Platelet-derived growth factor: mechanism of action and relation to oncogenes

Platelet-derived growth factor (PDGF) is the major mitogen in serum for connective tissue-derived cells (for a recent review on PDGF, see Heldin, Wasteson & Westermark, 1985). The in vivo function of PDGF is not known, but it has been speculated that it has a role in normal as well as pathological cell proliferation, e.g. in conjunction with wound healing, atherosclerosis, myelofibrosis and neoplasia. Our major interest in PDGF research comes from studies on the control of growth of cells in culture. Several recent findings link growth factors and components involved in their mechanism of action to oncogene products. As will be reviewed in this article, studies on PDGF have been particularly illustrative in this context.

The native PDGF-molecule has an M_r of about 30×10³ and consists of two different disulphide-bonded polypeptide chains of similar size, designated A and B (Johnsson, Heldin, Westermark & Wasteson, 1982). The molecule is probably a heterodimer of one A chain and one B chain, but the existence of homodimers A—A and B-B has not been ruled out. A dimer structure for PDGF appears to be important for the mitogenic activity, since reduction irreversibly inactivates PDGF. Determination of a partial amino acid sequence for PDGF has shown that the two polypeptide chains are homologous with each other and that one of them is almost identical to part of p28^sis, the transforming protein of simian sarcoma virus (SSV) (Waterfield et al. 1983; Doolittle et al. 1983; Devare et al. 1983; Johnsson et al. 1984). Sequence analysis of the human c-sis gene has established that it is identical to the gene for the B chain of PDGF (Josephs, Guo, Ratner & Wong-Staal, 1984; Chiu et al. 1984; Johnsson et al. 1984).

The close structural relationship between the B chain of PDGFand p28^sis infers a corresponding functional homology; i.e. a PDGF-like growth factor may operate in SSV-induced transformation via interaction with the PDGF receptor and autocrine stimulation of growth. In support of this hypothesis, cell lysates and conditioned media of SSV-transformed cells have been found to contain growth factor activity that can be inhibited by PDGF antibodies (Deuel et al. 1983; Bowen-Pope, Vogel & Ross, 1984; Owen, Pantazis & Antoniades, 1984; Garrettetal. 1984; Johnssonetal. 19856). Studies on the biosynthesis and maturation of p28^sis have revealed that the product is rapidly dimerized and then proteolytically cleaved at both the N terminus and the C terminus (Robbins et al. 1983). The structure of the active protein is thus probably very similar to a PDGF B-B homodimer.

If a secreted PDGF-like growth factor causes autocrine stimulation of growth via binding to cell surface PDGF receptors, one would expect that exogenously added PDGF antibodies would inhibit the growth of the producer cells. In order to test this prediction we have used human foreskin fibroblasts infected with SSV. Indeed, PDGF antibodies added to these cells inhibit their growth as well as normalize their transformed phenotype (Johnsson et al. 1985a). Similarly PDGF antibodies have been found to inhibit the growth of other SSV-transformed cells (Huang, Huang & Deuel, 1984). These data thus support the hypothesis that a PDGF-like growth factor is involved in SSV-induced cell transformation.

Several recent observations indicate that PDGF-like growth factors may be implicated in autocrine stimulation of growth also in cell types other than SSV-transformed cells. Thus PDGF-like growth factors have been identified in the conditioned media of a variety of different cell types, normal as well as malignant (Table 1). We have partially purified and characterized the factors produced by human osteosarcoma and glioma cell lines. Their structural and functional characteristics are similar and possibly identical to PDGF (Heldin, Westermark & Wasteson, 1980; Nister, Heldin, Wasteson & Westermark, 1984; Betsholtzet al. 1983). Thus it is possible that the normal gene(s) for PDGF are aberrantly expressed in these cell lines. An involvement of PDGF-like growth factors in certain forms of neoplasia may not be uncommon since sA-related 4·2×10³ base-pair transcripts have been found in a high proportion of sarcoma and glioma cell lines, i.e. cell types that carry PDGF receptors (Eva et al. 1982).

In order to investigate whether the endogenous production of a PDGF-like growth factor causes autocrine stimulation of growth, we studied the human osteosarcoma cell line U-2 OS. These cells do not show any specific binding of PDGF (Heldin, Westermark & Wasteson, 1981). Though this is probably due to receptor blocking and down regulation, and may as such be a sign of functional activity of the endogenously produced growth factor, it poses a practical problem when one wants to prove that the same cells that produce the growth factor also respond to it. If there is an inverse relationship between growth factor production and receptor number, one would expect to find receptors on cells that produce less growth factor. Indeed a low-producer clone of the osteosarcoma cells had a certain number of specific binding sites for [¹²⁵I]PDGF (Betsholtz, Westermark, Ek & Heldin, 1984). These receptors were functional; PDGF stimulated tyrosine phosphorylation of a 185×10³M_r component, which is probably the PDGF receptor itself (see below), in membranes as well as in intact cells (Betsholtz et al. 1984). Immunoprecipitation of metabolically labelled osteosarcoma cells using an antiserum against phosphotyrosine revealed that a 115×10³M_r component was constitutively phosphorylated in these cells. Since a component of similar M_r is also phosphorylated in human fibroblasts, but only after stimulation with PDGF (Ek & Heldin, 1984), this may reflect the fact that the postreceptor pathway is permanently activated in these cells.

Evidence for autocrine receptor activation in this clone of osteosarcoma cells was thus obtained. However, contrary to our findings with SSV-transformed cells, antibodies against PDGF had no effect on the growth rate of these cells (Betsholtz et al. 1984, and unpublished). Similarly PDGF antibodies have no effect on the growth of human glioma cell lines producing PDGF-like growth factors (Nister et al., unpublished). Therefore, it is possible that the endogenous production of PDGF-like growth factors has no significance for the growth of the cells. Alternatively, the endogenously produced growth factor may activate the PDGF receptor at a site where it is inaccessible to exogenously added antibodies, e.g. the interaction may take place inside the cell and involve newly synthesized receptors before they are inserted in the membrane.

One of the PDGF-producing glioma cell lines was cloned and individual cell clones analysed with regard to PDGF production and PDGF binding. Examination of about 80 different clones revealed a remarkable variability between different clones in both these parameters (Nister et al. 1985). Clones producing much of the PDGF-like growth factor were, as expected, found to have a low number of PDGF receptors. However, several low producers also showed a low binding of [¹²⁵I]PDGF, suggesting that a clonal variation in the expression of the genes for both the growth factor and the PDGF receptor. A correlation was found between the amount of PDGF produced and the growth rate of the cells under serum-free conditions, supporting the assumption that the endogenous growth factor production is of significance for autocrine growth stimulation. Furthermore, a correlation was noticed between PDGF production and passage level, indicating that high-producer cells may have a growth advantage over low-producer cells.

Taken together, the available data support the assumption that PDGF-like growth factors may participate in autocrine stimulation of growth. This may then occur via two different mechanisms. One is represented by SSV-transformed human fibroblasts, where the growth factor is secreted from the cell and where exogenously added PDGF antibodies inhibit the stimulation (Fig. 1A). The other one is represented by certain human osteosarcoma and glioma cell lines, where a significant portion of the receptor activation occurs in a compartment inaccessible to PDGF antibodies (Fig. 1B).

PDGF exerts its mitogenic action via binding to specific cell surface receptors on responsive cells (Heldin et al. 1981). The receptor is a transmembrane glycoprotein of M_r 185×10³ (Glenn, Bowen-Pope & Ross, 1982; Heldin, Ek & Ronnstrand, 1983). It is composed of two functional parts, an extracellular ligand binding domain, and an intracellular effector domain with an associated PDGF-stimulatable tyrosine kinase activity (Ek, Westermark, Wasteson & Heldin, 1982; Nishimura, Huang & Deuel, 1982; Pike, Bowen-Pope, Ross & Krebs, 1983). The kinase activity is likely to be an integral part of the receptor molecule, since it has affinity for PDGF-Sepharose (Heldin et al. 1983) and since the kinase activity remains associated with highly purified receptor preparations (Rônnstrand et al., unpublished). The functional organization of the PDGF receptor is thus very similar to that of other tyrosine kinase associated growth factor receptors, e.g. the receptors for epidermal growth factor (EGF) (Cohen, Ushiro, Stoscheck & Chinkers, 1982), insulin (Kasuga et al. 1983) and insulin-like growth factor I (Jacobs et al. 1983). That tyrosine phosphorylation is involved in control of cell proliferation is further supported by the findings that several oncogene products are tyrosine kinases (Bishop, 1983).

Incubation of membranes from human fibroblasts with PDGF and radioactively labelled ATP led to the phosphorylation of components of 185×10³ and 130X10³M_r (Ek et al. 1982; Ek & Heldin, 1982). The 185×10³M_r component represents the PDGF receptor proper, which undergoes autophosphorylation (Heldin et al. 1983). The 130×10³M_r component is a fragment of the receptor, formed after cleavage by an endogenous Ca²⁺-dependent SH-protease (Ek & Heldin, unpublished). The degraded form of the receptor retains kinase activity. It is not known whether this proteolysis has any significance in the mechanism of action of PDGF.

Autophosphorylation of the PDGF receptor on tyrosine residues also occurs in intact cells (Ek & Heldin, 1984). Furthermore, the receptor of intact fibroblasts was also found to contain phosphate bound to serine residues. This suggests that the various functional properties of the receptor, i.e. ligand binding, kinase activity and receptor turnover, may be regulated by phosphorylations involving two different types of kinases: autophosphorylation on tyrosine residues and phosphorylation on serine residues by as yet unidentified kinases.

The mechanism whereby the mitogenic signal is transmitted from the activated PDGF receptor further into the cell is largely unknown. In view of the functional homology between several growth factor receptors and certain oncogene products, it is likely that the receptor-associated tyrosine kinase is involved. Several methods have been used to identify substrates for the PDGF receptor kinase. Cooper et al. (1982) analysed lysates from metabolically labelled, PDGF-stimulated cells by twodimensional electrophoresis; gels were then treated with alkali in order to hydrolyse phosphoserine. This method revealed certain components in the M_r region 45×10³ to 43×10³ in both 3T3 cells and human fibroblasts. Nakamura, Martinez & Weber (1983) found the major increase in tyrosine phosphorylation in PDGF-stimulated chick cells in the same M_r region. Another approach was taken by Ek & Heldin (1984) and Frackelton, Tremble & Williams (1984), who used antisera that specifically recognize phosphotyrosine. This method identified the PDGF receptor as the major tyrosine-phosphorylated component in PDGF-stimulated cells. In addition, several other components were found, including a 115×10³M_r component. The function of these components, or their involvement in PDGF-stimulated mito-genesis, is not known.

Other rapid effects of PDGF on cells include activation of protein kinase C (Rozengurt, Rodriguez-Pena & Smith, 1983), a serine/threonine kinase, which is the target for certain tumour promoters (Castagna et al. 1982), as well as mobilization of Ca²⁺ from intracellular pools (Moolenar, Tertoolen & de Laat, 1984). Both these effects, which are likely to be important in the mitogenic pathway, can be explained by the stimulatory effect of PDGF on phosphatidylinositol metabolism (Habenicht et al. 1981; Berridge, Heslop, Irvine & Brown, 1984). This results in breakdown of phosphatidylinositol bisphosphate to diacylglycerol and inositol trisphosphate, which activates protein kinase C (Nishizuka, 1984) and mobilizes Ca²⁺ (Berridge & Irvine, 1984), respectively. PDGF also rapidly stimulates the amiloride-sensitive NA⁺/H⁺ exchanger (Burns & Rozengurt, 1983; Cassel et al. 1983), leading to an alkalinization of the cytoplasm.

It is known that gene expression is a necessary event in PDGF-stimulated mitogenesis since inhibitors of mRNA synthesis block cell proliferation (Smith & Stiles, 1981). Recent studies have indicated that PDGF induces the genes of several proteins (Pledger, Hart, Locatell & Scher, 1981; Cochran, Reffel & Stiles, 1983; Linzer & Nathans, 1983), including c-myc (Kelly, Cochran, Stiles & Leder, 1983) and c-fos (Greenberg & Ziff, 1984). It is possible that the products of c-myc and c-fos, which are localized in the nucleus, function as regulators of the expression of the genetic program that controls cell proliferation.

In conclusion, studies on PDGF and cellular components linked to its action have revealed several connections with retroviral oncogenes. Together with additional information from studies on other growth factors, this suggests that proto-oncogenes encode proteins that operate in the regulation of normal mitogenesis or differentiation; these genes may obtain transforming potential by aberrant expression, mutation or recombination, leading to the synthesis of proteins that perturb the mitogenic pathway at a regulatory point (see Heldin & Westermark, 1984).

Fig. 2 illustrates schematically the mitogenic pathway of growth factors and how one can envisage that various retroviral oncogenes interfere with this pathway at different levels. The basis for the assignment of the oncogenes to various levels of growth factor action is the current knowledge of their amino acid sequence, enzymic activity and subcellular localization (for references, see reviews by Bishop, 1983; Land, Parada & Weinberg, 1983; Heldin & Westermark, 1984; Hunter, 1984).

1. Oncogenes may code for a growth factor that acts by autocrine stimulation of growth. The sis product, which is a PDGF agonist, is an example in this category.

2. Oncogenes may code for proteins that mimic the action of activated growth factor receptors. The erbB product, which corresponds to a truncated EGF receptor (Downward et al. 1984), is an example in this group. Other oncogene products with tyrosine kinase activity may also belong to this category. Thefms product has not yet been shown to be a tyrosine kinase, but since it has an extensive amino acid homology with the tyrosine kinases and is probably a membrane protein (Hampe, Gobet, Sherr & Galibert, 1984), it has been placed in this group.

3. The oncogene products of mos (Kloetzer, Maxwell & Arlinghaus, 1983) and raf/mil (Moelling et al. 1984) have been shown to have serine/threonine kinase activity. These proteins may thus be involved in growth stimulation in a manner similar to protein kinase C, which is a serine/threonine kinase that is activated by PDGF and other growth factors.

4. Of the group of nuclear oncogene products, at least two, the products of myc and fos, have normal counterparts that are induced by growth factors.

5. The ras family of oncogene products constitute a separate group. Their products are localized at the cell surface, have GTPase activity and have a structural homology with signal transducing G proteins linked to the adenylate cyclase or other effector systems (Hurley et al. 1984). Their activity is required late in Gi, immediately before the S phase (Mulcahy, Smith & Stacey, 1985).

The categorization of oncogenes above is not meant to be definitive but merely as an attempt to summarize the present knowledge. Future work is likely to lead to extensions and alterations of this model.

Our work cited in the text was supported by the Swedish Cancer Society, the Swedish Medical Research Council and The University of Uppsala.