Human Antitumor CD8⁺ T Cells Producing Th1 Polycytokines Show Superior Antigen Sensitivity and Tumor Recognition

Adoptive transfer of T cells is an effective therapy for treatment of patients with life-threatening viral infections (1–3), various forms of leukemia and lymphoma (4), as well as some solid tumors (5–7). Successful therapy requires transfer of Ag-specific T cells with sufficient functional avidity to eliminate infected or transformed cells. Poor tumor recognition by lower avidity T cells can be bypassed using patient lymphocytes that are engineered to express high-affinity transgenic TCR (tgTCR) (8). Designer lymphocytes can be rapidly prepared for cancer patients based on matched HLA allotype and tumor profile. Adoptive transfer of designer lymphocytes with higher affinity tgTCR improved clinical efficacy (9, 10) but carried the risk for greater on-target toxicity, dependent on Ag specificity (11). Combined use of tgTCR with different specificities will likely be needed to avoid selection of tumor escape variants (10).

Libraries of TCR with suitable specificities and affinities must be collected for different tumor entities to fulfill broad clinical needs. A critical bottleneck exists in obtaining TCR with sufficient affinities that are specific for nonmutated self-proteins. Several approaches can be used, including obtaining T cells from mice that recognize human target proteins (12); however, xenogeneic TCR may not persist in vivo (13). This hindrance can be bypassed with a transgenic mouse carrying the full human TCR repertoire (14). TCR can be genetically modified to improve their affinity (15, 16) or higher affinity TCR can be acquired from allo-restricted, peptide-specific T cells (17–19).

Regardless of approach, many T cell clones must be analyzed in a time-consuming and costly process, further complicated by variations in proliferation and survival of individual clones in vitro. Therefore, rapid screening assays and selection parameters are needed to identify T cell clones whose TCR merit more study, despite limited T cell growth in vitro. Various functional parameters (20–22) and structural MHC-multimer–binding properties (23, 24) have been used for T cell characterization. Systematic comparisons, however, have used only a few clones with different specificities, leaving gaps in understanding the best overall measures for T cell selection.

Virus-specific T cells of humans and mice had proinflammatory polycytokine profiles that were critical for virus control (22, 25). Importantly, polyfunctional CD8⁺ T cells displayed increased Ag sensitivity (22). In this study, we performed a systematic comparison of 12 CD8⁺ antitumor CTL clones of identical peptide-MHC (pMHC) specificity to determine whether particular polycytokine patterns would also correlate with greater Ag sensitivity and superior tumor recognition. Transfer of function was assessed using TCR cloned from lower and higher avidity CTL. Functional and multimer-binding characteristics were compared to define the parameters best suited for rapid T cell selection in primary screening assays.

The human melanoma cell lines Mel-A375 (HLA-A2⁺, tyrosinase⁻), CRL-1619, American Type Culture Collection (ATCC); Mel 04A12 (HLA-A2⁺, tyrosinase⁺), transduced to secrete IL-2 (gift of P. Schrier, Department of Immunohematology, Leiden University Hospital, Leiden, The Netherlands); Mel-624.38 (26) and SK-Mel-23 (both HLA-A2⁺, tyrosinase⁺; gifts of M.C. Panelli, National Institutes of Health, Bethesda, MD); SK-Mel-29 (HLA-A2⁺, tyrosinase⁺; gift of P. Rieber, Institute of Immunology, Technical University Dresden, Dresden, Germany); WM 266.4 (HLA-A2⁺, tyrosinase⁺), CRL-1676, ATCC; as well as the lymphoid cell line T2 (HLA-A2⁺), CRL-1992, ATCC were cultured in RPMI 1640 medium supplemented with 12% FCS, 2 mM l-glutamine, 1 mM sodium-pyruvate, and 1 mM nonessential amino acids.

Effector CTL were derived by de novo priming with tyrosinase RNA-loaded dendritic cells, sorted using an HLA-A*02:01–tyrosinase_369–377 multimer (27), cloned by limiting dilution, and maintained in culture as previously described (19). All CTL are HLA-A*02:01 restricted and recognize the tyrosinase_369–377 peptide.

T cells and tgTCR-modified PBL were incubated with 100 pg/ml multimer (kindly provided by D.H. Busch, Institute of Medical Microbiology and Hygiene, Technical University Munich) in PBS + 0.5% human serum for 20 min. Allophycocyanin-labeled, CD8-specific Ab (Sk-1; BD) was added for a further 25 min. After washing, cells were analyzed on a FACSCalibur (BD Biosciences). MHC-multimer on rates were made at 4°C using titrated multimer (6–100 pg/ml). For assay of MHC-multimer off rates, cells were washed after multimer binding at 4°C, resuspended in FACS buffer containing BB7.2 mAb (HB-82; ATCC), which captures detached multimers to prevent rebinding. Cells were incubated at 4°C, and samples were removed, fixed, and analyzed after 0–2 h by flow cytometry. PE-labeled HLA-B7 or HLA-A*02:01 multimers loaded with pp65 peptide served as specificity controls (data not shown).

T2 cells (1 × 10⁶/ml) were incubated at 37°C and 5% CO₂ for 2 h with 10 μg/ml human β₂-microglobulin (Calbiochem) and titrated (10⁻⁵–10⁻¹¹ M) tyrosinase_369–377 peptide (YMDGTMSQV; Metabion). After washing, T2 cells were used in cytotoxicity, cytokine release, or intracellular cytokine staining (ICS) assays.

CTL (2 × 10³ cells/100 μl) or TCR-modified PBL (5 × 10⁴ cells/100 μl) were incubated with the melanoma cell lines (1 × 10⁴ cells/100 μl). PMA (5 ng/ml) and ionomycin (750 ng/ml; both Sigma) were used for TCR-independent stimulation. Culture supernatants at 24 h were assessed by standard ELISA (OptEIA Human IFN-γ Set; BD Biosciences). Data represent mean values. Multiplex protein assays for IFN-γ, IL-2, TNF-α, IL-5, IL-13, GM-CSF, MIP-1β, IL-4, IL-6, and IL-12p70 were measured in supernatants of CTL cocultures with or without tyrosinase_369–377–pulsed T2 cells (10⁻⁵ M), as specified by the manufacturer (Bio-Rad Laboratories). Standard curves and concentrations were calculated with Bio-Plex Manager 4.1.1, as described previously (28). Th1 cytokines (IFN-γ, IL-2, and TNF-α) released by tgTCR-transduced PBL after coculture with tumor cells or peptide-pulsed T2 cells were measured by Cytometric Bead Array as specified (BD Biosciences). FACSCalibur data were acquired and analyzed with FlowJo software (Tree Star) and FCAP Array Software (BD Biosciences).

CD107a/b mobilization and ICS assays were performed as described previously (29) using FITC-labeled Abs against CD107a and CD107b (clones H4A3 and H4B4; BD) and Golgi-Stop (brefeldin A at 10 μg/ml and monensin at 25 μM; both Sigma) at 37°C and 5% CO₂ for 6 h with CTL and unpulsed cells or T2 cells pulsed with 10⁻⁵, 10⁻⁷, and 10⁻⁹ M tyrosinase_369–377 peptide at a ratio of 2:1. Washed CTL were stained with CD8-Amcyan Ab (Sk1; BD) and 7-aminoactinomycin (Sigma) or live/dead fixable blue (Invitrogen) for 25 min at 4°C to detect dead cells. After washing, CTL were fixed with 1% paraformaldehyde (PFA) in PBS for 20 min at 4°C, washed thrice, and stored in PBS/1% FCS overnight. Cells were permeabilized by successive washes with 0.1 and 0.35% saponin (Sigma). ICS used V450-conjugated IFN-γ (B27; BD Biosciences), allophycocyanin-labeled IL-2 (MQ1-17H12; BD), A700-conjugated TNF-α (MAb11; BD), and PE-labeled TCRβ (BMA031; Immunotech) or murine TCRβ (H57-597; BD) Abs for 20 min. Data acquisition and analysis used an LSRII (BD Biosciences) and FlowJo software (Tree Star). CD8⁺TCRβ⁺ cells and CD8⁺murineTCRβ⁺ cells were gated for CTL and TCR-transduced PBL, respectively.

Cytotoxic activity was analyzed in a standard 4-h [⁵¹Cr]-release assay as published previously (19), using melanoma cell lines and peptide-loaded T2 cells as target cells. Percentages of specific, relative cytotoxicity and half-maximum cytotoxicity (EC₅₀) values were calculated as described previously (19).

Untreated and IFN-γ–stimulated (2 ng/ml; Imukin; Boehringer Ingelheim) melanoma cells were stained for MHC class I with W6/32 mAb (HB-95; ATCC) followed by detection with rat anti-mouse secondary Ab. Anti–IFN-γ Ab (50 μg/ml; MMHG-1; Calbiochem) was added to cocultures of CTL and melanoma cells to block CTL-derived IFN-γ. Alternatively, tumor cells were prefixed with 1% PFA to block response to IFN-γ.

Retroviral vector plasmids (19) were used for production of retroviral particles and subsequent transduction of T cells (30). Ten days after the second transduction, PBL were stained using allophycocyanin-labeled HLA-A2–tyrosinase_369–377 multimer (Coulter) and FITC-labeled CD8-specific Ab (BD Biosciences). On day 15, TCR-transduced PBL were used in cocultures with target cells to assess cytokine secretion, cytotoxicity, and CD107a/b mobilization or ICS.

The nonparametric, two-tailed Mann–Whitney U test was used to evaluate statistical differences between data sets derived from the lower and higher avidity groups in functional assays using GraphPad Prism software version 5.0 for Windows (GraphPad). Statistical significance is indicated by *p < 0.05 and **p < 0.01.

We previously described 12 HLA-A2–restricted tyrosinase-specific CD8⁺ CTL clones derived through self- or allo-restricted dendritic cell priming in vitro (19). Specific CTL sorted with MHC-multimer and peptide screening confirmed that all CTL recognized the HLA-A2–tyrosinase_369–377 pMHC epitope. The CTL displayed different peptide sensitivities in [⁵¹Cr]-release assays of T2 cells (HLA-A*02:01⁺) loaded with tyrosinase_369–377 peptide. For reference, peptide concentrations yielding EC₅₀ are depicted in Fig. 1A. The CTL clustered in two groups, with six in a lower avidity group showing EC₅₀ values near 1 × 10⁻⁸ M peptide and six in a higher avidity group with EC₅₀ values between 2 × 10⁻⁹ and 4 × 10⁻¹⁰ M.

The CTL killed the melanoma cell line Mel-93.04A12 (HLA-A2⁺tyrosinase⁺) to varying degrees (Fig. 1B) but did not kill Mel-A375 cells (HLA-A2⁺tyrosinase⁻) (data not shown). Higher avidity CTL mediated ≥50% tumor kill, whereas only two of the lower avidity CTL reached this level. IFN-γ secretion also varied, but always exceeded 1.0 ng/ml in the higher avidity group. Only two of the lower avidity CTL reached this level (Fig. 1C). Therefore, superior tumor cell recognition seen by greater cytotoxicity and higher IFN-γ secretion was associated with greater peptide sensitivity and distinguished the two groups, with only one exception (C43), when both parameters were used for CTL classification (Fig. 1D). Group differences in IFN-γ secretion were seen with additional melanoma cell lines (Supplemental Fig. 1A).

Because many factors including variations in levels of pMHC on tumor cells may influence CTL recognition, we moved to a standardized assay using T2 cells pulsed with saturating amounts (10⁻⁵ M) of tyrosinase_369–377 peptide. After 24-h coculture, cytokine secretion patterns were measured with a multiplex protein array detecting IL-2, IL-4, IL-5, IL-6, IL-12p70, IL-13, GM-CSF, IFN-γ, MIP-1β, and TNF-α. Only IFN-γ, IL-2, TNF-α, IL-5, IL-13, GM-CSF (Fig. 2), and MIP1-β (data not shown) were detected among the CTL. All higher avidity CTL made IFN-γ, but only four lower avidity CTL surpassed background levels (Fig. 2A). The mean group differences were significant (Mann–Whitney U test, **p < 0.01). Three lower avidity CTL secreted IL-2 and four released TNF-α above background, whereas all higher avidity CTL exceeded background for both cytokines. Although tending toward significance, these mean group differences did not reach significance because of the exceptionally high values of C125, which, however, showed very poor tumor cell kill (Fig. 1B).

The percentages of CTL in the two groups secreting one or more of the three Th1/Tc1 cytokines (hereafter Tc1-triplet) are depicted in Fig. 2B. All higher avidity CTL secreted the Tc1-triplet above background (Fig. 2B, right); only three lower avidity CTL produced this Tc1-triplet (Fig. 2B, left). All CTL released MIP-1β, but in greater amounts by the higher avidity group (data not shown). Reciprocal patterns were seen for IL-5, IL-13, and GM-CSF (Fig. 2C). In particular, only lower avidity CTL produced IL-5 (Mann–Whitney U test, **p < 0.01), and levels of IL-13 levels were greater in this group (Mann–Whitney U test, **p < 0.01). GM-CSF release was very low throughout and was poor for discrimination. No higher avidity CTL showed a Th2/Tc2 (hereafter Tc2) pattern after peptide-specific stimulation (Fig. 2D). The CTL showed different capacities to release various cytokines after TCR-independent activation with PMA and ionomycin; these mostly followed the Tc1/Tc2-qualitative, but not -quantitative, patterns seen after TCR-dependent, peptide-specific stimulation (Supplemental Fig. 1B, 1C).

To determine whether CTL-derived IFN-γ directly increased MHC class I on tumor cells, and thereby influenced recognition, we analyzed two CTL (C115 and C58) that released different amounts of IFN-γ after tumor cell contact (Fig. 1C). Direct stimulation with exogenous IFN-γ (2 ng/ml) did not alter Mel-624.38, but Mel-A375 had increased class I expression (Fig. 3A, top row). Tumor cells prefixed with 1% PFA were blocked for IFN-γ response but still retained their normal MHC expression (Fig. 3A, bottom row). Thereafter, C115 and C58 were cocultured with tumor cells, either in the presence of IFN-γ–blocking Ab (Fig. 3B, top row) or using PFA-fixed tumor cells (Fig. 3B, bottom row). IFN-γ could not be detected in cultures with IFN-γ–blocking Ab, but IL-2 and TNF-α secretion remained greater for C58 despite the IFN-γ block. Likewise, recognition of PFA-treated tumor cells was better with C58 for the full Tc1-triplet. Thus, superior recognition by C58 did not depend on its capacity to modulate MHC class I via higher IFN-γ secretion.

To determine whether characteristics of original CTL would be recapitulated by TCR gene transfer, we modified recipient PBL to express TCR of C115 or C58, as described previously (19). CD8⁺ cells expressing tgTCR were enumerated by HLA-A2–tyrosinase_369–377 multimer binding, and representative results for one matched pair are shown in Fig. 4A. EC₅₀ values of cytotoxicity against peptide-pulsed T2 cells were determined for matched pairs in independent experiments (Fig. 4B). PBL expressing TCR-C58 (PBL-TCR-C58) always displayed greater peptide sensitivity than PBL expressing TCR-C115 (PBL-TCR-C115). EC₅₀ cytotoxicity values differed by almost one log (mean 5.1 × 10⁻⁹ versus 3.7 × 10⁻⁸ M), but these only trended toward significance (Mann–Whitney U test, p = 0.11). This paralleled the original CTL, both of which showed good tumor cell killing (Fig. 1B). In matched pairs, Tc1-triplet secretion was always induced in PBL-TCR-C58 with less peptide (Fig. 4C). The EC₅₀ values for IFN-γ (**p < 0.01) and IL-2 (*p < 0.05), but not TNF-α, were significantly different (Mann–Whitney U test). Only PBL-TCR-C115 produced IL-5 and IL-13 above background levels after peptide stimulation. They also secreted more GM-CSF (Supplemental Fig. 2A).

Differences in peptide sensitivity for cytotoxicity and cytokine release could be directly ascribed to the tgTCR because they were introduced into the same recipient PBL populations, expressed at comparable levels, and assessed at the same time for each matched pair depicted in Fig. 4. Responses were Ag specific and required tgTCR expression because cytokine release required stimulation with specific tyrosinase_369–377 peptide and unmodified PBL did not release these cytokines (data not shown).

PBL-TCR-C58 also secreted more Tc1-triplet cytokines than PBL-TCR-C115 after tumor cell stimulation. Again, recognition was tgTCR dependent and Ag specific because untransduced PBL did not respond and Mel-A375 (HLA-A2⁺ tyrosinase⁻) was not recognized (Fig. 4D). Differences were seen with additional melanoma lines (HLA-A2⁺ tyrosinase⁺; Fig. 4E), but these were not significant, perhaps because of wide fluctuations in pMHC. PBL-TCR-C115 also released more IL-5, IL-13, and GM-CSF after tumor cell stimulation (Supplemental Fig. 2B). However, mean differences were only significant when PBL expressing tgTCR were stimulated with specific peptide, likely because of complex variations in tumor cells. Importantly, however, was the finding that the reciprocal Tc1/Tc2 patterns of the original CTL were recapitulated by corresponding TCR transfer into PBL. TCR-transduced PBL showed comparable capacities to release Tc1 cytokines after TCR-independent activation with PMA and ionomycin (data not shown).

ICS was tested as an alternative to analyze Tc1 polycytokine responses of the CTL, also allowing percentages of positive cells to be determined (Fig. 5, data not shown). Original C115 and C58 are shown in Fig. 5A, and PBL-TCR-C115 and PBL-TCR-C58 are shown in Fig. 5B, depicting the specific fraction of cells expressing the corresponding tgTCR. CTL were stimulated with T2 cells pulsed with excess (10⁻⁵ M) and two lower concentrations (10⁻⁷ and 10⁻⁹ M) of tyrosinase_369–377 peptide. Background staining of nonstimulated cells served as a negative control. The original CTL stained for the Tc1-triplet after stimulation with 10⁻⁵ M peptide, with comparable mean fluorescence intensities (MFIs). In contrast, greater MFI was detected in C58 for the Tc1-triplet using 10⁻⁷ M peptide. Only a small fraction of C58 cells showed IFN-γ and TNF-α with 10⁻⁹ M peptide. PBL expressing the tgTCR showed similar trends, with no clear distinction at 10⁻⁵ M peptide and superior ICS for the Tc1-triplet in PBL-TCR-C58 at 10⁻⁷ M peptide. Minimal IFN-γ and TNF-α was seen in PBL-TCR-C58 at 10⁻⁹ M peptide, whereas no ICS was seen for PBL-TCR-C115.

In addition, cytotoxic potential was measured by staining for CD107a/b (LAMP-1), a protein associated with release of cytolytic granules upon target cell contact (29). Original CTL expressed CD107 after stimulation with excess peptide (10⁻⁵ M), and C58 showed somewhat greater ICS at lower peptide concentrations (Fig. 5C). These distinctions were well recapitulated with the tgTCR populations (Fig. 5D). Although phenotypic shifts were seen in ICS and CD107 assays for both CTL and transgenic PBL, the MFIs were not significantly different (data not shown). Thus, measures of MFI in this assay did not match the significant quantitative differences detected by multiplex protein arrays.

Because the CTL were isolated via HLA-A2–tyrosinase_369–377 multimer, they could also be compared for multimer-binding parameters. Incubation with titrated amounts of multimer depicted multimer on rates. Loss of multimer over time, both with respect to percentage of positive cells and MFI, indicated the multimer off rates. All CTL showed variations in both measures (Supplemental Fig. 3, left). Importantly, ranking of CTL with the best to worst on or off rates did not correlate with their lower or higher avidity defined by peptide sensitivity (Supplemental Fig. 3, right). These results showed that multimer binding and functional parameters did not distinguish the same CTL groups.

Of particular interest were comparisons of C58 and C115, and the corresponding tgTCR-PBL (Fig. 6). Based on superior function, we expected C58 and PBL-TCR-C58 to have faster on rates and slower off rates. Surprisingly, opposite results were obtained whereby C115 and PBL-TCR-C115 showed faster on rates and slower off rates (Fig. 6A), indicating that they displayed greater TCR-pMHC stability (Fig. 6B, 6C).

The CTL were assessed for various surface markers, including CD3, CD25, CD28, CD40L, and CD62L, but discrete patterns that demarked the lower and higher avidity groups were not found (data not shown). Likewise, multimer staining (MFI) did not correlate with peptide sensitivity and superior tumor cell recognition. Rather, patterns of peptide-specific cytokine secretion and levels of tumor cell kill represented the prominent characteristics that distinguished the two groups, as illustrated in Table I. All higher avidity CTL simultaneously secreted the Tc1-triplet, but they produced no IL-5 and low IL-13 after peptide-specific stimulation. Fewer lower avidity CTL secreted the Tc1-triplet, but they always produced IL-5 and more IL-13 after peptide-specific stimulation. Also, higher avidity CTL mostly showed greater kill of tumor cells compared with lower avidity CTL. Thus, a functional biomarker of six parameters, including Tc1-triplet cytokine production with missing or modest Tc2 cytokine secretion, and greater cytotoxicity served well to demark CTL of higher avidity. A summed rank analysis of the 12 CTL, combining Tc1/Tc2 cytokine patterns and cytotoxicity, led to clustering of the lower avidity CTL in the lowest six positions, whereas the higher avidity CTL ranked in the top six positions. The cytokine biomarker analysis could be performed using 8 × 10³ CTL, which can normally be cultured in 3–4 wk, starting from a single cell. A second level screen for cytotoxicity requires 4 × 10⁴ cells (Supplemental Table I).

The goal of these studies was to identify functional or MHC multimer-binding parameters for primary screening to rapidly identify CD8⁺ T cell clones with superior antitumor function. We sought to determine whether the capacity of CTL to simultaneously produce particular cytokines would correlate with greater peptide sensitivity and improved tumor recognition. We also assessed the relationship between functional parameters and MHC-multimer binding. Furthermore, we determined whether characteristics of representative CTL would be recapitulated by TCR gene transfer, thereby improving development of designer lymphocytes. These questions were addressed using 12 CTL with identical pMHC specificity but varying Ag sensitivity that were previously characterized extensively for tumor specificity (19). The CTL were cryopreserved soon after isolation so that retrospective assays performed in this study reflect responses expected for newly emerging clones. The TCR of two CTL with different peptide sensitivities were used to assess transfer of function to recipient PBL.

Sensitive ELISA measuring either GM-CSF or IFN-γ have been used routinely to assess emerging T cell clones (16, 19, 22). In our studies, low levels of GM-CSF made it poorly suited to demark CTL with greater peptide sensitivity. IFN-γ secretion was more reliable, but inclusion in a multiplex-cytokine protein array delivered substantially more information. Comparisons of 10 factors showed that the Tc1-triplet of cytokines (IFN-γ, IL-2, TNF-α) successfully earmarked all higher avidity CTL but also identified several lower avidity CTL. Better group discrimination was achieved by including IL-5 and IL-13.

ICS assays of Tc1 cytokines also showed differences in lower and higher avidity CTL, but only when at least two concentrations of peptide were compared: excess peptide (10⁻⁵ M) demonstrated CTL capacity to produce Tc1 cytokines, whereas 10⁻⁷ M peptide showed greater ICS in higher avidity cells. CD107 mobilization also showed CTL differences after stimulation with 10⁻⁵ and 10⁻⁷ M peptide. Importantly, ICS and CD107 staining were recapitulated in transgenic PBL expressing TCR-C58 and TCR-C115, respectively. However, MFI values measured in ICS were not statistically significant, in contrast with values detected in multiplex assays. Based on this fact, as well as the need for many fewer cells, the multiplex protein array provided a faster and more reliable method for primary T cell selection. Preliminary studies suggest that further improvement is possible if two concentrations of specific peptide (10⁻⁵ and 10⁻⁷ M) are used for stimulation in multiplex assays (data not shown). Upon further expansion of Tc1-selected clones, a second level assay for tumor cell cytotoxicity would earmark CTL with presumed optimal antitumor functions.

Because these CTL were isolated by MHC-multimers, they could be compared for multimer-binding parameters. TCR–pMHC interactions can be judged by MFI of multimer binding as well as multimer on and off rates. All three measures correlated significantly with each other; however, none of these structural parameters correlated significantly with better CTL function assessed by the various functional assays (data not shown). This was clearly exemplified by analysis of C115 and C58, where C58 showed greater peptide sensitivity and superior function, but C115 had better multimer-binding parameters. Importantly, this reciprocal pattern was retained when the corresponding TCR were expressed in PBL. Multimer binding was useful for initial T cell isolation from priming cultures but failed to identify the best CTL with respect to Ag sensitivity and tumor recognition. Thus, functional parameters were needed for primary clone selection. New multimer-binding assays may overcome this deficit in the future. Biocore measurements are more precise (31, 32) but clearly not feasible for primary CTL screening. Others have also noted discrepancies in multimer binding and tumor cell recognition (33–36). Our studies add further support through analysis of multiple CTL with identical specificities. Preliminary live-cell imaging indicates that lower and higher avidity CTL differ in time to attachment and detachment from tumor cells, which may contribute to these discrepancies (data not shown).

The transfer of TCR from representative lower and higher avidity CTL into the same PBL populations allowed assessment of the therapeutic form needed for TCR gene therapy. Importantly, TCR-transgenic PBL recapitulated the properties of original CTL. Thus, polyfunctional Tc1 cytokine secretion rather than multimer-binding parameters served best for early clone selection and concurrently pinpointed important characteristics needed for designer lymphocytes. In this study, the capacity to secrete IL-2 may be particularly important because CD8⁺ TCR-transgenic PBL producing this IL may survive longer in vivo in the absence of CD4⁺ helper cells.

The association of Tc1- and Tc2 cytokine patterns with higher and lower avidity CTL, respectively, harken back to earlier studies of murine CD4⁺ T cells showing that potency of TCR–pMHC interactions regulated cytokine gene transcription and impacted on emergence of Th1 or Th2 responses (37–39). These original studies used altered peptide ligands to assess single transgenic TCR or mutants thereof. Surprisingly, our CD8⁺ CTL with naturally occurring TCR variations displayed Tc1 or Tc2 cytokine patterns after stimulation with an identical pMHC ligand, dependent on the peptide sensitivity of the individual CTL. These results suggest that selection of TCR based on peptide sensitivity may allow differential cytokine secretion to be manipulated in TCR gene therapy.

Studies of virus-specific CD8⁺ T cells demonstrated that polycytokine production of IFN-γ, IL-2, and TNF-α was associated with superior virus control in vitro and in vivo (22, 25, 40). Furthermore, the CD8⁺ HIV-specific clones that secreted these cytokines displayed greater peptide sensitivity (22). Our results extend these findings to antitumor responses, whereby greater peptide sensitivity and superior tumor recognition were also found for CTL that secreted this Tc1-triplet of cytokines. Thus, human CD8⁺ T cells with greater Ag sensitivity and superior antivirus or antitumor function share the common feature of Tc1 polycytokine secretion. Future clinical studies will show whether transfer of these characteristics to patient PBL via selected TCR will improve clinical outcomes in TCR gene therapy of tumors.

We thank S. Eichenlaub, K. Hummel, S. Kresse, and A. Slusarski for auxiliary work, J. Eckl, S. Krauss-Etschmann, and C. Heumann for help with statistical analysis, D.H. Busch for providing the multimer technology, and J. Mysliwietz, A. Turqueti-Neves, and J. Eckl for help with flow cytometry.

The authors have no financial conflicts of interest.