Grb2 Forms an Inducible Protein Complex with CD28 through a Src Homology 3 Domain-Proline Interaction*

CD28 provides a costimulatory signal that results in optimal activation of T cells. The signal transduction pathways necessary for CD28-mediated costimulation are presently unknown. Engagement of CD28 leads to its tyrosine phosphorylation and subsequent binding to Src homology 2 (SH2)-containing proteins including the p85 subunit of phosphatidylinositol 3′-kinase (PI3K); however, the contribution of PI3K to CD28-dependent costimulation remains controversial. Here we show that CD28 is capable of binding the Src homology 3 (SH3) domains of several proteins, including Grb2. The interaction between Grb2 and CD28 is mediated by the binding of Grb2-SH3 domains to the C-terminal diproline motif present in the cytoplasmic domain of CD28. While the affinity of the C-terminal SH3 domain of Grb2 for CD28 is greater than that of the N-terminal SH3 domain, optimal binding requires both SH3 domains. Ligation of CD28, but not tyrosine-phosphorylation, is required for the SH3-mediated binding of Grb2 to CD28. We propose a model whereby the association of Grb2 with CD28 occurs via an inducible SH3-mediated interaction and leads to the recruitment of tyrosine-phosphorylated proteins such as p52 shc bound to the SH2 domain of Grb2. The inducible interaction of Grb2 to the C-terminal region of CD28 may form the basis for PI3K-independent signaling through CD28.

Engagement of the T cell receptor (TCR) 1 by the major histocompatibility complex-peptide complex in the absence of costimulatory molecules is insufficient to induce production of cytokines and can render the T cells unresponsive to further antigenic challenge (1). CD28 is preeminent among a group of receptors, including 4 -1BB and CD43, that can provide costimulatory signals to T cells (2)(3)(4). CD28 is a type 1 transmembrane protein of the Ig superfamily, which is expressed on the cell surface as a glycosylated homodimer (5). CD28 costimulation of TCR-dependent responses increases IL-2 production (6,7), prevents the induction of anergy (8), and renders T cells resistant to apoptotic cell death (9 -11). These effects are me-diated by increased transcription of cytokine genes through the activation of a composite NF-B and AP-1 transcriptional element (12,13), the stabilization of cytokine mRNAs (14), and the expression of the survival protein, Bclx L (9). TCR-mediated T cell activation is sensitive to the immunosuppressive drug cyclosporin, while those pathways activated by CD28 are not, suggesting that costimulatory pathways are distinct from those activated by the TCR (7). Early biochemical events induced through CD28 include tyrosine phosphorylation (15,16) and activation of PI3K (17)(18)(19) and acidic sphingomyelinase (20). The identity of the signal transduction pathways that are required for CD28-mediated costimulation are presently unknown.
The cytoplasmic domain of CD28 contains no recognized intrinsic enzymatic activity; however, CD28 has been reported to associate with signaling proteins following ligation. Phosphorylation of CD28 on tyrosine 173 within the motif YMNM present in the cytoplasmic domain provides a binding site for the SH2 domain of the p85 subunit of PI3K (17)(18)(19); however, conflicting results regarding the requirement for PI3K in CD28dependent costimulation have been published. Mutant forms of CD28 that are unable to bind to PI3K demonstrate an absolute requirement for PI3K in mediating CD28 signals in mouse T cell hybridoma cell lines (17). Further, wortmannin, a potent inhibitor of PI3K, inhibits costimulation through CD28 in human peripheral T cells (21,22). In contrast to these studies, CD28 can provide costimulatory signals in the absence of PI3K activation in Jurkat cells and purified mouse T cells (23)(24)(25)(26). Moreover, activation of PI3K by ectopically expressed CD19, a potent activator of PI3K in B cells, in conjunction with TCR ligation is insufficient to induce IL-2 transcription in Jurkat cells (23). These studies demonstrate that PI3K activation is neither necessary nor sufficient for CD28-mediated costimulation in certain cellular systems and suggest that alternative signaling pathways are involved in costimulation. In support of this view, deletion of 10 amino acids in the C-terminal portion of CD28, remote from the PI3K binding site, attenuates costimulation (26,27). Until now, no signaling proteins that bind to this site have been identified.
More recently, Grb2 has been implicated in CD28 signaling (28,29). Grb2 is a linker protein that utilizes both SH2-and SH3-dependent interactions to bind to a diverse repertoire of signaling proteins. The canonical function of Grb2 is to stabilize an intermolecular complex between receptor tyrosine kinases, such as epidermal growth factor receptor, Met, and Flt3, and the positive regulator of the Ras pathway, Sos. Grb2 binds constitutively to Sos through its SH3 domains and inducibly binds to tyrosine-phosphorylated receptors via its SH2 domain (30 -35). Activation of the Ras signaling pathway is a critical step during T cell activation (36,37). In TCR-stimulated cells, there is a rapid formation of a complex between Sos/Grb2 and a 36-kDa membrane protein, LAT, that is a substrate for the TCR-induced tyrosine kinases (38 -40). Antibody-mediated aggregation of CD28 can also activate Ras (41). Grb2 has been shown to bind to CD28 following ligation of CD28 (28,29). Under these circumstances, the Grb2-CD28 association was in part mediated by the Grb2-SH2 domain binding to the CD28 PI3K binding site, Tyr 173 (28,29). This interaction may be responsible for CD28-dependent Ras activation (29). The SH3 domains of Grb2 also bind the product of the protooncogene c-cbl in T cells; however, the role of this interaction during T cell activation is not known (42)(43)(44)(45).
CD28 contains two potential SH3-binding diproline motifs, one of which is contained in part by the C-terminal region required for costimulation. SH3 domains bind to short peptide sequences rich in proline residues, which adopt a left-handed type II polyproline helix conformation. Two proline residues are presented as a bidentate hydrophobic contact surface that binds to a shallow hydrophobic groove common to SH3 domains (46). In this report, we demonstrate that in addition to binding SH2-containing signaling molecules, CD28 is an SH3-binding protein. Following ligation, CD28 binds to the SH3 domains of Grb2 and Itk as well as to the WW domain of YAP.
The association between Grb2 and CD28 occurs via an SH3proline interaction involving the diproline motif embedded in the C-terminal portion of the cytoplasmic domain of CD28. The interaction between CD28 and Grb2-SH3 domains is phosphotyrosine-independent and does not require a functional SH2 domain. The SH3-mediated interaction between CD28 and Grb2 allows the SH2 domain of Grb2 to bind to phosphotyrosine-containing proteins such as p52 shc . We propose a model in which Grb2 functions in a heretofore uncharacterized manner to couple CD28 to tyrosine-phosphorylated proteins involved in CD28-mediated costimulation.

EXPERIMENTAL PROCEDURES
Cell Lines-The murine thymoma cell lines VCD28 and VCD28⌬10 expressing human CD28 were a kind gift from D. Couez (INSERM U298, Angers, France). COS-7 and 293T cells were purchased from ATCC. All cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Life Technologies, Inc.) at 37°C in a humidified atmosphere containing 5% CO 2. G418 (Life Technologies) was added at 2 mg/ml to cells transfected with CD28. Rat2 cells expressing the Fms-Flt3 chimeric receptor (FF3) have been described elsewhere (47).
Antibodies-Anti-human CD28 mAb 9.3 ascites was a kind gift from P. Linsley (Bristol-Meyers Squibb Pharmaceutical Co., Seattle, WA). Anti-CD28 serum was produced in rabbits by immunizing with a synthetic peptide corresponding to the C-terminal 18 amino acids of CD28 coupled to keyhole limpet hemocyanin. Anti-Grb2-SH2 serum was produced in rabbits immunized with GST-Grb2-SH2. The antibodies were purified on protein A-Sepharose (Amersham Pharmacia Biotech), and coupled to CNBr-activated beads (Bio-Rad) according to the manufacturer's directions. Uncoupled CNBr sites were quenched by alternate washes in high and low pH followed by incubation in 100 mM Tris-HCl, pH 8.0. Purified rabbit Ig purchased from Jackson Laboratories was also coupled to CNBr beads and was used as a nonspecific Ig control in Fig. 4B. Anti-Flt3 serum was produced in rabbits against a TrpE-Flt3 fusion protein and has been described previously (47). Anti-Grb2, anti-Cbl, and anti-Myc were purchased from Santa Cruz Biotechnology (Santa Cruz, CA); anti-Sos was purchased from Transduction Laboratory; and the anti-phosphotyrosine mAb 4G10 was purchased from Upstate Biotechnology (Lake Placid, NY).
Plasmid cDNAs-Human CD28 cDNA (clone A53, gift from B. Seed, Massachusetts General Hospital, Boston, MA) was cloned into the EcoRI site of pME18 -226Neo, which contains the SR␣ promoter (gift from Gerrard Zurawski, DNAX, Palo Alto, CA). Polymerase chain reaction-based overlap extension mutagenesis was used to create point mutations within the cytoplasmic domain of CD28. The plasmids containing Myc epitope-tagged Grb2 under control of the cytomegalovirus promoter were a gift from David Pot (Chiron Corp., Emeryville, CA) and have been described previously (48).
Transfections of Cell Lines-293T cells were transfected using a standard calcium phosphate method. COS cells were transfected using Lipofectamine (Life Technologies), according to the manufacturer's instructions.
Cell Stimulation, Lysis, and Coprecipitation-VCD28 cells were harvested and resuspended at 2-4 ϫ 10 7 cells/ml in PBS. 1 g of mAb 9.3 was added to each stimulated sample (0.5 ml) in 1.5-ml Eppendorf tubes and incubated at 37°C for 5 min. The cells were transferred to ice, and 0.5 ml of 2ϫ PLC␥ lysis buffer was added. Following lysis, 1 g of mAb 9.3 was added to unstimulated samples. Confluent Rat2-FF3 cells were stimulated with 500 ng of colony-stimulating factor-1/ml for 5 min at 37°C, and lysed in 1ϫ PLC␥ lysis buffer. The lysate from one 100-mm plate was used per condition. 293T cells were lysed directly in 1ϫ PLC␥ lysis buffer in 100-mm tissue culture dishes, and 1 ⁄10th of each lysate was used per coprecipitation. Lysates were centrifuged at 21,000 ϫ g for 15 min, and the supernatant was incubated with protein A-Sepharose beads (Amersham Pharmacia Biotech) or with immobilized GST fusion proteins on glutathione beads for 2-4 h at 4°C. The beads were washed three times with cold lysis buffer and boiled in the presence of SDS sample buffer containing ␤-mercaptoethanol (Laemmli buffer). The protein complexes were resolved by SDS-PAGE and transferred to polyvinylidene difluoride membranes (Immobilon). CD28 was immunoblotted using anti-CD28 serum in PBS with 0.05% Tween and 5% skim milk powder, followed by incubation with horseradish peroxidase-conjugated donkey anti-rabbit antibody or protein A (The Jackson Laboratory). Protein bands were detected by Renaissance enhanced luminol reagent (NEN Life Science Products).
Biotinylation of Cell Surface Proteins-Before stimulation, cells were washed in PBS and resuspended in 10 mM sodium borate, 150 mM NaCl, pH 8.8. The biotinylation reaction was initiated by the addition of 5 l of 10 mg/ml sulfosuccinimidyl-6-(biotinamido)hexanoate (Pierce) in Me 2 SO. After 15 min, the biotinylation reaction was quenched by the addition of 10 l of 1 M NH 4 Cl. The cells were washed three times with PBS plus 10 mM Tris-Cl, 1 mM EDTA before stimulation and lysed as described above. The protein blots were probed with avidin-horseradish peroxidase (Amersham Pharmacia Biotech) diluted in 1% skim milk powder in PBS, 0.05% Tween. To verify that the cells were uniformly labeled, an aliquot of cells was incubated with avidin-Spectral Red (Southern Biotechnology Associates Inc.) and analyzed by flow cytometry.

RESULTS
CD28 Is an SH3-binding Protein-SH3 and WW domains are distinct polypeptide structures that function to form multimeric protein complexes as a result of their capacity to bind to proline-rich sequences. The structure of SH3 domains consists of a ␤-barrel of two three-stranded, antiparallel ␤-sheets, which presents an array of conserved hydrophobic side chains appropriately spaced for interaction with polyproline helices (49). WW domains are composed of three anti-parallel ␤-sheets rich in aromatic amino acids that form a hydrophobic ligand pocket (50). The ligand specificity of SH3 and WW domains may in some instances be overlapping (51). The cytoplasmic domain of CD28 contains two polyproline sequences and therefore may form complexes with SH3-and WW-containing proteins.
We tested whether CD28 could function as an SH3-or WWbinding protein. A set of 15 GST-SH3 domains and two GST-WW domains were screened for their capacity to coprecipitate CD28 from cellular lysates. VCD28 cells were stimulated with CD28-specific mAb and lysed. Lysates from 10 7 cells were incubated with immobilized GST fusion proteins. The protein complexes were resolved by SDS-PAGE and immunoblotted with CD28-specific antibodies. Fig. 1 shows that CD28 forms a complex with the SH3 domain of Itk (Fig. 1, lane 5). CD28 was also detected in a complex with the recombinant full-length Grb2 fusion protein (Fig. 1, lane 14). The recombinant p85 GST-SH3 has been reported to bind in vitro to a synthetic peptide derived from the C-terminal 17 amino acids of CD28 (27). We did not detect an interaction between immobilized p85 GST-SH3 and native CD28 from cellular lysates (Fig. 1, lane 17). Although the Nck-SH3 domains may bind CD28 weakly (Fig. 1, lane 16), no other SH3 domains were found to bind CD28. The WW domain of YAP bound to CD28, whereas the WW domain derived from the dystrophin gene product did not (Fig. 1, lanes 19 and 20). CD28 was not effectively phosphorylated in response mAb 9.3 alone, as was evidenced by the lack of CD28 precipitation by the p85 C-terminal SH2 domain (Fig. 1, lane 18). Phosphorylation of CD28 was achieved either by preincubating the cells with the tyrosine phosphatase inhibitor pervanadate or by further aggregation of CD28 with a secondary antibody (anti-mouse IgG), as others have shown (19, 52) (data not shown). The lack of CD28 phosphorylation following ligation with mAb 9.3 was exploited to detect CD28-SH3 interactions in the absence of SH2-mediated interactions.
The Interaction between Grb2 and CD28 Is Inducible and Is Mediated by the SH3 Domains of Grb2-We next examined which of the Grb2 domains were capable of binding to CD28 using recombinant GST fusion proteins. VCD28 lysates were prepared from resting or mAb 9.3-stimulated cells and incubated with equivalent amounts of immobilized GST, GST-Nterminal SH3 (GST-SH3N), GST-SH2, or GST-C-terminal SH3 (GST-SH3C) fusion proteins. The protein complexes were resolved by SDS-PAGE and immunoblotted with CD28-specific antibodies. CD28 from unstimulated cells did not complex to any of the Grb2-derived constructs ( Fig. 2A, lanes 1, 3, and 5). GST-SH3C demonstrated strong and inducible binding to CD28, while neither the GST-SH3N nor the GST-SH2 domain coprecipitated detectable amounts of CD28 ( Fig. 2A, compare  lane 6 to lanes 2 and 4). In other experiments, weak binding of the N-terminal SH3 domain, but not the SH2 domain, of Grb2 to CD28 was detected (data not shown). The isolated C-terminal SH3 domain was less effective than the full-length Grb2 molecule in coprecipitating CD28 (data not shown). The capacity of the Grb2-SH2 fusion protein to efficiently bind to tyrosine-phosphorylated proteins was tested in Fig. 2A, lower part. The Flt3 receptor tyrosine kinase binds to Grb2 at tyrosine 958 within the carboxyl tail in a manner similar to epidermal growth factor receptor or Met (53). In contrast to CD28, the activated Flt3 receptor was coprecipitated with the SH2 domain of Grb2, while there was no detectable interaction with either of the Grb2-SH3 domains ( Fig. 2A, compare lane 13 to lanes [11][12][13][14][15]. Sos was detected in complex with both the N-and C-terminal SH3 domains but not with the SH2 domain of Grb2 (Fig. 2C, lanes 1-6).
We next used loss-of-function mutations in the SH3 and SH2 domains of Grb2 to verify the domain requirements for the Grb2-CD28 interaction. Molecular characterization of mutant alleles of the Caenorhabditis elegans Grb2 homologue sem5 have identified single point mutations that diminish the binding capacity of the SH3 domains to their target proteins (34). GST fusion proteins carrying mutations in either the N-terminal (P49L) or C-terminal (G203R) SH3 domains resulted in a severe reduction in CD28 binding (Fig. 2B, lanes 4 and 6). No detectable CD28 was coprecipitated when both SH3 domains were mutated (Fig. 2B, lane 13). Grb2 constructs in which the SH2 domain was mutated within the conserved FLVRES motif within the SH2 domain (R86K) bound CD28 at levels comparable with the full-length wild type Grb2 (Grb2FL) protein (Fig.  2B, lane 11). In contrast to the interactions observed with CD28, the single or double Grb2-SH3 mutations had no effect on binding to activated Flt3 receptors (Fig. 2B, lanes 19, 21,  and 28), whereas a single point mutation in the SH2 domain resulted in complete loss in Flt3 binding (Fig. 2B, lane 26). Sos binding was only abrogated by the double SH3 mutant of Grb2, confirming that the mutation in either one of the SH3 domains did not affect the function of the other (Fig. 2C, lanes 7-16).
Together, these results show that CD28 can inducibly bind the SH3 domains of Grb2. Under these experimental conditions, the interaction between CD28 and Grb2 does not require the Grb2-SH2 domain.
Grb2 Binds to Diproline Motifs in the Cytoplasmic Domain of CD28 -Structural studies of the Grb2-SH3 domains bound to their ligands show that the two prolines in the PXXP motif represent contact residues with the SH3 hydrophobic binding groove. Mutation of either of these prolines results in significant attenuation of the binding interaction (54). We changed the first proline of each PXXP motif present in the cytoplasmic tail of CD28 to alanine by mutagenesis at codons 178 (P178A) and 190 (P190A) (Fig. 3A). The wild type and mutant forms of FIG. 1. CD28 binds SH3 and WW domains in vitro. VCD28 cells were stimulated with anti-CD28 mAb. Lysates from 10 7 cells were incubated with 10 g of immobilized GST fusion proteins, as indicated, and the resulting protein complexes were analyzed by Western blot analysis using antisera to CD28, Cbl, and Sos. In lanes 1-11, the GST-SH3 domain of Abl, Src, Lck, Fyn, Itk, Fgr, HS1, PLC␥, Vav, Gap, or spectrin was used, respectively, as an affinity reagent to coprecipitate CD28 from cellular lysates. GST alone, full-length GST-Grb2, GST-Crk-SH3, and GST-Nck-(SH3) 3 GST-p85-SH3, GST-p85-SH2, and the GST-WW domains of dystrophin and Yap are shown in lanes 12, and lanes 14 -20, respectively. Lanes 13 and 21 show CD28 immunoprecipitations as positive controls to indicate the electrophoretic mobility of CD28, which appears as a strongly glycosylated broad smear around 44 kDa.
CD28 were expressed in COS cells at similar levels as demonstrated by Western blotting (Fig. 3B, lanes 1, 3, and 5). Cells were stimulated and lysed as above. Wild type and the P178A mutant forms of CD28 from stimulated cells bound to GST-Grb2 at similar efficiency (Fig. 3B, lanes 2 and 4). Substitution within the C-terminal diproline motif (P190A) reduced the amount of CD28 that coprecipitated with Grb2 (Fig. 3B, lane 6).
These data suggest that the C-terminal diproline motif is the primary binding site for the SH3 domains of Grb2.
We next investigated whether Grb2 can associate with a 10-amino acid C-terminal deletion mutant of CD28, which lacks part of the C-terminal diproline motif (Fig. 3A). Because the ⌬10 mutant cannot be detected using antiserum raised against the C terminus of CD28, cell surface proteins were biotinylated before lysing the cells, and immunoprecipitated CD28 was detected using avidin-horseradish peroxidase. The efficiency of biotinylation was determined by flow cytometry (Fig. 3C, right parts). Flow cytometric analysis using mAb 9.3, which binds the extracellular domain of CD28, determined that CD28WT and CD28⌬10 were expressed at similar levels (Fig.  3C, left parts). A smeared band around 44 kDa, consistent with the appearance of CD28 on Western blots, was detected in CD28WT and CD28⌬10 immune complexes (Fig. 3D, lanes 5-8) but was not found in complex with immobilized GST-Grb2-SH2 (Fig. 3D, lanes 1-4). Next, lysates that had been precleared either by GST-Grb2-SH2 or by anti-CD28 were incubated with GST-Grb2FL. GST-Grb2FL coprecipitated a major 44-kDa band from the lysates of VCD28WT but not VCD28⌬10 cells (Fig. 3E, lanes 1-4). Immune-specific preclearing of CD28 resulted in the loss of binding of the 44-kDa band to GST-Grb2FL, confirming the identity of this band as CD28 (Fig. 3E, lanes 5-8). We did not observe inducible binding of Grb2 to CD28 under these conditions, which may be a consequence of the biotinylation procedure. These data demonstrate that the interaction of Grb2 with CD28 is specified by determinants contained within the C-terminal 10 amino acid residues of CD28, corresponding to the second diproline motif.
In Vivo Association between CD28 and Grb2-In order to determine whether CD28 and Grb2 formed a protein complex in vivo, as suggested by our in vitro experiments, cellular lysates from VCD28 cells were incubated with purified Grb2specific antiserum or with nonspecific Ig covalently linked to CNBr-coupled Sepharose beads. CNBr-coupled beads were used instead of protein A-Sepharose beads so that CD28 could not be immunoprecipitated by the stimulating antibody. Protein complexes present in Grb2 immunoprecipitates were resolved by gel electrophoresis and immunoblotted with CD28 antibodies. Grb2-specific anti-serum co-immunoprecipitated CD28 (Figs. 4, A and B, lanes 1 and 2). Ligation of CD28 increased the amount of CD28 present in Grb2 immune complexes (Fig. 4, A and B, compare lane 2 with lane 1), consistent with the results from Fig. 2.
We also determined the effect of Grb2 binding to CD28 on the distribution of Grb2 present in other protein complexes such as with Sos and Cbl. The Grb2 immune complexes shown in Fig.  4A were probed for Cbl and Sos. Ligation of CD28 resulted in a reduction in the amount of Cbl bound to Grb2, while the stoichiometry of the Grb2-Sos complex was largely unchanged (Fig.  4A, top, compare lanes 1 and 2). The reduced binding of Grb2 to Cbl following CD28 ligation may reflect competition between Cbl and CD28 for the SH3 domains of a common and limited pool of Grb2 protein. Recent published accounts similarly show that signaling through TCR results in redistribution of Grb2 pools with a decreased level of Grb2 bound to Cbl (42,44).

FIG. 3. Grb2 binds the C-terminal diproline motif of CD28.
A, sequence of CD28 cytoplasmic domain and diagram of CD28 mutants used in this study. B, cDNAs encoding wild type and mutant forms of CD28 were transfected into COS-7 cells. Lysates from transfected COS cells were equally split and analyzed for expression by immunoprecipitation using anti-CD28 antibodies or were incubated with immobilized GST-Grb2 to assess the capacity of the mutant forms of CD28 to coprecipitate with Grb2. The amount of CD28 bound to Grb2 (lanes 2, 4, and 6) is shown in comparison with the CD28 expression levels of wild type, P178A, and P190A mutants (lanes 1, 3, and 5). C, flow cytometric analysis of CD28 expression and surface biotinylation. VCD28 WT and VCD28⌬10 were stained with anti-CD28 followed by fluorescein isothiocyanate-conjugated antimouse Ig and streptavidin-Spectra Red and analyzed by flow cytometry. The negative population represents unstained samples. D and E, lysates from biotinylated VCD28WT (lanes 1, 2, 5, and 6) or VCD28⌬10 (lanes 3, 4, 7, and 8) cells were incubated first with GST Grb2-SH2 (D, lanes 1-4) or with protein A beads in the presence of CD28-specific mAb (D, lanes 5-8). Following 1-h incubations, the lysates were added to GST Grb2FL (E, lanes 1-8) immobilized on beads for a second incubation. Anti-CD28-stimulated cells are shown in lanes 2, 4, 6, and 8, whereas unstimulated cells are in lanes 1, 3, 5, and 7. Biotinylated cell surface proteins that coprecipitated with Grb2-SH2, anti-CD28, or subsequently with fulllength Grb2 were revealed by Western blot analysis using streptavidin-horseradish peroxidase. CD28 is seen as a smear around 44 kDa. plasmid encoding an inducible tyrosine kinase. Recombinant GST-CD28 expressed in the BL21 bacteria (CD28BL) was not tyrosine-phosphorylated, whereas GST-CD28 protein expressed in the BL21TK strain (CD28TK) was quantitatively tyrosine-phosphorylated, as detected by phosphotyrosine-specific antibodies and a shift in electrophoretic mobility (Fig. 5D). GST does not become tyrosine-phosphorylated under these conditions (data not shown). GST-CD28BL or phosphorylated GST-CD28TK fusion proteins were used to coprecipitate transiently expressed Grb2 from 293T cells. Wild-type and mutant Grb2 constructs carrying point mutations in the SH3 or SH2 domains were Myc epitope-tagged to distinguish them from the endogenous Grb2 protein (48). Grb3-3, a Grb2 isoform with a nonfunctional SH2 domain resulting from an internal deletion within the SH2 coding sequence (55), was also used in these experiments. The unphosphorylated GST-CD28BL protein formed a complex with wild type Grb2, the mutant N-terminal SH3 form, and Grb3-3 (Fig. 5A, lanes 1, 2, and 5). These data support the observation that the association of CD28 with Grb2 can occur in the absence of tyrosine phosphorylation and does not require a functional Grb2-SH2 domain. GST-CD28BL did not coprecipitate Grb2 when either the C-terminal or both SH3 domains were mutated, which, in accordance with the data presented in Fig. 2, indicates that the C-terminal SH3 domain of Grb2 is the dominant binding domain to CD28. As shown in Fig. 5C, the coprecipitation of Sos with Grb2 was also SH3-dependent, but in contrast to CD28, the N-terminal SH3 domain rather than the C-terminal SH3 domain defines the high affinity interaction (34).
In contrast to the unphosphorylated form, tyrosine-phosphorylated GST-CD28TK bound to all the mutant forms of Grb2, indicating that when CD28 is tyrosine-phosphorylated it can bind to either the SH2 or the SH3 domains of Grb2 (Fig. 5B,  lanes 1-5). The capacity of Myc-Grb3-3 to bind to GST-CD28TK demonstrates that a functional Grb2-SH2 domain is not required to bind to the phosphorylated form of CD28 (Fig.  5B, lane 5). GST-CD28TK bound endogenous Grb2 more effectively than GST-Grb2BL, possibly reflecting a higher affinity of the SH2 domain for the phosphotyrosine than the SH3 domains for the diproline motifs (Fig. 5, and data not shown).
Tyrosine-phosphorylated Shc Binds to CD28 -We have shown that Grb2 forms a protein complex with CD28 via an SH3-proline interaction. Grb2 may thus function to link CD28 with tyrosine-phosphorylated proteins present in activated T cells. The SH2 domain of Grb2 can bind to proteins that contain a common consensus binding site, pYXNX, where pY represents phosphotyrosine (56). Tyrosine-phosphorylated proteins present in activated T cells that bind to the Grb2-SH2 domain include p36 LAT , p62/68, Shc, the -chain of the TCR complex, and SHP-2 (38 -40, 57-59). The multiplicity of Grb2 binding partners suggests that Grb2 may have distinct and varied functions during T cell activation.
In order to identify the intracellular ligand for the SH2 domain of CD28-associated Grb2, we incubated lysates derived from T cells activated by ligation of both CD3 and CD28 with immobilized unphosphorylated GST-CD28 as an affinity trap.
In parallel experiments, we incubated these cellular lysates with the GST-Grb2-SH2 domain or with GST alone. Protein complexes were separated by SDS-PAGE, transferred to mem-  branes, and blotted with anti-phosphotyrosine antibodies (Fig.  6A). Both GST-Grb2-SH2 and GST-CD28, but not GST, coprecipitated a major phosphoprotein with an apparent molecular mass of 52 kDa, corresponding to the molecular weight of one of the isoforms of Shc expressed in T cells. We therefore repeated the experiment and immunoblotted the membrane with an anti-Shc monoclonal antibody (Fig. 6B). Both the GST-Grb2-SH2 domain and GST-CD28 precipitated Shc from activated lysates, whereas no hybridizing signal was seen in the GST lanes. In addition to Shc, GST-Grb2-SH2 coprecipitated a tyrosine-phosphorylated protein of 36 kDa, consistent with previous reports (38 -40). This protein was not observed in GST-CD28 coprecipitations, however. This raises the possibility that only a subset of Grb2-SH2-binding proteins are accessible to CD28-bound Grb2. The amount of Grb2 bound to GST-CD28 was equivalent from both quiescent and activated lysates (Fig.  6C), demonstrating that the inducible regulation of CD28 with Grb2 is lost when the cytoplasmic domain of CD28 is removed from the protein context of the extracellular and transmembrane domains and expressed as a GST fusion protein. These studies show that the GSTCD28-associated Grb2 may be present in a configuration capable of binding to the phosphorylated form of Shc. DISCUSSION We have demonstrated that CD28 is a binding target for a limited number of SH3-containing proteins. The isolated SH3 domains from Itk and Grb2 as well as the WW domain of YAP coprecipitated CD28 from cellular lysates. We have recently shown that the association of Itk with CD28 through the Itk-SH3 domain and the N-terminal diproline motif of CD28 results in the partial activation of the Itk kinase (60). In this report, we have presented data that demonstrate that Grb2 forms an inducible complex with CD28 via its SH3 domains binding to the C-terminal diproline motif of CD28. The preferred binding sites of Itk and Grb2 are therefore distinct and correspond to the two respective diproline motifs present in the cytoplasmic domain of CD28. Binding of Grb2 to CD28 does not require tyrosine phosphorylation; nor does tyrosine phosphorylation preclude Grb2-SH3-mediated interactions with CD28. We have shown that the nonphosphorylated CD28 cytoplasmic domain expressed as a recombinant fusion protein in prokaryotic cells bound Grb2 from cellular lysates in an SH3-specific manner and was independent of the Grb2-SH2 domain. Tyrosine 173 in the cytoplasmic tail of CD28 is imbedded in a motif that has been shown to be the common binding site for both the Grb2 and the C-terminal p85-SH2 domains, although the Cterminal p85-SH2 domain binds to this site with 100-fold greater affinity than does the Grb2-SH2 domain (29). We observed that when GST-CD28 was quantitatively phosphorylated, the isolated Grb2-SH2 domain could bind to CD28 presumably through this site. However, even under these conditions the SH2 domain was not required for binding as evidenced by the capacity of Grb3-3, an isoform of Grb2 lacking a functional SH2 domain, to bind to both the unphosphorylated and phosphorylated forms of CD28. Furthermore, we observed that mAb-mediated ligation of CD28 that was insufficient to induce tyrosine phosphorylation necessary for p85 binding was nonetheless sufficient to induce Grb2 binding. These data suggest that Grb2 can bind to CD28 in two distinct configurations depending on the degree of receptor clustering and the state of CD28-tyrosine phosphorylation. Initial receptor aggregation induces a CD28-Grb2 complex, which is mediated by proline-SH3 interactions. Under conditions where tyrosine 173 is phosphorylated, a second Grb2 binding site that requires the SH2 domain is created. However, CD28 can generate signals to induce IL-2 transcription in the absence of P85-SH2 association to tyrosine 173 (13,23,26). It is therefore of considerable interest to identify other regions of the CD28 cytoplasmic domain involved in protein interactions. We have demonstrated that CD28 utilizes proline-rich motifs to recruit SH3-containing proteins, providing an alternate mechanism for the initiation of signaling through CD28.
The Inducibility of Grb2-SH3 Binding to CD28 Correlates with Decreased Binding of Grb2 to Cbl-The factors that regulate the inducible interaction between Grb2 and CD28 described in this paper are presently unknown. SH3-diproline interactions are generally considered to be constitutive. For example, cytosolic Grb2 is bound to Sos in a preformed heteromeric complex (61). The interaction of Grb2 with Sos can be modulated, however, since serine/threonine phosphorylation of Sos diminishes Grb2 binding (62), whereas engagement of the Grb2-SH2 domain by a phosphopeptide can enhance the association between Grb2 and Sos (63).
The enhanced binding of CD28 to SH3-containing proteins such as Grb2 following ligation of the receptor could be a result of phosphorylation, allosteric changes in the cytoplasmic domain, or the release of another protein that blocks the interaction between CD28 and Grb2. We observed that tyrosine phosphorylation of GST-CD28 did not enhance the binding of Grb3-3 to the cytoplasmic tail of CD28, suggesting that tyrosine phosphorylation does not alter this association. Threonine phosphorylation of CD28 following phorbol 12-myristate 13acetate treatment (64) also did not affect the binding of Grb2 to CD28 (data not shown). There is currently no evidence for a constitutive interaction between CD28 and another molecule that could block the binding of CD28 to Grb2. We therefore support the possibility that an allosteric modification of the FIG. 6. Binding of tyrosine-phosphorylated Shc to GST-CD28. 2 ϫ 10 7 VCD28 cells were stimulated with anti-CD28 and anti-CD3, and lysates were incubated with immobilized GST, GST-CD28, or GST-Grb2-SH2. A, phosphotyrosine-containing proteins purified with GST-CD28 or GST-Grb2-SH2 were analyzed by immunoblotting using phosphotyrosine-specific mAb. A protein migrating with an apparent molecular mass of 52 kDa was coprecipitated from activated lysates with GST-CD28 (lane 4) and GST-Grb2-SH2 (lane 6) but not with GST alone (lane 2). B, as in A, except the blot was probed with a Shc-specific mAb. C, anti-CD3/CD28 stimulation does not affect the amount of Grb2 coimmunoprecipitated with GST-CD28. Cells were treated as in A, and the blot probed with anti-Grb2. Grb2 present in lysates from 4 ϫ 10 5 cells is shown as control. intracellular domain of CD28 exposes the diproline motif to SH3 domains following ligation of the receptor.
We provide evidence that CD28 ligation results in the redistribution of Grb2 within intracellular protein pools. The inducible interaction between Grb2 and CD28 corresponds to a concomitant decline in the amount of Cbl coimmunoprecipitated with Grb2. Cbl has previously been shown to bind to Grb2 through SH3-dependent interaction (43). These results raise the possibility that CD28 and Cbl compete for limited access to SH3 domains of Grb2 and that these two proteins bind Grb2 in a mutually exclusive manner. Cbl functions as a suppressor of Fc⑀RI signaling (65) and may regulate the anergic state in T cells (66). One positive effector function of CD28 may be to alter the amount of Grb2 present in the Cbl protein complex.
In a manner similar to the interaction of Grb2 with CD28, Grb2 has recently been reported to form a complex with the transmembrane receptor protein-tyrosine phosphatase ␣ via the Grb2 C-terminal SH3 domain and an 18-amino acid stretch localized close to the catalytic cleft of the N-terminal proteintyrosine phosphatase domain (67). In vitro binding experiments indicated that the binding of Grb2 to receptor proteintyrosine phosphatase ␣ and Sos/dynamin are mutually exclusive, reminiscent of our findings regarding the interaction of Grb2 with CD28 and Cbl.
Potential SH2 Binding Targets of the CD28-associated Grb2-The structural requirements for CD28-mediated costimulation are controversial. In some cellular systems, mutation of the PI3K binding site at 173 abrogates CD28-dependent IL-2 production, while in other systems it does not. The cytoplasmic tail of CD28 contains four conserved tyrosine residues, which, when all are mutated to phenylalanine (ALL F mutant), impairs signaling. Reconstitution of PI3K binding by a single add-back mutation at Tyr 173 in the ALL F mutant is insufficient to reconstitute costimulation (26). Add-back of tyrosine 191 within the motif PY 191 APPR that mediates binding of the SH3 domains of Grb2 to CD28 is sufficient to completely reconstitute CD28-dependent IL-2 production. Furthermore, deletion of the C-terminal 10 or 17 amino acids of CD28, which disrupts or deletes this diproline motif, profoundly impairs IL-2 production, whereas a seven-amino acid deletion, which leaves this motif intact, leads to enhanced costimulation (26,27). This region has more recently been shown to be required for the CD28-dependent tyrosine phosphorylation of the GTPase-activating protein-associated p62 protein (p62 DOK ) (68). We have demonstrated that a 10-amino acid C-terminal truncation mutant of CD28 no longer binds to Grb2 (Fig. 3D). Thus, a limited peptide sequence, which includes the proposed Grb2-SH3 binding site, is required for costimulation.
We have shown that the unphosphorylated form of bacterially expressed CD28 can bind to both Grb2 and to the phosphorylated form of Shc derived from activated T cell lysates. We propose that CD28-bound Grb2 links phosphorylated Shc to the CD28 cytoplasmic domain. The role of Shc remains elusive in T cell receptor signaling. Grb2-Sos binds to phosphorylated Shc at tyrosine 317 and thereby stimulates Ras activation in response to growth factor stimulation (69,70). Cross-linking of TCR and CD4 was observed to induce phosphorylation of both the 48-and 52-kDa isoforms of Shc (57). Grb2 has been detected in the phosphorylated Shc complexes in T cells. Shc has also been observed to bind to phosphorylated -chain of the TCR complex (57), while others have shown that this interaction is relatively inefficient (71). Recently, phosphopeptide mapping has identified a second site of Shc phosphorylation involving two adjacent tyrosines at positions 239 and 240 (72). Mutational analysis shows that these sites are not required for Ras/mitogen-activated protein kinase activation in response to IL-3 but rather may couple to pathways that regulate c-myc transcription (73). The Grb2-Shc complex may therefore link CD28 to the induction of c-myc message or to other phosphoproteins through the SH2 and PTB domains of Shc.
Creation of the Signaling Patch: Recruitment of CD28 into the TCR-CD3 Complex-The Grb2-SH2 domain can bind to the doubly phosphorylated -immunereceptor tyrosine-based activation motif, although this interaction is of lower affinity than that between Zap-70 and the -chain (71). No Sos was detected in -immunereceptor tyrosine-based activation motif precipitates, suggesting that the stoichiometry of this interaction is low or that Grb2 may be bound to a protein distinct from Sos (71). Stimulation of T cell clones by alloantigen and B7 on an APC induces a physical association between CD28 and the phospho--chain (74). This result is consistent with the observation that optimal costimulation occurs when both antigen and B7 are expressed on the same APC, which would allow this complex to form (75,76). Therefore, CD28 may be brought into the TCR-CD3 complex through a Grb2 bridge linking CD28 via its SH3 domains and the -chain by way of its SH2 domain. Alternatively, this bridge may be formed through Shc, which may bind the -chain through its SH2 domain (57) and which in the phosphorylated state binds the SH2 domain of Grb2 (69). This would avail CD28 to Src family kinases associated with the TCR complex including Lck and Fyn, which may be required for activating events such as the phosphorylation of tyrosine 173 within the PI3K binding site (77) and/or phosphorylation of the CD28-associated kinase Itk, a step required for its full activation (78). The formation of a multimeric protein complex composed of TCR, CD3, and CD28 within the contact patch between the T cell and the APC raises the possibility that integration between the TCR and CD28 signaling may occur near the plasma membrane.
While this manuscript was in preparation, Kim et. al. also reported that Grb2 binds to CD28 through the SH3 domains as well as through the SH2 domain (79). Our results are consistent with this report but extend the observations in several important ways. Kim et al. suggest that the SH3 domains of Grb2 function to stabilize the SH2-mediated interaction with CD28. We show that the interaction between CD28 and the Grb2-SH3 domains is inducible. We further propose that Grb2 may couple CD28 to tyrosine-phosphorylated signaling proteins, such as Shc, under circumstances where Grb2 is bound to CD28 strictly through the SH3 domains. There is mounting evidence that tyrosine phosphorylation of Tyr 173 , the docking site for the P85 and Grb2-SH2 domains, is dispensable for signaling (13,23,26), suggesting that protein interactions that occur independently of Tyr 173 may be required for CD28 costimulation.
We have demonstrated that Grb2 inducibly binds to CD28 through SH3 domains and propose a novel role for Grb2 to couple CD28 to tyrosine-phosphorylated proteins such as Shc or components of the TCR complex during T cell activation.