Quantitative TCR:pMHC Dissociation Rate Assessment by NTAmers Reveals Antimelanoma T Cell Repertoires Enriched for High Functional Competence

Protective T cell responses are characterized by strong initial clonal T cell bursts arising from a relatively small number of Ag-specific naive precursors. Following the encounter with their cognate peptide presented by MHC (pMHC), T cells proliferate and acquire various cellular survival and effector functions assuring long-term persistence, homing to inflamed tissues, and cytotoxic capacities. T cells can differentiate into memory and effector cells, with the extent of T cell differentiation being dependent, in part, on the strength of signals emanating from the TCR (1). Compared with other molecular interactions, the TCR:pMHC-binding strength is relatively weak (∼1–100 μM) with fast dissociation kinetics (t_1/2 <60 s) (2). TCR:pMHC interactions can be measured in terms of affinity or avidity, both of which can directly impact the overall functional T cell response (e.g., cytokine secretion, target cell killing) (3). The physical strength with which a TCR binds to a given pMHC complex is referred as the affinity of the TCR:pMHC interaction. In contrast, in the cellular context, TCR/CD8:pMHC avidity defines the strength of binding of monomeric TCR to the pMHC with the inclusion of CD8 coreceptor binding.

In cancer, CD8 T cells often recognize tumor-associated Ags of self-origins, and, due to thymic negative selection, tumor-specific T cells express TCRs of lower affinity/avidity compared with TCRs specific for viral, nonself epitopes, or mutated epitopes (4). Consequently, there is an expanding body of work on engineering TCRs with increased avidity, which endow T cells with improved functionality in clinical settings (5). As the challenge of cancer immunotherapy lies in the identification and activation of tumor-specific CD8 T cells offering a maximal functional avidity, T cell selection for immunotherapy is currently based on assays measuring Ag specificity, proliferation, and functional capacity (e.g., IFN-γ production, release of granzyme B and perforin) (6–9). However, technical limitations are the major reasons that TCR:pMHC-binding kinetics have yet to be considered as a determining factor in the selection of Ag-specific T cells for clinical protocols. Furthermore, although the TCR:pMHC affinity/avidity for pMHC is a major determinant of thymic selection (10), autoimmune pathogenicity (11–13), and T cell responsiveness (2), the precise impact of TCR:pMHC-binding parameters (k_on and k_off) on CD8 T cell repertoire selection and differentiation within the periphery in well-defined clinical settings, such as therapeutic vaccination, remains to be directly evaluated.

To this day, a debate remains regarding the accurate method to measure TCR:pMHC-binding kinetics (k_on and k_off) and which parameter(s) could better predict T cell activation and function (14). Whereas pMHC multimers were often used in binding and dissociation assays, results are biased by the multivalent nature of multimers and the ensuing use of rebinding antagonists (15, 16). Conversely, three-dimensional soluble TCR:pMHC affinity measurements by surface plasmon resonance (SPR) omit the impact of CD8 coreceptor binding and membrane fluidity (15). The two-dimensional techniques have offered important membrane-associated kinetics insights (17) and led to the elaboration of mathematical models (14) predicting TCR:pMHC interactions and their association with T cell functions. However, these labor-intensive techniques are not well suited for the high-throughput characterization of living Ag-specific T cells. Recently, novel approaches have allowed the direct and precise quantification of TCR:pMHC dissociation rates (k_off) on living CD8 T cells by real-time flow cytometry (15, 18) or microscopy (19). Among these, the reversible NTA-His tag–containing multimers (NTAmers), which can be rapidly dissociated into monomeric pMHC (15, 18), allow the accurate measurements of dissociation rates (k_off), even for weak TCR:pMHC interactions, such as those found in tumor-specific CD8 T cell repertoires (20). This technique allows for the quantification of dissociation kinetics (k_off) with higher sensitivity than multimer-based assays (20). Importantly, the monomeric dissociations of NTAmers account for the impact of the CD8 coreceptor in the overall avidity of the TCR for its cognate pMHC, an aspect not taken into consideration during SPR experiments and defined thereafter as TCR/CD8:pMHC avidity. Nevertheless, the k_off values measured by NTAmers correlate significantly to those derived from SPR assays (20). Together, the NTAmer technology represents a powerful tool to precisely quantify TCR:pMHC dissociation rates (k_off) on living CD8 T cells by real-time flow cytometry and at monomeric resolution.

Most groups use transgenically modified TCR or altered peptide libraries to establish an affinity range on which one can then evaluate the impact of TCR/CD8:pMHC avidity on T cell activation and functionality. Few studies are based on quantifying the association between TCR/CD8:pMHC avidity and T cell responsiveness in naturally occurring TCR repertoires and in humans. Our group has thoroughly characterized a cohort of melanoma patients who received peptide-based vaccination in combination with CpG and IFA (21). Interestingly, patients vaccinated with the native HLA-A2/Melan-A^MART-1_26–35 peptide developed antitumor T cell responses with improved functionality compared with patients vaccinated with the analog Melan-A^MART-1_{26–35 (A27L)} peptide (22–24). We will define hereafter the analog peptide as high affinity in comparison with the low-affinity native peptide as the analog Melan-A^MART-1_{26–35 (A27L)} peptide has a 10-fold increased binding to HLA-A2 (25). The vaccine-induced tumor-specific CD8 T cell clones derived from these patients thus represent valuable tools to inquire the relationship between TCR:pMHC kinetics, T cell repertoire selection, and functionality in a clinically relevant cancer setting.

In this study, we show that the use of a low-affinity (native) peptide in a therapeutic vaccination setting of melanoma patients favored the selection/differentiation of tumor Ag-specific T cells bearing TCRs with high TCR/CD8:pMHC avidity. Using a large series of Melan-A^MART-1_26–35–specific clones (n = 92), we demonstrate that CD8 T cells with slower dissociation rates (k_off) and longer t_1/2 had improved functional avidity (Ca²⁺ flux and killing potential). Together, our work shows the feasibility of TCR/CD8:pMHC avidity assessment in patients using NTAmers and offers novel indications that low-affinity (native peptide) vaccination increases anti-cancer TCR/CD8:pMHC avidity and CD8 T cell functionality.

We acknowledge the concept of the minimal information about T cell assays framework that was recently published (26, 27). Detailed information is structured in the modules proposed by the minimal information about T cell assays: samples, assays and data acquisition, data analysis, and laboratory environment in which the T cell assays were performed.

HLA-A*0201–positive patients with stage III/IV metastatic melanoma were included in a phase I prospective trial (ClinicalTrials.gov Identifier: NCT00112229) (21, 22). Study protocols were designed and conducted according to the relevant regulatory standards, upon approval by the ethical commissions and regulatory agency of Switzerland. Patient recruitment, study procedures, and blood withdrawal were carried upon written informed consent from all participating individuals (n = 7) prior to their inclusion in the study. The patient characteristics were previously described in details (21, 22).

Patients received monthly low-dose vaccinations injected s.c. with 100 μg either the low-affinity Melan-A^MART-1_26–35 unmodified native peptide (EAAGIGILTV) or the high-affinity Melan-A^MART-1_{26–35 (A27L)} analog peptide (ELAGIGILTV), mixed with 0.5 mg CpG 7909/PF-3512676 (Pfizer and Coley Pharmaceutical Group) and emulsified in IFA (Montanide ISA-51; Seppic) (21).

Ficoll-Hypaque (Pharmacia, Uppsala, Sweden) centrifuged PBMCs were cryopreserved in RPMI 1640, 40% FCS, and 10% DMSO at a concentration of 1 × 10⁷ PBMCs per ml for 72 h at −80°C and then transferred into the vapor phase of liquid nitrogen until further use.

After thawing, PBMCs were positively enriched using anti-CD8–coated magnetic microbeads and a MiniMACS device (Miltenyi Biotec, Bergish Gladbach, Germany) resulting in >90% CD3^pos/CD8^pos cells. Cells were stained in PBS, 0.2% BSA, and 50 μM EDTA with multimers (PE-labeled HLA-A*0201 analog Melan-A^MART-1_{26–35(A27L)} [TCMetrix Sàrl]) at 4°C for 45 min and surface markers staining at 4°C for 30 min, and then sorted into defined subpopulations of early differentiated effector memory (EM) multimer⁺/CD8⁺/CD45RA⁻/CCR7⁻/CD28⁺ (EM28⁺) and late differentiated multimer⁺/CD8⁺/CD45RA⁻/CCR7⁻/CD28⁻ (EM28⁻) used on a FACSAria (BD Biosciences) or Astrios (Beckman Coulter) cytometer. Sorted multimer^pos CD8 T cell subsets were cloned by limiting dilution in Terasaki plates and expanded in RPMI 1640 medium supplemented with 8% human serum, 150 U/ml human rIL-2 (gift of GlaxoSmithKline), 1 μg/ml PHA (Sodiag), and 1 × 10⁶/ml irradiated allogeneic PBMC (3000 rad) as feeder cells. The human serum lot was pretested for performance and mycoplasma contamination. T cell clones were expanded by periodic (every 14–21 d) restimulation with PHA, human rIL-2, and irradiated feeder cells. T cell clones were selected for the various assays following expansion and validation of Ag specificity by analog Melan-A^MART-1_{26–35 (A27L)} multimer staining. We have previously compared the sorting of Melan-A–specific CD8 T cells either with the analog-peptide or the native-peptide multimers. Compared with the native-specific multimer, the use of the analog-specific multimer does not introduce qualitative bias in the TCR Vβ repertoires nor quantitative differences in multimer-specific T cell frequencies (22, 23).

FACS experiments were performed using the following six-color stain combination: PE HLA-A2/ELA/NTAmer (TC Metrix), FITC anti-CD45RA (BD Biosciences), PE-Cy7 anti-CCR7 (BioLegend), allophycocyanin anti-CD28 (BD Biosciences), allophycocyanin A750 anti-CD8 (Beckman Coulter), and DAPI reagents. Staining was performed in a 50 μl FACS buffer (PBS, 5 mM EDTA [Life Technologies], 0.2% BSA [Merck], and 0.2% NaN₃ [Merck]) for 30 min at 4°C. The intracellular content of granzyme B and perforin was measured in Melan-A–specific CD8 T cell clones under unstimulated (baseline) conditions. Cells were fixed with 1% paraformaldehyde, 2% glucose, and 5 mM NaN₃ in PBS at room temperature during 20 min, before intracellular staining at 4°C during 20 min in FACS buffer plus 0.1% saponin (Sigma-Aldrich) with FITC antigranzyme B (Hölzel Diagnostica) or FITC antiperforin (Ancell). Vivid Aqua (Life Technologies) was used for live cell discrimination. All data were analyzed using FlowJo 9.7.6 software (Tree Star).

Samples were acquired using a Gallios or LSR-SORP (BD Biosciences) flow cytometer, using the Cytometry List Mode Data Acquisition & Analysis Software (v.1.2; Beckman Coulter) or FACSDiva (v.1.6, BD Biosciences) software, respectively. Photo Multiplier Tube voltages were set using unstained PBMCs, and compensations were calculated using single-stained PBMCs. A representative data set demonstrating the gating strategy is shown in Supplemental Fig. 1.

cDNA preparation and CDR3 spectratyping, sequencing, and clonotyping were performed, as described (24, 28). Briefly, each cDNA sample was subjected to individual PCR using a set of previously validated fluorescent-labeled forward primers specific for the 22 TRBV subfamilies and one unlabeled reverse primer specific for the corresponding Cβ gene segment. Additionally, we characterized the α-chain repertoire by targeting the highly dominant TRAV12-1 sequence. PCR products of interest were sequenced from the reverse primer (Fasteris SA). TRAV and TRBV segments were described according to the Lefranc nomenclature (29). Term of “dominant clonotype” refers to a nucleotide sequence found with a frequency >5% within a given patient.

Chromium release cytolytic assays were performed, as previously described (28). Briefly, ⁵¹Cr-labeled TAP^−/−-deficient T2 (HLA-A*0201⁺ Melan-A⁻) or C1R (HLA-A,B⁻) stably transfected with wild-type or D227K/T228A mutant versions of HLA-A2 were pulsed with serial dilutions (10⁻⁶ M to 10⁻¹³ M) of the native or analog Melan-A^MART-1_26–35 peptide with an E:T ratio of 10:1. The melanoma cell lines Me290 and NA8 were also used as targets in the absence or presence of the analog Melan-A^MART-1_26–35 peptide (1 μM) at an E:T ratio of 30:1 (23). An irrelevant Flu peptide (MA, 1 μM) and a no-peptide control were used as negative controls.

The percentage of specific lysis was calculated as 100 × (experimental − spontaneous release)/(total − spontaneous release). The EC₅₀ value was defined as the peptide concentration required for half of maximum target cell lysis. Nonkiller clones were defined as clones presenting a maximal lysis <25% and for which an EC₅₀ value could not be determined accurately. These nonkiller clones were not included in the statistical analyses.

Calcium mobilization assays were performed as previously described (30). Briefly, after loading with 2 μM Indo 1-AM (Sigma-Aldrich), baseline signal was recorded for 30 s before 1 μg/ml HLA-A2/Melan-A^MART-1_26–35 native/EAA-specific NTAmer was added and intracellular Ca²⁺ flux was assessed for 5 min under UV excitation and constant temperature (37°C) using a thermostat device on a LSR II SORP (BD Biosciences) flow cytometer.

Indo-1 (violet)/Indo-1 (blue) 405/425 nm emission ratio was analyzed by the FlowJo kinetics module software (v.9.7.6; Tree Star). Non-Ca²⁺ flux clones were defined as clones that displayed an Indo-1 emission ratio peak after 150 s and a mean fluorescence intensity <30,000 U. These non-Ca²⁺ flux clones were not included in the statistical analyses.

Dually labeled pMHC multimers built on NTA-Ni²⁺-His-tag interactions called NTAmers were used for dissociation kinetic measurements and synthetized by TCMetrix Sàrl, as described previously (15, 18). Stainings with 8 nM dually labeled EAA Melan-A^MART-1_26–35 (native)-specific NTAmers and data analysis were done, as previously described (18, 20). Briefly, staining was measured at 4°C using a thermostat device on a SORP-LSR II flow cytometer (BD Biosciences). Following 1 min of baseline acquisition, Imidazole was added (100 mM), and PE and Cy5 fluorescence were measured during the following 10 min. A schematic representation of the NTAmer technology is depicted in Supplemental Fig. 1C.

Data were processed using the kinetic module of the FlowJo kinetics module software (v.9.7.6; Tree Star), and corrected mean fluorescence intensity values were plotted and analyzed using the GraphPad Prism software (v.6; GraphPad).

Data were analyzed using GraphPad Prism (v.6; GraphPad) by F-test as well as nonparametric Mann–Whitney, Wilcoxon paired t test, Kruskal–Wallis, and Spearman’s correlations, as indicated throughout the manuscript. For all experiments, raw data can be provided upon request.

These studies were conducted in a laboratory that operates under Good Laboratory Practices principles. The laboratory participates in external ELISPOT and immune-monitoring proficiency panels. The study was performed using established laboratory protocols. These studies were performed using a qualified assay.

In previous publications in which Melan A–specific CD8 T cells were analyzed directly ex vivo, we have demonstrated an improved T cell functionality after vaccination with the low-affinity (native) peptide as compared with vaccination with the high-affinity (analog) peptide. These results were based on various assays including ex vivo single cell gene expression profiling (23, 24), ex vivo intracellular cytokine staining (22, 23), ex vivo IFN-γ ELISPOT (22, 24), and ⁵¹Cr-release cytotoxicity using T2-pulsed target (22, 24) or melanoma cell lines (23). Overall, these data on polyclonal Melan-A–specific CD8⁺ T cell populations provided convincing results demonstrating the superior antitumoral immune response generated by vaccination with the low-affinity (native) peptide formulation.

For this study, we generated a large library of T cell clones (n = 150) bearing well-defined TCR-αβ clonotypes isolated from melanoma patients vaccinated with either the low-affinity (native/EAA, n = 3) or the high-affinity (analog/ELA, n = 4) peptides, and derived from both the early and late differentiated subsets (defined as EM multimer⁺/CD8⁺/CD45RA⁻/CCR7⁻/CD28⁺ [EM28⁺] and CD28⁻ [EM28⁻], respectively). To validate this collection of tumor Ag-specific T cell clones with the previously gathered ex vivo data (22–24), we first assessed their functional efficiency in terms of target cell killing, expression of effector molecules, and Ca²⁺ mobilization capacity.

Extending on previous reports (23, 24), we observed that EM28⁺ and EM28⁻ T cell clones of low-affinity (native) peptide-vaccinated patients had a 10-fold increase in functional avidity compared with killing responses obtained from the corresponding subsets derived from high-affinity (analog) peptide-vaccinated patients (p = 0.002 and p < 0.001, respectively) (Fig. 1A, left panel). We defined functional avidity in relation with the EC₅₀ (50% maximal lysis effective peptide concentration) of T cell clones from the killing assays. Moreover, in contrast to analog ELA/EM28⁺ tumor-specific T cell clones showing a significant fraction of clones with poor cytolytic activity (defined as nonkillers), almost all native EAA/EM28⁺ clones were able to lyse peptide-pulsed target cells (Fig. 1A, right panel). Abrogating CD8 binding to pMHC by using target cells transfected with D227K/T228A HLA-A2 mutant led to a >10-fold decrease in target recognition capacity for both EAA and ELA clones (p < 0.001 and p = 0.006, respectively) (Fig. 1B, left panel) with an increased frequency of nonkiller ELA clones (Fig. 1B, right panel). Importantly, these functional avidity differences between tumor-specific CD8 T cell clones derived from the two cohorts of patients were discernable only when the low-affinity (native) peptide was used during the in vitro stimulation. Due to its 10-fold higher HLA-A2–binding affinity (25), the high-affinity (analog) peptide stimulation yielded comparable EC₅₀ values between both vaccination cohorts (Supplemental Fig. 2A, 2B, left panel). We also evaluated the tumor reactivity of the CD8 T cell clones for melanoma tumor cell lines. Against the HLA-A2⁺ Me290 cell line expressing the endogenous Melan-A Ag, the EAA clones in both EM28⁺ and EM28⁻ subsets had a slightly higher killing efficiency than their ELA counterparts, with again a large proportion of ELA-derived T cell clones that were deficient in cytotoxic function and were thus defined as nonkillers (Fig. 1C, left panel). The same assay using the HLA-A2⁺/Melan-A–deficient NA8 cell line demonstrated the Ag specificity of the selected clones, as most clones were unable to recognize and kill these target cells (Fig. 1D). By pulsing the Me290 or the NA8 cell lines with the analog ELA peptide, we again could observe a higher killing efficiency for the EAA-derived clones (Fig. 1C, 1D, right panels).

We then assessed whether tumor-specific CD8 T cells from low-affinity (native) peptide-vaccinated patients had qualitative biological differences promoting their increased killing capacity. Clones showed similar maximum Ca²⁺ fluxing potential (Fig. 1E, left panel) independently of the subset (EM28⁺ versus EM28⁻) or vaccination cohort (native versus analog). Nonetheless, we continued to find unresponsive tumor-specific T cell clones within the ELA cohort and a higher prevalence of clones with deficient Ca²⁺ flux within both ELA/EM28⁺ and ELA/EM28⁻-derived clones (Fig. 1E, right panel). By intracellular staining, we observed that the EAA/EM28⁺ clones expressed significantly more granzyme B and perforin than their ELA/EM28⁺ counterparts at resting state (Fig. 1F).

Together, the data obtained from this large collection of vaccine-induced, tumor-specific CD8 T cell clones extended on our recent ex vivo analyses (23, 24) by showing that the type of peptide vaccine (low or high affinity) qualitatively impacted the T cell cytotoxicity, the expression of effector molecules, to a lesser extent Ca²⁺ signaling, and with significant differences observed in the frequencies of nonfunctional clones (nonkiller or non-Ca²⁺ flux).

The TCR repertoire of the tumor-specific CD8 T cell clone library was, within the early differentiated EM28⁺ subset, predominantly composed of individual and nondominant clonotypes, which contrasted with the preferential selection of many codominant clonotypes in the differentiated EM28⁻ subset (Supplemental Fig. 3). In line with our previous report (23), we observed that the functional killing avidity, as defined by EC₅₀ values, was independent of clonotype frequency and instead was associated with the vaccination cohort (Fig. 2A). To exclude the potential bias due to the heterogeneous nature of the clonotype repertoire among T cell subsets, we further tested the functional avidity of EM28⁺ and EM28⁻ T cell clones bearing the identical TCR (same TRAV and TRBV sequences) (Fig. 2B, Supplemental Fig. 2C). The TRBV19-1 clones from patient LAU 618, vaccinated with the high-affinity (analog) peptide, showed decreased functional avidity with a higher EC₅₀ than the TRBV6c1 clones from patient LAU 1013 vaccinated with the low-affinity (native) peptide vaccination. Moreover, the EM28⁺ TRBV19c1 clones from LAU 618 had lower maximum lysis potential than their EM28⁻ counterparts (Fig. 2B, right panel), contrasting to the similar maximum lysis found for both subsets from LAU 1013 (Fig. 2B, left panel). Together, the functional heterogeneity present in our large clonal library yielded similar results to our previous ex vivo analyses based on the same cohorts of patients (22–24, 31), thereby indicating that these clones provide a powerful model to study the impact of TCR/CD8:pMHC avidity on vaccine-induced tumor-specific T cell responses in a clinically relevant setting.

To gain insight into the TCR/CD8-binding avidity for pMHC, we relied on a novel technology based on dual-color reversible multimers composed of multivalent NTAmer, which can be used to quantitatively assess TCR:pMHC dissociation (k_off) rates on living T cells (15, 18, 20). Following the addition of nontoxic low concentration of imidazole, the multimers decayed into Cy5-labeled pMHC monomers and PE-labeled streptavidin NTA₄ molecules within seconds (Fig. 3, Supplemental Fig. 1C). The Cy5 signal from the pMHC complexes then allowed for the accurate measurement of pMHC dissociation kinetics on live CD8 T cells directly by real-time flow cytometry and at monomeric resolution. Using the sensitive NTAmer technology on our tumor-specific CD8 T cell clone library, we were able to reliably quantify the dissociation kinetics (k_off) of TCR:pMHC complexes along a relatively wide range (t_1/2 from 0.22 to 974 s), which most probably correlates with a broad spectrum of TCR affinity given their biophysical interdependency (K_D = k_on/k_off) of naturally occurring physiological TCRs. Importantly, we used NTAmer loaded with the low-affinity (native) Melan-A peptide as they provided a more physiological assessment of the TCR:pMHC recognition efficacy (Fig. 1, Supplemental Fig. 2).

We observed that the type of peptide used for vaccination could impact the TCR/CD8:pMHC avidity of Ag-specific T cells. The EAA clones had significantly slower dissociation rates than the ELA clones (k_off of 0.041 s⁻¹ and 0.130 s⁻¹, respectively, p < 0.0001). We did not detect any significant difference between EM28⁺ and EM28⁻ clones when pooling all T cell clones generated from both vaccination cohorts (k_off of 0.092 s⁻¹ and 0.120 s⁻¹, respectively, p > 0.05) (Fig. 4A, left panel). We then characterized k_off rates according to subset and peptide used during the vaccination. For the early differentiated EM28⁺ clones, we observed a tendency toward slower dissociation rates within the EAA/EM28⁺ clones compared with ELA/EM28⁺ clones (k_off of 0.052 s⁻¹ and 0.111 s⁻¹, respectively, p = 0.100) (Fig. 4A, right panel). Importantly, these slower dissociation rates were more prevalent within the differentiated EAA/EM28⁻ clones than in their ELA/EM28⁻ counterparts (k_off of 0.020 s⁻¹ versus 0.155 s⁻¹, respectively, p < 0.0001). Similar differences were obtained when using the NTAmer t_1/2 values instead of the k_off rates (Supplemental Fig. 4).

By classifying the tumor-specific T cell clones according to their respective patient, we observed a higher variance in patients vaccinated with the high-affinity (analog) peptide (p < 0.0001, F-test), with the exception of patient LAU 444 (Fig. 4B). This particular patient was studied in detail by our group (32) and had a pre-existing and persistent clonotype (TRBV6c2, also named “ELGTASY”), which dominated the differentiated EM28⁻ subset and was boosted efficiently after vaccination. We further witnessed the robustness of the NTAmer technique through the analysis of tumor-specific T cell clones bearing the same TCR clonotype (identical TRAV and TRBV sequences), which displayed highly similar k_off rates (Fig. 4C). Together, these results indicated that the low-affinity (native) peptide vaccination promoted the selective enrichment of differentiated EM28⁻ tumor-specific T cells expressing TCRs with increased TCR/CD8:pMHC avidity for the pMHC complexes.

We next aimed to correlate the TCR:pMHC dissociation kinetic values with T cell functionality. When all clones were combined together, the NTAmer k_off rates correlated with Ca²⁺ flux peak (n = 73, R = 0.280, p = 0.016) (Fig. 5A, left panel). No correlation was observed when using the Ca²⁺ time-to-peak values. The relationship between Ca²⁺ flux peak and k_off rates illustrated significant differences between tumor-specific EAA and ELA T cell clones with a tendency toward slower TCR:pMHC k_off rates correlating with increased Ca²⁺ flux only within the ELA clones (Fig. 5B).

Furthermore, we observed a positive correlation between k_off rates and killing potency (EC₅₀) following low-affinity (native) peptide stimulation (R = 0.662, p < 0.0001) (Fig. 6A, left panel). Importantly, this correlation was lost when the high-affinity analog peptide was used during the killing assays, further emphasizing the usage of the low-affinity (native) Melan-A peptide as a more sensitive tool to monitor T cell function (Fig. 6A, right panel). Of note, nonkiller clones displayed a relatively diverse range of k_off rates, with a tendency for rapid off-rates. The classification of clones according to subset (EM28⁺ versus EM28⁻) and vaccination cohort (native versus analog) highlighted that the k_off rates/EC₅₀ and t_1/2 values/EC₅₀ correlations were maintained among all subsets, albeit at different strengths (Fig. 6B, Supplemental Fig. 4). The strongest correlation was observed within the late differentiated ELA/EM28⁻ clones (R = 0.801, p < 0.0001), with the EAA clones displaying weaker correlations, although statistically significant, mostly as a consequence of the restricted range of EC₅₀ values from the killing assays. Collectively, these results demonstrated a positive relationship between TCR:pMHC dissociation rates and functional avidity as monitored by Ca²⁺ flux and target cell killing. Furthermore, the type of peptide used for vaccination revealed differences in the impact of TCR/CD8:pMHC avidity on the functional potency as well as the quality of the late differentiated EM28⁻ tumor-specific T cells.

TCR affinity/avidity is an important determinant of thymic T cell selection (10) and T cell function (2). Reports have shown that stronger TCR:pMHC interactions generate superior effector functions in mice (19, 33–35) and humans (36–38). However, the impact of TCR/CD8:pMHC avidity on CD8 T cell differentiation and clonotype repertoire selection within the periphery remains poorly understood. In this report, we took advantage of a large library of Melan-A–specific CD8 T cell clones (n = 139) derived from melanoma patients who received vaccination with either low (native/EAA)- or high (analog/ELA)-affinity Melan-A^MART-1_26–35 decapeptide in combination with CpG and IFA (22). This clonal library, composed of EM28⁺ and EM28⁻ clones (defined as CD8⁺/CD45RA⁺/CCR7⁺/CD28⁺ or CD28⁻, respectively), served as a unique tool to study the influence of TCR:pMHC kinetics on T cell differentiation and T cell function in a well-defined, clinically relevant setting (Figs. 1, 2).

Until recently, the precise quantification of TCR:pMHC kinetics at the surface of live T cells was technically challenging. Different techniques have been developed for TCR:pMHC kinetic determination based on micropipetting and thermal fluctuation assays (39, 40). To our knowledge, the use of the novel NTAmer technology (15, 18) allowed, for the first time, the real-time quantification of dissociation along a broad dynamic range of TCR avidities (>3 logs), on live human CD8 T cells and with monomeric resolution (Fig. 3). Moreover, recently published data from our group demonstrate that TCR:pMHC dissociation (k_off) values derived using the NTAmer technology strongly correlate with k_off values obtained by SPR (20). Nauerth et al. (19) have elegantly described a Streptamer-based approach for the monomeric TCR:pMHC dissociation quantification using viral Ag. It must be noted that TCRs specific for viral Ags have a higher-affinity range than tumor-associated Ags, such as Melan-A (4). Due to their high stability and rapid reversibility (<5 s) upon addition of imidazole, NTAmers offer increased sensitivity for low-affinity TCRs (for instance with t_1/2 <60 s), which represent the large majority of naturally occurring TCRs specific for nonmutated tumor-associated Ags. Therefore, we chose the NTAmer technology as it eliminates the measurement biases conferred by the multivalent binding of multimers to the TCR (15, 41), allows a broad dynamic range of dissociation rates in the low-affinity spectrum, and provides high-throughput capacity with an easy-to-use experimental procedure not necessitating the generation of soluble TCRs.

Together with our unique CD8 T cell tumor Ag-specific clonal library, the NTAmer technology allowed us to accurately evaluate the impact of immunotherapy on TCR/CD8:pMHC avidity with regard to clonotype selection, differentiation, and T cell functions. By segregating the clones according to patient cohort (vaccination with native/EAA versus analog/ELA peptide) and T cell differentiation (EM28⁺ versus EM28⁻), we observed that the low-affinity (native) peptide vaccine promoted strong effector differentiation readily within the early differentiated EM28⁺ subset, thereby highlighting the impact of the peptide:MHC affinity during CD8 T cell priming and activation (23, 24). Moreover, our data further corroborate our previous results demonstrating that a large proportion of CD8 T cell clones derived from high-affinity (analog) vaccinated patients was unable to lyse T2-pulsed target cells (28, 42) and melanoma cell lines (22). To our knowledge, for the first time, we were able to quantify that vaccination with the low-affinity (native) peptide was associated with the presence of Ag-specific T cells bearing predominantly TCRs with high TCR/CD8:pMHC avidity (Fig. 4). Of note, although the TCR/CD8:pMHC avidity range was patient dependent, some clones derived from high-affinity (analog) peptide-vaccinated patients (LAU672 and LAU936; Fig. 4B) displayed TCR:pMHC dissociation kinetics associated with the high TCR/CD8:pMHC avidity of EAA clones. Our data support the notion that the high-affinity (analog) peptide vaccination can induce particular clones bearing high TCR/CD8:pMHC and functional avidity. However, our observations demonstrate that the low-affinity (native) peptide vaccination favors an enrichment/selection of Ag-specific clones bearing both high TCR/CD8:pMHC avidity and high functional competence. This was particularly evident in the differentiated EM28⁻ subset, thereby suggesting that repeated prime-boost vaccination with low-affinity peptide promoted the enrichment of high-avidity CD8 T cells during effector T cell differentiation.

Lower TCR affinities (43–45) and reduced functional avidities are typically seen in primary T cell–based immune responses compared with secondary responses. Moreover, the primary response contains a heterogeneous pool of TCR affinities, which further matures (46). Baumgartner and Malherbe (47) have found that vaccination with peptides forming pMHC complexes of low stability yielded high functional avidity CD4⁺ T cell responses, whereas pMHC of higher stability recruited lower-avidity clonotypes. Our data are in line with previously documented observations (33, 34, 48–50) that repeated vaccination with low concentration of a low-affinity peptide (e.g., native/EAA) promotes TCR/CD8:pMHC avidity maturation, which was absent with the high-affinity peptide (e.g., analog/ELA) cohort. We also observed that the dose of peptide could influence the quality of effector cells in vaccinated melanoma patients. Indeed, low peptide dose vaccination resulted in enhanced cytotoxicity and reduced CD8 dependence of Melan-A–specific CD8 T cells (51). Furthermore, a recent report in the RIP-OVA model suggests that high-affinity peptides could prolong the contact between T cell and APC, which favors asymmetric cell division of the daughter T cell proximal to the APC differentiating into an effector T cell (52). Conversely, CD8 T cell stimulation with low-affinity peptides results in symmetric division and reduces effector T cell differentiation. Moreover, Cole et al. (38) reported that tumor-specific T cell clones against Melan-A could have a strong preference for either the analog/ELA or the native/EAA peptide, impacting T cell function in response to either analog/ELA or native/EAA peptide stimulation. They further suggest that priming with a heteroclitic high-affinity peptide may generate tumor-specific CD8 T cells that would recognize the native Melan-A peptide suboptimally. These data together with ours (22) indicate that vaccination trials using heteroclitic analog peptides require careful re-evaluation with regard to the risk of priming T cells with imprecise Ag specificity. A clinical trial is currently under way (http://www.clinicaltrails.gov, NCT01308294) in which melanoma patients are primed with a low-affinity (native) peptide vaccine to induce high-quality Ag-specific CD8 T cells and then boosted with the high-affinity (analog) peptide to favor a maximal response (data not shown).

Finally, a major finding of this report was the demonstration of a robust correlation between TCR avidity and T cell responsiveness, in our case functional assays with Ca²⁺ flux (Fig. 5) and target cell killing (Fig. 6), using a large panel of tumor-specific CD8 T cell clones bearing naturally occurring TCRs. This observation has been reported by groups using either transgenically modified TCRs (37, 53–56) or pMHC variants or altered ligands (46, 48, 57, 58). In a viral setting using a microscopy and reversible Streptamer-based assay, slower k_off rates were associated with improved function and better in vivo protection (19). With a relatively large number of clones (n = 92) expressing naturally occurring TCRs and displaying a large spectrum of TCR/CD8:pMHC avidity and functionality values, we observed that tumor-specific vaccine-induced T cell clones with high TCR/CD8:pMHC avidities were also the best killers. A correlation, albeit weaker, was also observed between k_off rates and Ca²⁺ flux. Our data are therefore in accordance with the dissociation rate model in which slower k_off rates yield better functionality, and that different stimulation threshold can trigger different functional outputs (e.g., signaling, cytokines, killing) (3).

Altogether, our work offers clear evidence that TCR/CD8:pMHC avidity is a major determinant of T cell function. The use of the novel NTAmer technology permits real-time quantification of TCR dissociation kinetics directly with monomeric resolution at the surface of primary CD8 T cells in a convenient and accessible flow cytometry–based assay. Within the scope of cancer immunotherapy based on the adoptive cell transfer of tumor Ag-specific CD8 T cells, the quantification of T cell functions in standardized assays could easily be complemented with TCR kinetics assessment with the aim to determine which cells would offer the best clinical efficiency in patients. Furthermore, this assay may identify naturally occurring TCRs with high TCR/CD8:pMHC avidity, which could be used for T cell engineering.

We are grateful to the patients and blood donors for dedicated collaboration in this study, and Pfizer and Coley Pharmaceutical Group for providing CpG PF-3512676/7909. We thank B. Gupta, E. M. Iancu, D. Labbes, S. Winkler, J. Schmidt, I. Luescher, and P. Romero for essential collaboration and advice. We are also thankful for the excellent help from L. Cagnon, L. Leyvraz, N. Montandon, and M. van Overloop.

The authors have no financial conflicts of interest.