Identification of Tek/Tie2 Binding Partners

The Tek/Tie2 receptor tyrosine kinase plays a pivotal role in vascular and hematopoietic development. To study the signal transduction pathways that are mediated by this receptor, we have used the yeast two-hybrid system to identify signaling molecules that associate with the phosphorylated Tek receptor. Using this approach, we demonstrate that five molecules, Grb2, Grb7, Grb14, Shp2, and the p85 subunit of phosphatidylinositol 3-kinase can interact with Tek in a phosphotyrosine-dependent manner through their SH2 domains. Mapping of the binding sites of these molecules on Tek reveals the presence of a multisubstrate docking site in the carboxyl tail of Tek (Tyr1100). Mutation of this site abrogates binding of Grb2 and Grb7 to Tek in vivo, and this site is required for tyrosine phosphorylation of Grb7 and p85 in vivo. Furthermore, stimulation of Tek-expressing cells with Angiopoietin-1 results in phosphorylation of both Tek and p85 and in activation of endothelial cell migration and survival pathways that are dependent in part on phosphatidylinositol 3-kinase. Taken together, these results demonstrate that Angiopoietin-1-induced signaling from the Tek receptor is mediated by a multifunctional docking site that is responsible for activation of both cell migration and cell survival pathways.

Receptor tyrosine kinases (RTKs) 1 are cell surface proteins that receive cues from extracellular growth factors to ultimately control biological responses such as cell growth, differentiation, and survival. Signal transduction pathways mediated by RTKs are initiated upon ligand binding with subsequent receptor-mediated dimerization and autophosphorylation on specific tyrosine residues (1). These phosphoty-rosine residues serve as high affinity binding sites for numerous intracellular signaling molecules that contain Src homology 2 (SH2) or phosphotyrosine binding domains (2). Such specialized protein modules are found in various classes of RTK targets including the p85 subunit of the lipid kinase phosphatidylinositol (PI) 3-kinase, the protein tyrosine phosphatase Shp2, the docking molecules Shc and insulin receptor substrate-1, and the adaptor proteins Grb2, Grb7, and Grb14. Additional modular units such as Src homology 3 and pleckstrin homology (PH) domains are also found in Grb2 and the Grb7/Grb14 family, respectively. A regulated cascade of protein-protein interactions allows other signaling molecules to be recruited and assembled into an organized biochemical network that leads to the activation of important signaling pathways within the cell.
Two subfamilies of RTKs have been identified whose expression is almost exclusively restricted to cells of the endothelial lineage (3,4). The first subfamily contains the high affinity vascular endothelial growth factor (VEGF) receptors (VEGFRs) known as VEGFR-1 (Flt-1), VEGFR-2 (KDR/Flk-1), and VEGFR-3 (Flt-4). To date, there are four distinct isoforms of VEGF denoted VEGF-A through VEGF-D that bind and activate different subsets of the VEGFRs (5). The second subfamily consists of the TIE receptors known as Tie (Tie1) and Tek (Tie2), which are structurally dissimilar from the VEGFRs. Whereas the ligand for Tie has yet to be identified, a family of ligands known as the angiopoietins have recently been shown to bind Tek (6 -8). Interestingly, these ligands appear to have opposing actions as Angiopoietin-1 (Ang1) and Angiopoietin-4 (Ang4) stimulate tyrosine phosphorylation of Tek, whereas Angiopoietin-2 (Ang2) and Angiopoietin-3 (Ang3) can inhibit this phosphorylation.
Genetic experiments have demonstrated that the signal transduction cascades that are mediated by each of these RTKs are distinct and that each pathway is critical for normal embryonic blood vessel development. VEGFR-1 and VEGFR-2 appear to be required during the initial developmental process of blood vessel formation known as vasculogenesis (9,10), whereby endothelial cell precursors differentiate and assemble into a primitive vascular network (11). This network is subsequently remodeled through the sprouting of new capillaries from pre-existing larger vessels by a process known as angiogenesis (11). VEGFR-3, Tie, and Tek have all been shown to be required for this later stage of blood vessel maturation (12)(13)(14)(15).
Inactivation of Tek signaling pathways results in early embryonic lethality as a result of angiogenic defects throughout the vascular system including a reduction in blood vessel integrity, decreased vessel sprouting, and a loss of endothelial cells (7,13,14,16). These analyses have revealed that activation of Tek by Ang1 appears to control at least two different signal transduction cascades that can effect changes in angiogenic sprout formation and survival of endothelial cells. Consistent with the role of Tek signaling pathways in vessel sprouting, a modified form of Ang1 known as Ang1* has recently been shown to initiate endothelial cell migration and sprouting in vitro (17)(18)(19). Furthermore, Tek has been shown to signal through a novel phosphotyrosine binding domain-containing docking molecule known as Dok-R that can associate with numerous signaling molecules that are thought to be involved in cell migration (20). Collectively these results suggest that Tek signaling through Dok-R may lead to alterations in the intracellular architecture of endothelial cells, which is critical for their directed migration during sprouting angiogenesis.
The identification of a unique family of Tek ligands with opposing functions implies that the signal transduction pathways downstream of Tek are tightly regulated; however, these pathways are currently not well understood. The phosphorylated Tek receptor has been shown to associate with Grb2 and Shp2 in vitro (21), and more recently, Korpelainen et al. (22) have shown that Tek can activate the signal transducers and activators of transcription 1, 3 and 5. Tek has also been shown to initiate signal transduction pathways downstream of PI 3-kinase and Dok-R (20,23). Here we report that Grb7, Grb14, and the p85 subunit of PI 3-kinase, as well as Grb2 and Shp2, can associate with Tek. These five molecules can interact specifically with the phosphorylated Tek receptor through their SH2 domains in yeast and in vitro. Using synthetic phosphopeptides, we have mapped the potential binding sites of these molecules on Tek and found that one tyrosine residue, Tyr 1100 , may serve as a multisubstrate docking site. We provide evidence to suggest that Grb2, Grb7, and p85 can associate with Tek in vivo and that these interactions are mediated through this multidocking site. We also show that Tek can use both Grb7 and p85 as substrates and that tyrosine phosphorylation of these proteins is abrogated when Tyr 1100 is mutated. Furthermore, we demonstrate that p85 becomes tyrosine phosphorylated following Ang1 stimulation of Tek and that a PI 3-kinase signaling pathway is required for Ang1-induced endothelial cell migration and cell survival.

EXPERIMENTAL PROCEDURES
cDNA Library Construction and Yeast Two-hybrid Screening-Construction of the mouse day 12.5 embryonic heart and lung cDNA library and the yeast two-hybrid screening approach have been extensively described in Ref. 20.
Production of GST Fusion Proteins-GST fusion proteins were prepared from Escherichia coli using standard procedures, and the recombinant fusion proteins were purified following immobilization on glutathione-Sepharose beads (Amersham Pharmacia Biotech). Proteins were eluted from the beads upon treatment with free (10 mM) glutathione for 30 min at 4°C. Purified proteins were analyzed by SDS-polyacrylamide gel electrophoresis followed by Coomassie Blue staining. The concentrations of the proteins were estimated by comparison with bovine serum albumin standards.
Real Time Binding Measurements Using BIAcore-N-terminally biotinylated peptides were synthesized by the AMGEN peptide synthesis group (Boulder, CO), and purity was confirmed by mass spectral and amino acid composition analysis. Relative binding of purified GST-SH2 domains to biotinylated phosphopeptides was measured using surface plasmon resonance (BIAcore 2000, Biacore Inc., Piscataway, NJ). Peptides were coupled to streptavidin-coated sensor surfaces (Sensor Chip SA) to a density of 400 response units as per manufacturer's instructions. GST fusion proteins were serially diluted in running buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.005% (v/v) surfactant P20) and injected at a 5 l/min flow rate. Surface regeneration was performed by 10-l injections of 1 M NaCl in 50 mM NaOH at a 20 l/min flow rate. Relative binding was measured as an increase in arbitrary response units, and kinetic parameters were determined using the BIA evaluation 3.0 software.
GST Binding and Coimmunoprecipitation Assays-The cDNAs representing the cytoplasmic domain of Tek, both wild type and kinase inactive, were subcloned into pACTag2 (20) to generate hemagglutinintagged proteins. The cDNA representing the full-length Tek receptor was cloned into the pRc/RSV vector (Invitrogen), and the corresponding full-length Tek A853 and Tek F1100 cDNAs were cloned into pcDNA3.1(ϩ) (Invitrogen). Grb7 cDNA has been previously described (24), and p85 cDNA was a gift of Anke Klippel (Chiron, Emeryville, California). 10 g of each DNA was used to transfect a 10-cm culture dish of human embryonic kidney cells (HEK293T cells) using Lipofectin reagent (Life Technologies, Inc.) according to the manufacturer's instructions. For Ang1 stimulations on endothelial cells, cells were serum-starved for 6 h, pretreated with 1 mM sodium orthovanadate, pH 8.0, for 10 min at 37°C, and stimulated with 5 ml of conditioned medium in the presence of sodium orthovanadate for 10 min. Alternatively, for Ang1 stimulations on HEK293Tek cells, cells were serum-starved for 24 h and were stimulated in the absence of phosphatase inhibitors. GST mixes, immunoprecipitations, and Western blotting were performed as described previously (20).
Peptide Association Assays-1 g of eluted fusion protein was incubated with 1 g of biotinylated peptide in 1% Triton X-100 lysis buffer for 2 h at 4°C. Complexes were recovered on streptavidin agarose beads (Pierce), eluted in sample buffer, and processed as described in Ref. 20.
Antibodies Used for Immunoprecipitation and Western Blotting-Commercially available antibodies used were as follows: for immunoprecipitation, polyclonal anti-Tek C-20 (Santa Cruz), polyclonal anti-Grb7 N-20 (Santa Cruz), and monoclonal anti-Myc (Invitrogen), and for Western blotting, monoclonal anti-phosphotyrosine 4G10 (Upstate Biotechnology, Inc.), monoclonal anti-hemagglutinin-horseradish peroxidase clone 12CA5 (Roche Molecular Biochemicals), monoclonal anti-Grb2 (Transduction Laboratories), monoclonal anti-Grb7 (Transduction Laboratories), monoclonal anti-Myc (Invitrogen), and horseradish peroxidase-conjugated donkey anti-human IgG HϩL (Jackson ImmunoResearch). Monoclonal and polyclonal anti-Tek antibodies specific to the extracellular domain were a kind gift of Fu-Kuen Lin (AMGEN, Thousand Oaks, California). A peptide specific to the kinase insert region of Tek (NH 2 -CRKSRVLETDPAFAVANSTAST-COOH) was used to raise a polyclonal anti-Tek antiserum in rabbits, and this affinity purified antibody is referred to as anti-Tek KI . Polyclonal anti-p85 antibody was generously provided by Anke Klippel.
Cell Culture and Production of Conditioned Medium-Human umbilical vein endothelial cells (HUVEC) were obtained from ATCC and cultured in F12 medium supplemented with 15% fetal bovine serum (FBS), 1% penicillin, 1% streptomycin, 200 mM L-glutamine (all Life Technologies, Inc.), 0.1 mg/ml heparin (ICN Biomedicals), and 0.02 mg/ml bovine endothelial cell growth factor (Roche Molecular Biochemicals) in a 37°C, 5% CO 2 incubator. HEK293, HEK293T, HEK293Tek (a gift of Fu-Kuen Lin), Py4-1 (a gift of V. Bautch, North Carolina), and EA.hy926 (a gift of Cora Edgell, North Carolina) were grown on 10-cm plates in Dulbecco's modified Eagle's medium (DMEM) (Life Technologies, Inc.) supplemented with 10% FBS, 1% penicillin, 1% streptomycin, and 200 mM L-glutamine. HEK293Tek cells were further supplemented with 250 g/ml G418 (Life Technologies, Inc.), and 100 nM methotrexate (Sigma) and EA.hy926 were further supplemented with hypoxanthine, aminopterin, and thymidine (Sigma). Ang1 cDNA was subcloned into either SignalpIg-plus (Novagen) or pSecTagB (Invitrogen), and HEK293T cells stably expressing Ang1-F c or Ang1-MH were generated by selection of transfected cells with either 1.5 mg/ml G418 (Life Technologies, Inc.) or 1 mg/ml zeocin (Invitrogen), respectively, and resistant cells were pooled. Nontransfected HEK293T cells and stable transfectants were grown to confluence on 15-cm dishes, and conditioned medium was collected for 24 h in DMEM supplemented with 0.1% or 10% FBS. Harvested medium was clarified by centrifugation, and secretion of Ang1-F c or Ang1-MH into the medium was confirmed by Western analysis. Ang1-F c was depleted from 10 ml of conditioned medium by incubation with 500 l of 20% protein A-Sepharose slurry for 1 h at 4°C.
Cell Migration Assay-HUVEC, Py4-1, HEK293, and HEK293Tek cells were seeded at a density of 8.4 ϫ 10 4 cells in 500 l of DMEM ϩ 0.1% FBS, with or without inhibitor, in the upper chamber of an 8 m-pore modified Boyden chamber (Falcon). Conditioned medium was placed in the lower chamber, and wortmannin and LY299402 (both Sigma) were added at a final concentration of either 10, 50, or 100 nM or 10, 20, or 40 M, respectively. Cells were allowed to migrate for 4 h in a 37°C, 5% CO 2 incubator. Nonmigrating cells were scraped off, and filters were fixed in 100% methanol for 5 min, stained with Harris' Hematoxylin (BDH) for 10 min, and washed twice with tap water for 3 min each. Filters were then mounted using 100% glycerol (Fisher Biotech) and counted using a light microscope (Leica) at 400ϫ magnification.
Endothelial Cell Survival Assay-HUVEC were seeded at 1.2 ϫ 10 5 cells/well in 24-well dishes in DMEM ϩ10% FBS. After 24 h in culture, cells were washed once in 1ϫ phosphate-buffered saline and incubated in 1 ml of fresh DMEM ϩ 10% FBS or Mock, Ang1-F c , depleted or Ang1-F c ϩ 10 nM wortmannin conditioned medium. After 5 days of incubation, viable cells were stained with trypan blue (Life Technologies, Inc.) and counted using a hemacytometer (Bright-Line).
Statistical Analysis and Image Processing-Statistical significance was determined using Student's t test for comparisons between two means. All experiments were performed in triplicate. A p value of less than 0.05 was interpreted as statistically significant. Results are graphically expressed as the mean Ϯ S.E. Gels were digitally scanned and processed using Photoshop 5.0 on a Macintosh G3 computer.

Identification of
Tek Binding Partners-To identify proteins that could participate in Tek signaling pathways, we have used the yeast two-hybrid system to find molecules that associate with the cytoplasmic domain of Tek in a phosphotyrosine-dependent manner (25,26). Using this approach, we have previously reported the identification of Dok-R as a Tek binding partner (20). Initially, we tested whether Tek could associate with two known binding partners, Grb2 and Shp2 (21), in a phosphotyrosine-dependent manner in yeast. The entire intracellular domain of the wild type Tek receptor (Tek IC ), the catalytically inactive Tek receptor (Tek A853IC ), or a mutant receptor bearing a tyrosine to phenylalanine mutation in the predicted Grb2 binding site in the tail (Tek F1100IC ) (21) was coexpressed in yeast with the SH2 domains of Grb2 and Shp2. Expression of the truncated receptors in yeast results in constitutive tyrosine phosphorylation of both Tek IC and Tek F1100IC , whereas Tek A853IC remains unphosphorylated (Fig.  1A). Yeast expressing both Tek IC and Grb2 was able to activate the lacZ reporter gene and grow on selective medium ( Fig. 1B and data not shown). In contrast, yeast expressing either Tek A853IC and Grb2 or Tek F1100IC and Grb2 produced very low ␤-galactosidase activity and sparse growth on selective medium, illustrating that this interaction was dependent upon Tek kinase activity and an intact Tyr 1100 residue ( Fig. 1B and data not shown). Similar results were obtained with Shp2 (data not shown), indicating that the yeast two-hybrid system could be used to identify putative Tek binding partners.
A yeast strain expressing Tek IC was then used to screen for interacting proteins expressed from cDNAs obtained from a day 12.5 mouse embryonic heart and lung library as described previously (20). Early embryonic heart and lung express high levels of Tek (27); thus it was reasoned that cDNAs derived from these tissues should be rich in Tek signaling partners. Several clones were identified that interacted with Tek in a phosphotyrosine-dependent manner. Fig. 1C lists the names of the signaling molecules that were either obtained in the screen or were shown to interact with Tek IC or Tek IC mutants in yeast. Grb2, Shp2, and the p85 subunit of PI 3-kinase have previously been reported to serve as targets for numerous activated RTKs, while Grb7 and Grb14 are members of an emerging family of PH domain-containing adaptor molecules. Importantly, we confirmed that all of these signaling molecules are coexpressed with Tek in cultured endothelial cells (data not shown), suggesting that Grb2, Grb7, Grb14, Shp2, and p85 could indeed be true signaling partners of Tek.
The SH2 Domains of Grb2, Grb7, Grb14, Shp2, and p85 Mediate Binding to Tek-Sequence analysis of the partial Grb7, Grb14, and p85 cDNAs obtained in the screen demonstrated that they contained the regions coding for the SH2 domains of these proteins. To determine whether it was indeed the SH2 domains that mediated these interactions and also to test whether these interactions could occur outside of the yeast environment, we fused the SH2 domains of Grb2, Grb7, Grb14, Shp2, and p85 to GST and purified these fusion proteins from E. coli. Immobilized GST-SH2 domains were incubated with lysates prepared from HEK293T cells transiently expressing the hemagglutinin-tagged cytoplasmic domains of either wild type Tek (Tek IC ) or catalytically inactive Tek (Tek A853IC ). In a manner similar to that seen in yeast, overexpression of Tek IC in mammalian cells results in its autophosphorylation, and the kinase inactive form of Tek IC remains unphosphorylated ( Fig.  2A). The isolated SH2 domains of Grb2, Grb7, and Grb14 could all precipitate Tek IC but were unable to precipitate Tek A853IC (Fig. 2B). Both Shp2 and p85 contain two SH2 domains; thus we constructed GST-SH2 fusions of each domain individually to determine which of these domains mediated the interaction with Tek. Both SH2 domains of Shp2 and p85 were able to interact with Tek IC ; however, the C-terminal SH2 domain of Shp2 was the least effective at precipitating Tek IC (Fig. 2B). These in vitro binding data supported our results obtained in yeast (Fig. 1C) where the Shp2 N-terminal SH2 domain mediated stronger binding to Tek IC than the C-terminal SH2 do- Tek bearing a tyrosine to phenylalanine mutation in the predicted Grb2 binding site in the carboxyl tail (Tek F1100IC ). All intracellular domains are expressed, and Tek IC and Tek F1100IC are tyrosine phosphorylated. IP, immunoprecipitation. B, yeast coexpressing the SH2 domain of Grb2 and Tek IC , Tek A853IC (A853 IC ), or Tek F1100IC (F1100 IC ) were assayed for lacZ reporter gene activation. Only yeast expressing both Grb2 and wild type Tek IC were able to activate the reporter gene as indicated by a chromogenic assay (blue precipitate in yeast). C, summary of Tek binding partners isolated in yeast two-hybrid screen or found to associate with Tek in yeast. For Shp2, the individual SH2 domains (N or C) or the tandem SH2 domains (NϩC) were introduced into yeast separately. The relative intensities of the interactions between the putative Tek binding partners and the various Tek intracellular domains are indicated with ϩ or Ϫ.
main. Taken together, these results demonstrate that the SH2 domains of Grb2, Grb7, Grb14, and p85 and the N-terminal SH2 domain of Shp2 can all bind specifically to Tek in vitro and that these interactions are dependent on Tek tyrosine kinase activity.
Determination of the Receptor Binding Sites in Vitro-To this point, we have shown that the SH2 domains of these signaling molecules can interact with the receptor in a phosphotyrosinedependent manner both in yeast and in vitro, and we next wanted to determine the putative binding sites of these SH2 domains on Tek. Because the in vivo autophosphorylation sites of Tek have yet to be defined, we designed synthetic phosphopeptides representing all 19 possible tyrosine phosphorylation sites on Tek (Fig. 3A). Although it is highly unlikely that all of these tyrosine residues are phosphorylated in response to Ang1 stimulation, these phosphopeptides were used as a starting point to reveal the sites that could potentially mediate an interaction. The interactions between the various GST-SH2 domains and the 19 potential binding sites were investigated using real time biosensor (BIAcore) analysis, where the phosphopeptides were immobilized on sensor chips. GST alone did not bind strongly to any of the immobilized phosphopeptides, whereas the SH2 domains of both Grb2 and Grb7 bound most strongly to the peptide comprising phosphorylated tyrosine residue 1100 (Tyr(P) 1100 ) and not to the unphosphorylated counterpart ( Fig. 3B and data not shown). Tyr(P) 1100 is within a Y 1100 VNT context, and this sequence fits the consensus binding site for the SH2 domains of both Grb2 and Grb7 that predicts the presence of asparagine at the ϩ2 position following the tyrosine (28 -30). In contrast to Grb2 and Grb7, the SH2 domain of Grb14 did not bind strongly to Tyr(P) 1100 ; rather this SH2 domain was found to bind most effectively to peptides representing Tyr(P) 814 and Tyr(P) 1106 (Fig. 3B). Tyr(P) 814 is found within the context Y 814 PVL, whereas Tyr(P) 1106 is found within the context Y 1106 EKF, suggesting that Grb14 may bind preferentially to sites where the ϩ3 position is occupied by a large hydrophobic residue.
The binding of Shp2 to these immobilized peptides was somewhat more complex because peptides representing Tyr(P) 814 , Tyr(P) 895 , Tyr(P) 897 , Tyr(P) 1010 , Tyr(P) 1022 , and Tyr(P) 1111 were all found to interact strongly with the Shp2 tandem SH2 domains (Fig. 3B). These residues are found within the contexts Y 814 PVL, Y 895 LYL, Y 897 LAI, Y 1010 SVY, Y 1022 GVL, and Y 1111 AGI, respectively. Interestingly, when the individual SH2 domains were tested in these binding experiments, the Nterminal SH2 domain was able to bind to nearly all of the tyrosine residues initially detected using the tandem SH2 domains, whereas the C-terminal SH2 domain bound very poorly to these residues (Fig. 3B). This result suggested that the binding detected when using the tandem SH2 domains can be mostly attributed to the N-terminal SH2 domain. The consensus binding site for the Shp2 N-terminal SH2 domain predicts a hydrophobic valine or isoleucine at the ϩ1 and ϩ3 positions following the tyrosine (28), and these tyrosine residues conform closely to this consensus. In contrast to Shp2, when the individual SH2 domains of p85 were tested, the C-terminal SH2 domain displayed consistently higher binding to the various phosphopeptides than the N-terminal SH2 domain (Fig. 3B). The p85 SH2 domains were found to bind to peptides representing Tyr(P) 895 , Tyr(P) 1037 , Tyr(P) 1066 , Tyr(P) 1100 , and Tyr(P) 1106 , and these residues are within the contexts Y 895 LYL, Y 1037 CGM, Y 1066 DLM, Y 1100 VNT, and Y 1106 EKF, respectively. The consensus binding sites for both the N-and C-terminal SH2 domains of p85 predict the presence of methionine at the ϩ3 position following the tyrosine (28), and both Tyr 1037 and Tyr 1066 are followed by this predicted consensus.
To determine whether the interactions detected by BIAcore analysis were sufficient to allow coprecipitation, we performed in vitro mixing experiments with purified GST-SH2 domain fusion proteins and the various phosphopeptides. BIAcore analysis indicated that the SH2 domains of both Grb2 and Grb7 bound most strongly to the peptide representing Tyr(P) 1100 , and these findings were confirmed in the in vitro mixing experiments (Fig. 3C). Collectively, these results demonstrate that Grb2 and Grb7 can associate with the same tyrosine residue located in the C-terminal tail of Tek. In contrast, Grb14 appeared to bind preferentially to Tyr(P) 814 in the juxtamembrane region of the receptor (Fig. 3C), supporting the data from the BIAcore analysis that the Grb14 SH2 domain requires the presence of a hydrophobic residue at the ϩ3 position.
In the case of Shp2, the tandem SH2 domains exhibited a complex binding pattern upon BIAcore analysis, and similar results were obtained in the in vitro coprecipitation experiments (Fig. 3C). The tandem and N-terminal SH2 domains could be coprecipitated with the Tyr(P) 814 , Tyr(P) 895 , Tyr(P) 897 , Tyr(P) 1022 , and Tyr(P) 1111 peptides, whereas the C-terminal SH2 domain could only be coprecipitated by the Tyr(P) 1022 and Tyr(P) 1111 peptides. However, Tyr 895 , Tyr 897 , and Tyr 1022 are found within the conserved activation and catalytic loops of the Tek kinase domain, suggesting that they probably do not confer receptor binding specificity in vivo. Taken together, the results from the BIAcore analysis and the in vitro coprecipitation putative tyrosine phosphorylation sites and the corresponding synthetic phosphopeptides that were used in the mapping studies. TM indicates the transmembrane domain, whereas TK1 and TK2 represent the two lobes of the kinase domain. Note the absence of tyrosine residues in the region that separates TK1 from TK2. Amino acid residues are designated according to single-letter code, and the position of the phosphotyrosine (pY) is in bold type. B, graphic representation of the results obtained from BIAcore analysis using purified GST fusion proteins and immobilized phosphopeptides. Relative response units were measured and interpreted from sensograms. C, associations between various Tek phosphopeptides and purified GST fusion proteins were assayed in vitro. The interactions tested are summarized, and relative binding is indicated with ϩ or Ϫ. experiments demonstrate that the majority of binding of Shp2 to Tek is mediated by the N-terminal SH2 domain, and this interaction is likely through a consensus Shp2 binding motif in either the juxtamembrane or C-terminal tail of the receptor.
Lastly, coprecipitation experiments demonstrated that the N-terminal SH2 domain of p85 could interact with Tyr(P) 1066 and Tyr(P) 1111 peptides, whereas the C-terminal SH2 domain could associate with Tyr(P) 1037 , Tyr(P) 1100 , and Tyr(P) 1106 peptides (Fig. 3C). Although Tyr 1037 and Tyr 1066 are followed by the predicted p85 SH2 binding consensus, these residues are also found within the conserved kinase domain of the receptor. Collectively, these experiments demonstrate that the interaction of p85 with Tek is likely mediated through the C-terminal SH2 domain of p85 and a nonconsensus binding motif in the C-terminal tail of Tek.
These analyses have demonstrated that the Grb2, Grb7, and p85C SH2 domains all associate most strongly with the Tyr(P) 1100 peptide (Fig. 3, B and C), suggesting that these three signaling molecules may compete for the same binding site on Tek. Using BIAcore software, we have determined the relative dissociation rate constants (K d ) for each SH2 domain binding to the Tyr(P) 1100 peptide. The SH2 domains of Grb2, Grb7, and p85 were found to bind to Tyr(P) 1100 with apparent rate constants of 11, 91, and 38 nM, respectively (Table I). These values suggest that the SH2 domain of Grb2 has the greatest relative affinity for Tyr(P) 1100 , whereas that of Grb7 has the lowest relative affinity for this phosphopeptide.
In Vivo Association of Grb2, Grb7, and p85 with Tek in 293T Cells-To study the signal transduction pathways mediated by the Tek receptor in the absence of Ang1, we first engineered mammalian expression vectors to contain full-length Tek, Tek A853 , or Tek F1100 . When these proteins are transiently expressed in HEK293T cells, Tek becomes highly tyrosine phosphorylated, whereas Tek A853 remains unphosphorylated (Fig.  4A). Tek F1100 is only weakly tyrosine phosphorylated compared with Tek (Fig. 4A); however, we have been able to show that mutagenesis of this site does not affect the catalytic function of the receptor because Dok-R can still become tyrosine phosphorylated when coexpressed with this mutant. 2 These results suggest that Tyr 1100 may represent a major autophosphorylation site on Tek in vivo, although there are additional sites of tyrosine autophosphorylation.
To determine whether the associations we had detected in yeast and in vitro could occur in vivo, coimmunoprecipitation experiments were performed using lysates prepared from these transfected cells. Coimmunoprecipitation of endogenous Grb2 with Tek is only observed when Tek is tyrosine phosphorylated (Fig. 4B), demonstrating that Grb2 can associate with Tek in vivo and that this interaction is dependent on Tek kinase activity. Furthermore, this interaction does not occur when Tyr 1100 is mutated (Fig. 4B), illustrating that this site is required for the association of Tek with Grb2. Similarly, Grb7 is coexpressed with either Tek or Tek mutants and coimmunoprecipitation of Tek with Grb7 is also only observed when Tek is phosphorylated at Tyr 1100 ( Fig. 4C and data not shown). Furthermore, Grb7 becomes detectably tyrosine phosphorylated when coexpressed with wild type Tek, and this results in the association of Grb7 with two other phosphoproteins with relative sizes of approximately 70 (pp70) and 85 kDa (pp85) (Fig. 4C). It has been reported that Grb7 can associate with Shp2, which has a relative molecular mass of 70 kDa (32), and phosphorylated Shp2 may also associate with the p85 subunit of PI 3-kinase (33). However, these two unidentified proteins did not react with anti-Shp2 or anti-p85 antibodies (data not shown), and it remains to be determined whether these proteins are either Shp2 or p85 or novel Grb7 binding proteins. Upon coexpression of the p85 subunit of PI 3-kinase with Tek or Tek mutants, immunoprecipitated p85 is tyrosine phosphorylated, and this phosphorylation is also dependent on Tek kinase activity and an intact Tyr 1100 site (Fig. 4D). Because we could not detect Tek in p85 immunoprecipitates (data not shown), it is possible that the interaction between these two proteins may be either transient or indirect. In similar studies, we were unable to detect any in vivo associations of Tek with either Grb14 or Shp2 (data not shown). In summary, these

FIG. 4. In vivo associations between
Tek and putative downstream binding partners. A, full-length Tek, Tek A853 , or Tek F1100 were transiently expressed in 293T cells, and lysates were immunoprecipitated (IP) with polyclonal anti-Tek antibodies specific to the intracellular domain of Tek (Tek KI ). Expression of full-length Tek in 293T cells results in tyrosine phosphorylation of Tek, whereas Tek F1100 is only weakly tyrosine phosphorylated and Tek A853 is not phosphorylated. B, Tek immunoprecipitates from Tek, Tek A853 , or Tek F1100 expressing cell lysates were resolved and immunoblotted with anti-Grb2 antibodies. Coimmunoprecipitation of Grb2 with Tek could only be seen with the tyrosine phosphorylated receptor, and no association was seen with the Tek F1100 mutant. Equal expression of endogenous Grb2 was detected in all lysates. C, Tek, Tek A853 , and Tek F1100 were coexpressed with Grb7 in 293T cells, and lysates from these cells were immunoprecipitated with anti-Grb7 N-20 antibodies. Grb7 could coimmunoprecipitate the phosphorylated Tek receptor, and Grb7 was found to be tyrosine phosphorylated when coexpressed with Tek but not with Tek A853 or Tek F1100 . Furthermore, phosphorylated Grb7 was also found to coimmunoprecipitate two additional proteins with relative molecular masses of 70 (pp70) and 85 kDa (pp85). Nonimmunoprecipitated lysates show equal amounts of Grb7 in the initial lysates following immunoblotting with anti-Grb7 antibodies. D, p85 was coexpressed with Tek or Tek mutants in 293T cells, and lysates were immunoprecipitated with anti-p85 antibodies, resolved by SDS-polyacrylamide gel electrophoresis, and immunoblotted with anti-phosphotyrosine antibodies. Immunoprecipitated p85 was tyrosine phosphorylated in lysates from cells coexpressing Tek but not Tek A853 or Tek F1100 . Equal expression of p85 was detected in all lysates. results demonstrate that a multisubstrate docking site on Tek, Tyr 1100 , mediates an association with Grb2 and Grb7 in vivo and that phosphorylation at this site is required for tyrosine phosphorylation of Grb7 and p85.
Ang1 Induces PI 3-kinase-dependent Endothelial Cell Migration in Vivo-While these experiments were ongoing, Ang1 was identified as an activating ligand for Tek (6). To obtain sufficient quantities of this ligand for biochemical manipulation, we generated two HEK293T cell lines stably secreting Ang1, and to facilitate the identification of the ligand in these cell lines, Ang1 was tagged at the C terminus with either a Myc epitope and poly-histidine (Ang1-MH) or the F c region of human immunoglobulin (Ang1-F c ). Because this ligand is reportedly difficult to purify in its native state (6), we proceeded to use Ang1-containing conditioned medium collected from these cells in our experiments. Similar expression levels of Ang1-MH and Ang1-F c can be detected in conditioned medium harvested from transfected cells but not from nontransfected control cells (Mock), and Ang1-F c can be specifically depleted from the conditioned medium using protein A-Sepharose (Depleted) (Fig.  5A). Moreover, we have previously shown that both Ang1 ligands function in a similar manner in in vitro studies. 3 We tested the bioactivity of the native ligand on various endothelial cell lines that have been shown to express Tek and found that, in the presence of sodium orthovanadate, conditioned medium containing Ang1-MH could stimulate tyrosine phosphorylation of Tek (Fig. 5B). However, when we looked at various markers of cellular activation, we found that sodium orthovanadate alone had many of the same effects as Ang1-MH conditioned medium (data not shown). To effectively study the intracellular changes caused by Ang1 in the absence of sodium orthovanadate, we have used an HEK293 cell line that stably expresses the full-length Tek receptor (HEK293Tek). Interestingly, stable overexpression of Tek in these cells does not result in constitutive activation of the receptor, and we are able to induce tyrosine phosphorylation of Tek with Ang1-MH conditioned medium without any addition of sodium orthovanadate (Fig. 5B). Furthermore, we found that p85 transiently expressed in HEK293Tek cells could become tyrosine phosphorylated following Ang1 stimulation (Fig. 5B). Collectively these results show that the intact Ang1 ligand can induce tyrosine phosphorylation of p85 downstream of the Tek receptor.
Cell migration has previously been shown to require PI 3-kinase signal transduction pathways (35); thus we set out to determine whether Ang1 could mediate endothelial cell migration through PI 3-kinase. A modified form of Ang1 (Ang1*) has previously been reported to be a chemotactic agent for endothelial cells (17). Using a modified Boyden chamber approach, we assessed the chemotactic response of HUVEC, Py4-1, HEK293Tek, and HEK293 cells to conditioned medium containing Ang1-F c collected in low serum. Ang1-F c medium produced a significant migration of HUVEC, Py4-1, and HEK293Tek cells when compared with mock or depleted medium and as anticipated, HEK293 cells displayed a very limited response to Ang1-F c (Fig. 5C). Interestingly, the PI 3-kinase inhibitors wortmannin and LY294002 could significantly block, although not completely, the chemotactic responses of Ang1-F c (Fig. 5C). These results clearly demonstrate the ability of Ang1 to mediate chemotaxis of Tek-expressing endothelial and nonendothelial cells, and they support a role for PI 3-kinase activity in endothelial cell migration.
Angiopoietin-1 Functions as a Survival Factor for Endothelial Cells-Signal transduction pathways that regulate the process of cell survival are beginning to be elucidated, and PI 3-kinase has been shown to promote cell survival through activation of the PKB/Akt pathway (36). To assess whether Ang1 could initiate PI 3-kinase-dependent survival pathways downstream of Tek, HUVEC were seeded in conditioned medium containing Ang1-F c collected in 10% serum, and following 5 days in culture, the number of viable cells were counted. Endothelial cells cultured in mock medium exhibited a marked decrease in the number of cells surviving after 5 days; however, Ang1-F c was able to prevent cell death in HUVEC, albeit at lower levels than those seen in cells cultured in medium containing ECGF (Fig. 5D). Depletion of the Ang1-F c from the conditioned medium abolished this Ang1-F c -dependent cell survival to levels that were equivalent to cells grown in mock medium (Fig. 5D). The addition of wortmannin to the Ang1-F c conditioned medium also resulted in a dramatic reduction in cell survival (Fig. 5D), demonstrating that activation of the PI 3-kinase pathway may also be required for Ang1-dependent endothelial cell survival.

DISCUSSION
In this study, we report the identification of a number of SH2 domain-containing signaling molecules, namely Grb2, Grb7, Grb14, Shp2, and p85, that can associate specifically with phosphorylated Tek in yeast and in vitro. We have mapped potential binding sites of the SH2 domains of these molecules on Tek using synthetic phosphopeptides and found that one tyrosine residue, Tyr 1100 , can serve as a multifunctional docking site. This multisubstrate docking site is required for the association of Tek with Grb2 and Grb7 in vivo, and tyrosine phosphorylation of Grb7 and p85 is abrogated when Tyr 1100 is mutated. Furthermore, tyrosine phosphorylation of both Tek and p85 is rapidly detected following Ang1 stimulation, and PI 3-kinase activity is required in part for Ang1-induced endothelial cell migration and survival. In summary, these results present evidence to suggest that signaling from Tek is mediated by a host of signaling molecules that can associate with Tek through a multifunctional docking site that is ultimately responsible for activation of endothelial cell migration and survival pathways.
The Tek binding partners that were identified in this screen contain SH2 domains that we have shown to be essential for the interaction of these proteins with phosphorylated Tek in vitro. Moreover, we have identified putative key phosphotyrosine residues on Tek that mediate binding to the SH2 domains of these proteins. Grb2 and Grb7 bind almost exclusively to Tyr 1100 located in the C-terminal tail of Tek, whereas Grb14 interacts preferentially with Tyr 814 and Tyr 1106 in the juxtamembrane and C-terminal tail regions, respectively (Fig. 6). Although the binding specificity of the Grb14 SH2 domain has yet to be determined, these findings suggest that the preferred binding site for the SH2 domain of Grb14 may include the presence of a large hydrophobic residue at the ϩ3 position.
Binding studies on proteins such as Shp2 and p85 that contain tandem SH2 domains have suggested that cooperativity between the two domains confers high biological specificity (38). Both SH2 domains of p85 are required for a stable interaction (39), although the C-terminal SH2 domain has been shown to be responsible for the high affinity and high specificity interaction of p85 with the platelet-derived growth factor receptor (40). The distance between p85 binding motifs in bisphosphorylated target peptides can be quite short (38), and based on our binding data, it is possible that the C-terminal SH2 domain of p85 would bind to Tyr 1100 , whereas the Nterminal SH2 domain would stabilize this interaction through binding Tyr 1106 or Tyr 1111 . These data are in agreement with the data from Kontos et al. (23), which suggested that p85 binds FIG. 5. Ang1 stimulates PI 3-kinase-dependent endothelial cell migration and survival. A, cell pellets and conditioned medium were obtained from mock-transfected HEK293T cells (mock) or cells expressing either Ang1-MH or Ang1-F c . Ang1-F c was depleted from the conditioned medium using protein A-Sepharose (depleted). Conditioned media were immunoprecipitated (IP) with anti-Myc antibodies or precipitated with protein A-Sepharose (Pr-A) followed by immunoblotting with either anti-Myc or donkey anti-human IgG (␣-Fc), respectively. Ang1-F c and Ang1-MH can be detected in conditioned medium collected from transfected cells but not from mock or depleted medium. B, lysates from Py4-1, a polyoma transformed murine endothelial cell line, and EA.hy926, an immortalized HUVEC line, were stimulated with either Mock or Ang1-MH conditioned medium in the presence of orthovanadate, or lysates from HEK293Tek cells stimulated in the absence of orthovanadate were immunoprecipitated with anti-Tek polyclonal antibodies and immunoblotted with anti-phosphotyrosine antibodies. Phosphorylation of Tek could be detected following stimulation with Ang1-MH, whereas no phosphorylation was seen following stimulation of cells with Mock conditioned medium. Reprobing of these membranes with anti-Tek antibodies confirmed the presence of Tek in all lanes. Phosphorylation of transfected p85 could also be detected in HEK293Tek cells following stimulation of Tek with Ang1-MH. C, test conditioned media were placed in the lower chambers of a modified Boyden apparatus, and cells were seeded at a density of 8.4 ϫ 10 4 cells/well in the upper chambers. Cells were allowed to migrate for 4 h, and migrated cells were counted. Ang1-F c resulted in an ϳ4-fold increase in migration in all Tek-expressing cell types when compared with depleted medium, and Ang1-F c -induced migration was significantly inhibited in the presence of the PI 3-kinase inhibitors wortmannin (Wort) and LY294002 (LY). Data points for HUVEC and Py4-1 represent the average number of cells migrated relative to VEGF treatment, whereas HEK293Tek and HEK293 were compared with HEK293Tek treated with Ang1-F c . All experiments were performed in triplicate, and differences were found to be statistically significant (p Ͻ 0.05). D, HUVEC were seeded at 1.2 ϫ 10 5 cells/well and cultured in DMEM with 10% serum supplemented with ECGF, Mock, Ang1-F c , depleted, or Ang1-F c ϩ 10 nM wortmannin conditioned medium. Ang1-F c promotes endothelial cell survival, albeit at levels somewhat lower than those seen in ECGF-containing medium. In contrast, cells cultured in Mock, depleted, or Ang1-F c in the presence of wortmannin displayed significantly lower levels of survival. Statistical analysis revealed that all differences were statistically significant.
to Tek through a nonconsensus binding motif similar to those seen in other RTKs (41,42).
Grb2, Grb7, and p85 were all found to bind to the same phosphorylated tyrosine residue on Tek (Tyr 1100 ) both in vitro and in vivo, suggesting the existence of a multidocking site on Tek. Several precedents have been set in other RTKs where different SH2 domain containing proteins interact with the same phosphotyrosine residue (43,44). The ability of different SH2 domain-containing proteins to bind transiently to the same site may be affected by intracellular localization of the proteins as well as binding affinities between the SH2 domains and particular phosphotyrosine residues. The Grb2 SH2 domain exhibited the strongest relative binding for Tyr(P) 1100 , suggesting that Grb2 would effectively out-compete Grb7 or p85 for binding to this residue. However, Grb7 also contains a PH domain (45), and targeting of the PH domain to the membrane in conjunction with the SH2 domain would effectively increase the amount of Grb7 at the receptor, allowing it to compete with Grb2 for binding. Furthermore, a region between the PH and SH2 domains of Grb10 and Grb14 can serve as an alternate receptor binding domain (37,46), and this homologous region in Grb7 may further contribute to Tek binding. A similar multi-domain scenario exists for p85 where both SH2 domains could potentially bind simultaneously to two phosphorylated tyrosine residues in the C-terminal tail of Tek. Because we did not determine the binding constants for the tandem SH2 domains with either mono-or bisphosphorylated target peptides, it is possible that the paired SH2 domains would display a higher relative affinity for peptides containing Tyr(P) 1100 than Grb2.
Our inability to demonstrate an in vivo interaction between p85 and Tek may suggest that the association between p85 and Tek is not direct and that p85 is actually brought to the recep-tor in a complex with an adaptor molecule such as Grb2 or a docking molecule such as Gab1. However, the identification of p85 as a putative Tek binding partner in yeast suggests that this interaction is likely to occur in the absence of a bridging molecule, and we were able to show that the SH2 domains of p85 can associate with the activated Tek receptor in vitro. p85 became highly phosphorylated when coexpressed with activated Tek, and this phosphorylation was abrogated following disruption of the Tyr 1100 docking site. Importantly, we have also demonstrated that p85 becomes tyrosine phosphorylated following Ang1 stimulation of Tek. Moreover, it has been reported that activation of a chimeric Tek receptor in vivo results in downstream activation of PI 3-kinase, and this activation is dependent on an intact Tyr 1100 but not an intact Tyr 1111 (23). Collectively, these results suggest that p85 is a true downstream signaling partner of the Ang1-activated Tek receptor and that its interaction is mediated through a multidocking site on Tek.
The identification of a host of signaling molecules that can associate with Tek allows us to speculate on the role of Tek and its respective ligands in angiogenesis. For instance, Ang1 has recently been shown to cause endothelial cell sprouting in vitro (18). This role for Ang1 is consistent with the findings that Tekand Ang1-null mice have an angiogenic or migratory defect that is manifested as a lack of vessel sprouting into the neuroepithelium (14,16). Sprouting angiogenesis requires migration and proliferation of endothelial cells from pre-existing vessels, and in fact our data confirms that of Witzenbichler et al. (17) demonstrating that Ang1 is chemotactic for endothelial cells.
Signaling through p85 appears to mediate endothelial cell migration in response to Ang1, and PI 3-kinase signal transduction pathways have previously been shown to alter the shape and migratory properties of numerous cell types. However, because inhibition of PI 3-kinase can only partially abrogate the chemotactic effect of Ang1 on endothelial cells, it suggests that additional Tek binding partners such as Dok-R and Grb7 may contribute to Ang1-mediated endothelial cell migration. In support of this, a Caenorhabditis elegans homolog of Grb7 known as Mig-10 appears to function in pathways that modulate neuronal cell migration or changes in reorganization of the cytoskeleton (47), suggesting that Grb7 may play a similar role. Since phosphorylation of both p85 and Grb7 is dependent on Tyr 1100 , this multisubstrate docking site may play a critical role in Tek-mediated endothelial cell sprouting.
Our results further demonstrate a role for Ang1 in mediating endothelial cell survival, and this effect also appears to be dependent in part on intact PI 3-kinase signal transduction pathways. The finding that Ang1 is a survival factor for endothelial cells in vitro is in marked contrast to the findings of Witzenbichler et al. (17), where they demonstrate that Ang1* does not protect cells from apoptosis. One factor that may account for this discrepancy is the difference between the epitope-tagged native ligand used in these experiments and the chimeric ligand Ang1* (18) used in the previous experiments. PI 3-kinase has recently been shown to play a role in promoting cell survival through regulation of the PH domain-containing serine/threonine kinase PKB/Akt (36). In fact, activation of a chimeric Tek receptor in vivo results in downstream activation of PKB/Akt (23), and Ang1 has been shown to protect cultured endothelial cells from apoptosis (31). These results collectively suggest that Ang1 stimulation can activate PI 3-kinase, which results in induction of anti-apoptotic pathways that are controlled by PKB/Akt. The role of Tek in endothelial cell survival is consistent with the dramatic reduction in the number of endothelial cells seen in mice lacking Tek (13). Furthermore, Tek has been shown to be constitutively phosphorylated in FIG. 6. Summary of putative docking sites on Tek for SH2 domain-containing Tek binding partners. Grb14 and Shp2 both interact with Tyr 814 in the juxtamembrane region and Tyr 1106 and Tyr 1111 , respectively, in the C-terminal tail region of Tek. A multifunctional docking site, Tyr, is also present in the C-terminal tail of Tek, and this site is required for binding of Grb2 and Grb7 as well as phosphorylation of both Grb7 and p85. Phosphorylated Grb7 associates with two unknown proteins, pp70 and pp85, and the function of Grb7 remains to be elucidated. In contrast, PI 3-kinase activity, which is likely conferred by p85 binding to Tyr 1100 , is required for Ang1-induced endothelial cell migration and survival. quiescent endothelium (34), suggesting that constitutive activation of the PI 3-kinase signaling pathway through Tek is required for the maintenance of endothelial cells.
In addition to the unique pairs of ligands that are specific for both phosphorylation and dephosphorylation of Tek, the existence of a multifunctional docking site on Tek that appears to be required for both endothelial cell migration and survival suggests that there is exquisite control over the signaling pathways mediated by this receptor. The identification of the signaling elements and mechanisms that are controlled by Tek allows further insight into the complex biology of endothelial cells.