Transforming Growth Factor-β (TGF-β)-induced Down-regulation of Cyclin A Expression Requires a Functional TGF-β Receptor Complex CHARACTERIZATION OF CHIMERIC AND TRUNCATED TYPE I AND TYPE II RECEPTORS

Abstract Transforming growth factor-β (TGF-β) inhibits the proliferation of epithelial cells by altering the expression or function of various components of the cell cycle machinery. Expression of one of these components, cyclin A, is inhibited by TGF-β treatment. We have identified a 760-base pair fragment of the human cyclin A gene promoter that is sufficient to confer TGF-β responsiveness. Using this promoter fragment, we have developed a cyclin A-based luciferase reporter assay that quantitates the growth inhibitory effect of TGF-β in transient transfection assays. This assay was used to determine which domains of the type I (RI) and type II (RII) receptors were required for the antiproliferative effect of TGF-β. In parallel, the functionality of chimeric receptors, between RI and RII (RI-RII or RII-RI), was tested for TGF-β effect on gene expression using a reporter assay based on the plasminogen activator inhibitor type 1 (PAI-1) promoter. We found that TGF-β-induced inhibition of cyclin A expression was absent in RI or RII-deficient Mv1Lu cells and that this response was restored by expression of wild-type type I or type II receptors in these cells. Furthermore, expression of a single chimeric receptor, either RI-RII or RII-RI, did not confer cyclin A regulation by TGF-β. However, expression of two reciprocal chimeras (RI-RII and RII-RI) resulted in growth inhibition, similarly to wild-type receptors. In addition, chimeric receptors as well as mutant receptors with a deleted cytoplasmic domain and kinase-negative receptors inhibited TGF-β responsiveness in the cyclin A reporter assay in a dominant negative fashion. Finally, in both receptor types, the juxtamembrane domain preceding the kinase domain was essential for receptor function but the cytoplasmic tail was dispensable. Our results suggest that a functional TGF-β receptor complex is required for TGF-β-dependent down-regulation of cyclin A gene expression and illustrate the identical receptor requirements for TGF-β-induced growth inhibition and gene expression.

Transforming growth factor-␤ (TGF-␤) 1 is a secreted protein that induces a range of cell responses which affect growth and differentiation (for recent reviews see Refs. 1 and 2). TGF-␤ treatment of epithelial cells results in cell cycle arrest in late G 1 coincident with an inhibition of synthesis and/or activity of some cyclins and cyclin-dependent kinases (3)(4)(5)(6). In mink lung epithelial cells, TGF-␤ decreases the activities of two G 1 cyclindependent kinases, cdk2 and cdk4 (4,6,7). TGF-␤ down-regulates the synthesis of cdk4 and constitutive expression of cdk4 results in resistance to TGF-␤-induced growth arrest, suggesting that cdk4 is a major downstream target in TGF-␤-induced growth inhibition (4). The inhibition of cdk2 activity by TGF-␤ may involve the cdk inhibitor p27, which prevents complex formation of cyclin E/cdk2 as well as cyclin D/cdk 4 (8 -10). Another cdk inhibitor p15, that is TGF-␤-inducible, interferes with the cyclin D/cdk 4 complex in keratinocytes (11).
While down-regulation of cdk4 and activation of the cdk inhibitors correlates with TGF-␤-induced cell cycle arrest, other components of the cell cycle machinery are also inhibited by TGF-␤. Indeed, late G 1 arrest by TGF-␤ is associated with decreased synthesis of cdc2 (12) and cyclin A (13)(14)(15)(16)(17). Cyclin A plays a critical role in cell cycle progression by complexing with and regulating the activities of cdc2 and cdk2. Cyclin A/cdc2 and cyclin A/cdk2 complexes also interact with other proteins involved in the regulation of the G 1 to S transition, including pRB, the pRB-like proteins p107 and p130, and the transcription factor E2F (18 -22). Inhibition of pRB hyperphosphorylation by cyclin A/cdk2 may be involved in control of G 1 progression by TGF-␤ (23).
In addition to its antiproliferative effect, TGF-␤ induces the expression of many genes, including those for extracellular matrix proteins. In fact, the induction of plasminogen activator inhibitor type 1 (PAI-1) expression is often used as a biochemical marker for TGF-␤ responsiveness. Induction of PAI-1 expression can be easily measured using a reporter plasmid in which luciferase expression is driven from the PAI-1 promoter in a TGF-␤-dependent manner (24,25). This assay is often used in combination with transient transfection experiments to measure cellular responsiveness to TGF-␤ and the corresponding receptor requirements. In contrast, no similarly convenient reporter assay has been developed to measure the growth inhibitory response of TGF-␤, even though considerable evidence suggests that TGF-␤'s effects on growth inhibition and gene expression result from divergent signaling cascades (for review, see Ref. 26).
The early signaling events that lead to growth inhibition or extracellular matrix production by TGF-␤ are largely unknown. TGF-␤ interacts with a set of cell surface receptors (for review, see Refs. 26,27), including the type I and type II receptors which are essential for signal transduction. Both receptor types belong to an expanding family of transmembrane serine/threonine kinases (26, 28 -38). Several structural features distinguish type I from type II receptors. The extracellular domain of the type I receptor is shorter than that of the type II receptor. The type I receptor also has a highly conserved GS motif (SGSGSGLP) in the juxtamembrane region preceding the kinase domain, which is absent in the type II receptor (for review, see Ref. 26). The type II receptor can bind TGF-␤ by itself, but the type I receptors bind TGF-␤ only when coexpressed with the type II receptor, presumably as a result of a physical interaction of these receptors. To date, only one type II receptor for TGF-␤ has been identified (32), while several type I receptors, i.e. R4/ALK5 (34), Tsk7L (33), TSR1 (31), have been shown to bind TGF-␤ when coexpressed with the type II TGF-␤ receptor. Despite the structural similarity, certain functional differences among the different type I receptors have been found. For instance, only R4/ALK-5 transduces TGF-␤ signals in Mv1Lu epithelial cells (34,39), while Tsk7L mediates TGF-␤ signaling in another epithelial cell line (40).
The type II receptors (41,42), and presumably also the type I receptors, constitutively homodimerize independent of ligand binding. In addition, these two receptor types interact with each other (25,43,44) and form a heteromeric (most likely tetrameric) complex which is stabilized by ligand binding (45). The cytoplasmic domain of the type II receptor is constitutively phosphorylated, independent of ligand binding, due to both autophosphorylation and the activity of cytoplasmic kinases (46,47). In the heteromeric complex, the type II receptor kinase phosphorylates the type I receptor (46,47) and the two types of cytoplasmic domains associate with each other (48), further stabilizing the heteromeric interaction. This heteromeric complex is likely to be the signaling receptor unit which mediates the biological responses of TGF-␤.
Epithelial Mv1Lu cells lacking either type I or type II receptor are completely unresponsive to TGF-␤ to induce growth inhibition and gene expression (49,50). But the TGF-␤ responsiveness can be restored to these mutant cell lines by transfection with the missing receptor type (25,39). These findings support the notion that the two receptors are needed for both types of responses to TGF-␤. However, separate signaling pathways for the two types of responses have been observed in several studies (51)(52)(53). For example, cells in which the type II receptor has been functionally inhibited by a truncated type II receptor are resistant to the antiproliferative effect of TGF-␤, yet still induce the expression of PAI-1 and other gene products (51). Furthermore, a variety of cell lines, some of which contain defective type II receptors (e.g. Ref. 53), are resistant to TGF-␤-induced antiproliferation, but remain able to induce various genes including PAI-1 in response to TGF-␤ (see "Discussion").
To gain insight into the role of the type I and II receptors in TGF-␤-mediated growth inhibition, we have developed a convenient assay to score the antiproliferative effect of TGF-␤ using an expression plasmid in which a TGF-␤-responsive cyclin A promoter segment drives a luciferase reporter gene. We used this reporter system in transient transfection experiments to examine the function of the TGF-␤ receptors in the growth inhibitory response of Mv1Lu cells. Using chimeric and mutant receptors, we have dissected the role of the two receptors in the growth inhibitory response of TGF-␤ and compared the cyclin A reporter assay with the PAI-1 reporter assay with respect to TGF-␤ responsiveness.

Construction of the Cyclin A-Luciferase
Reporter Plasmid-Plasmid pCAL2 was made by fusing a PCR-amplified cyclin A promoter (Ϫ516 to ϩ245; for sequence see Ref. 54) to the luciferase coding sequence in the SstI-HindIII site of plasmid pGL (Promega). The sequences of primers used to amplify the cyclin A promoter from genomic DNA are: GACCGGTGAGCTCCGTGTTAAATAATTTA (sense) and AAGAAGCTTCACTGCTCACGGGAGTG (antisense). We also constructed plasmid pRK␤Gal, which expresses ␤-galactosidase under the control of the cytomegalovirus promoter and can be in-cluded in all transfections as an internal standard to monitor transfection efficiency.

Construction of Mutant and Chimeric TGF-␤ Receptors-
The cDNAs encoding the type II receptor (RII) (55) and type I receptor Tsk7L (designated RI 7L ) (33) were of mouse origin, whereas the type I receptor R4 (designated RI R4 ) cDNA was a rat cDNA (35). Specific domains of RI R4 , RI 7L , and RII were amplified using PCR methodology. In PCR primers listed below, restriction sites were incorporated to allow subcloning and fusion of different receptor domains. The underlined sequences indicate incorporated restriction sites and those in bold type are start or stop codons.
Amplified receptor cDNA sequences were cloned into the cytomegalovirus promoter-driven mammalian expression vectors pRK5F or pRK5M. These expression vectors are derived from pRK5 (56) but contain the sequences for a myc (57) or FLAG epitope tag (58) between the SalI and HindIII sites, respectively (data not shown). Three truncated receptors lacking the entire extracellular/transmembrane domain (ET) were generated. For example, RII(ET) contains residues  corresponding to the ET domain of RII, while RI R4 (ET) and RI 7L (ET) include the first 148 residues of RI R4 or RI 7L . Sequences containing ET domains were cloned between the EcoRI and BamHI sites, in frame with the downstream epitope tag, either myc for RI R4 (ET) and RI 7L (ET) or FLAG tag for RII(ET).
In the hybrid receptors, the cytoplasmic (C) domain of one receptor, as a BamHI-SalI fragment (in case of RI R4 , the BglII site substituted for a BamHI site), was placed between the ET domain of a different receptor and the epitope tag. Considering the large number of plasmids, a detailed description of how they were made is not provided in this report; however, further details of the cloning strategies can be obtained upon request. A total of four chimeric receptors between RI and RII were created from the fusion of ET domain of one type to C domain of another. At the junction of these chimeric receptors, three extra amino acid residues (Gly Ser Met) were created corresponding to the restriction sites (BamHI/BglII and the introduced start codon ATG preceding cytoplasmic domain). The nomenclature of the chimeric receptors reflects the origin of the two fused domains. For example, RI R4 (ET)-RII(C) contains ET domain of RI R4 and C domain of RII receptor. All receptor clones were sequenced to ensure that there were no PCR-induced mutations.
Cell Culture-COS-1 cells (CRL 1650, American Type Culture Collection) were maintained in the Dulbecco's modified Eagle's medium containing 10% fetal bovine serum (FBS) at 37°C in a 5% CO 2 incubator and the medium was changed every 3 days. Mink lung epithelial cells (Mv1Lu; CCL-64, American Type Culture Collection) and the derivative mutant lines DR26 and R1B, provided by Drs. J. Massagué and H. Lodish, were maintained in minimal essential Eagle's medium with Earle's balanced salt solution supplemented with non-essential amino acids and 10% FBS at 37°C in a 5% CO 2 incubator.
Transfection of COS-1 Cells, Immunoprecipitations, and in Vitro Kinase Reactions-COS-1 cells were transfected using LipofectAMINE (Life Technologies, Inc.) according to the manufacturer's instruction. Forty-eight h after transfection, cells were starved for 1 h in methionine-and cysteine-free Dulbecco's modified Eagle's medium and then metabolically labeled overnight in the same medium containing 0.5 mCi Pro-mix (Amersham Corp.). Labeled cells were lysed in MLB lysis buffer (20 mM Tris-HCl, pH 8, 137 mM NaCl, 1% Nonidet P-40) for anti-myc antibody 9E10 (gift of J. M. Bishop), or FLB lysis buffer (25 mM Tris-HCl, 300 mM NaCl, 1% Triton X-100) for anti-FLAG antibody M2 (Kodak). Lysates were precleared in a mixture of rabbit anti-mouse IgG (Jackson Laboratories) and protein A-Sepharose CL4B (Pharmacia), and immunoprecipitated using monoclonal antibodies 9E10 (antimyc) or M2 (anti-FLAG). Immunoprecipitated proteins were subjected to electrophoresis on SDS-PAGE and visualized by autoradiography.
For in vitro kinase assays, immunoprecipitated receptors with specific deletions in the cytoplasmic domains of RI R4 and RII were split into two halves. While half was directly subjected to SDS-PAGE, the other half was used in a kinase autophosphorylation assay. The kinase assay was carried out at room temperature for 30 min in 1 ϫ kinase buffer (10 mM HEPES-KOH, pH 7.5, 5 mM MgCl 2 , and 5 mM CaCl 2 ) containing 5 Ci of [␥-32 P]ATP (5000 Ci/mmol, Amersham) and then stopped by adding equal volume of 2 ϫ SDS sample buffer (80 mM Tris, pH 6.8, 3.2% SDS, 16% glycerol, 200 mM dithiothreitol, 0.02% bromphenol blue).
125 I-TGF-␤ Cross-linking of Receptor Proteins-COS-1 cells were transfected, with expression plasmids containing cDNAs of a single receptor or a combination of two receptors, as described above for immunoprecipitation. Forty-eight h after transfection, COS-1 cells were cross-linked with 125 I-labeled TGF-␤1 (Amersham) as described (59).
Functional Assays-Reporter assays were performed following transient co-expression of receptor and reporter expression plasmids in Mv1Lu cells, either the wild-type or the mutant lines. Reporter plasmid pCAL2 was used to monitor TGF-␤-dependent cyclin A down-regulation and the plasmid p800Luc to assay TGF-␤-induced PAI-1 induction. The latter plasmid, obtained from Dr. D. Loskutoff, contains a luciferase expression unit under the control of PAI-1 promoter (24). Cells (ϳ3 ϫ 10 6 ) were harvested and electroporated at 960 microfarads and 750 V/cm in a Bio-Rad electroporation apparatus, typically using 10 g of pCAL2 or p800luc, 10 g of pRK␤GAL, 10 g of receptor expression plasmid(s), and 10 g of carrier DNA (pBluescript II SKϩ). Usually, 10 g of receptor expression plasmid mix contained 5 g of each receptor construct in co-transfections of two expression plasmids or, when only a single receptor expression plasmid was transfected, 5 g of the mix was pRK5. After electroporation, cells were recovered for 4 h in MEM Eagle's with Earle's balance salt solution supplemented with non-essential amino acids and 10% FBS. The cells were then treated with or without 10 ng/ml TGF-␤1 in the same medium but containing 0.2% FBS. After 24 h (for PAI-1 assay) or 48 h (for cyclin A assay), the cells were harvested and lysed in reporter lysis buffer (Promega), and cell lysates were assayed for luciferase and ␤-galactosidase activities. The luciferase assay was carried out using Analytic Luminescence Laboratory's assay reagents and luminometer monolight 2010. ␤-Galactosidase was assayed in an assay buffer (Promega), and the activity was measured at 420 nm in a spectrophotometer. The luciferase activity which reflects the promoter activity of cyclin A or PAI-1 was normalized to ␤-galactosidase to account for transfection efficiency.
R1B-L17 cells, a highly transfectable line of R1B cells were provided by Dr. J. Massagué. Transfection of L17 cells was carried out using the DEAE-dextran method as described (46) at 40% confluence in 6-well plates. For each transfection, 1 g of receptor expression plasmid or vector control and 3 g of reporter constructs (1.5 g of pRK␤GAL ϩ 1.5 g of luciferase expression plasmid) were used.

Down-regulation of Cyclin A Expression by TGF-␤ in Mv1Lu
Cells-Transcription of the PAI-1 gene has frequently been used as a marker for TGF-␤ responsiveness. A reporter system in which the TGF-␤-inducible PAI-1 promoter drives luciferase expression has allowed the rapid evaluation of TGF-␤ receptor function in transient transfection assays (25). However, TGF-␤ is often primarily studied for its growth inhibitory effect, and considerable evidence suggests that the effects on growth inhibition and gene expression result from different signaling cascades. Unfortunately, no reporter assay has been available to monitor the antiproliferative effect of TGF-␤. Since TGF-␤ arrests epithelial cell proliferation in late G 1 phase, thereby preventing the induction of cyclin A gene expression (17), we developed a cyclin A promoter-based reporter system to monitor the TGF-␤-induced growth inhibition response in transient transfection assays.
As a first step in developing this assay, we identified the 5Ј promoter fragment of the cyclin A gene that mediates down-regulation of cyclin A gene expression in response to TGF-␤. The upstream region of the human cyclin A gene corresponding to nucleotides Ϫ516 to ϩ245 contains many potential regulatory elements (54) and was isolated by PCR-based amplification from human genomic DNA. We linked this promoter fragment to the luciferase coding sequence to generate plasmid pCAL2. The luciferase expression driven by this 760-bp fragment of the cyclin A promoter was compared with the luciferase activity from the 7.5-kilobase pair promoter region (nucleotides Ϫ7300 to ϩ245) in plasmid CD1 (54). The cyclin A reporter plasmid and a ␤-galactosidase expression plasmid pRK␤Gal were transiently transfected into TGF-␤-responsive Mv1Lu cells, and the luciferase activity was measured 48 h after transfection. Inclusion of pRK␤Gal allowed calibration of the luciferase activity against the ␤-galactosidase activity, providing normalization of transfection efficiency. Both cyclin A plasmids pCAL2 and pCD1 were found to express luciferase, and the luciferase activity was even somewhat higher when using plasmid pCAL2 (see 0 h point in Fig. 1A), indicating that this 760-bp cyclin A promoter fragment is functional in Mv1Lu cells.
FIG. 1. A, cyclin A-luciferase reporter assay. Cyclin A-luciferase plasmid pCAL2 or pCD1, together with a ␤-galactosidase expression plasmid pRK␤Gal, were transfected into Mv1Lu cells. Four h after transfection, TGF-␤ was added at 400 pM to the media for the defined times. Luciferase and ␤-galactosidase activity were determined as described under "Materials and Methods." Data are from two experiments with each point determined in triplicates. The vertical bars show the standard deviations. pCAL2 and pCD1 contain the human cyclin A promoter sequence from nucleotide Ϫ516 to ϩ245 and from Ϫ7300 to ϩ245, respectively. B, regulation of cyclin A-luciferase activity by TGF-␤ in wild-type and mutant Mv1Lu cells. Untransfected cells or DR26 cells transfected with RII or R1B cells transfected with RI R4 were treated (ϩ) or untreated (Ϫ) with TGF-␤. Transfections, TGF-␤ treatment (48 h), and reporter assays were carried out as in panel A. RII, type II TGF-␤ receptor; RI R4 , type I TGF-␤ receptor; Mv1Lu, wild-type Mv1Lu cells; DR26, RII-deficient cells; R1B, RI R4 -deficient cells.
To evaluate the TGF-␤ responsiveness of the cyclin A promoter, both luciferase plasmids were transfected into Mv1Lu cells, and the normalized luciferase activity was calculated at defined time points after initiation of TGF-␤ treatment. In Mv1Lu cells transfected with pCAL2, the luciferase activity was down-regulated by TGF-␤ (Fig. 1A) in a manner similar to that in pCD1-transfected cells. This decreased expression from the cyclin A promoter parallels the inhibition of cyclin A expression by TGF-␤ in Mv1Lu (15)(16)(17) and other epithelial cell lines (13). The TGF-␤-induced inhibition of luciferase expression was most evident when TGF-␤ treatment was initiated within 6 h after release from contact inhibition (data not shown), and the cyclin A promoter response was maximal when the cells were exposed to TGF-␤ for 48 h (Fig. 1A). These results indicate that the 760-bp cyclin A promoter fragment in pCAL2 contains the TGF-␤-responsive element(s) and that this cyclin A-luciferase assay can be used as transcriptional reporter system to measure the antiproliferative effect of TGF-␤ in transient transfection assays. Fig. 1A, cyclin A transcriptional down-regulation by TGF-␤ was apparent in wild-type Mv1Lu cells, which possess functional type I (RI) and type II (RII) receptors. To further characterize the requirement of receptors in TGF-␤-dependent cyclin A expression, we investigated whether the luciferase activity driven from the cyclin A promoter in pCAL2 is down-regulated by TGF-␤ treatment in Mv1Lu mutant cell lines lacking functional receptors (49,50). In both the RII-deficient DR26 cells and RI R4 -deficient R1B cells, the cyclin A promoter-driven luciferase expression was unchanged upon TGF-␤ treatment (Fig. 1B). In contrast, reintroduction of RII in DR26 cells or RI R4 , but not RI 7L , in R1B cells restored the response to TGF-␤ in the cyclin A-luciferase assay (Fig. 1B). These results suggest that the TGF-␤-induced growth inhibition requires both RI R4 and RII receptors.

Down-regulation of Cyclin A Expression by TGF-␤ Requires Both Type I and Type II TGF-␤ Receptors-As shown in
The Cytoplasmic Domains of Both Receptors Are Required for TGF-␤ Responsiveness in the Cyclin A-Luciferase Assay-To further examine the role of specific domains of TGF-␤ receptors in TGF-␤-dependent down-regulation of cyclin A expression, we generated chimeras between the extracellular/transmembrane (ET) and cytoplasmic (C) domains of RII and RI R4 or RI 7L as well as mutant receptors with cytoplasmic truncations of RI R4 , RI 7L and RII, as shown in Fig. 2. These mutant receptors were engineered to incoporate a C-terminal myc-or FLAG-epitope tag which allows immunoprecipitation with tag-specific antisera. We first examined whether these chimeric and truncated receptors were efficiently expressed in COS-1 or 293 cells. Following metabolic 35 S-labeling of the cells, the receptors were immunoprecipitated and analyzed by SDS-PAGE followed by autoradiography. All truncated and chimeric receptors were expressed at significant levels and had the expected molecular weights, although degradation products were often apparent (Fig. 3A). According to the mobility on SDS-PAGE gels, RII had the expected size of ϳ80 kDa, and the R4 and Tsk7L type I receptors, expressed by plasmids RI R4 and RI 7L , respectively, had the expected size of ϳ70 kDa. Transfection of the plasmids for the truncated receptors lacking the C domain generated a polypeptide of ϳ35 kDa for RII(ET) and ϳ30 kDa for both RI R4 (ET) and RI 7L (ET). The hybrids between RI (either RI R4 or RI 7L ) and RII comigrated with a size of ϳ80 kDa for the RII(ET)-RI 7L (C) chimeras and ϳ75 kDa for RI(ET)-RII(C) chimeras.
The ability of the mutant receptors to bind TGF-␤ at the cell surface was determined by binding and cross-linking of 125 I-TGF-␤. As shown in Fig. 3B, both RII and RII(ET)-RI(C) chimeras with their type II extracellular domain bound 125 I-TGF-␤ (lanes 2, 7, and 8). In contrast, RI and RI(ET)-RII(C) chimeras which have RI extracellular domains did not bind TGF-␤ when expressed alone (lanes 3, 5, 9, and 11). However, these receptors bound ligand when cotransfected with RII (lanes 4, 6, 10, and 12). The ability of the mutant TGF-␤ receptor to signal was then examined in the cyclin A-luciferase assay using receptor-deficient Mv1Lu mutant cells. Although wild-type RII restored TGF-␤-dependent down-regulation of cyclin A transcription in RII-deficient DR26 cells, none of the chimeric receptors (RI R4 -RII, RI 7L -RII, RII-RI R4 or RII-RI 7L ), when expressed alone, induced transcriptional regulation of cyclin A-luciferase in response to TGF-␤ (Fig. 4, left panel). Similarly, the chimeric receptors did not restore TGF-␤ responsiveness in the RI-deficient R1B cells which contain endogenous RII, even though RI R4 was able to restore the TGF-␤-induced inhibition of cyclin A-luciferase activity to the cells (Fig. 4, right panel). While a single chimeric receptor was unable to confer TGF-␤ responsiveness in the mutant cell lines, coexpression of the two chimeric receptors RII(ET)-RI R4 (C) and RI R4 (ET)-RII(C) resulted in TGF-␤-dependent inhibition of luciferase expression in these cells (Fig. 4). Similarly, coexpression of both chimeric receptors led to TGF-␤-dependent increases of transcription from the PAI-1 promoter in a PAI-1-luciferase assay (data not shown; Table I). While the coexpression of both chimeras in mutant cells restored the TGF-␤-dependent responses, significant levels of cyclin A inhibition and PAI-1 up-regulation were already observed in the absence of TGF-␤ in these cells. This constitutive signaling in the absence of TGF-␤ resulted in the lower relative level of inhibition of cyclin A expression in the presence of TGF-␤, when compared with that in the absence of TGF-␤ (Fig. 4). In fact, TGF-␤-independent signaling was also apparent when the two types of cytoplasmic domains (RII and RI R4 ) were expressed regardless of what extracellular domains they were fused to (data not shown). These results suggest that the high level expression of both cytoplasmic domains is responsi-ble for ligand-independent signaling and that the two receptor types have an intrinsic heteromeric affinity for each other which is consistent with previous findings (48). In contrast to the chimeric receptors derived from RI R4 , cotransfection of RII(ET)-RI 7L (C) with RI 7L (ET)-RII(C) did not lead to TGF-␤ responsiveness in the cyclin A-luciferase assay, consistent with the inactivity of the Tsk7L type I receptor in Mv1Lu cells (31,39).

Dominant Negative Effect of Chimeric and Truncated Receptors on TGF-␤-dependent Down-regulation of Cyclin A Expression-We further characterized the role of RI and RII in TGF-
␤-induced down-regulation of cyclin A expression by examining the function of the chimeras in wild-type Mv1Lu cells which contain endogenous RI and RII. Because of the TGF-␤ responsiveness associated with endogenous receptors, overexpression of wild-type RI or RII did not alter cyclin A transcription. In contrast, expression of chimeric receptors RI R4 (ET)-RII(C), RII-(ET)-RI R4 (C) or RII(ET)-RI 7L (C), but not RI 7L (ET)-RII(C), abolished TGF-␤-dependent inhibition of cyclin A transcription (Fig. 5, left panel), indicating that chimeras with either RI R4 -or RII-derived ET domains inhibited the activities of the endogenous receptors in a dominant negative fashion. Furthermore, in transfections with the RII(ET)-RI R4 (C) or RII(ET)-RI 7L (C) chimeric receptors, the cyclin A activity was even higher after TGF-␤ treatment (resulting in the negative value for the percentage inhibition, Fig. 5). Dominant negative inhibition was also observed with the kinase-deficient point mutants RIIKR and RI R4 KR, in which the Lys in the ATP-binding sites of RII and RI R4 was replaced by Arg (Fig. 5, middle panel). The

FIG. 4. Signaling activity of chimeric receptors in mutant Mv1Lu cells using TGF-␤-dependent cyclin A-luciferase reporter assay.
Receptor expression plasmids were transfected in DR26 or R1B cells together with the cyclin A-luciferase plasmid pCAL2. Cyclin A-luciferase data are shown as the percentage decrease in luciferase activity relative to cells that did not receive TGF-␤. In the cases when the cyclin A activity was higher after TGF-␤ treatment, a negative value for the percentage inhibition was observed. The vertical bars show the standard deviations. Receptors are abbreviated as in Fig. 2 except that hRII represents wild-type human RII.
dominant negative effect of these mutant receptors in wild type Mv1Lu cells was also apparent in the PAI-1-luciferase assay (data not shown; Table I).
Using the TGF-␤-responsive wild-type Mv1Lu cells and the cyclin A-luciferase assay, we also evaluated the effect of expression of truncated receptors lacking the cytoplasmic domain (Fig. 5, right panel). Consistent with the results using a truncated human type II receptor (51,60), expression of the mouse RII lacking its cytoplasmic domain, RII(ET), also blocked TGF-␤-dependent inhibition of cyclin A transcription in a dominant negative manner. Overexpression of the truncated RI R4 (ET) type I receptor lacking its cytoplasmic domain, but not RI 7L (ET), in Mv1Lu cells also abolished TGF-␤ responsiveness (Fig. 5, right panel). This effect was also seen in the PAI-1 assay (Table I). However, the inhibitory effect of RI R4 (ET) was less effective than RII(ET) in the PAI-1 assay (Table I, data not  shown).
Juxtamembrane Domains of Type I and Type II TGF-␤ Receptors Are Required for TGF-␤ Responsiveness-Finally, we evaluated the signaling ability of RII and RI R4 receptor mutants in which the juxtamembrane segment (between the transmembrane and the kinase domain) or the C-terminal tail following the kinase domain, or both, were deleted. We also included kinase-deficient point mutants of RII and RI R4 , i.e. RIIKR and RI R4 KR, for comparison. Expression plasmids for the RII mutants were transfected into DR26 cells and plasmids for the RI R4 mutants into R1B cells. These receptors were then tested for their ability to restore TGF-␤ responsiveness using both cyclin A-and PAI-1-luciferase assays ( Fig. 6; Table I). Both RI R4 and RII with the deletion of the cytoplasmic tail (amino acids 491-501 in RI R4 ⌬T and 546 -567 in RII⌬T) conferred TGF-␤ responsiveness to the corresponding mutant cell lines. In contrast, deletion of the juxtamembrane domains inactivated the signaling activity of both receptor types in the two assays, as shown by transfecting the mutant receptors RI R4 ⌬JM in R1B cells or RII⌬JM in DR26 cells. Accordingly, deletion of both the juxtamembrane and the cytoplasmic tail in both receptors also inactivated both receptor types. As expected, RI R4 KR and RIIKR were also biologically inactive (Fig.  6, Table I).
When expression plasmids for these mutant receptors were transfected into COS-1 cells and the kinase activity of the immunoprecipitated receptors was tested in vitro, we observed a correlation between the receptor kinase activity and their biological functions (Fig. 7). While RI R4 ⌬T and RII⌬T maintained their kinase and signaling activities similarly to wild-type receptors, RI R4 ⌬JM and RII⌬JM, like the kinase-negative mutants RI R4 KR and RIIKR, were also kinase inactive and unable to autophosphorylate. Thus, mutant receptors which were inactive in the biological assays were also inactive in the kinase assay.

Roles of TGF-␤ Receptors in TGF-␤-dependent Down-regulation of Cyclin
A Transcription-During the cell cycle, cyclin A transcription is suppressed in late G 1 phase and activated during S phase (16,17). Thus, cyclin A expression is considered to be an indicator of the cell growth status. Inhibition of cyclin A expression is sufficient to cause cell cycle arrest before G 1 /S transition (61, 62), and deregulated cyclin A expression is ap-  Fig. 2 except that plasmids RI R4 KR and RIIKR represent the kinase-negative mutants of RI R4 and RII, respectively. Data are presented as in Fig. 4.

FIG. 6. Biological effects of deletion of the cytoplasmic segments of TGF-␤ receptors on TGF-␤-dependent down-regulation of cyclin A-luciferase expression.
Cytoplasmic deletion mutants of both RI R4 and RII were as shown in Fig. 2 and in legend to Fig. 5. Data are presented as in Fig. 4.   FIG. 7. Expression and in vitro kinase activity of mutant receptors with cytoplasmic deletions. COS-1 cells were transfected with the plasmids indicated and the receptors were immunoprecipitated from 35 S-lysates as described in Fig. 3A. Half of the immunoprecipitated proteins were analyzed by SDS-PAGE (left panels), whereas the other half was subjected to in vitro kinase assay using [gamma] 32 P]ATP prior to gel analysis (right panels). The positions of the receptors and molecular weight markers are shown. parent in many cancers and transformed cells (14,62,63). Cyclin A expression is down-regulated by many factors, including TGF-␤ (13-17) and contact inhibition (64), which both arrest cell growth in late G 1 . The TGF-␤-dependent downregulation may play a critical role in TGF-␤-induced growth arrest (17). To generate a convenient and rapid reporter system that allows a measurement of growth inhibition by TGF-␤ and other factors, we created a plasmid in which luciferase expression is driven by a 760-bp segment of the cyclin A promoter. Treatment of Mv1Lu cells with TGF-␤ resulted in a rapid inhibition of luciferase expression, indicating that this short sequence confers TGF-␤ responsiveness. Several potential regulatory sequence elements, including four Sp1 sites, one cAMPresponsive element (CRE), two Yi sites, two E2F-binding sites, one AP1 site, and one p53 binding site, are contained within this short promoter sequence (54). Although the CRE sequence (also called activating transcription factor site, or ATF site) is critical in the down-regulation of cyclin A expression during contact inhibition (64), the promoter elements which mediate TGF-␤-induced inhibition of cyclin A expression in cultured cells are not known. However, contact inhibition and TGF-␤mediated growth arrest have several features in common, e.g. p27, an inhibitor of cdk2 and cdk4, is activated in both contactinhibited and TGF-␤-inhibited cells (9), implicating a similar mechanistic basis for down-regulation of cyclin A expression under both conditions.
Based on the TGF-␤ responsiveness of the 760-bp cyclin A promoter element, we have developed a functional assay to monitor cyclin A regulation in response to TGF-␤. This assay can be used in transient transfection assays to allow a functional evaluation of TGF-␤ receptors and other signaling proteins involved in the antiproliferative effect of TGF-␤. In spite of the many TGF-␤-induced responses, transcriptional activation of the PAI-1 gene is frequently used as the only indicator for TGF-␤ signaling and receptor function, primarily because of the availability of a highly sensitive and convenient reporter assay that measures luciferase expression from a TGF-␤-responsive promoter element (25). Before this study, no facile reporter assay has been available to measure the antiproliferative effect of TGF-␤, even though considerable evidence suggests that the growth inhibition and the induction of gene expression by TGF-␤ result from divergent signaling cascades. The availability of a cyclin A luciferase reporter system allowed us to characterize the role of specific domains of TGF-␤ receptors during TGF-␤-mediated growth inhibition using transient transfection assays.
Our results demonstrate that inhibition of cyclin A transcription by TGF-␤ requires both RII and RI R4 receptors. The cyclin A transcription is inhibited by TGF-␤ in Mv1Lu cells which contain endogenous RI R4 and RII receptors but not in the two mutant cell lines lacking either RI R4 or RII. Introduction of wild-type RI R4 into RI-deficient cells or RII into RII-deficient cells restores TGF-␤-dependent cyclin A transcription. In contrast, various mutant receptors including the chimeric receptors RII(ET)-RI R4 (C), RI R4 (ET)-RII(C), RII(ET)-RI 7L (C), and RI 7L (ET)-RII(C) as well as the kinase-negative point mutants RI R4 KR and RIIKR, and receptors lacking the juxtamembrane domain RI R4 ⌬JM and RII⌬JM, are unable to rescue the ability of mutant cells to respond to TGF-␤. However, coexpression of RII(ET)-RI R4 (C) and RI R4 (ET)-RII(C), which allows heteromerization of both extracellular and intracellular domains (48), restores TGF-␤-dependent down-regulation of cyclin A expression in either mutant cell line. In wild-type Mv1Lu cells, individual expression of these mutant receptors inhibits TGF-␤induced cyclin A down-regulation in a dominant negative fashion. These results suggest that cyclin A expression and, thus, the growth inhibitory effect of TGF-␤ need the cooperative interaction between type I and type II TGF-␤ receptors.
Functional Analysis of TGF-␤ Receptor Domains Using Both Cyclin A and PAI-1 Assays-Several recent studies have addressed the functions of the two receptor types (25,34,39) and their cytoplasmic domains in TGF-␤ signaling (46,66). Transfection of RII receptor in DR26 cells or R4/ALK-5 type I receptor (termed RI R4 in this study) in R1B cells restores TGF-␤ responsiveness, suggesting that both receptor types are required for TGF-␤ responsiveness (25,34,39). The ability of the two receptor types to undergo heteromeric interactions and the phosphorylation of the type I receptor cytoplasmic domain by the type II receptor kinase further reinforces the assumption that both cytoplasmic domains are required for TGF-␤ responsiveness (46,47). Thus a heteromeric complex is thought to be the signaling unit for the multiple responses to TGF-␤ (25,44).
In our study, we tested in parallel the TGF-␤-dependent cyclin A response and PAI-1 transcriptional response and did not observe an uncoupling of these responses (see Table I), which is consistent with the proposed requirement of both cytoplasmic domains for full TGF-␤ responsiveness. More specifically, homomerization of cytoplasmic domains, resulting from expression of the RII(ET)-RI R4 (C) chimera in DR26 cells or the RI R4 (ET)-RII(C) chimera in R1B cells, is not sufficient for TGF-␤-induced regulation of cyclin A and PAI-1 expression, suggesting that a functional interaction of the two types of cytoplasmic domains is a prerequisite for TGF-␤ responsiveness. In addition, despite the heteromeric combination of both cytoplasmic domains, homomeric combination of the extracellular/transmembrane domains of type I or type II receptors also fails to restore TGF-␤ responsiveness. In the case of the RI R4 (ET)-RII(C) chimera in DR26 cells which express endogenous RI R4 , this lack of signaling might be due to the inability of the RI R4 extracellular domains to bind ligand in the absence of RII. However, expression of the RII(ET)-RI R4 (C) chimera in R1B cells which have endogenous RII yet lack functional RI R4 also does not activate TGF-␤ responsiveness even though these receptors bind TGF-␤ efficiently. This suggests an active function of the RI R4 extracellular domain in TGF-␤ signaling.
Coexpression of the RII(ET)-RI R4 (C) and RI R4 (ET)-RII(C) chimeras in DR26 or R1B cells results in TGF-␤ responsiveness in both reporter assays, suggesting that the coexistence of both extracellular and cytoplasmic domains in trans provides a functional TGF-␤ receptor unit. TGF-␤ is likely to induce a conformational change in the heteromeric complex of these chimeras in a manner similar to the complex of wild-type RI R4 and RII receptors. The resulting activation promotes the interaction of the receptors with downstream effectors. Our data are in agreement with the recent findings of Okadame et al. (66), who showed the requirement of both cytoplasmic domains for TGF-␤-induced PAI-1 transcription. We extended their findings by using the cyclin A reporter assay and by examining the effects of chimeric receptors in wild-type Mv1Lu cells.
In wild-type Mv1Lu cells, expression of the RI R4 (ET)-RII(C) or RII(ET)-RI R4 (C) chimera inhibits TGF-␤-induced regulation of cyclin A and PAI-1 expression in a dominant negative manner. This inhibition is likely due to the fact that overexpression of the RI R4 (ET)-RII(C) or RII(ET)-RI R4 (C) chimera sequesters endogenous RII and/or RI R4 away from functional RI R4 -RII receptor complexes. The resulting homomeric complexes of chimeric receptors and the heteromeric complexes between endogenous receptors and the chimeras are then unable to transduce the TGF-␤ signals. This dominant negative interference resembles the inhibitory effect of truncated receptors lacking their cytoplasmic domain and the kinase-negative mutant receptors. The inhibition of TGF-␤ responsiveness by RII(ET) is in accord-ance with previous results (51,60,67). A dominant negative inhibition, albeit to a lesser extent, was also apparent with overexpression of RI R4 (ET) in Mv1Lu cells, but not with RI 7L (ET) even though the truncated form of Tsk7L inhibited TGF-␤-induced mesenchymal transdifferentiation in NMuMG cells (40). Thus, overexpression of chimeric, truncated, or kinase-defective receptors derived from RII or RI R4 results in dominant negative interference with TGF-␤ receptor signaling, most likely by interfering with the assembly of functional homo-and heteromeric complexes. After the submission of this manuscript, Vivien et al. (68) and Brand and Schneider (69) reported dominant negative inhibition of TGF-␤ signaling by chimeric and kinase-defective receptors, using assays for TGF-␤-induced gene expression. Their conclusions and our findings based on cyclin A and PAI-1 assays are in agreement; moreover, our cyclin A assay results further extend their observations to TGF-␤-induced growth inhibition.
Finally, we characterized the function of the RI R4 and RII receptor cytoplasmic domains containing defined deletions. Deletion of the C-terminal tail of either receptor did not affect the kinase and signaling activity of RI R4 and RII in both cyclin A and PAI-1 reporter assays. A similar result has been observed in the case of RII using the PAI-1 reporter assay (60). We also demonstrated that deletion of the juxtamembrane domain inactivates the signaling function of both receptor types. This correlates with a loss of kinase activity as measured by autophosphorylation. In the case of the type I receptor, the juxtamembrane domain contains a conserved SGSGSGLP sequence. This sequence is a major site of phosphorylation by the type II receptor, and its deletion has been shown to inactivate TGF-␤ signaling (46). Thus, our results indicate the requirement of the juxtamembrane domains for the function and kinase activity of both receptors. These domains may also represent binding sites for cytoplasmic effector proteins.
Coincident Responsiveness in the Cyclin A and PAI-1 Reporter Assays-In this study we did not observe a divergence of the pathways leading to growth inhibition and gene expression using the cyclin A and PAI-1 assays. This supports the conclusion of Wrana et al. (25) and Bassing et al. (44) who showed that the activities of both RI R4 and RII are required for TGF-␤induced gene expression and growth inhibition. Solely based on these studies, we would conclude that cytoplasmic domains of both RI R4 and RII are required for both types of TGF-␤ responses. However, uncoupling of the two types of responses has been observed in several systems. We have previously shown that stable expression of a truncated RII in Mv1Lu cells abolished the antiproliferative effect of TGF-␤ but not its induction of gene expression (51). Also, a Phe to Ala mutation at position 498 in RII does not affect the induction of transcription by TGF-␤, yet reduces significantly its antiproliferative effect (60). In addition, several cell lines lacking detectable levels of RII expression or with a defective RII display TGF-␤ responsiveness in gene expression assays but are not growth inhibited by TGF-␤ (53,70). Furthermore, various tumor cell lines also lack a growth inhibitory response to TGF-␤ yet respond to TGF-␤ by the induction of gene expression (52). Finally, differentiating osteoblasts undergo alterations in RI and RII expression and ligand binding and concomitantly show phenotypic changes in their responsiveness to TGF-␤ (71). Thus, these studies suggest that both signaling pathways should be considered as distinct and that in some cases, but not all, uncoupling can occur at the receptor level and that specific changes in functionality of TGF-␤ receptors correlate with induction of specific downstream signaling pathways. How all these observations can be reconciled in a simple model is unclear. However, the apparent discrepancy between both sets of observations might be ex-plained by a threshold model. In such a model, both receptor types are required but the TGF-␤-induced growth arrest requires full activity of endogenous RII (and maybe RI receptor), whereas changes in TGF-␤-induced gene expression can be mediated at a much reduced level of receptor activity, i.e. at a low threshold level. For example, overexpression of a truncated RII in Mv1Lu cells (51) strongly interferes with endogenous RII, but the remaining activity might still suffice to confer the transcriptional response to TGF-␤ but not growth inhibition. This threshold model could also explain the diminished activity of RII defective in N-glycosylation (53) and RII mutated at position 498 (60) in the antiproliferative response but not in TGF-␤-induced gene expression. Accordingly, we have observed that some chimeric and kinase-negative mutants derived from RI R4 and RII completely abolished TGF-␤-dependent cyclin A regulation but only reduced TGF-␤-induced PAI-1 transcription (data not shown). Thus, changes in expression or activity of receptors could possibly determine which downstream pathways are activated. Recently, a threshold model has also been proposed for the response of PC12 cells to receptor tyrosine kinase activation (for review see Ref. 72). In PC12 cells, transient versus sustained ERK activation by growth factors has very different consequences for nuclear translocation of ERK, gene expression, and subsequent proliferation or differentiation (72). An additional complexity in the current models for TGF-␤ receptor signaling is imposed by evidence that ligand binding to RI and RII is differentially regulated. This is apparent from the altered receptor binding in differentiating osteoblasts (71) as well as the observation that a mutated, monomeric TGF-␤ binds much less efficiently to type II than type I receptor. This mutated TGF-␤ is still 20% as active as native TGF-␤1 in PAI-1 induction, but it is less than 1% potent as native TGF-␤1 in growth inhibition in Mv1Lu cells (73), which may be consistent with the threshold model. Further assessment of receptor expression patterns, characterization of the cytoplasmic domains of the TGF-␤ receptors, and their interactions as well as the identification of interacting cytoplasmic proteins should bring insight into the mechanism of TGF-␤ receptor signaling.