Ligand Activation of ELK Receptor Tyrosine Kinase Promotes Its Association with Grb10 and Grb2 in Vascular Endothelial Cells*

ELK is a member of the Eph-related tyrosine kinase family that includes receptors signaling axonal guid-ance, neuronal bundling, and angiogenesis. We recently identified ELK expression in human renal microvascular endothelial cells and sought to identify intracellular proteins through which it signals responses. The cytoplasmic domain of ELK was used as “bait” in a yeast two-hybrid screen to identify interactive proteins expressed from a randomly primed embryonic murine li- brary (E9.5–10.5). Among interactive products of 76 cDNAs characterized, 10 nonidentical, overlapping clones encoded the SH2 domain of the recently reported Grb10 adapter protein, and an additional 3 encoded Grb2. A self-phosphorylated recombinant, baculovirus-expressed GST-ELK cy fusion protein bound Grb10 and Grb2 from human renal microvascular endothelial cell extracts, while the unphosphorylated fusion form did not. Site-directed mutation identified Tyr-929 as a putative phosphorylation site required for Grb10, but not Grb2, interaction in yeast and recombinant protein assays. The ELK ligand, LERK-2/Fc, stimulated tyrosine phosphorylation of ELK, and recruitment of Grb10 and Grb2 to endothelial ELK receptors recovered by wheat germ agglutinin lectin and immunoprecipitation. These findings define ligand-activated interaction between ELK and the SH2 domains of Grb2 and the newly identified Grb10 protein that shares homology with a Cae- norhabditis elegans

Members of the Eph-related receptor tyrosine kinase family have been assigned important roles in signaling axonal guidance, neuronal bundling, and, more recently, angiogenesis (1)(2)(3)(4). There are now at least 13 distinct family members that are expressed in species ranging from Xenopus laevis to humans, in cell-type and tissue-restricted patterns that suggest that they signal targeting or differentiation (5,6). This idea was supported by recent evidence that the Mek4 receptor targets retinal neurons to their posterior tectal projections in response to its ligand, AL-1 (RAGS) (1,2,4).
Other data have shown that Eph family receptors are important in vascular biology. The endothelial cell product, B61, is the ligand for the Eck receptor and mediates tumor necrosis factor ␣-induced angiogenesis (3). We recently identified expression of a second Eph family receptor, ELK, in vascular endothelial cells. Originally cloned from a rat brain cDNA library, ELK is also expressed in testis, fetal kidney, and skeletal muscle (7). It interacts with two membrane-bound ligands, LERK-2 and LERK-5, that are expressed in a wide range of tissues (8 -11). LERK-2 stimulates assembly of microvascular endothelial cells into capillary-like cords and tubes, through effects mediated by ELK. 1 Available data show that natively expressed Eph-related receptors, including ELK, Mek4, and Eck, do not signal proliferative responses upon ligand binding (3,12). 2 Because it mediates nonproliferative signals in microvascular endothelial cells, we were motivated to identify the intracellular interactive proteins through which ELK signals, anticipating that these ELK-interactive proteins may discriminate proliferative from nonproliferative responses.
To this end, we utilized a yeast two-hybrid screen to identify protein partners that interact with the ELK cytoplasmic domain (ELK cy ). 3 We further characterized determinants of those interactions using native and truncated forms of recombinant ELK cy proteins to address requirements for specific domains and tyrosine kinase function. We present evidence for ELK interactions with two distinct members of the Grb family, Grb10 and Grb2, in microvascular endothelial cells.

Construction of Recombinant Fusion Proteins and Baits-Fusion
plasmids were constructed to permit shuttling of ELK-encoding inserts from the baculovirus expression vector pAC-GST to the yeast twohybrid "bait" LexA fusion plasmid pBTM116. Parent sequences were derived from the predominant human ELK cDNA recovered from a human renal microvascular endothelial cell library (HRMEC) (HuELKI), and amino acid designations refer to that sequence. 1 The pAC-GST/ELK cy plasmid was constructed by PCR/ligation-independent cloning (LIC) (13,14) to place a HuELKI cDNA fragment encoding the complete cytoplasmic domain (amino acids 566 -984) in frame and downstream of GST-encoding sequences (PharMingen, San Diego, CA). To permit insert "shuttling," an in-frame 5Ј SmaI site and a customized SalI site 3Ј of translation termination codons were placed in positions flanking ELK cytoplasmic domain coding sequences by PCR amplification, using oligonucleotides containing the PCR/ligation-independent cloning sequence, a SmaI site, and HuELKI cDNA sequences (underlined) representing ELK cy sequences (oligonucleotide 1, 5Ј primer: (5Ј-CTG GTT CCG GCG ATC CCG GGG AGG AAA CGG GCT TAT  AGC-3Ј) and oligonucleotide 2, 3Јprimer: (5Ј-CTC GCT CCG GCG AGG  TCG ACG TCA TGC CAT TGC CGT TGG-3Ј)). PCR product was gel purified and ligated into pAC-GST-LIC (PharMingen). Constructs encoding regional domain deletions of the ELK cytoplasmic domain as GST fusions were generated in a similar manner and include: 1) the catalytic domain construct, pAC-GST/ELK cy ⌬JM/Cterm (HuELKI amino acids 618 -883); 2) pAC-GST-ELK cy ⌬Cterm (amino acids 566 -883); and 3) pAC-GST-ELK cy ⌬JM (amino acids 618 -984).
The LexA-ELK cy bait plasmids were constructed by digesting the respective pAC-GST-ELKcy plasmids with SmaI and SalI, recovering ELK cy fragments, and subcloning them into SmaI-SalI-digested pBTM116 vector (15). The resulting plasmids drive expression of the indicated cytoplasmic domains of HuELKI as LexA fusion proteins, under control of the Saccharomyces cerevisiae alcohol dehydrogenase promotor.
Site-directed Mutagenesis-Overlap extension PCR was used to generate a kinase-inactivating ELK cy mutation (K652R) by the method of Ho et al. (16). Overlapping oligonucleotides encoding the mutation (5Ј-CTT CAG GGT CCT GAT GGC CAC GTA-3Ј and 5Ј-GTG GCC ATC AGG ACC GTG AAG GCA-3Ј) were used to generate 5Ј-and 3Ј-overlapping PCR products that were reannealed, PCR amplified, and digested with BglII and Bsu36I. The recovered fragment was cloned into the BglII-Bsu36Idigested pAC-GST-ELK cy to generate pAC-GST/ELK cy (K652R).
pAC-GST-ELK cy Y929F was generated as described above, using oligonucleotide 1 (above) and a specific 3Ј-primer encoding the mutation (5Ј-GAA GCT GTT CCT GAA CTG GAC CAT TTT-3Ј) to generate a PCR product overlapping that generated using the 5Ј-primer (5Ј-AAA ATG GTC CAG TTC AGG AAC AGC TTC-3Ј) and oligonucleotide 2 (see above). Overlap-extended PCR-amplified products were digested with EspI-SalI and subcloned into EspI-SalI-digested pAC-GST-ELK cy . Products were confirmed by sequence analysis, and the SmaI-SalI fragments of ELK cy K652R and ELK cy Y929F were substituted for the SmaI-SalI fragment of ELK cy pBTM116-ELK cy to generate pBTM116-ELK cy K652R and pBTM116-ELK cy Y929F.
Two-hybrid Screen-The yeast reporter strain L40, containing the reporter genes lacZ and HIS3 downstream of a LexA promoter, was sequentially transformed with the pLexA-ELK cy (or derivative) plasmids, then with a cDNA library encoding VP16 (transcription activation domain) fusions with peptide sequences expressed in murine embryos (E9.5-10.5), using the lithium acetate method (19,20). The library used was size selected (350 -700 nucleotides) and random-primed to permit unweighted representation of NH 2 -and COOH-terminal domains and to permit recovery of independent clones of overlapping, but nonidentical sequences as independent confirmation of significant interactions. An estimated 2 ϫ 10 7 yeast transformants were grown for 4 h at 30°C in synthetic medium lacking leucine and tryptophan to maintain selection for the ELK cy bait and the cDNA library plasmids, respectively; transformants were then plated on synthetic medium lacking histidine, leucine, tryptophan, uracil, and lysine; supplemented with 1.5 mM 3-aminotriazole; and incubated at 30°C for 3 days (18).
Selected yeast clones (HIS ϩ ) were assayed for ␤-galactosidase activity (lacZ ϩ ), and yeast DNA was prepared from the His ϩ /lacZ ϩ colonies. Putative ELK-interactive protein (EIP) "prey" cDNAs were recovered by electroporation of yeast DNA into competent HB101 Escherichia coli (21,22) and tested to exclude nonspecific transactivation independent of the pBTM116-ELK cy bait by cotransfection with pBTM116 (no insert) and with pLexA-Lamin (15). Putative EIP cDNAs surviving this exclusion were sequenced, and the identity of each cDNA sequence was ascertained using the BLAST search engine (23).
Recombinant Baculovirus Protein Production and Affinity Binding Assays-Individual recombinant Baculovirus stocks were generated by cotransfecting a monolayer of Sf9 cells with a recombinant pAC-GST-ELK fusion construct and BaculoGold Baculovirus DNA (PharMingen) according the manufacture's protocol (24). Five days after transfection, supernatants were collected and subjected to two rounds of amplifica-tion (P1 lysate). For affinity binding assays, a monolayer of Sf9 cells was infected with the P1 lysate, and 3 days later, cells were lysed in triple detergent lysis buffer (50 mM Tris-Cl, pH 8.0, 150 mM NaCl, 0.02% sodium azide, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% Tween 20, 1 mM PMSF, 2 mg/ml aprotonin) for 30 min on ice. Lysates were cleared by centrifugation; the precleared lysates were incubated for 2 h at 4°C with glutathione-agarose beads (Sigma) washed twice with lysis buffer, followed by two washes with in vitro kinase buffer (20 mM HEPES, 20 mM MnCl 2 , 150 mM NaCl, 0.1% Triton X-100, 1 mM PMSF) at 4°C. As indicated in the legend to Fig. 3, glutathione-agarose beads were incubated in the presence or absence of 30 M ATP for 30 min at 30°C (25).
Confluent monolayers of HRMECs (ϳ5 ϫ 10 6 cells/P100) were washed twice with phosphate-buffered saline, 1 mM PMSF and then incubated for 20 min on ice with 3 ml of buffer D (50 mM Tris-Cl pH 8.0, 300 mM NaCl, 0.1% Triton X-100, 1 mM PMSF). Lysates were cleared of particulate by centrifugation at 4°C and incubated overnight at 4°C with ϳ50 g of glutathione-agarose-immobilized (phosphorylated or unphosphorylated) GST/ELK cy , GST/ELK cy ⌬JM, or GST/ELK cy Y929F fusion proteins, as indicated in the legend to Fig. 3. Immobilized GST/ fusion protein complexes were washed three times with buffer D, omitting bovine albumin from the last wash. Bound proteins were resolved on 15% SDS-polyacrylamide gel electrophoresis gels and detected by Western blot analysis using the ECL system (Amersham, Arlington Heights, IL).
Copurification of Ligand-stimulated ELK with Grb10 and Grb2-Plates of 90 -100% confluent HRMECs were washed and then incubated in OPTI-MEM medium (Life Technologies, Inc.), supplemented with 0.5 mM sodium suramin (Calbiochem) for 16 h to dissociate any endogenous ligand from ELK receptors. For 60 min prior to stimulation, 10 M sodium o-vanadate was added to the medium. Cells were washed and then exposed for 10 min at 37°C to 500 ng/ml of highly purified Fc fusion forms of either the best characterized ligand for ELK receptors, LERK-2/Fc (10), or an unrelated open reading frame protein, ORF/Fc, both complexed with 50 ng/ml IgG (anti-Fc, The Binding Site, Birmingham, United Kingdom) as indicated in the legend to Fig. 4. Cells were lysed in buffer WG (Triticum vulgaris-lectin purification), containing 50 mM HEPES, pH 7.4, 50 mM NaCl, 5 mM EDTA, 1 mg/ml bovine serum albumin, 1 mM sodium o-vanadate, 1% Triton X-100, and 1 mM PMSF; lysates were clarified by centrifugation and incubated at 4°C with agarose-conjugated lectin from T. vulgaris (Sigma). Agarose beads were washed repeatedly, and specifically adsorbed proteins were eluted in buffer WG containing 3 mM NЈ,NЉ,Nٞ-triacetylchitotriose (Sigma), as described.

RESULTS
Interaction Screening of an Embryonic Mouse Library with the Cytoplasmic Domain of ELK (ELK cy )-To identify EIPs expressed in an embryonic (E9.5-10.5) murine library (15), we utilized a yeast two-hybrid system. The yeast reporter strain L40 was initially transformed with the pLexA-ELK cy sequences encoding the entire ELK cytoplasmic domain (ELK cybait) protein fused to the LexA DNA-binding domain (LexBD). We determined that this bait did not transactivate and express the HIS3 gene product in the presence of 1.5 mM 3-aminotriazole (data not shown). On the basis of our previous observation that ELK is expressed in E9.5 murine embyros, 2 we screened a murine embryonic library (E9.5-10.5) for EIPs, anticipating that developmental expression of EIPs would coincide with ELK expression.
Yeast from a single transfected colony carrying the bait were selected on Trp Ϫ plates, made competent, and then cotrans-fected with the E9.5-10.5 embryonic murine cDNA library encoding potential prey products as fusion proteins with the VP16 transcription activation domain (Fig. 1). Formation of a complex between LexA-ELK cy and VP16-EIPs should confer histidine auxotrophy and ␤-galactosidase activity. Among ϳ2 ϫ 10 7 yeast transformants screened, 320 colonies grew on His Ϫ plates; among these, 290 displayed ␤-galactosidase activity. Plasmid DNA was prepared from 80 of the colonies surviving both selections, and of these, 4 were shown to transactivate as cotransfectants with either pLexA-Lamin fusion protein or pBTM116 alone. The remaining 76 were subjected to sequence analysis.
Among these 76 potential EIPs, multiple distinct cDNAs encoding at least 19 different previously described proteins and 3 novel proteins were identified. 4 Among the most frequently represented cDNA sequences recovered in this screen were those encoding SH2 domains of Grb10 (10 independent clones) and Grb2 (3 independent clones). Fig. 1B depicts the comparative positions of the peptides encoded by the partial cDNA sequences compared with the parent proteins.
ELK Self-phosphorylation Is Required for Its Interaction with Grb10 and Grb2-To determine whether the ELK cy interactions with Grb10 and Grb2 require ELK self-phosphoryl-ation, we inactivated the tyrosine kinase of the ELK cy fusion proteins by site-directed mutagenesis of the ATP-binding lysine (K652R). Fig. 2 shows that the kinase-inactive form of the ELK cy bait does not support interaction with Grb10 or Grb2 SH2 domain interactors. This finding implies that tyrosine phosphorylation of the ELK cy domain plays a critical role in its interaction with Grb10 and Grb2, as is predicted by the phosphotyrosine-binding function of SH2 domains (28).
To identify protein domains through which ELK cy interacts with Grb10 and Grb2, we examined their interactions with truncations of the original ELK cy bait that deleted the juxtamembrane domain (pLexA-ELK cy ⌬JM), the COOH-terminal domain (pLexA-ELK cy ⌬Cterm), or both (pLexA-ELK cy ⌬JM/ Cterm), as described under "Experimental Procedures." The catalytic domain was retained in each of these constructs to facilitate the tyrosine self-phosphorylation required to provide phosphotyrosine-containing SH2 binding sites. We found that while Grb2 retained interaction with the isolated ELK cy tyrosine kinase catalytic domain (pLexA-ELK cy ⌬JM/Cterm) (not shown here), Grb10 interaction required that ELK cy constructs retain the COOH-terminal domain (amino acids 883-984). This finding indicates that the sites of ELK interaction with Grb10 and Grb2 are distinct.
Because data were not available to define SH2 binding consensus sites for Grb10, we proceeded with site-directed muta-4 E. Stein and T. O. Daniel, manuscript in preparation.

FIG. 1.
Two-hybrid interaction screen for EIPs expressed in a murine embryonic cDNA library. A, fusion proteins encoded by the ELK cy bait and murine cDNA prey plasmids, indicating the numbers of colonies selected by the sequential process at each step and the fraction of putative EIP clones characterized. B, regions of overlap between individual EIP cDNA clones recovered and their respective Grb10 and Grb2 proteins. VP16AD, VP16 transcription activation domain.
tion of the single tyrosine residue in this COOH-terminal domain (Tyr-929). This single substitution disrupted the twohybrid interaction of ELK cy with the Grb10 SH2 domain represented in EIP 23 ( Fig. 1B) (data not shown), providing evidence that phosphorylated Tyr-929 is part of the binding domain with which Grb10 interacts.
Ligand-activated ELK Associates with Grb10 and Grb2-We used an in vitro affinity binding assay to verify that ELK cy is capable of binding endogenous endothelial Grb10 and Grb2 proteins, depending upon its state of tyrosine phosphorylation. GST fusion forms of ELK cy , GST/ELK cy , GST/ELK cy ⌬JM, and GST/ELK cy Y929F were generated in an insect expression system (see "Experimental Procedures"), bound to GSH-agarose beads, and incubated in kinase buffer in the presence (or absence) of ATP to determine if self-phosphorylation was required for binding. Incubation with ATP was required for tyrosine phosphorylation of recombinant, baculovirus-expressed proteins under these conditions (data not shown). Each of the GST/ELK cy fusion proteins were then incubated with HRMEC (endothelial) lysates, and washed extensively in high salt (300 mM NaCl) buffer.
In Fig. 3, Western analysis of proteins adsorbed to each immobilized ELK cy fusion protein evaluated binding of endogenous Grb10 and Grb2. Grb2 bound each of the tyrosine selfphosphorylated cytoplasmic domain ELK fusions but did not bind the unphosphorylated fusions. While Grb10 bound phosphorylated fusions including the full cytoplasmic domain (ELK cy ) and one lacking the juxtamembrane domain (ELK cy ⌬JM), it failed to bind the fusion containing the point mutation, Y929F (Fig. 3). This finding defines the site through which Grb10 interacts with ELK (Tyr-929) as distinct from that mediating ELK-Grb2 association. These data do not exclude the possibility that these interactions involve participation of other proteins that bridge between ELK and Grb2 or Grb10. However, that bridging function must be conferred by constitutively expressed yeast proteins to support the two hybrid interactions that we identified.
While these data show that in vitro phosphorylated ELK has the capacity to interact with endothelial Grb10 and Grb2, biologically significant interactions should be promoted by ligand activation of endogenous ELK receptors in intact cells. To explore this issue, we exposed HRMEC to the ELK ligand, LERK-2/Fc, or an irrelevant open reading frame Fc fusion, ORF/Fc, and recovered ELK receptors by lectin affinity chromatography (Fig. 4A) or immunoprecipitation (Fig. 4B). As shown in Fig.  4A, LERK-2/Fc-stimulated tyrosine phosphorylation of a wheat germ agglutinin-lectin-recovered protein that comigrated with ELK receptors identified by Western analysis (Fig. 4A, upper panels); increased amounts of Grb2 and Grb10 immunoreactive proteins were recovered in the wheat germ agglutinin-agarose eluate from LERK-2/Fc-stimulated cells, compared with cells exposed ORF/Fc at the same concentration.
To confirm that ELK receptors form coprecipitable complexes with Grb2 and Grb10 following ligand activation, we immunoprecipitated ELK receptors from cells exposed to either ORF/Fc or the ELK ligand, LERK-2/Fc, and analyzed immunoprecipitates for coprecipitating Grb2 and Grb10. In Fig. 4B (left), ELK immunoprecipitates recovered immunoreactive Grb2 (23 kDa) following stimulation of cells with LERK-2/Fc, but not ORF/Fc. The prominent, nonspecific 25-kDa band in each lane represents the immunoglobin light chain of the precipitating anti-ELK or preimmune antibodies. In parallel (Fig.  4B, right), ELK antisera coprecipitated Grb10-immunoreactive material from LERK-2/Fc-, but not ORF/Fc-, stimulated cells. Again, the common, 50-kDa band at the bottom of the figure represents precipitating immunoglobin heavy chain. The multiple forms of immunoreactive Grb10 protein, ranging in size from 65 to 80 kDa, were recovered in ELK immunoprecipitates from LERK-2/Fc-, but not ORF/Fc-, stimulated cells. These multiple Grb10 forms represent proteins generated in transfected cells through use of alternative translation start sites and modified by phosphorylation (27).

FIG. 2.
A tyrosine kinase-active form of ELK cy is required for interaction with Grb10 and Grb2. The reporter strain L40 was cotransformed with the indicated plasmids, plated on medium lacking tryptophan, leucine, lysine, and uracil, and grown for 3 days at 30°C. Growth of the same yeast cotransfectants was compared on medium containing (ϩ) or lacking (Ϫ) histidine (H). Plates were photographed after 3 days at 30°C.

FIG. 3. Binding of recombinant GST-ELK cy fusion proteins to Grb10 and Grb2 requires tyrosine self-phosphorylation.
Recombinant GST-ELK cy , GST-ELK cy ⌬JM, and GST-ELK cy Y929F fusion proteins were purified by adsorption to GSH-agarose beads and repeated washing and then incubated for 30 min in kinase buffer containing (ϩ) or lacking (Ϫ) ATP, as described under "Experimental Procedures." Immobilized, adsorbed phosphorylated, or unphosphorylated, GST-ELK cy fusion proteins were incubated with HRMEC extracts, washed extensively, and analyzed by Western blot, using monospecific antibodies against GST, Grb10, or Grb2, as indicated. Bound antibodies were detected using the ECL chemiluminescent system.

DISCUSSION
Our findings identify two previously described proteins, Grb2 and Grb10, as interactive partners that bind to ligandactivated, self-phosphorylated cytoplasmic domain sequences of ELK. This screen yielded multiple independent, but overlapping, clones that encode the SH2 domains of both Grb10 and Grb2; in aggregate, these clones account for some 20% of recovered sequences. We surmise this reflects some preference for high affinity ELK cy interaction with these particular SH2 domain-containing proteins, among many expressed in this library. This prediction was supported by independent demonstration that recombinant GST/ELK cy binds Grb10 and Grb2 in crude endothelial lysates (Fig. 3).
Further, we showed that ELK ligand stimulates ELK tyrosine phosphorylation and recruitment of Grb10 and Grb2 to copurify with ELK recovered by either a wheat germ agglutinin-lectin-binding step or immunoprecipitation (Fig. 4). Interactive properties of ELK cy subdomain deletions were evaluated in both two-hybrid (data not shown) and direct binding assays (Fig. 3). Although both Grb10 and Grb2 required that ELK kinase catalyze phosphorylation of intrinsic tyrosine residues, the Grb10 interaction required an intact COOH-terminal domain that is not required for the Grb2 interaction. Site-directed mutagenesis of Y929, the only tyrosine residue in this domain, eliminated both the two hybrid interaction between ELK cy and Grb10 (not shown), and binding of endothelial Grb10 to the GST fusion protein, GST/ELK cy Y929F (Fig. 3). In aggregate, these data provide strong support for ligand-coupled assembly of complexes that associate tyrosine-phosphorylated ELK with Grb10 and Grb2 at independent sites of interaction.
The cytoplasmic domain of ELK includes 17 tyrosine residues, 14 of which are represented at analogous sites in other Eph family kinases. Moreover, ELK shares striking sequence identity with mammalian Eck, Erk, and Ehk, as representatives of other family members. Of the 428 ELK cytoplasmic amino acids, 202 are identical among these receptors, and 79 are conserved. To date, information about intracellular proteins that mediate responses to Eph family receptors has been limited to data on Eck interactive proteins (29,30). Pandey et al. (29) identified Eck interaction with phosphatidylinositol 3-kinase (p85 subunit) in a yeast two-hybrid screen of a T lymphocyte library and confirmed that ligand activation of the Eck receptor activates PI3K in intact endothelial cells. More recently, they identified a novel c-src homologous protein they named SLAP as an interactive protein that binds Eck on ligand activation, similar to the PI3K findings (30).
Among the 76 sequenced EIP clones that we have evaluated to date, we did not recover either phosphatidylinositol 3-kinase or SLAP. This may reflect differences in expression of these and other transcripts in the two cDNA libraries that were screened; however, it appears more likely that sequence differences or tyrosine phosphorylation sites within cytoplasmic domain sequences of ELK and Eck define distinct SH2 recognition domains. ELK lacks a consensus sequence for binding the phosphatidylinositol 3-kinase p85 subunit (Y-hydrophobic-Xhydrophobic), while Eck and Ehk both have these motifs (28).
The Grb2 protein has been assigned important roles in signaling cell proliferation through the p21-RAS pathway as an adapter molecule binding activated platelet-derived growth factor, epidermal growth factor, and other receptors directly (31). However, Grb2 has also been recognized to participate in nerve growth factor-mediated differentiation of PC12 cells, where nerve growth factor receptor activation promotes assembly of SHC-Grb2-SOS complexes (32)(33)(34). We have ascertained that ELK receptor activation does not stimulate proliferation of the microvascular endothelial cells used in this study (data not shown). In this context, it appears that Grb2 binding to ELK is likely to signal nonproliferative downstream responses, such as cell-cell aggregation behavior.
Grb10 is a recently identified adapter protein family member with high level sequence similarity to Grb7, a known interacter with HER2 and epidermal growth factor receptors (27). Grb10 and Grb7 share COOH-terminal SH2 domains and a central domain of ϳ350 amino acids that is encoded by the C. elegans gene, F10E9.6 (27). That locus is mutated in mig-10 worms and is implicated in their defects in embryonic neural migration. 5 At present, it is attractive to speculate that Grb10, through its interaction with ELK, may participate similarly in neural cell targeting during embryonic mammalian central nervous system development, because ELK is highly expressed in human fetal brain (7). Grb10 participation in endothelial cell migration and targeting mediated through ELK activation is now accessible for study, using this cultured microvascular endothelial system.
Previous data have identified interactions between Grb10 5 J. Manser, unpublished results.
FIG. 4. ELK ligand, LERK-2/Fc, stimulates ELK tyrosine phosphorylation and copurification of Grb10 and Grb2 with ELK. HRMEC were treated as indicated under "Experimental Procedures," exposed for 10 min at 37°C to LERK-2/Fc or an unrelated Fc fusion protein, ORF/Fc, and then lysed and subjected to lectin affinity chromatography (A) or immunoprecipitation with ELK antibodies (B), as described. Lectin-adsorbed proteins were selectively eluted using 3 mM NЈ,NЉ,Nٞ-triacetylchitotriose and analyzed by Western blot for ELK, phosphotyrosine, Grb10, and Grb2, as indicated (A). Anti-ELK or preimmune immunoprecipitated proteins were analyzed by Western blot for Grb10 and Grb2, as indicated. and c-ret, a receptor tyrosine kinase involved in renal and enteric neuron development, thyroid papillary carcinomas, and multiple endocrine neoplasia (types 2A and 2B) (35). However, sufficient data are not yet available to identify the sequence preferences for Grb10 SH2 domains. Comparison of the primary amino acid sequences surrounding the putative Tyr-929 binding site in ELK to those surrounding cytoplasmic tyrosine residues of c-ret do not yet readily identify a consensus for Grb10 SH2 interaction (data not shown). Nevertheless, our data using ELK cy deletion and site-directed mutation constructs (Fig. 3) lead us to conclude that Tyr-929 is the site for phosphorylation-dependent binding of ELK to the Grb10 SH2 domain. Although tyrosine residues are conserved at that position among multiple Eph family receptors, including Eck, only 4 of 12 residues adjacent to Tyr-929 are shared betweeen ELK and Eck. Eck has been shown not to bind Grb10 under conditions where the c-ret kinase does (35). These limited findings suggest that, although receptors of the Eph family share striking cytoplasmic domain sequence identity, significant differences in coupled responses may be expected as individual receptors interact with distinct intracellular signaling partners.