A novel protein distinguishes between quiescent and activated forms of the type I transforming growth factor beta receptor.

Transforming growth factor beta (TGFbeta) signal transduction is mediated by two receptor Ser/Thr kinases acting in series, type II TGFbeta receptor (TbetaR-II) phosphorylating type I TGFbeta receptor (TbetaR-I). Because the failure of interaction cloning, thus far, to identify bona fide TbetaR-I substrates might reasonably have been due to the use of inactive TbetaR-I as bait, we sought to identify molecules that interact specifically with active TbetaR-I, employing the triple mutation L193A,P194A,T204D in a yeast two-hybrid system. The Leu-Pro substitutions prevent interaction with FK506-binding protein 12 (FKBP12), whose putative function in TGFbeta signaling we have previously disproved; the charge substitution at Thr204 constitutively activates TbetaR-I. Unlike previous screens using wild-type TbetaR-I, where FKBP12 predominated, none of the resulting colonies encoded FKBP12. A novel protein was identified, TbetaR-I-associated protein-1 (TRAP-1), that interacts in yeast specifically with mutationally activated TbetaR-I, but not wild-type TbetaR-I, TbetaR-II, or irrelevant proteins. In mammalian cells, TRAP-1 was co-precipitated only by mutationally activated TbetaR-I and ligand-activated TbetaR-I, but not wild-type TbetaR-I in the absence of TGFbeta. The partial TRAP-1 protein that specifically binds these mutationally and ligand-activated forms of TbetaR-I can inhibit signaling by the native receptor after stimulation with TGFbeta or by the constitutively activated receptor mutation, as measured by a TGFbeta-dependent reporter gene. Thus, TRAP-1 can distinguish activated forms of the receptor from wild-type receptor in the absence of TGFbeta and may potentially have a functional role in TGFbeta signaling.


Transforming growth factor ␤ (TGF␤) signal transduction is mediated by two receptor Ser/Thr kinases acting in series, type II TGF␤ receptor (T␤R-II) phosphorylating type I TGF␤ receptor (T␤R-I).
Because the failure of interaction cloning, thus far, to identify bona fide T␤R-I substrates might reasonably have been due to the use of inactive T␤R-I as bait, we sought to identify molecules that interact specifically with active T␤R-I, employing the triple mutation L193A,P194A,T204D in a yeast two-hybrid system. The Leu-Pro substitutions prevent interaction with FK506-binding protein 12 (FKBP12), whose putative function in TGF␤ signaling we have previously disproved; the charge substitution at Thr 204 constitutively activates T␤R-I. Unlike previous screens using wild-type T␤R-I, where FKBP12 predominated, none of the resulting colonies encoded FKBP12. A novel protein was identified, T␤R-I-associated protein-1 (TRAP-1), that interacts in yeast specifically with mutationally activated T␤R-I, but not wild-type T␤R-I, T␤R-II, or irrelevant proteins. In mammalian cells, TRAP-1 was co-precipitated only by mutationally activated T␤R-I and ligand-activated T␤R-I, but not wild-type T␤R-I in the absence of TGF␤. The partial TRAP-1 protein that specifically binds these mutationally and ligand-activated forms of T␤R-I can inhibit signaling by the native receptor after stimulation with TGF␤ or by the constitutively activated receptor mutation, as measured by a TGF␤-dependent reporter gene. Thus, TRAP-1 can distinguish activated forms of the receptor from wild-type receptor in the absence of TGF␤ and may potentially have a functional role in TGF␤ signaling.
The signal transduction events coupling receptors for the TGF␤ 1 superfamily to TGF␤-dependent responses remain incompletely understood. TGF␤ signaling requires two transmembrane receptors acting in series, T␤R-II phosphorylating T␤R-I, each characterized by a cytoplasmic serine/threonine kinase domain (1), but the subsequent signaling pathways are less clear-cut (2). Several candidates have been identified by interaction cloning in yeast "two-hybrid" systems, which interact with the cytoplasmic domain of T␤R-I. These include both the immunophilin-binding protein, FKBP12 (3), a target for the macrolides FK506 and rapamycin, and the ␣ subunit of Ras farnesyltransferase (FNTA) (4,5). T␤R-II-interacting protein 1, a Trp-Asp domain protein, was isolated analogously (6). FKBP12 associates with inactive ALK5 and is released from receptor complexes upon ligand binding (3). Although FKBP12 might inhibit TGF␤ signaling, at least at threshold concentrations of ligand (7,8), mutational analysis by ourselves (9) and others (8,10) has proven that FKBP12 recognition is dispensable for signal generation by T␤R-I; FNTA likewise is unnecessary for TGF␤ signaling (11). However, genetic screening in Drosophila identified the transcription factor, mothers against decapentaplegic (MAD), as acting downstream from the TGF␤ homologue, decapentaplegic (12). In vertebrates, multiple MAD-related proteins exist (Smads) that are thought to mediate signaling by TGF␤ family members via phosphorylationdependent nuclear translocation (13)(14)(15)(16)(17). TGF␤ responses required Smad4 in concert with Smad2 (14) or Smad3 (14,15). Because Smad proteins are substrates for type I receptors (16,17), these transcription factors may provide a direct link between type I receptors and the nucleus. Using genetic complementation in yeast, a novel member of the mitogen-activated protein kinase family, the TGF␤-activated kinase, TAK1, was identified as an alternative mediator of TGF␤ signaling, which may be necessary for at least a subset of TGF␤ effects (18). Thus, functional pathways appear to exist distinct from direct phosphorylation of Smad proteins.
Because the failure of interaction cloning to identify bona fide TGF␤ receptor substrates might reasonably have been due, in part, to the use of inactive receptor as bait, we sought to identify molecules that interact specifically with active T␤R-I, employing T␤R-I L193A,P194A,T204D as the bait. The charged amino acid substitution at Thr 204 confers constitutive activity in the absence of ligand and T␤R-II; disruption of the invariant Leu-Pro motif abrogates binding to FKBP12 (9). Using this strategy we identified a novel protein, TRAP-1, that discriminates between quiescent T␤R-I and T␤R-I that is activated in the presence of TGF␤.

EXPERIMENTAL PROCEDURES
Interaction Cloning and Two-hybrid Assays in Yeast-Constructions containing the cytoplasmic domains of T␤R-I and T␤R-II or point mutations of T␤R-I in the yeast expression plasmid pAS2-1 were previously described (9). The cytoplasmic domain of T␤R-I L193A,P194A,T204D in pAS2-1 was used as bait, to screen a human lymphocyte cDNA library in the pACT vector (CLONTECH). The bait and library DNA were * This work was supported in part by National Institutes of Health Grants R01 HL47567, R01 HL52555, P01 HL49953, and P50 HL42267 (to M. D. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
To obtain a more complete TRAP-1 cDNA, the 3Ј partial sequence was used to screen a human heart cDNA library in gt10 (CLON-TECH); 50,000 plaques were plated per 150-mm dish. Duplicate filters were hybridized with 32 P-labeled TRAP-1 cDNA (Rediprime DNA labeling kit, Amersham Pharmacia Biotech) in 7% SDS, 0.5 M NaH 2 PO 4 , and 1 mM EDTA at 55°C overnight. Filters were washed twice in 2 ϫ SSC and 0.05% SDS at room temperature for 15 min and then in 0.1 ϫ SSC and 0.1% SDS twice at 55°C for 15 min. Filters were subjected to autoradiography. Purification of positive clones was done as described (20). Inserts were subcloned into pBluescript SK (Strategene) and sequenced. The GenBank accession number for TRAP-1 is AF022795.
5Ј-Rapid Amplification of cDNA Ends (RACE)-The 5Ј sequence of TRAP-1 was completed using the 5Ј-RACE reaction. Human heart poly(A) ϩ RNA (CLONTECH) was reversed transcribed using the TRAP-1-specific primer 5Ј-GCC TGT GCT GTA ATT GTG GAT GAT GT-3Ј and Moloney murine leukemia virus reverse transcriptase. Second strand cDNA synthesis was performed using Escherichia coli DNA polymerase I, RNase H, and E. coli DNA ligase (21). Double-stranded cDNA was blunt-ended with T4 DNA polymerase, ligated to the Marathon cDNA adaptor (5Ј-CTA ATA CGA CTC ACT ATA GGG CTC GAG CGG CCG CCC GGG CAG GT-3Ј) (CLONTECH) and amplified by the polymerase chain reaction using a nested internal TRAP-1 specific primer (5Ј-ATG TTG ACC AGG CTG ATG GAG TTG TCA C-3Ј) and an adaptor primer (5Ј-CCA TCC TAA TAC GAC TCA CTA TAG GGC-3Ј). The polymerase chain reaction products were resolved by electrophoresis through 1% agarose. The resulting 460-bp band was excised, purified, subcloned into PCR-script (Strategene), and subjected to DNA sequencing.
Yeast Mating Assay-Yeast strain Y187 (MAT␣) was transformed with wild-type T␤R-I, point mutations of T␤R-I, or T␤R-II plasmids and was grown on SD/-Trp plates. Yeast strain Y190 (MATa) was transformed with TRAP-1 or TRAP-2 plasmids and grown on SD/ϪLeu plates. One colony from each mating type was picked, placed in 0.5 ml of YPD medium (20 g/liter peptone, 10 g/liter yeast extract), and incubated at 30°C with shaking at 250 rpm for 8 h. Mating cultures (15 l) were spread on SD/ϪLeu/ϪTrp and SD/ϪLeu/ϪTrp/-His ϩ25 mM 3-AT plates and incubated at 30°C for 5 days. Growth was scored on SD/ ϪLeu/ϪTrp/ϪHis ϩ25 mM plates. The ␤-galactosidase colony lift filter assay was performed on SD/ϪLeu/ϪTrp plates (22).
Northern Blot Analysis-Filters containing poly(A) ϩ RNA from multiple human tissues (CLONTECH) were prehybridized in ExpressHyb solution (CLONTECH) at 68°C for 30 min and hybridized with a 32 P-labeled TRAP-1 cDNA probe at 68°C for 1 h. Filters were washed in 2ϫ SSC and 0.05% SDS at room temperature for 30 min, washed in 0.1ϫ SSC and 0.1% SDS at 50°C for 40 min, and subjected to autoradiography.
Transcriptional Control-HepG2 hepatocellular carcinoma cells (ATCC) were grown in Dulbecco's modified Eagle's medium (DMEM) containing 20 mM HEPES, pH 7.4, 6 mM NaHCO 3 , 10% fetal bovine serum. For transfection, cells were seeded at 1 ϫ 10 5 cells/24-mm well in 12-well tissue culture dishes. Cells were transfected 36 h after plating by a calcium phosphate method (23), using 100 ng of p3TP-Lux (24), derived from the promoter for plasminogen activator inhibitor-1, 100 ng of the pCH110 ␤-galactosidase reporter gene, driven by the SV40 early promoter (Amersham Pharmacia Biotech) (25), 0 -600 ng of pFlag-⌬TRAP-1-CMV2, and 0 -100 ng of a CMV-driven T␤R-I L193A,P194A,T204D expression vector (9). Cells were fed 0.5 ml of DMEM with 10% fetal bovine serum 1 h before transfection. Calcium phosphate-DNA precipitates were formed by slowly mixing 20 l of 50 mM HEPES, pH 7.1, 280 mM NaCl, Na 2 HPO 4 ⅐7 H 2 0 with 20 l of 125 mM CaCl 2 containing 0.8 g of DNA. Cells were incubated with DNA precipitates (40 l/well) for 5 h and were cultured overnight in DMEM with 10% fetal bovine serum. Medium was replaced on the following day by DMEM with 0.03% fetal bovine serum in the absence or presence or 1 ng/ml TGF␤1 (R & D Systems). Cells were harvested 24 h later, and the luciferase activity and ␤-galactosidase activity were measured. For all comparisons, total DNA and promoter content were kept constant using equivalent amounts of vector. Luciferase activity was corrected for the internal constitutive lacZ plasmid, and results (mean Ϯ S.E.) are expressed relative to expression in vehicletreated, vector-transfected cells (six to nine cultures for each condition, from two or three independent experiments). Results were compared by analysis of variance and Scheffe's test using a significance level of p Ͻ 0.01.

RESULTS
Interaction Cloning of a Novel Protein, TRAP-1, Using Activated T␤R-I as Bait-A human lymphocyte cDNA library was screened using T␤R-I L193A,P194A,T204D as bait in the yeast twohybrid system (2.4 ϫ 10 6 colonies). Unlike previously reported screens using wild-type T␤R-I, where FKBP12 was the predominant protein detected, none of the resulting colonies encoded FKBP12, as expected from the obligatory role of the Leu-Pro dipeptide. Five clones corresponded to FNTA, which previously was shown to interact with both T␤R-I and T204D-T␤R-I (4). Two proteins were novel, designated T␤R-I-associated protein-1 and -2 (TRAP-1 and TRAP-2; Fig. 1). To define the binding specificity of TRAP-1 and TRAP-2, we employed the yeast mating assay to ensure co-introduction of plasmids encoding both chimeric proteins into a single yeast host (22). Yeast Y190 cells containing GAL4-BD-TRAP-1 or GAL4-BD-TRAP-2 plasmids were mated with yeast Y187 cells containing the "bait" (GAL4-AD-T␤R-I L193A,P194A,T204D ) or other test cDNAs, in frame with the GAL4 activator domain. In this assay system, interaction between two hybrid proteins induces two GAL4-dependent genes, HIS3 and lacZ, which allows both growth on histidine-deficient plates and induction of ␤-galactosidase. By both criteria, TRAP-1 and TRAP-2 interacted specifically with T␤R-I L193A,P194A,T204D but not with wild- type   FIG. 1. Features of the presumptive full-length TRAP-1 sequence. Putative phosphorylation sites are denoted for protein kinase C (underlined), casein kinase 2 (double underlined), and protein kinase A (bold); putative myristylation sites are indicated by dotted lines.

TRAP-1 Selectively Binds Active Forms of T␤R-I in Yeast
Two-hybrid Assays-Because three amino acid substitutions were incorporated in our T␤R-I bait, to establish which amino acid(s) conferred this interaction, the associations between TRAP-1 and TRAP-2 and each individual mutation of T␤R-I were tested. TRAP-1 also bound the T␤R-I T204D kinase domain, containing the substitution that activates T␤R-I signaling; TRAP-1 did not interact with T␤R-I L193A or T␤R-I P194A (Fig.  2B). Thus, activation of T␤R-I, not disruption of FKBP12 binding, was responsible for the conditional binding of TRAP-1. In contrast, TRAP-2 only interacted with T␤R-I L193A,P194A,T204D but not T␤R-I T204D or other single mutations. For this reason, TRAP-1 was selected for more extensive analysis.
To obtain the full-length coding sequence of TRAP-1, the 3Ј 1399-bp partial cDNA was used to screen a human heart cDNA library, and the 5Ј sequence of TRAP-1 was obtained by performing 5Ј-RACE on poly(A) ϩ RNA from human heart. The full-length cDNA was 3100 bp in length, comprising 860 amino acid residues (Fig. 1). An in-frame stop codon (TAA) was 153 bp upstream from the ATG start codon and Kozak consensus sequence (26). A search of GenBank using BLAST revealed no homologous proteins; a search for potential functional motifs using the BLAST Enhanced Alignment Utility identified one potential phosphorylation site for protein kinase A, nine for protein kinase C, and eight for casein kinase 2, and seven potential myristylation sites. No other known structural motifs were identified. TRAP-1 mRNA was detected as two transcripts of 4.4 and 6 kilobases in all tissues examined, with lesser abundance in lung and liver (Fig. 3); the mechanism for generating the two transcripts remains to be determined.
TRAP-1 Selectively Binds Ligand-activated T␤R-I in Mammalian Cells-To confirm the prediction that TRAP-1 might also associate specifically with activated T␤R-I in mammalian cells, the receptor-binding portion of TRAP-1 isolated in the yeast two-hybrid screen was epitope-tagged at its N terminus using the FLAG sequence and was cotransfected into human 293 cells together with wild-type T␤R-I or the activated mutation, T␤R-I L193A,P194A,T204D , each incorporating the Hemophilus influenzae hemagglutinin (HA) tag. Two days following transfection, the cells were incubated with TGF␤1 or the vehicle. To recover TRAP-1 itself, plus associated proteins, lysates were immunoprecipitated with antibody to the FLAG epitope  4. TRAP-1 specifically binds ligand-activated T␤R-I. Human 293 cells transfected with epitope-tagged T␤R-I and ⌬TRAP-1 constructs were incubated with vehicle versus TGF␤1. Cells were lysed, antibody to the FLAG epitope was added to immunoprecipitate (IP) TRAP-1 and associated proteins, immunoprecipitates were resolved by SDS-polyacrylamide gel electrophoresis, and TRAP-1-associated T␤R-I was detected with anti-HA antibody. Wt, wild type.

FIG. 5. A receptor-binding fragment of TRAP-1 inhibits TGFB signaling.
HepG2 cells were transfected with p3TP-Lux, the constitutive lacZ gene, and increasing amounts of the ⌬TRAP-1 expression vector. p3TP-Lux expression was triggered using the constitutively activated receptor, T␤R-I L193A,P194A,T204D (A) or 1 ng/ml TGF␤1 without exogenous receptor (B). Results in A and B are shown relative to expression in vector-transfected and vehicle-treated cultures (n ϭ 9 and 6, respectively). * and **, p ϭ 0.0001 and p ϭ 0.0002, in comparison with the absence of TRAP-1.
FIG. 2. TRAP-1 associates selectively with activating mutations of T␤R-I. The specificity for interaction of TRAP-1 and TRAP-2 with T␤R-I, mutations of T␤R-I, and control proteins was assessed by the yeast mating assay. Yeast strain Y190 was transformed with TRAP-1 or TRAP-2 in pAS2-1 and was mated with yeast strain Y187 carrying wild-type or mutant T␤R-I, T␤R-II, lamin C, or p53 in pACT2. In A and B, mating cultures were plated on nonselective medium (SD/ϪLeu/ϪTrp, a and d) to confirm the presence of both plasmids and on selective medium (SD/ϪLeu/ϪTrp/ϪHis/ϩ3-AT; b and e) to assess protein-protein interaction by two-hybrid induction of the HIS3 gene; lacZ activity was determined by a colony lift X-gal assay on filter replicas of the SD/ϪLeu/ϪTrp plates (c and f).
A Novel TGF␤ Receptor-associated Protein 9367 and then were blotted with antibody to the HA tag (Fig. 4). The constitutively activated receptor, T␤R-I L193A,P194A,T204D , was coprecipitated with TRAP-1 even in the absence of TGF␤1. However, wild-type T␤R-I-HA was coprecipitated only from TGF␤-treated cells. This preferential binding to the ligandactivated receptor contrasts markedly with FKBP12 and farnesyltransferase, both of which were released from T␤R-I in the presence of TGF␤1 (5). Thus, unlike these proteins, TRAP-1 is recruited selectively to the active forms of T␤R-I. TRAP-1 Can Inhibit TGF␤-dependent Transcription-To ascertain whether binding of TRAP-1 to the activated receptor might have functional consequences, HepG2 cells were transfected with the TGF␤-responsive p3TP-Lux reporter gene, an internal lacZ control, and 0 -600 ng of the partial TRAP-1 expression vector (Fig. 5). T␤R-I L193A,P194A,T204D was sufficient to up-regulate expression more than 30-fold (33.5 Ϯ 3.4; p ϭ 0.0001 versus vector-transfected cells). Co-transfection with increasing amounts of the ⌬TRAP-1 expression vector progressively decreased the induction of p3TP-Lux. Using 600 ng of this TRAP-1 construct, only 12-fold induction was seen (12.2 Ϯ 1.4; p ϭ 0.0001 versus the absence of TRAP-1). Analogous results were seen following stimulation of endogenous type I receptor with 1 ng/ml TGF␤1. Expression increased to 6.17 Ϯ 0.46, relative to vehicle-treated cells (p ϭ 0.0001), and this induction likewise was inhibited by ⌬TRAP-1 (2.51 Ϯ 0.24, using 600 ng; p ϭ 0.0001 versus the absence of TRAP-1). Thus, a partial TRAP-1 cDNA, which is sufficient to recognize the activated receptor selectively, can function as an inhibitor of TGF␤-dependent gene transcription. DISCUSSION Ligand-induced receptor homodimerization, causing reciprocal cross-phosphorylation in trans, is a strategy commonly employed by receptor tyrosine kinases for transmembrane signaling (27). Phosphorylated tyrosine residues of the receptor, in turn, recruit SH2 domain-containing molecules that are required for signaling. For receptor serine/threonine kinases and the TGF␤ superfamily, the signal transduction pathway is similar but distinct; TGF␤ ligands induce the heterodimerization of T␤R-II and T␤R-I, whereupon T␤R-II phosphorylates T␤R-I on serine and threonine residues of its Gly/Ser-rich domain. This directional phosphorylation event and the kinase activity of both receptors functioning sequentially are required for signal generation (1). Conversely, we and others have shown that negatively charged amino acids can serve as surrogates for the phosphorylation of T␤R-I and constitutively activate the receptor even in the absence of T␤R-II (9,28). According to this model, T␤R-I, being the substrate for T␤R-II, should interact with downstream mediators and effectors. Therefore, it has been discouraging that interaction cloning using the T␤R-I cytoplasmic domain as bait has not succeeded to date in identifying specific molecules that confer the T␤R-I signal (3,5,7,29). Based on precedents with receptor tyrosine kinases, one foreseeable explanation for this failure may be that some effector molecules acting downstream of T␤R-I might associate preferentially with phosphorylated (active) T␤R-I rather than with the unphosphorylated (inactive) T␤R-I used as bait in previously reported cloning efforts. Here, we have used T␤R-I L193A,P194A,T204D as bait, which has two potentially advantageous properties for this purpose: being autonomously active and not binding FKBP12. Beyond its ability to distinguish activated from wild-type T␤R-I in yeast two-hybrid assays, the novel protein, TRAP-1, binds only mutationally activated or ligand-activated T␤R-I in mammalian cells not the inactive type I receptor, reminiscent of the signaling tactics used by receptor tyrosine kinases. This novel protein is selec-tive in binding active forms of T␤R-I, which favors its involvement in the TGF␤ cascade. Similar binding specificity recently was demonstrated for Smad7, which preferentially associates with active T␤R-I, preventing receptor association with and phosphorylation of Smad2 (30). Thus, intracellular antagonists as well as mediators might be identified by use of the activated receptor as bait. Indeed, at least for Smad proteins, only the inhibitory forms like Smad7 have been shown to bind stably to the activated receptor. In keeping with this model, the Cterminal portion of TRAP-1 (which is sufficient for stable interaction with the activated receptor both in yeast and in mammalian cells) was found to inhibit TGF␤ signaling, as measured by the p3TP-Lux reporter gene. This does not preclude a more complex role for the full-length protein, which we have not yet expressed, or address the status of TGF␤ signaling in the absence of TRAP-1. Further work is required to validate the suggested function of TRAP-1 in T␤R-I signaling via Smad, TAK1, or an alternative pathway.