A Soluble Transforming Growth Factor-β (TGF-β) Type I Receptor Mimics TGF-β Responses

Transforming growth factor-β (TGF-β) signaling requires a ligand-dependent interaction of TGF-β receptors ΤβR-I and ΤβR-II. It has been previously demonstrated that a soluble TGF-β type II receptor could be used as a TGF-β antagonist. Here we have generated and investigated the biochemical and signaling properties of a soluble TGF-β type I receptor (ΤβRIs-Fc). As reported for the wild-type receptor, the soluble ΤβR-I does not bind TGF-β1 on its own. Surprisingly, in the absence of TGF-β1, the ΤβRIs-Fc mimicked TGF-β1-induced transcriptional and growth responses in mink lung epithelial cells (Mv1Lu). Signaling induced by the soluble TGF-β type I receptor is mediated via the obligatory presence of both TGF-β type I and type II receptors at the cell surface since no signal was observed in Mv1Lu-derivated mutants for TGF-β receptors R-1B and DR-26. The comparison between the structures of TGF-βs and a three-dimensional model of the extracellular domain of ΤβRI has shown that five residues of the supposed binding site of TGF-β1 (Lys31, His34, Glu5, Tyr91, and Lys94) were found with equivalent biochemical properties and similar spatial positions.

Transforming growth factor-␤ (TGF-␤) 1 is a multifunctional cytokine involved in the regulation of many biological processes (1). In mammalian cells, responses to TGF-␤ are mediated by type I and type II cell surface receptors, which are expressed in most tissues (1,2). TGF-␤1, which is the prototype of the TGF-␤ family, elicits its effects by binding to a heteromeric complex of transmembrane Ser/Thr kinase receptors cloned as type I and type II receptors. The TGF-␤ type II receptor (⌻␤R-II) binds TGF-␤ on its own and recruits and phosphorylates the TGF-␤ type I receptor (⌻␤R-I) in a juxtamembrane domain rich in glycine and serine called the GS box (3). Activation of the type I receptor leads to the phosphorylation of members of the Smaand Mad-related factors (Smads) (4,5). Following receptor-dependent phosphorylation, Smad2 and Smad3 interact with Smad4, a constitutively phosphorylated common mediator of all the members of TGF-␤ family (6 -8), to translocate to the nucleus and mediate TGF-␤ responses. By these mechanisms, TGF-␤1 activates the transcription of mammalian genes important for cell cycle regulation and for extracellular matrix formation and may also promote immunosuppressive responses.
Based on these functions, the use of TGF-␤ agonists or antagonists could be of considerable interest in human therapy to prevent or exacerbate TGF-␤ responses. A better understanding of the mechanisms by which TGF-␤ could have beneficial or deleterious effects in various physiological processes is obtained through the generation of molecular tools that allow novel routes of investigation. The recent generation of transgenic mice either lacking the functional gene for TGF-␤ or overexpressing TGF-␤ in brain has greatly aided our understanding of the role that TGF-␤ plays in many systems. For example, targeted disruption of the TGF-␤1 gene in mice causes enhanced inflammatory responses leading to early death (9,10). Overexpression of brain-derived TGF-␤ promotes vascular amyloidogenesis, a histopathological hallmark of Alzheimer's disease (11,12). However, the molecular mechanisms that lead to these phenomena remain poorly understood.
Recently a better understanding of the pathophysiological actions of TGF-␤ have been obtained through the use of recombinant soluble TGF-␤ type II receptors as TGF-␤ antagonists (13,14). In a previous study (13), we have generated a soluble type II receptor for TGF-␤ fused with the Fc region of human immunoglobulin and characterized its ability to prevent TGF-␤ signaling in mink lung epithelial cells by sequestering the TGF-␤ peptide. By using this tool, we have demonstrated that blockage of the activity of endogenously produced TGF-␤, in a model of cerebral ischemia in rats, enhanced the volume of the damaged tissue (15). These data suggest that the potentiation of the TGF-␤ signaling in ischemic brain tissues may represent an innovative approach for the treatment of stroke in man. With the aim to develop agonists or antagonists for TGF-␤, we have generated a soluble type I receptor for TGF-␤ and then tested its potential ability to bind TGF-␤1 and to influence both TGF-␤-induced transcriptional and cell growth responses.

EXPERIMENTAL PROCEDURES
Construction of the Vector-The cDNA encoding the extracellular domain of the human ⌻␤R-I was amplified by PCR from the plasmid pBlueScript (Stratagene), which contained a full-length cDNA of the receptor (with minimal 5Ј-and 3Ј-untranslated regions). The sense primer was 5Ј-CCGAAGCTTGGCGAGGCGAGGTTTGCTGGGGTGAG-* 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  GCAGCG-3Ј corresponding to a HindIII site (underlined) and to base pairs Ϫ76 to Ϫ46 of the cDNA preceding the start codon. The antisense primer had the sequence 5Ј-CGGGGATCCACTTACCTGTTTCCACAG-GACCAAGGCC-3Ј representing base pairs 357-375, which encoded residues Gly 120 to Glu 125 immediately preceded by a splice donator site and a BamHI site (underlined). The standard PCR conditions using Pfu DNA polymerase (Stratagene) were as follows: 92°C for 1 min, 55°C for 1 min, and 72°C for 1 min for 30 cycles with a final extension at 72°C for 7 min. The resulting amplified PCR product was digested with HindIII and BamHI and ligated into the respective sites in the polylinker of the vector pIg-Tail (R&D Systems). The pIg-Tail expression system enables the mammalian production of fusion proteins with a C-terminal Fc tail. The truncated cDNA derived from PCR was sequenced to confirm the fidelity of the reaction. The cDNA encoding the extracellular domain of human ⌻␤R-II was amplified by PCR and cloned in the polylinker between the EcoRI and BamHI sites of the pIg-Tail (R&D Systems) containing a hygromycin-selectable marker.
Expression of the Chimeric Receptors-As a first attempt, the chimeric protein ⌻␤RIs-Fc was expressed into the medium of transiently transfected COS-7 cells (American Type Culture Collection CRL1651). COS-7 cells were cultured in AIM V medium (Life Technologies, Inc.) and transfected as previously described (53) with the recombinant pIg-Tail vector.
Following biochemical characterization the chimeric receptor ⌻␤RIs-Fc was routinely expressed into the medium of cultured Chinese hamster ovary cells (American Type Culture Collection CRL 9618) after a stable double transfection with recombinant pIg-Tail vector and pTK-Hyg vector (CLONTECH). A stable clone was isolated by single-cell cloning. The selected clone was cultured in AIM V medium (Life Technologies, Inc.) in which it loses its adherent properties and forms spheroids.
Cell Inoculation-Stable transfected Chinese hamster ovary cells were adjusted to 1.3 ϫ 10 6 /ml, and 15 ml of the suspension were injected into the cell compartment of the CELLine production chamber (CL1000 Integra Biosciences). The nutrient medium reservoir was filled with 1 liter of AIM V medium without antibiotics supplemented with 4.5 g/liter glucose and 4 mM Glutamax (Life Technologies, Inc.). The CELLine was then placed into a 37°C humidified 5% CO 2 incubator. A silicone membrane, at the bottom of the production chamber, allowed gas exchange.
Cell Growth and Receptor Production-The production chamber in which cells were grown was separated from the nutrient medium reservoir by a dialysis membrane permeable to nutrients smaller than 10 kDa. Low molecular weight metabolic products diffused from the 15-ml production chamber, which, in turn, provided nutrients for cell proliferation and protein synthesis.
Nutrient medium was exchanged once every week. Cells were harvested 7 days after inoculation. The cell suspension (15 ml) was collected, and the concentration and percentage of viable cells were determined. Once more, 15 ml at 3 ϫ 10 7 cells/ml were injected into the cell compartment. Cells were harvested three times per week.
Protein Purification and Analysis-The human recombinant proteins ⌻␤RIs-Fc and T␤RIIs-Fc were purified by one-step protein A affinity chromatography. To preserve the activity of the purified protein, the pH was immediately neutralized by the addition of a 1 ⁄10 volume of 1 M Tris-HCl, pH 9.0 to the collected fractions. The eluted protein was dialyzed overnight against 0.1ϫ phosphate-buffered saline and lyophilized. The purified protein was then analyzed by 10% SDS-PAGE using a precast gel (Novex) followed by silver staining (Silver Staining kit, Novex). The amount of protein was quantified using the Bio-Rad Protein Assay kit (microtiter plate format).
Western Blotting Analysis-Purified recombinant protein was separated by 10% SDS-PAGE and then transferred to a polyvinylidene difluoride membrane (Schleicher & Schuell). The membrane was blocked for 30 min with 10% dry fat milk in pH 7.2 NaCl/P i supplemented with 0.1% Tween 20. For the analysis of fractions containing ⌻␤RIs-Fc, the membrane was incubated with goat anti-human IgG Fc (Caltag) then with swine anti-goat IgG conjugated to horseradish (Caltag) peroxidase. As indicated in the corresponding figure, some immunoblots were revealed either with a rabbit antibody raised against human TGF-␤1 (a generous gift from P. ten Dijke) or a goat antibody raised against human IgG Fc (Caltag) then with an appropriate secondary antibody conjugated to horseradish peroxidase. The chemiluminescence immunoassay was performed with Renaissance Western Blot Chemiluminescence Reagent (PerkinElmer Life Sciences) following the procedure given by the manufacturer.
Protein A FlashPlate Binding Assay-The FlashPlate (PerkinElmer Life Sciences) is a white 96-well polystyrene microtiter plate with plastic scintillator-coated wells. The FlashPlates were precoated with protein A to allow the immobilization of the chimeric proteins (100 l of affinity-purified protein at 2.5 g/ml in pH 7.2 NaCl/P i ) by the Fc portion for 2 h at room temperature. After two washes with binding buffer (128 mM NaCl; 5 mM KCl; 5 mM MgSO 4 ; 1.3 mM CaCl 2 ; 50 mM Hepes, pH 7.6) the wells were treated for 2 h at room temperature with a blocking solution consisting of binding buffer containing 5% bovine serum albumin. After incubation, the wells were washed three times with additional binding buffer containing 1% bovine serum albumin. 125 I-labeled TGF-␤1 (PerkinElmer Life Sciences) was diluted in binding buffer prior to its addition into the wells. The plates were incubated at room temperature for 2 h, then sealed, and counted on a Packard Top Count microplate scintillation counter. A 1-min counting period was used. All determinations were carried out at least in duplicate. Nonspecific binding was determined with an excess of unlabeled TGF-␤1 (100-fold). For the competition binding assay, the final concentration of radiolabeled ligand was 600 pM (specific activity, 300 -450 Ci/mmol) with final concentrations ranging over 0.001-3 g/ml for ⌻␤RIs-Fc and ⌻␤RIIs-Fc.
Competition of 125 I-TGF␤1 Binding on Mink Lung Epithelial (Mv1Lu) Cells-The Mv1Lu cell line (American Type culture collection CCL-64) was maintained in minimum essential medium (MEM) supplemented with 10% fetal calf serum (Life Technologies, Inc.). 125 I-TGF␤1 (PerkinElmer Life Sciences) binding to monolayer cells was performed following the published procedures (2). For the competition binding assay, the final concentration of radiolabeled ligand was 40 pM (specific activity, 3000 -4500 Ci/mmol) with final concentrations ranging over 0.001-10 g/ml for ⌻␤RIs-Fc and ⌻␤RIIs-Fc alone or associated.
Reporter Gene Assay-Mv1Lu cells were stable transfected (by using the lipofection protocol Transfast (Promega)) with the construct (caga) 12 MLP-Luc (16) and the selection vector pTK-Hyg (CLONTECH). CAGA reporter vectors were generated using pGL3 basic plasmid (Promega). Transfected cells were plated into 96-well plates (5 ϫ 10 4 cells/ well) using MEM containing 10% fetal calf serum for one night before the assay. The wells were washed three times in MEM, 0% fetal calf serum then cells were incubated at 37°C, 5% CO 2 in MEM, 1% fetal calf serum. After 1 h, increasing concentrations of TGF-␤1 (R&D Systems) (25-300 pg/ml) and ⌻␤RIs-Fc and ⌻␤RIIs-Fc (0.03-3 g/ml) were added to the medium. Cells were then harvested 6 h later and assayed for luciferase activity using the Luc-Lite Packard kit as described by the manufacturer. The total light emission was measured using a Packard TopCount microplate luminescence counter.
Similar experiments were performed following transient transfection of the (caga) 12 MLP-Luc in Mv1Lu, DR-26, and R-1B cell lines by using the lipofection protocol (Transfast, Promega). Cells were harvested 20 h after the addition of either TGF-␤1 (1 ng/ml) or ⌻␤RIs-Fc (3 g/ml) in serum-deficient MEM and assayed for luciferase activity by using a commercial system (luciferase reporter assay system, Promega).
DNA Synthesis Assay-Cells were plated 24 h after treatment in 24-well plates containing medium with 10% fetal bovine serum. Medium was replaced with fresh medium containing 1% fetal bovine serum in the presence or absence of TGF-␤1 (1 ng/ml) or ⌻␤RIs-Fc (3 g/ml) for 20 h. Cells were labeled with [ 3 H]thymidine (1 Ci/ml) (PerkinElmer Life Sciences) for the last 4 h of incubation. Cells were washed three times with cold phosphate-buffered saline, fixed with 5% trichloroacetic acid for 1 h at 4°C, washed twice with cold 5% trichloroacetic acid, and extracted with 1 N NaOH for 30 min at room temperature. The extracts were collected and counted in a ␤-counter.
Generation of a Knowledge-based Model for the Extracellular Domain of T␤R-I-The coordinates of the different three-dimensional (3D) structures are found in the Protein Data Bank (PDB) (17,18). The 3D structure of the extracellular domain (ECD) of bone morphogenetic protein receptor IA (BRIA-ECD) (19) has been recently determined (PDB entry 1es7). The software TITO (20) has been used to check and optimize the alignment and to build the C␣ chain of the extracellular domain of T␤R-I (T␤R-I-ECD) based on the coordinates of BRIA-ECD. The model has been built using Modeler (21). The 3D structure of TGF-␤ is known for the three isoforms (PDB entries 1klc for TGF-␤1 (22), 2tgi for TGF-␤2 (23), and 1tgj for TGF-␤3 (24)).

Generation of a Soluble TGF-␤ Type I Receptor Fused with the Fc Region of Human
IgG-COS-7 cells were transfected with the expression plasmid pIg-Tail containing the cDNA encoding for a truncated TGF-␤ type I receptor (amino acid residues 1-124) as described in Fig. 1. After one-step purifica-tion by affinity chromatography on protein A-Sepharose, a soluble pure chimeric protein was recovered from the medium of cells transiently transfected with the pIg-Tail/⌻␤RIs plasmid. Analysis of the purified chimeric receptor was performed by SDS-PAGE under either nonreducing or reducing conditions prior to silver staining or by immunoblotting performed with an antibody raised against human immunoglobulin (Fig. 2). A product of molecular mass of ϳ40 kDa was visualized in reducing conditions corresponding to the expected molecular mass of the extracellular domain of ⌻␤R-I fused with the Fc region of the human immunoglobulin (11 ϩ 30 kDa). However, in nonreducing conditions, immunoblotting revealed products of higher molecular mass (over 100 kDa), which may represent homomeric forms of a secreted chimeric soluble receptor due to the presence of the Fc fragment.
Binding Properties of the Soluble TGF-␤ Type I Receptor Fused with the Fc Region of Human IgG-The ligand binding activity of recombinant ⌻␤RIs-Fc receptor was tested and compared with the previously characterized ⌻␤RIIs-Fc (13) in a protein A FlashPlate binding assay. Protein A-coated plastic surfaces within the wells were coated with the purified recombinant ⌻␤RIs-Fc chimeric receptor allowing the immobilization of the chimeric protein by its Fc portion. Then 125 I-TGF-␤1 was added in the presence of increasing concentrations of soluble type II receptor. As shown in Fig. 3, increasing concentrations of soluble TGF-␤ type I receptor (from 3 ng/ml to 3 g/ml) failed to influence 125 I-TGF-␤1 binding (500 pM). However, co-incubation in the presence of similar concentrations of the previously characterized ⌻␤RIIs-Fc (13) revealed 125 I-TGF-␤1 binding to the coated type I receptor with a typical saturation curve. The kinetics of 125 I-labeled TGF-␤1 binding were performed at room temperature in conditions showing equilibrium binding, achieved after 2-3 h of incubation in the case of the ⌻␤RIIs-Fc-coated FlashPlate (data not shown). Nonspecific binding, obtained upon ⌻␤RIs-Fc coating, was determined for each condition in the presence of 200 nM cold TGF-␤1 and subtracted from total binding. Similarly, TGF-␤1 binding activity to the recombinant T␤RIIs-Fc was tested in a precoated protein A FlashPlate binding assay as described under "Experimental Procedures." The addition of increasing concentrations of the soluble type I receptor cannot compete binding of iodinated TGF-␤1 to the precoated T␤RIIs-Fc. Based on these observations, two concepts could explain these data. The first one could be that T␤RIs-Fc is not capable of binding TGF-␤1 on its own, and the second one may be that the T␤RIs-Fc is actually binding type II and does not directly contact the ligand.
To further characterize the binding properties of the soluble TGF-␤ type I receptor, 125 I-TGF-␤1 binding was evaluated in Mv1Lu cells in the presence of increasing concentrations of soluble TGF-␤ type I (Fig. 4). As expected, increasing concentrations of the soluble TGF-␤ type II receptor inhibited the binding ability of iodinated TGF-␤1 on the Mv1Lu cell line with an apparent IC 50 of 0.15 g/ml. In contrast, increasing concentrations of the soluble type I receptor failed to influence the binding of TGF-␤1 in Mv1Lu cells. These data show that extracellular domains of both ⌻␤R-I and ⌻␤R-II display the same binding properties that the full-length transmembrane receptors expressed in cell lines (3).
The Soluble TGF-␤ Type I Receptor Mimics TGF-␤ Responses-To test the properties of the recombinant chimeric TGF-␤ type I receptor on TGF-␤ signaling, this component was tested on a model of TGF-␤-responsive luciferase reporter gene assay in the presence of exogenous TGF-␤1. Following stable transfection in the Mv1Lu cell line, increasing concentrations of TGF-␤1 were evaluated for their ability to induce the activation of the previously proposed (16) TGF-␤-inducible element, named CAGA box (Fig. 5). As shown in Fig. 5A, TGF-␤1 enhances the activity of the CAGA-luciferase reporter gene in the Mv1Lu cell line in a dose-dependent manner. As expected, the soluble TGF-␤ type II receptor induced a dose-dependent inhibition of the TGF-␤1 response (TGF-␤1 at 80 pg/ml) (Fig.  5B). Surprisingly, although ⌻␤RIs-Fc did not influence the TGF-␤-dependent transcriptional activity for a range of concentrations between 30 ng/ml and 1 g/ml, at the highest concentrations tested (3 g/ml) it induced a reproducible enhancement of the TGF-␤-induced signal (Fig. 5B). Based on these results, similar experiments were performed in the absence of exogenous TGF-␤1. As observed in Fig. 5C, the ⌻␤RIs-Fc induced a marked dose-dependent activation of the TGF-␤1-inducible element CAGA box. Altogether these data demonstrate that while a soluble type II receptor could be characterized as a TGF-␤ antagonist, a soluble TGF-␤ type I receptor mimicked TGF-␤ transcriptional activity. Moreover, to address the possibility that immunoglobulin could drive TGF-␤ receptor heteromerization and subsequent signaling, increasing concentrations of purified whole human immunoglobulin were tested in a TGF-␤1-responsive reporter gene assay. As shown in Fig. 5D, whole IgG failed to induce TGF-␤1-like signaling. Moreover, the specificity of the T␤RIs-Fc-induced transcriptional response was confirmed by data showing that the same type of molecular construct containing the extracellular domain of the type II receptor did not induce any signal (Fig. 5C).

T␤RIs-Fc Mimics TGF-␤1-induced Growth Inhibition in Mv1Lu
Cell Line-Similar results to those obtained for the transcriptional activity of the TGF-␤1-dependent luciferase reporter gene were observed with the antimitogenic effect of  Fig. 7, although both TGF-␤1 and ⌻␤RIs-Fc enhance the CAGA-luciferase reporter gene activity in the Mv1Lu cell line (Fig. 7A), ⌻␤RIs-Fc is devoid of this activity in the DR-26 cell line (Fig. 7B). R1-B cells are unable to respond to TGF-␤ because of the absence of functional TGF-␤ type I receptor. As observed in Fig. 7C, the soluble h⌻␤RIs-Fc failed to restore TGF-␤ signaling in this model. Overall these results suggest that the soluble TGF-␤ type I receptor requires functional ⌻␤R-II and ⌻␤R-I to transduce its signal.
T␤RIs-Fc Shares Structural Similarities with TGF-␤-The soluble TGF-␤ type I receptor is able to form a stable complex with both ⌻␤R-II and ⌻␤R-I and to mimic TGF-␤1-mediated responsiveness. To further understand this process, we have generated a three-dimensional model of the extracellular domain of T␤R-I based on common sequence and structural fea- tures with the already known structure of BRIA-ECD (19). A model of T␤R-I-ECD (PDB entry 1tbi) was constructed using protectin CD59 as a template (25). The alignment between T␤R-I-ECD and CD59 was based on a conserved cysteine pairing pattern and a highly conserved cysteine box (sequence CCXXXXCN). No other similarities between theses two sequences emerged from the alignment. The alignment with BRIA-ECD shows a better conservation of hydrophobic properties and less gaps and insertions than the alignment with CD59 even if only 9 of the 10 cysteine residues are conserved. Four disulfide bridges among the five have been preserved. The new disulfide bridge is easily rebuilt within the model without important distortions. The potential of mean force and the surface potential appear correct according to the validation software Verify3D (data not shown) (26). A comparative analysis of the proposed T␤R-I-ECD model and the structures of TGF-␤s has shown interesting similarities. Although sequences and structures of TGF-␤s and T␤R-I-ECD are dissimilar and cannot be aligned, five residues with similar properties can be spatially superimposed (Fig. 8, A and B). We were able to superimpose an arginine (a lysine in TGF-␤2) on a lysine, a tyrosine on a phenylalanine, a histidine on a histidine, a glutamate on an aspartate, and a lysine on a lysine in TGF-␤s and in T␤R-I-ECD, respectively.

DISCUSSION
In the present study, we have reported that a soluble TGF-␤ type I receptor could mimic TGF-␤1 responses in mink lung epithelial cells. Moreover, a comparative analysis of both the T␤R-I extracellular domain and structures of TGF-␤s leads to the identification of a potential set of five amino acids directly involved in the binding site of TGF-␤s to its type II receptor.
To further understand the mechanisms of action of TGF-␤, we have generated a chimeric soluble TGF-␤ type I receptor and compared its properties with the previously generated soluble TGF-␤ type II receptor (13). Surprisingly, although the ⌻␤RIs-Fc does not bind TGF-␤1 on its own as previously characterized for the wild-type receptors (3,27), it was able to mimic TGF-␤ signaling in the Mv1Lu cell line. Moreover, although ⌻␤RIIs-Fc prevents TGF-␤1 binding to the Mv1Lu cell line, the soluble ⌻␤RI receptor does not. However, the observation that the extracellular domain of the TGF-␤ type I receptor could mimic TGF-␤ signaling was completely unexpected. The structure of the monomer of TGF-␤1 is composed of two antiparallel two-stranded ␤-sheets tightly bound by a cysteine knot and two disulfide bridges forming a ring with a third disulfide bridge that passes through both (22). Two loops protrude from the core of the protein, the first one between the first and second strand (amino acids [25][26][27][28][29][30][31][32][33][34][35] and the second one between the third and fourth strand (amino acids 90 -98). These two loops and an N-terminal helix have been proposed, according to the 3D structures of different TGF-␤ isoforms, to be involved in receptor recognition (24). In addition, mutations of amino acids 92-98 prevent an interaction of TGF-␤1 with its type II receptor (28,29). In the present study, we have found five residues of TGF-␤1 located in these two loops (Lys 31 , His 34 , Glu 35 , Tyr 91 , and Lys 94 ) with equivalent biophysical properties and similar spatial positions in T␤R-I-ECD. Mutational analysis has shown that residues 92-98 control the receptor (T␤R-II) affinity for each of the TGF-␤ isoforms (30). Interestingly, binding to T␤R-II is more efficient for TGF-␤1 and TGF-␤3 than for TGF-␤2 (3,31). Four of the five residues that we think are involved in the recognition process between TGF-␤1 and T␤R-II are conserved in the three TGF-␤ isoforms. In contrast, in position 94, an arginine in TGF-␤1 and TGF-␤3 is replaced by a lysine in TGF-␤2 as observed in the T␤R-I-ECD (Fig. 8C). These results could explain why T␤R-I-ECD forms a complex with T␤R-II but does not compete with TGF-␤1-bound T␤R-II. The binding of bone morphogenetic protein-2 (BMP-2), a member of the TGF-␤ superfamily, to its cellular receptors follows a mechanism differing from that established for TGF-␤s or activins (19). The ordered sequential binding mechanism established for BMP-2 involves, first, the binding of external ligand to the type I receptor chain and, second, within the membrane, the recruitment of the type II receptor into the complex. Interestingly, the crystal structure of the complex between BMP-2 and the type IA receptor for bone morphogenetic protein has shown that the receptor recognition site in the ligand was formed by two acidic groups, two basic groups, and a cluster of four hydrophobic groups (with one aromatic group among them) (19). The proposed recognition site of TGF-␤s for its type II receptor, based on the mimetic properties of T␤RIs-Fc, has similar characteristics: it is composed of three basic groups, one acidic group, and an aromatic hydrophobic group.
Recent studies have provided evidence for the involvement of TGF-␤1 and its signaling pathway in the pathogenesis of liver, kidney, lung, and brain diseases (32) and chronic inflammatory disorders such as cirrhosis, glomerulonephritis (33), chronic rejection (34), or pulmonary fibrosis (35). For example, transgenic mice overexpressing TGF-␤1 and adenovirus-mediated gene transfer of TGF-␤1 are responsible for widespread fibrosis (36). Moreover, perturbation of the TGF-␤ signaling pathway has also been involved in tumor genesis and related angiogenesis (37). Based on these observations, it has been hypothesized that inhibition of TGF-␤1 signaling may be a useful therapeutic strategy for the treatment of these pathologies.
TGF-␤1, more recent reports have also characterized TGF-␤ as a beneficial cytokine in other pathologies. TGF-␤ is antiproliferative with effects not only on the immune system as revealed by TGF-␤ knock-out mice (38,39) but also on epithelial cells (40). These observations are reinforced by findings showing that some epithelial cancers are resistant to TGF-␤; the mechanisms vary from mutation of type II receptor (41) to alteration of the TGF-␤ signaling pathway (42,43). Moreover, in a previous study, we have demonstrated that blocking the endogenously produced TGF-␤1 induced a large increase in the damaged tissue in a model of cerebral ischemia in rats (15,44).
Although soluble receptors able to bind their specific ligands have been characterized as potent antagonists, such as the tumor necrosis factor-␣ soluble receptor (45) or interleukin-1 receptor antagonist (46), no data so far published have demonstrated that a soluble receptor, unable to bind ligand by itself, could mimic specific signals. Nevertheless, it has been shown that soluble glial-derived neurotrophic factor receptor-␣1 (47) or soluble interleukin-6 receptor (48) can potentiate the signal of their respective ligands. All the members of the TGF-␤ family, including activins and BMPs transduce their signal through a set of type I and type II receptors (49). Truncated type II receptors (50) or soluble type II receptors for TGF-␤ (13) display dominant negative activities. These data suggest that soluble type I receptors for TGF-␤ family members could be used to mimic the biological actions of these cytokines.
Altogether our data show that chimeric soluble receptors are interesting tools for fundamental research. Moreover, the use of soluble receptors for cytokines fused with the Fc domain of immunoglobulin as tools for therapy may offer significant advantages over DNA-based strategies requiring adenoviral vectors because, as IgGs, these components have a half-life of several days in the circulation (51) and in the humanized form are unlikely to elicit an immune response. Nevertheless, the The first chain of TGF-␤1 is colored in pale blue, the second chain in blue, and the T␤R-I-ECD chain in gray. The disulfide bridges are colored in orange for both structures. The five residues that share similar properties and similar spatial positions are shown in atomic representation and are colored in dark blue for the positively charged residues, red for the negatively charged residues, and green for the aromatic residues. B, close-up view of the TGF-␤1 and T␤R-I-ECD recognition site for the TGF-␤ type II receptor. The residues belonging to the proposed T␤R-II recognition site are labeled in pale blue for TGF-␤1 and in gray for T␤R-I. The alternation of a positively charged residue (Arg 94 /Lys 19 ), an aromatic residue (Tyr 91 /Phe 13 ), a histidine (His 34 /His 81 ), and a negatively charged residue (Glu 35 /Asp 80 ) is similar in both proteins. In addition, another positively charged residue (Lys 31 /Lys 84 ) leads to an epitope size of approximatively 15-10Å. C, amino acid sequences for the three human TGF-␤ isoforms with their corresponding PDB codes in parentheses. The cysteine residues are circled in orange, and the disulfide bridges are shown with orange lines. The three zones previously characterized as belonging to the receptor binding site are highlighted with a red line. The two regions of T␤R-I-ECD that can be spatially superimposed to TGF-␤ are shown below the TGF-␤ sequences. The first region, located between the first and second strand in TGF-␤, is superimposed in reverse orientation, and the second region, located between the third and fourth strand, is superimposed in the same orientation as the TGF-␤s. The five predicted residues directly involved in the recognition process of the TGF-␤ type II receptor are circled in pale blue in the TGF-␤s and colored in gray in ⌻␤R-I. This figure was made using the software ESPript (52). use of soluble TGF-␤ type I or type II receptors in the treatment of TGF-␤-related pathologies should be further investigated in animal models.