Vascular Endothelial Growth Factor Signals Endothelial Cell Production of Nitric Oxide and Prostacyclin through Flk-1/KDR Activation of c-Src*

Vascular endothelial growth factor (VEGF) is a potent endothelial cell-specific mitogen that promotes angiogenesis, vascular hyperpermeability, and vasodilation by autocrine mechanisms involving nitric oxide (NO) and prostacyclin (PGI2) production. These experiments used immunoprecipitation and immunoassay procedures to characterize the signaling pathways by which VEGF induces NO and PGI2 formation in cultured endothelial cells. The data showed that VEGF stimulates complex formation of the flk-1/kinase-insert domain-containing receptor (KDR) VEGF receptor with c-Src and that Src activation is required for VEGF induction of phospholipase C γ1 activation and inositol 1,4,5-trisphosphate formation. Reporter cell assays showed that VEGF promotes a ∼50-fold increase in NO formation, which peaks at 5–20 min. This effect is mediated by a signaling cascade initiated by flk-1/KDR activation of c-Src, leading to phospholipase C γ1 activation, inositol 1,4,5-trisphosphate formation, release of [Ca2+] i and nitric oxide synthase activation. Immunoassays of VEGF-induced 6-keto prostaglandin F1α formation as an indicator of PGI2 production revealed a 3–4-fold increase that peaked at 45–60 min. The PGI2 signaling pathway follows the NO pathway through release of [Ca2+] i , but diverges prior to NOS activation and also requires activation of mitogen-activated protein kinase. These results suggest that NO and PGI2 function in parallel in mediating the effects of VEGF.

The physiological effects of VEGF on endothelial cells are well established, but the postreceptor signaling pathways are not yet fully understood. VEGF-receptor binding triggers a signaling cascade that results in tyrosine phosphorylation of phospholipase C␥1 (PLC␥1), leading to increases in intracellular levels of inositol 1,4,5-trisphosphate (1,4,5-IP 3 ) and elevation of intracellular calcium (6,7). The increase in intracellular calcium activates nitric oxide synthase (NOS) to produce nitric oxide (NO). NO formation activates guanylate cyclase within vascular smooth muscle cells and endothelial cells, causing cGMP production. This NO/cGMP cascade is thought to have an important role in the vasoactive effects of VEGF. NO production and elevation of cGMP levels have been found to contribute to the effects of VEGF on vascular tone (8). The effects of VEGF in stimulating angiogenesis and increasing vascular permeability also require NOS activity (9 -13). In addition, VEGF-induced activation of the mitogen-activated protein kinase (MAPK) cascade has recently been shown to involve NOSdependent signaling events (14,15). The NOS/guanylate cyclasedependent activation of the MAPK cascade is thought to lead to VEGF-induced proliferation of endothelial cells (16). VEGF has recently been found induce the nuclear translocation of endothelial NOS together with Flk-1/KDR and caveolin-1 (17), suggesting a possible role for NO in transcription factor activation.
Another pathway that appears to be involved in mediating the vasoactive effects of VEGF is the prostacyclin (PGI 2 ) release pathway. VEGF induces PGI 2 production via activation of phospholipase A 2 as a consequence of initiation of the MAPK cascade (18). Although the relationship between NO-mediated signaling events and PGI 2 production in the VEGF signal transduction cascade are not yet known, results of in vivo analysis have suggested that vascular hyperpermeability induced by VEGF results from the synergistic action of both NO and PGI 2 (12). Therefore, the fact that PGI 2 , like NO, has vasodilating effects and has been shown to stimulate vascular hyperpermeability and angiogenesis under some conditions (19,20) suggests that both molecules are likely to be involved in transducing the vascular effects of VEGF.
Finally, another group of signaling molecules that may be involved in the VEGF signaling cascade is the Src family tyrosine kinases. c-Src and Src family proteins have been shown to interact functionally with the transmembrane tyrosine kinase receptors for several growth factors, including platelet-derived growth factor (PDGF) (21), epidermal growth factor (EGF) (22), basic fibroblast growth factor (23), and colony-stimulating factor-I (24). These interactions are thought to modulate growth factor signaling due to the mutual stimulation of catalytic activity and enhanced phosphorylation of downstream targets of each protein tyrosine kinase. Phosphorylation of specific tyrosine residues in receptors promotes their interactions with a variety of proteins containing SH2 domains (the SH2 domain is a conserved sequence of ϳ100 amino acids with homology for region 2 of Src family proteins) (25). VEGF has been shown to promote tyrosine phosphorylation of several mediators of signal transduction that contain SH2 domains, including PLC␥1, phosphoinositide 3-kinase, GTPase activating protein, and the oncogenic adapter protein Nck (7,26,27). However, although the VEGF receptors have been shown to interact with SH2 domain-containing proteins, their specific interactions with Src family tyrosine kinases remain unclear. Studies with stably transfected cell lines expressing either flt-1 or flk-1/KDR have shown weak association between the Src family proteins Fyn and Yes with flt-1 but not flk-1/KDR (26). Work done using sinusoidal endothelial cells and flk-/KDR expressing NIH3T3 fibroblasts failed to show activation of the Src family members c-Src, Fyn, Lyn, or Yes upon VEGF stimulation (29).
The goal of the present study was to determine the role of Src family proteins in the VEGF signal transduction process and to characterize the signaling pathways underlying VEGF-activated production of NO and PGI 2 . Therefore, experiments in primary cultures of bovine aortic endothelial cells (BAECs) were designed to determine the effects of VEGF on NO and PGI 2 production in relation to the activity of Src family proteins, tyrosine kinases, PLC␥1, PKC, NOS, and the MAPK cascade.

EXPERIMENTAL PROCEDURES
Materials-GF 109203X (GFX) and U73122 were obtained from BI-OMOL (Plymouth Meeting, PA). AG-490, PP2, and PD98059 were from Calbiochem (La Jolla, CA). The cGMP radioimmunoassay kit was from PerSeptive Biosystem (Framingham, MA). The 1,4,5-IP 3 radioimmunoassay kit was from NEN Life Science Products. The 6-keto-prostaglandin F 1␣ enzyme immunoassay system was from Amersham Pharmacia Biotech. Recombinant human VEGF 165 was from R&D systems (Minneapolis, MN). Human VEGF 165 has been found to be highly potent as an angiogenic and permeability increasing factor in bovine endothelial cell systems (7,13,17). The anti-phosphotyrosine monoclonal antibody (PY 20 ) was from Transduction Laboratories (Lexington, KY). Antiflk-1 polyclonal antibody was from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-c-Src and anti-PLC␥1 antibody were from Upstate Biotechnology (Lake Placid, NY). Other reagents were from Sigma. Rat fetal lung fibroblast (RFL-6) cells were obtained from American Type Culture Collection.
Measurement of 1,4,5-IP 3 Production-Cells were lysed in 1 M trichloroacetic acid. Lysates were used to measure 1,4,5-IP 3 using a radioimmunoassay kit according to the manufacturer's recommended methods.
Measurement of 6-keto PGF 1␣ Release-The accumulation of 6-keto PGF 1␣ was used as an index of PGI 2 release. Confluent monolayers of BAECs were incubated overnight in serum-free M199. The next day, cultures were treated with inhibitors for 30 -60 min. Then, VEGF was added to the wells for 10 min, and then 500 l of conditioned medium aliquots were removed for 6-keto PGF 1␣ analysis. The amount of 6-keto PGF 1␣ in the medium was measured using an enzyme immunoassay kit according to the manufacturer's specifications.
Measurement of cGMP-RFL-6 reporter cell assays were done as described (31). Briefly, confluent BAEC cultures were maintained in serum-free medium overnight and then equilibrated in Locke's buffer (154.0 mM NaCl, 5.6 mM KCl, 2.0 mM CaCl 2 , 1.0 mM MgCl 2 , 3.6 mM NaHCO 3 , 5.6 mM glucose, and 10.0 mM Hepes (pH 7.4)). All treatment solutions were prepared with the same buffer. For experiments, BAECs were pretreated with inhibitors or vehicle alone and then with VEGF with or without inhibitors. Immediately before treatment, 0.3 mM 3-isobutyl-1-methylxanthine and 100 units/ml superoxide dismutase were added to the cultures. After treatment, the conditioned media were quickly transferred to RFL-6 cultures. After 3 min at 37°C, the conditioned medium was replaced by ice-cold 25 mM sodium acetate buffer (pH 3.3). Cells were frozen overnight at Ϫ20°C. The next day, the samples were collected and stored at Ϫ70°C until analyzed for cGMP using a cGMP 125 I radioimmunoassay kit.
Immunoprecipitation and Immunoblotting-BAECs were cultured to 95% confluence, switched to serum-free medium for 16 -18 h, and treated with VEGF in the presence or absence of PP2 or sodium orthovanadate (Na 3 VO 4 , 0.1 mM). Incubations were terminated by aspirating the medium and washing the cells twice with ice-cold phosphate buffer saline containing 2 mM Na 3 VO 4 and 1 mM phenylmethylsulfonyl fluoride. Immunoprecipitation and immunoblotting were done as described previously (32)(33)(34).
Data Analysis-The data were analyzed by analysis of variance and Dunnett's post hoc comparison test for comparing several treatment groups with a control (35). The results are expressed as the mean Ϯ S.E.

RESULTS AND DISCUSSION
VEGF-stimulated Activation of PLC␥1 Requires the Activity of Src Family Tyrosine Kinases-We first investigated the involvement of Src family tyrosine kinases in the VEGF signal transduction cascade. Because tyrosine phosphorylation of PLC␥1 is a well established early event in the VEGF signal transduction cascade (29,36), we tested the effects of the Src family tyrosine kinase inhibitor PP2 (37) on VEGF-induced tyrosine phosphorylation of PLC␥1 by using immunoprecipitation and immunoblot techniques. The data confirmed a VEGFinduced increase in PLC␥1 tyrosine phosphorylation. The response reached maximal levels at 5-20 min and returned to the basal level within 30 -60 min (Fig. 1). This effect was totally blocked by PP2, suggesting that the activity of members of the Src kinase family is required for the response.
To further test the role of the Src family proteins in the activation of VEGF of PLC␥1, we determined the effects of PP2 and the broad spectrum tyrosine kinase inhibitor genistein on VEGF-induced production of 1,4,5-IP 3 . These experiments showed a significant increase in 1,4,5-IP 3 production within 2 min of VEGF addition. The increase peaked at 10 min and then declined, returning to control levels at 60 min. This effect was blocked completely by both PP2 and genistein (Fig. 2), indicating that Src-dependent tyrosine phosphorylation is also required for 1,4,5-IP 3 production induced by VEGF, providing further evidence that a member or members of the Src family of tyrosine kinases plays a role in the VEGF signaling cascade.
Previous studies done with rat sinusoidal endothelial cells and NIH3T3 cells expressing flk-1/KDR have shown that PLC␥1 is a major substrate for flk-1/KDR (29). Because flk-1/ KDR is known to generate the major part of mitogenic signals in endothelial cells, we next tested whether c-Src interacts with flk-1/KDR by immunoprecipitating with anti-flt-1/KDR antibodies and immunoblotting with precipitated proteins with anti-c-Src antibodies. As shown in Fig. 3, this analysis revealed that VEGF treatment stimulated formation of a c-Src complex with flk-1/KDR. This association was evident within 1 min of VEGF addition and increased as the treatment continued. The early onset of this effect and its close temporal correspondence with that for the PLC␥1 activation strongly suggests that VEGF-induced tyrosine phosphorylation of flk-1/KDR directly activates c-Src.
Our data are the first to suggest that c-Src is activated by VEGF. The relationship between c-Src activation and the physiological actions of VEGF is not yet understood, but it appears very likely that c-Src activation has an important role in coordinating the effects of VEGF on growth with its effects on cell adhesion and cell motility. Tyrosine phosphorylation of receptors for both PDGF and EGF has also been found to promote formation of receptor complexes with Src family proteins, and consequent activation of Src family kinases has been implicated in cell proliferation events. For example, Src family kinases have been found to be required for PDGF-induced progression through the G1 stage of the cell cycle and entrance into S phase (38). Analyses of interactions between the EGF receptor and mutated Src family proteins indicate that c-Src is also required for EGF-induced mitogenesis and that Src may mediate regulation of a cell cycle checkpoint (39). Src activation has also been implicated in regulation of cell motility by both PDGF and EGF due to complex formation between Src and the focal adhesion tyrosine kinase FAK, together with the focal adhesion-associated protein paxillin (25). VEGF is also known to stimulate tyrosine phosphorylation of FAK and paxillin (40). Thus, c-Src may serve to link the VEGF receptor and adhesion receptor signaling pathways in vascular endothelial cells, as has been suggested for PDGF and EGF in other cell types.
VEGF Stimulation of NO Formation Requires the Activity of Tyrosine Kinases and Src Family Proteins-The next series of experiments analyzed the relationship between VEGF-induced c-Src activation and NO production. It is well established that NO-induced vasodilation occurs due to increased cGMP formation through activation of soluble guanylate cyclase in vascular smooth muscle cells. RFL-6 cells have been shown to respond to soluble guanylate cyclase activators and can be used as re-porter cells to detect NO release by other cells under various experimental conditions (31). Therefore, NO released by BAECs was analyzed by stimulating RFL-6 cells with BAEC conditioned medium and then measuring cGMP formation in the RFL-6 cultures.
Experiments testing the effects of different concentrations of VEGF on NO release showed that the VEGF-induced NO release was concentration-dependent (Fig. 4). The absolute measurements of cGMP levels varied somewhat between different endothelial cell isolations and batches of RFL-6 reporter cells, but VEGF concentrations of 10 ng/ml always induced a significant increase in NO release. Therefore, this concentration was used for all other experiments.
The release of NO induced by VEGF was also time-dependent (Fig. 5). NO levels were substantially increased within 5 min of VEGF treatment. The increase was sustained for 20 min and then declined, but NO remained significantly above basal levels for up to 60 min. Pretreatment with genistein or PP2 completely blocked this effect. Genistein or PP2 alone had no effect on basal NO release or on reporter cell cGMP production. Control studies also showed that cGMP production was completely blocked by L-NAME and that neither VEGF or L-NAME had any direct effect on cGMP production by RFL-6 cells. These data demonstrate that the VEGF-induced increase in NO release is dependent on tyrosine kinase activity and specifically on activity of the Src family kinases.
The results of the reporter cell assays are consistent with previous studies that directly measured VEGF-induced cGMP accumulation in human umbilical endothelial cells (41) and a cell line of coronary venule endothelial cells (9, 10). We have also obtained similar results using an NO electrode to directly measure NO accumulation within the media of VEGF-treated BAECs. 2 The change in the amount of NO in the BAEC conditioned medium must be the net result of formation, release, and degradation. The decline in NO at later times indicates that NO is continuously degraded, even in the presence of a high concentration of superoxide dismutase (100 units/ml). However, the fact that the VEGF-induced release of NO remains above the basal levels for up to 60 min after VEGF addition indicates a sustained effect of the growth factor on NOS activity. The overall pattern of VEGF release shown by our experiments is quite consistent with the pattern of VEGF-induced increases in intracellular calcium shown by previous work done in human umbilical vein endothelial cells. Those studies showed that VEGF induced an increase in intracellular calcium levels that was preceded by a latent period of 15 s, peaked at ϳ60 s, and was sustained for up to 90 min (6, 8). Specificity of the VEGF effects on NO release was demonstrated by experiments comparing the effects of different inhibitors on VEGF-induced formation of NO using the calcium ionophore A23187 as positive control. The NOS inhibitor L-NAME (42), the PKC inhibitor GFX (43), the PLC inhibitor U73122 (44), the specific MAPK kinase (MEK) inhibitor PD98059 (45), or the JAK2 inhibitor AG490 (46) was used to pretreat the cells before stimulation with VEGF. As expected, inhibition of NOS by L-NAME completely blocked NO formation (9,10,41,47). Furthermore, inhibition of PKC or PLC activity also completely blocked VEGF-stimulated NO release (Fig. 6). In contrast, the inhibition of MEK and JAK2 had no effect. These results suggest that PKC and PLC are involved in VEGF activation of the NOS/guanylate cyclase pathway, whereas MEK and JAK2 are not.
To determine which VEGF receptor is involved in NOS activation and NO production, we tested the effects of placental growth factor (PlGF). PlGF is a member of the VEGF family of proteins that binds flt-1 with high affinity but fails to bind flk-1/KDR (48). The results showed that PlGF did not increase cGMP production (Fig. 7). Thus, activation of the flk-1/KDR receptor is responsible for VEGF-induced NO production, whereas flt-1 is not involved.
Increased Intracellular Calcium Is Required for VEGF-induced cGMP Production-In endothelial cells, VEGF-induced activation of PLC␥1 leads to formation of 1,4,5-IP 3 , which induces increases in intracellular calcium levels and activation of calcium/calmodulin-dependent protein kinases (27,49). The interdependence of tyrosine phosphorylation and calcium signaling in endothelial cells is well established (50,51). Stimulation of NOS activation and NO formation by increased intracellular Ca 2ϩ is also well known (52). To determine the source of the Ca 2ϩ increase responsible for VEGF-induced NO production, we studied the effects of the following inhibitors on VEGFinduced cGMP production by reporter cells: TMB-8 (intracellular Ca 2ϩ channel blocker) (53), BAPTA/AM (intracellular Ca 2ϩ chelator) (54), EGTA (extracellular Ca 2ϩ chelator), and verapamil (extracellular Ca 2ϩ channel blocker) (55). As shown in FIG. 7. The effects of PlGF on NO production. BAECs were pretreated with or without L-NAME (1 mM) for 60 min and then treated with or without VEGF (10 ng/ml) or PlGF (30 ng/ml) for 10 min. NO formation in each treatment condition was analyzed by using a reporter cell assay of cGMP formation. Data shown are mean Ϯ S.E. for three experiments.  Fig. 8, TMB-8 and BAPTA/AM attenuated VEGF-induced NO production, whereas EGTA and verapamil had no effect. These data indicate that Ca 2ϩ release from internal stores is required for VEGF-induced activation of NOS. This conclusion is supported by results of previous experiments analyzing VEGF effects on cGMP formation in human umbilical vein endothelial cells, which showed that BAPTA/AM totally blocked the VEGFinduced increase in cGMP (41). Previous analysis of VEGF effects on intracellular Ca 2ϩ accumulation have shown that the initial VEGF-induced increase in intracellular Ca 2ϩ results from a combination of influx of external Ca 2ϩ and Ca 2ϩ release from internal stores and that sustained elevation of intracellular Ca 2ϩ levels requires the influx of external Ca 2ϩ (6). The lack of inhibitory effects of verapamil and EGTA shown in our experiments indicates that influx of extracellular Ca 2ϩ into the cell is not required for NOS activation under acute conditions (5 min of VEGF treatment). However, influx of external Ca 2ϩ would probably be needed to replenish the intracellular Ca 2ϩ if treatment were continued longer.
VEGF-induced PGI 2 Production Is Independent of NOS Activity-PGI 2 , like VEGF, is known to be involved in the regulation of vascular permeability and angiogenesis (19,20). VEGF has been shown to stimulate PGI 2 production in human umbilical vein endothelial cells via activation of flk-1/KDR (18). NOS activation in bovine coronary microvessel endothelial cells has also been found to increase their PGI 2 production (28). Thus, it was hypothesized that VEGF causes PGI 2 production via the NOS/guanylate cyclase. To test this, BAECs were pretreated for 30 -60 min with inhibitors (PP2, genistein, GFX, BAPTA/AM, PD98059, and L-NAME), incubated with VEGF, and analyzed for PGI 2 production. The data showed that VEGF stimulated a time-dependent increase in PGI 2 release that peaked at 45-60 min as described previously (18). The VEGF effect was inhibited by PP2, genistein, GFX and BAPTA/AM and by the specific MEK inhibitor PD98059 (Fig. 9), indicating that PGI 2 production is downstream from PKC activation in the VEGF signaling pathway and is also downstream from increased intracellular Ca 2ϩ and the MAPK cascade. However, the NOS inhibitor L-NAME had no effect on PGI 2 production, suggesting that VEGF signaling leading to PGI 2 production diverges subsequent to the increase in intracellular calcium and prior to NOS activation.
VEGF stimulation of the MAPK pathway is well established (14), but the relationship between the MAPK cascade and NOS activation is not yet clear. Results of work done using late passage coronary vein endothelial cells indicated that the NOS/ guanylate cyclase pathway lies upstream of the MAPK cascade (16). However, others have also shown that activation of the MAPK cascade by shear stress parallels NO production. Our observation that VEGF-induced production of PGI 2 was blocked by the MEK inhibitor but not by the NOS inhibitor indicates that this effect is mediated by the MAPK pathway independent of NOS.
Based on the above results, we propose a working model for VEGF signaling leading to increased NO and PGI 2 formation: VEGF binding to flk-1/KDR results in formation of receptor complexes with c-Src and c-Src activation. Activated c-Src activates PLC␥1, which leads to formation of diacylglycerol and 1,4,5-IP 3 . Diacylglycerol activates PKC, and 1,4,5-IP 3 stimulates Ca 2ϩ release from internal stores, which activates NOS, leading to NO production. The signaling pathway for VEGF stimulation of PGI 2 production follows the above pathway though the level of intracellular Ca 2ϩ release but diverges prior FIG. 9. The effect of VEGF on PGI 2 production. The accumulation of 6-keto PGF 1␣ in the supernatant of cells stimulated in six-well plates was used as an index of PGI 2 release. BAECs were pretreated with PP2, genistein, GFX, BAPTA/AM, PD98059, and L-NAME for 30 -60 min. Then VEGF (10 ng/ml) was added to the wells for 10 min. to initiation of the NOS/guanylate cyclase and MAPK cascades (Fig. 10). Our model, therefore, suggests that VEGF-induced NO and PGI 2 formation occur in parallel. This model, however, does not necessarily imply that direct phosphorylation or interactions occur between each of the indicated components in either pathway.