Peptide Modification or Blocking of CD8, Resulting in Weak TCR Signaling, Can Activate CTL for Fas- But Not Perforin-Dependent Cytotoxicity or Cytokine Production¹

CD8⁺ CTL recognize, by means of their TCR, antigenic peptides bound to MHC class I molecules on the surface of target cells (1, 2). Intercellular TCR-ligand engagement typically results in activation of the src tyrosine kinase p56^lck, and p59^fyn, which are primarily responsible for receptor phosphorylation events that initiate downstream signals, such as the recruitment, phosphorylation, and activation of ZAP-70, phospholipase C[γ (PLC-γ),³ phosphatidylinositol 3-kinase, and Ras (3, 4, 5). CTL activation generally elicits three different effector functions. The first and most prominent function is perforin-dependent cytotoxicity. This lytic principle involves exocytosis of preformed CTL granules containing pore-forming perforin and various esterases called granzymes, which in a synergistic manner induce cell death (1, 2). Second, CTL, upon activation, express Fas ligand (FasL), which binds to Fas, the receptor for FasL, present on most cells, and thus induce apoptosis via a well-characterized signaling mechanism (1, 2, 6, 7, 8). FasL, to exert its activity, can be cell associated, or secreted in soluble form (9). While the cellular mechanisms leading to apoptosis are universal, the induction of FasL expression is not, and in different systems different signaling pathways seem to exist (10, 11, 12, 13, 14). Third, activation of CTL usually results in production and secretion of cytokines, such as TNF-α, IFN-γ, and various ILs (1, 2, 7, 15, 16).

While in general CTL activation by TCR engagement elicits all three of these effector functions, modification of antigenic peptides can result in the activation of only some of them (7, 17, 18). For example, it has been shown that altered self peptide ligands can activate Fas-dependent killing in the absence of perforin-dependent cytotoxicity (7, 17, 18). This suggested that selective activation of Fas-dependent apoptosis may play a role in eliminating cells expressing altered autologous proteins (7, 17). There is, however, no evidence that altered peptide ligands preferentially induce this form of partial agonism in CTL (19). It also seems surprising that cytokine production was maintained in these studies, i.e., that perforin-dependent killing was lost selectively (7, 8). This cytotoxic pathway is very rapid and can be elicited by TCR signaling too weak to induce cytokine production, cell proliferation, or TCR down-modulation (20). It also has been shown that only very few MHC-peptide complexes on target cells, perhaps only one, are sufficient to induce perforin-dependent killing, which clearly is not the case for cytokine responses (20, 21). On the other hand, there exist several reports indicating that induction of Fas-dependent killing requires FasL transcription (1, 2, 7, 10, 22). At least for cytokine production, induction of gene transcription requires sustained TCR signaling for extended periods of time (20).

To address these apparent divergences, we now investigated the activation requirements for selective activation of Fas-dependent cytotoxicity. By screening 12 altered peptide ligands on seven CD8⁺CTL clones, we have previously observed one case in which perforin- but not Fas-dependent cytotoxicity was lost (18). The CTL clone concerned, S15, is H-2K^d restricted and specific for the Plasmodium berghei circumsporozoite peptide PbCS 252–260 (SYIPSAEKI), modified by replacing PbCS S252 with photoreactive iodo-4-azidosalicylic acid (IASA) and by conjugating PbCS K259 with 4-azidobenzoic acid (ABA) (23). The epitope recognized by this and related CTL clones was the ABA-modified PbCS peptide (23). Selective photoactivation of the IASA group allowed cross-linking to K^d and photoactivation of ABA to TCR (23, 24). This system thus permitted assessment of TCR-ligand binding on living CTL by TCR photoaffinity labeling with covalent soluble K^d-peptide complexes (18, 23, 24). Remarkably, S15 CTL replacement of the ABA group with 4-azidosalicylic acid (ASA, i.e., introduction of a hydroxy substituent in position 2 of ABA) resulted in a variant (K259(ASA)), which was only inefficiently recognized, possibly only via Fas-dependent cytotoxicity (18).

In this work, we report that this variant also failed to induce cytokine production by S15 CTL. Surprisingly, the same partial agonism was also observed for the wild-type peptide derivative in the presence of anti-CD8β mAb H35-17 or Fab′ fragments of anti-K^dα3 mAb SF1-1.1.1 (SF1-1.1.1 Fab′). These reagents prevent CD8 from binding MHC molecules that interact with TCR, which weakens TCR-ligand binding and accelerates TCR-ligand complex dissociation (25). As CD8 (and CD4) is associated with the T cell-specific tyrosine kinase p56^lck, which plays a key role in TCR signaling, this blocking of CD8 (and CD4) not only reduces the avidity of TCR-ligand binding but also prevents coreceptor-associated p56^lck from being brought to the TCR/CD3 complex (25, 26, 27). Consistent with this, we observed that blocking of CD8 substantially reduced phosphorylation of ζ-chain as well as ZAP-70 and mobilization of intracellular calcium [Ca²⁺]_i, all important early events in T cell activation (5). Essentially the same findings were obtained for variant K259(ASA), indicating that in this system Fas-dependent cytotoxicity was induced by TCR signaling too weak to elicit perforin-dependent killing or lymphokine production. Since S15 CTL constitutively expressed FasL, it appears that very limited TCR signaling can elicit Fas-dependent cytotoxicity, conceivably by inducing translocation of preformed, intracellular FasL to the cell surface, to become biologically active.

Amino acids and chemicals for peptide and conjugate synthesis were obtained from Sigma (Buchs, Switzerland) and Bachem Finechemicals AG (Bubendorf, Switzerland). Concanamycin A (CMA) was from Sigma. The synthesis and characterization of photoreactive PbCS peptide derivatives have been described previously (18, 23, 24). In brief, all peptide derivatives were synthesized on an ABI 431 peptide synthesizer (Applied Biosystems, Foster City, CA). K(ABA) and K(ASA) were incorporated as Fmoc-K(ABA)-OH and Fmoc-K(ASA)-OH, respectively. The peptide derivatives were deprotected and cleft from the resin and purified by HPLC on a C-18 column (1 × 25 cm, 5 μm particle size; Marcherey & Nagel, Oensingen, Switzerland). After reacting the purified peptide derivatives with ¹²⁵I or nonradioactive IASA-N-hydroxysuccinimide ester, they were HPLC purified. ¹²⁵I iodine was from New England Nuclear (Boston, MA) and had a specific radioactivity of approximately 2000 Ci/mMol. The molecular weight (M_r) of all compounds was verified by mass spectrometry (LDI 7000 mass spectrometer; Linear Scientific, Reno, CA).

The P815 mastocytoma cells and P815 transfected with Fas (P815Fas⁺) (28) were maintained in DMEM supplemented with FCS (5%) and HEPES (10 mM). A20 B lymphoma cells and a Fas⁻ variant, which lacks functional Fas (A20Fas⁻) (29), were cultured in the same medium supplemented with 1 mM 2-ME. Cloned S15 CTL were propagated by weekly restimulation with IASA-YIPSAEK(ABA)I-pulsed, irradiated P815 cells in the presence of irradiated BALB/c splenocytes and IL-2, as previously described (23). Hybridomas producing the following mAb were obtained from American Type Culture Collection (Manassas, VA): H35-17 (anti-CD8β), SF1-1.1.1 (anti-K^dα3), 20-8-4S (anti-K^dα1), and H57-597 (anti-TCR Cβ). Fab′ fragments were prepared following published procedures (30).

Cytolytic activities were assessed by a chromium release assay, as previously described (18, 23). In brief, ⁵¹Cr-labeled target cells (5 × 10³/well) were preincubated in 96-well plates for 15 min with 3- or 10-fold dilutions of peptide derivative in DMEM supplemented with 5% FCS and 10 mM HEPES. Cloned S15 CTL were added at the indicated E:T ratios, and after 4–6 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)]. In some experiments, mAb H35-17 (10 μg/ml) or SF1-1.1.1 Fab′ (20 μg/ml) was added with the peptide.

P815 target cells (5 × 10³/well) were sensitized with peptide derivatives in presence or absence of Abs, as described for the chromium release assay. After incubation with cloned S15 CTL (1 × 10⁴/well) at 37°C for 24 h, the content of IFN-γ in supernatants was determined by ELISA, using anti-IFN-γ mAb R4-6A2 as the first and biotinylated anti-IFN-γ mAb AN18 as the second Ab. Horseradish peroxidase-conjugated streptavidin (Amersham, Arlington Heights, IL) was used to detect the biotinylated Ab, followed by incubation with o-phenylenediamine hydrochloride (Sigma). The color was measured at 490 nm using an ELISA reader (MR7000; Dynatech Laboratories, Chantilly, VA).

Calcium flux was measured as previously described (16). In brief, S15 CTL were labeled with Indo-1-AM (Sigma) (2 mM, 1 × 10⁶ cells/ml) at 37°C for 45 min. After one wash, CTL were mixed at an E:T ratio of 1:3 with P815 cells, previously pulsed with the indicated concentrations of IASA-YIPSAEK(ABA)I or IASA-YIPSAEK(ASA)I. After centrifugation for 1 min at 1500 × g and incubation at 37°C for 1 min, calcium-dependent fluorescence of Indo-1 was assessed by flow cytometry on a FACStar cytofluorometer (Becton Dickinson), by gating on a forward light scattering corresponding to conjugates. The total time of recording was 8 min. For experiments with Abs, calcium flux was first acquired for 2 min, the sample tube was removed, and H35-17 mAb and SF1-1.1.1 Fab′ were added at 20 μg/ml. Then the calcium flux was reacquired for the remaining time.

To assess FasL message by reverse-transcriptase PCR, poly(A) mRNA was extracted from 10⁷ S15 cells previously incubated for 4 h either without or with 1 μM of IASA-YIPSAEK(ABA)I or IASA-YIPSAEK(ASA)I, using the mini message maker kit from Ingenius (R&D Systems, Wiesbaden, Germany), according to the protocol of the manufacturer. cDNA was generated from poly(A) mRNA and the respective downstream primers using the first-strand cDNA synthesis kit from Pharmacia (Uppsala, Sweden). For the subsequent PCR, the following primers were used: for FasL, 5′-TAG CTG ACC TGT TGG ACC TTG C and 5′-CAC TCA AGG TCC ATC CCT CTG; and for actin, 5′-ATC AAG ATC CTG ACC GAG CG and 5′-TAC TTG CGC TCA GGA GGA GC. Thirty-five PCR cycles were performed at 94°C for 1 min, at 55°C for 1 min, and at 72°C for 2 min in a volume of 25 μl. PCR samples (10 μl) were analyzed on a 2% agarose gel. Alternatively, FasL protein expression was assessed by Western blot analysis. S15 CTL, pretreated the same way, were lysed in 2% Triton X-114, and the proteins of the lower phase were recovered by precipitation with CHCl3/MeOH (1:4 v/v). After separation by SDS-PAGE (10%, reducing conditions) and transfer onto nitrocellulose membrane, FasL was detected with PE62 anti-FasL Ab (28) and horseradish peroxidase-conjugated goat anti-rabbit Igs and revealed by chemoluminescence (ECL; Amersham).

Photoaffinity labeling of soluble K^d molecules and TCR with ¹²⁵IASA-YIPSAEK(ABA)I or derivatives was performed as described previously (24). In brief, ¹²⁵IASA-YIPSAEK(ABA)I or ¹²⁵IASA-YIPSAEK(ASA)I (10⁸ cpm/ml) was incubated with purified soluble K^d (50 μg) in the presence of β₂-microglobulin (2, 5 μg) at room temperature for 2 h. After UV irradiation at ≥350 nm for 20 s with 3000 W UVA irradiator (Mutzhaas, Munich, Germany), the covalent K^d-peptide derivative complexes were purified by gel-filtration fast protein liquid chromatography. For TCR photoaffinity labeling, S15 CTL (3 × 10⁶ cells) were incubated with soluble covalent monomeric K^d-peptide derivative complexes (5–10 × 10⁶ cpm) for 2 h at 0°C in presence or absence of H35-17 mAb (10 μg/ml) or SF1-1.1.1 Fab′ (20 μg/ml), followed by UV irradiation at 312 ± 40 nm for 25 s with a 90 W mercury fluorescence lamp (BioBlock Scientific, Illkirch, France). TCR immunoprecipitation with mAb H57-597 and SDS-PAGE analysis (10%, reducing conditions) were performed as described (18, 23, 24). For quantification of covalent TCR-ligand complexes, the dried gels were evaluated by phosphor imaging using a PhosphorImager and Image Quant software (Molecular Dynamics, Sunnyvale, CA). Mean values and SDs were calculated from at least three independent experiments.

Phosphorylation of ζ-chain was assessed as described previously (16). Briefly, P815 cells (1 × 10⁶/ml), preincubated with peptide derivatives (1 μM) in presence or absence of mAb H35-17 (10 μg/ml), or SF1-1.1.1 Fab′ (20 μg/ml) or mAb 20-8-4S (10 μg/ml) for 15 min at 37°C, were added to S15 CTL (2 × 10⁷/ml). After centrifugation for 1 min at 1000 × g and incubation for 8 min at 37°C, cells were lysed in 0.5 ml ice-cold lysis buffer (1% Triton X-100, 50 mM Tris, pH 7.5, 150 mM NaCl, 10 mM EDTA, 1 mM PMSF, 10 μg/ml aprotinin, and 10 mM Na₃VO₄). Detergent-insoluble material was removed, and the supernatants were incubated with anti-ζ mAb H146 absorbed on protein A-Sepharose for 4 h at 4°C. The immunoprecipitates were washed three times and subjected to SDS-PAGE (15%, reducing conditions). After transfer on poly(vinylidene difluoride) membranes (New England Nuclear) and immunoblotting with antiphosphotyrosine mAb 4G10 (Upstate Biotechnology, Lake Placid, NY), the blots were developed using the enhanced chemoluminescence technique (ECL; Amersham). Phosphorylation of ZAP-70 was determined likewise, except that for immunoprecipitation, polyclonal anti-ZAP-70 Abs (Upstate Biotechnology) were used and for SDS-PAGE, 8% gels.

We have observed previously that cloned S15 CTL only inefficiently lysed target cells sensitized with IASA-YIPSAEK(ASA)I (18). Further analysis now showed that this was also true for the derivative that either had an additional iodine substituent (IASA-YIPSAEK(IASA)I) or lacked the N-terminal IASA group (YIPSAEK(ASA)I) (data not shown). To verify whether this inefficient lysis was accounted for by loss of perforin-dependent cytotoxicity, the ⁵¹Cr release assay shown in Fig. 1 was performed. S15 CTL efficiently killed A20 target cells pulsed with IASA-YIPSAEK(ABA)I but inefficiently lysed A20 cells pulsed with IASA-YIPSAEK(ASA)I (Fig. 1,A). The same inefficient lysis of A20 cells was observed for the wild-type peptide derivative in the presence of CMA, a drug that specifically inhibits perforin-dependent cytolysis (31). Together with the finding that CMA barely affected lysis of A20 cells pulsed with K259(ASA), this indicates that the low lysis observed for the variant epitope was not accounted for by perforin-dependent lysis. Consistent with this, we observed that A20Fas⁻ cells, which can only be lysed by perforin-mediated lysis (29), were efficiently lysed when pulsed with IASA-YIPSAEK(ABA)I but not when pulsed with IASA-YIPSAEK(ASA)I (Fig. 1 B). Again this lysis was inhibited by CMA to near background levels. The same findings were obtained when esterase release by S15 CTL was assessed, which is in agreement with the fact that perforin-dependent lysis involves exocytosis of CTL granules, which contain perforin and a mixture of serine/threonine esterases (granzymes) (data not shown and 2 .

Similar findings were obtained when P815 cells were used as targets. Again CMA reduced lysis in the presence of IASA-YIPSAEK(ABA) by more than threefold, to approximately the same level as observed for variant K259(ASA) (Fig. 1,C). Because P815 cells were more efficiently lysed than A20 cells, a shorter incubation period (4 instead of 6 h) and a lower E:T ratio (1:4 instead of 1:20) were used in this experiment. Since P815 cells were efficiently killed via the perforin but not the Fas pathway, we repeated this experiment using P815Fas⁺ cells that express high levels of Fas (28). These cells were efficiently killed in the presence of either the wild-type or the variant peptide derivative (Fig. 1 D). Moreover, CMA barely affected either lysis, indicating that P815Fas⁺, unlike normal P815, were killed efficiently via the Fas pathway. These results demonstrate that S15 CTL lysed IASA-YIPSAEK(ASA)I- pulsed A20 or P815 cells only via Fas-dependent cytotoxicity. This is consistent with reports showing that certain altered peptide ligands can activate Fas- but not perforin-dependent cytotoxicity (7, 17).

In the presence of anti-CD8β mAb H35-17 S15, CTL inefficiently recognized IASA-YIPSAEK(ABA)I (18). To find out whether this was accounted for by selective Fas-dependent killing, we performed a chromium release assay using A20 and A20Fas⁻ cells as targets (Fig. 2). In the presence of mAb H35-17, S15 CTL lysed A20 target cells sensitized with IASA-YIPSAEK(ABA)I as inefficiently as they lysed A20 cells sensitized with IASA-YIPSAEK(ASA)I (Fig. 2, A and B). Since variant K259(ASA) was recognized by S15 CTL only in a Fas-dependent manner (Fig. 1), this suggested that H35-17 mAb abolished perforin-dependent cytotoxicity by S15 CTL. Consistent with this, we observed that mAb H35-17 abolished perforin-dependent lysis of IASA-YIPSAEK(ABA)I-sensitized A20Fas⁻ targets (Fig. 2,C). Conversely, this Ab had no significant effect on Fas-dependent killing of A20 cells sensitized with K259(ASA) (Fig. 1 B). The same results were obtained when H35-17 Fab′ were used (data not shown).

This anti-CD8 mAb blocks participation of CD8 in TCR-ligand binding and CD8-dependent adhesion (25). To find out which of these two CD8 functions was required for perforin-dependent killing, we tested likewise Fab′ fragments of anti-K^dα3 mAb SF1-1.1.1 (SF1-1.1.1 Fab′), which block only participation of CD8 in TCR-ligand binding but not CD8-mediated adhesion (25). This reagent inhibited partially lysis of A20 cells sensitized with IASA-YIPSAEK(ABA) but abrogated lysis of A20Fas⁻ cells (Fig. 2, A and C). The reason for this differential effect is not clear but may be related to the more efficient killing of A20 as compared with A20Fas⁻ cells (half-maximal lysis of A20 cells required about 100-fold lower concentrations of IASA-YIPSAEK(ABA)I) (Fig. 2, A and C). As mAb H35-17, SF1-1.1.1 Fab′ had no effect on the Fas-dependent killing of A20 cells sensitized with IASA-YIPSAEK(ASA)I (Fig. 2 A–C). Together these results indicate that anti-CD8β mAb H35-17 and less efficiently SF1-1.1.1 Fab′ inhibited perforin-dependent cytotoxicity of S15 CTL but had no effect on Fas-dependent mediated killing, i.e., had the same effect as the epitope modification under study. The finding that SF1-1.1.1 Fab′ had a similar effect as mAb H35-17 suggests that for perforin- but not Fas-dependent cytotoxicity, participation of CD8 in TCR-ligand binding was important.

We next examined what effect epitope modification and blocking of CD8 have on IFN-γ production by S15 CTL. As shown in Fig. 3,A, S15 CTL incubated with A20 in the presence of graded concentrations of IASA-YIPSAEK(ABA)I produced IFN-γ in a dose-dependent manner. In contrast, no IFN-γ production was detectable when variant K259(ASA) was used (Fig. 3,B). The IFN-γ response elicited by IASA-YIPSAEK(ABA)I was abolished by mAb H35-17 and substantially diminished by SF1-1.1.1 Fab′ (Fig. 3 A). The same findings were obtained when P815 or P815Fas⁺ cells were used as targets (data not shown). Thus, in contrast to previous studies, in which either altered peptide ligands or mutant CTL exhibited Fas-dependent killing and IFN-γ response, but no perforin-dependent cytotoxicity (7, 8), only Fas-dependent killing was preserved.

A unique feature of our system is that TCR-ligand binding can be assessed by TCR photoaffinity labeling (18, 23, 24). To this end, soluble monovalent K^d was photo-cross-linked with IASA-YIPSAEK(ABA)I and IASA-YIPSAEK(ASA)I, respectively, and incubated with S15 CTL in the absence or presence of mAb H35-17 SF1-1.1.1 Fab′. As shown in Fig. 4, S15 TCR photoaffinity labeling by the variant ligand was approximately ninefold less efficient as compared with the wild-type ligand. In the presence of mAb H35-17 or SF1-1.1.1 Fab′, S15 TCR photoaffinity labeling by either ligand was reduced to background levels, observed in the presence of anti-K^dα1 mAb 20-8-4S, which blocks specific TCR-ligand binding (Fig. 4 and 24 . Thus, either the epitope modification or blocking of CD8 reduced the avidity of TCR-ligand binding by ≥8-fold.

We next examined what effect the epitope modification and blocking of CD8 had on phosphorylation of ζ-chain and ZAP-70. To this end, S15 CTL were incubated in the absence or presence of mAb H35-17 or SF1-1.1.1 Fab′ with P815 cells, previously pulsed with IASA-YIPSAEK(ABA)I and IASA-YIPSAEK(ASA)I, respectively. P815 rather than A20 cells were used as targets, because they express higher levels of K^d, and hence induce more extensive phosphorylations. As shown in Fig. 5,A, the wild-type epitope elicited considerably more extensive ζ-chain phosphorylation than variant K259(ASA) (lanes 1 and 6), and the ratio between the pp23 and pp21 phospho-ζ was reduced by about 40% (Fig. 5,B). The pp23 phospho-ζ is typically elicited by peptide agonists, and the pp21 ζ-phospho by partial agonists or antagonists (26, 32). While ζ-chain phosphorylation was barely detectable in the presence of anti-K^dα1 mAb 20-8-4S (lane 5), weak ζ-chain phosphorylation, mainly of the pp21 form, was observed in the presence of mAb H35-17 or SF1-1.1.1 Fab′ (lanes 3 and 4). Remarkably, this ζ-chain phosphorylation was little increased as compared with the one observed in the absence of antigenic peptide (lane 1). Similarly, significant phosphorylation of ZAP-70 on S15 CTL was only observed following incubation with IASA-YIPSAEK(ABA)I-pulsed P815 cells but not for the K(ASA) variant or upon blocking of CD8 (Fig. 5,C, lanes 3–5). The phosphorylated ZAP-70 observed in case of the wild-type peptide was associated with TCR/CD3, as seen by the coprecipitation of phosphorylated CD3ε and ζ-chains (Fig. 5 C, lane 3); thus, in this case, ZAP-70 was significantly recruited to TCR/CD3 and activated by phosphorylation.

Together these results indicate that the epitope modification or blocking of CD8 significantly decreased phosphorylation of ζ-chain, mainly of the pp23 phospho-form and recruitment, as well as activation of ZAP-70.

T cell activation typically provokes a rapid rise in [Ca²⁺]_i, which can be detected by calcium-dependent fluorescence of Indo-1 and flow cytometry (16, 20). As shown in Fig. 6,B, incubation of S15 CTL with P815 cells, previously pulsed with 1 nM IASA-YIPSAEK(ABA)I, caused a rapid increase of [Ca²⁺]_i. This calcium level was stable during the assayed 5 min, and was the same whether APC were pulsed with 1 or 100 nM IASA-YIPSAEK(ABA)I (data not shown). In contrast, S15 CTL incubated with P815 cells pulsed with 100 nM IASA-YIPSAEK(ASA)I exhibited [Ca²⁺]_i levels just slightly above background levels (Fig. 6,C). A similar marginal increase in [Ca²⁺]_i was observed when S15 CTL were incubated with IASA-YIPSAEK(ABA)I-pulsed P815 cells in the presence of mAb H35-17, 20-8-4S, or SF1-1.1.1 Fab′ (Fig. 6, D–F). These results indicate that the epitope modification or blocking of CD8 resulted in a very small, yet significant, increase in [Ca²⁺]_i in S15 CTL.

Overnight incubation of CTL with herbimycin A results in depletion of p56^lck in a dose-dependent manner (33). As shown in Fig. 7,A, gradual depletion of p56^lck severely impaired perforin-dependent killing of P815 cells sensitized with IASA-YIPSAEK(ABA)I by S15 CTL. Fifty percent inhibition was observed when using 1 μM of herbimycin A. In contrast, S15 CTL preincubated with 3 μM of herbimycin still efficiently killed P815Fas⁺ cells pulsed with IASA-YIPSAEK(ASA)I (Fig. 7,B). This Fas-dependent lysis was reduced by less than 60% when S15 CTL were preincubated with 10 μM of herbimycin A, a treatment that reduced the mainly perforin-dependent lysis of P815 cells by 85%. These findings suggest that the perforin/granzyme-mediated cytotoxicity in this system is more p56^lck-dependent than the Fas-dependent one. Since p56^lck in part is CD8 associated, this is consistent with the finding that blocking of CD8 impaired perforin- but not Fas-dependent cytotoxicity (Fig. 2). It must be noted, however, that herbimycin as well as other tyrosine kinase inhibitors have limited specificity.

To find out whether the Fas-dependent killing observed in this study requires induction of FasL, we assessed FasL mRNA by reverse-transcriptase PCR. As shown in Fig. 8,A, nonactivated S15 CTL constitutively expressed substantial levels of FasL message. Upon incubation of S15 CTL with P815 cells pulsed with IASA-YIPSAEK(ABA)I, FasL message slightly increased, and even less when IASA-YIPSAEK(ASA)I was used. The messages for actin or Fas were similar for S15 CTL that were either untreated or incubated with P815 cells in the presence of either peptide. Essentially, the same findings were obtained when FasL protein was assessed by Western blotting (Fig. 8 B, lanes 2, 4, and 5). This analysis also showed that S15 CTL express significantly more FasL as compared with P815 cells, and that this expression was not significantly increased upon mixing the cells in either the absence or presence of either peptide derivative. Similar findings were obtained when the incubation period was increased from 4 to 12 h, or when S15 CTL were used 1 or 2 wk after restimulation (data not shown). These findings suggest that S15 CTL have a pool of intracellular FasL, which upon activation, is either translocated to the cell surface or released in soluble form.

To find out whether the Fas-dependent killing observed in this system was mediated by soluble or cell-associated FasL, we performed a bystander cytolytic assay, in which IASA-YIPSAEK(ASA)I was photo-cross-linked with K^d molecules on P815Fas⁺ cells. In the experiment shown in Fig. 9,A, ⁵¹Cr-labeled sensitized targets were mixed with equal numbers of untreated P815Fas⁺ cells and incubated with S15 CTL for 4 h. In contrast to the high specific lysis observed in this experiment, no lysis was detectable when the peptide was on the cold targets (Fig. 9 B). The lack of bystander killing indicates that the Fas-dependent killing of P815Fas⁺ cells required CTL/target cell contact, i.e., did not involve soluble FasL, nor other soluble cytotoxic molecules, such as TNF-α.

We investigated a form of partial agonism for mature CD8⁺CTL, in which an altered peptide ligand selectively induces Fas-dependent cytotoxicity. While CTL usually primarily kill via perforin/granzyme-mediated cytolysis and concomitantly produce IFN-γ, TNF-α, and various ILs, we described in this work a case in which the only detectable CTL function was Fas-dependent cytotoxicity. Selective activation of Fas-dependent killing has been observed previously, either for altered peptide ligands on CD8⁺CTL and CD4⁺Th clones or for mutant CTL clones (7, 8, 17, 18, 34). However, in contrast to these studies, we observed no cytokine production, i.e., no IFN-γ response and, as deduced from the lack of bystander killing, also no TNF-α production (Figs. 3 and 9).

The lack of cytokine production and perforin-dependent killing observed in our study is consistent with the barely detectable increase in [Ca²⁺]_i and the minimal ζ-chain phosphorylation, mainly of the pp21 form, as well as the lack of significant phosphorylation of ZAP-70 (Figs. 5 and 6). While activation of perforin-dependent cytotoxicity is very rapid and induced by brief TCR engagement, induction of cytokine production requires sustained TCR signaling; yet activation of both CTL functions relies on similar, often overlapping, signaling pathways, which involve mobilization of [Ca²⁺]_i, significant ζ-chain phosphorylation, namely of the pp23 form, and recruitment and phosphorylation/activation of ZAP-70 (3, 4, 5, 32). Consistent with this, we observed that on closely related CTL clones, peptide modifications generally affected perforin-mediated cytotoxicity and IFN-γ production in good correlation, although occasional divergences can occur (15, 16).

A main finding of the present study was that upon blocking of CD8, the wild-type peptide exhibited the same selective Fas-dependent cytotoxicity as observed for variant K259(ASA) (Fig. 2). Similar results were obtained when anti-CD8β mAb H35-17, its Fab′ fragments, or Fab′ fragments of anti-K^dα3 mAb SF1-1.1.1 were used (Figs. 2 and 3, and unpublished results), which have in common that they block coordinate binding of CD8 to K^d molecules that interact with TCR (25). This blocking of CD8 substantially decreased the avidity of TCR-ligand binding (Fig. 4). A similar, although less dramatic reduction of TCR-ligand binding was observed for the epitope modification (Fig. 4). Based on the observation that blocking of CD8 or weakening of TCR-ligand binding by peptide modification typically accelerates TCR-ligand complex dissociation, one would expect this also to be the case in these situations (16, 25, 26, 27). Since the coreceptor is associated with p56^lck, its coordinate binding to TCR-associated MHC brings this src kinase to the CD3/ζ complex, which is hampered by acceleration of TCR-ligand complex dissociation or blocking of the coreceptor (16, 25, 26). As this tyrosine kinase plays an essential role in phosphorylation of CD3/ζ and ZAP-70, this may explain the faint ζ-chain phosphorylation observed upon blocking of CD8 or for the low affinity variant K259(ASA) and the lack of significant recruitment and phosphorylation of ZAP-70 (Fig. 5). Together with the observation that depletion of p56^lck by herbimycin A little affected Fas-dependent cytotoxicity (Fig. 7), these findings suggest that p56^lck, especially CD8-associated one, is not important for activating S15 CTL for this cytotoxicity. This is in accordance with reports showing that p56^lck and ZAP-70 are not required for induction of Fas-dependent apoptosis and that minimal ζ-chain phosphorylation can be sufficient for its activation (14, 34).

Different immunoreceptor tyrosine-based activation motifs of CD3/ζ are clearly not identical in terms of initiating downstream signaling cascades (35, 36). Importantly, it has been shown that induction of Fas-dependent killing requires phosphorylation of the first but not the second or third immunoreceptor tyrosine-based activation motif of ζ-chain (37). This is consistent with the finding that selective activation of S15 CTL for Fas-dependent killing involved limited ζ-chain phosphorylation, mainly of the pp21 form, and that this phosphorylation was different from the one elicited by peptide antagonists, because variant K259(ASA) is not an antagonist for S15 CTL (Fig. 5 and 18 . Conceivably, this form of ζ-chain phosphorylation is mediated either by p56^lck, which is not associated with CD8, or by another tyrosine kinase, possibly p59^fyn.

Moreover, the finding that blocking of CD8 or the peptide variant resulted in barely detectable increases in [Ca²⁺]_i suggests also that phosphorylation of PLC-[γ was impaired (Fig. 6). This lipase, upon activation by phosphorylation, releases from membrane lipids inositol triphosphate and diacylglycerol (5). While these secondary messengers have various effects, inositol triphosphate stimulates mobilization of calcium from intracellular stores (5). This is consistent with the observation that activation of Fas-dependent killing by CTL requires limited influx of extracellular but not mobilization of intracellular Ca²⁺ (38). The indifference of the selective Fas-dependent cytotoxicity observed in this study to depletion of p56^lck correlates with the lack of ZAP-70 phosphorylation, which in turn correlates with the very marginal increase in [Ca²⁺]_i, i.e., activation of PLC-γ, as both kinases are needed for its activation ( Figs. 5–7 and Refs. 3–5).

It is interesting to note that in several systems there is a clear divergence between the activation requirements for induction of FasL expression and Fas-dependent cytotoxicity. While several studies have clearly demonstrated that p56^lck and ZAP-70 are required for FasL induction (13, 39, 40, 42), other studies show that induction of Fas-dependent apoptosis does not, or minimally, require p56^lck (14, 41). In agreement with our findings, Chung et al. observed that induction of TCR-induced cell death, in contrast to lymphokine production, is relatively lck independent and in poor correlation with FasL expression (42).

It is important to realize that Fas-dependent killing can be mediated either by soluble or cell-associated FasL, and that for some cell types, Fas-dependent killing requires FasL transcription, whereas other cells, including macrophages, CTL, or melanocytes, have significant amounts of intracellular FasL, which upon activation, can be translocated to the plasma membrane (29, 43, 44). This latter pathway seems to be realized in our system. On one hand, cloned S15 CTL clearly killed via cell-associated and not soluble FasL (Fig. 9). On the other hand, S15 CTL expressed constitutively a high level of FasL message and protein that was little increased upon TCR triggering, especially in the case of variant K259(ASA) (Fig. 8). The observation that S15 CTL, which for propagation need to be stimulated periodically by Ag, exhibited significant ζ-chain phosphorylation, even after 2 wk of stimulation, suggests that these cells are sufficiently activated to express significant levels of FasL (Figs. 5 and 8).

Even though S15 CTL are very CD8 dependent, i.e., their perforin-dependent cytotoxicity and IFN-γ production were effectively blocked by anti-CD8 mAb, their ability to induce Fas-dependent killing was not (Figs. 2 and 3, and 18 . Our study shows that induction of Fas-dependent killing by CTL can occur even when CD8 is blocked or by low affinity altered peptide ligands. The TCR signaling required for this activation is extremely weak, and is unable to induce significant IFN-γ or FasL production, and not even perforin-dependent cytotoxicity or esterase release. This mechanism may allow eradication of cells in the absence of apparently any other CTL function and may play a role in eliminating cells expressing altered self proteins, e.g., proteins that were modified in vivo by reactive chemicals but also of cells expressing MHC class I molecules with defect CD8 binding.

We are indebted to Clotilde Horvath for expert technical help, and Anna Zoppi for preparing the manuscript.