Stimulation of myogenic differentiation by a neuregulin, glial growth factor 2. Are neuregulins the long-sought muscle trophic factors secreted by nerves?

It has long been known that nerves stimulate growth and maintenance of skeletal muscles in ways not dependent on physical contacts, but numerous attempts to identify and characterize the myotrophic agent(s) secreted by nerves have been unsuccessful. We here suggest that products of the neuregulin gene may be these agents. The neuregulins are a family of proteins made by alternative splicing of a single transcript to give as many as 15 protein products. One member of this family, glial growth factor 2 (rhGGF2) is a very potent stimulator of myogenesis in L6A1 myoblasts, giving a maximal stimulation of cell fusion and creatine kinase elevation at a concentration of 1 ng/ml (18 pM). The stimulation of myogenesis is not rapid, but it is prolonged, continuing over a period of at least 6 days. The effects of rhGGF2 are additive with those of insulin-like growth factor I (IGF-I) or its analog R3-IGF-I, suggesting that the actions of these two myotrophic agents differ in at least one rate-limiting step. We have observed one possible difference; unlike the IGFs, rhGGF2 does not induce elevation of the steady state level of myogenin mRNA.

It has long been known that nerves stimulate growth and maintenance of skeletal muscles in ways not dependent on physical contacts, but numerous attempts to identify and characterize the myotrophic agent(s) secreted by nerves have been unsuccessful. We here suggest that products of the neuregulin gene may be these agents. The neuregulins are a family of proteins made by alternative splicing of a single transcript to give as many as 15 protein products. One member of this family, glial growth factor 2 (rhGGF2) is a very potent stimulator of myogenesis in L6A1 myoblasts, giving a maximal stimulation of cell fusion and creatine kinase elevation at a concentration of 1 ng/ml (18 pM). The stimulation of myogenesis is not rapid, but it is prolonged, continuing over a period of at least 6 days. The effects of rhGGF2 are additive with those of insulin-like growth factor I (IGF-I) or its analog R3-IGF-I, suggesting that the actions of these two myotrophic agents differ in at least one rate-limiting step. We have observed one possible difference; unlike the IGFs, rhGGF2 does not induce elevation of the steady state level of myogenin mRNA.
The myotrophic actions of nerves have been known for a long time, and it has been widely recognized that nerves stimulate muscle formation and maintenance in ways not entirely dependent on direct physical contact. A review by Gutman (1) 20 years ago discusses a large body of early evidence for the existence of "long term maintenance regulations not mediated by nerve impulses" and states that "motor, sensory, or central neurons can supply the . . . agent." Subsequent searches for such myotrophic agents have not been successful (at least twice transferrin was isolated when a "muscle trophic factor" or "sciatin" was sought (2, 3)), but we have recently uncovered a strong candidate for the role of the myotrophic agent from nerve cells, glial growth factor 2 (GGF2 1 ) and presumably its fellow members of the neuregulin family of neurotrophic proteins.
GGF2 is one of a recently characterized family of closely related products of a single gene, made by alternative splicing to give as many as 15 distinct protein products (4 -7). The members of this family have a number of different names, including heregulin, neu differentiation factor, acetylcholine receptor inducing activity, and glial growth factors. Most contain transmembrane and cytoplasmic domains; they are primarily glycosylated, transported to the cell surface, and reside as transmembrane proteins (although there is some hydrolytic processing of a portion within the cell) (8). This group also found that cell-associated molecules can be released by regulated proteolytic cleavage; for example, phorbol esters cause this release. A new member, sensory and motor neuron-derived factor, has recently been added to the neuregulin family (9); like GGF2, it lacks transmembrane and cytoplasmic regions and is presumably secreted by cells that express it.
Most attention has been focused on the actions of these agents in the nervous system, but they also have major effects on growth and development of the mammary epithelial cells (10,11), and knockout experiments have demonstrated that they are essential for early development of the heart (12). Insofar as we are aware, the only studies of effects of neuregulins on skeletal muscle have been concerned with induction of the acetylcholine receptor (13,14); actions on other aspects of myogenic differentiation have not been reported. Two of the current authors have previously published an abstract describing stimulatory actions of rhGGF2 on a human muscle cell line (15); this prompted the current study, and it demonstrates that the response we see in L6A1 myoblasts is not limited to that cell line.

EXPERIMENTAL PROCEDURES
Materials-Unless otherwise specified, all cell culture materials were purchased from Life Technologies, Inc. rhGGF2 was prepared from stably transfected CHO cells essentially as described by Marchionni et al. (7). IGF-I was a gift from Ciba-Geigy, R3-IGF-I a gift from GroPep, and the myogenin probe from E. N. Olson. All materials used in the creatine kinase assay were purchased from Sigma.
Cell Cultures and Analyses-L6A1 myoblasts were plated in DMEM containing 10% horse serum and 1% chick embryo extract. Standard cultures were plated at 1.2 ϫ 10 5 cells/35-mm dish for creatine kinase or cell number determinations, and 1 ϫ 10 6 cells/100-mm dish for RNA isolations. After incubation overnight, the cultures were washed with DMEM before the addition of rhGGF2 or other agents in DMEM containing 0.05% bovine serum albumin (pretested as free of mitogenic activity). To quantitate cell proliferation, cells were trypsinized and counted in a Model ZBI Coulter Counter 24 to 48 h after the addition of growth factors. For measurement of the extent of cell differentiation, at 2 to 6 days, the cell monolayers were washed with phosphate-buffered saline, 0.25 ml of 0.05 M glycylglycine (pH 6.75) was added, and the dishes were stored at Ϫ70°C. The cultures were thawed on ice just prior to assay, the cells were scraped off, and the cell lysate was assayed for creatine kinase using a NAD-coupled microtiter assay (16), which allows simultaneous measurement of CK activity of as many as 96 samples, and for which all calculations are fully automated. DNA in * This work was supported in part by Grant HL11551 from the National Institutes of Health, United States Public Health Service. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
another aliquot of the same cell suspension was measured fluorometrically as described by Ewton et al. (17). Fig. 1 illustrates the results of an experiment (typical of four similar ones, each including triplicate incubations) in which an extended series of rhGGF2 concentrations was used to stimulate myogenesis (quantitated as the elevation of creatine kinase levels). It demonstrates very clearly that a biphasic curve characterizes the myogenic action of rhGGF2. (It should be understood that fusion to form myotubes closely parallels elevation of creatine kinase levels in all of these experiments.) The optimal concentration was 1 ng/ml; for a molecular weight of 55,000, this represents a molar concentration of approximately 18 pM, the lowest effective concentration of any growth factor having a major effect on muscle cells that has been reported. By contrast, IGF-I (M r ϭ 7500) gives optimal stimulation of myogenesis at about 40 ng/ml, or about 6 nM, a full 333-fold lower potency. To be sure, much of the lower potency of IGF-I can be attributed to secretion of inhibitory binding proteins by myoblasts; analogs with substantially reduced affinity for the IGF binding proteins (such as R3-IGF-I, in which an arginine has been substituted for glutamate as the third amino acid in the chain) exhibit potencies as much as 100 times as great as that of native IGF-I (18), but that still leaves them only one-third as potent as rhGGF2 in stimulating myogenesis.

Concentration Dependence of the Stimulation of L6A1 Myoblast Differentiation by rhGGF2-
It should be noted that rhGGF2 in the serum-free medium used here had little or no effect on myoblast proliferation whether measured as Coulter-counted cells at 24 or 48 h or DNA/dish at 72 h (data not shown). (The flat cell number and DNA concentration dependence curves for both rhGGF2 and DMEM treatments, as well as our earlier experiments with IGFs, show that rhGGF2 does not simply enhance survival of the cells.) In this way, it differs dramatically from IGF-I, which is an active mitogen for skeletal myoblasts in serum-free medium. However, in the presence of serum components, rhGGF2 does exhibit a modest mitogenic activity, with approximately a doubling in cell numbers after a 48-h incubation; under these conditions, R3-IGF-I gave a 5-fold increase in cell numbers (data not shown).
Time Course of CK Elevation by rhGGF2 and IGF-I-In our "standard" 72-h incubations such as that presented in Fig. 1, rhGGF2 caused substantially less elevation of creatine kinase and increase in cell fusion than did IGF-I or its R3 analog. However, when the incubation was extended to longer periods, the amount of fusion observed microscopically and the levels of CK attained were nearly as great in rhGGF2-treated cells as in those incubated with IGF-I at 80 ng/ml, as shown in Fig. 2. (L6 myotubes formed in the presence of IGF-I tend to lift from the culture dish after 4 or 5 days, so extended incubations with IGF-I are not feasible.) It is not clear why the response to rhGGF2 is so much slower than that to IGF-I, but this has been a consistent result in our experiments.
Additivity of Effects of rhGGF2 and R3-IGF-I-The stimulation of differentiation by rhGGF2 adds it to the handful of agents known to stimulate myogenesis (the others are IGFs, thyroid hormones, and retinoic acid, see Ref. 20). Fig. 3 presents the results obtained when we examined the effects of two of these stimulatory agents, IGF-I and rhGGF2, in combination. As shown in the figure, whether IGF or R3-IGF-I was varied in the presence of optimal rhGGF2 (Fig. 3, A and B) or rhGGF2 was varied in the presence of optimal R3-IGF-I (Fig.  3C), the levels of CK attained with the two agents were clearly greater in combination than with either alone. Such observations are usually interpreted to mean that the agents operate by different mechanisms (i.e. different processes become ratelimiting at optimal concentrations of each hormone), but in this case interpretation is slightly clouded by the biphasic nature of the response curves of both hormones. This biphasic effect may be attributable to a requirement for interaction of a single ligand with two receptors (19) or induction of myogenesisinhibiting oncogenes such as Fos and Jun (20).
Differences in Actions of rhGGF2 and IGF-I-We (21) demonstrated some time ago that the induction of differentiation by IGF-I involves increased expression of the myogenin gene, although later results have made us aware that a simple increase in myogenin mRNA does not completely account for the increased myogenesis. We have found that the R3 analog of IGF-I (at 1 ng/ml, a concentration that stimulates differentiation but has little effect on cell proliferation, see Ref. 18), gives a substantial increase in myogenin mRNA, but we found little or no elevation of myogenin mRNA levels in rhGGF2-treated cells (data not shown). Another difference between rhGGF2 and IGF-I was detected when we measured phosphorylation of the insulin receptor substrate 1, which we observed in IGF-Itreated myoblasts but not in cells incubated with rhGGF2 (data not shown). It has recently been shown that knockout of the neuregulin gene is embryo-lethal at days 10.5 to 11.5, when development of the heart begins (12). (Marchionni et al. (7) found that neuregulin mRNA was first detectable in the mouse embryo at embryonic day 10.5 to 11.) This lethality is attributed to failure of the heart to beat, an observation that suggests that at least one product of the neuregulin gene is essential for cardiac development. Although this observation suggests a major role of neuregulin gene products in development of one kind of muscle, its relevance to our suggestion that neuregulin products are essential myotrophic agents is limited by the fact that this embryonic lethality precedes significant skeletal muscle formation. So it is not possible to know from knockout animals whether or not neuregulin gene products are required for skeletal myogenesis. Although myogenin mRNA is detectable as early as day 9.5 (22), the protein is not detectable at day 10.5 (23), apparently because the mRNA is not processed (24); the latter authors cite reports that formation of skeletal muscle is not detectable until day 14. It is well established that myogenin protein is essential for skeletal muscle formation (25,26).
There have been no published reports of rhGGF2 actions on myogenesis or cell division in skeletal muscle cells, but studies with ARIA have given results consonant with our observations. ARIA has been shown to act on muscle to induce expression of the ⑀ subunit of the acetylcholine receptor (13) in mouse myotube cultures, as well as sodium channels (27); this is the best-characterized action of a member of the neuregulin family on skeletal muscle. It has been shown by Sandrock et al. (28) that ARIA is concentrated at neuromuscular junctions, and thus it could be available to function as a general stimulator of muscle cells if released from its membrane-bound form by proteases.
There are obvious candidates for the receptor(s) that mediate the myogenic effects of rhGGF2. Our initial determinations of total tyrosine-phosphorylated proteins in rhGGF2 cells revealed the presence of a phosphorylated band approximately the size of the ErbBs, i.e. about 185 kDa (data not shown). The neu differentiation factors were initially characterized as ligands for the c-ErbB-2 receptor, but more recent results (29) have indicated that ErbB-3 and ErbB-4 (often complexed with ErbB-2) are more direct receptors for the neuregulins. Jo et al. (30) found that there is detectable ErbB-2 and ErbB-3 mRNA in skeletal muscle and in the C2 myoblast cell line, with substantially increased ErbB-3 mRNA in myotubes compared to myoblasts. Beerli et al. (29) have shown very recently that the differences between ErbB-3 and ErbB-4 expression is cell typespecific in various mammary tumor cell lines, with substantial variation in the requirements for ErbB-2 for biological responses.
On the other hand, there are indications that ErbB-4 may play an important role, at least in cardiac muscle. Recent reports that knockouts of ErbB-2 (31) and ErbB-4 (32) gave the same embryonic lethality (i.e. failure to form normal cardiac trabecules and thus begin beating) as did knockout of the neuregulin gene (12) strongly suggest that this aspect of cardiac differentiation results from an interaction of a neuregulin gene product with a heterodimeric receptor composed of ErbB-2 and ErbB-4. No results of knockouts of the ErbB-3 gene have been reported, so it is not yet possible to evaluate its possible role. Because the most likely receptor binding sequence in the neuregulins is identical in all members of the family (i.e. amino acids 177-226 of heregulin ␤, see Ref. 33), it seems likely that all neuregulins share the myotrophic activities we have found for rhGGF2, although it is possible that other parts of the molecule may modulate the relative potencies of the members of the family.
It is not obvious how rhGGF2 could be effective in stimulating myogenesis in the absence of any apparent effect on expression of the myogenin gene, as myogenin knockouts are postpartum lethal because of the failure to develop functioning diaphragm muscle (25,26). Possibilities currently under investigation by us include increased levels of myogenin protein resulting from increased translation of the basal levels of myogenin mRNA found in growing myoblasts and increased halflife of myogenin protein, as well as elevated levels of MEF2, which has been shown by Olson's group (34) to enhance myogenic responses to the MyoD family.
Are we correct in suggesting that the neuregulins are the long-sought myotrophic factors secreted by nerves? As mentioned above, the timing of developmental events makes it difficult if not impossible to evaluate this point with knockout experiments, but there are several points which require that this possibility be considered seriously. They include the following.
1) The neuregulins have been shown to be produced by at least some cells of the nervous system (7,14). 2) Within nerves, at least some neuregulins are localized at or near the neuromuscular junction (28,30).
3) The actions of neuregulins on skeletal muscle cells are associated with growth or differentiation, i.e. they are myotrophic effects. (a) We found that rh-GGF2 causes an increase in myotube formation and elevated CK levels (Fig. 1). (b) Induction of acetylcholine receptors (i.e. ARIA activity) is generally associated with terminal differentiation, and it is often used (as we use CK levels) to quantitate myogenic differentiation. (c) rhGGF2 has at least some mito- FIG. 3. Additive effects of rhGGF2 and two IGFs on L6A1 myoblast differentiation. Conditions used were the same as those specified under Fig. 1. For the constant rhGGF2 additions, the concentration was 1 ng/ml, which gives maximal stimulation of differentiation. For constant IGFs, the concentration used was 31 ng/ml for IGF-I and 1 ng/ml for R3-IGF-I.

Myogenic Actions of Neuregulins 12701
genic activity in myoblasts in the presence of a serum fraction that is not itself mitogenic. 4) Knockouts of the neuregulin gene and two ErbB receptors are embryo-lethal because of failure of the heart to differentiate properly; there are some parallels between myogenesis in cardiac and in skeletal muscle, for example, ␣-cardiac actin is expressed early but transiently in skeletal muscle differentiation (35). We believe that these considerations make a substantial case for identification of products of the neuregulin gene as the long-sought myotrophic agents secreted by nerves.