Interaction of the Grb10 Adapter Protein with the Raf1 and MEK1 Kinases*

Grb10 and its close homologues Grb7 and Grb14, belong to a family of adapter proteins characterized by a proline-rich region, a central PH domain, and a carboxyl-terminal Src homology 2 (SH2) domain. Their interaction with a variety of activated tyrosine kinase receptors is well documented, but their actual function remains a mystery. The Grb10 SH2 domain was isolated from a two-hybrid screen using the MEK1 kinase as a bait. We show that this unusual SH2 domain interacts, in a phosphotyrosine-independent manner, with both the Raf1 and MEK1 kinases. Mutation of the MEK1 Thr-386 residue, which is phosphorylated by mitogen-activated protein kinase in vitro, reduces binding to Grb10 in a two-hybrid assay. Interaction of Grb10 with Raf1 is constitutive, while interaction between Grb10 and MEK1 needs insulin treatment of the cells and follows mitogen-activated protein kinase activation. Random mutagenesis of the SH2 domain demonstrated that the Arg-βB5 and Asp-EF2 residues are necessary for binding to the epidermal growth factor and insulin receptors as well as to the two kinases. In addition, we show that a mutation in Ser-βB7 affects binding only to the receptors, while a mutation in Thr-βC5 abrogates binding only to MEK1. Finally, transfection of Grb10 genes with specific mutations in their SH2 domains induces apoptosis in HTC-IR and COS-7 cells. These effects can be competed by co-expression of the wild type protein, suggesting that these mutants act by sequestering necessary signaling components.

The three members of the Grb7 family of signaling proteins, Grb7, Grb10 and Grb14, were first isolated through the ability of their SH2 1 domains to recognize phosphotyrosine-containing sequences on activated tyrosine kinase receptors. Grb7 has been shown to interact with the EGF, ErbB2/Her2, platelet-derived growth factor, and Fc⑀RI receptors, the Ret protooncogene, the Syp phosphatase, and the SHC adapter protein (1)(2)(3)(4)(5). Grb10 has a much reduced affinity for the EGF receptor, but its interactions with the Ret proto-oncogene, ELK receptor, insulin-like growth factor-I receptor, and insulin receptor (IR) have been demonstrated in vivo (6 -11). Grb14, the remaining member of this family, interacts in vitro with the plateletderived growth factor receptor (12).
The function of the Grb7/10/14 proteins remains obscure. They are currently classified as adapter proteins, since they have no obvious enzymatic function and contain several conserved polypeptide-binding regions in addition to their carboxyl-terminal SH2 domain (see Fig. 1). A small proline-rich region in the NH 2 terminus fits the consensus sequence for SH3binding domains. In fact, in vitro binding by the Grb10 prolinerich domain to the cAbl SH3 domain has been demonstrated (10). Furthermore, all three of these proteins contain a central Pleckstrin homology domain. Similar elements have been shown to mediate protein-protein and/or protein-lipid interactions (for a review, see Ref. 13). The Grb7/10/14 genes are expressed in a tissue-specific pattern, and there is ample evidence of alternative splicing of the transcripts (1, 6, 7, 10 -12, 14). Some of these splicing events result in the expression of proteins with altered properties; hGrb10␤ 2 contains a deletion in the amino-terminal half of its Pleckstrin homology domain (6). Finally, the Grb7/10/14 proteins can be phosphorylated in vivo, and their levels of phosphorylation are regulated by hormone treatment (2,4,7,12).
It is still unclear whether Grb10 acts as an inhibitor or an activator of signal transduction. Overexpression of hGrb10␤ in Chinese hamster ovary-IR cells reduces insulin-dependent pp60 and insulin receptor substrate-1 phosphorylation and diminishes PI-3 kinase activation (6). Others have observed that microinjection of the Grb10 SH2 domain partially inhibits mitogenesis in insulin-or insulin-like growth factor-I-treated Rat1 fibroblasts but not in cells treated with EGF or serum (14). These results appear to be in contradiction with the work of Morrione et al. (15), whose cell lines overexpressing mGrb10␣ show growth reduction in the presence of insulin growth factor-I but not insulin. Unlike the microinjection experiments, these cells do not show an inhibition in S phase entry but rather an accumulation in the S and G 2 phases. These contradictions might be explained by the use of different cell lines, experimental procedures, or Grb10 splicing variants. As for Grb7 and Grb14, both have been shown to be overexpressed in breast and prostate cancer tumors and cell lines (2,12,16).
The receptors that are recognized by the Grb7/10/14 family share the ability to activate the mitogenic MAP kinase signal transduction pathway (reviewed in Ref. 17). In this report, we present evidence that at least two members this pathway, Raf1 and MEK1, can interact with the SH2 domain of the Grb10 adapter protein. In some cultured cells, binding to Raf1 is constitutive, while the interaction of Grb10 with MEK1 is insulin-dependent. Interestingly, these interactions are phosphotyrosine-independent, and binding of MEK1 to Grb10 appears to be regulated through the retrophosphorylation of MEK1 by the MAP kinases. Finally, we have identified point mutations that affect the specificity of the Grb10 SH2 domain. Overexpression of these mutants was found to induce apoptosis in cultured cells.

EXPERIMENTAL PROCEDURES
Materials-Most of our chemicals were purchased from Sigma. Restriction and modification enzymes as well as the plasmids and resins used for the expression and purification of GST and MBP fusions proteins were obtained from New England Biolabs or from Amersham Pharmacia Biotech. Sequencing reactions were performed with a DNA sequencing kit from Perkin-Elmer and separated on an Applied Biosystems DNA Sequencer model 370A. PCRs used the Expand High Fidelity PCR System (Boehringer Mannheim). Western blots were performed on Immobilon membranes (Millipore Corp.). The Grb10 (K-20), and Raf1 (C-12) antibodies were from Santa Cruz Biotechnology, the MEK1-NT antibody was from Upstate Biotechnology, and antibodies against the maltose-binding protein and activated ERK came from New England Biolabs. The anti-Flag (M2) monoclonal antibody was from Eastman Kodak Co. Horseradish peroxidase-conjugated secondary antibodies were purchased from Bio-Rad, and the ECL Western blotting detection reagents were from Amersham Pharmacia Biotech. Protein-A and Protein-G Sepharose were from Amersham Pharmacia Biotech. Finally, the HeLa cell cDNA library and all of our cell culture reagents were obtained from Life Technologies, Inc.
Two-hybrid Assays-Following PCR amplification with the ANO-35 (GGGGGATCCAAATGCCCAAGAAGCCG) and ANO-36 (GCGCTCG-AGGCTCTTTTGTTGCTTCCC) primers, the coding sequences of the full-length human MEK1 gene (a gift from S. Pelech) were introduced between the BamHI and XhoI sites of the pEG202 LexA fusion plasmid, resulting in the pAN104 construct. Two-hybrid screening of 2 ϫ 10 6 primary transformants, from a human fetal brain cDNA library, was then performed exactly as described (18,19).
EcoRI and XhoI sites were introduced around the coding sequence of the human Raf1 gene (a gift from S. Meloche) by PCR amplification. The fragment was then subcloned in the same sites of the pEG202 vector, yielding pAN130. A similar PCR amplification was also used to insert the regulatory (aa 1-330) or catalytic (aa 331-649) domains of Raf1, as well as the cytoplasmic domain of the insulin (aa 974 -1370) and EGF (aa 672-1210) receptors, between the EcoRI and XhoI sites of pEG202. Subcloning of the partial MEK1 fusions in the pEG202 vector were performed as follows; N308, PCR amplification of the partial MEK1 gene with the ANO-35 and CTCGAGCCATGGGAGGTCGGCTGTCCT-TCC primers; N293, subcloned 5Ј SmaI fragment of pAN104 plasmid; N220, removed 3Ј NcoI/XhoI fragments from pAN104; C304, subcloned the EcoRI/XhoI insert of the pMB6 cDNA (which encodes aa 304 -392 of MEK1).
Quantitative ␤-galactosidase assays were performed using a modification of the permeabilized cell assay of Guarente (20). Briefly, cells were grown in 1.5 ml of media to an approximate A 600 of 0.5-1.0. They were then spun down for 2 min at 5000 rpm in a microcentrifuge and resuspended in 1.25 ml of Z-buffer (60 mM Na 2 HPO 4 ⅐7H 2 O, 40 mM NaH 2 PO 4 ⅐H 2 O, 10 mM KCl, 1 mM MgSO 4 ⅐7H 2 O, 50 mM ␤-mercaptoethanol, pH 7.0). We used 750 l of the suspension to determine cell density by measuring the A 600 (measuring cell density at this step reduced the introduction of errors due to loss of pelleted cells). We then added 25 l of CHCl 3 and 10 l of 0.1% SDS to the rest of the cell suspension and vortexed at top speed for 10 s. The enzymatic reaction was started by the addition of 100 l of ONPG solution (4 mg/ml o-nitrophenyl-␤-D-galactoside in Z-buffer) followed by incubation at 28°C. Upon development of a pale yellow color, the reactions were stopped by the addition of 250 l of 1 M Na 2 CO 3 . The samples were centrifuged at top speed for 2 min, and the amount of product was measured at A 420 . The activity, expressed as ␤-galactosidase units, is calculated with the formula, (1000 ϫ A 420 )/(A 600 ϫ reaction time (min)).
Amplification of a Full-length Grb10 Gene-Two g of DNA from a HeLa Cell cDNA library was used as a template in a PCR reaction with the GACGAATTCGAACCCATGGCTTTAGCCGGCTGCCCAG and GACCTCGAGCACAGACCGCTTCTTCACTCCAG primers. The resulting 2.2-kilobase pair fragment was fully sequenced on both strands and contains the coding sequences of the hGrb10 gene flanked by EcoRI and XhoI sites.
Production of Protein Fusions and Resin-binding Assay-EcoRI/XhoI fragments containing the hGrb10 gene or the SH2 domain encoded by the pMB58 cDNA were subcloned in the pMAL-c2 plasmid. The MBP-hGrb10 and MBP-SH2 fusion proteins were then purified from Escherichia coli UT5600 cells by affinity chromatography on amylose resin columns as described by the manufacturer (New England Biolabs). Plasmids containing the GST-MEK1 and GST-ERK1 constructs were obtained from S. Pelech, while the SEK1 gene was obtained from Jim Woodgett. Expression and purification of GST-MEK1, GST-SEK1, and GST-ERK1 were performed as described (21). GST-Raf1 was constructed by inserting the EcoRI/XhoI insert of pAN130 in pGEX-5X-1 and purified as described by Zhang et al. (22).
For the resin-binding assays, one g of the MBP fusion protein was incubated with 10 -50 l of glutathione-Sepharose resin, loaded with the GST-kinase fusions, in 1 ml of NETN buffer (20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40, 5 mM dithiothreitol, 10 mg/ml bovine serum albumin) for 30 min at 4°C. The resins were then spun down and washed three times with 1 ml of cold PBS containing 0.5% Nonidet P-40. Bound proteins were eluted by boiling in SDSpolyacrylamide gel electrophoresis buffer and detected by immunoblotting with an anti-MBP antibody. For the phosphatase assays, the GST-Raf1 and GST-MEK1 resins were first incubated with 0 -4 units of potato acid phosphatase (Boehringer Mannheim) for 20 min at 30°C in 40 mM Pipes, 50 mM NaCl, 1 mM ␤-mercaptoethanol. The resins were then washed three times with 1 ml of NETN before being used in the resin-binding assays.
Grb10 Mutagenesis-Site-directed mutagenesis of the Arg-␤B5 residue was done by overlap extension PCR (23). Random PCR mutagenesis of the pMB58 insert was performed according to Fromant et al. (24). The amplified fragment was then cleaved with EcoRI and XhoI and ligated back into pJG4 -5. Following transformation in E. coli, colonies from Ͼ10 4 transformants were inoculated in 100 ml of 2YT plus 100 g/ml ampicillin and grown for 4 h at 37°C. Plasmids from this mutagenized library were then purified and used to transform Saccharomyces cerevisiae RFY206. One thousand yeast colonies were patched on glucose Trp Ϫ plates, which were subsequently replica-plated to generate four copies. One plate was kept as a master, while the other three were mated to lawns of S. cerevisiae EGY48 cells previously transformed with the pSH18 -34 ␤-galactosidase reporter plasmid as well as one of either the pAN129 (LexA-EGF receptor COOH terminus), pAN138 (LexA-IR COOH terminus), pAN168 (LexA-RAF NH 2 terminus), or pAN104 (LexA-MEK1) bait plasmids. Following selection of the diploids on glucose His Ϫ Trp Ϫ Ura Ϫ plates, the cells were replica-plated to galactose/raffinose His Ϫ Leu Ϫ Trp Ϫ Ura Ϫ plates and scored for interaction. Plasmids from potentially interesting mutants were reisolated from the RFY206 master plates, sequenced, and retransformed in haploid EGY48 cells along with the pSH18 -34 reporter and LexA fusion plasmids. Binding specificity was confirmed by quantitative ␤-galactosidase assays on at least four independent transformants.
Cell Culture and Transfections-293 HEK, COS-7, and HTC-IR (25) cell lines were maintained in DMEM, supplemented with 10% fetal bovine serum (along with 400 g/ml G418 for the HTC-IR), at 37°C in a 5% CO 2 atmosphere. The 8-aa Flag immunological tag was added at the carboxyl-terminal end of the hGrb10 polypeptide by PCR amplification followed by subcloning in the EcoRI/XhoI sites of the pcDNA3 expression vector (Invitrogen), yielding pAN185. Transient transfections with lipofectamine were performed according to the manufacturer's instructions (Life Technologies, Inc.). Incubation with the DNAlipid complexes was carried on for 5 h followed by the addition of one volume of DMEM plus 20% fetal bovine serum. Twenty-four to fortyeight hours after transfection, the cells were starved overnight in DMEM plus 0.1% bovine serum albumin and then induced for 2-30 min with 100 nM insulin. Following hormone treatment, cells were washed with cold PBS, lysed for 10 min in lysis buffer (50 mM Hepes, pH 7.5, 150 mM NaCl, 100 mM NaF, 10 mM sodium pyrophosphate, 10% glycerol, 1.5% Triton X-100, 1.5 mM MgCl 2 , 1 mM EGTA, 10 g/ml aprotinin, 10 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride, 2 mM sodium vanadate) and spun down at maximum speed in a cold microcentrifuge for 10 min. Protein concentration was estimated with the Bio-Rad protein assay.
Production of Recombinant Adenovirus-The adenovirus (AV) vector used in this report is a modification of the human strain 5 in which the E1a, E1b, and E3 regions are deleted (26). The insulin receptor cDNA was isolated from the CMVneo plasmid (a gift of A. Ullrich). The resulting 4.2-kilobase pair ClaI-SpeII fragment was blunted with Klenow and cloned in the EcoRV site of the pAdcmvpoly(A) transfer vector (a gift of B. Massie). The 1.6-kilobase pair EcoRI-XhoI fragment encoding the full-length hGrb10 was also blunted with Klenow and cloned in the EcoRV site of the transfer vector. The homologous recombination of the replication-defective human type 5 AV, large scale production, purification, and titration of the recombinant AV have been described in detail (27). All recombinant AV were stored at a concentration of 1-4 ϫ 10 8 plaque-forming units/ml in DMEM supplemented with 10% fetal bovine serum. For the infections, 293 cells were transduced at a multiplicity of infection of 1 (1 plaque-forming unit/cell) for 2 h with stocks of either a control recombinant AV (Ad DE1/DE3) or recombinant AV (Ad-IR/Ad-Grb). Transduced cells were incubated for 24 h at 37°C in 5% CO 2 and then starved for 24 h in DMEM containing 0.1% bovine serum albumin. Cells were stimulated for 5 min with 100 nM insulin. The medium was then aspirated, and the cells were washed twice with PBS and solubilized in lysis buffer.
Immunoprecipitations-Protein samples were diluted to equal concentration with lysis buffer and then incubated with the antibodies for 1 h at 4°C. Protein A-or Protein G-Sepharose beads were then added, and the samples were incubated for another 1 h. The beads were then washed three times with PBS, and the bound proteins were released by boiling in SDS-polyacrylamide gel electrophoresis sample buffer.
Detection of Apoptosis by Microscopy and Flow Cytometry-The RL (GCTTTTTCTCCTCCTTGACAGCCAGAG), TS (GCATTTGTACTCT-CACTGTGTCATCACC) and SC (CTCCTCCGTGACTGCCAGAGTA-ATCC) primers were used to introduce SH2 domain mutations in the pAN185 plasmid using the Chameleon kit from Stratagene. The resulting pAN200, pAN208, and pAN209 plasmids express Grb10-R520L, Grb10-T531S, and Grb10-S522C, respectively. One million HTC-IR cells or 1-2 ϫ 10 5 COS-7 cells were inoculated in 35-mm wells and co-transfected with 0.5 g of pAdcMV5GFP Q (a gift from B. Massie) along with 1 g of either the pcDNA3 vector alone or one of the Grb10 expression plasmids. Twenty-four hours later, the cells were observed in a fluorescence microscope. Cells that were small, round, and showed nuclear fragmentation or condensation following Hoecht staining were counted as apoptotic. HTC-IR cells were also washed in PBS, trypsinized, and analyzed by flow cytometry. For the detection of apoptotic cells by the TUNEL assay, we omitted the GFP plasmid and used the ApoBrdU kit and MPlus software (Phoenix Flow Systems).

Interaction of Grb10 with MEK1 in a Two-hybrid Screen-
The SH2 domain of Grb10 was isolated from the two-hybrid screen of a human fetal brain cDNA library using the fulllength MAP kinase kinase MEK1. Our screen of 2 ϫ 10 6 primary transformants resulted in the isolation of five independent genes. The pMB58 cDNA 3 is almost identical to the 3Ј-end of the Grb10 gene (6,10,14) including the region that encodes the SH2 domain (Fig. 1A). A sequence comparison (Fig. 1B) between pMB58 and published Grb10 sequences suggests that this cDNA represents a novel splice variant with modifications in the extreme amino-terminal end of the SH2 domain. The borders between the regions of high and low sequence homology all contain sequences that are consistent with intron-exon borders. The variable region encoded by pMB58 still contains all the residues that have been shown to be important for SH2 domain function, namely a Trp-Phe-His motif in the first ␤-sheet as well as a conserved arginine residue 2 amino acids into the first ␣-helix (Fig. 1C).
Isolation of a Novel Grb10 Splice Variant-Unfortunately, we were unable to isolate longer Grb10 cDNAs from the fetal brain library. Using primers based on published sequence data 3 GenBank™ accession number AF000018.
FIG. 1. Structure of the Grb10 proteins. A, three domains, conserved among Grb7, Grb10, and Grb14 are shown as boxes. The lines below represent the splicing variants identified in humans as well as the extent of the partial polypeptide encoded by the MEK1-binding cDNA (pMB58). Regions whose aa sequences differ significantly among the variants are illustrated as small boxes. B, DNA sequence alignment between the 5Јend of the pMB58 cDNA and the other hGrb10 cDNAs (base pair labeling is that of the hGrb10 variant). Nucleotide sequences from the linker are shown in lowercase type up to the 5Ј EcoRI site. Putative intron-exon borders in the hGrb10 sequence are underlined. C, amino acid sequence alignment of the beginning of the polypeptide encoded by the pMB58 cDNA with the SH2 domain of the hGrb10 proteins (aa labeling is that of the hGrb10 variant). Amino acids derived from vectorial sequences are shown in lowercase type, while dashes represent insertions. Identities are marked as a vertical dash. The cylinder and arrow define the position of major features in the SH2 domain secondary structure, while the asterisk denotes residues thought to be important for interaction with tyrosinephosphorylated targets. The remaining pMB58 sequences are identical to the other hGrb10 genes.
(6), we isolated by PCR the coding sequence of a full-length human Grb10 gene from a HeLa cell cDNA library. Sequencing of the 2.2-kilobase pair fragment revealed that this clone is a hybrid between hGrb10␤ (6) and hGrb10␥ (10, 14) (see Fig. 1A). We named this new human splice variant, simultaneously isolated by Dong et al. (28), hGrb10. 2,4 It contains the longer amino-terminal sequences found in hGrb10␤ as well as the complete Pleckstrin homology domain seen in the other human and mouse variants. Inclusion of the pMB58 SH2 domain variant into this nomenclature will have to await the isolation of a full-length cDNA. Two forms of Grb10 were thus used in our experiments. The full-length Grb10 protein encoded by the hGRB10 variant was used in most experiments, while for simplicity, experiments using the SH2 domain used the sequences of the pMB58 clone.
Grb10 Interacts with Phosphorylated Raf1 and MEK1-Both the full-length hGrb10 protein and the SH2 domain initially isolated in the pMB58 cDNA were purified from bacterial extracts as fusions with the E. coli MBP. To confirm that the full-length hGrb10 could also bind to MEK1, we incubated the MBP-SH2 and the MBP-hGrb10 fusion proteins with glutathione-Sepharose resins loaded either with GST control protein or with GST fusions of various kinases. Interestingly, we found that MBP-SH2 was not only retained by the GST-MEK1 resin but also by GST-Raf1. We also observed binding by the fulllength MBP-hGrb10 protein to the GST-MEK1 (Fig. 2C) and GST-Raf1 resins (Fig. 3), indicating that the modifications in the pMB58 SH2 domain are not solely responsible for this binding specificity.
To determine if Grb10 will interact with any available kinases, we repeated the resin-binding assays using SEK1, a stress-activated kinase (also called JNKK, MEK4, or MKK4) (29 -32) that is 44% identical and 63% homologous to MEK1. Although both the GST-MEK1 and the GST-SEK1 resins contained equivalent amounts of bound proteins and both kinases had roughly similar autophosphorylation activities, only the MEK1 resin could effectively retain the MBP-hGrb10 protein (Fig. 2C). Thus, binding by Grb10 shows some specificity to members of the mitogenic response pathway.
To confirm whether the SH2 domain recognizes a phosphotyrosine-containing sequence on the bacterially expressed kinases, we probed the resin-bound GST-Raf1, GST-MEK1, and a GST fusion to the MAP kinase ERK1 with an anti-phosphotyrosine antibody. Fig. 2D clearly shows that, as described previously (21,33,34), only the bacterially expressed GST-ERK1 contains phosphotyrosine residues. This suggests a novel mode of interaction between the Grb10 SH2 domain and the Raf1 and MEK1 kinases that does not depend on phosphotyrosine residues.
There is growing evidence that certain SH2 domains can recognize their targets through non-phosphotyrosine residues (35)(36)(37)(38)(39)(40). To determine if the Grb10 SH2 domain recognizes another phosphorylated amino acid on Raf1 and MEK1, we incubated the appropriate GST resins with increasing concentrations of potato acid phosphatase. These resins were then washed and used in a binding assay with the full-length MBP-hGrb10. As shown in Fig. 3, binding of MBP-hGrb10 to both Raf1 and MEK1 was greatly reduced when the resins were pretreated with the phosphatase. A Coomassie-stained gel of the same samples confirmed that equal amounts of resin-bound proteins were used and that the reduced binding is not the result of proteolytic degradation of the kinases. Thus, either the Grb10 SH2 domain recognizes a phosphothreonine-or phosphoserine-containing sequence, or the modified structures of the dephosphorylated kinases are less desirable targets for Grb10 binding.
Localization of the Grb10-binding Site on Raf1 and MEK1-We returned to the two-hybrid assay both to confirm the observed Grb10-Raf1 interaction and to map the Grb10binding sites on the Raf1 and MEK1 kinases. Neither the full-length Raf1 nor its carboxyl-terminal catalytic domain (aa 331-648) interacted with the SH2 domain in the two-hybrid assay. The SH2-binding domain in Raf1 was located in its NH 2 -terminal regulatory domain (aa 1-330), since a LexA fusion of this region was able to confer growth on media lacking leucine to cells containing the pMB58 cDNA (Table I). A deletion of the 88 amino acids in the carboxyl terminus of MEK1 abolished interaction with the SH2 domain. We established that this absence of interaction is not the result of a two-hybrid artifact, since two of the COOH-terminal mutants (1-308 and 4 GenBank TM accession number AF000017. 1-293) were still capable of interacting with other MEK1binding clones that were isolated in our initial screen (results not shown). Due to the high background, most of the results from the amino-terminal deletions were inconclusive except for the 304 -393 construct in which growth with the Grb10 SH2 domain was clearly faster than with the vector alone. The carboxyl-terminal domain of MEK1 (aa 304 -393) contains 9 Ser/Thr residues including Thr-386, which is phosphorylated by MAP kinases in vitro (41,42). Another target for MAP kinase phosphorylation, Thr-292, lies very close to the end of the 1-308 mutant. After mutating either residue to alanine, a quantitative two-hybrid assay (Table II) demonstrated that Thr-292 is not involved in the interaction, while the T386A mutation reduced the affinity of MEK1 for Grb10 by almost half. These results thus locate the Grb10-binding site in the extreme COOH terminus of MEK1, a region that, incidentally, is not conserved in the sequence of the SEK1 kinase.
In Vivo Interaction among Grb10, Raf1, and MEK1-To determine whether the interaction of Grb10 with Raf1 and MEK1 also occurs in vivo, we performed immunoprecipitations in Triton-soluble protein extracts taken from a variety of sources. Fig. 4A shows the results obtained from 293 HEK cells in which both hGrb10 and the insulin receptor were expressed using adenovirus vectors. In these cells, Grb10 was co-immunoprecipitated by an anti-Raf1 antibody in both starved and insulintreated cells (lanes 3 and 4). We also used the anti-Flag monoclonal antibody to efficiently co-immunoprecipitate p74 raf1 from extracts of 293 HEK cells transiently expressing a hGrb10-Flag protein (Fig. 4B). As in the virus-infected 293 HEK cells, this interaction occurred irrespective of the presence of mitogenic agents. However, the Grb10-Raf1 interaction appears to be cell type-specific, since we were unable to coimmunoprecipitate these proteins from transfected NIH-3T3 or HTC-IR cells (not shown).
The binding of Grb10 to MEK1 in adenovirus-infected 293 HEK cells was greatly increased following treatment with insulin (Fig. 4A, lanes 5 and 6). In vivo binding of Grb10 to MEK1 was also observed in NIH-3T3 cells transfected with an hGrb10-Flag construct (not shown). Following the observation that the Thr-386 residue of MEK1 might be involved in interactions with Grb10 (Table II), we compared the kinetics of insulin-dependent MAP kinase activation with those of the Grb10-MEK1 interaction. HTC-IR cells were transfected with the hGrb10-Flag construct, starved overnight, and treated with 100 nM insulin for 2, 5, 15, or 30 min. As seen in Fig. 4C, co-immunoprecipitation of MEK1 by the Flag antibody was first detectable after 5 min of treatment and peaked at 15 min. MAP kinase activation, as detected with an antibody that specifically recognizes phosphorylated ERK, was first observed after 2 min and peaked at 5 min. The maximum levels of Grb10-MEK1 interaction thus follow MAP kinase activation by insulin.
Mutagenesis of the Grb10 SH2 Domain-The structure of several SH2 domains has revealed that an invariant arginine (Arg-520 in the case of hGrb10) interacts with the phosphate group of phosphotyrosine residues. We introduced an Arg-520 to leucine mutation in the pMB58 SH2 domain and observed that this mutant failed to interact, in a two-hybrid assay, with LexA fusions of the carboxyl-terminal domain of the insulin and EGF receptors, the regulatory domain of Raf1, or the full-length MEK1 kinase. Immunoblotting of total yeast cell extracts confirmed that expression of this mutant is not significantly different from that of the wild type SH2 domain (result not shown). This confirmed that a conserved residue, which is

TABLE I
Mapping of the Grb10-binding sites by the two-hybrid assay Full-length or partial Raf1 and MEK1 genes were fused to the LexA DNA-binding domain of pEG202 and tested in a two-hybrid assay against either the pMB58 cDNA, which encodes the Grb10 SH2 domain, or an empty pJG4 -5 vector. Binding was detected by the appearance of colonies on leucine-deficient media after 2 (ϩϩ) or 4 (ϩ) days of growth.  involved in the interaction of SH2 domains with the phosphate group of phosphotyrosine residues, is also necessary for the recognition of Raf1 and MEK1 by Grb10 although neither appears to be tyrosine-phosphorylated.
To identify additional residues involved in the interaction of Grb10 with the kinases, a library of mutagenized SH2 domains was generated by PCR amplification and screened for mutants defective in the interaction only with tyrosine kinase receptors, the amino-terminal domain of Raf1, or the full-length MEK1 kinase. Because of the high background resulting from expression of the LexA-IR fusion, we also tested our mutants with a LexA fusion of the carboxyl-terminal catalytic domain of the EGF receptor.
The positions of the point mutations are identified using the nomenclature for SH2 domain residues, first described by Waksman et al. (43), in which amino acids are labeled according to their position relative to major structural domains. Detailed structural data are available from three different SH2 domains, Src (43,44), Lck (45), and the NH 2 -terminal SH2 domain of the mouse Syp phosphatase (46). The structural features of these domains being relatively well conserved, we used an alignment of the pMB58 SH2 domain with the NH 2 -  (n ϭ 4). Ventral (C) and side views (D) of the three-dimensional structure of the Syp amino-terminal SH2 domain complexed with a phosphotyrosine-containing high affinity peptide are shown (47). In D, some of the frontal residues were sliced off to give a better view of the phosphate-binding pocket. Amino acids from Syp are colored blue, while the atoms of the target molecule are colored yellow (phosphate group) and green (peptide). The atoms equivalent to the Grb10 SH2 domain mutations have been labeled according to their binding specificities (red, inactivating mutation; light blue, impaired binding to tyrosine kinase receptors; orange, impaired binding to all targets but especially Raf1; magenta, impaired binding to MEK1).
terminal SH2 domain of the Syp phosphatase (47) (Fig. 5A) to better interpret our mutagenesis data.
The binding specificities of the mutant SH2 domains are shown in Fig. 5B. In addition to the Arg-␤B5 3 Leu mutant, the Arg-␤B5 3 Cys and Asp-EF2 3 Val mutants (shown in red in Fig. 5, C and D) also failed to interact with the LexA fusions. The Ser-␤B7 3 Cys mutation (colored in light blue) shows a greatly reduced affinity for the EGF and insulin receptor, moderate effects on binding to MEK1, and an increased affinity for Raf1. We failed to isolate any mutant that unequivocally affected binding only to Raf1. The best of such mutants is the Phe-␣B7 3 Tyr mutation (colored in orange), which shows reduced affinity to the receptors and the MEK1 kinase along with its inability to recognize Raf1. In Syp, the terminal hydroxyl group of this tyrosine residue protrudes into the proteinbinding groove (see Fig. 5C). Finally, the Thr-␤C5 3 Ser mutant (colored magenta) shows a relatively normal affinity for the EGF receptor, greatly increased binding to Raf1 and the insulin receptor, and no affinity for MEK1. A serine at position ␤C5 is normally seen in most of the other conventional SH2 domains, including those found in Syp, Grb7, and Grb14. Although this residue in Syp does not lie next to the tyrosinephosphorylated peptide, it makes extensive contacts with, and is believed to orient, the Arg-␤B5 residue, a major constituent of the phosphate-binding pocket (Fig. 5D).
Induction of Apoptosis by Grb10 Mutants-The Arg-␤B5 3 Leu, Ser-␤B7 3 Cys, and Thr-␤C5 3 Ser mutations were introduced in the full-length hGrb10 gene. These constructs (named Grb10-RL, Grb10-SC, and Grb10-TS) were co-transfected in HTC-IR cells along with a reduced amount of a second plasmid containing the gene for the green fluorescent protein (GFP). Consistent with the results of Morrione et al. (15), overexpression of wild type Grb10 had little effect on cell morphology and survival, but initial observations by fluorescence microscopy, followed by more stringent analysis by flow cytometry, showed a strong reduction in the number of GFP-positive adherent cells cotransfected with either of the three SH2 do-main mutants (Fig. 6A). Most of the GFP-positive cells were dead and could be found floating in the media, while the remaining adherent cells were smaller, rounder, and more refractive than normal (results not shown, but see Fig. 8). A TUNEL assay (49), which detects DNA strand breaks induced by programmed cell death, also showed an increase in the number of apoptotic cells following transfection with the Grb10-RL, Grb10-TS, and Grb10-SC plasmids (Fig. 6B). Because the small size and low transfection efficiencies of the HTC-IR cells made detailed analysis difficult, we repeated the transfections in COS-7. As seen in Fig. 7A, transfection of these normally large cells with SH2 domain mutants of Grb10 increased the proportion of small round cells from 17-23 to 42-53%. Co-staining of these cells with the DNA-binding dye Hoecht 33258 revealed the nuclear fragmentation that is one of the structural hallmarks of apoptosis (Fig. 8). Cells transfected with the pcDNA3 vector were similar to those expressing Grb10, while those transfected with Grb10-RL or Grb10-SC looked like cells transfected with Grb10-TS (results not shown). Finally, the effects of the Grb10-RL mutant can be reversed by co-expression of wild type Grb10 (Fig. 7B). DISCUSSION We used the yeast two-hybrid assay, resin-binding assays, and co-immunoprecipitations to demonstrate that Grb10 interacts with at least two members of the mitogenic MAP kinase cascade: Raf1 and MEK1. The extreme 5Ј-end of the cDNA initially isolated in the two-hybrid screen differs slightly from other Grb10 clones, possibly the result of splicing variation. To confirm that the interactions with the kinases are not limited to this specific SH2 domain variant, we used PCR to isolate a full-length human Grb10 cDNA, hGrb10, and observed, through resin-binding assays and immunoprecipitations, that it too can interact with Raf1 and MEK1. Phosphotyrosine blots, phosphatase treatment, and mutagenesis of the kinases have demonstrated that the recognition of Raf1 and MEK1 by the Grb10 SH2 domain is mediated by phosphothreonine or phos- phoserine residues. Examples of phosphotyrosine-independent binding by SH2 domains have been described before. These include the recognition of Bcr by the SH2 domains of cAbl, phospholipase C␥, Src, and GTPase-activating protein (35,36) and the binding of vAlb to Shc (40). It has also been reported that the SH2 domains of Fyn and Src interact with a phosphoserine residue on Raf (39), while a 62-kDa ubiquitin-binding protein is a phosphotyrosine-independent ligand of the p56 lck SH2 domain (37). There is also evidence for the interaction of the SH2 domain of the Syp phosphatase with its own catalytic domain in the absence of phosphotyrosine (38).
In a two-hybrid assay, only the isolated amino-terminal domain of Raf1 binds to the Grb10 SH2 domain. The absence of interaction with the full-length kinase is either an indication that the binding site is masked in the context of the whole protein or simply an artifact (our laboratory and others have frequently observed in two-hybrid assays that, while individual domains will interact with a given protein, the full-length protein will not (49)). In vivo interaction between Raf1 and Grb10 was observed in resting 293 HEK cells, which suggests that binding is mediated via a constitutively phosphorylated residue. For example, in unstimulated Balb/3T3 cells, Ser-43 and Ser-259, both of which are located in the regulatory domain of Raf1, are phosphorylated by means of a mechanism other than autophosphorylation (34). Binding between Grb10 and Raf1 is cell-type specific, since we failed to detect this interaction in NIH-3T3 and HTC-IR cells. The Grb10-binding site on MEK1 appears to be located in its carboxyl-terminal tail. Threonine 386, a residue that is phosphorylated in vivo by MAK kinases (41,42,50,51), was shown to play a partial role in Grb10 recognition. The residual binding of the SH2 domain to the MEK1-T386A mutant indicates that phosphorylation at this residue might not be as critical as it is for phosphotyrosinecontaining targets. The difficulty in synthesizing threoninephosphorylated peptides has greatly hampered further research into this area.
Based on the published structures of SH2 domains bound to phosphotyrosine targets, we know that Arg-␤B5 is positioned at the heart of the phosphate-binding pocket and makes hydrogen bonds with the phosphate residue. This interaction is also critical in the binding of the SH2 domain to Raf1 and MEK1, thus reinforcing our hypothesis for the recognition of a phosphoamino acid by Grb10. The equivalent of another SH2 domain residue necessary for interaction with the tyrosine kinase receptors and the two kinases, Asp-EF2, is positioned in Syp next to the polypeptide-binding groove and can theoretically make interactions with residues downstream of the phosphorylated amino acid. Structural analysis (43,44,46) and random mutagenesis of the phosphatidylinositol 3-kinase SH2 domain (52) have previously demonstrated the importance of the EF loop in binding specificity. The positions of the ␤B5 and EF2 mutants therefore suggest that the tyrosine-phosphorylated receptors and the Raf1 and MEK1 kinases interact with the Grb10 SH2 domain in a very similar fashion, involving interactions with a phosphorylated amino acid as well as with residues located further downstream. These results also suggest that the Grb10 SH2 domain cannot interact with more than one of these targets simultaneously, although this hypothesis has not been tested experimentally. Other residues in the Grb10 SH2 domain were shown to be involved in interaction with only some of its targets. One of these is Ser-␤B7, which is conserved between Grb10 and Syp and, in the latter case, makes hydrogen bonds with the phosphotyrosine. Mutagenesis of this residue into a cysteine has much greater effects on interactions with the receptors. If binding of the Raf1 and MEK1 kinases to Grb10 is coordinated by a much shorter phosphoserine or phosphothreonine residue, the phosphate group might not be in a position to make significant interactions with Ser-␤B7. Another mutant, Thr-␤C5 3 Ser, abrogates binding only to MEK1. Interestingly, almost all SH2 domains contain a serine at position ␤C5, including those found in Grb7 and Grb14. Other elements of the Grb10 SH2 domain have not been studied in detail but might also play a role in its binding specificity. For example, two residues that are critical for phosphotyrosine interactions in most SH2 domains, His-␤D4 and Lys-␤D6, are not conserved in Grb10. It was suggested, based on the Syp three-dimensional structure, that the ␤D-strand prevents shorter (3.5-Å) phosphothreonines and phosphoserines from coordinating with the Arg-␤B5 without an intervening water molecule, thus permitting interactions with only the 7.0-Å phosphotyrosine (46). In addition, the residue at position ␤D6 was recently shown to be a major determinant in the recognition of the tyrosine-phosphorylated ErbB2 receptor by Grb7 (53).
Fath et al. (54) have shown that expression of an isoform of the Grb2 adapter protein that is missing part of its SH2 domain induces apoptosis in Swiss 3T3 cells. We observed a similar phenomenon when the expression of Grb10 point mutants, whose SH2 domains are unable to interact with either MEK1 or with two tyrosine kinase receptors, increased the number of apoptotic cells following lipofectamine-mediated transfection. Reversal of the cell death phenotype by concomitant expression of the wild type Grb10 suggests that these mutants are acting by sequestering necessary signaling components. It has already been demonstrated that some tyrosine kinase receptors and the RAF-MEK-ERK pathway inhibit apoptosis (56,57) and that the equilibrium between the ERK and JNK pathways determines the choice between proliferation and programmed cell death (57,58).