Interaction of a GRB-IR Splice Variant (a Human GRB10 Homolog) with the Insulin and Insulin-like Growth Factor I Receptors EVIDENCE FOR A ROLE IN MITOGENIC SIGNALING*

We have utilized the yeast two-hybrid system to iden- tify proteins that interact with the cytoplasmic domain of the insulin receptor. We identified a human cDNA that is a splice variant of the human GRB10 homolog GRB-IR, which we term GRB10/IR-SV1 (for GRB10/ GRB-IR splice variant 1). The protein encoded by the GRB10/IR-SV1 cDNA contains an SH2 domain and a pleckstrin homology domain. Cloning of a full-length human cDNA revealed a predicted coding sequence that was similar to the mouse GRB10 protein, although GRB10/IR-SV1 contained an 80-amino acid deletion. The GRB10/IR-SV1 cDNA is a splice variant of the GRB-IR cDNA such that GRB10/IR-SV1 contains an intact pleck- strin homology domain and a distinct amino terminus. The interaction of GRB10/IR-SV1 with the insulin recep- tor and the insulin-like growth factor I (IGF-I) receptor is mediated by the SH2 domain, and we show that glu- tathione S -transferase-SH2 domain fusion proteins interact specifically in vitro with the insulin receptor de- rived from mammalian cells. The GRB10/IR-SV1 SH2 domain also interacted with an (cid:59) 135-kDa phosphoprotein from unstimulated cell lysates, an interaction that decreased after insulin

Insulin and insulin-like growth factor I (IGF-I) 1 mediate a variety of effects, including regulation of cellular metabolism, growth, differentiation, and apoptosis (3,4). These important hormones signal via structurally related receptor tyrosine kinases that interact with a number of cytoplasmic proteins that are believed to transduce signals to the interior of the cell. The best characterized proteins that interact with the insulin and IGF-I receptors include insulin receptor substrates 1 and 2 (IRS-1 and IRS-2) and the Src and collagen homologous (SHC) protein (3,5,6). These proteins have been shown to be substrates of these receptors, and phosphorylation of IRS-1 facilitates interaction with SH2 domain-containing proteins, including GRB2, Syp, and the p85 subunit of phosphatidylinositol (PI) 3-kinase, resulting in activation of various signaling cascades (7)(8)(9). Phosphorylation of SHC leads to its interaction with GRB2, which mediates activation of the Ras signaling pathway (10 -13). Both IRS-1 and SHC have been implicated in mitogenic signal transduction (14 -16). The molecular mechanisms by which the insulin and IGF-I receptors regulate other functions such as regulation of glucose transport are less clear, although IRS-1 has been suggested to play a significant role in this response (17). The leading candidate for a mediator of glucose transport regulation is PI 3-kinase, which interacts with and is activated by IRS-1 and, to a lesser extent, by the insulin receptor (18,19).
In addition to mitogenesis and glucose transporter regulation, many other cellular effects are thought to be regulated by insulin and IGF-I. These include regulation of amino acid uptake, regulation of a number of metabolic enzymes involved in glycogen, protein and lipid synthesis, regulation of ion transport, and protection from apoptosis (4,20). The regulation of such a diverse set of effects is likely to be mediated by a similarly diverse set of signaling proteins that can transduce these signals in a tissue-specific and tightly regulated manner. It is unlikely that the IRS and SHC proteins are sufficient to mediate all of these effects, particularly since SHC, IRS-1, and IRS-2 appear to be activated by a variety of receptors that do not appear to mediate the diverse signals that are regulated by the insulin receptor family of receptors (6,21,22). It is therefore possible that other proteins exist that interact with the insulin and IGF-I receptors and are responsible for the full range of effects outlined above. To begin to explore this possibility, we utilized the yeast two-hybrid system to screen a cDNA library to identify potential signaling proteins that interact with the cytoplasmic domains of the insulin and IGF-I receptors. We have previously shown the utility of this assay in the identification and characterization of the interactions that occur between these receptors and IRS-1, IRS-2, and SHC (23)(24)(25)(26)(27). In this paper, we report the cloning of a cDNA encoding a human homolog of the GRB10 protein (28) that interacts strongly with the insulin and IGF-I receptors in vitro and in vivo. This cDNA is an alternatively spliced version of the GRB-IR cDNA that was recently identified (29). Because of these alternative splicing events, the GRB10/IR-SV1 protein contains an intact PH domain, whereas the GRB-IR protein contains a deletion within this domain. In addition, the GRB10/ IR-SV1 protein has a distinct amino terminus compared with GRB-IR. We utilize microinjection techniques to show that the SH2 domain of GRB10/IR-SV1 can inhibit insulin-and IGF-Imediated (but not EGF-induced) mitogenesis, suggesting that the GRB10 family of proteins may play a role in mitogenesis by the insulin family of receptor tyrosine kinases.

EXPERIMENTAL PROCEDURES
Yeast Strains and Plasmids-The yeast strain EGY48 (␣, trp1, ura3-52, his3, leu2) and all yeast expression plasmids were provided by the laboratory of Roger Brent and have been previously described (26, 30 -32). All routine growth and maintenance of yeast strains were as described (34). Plasmid transformation of yeast was by the lithium acetate method (35). The insulin receptor, IGF-I receptor, p85, SHC, and IRS-1 cDNA fusions have been previously reported (23)(24)(25)(26). All site-directed mutants were generated using the method of Kunkel (36) using customized primers. 2 ␤-Galactosidase Assays-The colony color ␤-galactosidase assay was performed as described (26,37). The solution ␤-galactosidase assays were performed as described (38), and the units of ␤-galactosidase activity were calculated by the method of Miller (39).
Northern Blot Analysis-Human RNA blots were purchased from CLONTECH and hybridized to a GRB10/IR-SV1 probe using standard methodologies. The probe used corresponded to nucleotides 1309 -2286 of the GRB10/IR-SV1 cDNA. Rhesus monkey subcutaneous adipose tissue was obtained from Dr. Barbara Hansen, and RNA was prepared using the Trizol reagent (Life Technologies, Inc.). For the adipose RNA, 20 g of total RNA were run per lane.
Two-hybrid Library Screening-We screened a HeLa cDNA library provided by the Brent laboratory using the interaction trap methodology (30,32). We termed these proteins GRIPs for growth factor receptor-interacting proteins. The plasmids encoding the GRIP proteins were isolated from yeast and transferred to bacteria prior to sequencing. The initial clone contained amino acids 342-536 of the GRB10/IR-SV1 protein.
cDNA Library Screening and Reverse Transcription-PCR-We used the partial cDNA derived from the yeast two-hybrid library to screen a standard human skeletal muscle cDNA library (CLONTECH HL1124b) to identify a longer cDNA. We identified one cDNA that appeared to contain the full coding sequence of GRB10/IR-SV1 since multiple inframe stop codons were found upstream of the putative translational start site. To ensure that the 5Ј-end of the cDNA shown in Fig. 1A was in fact expressed in human skeletal muscle, we performed reverse transcription-PCR using human skeletal muscle mRNA (CLONTECH). We utilized the following primers to amplify a PCR product from human skeletal muscle poly(A) ϩ RNA: P1, 5Ј-CATCCTCCTTCTAAGTGG-TAG-3Ј (corresponding to nucleotides 27-47 on the sense strand); and P2, 5Ј-AGGAGGGGCACCGTGTCTGAC-3Ј (corresponding to nucleotides 330 -350 on the antisense strand). The predicted product was obtained from mRNA that had been reverse-transcribed using oligo(dT) primers as recommended by the manufacturer (Life Technologies, Inc.). A PCR product of the predicted size was obtained only when template cDNA was included, and this PCR product also contained predicted internal restriction sites (data not shown).
In Vitro Interaction Studies-In this study, we generated a GST fusion protein by introducing amino acids 358 -536 into the GST vector GEX5X (Pharmacia Biotech Inc.). This fusion protein, which contains the SH2 domain, was expressed in bacteria and purified on glutathioneagarose beads by standard methods (40). The beads that contained immobilized fusion protein were then incubated with cell lysates derived from CHO.T cells (which overexpress the insulin receptor) prior to or after insulin stimulation (10 min, 100 nM). Lysates were prepared by lysis for 30 min on ice in 50 mM HEPES (pH 7.6), 1% Triton X-100, 1 mM EGTA, 10 mM NaF, 20 mM sodium pyrophosphate, 1 mM phenylmethylsulfonyl fluoride, 1 mM sodium orthovanadate, and 10 g/ml each aprotinin and leupeptin, followed by spinning at 10,000 ϫ g for 10 min to remove insoluble material. The resulting lysate supernatants were incubated with the immobilized GST proteins for 4 h to overnight. After extensive washing with 50 mM HEPES (pH 7.6), 150 mM NaCl, and 0.1% Triton X-100, the proteins that coprecipitated with the GRB10/ IR-SV1 or control GST protein were analyzed by SDS-polyacrylamide gel electrophoresis, followed by immunoblotting with either an antiinsulin receptor (IR-CT1) (a gift from Ken Siddle) or anti-phosphotyrosine (PY20) (Transduction Laboratories) antibody. The experiments shown in Fig. 4B were performed in a similar manner using either CHO cells (⌬NPEY) (41) or Rat1 fibroblasts (wild type, YF2) (42) that stably overexpress each insulin receptor.
Microinjection Studies-Rat1 fibroblasts, which stably express the human insulin receptor (HIRcB) cells (43), were maintained as described previously (15) and grown on glass coverslips for microinjection experiments. The cells were rendered quiescent by serum deprivation for 24 -36 h prior to injection. GST-GRB10/IR-SV1 fusion proteins were expressed in DH5␣ cells and purified with glutathione-agarose beads by standard techniques (40). The fusion proteins were microinjected at a concentration of 2.5 mg/ml, and preimmune sheep IgG was added to the protein for detection purposes. Parallel injections of sheep IgG alone were also performed. One hour after injection, the cells were stimulated as indicated with growth factors at the following concentrations: insulin, 100 ng/ml; IGF-I, 100 ng/ml; EGF, 1 g/ml; or fetal bovine serum, 10%. A solution of 3-bromo-5Ј-deoxyuridine (BrdUrd) (Amersham Corp.) was also added to the medium for 16 h to label newly synthesized DNA. Following the labeling period, the cells were fixed and processed for immunofluorescence as described (44). Briefly, the cells were stained with a rat monoclonal anti-BrdUrd antibody (Accurate Scientific), followed by a tetramethylrhodamine B isothiocyanate-conjugated donkey anti-rat IgG antibody and a fluorescein isothiocyanate-conjugated donkey anti-sheep IgG antibody to detect the injected cells. This technique results in red nuclear staining in cells that have progressed to S phase of the cell cycle and green cytoplasmic staining in the injected cells. Results were analyzed and quantitated using an epifluorescence photomicroscope (Carl Zeiss, Inc.).

Identification of GRB10/IR-SV1 Using the Yeast Two-hybrid
System-We screened a HeLa cell-derived cDNA library with the intact insulin receptor cytoplasmic domain as the LexA (bait) hybrid. The cDNAs that were identified in this assay were termed GRIPs. One of these (GRIP-2) was found to contain an SH2 domain that was homologous to the GRB7 and GRB10 proteins, which had been previously described (28,45,46), and for the reasons outlined below was renamed GRB10/ IR-SV1 (for GRB10/GRB-IR splice variant 1) to denote its relationship to GRB10 and GRB-IR. We screened a human skeletal muscle library with this probe to identify a full-length cDNA. The sequence of the longest clone is shown in Fig. 1A. The insert from this clone encodes a 2286-base pair cDNA. Analysis of potential coding sequences suggests a protein product of 535 amino acids with a predicted molecular mass of 60,842 Da. It is likely that the ATG codon denoted as the start codon is correct since we noted two in-frame stop codons upstream of this Met codon with no intervening ATG codons, although it is possible that ATG residues farther downstream may be utilized. Analysis of the amino acid sequence revealed a PH domain encoded by amino acids 233-352 and an SH2 domain between amino acids 435 and 536.
GRB10/IR-SV1 Is Related to GRB10 and GRB-IR-Recently, a homolog of GRB7, termed GRB10, was identified (28) by virtue of its interaction with the EGF receptor using the CORT (cloning of receptor targets) methodology (46). Our GRB10/IR-SV1 clone showed a high degree of similarity to this cDNA especially within the carboxyl-terminal 400 amino acids. As shown in Fig. 1B, the PH and SH2 domains show 88 and 99% identity, respectively. The sequence between these domains is less conserved (83%). The most striking difference between GRB10/IR-SV1 and GRB10 is that GRB10 is 80 amino acids longer due to an internal insertion of 80 amino acids (specifically amino acids 117-196 of the murine GRB10 pro-FIG. 1. GRB10/IR-SV1 cDNA and predicted amino acid sequence and tissue specificity of expression. A, the cDNA shown was obtained as described under "Experimental Procedures." The predicted amino acid sequence is shown under the DNA sequence and is numbered on the right. The PH (amino acids 233-352) and SH2 (amino acids 435-536) domains are boxed. The two in-frame stop codons upstream of the putative translational start site are circled. The 5Ј-region of the GRB10/IR-SV1 cDNA that is distinct from that of the related GRB-IR cDNA (nucleotides 1-158) is underlined. The region of the GRB10/IR-SV1 PH domain that is excluded from the related GRB-IR protein is denoted by the heavy bars. tein) that would be found between amino acids 110 and 111 of the GRB10/IR-SV1 protein. Whether this difference is due to alternative splicing or represents a species difference is unclear. In addition, the amino termini of these proteins show only ϳ70% identity. Despite these differences, the sequence similarity, especially in the C-terminal regions, suggests that GRB10/IR-SV1 represents one human form of the murine GRB10 cDNA. More recently, a human homolog of GRB10 was identified using a two-hybrid system similar to ours and was termed GRB-IR (29). The GRB-IR cDNA is identical to the GRB10/IR-SV1 cDNA with two exceptions. First, the GRB10/ IR-SV1 cDNA contains a complete PH domain, while the GRB-IR cDNA has a 46-amino acid deletion within the PH domain. This difference is probably due to alternative splicing of an exon encoded by GRB10/IR-SV1 nucleotides 958-1095 that is included in GRB10/IR-SV1 and excluded from GRB-IR. Examination of the sequence surrounding this region is consistent with consensus splice sites. The second clear difference between GRB10/IR-SV1 and GRB-IR is that these cDNAs have completely different 5Ј-ends. Specifically, the initial 158 nucleotides of the GRB10/IR-SV1 cDNA are distinct from GRB-IR. This difference, presumably due to alternative splicing, would lead to the selection of a different start codon from that of GRB-IR since the start codon of GRB-IR is not present in the GRB10/IR-SV1 cDNA. Thus, the first potential start codon in GRB10/IR-SV1 is encoded by nucleotides 286 -288. This start codon, which contains a good "Kozak" consensus sequence with purines at positions Ϫ3 and ϩ1 in relation to the ATG codon, would therefore be in a similar position to that which has been predicted for the murine GRB10 protein (28). To be sure that the 5Ј-end of the GRB10/IR-SV1 cDNA was in fact expressed in human skeletal muscle and not a cloning artifact, we performed reverse transcription-PCR with human skeletal muscle mRNA as described under "Experimental Procedures." We chose two primers, one which was upstream of the region of difference between the GRB10/IR-SV1 and GRB-IR cDNAs (P1) and one which would be common to both (P2). These primers amplified a product of the predicted size that contained predicted internal restriction sites (data not shown). We therefore conclude that the cDNA depicted in Fig. 1A is expressed in human skeletal muscle. Comparison of the GRB10/IR-SV1 and GRB-IR cDNAs suggests that mRNA from this gene undergoes significant alternative splicing within the amino-terminal and PH domains, resulting in the production of distinct protein products. Thus, the GRB10/IR-SV1 protein, with its shorter amino terminus and an intact PH domain, might be expected to have distinct signaling functions compared with the GRB-IR protein.
Analysis of GRB10/IR-SV1 mRNA Expression in Human Tissues and Monkey Adipose Tissue-As shown in Fig. 1C, we hybridized a panel of human tissue mRNAs to the GRB10/IR-SV1 cDNA isolated using the two-hybrid system. Expression was highest in skeletal muscle and pancreas. Both of these tissues showed hybridization to an ϳ6.5-kb band. Interestingly, skeletal muscle showed hybridization to two additional bands, the most prevalent of which was at ϳ2.2 kb and another at ϳ5 kb. All tissues showed a minor band between 6 and 7 kb that showed slightly higher expression in heart and brain. These weakly hybridizing high molecular size bands may represent low level GRB10/IR-SV1 expression or hybridization to the related GRB7 mRNA, which in rodent tissues runs at ϳ7 kb (45,46). An additional band of ϳ3 kb was identified in some tissues with the following expression level: skeletal muscle Ͼ pancreas Ͼ Ͼ liver, kidney, heart. We also analyzed expression of GRB10/IR-SV1 in subcutaneous adipose tissue derived from rhesus monkey since human adipose RNA was unavailable. As shown in Fig. 1C, this tissue showed high level interaction with the GRB10/IR-SV1 probe and identified two diffuse bands at ϳ4 and 6 kb. The high level of expression in skeletal muscle and adipose tissue as well as the expression of unique bands within human muscle mRNA make this protein a candidate for a physiologically relevant effector protein for the insulin receptor. As expected, this pattern of expression is similar to that observed for GRB-IR by Liu and Roth (29), except that our longer exposure suggests some low level expression of this mRNA in other tissues, including heart and brain. It is unclear whether similar patterns of expression are seen in the mouse since skeletal muscle and pancreas were not examined (28).
GRB10/IR-SV1 Interacts with the Insulin and IGF-I Receptors in a Phosphotyrosine-dependent Manner-We next assayed the interaction of GRB10/IR-SV1 with the insulin or IGF-I receptor in the two-hybrid assay and compared the activity generated by GRB10/IR-SV1 with that observed with other insulin receptor-interacting proteins. We utilized a GRB10/IR-SV1 activation domain hybrid protein containing amino acids 342-536, which includes the SH2 domain. As shown in Fig. 2, coexpression of the GRB10/IR-SV1 hybrid protein with either the insulin or IGF-I receptor hybrid protein showed very high levels (500 -600 units) of activity. This activity was dependent on receptor autophosphorylation since mutation of the critical Lys (to Ala) within the ATP-binding sites of both receptors eliminated all activity. This is consistent with the likelihood that GRB10/IR-SV1 interacts with these receptors via its SH2 domain. GRB10/IR-SV1 activity in the two-hybrid assay was compared with that of other proteins known to interact with the insulin and IGF-I receptors. As shown in Fig. 2, GRB10/ IR-SV1 gave the highest level of activity compared with p85, The site of the 80-amino acid insertion within the mouse GRB10 protein is denoted with the arrow. B, the predicted proteins encoded by the human (h) GRB10/IR-SV1 and GRB-IR and murine (m) GRB10 cDNAs were compared. The numbers refer to percentage amino acid identity between each segment. aa, amino acids. C, a multiple-tissue human Northern blot was hybridized to a GRB10/IR-SV1 probe (left). Expression in rhesus monkey adipose tissue was also examined using 20 g of total RNA derived from adipose tissue (right). H, heart; Br, brain; Pl, placenta; Lu, lung; Li, liver; Sk, skeletal muscle; K, kidney; Pa, pancreas. SHC, and IRS-1. Although it has been reported that activities in the two-hybrid assay tend to reflect the affinity of interaction as measured by more conventional in vitro methodologies (47), we cannot conclude that the affinity of interaction of GRB10/ IR-SV1 and the insulin receptor is greater than that of the other proteins examined. However, the idea that the relative affinity of GRB10/IR-SV1 is higher than that of p85 is supported by the GST precipitations in vitro, where an equimolar amount of the GRB10/IR-SV1 SH2 domain precipitated more phosphorylated insulin receptor than did the p85 SH2 domain (see Fig. 4B). Taken together, these results are consistent with the idea that GRB10/IR-SV1 interacts efficiently with the insulin and IGF-I receptors.
In Vitro Interaction of the GRB10/IR-SV1 SH2 Domain with the Insulin Receptor-To further examine the interaction of GRB10/IR-SV1 and the insulin receptor, we produced a GST fusion protein that contained the GRB10/IR-SV1 SH2 domain. We examined the ability of this GST fusion protein to interact with proteins in lysates derived from unstimulated or insulinstimulated CHO cells that overexpress the insulin receptor. After incubation of the immobilized GST proteins with the CHO lysates, the complexes were washed extensively, and the coprecipitating proteins were analyzed by SDS-polyacrylamide gel electrophoresis and immunoblotted with an anti-insulin receptor or anti-phosphotyrosine antibody. As shown in Fig.  3A, the GST-GRB10/IR-SV1 fusion protein precipitated the insulin receptor from these lysates. The coprecipitation was dependent on prior insulin stimulation. This is consistent with the view that the SH2 domain of GRB10/IR-SV1 can interact with the autophosphorylated insulin receptor.
Interaction of the GRB10/IR-SV1 SH2 Domain with an Unidentified ϳ135-kDa Tyrosine-phosphorylated Cellular Protein-Interestingly, the GST-GRB10/IR-SV1 fusion protein also coprecipitated a protein of ϳ135-140 kDa from unstimulated cells (Fig. 3A). This interaction appeared to decrease after insulin stimulation, suggesting that tyrosine dephosphorylation of this protein may occur after insulin stimulation. Since it has been recently shown that one protein near this size range (125 kDa), focal adhesion kinase, showed decreased tyrosine phosphorylation after insulin stimulation (48), we tested whether this coprecipitating band might be focal adhesion kinase. Blotting of GST-GRB10/IR-SV1-associated proteins with an anti-focal adhesion kinase antibody suggested that this protein was not focal adhesion kinase (data not shown). We have also utilized antibodies against p130 cas and phospholipase C-␥, both of which were also negative (data not shown). Thus, the identity of the p135 protein remains unclear. Another protein was coprecipitated with the GST-GRB10/IR-SV1 fusion protein that migrated at ϳ185 kDa. This presumably represents IRS-1 and/or IRS-2. Whether this represents a direct interaction of GRB10/IR-SV1 and IRS-1 or is due to a coprecipitation with the insulin receptor is unclear. Two other protein bands (ϳ60 and 70 kDa) also coprecipitated with the GST-GRB10/IR-SV1 fusion protein in an insulin-dependent manner. The identities of these proteins are unclear, although neither band comigrates with SHC (data not shown).
To further examine the interaction of GRB10/IR-SV1 with the insulin receptor and with p135 and other bands, we examined the insulin dose dependence of these interactions. As shown in Fig. 3B, we stimulated CHO-IR cells with varying concentrations of insulin for 10 min and prepared lysates from these cells. The lysates were then mixed with immobilized GST-GRB10/IR-SV1 SH2 domain hybrid proteins. After extensive washing, coprecipitating proteins were examined by SDSpolyacrylamide gel electrophoresis and immunoblotted with either an anti-insulin receptor or anti-phosphotyrosine antibody. As shown in Fig. 3B, the insulin receptor was clearly coprecipitated beginning at a concentration of 1 nM. At higher insulin concentrations (5 nM), other proteins at ϳ185, 70, and 60 kDa were also coprecipitated. The identities of these proteins are unknown, although the 185-kDa band is presumed to be IRS-1. As discussed above, the p135 band, which coprecipitated with the GRB10/IR-SV1 SH2 domain in unstimulated cells, showed a clear concentration-dependent loss of coprecipitation beginning at insulin concentrations of between 1 and 5 nM.

GRB10/IR-SV1 Interaction with the Insulin Receptor Is Independent of Juxtamembrane and C-terminal Phosphoty-
rosines-To try to identify the site (or sites) of interaction within the insulin receptor that were responsible for GRB10/ IR-SV1 interaction, we used the two-hybrid assay to test GRB10/IR-SV1 for interaction with insulin receptor hybrid proteins containing deletions or substitutions of particular Tyr residues. As shown in Fig. 4A, GRB10/IR-SV1 showed undiminished interaction with the Y953A 3 and Y960F mutant receptors, suggesting that these juxtamembrane tyrosine residues are not essential for interaction. GRB10/IR-SV1 was also found to interact with an insulin receptor from which the 30 C-terminal amino acids had been deleted by introduction of a stop codon. Thus, the two tyrosine residues located within the carboxyl terminus of the insulin receptor (Tyr-1316 and Tyr- 3 The numbering of amino acids of the insulin receptor corresponds to the sequence of the receptor of Ullrich et al. (1). These differ from that of Ebina et al. (2) by being 12 amino acids less.

FIG. 3. In vitro interaction of the GRB10/IR-SV1 SH2 domain with the insulin receptor and other proteins.
A, cellular lysates were prepared from CHO cells that overexpress the insulin receptor before and after insulin stimulation (10 min, 100 nM). The lysates were then incubated with either a control GST protein or the GST-GRB10/ IR-SV1 SH2 domain fusion protein, and the associated proteins were examined by SDS-polyacrylamide gel electrophoresis, followed by immunoblotting with either an anti-insulin receptor (␣IR.CT1) or antiphosphotyrosine (␣PY20) antibody. B, lysates were prepared from insulin receptor (IR)-overexpressing CHO cells that had been treated for 10 min with the concentrations of insulin shown above the blot. These lysates were analyzed for proteins that interact with the GST-GRB10/IR-SV1 fusion protein with either an anti-phosphotyrosine or anti-insulin receptor antibody as described under "Experimental Procedures." 1322) are not essential for interaction in the two-hybrid assay. We also tested two insulin receptor mutants, which contained mutations within the "triple tyrosine" region of the kinase domain, that have been implicated in insulin receptor kinase activation (49,50). The Y1146F (FYY) insulin receptor mutant showed essentially wild-type GRB10/IR-SV1 interaction. Conversely, the Y1150F/Y1151F (YFF) mutant showed markedly reduced activities in the two-hybrid assay. However, this YFF receptor mutant has been found (data not shown) to have markedly reduced activities for all interacting proteins (the p85 subunit of PI 3-kinase, SHC, and IRS-1) and thus appears to be inefficiently autophosphorylated, a finding that is consistent with a previous report that examined these receptor mutants in CHO cells (50). We therefore cannot conclude nor rule out the possibility that Tyr-1150 and/or Tyr-1151 interacts with GRB10/IR-SV1. Nevertheless, it is clear that neither the juxtamembrane nor the carboxyl-terminal tyrosines are necessary for GRB10/IR-SV1 interaction, and we confirmed these findings in vitro by using the GST-GRB10/IR-SV1 SH2 domain fusion protein to assay interaction with the wild-type insulin receptor and two insulin receptor mutants expressed in fibroblasts. As shown in Fig. 4B, the GST-GRB10/IR-SV1 fusion protein interacted with an insulin receptor mutant in which the NPEY motif within the juxtamembrane region was deleted (41). Furthermore, GRB10/IR-SV1 showed interaction with an insulin receptor mutant in which the two C-terminal tyrosines (Tyr-1316 and Tyr-1322) had been changed to phenylalanine (CT-YF2) (42). As a control to validate the GST fusion precipitations using the GRB10/IR-SV1 SH2 domain, we showed that p85 interacted well with the wild type and the ⌬NPEY mutant, but not with the YF2 mutant. This is consistent with previous reports showing that p85 can interact with Tyr-1322 (51).

Microinjection of the GRB10/IR-SV1 SH2 Domain into Rat1 Fibroblasts Inhibits Mitogenesis Induced by Insulin and IGF-I, but
Not by EGF-To begin to test whether GRB10/IR-SV1 plays a functional role in insulin receptor signal transduction, we produced a GST fusion protein that contained the GRB10/ IR-SV1 SH2 domain. This fusion protein was microinjected into Rat1 fibroblasts that overexpress the insulin receptor, and the injected cells were monitored for mitogen-stimulated BrdUrd incorporation using previously described methods (15). As shown in Fig. 5, microinjection of the GRB10/IR-SV1 SH2 domain fusion protein inhibited insulin-and IGF-I-stimulated mitogenesis by ϳ50%. All microinjection samples contained preimmune IgG to allow the detection of injected cells, which were then scored as positive or negative for cell cycle progression based upon the presence or absence of nuclear rhodamine staining. Treatment of cells with growth factors resulted in an increase in the percentage of BrdUrd-positive cells from a basal level to the following values, which represent percent positive cells over background (untreated) cells: insulin, 56.1%; IGF-I, 37.0%; EGF, 35.2%; and serum, 48.7%. Microinjection of the GRB10/IR-SV1 fusion protein resulted in average decreases to 25.3% in insulin-treated cells and to 17.8% in cells treated with IGF-I, a decrease of ϳ50% for both insulin-and IGF-I-stimulated mitogenesis. Injection of the same fusion protein only marginally affected DNA synthesis in cells that were treated with EGF (from 35.2 to 31.1%) and had no effect on mitogenic stimulation by 10% fetal bovine serum. These results appear to rule out the possibility of a nonspecific toxic effect resulting from microinjection and indicate that the fusion protein demonstrates specificity in situ. Injection of a control GST protein that did not contain the GRB10/IR-SV1 SH2 domain had no effect compared with cells injected with preimmune IgG. Furthermore, microinjection of GST-SH2 domain fusion proteins derived from signaling molecules that are not required for insulin mitogenic signaling does not inhibit this assay even when they are injected at much higher concentrations (52,53). DISCUSSION Relationship of GRB10/IR-SV1 to GRB10 and GRB-IR-GRB10/IR-SV1 clearly represents a human homolog of the murine GRB10 cDNA as evidenced by the high degree of amino acid conservation in the PH and SH2 domains. It is unclear whether the additional 80 amino acids within the amino terminus of mouse GRB10 that are not present in human GRB10/ IR-SV1 play a functional role. The identity of the human GRB10/IR-SV1 and GRB-IR cDNAs (with the exception of the PH domain and amino terminus) clearly shows that these two cDNAs arise from the same gene, and their differences are almost certainly due to alternative RNA splicing. In fact, the existence of a GRB-IR cDNA such as ours, which contained an intact PH domain, was predicted by Liu and Roth (29) through the use of PCR, although no sequence was presented. The differential splicing of these RNAs suggests that the resulting GRB10/IR-SV1 and GRB-IR proteins might be expected to play distinct functional roles. The inclusion of an intact PH domain in the GRB10/IR-SV1 protein is of obvious potential importance to signal transduction by this molecule. The distinct amino termini predicted for these two proteins may also play a role in differential signaling by these proteins.
Interaction of GRB10/IR-SV1 with the Insulin and IGF-I Receptors-The interaction of GRB10/IR-SV1 with the insulin and IGF-I receptors clearly requires receptor kinase activity and almost certainly involves the SH2 domain of GRB10/IR-SV1. No interaction was observed in the two-hybrid system using receptor mutants that were kinase-inactive due to mutation of the critical lysine within the ATP-binding site (Fig. 2). As further evidence of this, no interaction of a GST-GRB10/IR-SV1 SH2 domain fusion protein could be detected with unstimulated mammalian cell-derived insulin receptors in vitro (Fig. 3). To try to identify the site(s) of interaction within the insulin receptor, we assayed the ability of GRB10/IR-SV1 to interact with a number of insulin receptor mutants containing specific tyrosine mutations or deletions. Specifically, we mutated or deleted each tyrosine that is thought to be phosphorylated during receptor activation, including the juxtamembrane Tyr-953 and Tyr-960; the activation loop (triple tyrosine) residues 1146, 1150, and 1151; and the two C-terminal residues 1316 and 1322 (5). In vitro interaction studies (Fig. 4B) support these two-hybrid findings since the GST-GRB10/IR-SV1 SH2 domain fusion protein was able to precipitate the CT-YF2 and NPEY insulin receptor mutants. We were unable to identify any mutants that significantly reduced the interaction of GRB10/IR-SV1, except for the Y1150F/Y1151F (YFF) mutant receptor, which did show a significantly reduced interaction. Although this suggests that these tyrosines may provide a binding site for GRB10/IR-SV1, it should be noted that this YFF receptor mutant showed markedly reduced interactions with the p85 subunit of PI 3-kinase, SHC, and IRS-1 (data not shown), suggesting that this mutant is probably unable to autophosphorylate properly. This is consistent with previous work showing that the FYY receptor autophosphorylates as well as the wild-type receptor, whereas the YFF receptor shows little or no phosphorylation (50). Despite these caveats, we cannot rule out that Tyr-1150 and Tyr-1151 may in fact interact with GRB10/IR-SV1. Another possibility is that GRB10/IR-SV1 may interact with more than one phosphotyrosine, and therefore, single mutations may not effectively reduce the interaction. Alternatively, GRB10/IR-SV1 may interact with a phosphotyrosine that has not been previously identified as an autophosphorylation site. In support of this possibility, it has been reported that at least one previously unidentified tyrosine may be phosphorylated within the insulin receptor after insulin stimulation of hepatoma cells (33). Related to this idea is the possibility that the insulin receptor hybrid proteins used in the two-hybrid system may phosphorylate tyrosines that are not typically phosphorylated in the normal in vivo situation. This possibility seems unlikely, however, since GST-GRB10/ IR-SV1 SH2 domain hybrid proteins were also shown to interact in vitro with activated insulin receptors derived from CHO cells. Thus, although we have been unable to clearly identify specific insulin receptor tyrosines that mediate the interaction with GRB10/IR-SV1, these tyrosines are definitely phosphorylated in an insulin-dependent manner in cells of mammalian origin.
Interaction of the GRB10/IR-SV1 SH2 Domain with Other Cellular Proteins-Our finding that the GRB10/IR-SV1 SH2 domain can recognize specific phosphoproteins in vitro before and after insulin stimulation of cells suggests that GRB10/IR-SV1 signaling may be complex. Of particular interest is the ϳ135-kDa phosphoprotein that is recognized by the GRB10/IR-SV1 SH2 domain prior to insulin stimulation and that shows decreased interaction after insulin stimulation. One simple interpretation of these data is that in the unstimulated cell, GRB10/IR-SV1 may interact with an unidentified protein that is constitutively phosphorylated on tyrosine residues. After insulin stimulation, detection of the interaction of GRB10/IR-SV1 with the pp135 protein decreases in vitro. This may be due to a rapid insulin-dependent dephosphorylation of the pp135 protein, or alternatively, the pp135 protein may be shifted to a relatively insoluble cellular compartment such that it is no longer present in the cell lysate supernatant. Further experiments will be required to address this issue. We also found that the GRB10/IR-SV1 SH2 domain coprecipitated at least three other tyrosine-phosphorylated proteins from lysates derived from insulin-stimulated cells. These proteins ran at ϳ60, 70, and 185 kDa. The identities of these proteins are unclear, although the 185-kDa band most likely represents IRS-1 and/or IRS-2.
Regulation of Cellular Mitogenesis by GRB10/IR-SV1-Our microinjection experiments provide the first evidence that the GRB10/IR-SV1 protein may play a role in mitogenic signaling by the insulin and IGF-I receptors. Our demonstration that the SH2 domain can inhibit insulin-and IGF-I-stimulated mitogenesis by ϳ50% whereas EGF stimulation was not significantly inhibited suggests that there is some specificity of the participation of GRB10/IR-SV1 in signaling pathways. These findings are consistent with previous observations that suggested that interactions of these proteins with the EGF receptor are relatively weak compared with those with the insulin receptor (29) and with the original observation that GRB10 interacts weakly with the EGF receptor compared with other known EGF receptor-binding proteins (28). It is possible therefore that the GRB10, GRB10/IR-SV1, and GRB-IR proteins play a physiological role in insulin receptor and IGF-I receptor signaling, but not in EGF signaling. The lack of an effect upon serum-stimulated mitogenesis shows that the GRB10/IR-SV1 SH2 domain fusion protein is not toxic to the cells and also that GRB10/IR-SV1 is not essential for cell cycle progression in response to all mitogens.
GRB10/IR-SV1, a Positive or Negative Signaling Pro-tein?-It has been suggested that the closely related GRB-IR protein may be an inhibitor of insulin receptor signaling (29). This conclusion is based upon studies in which a CHO cell line was generated by stable transformation with a GRB-IR expression construct. This cell line was reported to show reduced phosphorylation of IRS-1 and a 60-kDa substrate and to have an ϳ50% reduction in PI 3-kinase activity after insulin stimulation. Conversely, our study, which used mitogenesis as an end point, suggests that GRB10/IR-SV1 may be a positive mediator of insulin-stimulated mitogenesis. Several explanations for these seemingly contrary findings are possible. First, it is possible that results obtained from a stably transformed CHO cell line that overexpresses the entire GRB-IR protein could be affected by clonal variation especially since this study focused only on early insulin signaling end points and not on final effects such as mitogenesis. Second, it is possible that these two proteins may play different roles in signal transduction. In this regard, the GRB-IR and GRB10/IR-SV1 proteins have clear structural differences. These two proteins have distinct amino termini, and perhaps more important, the GRB-IR protein does not contain an intact PH domain. Thus, expression of full-length GRB-IR might act in a dominant-negative fashion since it can interact with the insulin receptor, but potentially not with other signaling components via the PH domain. Third, it is possible that binding of GRB-IR to the insulin receptor simply displaces other signaling proteins such as IRS-1 and thus leads to decreased IRS-1 phosphorylation and PI 3-kinase activation. Although our data show that GRB10/IR-SV1 binds to the insulin receptor despite deletions or mutations in the juxtamembrane domain known to bind IRS-1 and SHC, we cannot conclude that GRB10/IR-SV1 does not affect binding of these proteins to the receptor. Similarly, our microinjection data can be interpreted in multiple ways. It is possible that after microinjection, the GRB10/ IR-SV1 SH2 domain binds to the activated receptor and blocks interaction of the endogenous GRB10/IR-SV1 protein with the receptor and therefore blocks the activation of an uncharacterized mitogenic signaling pathway. Alternatively, as discussed above, the GRB10/IR-SV1 SH2 domain may block binding of known mitogenic signaling proteins such as IRS-1 and SHC to the insulin receptor. Our data suggest the possibility that GRB10/IR-SV1 may link the insulin family of receptor tyrosine kinases to a mitogenic pathway that is distinct from IRS-1 and SHC. Further work will be required to address these issues, including the generation of specific antibodies to these proteins. A better understanding of the signaling proteins that lie downstream of the GRB10/GRB7 family of signaling proteins will also be essential to resolve the role of these proteins in receptor-mediated signal transduction.