Upon activation, cytotoxic T-lymphocytes (CTL) differentiate into various subsets to execute their cytotoxic function. In this study, we used a widely accepted protocol for activation of T cells by staphylococcal enterotoxin A (SEA). PMBC were isolated from leukoreduction system (LRS) chambers and pulsed with SEA. CTL were positively isolated after the expansion of the mixed PBMC population for 5 days ( Figure 1A ).

CD8⁺ naïve and memory T cell subsets were identified by the surface molecules CD62L or CCR7 and CD45 isoforms RA and RO according to (14). Figure 1B illustrates that upon SEA stimulation, the CD8⁺ population was shifted towards a T_CM phenotype. The frequency of the T_CM subset increased from 19.8 ± 9.5% on day 0 to be the predominant subset (64 ± 5%) on day 5 of SEA-stimulation. Frequencies of T_EM and T_N were decreased by roughly 50% (T_EM 53.9%, T_N 46.7%) to 15.1% and 18.3%, while most of the T_EMRA cells disappeared during the expansion period ( Figure 1C ). Healthy human donors show a huge variation in subtype distribution but T_CM was always the prominent subtype after SEA stimulation whereas T_EM and T_N decreased, and T_EMRA was almost absent ( Figure 1D ). To test the SEA reactivity of stimulated CTL (SEA-CTL), we analyzed the degranulation after co-incubation with SEA-pulsed target cells by detection of surface-located CD107a ( Figures 1E, F ). On average, 47.3 ± 16.5% of SEA-CTL degranulated in case target cells were loaded with SEA compared to 4.4 ± 3.3% in case target cells were not loaded with SEA ( Figure 1F ).

To assess the cytotoxic potential of each individual subset, we determined the perforin expression in T_N, T_CM, T_EM and T_EMRA ( Figure 2A ). The antibody clone δG9v identifies granule-associated perforin ready to lyse the target cell (18). The content of perforin was higher in T_EM compared to T_CM. Although the frequencies of perforin expressing cells are comparable among T_EM (70.17 ± 12%) and T_CM (65.06 ± 8.9%) ( Figure 2A ), the analysis of the perforin MFI-ratios reveals that T_EM express 2.8 times more granular perforin than T_CM ( Figure 2B ). T_EMRA express a significantly increased amount of perforin ( Figure 2B ) but the subpopulation is almost absent after SEA-stimulation (2.14 ± 2.4% of all cells). The death receptors FasL and TRAIL are also expressed in T_CM and T_EM. However, only FasL ( Figures 2C, D ) but not TRAIL ( Figure 2E ) were moderately upregulated after SEA-loaded target cell contact. In conclusion, after expansion by SEA, T_EM and T_CM are the main cytotoxic-competent subpopulations.

To determine the cytotoxic potential of individual memory subsets we sorted bead-isolated SEA-stimulated CD8⁺ population (SEA-CTL) into CD62L⁺/CD45RO⁺ T_CM and CD62L^-/CD45RO⁺ T_EM subpopulations ( Figure 3A , pre- and post-sort). We also sorted CD62L⁺/CD45RO^- T_N from SEA-CTL of 4 different donors. We used these T_N as an internal negative control, although the cells cannot be considered as truly naïve since they were in the environment of SEA stimulation for 4 days. However, in terms of their surface markers, they are considered T_N. 24h after subpopulation sorting, we re-analyzed the sorted T_EM or T_CM cells and found a purity of >80% ( Figure 3B ). At the same time, we analyzed the killing capacity of the different subpopulations in a time-resolved population killing assay ( Figure 3C ). This assay allows quantification of the cytotoxic potential at different time points and quantification of the maximum lysis rate which describes the maximal difference in target cell lysis between two measurement points (10 min). As expected, T_N did not kill target cells. The T_EM subpopulation displayed the highest cytotoxic efficiency ( Figure 3C ) as also quantified by the maximal lysis rate ( Figure 3D ) and the target cell lysis at 120 min ( Figure 3E ) compared to T_CM, T_N and the entire SEA-CTL population as control. Interestingly, the change of target lysis was not different during the last 2 hours (between 120 min and 240 min) among T_EM, T_CM and SEA-CTL ( Figure 3F , also compare Figure 3C ). Thus, the main difference between T_EM and T_CM regarding cytotoxic efficiency is attributed to the initial phase of the cytotoxicity assay (within the first 2 hours).

Since perforin-mediated exocytosis is known to be rapidly executed (19–22), and since we and others have shown that perforin determines the initial killing phase in CTL and NK cells (22, 23), we explored the perforin content in sorted subsets in parallel to the population killing assay to confirm the increased expression of granule-associated perforin (using antibody clone ΔG9), which we observed for T_EM before in in non-sorted subsets ( Figures 2A, B ). Again, we found a significant difference in perforin content of T_EM compared to T_CM cells ( Figures 3G, H ). The SEA-CTL (CD8⁺) control-population expressed slightly more perforin than T_CM but less than T_EM, which is in good agreement with the subtype composition (T_CM>T_EM>T_N). T_N sorted cells have almost no perforin ( Figure 3H ) in good accordance with the negligible killing efficiency shown before ( Figure 3C ). The mean fluorescence intensity of perforin correlated almost perfectly with the maximal lysis rate of the subtypes and the entire SEA-CTL (CD8⁺) population ( Figure 3I , r² = 0.98). Therefore, it is likely that the differences of initial lysis of target cells during the first 120 min are caused by differences in perforin content.

When analyzing various effector molecules important for killing mechanisms on mRNA level, we also found interferon-γ, granzyme B and FasL but not TNFα or TRAIL significantly enhanced in T_EM ( Figure 3J ). The same was true for the surface expression of TRAIL analyzed by flow cytometry ( Figures 3K, L ), suggesting that additional, possibly slower killing mechanisms could contribute to the enhanced T_EM mediated target cell lysis.

So far, our results revealed that the differences in subset-specific cytotoxicity measured with the population killing assay were likely caused by perforin-mediated target cell lysis ( Figure 3C ), probably also involving granzyme B, which is expressed at a higher level in T_EM compared to T_CM Figure 3J ). To detect potential other mechanisms involved in the differential cytotoxic efficiency of T_EM compared to T_CM like target cell contact times and/or functional immune synapse formation, single cell resolution is required. Additionally, population-mediated assays usually do not allow to discriminate between apoptotic or necrotic target cell death. Numerous tools to dissect the different mechanisms during target cell lysis have been developed in the past years (20–22, 24) but the diversity between different subtypes was not in the focus of this research. The recently introduced microscope-based assay (24) allows the detailed analysis and identification of apoptotic and necrotic lysis events on a single cell level by using the FRET-based apoptosis sensor Casper-GR (25). This assay was up to now only applied to natural killer (NK) cells (24). To establish and validate this assay for human CTL, we first tried to stably transfect the clonal B cell lymphoma target cell line Raji (used for the population killing assay, Figure 3C ) with pCasper-GR. However, expression levels were highly variable compromising a detailed quantification. Therefore, we turned to the target cell line NALM6, a B cell line established from a patient with acute lymphoblastic leukemia (ALL) which can also be loaded with SEA.

To validate NALM6 and Raji cells as targets against each other, we compared CTL cytotoxicity in the population killing assay. Neither the kinetics of killing nor the maximal lysis rate were different ( Supplementary Figure 1 ) indicating that human CTL eliminate both targets equally efficient. We also compared FasR and TRAIL receptor expression. Both, NALM6 and Raji cells, express FasR and TRAIL receptors R1-R3 but only very little (Raji) or no (NALM6) TRAIL-R4. While the overall expression of NALM6 target cells is slightly lower compared to Raji cells, the number of positive cells is comparable ( Supplementary Figure 1 ).

Thus, we were confident to use stably-transfected Casper-GR NALM6 cells (from now on named NALM6 pCasper) in the following single cell experiments as target cells. 98.7% of the NALM6 pCasper population was homogenously and strongly positive for GFP and RFP signals ( Figures 4A, B ). To monitor the viability of NALM6 pCasper under experimental conditions (37°C, 5% CO₂) we recorded the GFP and FRET signals at 5 min intervals which remained stable over the complete 8 hours measurement period ( Figure 4C ). To test the functionality of the apoptosis sensor, staurosporine was applied to the targets (without using SEA-CTL effector cells), since staurosporine induces caspase-3 mediated apoptosis (26). In addition, we tested the induction of death receptor-mediated apoptosis by the anti-Fas antibody Apo1-1 mimicking FasL and by recombinant TRAIL to induce TRAIL-receptor mediated cell death over 10 hours. The effects of staurosporine, Apo1-1 and recombinant TRAIL are depicted in Figure 4D compared to untreated control. The quantitative analysis of target cell death is illustrated in a stacked diagram over time called death plots. These illustrate the kinetic distribution of lysed cells (dark grey), apoptotic (green) and living cells (light grey). After 10 hours, in comparison to the untreated living control, staurosporine and recombinant TRAIL significantly enhanced apoptotic cell deaths (95.7% ± 1.5% and 71.8% ± 12.3%, respectively) whereas Apo1-1 was relatively inefficient in inducing apoptotic cell death.

In contrast to staurosporine and TRAIL, the addition of Apo1-1 did not induce a significant increase in the proportional sum of apoptotic and lysed cells even after 10 hours compared with the control condition ( Figure 4E , left panel). Interestingly, when we counted the number of Apo1-1 treated cells, the calculated growth factor was significantly decreased compared to untreated control cells meaning the anti-Fas antibody Apo1-1 reduced the proliferation of NALM6 pCasper ( Figure 4E , right panel).

Next, we analyzed the modes of cell death occurring in NALM6 pCasper after contact with SEA-CTL. Target cell apoptosis and necrosis following CTL contact could be distinguished as previously established for NK cells (24). We confirmed that the assay allows to distinguish primary necrosis ( Figure 4F ), apoptosis ( Figure 4G ) and secondary necrosis ( Figure 4H ) of target cell induced after SEA-CTL contact. Primary necrosis of NALM6 pCasper target cells induced by SEA-CTL is characterized by the parallel loss of the GFP- and FRET fluorescence signals accompanied by morphological changes ( Figure 4F ). The induction of caspase-mediated apoptosis following SEA-CTL contact was identified by an increase of the GFP-fluorescence signal and a decrease of the FRET-donor ratio (DR) ( Figure 4G ). This is expected since GFP- and RFP-molecules of the pCasper FRET-construct are linked by the caspase recognition sequence DEVD which is cleaved by different caspases after induction of apoptosis (caspase-3, -7 and others). Secondary necrosis following initial apoptosis is characterized by a sudden drop of GFP- and FRET-fluorescence signals accompanied by an increase of the cell volume typically for necrotic cell death ( Figure 4H ). In summary, NALM6 pCasper are suitable targets to study different modes of CTL-mediated target cell death.

The population real-time killing assay revealed clear differences between T_EM and T_CM cytotoxic efficiencies ( Figure 3C ). However, it does not allow any conclusions on death modes. Therefore, we used the single cell killing assay to compare T_EM and T_CM to obtain insights into distinct target cell death modes.

Figure 5A shows representative examples of individual NALM6 pCasper target cells which are killed after contact with sorted T_EM or T_CM. For these experiments, only few T_EM or T_CM were used to favor single cell contacts. While the targets can be easily identified by their fluorescence, T_EM or T_CM are difficult to see because they are not stained. We therefore depicted an example in the insets to illustrate the contact between T_EM or T_CM with the targets. Initially all target cells are vital (orange color). The kinetic analysis revealed that T_EM and T_CM initially mainly induce apoptosis in NALM6 targets as indicated by the green color. We quantified the induction of target cell apoptosis in death plots indicated by the green areas in Figure 5B . T_EM were more efficient to induce apoptosis than T_CM, which is evident from the initial part of the curves ( Figure 5B ) and also at the time point of 2 hours ( Figure 5C ). Over longer times, both T_EM and T_CM eliminated almost all targets by apoptosis ( Figure 5B ) as quantified at 4 hours ( Figure 5D ) and 8 hours ( Figure 5E ). The faster apoptosis induction of T_EM compared to T_CM is also evident from the median time duration of 70 min until a target cell in the T_EM data set was apoptotic compared to 180 min in the T_CM data set ( Figure 5F ).

The death plots ( Figure 5B ) also indicate that after initial apoptosis, secondary necrosis (light grey) was induced in many of the targets. Again, T_EM were faster to induce secondary necrosis than T_CM ( Figure 5B ) which is quantified for the different time points of 2, 4, and 8 hours ( Figures 5C–E ). Finally, only T_EM but not T_CM, could induce primary necrosis (without preceding apoptosis) as indicated by the dark grey color ( Figures 5B–E ).Next to the perforin (and granzyme B) expression (see Figures 3G–J ), immune synapse formation or strength could also explain the different cytotoxic efficiency of T_EM compared to T_CM. Ca²⁺ influx has been used to define the initiation of degranulation and delivery of the lethal hit (i.e. 27, 28). Prior to using this approach, we first compared the amplitude of store-operated Ca²⁺ entry in T_CM and T_EM. Following the typical “re-addition” protocol after thapsigargin (TG) application, we found that TCM show higher Ca²⁺ signals compared to TEM as evident from averaged Ca²⁺ traces but also from the quantitative analysis of Ca²⁺ plateau ( Supplementary Figures 2A, B ). However, when we compared Ca²⁺ signals in T_EM and T_CM following contact with Raji cells after contact, we found that they were not much different ( Supplementary Figure 2C ). Both, T_EM and T_CM responded similarly with Ca²⁺ entry once Raji cell contact was initiated. However, T_CM showed slightly lower Ca²⁺ signals than T_EM, which is surprising considering the higher Ca²⁺ signals following store depletion by TG. This might be a hint that immune synapse formation in T_CM may be slightly less efficient than in T_EM. Thus, immune synapse formation might in principle contribute to decreased cytotoxicity of T_CM compared to T_EM, next to perforin (and granzyme B) expression. However, considering the large differences in cytotoxic efficiencies between T_EM and T_CM, the slight difference in immune synapse formation between T_EM and T_CM appears unlikely to the main mechanism to mediate the different efficiencies.

Differences of cytotoxic efficiencies between T_EM and T_CM against target cells may be caused by different mechanisms including, among others, CTL migration speed or persistence, search strategies, immune synapse formation, or different cytotoxic efficiencies during killer cell-target cell contacts. Considering that perforin, granzyme B and FasL (but not TRAIL) are expressed at higher levels in T_EM than T_CM, we wanted to use the single cell cytotoxicity assay to quantify their contributions to T_EM and T_CM cytotoxicity.

To test our single cell assay and compare it to other studies, we first quantified the role of perforin, granzyme B and FasL in SEA-CTL. In comparison to untreated NALM6 pCasper target cells, blocking either perforin alone (by CMA), or perforin and granzyme B together (CMA and z-AAD-CMK in combination), or perforin and FasL together (CMA and anti-FasL antibody in combination), decreased and delayed target cell apoptosis and secondary lysis significantly ( Supplementary Figure 3A ). Quantitative analysis illustrates a very strong inhibition of necrosis and apoptosis by CMA after 2 hours ( Supplementary Figure 3B ), which is not significantly influenced by the additional use of z-AAD-CMK or anti-FasL antibody in SEA-stimulated CTL. This confirms many other studies that the perforin/granzyme B pathway is of major importance for CTL cytotoxicity (20, 21, 23, 27–30). FasL is involved but does not play a major role and granzyme B has no further impact besides its well-described role in combination with perforin (28).

This analysis was now applied to T_EM and T_CM. Compared to control conditions ( Figure 6A , same plots as in Figure 5B ), anti-FasL blocking antibody had no effect on the induction of apoptosis and subsequent secondary necrosis by T_EM or T_CM mediated cell lysis ( Figure 6B ) as also evident from the quantification of apoptosis and necrosis at 2, 4, and 8 hours for T_EM ( Figure 6E ) and T_CM ( Figure 6F ). One could argue that the antibody does not work, but this seems unlikely, as we have shown previously that it works in NK cells (24) and it also works in CTL if lytic granule exocytosis is blocked (see below).

The release of perforin from lytic granules can be blocked very well by low nanomolar concentrations of concanamycin A (CMA, (24). CMA clearly reduced target cell apoptosis and it almost eliminated primary and secondary necrosis in both T_EM ( Figure 6C , quantification in Figure 6E ) and T_CM ( Figure 6C , quantification in Figure 6F ). We conclude that perforin is required for the initial fast phase of apoptotic killing by T_EM and also for the slightly slower initial phase of apoptotic killing of T_CM. The absence of perforin also eliminated necrosis. Without perforin the big difference in cytotoxic efficiency between T_EM and T_CM is diminished. Thus, the higher perforin content of T_EM compared to T_CM is involved in their increased cytotoxic efficiency.

The inhibition of perforin release by CMA in combination with anti-FasL blocking antibody significantly reduced the percentage of apoptotic target cells in the late phase of the measurement for both, T_EM and T_CM (slightly stronger in T_CM) ( Figure 6D , quantification in Figures 6E, F ). Thus, if the perforin/granzyme pathway is blocked, an increased role for FasL is unmasked.

To quantify the exact time points of apoptosis induction in the target cells, every individual cell was tracked over time. The respective starting point for quantification was the first contact of a killer cell to its target cell. The exact time point of apoptosis could be quantified by loss of FRET-signals and an increase of GFP intensities (causing a switch from orange to green fluorescence). Quantification of 89 (T_EM) or 122 (T_CM) cells treated with CMA drastically delayed the induction of apoptosis and as expected, even more dramatically in T_EM (by median increased by 220 min) compared to T_CM (median increased by 185 min) ( Figure 6G ). Inhibition of the Fas/FasL pathway with anti-FasL blocking antibody again indicated no significant difference compared to control conditions in T_EM and T_CM subtypes. However, the combined treatment with CMA and anti-FasL blocking antibody delayed the induction of apoptosis further but to a very similar value for both subtypes, 100 min for T_EM and 115 min for T_CM, respectively. Exploring the area under the curve (AUC) of the apoptotic and secondary necrotic events occurring over the whole experiment, allows the visualization and quantification of FasL-FasR interaction induced cytotoxicity ( Figure 6H ). The additional inhibition of FasL, decreases the AUC of combined apoptotic and secondary necrotic cells for T_EM by 35.3% and for T_CM by 59.2% compared to CMA alone.

Taken together these data indicate a minor role of Fas-mediated target cell lysis for both, T_EM and T_CM, in case perforin/granzyme-mediated cytotoxicity is functional. After blocking perforin function and subsequent induction of apoptosis by granzyme B, apoptotic target cell death induced through FasL-FasR interaction is increasingly important for T_EM and T_CM cytotoxicity against target cells. Perforin-mediated cytotoxicity against target cells appears to be prominent in both subsets, but the compensatory role of FasL-mediated cytotoxicity against target cells appears to have a higher impact in T_CM compared to T_EM.

Taken together this means that complete inhibition of perforin release by CMA and (at least) partial block of FasL-induced target cell death made both T_EM and T_CM less efficient serial killers. Nevertheless, T_EM were still slightly more efficient than T_CM.

It appears therefore likely that there are additional factors guiding T_EM efficiency in comparison to T_CM. To test this, we quantified T_EM and T_CM target cell killing in detail. First, we quantified how long T_EM or T_CM needed to find their first target cell ( Figure 7A ). We did no distinguish whether this contact was successful or not, i.e. resulting in target cell destruction or not after T_EM or T_CM left the target. There was no difference between the time to first contact ( Figure 7A , median T_EM 42.5 min to T_CM 40 min; n=96, 106 cells, respectively), indicating that T_EM and T_CM migration patterns or search strategies do not influence the likelihood of finding a target cell.

Next, we analyzed killing sequences in detail as displayed in Figure 7B . Examples for a T_EM and T_CM are shown in Figures 7C, D , respectively. We quantified the time to the first successful contact, i.e. the time it took a T_EM or T_CM needed to contact a target which was subsequently also killed by the CTL. The time to the first successful contact forming an IS was on average 1 hour for T_EM ( Figure 7C , statistics in 7E ). This time was significantly shorter than for T_CM (more than 2 hours, Figure 7D , statistics in 7E ). In addition, during a successful response the time from IS formation to apoptosis induction was significantly shorter for T_EM than for T_CM ( Figures 7D, E , statistics in 7F , median 10 versus 55 min, p<0.0001) and IS duration was also shorter in case of T_EM which is consistent with the results previously described by Jenkins et al. The authors discovered dramatically prolonged synapse times (up to fivefold) in CTL/NK cells failing perforin and granzyme cytotoxicity, which conversely means shorter IS duration for efficient killers (31).

The differences between T_EM and T_CM are summarized in Figure 7G . Time to first successful contact (dark grey), duration of IS, time between IS formation and apoptosis induction in targets and also the time between apoptosis and secondary necrosis are all significantly shorter in T_EM than T_CM.

In summary, whereas T_EM and T_CM need the same time to find a target cell T_CM need much longer than T_EM to establish the first contact with a subsequent successful killing. This means that T_CM have more failures, i.e. contacts with targets not resulting in killing. In addition, all measured time intervals (IS length, time to induce apoptosis, time to induce secondary necrosis) are shorter for T_EM compared to T_CM.

Research with human PBMC has been approved by the local ethic committee (84/15; Prof. Dr. Rettig-Stürmer). We got the leukocyte reduction system (LRS) chambers, a by-product of platelet collection from healthy blood donors, from the local blood bank (Institute of Clinical Hemostaseology and Transfusion Medicine, Saarland University Medical Center). All blood donors provided written consent to use their blood for research purposes.

Anti-human CD178 was obtained from BD Biosciences, anti-human CD95 Apo 1-1 was from Enzo Life Sciences. Flow cytometry antibodies (for details see Flow Cytometry) and apoptosis inducing TRAIL-Antibody were purchased from Biolegend. IL-7 was from Peprotech, IL-12 and IL-15 were from Miltenyi. Nucleofector Kit V was from Lonza. The original vector pCasper-GR (#FP971) was from Evrogen. BSA, fibronectin and staurosporine from Streptomyces sp. were from Sigma-Aldrich. The Dynabeads CD8 Positive Isolation Kit, fetal bovine serum (FBS) (EU-approved), IL-2, G418, PBS, RPMI-1640, AIM-V medium, penicillin/streptomycin (P/S) and were from ThermoFisher Scientific. Concanamycin A (CMA) was from Santa Cruz Biotechnology. All other chemicals and reagents not specifically mentioned were from VWR, ThermoFisher Scientific or Sigma-Aldrich.

Raji cells (ATCC, #CCL-86) and NALM6 cells (DSMZ, #ACC128) were cultured in RPMI-1640 (ThermoFisher Scientific) supplemented with 10% FBS and 1% Penicillin/Streptomycin at 37°C and 5% CO₂. Human peripheral blood mononuclear cells (PBMCs) were isolated from the leukoreduction system chamber (LRSC) as a by-product from healthy thrombocyte donation as described previously (55).

PBMCs were stimulated with staphylococcal enterotoxin A (SEA; 0.5 μg/ml) in a 24 well plate (density of 1–1.5 × 10⁸ cells/ml; 37°C for 1 h). PBMCs were resuspended at a density of 2–4×10⁶ cells/ml in AIM-V medium supplemented with 10% FBS and 100 U/ml recombinant human IL-2. 5 days after stimulation, SEA-specific CTL were positively isolated using a Dynabeads CD8 Positive Isolation Kit (ThermoFisher Scientific).

Purified CD8⁺ cells were cultured in AIMV medium supplemented with 10% FBS and 100 U/ml recombinant human IL-2 (ThermoFisher Scientific) and used for sorting of CD8⁺ T cell subsets and for further experiments three days after isolation.

The sequence encoding for Casper3-GR was amplified from pCasper3-GR (Evrogen # FP971) with the following primers introducing an XhoI recognition site at both ends of the amplicon. Forward primer: 5’-CTCGAGGCCACCATGGTGAGCGAG-3’, reverse primer: rev 5’-GACGAGCTGTACCGCT GACTCG AG-3’. The amplicon was subcloned into the XhoI site of a previously modified pGK-Puro-MO70 vector backbone were CITE-EGFP sequence was deleted (56). The generated plasmid encoded puromycin resistance gene and was used for selecting transfected NALM6 cells. 1 x 10⁶ NALM6 target cells were transfected with 1 µg plasmid using the Nucleofector 4D (SF cell line Kit, program CV-104, Lonza) following the manufacturer’s instruction. After 48 hours, selection started with 0.2 µl/ml puromycin (1 mg/ml stock solution). To enrich the pCasper GFP⁺/RFR⁺ population for clonal selection, GFP- and RFP-positive cells were sorted on a FACSARIAIII sorter and plated as 1 cell/per well in 96 well plates. GFP- and RFP-positive clones were selected by the ImageXpress Micro XLS screening image analysis system (Molecular Devices), expanded and frozen in 90%FBS/10%DMSO for further use.

The frequency of CD8⁺ T cell subpopulations was examined by intra- and extracellular stainings. 5x10⁵ cells were washed twice in PBS/0.5% BSA, stained with PerCP-conjugated anti-human CD3 (SK7, Biolegend), FITC-conjugated anti-human CD8 (SK1, Biolegend), PE/Cy7-conjugated anti-human CD45RA (Hl100), Biolegend) or CD45RO (UCHL-1, Biolegend), PerCp/Cy5.5-conjugated anti-human CD62L (DREG-56, Biolegend), Alexa Fluor-647-conjugated anti-human CD197 (150503, BD Biosciences), PE-conjugated anti-human CD253 (308205, Biolegend) antibodies in PBS/0.5% BSA, washed twice and recorded on a BD FACSVerse flow cytometer (BD Biosciences). For intracellular stainings, cells were fixated in 4% PFA in PBS for 20 min, washed in PBS/0.5% BSA, permeabilized for 10 min in PBS/0.1% saponin and stained with FITC-conjugated anti-human Perforin (δG9, Biolegend) in 0.1% PBS/0.1% saponin. After staining cells were washed and acquired on a FACSVerse flow cytometer.

For analysis of degranulation and staining of the death receptors FasL and TRAIL, 3x10⁵ SEA-pulsed target cells or target cells without SEA as control were seeded in wells of a 96-well plate. 6x10⁵ SEA-CTL were added to each well and cells were co-incubated in the presence of Brilliant Violet 421-conjugated anti-human CD107a antibody. After 4 h cells were collected on ice, washed (4°C) and a surface staining of CD45RA, CCR7, PE-conjugated anti-human CD178 (NOK-1, Biolegend) or PE-conjugated anti-human CD253 (RIK-2, Biolegend) was performed, cells were washed before analysis on a FACSVerse flow cytometer (BD Biosciences).

T_EM or T_CM were generated from human PBMC (peripheral blood mononuclear cell) using a widely accepted protocol for polyclonal activation of T cells by staphylococcal enterotoxin A (SEA) (47, 57) that importantly allows the subsequent analysis of SEA-dependent cytotoxicity (39, 48). For sorting 4x10⁷ SEA-CTL were stained with anti-CD62L and anti-CD45RO in PBS + 0.5% BSA, washed, strained (30 µm) and sorted on a BD FACSAria III (70 µm nozzle). Cells were collected in PBS/0.5% BSA/25% FBS, washed and cultured in AIM-V + FBS + 100 ng/ml IL-2. T_EM received 25 ng/ml IL-2, 5 ng/ml IL-12, 10 ng/ml IL-15, T_CM received 25 ng/ml IL-2, 5 ng/ml IL-7 and 10 ng/ml IL-15. Cells were measured on a BD FACSVerse flow cytometer and data analysis was performed using FlowJo.

To quantify the cytotoxicity of SEA-stimulated CD8⁺ T cells against SEA-pulsed Raji or NALM6 target cells we used a time resolved, real-time killing assay that was carried out essentially as previously described (39). Briefly, target cells (Raji or NALM6) were pulsed with 1 µg/ml SEA in AIM-V medium at 37°C and 5% CO₂ for 30 min. Next, cells were loaded with 500 nM Calcein-AM in AIM-V medium supplemented with 10 mM HEPES for 15 min at room temperature. After one washing step, 2.5 x 10⁴ target cells per well were pipetted into a 96-well black plate with clear-bottom (#7342480, VWR). After a short rest (15 min) to let target cells settle down, effector cells were added at the indicated effector to target ratio (E:T) and target cell lysis was measured in a Genios Pro (Tecan) reader using bottom reading function over 4 h at 37°C every 10 minutes. Maximal lysis rates were calculated as the maximum increase between two subsequent measured points (10 minutes).

Single cell killing assay measurements were run on an ImageXpress Micro XLS screening image analysis system (Molecular Devices). The detection of apoptotic and necrotic target cells with the apoptosis sensor Casper-GR construct was carried out as described before (24). Briefly, NALM6 pCasper target cells were pulsed with 1 µg/ml SEA in AIM-V medium at 37°C and 5% CO₂ for 30 min, washed taken up in 100 µl of phenol-free RPMI-1640/10% FBS and seeded into fibronectin coated wells (coating 30 min, 50 µl of 0.1 mg/ml fibronectin per well) of a 96-well plate (1x10⁴ targets per well). After a short rest (30 min) 2x10⁴ effector cells in 100 µl of phenol-free RPMI-1640/10% FBS were added per well and the measurement was started. Over 8h bright field, GFP- and FRET-signals were acquired with a 20x objective from 1-4 positions per well at a 5 min interval. Objects were excited via Spectra X LED illumination (Lumencor) using LED 470/24 for GFP. The filter sets for FRET were 472/30 nm for excitation and 520/35 nm for emission to measure GFP and 641/75 nm to measure RFP. A Nikon Super Fluor objective (20x/NA 0.75) was used.

A semi-automated analysis to quantify results shown in Supplementary Figure 3 was performed in the following way. The spot detection algorithm of the software Imaris (Bitplane, V8.1.2) was used to identify target cells based on their average size and GFP-fluorescence signal. To avoid potential errors of a completely automated tracking caused by collision or cell contacts the detected spots were not connected over time. The number of target cells and their fluorescence signals can only be quantified for each single time point separately. Due to this reason, it is not possible to determine if a necrotic cell death was due to primary or secondary necrosis (because the state of a cell in the previous time points is unknown). GFP- and FRET fluorescence values were exported into Excel and analyzed as described in the manual tracking section above. Necrotic target cells lose their fluorescence signal and cannot be detected by the spot detection algorithm. We set the total number of detected spots in the first pictures (maximum of total detected spots, before CTL-target cell contact set) to 100% and used the difference in number of detected spots at later time points to estimate the proportion of undetected necrotic target cells. The proportion of apoptotic, primary necrotic or secondary necrotic cells at any given time point was calculated and displayed in a color-coded death plot diagram over time (24).

To inhibit perforin-mediated target cell lysis effector cells were preincubated for 2 h in phenol-free RPMI-1640/10% FBS supplemented with 50 nM concanamycin A (CMA) and 50 nM CMA was also present during the measurements. 1x10⁴ targets per well were re-suspended in 100 µl phenol-free RPMI-1640/10% FBS/50 nM CMA. Effector cells were added in 100 µl of phenol-free RPMI-1640/10% FBS/50 nM CMA after preincubation.

To inhibit granzyme B-mediated target cell lysis, effector cells were preincubated for 2 h in phenol-free RPMI-1640/10% FBS supplemented with 10 µM z-AAD-CMK and 10 µM z-AAD-CMK was also present during the measurement. To this end 1x10⁴ targets per well were submitted in 100 µl of phenol-free RPMI-1640/10% FBS/10 µM z-AAD-CMK. Effector cells were added in 100 µl of phenol-free RPMI-1640/10% FBS/10 µM z-AAD-CMK after preincubation.

To block FasL- and TRAIL-mediated target cell lysis inhibitory anti-human antibodies were used to block the interaction of FasL or TRAIL with their respective apoptosis inducing receptors. To this end 1x10⁴ targets per well and 2x10⁴ effector cells per well were each preincubated for 2 h in 10 µg/ml anti-human CD178 antibodies (NOK-1 and NOK-2, BD Biosciences) or 5 µg/ml anti-human TRAIL antibody (RIK-2, Biolegend). After 2 h of preincubation effector cells were added to target cells.

In principle the manual analysis was carried out as described before (24). Briefly: Roughly 25 target cells were chosen randomly for each recorded position and tracked with Speckle Tracker J. The center of each target cell was marked at each time point manually. X,y positions and GFP- and FRET fluorescence signals were exported for every tracked cell with the speckle tracker intensity trajectories plugin. Data were imported into Excel to perform normalization to the starting value of a target cell and calculate the donor ratio for each cell. The GFP- and FRET-signal as well as the donor ratio were analyzed to determine the fate of a cell, viable (donor ratio > 0.75), apoptotic (donor ratio < 0.75) or necrotic (loss of both fluorescence signals, threshold FRET-value 10% above background signal; FRET< 150 units). This was used to identify the status of each tracked target cell at every time point. A cell being apoptotic prior to necrosis was defined as secondary necrosis. A necrotic event without a rise of the GFP-signal and an accompanying drop of the donor ratio was defined as primary necrosis. The sum of all detected cells per condition was set to 100% and the proportion of apoptotic, primary necrotic or secondary necrotic cells at any given time point was calculated and displayed in a color-coded death plot over time.

Images were analyzed as described in (24). Briefly, background correction via rolling ball, F/F0 normalization and donor ratio calculation were performed in ImageJ (NIH, version1.51d) and Microsoft Excel (Microsoft Office 2016). Following rolling ball background subtraction in ImageJ measurements were tracked manually with Speckle TrackerJ plugin for ImageJ.

Ca²⁺ imaging was essentially as described in Schwarz et al. (58). Briefly, 3 x 10⁵ T_EM or T_CM were loaded at 22–23 °C for 20 min with 1 µM fura-2/AM (Thermo Fisher Scientific) in medium, washed with fresh medium, stored at room temperature for 10 min, and immediately used. For Ca²⁺ re-addition experiments, T_EM or T_CM were allowed to adhere to poly-L-ornithine-coated (0.1 mg/mL) glass coverslip, which were assembled in self-build chambers on the stage of an Olympus IX 70 microscope. Measurements were carried out at 22–23 °C. A 20x objective (UApo/340, N.A. 0.75) was used, and cells were alternately illuminated at 340 and 380 nm with a Polychrome IV or V Monochromator (TILL Photonics) using SP 410 as excitation filter and DCLP 410 as dichroic mirror. The fluorescence emissions at >440 nm (LP 440) were captured with a CCD camera (TILL Imago), digitized, and analyzed using TILL Vision software. Ratio images and infrared images were recorded at intervals of 5 s. Background-corrected 340 nm/380 nm ratios were analyzed for each cell marked by “regions of interest” (ROIs) covering the whole cell. For analyzing Ca²⁺ signals of T_EM-Raji or T_CM-Raji contacts, the following modifications were made. Fura/AM-loaded T_EM or T_CM were settled down on Raji cells, already adhered to fibronectin-coated coverslips to allow T_EM or T_CM migration. All T_EM-Raji or T_CM-Raji contacts were analyzed. Ca²⁺ traces were aligned to the beginning of contact between T_EM or T_CM and Raji cells as indicated in the respective figure.

All molecular biology experiments were done as described before (59). Briefly, total RNA was isolated from 1.0 to 1.5 × 10⁶ CD8⁺ T-cells using TRIzol reagent (ThermoFisher Scientific) including 1 μl Glycogen (5 μg/μl, ThermoFisher Scientific). 0.8 μg total RNA was reverse transcribed to cDNA and 0.5 μl of cDNA was used for quantitative real time polymerase chain reaction (qRT-PCR). QRT-PCR was carried out in a CFX96™ Real-Time SystemC1000™Thermal Cycler (Biorad CFXManager). Primers used in this study: RNAPol and TBP as reference genes (59), FASL: forw 5′ GCACACAGC ATCATCTTT 3′, rev 5′ CAAGATTGACCCCGGAAGTA 3′; perforin (NM_005041 and NM_001083116) forw 5′ ACTCACAGGCAGCCAACTTT 3′, rev 5′ CTCTTGAAGTCAGGGTGCAG 3′; granzyme B, forw 5′ GAGACGACTTCGTGCTGACA 3′ and rev 5′ TCTGGGCCTTGTTGCTAGGTA 3′. QuantiTect primers for IFNγ, TNFα and TRAIL were purchased from Qiagen (QT00000525, QT00029162, QT00079212).

Data were tested for significance using one-way Anova, Mann-Whitney-Test, Kruskall-Wallis-Test with a Dunn´s multiple comparisons post-test: * p<0.05; ** p<0.01; *** p<0.001; ns, no significant difference as stated in the figure legends. Data analyses were performed using Microsoft Excel 2016, FlowJo, Imaris (Bitplane, V8.1.2), ImageJ, Igor Pro (WaveMetrics) and GraphPad Prism 7 software.