Pigment epithelium-derived factor inhibits fibroblast-growth-factor-2-induced capillary morphogenesis of endothelial cells through Fyn

Angiogenesis is a prerequisite for many physiological and pathological processes, such as wound healing, solid tumor growth and retinopathies (Folkman, 1995; Carmeliet and Jain, 2000; Bikfalvi and Bicknell, 2002). Angiogenesis comprises a series of cellular responses by endothelial cells. These include secretion of proteases to digest the basement membrane of vessel walls, migration into interstitial tissues, survival, proliferation and differentiation into newly established capillaries by endothelial cells. These cellular responses of endothelial cells are tightly regulated by a balance between proangiogenic factors and anti-angiogenic factors (Bouck et al., 1996; Hanahan and Folkman, 1996).

Pigment epithelium-derived factor (PEDF) (Bouck, 2002; Tombran-Tink and Barnstable, 2003), a noninhibitory member of the serpin superfamily, was first identified as a differentiation factor for neuronal cells (Tombran-Tink and Johnson, 1989). Later, its potential antiangiogenic action in vivo was found (Dawson et al., 1999). Previous in vitro studies showed that PEDF inhibits the migration, proliferation and capillary morphogenesis of cultured endothelial cells (Dawson et al., 1999; Duh et al., 2002; Hutchings et al., 2002; Wang et al., 2003). In vivo, PEDF inhibited experimental ocular neovascularization and tumor angiogenesis (Stellmach et al., 2001; Mori et al., 2001; Raisler et al., 2002; Mori et al., 2002). During the study, it was noted that apoptosis of endothelial cells was promoted by PEDF (Stellmach et al., 2001). PEDF upregulated Fas-ligand expression, which was in turn associated with proangiogenic-factor-stimulated expression of Fas, leading to endothelial cell apoptosis and inhibition of angiogenesis (Volpert et al., 2002). By using PEDF-deficient mice, it was found that the tissue mass and vasculature were regulated by PEDF in the prostate and the pancreas (Doll et al., 2003). Considered together, the above studies indicate that PEDF plays important roles in developmental and pathological angiogenesis.

The specific receptor for PEDF has not yet identified. However, it has been shown that type-I collagen, glycosaminoglycans and an 80 kDa cell surface protein bound to PEDF (Alberdi et al., 1998; Alberdi et al., 1999; Meyer et al., 2002). Likewise, only a limited number of studies have examined the intracellular signaling pathways regulated by PEDF. Nuclear factor κB (NF-κB) and its transcriptional target proteins were involved in PEDF-mediated neuroprotective activity (Yabe et al., 2001). PEDF promoted activation of mitogen-activated protein kinase in endothelial cells (Hutchings et al., 2002). More recently, it was shown that PEDF regulated endothelial-cell apoptosis by the inhibition of nuclear factor of activated T cells through Jun N-terminal kinases (Zaichuk et al., 2004). However, little is currently known about the PEDF-driven signal-transduction pathways leading to the inhibition of migration and capillary morphogenesis by endothelial cells.

We have previously shown that FGF-2 induced tube formation in a mouse-brain capillary endothelial cell line, IBE cells (Kanda et al., 1996), through Fyn (Tsuda et al., 2002). Src-family kinases were required for FGF-2-induced capillary morphogenesis of both IBE cells and human umbilical-vein endothelial cells (HUVECs) (Kanda et al., 2003b). Blocking of endogenous vascular-endothelial growth factor A (VEGF-A) secreted by endothelial cells attenuated FGF-2-induced capillary morphogenesis of IBE cells and HUVECs in association with downregulation of basal Akt activity (Kanda et al., 2004a). Addition of exogenous VEGF-A did not induce capillary morphogenesis and Akt activation. These observations suggest that the maintenance of basal Akt activity by endogenous VEGF-A in combination with upregulated Fyn activity were required for FGF-2-induced capillary morphogenesis by endothelial cells.

The aim of the present study was to elucidate the mechanisms underlying PEDF-induced inhibition of capillary morphogenesis by endothelial cells. PEDF inhibited FGF-2-induced capillary morphogenesis by both IBE cells and HUVECs. Inhibition of morphological change was associated with the decrease in Fyn activity. In IBE cells, Fyn activity was downregulated by phosphorylation of inhibitory tyrosine at the C-terminus, which was dependent on Fes tyrosine kinase activity.

Recombinant human PEDF was obtained from Chemicon International (Temecula, CA). Anti-Myc monoclonal antibody, anti-Fyn polyclonal antibody (FYN 3), anti-Src antibodies (SRC 2) and anti-phosphotyrosine monoclonal antibody (PY99) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-Fyn and anti-C-terminal Src kinase (CSK) monoclonal antibodies were from BD Biosciences (Bedford, MA). Anti-Src-[pY⁵²⁹] antibody was obtained from BioSource International (Camarillo, CA). Anti-FLAG mouse monoclonal antibody (M2), anti-FLAG rabbit polyclonal antibody, anti-vinculin mouse monoclonal antibody (hVIN-1), anti-CSK rabbit polyclonal antibody and anti-p190^Rho-GTPase-activating-protein (p190RhoGAP) monoclonal antibody were from Sigma-Aldrich (St Louis, MO). Recombinant human CSK and recombinant N-terminally six-His-tagged human Fyn were obtained from Upstate Biotechnologies (Lake Placid, NY). Growth-factor-reduced Matrigel matrix was purchased from BD Bioscience.

IBE cells were cultured in Ham's F-12 medium supplemented with 20% fetal bovine serum (FBS; Life technologies, Rockville, MD), 75 μg ml^–1 endothelial-cell growth supplement (Sigma), 10 ng ml^–1 human recombinant epidermal growth factor (EGF; Roche Molecular Systems, NJ) and 5 μg ml^–1 bovine insulin (Sigma). HUVECs and their culture medium were obtained from Cambrex Bio Science Walkersville (Walkersville, MD). HUVECs were cultured in endothelial cell medium 2 (EGM™-2) supplemented with 2% FBS, 20 ng ml^–1 FGF-2, 20 ng ml^–1 VEGF-A, 10 ng ml^–1 EGF, 20 ng ml^–1 insulin-like-growth-factor 1 and 1 μg ml^–1 hydrocortisone. Stable cell lines expressing FLAG-tagged wild-type Fes (WT6-8 cells), kinase-inactive Fes (KE5-15 cells) (Kanda et al., 2000) and kinase-inactive Fyn (KDFyn-8 cells) (Tsuda et al., 2002) have been described previously.

A cDNA encoding activated Fyn (Y531F) was constructed from the cDNA encoding kinase-inactive Fyn (K299M) subcloned into pBK-CMV phagemid (Tsuda et al., 2002) by using the QuikChange Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA) according to the protocol provided by the manufacturer. Sequence of the construct was confirmed by sequencing. The cDNA encoding Y531F Fyn was subcloned into pcDNA3.1/Myc-His (+) mammalian expression plasmid (Invitrogen, San Diego, CA) and was transfected into IBE cells by the use of TFL-5 liposome (a kind gift from H. Kikuchi, Daiichi Pharmaceuticals, Tokyo, Japan). 18 G418-resistant clones were picked, and expression of transfected protein in these clones was examined by immunoprecipitation followed by immunoblotting.

Capillary morphogenesis assay was described previously (Kanda et al., 2003b). In brief, IBE cells were suspended in Ham's F-12 medium containing 0.25% bovine serum albumin (BSA) and seeded onto growth-factor-reduced Matrigel® in wells of 24-well plates with or without the indicated samples. 24 hours later, capillary morphogenesis was examined under a phase-contrast microscope. For HUVECs, cells suspended in a culture medium containing 0.5% FBS were inoculated onto growth-factor-reduced Matrigel with or without treatment and were cultured for 24 hours. To quantify the length of capillaries, three different phase-contrast photomicrographs (10 × objectives) per well were taken and the length of each capillary was measured using NIH image software (version 1.64). Capillary length of cells treated with FGF-2 alone was set to 100.

Cells were lysed in Nonidet P-40 (NP-40) lysis buffer [50 mM N-2-hydroxyethylpiperazine-N-2-ethylanesulfonic acid (HEPES), pH 7.5, 150 mM NaCl, 10% glycerol, 1% NP-40, 0.1% sodium dodecyl sulfate (SDS), 100 U ml^–1 aprotinin, 1 mM phenylmethylsulfonyl fluoride, 1 mM ethylenediaminetetra-acetic acid (EDTA), 0.1 mM orthovanadate] and particular proteins were immunoprecipitated with indicated antibodies and separated by SDS polyacrylamide-gel electrophoresis (SDS-PAGE). After electronic transfer onto polyvinylidene difluoride (PVDF) membranes (Millipore, Bedford, MA), the blots were probed with indicated antibodies. To examine the expression of transfected activated Fyn in IBE cells, cells grown on 6 cm dishes in growth medium were used. To examine the phosphorylation status of Fyn or p190RhoGAP, cells were suspended in Ham's F-12 medium containing 0.25% BSA and seeded onto growth-factor-reduced Matrigel with or without the indicated reagents. After 2 hours, cells were lysed and particular proteins were immunoprecipitated, followed by immunoblotting.

The in-vitro kinase assay for Fyn has been described previously (Tsuda et al., 2002). In brief, cells were seeded onto Matrigel-coated wells of six-well plates or 6 cm dishes with or without the indicated samples. After 2 hours, cells were lysed in a buffer supplemented with 1% Triton X-100, 0.5% deoxycholic acid and 0.1% SDS on ice. Indicated kinases were immunoprecipitated by specific antibodies and incubated in kinase buffer (25 mM HEPES, pH 7.4, 2 mM MgCl₂, 10 mM MnCl₂, 0.1% 2-mercaptoethanol, 0.1% NP-40) supplemented with acid-denatured enolase and [γ-³²P] ATP, followed by separation by SDS-PAGE. Gels were washed, fixed, treated with 1 M KOH to remove phosphorylated serine residue and then examined using an Image Analyzer BAS 5000 (Fuji), followed by exposure on X-ray films. For the examination of Fes activity, autophosphorylation of FLAG-tagged Fes was determined without enolase.

To examine whether anti-Src-[pY⁵²⁹] antibody cross-reacted with phosphorylated Fyn, Src immunoprecipitated from HUVECs or Fyn from IBE cells was incubated in the kinase buffer (25 mM HEPES, pH 7.4, 0.1% NP-40, 5 mM MnCl₂, 1 mM MgCl₂, 0.1 mM orthovanadate, 0.1 mM ATP) supplemented with or without 500 ng of recombinant CSK at 30°C for 20 minutes. Reaction was stopped by adding 2× SDS sample buffer and proteins were separated by SDS-PAGE. Phosphorylation of tyrosine residue in Src or Fyn was examined by immunoblotting with anti-Src-[pY⁵²⁹] antibody. To determine whether Fes can phosphorylate tyrosine residues located at the C-terminus of Fyn, FLAG-tagged Fes was immunoprecipitated from cells grown on Matrigel with or without 200 ng ml^–1 PEDF for 2 hours and incubated in the kinase buffer supplemented with or without 30 ng of recombinant Fyn at 37°C for 20 minutes. As a positive control, 30 ng ml^–1 Fyn was incubated with 500 ng recombinant CSK. Phosphorylation of recombinant Fyn was examined by immunoblotting as described above.

Indirect immunofluorescent staining of cultured cells has been described previously (Kanda et al., 2004b). Cells were cultured on the surface of Matrigel-coated cover slips with or without PEDF for 2 hours. Cells were washed with PBS and fixed with 3.7% paraformaldehyde, followed by incubation with 0.1% Triton X-100. For the double staining, the primary antibodies from different species were combined as follows: anti-vinculin monoclonal antibody and anti-FLAG polyclonal antibody; anti-vinculin monoclonal antibody and anti-CSK polyclonal antibody; and anti-FLAG monoclonal antibody and anti-Fyn polyclonal antibody. Proteins recognized by rabbit polyclonal antibodies and by mouse monoclonal antibodies were visualized with secondary antibodies labeled with Rhodamine and fluorescein isothiocyanate, respectively. Two pictures of each field were taken at different wavelengths under fluorescent microscopic observations.

FGF-2 induces capillary morphogenesis of cultured endothelial cells. To examine whether PEDF inhibits FGF-2-induced capillary morphogenesis, we used a murine capillary endothelial cell line, IBE cells and HUVECs. As shown in Fig. 1A, FGF-2 efficiently induced capillary morphogenesis of IBE cells cultured on growth-factor-reduced Matrigel. PEDF dose-dependently inhibited FGF-2-induced capillary morphogenesis and the inhibition was maximal at a concentration of 200 ng ml^–1. PEDF inhibited elongation of IBE cells but did not increase the number of floating dead cells when FGF-2 was present, suggesting that PEDF inhibited the morphogenesis rather than the survival of endothelial cells. We also examined the effect of PEDF on FGF-2-induced capillary morphogenesis by HUVECs. As shown in Fig. 1B, 200 ng ml^–1 PEDF potently inhibited FGF-2-induced capillary morphogenesis by HUVECs without promoting cell death.

Fes is a non-receptor protein tyrosine kinase, which is distinct from Src-family kinases and Abl, and is exclusively expressed in hematopoietic cells, endothelial cells and certain neuronal cells (for review, see Smithgall et al., 1998). Fes is endogenously expressed in IBE cells and HUVECs (Kanda et al., 2000). PEDF acts on both endothelial cells and some neuronal cells, suggesting that Fes might be involved in PEDF-mediated cellular responses. We treated a stable IBE cell line expressing kinase-inactive Fes (clone KE5-15 cells) (Kanda et al., 2000) with PEDF. The cell line exhibited a dominant-negative effect on endogenous Fes in FGF-2-induced motility, stromal-cell-derived-factor-1α- and sonic-hedgehog-induced differentiation of IBE cells (Kanda et al., 2000; Kanda et al., 2003a; Kanda et al., 2003b). As shown in Fig. 2, FGF-2 induced capillary morphogenesis of KE5-15 cells. PEDF failed to inhibit the FGF-2-induced capillary morphogenesis of these cells, suggesting that PEDF might use Fes to inhibit capillary morphogenesis of endothelial cells.

To determine whether PEDF can increase kinase activity of Fes, FLAG-tagged Fes was immunoprecipitated from WT6-8 and KE5-15 cells grown on Matrigel, and an in-vitro kinase assay was performed. As shown in Fig. 3A, PEDF increased the autophosphorylation of FLAG-tagged wild type Fes. PEDF failed to upregulate Fes activity in KE5-15 cells. The results suggest that Fes acts as a downstream molecule of PEDF-driven signaling and expression of kinase-inactive Fes can exert a dominant negative effect. We previously showed that FGF-2-induced capillary morphogenesis by IBE cells and HUVECs was sensitive to the Src-family-kinase inhibitor PP2 (Kanda et al., 2003b). We examined FGF-2-induced capillary morphogenesis of IBE cell lines expressing either kinase-inactive Src or kinase-inactive Fyn (Tsuda et al., 2002). We observed that the latter, but not the former, failed to promote capillary morphogenesis by FGF-2 treatment (data not shown). We then examined the kinase activity of Fyn in parental IBE cells and KE5-15 cells. We examined the activity in cells cultured for 2 hours, because kinetic study of capillary morphogenesis (Kanda et al., 2004a) demonstrated that FGF-2-treated cells showed efficient extension, which is an important morphological change for this assay. As shown in Fig. 3B, FGF-2 activated Fyn in these cells. In the presence of PEDF, Fyn activity was downregulated in parental IBE cells but not in KE5-15 cells. This result supports the notion that PEDF might inhibit Fyn activity through Fes. We also examined the kinase activity of Fyn in HUVECs. As shown in Fig. 3C, FGF-2-induced Fyn activity was also inhibited by PEDF in HUVECs. One of the mechanisms involved in the downregulation of the activity of Src family kinases is the phosphorylation of tyrosine residue located at the C-terminus of kinases (the so-called inhibitory tyrosine) by CSK (Okada and Nakagawa, 1989). Phosphorylation of this residue allows C-terminal tail of kinases to associate with its own Src-homology-2 domain to form a closed, inactive conformation (Xu et al., 1997). The C-terminal amino acids around the inhibitory tyrosine of human Src and mouse Fyn have a close sequence homology. This suggests that phosphorylation-specific antibody against Y529-phosphorylated Src might also recognize the phosphorylated Y531 of mouse Fyn (which corresponds to Y529 of human Src). We incubated Src immunoprecipitated from HUVECs and Fyn immunoprecipitated from IBE cells with recombinant CSK and the tyrosine phosphorylation of C-termini was examined by immunoblot analysis using anti-Y529-phosphorylated-Src antibody. As shown in Fig. 3D, Src and Fyn phosphorylated by CSK were similarly recognized by the antibody. We then examined the phosphorylation status of Fyn in IBE and KE5-15 cells. As shown in Fig. 3E, PEDF induced phosphorylation of tyrosine at 531 of Fyn in IBE cells but not in KE5-15 cells. These results indicate that PEDF-mediated inhibition of Fyn was due to the phosphorylation of the inhibitory tyrosine at C-terminus and that the phosphorylation was dependent on Fes activity. We next examined whether Fes can phosphorylate the inhibitory tyrosine of Fyn. FLAG-tagged wild-type or kinase-inactive Fes was immunoprecipitated from cells cultured on Matrigel in the presence or absence of PEDF and phosphorylation of recombinant Fyn by Fes was examined in vitro. As shown in Fig. 3F, recombinant CSK phosphorylated the inhibitory tyrosine of Fyn. Kinase-inactive Fes did not phosphorylate the tyrosine. Wild-type Fes induced the phosphorylation of this tyrosine even in the absence of PEDF treatment. This result suggests that wild-type Fes can phosphorylate the inhibitory tyrosine of Fyn. Although Fes activity was increased in PEDF-treated cells (Fig. 3A), isolated Fes from PEDF-treated cells did not further enhance the tyrosine phosphorylation of Fyn. When compared with kinase-inactive Fes, wild-type Fes isolated from untreated cells showed higher autophosphorylation than kinase-inactive Fes, indicating that a part of isolated wild-type Fes is already active to some extent. To increase the tyrosine phosphorylation of Fyn by PEDF-treatment in vivo (Fig. 3E), certain mechanisms might be required, including complex formation by signaling molecules (some scaffold proteins and/or protein tyrosine phosphatases), which are not reconstituted in vitro.

We next examined the cellular distribution of Fes and Fyn. Cells were cultured on the surface of Matrigel-coated cover slips for 2 hours with or without FGF-2 or PEDF. Both wild-type and kinase-inactive Fes distributed in the cytoplasm. Neither FGF-2 nor PEDF affected the distribution (data not shown). Fes was also localized to the periphery of cells. Double staining with anti-vinculin antibody indicated that a proportion of Fes at the cellular periphery colocalized with the vinculin, suggesting that Fes was incorporated into focal adhesions (not shown). We next examined the localization of Fes and Fyn. FGF-2 treatment did not modulate the distribution of Fyn in IBE cells cultured on Matrigel-coated surface (not shown). However, PEDF promoted the localization of Fyn at the cellular periphery in a manner dependent on the kinase activity of Fes (Fig. 4A). This result might indicate that PEDF induced tyrosine phosphorylation of Fyn by modulation of its localization. We also examined the distribution of CSK in IBE cells. Localization of CSK was diffusely cytoplasmic and PEDF never changed its localization (Fig. 4B). In addition, CSK did not colocalize with vinculin, suggesting that CSK did not play its regulatory roles on Src family kinases at focal adhesions.

We constructed a cDNA encoding constitutively active human Fyn (Y531F) by site-directed mutagenesis and its sequence was confirmed by sequencing. The construct was subcloned into pcDNA3.1/Myc.His plasmid and transfected into IBE cells. 18 G418-resistant clones were obtained and two clones were found to express Myc-tagged Fyn, denoted CAFyn-6 and CAFyn-12 cells, respectively (Fig. 5A). Using these stable cell lines, we examined the effect of PEDF on capillary morphogenesis. As shown in Fig. 5B, these cells exhibited capillary morphogenesis in the absence of FGF-2. FGF-2 further induced capillary morphogenesis, suggesting the involvement of signaling pathways other than Fyn. PEDF failed to inhibit both spontaneous and FGF-2-induced capillary morphogenesis. This result reveals that Fyn is one of the targets of PEDF to inhibit capillary morphogenesis of IBE cells.

Recently, it has been shown that p190RhoGAP is tyrosine phosphorylated by Fyn, resulting in the increase in GTPase activity (Wolf et al., 2001). p190RhoGAP downregulates Rho activity by hydrolysing GTP. We have demonstrated that expression of dominant negative Rho or treatment with Rho-associated protein kinase (ROCK) inhibitor promoted FGF-2-independent capillary morphogenesis of IBE cells through inhibition of focal-adhesion disassembly (Kanda et al., 2004b). We then examined the tyrosine phosphorylation of p190RhoGAP in IBE cells and a stable cell line expressing kinase-inactive Fyn (KDFyn-8 cells). As shown in Fig. 6, FGF-2 enhanced tyrosine phosphorylation, which was inhibited by PEDF. In KDFyn-8 cells, FGF-2 failed to increase tyrosine phosphorylation of p190RhoGAP. In FGF-2-treated cells, downregulation of ROCK through Fyn might be responsible for capillary morphogenesis of IBE cells.

In the present study, we have demonstrated for the first time that PEDF inhibited FGF-2-induced capillary morphogenesis, possibly through Fes. This capillary morphogenesis assay involves elongated cellular morphogenesis, cell-cell contact and cell survival. PEDF did not induce endothelial cell death in the presence of FGF-2, indicating the lack of the involvement of apoptosis. In IBE cells, PEDF downregulated Fyn activity by phosphorylation of inhibitory tyrosine residue located at the C-terminus, which was not observed in cells expressing dominant negative Fes. These results suggest that PEDF inhibits capillary morphogenesis of endothelial cells by suppressing Fyn.

Fes is a non-receptor protein tyrosine kinase and its activation is strictly regulated. Extracellular stimuli induce oligomerization of Fes followed by autophosphorylation and activation (Smithgall et al., 1998). In the N-terminal unique region of Fes, there are two coiled-coil domains and mutation of the first coiled-coil domain resulted in enhanced kinase activity, suggesting that this domain might have autoregulatory function (Cheng et al., 1999; Cheng et al., 2001). In the present study, we show that PEDF also activated Fes in IBE cells cultured on Matrigel (Fig. 3A). In a cell line expressing kinase-inactive Fes (KE5-15 cells), PEDF failed to stimulate Fes activity. Ectopic expression of kinase-inactive Fes in endothelial cells exerted a dominant negative effect on endogenous Fes by oligomerization to block autophosphorylation in trans (Kanda et al., 2000). Thus, it seems likely that the mechanism underlying the activation of Fes by PEDF also involves oligomerization followed by autophosphorylation.

Our previous studies using stable endothelial cell lines expressing kinase-inactive Fes showed that Fes was involved in angiopoietin-2-, sonic-hedgehog- and stromal-cell-derived-factor-1α-induced activation of phosphoinositide-3-kinase (PI3-kinase) (Mochizuki et al., 2002; Kanda et al., 2003a; Kanda et al., 2003b). However, FGF-2-activated PI3-kinase was independent of Fes (Kanda et al., 2000) and PI3-kinase inhibitors did not inhibit FGF-2-induced capillary morphogenesis (Kanda et al., 2003b). In the present study, activation of PI3-kinase or phosphorylation of Akt was not observed in cells treated with PEDF (data not shown). In IBE cells cultured on collagen gels, angiopoietin 2 activated Fyn but not Fes (Mochizuki et al., 2002). By contrast, Fes was activated by FGF-2 and angiopoietin 2 in cells cultured on fibronectin-coated surface but not in cells cultured on type-I collagen gels or Matrigel (Kanda et al., 2000) (S.K. et al., unpublished). In cells on Matrigel, PEDF induced Fes activation. It is well known that different adhesion receptors (integrins) and growth factors produce unique sets of signaling molecules (Sastry and Horwitz, 1996). Thus, distinct compartmentalization of signaling molecules plausibly explains the difference in target molecule of Fes, such as PI3-kinase and Fyn.

The mechanism that increases the phosphorylation of the C-terminal inhibitory tyrosine residue of Fyn in PEDF-treated cells might involve the direct phosphorylation of Fyn by Fes. One well-recognized kinase responsible for this phosphorylation is CSK. Because CSK is constitutively active, the function of CSK is dependent on its localization. There are several proteins responsible for the recruitment of CSK to the particular sites, such as caveolin, paxillin and CSK-binding protein (Cbp)/phosphoprotein associated with glycosphingolipid-enriched microdomains (PAG) (Cao et al., 2002; Harder et al., 1998; Sabe et al., 1995; Shima et al., 2003). We examined the association of CSK and Fyn with these proteins by immunoprecipitation followed by immunoblotting. However, we could not observe the significant alteration of association in parental IBE cells and KE5-15 Fes cells by PEDF treatment (data not shown). CSK co-precipitated with caveolin was not affected by PEDF treatment and association of Fyn with caveolin was decreased in IBE cells and increased in KE5-15 Fes cells in response to PEDF treatment, suggesting that caveolin-bound CSK was dissociated from Fyn by PEDF treatment. Association of paxillin with Fyn was under the detectable level in these cells and binding of CSK to paxillin was not altered by PEDF treatment. In octylglucoside-extracted proteins from both IBE cells and KE5-15 Fes cells, tyrosine phosphorylation of Cbp/PAG was not observed by PEDF treatment and the association with neither CSK nor Fyn was affected by PEDF treatment. In addition, indirect immunofluorescent staining of CSK in IBE cells demonstrated that CSK diffusely distributed in the cytoplasm without any changes in its localization being caused by PEDF treatment (Fig. 4B). These results suggest that CSK did not contribute to PEDF-induced tyrosine phosphorylation of Fyn. On the contrary, PEDF treatment induced the colocalization of Fyn and Fes to focal adhesions in WT6-8 Fes cells but not KE5-15 Fes cells (Fig. 4A). Furthermore, wild-type Fes could phosphorylate the inhibitory tyrosine of Fyn (Fig. 3F). In proteins immunoprecipitated with anti-FLAG antibody, we did not detect co-precipitated CSK by immunoblotting (data not shown). These observations suggest that inhibition of Fyn activity might be due to the direct phosphorylation of tyrosine by Fes. It should be elucidated whether peripheral localization induced tyrosine phosphorylation of Fyn or whether phosphorylated Fyn localized to the periphery of cells in response to PEDF treatment.

The localization of Fes is rather controversial. Previous reports indicated that Fes localized to the nucleus (Yates et al., 1995). Recent studies have shown that Fes exists in the cytoplasm: one report demonstrated that Fes was present in the trans-Golgi network (Zirngibl et al., 2001), and another showed that Fes localized to the central region of microtubules (Kogata et al., 2003). Our study showed that Fes is located in the cytoplasm and the peripheral region of cellular protrusions and focal adhesions. The difference of the localization between these reports and ours might be explained by the difference of cell types and culture conditions. Fes was observed in the cytoplasm and especially at the trans-Golgi network in Cos-1 cells transfected with Fes (Zirngibl et al., 2001). Kinase activity was required to locate at the Golgi network. In human aortic endothelial cells, Fes also showed the cytoplasmic distribution and localized to microtubules (Kogata et al., 2003). In these studies, cells were examined that had been cultured on the surface of glass for a long time. By contrast, we studied cells cultured on a Matrigel-coated glass surface for only 2 hours in the absence of serum. These culture conditions might enhance the effect of integrin/extracellular-matrix interactions on the distribution of signaling molecules, such as accumulation of these molecules into focal adhesions (Giancotti and Ruoslahti, 1999). According to our previous study, FGF-2-induced capillary morphogenesis required early-phase elongated morphogenesis of endothelial cells on Matrigel (Kanda et al., 2004a). This morphological change is supported by the stabilization of focal-adhesion formation, which can be achieved by the inhibition of Rho-associated kinase (Kanda et al., 2004b). In the present study, Fes showed a cytoplasmic distribution and also localized to the periphery of cells, such as focal adhesions and the tip region of cellular protrusions (Fig. 4A). Neither PEDF nor FGF-2 modulated the localization of either wild-type or kinase-inactive Fes. However, PEDF, but not FGF-2 (not shown), induced the peripheral localization of Fyn in IBE cells cultured on Matrigel. Thus, it seems that PEDF might cause inactivation of Fyn by alterating its localization.

It has been shown that p190RhoGAP was tyrosine phosphorylated by Fyn. Rho-family GTPase proteins have intrinsic GTPase activity. p190RhoGAP stimulates Rho GTPase activity, which in turn hydrolyses GTP and inactivates Rho. Rho-family GTPases play crucial roles in the regulation of cell motility and morphogenesis (Etienne-Manneville and Hall, 2002; Burridge and Wennerberg, 2004). In migration, stress-fiber formation and focal-adhesion disassembly in the rear of cells are important, whereas morphogenesis requires stabilized focal adhesions. In endothelial cells, we and others have shown that the inhibition of Rho or ROCK induces capillary morphogenesis by impairing focal-adhesion disassembly (Connolly et al., 2002; Kanda et al., 2004b). Therefore, inhibition of Fyn kinase by PEDF followed by the impaired tyrosine phosphorylation of p190RhoGAP might not downregulate Rho, resulting in the lack of stability of focal adhesions.

PEDF-mediated downregulation of Fyn through Fes is a novel signaling pathway. Although the role of Fyn in FGF-2-induced angiogenesis in vivo has not yet been clarified, pharmacological inhibition of Src-family kinases has been shown to block FGF-2-induced angiogenesis in vivo (Kilarski et al., 2003). We suggest that PEDF-mediated downregulation of Fyn might be at least partly involved in its anti-angiogenic action in vivo.

We are grateful to T. Shimogama, M. Yoshimoto and members of Nagasaki University Radioisotope Center for their outstanding help. This work was supported by a grant from the Japan Society for the Promotion of Science.