Interaction of the Transforming Growth Factor- (cid:98) Type I Receptor with Farnesyl-protein Transferase- (cid:97) *

Transforming growth factor- (cid:98) 1 (TGF- (cid:98) 1) is the proto-type of a large family of molecules that regulate a variety of biological processes. The type I (T (cid:98) R-I) and type II (T (cid:98) R-II) receptors for TGF- (cid:98) 1 are transmembrane ser-ine/threonine kinases, forming a heteromeric signaling complex. Recent studies have shown that T (cid:98) R-II is a constitutively active kinase and phosphorylates T (cid:98) R-I upon ligand binding, suggesting that T (cid:98) R-I is the effec-tor subunit of the receptor complex, which transduces signals to intracellular targets. This model has been further confirmed by the identification of constitutively active T (cid:98) R-I that mediates TGF- (cid:98) 1-specific cellular responses in the absence of ligand and T (cid:98) R-II. To investi-gate signaling by TGF- (cid:98) 1, we have sought to isolate proteins that interact with the cytoplasmic region of T (cid:98) R-I. One of the proteins identified was the (cid:97) subunit of far-nesyl-protein transferase (FT (cid:97) ) that modifies a series of peptides including Ras. T (cid:98) R-I specifically interacts with FT (cid:97) in the yeast two-hybrid system. Glutathione S -transferase-T (cid:98) R-I fusion proteins bind FT (cid:97) translated in vitro . T (cid:98) R-I also phosphorylates FT (cid:97) . We further show that the constitutively active T (cid:98) R-I interacted with FT (cid:97) very strongly whereas an inactive form of T (cid:98) R-I did not. These results suggest that FT (cid:97) may be one of the substrates of the activated T (cid:98) R-I kinase.

Transforming growth factor-␤1 (TGF-␤1) 1 is a multifunctional cytokine that is involved in a variety of biological processes such as cell cycle progression, differentiation, adhesion, migration, extracellular matrix deposition, and immunoregulation (1,2). However, the mechanisms whereby TGF-␤1 exerts such pleiotropic effects have been elusive. TGF-␤1 binds to its specific receptors on the cell surface. Three distinct classes of receptors have been identified (2). The type III (T␤R-III) receptor termed betaglycan is involved in presenting ligand to other signaling receptors (3). The type I (T␤R-I) and type II (T␤R-II) receptors belong to a family of transmembrane serine/threonine kinases that include receptors for other TGF-␤1-related molecules such as activins, bone morphogenic proteins, and Mü llerian inhibiting substance (2). T␤R-II binds ligand without T␤R-I but requires T␤R-I to transduce signals. T␤R-I, on the other hand, requires T␤R-II to bind ligand (4). In the case of TGF-␤s and activins, the type II receptors determine ligand binding specificities (5), and the type I receptors specify cellular responses to ligand (6), suggesting that T␤R-I is the downstream component of the receptor system.
Recently a model for activation of the TGF-␤ receptor system has been proposed (7). T␤R-II is a constitutively active kinase. Upon ligand binding T␤R-II recruits and phosphorylates T␤R-I. The transphosphorylation sites in T␤R-I reside between the transmembrane and the kinase domain (8). This region contains a short peptide stretch called the GS domain, which is highly conserved among the type I receptors of the serine/ threonine kinase family. Mutations of certain serines and threonines in this domain impair phosphorylation and signaling activity of T␤R-I (8), indicating that phosphorylation of the GS domain by T␤R-II has a crucial role in TGF-␤ signaling. Replacement of threonine 204, between the core GS domain and the kinase domain, with aspartic acid resulted in constitutively active T␤R-I (8). This mutant has elevated in vitro kinase activity and signals both anti-proliferative and transcriptional responses in the absence of ligand and T␤R-II. These results further support the hypothesis that T␤R-I transmits signals by phosphorylating intracellular substrates.
We employed a yeast two-hybrid system (9) to identify proteins that interact with the cytoplasmic region of T␤R-I. One of the identified clones encoded farnesyl-protein transferase-␣ (FT␣), a component of an enzyme that modifies various proteins such as Ras. T␤R-I was shown to interact with and phosphorylate FT␣ in vitro. The constitutively active form of T␤R-I interacted with FT␣ more strongly than the wild type, and an inactive form of T␤R-I did not bind FT␣ at all. These results indicate that FT␣ may be one of the substrates of activated T␤R-I.

EXPERIMENTAL PROCEDURES
Screening and Interaction Assay-A HeLa cDNA expression library was screened exactly as described (10). Briefly, the yeast strain, EGY48 (9), was transformed with the reporter, pSH18 -34 (9), and the bait, pEGR4 (10), which contains the cytoplasmic region of the rat T␤R-I also called R4 (11). The library was then introduced into EGY48. The transformants were grown on appropriate media, and positive clones were selected depending on ␤-galactosidase activity and leucine prototrophy. Prey plasmids were rescued from EGY48, amplified in bacteria, and sequenced. Interaction assays using the interaction trap (9) were done as described before (10).
Plasmids-Construction of pEGR4 and pEGIIR containing the cytoplasmic region of T␤R-II was described (10). The wild type and mutant forms of the human T␤R-I, called ALK5 (12), were made using the polymerase chain reaction (PCR) and inserted into the yeast expression vector as follows. The internal EcoRI site in the human T␤R-I was removed with the peptide sequence unchanged. The cytoplasmic region * This work was supported in part by National Institutes of Health Grant CA 42572 (to H. L. M.) and grants-in-aid from the Ministry of Education, Science and Culture of Japan (to M. K. and K. M.). 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.
(amino acids 148 -503) was then amplified with an EcoRI site and a XhoI site attached at the 5Ј-and 3Ј-ends, respectively, and subsequently inserted into pJG4 -5(9), yielding pJG-T␤R-I. The bait plasmid was constructed by excising the EcoRI-XhoI fragment from pJG-T␤R-I and subcloning at the same restriction enzyme sites of pEG202 (9) (pEG-T␤R-I). T␤R-I mutants were constructed similarly. T␤R-I(⌬JM) lacks the juxtamembrane region (amino acids 148 -204) of the wild type. T␤R-I(T200V), T␤R-I(T204D), and T␤R-I(K232R) have valine instead of threonine 200, aspartic acid instead of threonine 204, and arginine instead of lysine 232, respectively. The cytoplasmic region of T␤R-II (amino acids 192-567) was subcloned into pJG4 -5 by PCR. The yeast expression plasmids of farnesyl transferase-␣ and -␤ (FT␤) were constructed by subcloning the entire coding region at the EcoRI and XhoI sites of pJG4 -5 and pEG202. All of the PCR products were sequenced. Glutathione S-transferase (GST) fusion protein expression plasmids were made by subcloning the EcoRI-XhoI insert in pJG4 -5 into pGEX-4T-1 (Pharmacia Biotech Inc.). pcDNA3-FT␣, the in vitro expression plasmid of FT␣, was constructed by subcloning the insert of pJG-FT␣ at the EcoRI and XhoI sites of pcDNA3 (Invitrogen). The detail of the subcloning procedures including the primer sequences can be obtained upon request.
In Vitro Binding Assay-GST fusion proteins were prepared as described (13) except that the induction with isopropyl thio-␤-D-galactoside was done at 30°C for 8 h. To synthesize the FT␣ proteins in vitro, the reticulocyte lysate system (Promega) was used. pcDNA3-FT␣ was transcribed with T7 RNA polymerase and translated in the presence of [ 35 S]methionine. Equal amounts of GST fusion proteins (approximately 3.5 g) bound to GST beads (50 l of 50% slurry), and 20 l of in vitro translation product were mixed with 500 l of NETN buffer (20 mM Tris, pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40), incubated at 4°C for 2 h, washed with NETN buffer 4 times, and subjected to SDS-PAGE (10% polyacrylamide gel) and autoradiography.
In Vitro Phosphorylation Assay-GST fusion proteins were released from GST beads by incubation with 15 mM reduced glutathione at 4°C for 15 min. The concentration of the released proteins was assayed according to the Bradford method (Bio-Rad). In the autophosphorylation assay, 0.44 g of released GST or GST-T␤R fusion proteins was incubated in 30 l of the kinase buffer (50 mM Tris, pH 7.5, 100 mM NaCl, 10 mM MnCl 2 , 10 mM dithiothreitol, 0.05% Triton X-100) with 10 Ci of [ 32 P-␥]ATP at room temperature for 30 min. In the transphosphorylation assay, the same amounts of GST-T␤R fusion proteins were mixed with 0.4 g of released GST proteins or GST-FT␣ proteins, preincubated at 4°C for 30 min in the kinase buffer, and incubated in the presence of [ 32 P-␥]ATP as in the autophosphorylation assay. Cleavage of GST fusion proteins was accomplished by incubation with 10 g of thrombin in 200 l of the cleavage buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 2.5 mM CaCl 2 , 1 mM dithiothreitol) at room temperature for 1 h. 0.6 g of cleaved FT␣ proteins was used in the phosphorylation reaction. Samples were subjected to SDS-PAGE (10% polyacrylamide gel) and autoradiography.
Quantitative ␤-Galactosidase Assay-␤-Galactosidase assay in liquid culture was done essentially as described (13). Four independent colonies of each transformation were first grown in appropriate selection media, and then the prey proteins were induced with galactose. ␤-Galactosidase activity was measured with o-nitrophenyl-␤-D-galactoside as substrate.

RESULTS
To search for proteins that interact with T␤R-I, we used the interaction trap screen developed by Brent and co-workers (9). A HeLa cell cDNA library was screened with the cytoplasmic region of the rat T␤R-I (11) as the bait. One of the clones encoded T-ALK, a novel type II serine/threonine kinase receptor (10), which was later shown to be the bone morphogenic protein type II receptor (14,15). Another subset of clones encoded FKBP12, a binding protein for FK506 and rapamycin (data not shown). Ten clones encoded varying portions of the ␣ subunit of farnesyl transferase.
Farnesyl transferase is a heterodimeric enzyme composed of an ␣ and ␤ subunit (16). Prey (pJG-FT␣ or pJG-FT␤) and bait (pEG-FT␣ or pEG-FT␤) plasmids containing the entire coding region of FT␣ or FT␤ were constructed. FT␣ and FT␤ interacted very strongly in the yeast assay (data not shown) as shown in vivo (17). To test the specificity of the interaction between T␤R-I and FT␣, pJG-FT␣ was introduced into EGY48 with pEG202, the bait vector, or a panel of unrelated baits. None of these were positive in the interaction assay (Fig. 1A). In addition, FT␣ did not interact with T␤R-II, and FT␤ did not associate with T␤R-I. Thus the interaction between T␤R-I and FT␣ was specific in the yeast system.
Only one of the FT␣ clones (CL69) contained the entire coding region whereas the others lacked part of the 5Ј-coding sequence (Fig. 1B). To compare the strength of interactions, all of the rescued FT␣ plasmids and pJG-FT␣ were tested in the interaction trap. The ␤-galactosidase production by the clones missing up to the first 45 amino acids of FT␣ was almost the same on X-Gal plates whereas the colony of CL74 starting from the 59th amino acid exhibited a lighter blue color. The color of CL24 starting from the 75th amino acid remained almost white (Fig. 1B). These data suggest that the first N-terminal 45 amino acids of FT␣ are dispensable for its interaction with T␤R-I. Interestingly, deletion of 51 amino acids at the N terminus of FT␣ has been reported to allow normal stabilization of FT␤ and production of enzyme activity, but deletion of 106 amino acids abolished both functions (17).
To confirm the specific interaction of FT␣ and T␤R-I in another system, we used GST fusion proteins of the cytoplasmic region of the TGF-␤ receptors and FT␣ translated in vitro. When FT␣ was translated in a reticulocyte system, two bands were obtained. A minor band with a faster mobility is probably a degradate of the entire protein since FT␣ is very unstable in mammalian cells (17). FT␣ bound to GST-T␤R-I but not to GST itself (Fig. 2). A minimal level of binding to GST-T␤R-II was observed in this assay.
We next tested whether T␤R-I phosphorylates FT␣ in vitro using GST fusion proteins (Fig. 3A).  7). A possibility that T␤R-I may phosphorylate GST through its interaction with FT␣ still remained. Therefore the FT␣ portion (44 kDa) was cleaved from the GST-FT␣ fusion proteins with thrombin (36 kDa) (Fig. 3B). Again FT␣ alone did not show any kinase activity (lane 1). When the cleaved FT␣ was mixed with GST-T␤R-I, a doublet-phosphorylated band the size of FT␣ appeared (lane 2) but not in GST-K232R (lane 4). Similar but much fainter bands were detected in GST-T␤R-II (lane 3). These data show that T␤R-I can phosphorylate FT␣. T␤R-I did not phosphorylate FKBP12 under the same condition (data not shown).
Recently Massagué and co-workers reported mutants of T␤R-I with various levels of signaling activity (8). One of the mutants, in which threonine 204 was replaced with aspartic acid (T204D), showed an elevated level of in vitro kinase activity and functioned as a constitutively active T␤R-I in vivo. Another mutation of threonine 200 to valine (T200V) yielded completely inactive T␤R-I. We tested the interaction of FT␣ with the wild type and mutant forms of T␤R-I using the interaction trap. In the X-Gal plate assay, colonies transformed with T204D showed an intense blue color whereas colonies of T200V remained completely white. The K232R transformants exhibited a faint blue color while the wild type and ⌬JM colonies showed intensity between K232R and T204D (data not shown). A quantitative ␤-galactosidase assay was performed (Table I).
As in plate assay, T204D showed the strongest interaction and T200V did not interact with FT␣. K232R interacted more strongly than T200V but less efficiently than the wild type. ⌬JM interacted more efficiently than the wild type but not as strongly as T204D. FKBP12 did not interact with ⌬JM but did interact with the wild type and all of the other mutants (data not shown). These results indicate that the activated T␤R-I binds FT␣ more efficiently than the wild type or loss of function mutant T␤R-I. DISCUSSION We screened a human cDNA library to identify proteins that bind to the cytoplasmic region of T␤R-I. One of the clones was a novel type II serine/threonine kinase receptor (10) showing that the two-hybrid system is useful in isolating one of the subunits of a membrane-bound receptor complex. The second clone was FKBP12. FKBP12 was reported to specifically associate with T␤R-I in vitro (18). However, its role in TGF-␤ signaling is yet unknown. Here we report FT␣ as another T␤R-I binding protein.
Farnesyl transferase is a heterodimeric enzyme composed of ␣ and ␤ subunits (16). Coexpression of both subunits seems to be necessary for stabilization and activity of the enzyme (17,19,20). The holoenzyme attaches a farnesyl isoprenoid to a cysteine residue (16,21). The CAAX (where A is an aliphatic amino acid and X is any amino acid) sequence is the target motif found at the C-terminal end of all Ras proteins and many other isoprenylated proteins (17,21). The ␣ subunit may perform the catalytic function (17), but the precise enzymatic mechanism is still elusive. Farnesylation plays a pivotal role in the subcellular localization of a variety of proteins including Ras (21,22). Heterologous proteins like protein A (23) and Raf (24) can be targeted to the plasma membrane by adding the C-terminal sequence of K-Ras. Membrane localization is critical for the activity of Ras. Inhibition of farnesylation leads to suppression of Ras-induced responses such as Xenopus oocyte maturation (25), cytoskeletal disorganization (26), cell growth  Interaction of T␤R-I with Farnesyl Transferase-␣ 29630 (25), and tumorigenesis in vivo (27). In Caenorhabditis elegans, the multivulva phenotype resulting from an activated let-60 Ras mutation was suppressed by FT inhibitors (28).
A number of investigations have suggested a relationship between TGF-␤ and Ras. TGF-␤1 treatment caused a rapid increase in GTP-bound Ras in several cell lines (29,30). In another report, however, microinjection of oncogenic Ha-Ras proteins overcame TGF-␤1-mediated growth inhibition and TGF-␤1 decreased the GTP-bound form of Ras (31). Although the effect of TGF-␤1 treatment on Ras status may vary depending on the context (31), the latter results are consistent with the fact that Ras is mitogenic whereas TGF-␤1 predominantly inhibits cell growth. Furthermore, TGF-␤1 inhibited the coupling of Ras to the activation of phosphatidylcholine hydrolysis (32). TGF-␤1 partially antagonized transformed phenotypes, including loss of organized actin cytoskeleton, caused by Ras transformation (33). Intriguingly, farnesyl transferase inhibition also caused actin stress fiber formation and morphological reversion in Ras-transformed cells (26). Thus TGF-␤1 negatively regulates Ras under certain conditions. It was also shown that cellular Ras activity is required for passage through the G 1 phase in Balb/c 3T3 cells while TGF-␤1 inhibits cell growth if added at any point in G 1 (34). These findings suggest that Ras may be directly involved in TGF-␤ signaling.
In the present report, we show that T␤R-I not only binds but phosphorylates FT␣ in vitro. FT␣ binds preferentially to an activated form of T␤R-I. These results indicate that FT␣ may be phosphorylated by T␤R-I upon ligand binding. Phosphorylation of FT␣ has not yet been reported. It is important to study whether TGF-␤1 induces phosphorylation of FT␣ and modulates its activity in vivo.