Adenovirus-mediated Transfer of a Truncated Transforming Growth Factor-β (TGF-β) Type II Receptor Completely and Specifically Abolishes Diverse Signaling by TGF-β in Vascular Wall Cells in Primary Culture

We constructed an adenoviral vector expressing a mutated human type II transforming growth factor-β (TGF-β) receptor that was truncated of its kinase domain (AdexCATβTR) and examined whether this truncated receptor could abolish signaling by TGF-β using arterial endothelial cells and smooth muscle cells, as well as a lung epithelial cell line (Mv1Lu). Infection of cells with AdexCATβTR induced expression of the truncated receptor, the amount of which would be excessive compared with those of both full-length type I and type II receptors, as assessed by levels of their mRNAs. The antiproliferative effect of TGF-β was completely eliminated in both endothelial cells and Mv1Lu that were infected with AdexCATβTR. The transcriptional activation by TGF-β of plasminogen activator inhibitor-1 and fibronectin was entirely suppressed. Abrogation of the TGF-β-enhanced production of type I collagen in infected smooth muscle cells was confirmed by immunocytostaining and by [14C]proline incorporation in a quantitative manner. Mitogenic response to other growth factors remained unaffected in infected cells. Our data demonstrated that the adenovirus-mediated transfer of a truncated type II TGF-β receptor completely and specifically abolishes the diverse effects of TGF-β as a dominant-negative mutation, supporting the hypothesis that both the type I and type II receptors are required for all signaling by TGF-β. This method may facilitate the clarification of the role of TGF-β both in vitro and in vivo.

Transforming growth factor-␤ (TGF-␤) 1 is a multifunctional cytokine that regulates cell proliferation and differentiation and extracellular matrix production (1)(2)(3). TGF-␤ seems to play pivotal roles in embryogenesis (4,5) and in chronic fibroproliferative pathophysiological conditions in adults such as liver cirrhosis, chronic glomerulonephritis, and atherosclerosis (3). However, partly due to its multiple functions, the exact roles of TGF-␤ remain controversial. To clarify the actual roles of TGF-␤ in physiological conditions, an effective and feasible method that inhibits the signaling by TGF-␤ in a specific manner is required.
TGF-␤ exerts its effects by binding to and activating specific receptors located on the cell membrane. Two signaling receptors, termed type I and type II receptors, have been cloned recently (6 -11). Each type possesses an extracellular region, a single transmembrane portion, and a serine/threonine kinase domain in its cytoplasmic region. Recent molecular and cytogenetic analyses have shown that the type II receptor can itself bind free TGF-␤ (8), whereas the type I receptor can only recognize TGF-␤ that is already bound with the type II receptor (7, 9 -13), and suggested that formation of a ligand-induced heterodimer (or a tetramer) involving both the type I and type II receptors is required for signaling (11,(13)(14)(15)(16). In this model, transphosphorylation of the type I receptor by the type II receptor, which is constitutively phosphorylated, seems to be essential for signal propagation (14 -16). Evidence supporting this idea is that a mutated type II receptor that was either truncated of its kinase domain (15) or substituted one critical amino acid resulting in the loss of its transphosphorylation activity (16) inhibited, albeit partially, many of the signaling responses induced by TGF-␤. However, other investigators have found that a similar truncated type II receptor inhibited only the antiproliferative effect of TGF-␤ and not the transcriptional activation of extracellular matrix proteins, suggesting that each receptor may have its own distinct signaling pathways (17). Further, in most studies reported so far, the signaling pathways were investigated either in an established epithelial cell line derived from the mink lung, Mv1Lu cells, which were either intact or chemically mutated, or in COS cells that were transiently transfected with the receptors. The use of cells in primary culture may be necessary to determine whether the truncated type II receptor could abolish all the diverse signaling pathways of TGF-␤ as a dominant-negative mutation, thus serving as a useful tool for the elucidation of the roles of TGF-␤.
It has been shown that an adenoviral vector is remarkably efficient for both in vitro and in vivo gene transfer in a wide variety of cells and species. In this study, we constructed a replication-defective adenovirus expressing the truncated human type II TGF-␤ receptor under a powerful constitutive promoter (AdexCAT␤TR) and investigated whether the adenovirus-mediated expression of the truncated receptor could eliminate the diverse effects of TGF-␤ as a dominant-negative mutation in cells derived from the arterial wall as well as in Mv1Lu cells. All signals tested elicited by three isoforms of TGF-␤ were completely suppressed in a TGF-␤-specific manner. Our data support the notion that the formation of ligandinduced heteromeric complexes involving both the type I and type II receptors is required for all diverse effects of TGF-␤. Furthermore, the results demonstrate that the adenovirusmediated transfer of the truncated type II TGF-␤ receptor should prove a useful tool for the clarification of the roles of TGF-␤ both in vitro and in vivo.

MATERIALS AND METHODS
Cell Culture-Mink lung epithelial cell line, Mv1Lu (CCL-64, American Type Culture Collection), was grown in minimum essential medium (Life Technologies, Inc.) with 10% fetal bovine serum (Bio Whittaker, Walkersville, MD). Arterial endothelial cells (EC) and smooth muscle cells (SMC) were primarily prepared from the bovine thoracic aorta as described previously (18,19) and maintained in Dulbecco's modified Eagle's medium (DMEM) (Life Technologies, Inc.) with 10% fetal calf serum (Hyclone Lab., Logan, UT). Each medium contained 100 IU/ml of penicillin and 50 g/ml of streptomycin. The EC expressed factor VIII antigen and the SMC showed a characteristic hill-and-valley morphology and proved positive to immunocytostaining using an antibody to ␣-smooth muscle actin. EC and SMC of passages 2-8 were employed in this study. Balb/3T3 cells (clone A31) were cultured in DMEM containing 10% fetal calf serum.
Construction of Replication-defective Recombinant Adenoviruses-A replication-defective E1 Ϫ and E3 Ϫ recombinant adenoviral vector expressing the truncated human type II TGF-␤ receptor was prepared as described previously (20 -22). Briefly, a human type II TGF-␤ receptor cDNA tagged with an influenza virus hemagglutinin (HA) epitope at its N terminus (15) (provided by J. Massague, Memorial Sloan-Kettering Cancer Center) had most of the cytoplasmic kinase region deleted (only 7 amino acids remained in the intracellular region), and it was placed into a cassette cosmid vector, pAdexCA1w (provided by I. Saito, University of Tokyo) under a CA promoter comprising a cytomegalovirus enhancer and a chicken ␤-actin promoter (23) (pAdexCAT␤TR). A recombinant adenovirus was constructed by in vitro homologous recombination in 293 cells (24) using pAdexCAT␤TR and the adenovirus DNA-terminal protein complex by a method previously described (25). The desired recombinant adenovirus, designated as AdexCAT␤TR, was purified by ultracentrifugation through a CsCl 2 gradient followed by extensive dialysis. The titer of the virus stock was assessed by a plaque formation assay using 293 cells and expressed as plaque formation unit. Two control adenoviruses were used: AdexCALacZ expressing bacterial ␤-galactosidase and Adex1w containing no exogenous gene (20 -22). Transfection of cells with adenoviral vectors was carried out by incubating confluent cells with vectors in serum-free medium for 2 h at room temperature under gentle agitation. Then the medium was exchanged with a complete medium, and the cells were incubated at 37°C in 5% CO 2 until assays.
Measurement of DNA Synthesis-Cells in 24-well plates were infected with AdexCAT␤TR, Adex1w, or AdexCALacZ at m.o.i. 20 or left uninfected and incubated for 24 h in medium containing either 0.2% fetal bovine serum for Mv1Lu cells or 1% fetal calf serum for EC. Mv1Lu cells and EC were incubated for 18 h with various concentrations of TGF-␤1 (provided by Kirin Brewery Co. Tokyo, Japan) in addition to serum (0.2% fetal bovine serum for Mv1Lu cells or 1% fetal calf serum for EC). Quiescent Balb/3T3 cells either infected with AdexCAT␤TR at m.o.i. 20 or left uninfected were stimulated for 18 h with either 5 ng/ml of platelet-derived growth factor (PDGF)-BB (Upstate Biotechnology Inc., Lake Placid, NY) or 5 ng/ml of basic fibroblast growth factor (bFGF) (Upstate Biotechnology Inc.). Then cells were pulsed for 4 h with 1 Ci/ml [ 3 H]thymidine (DuPont NEN). The incorporation of [ 3 H]thymidine into trichloroacetic acid-insoluble material was measured using a scintillation counter.
Immunocytochemistry-Expression of type I collagen in SMC were immunocytochemically examined using a monoclonal antibody recognizing human type I collagen (26). SMC on four-chamber slides (Nunc) were infected for 2 h with either AdexCAT␤TR or Adex1w at m.o.i. 20 or left uninfected and incubated in serum-free medium (DMEM containing 0.05% bovine serum albumin, 1 g/ml of insulin, 5 g/ml of transferrin and 25 mM of HEPES, pH 7.4) for 48 h. Then cells were challenged with 200 pM of TGF-␤1 for 60 h (the medium being refreshed every 24 h). Cells fixed with acetone were treated with the primary antibody overnight at 4°C. The primary antibody was not added to some cells as a negative control. Type I collagen was visualized using a biotinylated rabbit anti-mouse IgG antibody (Nitirei, Tokyo, Japan) as a secondary antibody, peroxidase-labeled streptavidin, and diaminobenzidine. The cells were lightly counterstained with hematoxylin.
Measurement of Collagen Synthesis-Quiescent SMC were infected with either AdexCAT␤TR or AdexCALacZ at m.o.i. 20, or left uninfected, and incubated in serum-free medium for 48 h. Cells were then stimulated with 160 pM of TGF-␤1 in serum-free medium for 24 h. Then the medium was replaced with DMEM containing 1.0 Ci/ml of L-[ 14 C(U)]-proline (266.4 mCi/mM) (DuPont NEN), 50 g/ml of ascorbic acid, and 50 g/ml of ␤-aminopropionitril fumarate in addition to 160 pM of TGF-␤1. Further 24 h later, the medium and cells were harvested, sonicated and centrifuged. Pellets were used for estimation of DNA content by the method of Burton (27). One tenth volume of trichloroacetic acid (55%) was added into the supernatant to precipitate protein.
Precipitates were washed in ice-cold trichloroacetic acid (5%), dissolved in 50 mM Tris, pH 7.4, and dialyzed extensively for 48 h at 4°C. The radioactivity in the precipitates was counted for total protein synthesis. Aliquots of each samples were digested with purified bacterial collagenase (Form III, Advance Biofactures Corp., Lynbrook, NY), and these collagenase-digestible fractions were counted for collagen synthesis as previously reported (28). Values were expressed as counts per minute (cpm)/g of DNA. Statistical analysis of values was performed by the variance and unpaired Student's t test with a P value of Ͻ0.05 considered significant.

Adenovirus-mediated Expression of Truncated Human Type II TGF-␤ Receptor in Mv1Lu
Cells-We first examined the expression levels of the truncated human TGF-␤ type II receptor in Mv1Lu cells exposed to AdexCAT␤TR at various m.o.i. As shown in Fig. 1A, mRNA corresponding to the truncated type II TGF-␤ receptor was detected by Northern blotting probed with a 5Ј end of the human TGF-␤ type II receptor cDNA. In Fig. 1B, a glycosylated protein of approximately 45 kDa was detected by immunoblotting analysis using an antibody against the HA epitope. The amount detected in both mRNA and protein levels seemed to be m.o.i.-dependent. The result suggested that transfection at m.o.i. 20 would achieve submaximal expression of the truncated TGF-␤ type II receptor.
We next wanted to know a relative expression level of truncated and full-length type II receptors and whether expression of the full-length receptor was affected by infection with AdexCAT␤TR. Because the TGF-␤ receptor is a low abundant protein (8) and not easily detected by Western blotting, we compared the levels of mRNA for the two receptors in EC that had been exposed to AdexCAT␤TR at various m.o.i. Two mRNAs were detected in Northern blotting using a 5Ј end of the human type II receptor cDNA as a probe (Fig. 2A). These mRNAs, of 5.5 and 0.9 kb, would correspond to the full-length bovine type II and the truncated human type II receptors, respectively. The mRNA of the truncated receptor in cells infected with AdexCAT␤TR at m.o.i. 3 was already extremely abundant compared with that of the bovine full-length receptor. Although we did not quantify the receptor proteins, and we might underestimate the mRNA level of the bovine full-length type II receptor due to using a human probe, these levels of mRNA may suggest that a large excess of the truncated receptor over the full-length receptor would be expressed in cells exposed to AdexCAT␤TR. Fig. 2B indicates that transcription of the full-length type II receptor was not significantly altered by the AdexCAT␤TR-mediated co-expression of the truncated type II receptor; the intensities of mRNA for full-length type II receptor normalized to signals of GAPDH or ␤-actin were virtually consistent. The mRNA level of the full-length type I receptor was also not changed by infection with AdexCAT␤TR (Fig. 2C).

Suppression of the TGF-␤-induced Antiproliferative Effect by Adenovirus-mediated Transfer of the Truncated Type II TGF-␤
Receptor-TGF-␤ has been shown to arrest the cells in the late G 1 phase of the cell cycle (1,29,30). We examined whether adenovirus-mediated transfer of the truncated type II TGF-␤ receptor could release the antiproliferative effect of TGF-␤ in either Mv1Lu cells (Fig. 3A) or EC (Fig. 3B). Cells infected with a control adenovirus, either Adex1w (3A) or AdexCALacZ (3B), as well as intact cells, exhibited an antiproliferative response to TGF-␤1 in a dose-dependent manner over the range of concentrations 0 -200 pM (Fig. 3). However, cells treated with AdexCAT␤TR at m.o.i. 20 completely lost the TGF-␤-induced antiproliferative effect even in the presence of a high concentration of TGF-␤1 (200 pM) (Fig. 3). This antiproliferative effect of TGF-␤1 and its abrogation by infection with AdexCAT␤TR were also observed with both TGF-␤2 and -␤3 (data not shown).

Inhibition of TGF-␤-induced Transcriptional Activation by Adenovirus-mediated Transfer of the Truncated Type II Receptor-It is well known that TGF-␤ stimulates the production of extracellular matrix proteins, which may lead to fibroproliferative disorders (3). To examine whether infection with
AdexCAT␤TR could abolish the TGF-␤-stimulated increase in transcription, we quantified the static levels of mRNA of PAI-I (Fig. 4A) and fibronectin (Fig. 4B) in EC. The levels of mRNA for both PAI-I and fibronectin were substantially increased in response to TGF-␤1 (100 pM) in EC either uninfected or infected with Adex1w. However, cells treated with AdexCAT␤TR became completely unresponsive to TGF-␤1 even at a higher concentration of TGF-␤1 (300 pM) (Fig. 4).

Inhibition by AdexCAT␤TR of the TGF-␤-induced Accumulation of Extracellular Matrix Protein in SMC-
To confirm the inhibitory effects of AdexCAT␤TR on the production of extracellular matrix at the protein level, we examined by immunocytochemistry the level of type I collagen in SMC that had been i. as indicated and subjected to Northern analysis using a 32 Plabeled 5Ј end of the type II receptor cDNA as a probe. The corresponding RNAs (0.9 kb) were detected by autoradiography. The same membrane was also probed with a cDNA coding GAPDH (1.3 kb) as an internal control to show the references of loaded RNA. B, cell lysates were harvested 72 h after infection with AdexCAT␤TR at various m.o.i. as indicated, subjected to SDS-polyacrylamide gel electrophoresis and transferred to a membrane. The membrane was probed with a monoclonal antibody against the HA epitope, and the signal was visualized using an alkaline phosphatase-conjugated anti-mouse IgG and chromogenic reagents. Molecular markers are in kilodaltons.

FIG. 2. Full-length bovine type I and type II and truncated human type II TGF-␤ receptor mRNA in bovine EC infected with AdexCAT␤TR.
A, polyadenylated RNA isolated from bovine EC infected with AdexCAT␤TR at various m.o.i. as indicated was probed with a 5Ј end of the human TGF-␤ type II receptor cDNA. Exposure time for autoradiography was 1 h. B, the membrane corresponding to the full-length bovine TGF-␤ type II receptor mRNA in A was exposed to a film for a prolonged period of time (16 h). C, the same polyadenylated RNA as in A were hybridized with the human TGF-␤ type I receptor cDNA, and a film was exposed for 16 h. Signals of both types of the full-length receptor (B and C) were normalized to their mRNA levels of both GAPDH and ␤-actin (B) or of GAPDH (C), and the values relative to the normalized value obtained from uninfected cells were shown in the bottom column (B and C). Two other independent experiments gave similar results. stimulated with TGF-␤1 (Fig. 5, B-E). In SMC infected with Adex1w as well as in intact SMC, the immunoreactivity for type I collagen observed mainly in the cytoplasm as fine granules, was significantly enhanced by treatment with TGF-␤1 (Fig. 5, B and C). However, in SMC exposed to AdexCAT␤TR, the immunoreactivity (Fig. 5D) was the same as that in quiescent cells (Fig. 5A), even in the presence of additional TGF-␤1. Cells stimulated with TGF-␤1 but not treated with a primary antibody showed no staining (Fig. 5E).
To confirm this further, we quantified the production of extracellular proteins by measuring incorporation of [ 14 C]proline into SMC. TGF-␤-stimulated incorporation of [ 14 C]proline into both collagenase-digestible protein (Fig. 6A), which represents collagen synthesis, and total protein (Fig. 6B) was completely abolished in SMC that had been infected with AdexCAT␤TR at m.o.i. 20.
Unaffected Signaling by Other Growth Factors in Cells Infected with AdexCAT␤TR-Finally, we examined whether infection with AdexCAT␤TR might affect signaling mediated by other growth factors. DNA synthesis in response to either PDGF-BB or bFGF was measured in Balb/3T3 cells that had been infected with AdexCAT␤TR. As shown in Fig. 7, neither PDGF-BB-nor bFGF-stimulated DNA synthesis was affected in cells infected with AdexCAT␤TR, indicating that AdexCAT␤TR specifically inhibits TGF-␤-mediated signal transduction but does not affect signaling by other growth factors.

DISCUSSION
For elucidating the roles of growth factors and cytokines in pathophysiological processes that develop in mature adults, the usefulness of targeted gene-disruption through a homologous recombination would be limited, because such mutations either have lethal effects at early embryonic stages (more often the more critical the molecules are) or show no effects if the target molecule belongs to a multi-gene family such as TGF-␤, which comprises at least three isoforms encoded by different genes. It has been demonstrated that a kinase-defective mutated form of receptor specifically abolished the receptor-mediated signaling by PDGF (31,32), FGF (33)(34)(35)(36)(37), epidermal growth factor (38), insulin (39,40), and vascular endothelial growth factor (41), for example. Based on these findings, in this study we constructed a recombinant adenovirus expressing a human truncated, kinase-defective type II TGF-␤ receptor (AdexCAT␤TR) and examined whether AdexCAT␤TR could serve as a dominant-negative mutation using primary cultures derived from arterial wall, rather than clonal cells that had been transfected and selected or chemically mutated cell lines.
Ligand-induced receptor dimerization and the subsequent autophosphorylation that recruits substrates for the receptor kinases are essential steps in the signal propagation by growth factor receptors with intrinsic tyrosine kinases such as the PDGF, FGF, epidermal growth factor, insulin, and vascular endothelial growth factor receptors. Homodimerization of a single kind of receptor is usual for those receptors, although it has been shown that heterodimers between related receptors can also be formed for signaling (32,34). In contrast, the molecular mechanism underlying the activation of the serine/ threonine kinase receptors for the TGF-␤ superfamily is still not fully established. It has been shown that the type II receptor binds to the free ligand, then forms a heteromeric complex with the type I receptor, phosphorylates, and activates the type I receptor kinase to initiate intracellular signaling (9, 10, 12-14, 16, 42). As a fact, a truncated type II receptor expressed in Mv1Lu cells formed a heteromeric complex with the type I and type II receptors and partially inhibited both the antiproliferative effect of TGF-␤ and transcriptional activation by TGF-␤  A, B, and E). After incubation for 2 days, cells were stimulated with 200 pM of TGF-␤1 for 60 h (B-E). Type I collagen protein was detected by immunocytostaining using a monoclonal antibody against human type I collagen. Specificity of type I collagen immunostaining was shown by lack of immmunoreactivity when primary antibody was absent (E). Cells were lightly counterstained with hematoxylin. The original magnification was ϫ100. either PAI-I or fibronectin (17). This result would seem to suggest that each type of TGF-␤ receptor may have its own distinct signaling pathway, each mediating a separate set of TGF-␤ actions and that the type I receptor-mediated signaling does not require the involvement of type II receptors (17). Similar observations have been made in other settings: 293 cells and some human carcinoma cell lines that lack detectable levels of type II receptors but do possess type I receptors were not responsive to the antiproliferative effect of TGF-␤ but did exhibit a TGF-␤-stimulated synthesis of type IV collagen and/or fibronectin (44,45). Double immunoprecipitation studies using type II receptors tagged with different epitopes showed that the majority of type II receptors are already homomeric even in the absence of TGF-␤ (46, 47). Chen and Derynck (46) proposed a model in which the type II receptor exists as a phosphorylated homodimer or as a heteromeric complex with the type I receptor on the cell membrane. Once the ligand binds, a conformational alteration would occur and critical residues would be exposed to be transphosphorylated by the adjacent kinase of either the type I or type II receptor, and primary substrates may be attached to the receptor kinases for downstream signaling (46). However, using chimeric receptors with different types of TGF-␤ receptor in their extracellular and intracellular regions, it has been shown that a homodimer of either the type I or type II kinase is not sufficient for signaling and that a heteromeric complex, at least in the intracellular region, is essential (48,49).
In summary, it seems likely that both type I and type II receptors should form a heteromeric complex for generating signal; however, it has been controversial in detailed mechanisms. One of the reasons of the disparity between studies may be due to the possibility that the expression level of the truncated receptor was not abundant enough to show its dominantnegative effects fully. In this study, using an adenoviral vector that enables an efficient gene transfer (virtually 100% in culture) and robust expression of the transferred gene, we showed that (i) an overwhelming amount of the truncated type II TGF-␤ receptor was expressed in primary cultures without affecting the mRNA levels of both the type I and type II fulllength receptors (Figs. 1 and 2) and (ii) expression of the truncated type II receptor completely and specifically abolished both the antiproliferative effect (Fig. 3) and the transcriptional activation (Fig. 4) and resultant extracellular matrix accumulation ( Fig. 5 and 6) effect of TGF-␤ in primary culture, as well as in an established epithelial cell line. Molecular interaction between type I and both truncated and full-length type II receptors in the presence of labeled ligand was demonstrated (15), although we did not confirm it in this study. Consequently, the data presented strongly support the hypothesis that both type I and type II TGF-␤ receptors are needed to form heteromeric complexes, which may be tri-or tetrameric (48), for activating all the known signaling pathways of TGF-␤. The study also demonstrated that AdexCAT␤TR would be a useful tool to help clarify the role of TGF-␤ both in vitro and in vivo.