Interaction of Transforming Growth Factor-β Receptor I with Farnesyl-protein Transferase-α in Yeast and Mammalian Cells

Abstract Transforming growth factor β (TGF-β) signals through two transmembrane serine/threonine kinases, known as TβR-I and TβR-II. Several lines of evidence suggest that TβR-II acts as a primary receptor, binding TGF-β and phosphorylating TβR-I whose kinase activity then propagates the signal to unknown substrates. We report an interaction between TβR-I and the farnesyl-protein transferase-α subunit (FT-α) both in a yeast two-hybrid system and in mammalian cells. These findings raise the possibility that TGF-β might regulate cellular functions by altering the ability of FT-α to catalyze isoprenylation of targets such as G proteins, lamins, or cytoskeletal components. However, we provide evidence that TGF-β action does not alter the overall protein isoprenyl transferase activity in Mv1Lu mink lung epithelial cells. In fact, the β subunits of farnesyl transferase and geranylgeranyl transferase, which are necessary for the activity of FT-α, prevent the association of FT-α with TβR-I. Furthermore, farnesyl transferase activity is shown to be dispensable for TGF-β signaling of growth inhibitory and transcriptional responses in these cells. These results suggest that the interaction between TβR-I and FT-α does not affect the known functions of these two proteins.

TGF-␤ 1 is a multifunctional cytokine that controls proliferation, differentiation, and many other functions in the cell (1)(2)(3)(4). The anti-mitogenic effects of TGF-␤ in particular have attracted much attention. By inhibiting cyclin-dependent kinases, TGF-␤ can override the action of mitogens without directly blocking their signal transduction pathways (5)(6)(7)(8). TGF-␤ initiates signaling at the membrane by contacting two types of transmembrane serine/threonine kinase receptors, the type I and II receptors (T␤R-I and T␤R-II) (9 -12). These receptors belong to a family that also includes receptors for other TGF-␤-related factors such as activins and bone morphogenic proteins (BMPs) (4). Type II receptors bind ligand present in the medium, and this complex associates with and phosphorylates type I receptors (12). Phosphorylation is at serine and threonine residues clustered in the GS domain, a region just upstream of the kinase domain and conserved in all type I receptors, and mutation of these sites blocks TGF-␤ signaling (13). Certain mutations in the GS domain generate a constitutively active T␤R-I that does not require the presence of ligand or T␤R-II for signaling (14). These observations suggest that T␤R-I, acting downstream of T␤R-II, is directly involved in transducing TGF-␤ signals to downstream substrates.
In order to search for proteins that interact with the cytoplasmic domain of T␤R-I, we used protein interaction cDNA cloning in yeast. This led to the identification of farnesyl transferase-␣ (FT-␣) as a T␤R-I interacting protein. FT-␣ is a shared subunit of heteromeric transferases that attach farnesyl or geranylgeranyl moieties to a variety of proteins that play key roles in signal transduction, protein secretion, and cytoskeleton assembly (15,16). We provide evidence that the interaction between T␤R-I and FT-␣ can also take place in mammalian cells. Similar observations were reported by two other groups (17,18) while this paper was in preparation. Additionally, we have analyzed the physiological relevance of this interaction.
Mammalian Expression Vectors and Transfections-The human FT-␣ cDNA (15) was modified by polymerase chain reaction to encode this protein with the Flag epitope sequence at the C terminus and subcloned into the mammalian expression vector pCMV5. Human FT-␤ and GGT-␤ cDNAs in the pcDNA3 vector were generous gifts of Drs. Joseph Goldstein and Michael Brown. The generation of mutant as well as chimeric receptors and reporter constructs had been previously described (14,22). The cell lines COS-1, Mv1Lu, and R-1B/L-17 were cultured and transiently transfected with the indicated vectors (11,22).
Immunoprecipitation and Western Blotting-Precipitation with anti-Flag antibodies (IBI-Kodak) or Ni 2ϩ -NTA-agarose (Qiagen) was done as described (20). Western immunoblotting with anti-Flag or anti-T␤R-I antibodies was done using a 1:2000 dilution of these antibodies, a 1:5000 dilution of secondary antibodies, and the ECL detection system (Amersham Corp.). Monoclonal antibodies against T␤R-I were raised using a juxtamembrane peptide sequence as described previously (10).
NEN) and 100 mM bacterially expressed human Ha-ras was added to cell lysates and incubated for 40 min at 30°C. Samples were analyzed by SDS-PAGE and autoradiography or by precipitation with trichloroacetic acid and counted. The farnesyl transferase inhibitor L-744-832 was a generous gift of Neal Rosen, MSKCC, and Allan Oliff, Merck (25). RESULTS We generated bait constructs for a yeast two-hybrid system (19) using cDNAs encoding the human T␤R-I cytoplasmic domain, either wild type or containing the T204D mutation (Fig.  1A). This mutation elevates the kinase activity of T␤R-I in vitro and endows this receptor with the ability to signal in the absence of TGF-␤ or T␤R-II (14). Screening of a HeLa cell cDNA library with either bait yielded three major classes of cDNAs. Two of these encoded, respectively, the FK506/rapamycin binding protein FKBP12 and the BMP type II receptor BMPR-II, both of which have been previously described as T␤R-I interacting proteins in the yeast two-hybrid system (20,26,27). The third class of cDNAs isolated in these screenings encodes FT-␣. This class accounted for 25% of all clones isolated with T␤R-I as bait, and 40% of those isolated with T␤R-I(T204D) as bait (of over 100 clones analyzed in each case). The isolated FT-␣ clones encoded the full-length protein or lacked no more than 79 amino acids at the N terminus, suggesting that the interaction with T␤R-I requires most of the FT-␣ protein.
FT-␣ interacted weakly or not at all with the cytoplasmic domains of T␤R-II (9), the activin type I receptor ActR-I (11), the mixed specificity type I receptor TSR-I (11), or with empty vector (Fig. 1B). However, FT-␣ interacted strongly with the cytoplasmic domain of the activin type I receptor ActR-IB (Fig.  1B), which has high sequence similarity (90% identity) to that of T␤R-I (22). To analyze the structural requirements of T␤R-I for its interaction with FT-␣, we used baits containing mutations that are known to alter the kinase activity of T␤R-I and its signaling ability in mammalian cells. The (K232R) bait, which contains a mutation that eliminates kinase activity in T␤R-I (13), interacted very weakly with FT-␣, whereas the (T204D) bait interacted with FT-␣ more strongly than the wild type bait (Fig. 1B). Curiously, mutations that eliminate T␤R-II phosphorylation sites in the T␤R-I GS domain and prevent signaling in mammalian cells (13,14) increased the interaction of the T␤R-I bait with FT-␣ (Fig. 1B). A similar result was obtained with a bait containing a full replacement of the GS domain with an unrelated juxtamembrane sequence of T␤R-II (construct T␤R-I (IIGS)) ( Fig. 1B), indicating that the GS domain is not directly involved in the interaction.
In order to investigate these interactions in mammalian cells, FT-␣ was tagged with the Flag epitope sequence at its N terminus and transfected into COS-1 cells alone or in combination with full-length T␤R-I constructs. The latter were tagged at the C terminus with a hexahistidine sequence that binds to Ni 2ϩ -NTA-agarose. An interaction between the transfected T␤R-I and FT-␣ was demonstrated by incubating the cell lysates with Ni 2ϩ -NTA-agarose beads followed by immunoblotting of the bound material using Flag antibody ( Fig. 2A). In contrast to the results in yeast, the interaction with FT-␣ was not detectably affected by the mutations K232R or T204D in T␤R-I. Cotransfection of T␤R-II and addition of TGF-␤ were also without effect on the association between T␤R-I and FT-␣, as judged from the amount of coprecipitated FT-␣ ( Fig. 2A). Metabolic labeling of these transfectants with [ 35 S]methionine and [ 32 P]phosphate followed by precipitation of total FT-␣ with anti-Flag antibody indicated that FT-␣ is a phosphoprotein in the absence of TGF-␤ or cotransfected receptors (Fig. 2B). No change in the ratio of 32 P/ 35 S labeling of FT-␣ was observed in cells cotransfected with the wild type, kinase-defective, or constitutively active T␤R-I constructs or with both T␤R-I and T␤R-II in the presence of TGF-␤ (Fig. 2B). Similar experiments using an untagged FT-␣ construct and precipitation with a FT-␣ monoclonal antibody (28) also yielded no evidence of TGF-␤-induced FT-␣ phosphorylation (data not shown).
The question was raised as to whether T␤R-I recognized FT-␣ as part of a holoenzyme with FT-␤ or GGT-␤. In experiments designed to test this possibility, we observed that cotransfection of the ␤ subunits inhibited the interaction of FT-␣ with T␤R-I (Fig. 3). Transfection of FT-␤ or GGT-␤ generated holoenzyme complexes with cotransfected FT-␣ while it inhibited the association of FT-␣ with T␤R-I. Controls showed that this effect was not due to a decreased expression of FT-␣ or T␤R-I (Fig. 3), suggesting that T␤R-I does not recognize the isoprenyl transferase holoenzymes.
In order to determine the functional consequence of the in- teraction between FT-␣ and T␤R-I, we tested the effect of TGF-␤ on protein isoprenylation. Mv1Lu epithelial cells were used since they are highly responsive to TGF-␤. Cells were metabolically labeled with [ 3 H]mevalonolactone, a biosynthetic precursor of both farnesyl and geranylgeranyl pyrophosphate. Incubation for 8 h with physiological concentrations of TGF-␤ did not modify the level of 3 H incorporation into Triton-soluble proteins detectable under these conditions (Fig. 3A). A labeled protein of 20 kDa (presumably a small G protein; Ref. 23) migrated slightly faster after cell treatment with TGF-␤ (Fig.  4A). T␤R-I has no consensus isoprenylation motif and did not become labeled in [ 3 H]mevalonolactone-labeled cells (data not shown). Farnesyl transferase activity assayed in cell lysates with recombinant Ha-ras as a substrate was similar in control and TGF-␤-treated cells (Fig. 4B). Therefore, TGF-␤ action does not significantly alter the overall protein isoprenyl transferase activity in the cell.  5A). Mv1Lu cells respond to TGF-␤ with increased transcription of plasminogen activator inhibitor-1 (22). Using a luciferase reporter construct (p3TP-lux) that contains the TGF-␤ response region of the plasminogen activator inhibitor-1 promoter (22), the luciferase response to TGF-␤ was not affected by addition of 20 M L-744-832 and was decreased by 40 M L-744-832 but only at the lower (Ͻ10 pM) TGF-␤ concentration range (Fig. 5B). In experiments designed to determine the effect of L-744-832 on the growth inhibitory response to TGF-␤, a partial inhibition of 125 I-deoxyuridine incorporation into DNA observed with 20 M L-744-832 was simply additive to the inhibitory effect of TGF-␤ (Fig. 5C). These results suggest that farnesyl transferase activity is not essential for TGF-␤ signaling in Mv1Lu cells and has little or no participation in the two TGF-␤ responses analyzed here. DISCUSSION We have used a yeast two-hybrid cloning system to search for proteins that interact with the cytoplasmic domain of T␤R-I since this is a downstream component of the TGF-␤ receptor system. Separate screenings of a HeLa cell cDNA library with either the wild type T␤R-I cytoplasmic domain or a constitutively active mutant version yielded multiple isolates of three different classes of clones. One class corresponds to the BMP type II receptor BMPR-II previously shown to interact with T␤R-I in yeast (20,27). Another class corresponds to the FK506-and rapamycin-binding protein FKBP-12 whose interaction with T␤R-I has also been previously documented and remains of unknown significance (26). The third class, as reported here, are clones encoding FT-␣. While this report was in preparation, both Kawabata et al. (17) and Wang et al. (18) using the same approach reported an interaction of T␤R-I with FT-␣ and proposed that this interaction might regulate the activity of the enzyme and explain the antiproliferative effects of TGF-␤. Here we provide evidence for this interaction in mammalian cells and address the question of its physiological significance.
T␤R-I has no consensus isoprenylation motif and was not isoprenylated in our assays. Therefore, it is unlikely that T␤R-I is a substrate of FT-␣. On the other hand, the kinase activity of certain T␤R-I constructs correlates with their ability to interact with FT-␣ in yeast, and similar findings have been made by Kawabata et al. (17). Recombinant T␤R-I kinase is able to phosphorylate FT-␣ in vitro (17). 3 However, some of our evidence challenges the notion that T␤R-I association with FT-␣ simply reflects a kinase-substrate recognition event. Mutations that eliminate ligand-dependent phosphorylation sites in the GS domain actually increase the interaction of T␤R-I with FT-␣ in yeast even though they decrease the kinase activity of T␤R-I and block T␤R-I signaling activity (14). Furthermore, in contrast to their effect on the receptor-FT␣ interaction in yeast,  The overall level of protein isoprenylation in intact cells or the Ras farnesylation activity in cell extracts is not affected by TGF-␤ addition to Mv1Lu cells, a cell line that is strongly growth-inhibited by this factor. TGF-␤ could have altered isoprenyl transferase activity in a transient manner that escaped detection in our experiments. However, this seems unlikely since protein isoprenylation has a relatively long half-life in the cell and is not known to undergo a highly dynamic regulation (16). We also considered the possibility that farnesyl transferase activity might be required for TGF-␤ signaling. This question was investigated with the use of the farnesyl transferase inhibitor L-744-832. This agent inhibits cell proliferation by inhibiting protein farnesylation (25). Yet, at a concentration that blocks cell proliferation, L-744-832 has little effect on the basal or TGF-␤-activated expression of a reported gene in these cells. On the other hand, the growth inhibitory effects of L-744-832 and TGF-␤ in Mv1Lu cells are additive with no evidence of synergy. These results suggest that farnesyl transferase activity does not participate in these TGF-␤ responses.
Significantly, in addition to forming complexes with FT-␣, the cotransfected FT-␤ or GGT-␤ subunits inhibit the interaction of FT-␣ with T␤R-I. Thus, the receptor can recognize isolated FT-␣ but not the holoenzymes and, therefore, may not be associated with isoprenyl transferase activity. It is possible that the receptor acts as a negative regulator of isoprenyl transferases by sequestering FT-␣. However, overexpression of T␤R-I alone does not cause the alterations in proliferation and other cellular functions that might be expected from an efficient sequestration of endogenous FT-␣. Alternatively, FT-␣ might have a high tendency to associate with ␤ subunits and, in their absence, with other proteins that for unknown reasons include T␤R-I. In any case, we find no evidence that T␤R-I or FT-␣ affects the known functions of each other. The hypothesis that Ras activity and cell proliferation are regulated through a direct interaction of the TGF-␤ receptor with farnesyl transferase is not supported by the evidence to date.