Characterization of the Retention Motif in the C-terminal Part of the Long Splice Form of Platelet-derived Growth Factor A-chain*

Platelet-derived growth factor (PDGF) A-chain is found in two different splice variants; the long A-chain variant differs from the short one in that it contains a stretch of basic amino acid residues in the C-terminal part that mediates retention of the growth factor inside the producer cell and to the cell matrix. By analyzing mutants in which different amino acid residues in the retention motif had been changed to alanine residues, we found that the total positive charge of the sequence is of importance for the function of the retention motif. Moreover, we showed that retention also occurs if only one of the polypeptides in the PDGF dimer carries the retention motif. Surface iodination and competition with a peptide having the sequence of the retention motif revealed that the long A-chain variant, in contrast to the short A-chain variant, is localized on the outside of the cells and is also associated to the cell matrix. The association is likely to be mediated partially through heparan sulfate proteoglycans since treatment of matrix with heparitinase released the long A-chain variant. Platelet-derived growth factor (PDGF)' is a 30-kDa dimeric protein composed of disulfide-linked A- and B-polypeptide chains that are 60% identical in their mature parts, with a perfect conservation of the 8 cysteine residues (reviewed in Refs. 1 and 2). The PDGF A-chain appears as long and short variants, as a result of differential splicing (3,


Platelet-derived growth factor
(PDGF) A-chain is found in two different splice variants; the long A-chain variant differs from the short one in that it contains a stretch of basic amino acid residues in the C-terminal part that mediates retention of the growth factor inside the producer cell and to the cell matrix. By analyzing mutants in which different amino acid residues in the retention motif had been changed to alanine residues, we found that the total positive charge of the sequence is of importance for the function of the retention motif. Moreover, we showed that retention also occurs if only one of the polypeptides in the PDGF dimer carries the retention motif. Surface iodination and competition with a peptide having the sequence of the retention motif revealed that the long A-chain variant, in contrast to the short A-chain variant, is localized on the outside of the cells and is also associated to the cell matrix. The association is likely to be mediated partially through heparan sulfate proteoglycans since treatment of matrix with heparitinase released the long A-chain variant.
Platelet-derived growth factor (PDGF)' is a 30-kDa dimeric protein composed of disulfide-linked A-and B-polypeptide chains that are 60% identical in their mature parts, with a perfect conservation of the 8 cysteine residues (reviewed in Refs. 1 and 2). The PDGF A-chain appears as long and short variants, as a result of differential splicing (3,4). All three isoforms, PDGF-AA, -AB, and -BB, have been purified from natural sources. The ability of PDGF to promote growth, chemotaxis, and matrix production of connective tissue cells has led to speculations that a normal function of PDGF is to stimulate wound healing. Recent findings suggest that PDGF also regulates cell growth and chemotaxis during embryonal development. Moreover, there is evidence for involvement of PDGF in pathological conditions including arthritis, atherosclerosis, and bone marrow fibrosis as well as in malignancies (reviewed in Ref. 5). The different biological effects of the PDGF isoforms are exerted via binding to two structurally similar protein tyrosine kinase receptors (6-8); the a-receptor binds both PDGF A-and B-chains, whereas the P-receptor only binds the B-chain with high afflnity (9-13). Binding of ligand leads to dimerization and autophosphorylation of the receptors, followed by ty- in v-sis-transformed cells also occur mainly as cell-associated forms (18,19). The cell retention has been found to be dependent on the presence of a motif of basic amino acids in the C-terminal propeptide of the PDGF B-chain and in the C-terminal part of the long variant of the A-chain (20)(21)(22)(23)(24). This motif has also been reported to mediate interaction with heparan sulfate proteoglycans of the extracellular matrix (20)(21)(22)25). A similar structural motif has also been found in the longer splice variants of vascular endothelial growth factor and placental growth factor, which are potent mitogens for endothelial cells and structurally related to PDGF (26-30).
The aim of the present study was to investigate the structural basis for the interaction between the retention motif and cellular components. For this purpose we analyzed mutants in which different amino acids in the motif in the long version of the PDGF A-chain were changed to alanine residues. In addition the cellular/pericellular localization of the retained material was determined.

MATERIALS AND METHODS
Construction of cDNAs Encoding PDGF Mutants-cDNAs encoding the long and the short variants of PDGF A-chain have been described (14). The corresponding expression vectors are named pSVA, and pSV&, respectively. Mutations of codons in the long PDGF A-chain variant corresponding to amino acid residues (numbering as in Ref. 14))  to alanine codons were made using the method of Kunkel et al. (31) on a uracil-containing template encoding the long variant of PDGF A-chain. Two double mutants and one triple mutant in which amino acid residues Lys-201/Lys-202, Lys-206/Arg-207, or Arg-203/Lys-204/ Arg-205 were replaced with alanine residues were also made. The expression vectors pSVA201, pSVA202, pSVA203, pSVA204, pSVA205, pSVA206, pSVA207, pSVA201/A202, pSVA206/A207, and pSVA203/ A204/A205 were generated by cloning of the mutated fragments into the EcoRI sites of the expression vector pSV7d (32). Two other mutants, in which the 2nd cysteine residue from the N terminus (Cys-123) in the long A-chain variant and the 4th cysteine residue (Cys-132) in the short A-chain variant were changed to serine residues, were made using the same method; the expression vectors pSVAL2 and pSVb4 were generated by cloning into the EcoRI sites of the expression vector pSV7d. All the plasmids were sequenced over the region encoding the mature parts of the proteins.
Expression and Immunoprecipitation of Recombinant Proteins-The pSV constructions encoding the mutant PDGF A-chains as well as pSVAs and pSVA, were transfected into COS cells as described (241, using 15 pg of plasmid DNA and 0.5-1 x lo6 cells in 60-mm culture dishes. In the case of double transfections, 7.5 pg of each plasmid DNA was used. Two days after transfection, metabolic labeling was performed by growing the cells overnight in 1.5 ml of cysteine-free MCDB 104 medium supplemented with 0.1 mCi of [35Slcysteine/ml, 10% dialyzed fetal calf serum, and antibiotics. After labeling, the media were collected and cleared of cell debris by centrifugation. The cells were washed once in PBS, collected by scraping, and lysed in 0.5 ml Of0.5 M NaCl, 20 mM Tris-HC1, pH 7.5, 0.5% Triton X-100, 1% Trasylol (Sigma), and 1 mM phenylmethylsulfonyl fluoride. The cell lysates were centrifuged for 15 min at 10,000 x g, and the supernatants, as well as the cell culture supernatants, were subjected to immunoprecipitations. Samples were precleared by incubation with 15 pl of normal rabbit serum at 4 "C for 1 h, followed by addition of 60 ml of a 50% Protein A-Sepharose slurry in PBS. After incubation a t 4 "C for 30 min, the beads were removed by centrifugation. The media and cell lysates then received 15 1. 11 of rabbit antisera raised against PDGF-AA(33) and were then incubated at 4 "C for 2 h. After incubation with Protein A-Sepharose as above, the beads were washed 5 times with 0.5 M NaCl, 20 II~M Tris, pH 7.5,5 mg/ml bovine serum albumin, 1% Triton X-100, and 0.1% SDS, and once with 20 ~TLM Tris-HC1, pH 7.5. The immunocomplexes were eluted by addition of 200 pl of nonreducing SDS sample buffer (34) and incubation at 95 "C for 3 min. Half of the eluted material was reduced by addition of dithiothreitol to a final concentration of 10 mM, followed by incubation at 95 "C for 2 min, and was then alkylated by addition of iodoacetamide to a final concentration of 50 mM. The samples were analyzed by SDS-gel electrophoresis (341, using 12-18% polyacrylamide gels, followed by fluorography. Iodination of Surface Proteins"IYansfections of COS cells were performed with pSV& and pSVAL as described above. A mock transfection was also made. Two days after transfections, the surface proteins were labeled with lZ5I using the lactoperoxidase method. The cells were washed 3 times with PBS-G (PBS supplemented with 0.9 mg/ml glucose) at 37 "C. To each dish was added 1 ml of PBS-G with 0.1 unitlml glucose oxidase, 6 unitdm1 lactoperoxidase, and 0.2 mCi of lZ5I, followed by incubation a t room temperature for 20 min. The cells were then washed 5 times with PBS, solubilized, immunoprecipitated, and analyzed by SDS-gel electrophoresis, as described above.

Release of the Long A-chain Variant from Labeled COS Cells after Incubation ulith
Peptide-COS cells were transfected with pSV& and pSVAL as described above. Two days after transfection, metabolic labeling was performed by growing the cells for 16 h in 1.5 ml of cysteine-free MCDB 104 medium supplemented with 0.1 mCi of [35Slcysteine/ml, 10% dialyzed fetal calf serum, and antibiotics. After 12 h, a peptide denoted RPR with the amino acid sequence RPRESGKKRKRKRLKRT, thus resembling the C terminus of the long A-chain, was added at a concentration of 40 p~. The conditioned media and cell lysates were then immunoprecipitated and analyzed, as described above.
Release of the Long A-chain Variant from Extracellular Matrix-COS cells were transfected and labeled as described above. The cells were then lysed using 0.5% deoxycholate, 0.15 M NaC1, 0.05 M Tris-HC1, pH 8.0, and the remaining matrix washed once with PBS. Each dish was then incubated for 4 h at 37 "C in 1 ml of PBS containing 40 p~ RPR peptide or 0.2 unit/ml heparitinase (Heparinase 111, Sigma). The eluate was then immunoprecipitated and analyzed, as described above.

Mutational Analysis of the Retention Motif of the Long PDGF
A-chain-In order to determine which of the amino acid residues in the C-terminal basic motif of the long version of the PDGFA-chain are important for retention, different residues in the motif were changed to alanine residues (Fig. 1B). The mutants were transfected into COS cells; analysis of immunoprecipitates from [35Slcysteine-labeled media and cell lysates by SDS-gel electrophoresis under nonreducing conditions followed by fluorography revealed precursor and mature PDGF species of 40-30 kDa. The figures in the lower part of Fig. 2 represent the average from three different experiments of secreted PDGF, expressed in percent of the total material. As can be seen, substitution of single amino acid residues had no effect. Thus, none of the residues were uniquely important for the retention (Fig. 2). When two or three amino acid residues were changed, some of the corresponding protein appeared in the medium (Fig. 2); PDGF-AL-A206/A207 and PDGF-AL-A203/A2041A205 were secreted as efficiently as the short form of PDGF-AA, whereas PDGF-AL-A201/A202 was still retained to some extent. Taken together, these observations indicate that the total charge of the retention motif, rather than single amino acid residues, is of importance for its function as a retention signal. However, there may also be somewhat of a positional effect with Lys-201 and Lys-202 being of lesser importance for the . . . P R E Name of slngle ammo acld mutants: illustrates the different single, double, and triple alanine residue mutants in the retention motif of the long A-chain variant that were used in the present study. Arrows and brackets indicate the amino acid residues that were substituted to alanine residues. Panel C shows the two serine residue mutants made. The cysteine residues in the A-chain are shown ( C ) ; S indicates that the corresponding cysteine residue was exchanged to a serine residue. retention than the other charged amino acids.
Retention of a PDGF Heterodimer-In view of the fact that PDGF occurs as homo-and heterodimeric isoforms, we wanted to determine whether retention requires targeting motifs in both chains of the PDGF dimer or whether one motif is sufficient. As we have shown elsewhere (35), the interchain disulfide bonds in PDGF are located between cysteine residue 2 in one chain and cysteine residue 4 in the other chain, and vice versa. In order to specifically assemble a heterodimer with only one retention motif, we mutated cysteine residue 2 in the long A-chain variant to a serine residue, and cysteine residue 4 in the short A-chain variant to a serine residue (Fig. 1C). When these two mutants were transfected individually into COS cells, the long chain variant was retained whereas the short chain was secreted.
Both proteins occurred as monomers, which is expected since in these mutants, interchain disulfide bonds cannot form. Co-transfection of the two plasmids yielded a dimeric molecule, presumably a heterodimer, which was only detected in the cell lysate (Fig. 3). Thus, we conclude that one retention motif in a PDGF dimer is sufficient for retention.
Localization of the Retained Material-Evidence has been presented that the retention motif of the long PDGF A-chain mediates interactions with components intracellularly (17,33) as well as extracellularly (20)(21)(22). In order to characterize further the localization of the retained PDGF, we transfected COS cells with pSVAs and PSVAL as described above. On the second day after transfection, the surface proteins were labeled with % PDGF secreted i 3

. Expression of the retention motif mutants in COS cells.
Fifteen pg of the pSV expression vectors containing the different long A-chain mutants, in which one or more amino acid residues in the retention motif were exchanged for alanine residues, was transfected into COS cells. A mock transfection (-) as well as transfections with pSVAL and pSV& were also camed out. After metabolic labeling with 135Slcysteine, the conditioned media and cell lysates were immunoprecipitated with antisera against PDGF-AA. The immunoprecipitates were then analyzed by SDS-gel electrophoresis under nonreducing conditions, followed by fluorography. The mature form of PDGF-AA is represented by the 30-kDa band, and the bands of 35 and 40 kDa represent precursor forms of PDGF-AA. In order to quantify the amount of released material, each lane was subjected to scanning with an Ultrascan XL enhanced laser densitometer; the amount of material in the region where the mature and precursor forms of PDGF are found (3040 kDa) was determined. The amount of PDGF in the medium fraction in relation to that of the corresponding cell lysate fraction (average of two or three determinations for each mutant) is given at the bottom of the figure. 1251 using the lactoperoxidase method. A metabolic labeling with [35S]cysteine was also performed on parallel transfected COS cells. Lysates from both the 1251-and [35Slcysteine-labeled cells were then immunoprecipitated with an antiserum against PDGF-AA and subjected to SDS-gel electrophoresis followed by autoradiography and fluorography, respectively. Analysis of 1251-labeled cells revealed that the long A-chain was present a t the surface of transfected COS cells as three different species of 30, 35, and 40 kDa; no 1251-labeled PDGF A-chain was immunoprecipitated from cells transfected with the short form of the A-chain (Fig. 4A). Interestingly, comparison of immunoprecipitated 1251-labeled exterior AL-chain with the AL-chain from [35S]cysteine-labeled cells revealed that the latter contained two additional species of 25 and 33 kDa (Fig. 4B ). We conclude that part of the long A-chain occurs at the cell surface or in the extracellular matrix where it is accessible to lactoperoxidase labeling, whereas additional species of 25 and 33 kDa of the long form of the A-chain occur inside the cell only.
In order to determine whether the long variant of PDGF-AA could be displaced from its extracellular binding sites, we transfected COS cells with pSVAL, labeled the cells with [35Slcysteine, and incubated them in the absence or the presence of 40 p~ of a 17-amino acid peptide (RPR), with a sequence corresponding to the C terminus of the long A-chain encompassing the retention motif during the last 4 h of labeling. The presence of the RPR peptide led to the release of components of 30 and 40 kDa into the medium, representing the mature and precursor form of PDGF-AAL, respectively; in the absence of peptide none of these components was seen (Fig. 5). This is likely to be due to a specific competition between the long A-chain and the peptide for binding sites on the cell surface or in the matrix. As expected, the amount of short A-chain in medium from COS cells transfected with pSVqS was similar in the absence or presence of the peptide (Fig. 5). In the presence of the RPR peptide relatively more of the precursor form of PDGF-& (40 kDa) compared with the mature form (30 kDa) was seen. This may be due to an inhibition by the basic peptide of extracellular N-terminal processing of the A-chain precursor at the -Arg-Arg-Lys-Arg-site (141, where cleavage normally occurs. To more specifically investigate the possibility that the long A-chain was associated with the cell matrix, COS cells transfected with PSVAL or pSVAs and labeled with [35Slcysteine were removed by incubation in a deoxycholate-containing buffer; the remaining matrix was then incubated with or without 40 p~ RPR peptide for 4 h. Analysis by immunoprecipitation with a PDGF-AA antiserum revealed that the long A-chain was released from the matrix after incubation with the peptide (Fig. 6).
Many growth factors have been shown to bind to heparin or heparan sulfate. In order to determine whether the retention of the long version of PDGF-AA involved binding to heparin or heparan sulfate, we incubated matrix from COS cells prepared in the same way as described above with 0.2 unitlml heparitinase for 4 h. The material released from the matrix after enzyme treatment was then immunoprecipitated with an antiserum against PDGF-AA. Incubation with heparitinase led to release of the long A-chain variant, whereas no release of the short variant was seen. Thus, part of the long A-chain is retained in the matrix bound to heparidheparan sulfate proteoglycans (Fig. 6).

DISCUSSION
The object of the present study was to explore the structure/ function relationship of the basic sequence in the C-terminal part of the longer splice variant of the PDGF A-chain that has been shown to serve as a retention motif. Our results suggest that the total negative charge rather than single amino acid residues in the retention motif is of importance for retention. Moreover, we demonstrate that the retention motif targets the protein to an intracellular destination as well as to the cell matrix, the latter in part through binding to heparinheparan sulfate proteoglycans.
A heterodimer of one long and one short PDGF A-chain was specifically assembled by coexpression in COS cells of PDGF A-chain mutants with the 2nd or 4th cysteine residues, respectively, substituted by serine residues. The heterodimer was found to remain associated with the producer cells; thus, in COS cells one retention motif in a dimer is sufficient to mediate retention. In contrast, a PDGF-AB heterodimer produced in CHO cells stably transfected with the B-chain and the short A-chain and thus having one retention motif per dimer was found to be secreted into the medium (16). It is thus possible that the retention motif in the long A-chain is more efficient than that in the B-chain. Alternatively, it is possible that the amounts of the components that bind the retention motif vary in different cell types. This notion is strengthened by the previous finding that PDGF-BB is efficiently secreted by human melanoma cells (36) but retained by several other cell types. COS cells may have a relative abundance of the components that associate with the retention motif, whereas CHO cells may have less, leading to saturation of the binding sites and thereby secretion of PDGF-AB heterodimers. Another possibility is that specific proteases cleave off the retention motif and release PDGF into the medium. CHO cells, in comparison to COS cells, may have relatively much of such proteases.
Analysis by immunoprecipitation of the different forms of the long A-chain in lysates of metabolically labeled cells compared with lz5I-1abeled A-chains in the matrix revealed two additional forms of 25 and 33 kDa in the cell lysate. In an analogous manner, the major part of PDGF-BB, which also contains the retention motif in the C-terminal prosequence, appears as a 24-kDa cell-associated form (16,17,37,38). Previous studies have shown that the 24-kDa form of PDGF-BB has undergone N-terminal as well as C-terminal processing. This processing takes place in the Golgi complex, from which PDGF-BB is then transported to lysosomes for additional degradation (17). It is not known whether this cell-associated form is biologically active or not. It is possible that the 25-kDa form of PDGF-AAL, seen only intracellularly, is analogous to the 24-kDa form of PDGF-BB and thus represents a species on its way to be degraded in the lysosomes.
So far no differences in biological activity between the longer and shorter variants of PDGF-AA have been found. In most investigated cell types that express the PDGF A-chain, around 10% of the mRNA appears as the longer splice variant. The functional significance of the presence of one secreted and one cell matrix-associated variant of PDGF-AA thus remains to be elucidated. It is possible that the shorter variant, PDGF-&, acts in a paracrine way and affects cells in the immediate environment of the producer cell, as well as at some distance, whereas PDGF-AAL stimulates cells by autocrine mechanisms and via its association with matrix. Matrix-associated PDGF-AAL could thus provide a scaffold of growth factor. The fact that also vascular endothelial growth factor and placental growth factor, which are structurally related to PDGF, appear as different splice variants, with or without retention motif (26,301, suggests that the compartmentalization of this growth factor family is an important feature. Moreover, several growth factors, including the stem cell factor, colony-stimulating factor-1, and transforming growth factor a, appear as membrane-anchored molecules, which are released in soluble forms by proteolysis (3941).
The present investigation in conjunction with previous studies indicates that the positively charged retention motif associates with heparinheparan sulfate proteoglycans, which are highly negatively charged molecules (2622,251. The data presented in this study support the notion that the total positive charge of the retention motif in PDGF-AAL correlates to the efficiency for retention. Future studies will be aimed at exploring whether the retention motif shows specificity for interaction with heparan sulfate proteoglycans or whether it also interacts with other proteoglycans and other negatively charged molecules.