Sox4-mediated caldesmon expression facilitates differentiation of skeletal myoblasts

Myoblast differentiation is a highly ordered process that requires a particular combination of gene expression and signaling (Clatworthy and Subramanian, 2001; Handwerger and Aronow, 2003; Marshall, 2003). Myogenic cells withdraw from the cell cycle and undergo dramatic morphological changes. Mononucleated myoblasts align and their membranes fuse to form multinucleated myotubes that then mature into muscle fibers. Concomitantly, transcriptional cascades regulated by multiple groups of muscle-specific transcription factors initiate the de novo synthesis of various muscle-specific proteins (Braun and Gautel, 2011). Several lines of evidence indicate that the migration of myoblast cells is required for membrane fusion (Biressi et al., 2007; Braun and Gautel, 2011). This dynamic cellular behavior is accompanied by massive changes in cytoskeletal proteins, including the dramatic rearrangement of actin filaments and related signaling proteins such as integrin and focal adhesion kinase (Burattini et al., 2004; Formigli et al., 2007).

Caldesmon (CaD) is an actin-binding protein found in almost all vertebrate cells. There are two types of CaD protein: the heavy form (h-CaD) is specifically detected in smooth muscle cells (SMCs) (Lehman et al., 1992; Sobue et al., 1981; Wang, 2001), whereas the light form (l-CaD) is ubiquitously expressed (Bretscher and Lynch, 1985; Matsumura and Yamashiro, 1993). The known functions of the CaD proteins suggest that they have crucial roles during muscle development. The expression of h-CaD in differentiated SMCs plays an important role in the maintenance of the actomyosin contractile structure (Bretscher and Lynch, 1985; Fürst et al., 1986). Moreover, the insufficient development of organs composed of SMCs and cardiac muscle in the CaD-knockout mouse suggests that CaD is required for myogenesis (Zheng et al., 2009a; Zheng et al., 2009b). Nevertheless, the function of CaD in skeletal muscle development is not clear.

The CaD isoforms originate from alternative splicing and/or promoter usage (Mayanagi and Sobue, 2011). Two independent CALD1 promoters have been identified in both chickens and humans: gizzard-type and brain-type in chickens and fibro-type and HeLa-type in humans (Hayashi et al., 1992; Yano et al., 1994). The activities of gizzard-type and fibro-type promoters are higher than the activities of brain-type and HeLa-type promoters (Fukumoto et al., 2009; Yano et al., 1995). Therefore, most studies of CALD1 promoter activity and associated transcription factors have focused on the gizzard-type or fibro-type promoter. Nonetheless, it is not clear why l-CaD expression is higher than h-CaD expression in mouse skeletal muscle (Yano et al., 1995).

Sox4, a transcription factor in the SRY (sex-determining region Y) family, contains a highly conserved high-mobility group (HMG) DNA-binding domain (DBD) that plays important roles in many developmental processes, including nervous system development (Cheung et al., 2000) and endocrine islet formation (Wilson et al., 2005). Cardiac development in Sox4^−/− mice showed defective outflow tract formation and valve development, thus indicating a role for Sox4 in myogenesis (Schilham et al., 1996). Moreover, Sox4 transcripts have been detected in skeletal myoblasts and in myocardium (Ling et al., 2009; Tomczak et al., 2004). These reports strongly suggest that Sox4 participates in myoblast differentiation. However, the molecular mechanisms underlying Sox4-mediated myogenesis have not been elucidated.

In this study, we investigated the molecular mechanism of CaD expression and its role in myoblast differentiation by using C2C12 murine skeletal myoblasts as a model system for skeletal muscle differentiation. CaD expression was induced before the onset of membrane fusion and gradually increased as myoblast differentiation proceeded. Silencing of CaD expression by siRNA significantly reduced membrane fusion and myotube formation and inhibited muscle-specific protein expression. Furthermore, the Cald1 gene in C2C12 myoblasts contained a HeLa-type promoter, and its activity was controlled by the direct interaction of the Sox4 transcription factor with cis-acting elements on the promoter. In conclusion, we show that CaD expression, regulated by Sox4, plays an important role in myofibril formation during muscle development in the mouse embryo.

First, we examined the expression of CaD during C2C12 skeletal myoblast differentiation. To induce differentiation, confluent C2C12 myoblasts were switched from growth medium (GM) to differentiation medium (DM). The level of CaD protein gradually increased along with the expression of skeletal muscle markers, including myogenin and myosin heavy chain (MHC), as C2C12 myoblast differentiation progressed (Fig. 1A,B). RT-PCR results indicated that the changes in CaD protein level during myoblast differentiation were controlled by transcriptional regulation (supplementary material Fig. S1). To examine the cellular localization of CaD in myoblasts, we performed immunocytochemistry with cells incubated in DM. Consistent with the immunoblot results, CaD fluorescence was low under proliferation conditions, and CaD became enriched in MHC-positive myotubes as differentiation proceeded (Fig. 1C).

To examine the effect of CaD on skeletal muscle differentiation, we constructed four small interfering RNA (siRNA) constructs specific for Cald1 mRNA and transfected control siRNA or CaD siRNA (#1 and #2) into C2C12 cells before inducing differentiation. Fig. 2A shows a representative western blot performed using cell lysates prepared from cells transfected with siRNA #1 or #2 and induced to differentiate for 2 days. The level of CaD in cells transfected with CaD-siRNA #2 was reduced by 80%. The level of β-actin was unaffected.

CaD is involved in actin-dependent cell morphology and motility in various cells (Castellino et al., 1995; Eppinga et al., 2006; Li et al., 2004; Mayanagi et al., 2008; Mirzapoiazova et al., 2005). Therefore, we examined the effect of depletion of CaD on the spreading of C2C12 cells. Microscopy observation of siRNA-transfected C2C12 myoblasts showed that Cald1 silencing decreased cell spreading by ∼50% (Fig. 2B). We confirmed the correlation between the effectiveness of Cald1 silencing and the myoblast phenotype with two additional CaD siRNA constructs (supplementary material Fig. S2A,B). Wild-type C2C12 cells were elongated, and CaD colocalized with actin-rich lamellar structures at the edges of the cells. By contrast, Cald1 silencing markedly inhibited cell polarity, decreased the cell area and disrupted lamellar formations (Fig. 2B and supplementary material Fig. S3). These results suggest that CaD is necessary for actin-mediated cell spreading and lamellipodia formation in C2C12 myoblast differentiation.

Because cell motility regulated by actin rearrangement is crucial for myoblast alignment and differentiation (Burattini et al., 2004; Formigli et al., 2007), we examined the effect of Cald1 silencing on cell migration in a scratch-wound healing assay. As shown in Fig. 1A, under proliferation conditions in which CaD expression was extremely low (D0), cell migration was not affected by CaD siRNA transfection, indicating that the siRNA used in the experiment had no side effects on cell migration. However, under differentiation conditions (D2) in which CaD expression increased transiently, silencing of Cald1 substantially reduced cell migration into the wound region (Fig. 2C and supplementary material Fig. S2C). These results suggest that the activity of CaD in C2C12 myoblasts is essential for the regulation of cell motility, which is required to initiate myoblast alignment in response to the differentiation signal.

Next, we investigated the role of CaD in C2C12 myoblast differentiation. In control C2C12 cells, myoblast fusion began on day 2 after the application of differentiation medium and gradually increased to form myotubes (Fig. 3A,B). By day 4, the frequency of MHC-positive myotubes increased to ∼50%. Membrane fusion was significantly lower in CaD-silenced C2C12 cells, thus reducing the frequency of MHC-positive myotubes to ∼10%. Measurement of myotube diameter, as an indicator of myotube development (Burattini et al., 2004; Doyle et al., 2011), suggested that CaD plays a major role in myoblast fusion by increasing myotube size during differentiation (Fig. 3C and supplementary material Fig. S2D).

Myoblast differentiation is orchestrated by the expression of muscle-specific proteins that initiate and guide the differentiation process. We, therefore, investigated the effect of Cald1 silencing on the expression of specific myogenic marker proteins. The expression of myogenin and MHC was markedly reduced in CaD-siRNA-transfected C2C12 myoblasts (Fig. 3D), whereas MyoD expression was similar in CaD-siRNA-transfected cells and control cells in the proliferation stage. These results, together with those in Fig. 2, clearly indicate that CaD functions in the early stage of myoblast differentiation to regulate initial cell spreading and motility.

The results described above demonstrated that the expression of CaD in response to the differentiation signal is an important event in C2C12 myoblast differentiation. To understand the molecular mechanism underlying CaD expression during C2C12 differentiation, we characterized the promoter of the gene encoding mouse CaD (Cald1). In previous studies, characterization of the promoters of avian and human CALD1 showed that a CArG box (Yano et al., 1995) and glucocorticoid-response element (GRE)-like sequences (Mayanagi et al., 2008), respectively, were cis-acting elements important for the regulation of promoter activity in each species. However, we found that these cis-acting elements are located −110 kbp upstream of the transcription initiation site of mouse Cald1. Moreover, the nucleotide sequence of the putative promoter region in mouse Cald1 is ∼70% identical to that of the HeLa-type promoter in human CALD1.

Because the HeLa-type promoter has not been extensively characterized, we first searched the mouse Cald1 promoter for putative cis-acting elements and their binding partners using TFSEARCH software (Searching Transcription Factor Binding Sites, Version 1.3; http://www.cbrc.jp/research/db/TFSEARCH.html). Sequence analysis identified several putative cis-acting elements in the mouse Cald1 promoter between −759 bp and +41 bp for the transcription factors GATA, Sox4, AP1, Nkx2.5, SP1 and CP2. The sequences are almost conserved in the human HeLa-type promoter (Fig. 4A). To determine which cis-acting elements are responsible for Cald1 promoter activity, we constructed reporter plasmids containing 800 bp of the Cald1 promoter (WT-Luc) or truncated fragments (mut-1-Luc–mut-4-Luc) (Fig. 4B). Each plasmid was introduced into C2C12 myoblasts and promoter activity was measured after culturing cells for 2 days in proliferation medium (D0) or differentiation medium (D2). With the WT-Luc plasmid, basal levels of promoter activity were observed in proliferating myoblasts (D0); promoter activity increased by ∼fourfold in differentiating C2C12 myoblasts (D2), indicating that Cald1 promoter activity is associated with C2C12 myoblast differentiation. However, the promoter activity in differentiating myoblasts was reduced considerably when cells were transfected with mut-3-Luc or mut-4-Luc, lacking 459 bp and 656 bp, respectively, at the 5′ end of the promoter, indicating that the region between −532 bp and −300 bp is crucial for mouse Cald1 promoter activity (Fig. 4C).

Because promoter analysis indicated that the nucleotide sequence between −421 bp and −415 bp contained a Sox4-binding site (Fig. 4A), we examined whether Sox4 directly binds this region of the Cald1 promoter in a electrophoretic mobility shift assay (EMSA) using a recombinant Sox4 DNA-binding domain (DBD) fused to glutathione S-transferase (GST). As shown in Fig. 5A, incubation of a radiolabeled Cald1 probe with the Sox4 DBD slowed the migration of the probe in a concentration-dependent manner (lanes 4, 5), indicating the formation of DNA–protein complexes. Incubation with GST alone did not produce DNA–protein complexes (lanes 2, 3). The DNA-protein complex was further shifted in the presence of anti-GST antibody (lane 6), whereas IgG had no effect on the mobility shift of the DNA–protein complex (lane 7). Moreover, the DNA-protein complex disappeared when unlabeled Cald1 probe was added (lane 8). The results indicate that Sox4 can bind directly to the Cald1 promoter region between −421 bp and −415 bp in vitro.

To examine whether Sox4 binds to the Cald1 promoter in vivo, we performed a ChIP analysis with C2C12 myoblast cells transfected with control or Sox4 siRNA and cultured in proliferation medium (D0) or differentiation medium (D2). As shown in Fig. 5B, neither the proximal nor distal region of the Cald1 promoter was specifically precipitated by the Sox4 antibody. The precipitation of the distal region of the Cald1 promoter was probably nonspecific because the region does not contain a Sox4-binding sequence. By contrast, anti-Sox4 antibody precipitated DNA corresponding to the Sox4-binding sequence in the Cald1 promoter. The amount of DNA precipitated was reduced considerably when the cells were transfected with Sox4 siRNA (Fig. 5B, middle panel). As a negative control, we performed ChIP analysis using IgG instead of Sox4 antibody; no significant signal was detected (supplementary material Fig. S4). In addition, acetylated histones H3 and H4, which are markers for active transcription, were associated with the Cald1 gene promoter in differentiating, but not proliferating, C2C12 cells, and their recruitment to the Cald1 promoter was reduced in Sox4-silenced C2C12 cells (supplementary material Fig. S4). Taken together, these results indicate that Sox4 can bind directly to the Cald1 promoter in the early stage of C2C12 myoblast differentiation.

To confirm that Sox4 induces CaD expression, HEK293 cells were co-transfected with the Cald1 WT-Luc construct and Sox4-GFP constructs. The activity of Cald1 WT-Luc increased up to ∼sixfold with the expression of GFP-Sox4 protein (Fig. 5C). A reporter assay with the 622 bp non-coding region of the mouse Tubb3 gene was used as a positive control for Sox4-mediated gene expression (Bergsland et al., 2006).

To determine whether Sox4-dependent Cald1 promoter activity is associated with C2C12 myoblast differentiation, constructs containing the full-length sequence of the Cald1 promoter (WT-Luc) or the Cald1 promoter with deletion of the Sox4-binding sequence (ΔSox4-Luc) were transfected into C2C12 myoblast cells. Newly synthesized luciferase protein was analyzed by immunofluorescence and western blot. At 3 days after differentiation, cells transfected with the WT-Luc construct showed strong expression of luciferase in myotubes, whereas luciferase expression was nearly absent in cells transfected with the ΔSox4-Luc construct (Fig. 5D, left panel). These results were confirmed by western blot analysis (Fig. 5D, right panel). As expected, none of the cells expressed luciferase when cultured in proliferation medium (D0). Meanwhile, cells transfected with WT-Luc, but not ΔSox4-Luc, synthesized luciferase when cultured in differentiation medium (D3). Myoblast differentiation was confirmed by the expression of myogenin. These results suggest that Sox4 is the core transcription factor regulating the promoter activity of Cald1 in myoblast differentiation.

Having shown that Sox4 regulates the expression of CaD in differentiated myoblasts, we next examined whether Sox4 silencing inhibits myoblast differentiation. As shown in Fig. 6A, Sox4 was expressed in proliferating C2C12 myoblasts, and the protein level increased modestly as differentiation progressed (Fig. 6A,B). To examine the effect of Sox4 silencing on myoblast differentiation, we transfected C2C12 myoblasts with Sox4 siRNA or with control siRNA and analyzed the lysates by western blot analysis 2 days later. In cells transfected with Sox4 siRNA (#1 and #2), but not in cells transfected with control siRNA, the level of Sox4 decreased by almost 90% (Fig. 6C). The expression of CaD and myogenic marker proteins was also reduced in the Sox4-silenced cells. The level of tubulin was unaffected. Consistent with this finding, the percentage of fused myoblasts also decreased in Sox4-silenced C2C12 myoblasts, which expressed MHC weakly (Fig. 6D,E). We then examined whether CaD overexpression could rescue the inhibitory effect of Sox4 siRNA on myoblast fusion. Sox4-silenced C2C12 cells were transfected with a FLAG-CaD construct and incubated for 3 days in differentiation medium. To verify the transfection efficiency, we transfected C2C12 myoblasts with an equal amount (3.0 µg) of GFP cDNA and examined GFP expression with florescence microscopy. GFP fluorescence was observed in almost 90% of the cells and this signal was detectable for more than 3 days. Thus, the transfection efficiency was high enough to perform the experiment with transient expression (supplementary material Fig. S5). Although overexpression of CaD in Sox4-silenced cells did not fully restore myoblast fusion to the level observed in mock-siRNA-transfected cells, both myotube formation and MHC expression were moderately rescued by CaD overexpression (Fig. 6E,F). To ensure that myoblast fusion and MHC expression in Sox4-silenced C2C12 cells was rescued only in cells that overexpressed FLAG-CaD, we co-transfected FLAG-CaD and GFP constructs at a 5∶1 molar ratio. Immunofluorescence analysis confirmed that myoblast differentiation and MHC expression were restored only in cells expressing CaD (supplementary material Fig. S6).

To understand the role of Sox4-mediated CaD protein expression in skeletal muscle differentiation in vivo, we introduced control or Sox4 siRNA, with or without a GFP expression vector, directly into the hindlimb muscle of mouse embryos at embryonic day (E) 14 by electroporation (Cardoso et al., 2004). To verify the transfection efficiency, we electroporated various amounts of GFP cDNA into tissue samples and quantitatively analyzed the extent of transfection. GFP fluorescence was detected in more than 90% of myofibers after transfection of 3 µg of GFP cDNA (supplementary material Fig. S7). Consistent with the results from C2C12 myoblasts, silencing of Sox4 inhibited CaD and MHC expression in the hindlimb muscle of mouse embryos (Fig. 7A).

Because decreased MHC expression during skeletal muscle differentiation results in skeletal muscle hypotrophy (smaller myofibers) or hypoplasia (fewer myofibers) (O'Rourke et al., 2007), we analyzed the morphological characteristics of myofibers in the Sox4-silenced mouse embryonic hindlimb muscle by staining for MHC and actin (Fig. 7B). As shown in Fig. 7C,D, the number of myofibers did not significantly differ between the samples, but the projected area of the myofibers decreased slightly upon silencing of Sox4. Of interest, MHC in control-siRNA-transfected muscle appeared as spot-like structures that completely colocalized with actin in the myofibers. However, in muscle transfected with Sox4 siRNA, distinct MHC structures were rarely detected, whereas some fibers still exhibited spot-like actin structures (Fig. 7B,E). Because actin and MHC assembly is crucial for myofibril formation in functional muscle development (Sanger et al., 2005), our results suggest that Sox4-mediated CaD expression is important for myofibril assembly during embryonic skeletal muscle differentiation.

During skeletal muscle differentiation, the regulation of cytoskeletal structure in myoblasts is essential for the initiation of membrane fusion at an early stage of differentiation and for the formation of a contractile unit in the myotube (Richardson et al., 2008). The establishment of actin cytoskeletal interactions with the extracellular matrix and between cells is important for myoblast differentiation (Kim et al., 2007). In addition, numerous studies have shown that actin-regulating proteins are involved in muscle differentiation and development (Bongiovanni et al., 2012; Nowak et al., 2009; Richardson et al., 2008). Suppression of Rho family small-GTPase activity by pharmacological inhibitors interferes with actin remodeling and the expression of myogenic factors such as myogenin and MRF4 (Bryan et al., 2005). However, little is known about the transcriptional regulation of actin-regulating proteins and its role in myoblast differentiation. In this study, we found that CaD regulates cell spreading and migration in the early stage of C2C12 myoblast differentiation. The expression of CaD in the early stage of differentiation is controlled at the transcriptional level by Sox4 transcription factor.

The cellular functions of CaD are controversial. There is evidence that h-CaD expression enhances stress fiber stability and reduces cell motility in non-muscle cells (Castellino et al., 1992; Castellino et al., 1995; Eppinga et al., 2006; Li et al., 2004; Mayanagi et al., 2008; Mirzapoiazova et al., 2005). Moreover, CaD expression induced by p53 suppresses Src-kinase-dependent podosome formation, which inhibits fibroblast and rat aortic SMC migration (Mukhopadhyay et al., 2009). By contrast, HeLa–l-CaD expression in endothelial cells and endothelial progenitor cells in the vasculature of various human tumors promotes cell migration for vasculogenesis and angiogenesis during tumor development (Zheng et al., 2004; Zheng et al., 2007). Of note, we observed that endogenous CaD in C2C12 myoblasts localized to the membrane ruffles, suggesting that CaD is involved in cell migration in C2C12 myoblasts. Consistent with this idea, silencing of CaD in C2C12 cells strongly inhibited cell spreading and migration (Fig. 2 and supplementary material Fig. S2). Disassembly of vinculin- and talin-containing focal adhesions was observed in HeLa–l-CaD-positive endothelial cells (Zheng et al., 2007). Overexpression of l-CaD promotes cell movement, but inhibits cell contractility with a decrease in focal adhesions (Helfman et al., 1999). Meanwhile, the increase in cell contractility with vinculin- and talin-mediated strong adhesions prevents the cell spreading of fibroblasts (Park et al., 2002). Thus, it is plausible that CaD in C2C12 myoblasts controls the dynamics of focal adhesion assembly, followed by the induction of cell migration in response to differentiation signaling.

CaD is expressed as two different isoforms, which are mainly generated by alternative splicing (Mayanagi and Sobue, 2011). Selective splicing of exon 1a or 1b produces the fibro-type or HeLa-type N-terminal sequence of CaD, respectively. Amino acid sequence comparison of mouse CaD and human fibro l-CaD-I or l-CaD-II showed that mouse CaD has no significant homology with human fibro l-CaD proteins in the N-terminal region, which corresponds to the first encoding exon. Instead, mouse CaD shares 78% identity with the N-terminus of HeLa l-CaD proteins (supplementary material Fig. S8A). Moreover, additional amino acid alignments of mouse CaD and human HeLa l-CaD-I or l-CaD-II confirmed that the CaD protein in C2C12 myoblasts is identical to HeLa l-CaD-II in humans (supplementary material Fig. S8B,C).

Furthermore, analysis of the genomic structure of Cald1 indicates that the transcripts encoding the CaD isoforms in chickens and humans are generated by independent promoters (Hayashi et al., 1992; Yano et al., 1994). In the human CALD1 gene, the promoter responsible for the fibro-type CaD protein is located in the proximal region of exon 1a-1, which generates h-CaD and l-CaD proteins; the HeLa-type promoter located upstream of exon 1b generates only l-CaD proteins (HeLa l-CaD-I and l-CaD-II) (Hayashi et al., 1992). Thus, it is reasonable to expect that the expression of mouse CaD is also regulated by the HeLa-type promoter. We also identified a fibro-type-like promoter for the mouse Cald1 gene within the coding region of an uncharacterized protein (LOC100861595; gene ID: 100861595). The promoter sequence is positioned approximately −110 kbp upstream from the transcription start site of the mouse Cald1 gene. Thus, it does not appear that the fibro-type-like promoter sequence is involved in mouse Cald1 gene expression. This was not a major concern for the analysis of HeLa-type promoter activity related to CaD expression because HeLa-type promoter activity is weaker than fibro-type promoter activity in various tissues (Fukumoto et al., 2009; Yano et al., 1995).

To date, several cis-acting elements such as the CArG box and GRE sequence, which are primarily responsible for the transcription of CALD1, have been identified (Fukumoto et al., 2009; Mayanagi et al., 2008; Momiyama et al., 1998; Yano et al., 1995). These conserved cis-acting elements are located in the fibro-type promoter. However, little is known about the cis-acting elements in the HeLa-type promoter. Sequence analysis of the mouse Cald1 promoter indicates that the cis-acting elements in the HeLa-type promoter are completely different from those in the fibro-type promoter. The promoter of mouse Cald1 is nearly identical to the HeLa-type promoter in humans. In particular, the Sox4-binding site is completely conserved in both promoters (Fig. 4A). Thus, the Sox4-binding site might serve as a crucial cis-acting element for HeLa-type promoter activity.

The promoter of the mouse Cald1 gene has several other conserved sequences that are capable of binding to transcription factors, including GATA2, GATA3, AP1, SP1 and Nkx2.5 (Fig. 4). However, these transcription factors repress myoblast differentiation, and their expression decreases during differentiation of C2C12 myoblasts (Itoh et al., 2008; Lehtinen et al., 1996; Riazi et al., 2005). Therefore, these trans-acting factors, except for Sox4, appear not to be involved in CaD expression during myoblast differentiation. Some Sox transcription factors might require binding partners to selectively recognize their target promoters (Wilson and Koopman, 2002). For example, Sox2, Sox8 and Sox10 interact with POU (Pit-1, Oct and Unc-86) transcription factors to recognize their target genes (Bernard and Harley, 2010). However, no POU transcription factors have yet been reported to bind Sox4 to regulate its transcriptional activity. Interestingly, we found that the mouse Cald1 promoter contains binding sequences for the Oct-1 transcription factor, a member of the POU family, near the Sox4-binding sequence, with lower score threshold (supplementary material Fig. S9). Oct-1 interacts with other Sox family proteins through their HMG domain, and the HMG domain in Sox4 is highly conserved with other Sox proteins. Thus, further study of the regulation of Sox4 activity by Oct-1 might provide new insight into Sox4-mediated differentiation of myoblasts.

Although the exact functions of Sox4 were not elucidated until a few years ago, Sox4 has been considered an important factor for the development of various organs and tissues, including the heart (Schilham et al., 1996), thymocytes (Schilham et al., 1997), nervous system (Cheung et al., 2000) and osteoblasts (Nissen-Meyer et al., 2007). Recently, functional studies of Sox4 have focused on its undefined roles in developmental processes. For instance, Sox4 is required for neural and glial cell development because it regulates pan-neuronal gene expression, including the transcription of the class-III β-tubulin gene Tubb3, which is required for neuronal maturation (Bergsland et al., 2006). Members of the SoxC group, including Sox11 and Sox12, in addition to Sox4, might have essential function in early mouse embryonic organogenesis; they directly target the Tead2 gene, a crucial transcription factor in the Hippo signaling pathway (Bhattaram et al., 2010). Recent reports have also shown that Sox4 and Sox11 are necessary for the control of corticospinal system development; Sox4 and Sox11 directly regulate Fezf2, which encodes a key factor for the specification of identity and the connectivity of corticospinal neurons (Shim et al., 2012). In C2C12 myoblasts, our quantitative RT-PCR results showed that the levels of Sox11 and Sox12 transcripts were extremely low compared with the level of Sox4; their expression pattern did not change during myoblast differentiation. Thus, of the Sox family members, Sox4 is probably the primary transcription factor for C2C12 myoblast differentiation (supplementary material Fig. S10).

The expression of myogenic regulatory factors (MRFs), including MyoD and myogenin, is believed to orchestrate the gene regulatory network that controls muscle differentiation. Proliferating myoblasts express the MyoD by which the cell can be committed to the muscle lineage. Myogenin expression increases dramatically at the onset of differentiation to regulate the expression of various downstream genes, including myosin (Braun and Gautel, 2011). Sox4 is expressed in proliferating cells, and its expression increases moderately upon serum deprivation. In this study, silencing of Sox4 by siRNA inhibited myoblast differentiation and reduced MyoD expression, indicating that Sox4 functions as a myogenic factor. Promoter sequence analysis showed that the promoter of the Myod gene contains Sox4-binding sites at various positions, although the score threshold was low. Thus, it is likely that Sox4 can control the transcription of Myod. Alternatively, Sox4 might have a similar function as a muscle-determining factor upstream of MyoD. The identification of target genes regulated by Sox4 activity will help define the molecular mechanism of skeletal muscle differentiation.

Consistent with our in vitro results, Sox4 siRNA decreased the levels of CaD and MHC when electroporated into the hindlimb of mouse embryos (Fig. 7). Interestingly, immunofluorescence staining with anti-MHC antibody and phalloidin demonstrated that the projected area of myofibers decreased in Sox4-silenced embryonic muscle tissue. Furthermore, MHC organization in individual myofibrils appeared defective, but the overall number of myofibers was similar to the number in control-siRNA-transfected muscle. CaD can interact with F-actin, myosin, calmodulin (CaM) and tropomyosin (TM) (Graceffa, 1987; Smith et al., 1987; Sobue et al., 1981; Sobue and Sellers, 1991; Wang et al., 1997). Moreover, CaD enhances the binding of F-actin to TM and competes with gelsolin, an actin-severing protein that binds to F-actin (Ishikawa et al., 1989; Warren et al., 1994). Consequently, because it promotes stable actin-myosin structures in smooth muscle, CaD is an important regulatory protein for thin filament organization (Sobue and Sellers, 1991). In developing skeletal muscle, the actin-myosin system is also required for normal muscle functions such as contraction or relaxation. Taken together, our results suggest that Sox4-mediated CaD expression is involved in actomyosin organization in myofibrils during skeletal muscle development.

Human embryonic kidney 293 (HEK293) cells and C2C12 cells, a myogenic cell line derived from mouse, were obtained from the American Type Culture Collection (Manassas, VA) and maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and penicillin-streptomycin (50 units/ml; Invitrogen) in a humidified atmosphere of 5% CO₂ at 37°C. For all experiments, C2C12 cells were seeded in proliferation medium (GM) at a density of 2×10⁴ cells/cm². Differentiation was induced 24 hours after seeding by exchanging GM for differentiation medium (DM; DMEM supplemented with 2% horse serum). Cells were transiently transfected with different plasmid DNA using a Neon electroporation system (Invitrogen) according to the manufacturer's instructions.

The full-length coding region of SOX4 was amplified from HEK293 cell cDNA generated by reverse transcriptase (Intron, Korea). Amplified SOX4 cDNA was introduced into a pEGFPC2 vector (6083; Clontech, Mountain View, CA). The full-length coding region of mouse Cald1 was amplified from C2C12 cell cDNA, generated by reverse transcriptase, and introduced into a pFLAG-CMV2 vector (E7033; Sigma-Aldrich). Tubb3 and mouse Cald1 promoter fragments were generated by PCR using genomic DNA extracted from C2C12 cells. Truncated mutants and Sox4-binding site deletion mutants were amplified from the WT Cald1 promoter using sequential PCR and were introduced into a pGL.4.12 basic vector (E6671; Promega, Madison, WI). For siRNA constructs, oligonucleotides for CaD siRNA and Sox4 siRNA were introduced into the pBabe-dual vector. All clones were verified by DNA sequencing. The PCR primers used in this study and their sequences are listed in supplementary material Table S1.

HEK293 or C2C12 cells were transfected with vectors containing the firefly luciferase reporter gene (0.1 µg) and pCMV-β-galactosidase (0.1 µg) together with GFP-Sox4 or siRNA Sox4 using an electroporation method. Thereafter, transfected cells were partially selected under puromycin or neomycin for 3–4 days. Transfected cells were lysed in reporter lysis buffer (Promega), and cell extracts were analyzed with a luciferase reporter assay system using a GloMax luminometer (Promega). Luciferase activities were normalized to the β-galactosidase activity of the co-transfected vector.

Polyclonal antibodies against Sox4 were produced in a rabbit using a Sox4-specific peptide (TNNAENTEALLAGESSDSGA) (Peptron, Korea) as an antigen. Pre-immune serum from the same rabbit was obtained before immunization. After boosting four times at 2 week intervals, the serum was tested by western blot analysis using purified GST-Sox4 DBD protein. The specificity of the purified antibodies was further analyzed by western blotting using serial dilution.

Oligonucleotide labeling and EMSA were performed as described by Hellman and Fried (Hellman and Fried, 2007). The oligonucleotides for the Cald1 promoter and their sequences are listed in supplementary material Table S1. The DNA–protein binding reaction was conducted in a mixture containing 10× binding buffer (100 mM Tris-HCl, pH 7.5, 10 mM EDTA, 1 M KCl, 1 mM DTT, 50% v/v glycerol, 0.1 mg/ml BSA), 4000 cpm of ³²P-labeled oligonucleotide and affinity-purified GST-Sox4 DBD protein for 30 minutes at 30°C, and resolved on a 6% acrylamide gel that had been pre-run at 150 V for 1 hour with 400 mM TAE (Tris, acetic acid and EDTA) buffer. The loaded gel was run at 200 V for 40 minutes, dried and placed on Kodak X-ray film (Eastman Kodak, Rochester, NY) for autoradiography. After 16 hours of exposure at −20°C, the film was developed.

The ChIP assay was performed following a protocol provided by Millipore (Temecula, CA). C2C12 cells were sonicated to shear chromatin to 500–1000 bp fragments. For immunoprecipitation, 2 mg of anti-Sox4 antibodies were incubated with the cell lysates overnight at 4°C with rotation. Lysates were immunoprecipitated with IgG as a control. Precipitated DNA was analyzed by quantitative real-time PCR using Cald1 promoter-specific primers, as indicated in supplementary material Table S1.

Anti-CaD (sc-15374; Santa Cruz Biotechnology), myogenin (sc-12732; Santa Cruz Biotechnology), MHC (MF20; Developmental Studies Hybridoma Bank) and luciferase (ab21176; Abcam) antibodies were used for immunoblotting. Horseradish-peroxidase-conjugated secondary antibodies were used for detection and immune complexes were visualized with ECL-chemiluminescence (sc-2048; Santa Cruz Biotechnology). Immunocytochemistry and immunohistochemistry were performed following a protocol provided by Abcam. Primary antibodies were used at 1∶200 for CaD, 1∶200 for MHC, and 1∶500 for luciferase. Cy3-conjugated donkey anti-mouse IgG (1∶200; Jackson ImmunoResearch Laboratories) and FITC-conjugated donkey anti-rabbit IgG (1∶500; Abcam) were used as secondary antibodies. DAPI (Sigma) was used for nuclear staining. Rhodamine-Phalloidin was used for polymerized actin staining (PHDR1; Cytoskeleton). Samples were incubated with 300 nM DAPI in PBS for 2 minutes at room temperature and visualized using a Nikon Eclipse 80i microscope (Melville, New York).

The DNA constructs indicated in Fig. 7 were prepared using a Qiagen maxiprep kit (Qiagen, Valencia, CA) and concentrated using an ethanol-precipitation method to a final concentration of 3.0 µg/µl DNA. A solution of DNA (1 µl) was injected into the hindlimb muscles using a pulled glass capillary connected to an IM6 microinjector. The injected muscle tissue was then placed between the electrodes of the electroporation chamber and covered with ice-cold PBS. Square-wave pulses (35 V; pulse length, 150 milliseconds; three pulses at 100 millisecond intervals) were delivered by the electroporator (BTX ECM 830; Harvard Apparatus, Holliston, MA). Electroporated hindlimb muscles of mouse embryos were incubated using in vitro muscle culture systems as described by Cardoso and co-workers (Cardoso et al., 2004).

The extent of myoblast fusion was expressed as the percentage of the number of nuclei in fused cells relative to the total number of nuclei in randomly chosen fields under a microscope. Cells were considered fused only when there was clear cytoplasmic continuity and at least three nuclei were present in each myotube.

We are grateful to Dr Kun Ho Lee (Chosun University, Korea) for helpful comments and suggestions.