Role of CD8 in Aberrant Function of Cytotoxic T Lymphocytes

The generation of the IASA-YIPSAEK(ABA)I-specific CTL clones S4, S14, S17, and T1 and their culture conditions have been described previously (4, 17). Antigenic activities of peptide derivatives were determined in a ⁵¹Cr-release assay as described (4, 17). In brief, ⁵¹Cr-labeled P815 cells (2 × 10³ per well) were incubated in DMEM supplemented with FCS (5%) and Hepes (10 mM) with 10-fold dilutions of peptide derivative ranging from 10⁻⁶–10⁻¹³ M. Cloned CTL (6 × 10³ per well) were added and after 4 h of incubation at 37°C the ⁵¹Cr content of supernatants was determined. The specific lysis was calculated as 100 × (experimental−spontaneous release)/(total− spontaneous release). The relative antigenic activities were calculated by dividing the concentrations of IASA-YIPSAEK(ABA)I required for half-maximal lysis by that required for the variant peptide derivatives. For sake of comparison the relative antigenic activities of the compounds were normalized by dividing by the relative K^d competitor activity, which expresses the ability of a peptide derivative to bind to K^d (4, 17). In experiments where Fab′ fragments of the anti-K^d α3 mAb SF1-1.1.1 were used, and in the antagonist assays, the peptide derivatives were covalently attached to P815 cell-associated K^d (see below).

The CTL were incubated with sensitized target cells under the same conditions as for the cytotoxic assay, except that peptide derivative concentrations ranged from 10⁻⁶– 10⁻¹¹ M. After 24 h incubation the IFN-γ content in supernatants was assessed by ELISA as previously described (18).

Synthesis and photoaffinity-labeling procedures were performed essentially as described (4, 17, 19). In brief, for photoaffinity labeling of cell-associate K^d molecules, P815 mastocytoma cells were incubated with the respective peptide derivatives for 60 min at 37°C, followed by UV irradiation at ⩾350 nm. Preparation of ¹²⁵I-labeled soluble covalent K^d-peptide derivative complexes was performed as described (4). The stock solutions had concentrations of 1–5 × 10⁸ cpm/ml and specific radioactivities of ∼2,000 Ci/mMol.

TCR photoaffinity labeling was performed as described (4, 15, 17, 19), except that the samples were incubated at 26°C for 30–60 min before UV irradiation at 312 ± 40 nm. For TCR-ligand complex dissociation experiments CTL (6 × 10⁶–2 × 10⁷ cells/ml) were preincubated in medium with ¹²⁵I-labeled K^d-peptide derivatives complexes (1–6 × 10⁷ cpm/ml for 90–150 min at 0–4°C or for 30–60 min in medium containing SF1-1.1.1 Fab′ (20 μg/ml). Aliquots (100–700 μl) were added to 10-ml volumes of 26°C medium containing 0.7% FCS and anti-K^d α1 mAb 20-8-4S (10 μg/ml), which prevents TCR-ligand binding (15), and incubated at 26°C for the indicated periods. After UV irradiation for 20 s with a 500 W UV B lamp (Dr. Hönle Inc., Munich, Germany, equipped with an infrared and an UV 296-nm filter), TCR were immunoprecipitated and analyzed by SDS-PAGE. TCR photoaffinity labeling was quantified by phosphor-imaging and normalized relative to the wild-type as described (4). Mean values and standard deviations were calculated from three to six independent experiments.

TCR-ligand binding was determined by a direct binding assay as previously described (4), except that the incubations of CTL (1.2 × 10⁶ cells/200 μl) with ¹²⁵I-labeled K^d-peptide derivative complexes (3–10 × 10⁶ cpm/incubation) were performed for 30–60 min at 26°C. Specific TCR-ligand binding was calculated by subtracting from the total cell-associated cpm, the cpm measured in presence of anti-K^d mAb 20-8-4S (10 μg/ml). The specific binding (cpm) for the wild-type (wt) ligand for a given clone was referred to as 1 and the binding of the variant ligands expressed relative to this value. Due to high background the difference between TCR-ligand binding in the presence and absence of SF1-1.1.1 Fab′ was assessed by TCR photoaffinity labeling. Each experiment was performed in triplicates and repeated at least twice.

To assess the significance of TCR-ligand complex dissociation for CTL function, we selected two cases, in which antigen recognition was more efficient than TCR-ligand binding (strong agonists) (peptide derivative variant P255A on T1 CTL and S256A on S17 CTL) and two examples where recognition was less efficient than binding (weak agonists) (P255S on S17 CTL and E258A on S14 CTL) (Fig. 1,a). Since perforin dependent cytotoxicity, as mainly assessed in the used chromium release assay, is a rapid CTL response and takes place at very low peptide concentrations (20, 21), we also assessed the interferon-γ (IFN-γ) response, which requires higher peptide concentrations and sustained TCR signaling for extended periods of time (21). In spite of these differences both CTL responses were remarkably similar, except that on S14 CTL for E258A the IFN-γ production was more efficient than cytotoxicity (Fig. 1 b). Occasional divergences between cytolytic and IFN-γ CTL responses among peptide variants have also been observed in other systems (24).

Assessment of the kinetics of TCR-ligand complex dissociation showed that for both strong agonists the dissociation was significantly faster than for the wild-type ligand (Fig. 2, a and c). In contrast, for the two weak agonists dissociation was markedly slower, even though in one case (E258A on S14 CTL) TCR-ligand binding was weak (Fig. 2, b and c). For sake of increased accuracy, these kinetics were assessed at 26°C. With S17 CTL kinetics were also measured at 37°C (Fig. 2 d), showing that on living CTL TCR-ligand complex dissociation increases considerably with temperature, as has been reported for T1 CTL (15). These findings are consistent with the concept of serial TCR engagement (21, 23, 24), namely that CTL activation depends on the frequency of serial TCR engagement, which is determined by the rate of TCR-ligand complex dissociation.

To test this hypothesis, we repeated these experiments in the presence of Fab′ fragments of the anti-K^d α3 mAb SF1-1.1.1, which prevents participation of CD8 in TCR-ligand binding, but not CD8 dependent adhesion (15). Remarkably, in the presence of this reagent the efficiency of antigen recognition (cytotoxicity) was often unchanged or even increased (Fig. 1,a). This was also true for IFN-γ production; indeed, in two instances blocking of CD8 substantially increased this response (wt and P255S on S17 CTL) (Fig. 1,b). However in other cases SF1-1.1.1 Fab′ dramatically reduced the cytotoxic or IFN-γ response (P255A on T1, wt and E258A on S14 CTL) (Fig. 1). With the exception of the latter cases, blocking of CD8 reduced TCR-ligand binding more than it decreased the efficiency of antigen recognition, as a result the efficiency of antigen recognition was increased relative to TCR-ligand binding; i.e., agonists or weak agonists were converted into strong agonists (Fig. 1). TCR-ligand complex dissociation was always considerably faster in the presence of SF1-1.1.1 Fab′ than in its absence (Fig. 2). This is consistent with reports showing that CD8 stabilizes TCR-ligand binding by decreasing dissociation of TCR-ligand complexes (15, 16). Our finding that acceleration of TCR-ligand complex dissociation increased the relative efficiency of antigen recognition supports the concept that CTL activation depends on the frequency of serial TCR engagement. It has been reported that CD8 expression on CTL can be downmodulated by antigen (25, 26), which is likely to have similar effects as the CD8 blocking experiments described here.

Exceptions to this rule were the cases where blocking of CD8 participation in TCR-ligand binding dramatically impaired antigen recognition. As it has been shown that blocking of CD4 can convert agonists into antagonists (7, 13), we examined whether, in the presence of SF1-1.1.1 Fab′, the variants P255A and E258A antagonized the cytotoxic responses of T1 and S14 CTL, respectively. As shown in Fig. 3, this was indeed the case. In the presence of this reagent the recognition of IASA-YIPSAEK(ABA)I by T1 (Fig. 3,a) and S14 CTL (Fig. 3,b) was significantly reduced on target cells that expressed covalent K^d-P255A or K^d-S256A complexes, compared to target cells that expressed irrelevant K^d-P255H complexes (4). Consistent with these findings we observed in both cases predominant pp21 ξ chain phosphorylation in the presence, but not in the absence of SF1-1.1.1 Fab′ (data not shown). This form of ξ chain phosphorylation has been shown to be preferentially or exclusively induced by TCR antagonists (1, 5–7). Remarkably, in both cases exceedingly fast TCR-ligand complex dissociations were observed (Fig. 2, e and f). The only other comparably fast dissociation was recorded for variant P255L on S14 CTL, which is an antagonist for this clone (Fig. 2 b and reference 4). These observations are in agreement with the kinetic proofreading concept (8, 9) and a study by Lyons et al. (10) demonstrating that TCR-ligand complex dissociation is more rapid for antagonists than for agonists. Taken collectively our findings indicate that acceleration of TCR-ligand complex dissociation increases the efficiency of antigen recognition by CTL up to a critical threshold, beyond which TCR engagement results in aberrant TCR signaling and TCR antagonism.

It is interesting to note that the half-life of TCR-ligand complexes varied considerably among different CTL clones (Fig. 2, a–c), which may explain why certain CTL clones are more prone to TCR antagonism than others (4). For example with S17 CTL, which exhibited remarkably slow dissociation kinetics, TCR antagonism was never observed, even when testing large panels of epitope variants, whereas the opposite was true for S14 CTL and, to a lesser extent, T1 CTL, especially in the presence of SF1-1.1.1 Fab′(Figs. 1 and 2, reference 4 and unpublished data).

A striking finding of our study is that the contribution of CD8 to TCR-ligand binding varied not only among CTL clones, but also among epitope variants on given clones (Fig. 4,a). While the former differences can be explained in part by variations in CD8 and TCR expression (Fig. 4,a), the latter indicate that epitope modifications can alter the avidity of CD8 participation in TCR-ligand binding. This explains, at least in part, why blocking of CD8 accelerated TCR-ligand complex dissociation in a diverse manner, and why one cannot reliably predict from the avidity of TCR-ligand binding the dissociation rates and hence the functional phenotype of epitope variants (Figs. 1 and 2 and reference 4). Moreover, several ligand variants displayed different ratios of TCR α versus β chain photoaffinity labeling (Fig. 4 b). Similar findings were obtained at 0–4°C (unpublished data), i.e., in the absence of metabolically active cellular processes, suggesting that epitope modification can induce either conformational changes in TCR or slightly alter the orientation of TCR-ligand binding, as a result of which CD8 may participate more or less avidly in TCR-ligand binding. This view is consistent with the observation that anti-TCR mAb and their Fab′ can substantially affect T cell responses (11, 12), and suggests that the conformational changes induced by such reagents, can affect T cell function by interfering with TCR-coreceptor cooperation. It is also interesting to note that TCR-ligand binding apparently can induce structural changes in the α3 domain of MHC class I molecules, which may alter their interaction with CD8 (27, 28).

The present study shows that the half-life of TCR-ligand complexes is of critical importance for CTL activation and is determined to a considerable degree by the coreceptor CD8. The finding that epitope modifications can alter the participation of CD8 in TCR-ligand binding, and hence TCR-ligand complex dissociation, reveals a new principle by which CD8 can affect CTL function. However CD8 plays other roles in CTL function as well, such as mediating CTL-target cell adhesion and participation in CTL activation by CD8-dependent signaling (29, 30), which must also be taken into account for a comprehensive understanding of the role of CD8 in normal and aberrant CTL function.

We thank C. Horvath and P. Bassanini for excellent technical assistance, Dr. C. Servis for peptide and conjugate synthesis, Dr. P. Romero for helpful discussions and Anna Zoppi for preparing the manuscript.