Adapter proteins such as Grb2 play a central role in the formation of signaling complexes through their association with multiple protein binding partners. These interactions are mediated by specialized domains such as the well-characterized Src homology SH2 and SH3 motifs. Using yeast three-hybrid technology, we have identified a novel adapter protein, expressed predominantly in T lymphocytes, that associates with the activated form of the costimulatory receptor, CD28. The protein is a member of the Grb2 family of adapter proteins and contains an SH3-SH2-SH3 domain structure. A unique glutamine/proline-rich domain (insert domain) of unknown function is situated between the SH2 and N-terminal SH3 domains. We term this protein GRID for Grb2-related protein with insert domain. GRID coimmunoprecipitates with CD28 from Jurkat cell lysates following activation of CD28. Using mutants of CD28 and GRID, we demonstrate that interaction between the proteins is dependent on phosphorylation of CD28 at tyrosine 173 and integrity of the GRID SH2 domain, although there are also subsidiary stabilizing contacts between the PXXP motifs of CD28 and the GRID C-terminal SH3 domain. In addition to CD28, GRID interacts with a number of other T cell signaling proteins, including SLP-76 (SH2 domain-containing leukocyte protein of 76 kDa), p62dok, and RACK-1 (receptor for activated protein kinase C-1). These findings suggest that GRID functions as an adapter protein in the CD28-mediated costimulatory pathway in T cells.
One of the emerging paradigms for signal transduction in lymphocytes is that receptors and other signaling molecules operate not in isolation but through the recruitment of a complex of other proteins (1, 2). These serve to amplify and diversify the signal into a number of biochemical cascades. This model is exemplified by the TCR, where the formation of the Ag/MHC/TCR trimolecular complex results in activation of an Src family kinase and phosphorylation of various chains of CD3, principally CD3ζ (3). ZAP70 recognizes this activated receptor and is itself activated by p56lck-mediated phosphorylation (4, 5). ZAP70 phosphorylates a number of downstream targets, such as SH2 domain-containing leukocyte protein of 76 kDa (SLP-76)2 and LAT (linker for activation of T cells) (6). These proteins are essential for directing the assembly of multiprotein signaling complexes, which are known to include other adapter proteins, such as Grb2 and Sos. Formation of these multiprotein complexes links proximal signaling events to downstream pathways such as Ras activation, Ca2+ mobilization, up-regulation of transcription factors such as NF-AT and AP-1, and ultimately the elevated expression of genes required for proliferation and differentiation.
Central to the formation of these complexes is a set of protein domains specialized for forming associations with other proteins. A large number of such motifs have been identified, including SH2 and PTB domains, which bind phosphotyrosine residues; SH3 domains, which bind PXXP motifs; leucine zippers; and many others with less well-defined ligand specificities (7). A common feature is that these individual sequence modules appear to behave as independent functional units, facilitating the evolution of proteins bearing various combinations and numbers of domains. These may be combined with a catalytic domain, such as a protein tyrosine kinase, as evidenced in the Src family of PTKs (2). Alternatively, in adapter proteins such as Grb2, the entire sequence is comprised of such modules (7, 8, 9). The sole function of these molecules appears to be the formation of associations or bridges between other proteins.
The stimulation of T lymphocytes by APC is known to require the activation of two intracellular signaling pathways. The primary signal is provided upon ligation of the TCR by MHC in conjunction with cognate peptide. Depending upon the presence or the absence of a second costimulatory signal, T cells can enter an activated or nonresponsive (anergic) state. The most potent and best characterized costimulatory signal arises from the interaction of TCR CD28 with its counter-receptors, CD80 and CD86, on the surface of APC (10, 11). Activation of CD28 is associated with the recruitment of nonreceptor PTKs, principally p56lck (12, 13, 14) and p72itk/emt (15, 16) and the subsequent phosphorylation of tyrosine residues in the CD28 cytoplasmic domain, creating ligands for proteins containing SH2 domains, formation of a multiprotein signaling complex, activation of the lipid kinase and Ras pathways, and eventually changes in gene transcription, including the up-regulation of survival factors such as IL-2 and Bcl-xL (17)
A number of studies have examined the molecular composition of the proximal CD28 signaling complex. These suggest that the major site for recruitment of downstream proteins appears to be a motif around Tyr173, which is phosphorylated by p56lck, leading to binding of the p85 subunit of PI-3-kinase (18) and Grb2 (19, 20). To identify other proteins that may be recruited to the activated CD28 receptor, we have used a CD28 cytoplasmic domain phosphorylated at Tyr173 as the “bait” in a yeast two-hybrid screen.
In this paper we describe the identification of a novel T cell adapter protein and a detailed characterization of its interaction with the costimulatory receptor CD28. This protein, termed GRID, possesses an SH3-SH2-SH3 domain structure and shares extensive homology with other adapter proteins, such as Grb2 and Grap (9, 21). GRID also contains a unique proline/glutamine-rich domain of unknown function. We demonstrate that GRID associates with CD28 in vivo following activation of the receptor and, furthermore, that it associates with a number of downstream signaling molecules, including RACK-1, p62dok, and SLP-76. Based on these findings, we postulate an important role for GRID in the CD28 costimulatory pathway.
Materials and Methods
Standard molecular biology techniques were employed throughout. The sequences of all constructs were verified by fluorescent dye terminator sequencing. Database searches were performed using the BLAST tool (22). Parental yeast trihybrid vectors, screening and selection methods, and the H9 cDNA library were essentially as described previously (23). For analysis of the CD28-GRID interaction, the cytoplasmic tail of CD28 was cloned into the binding domain yeast vector pAS1. All GRID constructs were cloned into the yeast activation domain vector pACT2. Both plasmids were cotransformed into a yeast strain (Y4.1lck) stably expressing the protein tyrosine kinase p56lck under an inducible promoter and selected on appropriate plates. Qualitative estimates of the association of CD28 and GRID were determined using methods described previously (24), while semiquantitative data were obtained from a modified liquid assay using an average of three to five independent clones (25). For identification of proteins that associated with GRID, GRID was cloned into the binding domain yeast vector pAS1 and stably transformed into the yeast strain Y190. This yeast strain was subsequently transformed with an H9 cDNA library inserted into the yeast activation domain vector.
Mutagenesis of CD28
Inserts encoding wild-type or various mutants of the CD28 cytoplasmic domain (summarized in Table I⇓) were prepared by PCR amplification from overlapping oligonucleotide templates (24). In the CD28F mutant, tyrosine 173, the principal target for p56lck phosphorylation and SH2 domain binding, was mutated to phenylalanine. The CD28-P1 mutant contains a double mutation at the first (most N-terminal) PXXP SH3 recognition motif, such that the PRRP sequence (aa 178–181) was mutated to ARRA. Similarly, the CD28-P2 mutant contains a double mutation at the second (most C-terminal) PXXP SH3 recognition motif, such that the sequence PYAP (aa 190–193) was mutated to AYAA. In the CD28-PP mutant, both PXXP motifs were mutated as described above.
Mutagenesis of GRID
Various different mutants of GRID (summarized in Table I⇑) were generated using an overlapping PCR strategy, based on domain-inactivating point mutations described by Liu and McGlade (26). The GRID amino-terminal and carboxyl-terminal SH3 were inactivated by single-point mutations converting proline residues at positions 47 and 321, respectively, to leucine residues (N-SH3* and C-SH3* constructs). A double mutant (N-SH3*/C-SH3*) was also constructed in which both SH3 domains were inactivated. The SH2 domain was inactivated by a single-point mutation converting arginine 83 to lysine (SH2*). Finally, a mutant form of GRID lacking the unique proline/glutamine domain (or insert domain) was generated in which aa 156–273 were deleted (Δinsert). The extent of the insert domain was determined by mapping the amino acid sequence of GRID onto the known three-dimensional crystal structure of Grb2 (A. Lewis, GlaxoWellcome, Stevenage, U.K., unpublished observation).
Spleen and lymph node cDNA libraries were gifts from Dr. E. Zanders (GlaxoWellcome); other cDNA libraries were obtained from Invitrogen (Groningen, The Netherlands). Each PCR reaction contained ∼2 ng of cDNA and 1 μM of each primer. GAPDH primers were: CE102, 5′-ACCACAGTCCATGCCATCAC; and CE103, 5′-TCCACCACCCTGTTGCTGTA. GRID amplifications used the following primers: CE 71, 5′-CATCGGATCCTTCCTTAGAGACAGAACCCGAGAA; and CE72, 5′-CATCGAATTCTTACCACCGCACTCGCCCTGCCGCCTG. For RT-PCR, 106 actively growing cells were lysed, and total RNA was prepared using the SV Total RNA Isolation kit (Promega, Madison, WI). RT-PCR was performed using the Access kit (Promega). Each reaction contained RNA from the equivalent of 8 × 104 cells and primers at 1 μM. Reactions were cycled using the supplied protocol for the indicated number of iterations. The primers were those described above. Analysis of GRID expression in activated T cells was performed using a variant of the T cell costimulation assay described previously (27). Briefly, CD4-positive T cells were isolated from peripheral blood, plated out into wells coated with activating anti-CD3 and anti-CD28 Abs (clones OKT3 and 9.3; both from American Type Culture Collection, Manassas, VA), and incubated at 37°C for various times up to 48 h before being harvested for RNA isolation and RT-PCR as described above. Cell activation and subsequent proliferation were assayed in parallel incubations by pulsing with [3H]thymidine after 48 h, and the incorporated radioactivity was quantitated by scintillation counting.
GST fusion proteins and mAbs
GRID wild-type and mutant constructs comprising the entire coding sequence, tagged with appropriate restriction enzyme sites, were PCR amplified and cloned into pGEX-4T3 (Pharmacia, St. Albans, U.K.) to form an in-frame fusion with the vector-encoded GST protein. Expression and purification of GST-GRID and control GST proteins were performed as previously described (24). One 8-wk-old female SJL mouse (The Jackson Laboratory, Bar Harbor, ME) was immunized on days 0, 2, 5, and 7 with 10 μg of recombinant GST-GRID using the RIMMS (repetitive immunizations, multiple sites) immunization and fusion strategy (28). On day 9 a total of 7.3 × 107 cells were isolated from pooled lymph nodes and used to prepare hybridomas. These were subjected to limiting dilution cloning, and clones were selected that secreted Ig immunoreactive for GST-GRID but not for control GST protein.
Protein complex precipitation
Actively growing Jurkat cells were washed in serum-free medium, activated by treatment with anti-CD28 and anti-CD3 Abs cross-linked with goat anti-mouse Ig antiserum for 4 min, and lysed at 2 × 107/ml in ice-cold RIPA buffer (1% Nonidet P-40, 0.5% sodium deoxycholate, and 0.1% SDS in PBS with added protease and phosphatase inhibitors). After centrifugation, the clarified lysate was precleared by tumbling with glutathione-Sepharose (Pharmacia) at 4°C for 1 h. The supernatant was transferred to tubes containing glutathione-Sepharose precharged with either GST-GRID or GST and incubated overnight at 4°C. After extensive washing in RIPA buffer, protein complexes were eluted by boiling in SDS-PAGE sample buffer, separated by SDS-PAGE, and blotted to a polyvinylidene difluoride membrane (Millipore, Bedford, MA). Precipitated proteins were visualized by enhanced chemiluminescent Western blot using antisera specific for human Sam68 (SC333, Santa Cruz Biotechnology, Santa Cruz, CA), Sos2 (SC258, Santa Cruz Biotechnology), and SLP-76 (SC1961, Santa Cruz Biotechnology).
Actively growing Jurkat (J6) cells were washed in PBS, resuspended in serum-free medium, and activated by treatment with anti-CD28 mAb (9.3) and cross-linking goat anti-mouse IgG. Cells were then lysed on ice for 15 min in Nonidet P-40 lysis buffer containing protease inhibitors (29). After centrifugation the clarified lysate was tumbled overnight with Sepharose-conjugated protein A (Sigma, P3391) at 4°C. After extensive washing with Nonidet P-40 lysis buffer, protein complexes were eluted by boiling in SDS-PAGE sample buffer, separated by SDS-PAGE, and blotted to polyvinylidene difluoride membrane. GRID or p85 association was visualized by enhanced chemiluminescent Western blotting reagents using the anti-GRID mAb (1-13.4) or polyclonal anti-p85 (Upstate Biotechnology, Lake Placid, NY; no. 06-195).
GRID-CD28 phosphopeptide ELISA
Two peptides were selected corresponding to the CD28 sequence around tyrosine 173. The control peptide, [biotin]-KLLHSDYMNMTPR, and the phosphorylated peptide, [biotin]-KLLHSDpYMNMT, were chemically synthesized ([biotin]-K indicates a lysyl residue bearing a biotin moiety, and pY indicates a phosphotyrosine residue). Nunc Maxisorp microtiter plates (Naperville, IL) were coated with 2 μg/ml of streptavidin (STAR1B, Serotec, Kidlington, U.K.), washed with Tris-buffered saline/0.1% Tween-20, and blocked with a 3% (w/v) solution of BSA in PBS. After washing as described above, a 5-μM solution of peptide was bound for 1 h at room temperature, washed, and then exposed to GST-GRID wild-type, GST-GRID SH2 domain, GST-GRID insert domain, or GST for 1 h at room temperature. Bound GST protein was quantified using a goat anti-GST primary Ab (Pharmacia) and an HRP-conjugated anti-goat secondary Ab (A5420, Sigma, Poole, U.K.). Bound peroxidase activity was visualized using a chromogenic substrate (Fast OPD, Sigma) according to the manufacturer’s instructions. The color reaction was terminated by the addition of 3 M sulfuric acid and was measured at 490 nm.
Cloning of GRID and sequence analysis
We have applied the yeast trihybrid system (23, 30) to screen for proteins capable of binding to the cytoplasmic domain of CD28 in which the tyrosine residue Y173 has been phosphorylated by coexpression of p56lck. The screen identified a number of clones from a library derived from the H9 T cell line, one of which is described in the present report.
The initial sequence obtained from the yeast two-hybrid screen was used to identify a number of ESTs from GenBank. These were used to build a composite sequence encoding an open reading frame of 993 nucleotides, starting with an ATG codon in a context suitable for translation initiation (31). PCR primers spanning the entire putative coding sequence were used to isolate full-length cDNA sequences from both a Jurkat J6 cell line and an H9 cDNA library. In both cases, a single band of ∼1 kb was obtained, cloned, and sequenced. The DNA for the Jurkat J6 cell line (J6 sequence) and the H9 cDNA library (H9 sequence) have been submitted to the GenBank database and assigned accession numbers AF236120 and AF236119.
Comparison of the J6 and H9 sequences revealed only two base changes. Although a conservative mutation was observed at position 388 of the J6 sequence, a second nonconservative mutation was also present at position 954. The codon spanning this position is CCT (leucine) in the J6 sequence and CTT (phenylalanine) in the H9 sequence.
The J6 cDNA sequence encodes a polypeptide of 38 kDa predicted Mr with extensive homology to adapter proteins such as Grb2 (9) and Grap (21, 32). This novel adapter protein was termed GRID (Grb2-related protein with insert domain). Like Grb2 and Grap, GRID consists of an N-terminal SH3 domain, an SH2 domain, and a C-terminal SH3 domain. Uniquely, however, the sequence also contains a proline and a glutamine region, lying between the SH2 and C-terminal SH3 domains. This domain, which we term the insert domain, has no strong similarity with other proteins.
During preparation of this work, a number of other workers have deposited sequences identical with GRID in the GenBank database: GrbX, accession number AF090456; GrbLG, AJ011736; Grf40, AF042380; Gads, Y18051; and Grap-2, AF102694 (26, 33, 34). Comparison of these cDNA sequences identifies a number of conservative point mutations in the coding sequence. Although the putative open reading frame from the J6 sequence is identical with those described above, the H9 sequence produces a protein sequence with a single mutation (L320F). Although the GenBank database does not contain any full-length cDNA clones that possess the GRID phenylalanine variant, two EST clones (accession no. R02185 and R08413) do contain this mutation, suggesting that the H9 sequence obtained in this study represents an authentic polymorphism in the GRID DNA sequence.
Identification of murine GRID
We subsequently used the full-length GRID cDNA sequence to rescreen GenBank for related murine nucleotide sequences. Two significant matches were obtained. The first was an EST ( accession no. AA537513; IMAGE Consortium Clone ID 949818) from the Washington University/Howard Hughes Medical Institute Mouse EST/IMAGE Consortium Project (35). The appropriate clone was obtained from the IMAGE Consortium, and the entire 1.4-kb insert was sequenced (accession no. AF236118). The results revealed an open reading frame of 966 bp encoding a protein 88% identical with GRID, possessing two SH3 domains, one SH2 domain, and, significantly, the insert domain. Based on the structure and degree of homology, we believe that this sequence represents the murine orthologue of GRID.
Human GRID gene structure
We also identified a 144-kb human genomic sequence (accession no. Z82206; submitted by the Sanger Centre Chromosome 22 Mapping Project) containing the GRID 5′ untranslated region, the entire coding region, and the 3′ untranslated region in a total of eight exons (Table II⇓). This clone is annotated as mapping to chromosomal band 22q12-22qter, whereas the other members of the family, Grb2 and Grap, lie on chromosome 17 (36) (J. H. Ellis, unpublished observations). These data in combination with a similarity dendrogram (data not shown) suggest that GRID probably diverged from a common proto-Grb2 ancestor before the divergence of Grb2 and Grap.
GRID associates with CD28 in vivo
An mAb (mAb 1-13.4) was raised against a full-length GST-GRID fusion protein as described in Materials and Methods. This Ab detected a single 39-kDa band in a Western blot analysis of Jurkat J6 total cell lysate (see Fig. 1⇓A), confirming that the mAb does not cross-react with the other family members, Grb2 and Grap.
To determine whether GRID interacts with CD28 in vivo, CD28 was immunoprecipitated from Jurkat cell lysates both before and after activation using an anti-CD28 Ab. Fig. 1⇑B shows that the amount of GRID present in anti-CD28 immunoprecipitations substantially increased upon CD28 activation, suggesting that GRID was actively recruited to the signaling complex following CD28 activation. As a positive control, CD28 immunoprecipitations were also probed with Abs to the p85 subunit of PI-3-kinase, a signaling protein that is recruited to activated CD28 (18). A substantial increase in CD28-associated p85 was also observed, although the kinetics of GRID and p85 association with CD28 were different (Fig. 1⇑C). GRID association with CD28 peaked at 1–4 min and declined thereafter, while p85 association continued to increase throughout the time course of activation.
Expression of GRID message
The most closely related genes, Grb2 and Grap, show markedly different expression patterns. Grb2 is ubiquitously expressed, whereas Grap expression is predominantly confined to hemopoietic cells (9, 21, 32, 37). We therefore investigated the tissue distribution of GRID expression. cDNAs isolated from a variety of normal human tissues were used as templates in PCRs with primers spanning a portion of the GRID sequence encoded by several exons (Table II⇑). The primers were positioned to detect any splice variants of GRID in which the exons encoding the unique insert domain were absent, which would produce a protein product very similar to Grb2 and Grap.
Specific signals were obtained only from spleen and lymph node cDNA, with extensive amplification failing to demonstrate expression in other tissues, including skin, liver, colon, muscle, and lung (Fig. 2⇓A). To refine these data further, we prepared RNA from four cell lines representing major cell types of lymphoid tissue: Jurkat T cells; two EBV-transformed B cell lines, OZZ and MAW (38); and the monocytic line Thp1. These were used in RT-PCR reactions with primers specific for GRID (Fig. 2⇓B). Of the cells tested, only the Jurkat T cell line showed expression of GRID message. These results were confirmed by Western blotting (data not shown), indicating that GRID is principally expressed in T lymphocytes. Only amplification products consistent with full-length GRID messages were observed, offering no support for the existence of splice variants.
To examine the modulation of GRID expression by external stimuli, purified human peripheral blood CD4-positive lymphocytes were activated in vitro by exposure to immobilized anti-CD3 and anti-CD28 Abs, and RNA samples obtained at a number of time points. These were analyzed for GRID mRNA content by RT-PCR (Fig. 3⇓) using conditions optimized to detect any increase in expression over the time-course of the experiment. The results show that GRID mRNA is present in resting CD4-positive lymphocytes, supported by the fact that GRID protein is also readily detectable by Western blotting (data not shown). Levels of GRID mRNA are substantially up-regulated over 48 h following activation compared with GAPDH levels, which show only a moderate increase over the same time course, consistent with cell proliferation. The results of RT-PCR and Western blotting indicate that while GRID is present in resting CD4-positive cells, consistent with its role in early activation events, the level of GRID transcript is significantly up-regulated upon activation over a period of 48 h.
Analysis of the GRID CD28 interaction
Having established that GRID is indeed a physiological binding partner for activated CD28 receptors, we undertook a detailed analysis of the molecular basis of the GRID:CD28 interaction. A series of CD28 and GRID mutants were generated and cloned into the yeast two-hybrid vectors (see Table I⇑ for details)
We used a modification of the yeast two-hybrid system in which the active form of p56lck was coexpressed under an inducible promoter to examine the importance of CD28 phosphorylation at Y173 on the binding of GRID (23). In the absence of p56lck activity no interaction between GRID and CD28 was observed (Fig. 4⇓C) while a strong interaction was detected when Lck activity was induced (Fig. 4⇓D). Comparable levels of association were detected when the C-terminal SH2 domain of the p85 subunit of PI-3-kinase was used (p85SH2-C, see Fig. 4⇓, I and J). The importance of the phosphorylation of Y173 was underlined by the inability of the CD28F mutant to bind GRID under any conditions tested (Fig. 4⇓, E and F). A quantitative form of this assay (25) demonstrated that the CD28-F mutation caused a 99% decrease in the association of GRID with CD28 (Fig. 5⇓A). Using purified GST fusion protein, we were able to detect binding of GRID to isolated tyrosine phosphopeptides that span the tyrosine 173 phosphorylation site (Fig. 6⇓). GRID was unable to bind to unphosphorylated peptides. Finally, using the yeast three-hybrid system, we were unable to observe binding of GRID to either the cytoplasmic tail of CTLA4 or CD3ζ, suggesting that GRID is specific to the CD28 signaling pathway.
Tyrosine 173 is contained within the sequence YMNM, a known SH2 (Src homology 2) recognition motif for adapter proteins such as Grb2 (19, 39, 40). The importance of the GRID SH2 domain for association with CD28 was evaluated using complementary approaches to those described above. Inactivation of the GRID SH2 domain (SH2*) by a single-point mutation completely abolished CD28 binding even in the presence of phosphorylated CD28 (Figs. 4⇑, G and H, and Fig. 5⇑). Because the isolated SH2 domain was also able to bind to CD28 phosphopeptides (Fig. 6⇑), this suggests that the SH2 domain was both necessary and sufficient for GRID-CD28 association.
SH2 domains from the Grb2 family of proteins represent a distinct structural class of SH2 domain, due to the presence of a tryptophan residue (W121) in the EF loop which sterically hinders the binding of the pTyr + 3 residue of the SH2 recognition motif (41, 42). This feature imposes constraints on the conformation of phosphopeptide ligands for the SH2 domain which are quite different from those associated with SH2 domains from Src family kinases (40). As a result of this structural feature, phosphopeptide ligands for the Grb2 SH2 fit the consensus sequence pYXNX, where pY indicates a phosphorylated tyrosine residue, and N is an invariant asparagine residue (19, 39). From a comparison of the sequences of the GRID, Grb2, and Grap SH2 domains, we predicted that the GRID SH2 domain would share this specificity. To test this hypothesis, we examined the target specificity of the GRID SH2 domain (in the context of a full-length protein) by mutating the CD28 SH2 binding motif around the Y173 phosphorylation site, aa 172-DYMNM-176. Mutations D172V and M174V had no effect on GRID binding, while the mutations N175K or M174V/N175K abolished GRID binding (data not shown). These results indicate that the GRID binding motif conforms to the Grb2 binding motif of pYXNX, consistent with our structural predictions and the finding that the asparagine at position Y+2 is essential for the SH2-mediated association of Grb2 and CD28, as reported by Kim and co-workers (19).
Contribution of the GRID SH3 domains and PXXP recognition motifs to GRID-CD28 binding
Although our data demonstrate that the SH2:phosphotyrosine 173 interaction was both necessary and sufficient to mediate association of GRID with the CD28 cytoplasmic domain, there are other possible interactions between these proteins. The association of the related adapter protein Grb2 and CD28 is stabilized by the interaction of the two SH3 domains with the PXXP motifs in the CD28 cytoplasmic tail (19, 20). To assess the contributions of these motifs to the overall interaction of GRID and CD28, we used mutants of these proteins in a quantitative form of the yeast two-hybrid system. Parallel experiments were conducted with the isolated C-terminal SH2 domain of the p85 subunit of PI-3-kinase, confirming that the mutations in the PXXP motifs did not interfere with phosphorylation of Tyr173 by p56lck or binding of this site by an SH2 domain (data not shown).
The contribution of the PXXP SH3 recognition motifs in the cytoplasmic tail of CD28 to GRID binding was evaluated using the CD28-P1, CD28-P2, and CD28-PP mutants. Fig. 5⇑B shows that whilst both PXXP motifs may contribute to GRID binding, the first motif PRRP (aa 178–181) is the most significant. We also conducted complementary experiments using a wild-type CD28 cytoplasmic domain, and variants of GRID with one or both SH3 domains inactivated. Disabling the N-terminal SH3 domain of GRID (N-SH3*) had no effect on the binding to phosphorylated CD28, while inactivation of the C-terminal SH3 domain (C-SH3* and N-SH3*/C-SH3*) reduced CD28 binding by 70% compared with wild-type GRID (Fig. 5⇑B). These data suggest that like Grb2, the overall association of GRID and CD28, was stabilized by interactions between C-terminal SH3 domain of GRID and the PXXP motifs in the CD28 cytoplasmic tail (19, 20). For GRID, the principal SH3 association site is the N-terminal PXXP motif (residues 178–181), which is in contrast to Grb2 where the most C-terminal PXXP motif (residues 190–193) is the most important.
GRID associates with other signaling proteins
To explore the spectrum of downstream proteins that GRID might serve to recruit to CD28, we investigated its ability to form complexes with other proteins known to be involved in T cell signaling pathways. Sos and Sam68 associate with the adapter protein Grb2 and are known to be involved in T-cell signaling events leading to Ras activation (43, 44, 45, 46). In an attempt to detect GRID-Sos or GRID-Sam68 interactions, we incubated Jurkat cell lysates with purified GST-GRID fusion protein. Whereas Sos2 and Sam68 were all detected when GRID was precipitated with glutathione Sepharose, none of these proteins was present when GST alone was used (Fig. 7⇓). The specific binding to Sos2 is supported by yeast two-hybrid data, which showed that GRID and GRID N-SH3* were able to associate with Sos2, while a single-point mutation in the C-terminal SH3 domain (GRID C-SH3*) completely abolished binding (data not shown).
Additionally, GRID was used as a bait in a yeast two-hybrid screen in an attempt to identify novel GRID interacting proteins. Using this approach we isolated a number of clones that were able to associate with GRID. Some of these clones represent sequences not present in the GenBank database and are currently the subject of further investigation. In addition to these novel sequences, three previously characterized clones were obtained that encoded SLP-76 (47), the GTPase-activating protein p62dok (48, 49) and RACK-1. GST and GST-GRID were incubated with lysates from either resting or activated Jurkat cells. Analysis of the coimmunoprecipitant revealed that GRID was able to associate with both SLP-76 and RACK-1 in either resting or activated cells (Fig. 7⇑).
Function of the GRID insert domain
To investigate the contribution of the GRID proline/glutamine-rich insert domain to CD28 association, binding of the GRID insert deletion mutant (Δinsert) to CD28 was quantitated and was comparable to that of wild-type GRID (Fig. 5⇑B). This observation was supported by the fact that the isolated GRID insert domain did not bind CD28 phosphopeptide (Fig. 6⇑), suggesting that the GRID insert domain did not contribute to CD28 binding. Furthermore, the absence of the insert domain had no effect on the ability of GST-GRID to immunoprecipitate SLP-76 from Jurkat cell lysates, whereas a single-point mutation in the C-terminal SH3 domain completely abolished binding to SLP-76 (data not shown).
Cloning of a novel CD28 binding protein
The activation of T cells is known to require two stimulatory signals. The primary signal is generated by the interaction of the TCR with a MHC-peptide complex on the surface of APC. Dependent upon the presence or the absence of a second costimulatory signal, T cells are either activated or anergized. The most potent costimulatory signal arises from the interaction of CD80/CD86 on APC with the T cell receptor CD28. Although some components of the CD28 signaling pathway have already been identified, we are interested in identifying novel components that contribute to the costimulatory signal. Using a phosphorylated form of the CD28 cytoplasmic domain to mimic an activated receptor, we have employed yeast three-hybrid technology to clone a novel CD28 binding protein that we term GRID.
GRID shares extensive homology with a small family of adapter proteins typified by Grb2 (50). Although it possesses an SH3-SH2-SH3 domain structure similar to that of Grb2 and the other known family member, Grap, GRID also has a proline/glutamine-rich domain (the insert domain) that lacks strong similarity to any other sequence in GenBank, located between the SH2 and C-terminal SH3 domains. Our analysis of the human genome sequence currently available indicates that GRID is encoded by at least eight exons in the 22q12 chromosomal band, in contrast to the Grb2 and Grap genes, which lie on chromosome 17. Although the exon structure of GRID is such that it would be possible to splice out the insert domain exons to leave a protein with the same architecture as Grb2 and Grap, our studies of GRID gene expression have only identified mRNA transcripts consistent with the full-length protein.
During the preparation of this study for publication, sequences identical with GRID were independently reported by a number of other groups (26, 33, 34, 51, 52). These reports confirm our finding that GRID is primarily expressed in T lymphocytes. In contrast to two reports (52, 53), using RT-PCR we find no evidence for the expression of GRID in monocytic/macrophage cell lines (Thp1) or freshly isolated human macrophage cells, either at rest or in an activated state (data not shown). This distribution is even more restricted than that of Grap (T and B cells) (32) and is very distinct from the ubiquitous expression of Grb2 (9).
GRID is a physiological part of the CD28 signaling complex
Our discovery that GRID interacts with the phosphorylated cytoplasmic tail of CD28 in vitro suggested that it may form a component of the CD28 signaling complex in vivo. Through RT-PCR studies, we have established that the GRID gene is expressed in resting naive CD4-positive peripheral blood T lymphocytes, and that expression is further up-regulated following cell activation. This profile is consistent with a gene product involved in the earliest events of T cell signaling. Conclusive demonstration of GRID’s involvement in CD28 signaling is provided by our observation (Fig. 1⇑B) that cross-linking of CD28 in intact T cells is accompanied by rapid recruitment of GRID to the costimulatory receptor signaling complex.
To dissect the GRID-CD28 interaction further, we conducted an extensive molecular analysis of the proteins using a variety of in vitro systems. Tyrosine phosphorylation at residue 173 of CD28 is absolutely essential to allow GRID to bind. Phosphorylation of this residue by p56lck (and possibly also p72itk/emt) is one of the first sequelae of CD28 activation (12, 13, 14, 15, 16). We note that this same phosphotyrosyl residue forms a key part of the binding site for the p85 adapter subunit of PI-3-kinase, raising the intriguing possibility of competition between the two adapter proteins for the activated receptor, and the scope for differential signals to arise. In support of this possibility, GRID and p85 recruitment to CD28 show different kinetics (Fig. 1⇑, B and C); the GRID content in CD28 immunoprecipitates peaks at 1 min after activation and declines to basal (or even subbasal) levels by 30 min, whereas the p85 content increases gradually over 30 min.
Targeted inactivation of the SH3 and SH2 domains of GRID by site directed mutagenesis showed that the SH2 domain of GRID is also essential for CD28 association. The SH3 domains appear to provide some additional sites of interaction, binding to one or both of the PXXP motifs in the CD28 cytoplasmic domain, but their role is subordinate to that of the SH2 domain, perhaps acting as stabilizing factors as has been reported for the Grb2-CD28 interaction (19). Deletion of the insert domain does not modulate the ability of GRID to bind CD28. In vitro, the isolated SH2 domain of GRID is sufficient to bind a phosphorylated CD28 peptide, although an isolated GRID insert domain was not. Taken together, these data indicate that GRID binds to phosphorylated CD28 primarily via a conventional SH2 domain-mediated mechanism.
What does GRID bring to the CD28 signaling complex?
To delineate the spectrum of signaling proteins that GRID may serve to recruit to a signaling complex, we and others have employed a number of biochemical techniques. GRID has been directly implicated in T cell signaling pathways due to its constitutive association with SLP-76 and its ability to enhance SLP-76-dependent signaling mediated by the TCR (34, 53). Using two independent approaches, we have also confirmed SLP-76 as a binding partner for GRID. SLP-76 was identified in a yeast two-hybrid screen using GRID as the bait. Furthermore, SLP-76 was coimmunoprecipitated with GST-GRID fusion protein from Jurkat cell lysates. Using GST fusion proteins encoding targeted mutants of GRID, we also found that the interaction of GRID with SLP-76 was dependent upon the integrity of the C-terminal SH3 domain. In contrast, mutations in the SH2 or N-terminal SH3 domains or deletion of the insert domain had no effect on the binding of GRID to SLP-76.
In addition to SLP-76, p62dok and RACK-1 were identified in a yeast two-hybrid screen using GRID as the bait. RACK-1 may anchor PKC to the cytoskeleton and function in T cell signaling pathways (54, 55). The role for PKC in CD28-mediated signaling remains controversial, with PKC inhibitors reported to have opposing effects on CD28-dependent activation of T cells (10). p62dok is a GTPase-activating protein and is phosphorylated upon activation of CD28 (49). The association between p62dok and GRID is discussed further below.
Grb2 has been shown to act as a link between activated receptors and downstream signaling pathways such as Ras activation by recruiting proteins such as the guanine nucleotide exchange factor Sos (56, 57). Using affinity chromatography and a yeast two-hybrid assay, we have also demonstrated that GRID can bind Sos2 via its C-terminal SH3 domains and the T cell signaling/adapter protein Sam68. These data suggest that GRID does have the capacity to bind Sos2, although the physiological relevance of this finding awaits further clarification.
Specificity within the Grb2 family
Because GRID shows such strong similarity to the other identified members of this family of adapter proteins, Grb2 and Grap, the question of specificity arises. Are the three proteins essentially redundant adapters, simply serving to recruit a broad set of common signaling molecules to activated receptors, or does each form complexes with a different (though perhaps overlapping) set of molecules?
Analysis of the GRID SH2 domain amino acid sequence reveals the presence of a tryptophan residue analogous to the tryptophan residue (W121) in Grb2 that defines the target specificity of the SH2 domain in this family of adapter proteins (19, 39). As a result of this structural feature, phosphotyrosine motifs bound by the Grb2 SH2 domain fit the consensus pYXNX.
Several of the reported GRID interacting proteins possess phosphotyrosine motifs that meet the Grb2 consensus, including CD28 (pYMNM) (the present study), Fms (pYKNI) (52), p66Shc (pYYND and pYVNV) (26), and LAT (pYVNV) (34, 53), suggesting that the SH2 recognition motif of GRID and that of Grb2 have broadly similar binding specificities. In our initial experimental analysis of GRID’s SH2 domain recognition properties, we found that mutation of the asparagine residue (N175) in the CD28 cytoplasmic domain abolishes GRID binding, whereas alterations at the preceding residue, Met174, have no effect, as predicted by the consensus.
However, although there are clearly similarities between the family members, there are also differences; for example, Grap does not bind the phosphorylated CD28 cytoplasmic domain in systems where GRID binding is clear (K. Fuller, unpublished observation). We also found significant differences in the spectrum of proteins that associate with the SH3 domains of these adapters. For example, p62dok and RACK-1 seem to interact specifically with GRID, showing no binding to Grb2 or Grap (C. Ashman, unpublished observations). In contrast, Sos2 appears to have the ability to bind to all family members.
These observations suggest that Grb2, Grap, and GRID are unlikely to form a redundant pool of adapters, but instead have unique properties to contribute to a signaling complex. Each provides a means to trigger common activation pathways such as those downstream of Ras, but gives a different context to the signal through the recruitment of highly specific partners. Such a multiplex approach appears highly suited to the complex signaling environment of T cells, where a number of different positive and negative signals (those arising from TCR, CD28, and CTLA4, for example), must be amplified, integrated, and interpreted before a final outcome emerges. It may therefore not be a coincidence that T cells uniquely express all three proteins, GRID, Grap, and Grb2, whereas most cells express only Grb2.
What is the role of GRID in T cells?
Because GRID expression is essentially restricted to T cells and is recruited to the CD28 costimulatory receptor upon activation, what is the specific role of GRID in T cells? We draw out two complementary hypotheses from the current information on this adapter protein.
First, we note that most of the signals associated with CD28 activation (activation of PI-3-kinase, Ras, and ceramide turnover) are common to other signaling pathways within the T cell. The best characterized proximal signal that appears to be specifically associated with CD28 triggering is the phosphorylation of p62dok; TCR activation does not result in this phosphorylation event (49). The present findings that GRID, but not Grap or Grb2, associates with both p62dok and CD28, suggest that GRID may recruit p62dok to CD28, where it becomes phosphorylated by the active Src and/or Tec family PTKs in the complex. A GRID-mediated p62dok activation may therefore provide a specific costimulatory context for interpretation of the other activation signals that arise as a consequence of receptor activation.
This model has intriguing parallels with observations from chronic myeloid leukemia cells that express a constitutively active PTK oncogene, p210bcr/abl. In one such cell line, p62dok is constitutively phosphorylated due to the activity of p210bcr/abl, even though it does not interact directly with the PTK (58). GRID has been shown to bind autophosphorylated p210bcr/abl and also to be expressed at high levels in another CML line (26), suggesting that in CML cells as well as T cells, GRID may serve as a bridge to recruit p62dok to an activating PTK environment.
Secondly, GRID has recently been reported to associate with LAT (linker for T cell activation, previously known as pp36) following activation of the T cells via cross-linking of the TCR (34, 53). Furthermore, activation of CD28 induces rapid tyrosine phosphorylation of LAT in the absence of ZAP-70 and Syk (59), which may be consistent with GRID recruiting LAT to a zone of PTK activity associated with an active costimulatory receptor. It is interesting to note that following CD28 activation GRID association with CD28 peaks at 1–4 min, while the peak in LAT phosphorylation occurs shortly thereafter (5–10 min). We postulate, therefore, that among other roles, GRID may provide the link between CD28 and LAT, acting as a crucial intermediary in the formation of a multiprotein signaling complex that serves as an integration point for signals from both the TCR and costimulatory receptors. Recent cellular studies showing that antisense oligomer-mediated down-regulation of GRID activity inhibits activation marker up-regulation following T cell stimulation (P. A. Hamblin et al., manuscript in preparation) are consistent with a role for GRID in T cell costimulatory signaling.
We are grateful to Dr. Michelle Young for advice on the GST pulldown experiments, to Dr. Paul Life for the EBV-transformed B cell lines, and to Dr. Alan Lewis for molecular modeling.
↵1 Address correspondence and reprint requests to Dr. Jonathan Ellis, Molecular Immunology Unit, GlaxoWellcome, Medicines Research Centre, Gunnels Wood Road, Stevenage, SG1 2NY, U.K. E-mail address:
↵2 Abbreviations used in this paper: SLP-76, SH2 domain-containing leukocyte protein of 76 kDa; PI-3-kinase, phosphoinositide 3′-kinase; GRID, Grb2-related protein with insert domain; EST, expressed sequence tag; PTK, protein tyrosine kinase; LAT, linker for activation of T cells; RACK-1, receptor for activated protein kinase C-1.
- Received August 30, 1999.
- Accepted March 21, 2000.
- Copyright © 2000 by The American Association of Immunologists