The Src Homology Domain 2-Containing Inositol Phosphatase SHIP Forms a Ternary Complex with Shc and Grb2 in Antigen Receptor-stimulated B Lymphocytes*

The inositol phosphatase SHIP has been implicated in signaling events downstream of a variety of receptors and is thought to play an inhibitory role in stimulated B cells. We and others have reported that SHIP is rapidly tyrosine phosphorylated upon B cell antigen receptor (BCR) cross-linking and forms a complex with the adapter protein Shc. Here, we report that cross-linking of the BCR induces association between Grb2 and SHIP as well as association between Shc and SHIP. We made use of a Grb2-deficient B cell line to demonstrate both in vitro and in vivo that Grb2 expression is required for the efficient association between Shc and SHIP. The results indicate that SHIP, Shc, and Grb2 form a ternary complex in stimulated B cells, with Grb2 stabilizing the interaction between Shc and SHIP. The interactions between Shc, Grb2, and SHIP are therefore analogous to the interactions between Shc, Grb2, and SOS. Shc and Grb2 may help to localize SHIP to the cell membrane, regulating SHIP’s inhibitory function following BCR stimulation.

The appropriate development and functioning of B lymphocytes is dependent upon signal transduction through the B cell antigen receptor (BCR). 1 BCR stimulation may cause B cells to differentiate, proliferate, arrest growth, or undergo apoptosis. The actual response of the cell to BCR cross-linking is determined by its developmental state and the context in which the signal is received (1). The biochemical changes evoked by stimulation of the BCR include many of the well described events observed upon growth factor stimulation of cells, such as the rapid activation of tyrosine kinases and the subsequent activation of multiple signaling enzymes including phospholipase C␥, phosphatidylinositol 3-kinase (PI 3-kinase), and Ras (2)(3)(4)(5)(6)(7).
The Ras signaling pathway regulates cell growth and differentiation in many systems (8), and has recently been found to be essential for BCR-induced proliferation of mature B cells. 2 The mechanism by which BCR cross-linking activates Ras is not established, although tyrosine kinase inhibitors block Ras activation in B cells (9). Part of this Ras activation is dependent upon protein kinase C (PKC) activity, while some is PKCindependent (10). The adapter protein Shc, which has been implicated in Ras activation in other systems, becomes tyrosine phosphorylated following BCR stimulation and has been suggested to play a role in Ras activation in B cells (11)(12)(13). Shc contains an SH2 domain and a PTB domain, allowing it to bind to proteins containing phosphotyrosines (14). It also contains two tyrosines, residues 239 and 313, that when phosphorylated act as binding sites for the SH2 domain of a second adapter protein, Grb2 (15)(16)(17). Grb2 possesses two SH3 domains that permit it to bind to proline-rich sequences in the C terminus of the Ras guanine nucleotide exchange factor SOS (18). Since Shc translocates to the plasma membrane of B cells after BCR stimulation (12), it is possible that it also recruits Grb2 and SOS to the membrane and thus mediates Ras activation by bringing SOS near its substrate Ras. Indeed, Shc forms a complex with both Grb2 and SOS after BCR stimulation (12,15).
Overexpression of SHIP inhibits cellular proliferation and/or induces apoptosis, suggesting that SHIP acts as a negative regulator of hematopoietic cells (29,30). In addition, SHIP has been implicated in inhibitory signaling induced by Fc␥RIIB, the B cell receptor for the Fc portion of soluble IgG. Co-crosslinking of the Fc␥RIIB with the BCR inhibits B cell proliferation, Ras activation, and the influx of calcium across the cell membrane normally observed after BCR stimulation (31)(32)(33)(34). SHIP binds to the cytoplasmic tail of the Fc␥RIIB after its co-ligation with the BCR (35,36) and has been shown to inhibit sustained influx of calcium into the cell (37).
It has been proposed that SHIP may indirectly inhibit Ras activation by competing with SOS for binding to Shc (20). To test this idea, we have examined the mechanism by which Shc and SHIP interact in stimulated B cells. We report that SHIP forms complexes with both Shc and Grb2 in BCR-stimulated cells and that mutation of the Grb2-binding sites on Shc perturbs both of these interactions. Significantly, SHIP and Shc associate poorly in Grb2-deficient cells and this defect can be complemented in vitro and in vivo by the addition of Grb2. These results demonstrate that Grb2 is important for the association between Shc and SHIP. We also found that a phosphopeptide that binds to the SH2 domain of SHIP brings down the intact SHIP⅐Shc⅐Grb2 complex, indicating that the SH2 domain of SHIP does not directly bind Shc or Grb2. We therefore suggest that SHIP, Shc, and Grb2 form a ternary complex similar to the Shc⅐Grb2⅐SOS complex in B cells stimulated through the BCR.

MATERIALS AND METHODS
Cell Lines and Culture Conditions-The murine B cell line WEHI-231 was originally obtained from Dr. L. Lanier (DNAX Research Institute, Palo Alto, CA). The WEHI-231 cells transfected with HA-Shc were previously described (15). The two BAL17 B cell lines were originally obtained from Dr. R. Asofsky (National Institutes of Health). It is not known when these cell lines diverged and BAL17.TR lost Grb2 expression. The human B cell line Ramos was obtained from the American Type Culture Collection (Rockville, MD), the murine pre-B cell line 70Z/3 was a gift from Dr. C. Sibley (University of Washington, Seattle, WA), the human B cell line J/Y was obtained from Dr. F. Brodsky (University of California, San Francisco, CA), and the murine B cell line A20 was provided by Dr. R. Asofsky (National Institutes of Health). All cell lines were cultured in RPMI 1640 medium supplemented with 5% fetal calf serum, 2 mM sodium pyruvate, 1 mM glutamine, and 50 M 2-mercaptoethanol. To induce maximal expression of transfected HA-Shc genes in WEHI-231, cells were incubated for 48 h in media plus 5 mM isopropyl-␤-D-thiogalactoside (Research Products International Corp., Mt. Prospect, IL). The cell lines used for vaccinia virus production (CV-1, 143B, and HeLa) were obtained from American Type Culture Collection (Manassas, VA) and maintained as directed.
Cell Stimulation and Lysate Preparation-B lymphocytes (5 ϫ 10 6 cells/ml in culture medium) were treated at 37°C for 3 min with medium alone, with goat anti-mouse IgM antibodies, rabbit anti-mouse IgM antibodies, or rabbit anti-mouse IgM F(abЈ) 2 fragments (all antibodies from Jackson Immunological Research Laboratories, West Grove, PA) at 20 g/ml unless otherwise indicated. Alternatively, cells were stimulated with 150 M pervanadate (a 50 mM pervanadate stock was prepared by adding equal volumes of 100 mM Na 3 VO 4 and 2% H 2 O 2 and incubating at 25°C for 30 min prior to use) for 4 min at 37°C. After stimulation, cell suspensions were diluted with cold phosphate-buffered saline containing 1 mM Na 3 VO 4 , pelleted by centrifugation, and resuspended in ice-cold lysis buffer consisting of 20 mM Tris (pH 8.0), 90 mM NaCl, 10% glycerol, 2 mM EDTA, 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 1 mM Na 3 VO 4 , and 1 mg/ml aprotinin. Detergent-insoluble material was removed by centrifugation and the protein concentration of the soluble fraction was determined using the bicinchoninic acid assay (Pierce Chemical Co., Rockford, IL).
GST-Grb2 Fusion Proteins-The GST-Grb2 and GST-Grb2R86K expression vectors were generously provided by Dr. G. Koretzky, University of Iowa (50). Unwanted alterations in the coding sequences of these clones were converted to the published sequence via site-directed mutagenesis. The fusion proteins were expressed in Escherichia coli by standard methods (40). Expression of the fusion proteins in bacteria was induced with 0.1 mM isopropyl-␤-D-thiogalactoside treatment for 3 h and the fusion proteins were affinity purified using glutathioneagarose beads (Sigma). Ten units of thrombin (Novagen, Madison, WI) were added per 500 g of GST-Grb2 fusion protein and proteolysis allowed to proceed for 3-4 h at 25°C. The Grb2-containing supernatant was removed and passed over a p-aminobenzamidine column (Sigma) to remove thrombin. Purified Grb2 was analyzed and quantitated by SDS-PAGE followed by Coomassie staining. Three to four g of recombinant wild type Grb2 or SH2-mutant Grb2 (R86K) was added to stimulated B cell lysates and incubated overnight at 4°C. Anti-Shc immunoprecipitations were then performed as described above.
Synthetic Peptides-The ITIM peptide corresponds to amino acids 303-315 of the murine Fc␥ receptor IIb (AEN TIT Y(p)SL LKH P), the Y(p)xNx peptide corresponds to amino acids 307-318 of murine p52 Shc (LFD DPS Y(p)VN IQN), the NPxY(p) peptide corresponds to residues 911-921 of murine SHIP (EMI NPN Y(p)IG MG), and the proline-rich peptide to amino acids 1149 -1162 of murine SOS1 (VPV PPP VPP RRR PE). All peptides were obtained from Quality Controlled Biochemicals, Inc. (Hopkinton, MA) and were biotinylated at the N terminus. For affinity purifications, 30 nmol of each peptide was pre-bound to 50 l of NeutrAvidin (Pierce) and then incubated with 40 ϫ 10 6 cell equivalents (stimulated and lysed as described above) for several hours at 4°C. The beads were then treated as described for immunoprecipitations. For peptide competition assays, 25 ϫ 10 6 cell equivalents were incubated with 100 M of each peptide for 3 h at 4°C. Anti-Shc immunoprecipitations were then carried out as described above.
Vaccinia Virus-Grb2 and Grb2R86K cDNAs (described above) were cloned into the vaccinia expression plasmid pSC65 (gift of Dr. B. Moss, National Institutes of Health, Bethesda, MD) and recombinant viruses generated using standard techniques (41). BAL17.TR B cells were infected at an multiplicity of infection of 20 and harvested 24 h after infection, before cytopathic effects were visible. Cells were stimulated and lysed as described above.

Association of the Inositol Polyphosphatase SHIP with the Adapter Proteins Shc and Grb2 in Stimulated B
Lymphocytes-We and others have previously noted that the inositol polyphosphate 5-phosphatase SHIP forms complexes with the adapter protein Shc in hematopoietic cells upon stimulation with antigen or cytokines (19 -21, 29, 42). This complex has been variously reported to include (42,43) and exclude (21) the adapter protein Grb2. To assess whether SHIP associates with Grb2 as well as Shc after cross-linking of the BCR, we immunoprecipitated SHIP, Shc, or Grb2 from unstimulated and stimulated B cells and then probed immunoblots with anti-SHIP, -Shc, or -Grb2 antibodies. As is shown in Fig. 1A, when any one of these three proteins was immunoprecipitated from stimulated cells, the other two were also precipitated. Thus anti-SHIP immunoprecipitates contained both Shc and Grb2, anti-Shc immunoprecipitates contained Grb2 and SHIP, and anti-Grb2 immunoprecipitates contained SHIP and Shc. These interactions were primarily stimulation dependent, although low levels of association were seen in unstimulated cells.
It was not clear from these results whether these associations represented the existence of separate Shc⅐Grb2, Shc⅐SHIP, and SHIP⅐Grb2 complexes or the existence of a ternary SHIP⅐Shc⅐Grb2 complex. To address this point, we examined whether depletion of Grb2 from stimulated cells also led to the depletion of the Shc⅐SHIP complex. Grb2 was removed from stimulated B cell lysates by serial immunoprecipitation, as was the tyrosine kinase Lyn as a control. Next, Shc was immunoprecipitated from these depleted lysates and association of SHIP and Grb2 assessed (Fig. 1B). Depletion of Grb2 led to a roughly 2-fold decrease in the amount of Shc immunoprecipitated from the lysate, indicating that at least half of the Shc in the stimulated lysate was associated with Grb2. SHIP could barely be detected in association with the Shc that did remain. SHIP therefore associates preferentially with the population of Shc that is bound to Grb2, indicating that SHIP, Shc, and Grb2 form a ternary complex upon BCR stimulation.
The Grb2-binding Sites of Shc Are Required for Its Efficient Association with SHIP-It has previously been shown that Shc contains two tyrosines that when phosphorylated act as binding sites for the SH2 domain of Grb2 (15)(16)(17) and that the PTB domain of Shc binds to phosphorylated tyrosines in the C terminus of SHIP (22,30,44,45). In addition, SHIP contains several proline-rich regions in its C terminus that can bind to the SH3 domains of Grb2 in vitro (22,42). Therefore a SHIP⅐Shc⅐Grb2 complex could be stabilized by SH2, SH3, and/or PTB interactions in a manner analogous to the SOS⅐Shc⅐Grb2 complex (15,17). The tyrosine phosphorylation of Shc and SHIP that occurs after BCR cross-linking (11)(12)(13)46) may trigger complex formation. To examine the importance of phosphorylation of the Grb2-binding sites for Shc association with SHIP, we examined the interactions of SHIP with mutated Shc proteins in which two of the three known tyrosine phosphorylation sites (15,16,47) had been changed to phenylalanine. These experiments were performed in the immature B cell line WEHI-231 which had been stably transfected with wild type and mutant Shc proteins tagged with the influenza virus hemagglutinin (HA) epitope (15). HA-Shc was immunoprecipitated from unstimulated or BCR-stimulated cells and associated SHIP protein was detected by immunoblotting with anti-SHIP antibodies. SHIP was readily detected in association with wild type HA-Shc isolated from stimulated cells ( Fig. 2A,  top panel). Less SHIP was found associated with HA-Shc in which tyrosine 313 was mutated to phenylalanine (HA-Shc Y313F), and little or no SHIP associated with HA-Shc in which the other Grb2-binding site, Tyr 239 , was mutated to phenylalanine (HA-Shc Y239F), or in which both Grb2-binding sites were mutated (HA-Shc Y239F/Y313F). The SHIP associated with Shc was also phosphorylated on tyrosine, as could be seen by reprobing these immunoblots with anti-phosphotyrosine antibodies ( Fig. 2A, middle panel). Detection of associated SHIP by this method appeared to be more sensitive, but otherwise gave similar results.
To confirm these results, the reciprocal experiment was per- formed, in which SHIP was immunoprecipitated from unstimulated and stimulated cells and Shc was detected by immunoblotting (Fig. 2B). HA-Shc can be distinguished from endogenous Shc protein by its slower migration on SDS-PAGE gels, due to the epitope tag addition. In BCR-stimulated cells, SHIP associated well with wild type HA-Shc, whereas mutation of Tyr 239 , Tyr 313 , or both led to a dramatic decrease in the amount of HA-Shc associated with SHIP. Similar amounts of endogenous Shc were associated with SHIP in each cell line, providing an internal control. These results indicate that the two Grb2-binding sites of Shc (15,16) are both required for the efficient formation of Shc⅐SHIP complexes, although a low level of association occurs in the absence of these sites.
Requirement of Grb2 for Formation of SHIP⅐Shc Complexes-Since Grb2 co-immunoprecipitated with Shc and with SHIP and mutation of the Grb2-binding sites on Shc decreased its ability to form complexes with SHIP, it seemed likely that Grb2 helps hold the Shc⅐SHIP complex together. An alternative explanation of the latter results would be that the SH2 domain of SHIP binds directly to the phosphorylated Tyr 239 and/or Tyr 313 regions of Shc. One way to test these possibilities is to examine the interaction between wild type Shc and SHIP in the absence of Grb2. This approach was made possible by our serendipitous finding that a variant of the mature B cell line BAL17, BAL17.TR, does not express Grb2. Expression of the proteins SHIP, Shc, and Grb2 was examined in the B cell lines BAL17 and BAL17.TR. Although both cell lines expressed equivalent amounts of SHIP and Shc, Grb2 could not be detected in BAL17.TR either by immunoblotting of whole cell lysates or immunoblotting of anti-Grb2 immunoprecipitates ( Fig. 3A and data not shown). By other criteria, these cell lines were similar to one another. They responded similarly to BCR stimulation, as judged by blotting whole cell lysates with antiphosphotyrosine antibodies and Ca 2ϩ flux assays, and they expressed similar amounts of other signaling proteins such as Syk and Fc␥R (data not shown).
Another adapter protein called Grap may act similarly to Grb2, as they are 59% identical, have highly related structures and can bind the same proteins (48,49). However, since the BAL17 and BAL17.TR cells do not express detectable levels of Grap, it is unlikely to play a significant role in these cells (Fig. 3B).
BAL17 and BAL17.TR cells were stimulated via the BCR and the amount of SHIP in anti-Shc immunoprecipitations was examined. As expected, SHIP was easily detected with both anti-SHIP antibodies and anti-phosphotyrosine antibodies in Shc immunoprecipitates from stimulated BAL17 (Grb2 ϩ ) cells (Fig. 3C, top two panels). In contrast, SHIP was not detected with anti-SHIP antibodies in association with Shc in the stimulated BAL17.TR (Grb2 Ϫ ) cells (Fig. 3C, first panel). The more sensitive anti-phosphotyrosine antibodies did reveal that a small amount of tyrosine-phosphorylated SHIP was associated with Shc in these Grb2 Ϫ cells (Fig. 3C, second panel). However, we estimated that Shc bound 20-fold less SHIP in the Grb2 Ϫ cells than in the Grb2 ϩ cells. Therefore, Grb2 expression seemed to be crucial for efficient association between SHIP and Shc following BCR stimulation.
The interpretation of these results is complicated by the poor tyrosine phosphorylation of Shc in the BAL17.TR cells following BCR stimulation (Fig. 3C, fourth panel; also visible by shift of Shc mobility in third panel). In contrast, SHIP was tyrosine phosphorylated to a similar extent in both the Grb2 Ϫ and the Grb2 ϩ cells (Fig. 3D). The defect in SHIP/Shc association in the Grb2 Ϫ cells might therefore have been due to a lack of Shc tyrosine phosphorylation rather than a lack of Grb2 expression per se. Shc tyrosine phosphorylation in BAL17.TR cells could be strongly induced by treatment with pervanadate, an inhibitor of tyrosine phosphatases (data not shown). Pervanadate stimulation of the Grb2 Ϫ cells did induce a low level of association between SHIP and Shc, but much less than was observed in the Grb2 ϩ cells (Fig. 4, top panel, compare lanes 7 and 10). Thus, in the absence of Grb2, tyrosine phosphorylation of Shc and SHIP was not sufficient for the efficient association of these two proteins.
Grb2 Strongly Enhances the Association between Shc and SHIP, Both in Vitro and in Vivo-To test more directly whether Grb2 stabilizes complexes of Shc and SHIP, we added recombinant, purified Grb2 to lysates made from Grb2 Ϫ cells and examined its effect on this association. We hypothesized that the complex between Shc and SHIP would require phosphorylation of Shc tyrosines 239 and 313. As Shc from BCR-stimulated BAL17.TR cells was poorly phosphorylated on tyrosine (Fig. 3C), we stimulated these Grb2 Ϫ cells and the Grb2 ϩ BAL17 cells with pervanadate to induce strong tyrosine phosphorylation of Shc (data not shown) and used these lysates as a source of Shc and SHIP. After addition of either wild type Grb2 (WT), or Grb2 in which the SH2 domain had been inactivated by the mutation of arginine 86 to lysine (R86K) (50) to these lysates, Shc was immunoprecipitated and the amount of associated SHIP was assessed by immunoblotting (Fig. 4). The

FIG. 3. Shc and SHIP associate poorly in Grb2-deficient cells.
A, lysates were made from unstimulated BAL17 and BAL17.TR cells. Approximately 1 ϫ 10 6 cell equivalents were separated on an 8% gel by SDS-PAGE, and transferred to nitrocellulose. The top portion of the blot was probed with anti-SHIP antiserum, the middle portion of the blot was probed with anti-Shc antibody, and the bottom portion with anti-Grb2 antibody. B, lysates were made from the human and murine B cell lines WEHI-231, BAL17, BAL17.TR, Ramos, A20, J/Y, and 70Z/3 and approximately 1 ϫ 10 6 cell equivalents of each was analyzed by immunoblotting with anti-Grap antiserum. C, BAL17 and BAL17.TR cells were left unstimulated, or were treated for 3 min with 13.3 g/ml goat anti-mouse IgM F(abЈ) 2 fragments, 20 g/ml goat anti-mouse IgM antibody, or 20 g/ml rabbit anti-mouse IgM antibody and then lysed. Anti-Shc immunoprecipitations were performed on 30 ϫ 10 6 cell equivalents of lysate and the samples were analyzed by immunoblotting as described above. D, anti-SHIP immunoprecipitations were performed on 35 ϫ 10 6 cell equivalents of lysate and then analyzed by immunoblotting.
addition of wild type Grb2 clearly enhanced the association between SHIP and Shc in lysates from pervanadate-stimulated BAL17.TR cells but not in lysates from unstimulated cells (Fig.  4, compare lanes 10, 11, and 12). This effect was apparently dependent upon the tyrosine phosphorylation of Shc, as the addition of wild type Grb2 did not enhance the association of SHIP and Shc derived from BAL17.TR cells that were stimulated through the BCR (data not shown). In contrast to wild type Grb2, the Grb2 mutant (R86K) that lacks a functional SH2 domain was unable to promote the formation of a complex between Shc and SHIP. The Grb2 SH2 domain was thus required for the formation of this complex, as was phosphorylation of Shc (presumably at its Grb2-binding sites). Interestingly, the addition of Grb2 did not increase the amount of SHIP associated with Shc in lysates of BAL17 cells, indicating that the amount of Grb2 present in those cells was not limiting for the formation of this complex.
To assess the importance of Grb2 for Shc/SHIP association in intact cells, BAL17.TR cells were infected with recombinant vaccinia viruses containing no insert (vector), wild type Grb2 (Grb2), or Grb2 with a mutated SH2 domain (Grb2R86K) and association between Shc and SHIP was examined. We found that expression of wild type Grb2, but not Grb2R86K, caused a robust association between Shc and SHIP upon BCR stimulation (Fig. 5A). This Shc/SHIP association correlated with the association between Shc and Grb2. Similar amounts of Grb2 and Grb2R86K were expressed in this experiment (Fig. 5B). Interestingly, the tyrosine phosphorylation of Shc upon BCR stimulation was strikingly enhanced in the cells expressing wild type Grb2 but not in cells expressing the SH2 domain mutant Grb2R86K (Fig. 5C), perhaps because binding of the Grb2 SH2 domain to Shc protected phosphorylated tyrosines 239 and 313 from intracellular phosphatases. These results indicate that Grb2 expression is essential for the formation of complexes between Shc and SHIP.
Grb2 has been implicated in the activation of the Ras/MAPK cascade in a number of systems (51). We therefore examined whether vaccinia virus-mediated expression of Grb2 in BAL17.TR cells increased activation of MAPK in response to BCR cross-linking. To our surprise, we found that expression of either wild type Grb2 or Grb2R86K caused little or no change in the extent of MAPK activation as judged by immunoblotting whole cell lysates with antibodies that recognize only phospho-rylated, activated MAPK (Fig. 5D). Since a substantial component of MAPK activation in BCR-stimulated B cells is PKC-dependent (10), we also stimulated these cells in the presence and absence of the PKC inhibitor compound 3. Even in compound 3-pretreated cells, we saw little or no difference in MAPK activation in BAL17.TR cells infected with the different Grb2 viruses (data not shown).
Interactions of SHIP, Shc, and Grb2 with Synthetic Peptides-The above data strongly suggested that SHIP, Shc, and Grb2 form a ternary complex upon B cell stimulation. To test this model further, we assessed the ability of synthetic peptides designed to interact with the SH2, SH3, and PTB domains of Shc, Grb2, and SHIP to precipitate the complex of these proteins. We first used a phosphopeptide corresponding to the phosphorylated ITIM sequences of the Fc␥RIIB (AEN TIT Y(p)SL LKH P), which has previously been reported to bind to the SH2 domain of SHIP (35,52), to affinity purify SHIP from unstimulated and stimulated BAL17 and BAL17.TR cell lysates (Fig. 6, lanes 1-4). Upon cell stimulation, Shc and Grb2 co-purified with SHIP isolated from the BAL17 cells. This required Grb2 expression, as Shc did not co-purify with SHIP in BAL17.TR cells. It therefore appears that the SH2 domain of SHIP is free to bind to the phosphorylated Fc␥RIIB even when SHIP is bound to Shc.
We performed similar experiments with peptides known to bind to Grb2 and Shc. A proline-rich peptide (VPV PPP VPP RRR PE), derived from the C-terminal sequences of SOS and which binds to the SH3 domains of Grb2 (53) purified Grb2 from unstimulated and stimulated BAL17 lysates (Fig. 6, lanes  5-8). We found that a small amount of Shc co-purified with Grb2 upon cell stimulation, probably due to binding of the Grb2 SH2 domain to phosphorylated Tyr 239 and Tyr 313 of Shc. A small amount of SHIP also co-purified with Grb2 upon cell stimulation, perhaps reflecting a Grb2-independent interaction between the Shc PTB domain and phosphorylated SHIP, analogous to the low level of Shc/SHIP association seen in the absence of Grb2 (Fig. 3C). In any case, most of the Shc⅐Grb2⅐SHIP complex was not pulled down by this peptide, indicating a relative lack of availability of the Grb2 SH3 domains in this complex.
We next used a phosphopeptide (EMI NPN Y(p)IG MG) derived from the C-terminal region of murine SHIP that was previously shown to bind the PTB domain of Shc (30,45) and that contains the consensus PTB-binding motif NPxY(p). This peptide indeed affinity-purified Shc from BAL17 and BAL17.TR cells (Fig. 6, lanes 9 -12). More Shc was isolated from unstimulated BAL17 cells than from stimulated BAL17 cells, suggesting that in the stimulated cells the PTB domains of many Shc proteins were not accessible for peptide binding. No such decrease in association upon cell stimulation was seen in the Grb2-deficient BAL17.TR cells. Interestingly, no SHIP or Grb2 was detected in association with this peptide, indicating that the Shc PTB is occupied in the Shc⅐Grb2⅐SHIP complex. Finally, we used a phosphopeptide (LFD DPS Y(p)VN IQN) corresponding to a region of Shc known to bind Grb2 and containing the consensus motif (Y(p)xNx) known to bind the Grb2 SH2 domain (17). This peptide did indeed purify Grb2 from unstimulated and stimulated BAL17 cell lysates (Fig. 6,  lanes 13-16). Surprisingly, this peptide also precipitated Shc from stimulated and unstimulated BAL17 and BAL17.TR cells, indicating that Shc binds to this peptide in a Grb2-independent manner. Inspection of this peptide's sequence revealed that it contains an isoleucine at the Y(p) ϩ 3 position, and thus also contains a consensus binding site for the SH2 domain of Shc (54). Shc, as well as Grb2, may therefore bind directly to this phosphopeptide. Shc association with the peptide decreased FIG. 4. Enhancement of the association between SHIP and Shc in BAL17.TR (Grb2 ؊ ) lysates by recombinant Grb2. BAL17 and BAL17.TR cells were left unstimulated or were treated with 150 M pervanadate for 4 min and then lysed. Three to four g of recombinant wild type Grb2, SH2-mutant Grb2 (R86K), or no protein was added to tubes containing 30 ϫ 10 6 cell equivalents of lysate. The lysates were blotted with anti-Grb2 antibodies to confirm that similar amounts of wild type Grb2 and Grb2R86K were added (data not shown). The lysates were incubated overnight at 4°C, after which anti-Shc antibody and protein A-Sepharose was added. Immunoprecipitations were performed for 2 h, and the samples analyzed by immunoblotting as described above. upon cell stimulation in the BAL17 cells, suggesting that the SH2 domain of Shc, like the PTB domain, is less accessible to peptides after stimulation of these cells. SHIP association with this peptide was detected in the BAL17 cells but not in the BAL17.TR cells, suggesting that SHIP was co-purified via its interaction with the SH3 domains of Grb2.
In related experiments, we used these synthetic peptides as competitive inhibitors to disrupt Shc⅐Grb2⅐SHIP association in lysates made from stimulated BAL17 cells. We found that the addition of the ITIM phosphopeptide, which binds to the SH2 domain of SHIP, had little or no effect on association between Shc, SHIP, and Grb2 (Fig. 7). This suggests that the SH2 domain of SHIP is not involved in stabilization of this complex. In contrast, the NPxY(p) peptide (that binds to the PTB domain of Shc), the Y(p)xNx peptide (that binds to the SH2 domain of Grb2), and the proline-rich peptide (that binds to the SH3 domains of Grb2) all disrupted the association between SHIP and Shc. The Y(p)xNx and proline-rich peptides caused the greatest decrease in SHIP/Shc association, correlating with a decreased association between Shc and Grb2. The NPx(p)Y peptide had no observable effect on Shc/Grb2 association, perhaps because it would not be expected to disrupt the Shc⅐Grb2⅐SOS complex whereas the proline-rich and Y(p)xNx peptides would probably disrupt both Shc⅐Grb2⅐SHIP and Shc⅐Grb2⅐SOS complexes. However, all three peptides disrupted Shc⅐Grb2⅐SHIP association, implicating the PTB domain of Shc and the SH2 and SH3 domains of Grb2 in stabilization of this complex. . Each batch of cells was then split in half; one-half was left unstimulated while the other was treated with 20 g/ml goat anti-(stim. Ab) for 3 min. A, 30 ϫ 10 6 cell equivalents were immunoprecipitated with anti-Shc antibodies and subject to immunoblotting as described above. B, 5 ϫ 10 5 cell equivalents of whole cell lysates (WCL) were blotted with anti-Grb2 antibody to demonstrate similar expression levels of Grb2 and Grb2R86K. The last two lanes in the panel were prepared from one batch of cells split in half just before harvesting, suggesting that the lighter Grb2 band in the last lane is due to the poor transfer of protein at the edge of the gel rather than differences in Grb2R86K expression. C, 30 ϫ 10 6 cell equivalents were immunoprecipitated with anti-Shc antibodies and immunoblots analyzed with anti-phosphotyrosine and anti-Shc antibodies. D, 5 ϫ 10 5 cell equivalents of whole cell lysates (WCL) were blotted with anti-Active MAPK antibody to assess levels of MAPK activation (second panel). This blot was then stripped and probed with anti-MAPK antibody to determine total level of MAPK present (third panel). Since all three vaccinia viruses express ␤-galactosidase, the blot was also probed with an anti-␤-galactosidase antibody to demonstrate that levels of vaccinia virus-mediated gene expression were similar in all cells (first panel).
FIG. 6. Binding of SHIP, Shc, and Grb2 to synthetic peptides. 30 nmol of the indicated peptides were pre-bound to NeutrAvidin-agarose. Beads were washed, and 40 ϫ 10 6 cell equivalents of BAL17 or BAL17.TR cells lysates (cells were stimulated for 3 min with 20 g/ml goat anti-mouse IgM) were added to each tube. Samples were incubated at 4°C for 7 h and then analyzed by immunoblotting as described above.

DISCUSSION
To better understand how the inositol phosphatase SHIP influences B cell activation, we have studied its mechanism of interaction with the adapter protein Shc. Four lines of evidence suggest that SHIP forms a ternary complex with Shc and Grb2 analogous to the SOS⅐Shc⅐Grb2 complex. First, SHIP co-immunoprecipitates with Grb2 and immunodepletion of Grb2 removes almost all of the Shc⅐SHIP complex from the lysate. Second, mutation of the Grb2-binding sites on Shc lead to loss of association with Grb2 and a commensurate loss of association with SHIP. Third, SHIP associates poorly with Shc in Grb2-deficient cells and this defect can be complemented in vitro and in vivo by exogenous Grb2. Finally, affinity purification and competitive inhibition experiments with synthetic peptides suggest that the SH2 and SH3 domains of Grb2 and the PTB domain of Shc, but not the SH2 domain of SHIP, act together to promote association between SHIP, Shc, and Grb2. Thus, the complex between Shc and SHIP exists as a ternary complex with Grb2.
It has previously been suggested that Shc and SHIP interact in a "bi-dentate" complex in which Grb2 plays no role (30,45) (Fig. 8A). In this model, the SH2 domain of SHIP binds to phosphorylated Shc and the PTB domain of Shc binds to phosphorylated SHIP. Although this model explains the failure of tyrosine-mutated forms of Shc to bind to SHIP (Fig. 2), it cannot explain the requirement for Grb2 in the formation of this complex (Figs. 4 and 5) or the ability of the ITIM peptide, which binds to the SH2 domain of SHIP (35,52), to pull down the intact SHIP⅐Shc⅐Grb2 complex (Fig. 6). Moreover, it has been reported that the SH2 domain of Grb2 binds to the Tyr 313 region of Shc with a 10-fold higher affinity than does the SH2 domain of SHIP (30), suggesting that the former interaction is more likely to be relevant in vivo. Although one study has reported that the SH2 domain of SHIP is essential for its association with Shc (30), this may be an indirect effect as mutation of the SH2 domain of SHIP abrogated SHIP tyrosine phosphorylation in these experiments and thus would preclude binding of the Shc PTB domain to SHIP. This interpretation is supported by the previous observation that the SH2 domain of SHIP is dispensable for Shc-SHIP interaction when SHIP is tyrosine phosphorylated (44).
We have combined our results summarized above with previously reported features of the association of SHIP with Shc to derive a new model for the mechanism by which these proteins associate (Fig. 8B). According to this ternary complex model, BCR stimulation leads to phosphorylation of tyrosines 239 and 313 of Shc, which are bound by two Grb2 molecules via their SH2 domains (15,16,44). The SH3 domains of Grb2 then bind to the three proline-rich regions of SHIP, as previously reported with in vitro binding assays (22,42). The formation of the complex between SHIP and Shc also appears to require association between the PTB domain of Shc and tyrosine phosphorylated sites on SHIP (30,44,55) (Fig. 7). This is a major difference between the Shc⅐Grb2⅐SHIP complex and the Shc⅐Grb2⅐SOS complex, as SOS does not become phosphorylated on tyrosine residues. The formation or maintenance of the Shc⅐Grb2⅐SHIP complex appears to require all of these proteinprotein interactions. We observed association between Shc, Grb2, and SHIP upon cell stimulation in all three murine B cell lines that we investigated (BAL17, WEHI-231, and A20, this report and data not shown), suggesting that this is a general mechanism in BCR-stimulated B cells. Other means of association may occur in other cell types, since it has been reported that SHIP does not associate with Grb2 in a hematopoietic cell line stimulated with IL-3 (21).
Association between Shc and SHIP in B cells has been reported to be optimal when the inhibitory receptor Fc␥RIIB is co-cross-linked with the BCR (20,36,43,46). However, a substantial amount of Shc⅐SHIP complex forms in B cells stimulated only through the BCR (19). In BAL17 cells, we found only modest differences in the amount of SHIP associated with Shc when the cells were stimulated with anti-IgM antibodies that differed in their ability to engage the Fc␥R (Fig. 3C and data  not shown). Similarly, a small decrease in SHIP-Shc association was observed when the Fc␥RIIB was blocked with a monoclonal antibody prior to stimulation with intact anti-IgM antibodies (data not shown). We therefore postulate that there are two mechanisms by which BCR stimulation causes tyrosine phosphorylation of SHIP, resulting in its association with Shc and Grb2. The first mechanism involves the binding of SHIP to the ITIM of the Fc␥RIIB, while the second is Fc␥RIIB-independent and may involve the association of SHIP with the BCR-activated tyrosine kinase Syk (19). Perhaps SHIP acts as a tonic negative regulator even in positively stimulated cells and negative stimulation, i.e. co-clustering of the BCR and the Fc␥RIIB, enhances SHIP-Shc-Grb2 interaction (and perhaps SHIP function) further. There are now several precedents for tonic negative regulators in tyrosine kinase receptor and cytokine receptor signaling (56 -59).
The effect that association of SHIP with Shc and Grb2 has on SHIP function is not yet clear. It has been suggested that SHIP binding to Shc inhibits Ras activation indirectly by preventing Shc association with SOS (20). In support of this model, Shc uses primarily the same protein-protein interaction sites to FIG. 7. Various peptides competitively inhibit the Shc-Grb2-SHIP interactions. 30 ϫ 10 6 cell equivalents of BAL17 lysates stimulated with goat anti-mouse IgM as described above were incubated with 100 M of the indicated peptides for 3 h at 4°C. Anti-Shc antibodies and protein A-Sepharose were then added and the immunoprecipitations allowed to proceed overnight at 4°C. Samples were then analyzed by immunoblotting as described above.
FIG. 8. Two models for Shc-SHIP interactions. A, the bi-dentate model of Shc-SHIP interaction, in which the SH2 domain of SHIP binds to phosphorylated Shc and the PTB domain of Shc binds to phosphorylated SHIP. B, the ternary complex model of Shc-SHIP interaction. In this model as in the bi-dentate model, the PTB domain of Shc binds to phosphorylated SHIP. However, we suggest that two Grb2 molecules bind to Shc with their SH2 domains and then to SHIP with their SH3 domains. According to this model, the SH2 domain of SHIP plays no role in stabilization of this complex. "PPP" represents proline-rich regions in SHIP while the encircled P's represent phosphorylated tyrosine residues.
interact with SOS and to interact with SHIP, indicating that Shc is unlikely to bind both of these molecules simultaneously. Indeed, we have never detected any association between SHIP and SOS (data not shown). However, we have not detected any decrease in association between Shc and SOS1 or Shc and SOS2 when cells were stimulated through the BCR and the Fc␥RIIB instead of stimulated through the BCR alone (data not shown). Also, the addition of recombinant SOS1 protein to stimulated B cell lysates did not lead to any decrease in association between Shc and SHIP (data not shown). These results suggest that the amount of Grb2 and tyrosine-phosphorylated Shc in stimulated B cells is not limiting and that Ras inhibition may occur via a mechanism other than the sequestration of Shc by SHIP. In support of this finding, it has recently been reported that ERK activation is unaffected in a SHIP-deficient mutant of the chicken B cell line DT-40, suggesting that SHIP does not modulate Ras activation in BCR-stimulated B cells (60). It should also be noted that there is currently no direct evidence showing that Shc and Grb2 play an important role in Ras activation in BCR-stimulated B cells. Indeed, we noted little difference in MAPK activation when comparing either the BAL17 and BAL17.TR cells or BAL17.TR cells that had been infected with Grb2-expressing vaccinia viruses or with control viruses (Fig. 5D). In addition, vaccinia virus-mediated overexpression of either wild type Shc or mutant Shc that no longer bound Grb2 (Shc Y239F/Y313F) had no obvious effect on MAPK activation in stimulated B cells (data not shown). These results suggest that BCR-induced Ras activation is not primarily mediated by Shc⅐Grb2⅐SOS complexes in BAL17 cells.
Another possibility is that formation of the Shc⅐Grb2⅐SHIP complex serves to regulate SHIP activity by controlling its intracellular localization. It has previously been reported that a Shc-associated, tyrosine-phosphorylated protein of 145 kDa translocated from the cytosol to the membrane fraction upon BCR stimulation (12). We also have found that complexes of Shc and a tyrosine-phosphorylated protein of 145 kDa are primarily membrane-bound in BCR-stimulated B cells (19). This 145-kDa protein is very likely to be SHIP, as tryptic fragments obtained from it match the published SHIP protein sequence. 3 Recruitment of SHIP to the plasma membrane by Shc and Grb2 might control its activity by bringing it near its substrate PI(3,4,5)P 3 .
SHIP activity has been shown to antagonize PI 3-kinase function in Xenopus oocytes and to inhibit calcium influx in stimulated B cells (23,37,60). Both of these effects could be mediated by the hydrolysis of PI(3,4,5)P 3 by SHIP, as PI(3,4,5)P 3 is thought to enhance the phospholipase C␥-catalyzed breakdown of PI(4,5)P 2 to inositol 1,4,5-trisphosphate and diacylglycerol (61). SHIP activity would therefore prevent the prolonged influx of calcium mediated by inositol 1,4,5trisphosphate and might inhibit calcium influx through other mechanisms as well (60). We therefore prefer the hypothesis that SHIP inhibits BCR signaling by its ability to hydrolyze PI(3,4,5)P 3 and that its association with Shc and Grb2 leads to SHIP localization at the plasma membrane, promoting its subsequent inhibitory actions.