In the Absence of Type III Receptor, the Transforming Growth Factor (TGF)-β Type II-B Receptor Requires the Type I Receptor to Bind TGF-β2*

Transforming growth factor β (TGF-β) ligands exert their biological effects through type II (TβRII) and type I receptors (TβRI). Unlike TGF-β1 and -β3, TGF-β2 appears to require the co-receptor betaglycan (type III receptor, TβRIII) for high affinity binding and signaling. Recently, the TβRIII null mouse was generated and revealed significant non-overlapping phenotypes with the TGF-β2 null mouse, implying the existence of TβRIII independent mechanisms for TGF-β2 signaling. Because a variant of the type II receptor, the type II-B receptor (TβRII-B), has been suggested to mediate TGF-β2 signaling in the absence of TβRIII, we directly tested the ability of TβRII-B to bind TGF-β2. Here we show that the soluble extracellular domain of the type II-B receptor (sTβRII-B.Fc) bound TGF-β1 and TGF-β3 with high affinity (Kd values = 31.7 ± 22.8 and 74.6 ± 15.8 pm, respectively), but TGF-β2 binding was undetectable at corresponding doses. Similar results were obtained for the soluble type II receptor (sTβRII.Fc). However, sTβRII.Fc or sTβRII-B.Fc in combination with soluble type I receptor (sTβRI.Fc) formed a high affinity complex that bound TGF-β2, and this complex inhibited TGF-β2 in a biological inhibition assay. These results show that TGF-β2 has the potential to signal in the absence of TβRIII when sufficient TGF-β2, TβRI, and TβRII or TβRII-B are present. Our data also support a cooperative model for receptor-ligand interactions, as has been suggested by crystallization studies of TGF-β receptors and ligands. Our cell-free binding assay system will allow for testing of models of receptor-ligand complexes prior to actual solution of crystal structures.

Transforming growth factor ␤ (TGF-␤) ligands exert their biological effects through type II (T␤RII) and type I receptors (T␤RI). Unlike TGF-␤1 and -␤3, TGF-␤2 appears to require the co-receptor betaglycan (type III receptor, T␤RIII) for high affinity binding and signaling. Recently, the T␤RIII null mouse was generated and revealed significant non-overlapping phenotypes with the TGF-␤2 null mouse, implying the existence of T␤RIII independent mechanisms for TGF-␤2 signaling. Because a variant of the type II receptor, the type II-B receptor (T␤RII-B), has been suggested to mediate TGF-␤2 signaling in the absence of T␤RIII, we directly tested the ability of T␤RII-B to bind TGF-␤2. Here we show that the soluble extracellular domain of the type II-B receptor (sT␤RII-B.Fc) bound TGF-␤1 and TGF-␤3 with high affinity (K d values ‫؍‬ 31.7 ؎ 22.8 and 74.6 ؎ 15.8 pM, respectively), but TGF-␤2 binding was undetectable at corresponding doses. Similar results were obtained for the soluble type II receptor (sT␤RII.Fc). However, sT␤RII.Fc or sT␤RII-B.Fc in combination with soluble type I receptor (sT␤RI.Fc) formed a high affinity complex that bound TGF-␤2, and this complex inhibited TGF-␤2 in a biological inhibition assay. These results show that TGF-␤2 has the potential to signal in the absence of T␤RIII when sufficient TGF-␤2, T␤RI, and T␤RII or T␤RII-B are present. Our data also support a cooperative model for receptor-ligand interactions, as has been suggested by crystallization studies of TGF-␤ receptors and ligands. Our cell-free binding assay system will allow for testing of models of receptor-ligand complexes prior to actual solution of crystal structures.
Transforming growth factor-␤ (TGF-␤) 1 represents a large superfamily of dimeric growth factors that include the TGF-␤s, inhibins, activins, Mullerian inhibiting substance, growth and differentiation factors, and bone morphogenetic proteins (BMPs) in mammals (1,2). These cytokines play important roles in an array of processes such as growth, differentiation, and development (3). There are three TGF-␤ isoforms that share a high degree of homology and overlapping biological activities (4). However, distinct expression patterns and unique, isoform-specific phenotypes of the corresponding knockout mice demonstrate significant non-redundancy of TGF-␤ function (5)(6)(7)(8)(9).
TGF-␤s exert their biological effects through three cell surface receptors designated as type I, II, and III (T␤RI, T␤RII, and T␤RIII) (2), all of which have been cloned (10 -13). In addition, the type II-B receptor (T␤RII-B), an alternatively spliced isoform of T␤RII containing an insert of 26 amino acids replacing Val 51 , has also been identified (14 -16). Type I and type II receptors have intracellular serine/threonine kinase domains, whereas the type III receptor has only a short intracellular domain. On binding of TGF-␤ ligands, constitutively active type II receptors recruit and phosphorylate type I receptors; the activated type I receptor kinase then interacts with and phosphorylates downstream signaling molecules, the R-Smads (13,17,18). The exact stoichiometry of the active receptor signaling complex is not known. Type II and type I receptors have been shown to form homodimers in the absence of ligand (19,20), thus raising the possibility that the active receptor signaling complex could be a large multimeric complex consisting of a minimum of two type II receptors, two type I receptors, and the TGF-␤ ligand homodimer. The crystal structures of TGF-␤2 (21,22), and TGF-␤3 (23) ligands have been solved. In addition, the extracellular domain of T␤RII alone (24), and in complex with TGF-␤3 (25), have also been crystallized. These structures yield qualitatively different information than the crystal structure of the BMP type IA receptor (BMPR1A)⅐ BMP-2 complex (26) and the activin type II receptor (ActRII)⅐ BMP-7 complex (27). In aggregate, these studies suggest that the TGF-␤ receptor and ligand interactions may involve a cooperative model of binding in which the extracellular domains of the type II and type I receptors make physical contact, whereas the BMP/ActRII and BMP ligands may utilize an allosteric model of binding in which the extracellular domains of the type II and type I receptors do not interact.
Of the three TGF-␤ isoforms, TGF-␤2 and TGF-␤3 have been less well investigated than TGF-␤1. TGF-␤3 appears to bind receptors and signal in a manner that is similar to TGF-␤1. In * 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.
The contrast, TGF-␤2 has much lower affinity for T␤RII than TGF-␤1 and -␤3. It has been demonstrated that an accessory receptor, T␤RIII, is necessary for efficient binding and crosslinking of TGF-␤2 and subsequent signaling (28). In this model, TGF-␤2 binds to T␤RIII, which then recruits T␤RII and T␤RI resulting in phosphorylation of T␤RI and downstream signaling (28). Interestingly, the recently published phenotype of the T␤RIII null mouse (29) is not completely overlapping with the phenotype of the TGF-␤2-deficient mouse (8), suggesting the existence of alternative methods for TGF-␤2 binding and signaling that do not involve T␤RIII. Cross-linking studies of cell-surface TGF-␤ receptors with 125 I-TGF-␤2 in transfected COS cells have suggested that high affinity TGF-␤2 binding and downstream signaling in these cells can occur via complexes of type I and type II receptors (30). Alternatively, Rotzer et al. (16) has proposed that T␤RII-B binds TGF-␤2 with ensuing signaling in the absence of T␤RIII. Of note, an earlier report showing the inability of unlabeled TGF-␤2 to compete for 125 I-TGF-␤1 binding to T␤RII-B inferred that T␤RII-B resembles T␤RII in its inability to bind TGF-␤2 (14). In general, these studies of TGF-␤ receptor binding to ligand were performed on receptors expressed at the cell surface where binding can only be measured indirectly via cross-linking to radioligand followed by autoradiography. These studies are therefore limited by an inability to directly quantify binding and thus obtain an accurate measurement of receptor-ligand affinities. In addition, the presence of other extracellular cell surface-associated proteins that may act as accessories to binding cannot be ruled out.
To overcome the limitations of cell surface expression studies and to specifically assess the ability of T␤RII-B to bind TGF-␤2, we have developed a cell-free system using soluble TGF-␤ receptors. We demonstrate that in comparison with TGF-␤1 and -␤3, TGF-␤2 bound poorly to soluble TGF-␤ type II (sT␤RII.Fc) or TGF-␤ type II-B receptors (sT␤RII-B.Fc) alone. However, TGF-␤2 did bind sT␤RII.Fc or sT␤RII-B.Fc in complex with the soluble type I receptor (sT␤RI.Fc) in solution, and cell-surface T␤RII or T␤RII-B together with T␤RI could mediate TGF-␤2 signaling in the absence of T␤RIII. Our heteromeric receptor binding assay system provides supporting evidence for a cooperative model of type II and type I receptor interactions with TGF-␤2 ligand, and provides a rapid and straightforward way to measure the binding of receptor and/or ligand mutants that arise from structure-function studies.

EXPERIMENTAL PROCEDURES
cDNA Subcloning-The cDNA encoding the extracellular domain of human T␤RII was amplified by PCR from human T␤RII cDNA (12). The PCR product was digested and ligated in-frame into the restriction sites BamHI (5Ј) and HindIII (3Ј) of the vector pIg-Tail (31) to generate the sT␤RII.Fc mammalian expression construct. The primers used were 5Ј-CCCAAGCTTATGCCGCTGCTGCTACTGCTG-3Ј (forward) and 3Ј-ATATTGTGGTCGTTAGGACTGCGCCTAGGG-5Ј (reverse). The cDNA was sequenced on both strands to confirm the fidelity of the construct.
To generate cDNA for the extracellular domain of human T␤RII-B, the 26-amino acid insert was generated by an overlapping primer strategy using PCR. The N-terminal half of the insert was generated by PCR using the following primers: 5Ј-CCCAAGCTTGCCGCCACCATG-GGTCGGGGGCTGCTCAGG-3Ј (forward) and 3Ј-CTGGGGCAGATGT-TCTGGGCCTCCATTTCCACATCCGACTTCTGAACGTGCGGT-5Ј (reverse). The C-terminal half of the insert and the rest of the extracellular domain was generated by PCR using the following primers: 5Ј-GGGG-GATCCGCGTCAGGATTGCTGGTGTTATA-3Ј (forward) and 3Ј-CTGT-AATAGGACTGCCCACTGAGAACATATATTAATAACGACATGATAG-TC-5Ј (reverse). Both PCR products were purified, mixed together, and a final round of PCR was performed using the following "outside" primers: 5Ј-CCCAAGCTTGCCGCCACCATGGGTCGGGGGCTGCTCAG-G-3Ј (forward) and the 3Ј-CTGTAATAGGACTGCCCACTGAGAACAT-ATATTAATAACGACATGATAGTC-5Ј (reverse). The resultant PCR product was purified, digested, and ligated in-frame into the restriction sites BamHI and HindIII (3Ј) of the vector pIg-Tail to generate the sT␤RII-B.Fc mammalian expression construct. The extracellular domain (ECD) of human T␤RII-B was then subcloned into full-length human T␤RII to generate full-length T␤RII-B. cDNAs were sequenced on both strands to confirm the fidelity of the construct.
Protein A Purification of sT␤RII.Fc and sT␤RII-B.Fc-The human recombinant receptors were purified by one-step Protein A affinity chromatography. Tissue culture medium was filtered through a vacuum-driven 0.22-m, Durapore Membrane Unit (Millipore Corp., Bedford, MA). The pH of the media was adjusted to pH 8.2 by addition of Tris base and the media was applied to HiTrap rProtein A FF columns (Amersham Biosciences) previously equilibrated with phosphate-buffered saline (Invitrogen). After protein loading, the columns were washed with binding buffer (phosphate-buffered saline) to remove nonspecifically bound proteins. Human soluble receptors were eluted with 3 volumes of 100 mM glycine buffer, pH 3.0. The pH of eluted fractions was immediately neutralized by addition of a 1/10 volume of 1 M Tris/ HCl, pH 9.0. Eluted protein was stored at Ϫ20°C. Purity of the protein was determined by 4 -12% SDS-PAGE using pre-cast mini gels (Novex) followed by silver staining (Bio-Rad). Amounts of proteins eluted were quantified by the bovine serum albumin protein assay (Pierce) and confirmed by amino acid analysis (MGH-Protein Core Facility). Yields were ϳ6 g of soluble protein per 10-cm tissue culture plates.
Soluble Receptor Deglycosylation-25 g of purified sT␤RII.Fc and sT␤RII-B.Fc were denatured in 0.5% SDS, 1% ␤-mercaptoethanol at 100°C for 10 min and incubated in 50 nM sodium phosphate supplemented with 1% Nonidet P-40 at 37°C. Denatured proteins were then incubated either in the presence or absence of 5,000 units of N-glycosidase F (New England Biolabs) at 37°C for 2 h and analyzed by SDS-PAGE and silver staining as described above.
Protein Analysis with Western Blot-Recombinant human receptors eluted from the HiTrap Protein A column were separated by 4 -12% gradient SDS-PAGE pre-cast minigels (Novex), then transferred to polyvinylidone difluoride transfer membrane (Schleicher & Shuell). After transfer, the membrane was washed in Tris-buffered saline supplemented with 0.1% Tween 20 (TBST), and blocked overnight in 8% powdered milk in TBST. The membrane was then incubated with goat anti-human T␤RII antibodies (R&D Systems) or goat anti-human Fcspecific IgG (Jackson ImmunoResearch Laboratories, West Grove, PA), followed by donkey anti-goat IgG conjugated to horseradish peroxidase (Santa Cruz Biotechnology). The chemiluminescence immunoassay was performed with Renaissance Western blot chemiluminescence reagent (Amersham Biosciences). Polyclonal antibodies directed against a peptide of human T␤RII-B (QKDEIICPSCNRTAHPLRHI, Peptide Core Facility, MGH, MA) were raised in goats (SIGMA-Gemosys, The Woodlands, TX) and employed to specifically detect human recombinant sT␤RII-B.Fc compared with human recombinant sT␤RII.Fc. Mouse sT␤RI.Fc was purchased from R&D Systems.
Binding Assays on Protein A Plates-Soluble recombinant human receptors were diluted in TBS/casein blocking buffer (BioFX, Owings Mills, MD) and incubated on Protein A-coated 96-well plates (Pierce) overnight. Plates were then washed with wash buffer (BioFX) and blocked 2 h at room temperature with TBS/casein buffer. For competition binding assays, fixed amounts of radioligands (50,000 -100,000 counts) were added to the plates together with increasing amounts (2 pM to 500 nM) of homologous or heterologous non-radioactive ligands.
Binding Assays in Solution-Soluble recombinant human receptors were diluted in TBS/casein blocking buffer (BioFX) and incubated overnight in the presence or absence of ligand. For competition binding assays, fixed amounts of radioligands (50,000 -100,000 counts) were added to the samples together with increasing amounts (2 pM to 500 nM) of homologous or heterologous non-radioactive ligands. Samples were then placed on Protein A-coated 96-well plates (Pierce) for 90 min, washed 3 times with wash buffer (BioFX), and counted using a ␥-counter.
Luciferase Reporter Assay-Mink Lung cells (Mv1Lu) were transiently transfected with the (CAGA) 12 MPL-Luc reporter construct (33) and with a pRL-TK vector (Promega) in a ratio of 10:1 to control for transfection efficiency. Cells were then serum starved for 6 h before treatment with varying amounts of TGF-␤ ligands in the presence or absence of varying amounts of soluble receptor for 16 h. Experiments were performed in triplicate wells. Cells were lysed and luciferase activity was determined with the Dual Reporter Assay (Promega). Relative light units were calculated as ratios of Firefly (reporter) and Renilla (transfection control) values. Alternatively, Mv1Lu cells were used that had been stably transfected with the (CAGA) 12 MPL-Luc reporter construct. In this case, relative light units were corrected for total amount of protein in the lysate as determined by a bovine serum albumin protein assay (Pierce). Rat myoblast L6 cells were transfected with the (CAGA) 12 MPL-Luc reporter construct and with a pRL-TK vector and in addition with empty vector or full-length TGF-␤ receptors constructs using LipofectAMINE 2000. The same protocol as for the Mv1Lu cells was then followed.
Data Analysis-Each experiment was repeated at least three times and different preparations of sT␤RII.Fc and sT␤RII-B.Fc were tested and used. Data are expressed as mean Ϯ S.E. The Ligand Program from the National Institutes of Health was used to fit binding curves for the binding data (34). The Student's t test was used with a p value of Ͻ0.05 to determine statistical significance.

Production and Characterization of Soluble Type II-B.Fc (sT␤RII-B.Fc) and Soluble Type II.Fc (sT␤RII.Fc) Chimeric
Proteins-cDNA encoding the ECDs of either T␤RII or T␤RII-B were fused to the Fc portion of human IgG and transfected into HEK 293 cells to generate sT␤RII.Fc and sT␤RII-B.Fc as described under "Experimental Procedures" and shown schematically in Fig. 1A. The ECD of human T␤RII-B contains the 26-amino acid insert that replaces Val 32 of T␤RII (14 -16).
Analysis of soluble receptor proteins purified by one-step Protein A affinity chromatography with SDS-PAGE and silver staining showed that the sT␤RII.Fc protein was ϳ50 kDa, whereas the sT␤RII-B.Fc protein was ϳ55 kDa (Fig. 1B, lanes  1 and 3), consistent with the presence of the 26-amino acid insert in sT␤RII-B.Fc. Under non-reducing conditions, protein bands of ϳ100 kDa for sT␤RII.Fc and ϳ110 kDa for sT␤RII-B.Fc were visualized, reflecting the disulfide bond formation of the dimeric Fc domain (data not shown).
Both sT␤RII.Fc and sT␤RII-B.Fc proteins were sensitive to N-glycosidase F treatment (Fig. 1B), indicating that both proteins are N-glycosylated. The molecular mass of the deglycosylated receptors, ϳ40 kDa (sT␤RII.Fc, lane 2) and ϳ42 kDa (sT␤RII-B.Fc, lane 4), correspond to the predicted molecular masses of the core protein of each chimeric protein.
Western blot analysis of soluble recombinant receptor chimeric proteins shown in Fig. 1C confirmed that the soluble receptor proteins contained both the human Fc domain (Fig.  1C, left), and the extracellular domain of the type II receptor (Fig. 1C, middle), using an anti-human Fc antibody (␣FC) and an anti-type II receptor ECD domain antibody (␣RII). As expected, a rabbit polyclonal antibody raised against the peptide encoding the 26-amino acid insertion sequence of T␤RII-B recognized only the sT␤RII-B.Fc protein with no detectable crossreactivity to sT␤RII.Fc protein (␣26aa, Fig. 1C, right). The anti-human Fc antibody also recognized s␤TRI.Fc (from R&D Systems, Fig. 1D).
sT␤RII.Fc and sT␤RII-B.Fc Can Bind TGF-␤1 and -␤3, but Not TGF-␤2-Radioligand competition experiments were performed to determine the selectivity and affinity of sT␤RII-B.Fc and sT␤RII.Fc proteins for different TGF-␤ isoforms. A nonsaturating amount of soluble receptor was incubated overnight with 125 I-labeled TGF-␤1, -␤2, or -␤3 with or without serial dilutions of unlabeled TGF-␤1, -␤2, or -␤3 at final concentrations from 2 pM to 500 nM, as indicated. The amount of com- Binding affinity values were calculated using Scatchard analysis of the binding data. A Scatchard analysis from one representative experiment is shown in Fig. 2, and averages from at least three separate experiments are summarized in Table I. sT␤RII-B.Fc and sT␤RII.Fc proteins had high affinity for TGF-␤1 and -␤3, with K d values in the picomolar range. sT␤RII-B.Fc had a severalfold higher affinity for TGF-␤1 and -␤3 than sT␤RII.Fc, but the difference in calculated affinities was not statistically significant. When 125 I-TGF-␤2 was employed, no binding could be detected, even when the amount of soluble receptor per well was increased to 100 ng/well. sT␤RII.Fc and sT␤RII-B.Fc Can Inhibit TGF-␤1 and -␤3, but Not TGF-␤2 Biological Activity-Next we tested whether sT␤RII-B.Fc and sT␤RII.Fc could block TGF-␤ activity by performing a biological inhibition assay using a TGF-␤ responsive luciferase reporter assay (Fig. 3). Mv1Lu cells transfected with the (CAGA) 12 MPL-Luc reporter construct were treated with 100 pM TGF-␤1, -␤2, or -␤3 in combination with increasing amounts (30 -1300 pM) of purified sT␤RII-B.Fc (Fig. 3A) or sT␤RII.Fc (Fig. 3B). The relative luciferase activity induced by TGF-␤1 (Fig. 3, diamonds) and TGF-␤3 (Fig. 3, triangles) was decreased in a dose-dependent manner by either sT␤RII-B.Fc or sT␤RII.Fc. The ED 50 was 360 pM for sT␤RII-B.Fc to both TGF-␤1 and TGF-␤3. For sT␤RII.Fc, the ED 50 was 664 pM for TGF-␤1 and 501 pM for TGF-␤3. In contrast, there was no inhibition of TGF-␤2-induced luciferase activity by either sT␤RII-B.Fc or sT␤RII.Fc (Fig. 3, squares).

Both Full-length T␤RII-B and T␤RII Expressed in L6 Cells That Lack T␤RIII Can Enhance TGF-␤2
Signaling Activity-We compared the effect of the three TGF-␤ ligands on the rat myoblast cell line L6 with or without transfection of fulllength T␤RII and T␤RII-B receptors. L6 cells lack T␤RIII (11,28) and T␤RII-B (16), but express T␤RII and T␤RI and are thus able to transduce TGF-␤1 signals (16,28). L6 cells were transfected with the (CAGA) 12 MPL-Luc reporter construct and treated with increasing amounts (0 -200 pM) of TGF-␤2 (Fig.  4A). The dose-response curve obtained with increasing amounts of TGF-␤2 suggested that the cells are able to signal through TGF-␤2 in the absence of T␤RIII and T␤RII-B, but require higher doses of TGF-␤2 compared with TGF-␤1 or -␤3 (data not shown), as has been previously shown (16).
Next, we examined the effect on TGF-␤2 signaling after transfecting L6 cells with increasing amounts (0 -5 g/6-cm plate) of T␤RII-B or T␤RII full-length cDNA (Fig. 4, panels B and C, respectively). Western analysis of cell lysates confirmed increased type II receptor protein expression with transfection of increasing amounts of cDNA (data not shown). The luciferase response of L6 cells transfected with reporter vectors did not differ when the cells were treated with 100 pM TGF-␤1, -␤2, or -␤3, confirming that L6 cells respond with similar potency to the three different TGF-␤ isoforms, in the absence of T␤RIII or TR␤II-B. Transfection of either type II receptor isoform into the L6 cells produced a dose-dependent increase in signaling to TGF-␤2 similar to that seen with TGF-␤1 and -␤3, indicating that both receptors are able to enhance signaling by TGF-␤2 with a similar potency to TGF-␤1 and -␤3.
sT␤RII-B.Fc and sT␤RII.Fc Require sT␤RI.Fc Protein for Binding to TGF-␤2-To investigate the possibility that sT␤RI.Fc and sT␤RII/II-B.Fc can form high affinity complexes, we mixed the soluble receptors and tested binding to ligands (Fig. 5). Different amounts of sT␤RII-B.Fc, sT␤RII.Fc (10 -50 ng), or sT␤RI.Fc (50 -100 ng) were incubated with 125 I-TGF-␤1 or 125 I-TGF-␤2. As expected, sT␤RII-B.Fc or sT␤RII.Fc alone significantly bound 125 I-TGF-␤1 in a dose-dependent manner, whereas there was no significant binding to 125 I-TGF-␤2. Also, sT␤RI.Fc by itself did not bind either 125 I-TGF␤1, as has pre-viously been reported (35), or 125 I-TGF-␤2. Next we mixed sT␤RII-B.Fc or sT␤RII.Fc at 10 and 50 ng together with 100 ng of sT␤RI.Fc in the presence of 125 I-TGF-␤2. The heterologous receptor complex was now able to bind 125 I-TGF-␤2 in a dosedependent fashion with increasing amounts of soluble type II receptors (Fig. 5A). A similar experiment was carried out by mixing increasing amounts of sT␤RI.Fc receptor (100 -500 ng) with a fixed amount of sT␤RII-B.Fc or sT␤RII.Fc (10 ng), which again lead to binding of 125 I-TGF-␤2 in a dose-dependent fashion (Fig. 5B).
sT␤RII-B.Fc and sT␤RII.Fc in Complex with sT␤RI.Fc Can Inhibit TGF-␤2 Signaling-To determine whether the complex composed of sT␤RII-B.Fc or sT␤RII.Fc with sT␤RI.Fc could effectively inhibit TGF-␤2 signaling, we incubated Mv1Lu cells transfected with the (CAGA) 12 MPL-Luc reporter construct, with 40 pM TGF-␤2 in the presence or absence of 5 g/ml sT␤RII-B.Fc, sT␤RII.Fc, or sT␤RI.Fc alone, or the mixture composed of 5 g/ml each of sT␤RII-B.Fc plus sT␤RI.Fc or sT␤RII.Fc plus sT␤RI.Fc. As shown in Fig. 6, when sT␤RII-B.Fc or sT␤RII.Fc were incubated with sT␤RI.Fc a functional heterocomplex could be reconstituted that significantly inhibited TGF-␤2-induced luciferase activity compared with each of the soluble receptors alone.

DISCUSSION
There are several advantages to using cell-free systems for analyzing binding properties of receptor complexes. First, the ability to quantitate specific binding is straightforward compared with indirect cell surface binding assays and thus allows an accurate measurement of binding affinity and specificity. This is especially helpful in cases such as the TGF-␤ family members in which the ligands often have multiple binding proteins of various affinities on or near the cell surface that lead to a high background of nonspecific binding. Second, the absence of any confounding co-expressed accessory proteins that might be present at the cell surface allows the determination of the binding properties of single types of receptors by themselves in isolation. Of note, the x-ray crystal structures of proteins are determined in a cell-free manner and the validity and utility of their data are well recognized and accepted.
The results presented here using a cell-free radioligand binding assay provide evidence that sT␤RII.Fc and sT␤RII-B.Fc by themselves can bind with high affinity to TGF-␤1 and -␤3 but do not bind to TGF-␤2. Corroborating data from a biological inhibition assay of TGF-␤ activity in cells were consistent with the selectivity and specificity of sT␤RII.Fc and sT␤RII-B.Fc seen in the binding data. Our measured binding affinities (30 to 500 pM) are consistent with previously published affinities for the ECD of T␤RII (31,36,37), and compare well with indirect estimates for membrane-bound complexes (1).
The precise role and effect on ligand binding of the 26-amino acid insert in T␤RII-B is unknown. It has been suggested that iodinated TGF-␤2 could be cross-linked to T␤RII-B but not T␤RII, thus indicating that the 26-amino acid insert might confer the ability to bind and signal via TGF-␤2 to the type II-B receptor (16). Indeed, the 26-amino acid insert is at the N  6 and 7) as indicated. Cells were lysed and lysates were assayed for luciferase activity. Luciferase values were normalized for transfection efficiency relative to Renilla activity, and data are presented as -fold increase in luciferase activity of cells treated with TGF-␤ ligand relative to untreated cells. Shown are results from one of three representative experiments. terminus of the type II-B receptor, and would be in the unresolved region of the crystal structure of the ECD of the type II receptor (25). This unresolved region is in close proximity to residues in the type II receptor that interact with TGF-␤3 (25), and could conceivably participate in altering ligand binding properties of the receptor. However, these prior studies were performed in COS cells, which express T␤RIII (11,12), and thus could not demonstrate the ability of T␤RII or T␤RII-B to bind TGF-␤2 in the absence of T␤RIII. In our cell-free system, we found no discernable difference in affinity or specificity between the TGF-␤ type II or type II-B receptors. Importantly, neither receptor by itself could bind TGF-␤2 with measurable affinity. Although there was a trend toward a higher affinity for TGF-␤1 and -␤3 for the sT␤RII-B.Fc than the sT␤RII.Fc protein, as well as a severalfold increase in the ED 50 in the biological inhibition assay, this increase was not statistically significant.
Interestingly, both full-length T␤RII and T␤RII-B, when expressed in L6 myoblasts that lack T␤RIII and native T␤RII-B, could increase TGF-␤2 signaling in a dose-dependent manner (Fig. 5), a finding that appeared paradoxical to the binding data. However, these results are consistent with a prior study showing responsiveness of L6 cells to increasing doses of TGF-␤2 and signaling augmentation by transfection with T␤RII-B (16). Although this prior study did not find augmentation of signal by transfection with T␤RII, a dose-response curve was not performed and so adequate doses of receptor may not have been used (16). Of note, another study demonstrated that type I and type II receptors could be cross-linked to 125 I-TGF-␤2 when they were overexpressed in COS cells, suggesting the possibility that TGF-␤2 can signal via complexes of type I and type II receptors (30). However, interpretation of this study is also limited by the presence of T␤RIII in COS cells as a potential confounding factor.
To understand the apparent discrepancy between our binding data showing high affinity binding of soluble type II receptors for TGF-␤1 and -␤3 but not ␤2, and the ability of the full-length receptor to mediate TGF-␤2 binding when expressed at the cell surface of L6 cells, we tested the ability of sT␤RII.Fc or sT␤RII-B.Fc and sT␤RI.Fc to form a functional complex with TGF-␤2 in vitro in the absence of T␤RIII. We found that the mixture of sT␤RI.Fc and sT␤RII.Fc or sT␤RII-B.Fc was sufficient to reconstitute binding to TGF-␤2. In addition, we demonstrate that the heterocomplex of sT␤RI.Fc and sT␤RII.Fc or sT␤RII-B.Fc was a functional inhibitor of TGF-␤2 in a biological inhibition assay compared with sT␤RI.Fc, sT␤RII.Fc, or sT␤RII-B.Fc alone.
These results imply that the complexed TGF-␤ type I and type II receptors are sufficient to bind TGF-␤2 in the absence of type III receptors when there are sufficient quantitities of type II receptors, type I receptors, and TGF-␤2 ligand present. This could result in TGF-␤2 binding and subsequent signaling in the absence of type III receptors, and could explain in part the difference between phenotypes of the type III receptor null mouse and the TGF-␤2 null mouse. For example, bone defects seen in the TGF-␤2 knockout phenotype were not evident in the T␤RIII null mutants, suggesting that bone cells are either exposed to sufficient TGF-␤2 ligand or express enough type II and type I receptors to allow TGF-␤2 signaling in the absence of the type III receptor.
Thus, we propose two pathways for TGF-␤2 signaling, one T␤RIII-dependent and one T␤RIII-independent (Fig. 7). As has previously been described, TGF-␤2 can bind to T␤RIII, which then recruits T␤RII and T␤RI, leading to phosphorylation of T␤RI and downstream signaling (Fig. 7, left; Ref. 28). Alternatively, in the absence of T␤RIII, TGF-␤2 can still bind and signal through the heterocomplex of T␤RII and T␤RI (Fig. 7,  right). Further studies will be needed to determine the relative affinities of TGF-␤2 for T␤RIII compared with the heterocomplex of T␤RII and T␤RI, and the relevance of these two pathways in vivo.
Our data also have important implications for the structure of the active TGF-␤2 signaling complex. The high resolution crystal structures of TGF-␤2 (21,22), TGF-␤3 (23), and more recently, the soluble ECD of the type II receptor (24), and the complex of the ECD-T␤RII and TGF-␤3 (25) have been solved, in addition to those of the BMPR1A⅐BMP2 (26) and ActRII⅐BMP7 complexes (27). In aggregate, these studies suggest that TGF-␤ receptor and ligand interactions may involve a cooperative model of binding with direct protein-protein contact between the type II and type I receptors as part of the assembly process, whereas the BMP/ActRII and BMP ligands may utilize an allosteric model of binding in which the type II and type I receptors do not necessarily make contact with each other. Our data strongly supports a cooperative model of receptor-ligand interaction for TGF-␤2 and predicts that a stable binding complex in the absence of T␤RIII will require the presence of the type II receptor, the type I receptor, and the TGF-␤2 ligand.
Many unanswered questions remain regarding the structure of the active TGF-␤2 signaling complex. Our model does not predict whether type II receptors, type I receptors, or both directly bind to TGF-␤2, only that the presence of both receptors is required for binding. One possibility is that both receptors bind directly to TGF-␤2. Another possibility is that the presence of one receptor causes a conformational change of the second, which then allows the latter to bind TGF-␤2. Perhaps most intriguing is the question of how the type III receptor is able to enhance TGF-␤2 binding to the type II and type I receptors in a multimeric ternary signaling complex. We have generated preliminary data using a soluble type III receptor-Fc construct suggesting that the interaction of the type III recep- FIG. 7. Model for TGF-␤2 binding to type II and type I receptors in a complex. TGF-␤2 can directly bind to type III receptors (RIII) but cannot directly bind to type II (RII) or type I (RI) receptors in isolation (top). Binding of TGF-␤2 to type III receptors is thought to lead to ensuing signaling by subsequent recruitment of type II and type I receptors (28) (bottom left). Alternatively, we present data showing that in the absence of type III receptors, TGF-␤2 can bind and signal through an heterocomplex of type II and type I receptors (bottom right).
tor and type II receptor are also cooperative. 2 Our assay system provides a rapid, potentially high-throughput and straightforward system to measure the binding of receptors or ligands to test hypotheses regarding receptorligand interactions. For example, it is unclear exactly why T␤RII and T␤RII-B cannot bind TGF-␤2 alone, because the amino acid residues that constitute the contact region of TGF-␤3 and TGF-␤2 are highly conserved with only conservative changes that do not suggest an obvious explanation for loss of binding to TGF-␤2 compared with TGF-␤3. We can now generate mutants in the corresponding amino acid residues that constitute the contact region of the type II and type II-B receptors to attempt to "restore" TGF-␤2 binding. Finally, mixing and matching experiments using different type I and type II receptors could allow the creation of new soluble complexes with novel specificities tailored to particular TGF-␤s to serve as antagonists.