Wnt/β-catenin signaling via Axin2 is required for myogenesis and, together with YAP/Taz and Tead1, active in IIa/IIx muscle fibers

The Wnt gene family encodes 19 secreted glycoproteins, which bind to the Frizzled (Fzd) transmembrane receptors on target cells. Wnt proteins regulate key processes such as development and differentiation (Bhanot et al., 1996; Cadigan and Nusse, 1997; Clevers, 2006; von Maltzahn et al., 2012) and are fundamental during embryonic myogenesis (von Maltzahn et al., 2012). For instance, they regulate the development of embryonic muscle in a spatiotemporal manner. Wnt1, Wnt3a and Wnt4 are expressed in the dorsal regions of the neural tube, whereas Wnt4, Wnt6 and Wnt7a are expressed in the dorsal ectoderm, and Wnt11 is expressed in the epaxial dermomyotome. In adult muscle, Wnt proteins are expressed upon regeneration (Polesskaya et al., 2003; von Maltzahn et al., 2012). Non-canonical Wnt signaling through Wnt7a regulates satellite cell number and the size of skeletal muscle (Le Grand et al., 2009; von Maltzahn et al., 2013). Generally, non-canonical Wnt signaling is required for skeletal muscle development, while canonical Wnt signaling, especially via Wnt3a, has been demonstrated to lead to increased fibrosis (Brack et al., 2007).

In canonical Wnt signaling the Wnt binds to the Fzd and Lrp5/6 receptor pairs, thereby leading to the inactivation of glycogen synthase kinase 3β (GSK3β) through dishevelled (Dsh). In the absence of Wnt stimulation, β-catenin forms a destruction complex with adenomatosis polyposis coli (APC), Axin1 and GSK3β (MacDonald et al., 2009). Phosphorylation of β-catenin by casein kinase I (CK1) and GSK3β causes ubiquitylation and proteasome-mediated degradation of β-catenin. The presence of Wnt ligand results in the activation of Dsh, which leads to phosphorylation-dependent recruitment of Axin1 to the low-density lipoprotein receptor-related protein (LRP) receptor and disassembly of the β-catenin degradation complex. Stabilized β-catenin accumulates in the cytoplasm and translocates to the nucleus. There, it complexes with T cell factor/lymphoid enhancer factor (TCF/LEF) transcription factors and acts as a transcriptional coactivator to induce the context-dependent expression of Wnt/β-catenin target genes (Eastman and Grosschedl, 1999). In contrast to canonical Wnt signaling, non-canonical Wnt signaling does not require the activity of β-catenin. Examples of non-canonical Wnt signaling pathways in skeletal muscle are the planar cell polarity (PCP) and the AKT/mTOR (Bentzinger et al., 2014; Le Grand et al., 2009; von Maltzahn et al., 2013) pathways.

Mechanical load or training can lead to an increase of myofiber size, accompanied by muscle enlargement. This condition is called hypertrophy and multiple studies identified different roles for canonical and non-canonical Wnt signaling in muscle fiber hypertrophy (Armstrong and Esser, 2005; Bernardi et al., 2011; Han et al., 2011; Rochat et al., 2004; von Maltzahn et al., 2011). Furthermore, expression of β-catenin is a prerequisite for the physiological growth of adult skeletal muscle, underscoring the importance of canonical Wnt signaling (Armstrong et al., 2006). A recent study employed TCF reporter mice to gain further insights into the role of canonical Wnt signaling (Kuroda et al., 2013). They found that canonical Wnt signaling is strongly activated during fetal myogenesis and only weakly activated in adult muscles, where it is limited to slow myofibers. Their data suggest that canonical Wnt signaling promotes the formation of, or switch to, slow fiber types and inhibits myogenesis (Kuroda et al., 2013). Furthermore, they demonstrated that Wnt1 and Wnt3a are potent activators of canonical Wnt signaling in myogenic progenitors using a TCF-luciferase reporter assay.

Axin1 and Axin2 are closely related (∼45% amino acid identity) negative regulators of canonical Wnt signaling (Behrens et al., 1998; Zeng et al., 1997). They have similar biochemical and cell biological properties but may differ in their in vivo functions. Whereas Axin1 is homogenously distributed in the mouse embryo (Zeng et al., 1997), Axin2 is a Wnt target gene and more selectively expressed in specific tissues (Behrens et al., 1998). Mutant embryos lacking Axin1 die at embryonic day (E) 9.5, with abnormalities including truncation of the forebrain, neural tube defects and embryonic axis duplications (Gluecksohn-Schoenheimer, 1949; Perry et al., 1995). Mice lacking Axin2 are viable but display craniofacial defects (Lustig et al., 2002; Yu et al., 2005). Interestingly, Axin2 has been shown to be upregulated in C2C12 mouse myoblast cells during differentiation (Bernardi et al., 2011; Figeac and Zammit, 2015). Importantly, no in vivo data on the function of Axin2 in skeletal muscle are available so far.

Another important pathway involved in the control of organ size, tissue regeneration and stem cell self-renewal is the Hippo pathway (Zhao et al., 2011). In mammals, the activation of kinases MST1/2 (also known as Stk4/3; homologs of Drosophila Hippo) and LATS1/2 leads to LATS-dependent phosphorylation of Taz (also known as Wwtr1) and YAP (also known as Yap1), thereby decreasing their stability, nuclear localization and transcriptional activity (Pan, 2010). Recently, YAP was identified as a crucial regulator of muscle fiber size (Watt et al., 2015). Moreover, YAP/Taz are incorporated into the β-catenin destruction complex and, thereby, orchestrate the Wnt response (Azzolin et al., 2014, 2012).

Here, we elucidated the role of canonical Wnt signaling activity in adult muscle fibers using a well-established Axin2-lacZ reporter mouse paradigm (Lustig et al., 2002). In these mice, canonical Wnt signaling is reflected by lacZ expression under control of the endogenous Axin2 promoter. We detected active canonical Wnt signaling (1) in myotubes derived from cultured C2C12 cells or murine primary myoblasts, (2) in muscle fibers of type IIa and, most likely, type IIx, (3) at neuromuscular synapses and (4) during regeneration of skeletal muscle after injury. Interestingly, YAP/Taz/Tead1-mediated signaling accompanied canonical Wnt signaling in adult muscle fibers. β-galactosidase-positive muscle fibers (reflecting canonical Wnt activity) were also positive for β-catenin, YAP/Taz and Tead1. In cultured muscle cells (1) the absence of Axin1 interfered with proliferation, (2) the absence of Axin2 slowed down differentiation into myotubes and treatment with Wnt3a had a similar effect, and (3) after knockdown of β-catenin or Tead1 myogenesis was increased. Moreover, canonical Wnt3a induced TOPflash and Tead1 reporters and, importantly, neither induction occurred in the presence of Dkk1, an inhibitor of canonical Wnt signaling.

Wnt signaling plays a crucial and complex role in myogenesis (von Maltzahn et al., 2012). We were interested in the biological role of Axin1 and Axin2, as negative regulators of canonical Wnt signaling, in muscle cells and analyzed their expression in cultured C2C12 cells, a well-established in vitro muscle differentiation paradigm that recapitulates myogenic differentiation (Blau et al., 1983). In situ hybridizations were performed for Axin1 and Axin2 transcripts in proliferating myoblasts and differentiating myotubes (Fig. 1A). Axin1 mRNA was detected in myoblasts and myotubes, while Axin2 was mainly found in myoblasts with extended processes and, even more prominently, in multi-nucleated myotubes (Fig. 1A). To follow the temporal expression profile and protein expression of Axin1 and Axin2, proliferating C2C12 myoblasts and differentiating myotubes were analyzed by immunoblot. Axin1 appeared to be constitutively expressed, whereas Axin2 protein was not detectable in myoblasts but increased gradually in differentiating myotubes (Fig. 1B,C). Expression of Axin1 and Axin2 in C2C12 cells was validated by quantitative PCR analysis (Fig. 1D). A TOPflash assay showed an increase in β-catenin/TCF-mediated gene transcription upon differentiation (Fig. 1E). Altogether, Axin2 expression and β-catenin-mediated canonical Wnt activity increased in differentiating myotubes.

The importance of canonical Wnt/β-catenin signaling in resting adult myofibers in vivo is not known. Therefore, we investigated the expression profile of Axin2 (using Axin2-lacZ mice) as a reporter for canonical Wnt signaling in different adult skeletal muscles. In Axin2-lacZ mice, the lacZ cassette is inserted in frame with the ATG start codon of Axin2 and thus β-galactosidase (β-gal) activity reflects Axin2 gene expression (Lustig et al., 2002). Diaphragm muscles of Axin2-lacZ reporter mice were dissected and stained with X-Gal (Fig. 2A). As expected, myofibers of diaphragms of wild-type littermates did not show any positive staining even after 24 h of incubation with X-Gal, whereas many myofibers of heterozygous Axin2-lacZ diaphragm muscles already turned blue after 4 h (Fig. 2A). Additionally, some blue spots appeared on the heterozygous Axin2-lacZ diaphragms, preferentially at the endplate zone, and were associated with neuromuscular synapses (Fig. 2A,B). After 1.5 h of staining, reporter expression was significantly higher in homozygous Axin2-lacZ diaphragms than in heterozygous diaphragms after 4 h of staining, indicating that the absence of Axin2 relieved the repression of β-catenin/TCF-mediated gene transcription (Fig. 2A).

Intriguingly, in the heterozygous and homozygous Axin2-lacZ diaphragms, canonical Wnt signaling was turned on only in a subset of muscle fibers (Fig. 2A). To understand whether there are muscle-specific differences regarding canonical Wnt signaling, X-Gal staining was performed on transverse sections of different hindlimb muscles. In heterozygous Axin2-lacZ mice, the most prominent β-gal-positive fibers appeared in the plantaris muscle and the extensor digitorum longus (Fig. 2C-F). In homozygous Axin2-lacZ muscles, positively stained fibers were apparent in most analyzed muscles of the calf and shinbone (Fig. 2D). Almost no staining was detected in the soleus muscle (Fig. 2D). Whereas transverse sections of heterozygous tibialis anterior muscle showed weak blue staining, homozygous Axin2-lacZ muscle fibers exhibited intense blue staining (Fig. 2F). Quantitative PCR analysis was performed with cDNAs of five different wild-type muscles to analyze endogenous Axin2 mRNA levels (Fig. 2G). In agreement with the results obtained by X-Gal staining using Axin2-lacZ reporter mice (Fig. 2A,D,F), the Axin2 transcript level was lowest in the soleus muscle (Fig. 2G).

To understand whether the higher β-gal activity in homozygous Axin2-lacZ muscles reflects a gene dosage effect, or is the consequence of stronger induction of canonical Wnt signaling activity due to absence of the negative regulator Axin2, we quantified the amount of β-gal protein in heterozygous and homozygous Axin2-lacZ muscles. In homozygous muscles, β-gal levels were 5-fold higher in diaphragm and 10-fold higher in the tibialis anterior muscle than in heterozygous muscles of the same type (Fig. 2H,I), clearly pointing to a β-catenin-mediated increase of reporter Axin2-lacZ expression rather than a gene dosage effect.

In order to determine which types of muscle fibers were β-gal positive, a series of co-stainings for β-gal and myosin heavy chain (MyHC) markers were performed using antibodies with specificities for fiber types I, IIa, and IIb. Any type II fibers that were not stained by fiber type IIa-specific or IIb-specific antibodies were considered to be type IIx. We analyzed four muscle types: (1) the plantaris muscle, which contains blue-stained fibers in heterozygotes and homozygotes and is mainly composed of type IIa, IIb and IIx fibers (Agbulut et al., 2003); (2) the extensor digitorum longus, which is mainly composed of type II fibers and also contains intensely stained fibers in heterozygotes and homozygotes; (3) the tibialis anterior, which contains mostly weakly stained fibers in heterozygotes (Fig. 2F); and (4) the gastrocnemius medialis, which contains type II and slow type I fibers.

In plantaris muscles from Axin2-lacZ mice almost all β-gal-positive fibers were also positive for MyHC type IIa but not IIb (Fig. 3A,C). The remaining β-gal-positive fibers are therefore most likely type IIx (Fig. 3C). We found a similar distribution of β-gal-expressing fast fiber types in the extensor digitorum longus and tibialis anterior, as that observed in the plantaris (Fig. 3A,C). We asked whether type I fibers might also be β-gal positive and analyzed the gastrocnemius medialis, which contains a few type I fibers. However, the type I fibers in this muscle are not β-gal positive (Fig. 3B). Finally, for each muscle we asked what proportion of the total population of specific subtype II fibers are β-gal positive. In fact, regardless of whether Axin2-lacZ heterozygous or homozygous, almost all type IIa fibers were β-gal positive in plantaris, extensor digitorum and tibialis anterior (Fig. 3D-F). Moreover, in all three muscles type IIb fibers were rarely β-gal positive (Fig. 3D-F). The population of other type II fibers that are β-gal positive is lower in plantaris and tibialis anterior than in extensor digitorum longus (Fig. 3D-F).

To confirm that β-gal-positive fibers in heterozygous and homozygous Axin2-lacZ muscles indeed show active canonical Wnt signaling, transverse sections of hindlimb muscle were stained for β-gal, β-catenin, laminin and with DAPI (Fig. 4A). Most of the β-gal-positive muscle fibers within the plantaris, extensor digitorum longus and tibialis anterior muscles also showed expression of nuclear β-catenin (Fig. 4B). To exclude any muscle fiber type switch in the absence of Axin2, the distribution of all type II fibers was quantified for wild-type, heterozygous and homozygous Axin2-lacZ plantaris, extensor digitorum longus and tibialis anterior muscles. Importantly, no difference in fiber type distribution between the different genotypes was observed (Fig. 4C).

Next, we measured the cross-sectional areas of β-gal-positive fibers to investigate whether canonical Wnt signaling was related to changes in fiber size. Intriguingly, in heterozygous and homozygous Axin2-lacZ genotypes the β-gal-positive fibers of plantaris, extensor digitorum longus and tibialis anterior muscles were significantly smaller in cross-sectional area than β-gal-negative fibers, and this was true for all type II fiber types (Fig. 4D). Furthermore, no difference in cross-sectional area of type II fibers was detected between β-gal-positive heterozygous or homozygous Axin2-lacZ reporter muscles (Fig. 4D). Additional histological stainings were performed to exclude muscle impairments due to the absence of Axin2 in homozygous Axin2-lacZ muscle fibers. Indeed, no obvious changes in general morphology or oxidative metabolism were detected when employing Hematoxylin and Eosin (H&E), Gomori trichrome or NADH staining (Fig. 4E). Finally, we did not observe any signs of myopathy in the homozygous mice, as the number of fibers with centrally located nuclei or apoptotic cells was similar in the plantaris muscles of heterozygous and homozygous Axin2-lacZ reporter mice (data not shown).

Previously, a role for Taz in Wnt signaling was identified and subsequently confirmed by YAP/Taz incorporation into the β-catenin-containing destruction complex, which orchestrates the Wnt response (Azzolin et al., 2014, 2012). Moreover, the Hippo pathway effector YAP was demonstrated to be a crucial regulator of skeletal muscle fiber size (Watt et al., 2015). To further examine the potential correlation between canonical Wnt and YAP/Taz signaling, transverse sections of plantaris, extensor digitorum longus and tibialis anterior muscles were immunostained for Tead1 and YAP/Taz, and co-stained for laminin, β-gal and with DAPI (Fig. 5A,B). We found that the majority of β-gal-positive small-diameter fibers were also positive for Tead1 and YAP/Taz, regardless of whether they were from heterozygous or homozygous Axin2-lacZ mice (Fig. 5A-D), supporting the interplay between YAP/Taz and canonical Wnt signaling.

To further investigate Wnt/β-catenin signaling in skeletal muscle, we performed X-Gal staining in satellite cells. For this we used floating single-fiber cultures of extensor digitorum longus muscle from heterozygous Axin2-lacZ reporter mice. Interestingly, all β-gal-positive nuclei belonged to muscle fibers (Fig. 6). In fact, no satellite cells (marked by Pax7 expression) were β-gal positive, regardless of whether they were attached to β-gal-positive or β-gal-negative muscle fibers and regardless of whether they were quiescent or proliferating, as observed after 3 days of culture. Furthermore, no change in the number of quiescent satellite cells (Pax7-positive cells at day 0 of culture) was detected in β-gal-positive (mean 5.80±1.16) or β-gal-negative muscle fibers (mean 5.47±0.44), suggesting that Axin2 expression in the muscle fiber does not influence satellite cell numbers.

Next, we analyzed the expression of Axin2 during regeneration of skeletal muscle after injury. For this, we injected cardiotoxin (CTX) into the tibialis anterior muscle of adult wild-type, heterozygous and homozygous Axin2-lacZ mice. We observed β-gal expression by X-Gal staining in centrally located nuclei of newly formed myofibers in heterozygous and homozygous Axin2-lacZ mice (Fig. 7A). We further asked whether the time required for the formation of new fibers after CTX-induced injury differs in the absence of Axin2. The proportion of regenerating fibers expressing developmental MyHC after CTX injury was similar in wild-type, heterozygous and homozygous Axin2-lacZ muscles (Fig. 7B). Mononucleated muscle cells, regardless of whether they expressed Pax7 or MyoD (Myod1), were β-gal negative, further supporting the notion that satellite cells do not exhibit Wnt/β-catenin signaling (Fig. 7C,D).

To examine the influence of canonical Wnt/β-catenin signaling and YAP/Taz/Tead1-mediated gene transcription in muscle cells, the roles of Axin1, Axin2, β-catenin and Tead1 were studied in a series of knockdown experiments. First, axin-specific shRNAs were designed, evaluated and used for transfection of cultured primary myoblasts. The number of proliferating myoblasts was significantly lower in cells treated with shRNA against Axin1 and both axins than in mock-transfected cells (Fig. 8A). The number of proliferating myoblast cells did not differ in cells transfected with shRNA against Axin2 versus a mock control (Fig. 8A).

We then differentiated the cells and first evaluated the total number of myotubes per area after 2, 3 and 4 days. Interestingly, whereas myotube numbers were comparable to those of mock-transfected cells in the absence of Axin1, the number of myotubes was significantly reduced in the absence of Axin2, and regardless of whether shRNA against Axin2 alone or together with shRNA against Axin1 was used (Fig. 8B). Second, we tested whether YAP/Taz/TEAD signaling is stimulated during muscle cell differentiation. We transfected a GTIIC-luciferase reporter plasmid (a synthetic YAP/TAZ-responsive luciferase reporter; Fig. 8C) into cultured primary myoblasts. Interestingly, we detected a stimulation of YAP/Taz/TEAD signaling during muscle cell differentiation (Fig. 8C). Third, we asked whether an increase of YAP/Taz/TEAD signaling in myotubes (Fig. 8C) is accompanied by an increase in the expression of Tead1 target genes. Indeed, transcript levels of Tead1 target genes, such as Ctgf, Ankrd1 and Cyr61, increased significantly during the differentiation of muscle cells (Fig. 8D). Fourth, we analyzed the specificity of canonical Wnt/β-catenin signaling activity in muscle cells during differentiation by knocking down either β-catenin or Tead1 and monitoring TOPflash activity (Fig. 8E). Knockdown of β-catenin, but not Tead1, significantly reduced TOPflash activity (Fig. 8E). Fifth, we asked how the GTIIC-luciferase reporter responds to the knockdown of either β-catenin or Tead1. Knockdown of Tead1 significantly reduced reporter activity, whereas knockdown of β-catenin stimulated the GTIIC-luciferase reporter (Fig. 8F). Sixth, we analyzed the influence of β-catenin or Tead1 knockdown on muscle cell differentiation. In the absence of either β-catenin or Tead1, there was a significant increase in myotube formation (Fig. 8G).

To understand whether canonical Wnts alter myogenic differentiation by regulating canonical Wnt and TEAD signaling, the response to dickkopf 1 (Dkk1) and stimulation by different Wnts was studied by TOPflash and GTIIC-luciferase reporter assays in transiently transfected cultured differentiating primary myoblasts and myotubes. The specificity of intrinsic Wnt-dependent TOPflash stimulation during the differentiation of cultured muscle cells was confirmed by inhibition of TOPflash stimulation in the presence of Dkk1, which specifically inhibits the canonical Wnt signaling pathway by the disruption of the Wnt-Fzd LRP complex (Krupnik et al., 1999) (Fig. 9A). Moreover, stimulation of TOPflash in differentiating myotubes was significantly accelerated by concomitant transfection of a plasmid encoding the canonical Wnt3a, but not if a plasmid encoding non-canonical Wnt7a was used (Fig. 9B). Next, we asked whether Wnts are able to stimulate the GTIIC-luciferase reporter as a means to monitor activation of YAP/Taz/TEAD signaling (Fig. 9C). Indeed, canonical Wnt3a induced GTIIC-luciferase activity during myotube differentiation. Importantly, this effect was inhibited by Dkk1, demonstrating the specificity of the signal (Fig. 9C).

Previously, it was shown that Axin2 is upregulated by Wnt signaling (Jho et al., 2002; Leung et al., 2002; Lustig et al., 2002), indicating that it is engaged in a negative-feedback loop controlling the cellular Wnt response. Indeed, knockdown of Axin2 was shown to increase β-catenin levels after Wnt stimulation (Bernkopf et al., 2015). To test whether the upregulation of Axin2 observed upon myotube formation is mediated by secreted canonical Wnt proteins we treated differentiating C2C12 cells with soluble Dkk1. Cultivation of myotubes, after 3 days of differentiation, with Dkk1 for 24 h resulted in near-complete loss of Axin2 expression (Fig. 9D). This demonstrates that the canonical Wnt/β-catenin pathway is activated in differentiating myocytes by canonical Wnt ligands, and that this activation is responsible for the expression of Axin2.

We next analyzed whether Wnt3a would also alter myogenic differentiation. Wnt3a cDNA was delivered via retroviral infection of C2C12 cells, which were subsequently selected for stable integration of the viral cDNA. All investigations were performed with these polyclonal cell populations. Overexpression of Wnt3a induced an elongated cell morphology in myoblasts, as compared with control cells infected with empty vector (Fig. 9E, top panel). Activation of the canonical Wnt pathway was demonstrated by increased levels of cytosolic β-catenin using western blot analysis (data not shown). Importantly, when Wnt3a-overexpressing cells were cultured in differentiation medium, cells did not form multinucleated myotubes, in contrast to control cells (Fig. 9E, bottom panel). Impaired differentiation of C2C12 cells by incubation with Wnt3a-conditioned medium was further confirmed by MyHC analysis (Fig. 9F,G). Lysates of C2C12 cells contained significant amounts of MyHC after 2 days of differentiation, but this increase did not occur in cells incubated with Wnt3a (Fig. 9F). Lack of MyHC expression and myotube formation was further confirmed by immunofluorescence staining of C2C12 cells (Fig. 9G).

The impairment of myotube formation by knockdown of Axin2 or treatment with canonical Wnts implies that overactivation of Wnt signaling is detrimental to myogenic differentiation. This is in line with several publications showing that activated β-catenin modulates myogenic differentiation (Gavard et al., 2004; Goichberg et al., 2001; Martin et al., 2002; Rudolf et al., 2016).

This is, to our knowledge, the first report describing active canonical Wnt together with YAP/Taz/TEAD signaling in adult skeletal muscle fibers. Using Axin2-lacZ reporter mice, we identified active canonical Wnt signaling in type II myofibers (Fig. 2) (Lustig et al., 2002). This mouse model is well established for tracing active canonical Wnt signaling since Axin2 is a direct target of β-catenin-mediated gene expression (Barolo, 2006). In fact, heterozygous Axin2-lacZ mice, without any known signs of haploinsufficiency, express muscular β-gal in type II muscle fibers, suggesting that active canonical Wnt signaling is present physiologically in adult muscle fibers (Fig. 2). In homozygous Axin2-lacZ mice, not only is the doubled lacZ gene dosage responsible for elevated Axin2-lacZ reporter expression, but also the derepression of canonical Wnt signaling, since Axin2 itself is a negative regulator and target of canonical Wnt signaling and therefore participates in a negative-feedback loop. In fact, we detected a 5- to 10-fold increase in β-gal protein in muscles of homozygotes as compared with heterozygous Axin2-lacZ reporter mice (Fig. 2H,I). However, in Axin2-lacZ reporter mice only muscle fiber types IIa and IIx (shown for IIx indirectly) are positive for β-gal activity, suggesting that only these fibers display active canonical Wnt signaling (Fig. 3). Interestingly, soleus muscle does not express any significant amounts of β-gal in heterozygous or homozygous Axin2-lacZ reporter mice, suggesting that Wnt signaling is inactive in this particular muscle (Fig. 2D). Furthermore, the specificity of muscle cell-specific canonical Wnt signaling was confirmed when the endogenous Wnt-dependent stimulation of TOPflash reporter in cultured primary muscle cells was inhibited by Dkk1, an extracellular inhibitor of canonical Wnt signaling (Fig. 9A).

Previously, the only report concerning canonical Wnt signaling in adult skeletal muscle fibers employed TCF-lacZ transgenic reporter mice, using a promoter with artificial multimerized TCF binding sites to drive transcription of the lacZ reporter gene (Kuroda et al., 2013). It was shown that canonical Wnt signaling is strongly activated during fetal myogenesis and weakly activated in slow myofibers of adult muscles (Kuroda et al., 2013). Intriguingly, in contrast to those data, we find that canonical Wnt signaling is absent from slow type I muscle fibers (Fig. 2C, Fig. 3B). These contradictory data might be explained by the fact that the Axin2-lacZ reporter mouse better represents the physiological condition, since the lacZ gene is driven by the endogenous Axin2 promoter.

Importantly, the absence of Axin2 in adult homozygous Axin2-lacZ muscles does not result in any obvious muscle phenotype. Grip strength appeared to be marginally affected in young homozygous Axin2-lacZ mice (data not shown), but this remains to be investigated further in aged mice, where typically a muscle weakness phenotype is more prominent. Still, neither fiber type switches nor any histological or metabolic impairments were detected in mutant muscles (Fig. 4C,E). Therefore, in adult muscle fibers Axin2 appears to reflect the localization of active canonical Wnt signaling but without itself being of any obvious importance for skeletal muscle under steady-state conditions.

Although general muscle biology is not affected by the absence of Axin2, muscle fibers with active Wnt signaling differ morphologically from neighboring fibers lacking canonical Wnt signaling. All muscle fibers that exhibited β-gal activity, regardless of whether they were from heterozygous or homozygous Axin2-lacZ reporter mice, had a significantly smaller cross-sectional area (Fig. 4D). As such, just by measuring the cross-sectional area it is, in principle, possible to identify muscle fibers active in canonical Wnt signaling. Previously, a change of fiber diameter in regenerated fibers in the absence of β-catenin and APC was reported (Jones et al., 2015; Murphy et al., 2014; Parisi et al., 2015). As a first step towards understanding why canonical Wnt signaling is active in those muscle fibers, we speculated that the Hippo pathway might be involved, since it has been demonstrated that the Hippo pathway effector YAP is also a crucial regulator of skeletal muscle fiber size (Watt et al., 2015). Moreover, it has been shown that YAP/Taz are part of the β-catenin destruction complex and are involved in orchestrating the Wnt response (Azzolin et al., 2014, 2012). Thus far, it had not been investigated whether YAP regulates skeletal muscle fiber size in a fiber-type dependent manner (Watt et al., 2015). Another player might be Tead1, since a potential role for Tead1 signaling in muscle is promotion of a generally slower skeletal muscle contractile phenotype (Tsika et al., 2008). We detected YAP/Taz and Tead1 in the same fiber types in which canonical Wnt signaling was active (Fig. 5A-D). Moreover, the expression of the Tead1 target genes Ctgf, Cyr61 and Ankrd1 increased together with the increase in Axin2 expression in muscle cells (Fig. 1A-D, Fig. 8D). Altogether, these findings indicate that canonical Wnt signaling might act together with YAP/Taz/Tead1-mediated signaling in the regulation of skeletal muscle fiber size. Interestingly, intrinsic gradual GTIIC-luciferase reporter activation, resembling the Axin2 expression profile and TOPflash activity, was detected during the differentiation of cultured primary muscle cells (Fig. 1, Fig. 8C). Knockdown of Tead1 did not influence TOPflash activity, suggesting that there is no feedback loop from Tead1 to canonical Wnt signaling (Fig. 8E). However, and in agreement with previously published data obtained in HEK293 cells showing that β-catenin stimulates Taz degradation (Azzolin et al., 2012), our data showed that knockdown of β-catenin increased GTIIC-luciferase activity. Furthermore, we show that the GTIIC-luciferase reporter responds to exogenous canonical Wnt stimulation (Fig. 9C), but is also sensitive to Dkk1, demonstrating the cross-talk between both pathways in skeletal muscle.

Interestingly, we also found that Axin2 is expressed in regenerating myofibers but not in quiescent or activated satellite cells (Figs 6, 7). Our experiments further show that endogenous Axin2 expression is absent in myoblasts (Fig. 1A,B,D), which is in accordance with recent data demonstrating that only very low levels of active β-catenin protein are found in undifferentiated muscle progenitors (Rudolf et al., 2016). We therefore speculate that canonical Wnt signaling is an important player in the differentiation process (from myoblasts to myotubes) but is not required for satellite cell maintenance. This fits well with previous reports describing Wnt/β-catenin activity as regulating satellite cell myogenic potential (Bernardi et al., 2011; Brack et al., 2008; Jones et al., 2015). However, Wnt-dependent TGFβ2 signaling was reported to be important for the commitment of satellite cells to a non-myogenic lineage, with a shift towards a more fibrotic phenotype that is associated with increased TGFβ signaling in regenerating or diseased muscle (Biressi et al., 2014; Brack et al., 2007). Data obtained in our laboratory do not indicate any change in Tgfb2 transcript level in Axin2-deficient muscles (data not shown).

Since Axin2 autoregulates its own expression by a negative-feedback loop, we suggest that the temporal activity and level of canonical Wnt signaling are controlled via Axin2. This tight regulation would halt activation of canonical Wnt signaling in the presence of Axin2, thereby avoiding transdifferentiation. To gain further insight into cell biological aspects of canonical Wnt signaling, changes in Axin expression were monitored in C2C12 and primary muscle cells (Figs 1 and 8). Previously, it was reported that Axin2 expression increases during muscle cell differentiation in vitro (Bernardi et al., 2011; Figeac and Zammit, 2015). Here, we confirmed these data in primary muscle cell cultures and additionally demonstrated that Axin1 is constitutively expressed in myoblasts and myotubes (Fig. 1A-D). Knockdown experiments showed that Axin1 is required for the proliferation of myoblasts (Fig. 8A), whereas Axin2 is necessary for myotube formation (Fig. 8B). In the absence of Axin2, myotube formation was inhibited but not fully blocked, suggesting delayed differentiation upon strong activation of Wnt/β-catenin signaling. Similar results were obtained by addition of canonical Wnt3a to differentiating muscle cell cultures (Fig. 9E-G), further supporting the notion of delayed differentiation upon strong activation of Wnt/β-catenin signaling. Recent studies employing genetic non-degradable β-catenin mutants in satellite cell progeny reported controversial results concerning the muscle regeneration phenotype in loss- versus gain-of-function mutants, but both underline the necessity for strict regulation of β-catenin signaling activity during muscle regeneration (Murphy et al., 2014; Rudolf et al., 2016).

We present a model describing the interplay between canonical Wnt signaling and YAP/Taz/TEAD signaling in myofibers, and hypothesize that the Wnt glycoproteins might be supplied by the muscle cells themselves, providing them in an autocrine/paracrine fashion, and/or by the motor nerve endings (Fig. 10). During differentiation or regeneration, Axin2-mediated canonical Wnt signaling might be involved in myotube formation, whereas in adult muscle fibers Axin2-mediated canonical Wnt signaling and YAP/Taz/Tead1-mediated signaling might ensure proper fiber size diameter in a subset of skeletal muscle fibers (Fig. 10).

Silencing of Axin1, Axin2, β-catenin and Tead1 was achieved using plasmid-derived shRNAs (see supplementary Materials and Methods). For the generation of in situ riboprobes, Axin1 and Axin2 fragments were PCR amplified using the primers described in Table S1 and subcloned into pGEM-T Easy (Promega) (for details, see the supplementary Materials and Methods). Recombinant Dkk1 was produced essentially as described previously (Krupnik et al., 1999) (see supplementary Materials and Methods). pHAN-puro, pHAN-Wnt3a and pHAN-Wnt7a vectors were kindly gifted by Michael A. Rudnicki (Ottawa Hospital Research Institute, Canada). Total RNA was extracted from mouse tissues with TRIzol reagent (Life Technologies) (Cheusova et al., 2006). After reverse transcription, cDNAs were used with mouse-specific primers (Table S1) for quantitative PCR reactions using the ABsolute QPCR SYBR Green Capillary Mix (Fisher Scientific), glass capillaries and LightCycler (Roche Diagnostics) according to the manufacturers' instructions. Data analysis was performed as described (Cheusova et al., 2006).

Primary skeletal muscle cells were prepared from muscles of adult BL6 wild-type mice using the MACS Satellite Cell Isolation Kit (Miltenyi Biotech) according to the manufacturer's instructions. C2C12 and primary skeletal myoblasts were transiently transfected by nucleofection according to the manufacturer's instructions (Nucleofector, Lonza). Single myofibers were isolated from the extensor digitorum longus muscle by collagenase digestion and cultured as described elsewhere (von Maltzahn et al., 2013). Culture conditions and transfection protocols are provided in the supplementary Materials and Methods.

In situ hybridization for Axin1 and Axin2 was performed as previously described (Schubert et al., 2008). A detailed protocol is provided in the supplementary Materials and Methods.

TOPflash or GTIIC-luciferase reporter assays were performed using primary skeletal muscle cells or C2C12 myoblasts transiently transfected with luciferase reporter and other expression plasmids (hDkk1-Flag, pHAN-puro, pHAN-Wnt3a, pHAN-Wnt7a). After transfection, myoblasts were differentiated to myotubes, harvested at the indicated time points, and cell extracts prepared for luciferase assays. Where indicated, cells were exposed to control- or Dkk1-conditioned medium for 24 h before harvesting. Each sample was transfected in duplicate, and each experiment was repeated at least three times.

Equal amounts of extracted proteins were solubilized in Laemmli buffer, boiled, separated by SDS-PAGE and blotted to nitrocellulose membrane (Protran BA85, Whatman) or Immobilon membranes (Amersham). Membranes were blocked and incubated with primary antibodies at 4°C overnight. Corresponding secondary antibodies conjugated with horseradish peroxidase were used to detect primary antibodies. Protein bands were detected by chemiluminescence reaction and exposed on a LAS-3000 luminescence image analyzer (Fujifilm). Western blot results were quantified by densitometric analysis using ImageJ (http://rsb.info.nih.gov/ij/). For further details, including antibodies used, see the supplementary Materials and Methods.

Mouse experiments were performed in accordance with animal welfare laws and approved by the responsible local committees (animal protection officer, Sachgebiet Tierschutzangelegenheiten, FAU Erlangen-Nürnberg, AZ: I/39/EE006 and TS-07/11) and government bodies (Regierung von Unterfranken). Mice were housed in cages that were maintained in a room with temperature 22±1°C and relative humidity 50-60% on a 12-h light/dark cycle. Water and food were provided ad libitum. Mouse mating and genotyping were performed as previously described (Yu et al., 2005).

For muscle injury experiments mice were anesthetized and 50 µl cardiotoxin (Sigma; 10 µM in 0.9% NaCl) directly injected into the tibialis anterior muscle. Mice were sacrificed 10 days after injury and muscles frozen in liquid nitrogen.

Detailed protocols for histochemical stainings, including X-Gal, H&E, Gomori trichrome and NADH, and for the TUNEL assay for apoptotic cells are provided in the supplementary Materials and Methods.

For immunofluorescence analysis, transverse sections of muscle were permeabilized for 5-10 min in 0.1% Triton X-100 in PBS, blocked in 10% (v/v) FCS, 1% (v/v) BSA in PBS or M.O.M. blocking reagent (Mouse Ig Blocking Reagent, Vector Laboratories) for 1 h at room temperature. For analysis of satellite cells, single isolated muscle fibers were blocked in 5% (v/v) horse serum in PBS. Incubation with primary antibodies was performed overnight at 4°C. Fluorophore-coupled secondary antibodies were used for detection of primary antibodies and nuclei were counterstained with DAPI. Information on antibodies is provided in the supplementary Materials and Methods.

Effects on proliferation and primary myotube formation after knocking down axins were quantified using a Leica DM IRB microscope. For assessment of proliferation, myoblasts were fixed 48 h after transfection. Total myoblast numbers per 3.5 cm plate were counted that were positive for the co-transfected GFP expression. For myogenic differentiation analysis, GFP-positive myotubes containing more than ten myonuclei were counted and related to the GFP-positive total myotube number.

pHAN-puro, pHAN-Wnt3a and pHAN-Wnt7a vectors were kindly gifted by Michael A. Rudnicki.