Oncogene (1999) 18, 5413 ± 5422 ã 1999 Stockton Press All rights reserved 0950 ± 9232/99 $15.00 http://www.stockton-press.co.uk/onc Phosphorylation at Tyr-838 in the kinase domain of EphA8 modulates Fyn binding to the Tyr-615 site by enhancing activity

Sunga Choi1 and Soochul Park*,1

1Institute of Environment and Life Science, Hallym University, 1 Okcheon-dong, Chuncheon, Kangwon-do, 200-702, Korea

Eph-related receptors and their ligands are highly and they are activated by cell-surface bound ligands conserved families which play important roles in termed (van der Geer et al., 1994; Eph targeting axons and migrating cells. In this study we Nomenclature Committee, 1997). Eph-related recep- have examined the functional roles of two major tors have been classi®ed into two subfamilies, EphA autophosphorylation sites, Tyr-615 and Tyr-838, in the and EphB, depending on their preferential binding to EphA8 receptor. Two-dimensional phosphopeptide map- either glycosylphosphatidylinositol (GPI)-anchored ping analysis demonstrated that Tyr-615 and Tyr-838 ephrin-A ligands or transmembrane ephrin-B ligands. constitute major autophosphorylation sites in EphA8. Since ephrins are membrane-bound that Tyr-615 was phosphorylated to the highest stoichiome- appear to require multimerization for full activity try, suggesting that phosphorylation at this site may have (Davis et al., 1994; Stein et al., 1998a), Eph-ephrin a physiologically important role. Upon conservative interactions are likely to mediate cell-cell signaling mutation of Tyr-838 located in the tyrosine kinase events. domain, the catalytic activity of EphA8 was strikingly Physiological roles for Eph-related receptors and reduced both in vitro and in vivo, whereas a mutation at ephrin ligands have been extensively studied in primary Tyr-615 in the juxtamembrane domain did not impair the culture systems and using in vivo genetic experiments. tyrosine kinase activity. In vitro binding experiments Repulsive modes of interaction between Eph receptors revealed that phosphorylation at Tyr-615 in EphA8 and their ligands have been proposed to set up mediates the preferential binding to Fyn-SH2 domain topographic projections in the retinotectal or retino- rather than Src and Ras GTPase-activating protein (Ras collicular systems (Cheng et al., 1995; Drescher et al., GAP)-SH2 domains. Additionally, a high level of EphA8 1995; Nakamoto et al., 1996; Monschau et al., 1997; was detected in Fyn immunoprecipitates in intact cells, Frisen et al., 1998), the hippocampal and thalamic indicating that EphA8 and Fyn can physically associate neurons (Gao et al., 1996, 1998), and the guidance of in vivo. In contrast, the association of full-length Fyn to migrating neural crest cells and motor axons to the EphA8 containing mutation at either Tyr-615 or Tyr-838 rostral halves of developing somites (Krull et al., 1997; was greatly reduced. These data indicate that phosphor- Wang and Anderson, 1997). Abnormalities in the ylation of Tyr-615 is critical for determining the formation of brain commissures have been observed association with Fyn whereas the integrity of Tyr-838 in mice lacking EphB2, EphB3 and EphA8 receptors phosphorylation is required for ecient phosphorylation (Henkemeyer et al., 1996; Orioli et al., 1996; Park et at Tyr-615 as well as other major sites. Finally, it was al., 1997). Although these studies suggest a major observed that cell attachment responses are attenuated function of Eph receptors and their ligands in the by overexpression of wild type EphA8 receptor but to neural patterning of the vertebrate nervous system, much less extent by EphA8 mutants lacking phosphor- Eph and ephrin function is also implicated in ylation at either Tyr-615 or Tyr-838. Furthermore, angiogenesis and vascular network assembly (Pandey transient expression of kinase-inactive Fyn in EphA8- et al., 1995b; Stein et al., 1998a; Wang et al., 1998). overexpressing cells blocked cell attachment responses Unlike other RTKs, a two-step mechanism of Eph attenuated by the EphA8 signaling. We therefore receptor activation was recently proposed; the ephrin propose that Fyn kinase is one of the major downstream ligand binding is restricted to the amino-terminal targets for the EphA8 signaling pathway leading to a globular domain of Eph receptors, whereas the modi®cation of cell adhesion, and that autophosphoryla- distinct domain in the carboxy-terminal region of the tion at Tyr-838 is critical for positively regulating the receptor ectodomain mediates ligand-independent EphA8 signaling event. receptor-receptor oligomerization leading to receptor transphosphorylation (Lackman et al., 1998). During Keywords: Eph; ; Fyn the past few years, several immediate downstream targets of Eph receptors have been identi®ed. Several known signaling proteins containing SH2 domains, such as Fyn, Src, and Ras GTPase-activating protein Introduction (Ras GAP) have been shown to interact with the conserved tyrosine phosphorylation site in the juxta- The Eph family of receptors represents the largest membrane domain of Eph receptors (Ellis et al., 1996; known subfamily of receptor tyrosine kinases (RTKs) Holland et al., 1997; Hock et al., 1998; Zisch et al., 1998). It has also been reported that direct interaction of EphB1 with the SH2 domain of the adapter protein Nck involves Tyr-594 phosphorylation in EphB1 (Stein *Correspondence: S Park Received 15 February 1999; revised 8 April 1999; accepted 15 April et al., 1998b). Phosphorylated Tyr-929, in the carboxy- 1999 terminal tail of EphB1, has been shown to associate Modulation of EphA8-Fyn interactions SChoiandSPark 5414 with the SH2 domain adapter protein Grb10 (Stein et al., 1998). Interestingly, EphA8 lacks the tyrosine al., 1996). Interestingly, the equivalent Tyr-929 in residue corresponding to Tyr-596 in EphA4, which is EphB1 was required for the recruitment of low also highly conserved and may modulate protein molecular weight protein tyrosine phosphatase binding to its neighboring tyrosine residue (Figure 1) (LMW-PTP) to receptor complexes upon treatment of (Ellis et al., 1996). multimeric ephrin ligands (Stein et al., 1998a). Tyr-838 is located in the tyrosine kinase domain of Although the interacting tyrosine sites in Eph the EphA8 receptor and its corresponding tyrosine receptors have not been de®ned yet, Grb2, SLAP, residues are found in all known members of the Eph and the regulatory subunit p85 of PI-3 kinase were also family receptors (Figure 1). Despite this fact, the implicated as immediate downstream targets for Eph functional role of this conserved tyrosine residue has receptors (Pandey et al., 1994, 1995a; Stein et al., not been determined yet. The ®nding that tyrosine 1996). Some studies have indicated that Eph receptors residues corresponding to Tyr-615 and Tyr-838 in do not exert pronounced mitogenic or transforming EphA8 are highly conserved in all Eph-related activities (Lhotak and Pawson, 1993; Brambilla et al., receptors suggested to us that phosphorylation of 1995). Rather than mediating mitogenic signaling, Eph these tyrosine residues may play a pivotal role in the receptors are likely to regulate the cell cytoskeleton Eph signaling pathway. reorganization, on the basis of their biological functions in axon and cell targeting, and further Mutation of Tyr-838 in the kinase domain of EphA8 supporting evidence (Meima et al., 1997; Stein et al., impairs kinase activity 1998a,b). It remains to be determined how the cytoplasmic targets downstream of Eph receptors To investigate the functional roles of phosphorylations regulate the cytoskeleton, leading to axon collapse or at Tyr-615 and Tyr-838 in EphA8, the codons for cell attachment. either Tyr-615 or Tyr-838 were substituted with those We have previously reported the molecular cloning for phenylalanine within the EphA8 cDNA as of a full-length cDNA for mouse EphA8, a member of described under Materials and methods (Figure 2a). the EphA subgroup (Park and Sanchez, 1997). Further At the COOH terminus of each EphA8 construct, a 9- studies using mice defective for EphA8 indicated that amino acid HA epitope was appended to facilitate EphA8 receptors play an important role in axonal immunoprecipitation of exogenously expressed EphA8 path®nding for a subset of tectal commissural axons with an anti-HA antibody. We also generated a rabbit (Park et al., 1997). In the present studies, we characterized two major autophosphorylation sites in EphA8. Phosphorylation of Tyr-615 in the juxtamem- brane domain mediated a high anity association with the SH2 domain of Fyn both in vitro and in vivo. Phosphorylation of Tyr-838 in the kinase domain was critical for maintaining its catalytic tyrosine kinase activity in an active state, thereby modulating Fyn binding to the Tyr-615 site in the juxtamembrane domain. Functional studies using kinase-inactive Fyn suggest that the recruitment of active Fyn to the activated EphA8 receptor plays an important role in attenuating cell attachment responses. Our ®ndings support a model that Tyr-838 in EphA8 is an important regulatory phosphorylation site determining positive Eph signaling events such as cellular cytoskeletal modi®cations mediated through the Fyn tyrosine kinase.

Results

Tyr-615 and Tyr-838 in EphA8 are highly conserved in all known Eph-related kinases The Eph-related receptors contain a variable number of tyrosine residues in their cytoplasmic region. Some of these tyrosine residues are highly conserved among all known members of Eph receptor family. Tyr-615 in EphA8 is located in the juxtamembrane domain and all Figure 1 Tyr-615 and Tyr-838 in EphA8 are highly conserved in known Eph-related receptors possess the corresponding all Eph-related receptors. The juxtamembrane and tyrosine kinase amino acid sequences of Eph family members are aligned in tyrosine residue in their juxtamembrane domain relation to the Tyr-615 and Tyr-838 residue of EphA8, (Figure 1). Previous studies indicated that phosphor- respectively. The horizontal bar indicates the identical amino ylation at this conserved tyrosine residue mediates high acid present in EphA1, with the exception that highly conserved anity association with the SH2 domains of either the tyrosine residues are indicated in bold type. A small letter nonreceptor tyrosine kinases or Ras-GAP (Ellis et al., representing chick (c), mouse (m), rat (r), and human (h) was included in a parenthesis next to each Eph receptor to indicate 1996; Holland et al., 1997; Hock et al., 1998; Zisch et their cloning origin Modulation of EphA8-Fyn interactions SChoiandSPark 5415 polyclonal antibody directed against the juxtamem- ciently autophosphorylated (Figure 2b, top panel, lanes brane region of the EphA8 receptor, which recognized 4 and 5). Thus our result strongly suggests that speci®cally EphA8 but not EphA4 (data not shown). blocking phosphorylation of Tyr-838 in the kinase These EphA8 expression constructs were transiently domain impairs the tyrosine kinase activity of the transfected into 293T cells and lysates from each EphA8 receptor in vivo. transfected cell line were prepared for immunoprecipi- To investigate more closely if defective phosphoryla- tation with anti-HA antibodies. Immunoprecipitates tion at Tyr-838 a€ects the kinase activity of EphA8 were then assayed by immunoblot using either anti- towards exogenous substrates, we performed an in vitro phosphotyrosine antibodies or anti-EphA8 antibodies kinase assay using enolase as a substrate. Lysates of as probes (Figure 2b). Similar amounts of HA-tagged 293T cells transfected with the indicated constructs EphA8 were present in lysates from cells expressing were prepared as described above. Immunoprecipitates wild type EphA8, EphA8F615, EphA8F838 and prepared with anti-HA antibodies were either directly EphA8F615,838, respectively, but none was detectable in incubated with [32P] ATP plus enolase or analysed by control cell lysate (Figure 2b, bottom panel). It was immunoblot using anti-EphA8 antibodies to compare evident that both wild type EphA8 and mutant EphA8 overall levels of HA-tagged EphA8 present in each containing Phe in the place of Tyr-615 are highly immunoprecipitate (Figure 3). In immunoprecipitates autophosphorylated without further stimulation with containing wild type EphA8 or EphA8F615, most of exogenous ligands (Figure 2b, top panel, lanes 2 and phosphorylated proteins were detected as either EphA8 3). The reason for exogenous ligand-independent or enolases (Figure 3, top panel, lanes 2 and 3). EphA8 phosphorylation may be an abundant expres- Unknown proteins migrating as 60 kDa were also sion of endogenous ephrin-A ligands since the eciently and reproducibly phosphorylated in this in extracellular domain of the EphA8 receptor bound to vitro kinase assay. In contrast, neither EphA8 nor the cell surfaces within nM anity range (data not enolases were eciently phosphorylated in immuno- shown). Further treatment of exogenous ligands as precipitates containing mutant EphA8F838 or either dimeric or multimeric forms did not stimulate EphA8F615,838 (Figure 3, top panel, lanes 4 and 5). additional phosphorylation of EphA8 in these experi- These results demonstrate that phosphorylation at Tyr- ments (data not shown). In contrast, the mutant 838 in the kinase domain of EphA8 is critical for the receptor EphA8F838 and EphA8F615,838 were not effi- kinase activity.

Figure 2 (a) Schematic diagrams of wild type EphA8, and the EphA8 mutants containing Phe in the place of either Tyr-615 or Tyr-838. For each protein, the solid boxes and hatched boxes represent the transmembrane domain and the catalytic tyrosine kinase domain, respectively. Tyr-615 (Y615) and Tyr-838 (Y838) of EphA8 are indicated together with its mutated residue, Phe-615 (F615) and Phe-838 (F838), respectively. The 9-amino acid HA epitope tag is fused to each COOH-terminal end, which is not indicated in this diagram. Proteins were expressed under the CMV promoter from the pcDNA3 vector. (b) In vivo tyrosine phosphorylation of EphA8, EphA8F615, EphA8F838 and EphA8F615,F838. The indicated construct was transfected into 293T cells and proteins from each cell lysate were immunoprecipitated using anti-HA antibodies. The arrows indicate the EphA8 receptor. (Top panel) 80% of each anti-HA immunoprecipitate were separated by 7.5% SDS ± PAGE and probed with anti-phosphotyrosine antibodies for ECL Western blot analysis. (Bottom panel) ECL Western blot analysis of 20% of each immunoprecipitate using anti-EphA8 antibodies elicited against a bacterial GST-EphA8 juxtamembrane fusion proteins Modulation of EphA8-Fyn interactions SChoiandSPark 5416

Figure 4 Comparison of tryptic phosphopeptide maps of wild type EphA8, EphA8F615 and EphA8F838.(a) 32P-labeled EphA8 or EphA8 phosphorylation mutants resolved by SDS ± PAGE as described in the legend to Figure 3 were electrophoretically transferred to Immobilon membranes. Phosphorylated EphA8 or EphA8 mutants immobilized on membranes were digested with Figure 3 Tyr-838 in the kinase domain of ephA8 is critical for trypsin and subjected to phosphopeptide mapping analysis. tyrosine kinase activity. Proteins from lysates of 293T cells Approximately 6000 c.p.m. of tryptic phosphopeptides were transfected with the indicated constructs were immunoprecipitated analysed on TLC plates by electrophoresis at pH 1.9 bu€er in with anti-HA antibodies. (Top panel) 80% of each anti-HA the horizontal dimension with the anode on the left and ascending immunoprecipitate were incubated in an in vitro kinase reaction chromatography in the vertical dimension. The origin is marked with [32P]ATP, and acid-denatured enolase as exogenous with an arrow, and the major phosphopeptides are labeled a, b, substrate. Entire 32P-labeled reaction products were analysed by b', c, and d (Left panel) tryptic phosphopeptide map of wild type 10% SDS ± PAGE and autoradiography. Molecular mass markers EphA8 phosphorylated in vitro. (Middle panel) tryptic phospho- are indicated in kDa. 60 kDa protein phosphorylated in vitro by peptide map of in vitro phosphorylated EphA8 mutant containing wild type EphA8 or EphA8F615 is indicated by*. (Bottom panel) Phe in the place of Tyr-615. (Right panel) tryptic phosphopeptide anti-EphA8 immunobot of 20% of each anti-HA immunopreci- map of in vitro phosphorylated EphA8 mutant containing Phe in pitate the place of Tyr-838. (b) Phosphopeptides b and b' were separately extracted from the TLC plate containing tryptic phosphopeptides of wild type EphA8, further digested with trypsin, and then analysed on a TLC plate by one-dimensional ascending chromatography. The arrow indicates the sample origin Tyr-615 and Tyr-838 are major autophosphorylation sites of the EphA8 receptor To further investigate whether Tyr-615 and Tyr-838 are site (Figure 4a, second panel). Indeed, amino acid major autophosphorylation sites in EphA8, we sequences ¯anking the trypsin cleavage site near to performed phosphopeptide mapping experiments de- Tyr-615 consists of -Lys-Phe-Pro-Glu-Thr-, which is scribed below. 32P-labeled EphA8 or EphA8 mutants known to be an inecient cleavage site for trypsin were isolated, digested with trypsin, and then analysed (Boyle et al., 1991). To further assess which by two-dimensional peptide mapping using electro- phosphopeptide is a partial cleavage product, phos- phoresis and chromatography on TLC plates (Figure phopeptide b and b' were separately extracted from 4a). Wild type EphA8 contained ®ve major labeled TLC plates and then subjected to further digestion tryptic phosphopeptides (a, b, b', c and d) as well as a with trypsin. These tryptic digests were analysed by few minor phosphopeptides (Figure 4a, ®rst panel). one-dimensional chromatography. This data revealed Most of the minor phosphopeptides that are variably that phosphopeptide b' are further cleaved to produce detected appear to arise from in vitro phosphorylated phosphopeptide b, demonstrating that phosphopeptide proteins nonspeci®cally immunoprecipitated by anti- b' is a partial cleavage product containing Tyr-615 HA antibodies because these phosphopeptides were (Figure 4b). Since two major phosphopeptides b and b' also observed from the 32P-labeled control immunopre- contained phosphorylated Tyr-615 residue, Tyr-615 is cipitate lacking the HA-tagged EphA8 receptors stoichiometrically the most abundant autophosphor- (Figure 3, lane 1; data not shown). In contrast, ylation site of the EphA8 receptor. EphA8F615 contained only three major tryptic phospho- For the EphA8F838 mutant protein, the phosphopep- peptides (a, c, and d), indicating that both phospho- tide map resembled that of the wild type EphA8 peptide b and b' include Tyr-615 as a phosphorylation receptor, with the exception that one of the ®ve major Modulation of EphA8-Fyn interactions SChoiandSPark 5417 phosphopeptides, spot d, is missing (Figure 4a, third domains directly interacting with the tyrosine phos- panel). Speci®c loss of phosphopeptide d in EphA8F838 phorylated EphA8 receptors, in vitro binding assays was con®rmed by mapping mixes containing similar were performed using the bacterially expressed GST- counts of tryptic digests from wild type EphA8, and SH2 fusion proteins of Fyn, Src, Ras-GAP or the PI-3 EphA8F838 (data not shown). The presence of additional kinase p85 subunit. In these experiments, approxi- phosphopeptides a and c indicates that there are at mately equal amounts of puri®ed Fyn-SH2, Src-SH2, least two additional autophosphorylation sites in RasGAP-SH2, p85-SH2 or GST alone, were bound to EphA8, which were not further characterized for these glutathione-Sepharose beads, then mixed with equiva- experiments. The stoichiometry of Tyr-838 phosphor- lent amounts of lysates prepared from 293 cells stably ylation was less than that of Tyr-615 phosphorylation, expressing HA-tagged EphA8. Proteins from the cell re¯ecting than phosphorylation of Tyr-838 may be lysates that bound to each SH2 fusion protein were unstable. Taken together, these results clearly demon- recovered with the glutathione-Sepharose beads, and strate that Tyr-615 in the juxtamembrane domain and then analysed by immunoblot using anti-EphA8 Tyr-838 in the kinase domain constitute major antibodies as probe (Figure 5a). Lower molecular autophosphorylation sites of the EphA8 receptor. weight species, which appear to be degradation products derived from either GST-fusion proteins of EphA8, were variably detected. Nevertheless, these EphA8 receptors associate with the SH2 domain of the data clearly showed that wild type EphA8 receptors Fyn kinase formed a stable complex with the Fyn-SH2 fusion Previous studies indicated that tyrosine phosphorylated proteins, but not with GST alone (Figure 5a, lanes 1 Eph-related receptors directly interact with the SH2 and 2). It was reproducibly observed that wild type domains of either nonreceptor tyrosine kinases or Ras EphA8 bound to the SH2 domain of Fyn with a much GAP (Ellis et al., 1996; Holland et al., 1997; Hock et higher anity than that of Src (Figure 5a, lanes 2 and al., 1998; Zisch et al., 1998). To identify the SH2 3). In contrast, for the SH2 domains of either the Ras-

Figure 5 (a) A preferential binding of EphA8 to the Fyn-SH2 domain. Approximately equal amounts of puri®ed GST alone, or GST-SH2 fusion proteins of Fyn, Src, Ras-GAP, and p85 subunit of PI-3 kinase, respectively, were bound to glutathione-Sepharose beads and then incubated with equal amounts of lysates from wild type EphA8 expressing 293 cells. Proteins that associated stably with various GST fusion proteins were separated by 7.5% SDS ± PAGE and then probed by anti-EphA8 antibodies for ECL Western blot analysis. The EphA8 receptor bound to the SH2 fusion protein is marked by an arrow. (b) Evidence that phosphorylation at Tyr-615 in EphA8 mediates high anity association with the Fyn-SH2 domain. (Top panel) equivalent amounts of puri®ed Fyn-SH2 fusion proteins were bound to glutathione-Sepharose beads and then incubated with the lysates of 293 cells stably transfected with the indicated expression constructs. The washed beads were separated by 7.5% SDS ± PAGE, and probed on a Western blot with anti-EphA8 antibodies, followed by detection with ECL reagents. (Bottom panel) a fraction of each cell lysate was separately fractionated by SDS ± PAGE and then analysed by immunoblot using anti-EphA8 antibodies as probe Modulation of EphA8-Fyn interactions SChoiandSPark 5418 GAP or the p85 subunit, the ability to bind to the (Figure 6a). As expected, a high level of wild type EphA8 receptor was minimal, compared with the Fyn- EphA8 was detected in Fyn immunoprecipitates from SH2 domain (Figure 5a, lanes 4 and 5). Together, these the cells expressing wild type EphA8 but not parental results demonstrate that, like some members of the cells (Figure 6a, lanes 1 and 2). In contrast, for Eph-related receptors, the EphA8 receptor is able to EphA8F615, EphA8F838 or EphA8F615,F838, the ability to form a stable complex with the Fyn-SH2 domain. bind full-length Fyn was abolished (Figure 6a, lanes 3, Similar experiments revealed no stable association with 4, and 5). The kinase-inactive Fyn protein retained its several other SH2-containing proteins, including PLC- ability to associate with the EphA8 receptor (Figure g, and (data not shown). 6a, lane 6). These data demonstrate that phosphoryla- tion at Tyr-615 in EphA8 mediates association with the Fyn kinase in vivo. Furthermore, our data indicate that Phosphorylation at Tyr-615 in EphA8 mediates constitutive phosphorylation at Tyr-838 in the kinase association with the Fyn kinase domain is required to mediate the association of It is likely that phosphorylation at Tyr-615 in EphA8 signaling proteins such as Fyn by escalating phosphor- mediates association with the Fyn kinase, on the basis of the reports that phosphorylation at the correspond- ing residue in other Eph-related receptors is required for binding to the SH2 domains of nonreceptor tyrosine kinases or Ras-GAP (Ellis et al., 1996; Holland et al., 1997; Hock et al., 1998; Zisch et al., 1998). To verify that phosphorylation at Tyr-615 mediates strong association with Fyn-SH2 fusion proteins, in vitro binding experiments were performed using lysates prepared from 293 cell lines stably expressing wild type EphA8, EphA8F615, EphA8F838 and EphA8F615,838, respectively. No changes in cell morphology were apparent in these stably transfected cells irrespective of EphA8 expression (data not shown). Immunoblotting analysis of whole cell lysates showed that similar amounts of EphA8 were present in lysates from cells stably expressing wild type EphA8, EphA8F615, EphA8F838 and EphA8F615,838, respectively. (Figure 5b, bottom panel). For in vitro binding experiments, equal amounts of Fyn-SH2 fusion proteins bound to glutathione-Sepharose beads were mixed with cell lysates containing EphA8, EphA8F615, EphA8F838 or EphA8F165,F838. The washed beads were assayed by immunoblot using anti-EphA8 antibodies as probe (Figure 5b, top panel). As expected, EphA8F615 did not exhibit strong association with Fyn-SH2 (Figure 5b, top panel, lane 3). Interestingly, neither did EphA8F838 show strong binding to Fyn-SH2 despite the high level of protein expression (Figure 5b, lane 4). As shown in phosphopeptide mapping experiments, the reduced binding of Fyn-SH2 to EphA8F838 is likely due to inecient phosphorylation of Tyr-615. Taken together, our results demonstrate that phosphorylation of tyrosine 615 in EphA8 is required for binding of the SH2 domain of Fyn. In order to investigate the biological relevance of the interaction between Fyn-SH2 domain and the EphA8 receptor detected in vitro, we assessed whether EphA8 associates with the Fyn kinase in intact cells. A Figure 6 (a) Evidence that phosphorylation at Tyr-615 in EphA8 fraction of cell lysates containing EphA8, EphA8F615, is critical for stable association with the full-length Fyn kinase in intact cells. Proteins from each cell lysate were immunoprecipi- EphA8F838 or EphA8F615,F838 were analysed by immuno- tated with anti-Fyn antibodies and then analysed by immunoblot blot using either anti-EphA8 antibodies or anti-Fyn using anti-EphA8 antibodies as probe. For expression of kinase- antibodies as probes in order to show the relative inactive Fyn (Lys-299?Met), the pcDNA3-derived expression amounts of EphA8 or Fyn present in those cell lysates plasmids were transiently transfected into the EphA8-expressing prepared for immunoprecipitation (Figure 6b). Tran- 293 cells using LipofectAMINE and then cells were lysed 48 h after transfection. (b) A fraction of each cell lysate was separately sient expression of kinase-inactive Fyn with a single fractionated by SDS ± PAGE and then analysed by immunoblot mutation in the ATP- (Twamley-Stein et using anti-EphA8 antibodies (top panel) or anti-Fyn antibodies al., 1993) in EphA8-expressing cells resulted in 4®ve- (bottom panel) as probes. (c) Comparison of cell attachment fold increase over the endogenous, active protein responses. Parental 293 cells or stably transfected cells expressing (Figure 6b, bottom panel, lane 6). Cell lysates were the indicated EphA8 receptors as described above were plated on dishes as described in Materials and methods. The percentage of immunoprecipitated with anti-Fyn antibodies and total cells attached after 60 min is displayed. Each data set immunoblotted with anti-EphA8 antibodies as probe represents the mean of three separate experiments +s.e.m. Modulation of EphA8-Fyn interactions SChoiandSPark 5419 ylation at Tyr-615 in the juxtamembrane domain of not altered by the treatment of PI-PLC, possibly due to EphA8. modi®cation of GPI-anchors in 293 cells. Nevertheless, our results based on the constitutive tyrosine phosphorylation processes in 293 cells are in a good Activation of the EphA8 receptor attenuates cell agreement with the notion that phosphorylation of attachment tyrosines within the catalytic domain is essential for The roles for Eph receptors and their ligands in cell-cell maintaining the tyrosine kinase in an active state, repulsion have been well established (Drescher et al., whilst tyrosine phosphorylation in noncatalytic regions 1995; Nakamoto et al., 1996; Monschau et al., 1997; leads to generation of docking sites for SH2 or PTB Frisen et al., 1998). To investigate whether the EphA8 domains of signaling molecules (van der Geer et al., signaling a€ects cell adhesion, the same cell lines used 1994; Schlessinger et al., 1997). for demonstrating Fyn association in vivo were tested The data presented here provide the ®rst character- for their properties to attach to culture dishes. After ization of an autophosphorylation site, Tyr-838, seeding the equivalent numbers of each cell line, the conserved in all Eph-related receptors. Our phospho- plates were washed with PBS after 60 min and the peptide mapping analysis indicated that the stoichio- fraction of attached cells were calculated. In multiple metry of Tyr-838 phosphorylation is less than that of experiments, cells expressing a high level of the wild Tyr-615 phosphorylation. This may re¯ect that type EphA8 receptor showed a signi®cant reduction of phosphorylation of Tyr-838 may be unstable during cell attachment, whereas the attachment responses of the preparation of phosphopeptides. Alternatively, Tyr- cell lines expressing EphA8F615, EphA8F838 or 838 may be highly phosphorylated sites in intact cells EphA8F615,838 were reduced to a much less extent and not be subject to incorporation of label. Initially, (Figure 6c). Furthermore, co-expression of kinase- the motif YWNM containing Tyr-838 led us to inactive Fyn with EphA8 resulted in marked increases investigate whether phosphorylation at this site in cell attachment, suggesting that the kinase-inactive mediates the association with the p85 subunit SH2 form of Fyn competed for the binding site for domain of PI-3 kinase. Although EphA8 bound to the endogenous Fyn and possibly other non-receptor p85 subunit SH2 domain at a low level, this interaction tyrosine kinases on the activated EphA8 receptor. was not dependent on the phosphorylation at Tyr-838 Similar experiments were carried out using two (unpublished data). Recently, it was proposed that the additionally independent clones expressing wild type Eph receptor activation takes place in a two-step EphA8, which gave virtually identical results (data not mechanism with distinct ligand binding domain and shown). Together, our ®ndings suggest that the EphA8 ligand independent receptor-receptor oligomerization signaling interferes with the adhesive properties of the events (Lackmann et al., 1998). Likewise, transphos- cells through the Fyn signaling. phorylation of Eph receptors may occur in a stepwise mechanism, in a way that Tyr-838 is primarily phosphorylated to maintain the kinase domain in an Discussion active state. Our ®nding that Tyr-838 in EphA8 is critical for the tyrosine kinase activity also suggests In this report, we examined the major sites of tyrosine that this tyrosine residue may be an important phosphorylation of EphA8 as well as its interactive autoregulatory site for Eph signaling pathway. For SH2 partners. Tryptic peptide mapping analyses of instance, dephosphorylation at Tyr-838 by yet un- various EphA8 mutants from in vitro kinase reactions known tyrosine phosphatase would lead to an inactive established that Tyr-615 and Tyr-838 in EphA8 state of the kinase domain, thereby turning o€ the constitute major autophosphorylation sites. The lack signal transduction. Interestingly, it was reported that of phosphorylation at Tyr-838 located in the kinase EphB1 receptor complexes recruit low molecular domain of EphA8 impaired the kinase catalytic weight phosphotyrosine phosphatase (LMW-PTP) activity. In contrast, the enzymatic activity of EphA8 upon treatment with multimeric ligands (Stein et al., containing Phe at the place of Tyr-615 was not 1998a). It will be interesting to examine whether signi®cantly impaired. Like some other Eph-related LMW-PTP is able to eliminate phosphorylation at receptors, phosphorylation at Tyr-615 in the juxta- Tyr-838 in EphA8. membrane domain of the EphA8 receptor was critical Our ®nding that Tyr-615 in the juxtamembrane for strong association with the Fyn kinase in a manner domain of EphA8 is the most abundant autophos- dependent of the SH2 domain. Furthermore, inecient phorylation site and also the predominant binding site phosphorylation at Tyr-615 due to the lack of Tyr-838 for Fyn suggests that the signaling mechanism of phosphorylation resulted in a great reduction of Fyn EphA8 is similar to that of EphA4. However, EphA8 association to EphA8. 293 cells are likely to express does not possess the tyrosine residue corresponding to endogenous ligands for the EphA8 receptor since the Tyr-596 in EphA4, which is another juxtamembrane extracellular domain of EphA8 binds to the parental autophosphorylation site conserved in most Eph- 293 cells with a signi®cant anity (data not shown). related receptors. It has been postulated that phos- This is consistent with the result that the EphA8 phorylation at Tyr-596 plays an important role in receptors expressed in 293 cells are constitutively modulating Fyn binding to the Tyr-602 site of EphA4, tyrosine-phosphorylated. Moreover, tyrosine phosphor- the residue corresponding to Tyr-615 in EphA8, since ylation status of EphA8 in 293 cells was not mutation of Tyr-596 to Phe led to a reduction in stable signi®cantly altered by further treatment of exogenous association between EphA4 and Fyn (Ellis et al., 1996). ephrin ligands regardless of being clustered by anti-Fc Other studies also indicated that phosphorylation of antibody. Interestingly, the capacity of the extracellular the tyrosine residue corresponding to Tyr-596 in domain of EphA8 to bind the parental 293 cells was EphA4 creates the formation of a Nck docking site Modulation of EphA8-Fyn interactions SChoiandSPark 5420 in EphB1 (Stein et al., 1998b). The absence of the term potentiation, spatial learning, hippocampal corresponding tyrosine residue in EphA8 suggests that development, and myelination (Grant et al., 1992; phosphorylation of Tyr-615 is sucient to create a Yagi et al., 1993b; Umemori et al., 1994). Our previous high anity binding motif for Fyn at least in the case studies using ephA87/7 mice indicated that EphA8 of the EphA8 receptor. It remains to be determined receptors play an important role in axonal path®nding whether, in addition to Tyr-615, there are other sites in for a subset of tectal commissural axons (Park et al., EphA8 to which Fyn-SH2 domain binds since the 1997). Many other studies also indicated that the EphA8F838 mutant retains the capacity to bind the Fyn- interaction between Eph receptors and their corre- SH2 domain albeit the greatly reduced endogenous sponding ligands in neuronal cells regulate growth cone tyrosine kinase activity of EphA8. Unlike EphB2 and motility or axon fasciculation, rather than cell other Eph-related receptors, phosphorylation at Tyr- proliferation (Orioli and Klein, 1997). Therefore, Fyn 615 in EphA8 resulted in the preferential recruitment may represent the functional link between the EphA8 of Fyn, rather than Src and Ras-GAP, to signaling receptor and the downstream signaling proteins that pathway of these receptors. It is not clear how the SH2 regulate these neuronal processes. The determination of domains of nonreceptor tyrosine kinases discriminate how the EphA8 and Fyn signaling regulates the cell between Eph receptors, but it is likely that additional cytoskeleton in vivo requires additional experiments, nonconserved amino acids, in addition to the sequence including actin ®lament staining and characterization motif identi®ed as the key site of interaction may of focal adhesion or cell adhesion proteins. Never- provide the speci®city observed. It was recently theless, our data on the functional roles of major proposed that the sequence divergence at position +5 tyrosine phosphorylation sites in EphA8 should relative to Tyr-614 in EphB3, the site corresponding to facilitate further studies directed toward how the Tyr-615 in EphA8, might be crucial for determining the EphA8 receptors supports the development of the binding properties of Ras-GAP (Hock et al., 1998). neuronal cells in the brain. For instance, the presence of an acidic amino acid such as glutamic or aspartic acid at position +5 determined the binding anity to the amino-terminal SH2 domain Materials and methods of Ras-GAP as demonstrated in the cases of EphB1 and EphB3. Like EphA4, our data is in a good Cell culture and transfection agreement with this ®nding as the nonconservative arginine residue at position +5 relative to the Tyr-615 293T cells were maintained in a humidi®ed atmosphere of 5% CO2 at 378C in Dulbecco's modi®ed Eagle's medium (Sigma) in EphA8 was consistent with the result that Ras-GAP supplemented with 10% heat-inactivated fetal bovine serum. SH2 weakly bound to the EphA8 receptor in an in vitro 293 cells were routinely cultured in alpha-MEM (Sigma) binding experiment. containing 10% heat-inactivated fetal bovine serum. For It seems that the association between EphA8 and stable transfection with pcDNA3-derived expression plas- Fyn is very stable in intact cells since it withstands mids, the calcium phosphate precipitation method was used stringent detergent bu€er conditions such as 1% Triton as described previously (Graham and van der Eb, 1973). X-100. Furthermore, the adhesive properties of the Stable G418-resistant clones were selected by supplementing cells were attenuated in a good correlation with the the culture medium with 250 mg/ml G418. The clones were kinase activity and Fyn association. Dominant negative periodically cultured in the same selection medium to approach using kinase-inactive Fyn revealed that the maintain stable expression. For transient transfections, cells were transfected using LipofectAMINE (Life Technologies, Fyn signaling plays an important role in linking Inc) as described by the manufacturer. EphA8 signaling to cell attachment responses. It is unlikely that phosphorylation at Tyr-615 in EphA8 is mediated by the cellular Fyn kinase associating with Construction of HA-tagged EphA8 EphA8 since tyrosine phosphorylation status of EphA8 To introduce the 9 amino acid HA epitope (YPYDVPDYA) in cells transfected with the kinase-inactive Fyn mutant at the COOH terminus of murine EphA8, two PCR was not altered (data not shown). Our data on EphA8 oligomers were synthesized. A 58- oligonucleotide function on cell adhesion appear contradictory to consisted of the extreme 3' end 18 bp nucleotides of the previous reports indicating that activation of Eph ephA8 coding sequence, the HA tag sequence, a TAA stop receptors promotes cell attachment response in codon, and a SalI site at the 3' end. A 20 bp oligonucleotide corresponded to nucleotides 2833 ± 2852 of the murine ephA8 endothelial- or teratocarcinoma-derived cell lines coding sequence. PCR using these amplimers was performed (Stein et al., 1998a,b). These di€erences may re¯ect with Taq polymerase (Boehringer Mannheim) to amplify 218 the cell adhesion properties of the di€erent cell-types bp DNA fragments. The PCR products were further digested used. Consistent with our data, ectopic expression of with BssHII and SalI and the isolated DNA fragment was EGFR-XEphA4 chimeric receptor in Xenopus embryos subcloned into full-length ephA8 in pcDNA3. DNA induces dissociation of embryonic tissues, suggesting sequencing in the region of the polymerase chain reaction that the Eph signaling interferes with cell adhesion was performed to exclude the possibility of errors by the (Winning et al., 1996). We therefore propose that Fyn polymerase. is the principal target of the EphA8 receptor in the To introduce phenylalanine at the place of tyrosine, site signal transduction leading to the modulation of the directed mutagenesis was performed using Altered Sites II in vitro mutagenesis systems (Promega). First, a 2134 bp XhoI cytoskeletal architecture of the epithelial cells or fragment of murine ephA8 cDNA containing the codons for possibly neuronal cells. Fyn in the brain is localized both Tyr-615 and Tyr-838 was isolated and subcloned into in the developing axonal tracts as well as subpopula- the SalI site of pAlter-I vector (Promega). To introduce a tions of adult neurons and glia (Ingraham et al., 1992; point mutation at the Tyr-615 codon, a 20 bp oligonucleotide Bare et al., 1993; Yagi et al., 1993a, 1994). Additional (nucleotides 1833 ± 1852 of the ephA8 coding sequence) studies on fyn knock-out mice reveals defects in long incorporating TTT (Phe) instead of TAT (Tyr-615) was Modulation of EphA8-Fyn interactions SChoiandSPark 5421 used together with amp repair and tet knock-out oligonucleo- room temperature, and the whole reaction was loaded on the tides, according to the manual. The point mutation at the gel. The labeled proteins were resolved by SDS ± PAGE, and Tyr-838 codon was generated using a 20 bp oligonucleotide dried gels were autoradiographed for 30 min at room (nucleotides 2503 ± 2522 of the ephA8 coding sequence) temperature. incorporating TTT (Phe) instead of TAT (Tyr-838). The For phosphopeptide mapping analysis, labeled proteins resulting inserts containing Phe-615 or Phe-838 codon were separated by SDS ± PAGE were transferred to Immobilon-P digested with NheI and EcoRI, and the isolated 1964-bp membranes and then localized by autoradiography. Immobi- fragments were subcloned into the corresponding region of lized proteins were directly digested with TPCK-trypsin full-length ephA8 in pcDNA3 vector. To generate the full- (20 mg) by incubating membrane pieces in 50 mM F615,F838 length EphA8 construct, a 1976 bp EcoRI and BamHI NH4HCO3, pH 7.5 at 378C for 4 h. Tryptic peptides were fragment containing the Phe-615 codon was subcloned into then analysed by electrophoresis in pH 1.9 bu€er (®rst the EcoRI site of pcDNA3 together with a 1269 bp BamHI dimension) and chromatography in regular bu€er (second and EcoRI fragment containing the Phe-838 codon. dimension) on TLC plates as described (Boyle et al., 1991). To elute the phosphate-containing peptides from TLC plates, the areas on TLC plates were aspirated using 1-ml blue tips Immunoprecipitation and immunoblotting plugged by polyethylene matrix and eluted in pH 1.9 bu€er. Con¯uent 10 cm plates of cells were washed two times with cold phosphate-bu€ered saline containing 1 m sodium M Cell attachment assays orthovanadate and then lysed in 1 ml of cold PLC lysis bu€er (50 mM HEPES, pH 7.5, 150 mM NaCI, 10% glycerol, Cells were collected by brief trypsinization, washed once with

1% Triton X-100, 1.5 mM MgCI2,1mM EGTA, 10 mM the complete medium (alpha-MEM containing 10% heat- sodium pyrophosphate, 100 mM sodium ¯uoride, 20 mg/ml inactivated fetal bovine serum) and then plated at 56105 cells Aprotinin, 1 mM PMSF, and 1 mM sodium orthovanadate). per 35-mm plate without coating. After 60 min, unattached Lysates were clari®ed by centrifugation in a microcentrifuge cells were collected and combined with the remaining for 15 min at 48C, and then incubated with the indicated unattached cells recovered by washing once with phosphate- antibodies for 1 h on ice. Protein A-Sepharose (Pharmacia) bu€ered saline containing calcium and magnesium. The was then added for 30 min, and immunoprecipitates were attached cells were collected by brief trypsinization after an washed three times by pelleting in a cold HNTG bu€er additional 60 min and the viable cells were counted. The ratio (20 mM HEPES, pH 7.5, 150 mM NaCl, 0.1% Triton X-100, of attached to total number of cells recovered was calculated. 10% glycerol, and 1 mM sodium orthovanadate). For GST- mixing experiments, 5 mg of fusion proteins were immobilized Antibodies on glutathione-Sepharose beads and then mixed with cell lysates for 1 h at 48C. Then, proteins bound to the beads Polyclonal rabbit antisera speci®c for the juxtamembrane were washed three times by pelleting in cold HNTG bu€er. domain of EphA8 were raised by injection of rabbits with the Immunoprecipitates were boiled in SDS-sample bu€er, GST fusion proteins puri®ed from E. coli culture. To loaded on 7.5% or 10% SDS-polyacrylamide gels and generate a GST-juxtamembrane fusion protein expression separated by electrophoresis. Proteins resolved by SDS ± vector, a 233 bp PCR ampli®ed DNA fragment, correspond- PAGE were electrophoretically transferred to Immobilon-P ing to the nucleotides 1684 ± 1916 of the ephA8 coding (Millipore) membranes. Membranes were blocked in TN-TX sequence, was subcloned into the GST-fusion vector pGEX- bu€er (50 mM Tris-HCl, pH 7.5, 200 mM NaCl, 0.2% Triton 5X-1 (Pharmacia). Monoclonal anti-phosphotyrosine anti- X-100) containing 3% BSA (for anti-phosphotyrosine anti- bodies (4G10) and polyclonal rabbit anti-Fyn antibodies body blots) or skimmed milk and immunoblotted according (raised against amino acid 35 ± 51 of human Fyn) were to standard protocols. Detection of primary antibodies was purchased from Upstate Biotechnology (Lake Placid, NY, performed using horseradish peroxidase-conjugated donkey USA). Polyclonal rabbit anti-HA antibodies were obtained anti-rabbit IgG or anti-mouse IgG followed by Enhanced from either Zymed Laboratories Inc. (San Francisco, CA, Chemiluminescence (Amersham). USA) or Santa Cruz Biotechnology (Santa Cruz, CA, USA).

In vitro labeling and phosphopeptide mapping analysis For in vitro phosphorylation of EphA8, immunoprecipitates described above were washed two times in cold HNTG bu€er Acknowledgments and two times in kinase reaction bu€er (20 mM HEPES, pH WeareindebtedtoDrJonasFrisenandDianaClarkefor

7.5, 2.5 mM MgCl2,4mM MnCl2,1mM sodium orthovana- their valuable discussions, Jaemin Jeong for his excellent date). The washed beads were incubated in 50 ml of kinase technical assistance, and Dr Dongeun Park for human Fyn reaction bu€er with 5 mCi of [g-32P]ATP (NEN) for 30 min at plasmids. The authors wish to acknowledge the ®nancial room temperature. Unincorporated ATP was removed by support of the Korea Research Foundation made in the washing 2 ± 3 times in cold HNTG bu€er. For the enolase Program Year 1997. This research was also supported by assay, 5 mg of acid-denatured enolase (Cooper et al., 1984) the Hallym Academy of Sciences, Hallym University, was added to the kinase reaction, incubated for 30 min at Korea.

References

Bare DJ, Lauder JM, Wilkie MB and Maness PF. (1993). Cooper JA, Esch FS, Taylor SS and Hunter T. (1984). J. Oncogene, 8, 1429 ± 1436. Biol. Chem., 259, 7835 ± 7841. Boyle WJ, van der Geer P and Hunter T. (1991). Methods Davis S, Gale NW, Aldrich TH, Maisonpierre PC, Lhotak V, Enzymol., 201, 110 ± 149. Pawson T, Goldfarb M and Yancopoulos GD. (1994). Brambilla R, Schnapp A, Casagranda F, Labrador JP, Science, 266, 816 ± 819. Bergemann AD, Flanagan JG, Pasquale EB and Klein R. Drescher U, Kremoser C, Handwerker C, Loschinger J, (1995). EMBO J., 14, 3116 ± 3126. Noda M and Bonhoe€er F. (1995). Cell, 82, 359 ± 370. Cheng HJ, Nakamoto M, Bergemann AD and Flanagan JG. Ellis C, Kasmi F, Ganju P, Walls E, Panayotou G and Reith (1995). Cell, 82, 371 ± 381. AD. (1996). Oncogene, 12, 1727 ± 1736. Modulation of EphA8-Fyn interactions SChoiandSPark 5422 Eph Nomenclature Committee. (1997). Cell, 90, 403 ± 404. Orioli D and Klein R. (1997). Trends Genet., 13, 354 ± 359. Frisen J, Yates PA, McLaughlin T, Friedman GC, O'Leary Park S and Sanchez MP. (1997). Oncogene, 14, 533 ± 542. DD and Barbacid M. (1998). Neuron, 20, 235 ± 243. Park S, Frisen J and Barbacid M. (1997). EMBO J., 16, Gao PP, Zhang JH, Yokoyama M, Racey B, Dreyfus CF, 3106 ± 3114. Black IB and Zhou R. (1996). Proc. Natl. Acad. Sci. USA, Pandey A, Lazar DF, Salteil AR and Dixit VM. (1994). J. 93, 11161 ± 11166. Biol. Chem., 269, 30154 ± 30157. GaoPP,YueY,ZhangJH,CerrettiDP,LevittPandZhou Pandey A, Duan H and Dixit VM. (1995a). J. Biol. Chem., R. (1998). Proc.Natl.Acad.Sci.USA,95, 5329 ± 5334. 270, 19201 ± 19204. Graham FL and van der Eb AJ. (1973). Virology, 54, 536 ± Pandey A, Shao H, Marks RM, Polverini PJ and Dixit VM. 539. (1995b). Science, 268, 567 ± 569. Grant SG, O'Dell TJ, Karl KA, Stein PL, Soriano P and Schlessinger J. (1997). Cell, 91, 869 ± 872. Kandel ER. (1992). Science, 258, 1903 ± 1910. Stein E, Cerretti DP and Daniel TO. (1996). J. Biol. Chem., Henkemeyer M, Orioli D, Henderson JT, Saxton TM, Roder 271, 23588 ± 23593. J, Pawson T and Klein R. (1996). Cell, 86, 35 ± 46. Stein E, Lane AA, Cerretti DP, Schoecklmann HO, Schro€ Hock B, Bohme B, Karn T, Feller S, Rubsamen-Waigmann AD, Van Etten RL and Daniel TO. (1998a). Dev., H and Strebhardt K. (1998). Oncogene, 16, 255 ± 260. 12, 667 ± 678. Holland SJ, Gale NW, Gish GD, Roth RA, Songyang Z, Stein E, Huynh-Do U, Lane AA, Cerretti DP and Daniel TO. Cantley LC, Henkemeyer M, Yancopoulos GD and (1998b). J. Biol. Chem., 273, 1303 ± 1308. Pawson T. (1997). EMBO J., 16, 3877 ± 3888. Twamley-Stein GM, Pepperkok R, Ansorge W and Ingraham CA, Cooke MP, Chuang YN, Perimutter RM and Courtneidge SA. (1993). Proc. Natl. Acad. Sci. USA, 90, Maness PF. (1992). Oncogene, 7, 95 ± 100. 7696 ± 7700. Krull CE, Lansford R, Gale NW, Collazo A, Marcelle C, UmemoriH,SatoS,YagiT,AizawaSandYamamotoT. Yancopoulos GD, Fraser SE and Bronner-Fraser M. (1994). Nature, 367, 572 ± 576. (1997). Curr. Biol., 7, 571 ± 580. van der Geer P, Hunter T and Lindberg RA. (1994). Annu. Lackmann M, Oates AC, Dottori M, Smith FM, Do C, Rev. Cell Biol., 10, 251 ± 337. Power M, Kravets L and Boyd AW. (1998). J. Biol. Chem., Wang HU and Anderson DJ. (1997). Neuron, 18, 383 ± 396. 273, 20228 ± 20237. Wang HU, Chen ZF and Anderson DJ. (1998). Cell, 93, Lhotak V and Pawson T. (1993). Mol. Cell. Biol., 13, 7071 ± 741 ± 753. 7079. Winning RS, Scales JB and Sargent TD. (1996). Dev. Biol., MeimaL,KljavinIJ,MoranP,ShihA,WinslowJWand 179, 309 ± 319. Caras IW. (1997). Eur. J. Neurosci., 9, 177 ± 188. Yagi T, Shigetani Y, Okado N, Tokunaga T, Ikawa Y and Monschau B, Kremoser C, Ohta K, Tanaka H, Kaneko T, Aizawa S. (1993a). Oncogene, 8, 3343 ± 3351. Yamada T, Handwerker C, Hornberger MR, Loschinger Yagi T, Aizawa S, Tokunaga T, Shigetani Y, Takeda N and J, Pasquale EB, Siever DA, Verderame MF, Muller BK, Ikawa Y. (1993b). Nature, 366, 742 ± 745. Bonhoe€er F and Drescher U. (1997). EMBO J., 16, YagiT,ShigetaniY,FurutaY,NadaS,OkadoN,IkawaY 1258 ± 1267. and Aizawa S. (1994). Oncogene, 9, 2433 ± 2440. Nakamoto M, Cheng HJ, Friedman GC, McLaughlin T, Zisch AH, Kalo MS, Chong LD and Pasquale EB. (1998). Hansen MJ, Yoon CH, O'Leary DD and Flanagan JG. Oncogene, 16, 2657 ± 2670. (1996). Cell, 86, 755 ± 766. Orioli D, Henkemeyer M, Lemke G, Klein R and Pawson T. (1996). EMBO J., 15, 6035 ± 6049.