CD28 and CTLA-4 are related receptors that differentially regulate T cell activation. Despite the fact that they bind the same ligands, CD28 is a classical costimulator enhancing proliferation whereas CTLA-4 appears to perform negative regulatory functions. In this study, we have utilized the natural ligand for CD28 and CTLA-4 (CD80) to determine under what circumstances positive and negative effects are operative. We show here that the stimulation of purified human T cells with phorbol ester and ionomycin is inhibited in the presence of Chinese hamster ovary (CHO) cells expressing CD80. This inhibition is reversed by blocking with both anti-CD80 or Fab fragments of anti-CTLA-4 but also requires CD28 engagement. Furthermore, we show that the inhibitory function of CD80 requires elevated intracellular calcium since inhibition was observed only in the presence of ionomycin. In the absence of intracellular calcium elevation, CTLA-4 was not expressed at the cell surface, and CD80 acted positively as a costimulator of T cells, via CD28. These results demonstrate that the natural ligand CD80 can either costimulate or inhibit T cell responses depending on the conditions of T cell stimulation.
In addition to providing antigenic targets for T cells, professional APCs deliver important costimulatory signals via the CD80/CD86 interactions with the CD28 receptor (1, 2, 3, 4). T cells activated in the absence of this costimulatory signal appear to be unresponsive and may become anergic to Ags or display increased susceptibility to apoptosis (5, 6, 7, 8). However correct CD28 ligation in the presence of TCR occupancy results in T cell proliferation, enhanced survival, and cytokine production and is important in the normal functioning of T cells (8, 9, 10, 11).
In contrast, CTLA-4 is a receptor, homologous to CD28, but with apparently opposing functions in that it appears to primarily inhibit T cell responses. In line with this view, experiments using fragments of anti-CTLA-4 Abs have been found to enhance T cell proliferation (12), consistent with the blocking of a negative regulatory function, whereas agonistic cross-linking of intact Abs has revealed potent inhibition of T cell proliferation (13). However, the most compelling evidence for a negative role of CTLA-4 has been demonstrated using CTLA-4 knockout mice (14, 15), which suffer from a fatal lymphoproliferative disorder characterized by the uncontrolled proliferation of mature CD25+ T cells.
Despite their opposing functions, CD28 and CTLA-4 bind to the same ligands, namely CD80 and CD86, which are found predominantly on APCs or activated lymphocytes (16, 17, 18, 19). The most notable difference in their binding of these ligands is that CTLA-4 binds to both with a considerably greater affinity than CD28, suggesting a strong preference for CTLA-4 interactions. However, the absolute affinities of these interactions are subject to debate (20, 21, 22). In spite of this apparent preference for CTLA-4 binding, current evidence suggests that cells transfected with CD80 or CD86 ligands provide effective costimulation via CD28 and provide little evidence for ligand-operated CTLA-4 function (1, 4, 23). This may be explained in part by the substantial differences in expression and location of CD28 and CTLA-4. In this respect, CD28 is found on resting T cells with relatively abundant surface expression (24) whereas CTLA-4 is undetectable on resting T cells and reaches peak expression approximately 48 h following activation but is still expressed at levels substantially lower than CD28 (25, 26). Furthermore, CTLA-4 expression appears to be predominately intracellular, located in vesicles that are directionally targeted to the cell surface upon activation, suggesting that CTLA-4 may act after CD28 costimulation (27, 28, 29, 30, 31) although the requirement for CD28 in this process is unclear. However, data from Ab studies indicate that CTLA-4 acts early to prevent cell cycle and proliferation, despite the apparent lack of surface CTLA-4 at these time points (32, 33).
Given the complex situation regarding CD28 and CTLA-4 receptor ligation, it is important to determine under what circumstances natural ligands preferentially operate either CD28 or CTLA-4 to better understand the basis of CTLA-4 function. In this report, we demonstrate for the first time negative regulation on purified human T cells, induced by a CD80 ligand transfected on Chinese hamster ovary (CHO)3 cells. We show that T cell activation induced by PMA plus ionomycin (PI) is inhibited by CD80 in a CTLA-4-dependent manner. This interaction appears to take place early after T cell activation, and appears dependent on T cell activation signals.
Materials and Methods
T cell purification
Purified resting T cells were isolated from peripheral blood of healthy human donors. PBMC were isolated by density gradient centrifugation at 800 × g (Nycomed 1.077 g/ml), and the buoyant layer was isolated. Following several washes with PBS, the cells were further purified by adhering to plastic for 1 h at 37°C in complete medium (RPMI 1640 + 10% FCS and antibiotics). Nonadherent cells containing the T cell population were recovered and incubated with mAbs against CD14 (UCHM1), HLA-DR (L243), and CD19 (BU12). Stained cells were magnetically removed using sheep anti-mouse IgG magnetic beads (Dynal, Oslo, Norway). The remaining cells that comprised resting T cells were used in subsequent experiments. For experiments using PBMC, cells were taken following density gradient centrifugation and washing and used directly in experiments.
Chinese hamster ovary (CHO) cells transfected with the CD80 receptor were generated as previously described (23) and maintained in glutamine-free DMEM containing 10% FCS, plus penicillin and streptomycin. When confluent, cells were trypsinized and removed from the plastic flasks and counted. Before use, the cells were fixed with 0.025% glutaraldehyde in PBS for 2 to 3 min, washed extensively with medium containing 10% FCS, recounted, and utilized.
T cell proliferation assays
Proliferation assays were performed in triplicate in 96-well plates in a final volume of 200 μl per well in RPMI 1640 medium containing 10% FCS and antibiotics. T cells (5 × 104) were cultured either alone or with a combination of PMA, ionomycin, or PI with or without CHO/CHO-CD80 cell additions. Doses of PMA and ionomycin and the ratio CHO:T cells were as specified in figure legends. Assays were incubated at 37°C for 72 h, at which time 50-μl aliquots were removed for IL-2 analysis and the cells were incubated for a further 18 h with the addition of 1 μCi [3H]thymidine per well. Cells were harvested onto 96-well filter plates using a Packard (Meriden, CT) 96-well harvester, and [3H]thymidine uptake was determined via liquid scintillation counting. Where blocking Abs were used, these (or isotype-matched mouse Ig) were added at the start of culture in 50-μl volumes to the final concentrations indicated in the figure legends. Data are plotted as mean values of triplicate wells. SEs were always less than 10% of the mean value.
T cells (2 × 105) were left unstimulated or stimulated with either PMA (5ng/ml) alone or PMA (5 ng/ml) plus ionomycin (1 μM). To allow for receptor recycling and maximize staining, anti-CTLA-4 Ab (11D4, a gift from Dr P. Linsley, Bristol-Myers Squibb, Seattle, WA) or control Ab (anti-CD14 UCHM1) were added at 1 μg/ml to cultures at 37°C during the stimulation period (4 h). Subsequently, the cells were fixed in 1% paraformaldehyde for 5 min, washed, and resuspended in 50 μl of PBS containing 0.1% saponin to permeabilize the cells. Primary Abs were detected using anti-mouse IgG-FITC (Sigma, Poole, U.K.) at 1/50 dilution in 0.1% saponin, and cells were analyzed by FACS. Traces shown are representative of three independent experiments.
CD80 negatively regulates T cell activation induced by PMA and ionomycin
To examine whether CD80 could play a negative regulatory role in T cell activation, we initially began by establishing models of activation that were independent of CD80 costimulation. The fact that many forms of T cell stimulation require CD28 engagement to establish proliferation makes the study of negative regulation by CD80 difficult. To avoid this problem, we stimulated resting T cells to proliferate by using the phorbol ester PMA and the calcium ionophore ionomycin, which activates T cells in a CD28-independent manner. As shown in Figure 1, individually PMA or ionomycin were unable to cause T cell proliferation; however, as expected, when both stimuli were used, a strong synergy was observed and substantial T cell proliferation occurred.
To study the influence of CD80 on this system, we also performed these stimulations in the presence of CHO cells transfected with the CD80 receptor to determine whether costimulatory or inhibitory effects were generated (Fig. 1). As expected, we observed that PMA could effectively synergize with CHO-CD80 cells and costimulate T cell proliferation, thereby confirming the costimulatory potential of our CD80 transfectants under these circumstances. However, surprisingly, when CHO-CD80 cells were added to T cells treated with PI, a substantial negative effect on T cell proliferation was observed. While a slight decrease in proliferation was occasionally observed when using control CHO cells, this was minimal and appeared due to CHO cell interference in T cell clustering. These experiments therefore demonstrated that the CD80 ligand could potently costimulate in the presence of PMA and yet was inhibitory in the presence of PI. To further confirm the specificity of CD80 down-regulation, we performed titration experiments of different numbers of CHO-CD80 cells compared with CHO controls (Fig. 2,A) and in addition used an anti-CD80 Ab (BB1) to block the CD80 receptor (Fig. 2 B). The results of these experiments supported our initial findings that PI stimulation was substantially, although not completely, inhibited by CHO-CD80 in a dose-dependent manner, and this down-regulation was reversed by anti-CD80 Abs, confirming the dependence of this response on the CD80 molecule.
Decreased proliferation is not due to more rapid kinetics of PI-CD80 responses
One possibility that might explain the decreased proliferative response was that responses to PI and CD80 are more rapid than those to PI alone. Thus, while 72 h represented the peak of proliferation to PI, the response to PI-CD80 may have been more rapid and therefore missed. To investigate this possibility, we performed kinetic experiments measuring proliferation at a number of time points. The results from these experiments, shown in Figure 3, indicated that responses to PI-CD80 were reduced at all time points measured. Thus down-regulation of responses by CD80 appeared due to a distinct mechanism that prevented T cells from proliferating to PI and not to a more rapid response.
PI but not PMA stimulation causes up-regulation of CTLA-4 surface expression
The above experiments suggested the possibility that, under conditions of PI stimulation, CD80 may be acting to inhibit responses via CTLA-4 whereas, using PMA alone, CD28 costimulation predominated. To test this possibility, we investigated whether stimulation with PI was capable of inducing the expression of CTLA-4 at the cell surface. Accordingly T cells were stimulated for 4 h with either PMA, PI, or left unstimulated and stained using an anti-CTLA-4 Ab. This analysis (Fig. 4) revealed that CTLA-4 expression was indeed detected only under conditions of PI stimulation but not in unstimulated cells or in cells treated with PMA alone. Despite the fact that surface expression of CTLA-4 was limited, this level of expression was in line with other studies indicating low levels of CTLA-4 at the cell surface. Thus these data supported the possibility that CD80-mediated inhibition in the presence of PI may be due to the expression of CTLA-4 and that PMA alone did not promote CTLA-4 surface expression, thereby allowing costimulation to predominate.
CTLA-4 and CD28 blockade increases PI-CD80 responses
To further test our hypothesis that CD80-mediated inhibition was via CTLA-4, we next attempted blocking experiments using Fab fragments of anti-CTLA-4 to determine whether the inhibition by CD80 could be reversed by blocking CTLA-4. This experiment (Fig. 5,A) revealed that the addition of CTLA-4 Ab resulted in enhanced proliferative responses in PI-CD80-treated cells and strongly supported the hypothesis that CD80-mediated inhibition was via CTLA-4. Furthermore, we also observed that the activation of T cells induced by PI alone was slightly enhanced in the presence of CTLA-4 Fab, suggesting that endogenous CD80/CD86 might also be present on activated T cells and down-regulate T cell activation via CTLA-4 under these circumstances. In addition to these experiments, we also investigated the effects of Fab fragments of CD28 Ab in this system to determine whether CD28 played any role in down-regulation of responses (Fig. 5 B). As predicted, anti-CD28 could entirely block PMA-CD80 responses, demonstrating that this response was CD28 dependent. However, the effects of CD28 Fab were also marked on PI-CD80 responses and significantly affected down-regulation of responses to PI-CD80. The most likely explanation for this result is that CD28 is either providing effective adhesion for the T cell-transfectant interactions and thereby also permitting CTLA-4 interactions, or CD28 signals are in some way required to promote CTLA-4 expression. Indeed, these mechanisms are not mutually exclusive. Nonetheless, from these experiments we concluded that both CTLA-4 and CD28 molecules were involved in allowing suppression of responses by natural ligand, possibly reflecting an initial requirement for CD28 engagement followed by CTLA-4 inhibition.
CD80 inhibition is most effective early after activation
Having established that PI-CD80 stimulation was capable of inhibiting proliferative responses as measured at 72 h, we next performed experiments aimed at establishing the time points at which CTLA-4 exerted its effects. The fact that cross-linked CTLA-4 Abs have been found to prevent T cell activation by CD3 and CD28 has suggested that CTLA-4 may act early; however, this contrasts with expression studies that reveal maximal expression approximately 48 to 72 h following activation. To investigate the effects of CD80 on this regulation, T cells were activated using PI, and the addition of CD80 transfectants was delayed to determine at which times inhibition was maximal. As shown in Figure 6, these experiments revealed that the addition of CHO-CD80 cells at the start of the culture resulted in negative regulation, as expected. Interestingly, delaying the addition of CD80 by 2 h resulted in a more profound inhibitory effect; however, this became less potent with further delays in the addition of CHO-CD80. These results suggested that CTLA4 inhibition acts at an early stage following T cell activation, but that there is a short time delay before maximal inhibition is seen, most likely allowing for CTLA-4 surface expression to be established.
Inhibition of endogenous CD80/CD86 enhances responses to PI
Based on the findings above, using transfected CD80, we wished to establish whether this phenomenon also occurred via endogenous ligands expressed by APCs. If so, this would predict that CD80/CD86 molecules expressed on APCs should be capable of down-regulating the response to PI via CTLA-4. To test this hypothesis, we next stimulated whole PBMC preparations containing endogenous APCs with PI and studied the effect of blocking CD80 and CD86 with CTLA4Ig. In accordance with our predictions, this experiment (Fig. 7) revealed a substantial increase in proliferation to PI when CTLA4Ig was present but not in the presence of control mouse Ig. Significantly, enhancement of responses was observed only at higher concentrations of ionomycin, suggesting that, at more modest levels of stimulation, CTLA-4 may not be effective. These results were in stark contrast to the immunosuppressive effects of CTLA4Ig seen during CD28-dependent costimulation but strongly supported our hypothesis that, under conditions of PI stimulation, a substantial response to natural ligands occurs via CTLA-4. Thus, under these circumstances, blocking of CD80 and CD86 reveals their negative regulatory functions.
The role of CTLA-4 as a negative regulator of T cell activation has been convincingly documented both in vivo and in vitro (13, 14, 15, 25). However, while it is clear that Abs directed to CTLA-4 can efficiently inhibit T cell activation, similar studies using natural ligands have not been forthcoming, despite this being a central prediction. On the contrary, where T cell activation has been studied using CD28/CTLA-4 ligands, only the costimulatory function of CD28 has been observed (1, 4, 23). In this study we have demonstrated that, during stimulation of resting T cells with PMA, CD80 provides potent, synergistic costimulatory signals via CD28 resulting in T cell proliferation while, in contrast, the addition of ionomycin allows inhibition of responses associated with CTLA-4 expression. Thus, this study represents the first demonstration of both positive and negative effects of the natural ligand CD80 on T cell proliferation.
One important feature of this system, which allows the study of negative regulation by CD80, is that CD28 costimulation is not absolutely required for the induction of proliferation. Consequently, stimuli for proliferation and cell cycle induction are provided by PI, and the effect of CD80 can be either to costimulate or inhibit proliferation according to the balance of CD28 vs CTLA-4 signals. Notably, only PI-, but not PMA-, stimulated T cells are negatively regulated by CD80. The most likely explanation for this, consistent with the data presented here, is that CTLA-4 expression is dependent on a sustained calcium signal provided by ionomycin. Thus, only under PI stimulation is CTLA-4 detectable at the cell surface. This is in line with data from at least two other studies (27, 34) that have demonstrated a requirement for calcium signaling in both messenger RNA induction and in surface expression. In particular, Linsley et al. (27) showed that ionomycin was the most potent signal at inducing surface expression of CTLA-4, while others have shown increased CTLA-4 messenger RNA following PI stimulation of resting T cells (35). Importantly, we have also observed that the addition of cyclosporin A (CsA) leads to a reversal of CD80-induced down-regulation and restores proliferation to the level of PMA and CD80 combined (G. Boulougouris, manuscript in preparation). This finding is highly consistent with the effects of cyclosporin A on CTLA-4 expression that have been described in previous studies (28, 36).
The preference of CD80 for CTLA-4 may be enhanced by a decrease in CD28 expression following PI stimulation that we observed was more pronounced in the presence of CD80 (L. S. K. Walker, unpublished observations). It has previously been demonstrated (37), and we have consistently found, that CD28 is down-regulated by CD80 ligand. Thus the combination of PI plus CD80 may serve to increase CTLA-4 expression while simultaneously decreasing CD28 expression, strongly indicating that the final outcome is influenced by the relative levels of the two receptors. While previous reports have indicated that CTLA-4 does not appear substantially at the surface until approximately 72 h after activation (25, 28), this does not exclude the possibility that CTLA-4 is present in functionally significant amounts at earlier time points. Indeed, recent data in support of this concept have shown functional CTLA-4 activity at time points where no CTLA-4 is detectable at the cell surface (32, 33). Thus, despite the limited expression of CTLA-4 observed in our experiments, our results appear generally compatible with the findings of others.
Our results of using Fab fragments were also informative about the balance between CD28 and CTLA-4. Our experiments reveal that, while CTLA-4 is clearly involved in suppressing PI responses, CD28 interactions are also important in promoting down-regulation. Given that CD80 binds to both ligands, this is perhaps not surprising although it might appear paradoxical that CD28 appears to be involved positively in PMA-CD80 responses yet negatively in PI-CD80 responses. However, we do not believe that CD28 is acting in a negative way in these experiments and prefer the explanation that CD28 interactions may be important in promoting CTLA-4 expression and function. There are several possible mechanisms for this. First, it is highly likely that CD28 is involved in enhancing contact between CD80 transfectants and T cells, and this can easily be demonstrated in adhesion assays. A second possibility is that CD28 signals may actually be a requirement for effective expression of CTLA-4, as has been previously suggested (28). Third, CD28 down-regulation as a result of ligand engagement may also be involved in shifting the balance toward CTLA-4. Clearly, additional experiments will be required to dissect these possibilities.
Our results also demonstrate that CD80 inhibition is most effective approximately 2 h after PI activation. This time delay may be explained by the requirement for de novo CTLA-4 synthesis and transport to the cell surface since CTLA-4 mRNA is not observed in resting T cells (34) but is rapidly induced following PI stimulation. After this period, CTLA-4 inhibition becomes less effective, with further delays of CD80 addition following PI stimulation, possibly indicating that once cells have been committed to cell cycle CTLA-4 ligation is less effective at inhibiting responses. Thus it appears that CTLA-4 may act as an “off switch” rather than a “brake” for T cell proliferation. In support of this interpretation, we have observed only stimulatory effects when challenging previously activated and cycling T cell blasts with CD80 transfectants (38), indicating that, once T cells are proliferating, CD28 costimulation may predominate.
Interestingly, during these studies, we observed that the level of inhibition of responses by CD80 was somewhat variable (between approximately 30% and 90%). One explanation for this is the fact that proliferation analysis measures the net result of a balance between costimulatory and inhibitory outcomes mediated by CD28 and CTLA-4. Thus, an overall inhibitory result is revealed only when inhibition outweighs proliferation, and thus even small decreases in proliferation may reflect inhibition of a significant population of cells. In contrast to CTLA-4 Ab-based studies, natural ligands are clearly agonists of both costimulation and inhibition. Our experience suggests that, under conditions of limited stimulation (as reflected by intracellular calcium concentration), addition of CD80 is generally costimulatory whereas under increased stimulation a negative effect is revealed. Thus, more potent signals appear to promote greater CTLA-4 expression and thus influence the balance in favor of CTLA-4 ligation, as has also been suggested by others (27, 28). This concept, that high intensity signals are more susceptible to modulation by CTLA-4, is somewhat counterintuitive since CTLA-4 might therefore potentially inhibit higher affinity T cell interactions. However, similar mechanisms for screening out high affinity interaction appear to be used in thymic selection (39) to remove potentially autoreactive clones, and such a use of CTLA-4 peripherally may represent a similar protective mechanism from the consequences of autoimmunity, as seen in CTLA-4-deficient mice (14, 15). Overall, our studies reveal both positive and negative influences of CD80 in the regulation of T cell proliferation and provide new insights as to how the decision between the use of CD28 and CTLA-4 is made.
We thank Drs. P. S. Linsley and C. H. June for generous gifts of anti-CTLA-4 and anti-CD28 Fab fragments, respectively.
This work was supported by the Wellcome Trust (J.D.M., L.S.K.W., and Y.P.). G.B. is a Yamanouchi student. D.M.S. and C.E. are funded by the Arthritis Research Campaign (ARC). D.M.S. is an ARC Senior Research Fellow.
Abbreviations used in this paper: CHO, Chinese hamster ovary; PI, PMA plus ionomycin.