. Consistent with the higher expression levels of talin 2 in skeletal muscle compared with those of talin 1(Conti et al., 2008; Senetar and McCann, 2005), Tln2-KO mice developed a more severe myopathy than talin 1 mutants,which is characterized by centrally nucleated myofibers and prominent MTJ defects. Nevertheless, MTJ defects were also present in muscle lacking talin 1(Conti et al., 2008). This result could be explained by two mechanisms. First, talin protein levels might be important. Talin 1 was redistributed to MTJs in Tln2-KO mice, but overall talin 1 levels were not changed. Therefore, a reduction of total talin levels due to loss of talin 2 might have caused MTJ instability. Alternatively, talin 1 and 2 at MTJs might not be entirely interchangeable. Although talin 1 shares 74% identity with talin 2, differences in the remaining amino acids could possibly affect protein function. This model is consistent with previous findings. Although myoblasts express integrinβ1A, muscle fibers express the β1D isoform, which binds with higher affinity to F-actin (Belkin et al.,1997; Belkin et al.,1996; van der Flier et al.,1997). The I/LWEQ module of talin 2 binds with higher affinity to muscle α-actin than the corresponding module in talin 1, which in turn binds with higher affinity to non-muscle α-actin(Senetar et al., 2004). These data delineate a model whereby the expression of β1D-integrin and talin 2 might be important to confer a strong mechanical link between integrins and the cytoskeleton. As ILK-deficient mice also show defects at MTJs(Wang et al., 2008), it appears that several integrin effectors cooperate in this process. Although speculative, talin 1 might be more important for dynamic connections in other cells such as fibroblasts and platelets.

Our studies also provide evidence that talin 1 and 2 cooperate to regulate muscle fiber development. Tln1/2dKO mice show defects in myoblast fusion and sarcomere assembly similar to β1 integrin-deficient mice(Schwander et al., 2003),suggesting that talin1/2 are required for integrin functions in muscle. Previous studies have shown that talin1/2 regulate both integrin activation and their linkage to the cytoskeleton. Although defects in either of these processes could lead to the skeletal muscle defects in Tln1/2dKOmice, our data suggest that defects in cytoskeletal linkage are the cause for the phenotype in the mutant mice. Consistent with this model, β1 integrin expression and activation were not affected in primary cultures of fetal myoblasts from Tln1/2-dKO mice. Similar to our findings, it has previously been shown that talin-mediated integrin activation is also not essential for the initial adhesion of fibroblasts to ECM substrates(Zhang et al., 2008). Although integrin activation by talin is essential in other cell types such as platelets (Petrich et al.,2007; Nieswandt et al.,2007), it might be less important in myoblasts. Kindlin-3 synergizes with talin in regulating integrin activation in platelets(Moser et al., 2009; Svensson et al., 2009), and other proteins, such as ILK, regulate this process(Honda et al., 2009). Therefore, the ligand binding activity of integrins might be regulated by different mechanisms in a cell-type-specific manner.

Instead, our findings suggest that defects in Tln1/2-deficient skeletal muscle fibers are caused by defects in the interaction of integrins with the cytoskeleton. In support of this model, we observed in the mutants detachment of the muscle fiber cytoskeleton from MTJs and defects in the linkage of integrin adhesion complexes to costameres. The fact that integrins in talin 1/2-deficient myoblasts effectively bound to ligands suggests that steps subsequent to adhesion also led to defects in myoblast fusion. Fusion defects in mice lacking β1 integrins are accompanied by reduced recruitment of the tetraspanin CD9 to the site of fusion(Schwander et al., 2003). However, CD9 was still recruited normally in talin1/2-deficient myoblasts,suggesting that talin1/2 act at a subsequent step. Interestingly, several proteins that regulate integrin function and actin dynamics are implicated in myoblast fusion. For example, the guanine-nucleotide exchange factors (GRFs)Dock180 and Brag2/GEP₁₀₀ are required for the fusion of myoblasts and macrophages (Laurin et al.,2008; Pajcini et al.,2008); genetic ablation of Dock180 in mice leads to impaired myoblast fusion (Laurin et al.,2008); ablation of filamin C leads to defects in myogenesis(Dalkilic et al., 2006);furthermore, ablation of FAK affects cell fusion and myofiber regeneration(Quach et al., 2009). This raises the interesting possibility that integrins, and their effectors such as talin1/2, FAK and filamin C, might cooperate to regulate actin reorganization during myoblast fusion.

Centrally nucleated skeletal muscle fibers, as observed in Tln2-KOmice, are also present in several myopathies(Pierson et al., 2005). In patients and mice with mutations that affect the dystrophin complex, centrally nucleated fibers are an indication of fiber degeneration and regeneration that is caused by plasma membrane breakdown(Straub et al., 1997). By contrast, centrally nucleated fibers in Tln2-KO mice accumulated without noticeable plasma membrane breakdown. This resembles the situation in mice and humans with mutations in the gene encoding the integrin α7 subunit (Hayashi et al., 1998; Mayer et al., 1997). The findings suggest that mechanisms associated with defects in the dystrophin and integrin adhesion complexes differ. Interestingly, mutations in genes encoding proteins that are indirectly involved in actin reorganization or membrane trafficking, such as myotubularin 1 (Mtm1), dynamin 2 (Dnm2)and γ-actin, also lead to centrally nucleated skeletal muscle fibers without plasma membrane breakdown (Bitoun et al., 2005; Buj-Bello et al.,2002; Sonnemann et al.,2006). The Tln2-KO mice presented here provide a useful model for studying the molecular mechanisms that lead to fiber degeneration in the absence of plasma membrane damage. Our data also suggest that it will be important to sequence the TLN1 and TLN2 genes in patients affected with genetically uncharacterized congenital myopathies.