Manipulation of EphB2 Regulatory Motifs and SH2 Binding Sites Switches MAPK Signaling and Biological Activity*

Signaling by the Eph family of receptor tyrosine ki- nases (RTKs) is complex, because they can interact with a variety of intracellular targets, and can potentially induce distinct responses in different cell types. In NG108 neuronal cells, activated EphB2 recruits p120RasGAP, in a fashion that is associated with down-regulation of the Ras-Erk mitogen-activated kinase (MAPK) pathway and neurite retraction. To pursue the role of the Ras-MAPK pathway in EphB2-mediated growth cone collapse, and to explore the biochemical and biological functions of Eph receptors, we sought to re-engineer the signaling properties of EphB2 by manip-ulating its regulatory motifs and SH2 binding sites. An EphB2 mutant that retained juxtamembrane (JM) RasGAP binding sites but incorporated a Grb2 binding mo- tif at an alternate RasGAP binding site within the kinase domain had little effect on basal Erk MAPK activation. In contrast, elimination of all RasGAP binding sites, accompanied by the addition of a Grb2 binding site within the kinase domain, led to an increase in phospho-Erk levels in NG108 cells following ephrin-B1 stimulation. Functional assays indicated a correlation between neurite retraction and the ability of the EphB2 mutants to down-regulate Ras-Erk MAPK signaling. These data suggest that EphB2 can be designed

The mechanisms by which signals are conveyed from receptor tyrosine kinases (RTKs) 1 at the plasma membrane to their intracellular targets in the cytoplasm and nucleus have been extensively explored (1)(2)(3). Ligand-induced autophosphorylation in the activation segment of the kinase domain, as well as phosphorylation of juxtamembrane (JM) tyrosines in RTKs such as Eph receptors, induce conformational changes that stimulate the activity of the kinase domain (4 -6). Furthermore, receptor tyrosine phosphorylation, usually within noncatalytic sequences flanking the kinase domain, creates docking sites for proteins with SH2 or PTB domains, leading to the recruitment of cytoplasmic targets that regulate downstream signaling (3,7).
This scheme is evident in the ability of RTKs to recruit negative or positive regulators of the Ras GTPase, such as the p120 Ras GTPase-activating protein (RasGAP) or Grb2 (8). RasGAP, which stimulates the hydrolysis of Ras-bound GTP, is a modular polypeptide with a C-terminal catalytic domain that down-regulates Ras, linked to two SH2 domains (surrounding an SH3 domain) at the N terminus (9,10). In contrast, Grb2 is an adaptor comprised of a central SH2 domain flanked by two SH3 domains that bind to targets with proline-rich motifs (11)(12)(13). In particular, through recruitment of the Ras guanine nucleotide exchange factor Sos1, Grb2 directs phosphotyrosineinduced exchange of GDP for GTP on Ras, thereby stimulating the Erk MAPK pathway (3). RasGAP and Grb2 are therefore SH2-containing proteins with opposing effects on the state of Ras activation, and data from both invertebrate and mammalian systems suggest that RTKs potentially employ Grb2 and RasGAP in combination to achieve sophisticated regulation of the Ras signaling pathway (14,15). SH2 domains generally bind short phosphotyrosine-containing peptide sequences but differ in their ability to recognize residues C-terminal to the phosphotyrosine (7,16,17). For example, the Grb2 SH2 domain binds preferentially to pTyr-X-Asn motifs (17), whereas the RasGAP SH2 domains binds optimally to pTyr-X-X-Pro motifs (18). We have employed these observations as the basis to explore signaling by EphB2, a member of the predominant family of mammalian RTKs. Eph receptors are activated by their association with cell surface ligands, ephrins, and regulate biological events involving cellcell interactions, including axon guidance, boundary formation, neural crest migration, angiogenesis, and synaptic function (19 -23). In their cytoplasmic regions, Eph receptors have a JM sequence, followed by the kinase domain, a SAM domain, and a C-terminal PDZ domain binding motif. Activated receptors become phosphorylated on multiple tyrosine residues, including a tyrosine in the activation loop of the kinase domain and two conserved tyrosines in the JM region that regulate kinase activity (24 -27). Phosphorylated tyrosines in the JM region and elsewhere can then directly engage cytoplasmic targets. In addition, the receptor can potentially phosphorylate docking proteins, such as p62Dok-1, with the ability to engage signaling proteins (18). Of interest, in contrast to other RTKs, which generally stimulate Ras, activation of EphB2 (28) and EphA2 (29) down-regulates the Ras-MAPK pathway in specific mammalian cell types (i.e. neuronal and endothelial cells). In the context of EphB2, RasGAP recruitment appears to inhibit Ras activation, which then leads to down-regulation of the MAPK pathway (28). Interfering with Ras down-regulation, either by expression of a dominant-negative form of RasGAP or a constitutively activated Ras variant, suppresses the ability of ephrin-B1 to induce neurite retraction in NG108 neuronal cells expressing EphB2.
In this study, we demonstrate that mutating the previously identified RasGAP binding sites in the EphB2 JM region is not sufficient to switch the EphB2 signal from down-regulation to up-regulation of the MAPK signaling. We identify a novel binding site for RasGAP and show that introduction of a Grb2 binding sequence in place of this second RasGAP binding site within the kinase domain stabilizes phospho-Erk levels in ephrin-B1-stimulated cells. Complete abrogation of RasGAP binding coupled with recruitment of Grb2, leads to an up-regulation of phospho-Erk, and a block in the ability of EphB2 to induce neurite retraction. These results demonstrate that it is possible to rewire RTK signaling to the Ras-MAPK pathway by the rational design of mutant receptors with modified sites for protein-protein interactions. Our results are consistent with an intricate regulation of EphB2 through the combined actions of the JM region and kinase domain.

EXPERIMENTAL PROCEDURES
DNA Constructs, Mutagenesis, and Reagents-Full-length murine EphB2 cDNA was cloned into the mammalian expression vector pcDNA3 (18). Mutagenesis of EphB2 was performed using the QuikChange system (Stratagene). Green fluorescent protein-tagged EphB2 (EphB2-GFP) was constructed by inserting the whole length of GFP cDNA into ApaI site in the C-terminal of EphB2 between Arg 983 and Ala 984 . For glutathione S-transferase (GST) fusion constructs, cDNA sequences corresponding to the mouse Grb2 SH2 domain (residues 79 -160) and human RasGAP SH2 domain (residues 169 -470) were cloned into pGEX vector (Amersham Biosciences) and expressed as GST fusion proteins.
For immunoprecipitation and Western blotting, anti-EphB2 (18, 30) and anti-HA mouse monoclonal (12CA5) (31) antibodies have been described previously. Anti-Grb2 and anti-phospho-Erk1/2 antibodies were purchased from Transduction Laboratories and Cell Signaling Technologies, respectively. Other antibodies were purchased from Santa Cruz Biotechnology.
Cell Culture, DNA Transfection, and EphB2 Stimulation-NG108-15 (NG108) cells were routinely cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum and 1ϫ hypoxanthineaminopterin-thymidine (Invitrogen). Stable cell lines in NG108 cells were produced as previously described (18,28). Briefly, parental NG108 cells were transfected with 20 g of cDNA using the calcium phosphate precipitation method (32). Cells were then grown in the presence of 400 g/ml G418 (Invitrogen) to select stable transfectants. Human embryonic kidney 293 (HEK-293) cells were maintained in Dulbecco's modified Eagle's medium with 10% fetal calf serum. 20 g of plasmid DNA was used to transfect 2 ϫ 10 6 HEK-293 cells/100-mm plate by the calcium phosphate precipitation method (32). Control cells were treated in the same way but with no DNA or with expression vector DNA only. Stimulation of EphB2-expressing cells was carried out as previously described (18,28) with 2 g/ml aggregated Ephrin-B1Fc. For Erk activity studies, EphB2 cells were seeded into six-well plates 24 h prior to stimulation at 3 ϫ 10 5 cells/well. Immunoprecipitation, GST Fusion Protein Pull-down, and Western Blotting-Unless otherwise indicated, NG108 cells were serum-starved overnight in DMEM containing 1ϫ hypoxanthine-aminopterin-thymidine. For Erk and anti-phosphotyrosine (anti-pTyr) Western blots, cells were stimulated with Ephrin-B1Fc as indicated and lysed directly in 2ϫ SDS loading buffer. For immunoprecipitation experiments, cells were rinsed twice in ice-cold phosphate-buffered saline and routinely lysed in PLC lysis buffer (50 mM Hepes, pH 7.5, 150 mM NaCl, 10% glycerol, 1% Triton X-100, 1.5 mM MgCl 2 , 1 mM EGTA, 10 mM NaPPi, 100 mM NaF, 1 mM vanadate, and a mixture of protease inhibitors) as previous described (18,30). Protein concentrations of cleared lysates were determined using a BCA assay (Pierce). Proteins were immunoprecipitated for 1-2 h at 4°C, and beads were routinely washed three times in HNTG buffer (20 mM Hepes, pH 7.5, 10% glycerol, 0.1% Triton X-100, and 150 mM NaCl). Proteins were separated on SDS-PAGE gels and transferred to an Immobilon-P membrane (Millipore). Membranes were blocked in TBST containing 5% bovine serum albumin for anti-phosphotyrosine blots or 5% skimmed milk for other blots, and immunoblotted as per standard protocols. Primary antibodies were detected with anti-mouse-or protein-A-horseradish peroxidase followed by treatment with Enhanced Chemiluminescence (Pierce). For the GST fusion protein pull-down experiment, pre-bound GST fusion protein beads (around 6 g of protein/ml) were used to pull down proteins from whole cell lysates in PLC lysis buffer at 4°C for 1 h and washed three times in HNTG buffer.
Peptide Spots Array Synthesis-Peptide arrays were constructed according to the Spots-synthesis method as previously described (33). Acid-hardened cellulose membranes pre-derivatized with polyethylene glycol (AbiMed, Langfield, Germany) were spotted with a grid of Fmoc ␤-alanine (Bachem) prior to peptide synthesis. Standard Fmoc chemistry was used throughout. Fmoc-protected and -activated amino acids were spotted in high density 24-ϫ 18-spot arrays on 130-ϫ 90-mm membranes using an AbiMed ASP422 robot. All washing, Fmoc, and side-chain deprotection steps were done manually in polypropylene containers. The amino acids were used at a concentration of 0.25 M and were twice spotted at a volume of 0.2 l for each coupling reaction. The peptides covering all tyrosine residues in the EphB2 intracellular region were synthesized (see sequence in Fig. 3 below). Following peptide synthesis and side-chain deprotection, membranes were blocked overnight in 5% skim milk. Purified GST or GST fusion proteins were added at 0.1 M in TBS and incubated for 1 h at 4°C. Membranes were washed three times in TBS and incubated with horseradish peroxidase-conjugated anti-GST antibody for 1 h in TBS. Detection was by SuperSignal enhanced chemiluminescence (Pierce).
Phosphopeptide Mapping by Two-dimensional Chromatography-Wild type and mutant (JX Grb2 ) forms of EphB2 were immunoprecipitated from transiently transfected NG108 cells, radiolabeled by in vitro kinase assay with [␥-32 P]ATP, and digested with trypsin. The digests were resolved in two dimensions on 100-m ϫ 20-cm ϫ 20-cm thin-layer cellulose plates by electrophoresis followed by ascending chromatography. Electrophoresis was performed at pH 1.9 in 50:156:1794 88% formic acid/glacial acetic acid/water for 40 min at 1000 V on a Hunter HTLE-7000 thin layer electrophoresis system, and ascending chromatography was carried out in 750:500:150:600 n-butanol/pyridine/glacial acetic acid/water. The plates were exposed to x-ray film for indicated days at Ϫ70°C.
Luciferase Assay-HEK-293 cells were transiently transfected with wild type EphB2, EphB2-EE, KIN Grb2 , KIN 667F , or JX EE -KIN Grb2 in the addition of pFA-Elk1 and pFR-Luciferase reporter constructs (Stratagene PathDetect system), and pSV2-␤-galactosidase. At 48 h posttransfection, the cells were lysed with reporter lysis buffer (Promega Corp.). The luciferase activity in lysates was determined by a microplate luminometer LB 96V (EG&G Berthold, Germany) and Luciferin reagent (Promega Corp.). ␤-Galactosidase assays were performed as per the standard protocol (34). Neurite Retraction Assay-NG108 cells were transiently transfected by using the calcium phosphate precipitation method with wild type or mutant EphB2-GFP. After 20 h, the transfected NG108 cells were differentiated by using 1 mM N 6 ,O 2 -dibutyryladenosine 3Ј:5Ј-cycle monophosphate (Sigma) for 5 h. Cells with neurite length of more than two times their cell body length were stimulated with 2 g/ml clustered ephrin-B1Fc and imaged every 1 min for 20 min on an inverted Leica DM IRB fluorescence microscope equipped with OpenLab software (Improvision Co., UK).

Endogenous RasGAP Is Required for Down-regulation of the Erk MAPK Pathway by Eph Receptor
Signaling-We have previously found that activation of ectopically expressed EphB2 in NG108 neuronal cells leads to a down-regulation of the Ras-MAPK pathway, in a fashion that is attenuated by a dominant negative RasGAP polypeptide (33). To explore whether this is a more general and physiological event, we have investigated the phosphorylation of Erk MAPK in wild type mouse embryo fibroblasts (MEFs) or MEFs derived from mouse embryos homozygous for a null mutation in the gene for RasGAP. Stimulation of wild type MEFs with clustered Fc-ephrin-A1 to activate EphA2 endogenously expressed in these fibroblasts led to down-regulation of Erk phosphorylation, whereas no change in Erk phosphorylation was detected in RasGAP Ϫ/Ϫ cells. These data suggest that the ability of endogenous Eph receptors to down-regulate the Erk MAPK pathway depends on the expression of RasGAP (Fig. 1).

Mutation of the EphB2 Juxtamembrane Region Reveals a Sequence Dependence for Receptor Regulation and Phosphorylation and a Secondary Binding
Site for RasGAP-To explore the determinants that regulate the coupling of Eph receptors to RasGAP, and the Ras-MAPK pathway, we turned to NG108 neuronal cells induced to express EphB2. Ephrin-B1 stimulation results in phosphorylation of EphB2 tyrosine residues, including two conserved JM tyrosines (Y 604 IDPFTY 610 EDP) (18,25). In their unphosphorylated form, these JM tyrosines suppress Eph kinase activity (6,24,27); this inhibition is relieved by autophosphorylation of the JM motifs, which can subsequently serve as docking sites for SH2 domain containing proteins, including RasGAP (18). RasGAP recruitment to EphB2 is associated with the down-regulation of Erk MAPK activity and may play a significant role in the axon guidance function of EphB2 (28). We hypothesized that substitution of the RasGAP SH2 domain binding sites in the JM region with known Grb2 binding motifs, the sequence Y 604 VNVFTY 610 VNV (see Fig. 5A), might lead to recruitment of Grb2 to the JM region of the resulting JX Grb2 mutant EphB2 and induce upregulation of MAPK activity, an effect opposite to that of wild type (WT) EphB2.
To study the Erk MAPK response to the EphB2 mutant JX Grb2 , we derived an NG108-15 (NG108) cell line stably expressing this mutant. The parental NG108 cells do not express detectable EphB2 and do not respond to ephrin-B1 stimulation. In contrast, ephrin-B1 stimulation of NG-EphB2 cells, which stably express wild type EphB2 receptor, leads to an increase in the tyrosine phosphorylation of EphB2 and cellular proteins such as p62Dok-1 and induces neurite retraction (18) and down-regulation of MAPK activity (28). Similarly, ephrin-B1 stimulation of cells expressing JX Grb2 caused an increase in cellular tyrosine phosphorylation, although it was attenuated relative to WT EphB2 ( Fig. 2A). We also observed ligand-dependent association of the mutant receptor with RasGAP in NG108 cells expressing JX Grb2 (Fig. 2C), albeit to a lesser extent than WT EphB2, and Erk1/2 dephosphorylation (Fig.  2B) in a manner similar to cells expressing WT receptor. Interestingly, we were unable to detect association of either the WT or JX Grb2 EphB2 with Grb2 following ephrin-B1 stimulation (Fig. 2C).
The observation that the JX Grb2 Eph receptor retains the ability to recruit RasGAP raises the possibility that tyrosine residues at the mutated Y 604 VNVFTY 610 VNV motif in JX Grb2 FIG. 2. Response of EphB2 wild type (WT) and a mutant containing a tandem potential Grb2 SH2 binding motif in the JM region (JX Grb2 ) to ephrin stimulation. NG108 cells were stably transfected with pcDNA-EphB2 or pcDNA-EphB2-JX Grb2 , and individual G418-resistant clones were isolated. A and B, parental, EphB2-, and JX Grb2expressing NG108 cells were serumstarved overnight and challenged with 2 g/ml clustered Ephrin-B1Fc for the indicated time points and lysed directly in 2ϫ SDS-PAGE sample buffer. The lysates were electrophoresed, blotted and probed with antibodies against phosphorylated tyrosine (A, top panel) or phosphorylated Erk1/2 (B, top panel). The blots were stripped and reprobed for EphB2 receptor and total Erk1/2 (bottom panels, A and B, respectively). C, serum-starved parental, EphB2-, and JX Grb2 -expressing NG108 cells were incubated in the presence or absence of 2 g/ml aggregated ephrin-B1 for 20 min and lysed in PLC lysis buffer. EphB2 immunoprecipitates were separated by SDS-PAGE, the upper portion of the blot was probed with anti-pTyr antibody (upper panel) and reprobed with anti-EphB2 (second panel) and anti-p120-RasGAP (third panel) antibodies. The lower portion of the blot was probed with anti-Grb2 antibody (bottom panel). WCL, whole cell lysate. may still be able to preferentially recruit RasGAP instead of Grb2, or the mutant receptor could directly or indirectly associate with RasGAP at other sites. In addition, the JM tyrosine residues in the JX Grb2 mutant may not be phosphorylated upon ephrin-B1 stimulation, and therefore binding sites for the Grb2 SH2 domain may never be generated in the JX Grb2 mutant. To test the first hypothesis, arrays of tyrosine-phosphorylated and unphosphorylated peptides encompassing all tyrosine residues in the intracellular EphB2 region were synthesized directly on a membrane and probed with a GST-Grb2 SH2 domain fusion protein (Fig. 3A). As controls, a YVNV-containing phosphopeptide from Shc, which binds the Grb2 SH2 domain, and a YSVPcontaining phosphopeptide from p190 RhoGAP, which binds the RasGAP SH2 domains, were included. The GST-Grb2 SH2 fusion protein bound specifically to the phosphorylated Shc peptide and to the JX Grb2 and JXЈ Grb2 peptides (Fig. 3A) as anticipated but not to their non-phosphorylated counterparts. A GST fusion protein containing both RasGAP SH2 domains associated with the p190 RhoGAP phosphopeptide but was unable to associate with either the JX Grb2 or JXЈ Grb2 peptide (Fig. 3A). These results suggest that the mutated Y 604 VNVFTY 610 VNV motif in JX Grb2 should bind to Grb2 and not RasGAP, if it were accessible and phosphorylated after ephrin-B1 stimulation. To investigate whether the mutated tyrosine sites in JX Grb2 were indeed phosphorylated, WT EphB2 and mutated JX Grb2 were radiolabeled by using [␥-32 P]ATP in an in vitro kinase assay, digested with trypsin, and analyzed by two-dimensional tryptic phosphopeptide mapping. The resulting maps indicated that the most prominent phosphopeptide present in WT EphB2, likely corresponding to the phosphorylated JM site (24,25), is missing from the mutant JX Grb2 (Fig. 3B). Taken together, these results indicate that the JX Grb2 mutant retains an ability to down-regulate Erk1/2 activity, and to recruit RasGAP, suggesting the presence of a secondary binding site for RasGAP. In addition, the JX Grb2 mutant was unable to recruit Grb2, apparently due to a lack of phosphorylation at the modified JM region. Because the JX Grb2 mutant receptor was still activated by ephrin-B1 stimulation, these results indicate that mutation of the residues C-terminal to the JM tyrosines interferes with both the inhibitory activity of the JM sequence as well as its ability to become phosphorylated upon receptor activation.
Introduction of a Grb2 Binding Site in the EphB2 Kinase Domain Stabilizes Erk Activity-To identify other potential RasGAP binding sites in mutant EphB2, we screened a peptide array encompassing all the intracellular phosphopeptides of EphB2 for alternative sites of interaction with a GST fusion protein containing both RasGAP SH2 domains. This revealed another possible RasGAP SH2 binding site, Tyr 667 , in the EphB2 kinase domain (Fig. 3A). Tyr 667 has been previously identified as a phosphorylation site in vivo (25) and is on the surface of the kinase domain (6). We therefore considered the possibility that the Tyr 667 site might be a suitable candidate site for introducing an ectopic Grb2-binding motif in EphB2. To test whether such a mutant EphB2 (KIN Grb2 ), engineered to contain the Grb2-binding motif at this site (Y 667 VNV), can be phosphorylated, we expressed a GST fusion protein containing the intracellular domain of the KIN Grb2 EphB2 mutant. The GST-KIN Grb2 fusion protein became tyrosine-phosphorylated, as detected by an anti-phosphotyrosine antibody (data not shown). The GST-KIN Grb2 protein was subsequently digested and analyzed by nanoelectrospray tandem mass spectrometry A, arrays of tyrosine-phosphorylated (upper row) and unphosphorylated (bottom row) peptides encompassing all tyrosine residues in the intracellular portion of EphB2 were synthesized directly on membrane. SHC, GAP190, JX Grb2 , and JXЈ Grb2 peptides contain the Grb2 SH2 binding motif of SHC, the RasGAP SH2 binding motif of p190-RhoGAP, and the two engineered Grb2 SH2 binding motifs of the JX Grb2 EphB2 mutants, respectively. Peptide sequences are also listed. Membranes were incubated with 0.1 M purified GST fusion proteins, followed by detection with horseradish peroxidase-anti-GST antibody. B, tryptic phosphopeptide maps of wild type (WT) and mutant JX Grb2 of EphB2. Wild type and mutant JX Grb2 receptors expressed in 293 cells were immunoprecipitated, autophosphorylated with [␥-32 P]ATP, and digested with trypsin, and the resulting phosphopeptides were separated in two dimensions and exposed to film for 2 days at Ϫ70°C for the WT and for 2 weeks at Ϫ70°C for JX Grb2 EphB2 mutant.
to map the phosphorylation sites. Ion peaks corresponding to the Tyr 667 -containing phosphopeptide (SGpY 667 VNVRQ) were identified by sequencing and, thus, confirmed the specific phosphorylation of Tyr 667 in mutant KIN Grb2 (data not shown). These results suggest that the KIN Grb2 binding site can be phosphorylated in vitro and may thus be able to recruit Grb2 in vivo.
To test whether the mutant KIN Grb2 receptor can bind to Grb2 and thereby influence Erk MAPK activity in vivo, we made an NG108 cell line stably expressing KIN Grb2 . Stimulation of cells expressing WT or KIN Grb2 EphB2 caused an increase in cellular tyrosine phosphorylation in both cases; the KIN Grb2 receptor itself was inducibly tyrosine-phosphorylated, although to a lesser extent than WT (Fig. 4, A and C). However, although we observed down-regulation of MAPK activity in WT EphB2 cells, cells expressing KIN Grb2 retained a constant level of phosphorylated Erk1/2 over time (Fig. 4B). Immunoprecipitation of EphB2 and blotting for either RasGAP or Grb2 revealed that, like WT EphB2, the KIN Grb2 mutant was able to associate with RasGAP, but in contrast to the WT receptor, KIN Grb2 also recruited endogenous Grb2 in a ligand-dependent manner (Fig. 4C). Thus, the mutant EphB2 receptor, KIN Grb2 , is able to bind to Grb2 and to neutralize the down-regulation of MAPK activity observed in response to WT EphB2, suggesting that the positive effect of Grb2 may balance the inhibitory function of RasGAP. However, we could not immediately ex-clude the possibility that the stable MAPK activity in cells expressing the KIN Grb2 mutant after ephrin stimulation was due to a lower level of receptor activation, rather than Grb2 recruitment. To address this issue, we sought to engineer an EphB2 receptor that could up-regulate MAPK activity through Grb2 recruitment.
Engineering an EphB2 Variant That Up-regulates Erk MAPK Activity in NG108 Cells-We speculated that the inability of the KIN Grb2 mutant to stimulate Erk activation might be due to its association with RasGAP as well as Grb2 and, therefore, sought to engineer a receptor that lacks RasGAP binding sites but possesses a phosphorylated Grb2 binding motif. To this end, a series of mutations were introduced into EphB2 at tyrosine phosphorylation sites in the JM sequence, the kinase domain (Tyr 667 ), and SAM domain (Tyr 938 ), either singly or in combination, as shown in Fig. 5A. Tyr 938 in the EphB2 SAM domain is conserved across all Eph receptors except EphA3 and has been implicated in binding to the Grb10 adaptor protein (35). WT and mutant EphB2 receptors were transiently expressed in 293 cells and analyzed for tyrosine phosphorylation (Fig. 5B). EphB2 mutants became tyrosine-phosphorylated to varying degrees, with the exception of JX FF , in which the two JM tyrosines are changed to phenylalanine, thus locking the receptor in its autoinhibited state (6,18). The phosphorylation of mutants with tandem Grb2 binding motifs in the JM region, and a YVNV site in either the kinase or SAM domain (JX Grb2 -KIN Grb2 and JX Grb2 -SAM Grb2 ) was defective (Fig. 5B, upper  panel), and indeed a GST-Grb2 SH2 fusion failed to bind JX Grb2 -KIN Grb2 or JX Grb2 -SAM Grb2 (Fig. 5C). We therefore sought to eliminate RasGAP binding in the JM region without significantly altering the surrounding amino acid sequence (24,28). To this end, we replaced the JM tyrosines with glutamates, which preserve the ability of the receptor to be activated but do not provide SH2 binding sites (27), and introduced a Grb2 binding motif into the kinase domain at Tyr 667 . We anticipated that this mutant (JX EE -KIN Grb2 ) might retain kinase activity and bind selectively to Grb2 but not to RasGAP. Indeed, the JX EE -KIN Grb2 receptor was tyrosine-phosphorylated in ephrin-B1-stimulated NG108 cells (Fig. 5B).
To test binding of the various EphB2 mutants to Grb2 or RasGAP, we used GST fusion proteins containing the SH2 domains of Grb2 or RasGAP in pull-down experiments with EphB2 mutants from whole cell lysates (Fig. 5C). The GST-Grb2-SH2 fusion protein bound to KIN Grb2 , but not the corresponding mutant with Tyr 667 replaced with Phe (KIN F-Grb2 ), indicating that the association of KIN Grb2 with Grb2 depends on Tyr 667 . As anticipated, both KIN Grb2 and KIN F-Grb2 bound the RasGAP SH2 domains, consistent with the presence of the JM RasGAP binding sites in these mutants. Significantly, the double-mutant JX EE -KIN Grb2 bound the Grb2 SH2 domain but not the RasGAP SH2 domains (Fig. 5C). A similar Grb2 binding profile was observed when EphB2 was co-expressed with HAtagged, full-length Grb2 (Fig. 5D), indicating that Grb2 can associate with both KIN Grb2 and JX EE -KIN Grb2 in vivo.
To test the effect of the JX EE -KIN Grb2 EphB2 on cellular signaling pathways, we constructed stable NG108 lines expressing this mutant receptor. Stimulation of cells expressing JX EE -KIN Grb2 with ephrin-B1 induced tyrosine phosphorylation, albeit with slower kinetics than WT EphB2 (Fig. 6A), and association of the mutant receptor with Grb2 ( Fig. 6C)  biphasic increase in phospho-Erk (Fig. 6B), suggesting that this mutant can up-regulate signaling through the Erk MAPK pathway.
To pursue the effect of JX EE -KIN Grb2 on MAPK signaling, HEK 293 cells were transiently transfected with WT EphB2, EphB2-EE, KIN Grb2 , KIN 667F , or JX EE -KIN Grb2 as well as pFA2-Elk1, pFR-Luciferase, and ␤-galactosidase reporter constructs. The pFA2-Elk1 trans-activator plasmid consists of the activation domain of the Elk transcription activator fused with the DNA binding domain of the yeast GAL4, and the pFR-Luc reporter plasmid contains a synthetic promoter with five tandem repeats of the yeast GAL4 binding sites that control expression of the Photinus pyralis (American firefly) luciferase gene (Stratagene PathDetect system). Control cells transfected with pcDNA, pFA-Elk1, and pFR-Luciferase reporter constructs had very low basal Elk1-induced luciferase activity (data not shown). Overexpression of KIN Grb2 and JX EE -KIN-Grb2 resulted in a significant increase of luciferase activity, as compared with WT EphB2 (Fig. 6D). In addition, EphB2-EE and KIN 667F expression resulted in luciferase activity higher than WT EphB2, suggesting that loss of RasGAP binding sites on EphB2 leads to an increase in MAPK activity. In this assay the transfected receptors are autoactivated by overexpression, and hence ephrin-B1 stimulation did not augment these effects (data not shown). Collectively, these results indicate that introduction of a Grb2 binding site into the EphB2 kinase domain can recruit Grb2, and, in conjunction with ablation of the RasGAP binding motifs, lead to ligand-dependent up-regulation of Erk MAPK activity. We have therefore been able to switch EphB2 signaling to the Erk MAPK pathway in neuronal cells from inhibitory to stimulatory by manipulation of the receptor's regulatory sites and SH2 binding motifs.
The Effect of EphB2 Mutants on Neurite Retraction-To assess the biological properties of these EphB2 mutants, we established a neurite retraction assay measuring ephrin-induced changes of neurite length in NG108 cells transiently transfected with WT or mutant forms of EphB2, tagged at the C terminus with green fluorescent protein (GFP) to visualize expressing cells. Ephrin-B1 stimulation of differentiated, neurite-bearing NG-EphB2 cells leads to the reorganization of polymerized actin structures and to neurite retraction (18,28). When transiently expressed in NG108 cells, WT EphB2-GFP was tyrosine-phosphorylated following ephrin-B1 stimulation and was bound by the RasGAP SH2 domains in vitro (data not shown), indicating that it has similar biochemical features as native EphB2. Transfected NG108 cells were therefore differentiated with cAMP, stimulated for 20 min with ephrin-B1 (Fig. 7A), and examined for neurite retraction in response to activation of WT or mutant EphB2-GFP receptors. Only the green fluorescent cells with neurite lengths more than two times their cell body length were included (Fig. 7B). After 20-min stimulation with 2 g/ml ephrin-B1, the neurites of cells expressing WT EphB2-GFP cells showed a strong retraction to 23.1 Ϯ 6.4% (n ϭ 13) of their original length, whereas transiently transfected with pcDNA-EphB2 or different mutant constructs as indicated. At 48 h post-transfection, the cells were lysed in PLC lysis buffer, and the lysates were subjected to immunoprecipitation with anti-EphB2 antibody, electrophoresed, and blotted with anti-pTyr antibody (top panel). Immunoblots were subsequently reprobed for EphB2 receptor (bottom panels). C, GST fusion protein pull-down assays of EphB2 mutants. Sepharose beads, containing 6 g/ml of the GST-Grb2 SH2 domain (upper panel) or the GST-GAP-both SH2 domain (bottom panel) fusion proteins, were used to pull down WT or mutant EphB2 from transiently transfected HEK-293 cells. The pull-down products were electrophoresed and blotted with anti-EphB2 antibody. D, co-immunoprecipitation of Grb2 with different EphB2 mutants. HA-tagged Grb2 and different EphB2 mutants were co-transfected in HEK-293 cells, and an anti-HA antibody was used to immunoprecipitate Grb2. Immunoprecipitates were subsequently separated by SDS-PAGE and probed with anti-EphB2 (upper panel) or anti-HA (bottom panel) antibodies, respectively. neurites of cells expressing the constitutively autoinhibited receptor JX FF -GFP were not significantly retracted (91.5 Ϯ 5.6% (n ϭ 10) of the original neurite length). Cells expressing the KIN Grb2 -GFP (83.7 Ϯ 6.6%, n ϭ 14) or JX EE -KIN Grb2 -GFP (92.6 Ϯ 3.7%, n ϭ 10) mutants exhibited a lack of neurite retraction similar to the JX FF -GFP-expressing cells (Fig. 7B). Thus, switching the binding and signaling properties of EphB2 from repressing Erk1/2 activation in neuronal cells, to stabilizing or stimulating this MAPK pathway, as in the case of KIN-Grb2 or JX EE -KIN Grb2 , also leads to a change in the receptor's biological activity in NG108 neuronal cells. Together, these data suggest that EphB2 signaling can be altered by the rational design of mutant receptors with modified regulatory sites and SH2 binding motifs. DISCUSSION Activated RTKs can engage both positive regulators of the Ras-MAPK pathway, such as the Grb2⅐Sos1 complex, and Ras inhibitors such as RasGAP. Although the primary effect of most RTKs is to enhance Ras-GTP loading, stimulation of EphB2 or EphA2 down-regulates GTP-bound Ras and the level of phospho-Erk1/2 in at least some cells (28,29). To probe EphB2 signaling, we have attempted to alter the binding properties of EphB2 to favor recruitment of Grb2 as compared with RasGAP. Surprisingly, YVNV Grb2 binding motifs incorporated in place of the JM YXXP RasGAP binding sites were apparently not phosphorylated, even though kinase activity of the mutant receptor was induced following ephrin-B1 stimulation, most likely due to autophosphorylation within the activation segment of the kinase domain (24,25). These substitutions therefore appear to interfere with the inhibitory interaction of the JM sequence with the kinase domain and with recognition of the JM tyrosines by the kinase domain active site. These data suggest that the sequence, and possibly conformation, of the JM sequence is important not only for repression of kinase activity but also for the phosphorylation of the JM tyrosines induced by ephrin stimulation. Consistent with this interpretation, a JX F-Grb2 mutant, in which the tyrosine residues in the JM YVNV motifs of JX Grb2 were changed to phenylalanine, behaved similarly to JX Grb2 (data not shown). Despite lacking the RasGAP binding sites in the JM region, the JX Grb2 mutant was effectively WT in activity, in the sense that ephrin-B1 stimulation of JX Grb2 NG108 cells still led to RasGAP recruitment (albeit less strongly than WT EphB2) and attenuation of Erk1/2 phosphorylation (Fig. 2).
We have previously found that an EphB2 mutant (EphB2-EE) in which the JM tyrosines are changed to glutamate also retained an ability to associate with RasGAP (28), raising the possibility of a secondary RasGAP binding site on EphB2.
Peptide binding experiments demonstrated that the RasGAP SH2 domains were unable to associate with the JM peptides The lysates were electrophoresed and probed with antibody against phosphorylated Erk1/2 (top panel) and reprobed for total Erk1/2 (bottom panels). C, serum-starved EphB2-and JX EE -KIN Grb2 -expressing NG108 cells were stimulated by ephrin-B1Fc for 20 min and lysed in PLC lysis buffer. Anti-Grb2 immunoprecipitates were separated by SDS-PAGE. The upper portion of the blot was probed with anti-EphB2 antibody (upper panel), and the lower portion of the blot was probed with anti-Grb2 antibody (bottom panel). D, HEK-293T cells were transiently transfected with wild type EphB2, EphB2-EE, KIN Grb2 , KIN 667F , or JX EE -KIN Grb2 in the addition of pFA-Elk1 and pFR-Luciferase reporter constructs, and pSV2-␤-galactosidase. At 48 h post-transfection, the cells were washed and lysed. The luciferase activity in lysates was determined using Promega's luciferase assay system. The assay was repeated several times, and a representative experiment is shown here. from either JX Grb2 or EphB2-EE but weakly associated with a peptide that includes phosphorylated Tyr 667 in the kinase domain (Fig. 3A). In the crystal structure of the EphB2 kinase domain, Tyr 667 is exposed on the upper surface of the N-terminal lobe, situated in the ␤3-␣C linker (6), and is therefore potentially accessible for binding. Indeed we were able to abrogate RasGAP association with EphB2 only when both the JM motifs and the Tyr 667 motif were mutated in concert, as in the JX EE -KIN Grb2 mutant (Fig. 5). Single substitutions in the JM region as in EphB2-EE (28) or JX Grb2 (Fig. 2) or in the kinase domain such as KIN Grb2 did not fully inhibit recruitment of RasGAP (Fig. 5), indicating EphB2 receptor may have multiple binding sites for RasGAP. These suggestions are further supported by luciferase assay results, which demonstrate that overexpression of EphB2 mutants lacking either of the RasGAP binding sites (EphB2-EE and KIN 667F ) increased luciferase activity compared with WT EphB2. Introduction of a Grb2 binding site to these mutants, to make JX EE -KIN Grb2 and KIN Grb2 , respectively, led to further increases in luciferase activity (Fig. 7).
Multiple sites of interaction between two signaling proteins are quite common. In the context of Eph receptors, Yu et al. (36) have recently reported that EphB2 and the Abl/Arg tyrosine kinases can interact by distinct mechanisms, at two independent sites on EphB2. Why RasGAP and Abl/Arg engage in such complex binding interactions with EphB2 remains to be established. Multiple binding sites may serve to finely regulate the association of an individual target with EphB2 or to allow the selective binding of specific proteins among multiple possible targets. Peptide competition studies have demonstrated a relatively weak binding affinity between a phosphorylated EphB2 JM peptide and the N-terminal RasGAP SH2 domain (18). The presence of multiple binding sites in EphB2 may therefore stabilize its interactions with targets such as RasGAP.
The unusual ability of Eph receptors to antagonize Erk MAPK activation has provided an opportunity to engineer EphB2 variants in which biochemical signaling to the Ras-MAPK pathway is switched through the incorporation of a Grb2 binding site at Tyr 667 , such that ephrin stimulation induces Erk activation. The ability of this mutant (JX EE -KIN Grb2 ) to up-regulate Erk phosphorylation in a ligand-inducible fashion was dependent on the ablation of RasGAP binding sites in the JM region (while at the same time suppressing the inhibitory effect of the JM sequence on kinase activity). Indeed a mutant with both Grb2 and RasGAP binding sites (KIN Grb2 ) was neutral in its effects on Erk phosphorylation. These data suggest that an Eph receptor's effects on the Erk MAPK pathway can be titrated from inhibitory to neutral to stimulatory by the relative incorporation of RasGAP and Grb2 binding sites. The switch in the biochemical output of the JX EE -KIN Grb2 mutant is accompanied by a loss of the ephrin-B1-induced neurite retraction observed in NG108 cells expressing WT EphB2. These data are consistent with the possibility that down-regulation of the Ras-MAPK pathway is involved in the process of neurite retraction induced by EphB2 (28); indeed, these observations may have a more general significance, because neuronal outgrowth induced by netrin-1 signaling through the attractive DCC receptor appears dependent on MAPK activation and is antagonized by inhibition of the MAPK pathway (37). RTKs have generally been viewed as being regulated by autophosphorylation within their kinase domains and forming docking sites for cytoplasmic targets within non-catalytic regions of the receptor. Although this scheme is generally applicable, it is apparent that Eph receptors are organized in a somewhat more complex fashion, in which the non-catalytic JM region can both regulate kinase activity and bind SH2-containing targets. The importance of the sequence and, likely, the conformation of the JM region in integrating Eph receptor signaling are revealed by the observation that altering the FIG. 7. Neurite retraction assay in neuronal cells expressing WT or mutant EphB2. A, images of representative NG108 cells transiently transfected with EphB2-GFP and stimulated with aggregated ephrin-B1Fc. NG108 cells were transiently transfected with WT and mutant EphB2-GFP and were differentiated by the addition of 1 mM cAMP for 5 h. Transfectants were stimulated with 2 g/ml clustered ephrin-B1Fc for 20 min, and cells with neurite length greater than twice the cell body length were monitored for neurite retraction. B, relative neurite length of WT and mutant EphB2 in response to ephrin-B1Fc. The data are expressed as a percentage of the original neurite length before the ligand stimulation. Each error bar represents the mean and S.E. of more than 10 cell experiments for each construct. sequence C-terminal to the JM tyrosines interferes with both its regulatory properties, as well as its ability to become phosphorylated and bind cytoplasmic polypeptides. Furthermore, we have identified a site within the kinase domain itself, Tyr 667 , which has the potential to act as a docking site for SH2 proteins. Although the physiological importance and general applicability of this observation remains to be fully explored, it emphasizes that RTK targets may potentially interact with the kinase domain at sequences removed from the active site. In a similar fashion, the kinase domain of the serine/threoninespecific protein kinase glycogen synthase kinase 3␤ binds the phosphoserine/threonine residues of primed substrates through a positively charged pocket within the kinase domain (38). The complex regulation and signaling properties of Eph receptors likely reflect their sophisticated ability to control cell-cell interactions in both embryonic and adult tissues.