The inositol 5'-phosphatase SHIP binds to immunoreceptor signaling motifs and responds to high affinity IgE receptor aggregation.

Immunoreceptors such as the high affinity IgE receptor, FcεRI, and T-cell receptor-associated proteins share a common motif, the immunoreceptor tyrosine-based activation motif (ITAM). We used the yeast tribrid system to identify downstream effectors of the phosphorylated FcεRI ITAM-containing subunits β and γ. One novel cDNA was isolated that encodes a protein that is phosphorylated on tyrosine, contains a Src-homology 2 (SH2) domain, inositolpolyphosphate 5-phosphatase activity, three NXXY motifs, several proline-rich regions, and is called SHIP. Mutation of the conserved tyrosine or leucine residues within the FcεRI β or γ ITAMs eliminates SHIP binding and indicates that the SHIP-ITAM interaction is specific. SHIP also binds to ITAMs from the CD3 complex and T cell receptor ζ chain in vitro. SHIP protein possesses both phosphatidylinositol-3,4,5-trisphosphate 5′-phosphatase and inositol-1,3,4,5-tetrakisphosphate 5′-phosphatase activity. Phosphorylation of SHIP by a protein-tyrosine kinase, Lck, results in a reduction in enzyme activity. FcεRI activation induces the association of several tyrosine phosphoproteins with SHIP. SHIP is constitutively tyrosine-phosphorylated and associated with Shc and Grb2. These data suggest that SHIP may serve as a multifunctional linker protein in receptor activation.

The aggregation of immunoreceptors by antigen initiates a complex response leading to cellular activation (1). Receptors on T-, B-, and mast cells each contain subunits with similar primary amino acid sequence within their cytosolic domains, comprising the immunoreceptor-based tyrosine activation motif (ITAM), 1 whose consensus is (D/E)X 2 YXX(L/I)X 6 -8 YXX(L/I) (2,3). Both tyrosine residues within the ITAM are rapidly phosphorylated by protein kinases after receptor aggregation. The bisphosphorylated ITAM then binds directly to cytosolic tyrosine kinases such as Syk in B-cells and mast cells and ZAP70 in T-cells, thereby activating their tyrosine kinase activity (4,5). In mast cells, the Fc⑀RI subunits ␤ and ␥ each possess a single ITAM, which, when bisphosphorylated on tyrosine, binds to Syk (6 -8).
We have used a novel genetic approach, the yeast tribrid system (8), to isolate cDNAs that encode proteins that interact with the tyrosine-phosphorylated Fc⑀RI ␥ ITAM. The yeast two-hybrid system facilitates the study of protein-protein interactions but is limited to the investigation of proteins that are properly expressed and modified in the host, Saccharomyces cerevisiae. S. cerevisiae does not employ tyrosine phosphorylation as a major regulatory modification of proteins (9,10). This limits the utility of the two-hybrid system, especially in the area of signal transduction, where tyrosine phosphorylation is a critical component of the process. In order to study protein-protein interactions that are dependent on tyrosine phosphorylation or on other post-translational or allosteric modifications, the yeast tribrid system was developed (8,11).
In the yeast tribrid system, a third plasmid is introduced, which directs the synthesis of a protein-tyrosine kinase, which catalyzes the phosphorylation of a tyrosine-containing protein, facilitating the interaction with an SH2 domain-containing protein. The interaction is detected through the use of a transcriptional activation assay, as is used in the two-hybrid system (12). The tribrid system lends itself to a variety of studies, including the characterization of phosphotyrosine-SH2 domain interactions, identification of ligands for SH2 domains, cloning of novel protein-tyrosine kinases, and the identification of novel SH2 domain-containing proteins, to name a few. In this study, the tribrid system was used to isolate novel cDNAs that encode proteins that interact with the tyrosine-phosphorylated Fc⑀RI ␥ chain.
One of the cDNAs isolated encodes SHIP, a novel ITAMbinding protein that is part of the Fc⑀RI activation pathway, as revealed by its association with several additional tyrosine phosphoproteins following Fc⑀RI stimulation. SHIP also forms complexes with the signaling molecules Shc and Grb2 in vivo. The cloned gene product is a phosphatidylinositol-3,4,5trisphosphate and inositol-1,3,4,5-tetrakisphosphate 5-phos-phatase. SHIP associates with tyrosine-phosphorylated ITAMs in the yeast tribrid system and with tyrosine-phosphorylated ITAM peptides in vitro.

EXPERIMENTAL PROCEDURES
Commercially Available Reagents-Monoclonal antibodies recognizing Shc and glutathione S-transferase fusion proteins were purchased from Santa Cruz Biotechnology. The anti-Grb2 monoclonal antibody and anti-RasGAP polyclonal antibody were obtained from Upstate Biotechnology, Inc. Antibodies to phospholipase C␥-1, Nck, Sos1, and Sam-68 were purchased from Transduction Laboratories. Peroxidaseconjugated anti-rabbit and anti-mouse secondary antibodies were obtained from Jackson Immunoresearch.
Isolation of SHIP-A cDNA library from RBL-2H3 cells (8) was screened using LexA-Fc⑀RI ␥ as bait (p4108; Ref. 8) in host strain S-260 as described (8). Three overlapping clones were identified (p4187, p4193, and p4195), and the DNA sequences were determined. The 3Ј-end was recovered by screening a ZAP cDNA library with a fragment of p4187. The 3Ј-end was then amplified from RBL-2H3 RNA by reverse transcription-PCR (Life Technologies, Inc.) using specific primers that hybridized to the coding region within p4187 and the 3Јuntranslated region. The fragment amplified was ϳ1-kilobase pair longer than that recovered from the clone. After subcloning into pCRII (Invitrogen), p4187 and the PCR-amplified fragment were ligated after KasI/XhoI digest to generate p4802. These sequence data are available from GenBank TM under accession number U55192. We refer to this cDNA as SHIP.
Plasmid Constructions-Two-step PCR was used to construct mutant LexA-ITAM fusions in a manner identical to that described previously (8). The T-cell receptor chain was PCR-amplified from peripheral blood lymphocyte cDNA. 2 The p4187 (⌬SH2) plasmid was generated by PCR and contains nucleotides 885-2570 of the complete SHIP sequence (encoding amino acids 219 -790). GST-SHIPN was generated by PCR and encodes amino acids 4 -356 of SHIP joined to glutathione S-transferase (pGEX3-X; Pharmacia Biotech Inc.). GST-Vmw65 was generated by restriction endonuclease cleavage of p4064 (8) with BamHI and EcoRI and ligated to pGEX 4T-1 (Pharmacia). The fusion proteins were purified from Escherichia coli strain MC1061 according to the manufacturer's instructions.
Immunoblotting-Expression of all LexA and Vmw65 fusion proteins were confirmed by induction of yeast cells bearing the appropriate plasmids and processing for immunoblot after SDS-PAGE as described (8). Anti-SHIP immunoblots were probed with 1 g/ml of affinity-purified polyclonal antibodies. All commercially available antibodies were used at the concentrations recommended by the manufacturer.
Northern Blotting-A rat multiple tissue Northern blot of poly(A) ϩ mRNA was purchased from Clontech and probed with the 32 P-labeled insert recovered from p4187 after restriction endonuclease cleavage with EcoRI and XbaI. All steps were carried out according to the supplier's instructions.
Antibody Purification-Antibodies were raised to GST-SHIPN in rabbits (Pocono rabbit farm, Canadensis, PA), and IgG was affinitypurified by passage over a GST-Vmw65 column to remove GST-reactive antibodies, followed by passage over immobilized GST-SHIPN in a manner similar to that described previously (8).
Peptide Binding Assays-ITAM-containing biotinylated 23-mer carboxamide peptides (Table I) were prepared as described (13). Each peptide (10 g/ml in 50 l of 50 mM Tris pH 7.5, 100 mM NaCl (TBS) was incubated with 60 liters of streptavidin-agarose beads (Sigma) with rotation for 1 h at 4°C. After washing with TBS, the beads were preincubated with 4% bovine serum albumin in TBS for 15 min, washed three times with TBS, and incubated with 0.25 g of purified GST-SHIPN in 200 l of TBS containing 1% Nonidet P-40 and 10% glycerol for 1 h with rotation. The beads were washed three times and then boiled in 1 ϫ Laemmli sample buffer and analyzed by immunoblot after separation by SDS-PAGE and immunoblotting with anti-GST antibodies (Santa Cruz Biotechnology). All immunoblots were developed using horseradish peroxidase-coupled secondary antibodies and enhanced chemiluminescence (Amersham Corp.).
Immunoprecipitation-RBL-2H3 monolayers were incubated with biotinylated rat IgE (a gift of M. Basu, Roche) at 1 g/ml for 30 min in Iscove's modified Dulbecco's medium, 10% fetal calf serum, the cells were rinsed twice with medium, and the receptors were aggregated with 10 g/ml egg white avidin for 10 min. The cells were rinsed twice with phosphate-buffered saline and lysed in 1 ml of 50 mM Tris pH 8, 150 mM NaCl, 1 mM Na 3 VO 4 , 10 mM NaF, and protease inhibitors aprotinin, leupeptin, and pepstatin A at 1 g/ml. Clarified extracts were incubated with unrelated rabbit serum for 30 min at 4°C and precipitated with protein G-Sepharose (Pharmacia), followed by incubation with 2 g of affinity-purified rabbit anti-SHIP antibodies for 1 h. Control immunoprecipitations utilized 2 g of affinity-purified anti-GST antibodies. Immunoprecipitates were adsorbed to protein G-Sepharose (Pharmacia) for 1 h, washed with buffer, boiled in Laemmli sample buffer, and electrophoresed on a 4 -20% SDS-PAGE.
GST Fusion Protein Chromatography-Extracts of RBL-2H3 cells that were treated with biotinylated IgE and egg white avidin or left untreated were prepared as above and incubated with 1 g of purified GST fusion proteins (Santa Cruz Biotechnology) for 1 h at 4°C. Glutathione-Sepharose beads (10 l) were added, and the incubation continued for an additional 30 min to 1 h. The beads were collected by centrifugation, washed in buffer, and boiled in sample buffer. The samples were then subjected to electrophoresis on a 4 -20% SDS-polyacrylamide gel, transferred to polyvinylidene difluoride or nitrocellulose, and processed for immunoblotting with affinity-purified anti-SHIPN antibodies (1 g/ml).
Inositolpolyphosphate 5-Phosphatase Activity-S. cerevisiae strain S-260 (8) was transformed with Vmw65 fusion protein plasmids p4187 (the original SHIP isolate from the tribrid screen, terminating at amino acid 790) or p4802 (encoding the full-length SHIP terminating at amino acid 1190) with p4140 (encoding Lck; Ref. 8) or pRS415 (vector for p4140). Cells were grown overnight in liquid Ura Ϫ Leu Ϫ medium containing 2% glucose. The next morning, the cells were centrifuged, washed with H 2 O, and diluted to an A 600 of 0.3 in 50 ml of Ura Ϫ Leu Ϫ liquid containing 2% galactose as the sole carbon source. After 4 h, the cells were harvested by centrifugation, enzymatically treated to generate spheroplasts with zymolyase and glusulase, and lysed in lysis buffer (10 mM imadizole, pH 7.2, 1 mM MgCl 2 , 1 mM EDTA, 0.3 M sucrose, 1% Triton X-100). After clarification by centrifugation, an unrelated rabbit antibody was added for 30 min, followed by protein G-Sepharose beads (Pharmacia). After centrifugation, anti-GST-Vmw65 antiserum raised against a GST fusion protein containing amino acids 410 -490 of HSV1 Vmw65 protein was added, followed by protein G-Sepharose beads. The beads were incubated for 1 h with the antibody/extract mixture and then washed twice with lysis buffer, twice with lysis buffer lacking detergent, and then twice with 5Ј-phosphatase assay buffer (50 mM Tris, pH 7.5, 3 mM MgCl 2 ). The beads were aspirated to dryness and stored frozen at Ϫ70°C until assayed.
Phosphopeptide Specificity of SHIP SH2 Domain-The specificity of the SHIP SH2 domain was studied using the peptide library GDGpYX-XXSPLLL, where pY indicates phosphotyrosine and X indicates any amino acids except for tryptophan and cysteine (19). Approximately 0.3 mg of degenerate peptide library was incubated with 200 -300 mg of bead-immobilized GST-SHIPN. Peptide purification, sequencing, and data analysis were as described previously (19).

RESULTS
Isolation and Analysis of SHIP cDNA-We utilized the yeast tribrid system in order to identify proteins that interact with the activated Fc⑀RI ␥ chain. A LexA-Fc⑀RI ␥ fusion protein (containing the 42 C-terminal amino acids of the Fc⑀RI ␥ chain) was used as a "bait," the protein-tyrosine kinase Lck was used to phosphorylate Fc⑀RI ␥ on tyrosine, and a Vmw65 activation domain fusion-cDNA expression library from the mast cell line RBL-2H3 was screened in S. cerevisiae strain S-260 for positive interactors. Of 500,000 colonies screened, five were identified that required the tyrosine kinase Lck for interaction with the LexA-␥ fusion protein and did not interact with other negative control baits such as LexA-lamin C, LexA-SNF1, and LexA-SIR4. Three of these clones contained similar sequences and were characterized further. When other LexA-ITAM fusions were used as baits, the cDNAs identified were found to interact with both the Fc⑀RI ␤ and ␥ subunits, as well as the T-cell receptor chain (Table I).
The three LexA-␥ interactors contained overlapping sequences encoding a novel protein (a schematic diagram is shown in Fig. 1). All three cDNAs had similar 5Ј-end points but differed in their 3Ј-end points (p4187 terminating at nucleotide 2570, p4193 at 1778, and p4195 at 1211, respectively). The cDNAs had open reading frames extending throughout the length of the insert. The 3Ј-end of the transcript was cloned by screening a library from RBL-2H3 cells with a 3Ј-fragment of the longest cDNA and assembled with reverse transcription-PCR-amplified cDNA from RBL-2H3 cells. The 5Ј-end was cloned by screening a similar cDNA library. The entire clone encompassed 4.9 kilobase pairs and encoded a protein of 1190 amino acids, with a predicted molecular weight of 133,000. The 5Ј-end contains an in-frame stop codon, suggesting that there is no additional coding sequence in this region. 3Ј to the TGA stop codon are several other stop codons in all three reading frames.
A comparison of the predicted peptide sequence of the cDNA to nucleic acid and protein sequence data bases revealed extensive similarities to two classes of proteins. The first region, amino acids 7-110, displayed 38% identity to Csk and other SH2-containing proteins of the Src family (group I; Refs. 20 and 21). The second region (amino acids 270 -715) is similar (ϳ40% identical) to a class of proteins possessing inositol-5-phosphatase activity. These proteins display a variety of substrate specificities, but all are able to hydrolyze the 5Ј-phosphate group from a particular species of inositol phosphate or inositol phosphlipid (14 -16, 22-24). The C-terminal half of the protein contains numerous tyrosine residues (14 between amino acids 558 and 867) and three NXXY motifs (at amino acids 555, 914, and 1017). NXXY motifs of other proteins, when phosphorylated on the tyrosine residue, have been shown to bind to phosphotyrosine-binding (PTB) domains in a manner different from that of SH2 domains (25)(26)(27)(28). The C terminus is also very proline-rich and has several regions that (by visual inspection) could make good Src-homology domain 3 (SH3) ligands. We refer to this cDNA as SHIP.
The tissue distribution of rat SHIP mRNA was determined by screening a Northern blot of poly(A) ϩ mRNA with the longest SHIP clone identified in the tribrid screen (Fig. 2). Lung and spleen tissue displayed strongly reactive bands at approximately 4.9 kilobase pairs, with lesser amounts present in all tissues tested (heart, brain, skeletal muscle, kidney, and testis). Extremely low mRNA levels were observed in liver tissue.
The SHIP SH2 Domain Binds to ITAMs-The fact that all three cDNAs isolated in the screen had similar 5Ј-end points suggested that the N terminus contained a region of the protein that was important for binding to the LexA-ITAM fusion protein. The shortest of these cDNAs identified (p4195) contained amino acids 1-338, excluding the possibility that any portion of the protein distal to amino acid 338 is required for an interaction to be detected (see Fig. 1). The most obvious region of SHIP that could facilitate the interaction with the phosphorylated LexA-ITAM fusion is the SH2 domain. To test this hypothesis, a deletion mutant of the longest of the three clones was constructed to remove just the SH2 domain. This SHIP(⌬SH2)- Vmw65 fusion protein (containing amino acids 219 -790) was then tested in the tribrid system for its ability to interact with the LexA-Fc⑀RI ␥ bait. As shown in Table I, removal of the SH2 domain results in the loss of blue colony color in the yeast tribrid assay, indicating a loss of interaction. This result demonstrated that the SH2 domain was essential for the interaction to be observed. To further characterize the interaction between the SHIP and the LexA-ITAM baits, conservative mutations were generated within the ITAM sequences. As shown in Table I, substitution of the conserved N-terminal tyrosine or leucine residue within the ITAM has no effect on the ability of SHIP to interact with LexA-␥ or LexA-␤. However, mutation of the conserved C-terminal tyrosine or leucine in the ␥ ITAM abolishes the interaction, while mutation of the C-terminal tyrosine in the ␤ ITAM reduces the interaction. Therefore, the C-terminal half of the ITAM, specifically both the tyrosine and leucine residues within the YXXL, are important for an interaction to be observed between SHIP and the ␤ or ␥ ITAMs. Consistent with these results, mutations of all six tyrosine residues within the T cell receptor chain results in elimination of SHIP binding. Mutation of two of the six tyrosines, eliminating any one of the three ITAMs, does not eliminate binding, suggesting that all three ITAMs are equally able to bind to SHIP in the tribrid system. The ability of SHIP to associate with ITAMs was further explored using synthetic peptides. Several biotinylated bis-, mono-, and nonphosphorylated peptides representing a number of different ITAMs derived from the CD3 ␥, ␦, and ⑀ chains and T-cell receptor-associated chains were synthesized and tested (Table II). A GST-SHIP (amino acids 4 -356) fusion protein containing the SH2 domain was incubated with the peptides, and proteins specifically bound were resolved by SDS-PAGE and detected with anti-GST antibodies. The nonphosphorylated ITAM peptides were unable to bind to the GST-SHIP fusion protein (Fig. 3). All of the phosphorylated ITAM peptides, including the monophosphorylated 3 peptide, were able to interact with the GST-SHIP protein. There was some differential binding to the various peptides, with the 1 and 2 peptides displaying the better ability to bind the GST-SHIP protein under the experimental conditions.
The SH2 domain of SHIP can bind to ITAMs, since direct peptide selection by the GST-SHIP-(4 -356) fusion protein indicated a preference for the motif Y(Y/D)X(L/I/V), in general agreement with the ITAM consensus of YXXL (Fig. 4).
SHIP Associates with Signaling Proteins in RBL-2H3 Cells-To determine the association of SHIP with other proteins in RBL-2H3 cells, extracts from untreated or Fc⑀RI-stimulated cells were immunoprecipitated with affinity-purified rabbit anti-SHIP antibodies. As shown in Fig. 5, proteins of 145, 45, and 52 kDa are observed co-precipitating with SHIP (but not with control antibody) in untreated cells. Longer exposures of the same gel reveal a protein of 90 kDa that is also specifically precipitated by anti-SHIP antibodies. After Fc⑀RI aggregation, several additional phosphorylated proteins of 30 -40 kDa and ϳ60 kDa are present in addition to the 145-, 52-, and 45-kDa proteins. The 145-kDa species co-migrates with SHIP (Fig. 5, anti-SHIP blot), and reprobing of this blot confirms this (data not shown), indicating that SHIP is itself a tyrosine phosphoprotein.
To determine whether any other proteins implicated in signal transduction were associated with SHIP, identical immunoprecipitates were probed with antibodies to Shc and Grb2. As shown in Fig. 5 (anti-Shc blot) the 45-and 52-kDa bands observed in the anti-phosphotyrosine blot include Shc, since monoclonal anti-Shc antibodies recognize species of this band. In addition, immunoprecipitation with anti-Shc followed by anti-SHIP immunoblotting reveals an association with SHIP (Fig. 5, anti-SHIP blot). In addition to Shc, Grb2 is also present in anti-SHIP immunoprecipitates (Fig. 5, anti-Grb2 blot), although Grb2 is not tyrosine-phosphorylated (compare anti-Grb2 blot with anti-phosphotyrosine blot). Identical immunoblots probed with antibodies to phosphoinositide 3Ј-kinase, Sam-68, Sos1, Nck, RasGAP, and phospholipase C␥-1 showed no evidence of co-immunoprecipitation of these proteins with SHIP (data not shown).
Immunoblot analysis of SHIP immunoprecipitates using an-   Table  II tibodies specific for the Fc⑀RI ␤ and ␥ did not reveal any association with either of these subunits, although tyrosine phosphoproteins were detected in these samples. Similarly, immunoprecipitation of extracts from unstimulated and stimulated RBL-2H3 cells with anti-Fc⑀RI-␤ or anti-Fc⑀RI-␥ anti-bodies did not reveal any SHIP associated with the Fc⑀RI (data not shown).
In order to further explore the interaction of SHIP and Grb2, GST fusion proteins to a variety of SH2 and SH3 domaincontaining proteins were used as affinity reagents to determine the spectrum of potential SHIP interactors. As shown in Fig. 6, the GST-Grb2 (SH3SH2SH3) and GST-phospholipase C␥ (SH2SH2SH3) fusion proteins specifically precipitate SHIP from RBL-2H3 extracts of both unstimulated and stimulated cells. GST fusion proteins containing the N-terminal SH2 domain of phosphoinositide-3-OH-kinase (PI 3-kinase), the SH2 domain of SHIP, and SH2 domains of SH-PTP2 (data not shown) did not display any binding to SHIP. A GST-Grb2 SH2 domain fusion and a GST-phospholipase C␥-1 SH2SH2 domain fusion do not precipitate SHIP, indicating that the interaction with these proteins is not mediated by an SH2 domain. In addition, a GST-phospholipase C␥-1 SH3 domain fusion protein does precipitate SHIP, suggesting that SHIP is associating with SH3 domains.
SHIP Is an Inositol-5-phosphatase-To test whether SHIP possessed inositol-5Ј-phosphatase activity, as was suggested by its similarity to other proteins, SHIP was expressed in S. cerevisiae. Immunoprecipitates from S. cerevisiae bearing either the original SHIP plasmid (encoding amino acids derived from the 5Ј-untranslated region through amino acid 790) or a reconstituted full-length clone (encoding amino acids from the 5Ј-untranslated region through amino acid 1190) were incubated with inositol 1,3,4,5-tetrakisphosphate or phosphatidylinositol 3,4,5-trisphosphate, and the products were analyzed. The full-length SHIP protein contains 5Ј-phosphatase activity (Fig. 7, A and B, column 4), while the truncated form does not (Fig. 7, A and B, column 2). No enzymatic activity was observed when inositol 1,4,5-trisphosphate was used as a substrate (data FIG. 4. Ligand specificity of the SHIP SH2 domain. Boxes A, B, and C show, respectively, the amino acid selectivity at positions ϩ1, ϩ2, and ϩ3 C-terminal to the phosphotyrosine. The degenerate phosphopeptide library GDGpYXXXSPLLL (19) was added to a GST-SHIPN or GST column. The columns were washed, and bound peptides were eluted with phenyl phosphate. The eluted peptide mixture from a GST-SHIP column was subjected to microsequence analysis, and results were compared with those from the eluate of a control GST column. The value represents the ratio of the amount of each amino acid eluted from GST-SHIPN columns divided by that of the control GST columns at the same cycle. The data were normalized such that a value of 1 or less indicates no selectivity for a given amino acid. To test the effect of tyrosine phosphorylation on SHIP enzymatic activity, a similar experiment was repeated with immunoprecipitates from yeast expressing SHIP in the presence or absence of the tyrosine kinase Lck. Co-expression of the tyrosine kinase Lck with SHIP results in the tyrosine phosphorylation of SHIP (data not shown) and a 2-3-fold reduction in the level of 5Ј-phosphatase activity (Fig. 7, A and B, compare lane  1 with lane 2 and lane 3 with lane 4). Similar amounts of SHIP were immunoprecipitated from both preparations (data not shown). The results clearly demonstrate a loss of 5Ј-phosphatase activity when SHIP is tyrosine-phosphorylated. Control immunoprecipitations (Fig. 7B, lanes 3c and 4c) display a small amount of associated 5Ј-phosphatase activity. This could be due to nonspecific binding of SHIP to the antibody beads, due to the large overexpression of SHIP from the pGAL promoter. DISCUSSION The molecular details of signaling through the IgE receptor are still being defined. To unravel the underlying mechanisms involved, it seems crucial to identify the molecules participating in the Fc⑀RI signal transduction cascade. To this end, we have employed a genetic strategy, the yeast tribrid system, to identify proteins that interact with the activated Fc⑀RI. We have identified one cDNA, encoding SHIP, a 145-kDa tyrosine phosphoprotein that specifically recognizes individual phosphotyrosine motifs within the Fc⑀RI ITAMs.
SHIP binds to a variety of ITAMs, both in the yeast tribrid system (Table I) and in vitro (Table II). The interaction is dependent on the tyrosine phosphorylation of the ITAM and is not due to nonspecific phosphotyrosine binding, since mutation of the conserved leucine residue within the YXXL abolishes interaction with SHIP, just as it does for ZAP70 (29). Direct peptide binding studies reveal that the SH2 domain of SHIP binds to the CD3 complex ␥, ␦, and ⑀ and T cell receptor chain phospho-ITAM peptides tested, while nonphosphorylated peptides do not. In addition, the results obtained with the yeast tribrid system clearly demonstrate a specificity for SHIP binding to the ITAM motif YXXL. Within the ITAM, SHIP displays preferential binding to the C-terminal phosphotyrosine of the Fc⑀RI ␤ and ␥ ITAMs ( Table I). The optimal ligand for the SHIP SH2 domain, Y(Y/D)X(L/I/V) is consistent with ITAM-binding ability. However, there may be other SHIP ligands present in the cell, since ITAMs typically do not contain an aromatic residue C-terminal to the phosphotyrosine (30). These results suggest a potential mechanism for SHIP binding to the hemiphosphorylated ITAM and participating in the IgE response in a manner distinct from that of Syk, which only binds tightly to doubly phosphorylated ITAMs.
SHIP is tyrosine-phosphorylated and bound to Shc in both resting and Fc⑀RI-stimulated RBL 2H3 cells (Fig. 5). Li and co-workers (31) noted a 145-kDa protein that acquires additional phosphate in response to Fc⑀RI aggregation. Our experiments cannot exclude the possibility that SHIP phosphorylation is increased following receptor stimulation. However, the high basal state of SHIP tyrosine phosphorylation would make this difficult to detect. Co-immunoprecipitating with SHIP are the p46 and p52 forms of the signaling adaptor protein Shc, which is also basally tyrosine-phosphorylated in RBL-2H3 cells (32), unlike other cell types (33,34). Shc contains two phosphotyrosine-binding motifs, an SH2 domain, and a PTB domain (25,35,36). SHIP may thus serve as a bridge between the ITAM-containing T cell receptor and Shc via a larger linking complex.
Also present in anti-SHIP immunoprecipitates is the adaptor protein Grb2 in both resting and Fc⑀RI-activated RBL-2H3 cells. Grb2 is typically bound to Sos, which serves as a GTP exchange factor for Ras. Although GST-Grb2 fusion proteins are able to precipitate Sos (32) from RBL-2H3 cells, anti-SHIP immunopreciptates containing Grb2 do not contain Sos. This may be a reflection of the sensitivity of the reagents available, since others have had difficulty in observing Sos-Shc co-immunoprecipitation (38). Alternatively, the Grb2-SHIP association may be regulatory in nature, as a means of dissociating the Grb2-Sos signaling complex. GST-Grb2 fusion proteins bind to SHIP, and this interaction requires more than the SH2 domain of Grb2, since the GST-Grb2 SH2 domain fusion protein does not precipitate SHIP (Fig. 6). The proline-rich region of the C terminus of SHIP is an excellent candidate for an SH3 ligand and may serve as a binding site for the Grb2 SH3 domain.
In RBL-2H3 cells, Shc, Grb2, and SHIP are constitutively associated with each other, while in the other reports (17, 18,   FIG. 7. SHIP is a phosphatidylinositol 3,4,5-trisphosphate-5phosphatase and an inositol 1,3,4,5-polyphosphate-5-phosphatase that is inhibited by tyrosine phosphorylation. A, phosphatidylinositol-3,4,5-trisphosphate-5-phosphatase activity of SHIP. Immunoprecipitations were carried out as described under "Experimental Procedures." Lane 1, SHIP truncated at amino acid 790, co-expressed with Lck; lane 2, SHIP truncated at amino acid 790, without Lck; lane 3, full-length SHIP, co-expressed with Lck; lane 4, full-length SHIP, without Lck. 1c, 2c, 3c, and 4c refer to immunoprecipitations with control antibodies from the same cells used for the experimental immunoprecipitations (lanes 1-4). The bottom part of the figure is an autoradiogram of the TLC plate displaying separation of the different phosphatidylinositol phosphates. The top is a histogram generated by densitometric scanning of the TLC plate. B, inositol-1,3,4,5-tetrakisphosphate-5Ј-phosphatase activity of SHIP. S. cerevisiae immunoprecipitates (the same as described in A) were assayed for enzymatic activity as described under "Experimental Procedures." Labeling of the histogram is the same as for A. 40), the association between Shc and SHIP is only detected after stimulation with growth factors. These observations may reflect a fundamental difference between RBL-2H3 cells and other cell types. When all data are considered together, a more complete story of SHIP activity emerges. Damen et al. (17) used the Grb2 N-terminal SH3 domain to purify mouse SHIP, and Kavanaugh et al. (18) used a strategy based on expression cloning via Grb2-binding ability to clone cDNAs encoding a class of proteins they term SIP. It thus appears that Grb2 binding is mediated by one (or more) of the polyproline stretches at the C terminus of SHIP, consistent with the observation that Grb2-SHIP binding is not detected with the Grb2 SH2 domain alone (Fig. 6). Furthermore Kavanaugh et al. (18) showed that binding of SIP to Grb2 required both SH3 domains but not the SH2 domain. Lioubin et al. (41), using a strategy similar to ours to identify Shc PTB domain ligands, showed that the region of SHIP containing the NPNY motif (at amino acid 914) was sufficient for Shc binding.
SHIP is basally tyrosine-phosphorylated in RBL-2H3 cells, in agreement with other reports (17,18,40). We demonstrate that Fc⑀RI stimulation results in SHIP association with several additional tyrosine phosphoproteins, which is not observed in growth factor-stimulated cells (17,18,40). These results demonstrate that SHIP can associate with receptor-activated proteins and that SHIP may play a role in the signaling events associated with Fc⑀RI activation.
Other investigators have observed a tyrosine phosphoprotein of 145 kDa (termed pp145) in a complex with tyrosine-phosphorylated Shc in cells treated with cytokines or stimulated through Fc receptors (33,34,38,(42)(43)(44)(45)(46). In B cells and macrophages, both IL-4 and antigen induce the association of Shc with pp145. It remains to be proven whether all of the 145-kDa phosphoproteins described in these experiments are identical, but their behavior suggests a common mechanism for association of Shc with SHIP in response to receptor activation.
SHIP is an inositol-5Ј-phosphatase that will hydrolyze phosphatidylinositol 3,4,5-trisphosphate in addition to the soluble inositol 1,3,4,5-tetrakisphosphate. The 5-phosphatase activity observed in this protein does not require tyrosine phosphorylation, since immunoprecipitations from S. cerevisiae (which has few tyrosine phosphoproteins) not expressing Lck do not interfere with its activity. Although it is difficult to quantify, the observation that the specific activity of SHIP is reduced when tyrosine-phosphorylated suggests that receptor activation (and concurrent SHIP phosphorylation) would lead to a decrease in SHIP activity.
SHIP is able to dephosphorylate phosphatidylinositol 3,4,5trisphosphate (PIP3), the product of PI 3-kinase, to phosphatidylinositol 3,4-bisphosphate (PIP2). PIP2 has been reported to stimulate protein kinase C isozymes (47), and SHIP activity would result in an increase in PIP2. The observation that SHIP is negatively regulated by tyrosine phosphorylation would be consistent with PI 3-kinase activation in response to receptor activation. Since SHIP activity is decreased, the 5-phosphatase may play a negative regulatory role in receptor activation, similar to protein-tyrosine phosphatases. Such a mechanism would lead to a synergistic increase in the levels of PIP3, since PI 3-kinase is activated and 5-phosphatase activity is inhibited. Given the importance of PI 3-kinase activation in cell signaling in RBL-2H3 cells (48), SHIP may function as a negative regulator of PI 3-kinase. Since SHIP is able to bind to hemiphosphorylated ITAMs, Fc receptor activation would provide a docking site for SHIP, facilitating its translocation to the membrane, where the substrate for its enzymatic activity is found. Its association with Grb2, but not with Sos, suggests that it may serve to sequester Grb2 during the signal shut-off period.
The recent report that the PI 3-kinase binds to PIP3 directly via its SH2 domain (21) also raises the possibility that SHIP, by virtue of its PIP3 to PIP2 catalytic activity, may act to alter the pattern of SH2 domain-containing proteins with the activated receptor complex at the membrane. The observation that SHIP will bind to the SH3 domain of phospholipase C␥-1 (Fig. 6) suggests that SHIP may allow SHIP to regulate or be regulated by phospholipase C activation and/or localization. The involvement of several different phospholipid modifying enzymes in signal transduction (PIP3, phospholipase C␥, and SHIP) may reflect the importance of membrane dynamics in signal transduction from cell surface receptors.
SHIP is also able to hydrolyze inositol 1,3,4,5-tetrakisphosphate, and the product of this enzymatic reaction, Ins(1,3,4)P, may be involved in cell signaling as well. The exact nature of the role of Ins(1,3,4)P 3 is not apparent at present.
The precise role of SHIP in signal transduction from Fc⑀RI is not yet clear. However, it is reasonable to conclude that SHIP is an important component of the receptor signal transduction apparatus in several hematopoetic lineages. Three distinct types of SHIP interactions, via SH2, SH3, and PTB domains have been identified, providing support that SHIP has a complex adapter function. Further experiments examining the role of SHIP in receptor activation or deactivation will reveal additional insights into the varied overlapping signal transduction pathways in the cell.