A Single Amino Acid Substitution Converts a Transmembrane Protein Activator of the Platelet-derived Growth Factor β Receptor into an Inhibitor*

Background: Certain artificial small transmembrane proteins can activate the PDGFβ receptor to promote cellular growth transformation. Results: A point mutant of one such protein binds to this receptor and inhibits activation induced by different ligands. Conclusion: Small transmembrane proteins can be engineered to inhibit PDGFβ receptor activity. Significance: This strategy could be generalized for designing novel inhibitors of growth factor receptors, which may have therapeutic implications. Receptors for PDGF play an important role in cell proliferation and migration and have been implicated in certain cancers. The 44-amino acid E5 protein of bovine papillomavirus binds to and activates the PDGFβ receptor (PDGFβR), resulting in oncogenic transformation of cultured fibroblasts. Previously, we isolated an artificial 36-amino acid transmembrane protein, pTM36-4, which transforms cells because of its ability to activate the PDGFβR despite limited sequence similarity to E5. Here, we demonstrated complex formation between the PDGFβR and three pTM36-4 mutants: T21E, T21Q, and T21N. T21Q retained wild type transforming activity and activated the PDGFβR in a ligand-independent manner as a consequence of binding to the transmembrane domain of the PDGFβR, but T21E and T21N were severely defective. In fact, T21N substantially inhibited E5-induced PDGFβR activation and transformation in both mouse and human fibroblasts. T21N did not prevent E5 from binding to the receptor, and genetic evidence suggested that T21N and E5 bind to nonidentical sites in the transmembrane domain of the receptor. T21N also inhibited transformation and PDGFβR activation induced by v-Sis, a viral homologue of PDGF-BB, as well as PDGF-induced mitogenesis and signaling by preventing phosphorylation of the PDGFβR at particular tyrosine residues. These results demonstrated that T21N acts as a novel inhibitor of the PDGFβR and validated a new strategy for designing highly specific short transmembrane protein inhibitors of growth factor receptors and possibly other transmembrane proteins.


Receptors for PDGF play an important role in cell proliferation and migration and have been implicated in certain cancers.
The 44-amino acid E5 protein of bovine papillomavirus binds to and activates the PDGF␤ receptor (PDGF␤R), resulting in oncogenic transformation of cultured fibroblasts. Previously, we isolated an artificial 36-amino acid transmembrane protein, pTM36-4, which transforms cells because of its ability to activate the PDGF␤R despite limited sequence similarity to E5. Here, we demonstrated complex formation between the PDGF␤R and three pTM36-4 mutants: T21E, T21Q, and T21N. T21Q retained wild type transforming activity and activated the PDGF␤R in a ligand-independent manner as a consequence of binding to the transmembrane domain of the PDGF␤R, but T21E and T21N were severely defective. In fact, T21N substantially inhibited E5-induced PDGF␤R activation and transformation in both mouse and human fibroblasts. T21N did not prevent E5 from binding to the receptor, and genetic evidence suggested that T21N and E5 bind to nonidentical sites in the transmembrane domain of the receptor. T21N also inhibited transformation and PDGF␤R activation induced by v-Sis, a viral homologue of PDGF-BB, as well as PDGF-induced mitogenesis and signaling by preventing phosphorylation of the PDGF␤R at particular tyrosine residues. These results demonstrated that T21N acts as a novel inhibitor of the PDGF␤R and validated a new strategy for designing highly specific short transmembrane protein inhibitors of growth factor receptors and possibly other transmembrane proteins.
Many aspects of cell behavior are regulated by cell surface receptors. Receptors for PDGF are transmembrane tyrosine kinases, which initiate signaling pathways that affect the proliferation, motility, and survival of fibroblasts, vascular smooth muscle cells, capillary endothelial cells, and neurons. There are two different forms of the PDGF receptor, ␣ and ␤, which differ in their ligand binding affinities and downstream signaling effects (1). When PDGF binds to the extracellular domains of two receptor molecules, it promotes receptor dimerization, which in turn results in autophosphorylation of key tyrosine residues in the cytoplasmic domain of the receptor (2). These phosphorylated tyrosines then recruit specific signaling or adaptor proteins containing Src homology 2 (SH2) 2 domains to the receptor (3). Once bound to the receptor, these proteins are phosphorylated and initiate intracellular signaling cascades culminating in cell proliferation, migration, or survival.
Uncontrolled activation of PDGF receptors has been associated with several cancers including glioblastomas, fibrosarcomas, hematological malignancies, and gastrointestinal stromal tumors (reviewed in Ref. 4). Receptor activation in these tumors is typically driven by activating mutations in the receptor or by an autocrine mechanism in which the receptor and/or ligand is overexpressed. In addition, PDGF produced by tumor cells can act in a paracrine manner and promote the proliferation of tumor blood vessels and stromal cells, which can contribute to tumor growth. Therefore, understanding how PDGF receptors and their downstream signaling pathways are regulated may allow the design of novel therapies for these cancers.
Studies of the 44-amino acid bovine papillomavirus (BPV) E5 oncoprotein demonstrated that the PDGF␤ receptor (PDGF␤R) could be modulated by proteins that target its transmembrane domain. E5 is a dimeric transmembrane protein that interacts with the transmembrane domain of the PDGF␤R and induces its activation by promoting receptor dimerization (5)(6)(7)(8)(9)(10). Sustained activation of the PDGF␤R by E5 results in tumorigenic transformation of mouse fibroblasts and partial transformation of mortal human fibroblasts (10 -12). Extensive mutational analysis combined with molecular modeling suggested that the two proteins interact through a salt bridge between Asp 33 of E5 and Lys 499 of the PDGF␤R, hydrogen bonding between Gln 17 of E5 and Thr 513 of the PDGF␤R, and presumably packing interactions (8,(13)(14)(15)(16)(17)(18)(19)(20)(21). In addition, covalent dimerization of E5 mediated by two C-terminal cysteine residues is required for interaction with the receptor (18,22), although the E5 transmembrane domain itself has intrinsic dimerization potential (23,24).
By screening retroviral libraries expressing hundreds of thousands of small proteins with randomized transmembrane domains, we identified artificial small transmembrane proteins with limited sequence identity to E5 that activate the PDGF␤R and transform cells (25)(26)(27)(28). One such 36-amino acid protein, pTM36-4, possessed a polyhistidine tag at the N terminus followed by 7 consecutive amino acids of E5 and a random 19amino acid sequence consisting of mostly hydrophobic residues (see Fig. 1A) (25). Importantly, this protein lacked the specific amino acids of E5 that are required for activation of the PDGF␤R (namely Gln 17 , Asp 33 , Cys 37 , and Cys 39 ), indicating that there are multiple different avenues to generate small transmembrane activators of the PDGF␤R. pTM36-4 induced morphologic transformation and focus formation of C127 mouse fibroblasts and growth factor independence in BaF3 cells in a manner that required the transmembrane domain of the PDGF␤R (25). Although pTM36-4 was able to homodimerize, we were unable to detect a physical interaction between pTM36-4 and the PDGF␤R. Nevertheless, we speculated that pTM36-4, like E5, binds to the transmembrane domain of the PDGF␤R.
By analogy to Gln 17 in E5, we postulated that one of the hydrophilic amino acids in the transmembrane domain of pTM36-4 might interact with Thr 513 in the PDGF␤R. In this study, we showed that Thr 21 is required for pTM36-4 activity and that replacing this residue with a large hydrophilic residue allowed us to detect a specific physical interaction between the pTM36-4 mutants and the transmembrane domain of the PDGF␤R. One of these mutants, T21N, was not only severely defective for cell transformation, but actually inhibited PDGF␤R activation and its associated signaling, mitogenic, and transforming effects. This study is the first to report a transmembrane protein inhibitor of the PDGF␤R and establishes the feasibility of a genetic approach to isolate similar inhibitors of growth factor receptors and possibly other transmembrane proteins.
Plasmid Constructs and Mutagenesis-The original pTM36-4 ORF was isolated in the pT2H-F13 retroviral plasmid (25). We inserted an in-frame C-terminal HA tag (YPYDVPDYA) onto pTM36-4 in the MSCVhyg retroviral plasmid. T21L, T21A, and T21N substitutions were introduced into the original pTM36-4-p2TH construct by site-directed mutagenesis using the QuikChange method (Agilent Technologies), and T21Q, T21E, and T21N substitutions were introduced into the pTM36-4-HA-MSCVhyg construct. HA-tagged T21Q, T21N, and T21E were also subcloned into the MSCVpuro vector for expression in BaF3 cells. A codon-optimized version of the T21N-HA mutant was cloned into the pBabepuro retroviral vector and designated T21N-HA* (details available from authors on request). pBabepuro constructs expressing E5 or v-Sis and the rat Neu*-LXSN retroviral construct were described previously (11), as was the RV-HYG R (RVY) retroviral construct expressing E5 (29). LXSN constructs expressing wild type or mutant (I506A and T513L) PDGF␤R were generated previously (8,16). The S516L PDGF␤R mutant in LXSN was created by site-directed mutagenesis. The ␤␣␤ chimeric receptor was expressed from LXSN and contains the transmembrane region of the human PDGF␣ receptor (amino acids 524 -528) flanked by extracellular and intracellular domains of the murine PDGF␤R (25). A truncated form of the human PDGF␤R (TPR), which lacks most of the extracellular domain (amino acids 38 -530) was expressed from LXSN and described previously (30).
Stable Expression of Foreign Genes in Cells-Recombinant retrovirus was produced by co-transfecting 293T cells with a retroviral plasmid encoding the gene of interest and the packaging plasmids pCL-Eco and pVSVG, which encodes the vesicular stomatitis virus G protein (Imgenex), as previously described (25). Retrovirus was used to introduce pTM36-4, PDGF␤R, E5, v-Sis, or their derivatives into C127, HFFs, or BaF3 cells as described (11,15). After 7-10 days of drug selection, cell lines stably expressing the desired transgene were established. Cells co-expressing two transgenes were established by sequential retroviral infection and selection.
Focus Forming Assay-Parental C127 cells or HFFs or C127 cells stably expressing T21N, T21N-HA, T21E-HA, or control retrovirus were seeded in 60-mm dishes. When the cells reached 60 -85% confluence, they were infected with control or recombinant retrovirus expressing pTM36-4 or its derivatives, E5, v-Sis, or Neu* at a low multiplicity of infection in the presence of 4 g/ml Polybrene. The next day, the cells were split 1:3 or 1:2 and then maintained at confluence with bi-weekly medium changes using DMEM-10 for C127 cells and MEM ␣-10 for HFFs. At 2-3 weeks after infection, cell monolayers were fixed in methanol and stained with a 5% dilution of a modified Giemsa solution (Sigma-Aldrich) for the visualization of foci. For the experiments shown in Fig. 1B and supplemental Fig. S1, the number of stained foci was normalized for virus titer (determined by counting drug-resistant colonies in parallel cultures).
IL-3 Independence Assay-To determine whether BaF3-derived cell lines could proliferate in the absence of IL-3, 5 ϫ 10 5 cells were washed twice in PBS and then resuspended in 10 ml of RPMI medium containing 1% FBS, 0.05 mM ␤-mercaptoethanol, and antibiotics but lacking IL-3. Cells were then transferred to a T25 flask and incubated at 37°C. At various times thereafter, live cells were counted using a hemacytometer.
DNA Synthesis Assay-C127 cells expressing T21N-HA* or control cells harboring the pBabepuro vector were seeded into a 24-well dish at 3 ϫ 10 4 cells/well. After reaching confluence, the cells were starved by replacing the medium with serum-free DMEM. Two days later, the medium in triplicate wells was replaced with DMEM (untreated) or DMEM containing 2.5 or 5 ng/ml PDGF-DD (R&D Systems) or 10% or 2% FBS. Approximately 24 h later, the medium was replaced with DMEM containing 1.5 Ci/ml [ 3 H]thymidine (PerkinElmer Life Sciences; specific activity, 70 -90 Ci/mmol) in the presence or absence of PDGF-DD or FBS. Five to six hours later, cellular nucleic acid was precipitated by washing three times in cold 10% trichloroacetic acid and then solubilized by heating in 3% perchloric acid at 95°C for 30 min. Acid-precipitable [ 3 H]thymidine incorporated into cellular DNA was measured using a liquid scintillation counter. Statistical analysis of data from multiple trials was performed using the dbplot program developed by Christopher Petti (Mountain View, CA).
Immunoprecipitation and Immunoblotting-To prepare extracts of C127 cells and HFFs, cell monolayers were typically grown to confluence and starved in serum-free medium overnight. For the experiment shown in Fig. 9, cells were either left untreated or treated by adding PDGF-DD directly to the medium (final concentration, 2.5 ng/ml) and incubating for 25-30 min at 37°C. Cell monolayers were washed twice with PBS and then lysed in cold radioimmune precipitation assay-MOPS buffer (20 mM MOPS, pH 7.0, 150 mM NaCl, 1% Nonidet P-40, 1 mM EDTA, 1% deoxycholic acid, 0.1% SDS) supplemented with inhibitors (1ϫ HALT protease and phosphatase inhibitor mixture (Thermo Scientific), 1 mM PMSF, and 0.5 mM sodium metavanadate). For BaF3 cell extracts, ϳ10 7 cells were pelleted, washed once with PBS, and lysed in 1 ml of radioimmune precipitation assay-MOPS buffer as above. Protein concentrations in clarified lysates then were determined using a bicinchoninic acid assay kit (Thermo Scientific).
To immunoprecipitate the PDGF␤R, E5, or T21N/Q-HA, the appropriate antibody was added to extracts at 5-10 l/mg of extracted protein, and the mixture was incubated for 3-16 h at 4°C. Immune complexes were then precipitated using protein A-Sepharose beads, washed, and eluted in 2ϫ Laemmli sample buffer as described previously (11). In Fig. 9B, radioimmune precipitation assay extracts were mixed 1:1 with 2ϫ sample buffer and analyzed directly by immunoblotting.
For PDGF receptor, phosphotyrosine, p85-PI3K, or SHP-2 blotting, samples were boiled, electrophoresed on a SDS-7.5% gel, and transferred to nitrocellulose in transfer buffer (25 mM Tris base, 192 mM glycine, and 20% methanol) containing 0.1% SDS for 1.5 h at 100 V. For phospho-SHP-2, phospho-AKT or phospho-ERK1/2 blotting, samples were electrophoresed on a SDS-10% gel and transferred without SDS to 0.45-m PVDF for 1.2 h. For E5 and HA blotting and the PDGF receptor blots shown in Figs. 3B and 10B, samples were run on a precast 4 -20% or 7.5% polyacrylamide gel (Bio-Rad) and transferred without SDS to 0.2-m PVDF for 1 h. The blots were blocked in blocking buffer (5% nonfat dry milk in TBST (10 mM Tris-HCl, pH 7.4, 167 mM NaCl, 1% Tween 20)) for 1-2 h and incubated in primary antibody overnight. Anti-PDGF receptor, -E5, and -HA antibodies were diluted 1:250, 1:250, and 1:500, respectively, in blocking buffer. All other primary antibodies were diluted in blocking buffer according to the recommendations of the manufacturer. The blots were washed five times in TBST and incubated in a 1:7000 -1:10,000 dilution of HRP-conjugated anti-rabbit or anti-mouse secondary antibody (Jackson ImmunoResearch Labs) or HRP-protein A (GE Healthcare). The blots were washed as above and then detected by ECL. In some cases blots were stripped using Restore Western blot stripping buffer (Thermo) and then reprobed with a different antibody. 21 for Activity-pTM36-4, like BPV E5, requires the transmembrane domain of the PDGF␤R for its transforming activity (25). Alignment of the amino acid sequences of pTM36-4 and E5 revealed that Thr 21 in pTM36-4 aligns with the essential Gln 17 in E5 (Fig. 1A), which is thought to hydrogen bond to Thr 513 of the PDGF␤R (8,13,17). To test whether Thr 21 is important for the transforming activity of pTM36-4, we used site-directed mutagenesis to replace Thr 21 with alanine, leucine, glutamine, glutamic acid, or asparagine to generate T21A, T21L, T21Q, T21E, and T21N, respectively. These mutants were tested for their ability to induce focus formation in C127 cells. As shown in Fig. 1B, T21E and T21N were nearly completely defective for inducing focus formation. The T21A and T21L mutants were also defective (supplemental Fig.  S1A). In contrast, T21Q was ϳ80% as active as pTM36-4. These results indicated that Thr 21 of pTM36-4, like Gln 17 of E5, plays an essential role in its transforming activity.

pTM36-4 Requires Thr
We also tested whether the Thr 21 mutants activated the PDGF␤R. Tyrosine phosphorylation of the endogenous PDGF␤R was assessed in C127 cells stably expressing wild type pTM36-4, one of the Thr 21 mutants, or no transgene (vector control). PDGF receptors were immunoprecipitated from cell extracts and then subjected to immunoblotting using an anti-phosphotyrosine antibody. Like E5, wild type pTM36-4 induced tyrosine phosphorylation of two forms of the PDGF␤R, the mature, 190-kDa, cell surface form and a faster migrating, intracellular precursor form (Fig. 1C). For the E5 protein and its mutants, tyrosine phosphorylation of the precursor form of the receptor is a reliable measure of receptor activation (18,31). Compared with wild type pTM36-4, the T21N and T21L mutants were impaired in their ability to induce tyrosine phosphorylation of both mature and precursor forms of the PDGF␤R, whereas the T21E and T21A mutants were defective for inducing tyrosine phosphorylation of primarily the precursor ( Fig. 1C and supplemental Fig. S1B). In contrast, T21Q induced robust tyrosine phosphorylation of the PDGF␤R precursor ( Fig. 1C), presumably accounting for the transforming activity of this mutant. These results indicated that the ability of pTM36-4 to activate the PDGF␤R requires a threonine or glutamine at position 21.
We also assessed the activity of two of the hydrophilic Thr 21 mutants, T21N and T21Q, in BaF3 cells, hematopoietic cells that do not express endogenous PDGF receptors. These cells require IL-3 for growth, but exogenous expression of the PDGF␤R and an activating protein such as E5, pTM36-4, or the PDGF-B homologue v-Sis can substitute for IL-3 to induce proliferation (25,30). Here, we expressed the murine PDGF␤R in the presence or absence of T21N or T21Q in BaF3 cells and then assayed the cells for IL-3-independent growth. T21Q but not T21N was able to cooperate with the PDGF␤R to confer substantial IL-3-independent proliferation of BaF3 cells (Fig.  1D). Moreover, in BaF3 cells, T21Q but not T21N induced substantial phosphorylation of the PDGF␤R at Tyr 857 , which is required for optimal receptor kinase activity (3) (Fig. 1E). Thus, T21Q but not T21N was able to activate the PDGF␤R to induce growth of these cells, consistent with the activities of these proteins in C127 cells.
Thr 21 Mutants Form a Stable Complex with the PDGF␤R-Because the transmembrane domain of the PDGF␤R is specifically required for pTM36-4 activity, it seemed likely that pTM36-4 either directly or indirectly binds to the transmembrane domain of the receptor (25). However, co-immunoprecipitation analysis failed to detect an interaction between pTM36-4 and the PDGF␤R. We postulated that the interaction between pTM36-4 and the PDGF␤R is not stable enough in detergent extracts to be detected by this method and that replacing Thr 21 with a larger hydrophilic amino acid might strengthen the interaction, thereby allowing its detection. Therefore, we assessed the ability of the hydrophilic substitution mutants of pTM36-4 to form a complex with the PDGF␤R.
To perform co-immunoprecipitation analysis, we took advantage of the influenza HA epitope tag inserted at the C terminus of pTM36-4, T21E, T21N, and T21Q. Detergent extracts were prepared from C127 cells stably expressing HAtagged pTM36-4, T21E, T21N, or T21Q or from control cells harboring an empty expression vector. Wild type or mutant pTM36-4 was immunoprecipitated using an antibody against the HA epitope, and immune complexes were analyzed for the presence of the PDGF␤R by SDS-PAGE followed by immunoblotting. As shown in Fig. 2A, the intracellular precursor form of the PDGF␤R co-precipitated with T21Q-HA but not with wild type pTM36-4-HA (compare lanes 2 and 5). No PDGF␤R co-precipitated if the HA antibody was blocked with a peptide The graph shows the results of multiple trials expressed as the percentage of the number of foci relative to that induced by wild type pTM36-4 (corrected for virus titer), with standard error of the mean. C, Thr 21 mutants differ in their ability to induce tyrosine phosphorylation of the PDGF␤R. The PDGF␤R was immunoprecipitated from extracts of C127 cells stably expressing wild type pTM36-4, the indicated Thr 21 mutant, or empty vector. Immunoprecipitates were subjected to anti-phosphotyrosine (PY) or anti-PDGF receptor (PR) immunoblotting, as indicated. D, ability of the T21Q and T21N mutants to cooperate with the PDGF␤R to induce IL-3 independence in BaF3 cells. The murine PDGF␤R was stably expressed with T21Q, T21N, or empty vector in BaF3 cells. Cells were then cultured in the absence of IL-3 for 7 days, and viable cells were counted. The graph shows results from two trials expressed as the percentage of the number of live cells relative to that in T21Q-expressing cultures, with standard error of the mean. E, T21Q but not T21N activates the PDGF␤R in BaF3 cells. PDGF␤R immunoprecipitates from extracts of BaF3 cells expressing the murine PDGF␤R with T21Q, T21N, or empty vector were immunoblotted using an antibody recognizing phosphorylated Tyr 857 of the PDGF␤R (PY-857). The blot was then stripped and reprobed for total PDGF receptor (PR). In C and E, the numbers on the left indicate the size of molecular mass markers in kilodaltons, and arrows point to the mature (m) and precursor (p) forms of the PDGF␤R. C-terminally HA-tagged pTM36-4 constructs were used in all experiments shown.
comprising the HA epitope ( Fig. 2A, lane 6) or if the cells did not express an HA-tagged protein ( Fig. 2A, lane 1). Thus, the presence of the PDGF␤R in T21Q-HA immunoprecipitates is not due to nonspecific binding of the receptor to the HA antibody or beads. Instead, these results indicated that T21Q and the precursor form of the PDGF␤R exist in a stable complex. Similarly, the precursor form of the PDGF␤R co-precipitated with T21E-HA and T21N-HA ( Fig. 2A, lanes 3 and 4), showing that these mutants also formed a stable complex with the PDGF␤R, even though they did not transform cells. We infer that wild type pTM36-4 also interacts with the PDGF␤R in transformed cells and that the T21Q, T21E, and T21N mutations either stabilize the interaction or facilitate its detection in some other way.
To determine whether the transmembrane domain of the PDGF␤R is required for the interaction with T21Q-HA and T21N-HA, we tested whether these mutants interacted with a chimeric receptor containing the extracellular and cytoplasmic domains of the PDGF␤R and the transmembrane domain of the PDGF␣ receptor (25). First, the chimeric receptor (designated ␤␣␤) or the wild type murine PDGF␤R was expressed in BaF3 cells. The ␤␣␤ receptor was functional in these cells, because it was able to respond to v-Sis and induce IL-3-independent growth (Fig. 2B). Next, HA-tagged pTM36-4, T21Q, T21N, or empty vector was also expressed in these cells and then immunoprecipitated from detergent extracts with the HA antibody. PDGF receptor immunoblotting of the HA immunoprecipitates revealed that the wild type PDGF␤R co-precipitated with T21Q-HA and T21N-HA but not with pTM36-4-HA, showing that T21Q and T21N form a stable complex with the PDGF␤R in BaF3 cells as well as in C127 cells (Fig. 2C, lanes 3 and 7). In contrast, the ␤␣␤ receptor was not detected in T21Q-HA or T21N-HA immune complexes (Fig. 2C, lanes 6 and 8), indicating that this chimeric receptor could not form a stable complex with either T21Q or T21N, even though both proteins were abundantly expressed in these cells (Fig. 2C, bottom panels). Because the ␤␣␤ receptor differs from the wild type receptor by only the sequence of the transmembrane domain, these results indicated that the transmembrane domain of the PDGF␤R is specifically required for a stable complex formation with T21Q and T21N. Consistent with these results, T21Q cooperated with the PDGF␤R but not with the ␤␣␤ receptor to allow BaF3 cells to proliferate in the absence of IL-3 (Fig. 2B). Therefore, the transmembrane domain of the PDGF␤R is required for T21Q to productively interact with the receptor.
Interaction between the PDGF␤R and pTM36-4 Mutants Is Ligand-independent-To determine whether the interaction between the PDGF␤R and T21Q or T21N requires the ligandbinding domain of the receptor, we assessed the ability of these mutants to interact with a truncated form of the human PDGF␤R, TPR, which lacks most of the extracellular domain (30). TPR or the wild type human PDGF receptor was expressed with or without HA-tagged T21Q or T21N in BaF3 cells. PDGF␤R immunoblotting showed that TPR was abundantly expressed in these cells and migrated as a doublet of ϳ75 kDa, which is the expected size of this truncated receptor (Fig. 3A, left panel). T21Q-HA induced abundant tyrosine phosphorylation of TPR, as assessed by anti-phosphotyrosine blotting of PDGF␤R immunoprecipitates (Fig. 3A, right panel). Anti-HA immunoprecipitation followed by PDGF␤R immunoblotting revealed that TPR co-immunoprecipitated with T21Q-HA as well as if not better than the wild type receptor (Fig. 3B). T21N-HA also co-immunoprecipitated with the truncated receptor, but detection of this interaction required the use of a codon-optimized version of T21N and a longer exposure time (Fig. 3B). Finally, T21Q-HA cooperated with TPR to induce robust IL-3-independent proliferation of BaF3 cells, whereas the transformation defective T21N was markedly impaired in this activity (Fig. 3C). Because T21Q was able to engage a receptor lacking the extracellular ligand-binding domain, its ability to productively interact with the PDGF␤R must be ligand-independent.
Identification of PDGF␤R Amino Acids Required for Interaction with pTM36-4 Mutants-To explore the basis for PDGF␤R recognition by the pTM36-4 mutants, we next tested whether the same amino acids in the transmembrane domain of the PDGF␤R were required for a stable interaction with the pTM36-4 mutants and E5. E5, T21Q-HA, or T21N-HA was co-expressed in BaF3 cells with PDGF␤R mutants containing an I506A, T513L, or S516L substitution in the transmembrane domain, and complex formation was assessed by anti-E5 or anti-HA immunoprecipitation followed by PDGF␤R immuno- blotting. Compared with the wild type PDGF␤R, the T513L mutant receptor was defective not only for binding to E5 as expected, but also for interacting with T21Q-HA and T21N-HA (Fig. 4A). This indicated that Thr 513 in the PDGF␤R is required for the interaction with the pTM36-4 mutants, consistent with our proposal that position 21 of pTM36-4 is functionally equivalent to position 17 of E5. On the other hand, the I506A receptor mutant, which was defective for an interaction with E5 as previously reported (16), formed a stable complex with T21Q-HA and T21N-HA (Fig. 4A), suggesting that Ile 506 plays a role in binding to E5 but not in binding to T21Q or T21N. Conversely, the S516L receptor mutant interacted with E5 but not with T21Q-HA or T21N-HA (Fig. 4A), suggesting that Ser 516 plays a role in binding T21Q and T21N but not E5. Consistent with these data, T21Q cooperated with the I506A but not the S516L or T513L receptor mutant to confer IL-3independent growth of BaF3 cells, whereas E5 cooperated with the S516L mutant but poorly with the other two receptor mutants (Fig. 4B). These results provide strong genetic evidence that the ability of E5 and T21Q to confer growth factor independence is dependent on their ability to bind to the transmembrane domain of the PDGF␤R. Furthermore, the PDGF␤R amino acid requirements for binding the pTM36-4 mutants and E5 differ, suggesting that pTM36-4 and E5 bind to overlapping but not identical sites in the transmembrane domain of the PDGF␤R.
T21N Inhibits E5-induced Focus Formation and Activation of the PDGF␤R in Fibroblasts-Although the T21E and T21N mutants interacted with the PDGF␤R, they were defective for inducing PDGF␤R tyrosine phosphorylation and cellular transformation, suggesting that these mutants interacted nonproductively with the receptor. If this is the case, the interaction of these mutants with the PDGF␤R might inhibit its ability to respond to E5 or PDGF.
To test whether T21N and T21E inhibited the activity of E5, we assessed their effect on E5-induced focus formation. We stably expressed T21N-HA and T21E-HA in normal HFFs and C127 cells and then infected these cells with control or E5-expressing retrovirus. Cells were maintained at confluence and monitored for the appearance of transformed foci. As expected, numerous foci appeared in naive HFFs and C127 cells after infection with the E5-expressing virus but not with the control virus (Fig. 5A). Strikingly, the E5 retrovirus induced few if any foci in HFFs or C127 cells expressing T21N-HA and T21E-HA. (These results are quantified in Fig. 5B.) Thus, T21N and T21E inhibited the ability of the E5 protein to induce focus formation in human and murine fibroblasts. Because T21N displayed greater inhibitory activity, the remainder of these studies focused on this mutant.
Next, we examined the effect of T21N on E5-induced activation of the PDGF␤R. The E5 gene was stably expressed in control and T21N-HA-expressing HFFs and C127 cells, and tyro- sine phosphorylation of the receptor was assessed by PDGF␤R immunoprecipitation followed by phosphotyrosine immunoblotting. As expected, the PDGF␤R was abundantly tyrosinephosphorylated in cells expressing E5 alone, but not in cells lacking E5 expression (Fig. 5C). T21N-HA expression caused a substantial reduction in E5-induced PDGF␤R tyrosine phosphorylation in both cell types without affecting overall expression of the receptor (Fig. 5C). Thus, T21N inhibited the ability of E5 to induce tyrosine phosphorylation of the PDGF␤R, as well as focus formation.
T21N Does Not Inhibit the Expression of E5 or the Interaction of E5 with the PDGF␤R in Fibroblasts-To test whether T21N affects E5 expression, extracts were prepared from C127 cells expressing E5 with or without T21N-HA. The extracts were immunoprecipitated with an antibody that recognizes the E5 protein, and immune complexes were immunoblotted with the same antibody. As shown in Fig. 6A (middle panel), T21N expression did not inhibit E5 expression. Thus, decreased E5 expression is not responsible for the inhibition of E5-induced focus formation or PDGF␤R tyrosine phosphorylation. In addition, probing the E5 immunoprecipitates with the HA antibody failed to detect T21N-HA (Fig.  6A, bottom panel, lane 4), suggesting that the E5 protein and T21N-HA do not exist in a stable complex. This conclusion was verified in the reciprocal experiment in which anti-HA immunoprecipitates did not contain the E5 protein, although they did contain T21N-HA and the PDGF␤R precursor, as expected (supplemental Fig. S2, lane 8).
To test whether T21N-HA impairs the binding of the E5 protein to the PDGF␤R, we compared the ability of E5 to coimmunoprecipitate with the PDGF␤R in the presence and absence of T21N-HA. E5 was immunoprecipitated from extracts of control and T21N-expressing C127 cells, and E5 immune complexes were analyzed by PDGF␤R immunoblotting. As expected, both mature and precursor forms of the receptor could be detected in E5 immunoprecipitates from extracts of cells expressing E5 alone but not of cells lacking E5 (Fig. 6A, upper panel, lanes 1 and 2). Abundant amounts of both forms of the receptor were also co-precipitated with E5 from cells that co-expressed T21N-HA (Fig. 6A, upper panel, lane 4). Similar results were observed in HFFs (data not shown). Therefore, T21N did not inhibit the ability of E5 to interact with the receptor. In fact, the amount of receptor present in E5 immune complexes was reproducibly greater in cells expressing T21N-HA (Fig. 6B).
T21N Inhibits v-Sis-induced PDGF␤R Activation-Because T21N inhibits E5 from activating the PDGF␤R without preventing E5 from associating with the receptor, we hypothesized that the inhibitory activity of T21N is not specific to E5 and that T21N would also inhibit the effects of the natural ligands of the PDGF␤R. To test this possibility, we first determined whether T21N inhibited focus formation by v-Sis, the viral homologue of PDGF-BB (32,33). C127 cells stably expressing T21N or control C127 cells harboring an empty vector were infected with retrovirus expressing E5, v-Sis, or Neu*, an activated receptor tyrosine kinase unrelated to the PDGF␤R. Cells were maintained at confluence and monitored for the appearance of transformed foci. As shown in Fig. 7A, few foci were induced by the pBabe retrovirus lacking a transgene. As expected, the E5 retrovirus induced numerous foci in the control cells but not in cells expressing T21N (Fig. 7, A and B), confirming that T21N inhibited E5-induced focus formation. Strikingly, in cells expressing T21N, focus formation by v-Sis was inhibited by ϳ10-fold (Fig. 7, A and B). In contrast, the Neu*-expressing retrovirus induced similar numbers of foci in the control cells and cells expressing T21N, demonstrating that T21N did not inhibit the ability of C127 cells to undergo transformation. These results indicated that T21N specifically inhibits focus formation mediated by the PDGF␤R.
The effect of T21N on v-Sis-induced PDGF␤R activation was also assessed. As expected, the PDGF␤R was heavily tyrosinephosphorylated in cells expressing either E5 or v-Sis alone but not in the control cells lacking a transgene (Fig. 7C). In contrast, T21N substantially reduced PDGF␤R tyrosine phosphorylation in cells expressing E5 or v-Sis without affecting the total level of the receptor (Fig. 7C). We also examined the recruitment of an important signaling substrate, the p85 regulatory subunit of PI3K, to the PDGF␤R in response to v-Sis or E5. As shown in Fig. 7C (bottom panel), E5 or v-Sis increased co-immunoprecipitation of p85-PI3K with the PDGF␤R in control cells, demonstrating that E5 and v-Sis caused the receptor to recruit this substrate. However, E5-or v-Sis-induced recruitment of p85-PI3K to the PDGF␤R was eliminated in cells expressing T21N. Taken together, these results indicated that T21N inhibits v-Sis-induced activation of the endogenous PDGF␤R in C127 cells.
T21N Inhibits PDGF-induced PDGF␤R Activation-We next asked whether T21N inhibited DNA synthesis induced by soluble PDGF-DD. This form of PDGF binds specifically to the PDGF␤R and not the PDGF␣ receptor (34,35). To maximize expression levels of T21N, we constructed a codon-optimized version encoding T21N with a C-terminal HA tag (T21N-HA*). C127 cells stably expressing T21N-HA* or control cells harboring an empty vector were grown to confluence, serum-starved for 2 days, and then treated with PDGF-DD or FBS or left untreated. Approximately 24 h later, DNA synthesis was evaluated by measuring [ 3 H]thymidine incorporation into DNA. As shown in Fig. 8A, treatment with 2.5 or 5 ng/ml of PDGF-DD FIGURE 5. T21N and T21E inhibit E5 action. A and B, C127 cells and HFFs expressing T21N-HA, T21E-HA, or the empty MSCVhyg vector were infected with control retrovirus (Ϫ) or retrovirus expressing the E5 protein (ϩ), and focus formation was measured as described under "Experimental Procedures." The representative experiment shown in A was quantified in B and expressed as the percentage of the number of foci relative to that in E5-infected cells expressing empty vector. Similar results were obtained in multiple independent experiments. C, PDGF receptors were immunoprecipitated from extracts of C127 cells and HFFs co-expressing T21N-HA or control retrovirus (vector) in the absence (Ϫ) or presence (ϩ) of E5 and then subjected to anti-phosphotyrosine (PY) or PDGF receptor (PR) immunoblotting. The set of panels analyzing HFFs (top panel) or C127 cells (bottom panel) were each derived from a single gel exposed the same amount of time. The numbers indicate the sizes of molecular mass markers in kilodaltons, and arrows point to the mature (m) and precursor (p) forms of the PDGF␤R.
caused a substantially greater dose-dependent increase in DNA synthesis in control cells than in cells expressing T21N-HA*. In multiple trials, PDGF-DD-induced DNA synthesis in the T21N-HA*-expressing cells was on average 40 -60% of that in the control cells, a statistically significant decrease (Fig. 8B). In contrast, FBS treatment induced DNA synthesis to a similar extent in control and T21N-HA*-expressing cells (Fig. 8). Thus, T21N specifically inhibits the DNA synthesis response to PDGF-DD.
Finally, we determined whether T21N inhibited tyrosine phosphorylation of the PDGF␤R and/or its downstream signaling effectors in response to PDGF. Control and T21N-HA*expressing C127 cells were either left untreated or treated with 2.5 ng/ml PDGF-DD. After 30 min, cells were lysed, and tyrosine phosphorylation of the PDGF␤R was assessed as described above. As shown in Fig. 9 (top panels), PDGF-DD increased PDGF␤R tyrosine phosphorylation to a similar level in the control and T21N-HA*-expressing cells. Thus, although T21N inhibited mitogenesis in response to PDGF, global PDGF-induced PDGF␤R tyrosine phosphorylation was not inhibited by T21N. This result also implied that T21N-HA* did not impair expression of cell surface, PDGF-accessible PDGF␤R. We next determined whether T21N affected PDGF-induced phosphorylation of specific tyrosine residues on the receptor, namely Tyr 1009 , a SHP-2-binding site (36,37), and Tyr 751 , a PI3K-binding site (38,39). For this purpose, phospho-specific antibodies FIGURE 6. T21N does not inhibit complex formation between E5 and the PDGF␤R. A, extracts of C127 cells expressing T21N-HA or control retrovirus (vector) in the absence (Ϫ) or presence (ϩ) of E5 were immunoprecipitated with anti-E5 antibodies. Immunoprecipitates were then immunoblotted for the PDGF receptor (PR), E5, or HA as indicated. The numbers indicate the sizes of molecular mass markers in kilodaltons, and arrows point to the mature (m) and precursor (p) forms of the PDGF␤R. B, band intensities of PDGF␤R and E5 as shown in A were quantitated from three trials by using the Image J program. The data were normalized for E5 levels and are expressed as the percentages of PDGF␤R that co-immunoprecipitated with E5 relative to that in cells expressing E5 alone, with standard error of the mean shown. that recognize these sites were used to probe immunoblots of PDGF receptor immunoprecipitates. Phosphorylation at the Tyr 1009 site was induced to a similar level by PDGF treatment in both the control and T21N-HA*-expressing cells (Fig. 9A). In contrast, phosphorylation at the Tyr 751 site was substantially reduced in the T21N-HA*-expressing cells with or without PDGF treatment. Accordingly, the amount of PI3K recruited to the PDGF␤R was also reduced in the T21NHA*-expressing cells (Fig. 9A). Thus, T21N specifically inhibited phosphorylation of Tyr 751 on the PDGF␤R and the subsequent recruitment of PI3K to the receptor.
To assess the consequences of this differential phosphorylation on PDGF␤R signaling, cell extracts were subjected to immunoblotting using phospho-specific antibodies recognizing the phosphorylated forms of AKT, ERK1/2, and SHP-2, downstream effectors in PDGF␤R signaling pathways. As shown in Fig. 9B, PDGF-DD induced similar levels of SHP-2 phosphorylation in the control and T21N-HA*-expressing cells, consistent with the analysis of Tyr 1009 phosphorylation described above. In contrast, although PDGF-DD induced abundant phosphorylation of AKT and ERK1/2 in the control cells, it failed to induce substantial phosphorylation of these substrates in the T21N-HA*-expressing cells (Fig. 9B). The overall abundance of AKT, ERK1/2, and SHP-2 was not affected by T21N-HA*. T21N-HA* also inhibited the phosphorylation of AKT and ERK1/2 but not SHP-2 in response to PDGF-BB treatment (supplemental Fig. S3). Taken together, these results show that T21N inhibits phosphorylation of specific tyrosine residues in the PDGF␤R and thus inhibits specific ligandinduced PDGF␤R signaling events, namely PI3K pathways involving AKT and ERK1/2 phosphorylation, that lead to cellular DNA synthesis.
T21N Minimally Inhibits PDGF␤R Signaling in BaF3 Cells-We also determined whether T21N inhibited the activity of E5  in BaF3 cells. As expected, E5 cooperated with the PDGF␤R to induce IL-3-indpendent growth (Fig. 10A). Co-expression of T21N-HA with E5 inhibited this activity by ϳ30% (Fig. 10A). Moreover, T21N did not inhibit IL-3 independence induced by v-Sis or PDGF (data not shown). To explore why T21N only minimally inhibited PDGF␤R signaling in BaF3 cells in contrast to the dramatic inhibition in fibroblasts expressing endogenous PDGF␤R, we compared the ability of T21N and T21Q to interact with the PDGF␤R. Co-immunoprecipitation analysis revealed that substantially less receptor co-precipitated with T21N than with T21Q, even though T21N and T21Q were expressed at similar levels in these cells, as was the PDGF␤R (Fig. 10B). This suggests that in BaF3 cells the interaction of T21N with the PDGF␤R is greatly reduced compared with the interaction of T21Q with the receptor, in contrast to the situation in C127 cells where T21N and T21Q appear to bind to the receptor with similar efficiency (Fig. 2A). Therefore, the limited ability of T21N to inhibit the PDGF␤R in BaF3 cells correlates with its reduced binding to the receptor in these cells.

DISCUSSION
We report that the residue at position 21 in the transmembrane domain of pTM36-4 determines its effect on the PDGF␤R and its ability to transform cells. Although the T21Q, T21E, and T21N pTM36-4 mutants all interacted with the PDGF␤R, only T21Q retained near wild type transforming activity, whereas T21E and T21N were severely defective. Strikingly, T21N inhibited activation of the PDGF␤R by E5, v-Sis, and PDGF itself. Therefore, a single amino acid substitution converted a small transmembrane protein activator of the PDGF␤R into an inhibitor.
Although we were not able to detect an interaction between pTM36-4 and the PDGF␤R, complex formation between the precursor form of the PDGF␤R and several pTM36-4 point mutants was readily detectable. Furthermore, the transmembrane domain but not the extracellular domain of the PDGF␤R was required for this interaction and for T21Q activity, consist-ent with our previous analysis of pTM36-4 activity (25). These results strongly suggest that wild type pTM36-4 and the PDGF␤R also exist in a stable complex, primarily in intracellular membranes. The T21Q, T21E, and T21N mutations may directly stabilize the interaction between pTM36-4 and the PDGF␤R, thereby enabling detection by co-immunoprecipitation, or these mutations may increase the propensity of pTM36-4 to oligomerize, which may indirectly stabilize the complex with the PDGF␤R or increase the number of receptor molecules bound to pTM36-4. Consistent with a possible effect on homo-oligomerization, the identity of the hydrophilic amino acid position 17 in the BPV E5 protein and a similar position in the Hendra virus F protein affects the ability of these proteins to self-associate via transmembrane domain interactions (13,40). It is also possible that the residue at position 21 of pTM36-4, like Gln 17 of E5, participates in both homo-oligomerization and binding to the PDGF␤R.
Analysis of the transformation competent mutant, T21Q, indicated that pTM36-4 is functionally similar to E5. Like E5, pTM36-4 and T21Q activate the PDGF␤R in a ligand-independent manner and require the transmembrane domain of the PDGF␤R for activity. Also, T21Q is likely to localize to intracellular membranes, because it binds and activates the intracellular precursor form of the receptor. Finally, there was an absolute correlation between the ability of T21Q to bind to various PDGF␤R mutants and to induce proliferation in BaF3 cells. Despite this functional similarity, the numerous sequence differences between pTM36-4 and E5 imply that they differ in the way they interact with the transmembrane domain of the PDGF␤R. In fact, overlapping but distinct sets of amino acids in the receptor transmembrane domain are required for an interaction with the pTM36-4 mutants compared with E5: only the pTM36-4 mutants required Ser 516 , and only E5 required Ile 506 . Ser 516 may hydrogen bond with the hydrophilic amino acid at position 21 or another hydrophilic amino acid in pTM36-4, such as Asn 18 , which is also required for focus forming activity. 3 The T21N mutant not only failed to activate the PDGF␤R, but it actually inhibited the cellular response to E5, v-Sis, and PDGF. Decreased tyrosine phosphorylation of the receptor and a reduction in the amount of PI3K recruited to the receptor provide biochemical evidence that T21N inhibited PDGF␤R activation by E5 and v-Sis. Similarly, T21N specifically inhibited PDGF-induced phosphorylation of the PI3K-binding site (Tyr 751 ) on the PDGF␤R, recruitment of PI3K to the receptor, and phosphorylation of AKT and ERK1/2, signaling effectors of PI3K. However, T21N did not inhibit PDGF-induced phosphorylation of the SHP-2-binding site (Tyr 1009 ) or of SHP-2 itself. The inhibition of the PI3K pathway most likely contributed to the T21N-mediated decrease in PDGF-induced DNA synthesis, because PI3K stimulates a number of pathways that contribute to mitogenesis. Although T21N inhibited global phosphorylation of the PDGF␤R in response to E5 or v-Sis, both of which activate intracellular forms of the receptor, it inhibited phosphorylation of only certain key tyrosine residues on the cell surface form of the receptor that binds to PDGF. Because T21N binds primarily to the precursor form of the receptor, it may have a more potent inhibitory effect on PDGF␤R that encounters E5 or v-Sis in intracellular membranes than it does on cell surface receptors.
Minor structural differences in the side chains at position 21 determine the activity of the pTM36-4 mutants. Even though glutamine differs from asparagine by only an additional methylene group, glutamine dictates that pTM36-4-T21Q is an activator of the PDGF␤R, whereas asparagine converts it into an inhibitor. The amino acid at position 21 may differentially affect the way in which pTM36-4 self-associates, which may alter its conformation in the membrane and/or the formation of an active receptor complex, or it might directly influence the manner in which pTM36-4 interacts with the PDGF␤R, which in turn may affect the proper positioning or activity of the receptor kinase domains. Previous studies of the EGF receptor family revealed that certain transmembrane domain interactions involved in receptor dimerization reposition the kinase domains to the specific orientation required for activation (41)(42)(43). Thus, threonine or glutamine but not asparagine or glutamic acid may contact the PDGF␤R in a manner that induces the proper coupling between the transmembrane and kinase domains. Biochemical and biophysical analyses of T21Q, T21N, and T21E protein complexes may help determine the basis for these differences.
Our results provide insight regarding the mechanism by which T21N inhibits PDGF␤R signaling. The ability of T21N to inhibit two completely different activators (E5 and v-Sis/ PDGF), which bind to distinct sites on the receptor, suggests that T21N does not compete with E5 or PDGF for binding to the receptor. In fact, because T21N binds to a truncated PDGF␤R lacking the ligand-binding domain, it must bind to a different domain of the receptor than PDGF. Similarly, different amino acids in the transmembrane domain of the PDGF␤R were required for the interaction with T21N or E5, and T21N did not inhibit co-immunoprecipitation of the PDGF␤R with E5, suggesting that T21N does not inhibit E5 action by displacing the viral protein. Finally, T21N and T21Q showed comparable binding to the PDGF␤R in C127 cells, whereas T21N showed reduced binding in BaF3 cells, correlating with the limited ability of T21N to inhibit PDGF␤R-mediated proliferation in BaF3 cells. This observation suggests that the ability of T21N to bind to the PDGF␤R plays a role in the mechanism of inhibition.
The forms of receptor inhibited by T21N may also provide insight into the inhibitory mechanism. T21N binds primarily to the intracellular form of the PDGF␤R, but it inhibits tyrosine phosphorylation of both mature and precursor forms in response to E5 or v-Sis, and it inhibits phosphorylation of the PI3K-binding site induced by PDGF, which activates only cell surface forms of the receptor. T21N might induce a change in the PDGF␤R precursor that persists in the mature form even if T21N dissociates from the receptor, or T21N-bound intracellular receptors might selectively mobilize an inhibitory signal, which acts in trans on other PDGF␤R molecules, including cell surface forms. Such a negative signal must be specific for the PDGF␤R, because T21N did not inhibit focus formation induced by the activated Neu* receptor or DNA synthesis induced by FBS. Furthermore, if T21N, like the E5 protein (6), binds only a small fraction of PDGF␤R molecules, a trans signaling event could explain how T21N inhibited signaling in response to PDGF, which stimulates the vast majority of receptor molecules.
Several other observations also may be important clues regarding the mechanism(s) by which binding of T21N to the PDGF␤R inhibits the response to several different ligands. Expression of T21N caused an increased amount of PDGF␤R to co-immunoprecipitate with E5 (Fig. 6). T21N may cause the recruitment of excess receptor molecules into the E5 complex, thereby accounting for increased co-immunoprecipitation. In such an aberrant complex, the receptor kinase domains might be inappropriately positioned or sterically hindered. However, our inability to co-immunoprecipitate E5 and T21N suggests that both proteins are not simultaneously present in the same PDGF␤R complexes. It is also interesting that the interaction between T21N and the exogenous PDGF␤R was reduced in BaF3 cells, suggesting that a cellular protein, expressed at low levels in BaF3 cells, may be required for the interaction and the inhibitory activity of T21N. For example, complex formation between the T21N and PDGF␤R might be facilitated by another protein that is bound by T21N, such as a transmembrane protein-tyrosine phosphatase. Such an interaction might recruit the phosphatase to the receptor, where it might catalyze inhibitory dephosphorylation globally or at specific sites. Finally, it is possible that T21N locks the PDGF␤R in an inactive conformation that inhibits receptor autophosphorylation. For example, recent studies of EGF receptor and Neu suggest that certain transmembrane domain interactions play a role in receptor activation by altering the positioning of the cytoplasmic juxtamembrane domain such that it dissociates from an inhibitory interaction with lipids in the membrane (44 -46). Similarly, T21N might alter the conformation of the transmembrane domain of the receptor to impose an inactive conformation on the PDGF␤R. In any case, the mechanism by which T21N inhibits the PDGF␤R is likely to involve complex, dynamic interactions of T21N with various receptor forms, and further analysis of this phenomenon may reveal additional interesting features of PDGF␤R trafficking or metabolism.
The rational design of inhibitory proteins is challenging. In some cases, it is possible to convert a naturally occurring protein into a dominant-negative version by deleting a domain of the protein required for its normal activity. For example, deletion of the C-terminal kinase domain of the transforming growth factor-␤ receptor and insulin-like growth factor-I receptor gives rise to a dominant-negative truncated receptor, which binds and sequesters ligand (47,48). This approach requires a detailed understanding of the structure-function relationships of the native protein. Here, we describe an alternative approach to construct artificial inhibitory proteins. We begin with the biological selection of a small transmembrane protein activator of the target of interest. Selections for activators are often more straightforward than selections for inhibitors and are sufficiently robust to screen large libraries encoding many proteins with totally randomized transmembrane domains. This activator can then be used as a starting point to introduce a limited number of substitutions, which can be screened on a much smaller scale to identify those that convert the activator into an inhibitor. Our results suggest that the initial selection of the activator allows the isolation of a protein that binds to the target of interest and that the mutations introduced later subtly alter the interaction between the small protein and its target to change its activity.
In conclusion, we describe an artificial 36-amino acid transmembrane protein inhibitor of the PDGF␤R. Further analysis of T21N may provide new insight into the activation and signaling mechanisms utilized by the PDGF␤R. Ultimately, T21N could serve as a prototype for developing hydrophobic peptidomimetic inhibitors of the PDGF␤R that could be used for therapeutic purposes. We constructed this inhibitor by introducing a single amino acid substitution into an activator, an approach that may be applicable to other growth factor receptors and transmembrane proteins.