Transforming Growth Factor-β Family Ligands Can Function as Antagonists by Competing for Type II Receptor Binding*

Transforming growth factor-β (TGF-β) family ligands are pleiotropic cytokines. Their physiological activities are not determined by a simple coupling of stimulus and response, but depend critically on context, i.e. the interplay of receptors, ligands, and regulators that form the TGF-β signal transduction system of a cell or tissue. How these different components combine to regulate signaling activities remains poorly understood. Here, we describe a ligand-mediated mechanism of signaling regulation. Based on the observation that the type II TGF-β family receptors ActRIIA, ActRIIB, and BMPRII interact with a large group of overlapping ligands at overlapping epitopes, we hypothesized high affinity ligands compete with low affinity ligands for receptor binding and signaling. We show activin A and other high affinity ligands directly inhibited signaling by the low affinity ligands BMP-2, BMP-7, and BMP-9. We demonstrate activin A functions as a competitive inhibitor that blocks the ligand binding epitope on type II receptors. We propose binding competition and signaling antagonism are integral functions of the TGF-β signal transduction system. These functions could help explain how activin A modulates physiological signaling during extraordinary cellular responses, such as injury and wound healing, and how activin A could elicit disease phenotypes such as cancer-related muscle wasting and fibrosis.

Transforming growth factor-␤ (TGF-␤) family signaling pathways play fundamentally important roles in stem cell fate determination, embryonic development, organogenesis, immunity, and cancer (1)(2)(3). The basic principles underlying TGF-␤ family action are well established. A dimeric ligand binds two type I and two type II receptors to form a hexameric complex, thus initiating a signaling cascade that leads to phosphorylation of SMAD transcription factors, their translocation to the nucleus, and expression of target genes (4 -9). Although this simple mechanism completely describes the molecular basis of signaling and response, it fails to explain the complex and sometimes opposite responses elicited by many TGF-␤ family ligands. For example, some TGF-␤ family ligands can both inhibit and promote cell growth, maintain pluripotency, and induce differentiation, and both suppress and activate tumor cells. These paradoxical effects have supported the idea that cellular responses to a TGF-␤ family ligand depend not only on the ligand-induced signaling cascade but also on the cellular context, i.e. the molecular interplay of all the components that form the TGF-␤ signal transduction system of a particular cell type or tissue (10 -14).
In humans, the TGF-␤ family consists of 33 ligand genes (TGF-␤s, activins, bone morphogenetic proteins (BMPs), 2 growth and differentiation factors (GDFs, nodal and lefty), seven type I receptors, (ALK1-7), five type II receptors (ActRIIA, ActRIIB, BMPRII, TGF␤RII, and AMHRII), as well as a number of co-receptors, regulators, and intracellular SMAD transcription factors (3,15). A distinct feature of the family is the promiscuity of its members. Ligands can bind several different receptors, and receptors can bind multiple ligands. Yet ligand-receptor binding affinities vary greatly. Activin A, activin B, GDF-8, GDF-11, and BMP-10 bind the type II receptors ActRIIA and ActRIIB with very high affinity (16 -18). By contrast, BMP-2 and BMP-4 bind ActRIIA and ActRIIB with low affinity, but they bind type I receptors with high affinity (19,20). These observations have supported a model of sequential signaling complex assembly where activins, GDF-8 and GDF-11, first bind type II receptors with high affinity and then recruit low affinity type I receptors (5,21). By contrast, BMPs and GDFs first bind type I receptors with high affinity and then recruit low affinity type II receptors (22). Exceptions include BMP-10, which binds both type I and type II receptors with high affinity (9,(23)(24)(25).
Significantly, high and low affinity ligands bind the same type II receptors at the same epitope (26,27). This raises the following question. What happens to low affinity BMP or GDF signaling when high affinity ligands like activin A, GDF-11, or BMP-10 are present at the same time? Thus far it has been suggested that low affinity BMP and GDF signaling is independent of high affinity ligands, because they uniquely utilize BMP-RII for signaling (4,7,20,27). But recent studies found Nodal, activin A, activin B, and BMP-10 bind BMPRII with much higher affinity than most BMPs and GDFs (9,18,28,29), indicating low affinity ligands do not have a dedicated type II receptor. Instead, low affinity ligands use the same type II receptors as high affinity ligands. We therefore hypothesized high affinity ligands compete with low affinity ligands for type II receptor binding and antagonize low affinity ligand signaling. In this model, high affinity ligands can function both as signal carriers and as signaling regulators that mediate the biological activities of ligands that bind type II receptors with lower affinities.
To test this hypothesis, we examined ligand-type II receptor binding and ligand signaling. Activins and activin-related ligands like GDF-8 and GDF-11 generally bound type II receptors with higher affinity than most BMPs and signaled via the SMAD2/3 pathway. By contrast, BMPs generally bound type II receptors with lower affinity and signaled via the SMAD1/5/8 pathway, as expected. Notably, high affinity ligands directly inhibited SMAD1/5/8 signaling by low affinity ligands, although they activated their canonical SMAD2/3 pathways. Cross-inhibition was not restricted to low affinity ligands. High affinity ligands also inhibited other high affinity ligands. Significantly, cross-inhibition could be prevented by blocking the activin A-type II receptor interaction but not by inhibiting the intracellular signal transduction pathway. These findings thus suggest cross-inhibition is due to competition for type II receptor binding. That ligands can act as antagonists has been suggested for BMP-3 (30 -32), activin A (33), , and inhibin (35)(36)(37). We propose ligand antagonism and signal transduction pathway switching is a general mechanism of TGF-␤ family regulation and an essential program during extraordinary cellular responses, such as wound healing (38). Cross-inhibition may also help explain how increased TGF-␤ family ligand expression can lead to pathophysiological responses, such as cancer cachexia (39,40).
Receptor-Fc Purification-Proteins were expressed using Chinese hamster ovary (CHO) cells. Secreted receptor-Fc fusion proteins were captured from condition medium using protein A affinity chromatography, eluted with 100 mM glycine, pH 3.0, and directly neutralized by adding 10% v/v 2 M Tris/ HCl, pH 9.0. Purified proteins were either dialyzed directly into phosphate-buffered saline (AMRESCO), pH 7.5, and stored at Ϫ80°C or further purified by size exclusion chromatography in phosphate-buffered saline, pH 7.5, and stored at Ϫ80°C. Purity of receptor-Fc fusion proteins was determined by SDS-PAGE under reducing and non-reducing conditions. Surface Plasmon Resonance-All experiments were performed using BIAcore 2000 and carried out at 25°C using HBS/ EPS (0.01 M HEPES, 0.5 M NaCl, 3 mM EDTA, 0.005% (v/v) Tween 20, pH 7.4) containing 0.1% BSA as running buffer. Experimental flow rate was 50 l/min. For receptor capture, anti-human IgG (Fc) antibody was immobilized on four channels of a CM5 chip using amine-coupling chemistry. 250 response units (RU) of purified ActRIIA-Fc, ActRIIB-Fc, BMP-RII-Fc, or TGF␤RII-Fc were loaded on the experimental flow channels. A reference channel was monitored to account for nonspecific binding, drift, and bulk shifts. For ligand binding studies, different ligands at a concentration of 80 nM were injected, including activin A, activin B, GDF-1, GDF-8, GDF-11, TGF-␤1, TGF-␤2, TGF-␤3, BMP-2, BMP-3, BMP-4, BMP-6, BMP-7, BMP-9, BMP-10, and nodal. After each binding cycle, the antibody surface was regenerated to baseline by injecting MgCl 2 . For binding competition studies, anti-BMP-7 monoclonal antibody (MAB3542, R&D Systems) was immobilized on two channels of a CM5 chip using amine-coupling chemistry. 300 RU of purified BMP-7 were captured on the experimental flow channel. ActRIIA-Fc (12 nM) preincubated with different amounts of activin A was injected. After each binding cycle, the antibody-BMP-7 surface was regenerated to base line by injecting MgCl 2 .
Cell Lines-A-204 rhabdomyosarcoma cells (HTB-82) and HepG2 hepatocellular carcinoma cells (HB-8065) were obtained from the ATCC (American Type Culture Collection). Cells were maintained according to standard ATCC culture conditions. Briefly, A-204 cells were grown in McCoy's 5A medium supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptavidin. HepG2 cells were grown in Eagle's minimum essential medium supplemented with 10% FBS and 1% penicillin/streptavidin. Cells were grown at 37°C under a humidified 5% CO 2 atmosphere. Freshly thawed cells were passaged at least three times before performing assays. A 40-m cell strainer (Greiner Bio-one) was used to obtain a uniform single-cell suspension of HepG2 cells for plating.
Reporter Assays-ϳ50,000 A-204 or ϳ10,000 HepG2 cells in complete medium were seeded in each well of a 96-well plate and grown overnight. For transfection, solutions containing 24 l of Lipofectamine 2000, assay medium (960 l growth medium supplemented with 0.1% BSA), 192 ng of pGL4.74 (Luc2P/hRluc/TK) vector (control luciferase reporter plasmid, Promega), and 19.2 g of the SMAD3 responsive reporter plasmid pGL4.48 (luc2P/SBE) or the SMAD1/5/8 responsive reporter plasmid pGL3 (luc2P/BRE) were prepared and incubated at room temperature for 30 min. After incubation, 3840 l of assay medium was added to the transfection solutions, and 50 l of this mixture were added to each well. Transfection medium was removed the following day, and medium was replaced with assay medium, which contained test proteins, including ligands (1-10 nM) and/or the receptor-Fc constructs (0 -250 nM). Assay medium containing test proteins was incubated at 37°C for 1 h before adding to cells. After addition of medium, A-204 cells were incubated for 6 h, and HepG2 cells were incubated for 16 h at 37°C. Luciferase activity was detected using a homemade dual-glow luciferase assay (41). Luminescence was determined using a FLUOstar Omega plate reader. Relative luciferase units were calculated by dividing firefly luciferase units with Renilla luciferase units (RLU). Data are expressed as mean of four independent measurements. Error bars correspond to S.E. of four independent measurements.
Statistics-Reporter gene assays were performed in quadruplicates and were repeated two to three different times. Statistical analysis was done with GraphPad Prism 6. To determine statistical significance for induction experiments (Fig. 2), oneway analysis of variance and Dunnett's post hoc test were used.

Type II Receptors Bind Multiple Ligands with High Affinity-
Most type II TGF-␤ receptors bind multiple ligands, and many ligands bind multiple type II receptors. Because of this promiscuity, ligand-receptor interactions in the TGF-␤ family remain poorly defined. To characterize the ligand-binding specificity of the four major type II TGF-␤ receptors, ActRIIA, ActRIIB, BMPRII, and TGF␤RII, we used a high throughput SPR-based binding assay (Fig. 1). We captured purified receptor-Fc fusion proteins on a BIAcore sensor chip that was cross-linked with an anti-human Fc antibody and injected 16 different ligands at concentrations that exceeded physiological levels (80 nM) (42)(43)(44). As anticipated, ActRIIA-Fc and ActRIIB-Fc had similar ligand binding profiles (Fig. 1). Both receptors bound activin A and activin B with very high affinity as seen in the very fast association and very slow dissociation rates (Fig. 1, A and B, red and cyan curves). Other ligands that bound both ActRIIA-Fc and ActRIIB-Fc with high affinity included GDF-8, GDF-11, BMP-6, and BMP-10 ( Fig. 1, A and B, green, blue, magenta, and maroon curves). However, either their association rates were slower or their dissociation rates were faster than those seen for activin A or B, resulting overall in a weaker interaction. Important specificity differences between these two highly homologous receptors were seen with BMP-7 and BMP-9, which bound ActRIIA-Fc and ActRIIB-Fc, respectively, with relatively high affinity. TGF␤RII and BMPRII had very distinctive binding profiles. TGF␤RII only bound TGF-␤1 and TGF-␤3 with very high affinity but not TGF-␤2 (Fig. 1D, light and dark gray curves) (45,46). The BMPRII ligand profile was unexpected, as we found it also has high affinity ligands, namely nodal, activin B, and to a lesser degree, BMP-10 ( Fig. 1C, olive green, red, and maroon curves). BMP-9, activin A, and BMP-6 also associate with BMPRII, but these complexes are much weaker. Notably, some ligands, including GDF-1, BMP-2, BMP-3, and TGF-␤2, did not bind any one of these type II receptors with high affinity. We estimated K d values based on single injections (supplemental Tables S1-S4). These are comparable with published K d values determined from multiple injections (9,16,45,46).

Competition and Antagonism in TGF-␤ Family Signaling
Responses to activin A, activin B, GDF-8, and GDF-11 diverged in HepG2 cells. The SMAD2/3-mediated activin B response remained constant in both A-204 and HepG2 cells (8-fold over background at 1 nM activin B). In contrast to A-204 cells, the activin A response in HepG2 cells was smaller at 1 nM concentration but could be induced strongly at 10 nM concentration (Fig. 2, B and C). Interestingly, GDF-8 signaling was reduced relative to A-204 cells, whereas GDF-11 signaling increased significantly (from 16-fold over background in A-204 cells to ϳ40-fold over background in HepG2 cells) (Fig. 2, B and  C). Notably, 1 nM activin B induced a strong SMAD1/5/8-mediated response in HepG2 cells, which exceeded the canonical SMAD2/3 response (Fig. 2C, 18-fold compared with 8-fold over background).
To explain the differences in signaling, we analyzed expression of TGF-␤ family components in both cell lines ( Fig. 2A). Levels of ActRIIB and TGF␤RII are low in A-204 cells. By contrast, HepG2 cells express type II receptors at much higher levels. Notably, TGF␤RII expression is so high in HepG2 cells that they rank in the top 8 percentile of 470 cell lines when comparing TGF␤RII levels (dataset id gse57083). We suggest that differences in ActRIIB and TGF␤RII expression could account for significant differences in signaling. But HepG2 cells also express high levels of ␤-glycan, which helps increase cellular sensitivity to BMPs (53) and enhance responsiveness to TGF-␤s (54). We speculate ␤-glycan could contribute significantly to the divergent signaling activities.
Activin A Antagonizes BMP-2 and BMP-7 Signaling-As BMPs, GDFs, and activins utilize the same type II receptors for signaling and bind these receptors at the same site (27), we hypothesized high affinity ligands like activin A could directly inhibit low affinity ligand signaling. To test this hypothesis, we first determined by titration the limits of activin A signaling in HepG2 cells (Fig. 3A). The greatest increase in activin A signaling occurred between 0 and 5 nM concentrations. Beyond that, signaling reached a plateau. Even a 10-fold higher concentration of activin A (50 nM) did not increase the SMAD2/3 luciferase signal further (Fig. 3A). To evaluate cross-inhibition, we incubated activin A with BMP-2 or BMP-7. Activin A strongly inhibited BMP-2-and BMP-7-mediated SMAD1/5/8 signaling (Fig. 3, B and C). At a 5-fold excess, activin A reduced the SMAD1/5/8 signal ϳ7-fold for BMP-2 and 10-fold for BMP-7. Even at equimolar concentrations, activin A attenuated the BMP-2-and BMP-7-mediated SMAD1/5/8 responses 4-fold. Notably, activin A also activated its canonical SMAD2/3 signaling pathway. The SMAD2/3 signal was the same, whether BMP-2 or BMP-7 was present or not (Fig. 3, A-C). However, a small and statistically significant residual SMAD1/5/8 signaling activity remained for both BMP-2 and BMP-7 at the highest activin A concentrations tested (Fig. 3, B and C).
We performed a similar inhibition experiment with TGF-␤2 or TGF-␤3 (Fig. 3D). Activin A did not inhibit the SMAD2/3mediated TGF-␤2 and TGF-␤3 signals, as expected. Instead, the activities of activin A and TGF-␤2 or TGF-␤3 were additive (Fig. 3D). We also examined whether activin A could inhibit ligands that bind the same type II receptors with high affinity, including BMP-9 and BMP-10 (Fig. 3, E and F). As we show for BMP-2 and BMP-7, activin A antagonized signaling by these two ligands, reducing their peak response ϳ2.5-fold for BMP-10 and 25-fold for BMP-9 at equimolar concentrations. Higher concentrations of activin A further reduced BMP-9 and BMP-10 signaling, but a small SMAD1/5/8 signal remained.
ActRIIA-Fc and Follistatin, but Not SB-431542, Inhibit Activin A Antagonism-To demonstrate that cross-inhibition is due to ligand-receptor binding competition, we examined how the extracellular inhibitors ActRIIA-Fc, follistatin, and the intracellular kinase inhibitor SB-431542 affected activin A antagonism (Figs. 5-7). We hypothesized ActRIIA-Fc and follistatin would block activin A antagonism, as both prevent binding of activin A to type II receptors and preclude activin A from competing with BMP-2 or BMP-7 (55). By contrast, SB-431542 would not prevent activin A antagonism (56,57), as activin A would still bind type II receptors and compete with BMP-2 and BMP-7 for type II receptor binding. To test this model, we examined BMP-7-mediated reporter gene expression in A-204 and HepG2 cells (Fig. 5, A and B), BMP-2-mediated reporter gene expression in HepG2 cells (Fig. 6A), and BMP-2-mediated SMAD phosphorylation in HepG2 cells (Fig.  6B).
ActRIIA-Fc inhibited SMAD2/3-mediated activin A signaling but not SMAD1/5/8-mediated BMP-2 or BMP-7 signaling (Figs. 5, A and B, and 6A). Significantly, activin A strongly inhibited BMP-7 and BMP-2 signaling when alone but not when ActRIIA-Fc was also present (Figs. 5, A and B, and 6A). Likewise, follistatin inhibited SMAD2/3-mediated activin A signaling and rescued BMP-7 signaling in the presence of activin A (Fig. 7A). However, follistatin also inhibited BMP-7 signaling at high concentrations (58,59). In sharp contrast to ActRIIA-Fc and follistatin, SB-431542 did not prevent activin A antagonism. Activin A reduced BMP-7 and BMP-2 signaling equally whether SB-431542 was added or not (Figs. 5, A and B, and 6A). Importantly, these findings were paralleled by SMAD phosphorylation (Fig. 6B), indicating that the response measured by reporter gene expression is directly linked to SMAD-mediated signal transduction pathways. Thus, our findings show inhibition of BMP-7 and BMP-2 signaling by activin A can be rescued by trapping activin A with ActRIIA-Fc or follistatin and pre- venting activin A-receptor binding. However, suppressing signal transduction pathway activation with intracellular kinase inhibitors does not block the activin A antagonist activity.
Activin A Blocks BMP-7-ActRIIA Binding-To demonstrate directly that activin A inhibits binding of BMP-7 to type II receptors, we developed a competition SPR binding assay (Fig.  7B). We captured BMP-7 on the sensor chip using a crosslinked BMP-7-specific monoclonal antibody, and we flowed ActRIIA-Fc preincubated with different concentrations of activin A. ActRIIA-Fc (12 nM) bound captured BMP-7 with an association rate that was nearly the same as that determined using our standard setup (supplemental Table S5). ActRIIA-Fc binding to BMP-7 decreased with increasing activin A concentrations, and a 4-fold molar excess of activin A (48 nM) completely inhibited the ActRIIA-Fc-BMP-7 interaction. Thus, activin A can function as a competitive inhibitor of BMP-7 binding to ActRIIA.
BMP-10 Is a Weak Activin A Antagonist-As the antagonist activity of TGF-␤ family ligands was not limited to activin A (Fig. 3), we hypothesized some ligands may antagonize activin A signaling. Previous studies have indicated BMP-3 inhibits activin A signaling (30 -32), and we speculated BMP-10 could potentially antagonize activin A as it is a high affinity ligand of the three type II activin A receptors (Fig. 1). To test this hypothesis, we examined activin A-mediated reporter gene expression in HepG2 cells and antagonism by BMP-3 and BMP-10 (Fig.  6C). Notably, 50 nM BMP-3 did not inhibit signaling by 5 nM activin A, even after we tested three different BMP-3 samples. By contrast, a 10-fold excess of BMP-10 reduced the activin A-mediated SMAD2/3 signal about 2-fold (from 37.4 to 21.7 RLU), while activin A attenuated the BMP-10-mediated SMAD1/5/8 response as expected (from 45.3 to 16.6 RLU). Although BMP-10 antagonism was not as potent as activin A antagonism, this finding shows activin A signaling could also be inhibited by some TGF-␤ family ligands.

Discussion
Our goal was to test the hypothesis that TGF-␤ family ligands can function as signaling antagonists by competing for type II receptor binding. We showed several TGF-␤ family ligands HepG2 were transfected with control plasmid and SMAD2/3-or SMAD1/5/8-responsive reporter plasmids. 1 nM BMP-2 or BMP-7 was added to induce signaling, and 10 nM GDF-8, GDF-11, activin B, BMP-9, or BMP-10 was added for inhibition. Dark gray bars correspond to SMAD2/3-mediated reporter gene expression, and light gray bars correspond to SMAD1/5/8-mediated reporter gene expression. RLU were calculated as the ratio of luminescence from the experimental reporter to luminescence from the control reporter. Data are expressed as mean Ϯ S.E. of four independent measurements. Horizontal light gray lines indicate statistically significant differences in SMAD1/5/8 signaling, and horizontal dark gray lines indicate statistically significant differences in SMAD2/3 signaling (p Յ 0.05) between relevant experimental pairs. directly inhibited signaling by other members of the family. The antagonist activity could be suppressed by blocking the type II receptor-binding site on antagonistic ligands, but not by inhibiting their signal transduction cascade. Moreover, we demonstrated ligand antagonism is mediated by direct competition for receptor binding. We propose ligand antagonism is a common mechanism that can be used to modulate TGF-␤ family signaling. We speculate the precise combination of ligands available to a cell or tissue will profoundly affect how they read a particular TGF-␤ family signal (60).
Ligand Competition Regulates Signaling-An enduring conundrum in the TGF-␤ field is the promiscuity of its ligandreceptor pairings. This promiscuity is particularly relevant to the "activin receptors" ActRIIA and ActRIIB and the "BMP receptor" BMPRII. They interact with a large group of overlapping ligands, including activins, BMPs, GDFs, and nodal (Fig.  1). As these ligands bind receptors at a common site (Fig. 8A), we hypothesized ligands can compete for receptor binding, and high affinity ligands can suppress low affinity ligand signaling ( Fig. 8B). We show the high affinity ligand activin A effectively inhibited signaling by the low affinity ligand BMP-2, the medium affinity ligand BMP-7, and the high affinity ligands BMP-9 and BMP-10 ( Figs. 3 and 5-7). The ability to inhibit signaling was not specific to activin A. GDF-8, GDF-11, BMP-9, BMP-10, and activin B also inhibited BMP-2 and BMP-7 signaling. However, activin A was the most powerful BMP-2 and BMP-7 antagonist. Notably, BMP-10 antagonized activin A signaling with some efficacy (Fig. 6C). The pervasiveness of competition and antagonism between ligands suggests they are a key feature of the TGF-␤ family. However, they depend on shared receptor utilization, as TGF-␤1 cannot inhibit BMP-6 or BMP-9 signaling (33). We speculate competition and antagonism could play an important role in the biological activities of ligands that are involved in acute biological responses, such as injury, infection, and inflammation (61)(62)(63). Thus, by acting as antagonist, activin A secreted in wound healing could shift the signaling equilibrium from a state that supports tissue homeostasis to a state that leads to regeneration (64 -66). Importantly, FIGURE 5. Extracellular activin A signaling inhibitor ActRIIA-Fc rescues BMP-7 activity. Luciferase reporter gene expression was induced with BMP-7 in A-204 (A) and HepG2 (B) cells transfected with SMAD2/3-or SMAD1/5/8-responsive reporter plasmids. 1 nM BMP-7 was added to induce signaling; 10 nM activin A was added to antagonize BMP-7 signaling. 250 nM ActRIIA-Fc was used to block binding of activin A to type II receptors; 5 M SB-431542 was added to block activin A-mediated kinase activation and SMAD2/3 phosphorylation. Dark gray bars correspond to SMAD2/3-mediated reporter gene expression, and light gray bars correspond to SMAD1/5/8-mediated reporter gene expression. Data are expressed as mean Ϯ S.E. of four independent measurements. Horizontal light gray lines indicate statistically significant differences in SMAD1/5/8 signaling, and horizontal dark gray lines indicate statistically significant differences in SMAD2/3 signaling (p Յ 0.05) between relevant experimental pairs. although activin A suppressed BMP-2-and BMP-7-mediated SMAD1/5/8 signaling, it also activated its canonical SMAD2/3 signaling pathway, causing a switch in signal transduction pathway utilization.
Molecular Basis of Ligand Competition-ActRIIA-Fc, follistatin, and SB-431542 are potent activin A signaling inhibitors with distinct modes of action. ActRIIA-Fc binds activin A and shields its type II receptor-binding site (26). Follistatin binds activin A and blocks both type I and type II receptor-binding sites (67). SB-431542 inhibits type I receptor kinase activity (57). Activin A trapped by ActRIIA-Fc or follistatin cannot bind type II receptors. By contrast, activin A can bind type II receptors when SB-431542 is present. We show ActRIIA-Fc and follistatin rescued BMP-2 and/or BMP-7 signaling in the presence of activin A, but SB-431542 did not (Figs. 5-7). The antagonist function of activin A thus can be suppressed by blocking activin A binding to type II receptors but not by inhibiting its signal transduction cascade. As activin A, BMP-2, BMP-7, and BMP-9 share type II receptors and bind receptors at the same site, we speculated it is a competitive BMP inhibitor that blocks the weak BMP-type II receptor interaction (Figs. 1 and 8). Indeed, activin A prevented BMP-7 binding to ActRIIA in vitro, providing direct evidence for a competitive inhibitor function (Fig.  7B). Notably, binding competition and signaling antagonism are not new concepts in the TGF-␤ family. Functional antagonism was described for the activin A/BMP-7 pairing in human embryonic carcinoma cells (30) and for the GDF-8/BMP-7 pairing in C3H10T1/2 cells (68). Inhibin A, an activin A-related ligand that lacks signaling activity, inhibited BMP responses by competing for type II receptor binding (35)(36)(37). Binding competition was also proposed for some BMPs together with type II receptors (69), for activin A combined with type I receptors (70), and for activin A with BMP-9 (33). We propose binding competition and signaling antagonism is a common regulatory FIGURE 6. Role of activin A in BMP-2 signal transduction and activin A inhibition with BMP-10. A, luciferase reporter gene expression was induced with BMP-2 in HepG2 cells transfected with SMAD2/3-or SMAD1/5/8-responsive reporter plasmids. 1 nM BMP-2 was added to induce signaling, and 10 nM activin A was added to antagonize BMP-2 signaling. 250 nM ActRIIA-Fc was used to block binding of activin A to type II receptors; 5 M SB-431542 was added to block activin A-mediated kinase activation and SMAD2/3 phosphorylation. Dark gray bars correspond to SMAD2/3-mediated reporter gene expression, and light gray bars correspond to SMAD1/5/8-mediated reporter gene expression. B, Western blots showing BMP-2-and activin A-mediated SMAD phosphorylation in HepG2 cells. 1 nM BMP-2 was added to induce SMAD1/5/8 phosphorylation; 10 nM activin A was added to antagonize BMP-2-mediated SMAD1/5/8 phosphorylation; 250 nM ActRIIA-Fc was used to block binding of activin A to type II receptors; and 5 M SB-431542 was added to block activin A-mediated kinase activation and SMAD2/3 phosphorylation. C, HepG2 cells were transfected with SMAD2/3-or SMAD1/5/8-responsive reporter plasmids. 5 nM activin A was added to induce signaling; 50 nM BMP-10 or BMP-3 was added to inhibit activin A-mediated SMAD2/3 reporter gene expression. Dark gray bars correspond to SMAD2/3-mediated reporter gene expression. Light gray bars correspond to SMAD1/5/8-mediated reporter gene expression. Data are expressed as mean Ϯ S.E. of four independent measurements. Horizontal light gray lines indicate statistically significant differences in SMAD1/5/8 signaling, and horizontal dark gray lines indicate statistically significant differences in SMAD2/3 signaling (p Յ 0.05) between relevant experimental pairs. mechanism in the TGF-␤ family enabled by receptor promiscuity and binding site conservation (19).
Role for Ligand Competition in Human Diseases-TGF-␤ family ligands play critical roles in many human diseases, including cancer, fibrosis, bone loss, and muscle loss (10). It is suggested that elevated ligand expression leads to hyper-activation of their canonical signaling pathways and, consequently, to disease onset and progression (71). However, the ability of a TGF-␤ family ligand to hyper-activate its signaling pathways can be limited; TGF-␤ family ligands can have a "signaling ceiling" (Fig. 3). Importantly, we found "high affinity" ligands like activin A attenuated or antagonized signaling pathways that were activated by "lower affinity" ligands like BMP-2 and BMP-7 (Figs. 3 and 5-7). Thus, although hyper-activated signaling could occur in the appropriate context and may be the root cause of a number of diseases, we speculate binding competition and signaling antagonism between ligands could also play a role in the pathobiology of activin A and other TGF-␤ family ligands. Significantly, a recent study indicated activin A-mediated progression of fibrodysplasia ossificans progressiva by competing for type I receptor binding (70).
Differential Effect of Inhibitors on Antagonism-Inhibiting activin A signaling has considerable therapeutic potential. Several approaches have been explored, including blocking the receptor kinase activity with SB-431542 and preventing formation of activin A-receptor signaling complexes using ActRIIA-Fc (55)(56)(57)72). Both SB-431542 and ActRIIA-Fc inhibited activin A signaling, as determined by SMAD2/3 phosphorylation and SMAD2/3-mediated luciferase expression. But they differed drastically in their ability to prevent activin A antagonism. Only extracellular activin A inhibitors rescued low affinity ligand signaling (Figs. 5-7). As activin A could either hyper-activate its canonical signaling pathway or antagonize low affinity BMP signaling in disease, our findings indicate selection of an appropriate therapeutic approach to inhibit activin A signaling should depend on the mechanism of pathogenesis. Even so, extracellular inhibitors that target ligands and prevent receptor binding offer a more FIGURE 7. Follistatin suppresses activin A antagonism. A, luciferase reporter gene expression was induced with BMP-7 in HepG2 cells transfected with SMAD2/3-or SMAD1/5/8-responsive reporter plasmids. Follistatin inhibited BMP-7 signaling at high concentrations (40 and 100 nM) but less at lower concentrations (10 and 20 nM). Activin A (10 nM) effectively inhibited BMP-7 signaling. Follistatin (10 -20 nM) inhibited activin A signaling and prevented antagonism of BMP-7. Dark gray bars correspond to SMAD2/3-mediated reporter gene expression, and light gray bars correspond to SMAD1/5/8-mediated reporter gene expression. Data are expressed as mean Ϯ S.E. of three independent measurements. Horizontal light gray lines indicate statistically significant differences in SMAD1/5/8 signaling, and horizontal dark gray lines indicate statistically significant differences in SMAD2/3 signaling (p Յ 0.05) between relevant experimental pairs. B, activin A prevents binding of ActRIIA-Fc to BMP-7. BMP-7 was captured on the SPR sensor chip using an anti-BMP-7 monoclonal antibody (top right panel). 12 nM ActRIIA-Fc preincubated with 0 nM (blue), 3 nM (red), 6 nM (magenta), 12 nM (green), 24 nM (maroon), and 48 nM (gray) activin A was injected over the sensor chip. Activin A prevented binding of ActRIIA-Fc to BMP-7 in a concentration-dependent manner. The bottom right panel shows a dose-response curve. RU values were taken at 290 s after starting injection of ActRIIA-Fc. MAY 13, 2016 • VOLUME 291 • NUMBER 20 defined approach for rescuing the effects of excess ligand in disease.

Competition and Antagonism in TGF-␤ Family Signaling
Conclusions-We show high affinity TGF-␤ family ligands like activin A compete with low affinity ligands for receptor binding and antagonize low affinity ligand signaling. Several diseases associate with activin A overexpression, including inflammation, fibrosis, and cancer-related muscle wasting (39,40,73). How elevated activin A levels lead to disease is not well understood. Hyper-activated signaling could lead to pathogenesis. However, our findings indicate a second possibility, namely that ectopic activin A could antagonize normal signaling and thus promote disease. Significantly, a recent study indicated activin A antagonism may be critical for progression of fibrodysplasia ossificans progressiva (70).