Ag recognition is an essential component for an effective T cell response. However, T cell activation is also subject to additional regulation by accessory molecules. CD28 provides essential costimulatory signals that allow T cells to proliferate, whereas molecules such as CTLA-4 and CD95 (Fas) appear to be negative regulators. Currently, which outcome predominates under conditions of antigenic challenge is poorly understood. In particular it has been suggested that one consequence of antigenic activation of T cells is the up-regulation of both CD95 and CD95 ligand, thereby exposing activated T cells to apoptotic death. We have investigated this possibility in normal human peripheral blood T cells triggered by the superantigen SEB either in the presence of endogenous APCs or transfectants expressing DR4 and CD80. In either case, we find that such activation does not expose the majority of T cells to anti-CD95-induced apoptosis as detected by annexin V externalization and DNA fragmentation. Furthermore, by phenotypically identifying, by flow cytometry, those cells that received both antigenic and costimulatory signals from those cells that did not, we observed that CD95-induced apoptosis was not seen in activated T cells receiving Ag and costimulatory signals via CD28. However, while not all T cells were stimulated by superantigen, CD95 expression was found to be homogeneously up-regulated, suggesting a mechanism whereby bystander cells might be made susceptible to CD95-induced death. We conclude that antigenic activation of T cells via the TCR and CD28 engagement provides protection from CD95-induced apoptosis.
Antigenic stimulation in T lymphocytes results in a number of functional outcomes, including activation, anergy, and apoptosis. The primary signal in T cell activation is the engagement of the TCR-CD3 complex by peptide-Ag bound to an MHC molecule present on the surface of an APC. However, in the absence of other molecular interactions, Ag-induced unresponsiveness or anergy may occur (1, 2). In addition, engagement of the TCR is also implicated in the apoptotic death of T cells via activation-induced cell death (AICD)3, which occurs upon ligation of the TCR and subsequently involves the interaction of CD95 ligand (CD95L) with its receptor CD95 (3, 4, 5). In contrast to this negative regulation, provision of additional signals via costimulatory molecules such as CD28 prevents anergy and is required for full activation and proliferation (6, 7, 8, 9). Thus the outcome of TCR engagement in terms of anergy, activation, or apoptosis via the CD95 receptor appears to result from a balance between positive and negative signals.
The CD95 receptor is a surface glycoprotein of 43,000 m.w. and is one of a growing family of molecules with homology to the TNF receptor (10, 11). The CD95 protein is widely expressed on different tissues and is up-regulated on the surface of human T cells upon activation (12). In contrast, its ligand, CD95L, has a much more limited distribution and is predominantly expressed on T lymphocytes as measured by Northern blot analysis (13). While CD95L is a type II transmembrane protein, its primary location appears to be intracellular vesicles that may be released rapidly on T cell stimulation with phorbol ester and ionophore, or anti-CD3 Abs (14). Subsequently CD95L is expressed on the cell surface and ultimately cleaved to a soluble form by metalloproteinases (15). The fact that such soluble CD95 ligand can also cause apoptosis implies a need for a level of death control above that of simple ligand-receptor interactions, since during T cell activation essentially all cells up-regulate CD95.
One receptor that may be important in coordinating T cell survival and proliferation is CD28. Stimulation via this receptor has been widely demonstrated to be important in T cell proliferation (9, 16) and recently several studies have suggested that CD28 may influence T cell survival (17, 18). While CD28 has been shown to up-regulate the antiapoptotic protein Bcl-XL, and expression of Bcl-XL in transfected cell lines may protect from CD95, a role for Bcl-2 and Bcl-XL in CD95 protection is not universally observed. Moreover, direct evidence that CD28 costimulation provides protection from CD95 in normal human T cells is lacking.
Induction of apoptosis through engagement of the CD95 receptor by soluble ligand or mAbs has been readily observed in numerous systems, including CD95 transfectants, mouse T cell hybridomas, and human T cell lines (4, 5, 19, 20). However, the situation regarding apoptosis in activated normal human T cells is somewhat less clear and studies have reported both sensitivity and resistance to CD95 (12, 21). It has been clearly demonstrated that TCR engagement induces the expression of CD95L on T cells, and that this is a major mechanism of AICD in CD95-sensitive T cells (4, 5, 20). However, given that the net result of T cell activation is largely a proliferative response and not apoptosis, generalized sensitivity to CD95-induced death is not consistent with generation of an effective immune response. To investigate both sensitivity and resistance to CD95-induced apoptosis of activated T cells, we used the superantigen staphylococcal enterotoxin B (SEB) to activate human peripheral blood T cells to allow comparisons between cells that have been triggered by Ag with cells that have not. Our studies reveal a picture of sensitivity to CD95 in which cells that have been stimulated by Ag and costimulation show characteristic enlargement, show expression of activation markers, and are resistant to CD95-induced apoptosis. In contrast, cell death via CD95 occurs predominantly in cells that do not display these characteristic changes. Thus our data suggest a model of apoptosis in normal T cells whereby correct activation, i.e., cells receiving antigenic activation and CD28 costimulation, provides protection from CD95-induced apoptosis.
Materials and Methods
T cell stimulation
Peripheral blood from healthy donors was obtained and cells isolated by Ficoll density gradient centrifugation. The buoyant layer was removed and, following several washes in PBS to remove platelets, PBMC were either stimulated by the addition of SEB (1 μg/ml) or further purified to isolate resting T cells as previously described (8). Cells were stimulated at 1 × 106 cells/ml of complete medium (RPMI 1640 plus 10% FCS and penicillin/streptomycin) in 24-well plates. Anti-CD3 (OKT3) and anti-CD28 (15-E9) Abs were coated onto plates at 10 and 2 μg/ml, respectively, and PHA was used at 1 μg/ml.
Surface receptor expression
FACS analysis was undertaken on each day of culture using the following Abs at 1 μg/ml: anti-CD3 (OKT3), anti-α/β TCR (BMA031; Immunotech, Marseille, France), anti-CD25 (HB8784; American Type Culture Collection (ATCC), Rockville, MD), anti-CD69 (55.3.1; Serotec, Oxford, U.K.), anti-CD28 (YTH913.12; Serotec), anti-CD95 (ZB4; Upstate Biotechnology, Lake Placid, NY), anti-HLA-DR (L243 HB55; ATCC), and anti-Vβ3 (8F10; Serotec). The primary Ab was added for 30 min at 4°C, cells were washed in PBS, and the secondary anti-mouse polyvalent FITC (Sigma Chemical Co., St. Louis, MO) was added for 30 min at 4°C. When dual staining was undertaken concomitantly with detection of apoptosis, an anti-mouse phycoerythrin (PE) secondary Ab was utilized to detect the surface Ag. In all, 10,000 events were analyzed by flow cytometry on a Becton Dickinson FACStarPlus using a 488-nm 100-mW laser.
Chinese hamster ovary (CHO) cells were transfected with cDNAs encoding human CD80, HLA DR4*0401β, and HLA DRα chains under the control of a CMV promoter. Cells were grown and selected as previously described (8). CHO (DR4) transfectants (5 × 106 cells/ml) were pulsed with SEB (10 μg/ml) for 4 h, washed, and fixed. All transfectants (5 × 106 cells/ml) were fixed before use by treating with 0.025% glutaraldehyde for 2 min and then washed.
T cell proliferation
Parallel proliferation studies were undertaken for all stimulation assays to verify activation. Cells were cultured in 96-well plates. A total of 1 × 105 cells were taken in triplicate each day, and 1 μCi of [3H]TdR was added overnight. The cells were harvested using a Skatron cell harvester (Skatron, Inc., Lier, Norway) onto glass fiber filters and [3H]TdR uptake was assessed by liquid scintillation counting. All experiments were done at least three times in triplicate.
Detection of apoptosis
DNA fragmentation analysis (terminal transferase dUTP nick end-labeling (TUNEL)) was conducted using an in situ cell death detection kit (Boehringer-Mannheim, Mannheim, Germany). Briefly, cells were fixed with 4% paraformaldehyde in PBS (pH 7.4) for 30 min, washed with PBS, permeabilized with 0.1% Triton X-100 in 0.1% sodium citrate (2 min, on ice), and then incubated for 60 min at 37°C with the reaction buffers provided to end label fragmented DNA with fluorescein-dUTP. Apoptosis was detected by FACS analysis where percentage increases in fluorescein-dUTP binding above control levels were measured.
For some experiments externalization of phosphatidylserine was measured by annexin V binding using the Apoptest (Nexins Research, Hoeven, The Netherlands). The apoptotic anti-CD95 Ab, CH11 (TCS Biologicals, Buckingham, U.K.), was incubated at 0.3 μg/ml with the cells for 18 h. At the termination of the experiment, the cells were pelleted at 4°C, washed in ice-cold PBS, and stained according to the manufacturer’s instructions before analysis by flow cytometry. Where surface staining was performed in conjunction with the Apoptest, the surface marker was detected using a PE-conjugated anti-mouse secondary Ab. Percent increases in annexin V-fluorescein binding above controls were measured.
Primary superantigen-activated normal T cells are predominantly resistant to CD95-induced apoptosis
To investigate the role of CD95 in the apoptotic death of normal human T cells, we used the superantigen SEB to study the effect of CD95 ligation on the resulting activated cells. We have previously characterized this system and shown that SEB activation is similar to normal antigenic activation in terms of requirements for APCs and costimulatory signals via CD28, indicating that this is a relevant model of T cell activation. As expected, we observed a robust proliferation in response to SEB stimulation (Fig. 1,A), which proceeded with the appropriate kinetics, establishing that the cultures were activated by the superantigen. In addition, T cells also expressed the expected activation markers following stimulation (Fig. 1 B), including up-regulation of CD25, CD69, CD28, and CD95 (Fas), demonstrating that these cultures expressed the characteristics of T cell activation, indicating recognition of Ag.
Since it has been suggested that T cells develop sensitivity to CD95 with time following activation, we investigated whether our activated T cells developed sensitivity during the first 8 days in culture. We therefore challenged SEB-activated T cells with the apoptosis-inducing anti-CD95 Ab (CH11) at various time points following activation and determined the percentage apoptosis by TUNEL analysis. Consistent with previous studies, we observed that before activation and in the first 48 h following activation very little apoptosis could be induced by anti-CD95 treatment (Fig. 2). Furthermore, we also observed an increase in sensitivity approximately 4 days after activation (to 25% of the culture), but more importantly we consistently observed that the majority of cells remained viable despite direct challenge with anti-CD95 Ab. These experiments were repeated with several methodologies for apoptosis detection including general viability staining with propidium iodide as well as phosphatidylserine exposure by annexin-FITC detection (data not shown). In none of these assays were we able to detect substantial CD95-induced apoptosis in normal activated T cells, yet in all cases apoptosis in control cells was seen in nearly 100% of the cells. We therefore concluded that under our conditions of normal T cell activation, the majority of T cells remained insensitive to apoptosis via CD95.
Up-regulation of CD95 in bystander activated cells
Since we observed CD95 resistance in our T cells, but also a small and reproducible apoptosis, we attempted to establish the characteristics of the cells that were either resistant or sensitive. Under normal conditions of stimulation by superantigens, only a percentage of T cells that express appropriate TCR rearrangements would be predicted to respond to stimulation. By investigating cells activated at early time points (up to approximately 96 h in vitro), it was possible by FACS analysis to distinguish cells with large forward scatter (FSC) and side scatter (SSC) characteristics (Fig. 3,A (R1)), which appeared to identify cells that had been activated by Ag, as well as cells with lower FSC and SSC characteristics (Fig. 3,A (R2), which could be considered bystander cells. To substantiate this hypothesis we compared surface receptor expression of several activation markers in these two populations (Fig. 3 B). As expected, the large scattering cells (R1) expressed the phenotypes of antigenically activated T cells, with increased levels of CD25, CD69, CD28, HLA-DR, and CD95 compared with cells in R2. Furthermore, consistent with the concept that the lower scattering cells (R2) had not been antigenically activated, the same analysis of the lower scattering cells revealed that these cells expressed little or no CD25 and had substantially lower levels of CD28 and HLA-DR molecules than R1. Surprisingly however, these smaller cells (R2) were found to have substantial expression of CD69 and virtually identical levels of CD95 expression when compared with the activated cells (R1). This indicated that cells within this lower scattering population (R2), while not expressing markers of antigenic activation, did up-regulate CD95 and CD69, suggesting that such cells might be sensitive to CD95. Overall we concluded that we could distinguish between correctly activated cells and bystander cells based on light scatter and cell surface criteria and, furthermore, that CD95 expression was up-regulated irrespective of antigenic activation.
CD28 up-regulation requires Ag and costimulation of T cells
During our experiments we also observed that the levels of CD28 expression were surprisingly heterogeneous within the first 96 h following Ag stimulation. In particular, CD28 expression appeared to be markedly up-regulated in cells productively stimulated by superantigens. To investigate whether CD28 up-regulation was a marker of T cells that had been activated correctly, we utilized transfectants where we could control both the antigenic and costimulatory properties of the APCs. Accordingly, we utilized CHO cells that had been stably transfected with either HLA-DR4 or HLA-DR4 and CD80 (B7) to provide costimulation. In this system, SEB was prebound onto the MHC molecule so that the only source of Ag was via the transfected cells. After 96 h, T cells in this system were analyzed for both Vβ3 expression (as a marker of those cells that should recognize SEB) and CD28 expression. This analysis (Fig. 4) revealed that following antigenic activation, CD28 expression was indeed heterogeneous. Cells that had been challenged by SEB in the absence of costimulation contained few CD28 “bright” cells, and those expressing Vβ3 were almost entirely confined to the population expressing resting levels of CD28. In striking contrast, the cells that had been stimulated with SEB and costimulated via CD80 revealed a substantial population of CD28 “bright” cells that now contained the entire Vβ3 population. This CD28 “bright” population also completely correlated with the large scattering activated cells seen previously. This experiment demonstrated important features of SEB-driven T cell activation. Firstly, efficient activation with SEB required costimulation via CD28, and secondly, costimulated cells responded by substantially up-regulating CD28, suggesting that cells that express high levels of CD28 are those that have been activated by both Ag and costimulation via CD28.
Activation by SEB and costimulatory molecules protects from CD95-induced apoptosis
Having established that within our SEB-activated cultures we could identify cells that had been correctly activated and costimulated, we proceeded to determine whether the cells were sensitive to CD95-induced apoptosis. As in the previous experiment, we stimulated T cells with and without costimulation using transfectants as APCs. After 72 h, we challenged these cells directly with anti-CD95 and measured apoptosis by TUNEL (Fig. 5) and at the same time stained the cells for expression of Vβ3. In accordance with our previous findings, we observed limited apoptosis in the overall culture in response to anti-CD95. However, in the cultures that were challenged with SEB in the presence of CD28 costimulation, we did not observe any apoptosis in the cells expressing Vβ3 either in the presence or absence of CD95 ligation. However, these cells were clearly activated according to the criteria described earlier, expressing high levels of CD28 and blast morphology. In contrast, cells that received no costimulation clearly showed signs of apoptosis within the Vβ3 population both in the absence of anti-CD95 Abs and following anti-CD95 treatment. Thus these data unequivocally demonstrated in a defined population that antigenic activation in the presence of CD28 costimulation provided absolute protection from CD95-driven apoptosis.
Since the percentage of Vβ3-positive cells stimulated in these SEB cultures is relatively small and to relate our findings to previous work, we also explored whether CD28 costimulation could protect in other systems. Since we had previously shown that both PHA and anti-CD3 activation of T cells was costimulation dependent (8), we utilized these systems to further explore CD95 sensitivity. T cells were therefore stimulated with PHA or anti-CD3 in the presence or absence of CD28 costimulation. Consistent with our SEB findings, this revealed (Fig. 6) that T cells that received costimulation were highly resistant to anti-CD95 with neither PHA-B7 nor anti-CD3/anti-CD28 treatment showing any CD95-induced apoptosis. In contrast, in the absence of costimulation, substantial apoptosis was observed both in the absence of and in response to direct anti-CD95 challenge. In particular, cells in the noncostimulated cultures did not up-regulate CD28, were predominantly small and displayed the nonactivated phenotypes similar to those shown in Figure 3 B (R2). Thus overall, these data confirmed our SEB findings that T cell stimulation with both antigenic and costimulatory signals resulted in activated T cells that were entirely resistant to CD95-induced apoptosis, whereas cells that did not receive these stimuli displayed sensitivity to CD95.
The discovery of the CD95 receptor as a potent inducer of apoptosis has highlighted the importance of apoptosis as a normal mechanism of homeostasis in many tissues. Within the immune system, mice that lack functioning CD95 or its ligand have a lymphoproliferative syndrome with autoimmune features providing strong evidence for the utilization of CD95 as a regulatory mechanism within the immune system (23, 24). More recently, several studies have suggested a role for CD95 as a mediator of AICD in Ag receptor-triggered T cells and the up-regulation of both CD95 and CD95L on T cells following antigenic activation, making this an attractive possibility (3, 5, 20). However, the concept that cells activated by Ag are then rapidly made susceptible to Ag-driven cell death is counterintuitive in the context of raising an effective T cell response to Ag. We therefore investigated the susceptibility of T cells to CD95-induced apoptosis in a well-characterized model of antigenic activation of primary human T cells. Our findings suggest that rather than inducing susceptibility, correct antigenic activation and costimulation are protective from CD95-induced apoptosis in the majority of T cells.
While our findings appear somewhat contrary to the current dogma concerning CD95-induced apoptosis in T cells, there are several explanations for these apparent discrepancies. Firstly, while it is clear that antigenic activation can induce apoptosis in T cells via CD95/CD95L interactions, many of these studies have been performed on cell lines that are highly CD95 sensitive and where CD95 may not be subject to normal regulation (20). In these studies it is also clear that while the majority of sensitive cells rapidly become apoptotic, the percentage of apoptotic death of normal T cells is normally substantially lower. This indicates that while some T cells are sensitive to AICD by Ag, the majority of T cells are in fact resistant, a concept consistent with our observations. Further support for this view of T cell sensitivity to CD95-induced apoptosis comes from a recent paper by Suda et al. (19), who have also demonstrated in murine studies that specifically activated T cells were CD95 resistant. Although this paper also found sensitivity among naive bystander T cells to a recombinant human ligand, these were not susceptible to apoptosis via anti-CD95 Abs.
Our studies also demonstrate that it is clearly necessary to define the phenotype of cells within T cell culture systems that are undergoing apoptosis in response to CD95. In the present study we have provided a careful phenotypic analysis of normal human activated T cells and their susceptibility to apoptosis. Our studies reveal that correctly activated and costimulated T cells undergo characteristic phenotypic changes including substantial increases in forward scatter, indicative of blastogenesis, and strong up-regulation of CD28. This is to our knowledge the first time that CD28 expression has been analyzed in the context of CD95 engagement, and our data clearly show that cells expressing high levels of CD28 are entirely resistant to apoptosis via CD95. This is consistent with the findings of Lu et al. (25), who also found that CD28 costimulation in the context of alloantigen presentation was protective from CD95L killing. Furthermore, while in many studies it is assumed that the addition of polyclonal activators results in homogeneous activation of all cells present, our experience indicates that these cultures contain cells in various states of activation, emphasizing the importance of identifying which cells are undergoing apoptosis. This is especially relevant since we have shown that both CD69 and CD95 can be up-regulated on cells that have not been correctly activated, potentially accounting for the findings of activation-induced CD95 sensitivity.
One major difference between our studies and those of others is the defined provision of costimulatory signals via CD28. Our experiments show that activation of T cells with superantigen results in strong up-regulation of CD28, provided that the T cells receive CD28 costimulation. In contrast, studies that reveal AICD via CD95L as a result of TCR stimulation are performed under circumstances in which CD28 interactions have not been provided (4, 5, 20). Thus, this lack of costimulation may promote conditions that allow CD95-induced apoptosis to proceed and further explain the lack of apoptosis seen under normal activation conditions. Consistent with such a hypothesis, we have found that under conditions of direct SEB challenge, T cells undergo a CD95-induced cell death. However, under conditions in which an appropriate APC was provided, especially in the context of CD28 costimulation, we observed proliferative responses (26). This again provides evidence that the decision between proliferation and death may be strongly influenced by the context of the TCR stimulation. This concept is also supported by the recent studies of Suda et al. (19), who reported that anti-CD3 stimulation of mouse splenocytes resulted in protection from CD95-induced death. Since spleen cells are a good source of costimulatory signals including CD28 ligands, these data are certainly consistent with our findings. It should be noted however that we have found no differences between using anti-CD95 Abs or soluble recombinant ligands in our assays. Increasingly there is emerging evidence that protection from CD95 exists in several cell types. In particular, studies on the effect of CD95 on B cells have also suggested that antigenic activation may provide protection from apoptosis (27). Furthermore, others have observed that under anergic conditions in which B cells may have been deprived of costimulatory signals, susceptibility to CD95-induced death was seen (28).
Our data are not, however, consistent with some interpretations of CD95 sensitivity. Previous studies that have investigated how T cell activation state influences apoptosis (21) have concluded that T cell activation generates an apoptosis-sensitive phenotype. However, in these studies no evidence of T cell activation was provided and the initial stimulation protocol used in these experiments involved purified T cells activated with PHA. We have shown that PHA activation is highly dependent on costimulation (8) and, in the present experiments, we observed that purified T cells responded very poorly to PHA or anti-CD3 stimulation alone, whereas costimulated cells responded vigorously. More importantly, in the presence of anti-CD95 Ab, cells that received CD28 costimulation did not undergo apoptosis compared with those for which costimulation was absent, which showed significant apoptosis. These experiments suggest that lack of costimulation in different culture systems may be a major explanation for the varying degrees of CD95 sensitivity observed.
A role for CD28 in costimulation in protection from apoptosis receives substantial support from several studies demonstrating that CD28 ligation provides important survival signals to T cells, including the up-regulation of Bcl-XL (18, 29). Furthermore, transfection studies have indicated that Bcl-XL can provide direct protection from CD95-induced apoptosis (17, 30). Other studies in TCR-transgenic mice have not demonstrated a role for Bcl-XL or CD28 in protection from AICD, but conclude that CD28 is protective from other forms of cell death (31). Overall, these studies indicate that CD28 is a potent survival factor, especially during early activation, and suggest that Bcl-XL is likely to play a role in this protection. However, our data for the first time reveal that CD28 costimulation can directly protect human T cells from CD95-induced apoptosis.
In the context of an immune response, CD95-induced apoptosis may serve to remove cells that have been nonspecifically, or incorrectly, activated while correctly activated cells are protected. Since CD95L exists largely in soluble form following activation it is possible that susceptible cells are constantly removed by CD95 engagement. Consistent with this concept, we observed much higher levels of background apoptosis in cultures that did not receive costimulation. Some of this death may be due to CD95-CD95L interactions as CD95 ligation has been demonstrated to be a major cause of cell death following SEB stimulation in vivo (32). Furthermore, SEB-induced death is more rapidly observed in mice that lack CD28, further supporting our observation that CD28 costimulation provides protection from SEB-driven apoptosis (33).
The surface up-regulation of CD28 in the generation of survival may also be significant in that we have previously demonstrated that activated T cells are capable of utilizing CD28 signals to enhance survival/proliferation in the absence of further TCR engagement (34). This is consistent with other observations within our cultures that indicate that cells susceptible to CD95-induced apoptosis express lower levels of CD28 than those that are protected, possibly indicating a role for CD28 down-regulation in predisposition to apoptosis. Interestingly, it has been observed that patients with HIV infection have both low CD28 expression as well as a highly increased susceptibility to apoptosis via CD95, as do aged T cell clones, which also down-regulate CD28 expression (35).
In summary, this study addresses the influence of CD28 costimulation on the CD95/CD95L system in T cells. We have found that under circumstances in which T cells are activated by TCR and CD28 coligation, these cells proliferate and are resistant to CD95-induced apoptosis. In contrast, “bystander cells” that up-regulate CD95 as a consequence of exposure to the local environment, or engagement of their Ag receptor in the absence of appropriate costimulatory signals acquire CD95 sensitivity. Thus in contrast to the current perception of activation-induced sensitivity to CD95-induced apoptosis, we suggest that correct activation and costimulation through CD28 generates protection from CD95-induced apoptosis.
J.D.M., L.S.K.W., and Y.P. are funded by The Wellcome Trust. D.M.S. is funded by the Arthritis and Rheumatism Council. G.B. is a Yamanouchi student.
Abbreviations used in this paper: AICD, activation-induced cell death; CD95L, CD95 ligand; SEB, staphylococcal enterotoxin B; PE, phycoerythrin; CHO, Chinese hamster ovary; TUNEL, terminal transferase dUTP nick end-labeling.