Critical Requirement for the Membrane-Proximal Cytosolic Tyrosine Residue for CD28-Mediated Costimulation In Vivo¹

The generation of Ag-specific T cell response requiresat least two distinct signals from the APC. The first signal is triggered by engagement of Ag-specific TCR withpeptide-bound MHC Ags, while the second or costimulatory signal is provided by a set of receptor-ligand interactions that occurs during T cell-APC interaction. In the absence of appropriate costimulatory signals, Ag stimulation can lead to a state of long-lasting Ag-specific unresponsiveness, termed anergy (1). One such costimulatory signal critical for the productive outcome of the immune response is provided by CD28. CD28 costimulation in the presence of a suboptimal TCR signal results in increased transcription and translation of multiple cytokines, T cell clonal expansion, and development of effector functions (2, 3, 4). Although the biological significance of CD28-mediated costimulation has been well established, the nature of intracellular signaling pathways triggered by CD28 ligation has yet to be clearly defined.

CD28 is a homodimer of glycosylated 44-kDa chains, each containing a single disulfide-linked extracellular Ig variable-like domain that belongs to the Ig superfamily. Sequence comparison among the human, mouse, and rat CD28 cytoplasmic domains demonstrates high interspecies conservation, suggesting a crucial role for this domain in signal transduction. The cytoplasmic domain of CD28 lacks any direct enzymatic activity and is therefore presumed to signal via the recruitment of other molecules. There is a YMNM motif within the cytoplasmic domain of CD28 that is considered to serve as a binding motif for the Src homology 2 domains of phosphatidylinositol 3-kinase (PI3-K)³ and Grb-2 after phosphorylation of the tyrosine residue upon CD28 ligation (5, 6, 7, 8, 9). However, the role of this YMNM motif in CD28-dependent costimulation has been a matter of controversy. For instance, it has been shown that the mutation of the human CD28 Tyr¹⁹¹ to Phe prevents CD28:PI3-K association and production of IL-2 in response to CD28 ligation in a murine T cell hybridoma (5, 10). In contrast, Truitt et al. reported that mutation of Tyr¹⁸⁹ of murine CD28 (which corresponds to Tyr¹⁹¹ of human CD28) to Phe abrogated the ligation-induced association of CD28 with PI3-K, but did not inhibit CD28-dependent IL-2 production in Jurkat cells (11). These discrepancies may reflect differences in the relative abundance or activation state of effector proteins that exist in the respective cell lines. However, as CD28 costimulation is most important for the activation of naive T cells, these studies in transformed cell lines may not reflect the biological roles of CD28 in primary T cells.

To address the role of the YMNM motif in CD28 cosignaling in primary T cells, we generated transgenic mice that express a CD28 point mutant lacking binding ability to PI3-K and Grb-2 due to the alteration of tyrosine to phenylalanine in YMNM. We assessed the function of the transgenic T cells in vitro and in vivo by breeding these mice with CD28 mutant mice. The results obtained uncovered a critical role of the pYMNM motif in vivo and further indicate that cell lines may be poor models of CD28-mediated signal transduction in naive T cells

C57BL/6 (B6) and (C57BL/6 × DBA/2)F₁ (BDF₁) mice were obtained from Sankyo Labo Service (Tokyo, Japan). CD28-deficient mice were generated as previously described (12). These mice were further backcrossed to the B6 background for seven generations and are described as B6CD28KO. Mice were bred in our facility. The experiments described herein were conducted according to the principles set forth in the Guide for the Care and Use of Laboratory Animals, Institute of Laboratory Animal Resources, National Research Council.

Anti-CD3 (145-2C11) was provided by J. Bluestone (University of California, San Francisco, CA). Anti-B7-1 (RM80) was provided by K. Okumura (Juntendo University School of Medicine, Tokyo, Japan). Anti-B7-2 (GL-1) was provided by R. J. Hodes (National Institutes of Health, Bethesda, MD). Anti-CD28 (PV-1) was described previously (13). B cell hybridomas producing anti-class II MHC, anti-B220 (RA3-2C2/1), and anti-Fc receptor (2.4G2) mAb were obtained from American Type Culture Collection (Manassas, VA), and their culture supernatants were used. Rabbit anti-mouse IgG was purchased from ICN (Costa Mesa, CA). Anti-H-2D^d (34-2-12), anti-CD4 (GK1.5), anti-CD8 (53-6.7), and anti-IgM (R6-60.2) were purchased from PharMingen (San Diego, CA). Chinese hamster ovary (CHO) cells transfected with murine B7-1 or B7-2 cDNA were provided by H. Nariuchi (University of Tokyo, Tokyo, Japan).

Murine CD28 cDNA was a gift from K. Lee (University of Miami, Miami FL). Murine CD28 cDNA was subcloned into pBluescript (Stratagene, La Jolla, CA). Mutant CD28 constructs were generated by oligonucleotide-directed site-specific mutagenesis and were verified by DNA sequencing. CD28 wild-type (WT) and mutant constructs were subcloned into a human CD2 expression cassette (14). These constructs were introduced into a fertilized mouse embryo ((BDF₁ × B6CD28KO)F₁). Each line was backcrossed with a B6CD28KO mouse five times.

Spleen cells were incubated with Abs specific for mouse class II MHC and B220 for 30 min at 4°C. Cells were applied to plates coated with rabbit Ab specific for mouse IgG and incubated for 60 min at 37°C, then nonadherent cells were collected and used as the T cell-enriched population.

Splenic T cell proliferation was assessed as described previously (15). Briefly, T cell-enriched populations (2 × 10⁵) were cultured in 200 μl of complete medium consisting of RPMI 1640 supplemented with 10% FCS, penicillin, streptomycin, 10 mM HEPES (pH 7.55), and 50 μM 2-ME in flat-bottom 96-well plates (Becton Dickinson, Franklin Lakes, NJ). The cells were stimulated in plates coated with anti-CD3 Ab (1 μg/ml) or PMA (1 ng/ml; LC Services, Woburn, MA) plus the indicated concentration of purified anti-CD28 Ab. In the experiments using CHO-B7 cells, T cell-enriched populations (1 × 10⁵) were cocultured with B7-transfected or nontransfected CHO cells that were treated with mitomycin C (50 μg/ml) on 96-well plates coated with anti-CD3 Ab (1 μg/ml). The specificity of stimulation with CHO-B7 cells was assayed by the addition of anti-B7-1 or B7-2 mAb to the cultures at 5 μg/ml. The cultures were pulsed with 0.5 μCi of tritiated thymidine for the final 8 h of the indicated culture period. In the graft-vs-host reaction (GVHR) experiments, spleen cells (2 × 10⁵) were stimulated with Con A (5 μg/ml). The cultures were pulsed with 0.5 μCi of tritiated thymidine for final 8 h of a 48-h culture.

T cell-enriched populations were suspended at 5 × 10⁵ cells/ml in complete medium and plated at 5 × 10⁵ cells/well in 48-well plates (Corning Costar, Corning, NY). The cells were cultured for the indicated culture period on plates coated with 3 μg/ml anti-CD3 Ab and soluble anti-CD28 Ab (10 μg/ml) or with PMA (10 ng/ml) and soluble anti-CD28 Ab (10 μg/ml). Supernatants were harvested, and IL-2 content was measured using an ELISA kit (PharMingen) according to the provided protocol.

Single-cell suspensions of donor spleen cells (5 × 10⁷) in PBS were injected i.v. via the tail vein into unirradiated BDF₁ hosts to induce GVHR; this time point was designated day 0. Mice were tested on day 14.

Spleen cell suspensions from donor cell-injected GVHR mice were prepared in FACS medium (PBS plus 0.1% BSA (Sigma, St. Louis, MO) and 0.1% sodium azide). Cells (10⁶/tube) were incubated first with unlabeled anti-Fc receptor (2.4G2, 100 μl of culture supernatant) to block nonspecific binding, then were stained with Abs specific for CD4, CD8, H-2D^d, B220, and CD28. Two-color flow cytometric analyses were performed on a FACSCalibur using CellQuest software (Becton Dickinson).

Anti-host cytotoxicity by GVHR spleen cells was determined directly, with no in vitro restimulation, by assessing the killing of tumor line from the H-2^d background. P815, a murine DBA/2-derived mastocytoma line, was incubated with 0.05 mCi of ⁵¹Cr for 90 min at 37°C in 33% FCS, washed three times, and diluted to 5 × 10⁴ cells to serve as a target. Single-cell suspensions of spleen cells (5 × 10⁶ cells/ml in complete medium) were serially diluted and incubated for 4 h at 37°C with target cells at four E:T cell ratios. Chromium release into the supernatant was measured by gamma counter (Wallac Oy, Turku, Finland). The percent cytotoxicity was calculated as (experimental chromium release − spontaneous chromium release)/(maximal chromium release − spontaneous chromium release) × 100%.

To study the role of the YMNM motif, a transgene encoding the mouse CD28 WT, the CD28 Y189→F mutant (Y189F), or a CD28 mutant lacking the CD28 cytoplasmic portion (transmembrane domain (TM)) was expressed using the human CD2 promoter/enhancer in CD28-deficient B6 mice (Fig. 1,A). Cell surface expression of the transgene WT, Y189F mutant, or TM mutant can be detected on splenic T cells from each of the transgenic mice at levels higher than those from B6 mice, but comparable with one another (Fig. 1 B). The experiments presented below were performed using these transgenic mice.

To test the function of T cells expressing the various CD28 transgenes, splenic T cells from each of the transgenic mice were stimulated with anti-CD28 Ab in the presence of either anti-CD3 Ab or PMA. T cells from the CD28 WT Tg proliferated weakly upon treatment with a suboptimal amount of anti-CD3 Ab (1 μg/ml) alone, and this effect was remarkably enhanced by coculture with Ab to CD28 at each time point (Fig. 2,A). At 24 h after stimulation the proliferative response of CD28 Y189F expressing T cells was minimally enhanced by anti-CD28 Ab stimulation. However, at 48 h the costimulatory effect of CD28 cross-linking on CD28 Y189F-expressing T cells became apparent (Fig. 2,A). After stimulation with anti-CD28 Abs in the presence of PMA, T cells from CD28 WT Tg showed strong proliferation, and CD28 Y189F expressing T cells also showed weak, but significant, proliferation. In contrast, CD28 TM T cells did not demonstrate CD28-mediated costimulation under any condition (Fig. 2 B).

The effect of Abs is often different from the effect of natural ligands for various receptors, and this has been shown to be the case with CD28-mediated signaling (16). Therefore, we examined the response of CD28 transgenic T cells to costimulation provided by B7-1- or B7-2-expressing transfectant cells. As shown in Fig. 2 C, the proliferative response of CD28 WT transgenic T cells was greatly enhanced by coculture with B7-expressing CHO cells at 24 and 48 h after stimulation. This costimulatory response was B7 specific, because addition of anti-B7 mAb completely blocked the proliferative response against the B7-1- and B7-2-CHO cells. Furthermore, CHO transfectants that did not express B7 failed to augment the proliferative response. In contrast, the proliferative response of CD28 Y189F-expressing T cells was weakly, but significantly, enhanced by CHO-B7 cells at 24 h, and at 48 h these T cells showed almost equivalent proliferation to CHO-B7 cells as WT transgenic T cells. These results indicated that Y189F transgenic T cells could respond not only to Ab cross-linking, but also to the natural CD28 ligand B7 in the presence of anti-CD3 Ab in vitro.

Next, we assessed the ability of CD28 to induce IL-2 production from these transgenic T cells. In response to CD28 Ab cross-linking with suboptimal CD3 stimulation, T cells expressing CD28 Y189F hardly produced IL-2 at 24 h, whereas their IL-2 production gradually increased and reached levels equivalent to the CD28 WT at 72 h after stimulation (Fig. 3,A). In contrast, CD28 Y189F T cells had a profound defect in IL-2 secretion at all time points after CD28 Ab stimulation in the presence of PMA (Fig. 3 B). As expected, CD28 TM T cells produced little or no IL-2 at any time point by anti-CD28 Ab stimulation in the presence of either anti-CD3 Ab or PMA.

We have previously shown that blockade of B7-CD28 interaction by CTLA4-Ig treatment completely aborts the development of acute GVHR (17). Analogous results were obtained when spleen cells derived from CD28-deficient mice were injected into F₁ mice (18). These results suggest that CD28-mediated costimulation plays a crucial role in the induction of acute GVHR. Thus, using the acute GVHR system, we assessed the role of the YMNM motif in CD28-mediated costimulation in vivo. Spleen cells from B6, CD28KO, or transgenic mice were injected into unirradiated BDF₁ recipients to compare the abilities of CD28 WT, CD28 Y189F, and CD28 TM T cells to induce an acute GVHR. Fourteen days after injection, the degree of donor/host chimerism in the spleen cells was assessed by flow cytometry. Following the adoptive transfer of CD28 WT spleen cells, the frequency of H-2^d-negative donor CD4 and CD8 cells increased dramatically in the spleen of the recipients, indicating the development of an acute GVHR. In contrast, donor cells did not increase in the spleen of the recipient transplanted with either CD28 Y189F cells or CD28 TM cells. These results indicate that the Y189F motif is required for T cell engraftment and alloproliferation in BDF₁ mice (Figs. 4 and 5). Furthermore, the failure of CD28 TM cells to induce GVHR is consistent with previous results with CD28 null T cells (18).

A reduction in the numbers of host B cells is another sign of acute GVHR development. As shown in Figs. 4 and 5, the frequency of host B cells (H-2^d/B220⁺) was reduced in the spleens of mice transplanted with CD28 WT spleen cells, whereas it remained intact in the spleens of mice transplanted with either CD28 Y189F cells or CD28 TM cells.

The loss of host B cells in CD28 WT Tg→BDF₁ spleens was suggestive of the presence of anti-host cytotoxic effector cells. Spleen cells were tested for anti-H-2^d activity in a cytotoxicity assay on an H-2^d tumor target (Fig. 6). The normal splenic control, BDF₁, showed no activity in this assay. The CD28 WT Tg→BDF₁ spleens had a high level of anti-H-2^d killing, equivalent to the B6 mice (Fig. 6). In contrast, both CD28 Y189F Tg→BDF₁ and CD28 TM Tg→BDF₁ spleen cells had little or no anti-host cytotoxic activity.

Acute GVHR is characterized by a T cell deficiency to respond to Ag or mitogen. Thus, the proliferative response of T cells obtained from GVHR-induced mice to Con A was tested. As shown in Fig. 7, spleen cells from CD28 WT Tg→BDF₁ mice showed a marked reduction of proliferative response. In contrast, the response was intact when CD28 Y189F Tg→BDF₁ and CD28 TM Tg→BDF₁ spleen cells were tested, as they showed mitogen responses comparable to those of untreated BDF₁ spleen cells. These results clearly indicate that the YMNM motif is absolutely required for the development of acute GVHR.

In this study we assessed the role of the YMNM motif in CD28 costimulation. We have previously shown that in vivo treatment of the host with CTLA4-Ig can completely abort the subsequent development of acute GVHR (17). Furthermore, we and others have shown that acute GVHR does not develop when CD28-deficient spleen cells are injected into F₁ mice (18). These results clearly demonstrate a pivotal role of CD28-mediated T cell costimulation for the alloantigen-mediated T cell proliferative response. The absence of GVHR development in BDF₁ mice transplanted with CD28 TM transgenic spleen cells is consistent with this idea. The fact that CD28 Y189F transgenic spleen cells cannot induce acute GVHR suggests that signaling molecules that bind to pYMNM are required for CD28-mediated intracellular costimulatory signaling following physiological receptor-ligand binding upon T cell-APC interaction.

How can one explain the observed discrepancy of the effect of the tyrosine residue mutation in the YMNM motif between the in vivo GVHR and the in vitro assay? We think that this may arise from a difference in the requirements for proliferation and IL-2 production examined in the in vitro experiment and in those for development of acute GVHR in vivo. In the in vitro system, T cell activation caused by ligation of CD28 directly assessed the effect of mutation in CD28 molecule. In contrast, in vivo during the GVHR, various interactions with other molecules on T cells and APCs control the activation and differentiation of effector T cells. It was reported that CD28 costimulation plays an important role in the up-regulation of integrin activity, and that the association of PI3-K with the YMNM motif is required for this event (19). It is also conceivable that mutation of Y189 may alter the in vivo production of various cytokines and chemokines that control the development of GVHR.

Our results suggested that the YMNM motif has critical roles in the early stages of T cell activation, whereas the requirement for the YMNM motif in CD28 costimulation diminished as T cell activation progressed. Unlike anti-CD3 and anti-CD28 Ab stimulation, PMA and anti-CD28 Ab stimulation barely induced IL-2 production from CD28 Y189F transgenic T cells at any time point. What causes the differences that were observed between PMA and anti-CD3 stimulation? TCR engagement induces many signaling events other than PKC activation. For example, cross-linking of CD3 by mAb induces increases in the intracellular calcium concentration leading to activation of calcineurin. Furthermore, upon TCR engagement, a number of signaling components appear to relocalize to lipid rafts, and this results in the efficient activation of many signaling events (20). Our results are most consistent with the idea that strong TCR-mediated signals triggered by anti-CD3 Ab cross-linking may overcome the defective costimulatory signal transmitted by Y189F mutant CD28. It has been known that TCR stimulation induces overlapping signaling events with those transmitted through the YMNM motif of CD28. For example, stimulation of TCR activates the p85-p110 type of PI3-K, leading to the production of phosphatidylinositol 3,4,5-trisphosphate (21). Grb-2 is known to associate other adapter proteins, linker for activated T cells (22) and SH2-domain-containing leukocyte-specific phosphoprotein of 76 kDa (23), through which Grb-2 couples TCR stimulation to downstream signaling events. Long-lasting TCR engagement may lead to sufficient activation of these signaling pathways, so that a strong TCR signal may overcome the defect in the CD28 Y189F T cells at the late activation stage.

For the past few years the role of this tyrosine residue in the CD28 costimulation has been extensively investigated by several groups. However, contradictory results have been obtained. For instance, it has been shown that the mutation of the tyrosine residue in the YMNM motif in human CD28 abrogated CD28-mediated IL-2 production when this mutant was transfected into mouse T cell hybridoma cell lines (5, 10). In contrast, Truitt et al. have reported that the same mutation has no effect on the ability of CD28 to deliver a costimulatory signal to Jurkat cells (11). These conflicting results may depend on the nature of the cell line used in respective studies. The result obtained from the mouse T cell hybridoma is consistent with our result in that CD28 Y189F T cells failed to produce IL-2 at 24 h after stimulation, whereas the result observed in Jurkat cells is compatible with our result in that the same T cells produced comparable amounts of IL-2 as CD28 WT T cells after 72 h. These observations may indicate that each cell line reflects different activation states of primary T cells. Thus, the biologic effects of CD28 signaling appear to be conditional to the cellular context in which they are received.

We found that CD28 Y189F transgenic T cells did not expand in the host on day 14 after the induction of GVHR, although at this time point we could not distinguish whether the CD28 Y189F T cells failed to engraft or were decreasing following transient expansion. It is conceivable that CD28 Y189F T cells may be initially activated and expanded in response to host alloantigens, but then die quickly through a process of apoptosis. As CD28 costimulation is shown to enhance T cell survival following TCR stimulation by increasing the expression of Bcl-x_L (24), the CD28 Y189F mutant may lack the ability to induce the expression of Bcl-x_L.

It has been recently shown that Akt, a serine/threonine kinase, is one of the targets of PI3-K, and that this PI3-K-Akt pathway blocks apoptosis through phosphorylation of several proteins (25, 26, 27, 28). Because the CD28 Y189F mutant no longer has the ability to bind PI3-K, this mutant may not be able to block apoptosis because of deficiency in the PI3-K-Akt pathway. We are presently studying the apoptotic response of CD28 Y189F mutant T cells.

Various other PI3-K functions in T cell activation have been proposed. Several groups have reported that actin polymerization is necessary for promoting the correct orientation and contacts between T cells and APCs (29, 30, 31), and that PI3-K is involved in cytoskeletal reorganization (32, 33). Deficient PI3-K activation in the CD28 Y189F mutant may fail to promote T cell-APC interaction and to induce efficient T cell activation through the cytoskeletal reorganization. Viola et al. demonstrated that CD28 engagement results in the redistribution of lipid rafts, membrane microdomains in which many molecules involved in TCR signaling are enriched, to the site of TCR engagements (34). As functional interaction between the cytoskeleton and lipid rafts in T cell activation has been shown (35, 36), PI3-K may be involved in the redistribution of lipid rafts by CD28 through cytoskeletal reorganization. The abnormal T cell function of CD28 Y189F transgenic mice described above can be explained by the deficiency in these roles of PI3-K in T cell activation.

It has been reported that besides PI3-K, Grb-2 binds to the YMNM motif of CD28 after the phosphorylation of tyrosine residue (9). It has been shown that mutation of Y189 to F disrupts binding of both PI3-K and Grb-2 to CD28 (9, 10). Because of this, lack of Grb-2 association to the YMNM motif may also influence the T cell function of CD28 Y189F transgenic mice. Because mutation of the methionine residue in the YMNM motif was found to selectively disrupt PI3-K binding without affecting the binding of Grb-2 (10), and mutation of arginine decreased only Grb-2 binding (37), we are generating transgenic mice expressing these CD28 mutations to distinguish the roles of PI3-K and Grb-2 in CD28 costimulation.

We thank Dr. Richard Hodes for critically reading the manuscript and for providing the GL-1 Ab. We also thank Dr. Ko Okumura for the RM80 Ab, Dr. Hideo Nariuchi for the CHO-B7 transfectants, and Dr. Kelvin Lee for the murine CD28 cDNA.