Phosphatidylinositol linkage of a truncated form of the platelet-derived growth factor receptor.

The platelet-derived growth factor (PDGF) receptor is usually anchored to the plasma membrane through a membrane-spanning hydrophobic amino acid sequence that splits the molecule into two approximately equal pieces, an amino-terminal external domain that contains the binding site for PDGF and a carboxyl-terminal cytoplasmic domain that includes the tyrosine kinase coding sequences. Here we report the expression of a truncated PDGF receptor that consists of the extracellular domain without the transmembrane and cytoplasmic domains. Unexpectedly, this form of the receptor that lacks a hydrophobic membrane-anchoring sequence was bound to the membrane and was not secreted into the culture media. Conventional methods to dissociate noncovalent protein-protein interactions failed to release the protein from the membrane. When the transmembrane and cytoplasmic sequences were artificially deleted from the PDGF receptor, the truncated extracellular domain was anchored to the membrane through phospholipids and could be released by phospholipase C treatment. This truncated form of the receptor bound PDGF with an affinity 5-20-fold lower than the full-length receptor.

The platelet-derived growth factor (PDGF) receptor is usually anchored to the plasma membrane through a membrane-spanning hydrophobic amino acid sequence that splits the molecule into two approximately equal pieces, an amino-terminal external domain that contains the binding site for PDGF and a carboxyl-terminal cytoplasmic domain that includes the tyrosine kinase coding sequences. Here we report the expression of a truncated PDGF receptor that consists of the extracellular domain without the transmembrane and cytoplasmic domains. Unexpectedly, this form of the receptor that lacks a hydrophobic membrane-anchoring sequence was bound to the membrane and was not secreted into the culture media. Conventional methods to dissociate noncovalent protein-protein interactions failed to release the protein from the membrane. When the transmembrane and cytoplasmic sequences were artificially deleted from the PDGF receptor, the truncated extracellular domain was anchored to the membrane through phospholipids and could be released by phospholipase C treatment. This truncated form of the receptor bound PDGF with an affinity 5-20-fold lower than the full-length receptor.
Platelet-derived growth factor (PDGF)' stimulates the proliferation of mesenchymal cells by binding to a 185-kDa receptor protein (1)(2)(3)(4)(5). The receptor for PDGF is a member of a class of transmembrane proteins capable of transducing an extracellular signal, growth factor binding into a number of intracellular events that subsequently lead to cell division. When PDGF binds to the extracellular domain of the receptor it elicits a change in the conformation of the receptor' and stimulates the receptor's tyrosine kinase activity, which appears to be essential for the mitogenic response to PDGF. This single membrane-spanning region of the receptor splits the molecule into two roughly equal pieces, an amino-terminal external domain that contains the binding site for PDGF and a carboxyl-terminal cytoplasmic domain that includes the tyrosine kinase coding sequences (6). The external PDGFbinding domain of the receptor consists of tandem repeats of * This work was supported by National Institutes of Health Grant R01 HL32898. 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.
The abbreviations used are: PDGF, platelet-derived growth factor; CHO, Chinese hamster ovary; FBS, fetal bovine serum; BSA, bovine serum albumin; SDS, sodium dodecyl sulfate; PI, Phosphatidylinositol; FITC, fluorescein isothiocyanate; PMSF, phenylmethylsulfonyl fluoride; kb, kilobase; PBS, phosphate-buffered saline. Biol. Chem. 263, in press five immunoglobulin-like domain^.^ Here we report the expression of a truncated PDGF receptor encoding the extracellular domain protein devoid of transmembrane and cytoplasmic sequences. Unexpectedly, this form of the receptor that lacked a hydrophobic membraneanchoring sequence was bound to the membrane and was not secreted into the culture media. A diverse group of membrane proteins have recently been found to be associated with the membrane through glycosyl PI that is attached to the carboxyl terminus of the protein (7-11). Here we show that a truncated form of the PDGF receptor that lacks membrane spanning and cytoplasmic sequences is anchored to the membrane through phospholipids and can be released from membranes by phospholipase C treatment.

EXPERIMENTAL PROCEDURES
Materials-CHO cells (clone Kl), Ham's F-12 media, fetal bovine serum (FBS), penicillin, streptomycin, and methionine-free media were obtained from the University of California, San Francisco, Tissue Culture Facility. Geneticin G-418 was obtained from GIBCO. FITC-goat anti-rabbit was obtained from Boehringer Mannheim. IODO-GEN was obtained from Pierce. The PD-10 column and Ficoll-Hypaque were obtained from Pharmacia LKB Biotechnology Inc. Bovine serum albumin (BSA), L-methionine, protein A-Sepharose, phospholipase C (type 111), and radioactive molecular weight standards were purchased from Sigma. [%]Methionine, EN3HANCE, and Na'*'I were purchased from Du Pont/New England Nuclear. Electrophoretic chemicals, nonradioactive molecular weight standards, horseradish peroxidase-labeled goat anti-rabbit IgG, and horseradish peroxidase color development reagent were obtained from Bio-Rad. Restriction enzymes were obtained from New England BioLabs. PDGF was produced from outdated platelets as previously described (4,5).
Recombinant Plusmid-The full-length cDNA sequence encoding the mouse PDGF receptor obtained from a A-16 clone was inserted at the EcoRI site of pSV7d (12) to form an expression vector, control of the SV40 early promoter (13). The cDNA encodes the designated pSVR1, in which the receptor gene is under transcriptional initiation methionine (amino acid -31), the signal peptide (amino acid -31 to -1, base pair 139-231), the external domain of the PDGF receptor (amino acid 1-499, base pair 232-1728), the transmembrane region (amino acid 500-524, base pair 1729-1803) and the intracellular domain (amino acid 525-1067, base pair 1804-3432). The cDNA also includes the untranslated 5' region (base pair 1-138) and the untranslated 3' region (base pair 3432-5134) (6). The receptor transmembrane and cytoplasmic domain coding sequences were removed from pSVRl by digestion with BstEII. Two fragments were generated, one of 2.9 kb from base pair 1696 to base pair 4623 of the PDGF receptor and another of 4.9 kb. The fragments were separated by electrophoresis on a 1% agarose gel and the 4.9-kb fragment that purified and self-ligated (pSVED). This plasmid was used to trans-lacked transmembrane and cytoplasmic receptor coding regions was form Escherichia coli HB101, and was tested for the presence of the desired deletions by digestion with appropriate restriction endonucleases. 15159 Cells and Transfections-CHO cells (clone K1) were grown in Ham's F-12 media supplemented with 10% FBS, penicillin (100 units/ ml) and streptomycin (100pg/ml) at 37 "C in 5% C02/95% air. pSVR1 plasmid DNA or pSVED plasmid DNA (10 pg) and pSV2Neo (1 pg) were used to cotransfect 1 X IO6 CHO cells by the calcium phosphate precipitation technique (14) with the addition of 100 pg of chloroquine diphosphate to prevent degradation of the transfected DNA (15). After 16 h of exposure to DNA, the cells were split 1:7, and 24 h later were put under selection with the antibiotic Geneticin G-418 (0.4 mg/ ml). After 2 weeks, independent colonies were picked and transferred to 24-well plates. Confluent cultures were assayed for the presence of the full-length or the truncated PDGF receptor. Cells were grown and maintained in the media specified here throughout all the experiments.
Flow Cytometric Analysis-Confluent monolayers of untransfected CHO cells, CHO cells expressing the full-length receptor, and CHO cells expressing the truncated receptor were washed twice with phosphate-buffered saline (PBS) and detached with 2 mM EDTA in PBS. Cells were washed twice with cold PBS and adjusted to 10 X 10' cells/ ml in Ham's F-12 media supplemented with 5 mg/ml BSA. Cells were incubated with antireceptor antibody for 1 h at 4 "C, washed three times with cold PBS and labeled with FITC-goat anti-rabbit antibody for 30 min at 4 "C. After three washes with cold PBS, stained cells were adjusted to 1 X IO6 cells/ml Ham's F-12 media and analyzed for immunofluorescence using a FACS IV instrument equipped with a log amplifier.
Metabolic Labeling and Immunoprecipitation of Cell Extracts-Cells were labeled with [%3]methionine and immunoprecipitated as described previously (16). Briefly, untransfected CHO cells, CHO cells expressing the full-length receptor, and CHO cells expressing the truncated receptor were grown in complete media on Costar six-well plates (35 mm) and used 5 days after plating. Cell monolayers were washed and incubated for 30 min in serum-free, methionine-free media at 37 "C. After methionine starvation, cells were labeled to equilibrium in 1 ml of methionine-free media supplemented with 10% dialyzed FBS, 1 pg/ml L-methionine, and 250 pCi/ml [36S]methionine for 20 h at 37°C. After labeling, cells were washed with cold PBS and solubilized at 4 "C in 400 fillwell Tris-buffered Triton containing 10 mM Tris (pH 7.4), 50 mM NaCl, 50 mM NaF, 30 mM sodium pyrophosphate, 100 p~ sodium metavanadate, 5 mM EDTA, 0.1% Triton X-100, 1 mg/ml BSA, and 1 mM phenylmethylsulfonyl fluoride (PMSF). Lysates were cleared by centrifugation. Immunoprecipitation with antireceptor antibody was performed as described (16). Lysates were then incubated with 50 pl of a 50% (v/v) suspension of protein A-Sepharose in immunoprecipitation buffer for 45 min. Immunoprecipitates were washed three times in immunoprecipitation buffer containing decreasing concentrations of salt.
Electrophoresis, Autoradiography, and Western Blots-SDS-polyacrylamide gel electrophoresis was performed according to the method of Laemmli (17). Samples for electrophoresis were suspended in sample buffer to a final concentration of 10% glycerol, 2% SDS, and 50 mM dithiothreitol in 35 mM Tris (pH 6.8). Samples were heated to 95 "C for 5 min and loaded onto a vertical-slab 7% polyacrylamide gel. Gels containing radioactive samples were treated with EN3HANCE for 1 h, dried, and exposed to Kodak XAR-5 film at -70 "C for 1-5 days. Gels containing nonradioactive samples were transferred to nitrocellullose by electrophoresis toward the anode in 20 mM Tris (pH 8.3), 150 mM glycine, 20% methanol, and 0.1% SDS (18). Nitrocellulose blots were soaked for 1 h in 50 mM Tris (pH 7.4), 150 mM NaCl, 5% BSA, and 0.05% Tween 20, and were then incubated with antireceptor antisera for 2 h at room temperature. Blots were washed with three changes of Tris-buffered saline and incubated with horseradish peroxidase conjugated to goat anti-rabbit IgG for 1 h at room temperature. Blots were washed as above and horseradish peroxidase-labeled proteins were localized by incubation with horseradish peroxidase-color development reagent containing the enzyme substrate 4-chloro-I-naphthol and hydrogen peroxide.
Iodination of PDGF-Purified PDGF was iodinated by the IODO-GEN method as previously described (4). Briefly, a 15-pl aliquot of IODO-GEN (0.2 mg/ml methylene chloride) was dispensed into a prewashed dry tube (solvent was removed under a stream of nitrogen). Sodium phosphate buffer (0.2 M, pH 7.4) (100 pl) was added to 1 mCi of Na'T and transferred to a siliconized tube containing 5 pg of PDGF in 50 pl of 0.2 M phosphate buffer. The mixture was transferred to the IODO-GEN tube and allowed to incubate for 25 min at 4 "C, agitating every minute. The reaction was stopped by transferring to a siliconized tube containing 100 p1 of 2 M acetic acid and 100 mM NaI for 10 min at 4 "C. Free iodine was separated from 'T-PDGF by excluding over a prepacked PD-10 column preequilibrated and eluted with 0.1% Pentax BSA in 1 M acetic acid. The purity of the iodinated product was documented by SDS-polyacrylamide gel electrophoresis under reducing and nonreducing conditions. The radiospecific activity of "'I-PDGF was 140 pCi/pg of protein.
Binding Studies-CHO cells expressing the full-length PDGF receptor and CHO cells expressing the truncated PDGF receptor were grown to confluence in complete media in Costar T-150 flasks. Five days after seeding and 2 days after confluence was reached, monolayers were washed with PBS and detached with 2 mM EDTA in PBS. Cells were washed twice with PBS and adjusted to 15 X IO6 cells/ml. For a saturation binding study, 1.5 X IO6 cells were incubated with increasing concentrations of "'1-PDGF in a final volume of 0.5 ml in PBS containing platelet-poor plasma. Incubations proceeded for 45 min (full-length receptor) and 2 h (truncated receptor) at 37 "C with gentle shaking. Cell suspensions were layered on top of cold diluted Ficoll-Hypaque (28.5% in PBS, 700 pl) and were spun for 2 min at 13,000 X g. The supernatants were aspirated and the tube tip with the cell pellet was cut and counted. For competition experiments, 1.5 X IO6 cells were incubated at steady-state conditions with "' 1-PDGF (40,000 cpm) together with various concentrations of unlabeled PDGF in PBS containing 10% platelet-poor plasma (final volume, 0.5 ml). Free radioactivity was separated from cell-bound radioactivity as described above. Nonspecific binding, defined as counts remaining in the presence of 1,000-fold excess of unlabeled PDGF, was subtracted from all the readings. The remaining counts were defined as specific binding. The affinity constants and the number of receptors were calculated from the binding curves by Scatchard analysis (19).
Membrane Preparations-Membranes were prepared from CHO cells expressing the full-length or the truncated PDGF receptor according to a procedure described previously (5). Briefly, cells were washed with PBS and removed from roller bottles using 2 mM EDTA in PBS. The suspended cells were washed with cold borate-buffered saline (0.05 M borate, pH 7.4, 0.15 M NaCl, 2 mM EDTA) and were lysed with 2 ml of hypotonic borate buffer (0.02 M borate, pH 10.2, 2 mM EDTA, 1 mM PMSF). The lysates were centrifuged at 40,000 X g at 4 "C for 20 min and the pellets were resuspended in PBS with 2 mM EDTA and 1 mM PMSF in a Potter homogenizer (30 strokes). The suspension was centrifuged at 300 X g for 10 min. The supernatant devoid of intact cells, nuclei, mitochondria, and cell debris was centrifuged at 40,000 X g at 4 "C for 20 min. The pellet was washed twice by resuspension in PBS with 2 mM EDTA, followed by centrifugation. The final pellet was resuspended in 50 mM Tris (pH 7.4) containing 0.25 M sucrose. Membranes were frozen and stored at -70 "C until use.
Membrane Treatments-Membranes prepared as described above were thawed and washed twice with 25 mM Tris (pH 7.5) by centrifugation at 13,000 X g at 4 'C. The final pellet was resuspended and incubated in 100 pl of 25 mM Tris for 30 min at 4 "C, or 0.5 M NaCl for 30 min at 4 "C, or 100 mM sodium-carbonate (pH 11) for 30 min at 4 "C (20), or 0.2 M acetic acid, pH 2.5, containing 0.15 M NaCl for 6 min at 4 "C. Alternatively, membranes were suspended in 100 pl of 25 mM Tris (pH 7.5) and mixed with 3 units of phospholipase C that had been preincubated for 16 h at 4 "C in 25 mM Tris (pH 7.5) or the same buffer with 2 mM ZnCla, or the same buffer with 25 mM 1,lOphenanthroline. The samples were incubated for 2 h at 37 'C and the membranes were pelleted at 13,000 X g. Membrane extracts and supernatants were analyzed by SDS-polyacrylamide gel electrophoresis and immunoblotting.

RESULTS
The expression plasmid pSVRl containing the coding sequences for the full-length PDGF receptor cDNA (6) was digested with BstEII to create the expression plasmid pSVED containing the external domain sequences of the PDGF receptor cDNA under the transcriptional control of the SV40 early promoter (Fig. 1). The truncated and full-length receptor expression plasmids were used to transfect CHO cells that do not express detectable levels of the PDGF receptor protein or mRNA. It was anticipated that the truncated receptor would be secreted into the media because it has the amino-terminal signal sequence of the mature receptor and lacks the hydrophobic transmembrane region. However, no receptor protein was found in conditioned media of cells transfected with the truncated receptor cDNA as assayed by immunoprecipitation  or by Western blot analysis using antireceptor antibody. Both the full-length and truncated receptors were readily detected on the surface of their respective transfected cells by fluorometric studies using antireceptor antibodies (16) and FITC conjugated to goat anti-rabbit antibody (Fig. 2). Approximately equal amounts of surface protein were detected in the transfectants expressing either the full-length or truncated form of the receptor (Fig. 2). The apparent molecular mass of the truncated receptor was 84 kDa as assessed by immunoblot analysis in cell extracts using antireceptor antibody (Fig. 3,  lane 5). As previously reported by our laboratory, the fulllength receptor was expressed as a 195-kDa protein (13). These results were confirmed by experiments in which metabolically labeled receptor was immunoprecipitated using antisera directed against the external domain of the PDGF receptor (Fig. 3, lunes 2 and 3 ) . The level of truncated protein was &fold higher than that of the full-length receptor protein in both the metabolic labeling and immunoblot experiments. Because this difference in level of expression was not reflected by the fluorometric studies (Fig. 2), it is likely that 80% of were labeled to equilibrium in 1 ml of methioninefree media supplemented with 10% dialyzed FBS, 1 pg/ml L-methionine, and 250 pCi/ml[35S]methionine for 20 h at 37 "C. After labeling, cells were solubilized and lysates were immunoprecipitated with antireceptor antibody and protein A-Sepharose (16). Samples were electrophoresed on a 7% SDS-polyacrylamide gel (17) and exposed to Kodak XAR film for 1-3 days at -70 "C using an intensifying screen.

CHO cells transfected with plasmid pSVRl (lane 4 ) and CHO cells transfected with plasmid pSVED (lane 5)
were lysed, electrophoresed, and transferred to nitrocellulose paper. The filter sheet was seauentially incubated with antireceptor antibody, goat anti-rabbit antibody conjugated to horseradish peroxidase and horseradish peroxidasecolor reagent. Molecular weight standards from top to bottom: myosin, 0-galactosidase, phosphorylase b, bovine serum albumin, ovalbumin, and carbonic anhydrase.
20% is expressed at the surface. By contrast, the full-length receptor is expressed almost exclusively at the surface (16, 21). Membrane preparations of the cells expressing truncated and full-length receptors showed the same 5-fold difference in level of protein (not shown), indicating that the intracellular pool of expressed external domain was associated with membranes. Moreover, when intact cells were treated with the truncated receptor protein is present inside the cell and trypsin for 2 h at 4 "C, lysed, and immunoblotted with anti- (1.5 X lo6 cells) were incugether with several concentrations of unlabeled PDGF in PBS containing platelet poor plasma at 37 "C (final volume, 0.5 ml). Free radioactivity was separated from cell-bound radioactivity as described under "Experimental Procedures." Specific binding was 75% for CHO full-length receptor cells and 50% for CHO truncated receptor cells. The data shown represent averages of three determinations. Standard deviations are less than 7%. and CHO truncated receptor ( U ) were incubated with increasing concentrations of '=I-PDGF at the same conditions described above. The dissociation constant and the number of receptors was calculated according to Scatchard (19). The data shown represent averages of three determinations. Standard deviations are less than 7%. receptor antibody directed against the external domain of the PDGF receptor (16), the 195-kDa band corresponding to the full-length receptor disappeared, whereas the 84-kDa band corresponding to the truncated receptor was only partially diminished (data not shown).
The transfected CHO cells that expressed the truncated receptor and the cells that expressed the full-length receptor bound lZ5I-PDGF with high affinity. Kinetics studies showed that the 'T-PDGF binding to full-length receptor was maximal a t 45 min, whereas the truncated receptor required 2 h to bind a maximal amount of lZ5I-PDGF a t 37 "C (data not shown). Thus, all binding studies were performed under these conditions in subsequent experiments. To assess the apparent affinities of the full-length and truncated receptors, competition studies were performed using unlabeled PDGF to inhibit binding of lZ5I-PDGF to both types of cells (Fig. 4). In CHO cells transfected with a full-length receptor, purified PDGF displaced 50% of specifically bound lZ5I-PDGF a t a concentration of 0.1 nM. In CHO cells transfected with the truncated receptor, 1-2 nM unlabeled PDGF was required to displace 50% of the specific binding of '251-PDGF. By Scatchard analysis (Fig. 5 ) the apparent Kd of the full-length receptor was 0.114 nM and the apparent Kd of the truncated receptor was 0.5 nM. By this analysis there were 4000 receptors on the cell surface of cells bearing the full-length receptor and 5200 truncated receptor molecules per cell.
The predicted amino acid sequence of the truncated receptor does not contain hydrophobic regions that should provide a membrane anchor to the protein. To clarify whether this protein was bound to cell membranes by noncovalent interactions, membrane preparations were exposed to high pH, low pH, and high salt concentrations. As shown in Fig. 6, none of these treatments dissociated the truncated receptor protein from the membranes. However, when membrane preparations were treated with phospholipase C, the external domain protein was released from the membranes and could be readily detected in the soluble phase (Fig. 7). Pretreatment of the enzyme with ZnClz, which inhibits the phosphatidylinositol-specific phospholipase C activity, prevented release of truncated receptor protein from the membranes, indicating that the external domain receptor protein was anchored to the membrane through a glycosyl-phosphatidylinositol moiety. By contrast, pretreatment of phospholipase C with 1,lO-phenanthroline did not inhibit the ability of phospholipase C to release the external domain protein. This chelating agent does not inhibit phosphatidylinositol-specific phospholipase C (22) but inhibits phosphatidylcholine-specific enzyme activity (23). Thus the activity of phospholipase C was specific for a phosphatidylinositol linkage of the external domain protein.
In control experiments on membranes that express the full-length receptor, no receptor protein was released from the membrane by phospholipase C treatment (Fig. 7). Thus the effect of phospholipase C was not due to disruptive effects on membrane structures.

DISCUSSION
Membrane proteins vary markedly in the nature and extent of their interaction with the lipid bilayer. Some of these proteins are associated with lipids through noncovalent interactions involving one or more sequences of hydrophobic amino acids that span the membrane. Other proteins in both prokaryotic and eukaryotic cell membranes contain covalently attached lipid that is incorporated into the bilayer. Among this group of proteins are those containing a covalently attached glycosylated phosphatidylinositol moiety that is linked to the protein through 0-acyl bonds (7-9). The diverse group of proteins anchored to the plasma membrane through an acarboxyl-glycosyl-phosphatidylinositol group includes alkaline phosphatase (24- (44), T-cellactivating protein (45), N-CAM120 (46, 47), heparan sulfate proteoglycan (48), transferrin receptor, and p195 surface antigen of Plasmodium falciparum (49,50).
We have constructed a mutant cDNA clone that encodes the extracellular domain of the mouse PDGF receptor. The deletion of the fragment corresponding to base pair 1696-4623 produces a mutant that lacks 33 base pairs of the extracellular domain upstream of the transmembrane region, the transmembrane region, and the entire intracellular domain (Fig. l). In the construction of the vector, a sequence of 51 base pairs from the 3"untranslated region of the receptor cDNA (nucleotides 4623-4674) was linked in frame with the external domain coding sequences. (A termination codon is located at position 4675.) Of these 17 amino acids, 13 are polar and 4 are hydrophobic amino acid residues, distributed randomly through the region. The truncated receptor protein was covalently attached to the membrane. Conventional methods to dissociate noncovalent protein-protein interactions failed in releasing the protein from the membrane. However, phospholipase C treatment of the membranes readily released the truncated receptor, indicating that the protein was anchored through phospholipids. Even though the phospholipase C used in this work is of broad specificity, it was specifically inhibited by ZnC12, which inhibits the phosphatidylinositol-specific enzyme. Conversely, release was not inhibited by 1,lO-phenanthroline, a specific inhibitor of phosphatidylcholine (23). A control preparation of membranes with the full-length receptor was not affected by phospholipase C treatment, indicating that the releasing effect was not due to a disruptive effect on the membrane. Although we cannot make out the possibility that the full-length receptor has a PI linkage as well as a transmembrane sequence, there is no known precedent for a dual anchor to the membrane.
The truncated receptor was expressed at a &fold higher level than the full-length receptor, as determined by autoradiography and Western blots (Fig. 3). However, approximately equal amounts of truncated and full-length receptors were expressed at the cell surface, as indicated by fluorometric studies (Fig. 2). Trypsinization of intact cells slightly diminished (20%) the amount of truncated receptor detectable by immunoblot, whereas the full-length receptor was almost completely cleaved by this treatment (not shown). These data suggest that approximately 80% of the truncated receptor protein is not expressed on the cell surface but is bound to intracellular membranes. However, all the receptor protein appears to be anchored by a phospholipid tail (Fig. 7).
The apparent affinity of the truncated PDGF receptor was lower than the affinity of the full-length receptor (Fig. 4); it is possible that the abnormal attachment to the membrane decreases the affinity by changing the conformation of the binding domain of the receptor. A truncated external domain of the interleukin 2 receptor binds interleukin 2 with an affinity 100-fold lower than that of the wild-type receptor (51). Similarly, the 105-kDa truncated epidermal growth factor receptor external form secreted by A431 cells has an affinity for polyclonal and monoclonal antibodies directed against the epidermal growth factor receptor that is 10-fold lower than that of the full-length receptor (52).
The signal that designates a protein for attachment to glycosyl-phosphatidylinositol and the mechanism by which this attachment occurs are not known. Several studies have suggested that the amino acid residues at the COOH terminus are important in the linkage. The decay-accelerating factor is a protein linked through its COOH terminus to PI (53). When the COOH-terminal31 amino acids of the decay-accelerating factor are linked to the COOH terminus of a truncated version of the herpes simplex glycoprotein, the viral protein is no longer secreted but is anchored in the membrane through a PI linkage (54). Similar experiments have shown that the COOH terminus of the Qa-2 antigen confers lipid anchorage to the H-2 protein (55). In the cases of Thy-1 antigen (56) and variant surface glycoprotein of T. brucei (57), the cDNA sequence predicts the presence of a COOH-terminal sequence of 15-20 hydrophobic residues that are not found in the mature PI-linked protein. For these two proteins the attachment of the glycosyl-PI may occur in conjunction with proteolysis of the COOH-terminal hydrophobic region. A similar situation occurs with the neural cell adhesion molecule, N-CAM, that may be anchored either by glycosyl-PI or a transmembrane amino acid sequence, depending on the form of the N-CAM mRNA that is expressed; the glycosyl-PI anchored form of the protein is encoded by a mRNA that is processed to include a unique 3' sequence that codes for a short hydrophobic COOH-terminal peptide. Differential RNA splicing gives rise to a form of the protein that has a different COOH terminus that includes a full transmembrane amino acid sequence (46,58). Although these studies have indicated the importance of the COOH terminus in directing a protein to become a glycosyl-PI anchored molecule, the specific amino acid sequences of these proteins have not provided a recognizable consensus for this process. In the experiments reported here, the predicted COOH-terminal sequences encoded by the truncated PDGF receptor cDNA are not similar to the known COOH-terminal sequences of other proteins that are linked to PI and are not even particularly hydrophobic. The extreme COOH-terminal sequence of the expressed truncated receptor consisted of a 17-amino acid region that is derived from the 3'-untranslated portion of the native receptor (see Fig. 1 and "Experimental Procedures"). This sequence, VTSGHCHEERVDRHDGE, includes only 4 hydrophobic amino acids. Transfection experiments on hybrid molecules containing the sequence at the COOH terminus are now in progress. From these studies it should be possible to determine the structural features of the truncated PDGF receptor that were responsible for forming a PI linkage in a protein that is normally anchored through a transmembrane sequence of hydrophobic amino acids.