Inhibition of myogenic differentiation in myoblasts expressing a truncated type II TGF-β receptor

The decision of a cell to proliferate, differentiate and migrate during development is intimately connected to its environment. Transplantation and ablation studies have shown the importance of cell-cell interactions and the extracellular milieu in the determination of cell fate (Greenwald and Rubin, 1992; Gurdon, 1992). Several peptide growth factors have been shown to play a role in inductive processes that give rise to differentiated cell types (Jessell and Melton, 1992). Transforming growth factor-β (TGF-β), one such factor, affects the growth and differentiation of many cell types in vitro, especially those of mesenchymal origin (Ignotz and Massagué, 1985; Rosen et al., 1988; Torti et al., 1989). Its expression pattern during mouse development also suggests an important function in specific morphogenetic and differentiation events in vivo (Heine et al., 1987; Millan et al., 1991; Pelton et al., 1991). In particular, TGF-β is highly expressed during periods of morphogenesis or remodeling of mesenchyme and at sites of mesenchymal-epithelial interactions (Heine et al., 1987).

TGF-β may play an important role in myogenic differentiation (Olson et al., 1986). TGF-β is thought to be involved in cardiomyogenesis (Potts et al., 1989, 1991) and its expression is induced during experimental myocardial infarction (Thompson et al., 1988; Lefer et al., 1990). However, the role of TGF-β in skeletal myoblast differentiation is unclear, since exogenous TGF-β has both positive and negative effects on muscle cell development in vitro. On the one hand, treatment of skeletal myoblast cell lines or primary muscle cells with TGF-β in low serum inhibits terminal differentiation (Massagué et al., 1986). In addition, TGF-β blocks the expression and function of two muscle-specific transcription factors, MyoD (Vaidya et al., 1989) and myogenin (Brennan et al., 1991), thereby preventing expression of downstream muscle transcripts. On the other hand, TGF-β in normal serum can induce differentiation of myoblasts (Zentella and Massagué, 1992) and treatment of embryonic stem cells with TGF-β results in preferential differentiation of cells into muscle (Slager et al., 1993).

TGF-β acts by binding to a set of specific receptors (Massagué, 1992). Of these, the type I and type II TGF-β receptors mediate most of its biological effects (Laiho et al., 1990a, 1991; Geiser et al., 1992). The cytoplasmic domains of the cloned type I and type II TGF-β receptors have sequences characteristic of serine-threonine kinases (Lin et al., 1992; Ebner et al., 1993a, Franzén et al., 1993), and the type II receptor has been shown to be a functional kinase (Wrana et al., 1992). A physical association between the type I and type II receptors has been proposed (Wrana et al., 1992), and the type I receptor may require the type II receptor for its activity (Ebner et al., 1993a, b, Franzén et al., 1993, Bassing et al., 1994). Functional inactivation of the type II receptor in an epithelial cell line has indicated that the type II receptor is required for the antiproliferative effect of TGF-β and that its effect on extracellular matrix protein synthesis may be mediated through the type I receptor (Chen et al., 1993). Which of these two receptors mediates the effects of TGF-β on myogenesis is not yet known. Whereas most cells in culture have both receptors, the type II receptor is expressed in vivo at high levels in undifferentiated mesenchyme and in differentiated muscle tissue (Lawler et al., 1994). The co-localization of TGF-β expression suggests a developmental role for TGF-β during muscle cell differentiation (Heine et al., 1987; Pelton et al., 1991; Lawler et al., 1994).

To address the seemingly paradoxical observations on the effects of TGF-β on myoblasts and to gain insight into its role during myogenic differentiation, we transfected C2C12 myoblasts (Yaffe and Saxel, 1977; Blau et al., 1983) with a truncated form of the type II TGF-β receptor. This truncated receptor inhibits signaling by the type II, but not the type I, TGF-β receptor in a dominant negative fashion (Brand et al., 1993; Chen et al., 1993). We show that myoblasts expressing the truncated type II TGF-β receptor do not undergo morphological or biochemical differentiation. This inhibition of differentiation is associated with decreased expression of the myogenic determining genes, MyoD and myogenin, and can be rescued following infection with a MyoD-expressing retrovirus. These studies suggest that some, but not all, changes associated with myotube formation require signaling through the TGF-β receptor, and that TGF-β functions in an autocrine or paracrine fashion as a competence factor for myogenic differentiation.

C2C12 cells which have been clonally purified for reproducible myogenic differentiation (Blau et al., 1983) from C2 cells (Yaffe and Saxel, 1977), were obtained from Dr H. Blau and grown in 20% fetal calf serum (FCS) in DMEM. To induce differentiation into myotubes, the cells were switched into DMEM containing 2% horse serum for 1 to 3 days. C2C12 cells were transfected with pcDNA1Neo (InVitrogen) expressing the truncated type II TGF-β receptor and encoding neomycin resistance (Chen et al., 1993) using the calcium phosphate precipitation method (Sambrook, 1989). The transfected cells were then cultured in medium containing G418 (400 μg/ml). After 14 to 21 days, G418 resistant clones were isolated, expanded and screened by northern hybridization for expression of mRNA for the truncated receptor. The three clones selected for this study were chosen on the basis of cell surface expression of the truncated receptor, as determined by chemical cross-linking using ¹²⁵I-TGF-β. Five randomly selected neomycin-resistant, controltransfected clones were also analyzed and found to undergo normal myogenic differentiation.

3 days after plating equal numbers of cells (determined using a Coulter counter), cells were switched to differentiation media (2% horse serum) or were kept in growth media (20% fetal calf serum) for three more days. RNA was isolated (Chomczynski and Sacchi, 1987) and analyzed by northern analysis (Sambrook, 1989) using cDNA probes radiolabelled by random priming using a commercial kit (Boehringer Mannheim). Following hybridization at 42°C for 24 hours, the nitrocellulose blots were washed in 0.5× SSC, 0.5% SDS for 20 minutes at 42°C and 0.1× SSC, 0.5% SDS for 30 minutes at 55°C. cDNAs for myosin light chain 2, MyoD (Davis et al., 1987) and myosin light chain 1/3 (Periasamy et al., 1984, modified by Dr N. Rosenthal), the acetylcholine receptor α subunit (Isenberg et al., 1986) and myogenin (Wright et al., 1989) were used as hybridization probes to assess the differentiation state.

Recombinant human TGF-β1 was ¹²⁵I-labelled using a slightly modified chloramine T method (Frolik et al., 1984). Cross-linking of parental and transfected C2C12 cells was carried out as described (Gazit et al., 1993).

Equal quantities of protein (determined by the Biorad colorimetric assay) were electrophoresed in 12% (for troponin T or desmin) or 7.5% (for pRB or myosin heavy chain) denaturing polyacrylamide gels and transferred to nitrocellulose membranes at 40 V for 2 hours. The blots were washed for 10 minutes in TBST (25 mM Tris pH 8.0, 125 mM NaCl, 0.025% Tween 20), incubated in blocking buffer (25 mM Tris pH 8.0, 125 mM NaCl, 0.1% Tween 20, 1% BSA, 0.1% NaN₃) for 2 hours at room temperature or 24 hours at 4° C, and then incubated with the antibody for 2 hours at room temperature. They were then washed three times with blocking buffer (15 minutes per wash) and incubated with a secondary, alkaline phosphatase-conjugated antibody (Promega) for 1 hour. After three 15 minute washes with blocking buffer and three 5 minute washes with TBST, the blots were incubated in phosphatase substrate (BCIP/NBT, Kirkegaard and Perry Laboratories) for 5-30 minutes and, once developed, washed with distilled water. Antibodies for troponin T, myosin heavy chain (i.e. MF20 antibody, Sigma) and desmin were provided by Dr Charles Ordahl (UCSF). The antibodies to pRB (DeCaprio et al., 1988) were purchased from Pharmingen.

Cells on coverslips were washed three times with phosphate-buffered saline (PBS) and fixed in methanol (for myosin heavy chain antibodies) or in 4% formaldehyde in PBS (for troponin T or desmin antibodies) for 10 minutes at room temperature. After washing with PBS, cells were incubated in 0.1% Triton in PBS for 10 minutes and rinsed again with PBS. The samples were then treated with 3% bovine serum albumin (BSA) in PBS for 15–30 minutes and incubated with primary antibody for 45 minutes at 37° C. After three washes in PBS, the cells were incubated with a rhodamine-conjugated secondary antibody for 45 minutes at 37° C. The cells were then washed three times with PBS, hydrated in 70% then 100% ethanol (3 minute washes each), air dried, mounted with Fluoromount-G (Fisher Scientific) and a coverslip, and photographed with a Zeiss Axioplan microscope..

On day 0, the cells were counted and plated such that by the day of treatment, all samples would have approximately the same cell number. On day 4, one set of plates was changed to differentiation media (2% horse serum), while the other set remained in growth media (20% serum). All cells were changed and counted every other day using a Coulter counter.

To construct the retrovirus encoding hygromycin resistance and βgalactosidase, the β-galactosidase gene was cut from plasmid pML62 (constructed by Drs M. Landowski and G. Martin) with SalI and BamHI and ligated into the SalI and BamHI sites of a derivative of pLXSH (Miller et al., 1993), obtained from Drs M. Lochric and H. Varmus. The packaging cell line, PA317 (Miller et al., 1993), was transfected with this plasmid and the retrovirus-containing conditioned media, harvested 48 hours later, was used to infect another packaging cell line, PE501. These cells were grown for 2 weeks in DMEM/10% serum with hygromycin, and individual colonies were isolated and grown. Viruses produced from these clones were titered using NIH-3T3 fibroblasts, and the highest titer virus stock was used to infect C2C12 cells. 48 hours after infection, C2C12 cells were switched to growth media with hygromycin (500 μg/ml) and selected for two weeks prior to X-gal staining.

The high titer MyoD retrovirus was kindly supplied by Dr Dusty Miller (Weintraub et al., 1989). Infections were performed by incubating growing cells overnight with viral stock plus 4 μg/ml of polybrene.

For intramuscular injections into mice, the cells were trypsinized, washed twice and resuspended in PBS at 4°C. Approximately 4×10⁵ myoblasts (in 20 μl) were delivered via 4 injections (5 μl each) into the hind limbs of anesthetized 3to 4-week-old nude mice. 2 weeks later, these limbs were removed from euthanized mice and frozen in isopentane, serially sectioned, and fixed and stained for β-galactosidase activity as described (Dhawan et al., 1991). Muscle fibers were examined across multiple serial sections throughout the length of the limb. Sections were mounted in Gel/Mount (Biomeda Corporation) and photographed with a Zeiss Axioplan microscope.

The mouse C2C12 myoblast cell line (Yaffe and Saxel, 1977; Blau et al., 1983) is frequently used as a model to study the process of myogenic differentiation in vitro. This clonal cell population has highly reproducible differentiation properties (Blau et al., 1983). These cells proliferate as mononuclear myoblasts in high (20%) serum media (growth media), but become growth arrested and undergo morphological and biochemical differentiation after a switch to low (2%) serum media (differentiation media) (Blau et al., 1983). To examine the role of TGF-β and its signaling through the type II receptor during myogenesis, C2C12 cells were transfected with a truncated type II TGF-β receptor expression vector containing a neomycin-resistance marker. This truncated receptor, which lacks its cytoplasmic domain, has been shown to inhibit signaling specifically through the type II TGF-β receptor in a dominant negative fashion (Chen et al., 1993). After 2 weeks, G418-resistant colonies were isolated, propagated as stable cell lines, and tested for expression of the truncated receptor mRNA by northern blotting (data not shown). To verify expression of the truncated TGF-β receptor protein, the transfected cells were incubated with ¹²⁵I-labelled TGF-β, and the cross-linked receptors were analyzed by polyacrylamide gel electrophoresis. As shown in Fig. 1A, the truncated receptor was expressed at the cell surface of transfected clones and bound TGF-β. The ¹²⁵I-TGF-β cross-linked receptor could be immunoprecipitated with antibodies against an epitope tag engineered at its carboxy terminus (Fig. 1B). Of the five clones that expressed the truncated receptor mRNA, three showed expression of the corresponding protein at the cell surface. These three clones were used for further analysis and compared with parental and neomycin-resistant control C2C12 cells.

Under proliferative conditions, cells transfected with the truncated receptor were morphologically similar to the parental C2C12 cells (Fig. 2A−). To induce differentiation, cells were switched to low serum (2%) media. Within 3 days, the parental C2C12 cells ceased proliferating and fused to form multinucleated fibers. In contrast, in the three C2C12 clones expressing the truncated type II TGF-β receptor? very few (< 5%) of the cells formed myotubes (Fig. 2A+). Their lack of differentiation was not due to delayed onset of differentiation, since transfectants maintained for up to 9 days in low serum did not show significant myotube fusion (Fig. 2B).

These clones are unlikely to be spontaneous, nondifferentiating mutants, since the probability of picking a single natural, non-differentiating clone of these cells is less than 10⁻³ (Lassar et al., 1989). Therefore, the probability that all three clones are natural, non-differentiating mutants is less than 10⁻⁹. Furthermore, neither we nor others (Lassar et al., 1989) have detected a non-differentiating phenotype in control, neomycin-resistant colonies. Our results therefore suggest that expression of a truncated type II TGF-β receptor inhibits morphological differentiation of C2C12 myoblasts.

Since myoblast fusion is not required for the induction of differentiation-specific gene expression (Endo and Nadal-Ginard, 1987; Konieczny et al., 1989; Hu and Olson, 1990), we tested whether our myotube fusion-defective clones were capable of biochemical differentiation following a switch to low serum medium. Northern hybridizations showed that, as expected, the parental C2C12 cells induced mRNA for myosin light chains 1/3 and myosin light chain 2 after a switch to differentiation media. The expression of myosin mRNAs in the parental C2C12 cells in growth media is due to spontaneous differentiation of a fraction of the confluent cells, as previously observed (Lassar et al., 1989). In contrast, when cultured in growth media, cells expressing the truncated receptor did not have detectable myosin mRNA levels, and a change to differentiation media resulted in little induction of these mRNAs (Fig. 3). These clones, however, responded to the change in serum concentration by inducing the α chain of the acetylcholine receptor mRNA to levels comparable to the parental C2C12 cells (Fig. 3).

Western analysis showed the expected induction of troponin T and myosin heavy chain in the parental C2C12, cultured in low serum for 3 days. In contrast, all three transfected clones showed undetectable levels of these proteins in growth media, and had reduced levels of troponin T, and little, if any, myosin heavy chain expression following incubation in differentiation media (Fig. 4). The transfected clones contained significant desmin expression under both conditions. To analyze the expression of these proteins in individual cells, the cells were grown in differentiation media for 3 days and stained with antibodies to troponin T, myosin heavy chain and desmin. The parental C2C12 cells exhibited myotubes which stained strongly with antibodies to all three proteins. In contrast, cells expressing the truncated receptor expressed little if any troponin T or myosin heavy chain (Fig. 5). Using antibodies to desmin, a low level background, punctuated by several brightly staining mononuclear cells, was observed in the transfected clones (Fig. 5). Thus, C2C12 myoblasts expressing the truncated receptor do not exhibit normal morphological or biochemical differentiation. However, not all differentiation markers are equally affected by expression of the truncated receptor.

Proliferating myoblasts express the protein Id, an inhibitor of differentiation (Benezra et al., 1990). In high serum media, Id prevents MyoD activation of downstream genes by competing with MyoD for binding to the E12 protein (Murre et al., 1989a, b). Upon a switch to low serum media, Id expression is greatly decreased in cultured myoblasts (Benezra et al., 1990). Since down-regulation of Id expression is required for normal myoblast differentation (Weintraub et al., 1991), we compared the level of Id mRNA in the parental and transfected C2C12 cells. As expected, exposure of C2C12 cells to low serum decreased Id mRNA levels (Fig. 3). Surprisingly, cells expressing the truncated receptor had a similar downregulation of Id mRNA levels when switched to low serum, suggesting that the regulation of this gene is not affected by expression of the truncated TGF-β receptor. This result agrees with studies showing that TGF-β does not affect Id expression (Brennan et al., 1991). Furthermore, the inability of the transfected clones to differentiate cannot be explained by a failure of Id to be downregulated in low serum.

MyoD and myogenin are expressed only in skeletal muscle cells, and their expression is upregulated during myogenic differentiation in vitro (Davis et al., 1987; Edmondson and Olson, 1989; Wright et al., 1989). In addition, expression of transfected cDNAs for these transcription factors converts various cell types to a muscle phenotype (Davis et al., 1987; Weintraub et al., 1989). Thus, MyoD and myogenin are believed to be involved in the commitment of cells to the myogenic lineage and in the establishment and maintenance of the differentiated phenotype (Weintraub et al., 1991). To ascertain whether the truncated TGF-β receptor affected the determination of cell identity, we examined the expression of MyoD and myogenin. As expected, C2C12 cells induced transcripts for both MyoD and myogenin after a switch to differentiation media. Expression of myogenin in these cells in growth media is most likely due to spontaneous differentiation which occurs at high cell density (Lassar et al., 1989). Unlike the parental C2C12, the cells expressing the truncated receptor showed no MyoD or myogenin expression under growing conditions, and exhibited little induction of these transcripts in differentiation media (Fig. 3). The failure of these cells to differentiate may be due to inadequate levels of MyoD and myogenin, and raises the possibility that these clones are no longer committed to the myogenic lineage.

The inability of the transfected cells to differentiate could, in theory, be due to an increase in cell growth, since proliferation and differentiation are assumed to be mutually exclusive in myoblasts (Olson, 1992). Furthermore, the differentationresistant phenotype of myoblasts, created by introduction and expression of some oncogenes, such as v-myc (Falcone et al., 1985), v-ras (Olson et al., 1987; Konieczny et al., 1989; Lassar et al., 1989) and v-fos (Lassar et al., 1989), or the Hox 7.1 homeobox gene (Song et al., 1992), is associated with cellular transformation. Thus, we evaluated the proliferation rate and transformation state of the transfected clones.

Neither our transfected cells nor the parental cells displayed anchorage-independent growth in a soft agar colony formation assay unlike our positive control, Cos-1 cells (data not shown). Therefore, inhibition of TGF-β responsiveness through the type II receptor does not induce transformation. Furthermore, the proliferation rate of the cells expressing the truncated receptor was significantly slower than of the parental C2C12 cells (Fig. 6). Thus, the inability of the transfected cells to differentiate into myotubes can not be explained by increased or uncontrolled growth. In fact, a decreased proliferation rate might be considered an advantage for differentiation.

Another possible way in which expression of the truncated TGF-β receptor could block differentiation would be to prevent the growth arrest that normally occurs in myoblasts in low serum media (Olson, 1992). The proliferation rate before and after a switch to low serum media indicated that the transfected cells ceased proliferation similarly to the parental C2C12 cells (Fig. 6). Thus, expression of the truncated type II TGF-β receptor does not prevent growth arrest of the cells in differentiation media.

Phosphorylation of the growth suppressor gene product pRB has been used as a cell cycle marker, since this protein becomes phosphorylated at the G₁/S boundary and dephosphorylated during G₂/M phase (reviewed in Marshall, 1991; Weinberg, 1991). Thus, in growth-arrested myoblasts, only the unphosphorylated form of pRB can be detected (Gu et al., 1993). In addition, pRB has been implicated in the antiproliferative effect of TGF-β (Laiho et al., 1990b; Moses et al., 1990) and plays a key role in myogenic differentiation by interacting with MyoD (Gu et al., 1993). We therefore examined the phosphorylation state of pRB under growing or differentiating conditions. Withdrawal of high serum shifted pRB to its faster migrating, dephosphorylated form in the parental C2C12 cells, as previously observed (Gu et al., 1993). A similar shift in pRB migration was observed in the transfected clones (Fig. 6). Thus, expression of the truncated TGF-β receptor does not inhibit the change in pRB phosphorylation induced by low serum in C2C12 cells, nor, by extrapolation, their withdrawal from the cell cycle and subsequent growth arrest.

When injected into skeletal muscle, C2C12 myoblasts incorporate into and fuse to form myofibers (Barr and Leiden, 1991; Dhawan et al., 1991). To evaluate the effect of expression of the truncated receptor on myoblast differentiation in vivo, we injected the transfected and control myoblasts into the skeletal muscle of hind limbs of nude mice. Prior to injection, the cells were infected with a retrovirus expressing β-galactosidase and hygromycin-resistance. Following selection in hygromycin, more than 90% of the cells expressed β-galactosidase, as assessed by histochemical staining. Furthermore, retroviral infection and β-galactosidase expression did not affect myotube formation of the parental C2C12 cells in vitro, nor did myotube formation decrease βgalactosidase expression (data not shown). 2 weeks after injection, mouse limbs were serially sectioned and analyzed for β-galactosidase expression by histochemical staining with X-gal. Sections from muscle tissue injected with the parental C2C12 cells showed many β-galactosidase-expressing muscle fibers, which resembled in morphology and size the surrounding muscle tissue (Fig. 7). Thus, as previously shown (Barr and Leiden, 1991; Dhawan et al., 1991), the parental C2C12 cells fused and incorporated into the myofibers. In contrast, tissue injected with two of the transfected cell lines contained few β-galactosidase-positive cells, and those cells appeared primarily between the endogenous muscle fibers. Similarly, fibroblasts injected into muscle do not incorporate into myofibers and gradually lose their ability to proliferate (G. Pavlath and H. Blau, personal communication). Tissue injected with the third clone showed a significant number of β-galactosidase-positive cells adjacent to, but generally not within, muscle fibers. The vast majority of these cells were unicellular and were part of a disorganized tissue (Fig. 7). Thus in contrast to the parental cells, cells expressing the truncated receptor did not fuse into endogenous muscle tissue.

The viability of the cells in vivo was verified by trypan blue staining, showing that the transfected and parental cells were not differentially sensitive to the preinjection conditions. Furthermore, injected limbs sectioned one day after injection of the clones showed many β-galactosidase-positive cells, suggesting similar early survival and β-galactosidase expression of the parental and transfected cells in vivo (data not shown). Thus, our results suggest that the in vivo environment is unable to overcome the block in differentiation observed in the transfected cells in vitro. Furthermore, the inability of the clones to mature in vivo suggests that their in vitro differences from the parental C2C12 cells have physiological relevance.

To determine whether downregulation of MyoD was required for the inhibition of myoblast differentiation, the transfected cells expressing the truncated TGF-β receptor were infected with a high titer MyoD-expressing retrovirus and tested in transient assays for their ability to differentiate. Northern analysis showed that exogenous MyoD was expressed in the infected cell population (data not shown). Based on the ratio of G418-resistant to G418-sensitive C2C12 cells, we estimate an infection efficiency of 50–70%. Of those cells expressing MyoD, we would expect approximately 50% to undergo myogenic differentiation (Davis et al., 1987; Koniezcny et al., 1989). Accordingly, 10–50% of the transfected cells infected with the MyoD retrovirus formed myotubes when cultured in high or low serum media (Fig. 8A). Furthermore, cells infected with the MyoD virus showed a similar frequency of myosin heavy chain-expressing cells, as visualized by immunofluorescence (Fig. 8B). Thus, forced expression of MyoD allows cells expressing the truncated receptor to differentiate, suggesting that downregulation of MyoD is essential for inhibition of myogenic differentiation by the truncated TGF-β receptor.

We have examined the function of TGF-β and its type II receptor-mediated signaling system during the growth and differentiation of myoblasts using C2C12 cells as a model system. Our approach was to generate cells expressing a truncated type II receptor which inhibits signaling in a dominant negative fashion (Brand et al., 1993; Chen et al., 1993). The resulting changes in the ability of myoblasts to differentiate in vitro and in vivo led to several conclusions.

Responsiveness to TGF-β, endogenously synthesized by the cells or in the media, has positive effects on the growth and differentiation of myoblasts. TGF-β may assist in myoblast proliferation as suggested by the stimulatory effect of TGF-β on C2C12 proliferation (data not shown) and the considerably slower growth rate of cells expressing the truncated type II TGF-β receptor. These transfected cells do not differentiate when cultured in low serum, as evidenced by their lack of fusion and inability to induce expression of several myoblast differentiation markers. In contrast, control transfected C2C12 cells differentiated normally. The C2C12 cells expressing the truncated receptor were still differentiation competent, since, as the parental cells, they formed myotubes following treatment with dexamethasone (data not shown). Thus, TGF-β responsiveness appears to be required for muscle-specific gene transcription and fusion during low serum-induced myogenic differentiation. Interestingly, attempts to create myoblasts overexpressing the full-length type II TGF-β receptor were unsuccessful. This would be expected if type II TGF-β receptor overexpression favored myotube formation, just as functional inactivation of the endogenous receptors blocked differentiation.

The growth arrest and change in phosphorylation of pRB, which occurred in the transfected cells after a switch to low serum, indicate that the inhibition of differentiation cannot be explained by an inability to exit the cell cycle. Transfection of myoblasts with v-myc (Falcone et al., 1985), v-ras (Olson et al., 1987; Konieczny et al., 1989; Lassar et al., 1989), v-fos (Lassar et al., 1989) or Hox 7.1 (Song et al., 1992), which block low serum-induced differentiation, also results in cellular transformation. In contrast, cells expressing the truncated receptor were not transformed, as determined by a soft agar assay.

The parental C2C12 cells express both MyoD and myogenin, and these proteins stimulate expression of themselves as well as of each other (Thayer et al., 1989; Edmondson et al., 1991). In contrast, C2C12 cells transfected with the truncated type II receptor no longer express MyoD and myogenin suggesting an involvement of TGF-β in the maintenance of their expression, perhaps by affecting their positive autoregulatory loops. Since MyoD and myogenin induce a program leading to myogenic differentiation (Weintraub et al., 1991), their suppression in the transfected cells may be the basis of the block in differentiation. This hypothesis is supported by our finding that infection of our transfected C2C12 cells with a MyoD retrovirus permits expression of myosin heavy chain and myotube formation, as observed in vras or v-fos transformed C2C12 cells (Lassar et al., 1989). This similarity is intriguing since c-ras (Mulder and Morris, 1992) and c-fos (Kim et al., 1990; Matrisian et al., 1992) may be part of the signaling pathways induced by TGF-β.

The effect of TGF-β on differentiation could depend on the presence of other growth factors (Allen and Boxhorn, 1989; Zentella and Massagué, 1992) and the immediate cellular environment. However, the inhibition of myogenic differentiation also occurred in vivo, in the context of the muscle tissue from which C2C12 cells were derived (Yaffe and Saxel, 1977). Thus, the in vivo environment could not overcome the block in differentiation, and the inhibitory effect of the truncated receptor appears to predominate. Furthermore, these in vivo results indicate that the inhibition of differentiation observed in vitro may have physiological relevance.

The dominant negative mutant of the TGF-β receptor does not block all aspects of differentiation, suggesting that at least two pathways are involved in muscle cell differentiation in vitro. One pathway regulates the induction of MyoD, myogenin, myosin heavy chain, myosin light chain and troponin T. This pathway is inhibited in the transfected clones and, thus, requires signaling through the type II TGF-β receptor. The other pathway(s) involves basal desmin expression and other low-serum-induced changes — such as induction of the α chain of the acetylcholine receptor, downregulation of Id, hypophosphorylation of pRB and growth arrest — and is not affected by inhibition of type II receptor signaling. Induction of the latter pathway in the transfected cells is not sufficient to cause full myogenic differentiation. Similarly, v-ras transformed myoblasts induce expression of the acetylcholine receptor and maintain desmin expression but are blocked in expression of other markers including MyoD (Lassar et al., 1989). Thus, one pathway may be independent of MyoD expression, as cells expressing desmin and other specific contractile proteins, but not MyoD, have been described (Kaufman et al., 1991; Peterson et al., 1990; CusellaDe Angelis et al., 1992). Our results also imply that myoblast differentiation is a multistep process. Our transfected cells, which undergo growth arrest but do not differentiate, may allow dissection of the pathways involved in myoblast differentiation.

The seemingly conflicting reports on the inhibitory and stimulatory effects of TGF-β in muscle formation could be explained in several ways. One possibility is that the effect of TGF-β depends on the intrinsic, developmental state of the cell. In this scenario, TGF-β could help maintain MyoD expression in immature myoblasts, thus providing a state of competence for differentiation. Further differentiation would then be prevented by the inhibitory effect of TGF-β until an appropriate signal to differentiate occurred, or until cells have migrated away from areas of TGF-β production. Thus, TGF-β could maintain a proliferating population of committed cells, poised to become myotubes, while also preventing premature myoblast differentiation until sufficient muscle mass exists (Olson et al., 1986; Massagué et al., 1986). Interestingly, the expression pattern of TGF-β during mouse muscle development shows early expression in somites at day 11 (Heine et al., 1987), which is consistent with the above model.

Another possibility is that the two effects of TGF-β, stimulation versus inhibition of myoblast differentiation, are mediated by two different receptors. We have used a truncated type II receptor, which specifically inhibits signaling by the type II, but not the type I, TGF-β receptor in a dominant negative fashion (Chen et al., 1993). Based on that study in epithelial cells, the type II receptor is believed to be required for the growth inhibition induced by TGF-β? and the type I receptor may be involved in TGF-β induction of the extracellular matrix (Chen et al., 1993). Since collagen (Heino and Massagué, 1990), and c-jun (Bengal et al., 1992) inhibit myogenesis, signaling through the type I receptor could indirectly prevent myoblast differentiation by inducing expression of these inhibitory factors. In contrast, since unphosphorylated pRB may play a critical role in myogenesis by interacting with MyoD (Gu et al., 1993), signaling through the type II receptor could have a positive effect on skeletal muscle cell differentiation. Thus, the pleiotropic effects of TGF-β on myoblasts might be mediated by at least two pathways involving two TGF-β receptor types. Under normal circumstances, a balance may exist between the negative and positive signaling through the type I and type II receptors. Introduction of a truncated type II receptor into C2C12 cells may upset this balance, such that the negative effect of signaling through the type I receptor predominates. C2C12 cells in culture produce all three TGF-β isoforms (Lafyatis et al., 1991). According to this model, the inability of cells expressing the truncated receptor to differentiate could be due to autocrine stimulation of the type I TGF-β receptor.

The possibility that TGF-β can have mutually distinct effects that result in the synchronization of cells in a specific developmental stage is an intriguing one. Furthermore, the effect of TGF-β on a cell may depend as much on the balance of expression of specific receptors within the cell, as on the extracellular concentration of the ligand.

We thank Dr Charles Ordahl and the members of his lab for valuable reagents and advice, Drs Helen Blau and Andrew Lassar for critical reading of the manuscript, Drs Nadia Rosenthal, Harold Weintraub, Harold Varmus and Woody Wright for plasmids, Dr Dusty Miller for the high titer MyoD retrovirus, and Dr Grace Pavlath for teaching us the in vivo experiments. We are also grateful to Adrian Erlebacher and Dr Ruey-Hwa Chen for helpful suggestions and support. This work was supported by an NIH postdoctoral training grant to E. H. F., by a grant from the NIH to R. D. and by the Markey Charitable Trust to the Programs in Biological Sciences.