Mechanisms of platelet-derived growth factor-induced chemotaxis

The ability of cells to migrate is of vital importance in many biological processes, including angiogenesis, wound healing and embryogenesis. Cell movement is also a crucial step for the invasion of cancer cells and during metastasis, when tumor cells leave the primary tumor to spread through blood and lymph vessels and invade other tissues. Cell migration is an intricately orchestrated process involving several steps. Membrane protrusions, so-called lamellipodia, are formed in the anterior part of the cell and adhere to the substrate. The cell translocates by actin filament contraction, and finally, the posterior part of the cell detaches from the substratum (reviewed by Lauffenburger and Horwitz1). Of particular interest is the ability of cells to migrate toward a chemical gradient, i.e., chemotaxis. Several growth-regulatory factors and cytokines stimulate chemotaxis of different cell types, one example being platelet-derived growth factor (PDGF), which primarily affects connective tissue cells. In this review, we discuss the present knowledge of the signal-transduction pathways by which PDGF stimulates chemotaxis. PDGF was originally discovered as a factor present in the α granules of platelets that had the ability to induce a mitogenic response in connective tissue–derived cells (for review, see Heldin et al.2). In addition to its mitogenic action, PDGF induces chemotaxis and actin re-organization and prevents cells from dying by apoptosis. PDGF is a family of dimeric isoforms of structurally related A-, B- and C-polypeptide chains which exert their action on responsive cells through differential binding to 2 related tyrosine kinase receptors; the PDGF α receptor binds all 3 PDGF chains, whereas the β receptor binds only the B chain. Thus, depending on which receptor types the target cell expresses and on the particular PDGF isoform present, different signaling receptor complexes are formed; PDGF ββ homodimeric receptors are induced by PDGF-BB, αβ heterodimeric receptors by PDGF-AB or -BB and αα homodimeric receptors by all PDGF isoforms. PDGF receptors have an extracellular ligand-binding domain, consisting of 5 immunoglobulin-like domains, followed by a transmembrane domain and an intracellular part. The intracellularly located tyrosine kinase domain contains a characteristic insert of approximately 100 amino acid residues. Upon ligand binding, the receptors dimerize, whereby the 2 receptor molecules are able to cross-phosphorylate each other. This leads to kinase activation, autophosphorylation of key tyrosine residues in the receptors and phosphorylation of downstream substrates. One autophosphorylated residue, Tyr857, which is located in the activation loop of the tyrosine kinase domain, is involved in the regulation of kinase activity.3, 4 Other autophosphorylation sites constitute binding sites for a class of signal-transduction molecules containing Src homology 2 (SH2) domains, which bind phosphorylated tyrosine residues in the context of a specific amino acid sequence. In the PDGF β receptor, so far 13 tyrosine autophosphorylation sites have been identified in the juxtamembrane, kinase insert and C-terminal tail regions; several of their docking partners have been identified, each one potentially initiating an intracellular signaling pathway (Table I). The PDGF β receptor induces a potent stimulation of chemotaxis, either in the homodimeric form or in the heterodimeric form with the α receptor. In contrast, the homodimeric PDGF α receptor does not mediate chemotaxis, at least not in certain cell types. Thus, using porcine aortic endothelial (PAE) cells transfected with the PDGF α receptor,5 human foreskin fibroblasts6 or vascular smooth-muscle cells,7 no stimulation of chemotaxis was seen after treatment of cells with PDGF-AA; indeed, PDGF-AA inhibited chemotaxis induced by other agents. In contrast, Hosang et al.8 found that Swiss 3T3 cells and human dermal fibroblasts responded to all 3 isoforms of PDGF with chemotaxis, as did Ferns et al.,9 using human arterial smooth muscle cells. It is possible that these differences reflect cell type-specific or context-dependent signaling mechanisms. Another possibility is that differences in methodology have contributed to the different results. In the work of Eriksson et al.5 and Nistér et al.,6 a thick filter assay was used, in which cells are allowed to migrate in a 150-μm-thick filter and the distance the cells migrate is recorded. In other studies, a thin filter assay was used, in which the number of cells migrating through the filter is scored. Taken together, the available data show that the different dimeric PDGF receptor complexes affect chemotaxis differently; however, the implication of these in vitro studies for the effects of different PDGF isoforms on chemotaxis in vivo remains to be elucidated. Binding of SH2 domain–containing signaling molecules initiates a number of different signaling pathways (Table I). Attempts to elucidate the importance of these pathways for the various cellular effects of PDGF have involved use of tyrosine-to-phenylalanine mutants specifically defective in the activation of individual pathways, dominant negative approaches and specific inhibitors. Surprisingly, several of the signaling pathways have been shown to be involved in mitogenicity as well as chemotaxis. We will discuss below signaling mechanisms claimed to be important for chemotaxis via the PDGF β receptor (Fig. 1). PDGF β receptor–induced pathways involved in chemotaxis. A ligand-bound dimeric PDGF β receptor is depicted together with some of the components of signal-transduction pathways implicated in chemotaxis. Dashed lines indicate that the intermediate steps are unknown. For explanation, see text. Phosphatidylinositol 3′-kinase (PI3-kinase) consists of 2 subunits, the SH2 domain–containing p85 subunit and the p110 catalytic subunit; both p85 and p110 occur in several isoforms (reviewed by Vanhaesebroeck and Waterfield10). Docking of PI3-kinase to phosphorylated Tyr740 and Tyr751 in the PDGF β receptor leads to activation of the lipid kinase of PI3-kinase, which phosphorylates phosphatidylinositol 4,5-bisphosphate (PIP2) in the D3 position, yielding phosphatidylinositol-3,4,5-trisphosphate (PIP3). PI3-kinase plays a pivotal role in mediating motility responses, including PDGF-induced chemotaxis. A mutant PDGF β receptor in which Tyr740 and Tyr751, sites that mediate binding and activation of PI3-kinase, were mutated to phenylalanine residues failed to mediate chemotaxis when expressed in PAE cells.11 These data were supported by the finding that the PI3-kinase inhibitors wortmannin and LY294002 inhibit PDGF-mediated chemotaxis. PI3-kinase also plays an important role in chemotaxis induced by other factors, including hepatocyte growth factor,12 nerve growth factor13 and insulin-like growth factor-I13a. Interestingly, however, the dependence on PI3-kinase for chemotaxis appears to be cell type-specific; Higaki et al.14 were not able to inhibit PDGF-dependent chemotaxis of vascular smooth-muscle cells and Swiss 3T3 cells using the PI3-kinase inhibitors wortmannin and LY294002, suggesting that other pathways may limit chemotaxis in those cells. PI3-kinase has a number of downstream targets, including Rac, a member of the Rho family of small GTP-binding proteins; Ras, another small GTP-binding protein; and the serine/threonine kinase PDK1, which after activation phosphorylates and activates a number of serine/threonine kinases, such as the kinases PKB/Akt, p70S6K and various protein kinase C (PKC) isoforms (reviewed by Vanhaesebroeck and Alessi15). There are indications that several of these pathways are involved in chemotactic signaling, i.e., Ras (see below), PKC isoforms and Rac. Exogenously added PIP3 stimulated chemotaxis of NIH-3T3 cells15a. The chemotactic response to PIP3 was inhibited by the PKC inhibitor calphostin C. Furthermore, down-regulation of PKC by long-term treatment with PMA resulted in inhibition of PIP3-induced as well as PDGF-induced chemotaxis. Thus, it is likely that PI3-kinase mediated activation of PKC isoforms is an important pathway for PDGF-induced chemotactic signaling. Rac and other members of the Rho family of GTPases are involved in various types of actin re-arrangement, which may be important for cell motility; Rho is involved in stress fiber formation, Rac in the formation of lamellipodia, while Cdc42 is involved in the formation of filopodia (reviewed by Kjøller and Hall16). Rac is dependent on the activity of PI3-kinase17 and is important in PDGF-BB–induced chemotaxis; over-expression of wild-type Rac in PAE cells stably transfected with the PDGF β receptor led to increased chemotaxis toward PDGF-BB.18 Moreover, Anand-Apte et al.19 found that migration of Rat1 fibroblasts toward PDGF-BB was inhibited by expression of dominant negative Asn17 Rac1. However, the intermediate steps between PI3-kinase and Rac are not known. Recruitment of the Grb2–Sos1 complex directly to Tyr716 of the PDGF β receptor20 or indirectly to receptor-bound and phosphorylated Shc21 or SHP-222 leads to conversion of inactive GDP-bound Ras to active GTP-bound Ras. This initiates a MAP kinase pathway, which, through sequential activation of Raf-1, MEK and Erk1/2, induces phosphorylation of certain transcription factors and transcription of a number of genes. In addition, several other signal-transduction molecules lie downstream of activated Ras, e.g., PI3-kinase, the MAP kinase p38 and the small GTP-binding proteins Rac and Ral. Using both the constitutively active and dominant negative forms of Ras, Kundra et al.23 demonstrated that PDGF-stimulated chemotaxis was dependent on Ras; however, the signal-transduction molecules downstream of Ras involved in the response, were not identified. The role of Erk1/2 in PDGF-induced chemotaxis is debated and could depend on the type of cells studied. Anand-Apte et al.19 showed that blocking the action of MEK1 by either over-expressing a dominant negative form of MEK1 or incubating with a MEK-specific inhibitor, PD98059, did not abrogate PDGF-BB–induced chemotaxis of Rat1 fibroblasts. In contrast, PDGF-BB–induced chemotaxis of human mesangial cells24 and retinal pigment epithelial cells25 was partly dependent on the activity of MEK1. Originally identified as a stress-activated kinase, p38 has also been reported to be activated in response to growth factors and cytokines (for review, see Ichijo26). A role of p38 in PDGF-mediated chemotaxis in PAE cells was demonstrated by Matsumoto et al.,27 using the specific p38 inhibitor SB203580. Furthermore, expression of dominant negative Ras, but not dominant negative Rac, inhibited p38 phosphorylation, suggesting that p38 is activated downstream of Ras and is independent of Rac for its activation. Similar findings were made by Hedges et al.,28 who found that SB203580 inhibited p38 activation, phosphorylation of HSP27 and PDGF-mediated chemotaxis in tracheal myocytes. Since HSP-27 has been implicated as a modulator of F-actin polymerization,29 it may contribute to the effect of p38 in chemotaxis. Whereas the activity of Ras is positively regulated by the nucleotide-exchange molecule Sos1 in complex with the adapter Grb2, it is negatively regulated by the Ras GTPase-activating protein (RasGAP), which converts active GTP-bound Ras to inactive GDP-bound Ras.30 RasGAP is composed of 2 N-terminally located SH2 domains and a carboxy-terminal region containing the catalytic domain, which drives the conversion of GTP-bound Ras to GDP-bound Ras. The SH2 domains of RasGAP associate with the PDGF β receptor at Tyr771 and can also mediate binding to p190RhoGAP and p62DOK in a phosphorylation-dependent manner.31, 32 p190RhoGAP negatively regulates signaling via the small GTPase Rho,33 thus counteracting Rho-mediated stress fiber formation. It is not known whether Rho is involved in the regulation of PDGF-mediated chemotaxis, but RhoA is implicated in chemotaxis mediated by the closely related CSF-1 receptor.34 A role for RasGAP in PDGF-mediated chemotaxis was first suggested by Kundra et al.,35 who demonstrated reduced migration toward a gradient of PDGF-BB of cells expressing a PDGF β receptor mutant in which Tyr771, the site of interaction with RasGAP, was replaced with a phenylalanine residue. Kundra et al.35 did not exclude that another signaling molecule, which also bind to phosphorylated Tyr771, is responsible for the positive effect. As an example, the adapter molecule Grb10 also associates with Tyr771 in a phosphorylation-dependent manner (see below). However, there are also other observations supporting a role of RasGAP in PDGF-induced migration; using fibroblasts carrying a targeted deletion of RasGAP, Kulkarni et al.36 found that PDGF-stimulated random migration was inhibited in cells lacking RasGAP, while RasGAP-expressing cells responded to PDGF with increased migration. Since loss of RasGAP binding would lead to increased activation of Ras, which, as mentioned above, is associated with increased chemotaxis, the positive effect of RasGAP is likely to be exerted via some mechanism other than down-regulation of Ras signaling. Phospholipase C-γ1 (PLC-γ1) binds to Tyr1021 of the activated PDGF β receptor, leading to phosphorylation and activation of PLC-γ1.37 Hydrolysis of PIP2 by PLC-γ1 leads to formation of inositol trisphosphate (IP3) and diacylglycerol (DAG). IP3 binds to receptors on the endoplasmic reticulum, leading to release of calcium from internal stores, whereas DAG is a potent activator of the classical and novel forms of PKC. PLC-γ1 is important in PDGF-induced motility responses in CHO cells and canine kidney epithelial cells,35 whereas a PDGF receptor mutant unable to bind and activate PLC-γ1 had an unperturbed ability to mediate chemotaxis of transfected PAE cells.11 That PLC-γ1, under certain circumstances, can be important for chemotaxis was further demonstrated by Hansen et al.,38 when investigating a mutant PDGF β receptor in which Tyr934, a Src phosphorylation site, was mutated. This receptor mutant induced phosphorylation and activation of PLC-γ1 much more efficiently than the wild-type receptor, and it mediated a more potent chemotactic response. These findings were corroborated by Rönnstrand et al.,39 using cell lines expressing either wild-type or a catalytically compromised form of PLC-γ1 under control of the tetracycline promoter; over-expression of PLC-γ1 led to a hyperchemotactic response to PDGF-BB, while cells over-expressing a catalytically compromised form of PLC-γ1 did not show this effect. Thus, under certain conditions, PLC-γ1 affects PDGF-mediated chemotaxis; it is possible that the activation, via DAG production, of different PKC isoforms, mediates the chemotactic signal farther downstream, as discussed above. However, it is possible that the IP3-mediated increase in cytoplasmic Ca2+ also contributes. The Src family kinases expressed in fibroblasts (Src, Yes and Fyn) associate with Tyr579 and Tyr581 located in the juxtamembrane region of the PDGF β receptor.40 After association, Src becomes phosphorylated on key residues, leading to activation of its intrinsic tyrosine kinase activity.41 The role of Src in PDGF-induced chemotaxis is not clear. Using Ph cells, which lack endogenous PDGF α receptor and were transfected with PDGF α/β receptor chimeras either containing or lacking the Src association sites, DeMali et al.42 demonstrated dependence of the Src association sites for chemotaxis. However, it has not been excluded that the effect involved other signal-transduction molecules that also dock to Tyr579 and/or Tyr581. Interestingly, fibroblasts carrying targeted deletions of Src, Yes and Fyn migrated as wild-type cells toward a gradient of PDGF.43 However, the fibroblasts used by Klinghoffer et al.43 were immortalized with polymer large T antigen. It has been shown by Broome and Courtneidge44 that cells immortalized with large T antigen, in contrast to wild-type cells, are not dependent on Src family kinases for PDGF-induced mitogenicity; it is not known whether large T immortalization influences their chemotactic behavior. In conclusion, it is still not clear whether members of the Src family of tyrosine kinases are involved in PDGF-stimulated chemotaxis. Focal adhesion kinase (FAK) is a non-receptor protein tyrosine kinase that has been implicated in linking integrin receptors to intracellular pathways (for review, see Schlaepfer et al.45). Several lines of evidence indicate a role of FAK in integrin-stimulated cell migration. Fibroblasts derived from mice with a targeted deletion of FAK showed reduced cell motility and enhanced focal adhesion contact formation,46 demonstrating the importance of FAK in focal contact turnover. Thus, there is good evidence that FAK has an important role in cell migration through its ability to regulate focal contact turnover in an integrin-dependent manner. However, there are data suggesting that FAK also has another role in PDGF-mediated chemotaxis. Sieg et al.47 showed that FAK forms a complex with the activated PDGF β receptor and that phosphorylation of FAK at Tyr397 is required for PDGF-stimulated cell motility. In contrast, neither the kinase activity of FAK nor the phosphorylation of FAK at Tyr925, a Grb2 association site, was necessary for PDGF-stimulated chemotaxis. Thus, FAK may have a kinase-dependent role in PDGF-stimulated chemotaxis via the regulation of focal contact turnover, as well as a kinase-independent role as an adapter. Grb7 and Grb10 belong to a family of adapter proteins that are structurally related to mig-10 of Caenorhabditis elegans. Since mig-10 is essential for migration during embryonal development (reviewed by Margolis48), it is an interesting possibility that Grb7 or Grb10 are involved in PDGF-induced chemotaxis. However, PAE cells expressing a mutant PDGF β receptor lacking the Grb7-binding site Tyr775 did not show any effect on PDGF-mediated chemotaxis compared to wild-type receptor–expressing cells.49 Grb10 has been reported to associate with Tyr771, i.e., the same phosphorylation site as for RasGAP.50 Experiments using over-expression of full-length Grb10 or the Grb10 SH2 domain have suggested involvement in the mitogenic response to PDGF. It remains to be elucidated whether Grb10 has a role in PDGF-induced chemotaxis. Two tyrosine phosphatases, SHP-2 and low m.w. protein tyrosine phosphatase (LMW-PTP), are involved in PDGF signaling. SHP-2 is ubiquitously expressed and contains 2 SH2 domains, binding of which to phosphorylated tyrosine residues leads to increased phosphatase activity of SHP-2.51 SHP-2 associates with phosphorylated Tyr763 and Tyr1009 in the PDGF β receptor,52, 53 after which it becomes phosphorylated on tyrosine residues that can be recognized by the Grb2 SH2 domain.54 The PDGF β receptor has been suggested to be a target for the phosphatase activity of SHP-2, which thereby would attenuate PDGF receptor signaling.55 A negative role of SHP-2 in PDGF signaling was suggested by the finding that a phosphatase-negative mutant of SHP-2 prolonged the membrane ruffling activity seen after PDGF-BB stimulation.55 SHP-2 is also involved in positive signaling from PDGF receptors, including activation of the Ras/Erk kinase pathway. It has been suggested that this is due not only to binding of Grb2–Sos1 to phosphorylated SHP-2 since over-expression of a catalytically compromised mutant of SHP-2 suppresses PDGF-stimulated mitogenesis while retaining the Grb2-binding site.56 Furthermore, PDGF β receptors in which the SHP-2 association sites Tyr763 and Tyr1009 had been mutated to phenylalanine residues did not mediate a chemotactic response and showed a much lower ability to mediate the PDGF-stimulated increase of Ras GTP loading and activation of Erk2.53 Similarly, using a PDGF β receptor mutant lacking the main site of association of SHP-2, Tyr1009, Qi et al.57 found a decreased chemotactic response. Furthermore, PDGF stimulation of cells expressing the wild-type PDGF β receptor led to induction of FAK activity at low concentrations of PDGF and a decrease at higher concentrations. In contrast, in cells expressing the PDGF β receptor mutant in which Tyr1009 had been mutated to a phenylalanine residue, PDGF-BB failed to induce activation of FAK, suggesting that FAK activity was involved in the regulation of the chemotactic response. Fibroblasts derived from mice with a targeted deletion of SHP-2 were impaired in their ability to spread and migrate on fibronectin compared to wild-type cells.58 Thus, several observations suggest involvement of SHP-2 in PDGF-induced chemotaxis, though the precise mechanisms remain to be determined. LMW-PTP is an evolutionarily conserved 18 kDa enzyme that is expressed in many mammalian tissues (reviewed by Ramponi and Stefani59). The PDGF β receptor associates with LMW-PTP in a phosphorylation-dependent manner through the phosphatase domain of LMW-PTP. Furthermore, over-expression of LMW-PTP led to a strong reduction of cell growth rate in response to PDGF, which could be explained partly by dephosphorylation of the PDGF receptor.60 After PDGF stimulation, a cytoskeleton-associated pool of LMW-PTP is specifically phosphorylated by c-Src, leading to a 20-fold increase in phosphatase activity of LMW-PTP.61 Furthermore, expression of a catalytically compromised mutant of LMW-PTP led to decreased PDGF-mediated chemotaxis, which coincided with increased phosphorylation of p190RhoGAP, a target of LMW-PTP.62 Thus, it was suggested that LMW-PTP–mediated dephosphorylation could inactivate p190RhoGAP, leading to increased activation of Rho and, hence, cytoskeletal re-arrangements. A number of sphingolipid metabolites, including sphingosine 1-phosphate (SPP), formed by sphingosine kinase, may enhance tyrosine phosphorylation of FAK. This leads to Rho-dependent stress fiber formation and focal contact assembly in Swiss 3T3 cells.63 Furthermore, in human arterial smooth-muscle cells, SPP generated in response to PDGF interfered with actin re-organization, resulting in a marked inhibition of cell spreading, extension of lamellipodia and chemotaxis toward PDGF.64 This suggests that endogenous SPP may play an important role in regulating chemotaxis. The mechanisms whereby PDGF stimulates sphingosine kinase activity, however, remain to be elucidated. Over the last decade, more than a dozen tyrosine-phosphorylation sites within the PDGF β receptor and their corresponding docking partners have been identified. Many pathways that are initiated by binding of these signaling molecules influence both the mitogenic and chemotactic responses of PDGF. This may seem surprising but probably reflects the fact that many of these signal-transduction pathways are interconnected and influence each other. Furthermore, under physiological conditions in vivo, the receptors are likely to be challenged with much lower concentrations of PDGF than under experimental conditions. Thus, it is not unlikely that signal-transduction pathways are induced under experimental conditions after maximal receptor activation following stimulation with high doses of ligand, which under physiological conditions play a minor role. Several observations implicate both PI3-kinase and PLC-γ1 in chemotactic signaling. The candidate downstream targets include Rac and various isoforms of PKC; however, the detailed mechanisms remain to be identified (Fig. 1). Furthermore, Ras appears to play an important role in the chemotactic response to PDGF in many cell types. Several downstream effectors of Ras are necessary for a full chemotactic response, e.g., the Erk and p38 MAP kinases and Rac. Notably, Ras is able to activate PI3-kinase and vice versa, providing one example of cross-talk between signaling pathways. The information we have so far obtained about the signaling mechanisms involved in PDGF-stimulated chemotaxis has come mainly from the use of tyrosine-to-phenylalanine mutants of the receptor, dominant negative forms of the signal-transduction molecules or more or less specific pharmacological inhibitors. These methods have their limitations: A phosphorylated tyrosine residue might bind more than one SH2 domain–containing signal-transduction molecule, a dominant negative construct might interfere with closely related pathways and the specific inhibitors many times show a broader specificity than originally anticipated. A refinement of the techniques is necessary to enable a more precise dissection of the pathways involved in chemotactic signaling. Targeted deletion of specific genes in mice as well as replacement of wild-type protein in the mouse with a mutated version, so-called knock-in, have already generated important data and are likely to be important methods for future studies. Furthermore, methods to visualize interactions between molecules in real time in living cells, such as fluorescence resonance energy transfer, will add to the available arsenal of tools for future studies of signal-transduction pathways involved in chemotaxis, as well as in other cellular events.