Human GRB-IRbeta/GRB10. Splice variants of an insulin and growth factor receptor-binding protein with PH and SH2 domains.

cDNA clones encoding human (h) Grb7 and a previously unknown protein with high homology to hGrb-IR and mGrb10 (where m indicates mouse) were found by screening expressed sequence tag data bases. hGrb7 mRNA expression is greatest in pancreas and restricted to a few other tissues. The second protein termed hGrbIRb/Grb10 contains an intact PH domain and lacks the 80-residue mGrb10 insertion. Expression is greatest in pancreas and muscle but occurs in nearly all tissues. hGrb-IRb/Grb10 and hGrb-IR likely arise as alternative mRNA splicing products of a common gene. Reverse transcriptase-coupled polymerase chain reaction shows both mRNAs in muscle. In cells, Grb-IRb/Grb10 protein translocates from cytosol to membrane upon insulin stimulation, most likely due to direct interactions with the insulin receptor. These interactions are mediated by the SH2 domain and additional regions of the protein. Studies with mutated receptors and synthetic phosphopeptides show that the hGrb-IRb/Grb10 SH2 domain binds at least two sites in the insulin receptor: the kinase activation loop > the juxtamembrane site. hGrbIRb/Grb10 also binds a 135-kDa phosphoprotein in unstimulated 3T3-L1 adipocytes; binding is reduced upon insulin stimulation. In addition, the c-Abl SH3 domain binds Grb-IR/Grb10, whereas Fyn, phosphatidylinositol 3-kinase p85, and Grb2 SH3 domains do not. The site of c-Abl SH3 domain interaction is highly conservedwithin the Grb-IR/Grb10/Grb7/Grb14 family. hGrb-IRb/Grb10 also binds platelet-derived growth factor and epidermal growth factor receptors, suggesting a broader role in the signaling pathways of numerous receptors. We conclude that hGrb-IRb/Grb10 is a widely expressed, PH and SH2 domain-containing, SH3 domain-binding protein that functions downstream from activated insulin and growth factor receptors.

Many of the effects of activated tyrosine kinase-linked receptors are mediated by cascades of intracellular tyrosine phosphorylation reactions. Receptors with intrinsic kinase activity typically phosphorylate themselves, and in many cases the phosphorylated receptor tyrosines serve as docking sites for SH2 domain proteins (1,2). Since many SH2 domain proteins either are enzymes or associate with enzymes, these interactions provide a mechanism for recruiting catalytic effectors to the activated receptor. In the case of insulin signaling, autophosphorylation activates the receptor kinase (3,4) and creates a docking site for substrate protein PTB domains (5,6). Both effects are necessary to trigger intracellular pathways via the substrates IRS-1 and Shc. However, the SH2 domain effectors of insulin action bind primarily to the phosphorylated substrate proteins rather than the insulin receptor itself.
The recent discovery of an SH2 domain protein called Grb-IR was met with considerable interest because it binds the insulin receptor and not its substrates (7). However, this Grb7-like protein reportedly inhibits insulin signaling and contains an unusual 46-residue deletion within its apparent PH domain. Therefore, we have considered the possibility that additional related proteins (potentially with intact domains) might exist to provide positive signals downstream from the insulin receptor. Three members of the Grb7 family (Grb7, Grb10, and Grb14) have been identified by screening cDNA expression libraries with phosphorylated fragments of the EGF 1 receptor (8 -10). Although each is derived from a distinct genetic locus, they share a common domain architecture: a C-terminal SH2 domain and Ͼ300 residues of extended homology that encompasses a PH domain. mGrb10 also contains an 80-residue insertion, relative to mGrb7 and hGrb14. Its function is unknown. Although the presence of SH2 and PH domains strongly implies a role for these proteins in cellular signaling, their physiologic functions remain vague. Nevertheless, all have been implicated in neoplastic conditions. mGrb7 binds and is coamplified with HER2/neu in certain types of breast cancer (11). mGrb10 binds the Ret receptor (12), whose gene (the ret protooncogene) is rearranged and activated in certain thyroid carcinomas and contains germ line mutations associated with syndromes of multiple endocrine neoplasias (13,14). And expression of Grb14 may be elevated in estrogen receptorpositive breast cancer cell lines and certain prostate cancer cells (10). hGrb-IR and mGrb10 proteins share regions of high homology, although the two are not simple species variants. hGrb-IR has not been implicated in oncogenesis.
Expressed sequence tag data bases (dbEST) were screened as a strategy to identify related proteins. Two clones were found. 2 One encodes human Grb7. Further sequencing of the second clone revealed a previously unknown protein with high homology to hGrb-IR and mGrb10, which we refer to as hGrb-IR␤/Grb10. It has an intact PH domain and lacks the mGrb10 insertion. hGrb-IR␤/Grb10 and hGrb-IR probably represent alternative mRNA splicing products of a common gene. Both mRNAs are expressed in muscle, a major site of insulin action. hGrb-IR␤/Grb10 protein is present in the cytosol of unstimulated Rat1 fibroblasts and translocates to the membrane following insulin stimulation. We have characterized hGrb-IR␤/ Grb10 interactions with activated insulin, EGF, and PDGF receptors and SH3 domain proteins. We conclude that Grb-IR␤/ Grb10 is a previously unknown signaling protein that may function downstream from activated insulin and growth factor receptors.

MATERIALS AND METHODS
DNA Sequencing-Automated DNA sequencing was carried out at the Joslin Diabetes Center DNA Core Facility with an Applied Biosystems Model 373 DNA sequencer. Sequencing substrates were produced by unidirectional nested deletions of plasmid substrates, with ExoIII/S1 treatment. Sequence assembly and analysis was carried out with Genetics Computer Group Wisconsin Package version 8.1 software.
Northern Blot Analyses-The inserts of the plasmid cDNA clones HUKCW90 and HCEEI20 were amplified by PCR with the T3 and T7 primers. Purified PCR fragments, labeled with [ 32 P]dATP to greater than 2 ϫ 10 9 cpm/g by the random hexamer method, were hybridized to human multiple tissue Northern blots (Clontech). The membranes were washed at high stringency and exposed to storage phosphor screens (Molecular Dynamics). Northern blot manipulations were carried out according to procedures recommended by the manufacturer.
hGrb-IR/Grb10 Fusion Proteins-EcoRI and XhoI sites were introduced by PCR immediately upstream and downstream of the regions of the hGrb-IR␤/Grb10 cDNA encoding residues 1-536 and 435-536. The DNA fragments were subcloned into the corresponding sites in a pGEX4T-3 (Pharmacia Biotech Inc.) plasmid and used to transform Escherichia coli strains LE392 and XL-1 Blue (Stratagene). Following induction of protein expression with isopropylthio-␤-D-galactoside and cell collection and lysis, the protein was purified by affinity chromatography using an immobilized glutathione-agarose column (Molecular Probes). GST (glutathione S-transferase) fusion proteins were eluted with 50 mM glutathione and dialyzed against 100 mM ammonium bicarbonate containing 1.0 mM dithiothreitol. The proteins were concentrated using a Centricon-10 device (Amicon).
Cell Lines-3T3-L1 cells were grown in DMEM containing 10% calf serum. To induce the differentiation of 3T3-L1 fibroblasts into adipocytes, cells in DMEM containing 10% fetal bovine serum were treated for 3 days with 0.5 mM 1-methyl-3-isobutylxanthine, 0.4 g/ml dexa-methasone, and 5.0 g/ml insulin. Cells were then maintained for an additional 10 -15 days in DMEM containing 10% fetal bovine serum and 5.0 g/ml insulin. Prior to experiments, the 3T3-L1 adipocytes were serum-deprived for 48 h in DMEM containing 0.2% bovine serum albumin. Rat1 fibroblasts overexpressing the wild-type human insulin receptor (HIRc) were kindly provided by J. Olefsky, University of California, San Diego. Mutated human insulin receptors (15)(16)(17) were obtained by solubilizing transfected Chinese hamster ovary cells as described (18) (the cells were generously provided by C. R. Kahn and M. White, Joslin Diabetes Center).
Cellular Fractionation-HIRc fibroblasts were serum-deprived for 16 h prior to stimulation with 10 Ϫ7 M insulin for 5 min at 37°C. The cells were collected by scrapping in ice-cold buffer A (50 mM Hepes, 150 mM NaCl, 10 mM EDTA, 10 mM Na 4 P 2 O 7 , 100 mM NaF, 2 mM sodium vanadate, 0.5 mM phenylmethanesulfonyl fluoride, 10 g/ml aprotinin, 10 g/ml leupeptin, pH 7.4) and further disrupted by multiple passages through a 26-gauge needle. After a slow speed centrifugation (600 ϫ g for 5 min) to remove nuclei and cellular debris, the mixtures were centrifuged 1 h at 10 5 ϫ g and 4°C. Supernatants are considered to be cytosolic fractions. The pellets were resuspended in buffer A containing 1% Triton X-100 and mixed for 45 min at 4°C. These solutions were centrifuged 1 h at 10 5 ϫ g and 4°C. Supernatant solutions are considered to be the membrane fraction and pellets are Triton-insoluble fractions. Proteins were separated by SDS-PAGE and identified by immunoblotting with specific antibodies.
Phosphopeptide Competition-Phosphopeptides corresponding to phosphorylation sites within the insulin receptor were synthesized manually using Fmoc-protected amino acids and O-benzotriazolyl-N,N,NЈ,NЈ-tetramethyluronium hexafluorophosphate as the coupling reagent: the monophosphoryl juxtamembrane site, pY960 (SSNPEpYL-SASDVE-NH 2 ); the bisphosphoryl C terminus, pY2CT (GFKRSpYEE-HIPpYTHMNG-NH 2 ); and the trisphosphoryl activation loop, pY3Loop (MTRDIpYETDpYpYRKGGKG-NH 2 ) (19). Peptides were purified by preparative high performance liquid chromatography and characterized by analytical high performance liquid chromatography and electrospray mass spectrometry. Solubilized, wheat germ agglutinin-agarose-purified insulin receptors (19,20) were reacted with 10 Ϫ7 M insulin for 45 min and phosphorylated in the presence of 15 M ATP, 8 mM MgCl 2 , and 4 mM MnCl 2 for 30 min, all at 22°C. Immobilized hGrb-IR␤/ Grb10 GST-SH2 domain or GST full-length protein (2 g) was incubated at 22°C for 30 min with the phosphorylated insulin receptor. Incubations were conducted in the presence and absence of the phosphopeptides, and bound insulin receptor was identified by immunoblotting.
SH3 Domain Binding-Human Fyn SH3 domain (residues 84 -148) (provided by H. Band, Brigham and Women's Hospital) (21), murine c-Abl type IV SH3 domain (residues 84 -138) (provided by B. Mayer, Children's Hospital, Boston), and PI 3-kinase p85 SH3 domain (residues 1-80) (provided by L. Cantley, Beth Israel Hospital) (22) were expressed as GST fusion proteins using pGEX-2T vectors (Pharmacia), as described. Full-length murine Grb2 (provided by M. Moran, University of Toronto) was expressed as a GST fusion protein using pGEX-3X vector (Pharmacia). Methods for fusion protein expression and purifi-cation were as described above. HIRc fibroblasts were solubilized, and clarified cell lysates were incubated with 2 g of immobilized SH3 domain or Grb2 for 1 h at 4°C. Incubations were conducted in the presence or absence of 3BS peptide (SLPAIPNPFPEL). Pellets were washed, proteins were separated by SDS-PAGE, and Grb-IR/Grb10 isoforms were identified by immunoblotting.

Identification of Human cDNAs Encoding Grb7 and Grb-IR/ Grb10
Related Proteins-Genome sequencing initiatives have generated many millions of nucleotides of human DNA sequence. Much of this information was derived from expressed sequence tags (23). Although expressed sequence tags typically represent incomplete gene sequences, partial protein coding sequences can be deduced from the data, and in some cases it is possible to predict the function of the encoded protein based on sequence homology. This is particularly true for proteins with relatively short but characteristic homology domains. Expressed sequence tags 370184 and 201358 (TIGR HCD Accession D70184 and C01358) were identified by the BLASTX homology search program as potential SH2 domain proteins. 2 Further inspection and alignment with all known SH2 domains suggested that these proteins might be members of the Grb7/ Grb10 family of SH2 domain protein, as these are the only known proteins with C-terminal SH2 domains that end with the residues Val-Ala-Leu (e.g. Fig. 1). The corresponding plasmid cDNA clones (HUKCW90 and HCEEI20) were obtained from Damien Dunnington (SmithKline Beecham Pharmaceuti-cals), and the nucleotide sequences of the inserts were refined and extended. HUKCW90 was derived from a human uterine cancer cDNA library. The insert contains a 2.0-kb cDNA encoding an intact protein identical to human Grb7. The sequence of hGrb7 has not been published in journal format but is available (GenBank D43772). HCEEI20 was derived from a human cerebellum cDNA library. It contains a 3.5-kb cDNA encoding a protein fragment whose C-terminal sequence (residues 388 -548) is identical to human Grb-IR (7) and very similar to murine Grb10 (9).
Isolation and Sequence Characterization of the Full-length cDNA-A gt11-specific primer and an HCEEI20-specific primer were used to amplify cDNAs carrying 5Ј extensions of HCEEI20 from a human skeletal muscle cDNA library by PCR (see "Tissue Distribution," below). Three 5Ј-RACE (rapid amplification of cDNA ends) candidates were isolated, subcloned, and sequenced. The entire sequences of the 1.0-and 0.4-kb products were found within the 1.7-kb product. The combination of sequences of HCEEI20 and the 1.7-kb RACE product yielded an open reading frame which potentially encodes a 536-amino acid protein (Fig. 1). The putative initiation site is preceded in-frame by two termination codons. The termination codon after residue 536 is followed by a series of additional termination codons (data not shown, cDNA sequence deposited, GenBank accession number U69276). Residues 233-355 fit the consensus for an intact PH domain (24,25). Residues 434 -536 form an SH2 domain. The structure and function of the region encompassed by residues 1-232 is difficult to predict, other than a conserved site with potential for SH3 domain binding (3BS, residues 76 -87).
This new, putative PH/SH2 domain protein is related in sequence and domain architecture to mouse Grb10 and human Grb-IR (Fig. 1). mGrb10 contains a similar PH domain (85% identity and 93% similarity, as analyzed by the program GAP), and its SH2 domain is essentially identical (99%). Our protein does not contain the 80-residue mGrb10 insertion. Where the cDNAs encoding human Grb-IR and our protein are common they are identical, suggesting that the two arise from the same genetic locus. However, the Grb-IR cDNA contains a distinct 5Ј end and a 144-nucleotide deletion in its center, relative to our sequence. These result in a 58-amino acid N-terminal extension and a 48-residue PH domain deletion within Grb-IR, relative to our protein (Fig. 1, A and B). The 5Ј-untranslated sequences are also distinct, while 3Ј-untranslated sequences are common. Potential consequences of the insertion and deletion differences are discussed in subsequent sections. Since this new protein is distinct from Grb-IR and mGrb10, and interacts with the insulin receptor, we have called it human Grb-IR␤/Grb10.
Tissue Distribution-Radiolabeled inserts derived from clones HUKCW90 and HCEEI20 were used to probe multiple tissue Northern blots to assess the tissue distributions of corresponding mRNAs. The largest amount of a 2.3-kb human Grb7 transcript was in pancreas ( Fig. 2A). Lesser amounts were detected in kidney, prostate, small intestine, and placenta. In contrast, Margolis and co-workers (8) found predominant expression of mouse Grb7 mRNA in the liver and kidneys of a 6-week-old mouse, with less in the gonads. This may reflect variations in Grb7 expression due to age or species (adult human versus immature mouse) or differences in protein function.
The HCEEI20 probe (hGrb-IR␤/Grb10) identified a 5.6-kb mRNA with broad tissue distribution (Fig. 2B). Variable amounts of the transcript were detected in 15 out of 16 tissues examined. The abundance is highest in skeletal muscle and pancreas and relatively high in cardiac muscle and brain. Intermediate levels were detected in placenta, lung, liver, kidney, spleen, prostate, testis, ovary, small intestine, and colon. Only thymus and peripheral blood leukocytes showed very low or no detectable hGrb-IR␤/Grb10 mRNA. The transcript in brain is shifted slightly upwards, perhaps indicating a variant mRNA. Two additional 4.8-and 3.1-kb transcripts were detected in skeletal muscle, suggesting alternative polyadenylation sites or differential splicing. The reported tissue distribution for hGrb-IR mRNA should be identical because the probes in the studies were derived from common regions of the two cDNAs (7). hGrb-IR was detected primarily in skeletal muscle and pancreas, and there were three transcripts in skeletal muscle (reportedly 6.5, 5.0, and 2.2 kb). The relative abundance of the message in alternative tissues is less clear. The mGrb10 expression pattern is distinct. A single 6.0-kb transcript is predominant in heart and kidney, with lesser amounts detected in brain and lung (9).
Presence of Two Distinct Transcripts Confirmed by RT-PCR-The coexistence of distinct mRNAs for hGrb-IR and hGrb-IR␤/Grb10 was confirmed by reverse transcriptase-coupled PCR (RT-PCR) on polyadenylated mRNA derived from human skeletal muscle (Fig. 3). Two upper primers (A1 and A2) specific for Grb-IR␤/Grb10, and a lower primer (P1) annealing to a region common to both Grb-IR␤/Grb10 and Grb-IR, gave the predicted PCR fragments of 288 and 203 bp, respectively (see Fig. 1B for a projection of primer binding sites on Grb-IR␤/ Grb10 and Grb-IR). This result independently confirms the presence of an mRNA corresponding to 5Ј-untranslated and coding sequences for Grb-IR␤/Grb10. In parallel, an upper primer (B1) specific for Grb-IR and common primer P1 gave the predicted PCR fragment of 231 bp. Therefore, mRNAs with elements unique to Grb-IR and Grb-IR␤/Grb10 coexist in human skeletal muscle. The yield of Grb-IR products was significantly lower than that of the Grb-IR␤/Grb10 products, consistent with the possibility that the Grb-IR␤/Grb10 transcript is more stable or that Grb-IR␤/Grb10 is expressed more abundantly.
A third isoform of the Grb10/Grb-IR family in muscle is indicated by the presence of an additional 450-bp PCR product in the Grb-IR reaction. The primers used for its identification (B1 and P1) suggest that this isoform could have a third Nterminal sequence.
Additional experiments confirmed the domain architecture of hGrb-IR␤/Grb10 and validated our sequencing strategy (Fig.  3). A second lower primer (P2) was designed to anneal mRNA encoding the segment of hGrb-IR␤/Grb10 PH domain that is lacking in hGrb-IR. When used in combination with upper primer A2, the predicted PCR fragment of 954 bp indicates that the hGrb-IR␤/Grb10 mRNA encodes an intact PH domain. The two portions of our complete cDNA sequence were derived from distinct sources; the original insert (HCEEI20) was from a cerebellum cDNA library, and the rest was pulled from skeletal muscle mRNA. An additional RT-PCR experiment with lower primer 7-2 and upper primer A2 gave the predicted PCR fragment of 1461 bp, confirming the presence of appropriate intact transcripts in skeletal muscle.
Grb10 Moves from Cytosol to Membrane Fractions Upon Insulin Stimulation-Many signaling events occur at discrete locations in cells, and modular elements of protein structure domains (e.g. SH3, PH, and SH2 domains) frequently participate in subcellular compartmentalization (1,2,26). HIRc fibroblasts that overexpress human insulin receptors were fractionated to determine the cellular location of mGrb-IR/Grb10 isotypes. In unstimulated HIRc cells the protein is in the cytosol (Fig. 4). However, upon insulin stimulation a remarkable amount of mGrb-IR/Grb10 redistributes to the membrane fraction. This may be through direct binding to the insulin receptor or interactions with additional cellular constituents. While the polyclonal anti-Grb10 antibody (from B. Margolis) recognizes three predominant species, only intact p65 mGrb10/Grb-IR redistributes upon insulin activation. Furthermore, only p65 binds SH3 domains (see below), suggesting that the lower molecular weight immunoreactive species are functionally (and perhaps structurally) unrelated. Control studies show that insulin receptors present in cell membranes do not redistribute. Under more physiologic conditions where fewer receptors are expressed, less of the protein may translocate to the plasma membrane. However, it is clear that insulin receptor activation drives translocation. This may be mediated by one or more of the functional domains within Grb10/Grb-IR.

Grb10/Grb-IR Isoform Binds Activated Insulin Receptors in Fibroblasts-
In initial studies to determine the mechanism of translocation and the role of Grb-IR/Grb10 in signaling, HIRc cells were stimulated with insulin. The cells were lysed and proteins were immunoprecipitated with the anti-Grb10/Grb-IR antibodies. An anti-Tyr(P) immunoblot shows predominant phosphorylation of a 95-kDa protein migrating at the position of the insulin receptor ␤-subunit (Fig. 5A). The protein remains heavily phosphorylated following 5 and 20 min of insulin stimulation. Secondary immunoblotting confirmed that this is the insulin receptor ␤-subunit (Fig. 5B). A 65-kDa protein is transiently phosphorylated upon insulin stimulation. Judging from its immunoreactivity (Fig. 5C) and molecular size, it is probably mGrb10. While its phosphorylation varied between exper-iments, the intensity never approached that of the insulin receptor. These studies confirm that Grb10/Grb-IR proteins interact with activated insulin receptors and are variably phosphorylated (7).
hGrb-IR␤/Grb10 Binds Insulin, EGF, and PDGF Receptors-Differentiated 3T3-L1 adipocytes are known to contain insulin, EGF, and PDGF receptors. The appropriate ligands stimulate autophosphorylation of each receptor (Fig. 6, right  panel). Immobilized hGrb-IR␤/Grb10 binds endogenous, activated insulin receptors (Fig. 6, center), as has been seen previously with Grb-IR (7). Intact hGrb-IR␤/Grb10 also binds activated EGF and PDGF receptors. Therefore, in addition to having structural similarities with Grb-IR and Grb10, hGrb-IR␤/Grb10 has functions common to both proteins (9). The specificity of hGrb-IR␤/Grb10 toward the receptors was further assessed by estimating amounts precipitated by the immobilized protein and comparing this to the amounts of each receptor present in cell lysates. The insulin and PDGF receptors are both efficiently precipitated (Ϸ10% of the total receptor is precipitated). In comparison, the EGF receptor binds in these in vitro pull-downs with significantly lower avidity (Ͻ1% of the total receptor is precipitated). Grb-IR also appears to have a significantly lower affinity for the EGF receptor than the insulin receptor (7). hGrb-IR␤/Grb10 does not appear to bind IRS-1 ( Fig. 6 and data not shown). Due to high backgrounds on the immunoblots, we have not been able to determine whether Grb-IR/Grb10 isoforms co-immunoprecipitate in 3T3-L1 adipocytes with activated insulin, PDGF, or EGF receptors. It has also been difficult to show co-immunoprecipitation of mGrb10 or Grb14 with activated PDGF and EGF receptors, even though these proteins are serine-phosphorylated in EGF-and PDGFstimulated cells (9,10). Grb10 does appear to bind the activated Ret cytoplasmic domain in vivo (12).
Intact hGrb-IR␤/Grb10 Binds More Avidly Than Its Isolated SH2 Domain-To further differentiate potential mechanisms of interaction, comparisons were made using intact hGrb-IR␤/ Grb10 and its isolated SH2 domain. Immobilized hGrb-IR␤/ Grb10 precipitated insulin receptors from lysates of insulinstimulated 3T3-L1 adipocytes, whereas receptors from unstimulated cells were not detected (Fig. 7A). In contrast, the immobilized hGrb-IR␤/Grb10 SH2 domain did not precipitate insulin receptors from these cells. Because many phosphoprotein-binding interactions are mediated by SH2 domains, and the mGrb10 SH2 domain (which is essentially identical to the hGrb-IR␤/Grb10 SH2 domain) binds insulin receptors from transfected cells (27), additional experiments were conducted with lysates from transfected HIRc fibroblasts. The full-length hGrb-IR␤/Grb10 protein was again found to precipitate activated insulin receptors (Fig. 7B). In addition, the immobilized SH2 domain precipitated significantly less but readily detectable amounts of insulin receptor. HIRc cells express considerably greater numbers of receptors per cell than differentiated 3T3-L1 adipocytes. Although higher receptor numbers may facilitate SH2 domain binding in the latter case, it certainly appears that in addition to its SH2 domain other regions of hGrb-IR␤/Grb10 participate in interactions with activated insulin receptors.
Grb-IR␤/Grb10 Binds pp135 via Its SH2 Domain-In addition to binding the activated insulin receptor, hGrb-IR␤/Grb10 and the SH2 domain bind a 135-kDa protein present in 3T3-L1 adipocyte lysates (Fig. 7A). As pp135 was detected in anti-Tyr(P) immunoblots, it is likely to be tyrosine-phosphorylated. Equivalent amounts of pp135 bind the SH2 domain and hGrb-IR␤/Grb10, indicating that this interaction is mediated predominantly by the SH2 domain. Unlike the insulin receptor, pp135 was precipitated from lysates of unstimulated 3T3-L1 adipocytes. Much less was precipitated from insulin-stimulated cells, suggesting that its interaction with hGrb-IR␤/Grb10 decreased following insulin stimulation. This is likely due to a decrease in p135 tyrosine phosphorylation. Initial attempts to characterize the molecular nature of pp135 have failed. It does not immunoreact with antibodies to known proteins of similar size, such as phospholipase C-␥1 and p125 FAK (data not shown). pp135 was not identified in HIRc fibroblasts (Fig. 7B).
Binding Sites for hGrb-IR␤/Grb10 on the Insulin Receptor-At least six tyrosine residues are phosphorylated in activated insulin receptors (3,28). These are located in three discrete regions of the protein: Tyr-960 in the juxtamembrane domain serves as a docking site for substrate protein PTB domains (5, 6, 29 -31), phosphorylation of a cluster of three tyrosines (1146, 1150, and 1151) within the kinase loop is necessary for receptor activation (3, 4, 16, 32, 33), and Tyr-1316 and Tyr-1322 within the C-terminal tail. Biological functions for the latter two sites have been debated (17, 34 -36) and have been proposed to serve as potential docking sites for numerous SH2 domains, including Grb10 (27,37,38). Solubilized and partially purified mutant receptors were used to investigate which of these sites might bind hGrb-IR␤/Grb10. Immobilized hGrb-IR␤/Grb10 bound wild-type receptors and receptors containing a single Tyr-960 3 Phe substitution (Y960F), a deletion of 43 residues at the C terminus including Tyr-1316 and Tyr-1322 (⌬CT), or substitution of all three tyrosines (1146, 1150, and 1151) within the kinase activation loop (YF3) (Fig. 8A). High levels of basal phosphorylation seen here with YF3 receptors have been observed previously with related mutated receptors (32,33). Although insulin receptors were visualized with anti-Tyr(P) antibodies in the experiment shown, similar results have been obtained with anti-insulin receptor antibodies (e.g. Fig. 5 and data not shown). These findings indicate that binding of intact hGrb-IR␤/Grb10 is not abolished by eliminating phosphorylation within any one of the three domains.
Additional studies tested the role of the hGrb-IR␤/Grb10 SH2 domain in mediating this interaction. The isolated SH2 domain precipitated equivalent amounts of activated, solubilized wild-type, Y960F, ⌬CT, and YF3 receptors (Fig. 8B). These results support the conclusion that binding is not abolished by eliminating phosphorylation within any one of the three insulin receptor domains and further demonstrate that the SH2 domain is involved in the interaction. These results suggest that hGrb-IR␤/Grb10 and its SH2 domain might bind the insulin receptor at more than one site, so elimination of any one site should not abolish binding. As an alternative explanation, hGrb-IR␤/Grb10 may bind the receptor at a previously unrecognized phosphorylation site (these exon 11 Ϫ receptors should not be phosphorylated at sites proposed by Webster and colleagues (39)).
Phosphopeptide Competition Studies-To further investigate potential modes of interaction, synthetic phosphopeptides corresponding to each of the major insulin receptor phosphorylation domains were used to compete with interactions between wild-type insulin receptors and either intact hGrb-IR␤/Grb10 or its SH2 domain. The sequences include the monophosphoryl juxtamembrane site surrounding Tyr-960 (pY960), the bisphosphoryl C terminus encompassing Tyr-1316 and Tyr-1322 (pY2CT), and the trisphosphoryl kinase activation loop (pY3Loop). These peptides were designed to bind SH2 domains and each extends at least 5 residues past the N-and C-terminal phosphotyrosine. Nevertheless, the peptides do not block inter- actions between intact hGrb-IR␤/Grb10 and the insulin receptor, even at 1.0 mM concentrations (Fig. 9A). In contrast, SH2 domain binding was inhibited by two of the three peptides (Fig.  9B). Peptide pY3Loop had slightly higher inhibitory potency than pY960, whereas pY2CT had no effect. A previous study suggested that the mGrb10 SH2 domain binds the insulin receptor C terminus (27). Since the SH2 domain sequences of mGrb10 and hGrb-IR␤/Grb10 are essentially identical (Fig. 1), results should be consistent. However, these investigators attempted to attach seven monophosphoryl peptides to Affi-Gel and use these reagents for pull-down studies. Due to large differences in chemical reactivity, levels of covalent peptide attachment typically vary many-fold. These levels were not determined. We conclude that hGrb-IR/Grb10 interacts with the insulin receptor via its SH2 domain and by an additional unidentified mechanism. The SH2 domain has potential for binding at least two sites, as has been shown previously for SH2 domain interactions with the EGF receptor (40). It appears to bind the phosphorylated insulin receptor kinase loop and juxtamembrane region with greatest avidity.
SH3 Domain Binding to the 3BS Site of Grb10 -hGrb-IR␤/ Grb10, hGrb-IR, and mGrb10 have potential SH3 domain binding sites (Fig. 1A). To test whether interactions occur, immobilized SH3 domains from PI 3-kinase p85, c-Abl, and Fyn, and immobilized full-length Grb2 were used to precipitate proteins from Rat1 fibroblast lysates. Immunoblotting with anti-Grb10 antibodies revealed strong binding to the Abl SH3 domain but not the other proteins (Fig. 10). The negative results in these assays suggest that Grb-IR/Grb10 isotypes probably don't bind PI 3-kinase p85, Fyn, or Grb2 via their SH3 domains in cells. While binding the Abl SH3 domain shows that Grb-IR/Grb10 isotypes can bind SH3 domain proteins, these results do not necessarily mean that this particular interaction has physiological relevance. The possibility that Grb-IR/Grb10 isotypes interact with c-Abl or the Abelson oncogene product is a subject for future investigation.
One potential SH3 domain binding site (3BS) is common to Grb7 and Grb14, in addition to the Grb-IR/Grb10 isoforms (Fig.  1A). A peptide corresponding to the 3BS site (SLPAIPNPFPEL) abolished binding to the Abl SH3 domain at a peptide concentration of 100 M (Fig. 10). These findings strongly support the notion that SH3 domain proteins bind Grb-IR/Grb10 isoforms and that 3BS represents a site for binding. DISCUSSION We have identified a variant transcript of the human Grb-IR/Grb10 gene. The encoded protein has high sequence homology with hGrb-IR and mGrb10, although its domain architecture is more similar to that of Grb7 and Grb14. This is because hGrb-IR contains a 46-residue deletion within its PH domain and 58-residue extension at its N terminus, relative to our protein, Grb7 and Grb14. mGrb10 contains an 80-residue insertion, relative to the other proteins. The functions of these insertions, deletions, and extensions are unknown. We have analyzed functions of Grb-IR/Grb10 proteins in vitro and in cells. The protein is present in the cytosol of quiescent cells but translocates to the membrane upon insulin stimulation. This redistribution is likely through interactions with the insulin receptor. Receptor interactions with hGrb-IR␤/Grb10 are mediated in part through the SH2 domain, which binds to the phosphorylated kinase activation loop and the phosphorylated juxtamembrane region. This may explain why overexpression of Grb-IR inhibits insulin signaling (hGrb-IR and hGrb-IR␤/ Grb10 SH2 domains are identical) (7). Binding at juxtamembrane Tyr-960 must block PTB domain interactions and prevent phosphorylation of substrates such as IRS-1 and Shc. Binding at the kinase loop would prevent substrate phosphorylation, as well.
It is also clear from our results that regions outside of the SH2 domain participate in receptor binding. The PH domain may have a role, perhaps analogous to the way tandem PH and PTB domains of IRS proteins participate in receptor recognition (5,6,29,41,42). 3 However, it is difficult to predict the function of PH domains (26). This one is located within a longer (Ϸ320-residue) region of extended homology with the Grb7/ Grb10/Grb14 family and the product of the Caenorhabditis elegans gene mig10 (9). This may suggest that the PH domain is embedded in a larger structure or flanked by one or two other functional domains. It has not been recognized previously that regions outside of the SH2 domains of Grb7, Grb10, Grb14, or Grb-IR participate in their interactions with phosphoproteins.
Grb-IR lacks 40 residues of its apparent PH domain yet interacts avidly with the insulin receptor. This may imply that the PH is not involved. However, the N-terminal Grb-IR extension may compensate by completing the PH domain. Phospholipase C-␥1 contains such a "split" PH domain with its entire SH2-SH2-SH3 domain region inserted between strands ␤3 and 3 S. Dhe-Paganon and S. E. Shoelson, unpublished observations. ␤4 of the PH domain ␤-sandwich. Strands ␤1-␤3 are separated from the remainder of the domain by over 300 residues. Even though Grb-IR is a stable, folded protein, its PH domain appears to be missing strands ␤1-␤3. Since one cannot generally remove large pieces of a stable protein fold without denaturing the protein and altering its physicochemical characteristics, something seems to have taken the place of strands ␤1-␤3. The N-terminal extension would be one obvious candidate, although its sequence does not match the PH domain consensus. 4 Our findings also show selective in vitro interactions between SH3 domains and Grb-IR/Grb10. We identified a high affinity site for c-Abl SH3 domain binding that is common to the known members of the Grb7/Grb10/Grb14/Grb-IR family. While further studies are needed to determine if these proteins bind c-Abl or the Abelson oncoprotein in cells, and which additional SH3 domain proteins bind Grb7/Grb10/Grb14/Grb-IR family members, it is tempting to speculate that these interactions might have potential roles in normal signaling and oncogenesis. Interestingly, the 3BS sequence (SLPAIPNPFPEL) does not conform to the known class I specificity of the Abl SH3 domain: (N)PX⌰X⌿PX⌿P(C) (⌰ and ⌿ represent residues with aromatic and hydrophobic side chains, respectively) (43,44). If the orientation of 3BS is flipped, however, then it fits a class II consensus remarkably well: (C)P⌰XXP⌿XP⌿(N). Several SH3 domains (e.g. Src and Grb2) bind polyproline peptides in two orientations (43)(44)(45). While Abl has not been shown to do so previously, our data suggest that it can. The high resolution structure of the liganded Abl SH3 domain supports this possibility (46). 5 While this manuscript was being reviewed a related study was published (47). The two studies agree on most points. The sequences and domain architectures of the two predicted proteins are identical, and both studies show prominent expression of hGrb-IR␤/Grb10 mRNA in skeletal muscle and pancreas, an activation-dependent interaction of Grb-IR␤/Grb10 with the insulin receptor, and a decrease in binding to pp135 with insulin stimulation. Both studies also conclude that the activation loop of the insulin receptor kinase is a primary site of binding to the hGrb-IR␤/Grb10 SH2 domain. We report an additional interaction with the juxtamembrane region of the receptor, but our results are not in conflict. 6 O'Neill et al. (47) demonstrated interactions of Grb-IR␤/Grb10 with insulin-like growth factor-1 receptors, whereas we showed binding to activated PDGF and EGF receptors. They showed that the mitogenic effects of insulin and insulin-like growth factor-1 were blocked by microinjecting the Grb-IR␤/Grb10 SH2 domain into fibroblasts. We did not study this effect. Additional findings that we report include the Northern analyses of hGrb7, the presence of distinct transcripts encoding Grb-IR and hGrb-IR␤/ Grb10, the insulin-induced translocation of Grb-IR␤/Grb10 from cytosol to membrane, the interaction of hGrb-IR␤/Grb10 with the insulin receptor at a site outside of its SH2 domain, the interaction of Grb-IR/Grb10 with selected SH3 domains, and the demonstration that the conserved 3BS sequence binds SH3 domains.
Many SH2 domain proteins are enzymes, such as the Src, Abl, and ZAP70/Syk kinases, the SHP1/SHP2 phosphatases, the phospholipase C-␥1, and RasGAP. Others associate with enzymes such as the p85 subunit of PI 3-kinase and Grb2 and Crk binding to guanyl nucleotide exchange factors Sos and C3G, respectively. Still other SH2 domain proteins like Shc and Shb act as substrates of receptor tyrosine kinases. The identification of intrinsic or associated activity provides valuable clues about function, but these are lacking for the entire Grb7 family of proteins. Potential clues to Grb-IR/Grb10 functions should be provided by cellular studies. Stable expression of Grb-IR inhibited insulin signaling in Chinese hamster ovary cells, as assessed by substrate phosphorylation and associated PI 3-kinase activity (7). In contrast, microinjection of its SH2 domain (a potential dominant-negative) inhibits DNA synthesis in fibroblasts, suggesting that Grb-IR/Grb10 is a positive mediator of mitogenesis (47). Although Grb-IR and Grb-IR␤/ Grb10 differ in domain structure and may have distinct or even opposing biological functions, it is not easy to reconcile these findings. There also appears to be a third related transcript in skeletal muscle, suggesting that yet another form of the protein may be involved in signaling downstream from insulin and possibly growth factor receptors. Additional studies are needed to elucidate functions of this intriguing family of proteins.