A T Cell Clone’s Avidity Is a Function of Its Activation State¹

A T cell’s fate and function revolve around “affinity.” The immature T cell dies by neglect (is negatively selected) in the thymus if it does not recognize, with a minimal affinity, the MHC molecule/self-peptides present in this organ (reviewed in Refs. 1 and 2). Also, the naive T cell’s survival depends on the continued minimal activation provided by the positively selecting MHC-peptide ligand (3, 4). In contrast, MHC-peptide recognition above a certain stimulatory threshold results in negative selection by activation-driven cell death, which occurs in the thymus or in the immune periphery for immature or mature T cells. In between these two cutoffs for what could be called the lower and upper activation thresholds, TCR engagement can induce the maturation of naive T cells into effector/memory T cells. Hereby the signal strength has been shown to codefine whether the T cell differentiates along the Th1 or Th2 pathway; that is, to codefine the long-term engagement of effector class (5, 6). The type of cytokine/effector function engaged by the committed Th1/Th2 memory cell can also depend on the signal strength (7, 8). Therefore, different signal strengths/thresholds may define whether autoreactive T cells become deleted, inactivated, remain ignorant, or are induced to mediate effector functions and, if so, what type of effector functions are engaged. In the context of autoimmunity, it remains largely unknown why T cells that evaded negative selection and persist as naive autoreactive T cells apparently ignorant of the autoantigen can attack the same autoantigen once they become primed by environmental stimulation (9). Do memory cells lower their activation threshold such that they become stimulated by a signal strength (the actual level of autoantigen presentation) that was nonstimulatory to the naive T cell?

In general, is T cell affinity a constant feature of a clone, or one that can change dependent on the T cell’s state of activation? Therefore, while affinity is at the heart of understanding T cell function in general, and that of autoreactive T cells in particular, and while this term has been widely used to describe different signal strength-dependent reactions of T cells, it has remained challenging to conceptualize what affinity actually means in terms of T cell function.

As the original chemical term, affinity reflects the strength (free energy) of the binding between a monovalent receptor and a monovalent ligand at the state of equilibrium. In the context of Ag recognition by T cells, affinity could be defined as the strength of binding between a given (soluble) TCR with a given MHC-peptide ligand (1). The TCR-MHC-peptide complex being trimolecular, the peptide’s affinity for MHC codefines the overall strength of the TCR-MHC interaction (10). Therefore, even monovalent TCR-MHC-peptide binding reflects a higher-order reaction that, involving two equilibrium parameters, might be better described by the term “effective affinity.” The effective affinity of the TCR-MHC-peptide binding can be measured, for instance, by using surface plasmon resonance. In such experiments, using TCR-peptide-MHC pairs with known thymic selection outcome, a correlation was shown between low-avidity binding and positive selection as well as between high-avidity binding and negative selection (11).

However, defining the effective affinity of the TCR-MHC-peptide binding does not suffice to describe TCR-triggered T cell functions. First, when the TCR-MHC-peptide ligations are considered during the T cell’s interaction with the APC, the actual dynamics of this process in the cell-cell contact area are fundamentally different from the same reaction in a solution (12). Accordingly, the multivalent binding of TCR-MHC-peptide pairs will define the overall avidity of the binding, being a function of both the densities of the TCRs on the T cell and of the MHC molecules bearing the specific peptide on the surface of the APC (13). Second, several additional cell surface molecule interactions contribute to the overall avidity of this interaction, including TCR dimerization (14), the lateral binding of the CD4/CD8 coreceptor molecules to the TCR-MHC-peptide complex (15), and attachments by general cell adhesion molecules constitutively expressed and induced during the cognate T cell-APC interaction. Third, as far as the functional consequences of these molecular interactions within the immunological synapse are concerned, the transduction of the T cell activating signal depends on serial triggering, whereby an optimal half-life of the TCR-MHC-peptide complex seems to be more important than the actual strength of the equilibrium binding (16, 17). A low off-rate associated with high avidity of TCR binding does not favor multiple TCR engagements and hence serial triggering by the same MHC-peptide ligand. Fourth, creating an additional level of complexity, the multiple coreceptor molecules involved in the T cell-APC interaction coaggregate with TCR molecules into supramolecular activation clusters (18) and, by contributing intracytoplasmic domains for the docking of kinases of tyrosine phosphorylation, they participate in the signaling events that lead to T cell activation (reviewed in Refs. 19 and 20). The cytoskeleton and plasma lipids actively contribute to the generation of the T cell-activating signal (21). Finally, the multiple levels of the signal transducing cascades are actively regulated. Because signal transduction events start immediately after cell-cell contact (22), neither of the above-mentioned processes actually reach equilibrium. Therefore, T cell avidity (describing the total binding energy generated by all the cell surface molecules involved in the T cell-APC interaction) cannot be directly translated into biochemical and biological reactions triggered in T cells; strictly speaking, the complexity of the signal transduction machinery and of the gene regulation processes involved preclude the description of T cell activation in terms of affinity/avidity.

However, T cells clearly have Ag dose-response characteristics that define their reactions and fates. Therefore, it might be more appropriate to define “T cell avidity” in functional terms, by measuring the Ag dose-dependent activation of effector function such as the induction of proliferation or cytokine production, and to refer to it as “functional avidity” (23). In such dose-response experiments, naive and memory cells of the same specificity have different sensitivity for Ag stimulation (24), and memory cells were shown to display variable “tunable” activation thresholds dependent on different costimulatory molecules expressed on different types of APC (25, 26, 27, 28). Moreover, evidence emerged that different responses of a T cell—such as killing, proliferation, and the production of different cytokines—is each achieved at different activation thresholds (7, 8). Therefore, different functions of the same T cell may have different functional avidities.

Using ELISPOT analysis, we provide here the first detailed single-cell analysis of a T cell clone’s secretory IFN-γ response, including its kinetics and the quantification of the cytokine produced per cell at various signal strengths and at various stages of the clone’s Ag history. As opposed to intracytoplasmic cytokine staining that requires the pharmacologic treatment of the cells with brefeldin A or monensin to disrupt the Golgi apparatus and that measures the synthesized cytokine inside the cell, the ELISPOT assay directly visualizes the cytokine actually secreted by pharmacologically untreated cells. In the ELISPOT assay, the released cytokine is being continuously captured on a membrane around the secreting cell over the assay’s duration. Therefore, the size of spots reflects an integrated amount of cytokine produced over the assay time period, allowing quantification of the per-cell cytokine output for the individual cells (29). New image analysis capabilities permitted us to measure the per-cell cytokine production in thousands of individual cells as a function of time and of signal strength. By obtaining high-resolution data on these parameters of the cytokine response of a T cell clone, we attempted to gain insight into T cell activation thresholds and functional avidity.

For this study, we selected a previously well-defined human CD4⁺ T cell clone that is specific for hemagglutinin (HA)⁴ peptide 307–319. This T cell clone was found to undergo cyclic changes (30). With increasing time since the last restimulation, the clone’s proliferative responsiveness increased. In the same report, such cyclic behavior of proliferative responsiveness was seen for all four randomly chosen clones studied, suggesting that this response pattern might be generalizable. Studies of the expression of 24 different cell surface molecules showed consistently increased expression of CD26, LFA, and very late Ag-1 on the “late memory cells” (21 days after the last Ag restimulation) vs “early memory cells” (5 days after Ag stimulation), while the expression of other molecules including TCR and CD4 were unchanged at these time points. We were interested in learning whether such cyclic changes in the proliferative potential of the T cell also reflect changes in the clone’s activation threshold for IFN-γ-production. If the activation threshold/functional avidity for engaging an effector function is not a constant feature of a T cell clone, but one that undergoes cyclic changes, then this finding might need to be accounted for in models of positive and negative selection and might contribute to a better understanding of mechanisms underlying T cell-mediated autoimmune diseases.

HA peptide 307–319 (PKYVKQNTLKLAT) was synthesized by the solid-phase method and was purified by reverse-phase HPLC. NBHAC25 is a human CD4⁺ T cell clone that specifically recognizes this HA peptide on DRB1*0101 as the restricting class II molecule; NBHAC25 expresses Vβ3 and an undefined Vα (30). NBHAC25 T cell clone was maintained in complete RPMI 1640 medium (94% RPMI 1640 + 5% human serum AB + 1% l-glutamine) that was supplemented with 10 ng/ml recombinant human IL-2. The medium was changed every 5 days. The T cell assays were done on the respective 5th day, before feeding with IL-2 medium. For the periodic restimulation with Ag (monthly for routine propagation of the clone), NBHAC25 cells (3 × 10⁵/ml) were cultured with HA peptide (1 μg/ml) and irradiated (10,000 rad) LG2 cells (EBV-transformed, DRB1*0101-positive B cell clone, at 6 × 10⁵/ml) in recombinant human IL-2 (2 ng/ml)-containing medium. LG2 cells were grown in complete RPMI 1640 medium, splitting the cultures every 5 days. Both the NBHAC25 and the LG2 clone, as well as the HA peptide, were a gift from Dr. Z. A Nagy (Hoffmann-La Roche, Nutley, NJ).

These assays were performed as described previously (31). Briefly, ImmunoSpot plates (Cellular Technology, Cleveland, OH) were coated overnight at 4°C with the IFN-γ-specific capture Ab (M700A; Endogen, Woburn, MA) at 2 μg/ml in PBS. Two types of ELISPOT plates were used. One plate had the regular format of 96 wells, 200 μl per well (ImmunoSpot M200). The membrane in the other plate is downsized 1:4 (ImmunoSpot P50), requiring 1/4 of the reagents. The plates were then blocked with BSA (10 g/L in PBS) for 1 h and washed three times with PBS. Cells were plated in complete RPMI 1640 medium. Irradiated LG2 cells (10,000 rad) were plated at 5 × 10⁴ cells per well in the ImmunoSpot M200 plates and at 1.25 × 10⁴ per well in the ImmunoSpot P50 plates. HA peptide was added as specified in the figures. After 30 min incubation (to permit equilibrium binding of the peptide to DRB1*0101; Ref. 30), the NBHAC25 cells were added at 300 cells per well in the ImmunoSpot M200 plates or at 100 cells per well in the ImmunoSpot P50 plates (the exception being Fig. 1 B, where the T cells were serially diluted). After 24 h, or as specified, the plates were washed and the biotinylated anti-IFN-γ-detection Ab (M701; Endogen; biotinylated in our laboratory) was added at 2 μg/ml in PBS/BSA/Tween (10 g/L BSA with 0.5% Tween). After an overnight incubation at 4°C, and washing three times with PBS/Tween, streptavidin-HRP (1/2000 dilution; Dako, Roskilde, Denmark) in PBS/BSA/Tween was added for 2 h at room temperature. After washing with PBS/Tween, followed by PBS, the spots were developed using a 3-amino-9-ethylcarbazole solution (Pierce, Rockford, IL). The AEC stock solution was prepared by dissolving 10 mg AEC in 1 ml N,N-dimethyl formamide (Fischer Scientific, Fair Lawn, NJ). For the actual development, 1 ml of this AEC stock solution was freshly diluted into 30 ml of 0.1 M sodium-acetate buffer (pH 5.0), filtered (0.45 μm), and mixed with 15 μl H₂O₂, 200 μl (ImmunoSpot M200), or 100 μl (ImmunoSpot P50) of which was plated per well. The plates were developed for 15 min, after which the reaction was stopped by rinsing with tap water. This enzymatic reaction yields 1- to 3-fold interassay variations in spot size generated; therefore, spot sizes were internally controlled for each experiment, and data points to be compared were generated in the same experiment. The plates were air-dried overnight before subjecting them to image analysis.

We used a Series 1 ImmunoSpot Image Analyzer (Cellular Technology) specifically designed for the ELISPOT assay. Digitized images were analyzed for the presence of areas in which color density exceeds background by an amount set on the basis of the comparison of experimental wells (containing T cells, APC, and Ag) and control wells (containing T cells and APC only). After background and noise subtraction, custom software is used to analyze spot morphology for circularity and density distribution to identify and separate touching and overlaying spots. Objects that meet these criteria are recognized as spots and counted. The measurement of spot size distribution is also a built-in function of the software; it is based on the array of spot sizes in a given well sorted according to distinct size categories.

Staining for flow cytometry analysis was performed as described (32). T cell clone NBHAC25 (5 × 10⁵/ml) was cultured with 10,000 rad irradiated LG2 cells (1 × 10⁶/ml) with or without HA peptide (0, 4 μg/ml) in complete RPMI 1640 medium. After 2 h cell culture, brefeldin A (BD PharMingen, San Diego, CA; 10 μg/10⁶ cells) was added, and another 2 h later the cells were fixed, permeabilized with 0.5% saponin in PBS/BSA, and stained with PE-labeled anti-IFN-γ (BD Biosciences, San Jose, CA) and PerCP labeled anti-CD3 Ab (BD Biosciences). Two-color flow cytometry was performed on FACScan (BD Biosciences). The isotype-matched control mAbs were obtained from BD Biosciences.

In general, this assay was performed as described previously (34). In particular, NBHAC25 and LG2 were cultured with HA peptide in 96-well regular round-bottom tissue culture plates (Nunc Immunoplate; Fisher Scientific, Pittsburgh, PA) in parallel to the ELISPOT assays. Supernatants were collected after 24 h. Recombinant human IFN-γ (BD PharMingen) was serially diluted as the standard. ELISA were done using mAb M700A (at 2 μg/ml) as the capture Ab and the biotinylated mAb M701 (1 μg/ml) as the detection Ab. Streptavidin-alkaline phosphatase (1/2000 dilution; Dako) followed by p-nitrophenylphosphate (1.6 mg/ml; Research Organics, Cleveland, OH) was used for the colorimetric reaction. The reaction product was measured at 405 nm.

Irradiated (10⁴ rad) LG2-B cells (4 × 10⁴/well) were preincubated with HA peptide (in different concentrations as specified in Results) in microtiter plates (U-bottom shape; Nunc) for 30 min. Subsequently 2 × 10⁴ NBHAC25 cells were added per well. After 72 h of culture, the proliferation rate was measured as [³H]thymidine incorporation for 12 h.

The CD4 T cell clone that we used is specific for HA peptide 307–319. The clone recognizes this peptide in an HLA-DRB1*0101-restricted manner: HA_307–319 binds to the HLA-DRB1*0101 molecule with an IC₅₀ of 0.33 μM. A cloned HLA-DRB1*0101-positive EBV-transformed B cell line was used to provide a uniform population of APC. A monolayer of such clonal APC was used in a 100-fold excess over the T cells to assure the synchronized and immediate Ag contact by the T cell. Because the peptide used does not require processing, it can be directly loaded onto the APC’s surface on the HLA molecule (30). Therefore, the amount of peptide added is proportional to the number of peptide-MHC complexes formed. By performing IFN-γ ELISPOT assays with this setup, the cytokine response of individual T cells was studied under highly defined conditions, with the MHC-peptide ligand density being the primary assay variable.

In the first set of experiments, we tested whether our IFN-γ ELISPOT assay—within the experimental setup used—would permit the direct visualization of cytokine production by NBHAC25 cells at the single-cell level, whether there were variations in the amount of cytokine produced by individual cells within the clone, and where the lower detection limit is for the per-cell-produced cytokine. We tested T cell clone NBHAC25 by intracytoplasmic IFN-γ staining, and by ELISPOT analysis, in parallel. Fig. 1,A shows that by intracytoplasmic staining and flow cytometry, ∼50% of the individual T cells were induced to produce IFN-γ at 0.1 μg/ml HA peptide. When the T cell clone was plated in serial dilutions with a fixed number of the same APC (clonal EBV-transformed B cells) and the peptide was kept at the same concentration (0.1 μg/ml), the number of spots detected in the ELISPOT assay was 59 ± 8% of the T cells plated (Fig. 1 B); this frequency was seen independent of the number of T cells plated per well, resulting in a linear function between T cells/well and peptide-induced spots/well that passes through the origin (r = 0.98). Therefore, at the peptide concentration tested, the number of T cells that were found to be IFN-γ positive by intracytoplasmic staining were also detected by ELISPOT, and the lower detection limit for IFN-γ expressed per cell was comparable for the two assays, moderately in favor of the ELISPOT approach.

Next, we kept constant both the number of T cells (300 cells/well) and the number of the EBV-B cells (5 × 10⁴ cells/well) and titrated the HA peptide (in this experiment, the T cells were tested 20 days after their last stimulation with Ag) (Fig. 2,A). At peptide concentrations exceeding 0.01 μg/ml, ∼300 spots were counted, closely matching the number of T cells plated per well (Fig. 2 A). Therefore, above this peptide concentration, all T cells plated were detected as IFN-γ-secreting cells. Jointly, these data show that the ELISPOT measurements are made at single-cell resolution.

Functional avidity can be defined as the concentration of peptide that leads to activation of the 50% maximal number of T cells (22, 23). Fig. 2,A shows such a dose-response curve, with the frequency of IFN-γ-producing NBHAC25 cells vs the corresponding peptide concentration. The curve obtained was closely approximated by a sigmoidal function in the log scale (Fig. 2 A) or a hyperbolic function on a linear peptide scale (data not shown), asymptotically reaching a plateau at high concentrations of the peptide. Because the minimally activating and maximally activating peptide concentration cannot be accurately defined in such curves, we used the 5 and 95% activation points to evaluate the respective Ag dose.

We studied the kinetics of IFN-γ spot formation at increasing Ag doses to test whether the low frequency of cytokine-producing cells detected by 24 h at low peptide concentrations reflects delayed activation of T cells or whether only a fraction of cells within the clonal population becomes activated at such conditions. At maximally stimulating peptide doses, spot formation started within 2 h after the Ag challenge, and the number of spots gradually increased until all of the plated T cells became activated, which occurred by ∼10 h after the initial Ag contact (Fig. 2 B). In the presence of submaximally stimulating peptide doses, spot formation started with a considerable delay (after up to 6 h), then increased to a plateau at 18 h, reaching numbers that corresponded to only a fraction of the T cells plated. The data suggest that, while the kinetics of the per-cell IFN-γ production is Ag dose dependent, only a fraction of the T cells within a clone become activated at a submaximally stimulating peptide dose.

In all experiments, a high level of heterogeneity of IFN-γ ELISPOT spot sizes was observed, closely mirroring the heterogeneity seen when this T cell clone was studied by intracytoplasmic staining by us (Fig. 1,A) or when different clones were studied by others (36). Fig. 3 provides an example for the range of spot sizes seen in a 24-h assay at the maximally stimulating HA peptide dose of 0.4 μg/ml vs the medium control well. The size of ELISPOTs is proportional to the total amount of protein produced by the cell during the assay’s duration (36). Therefore, the spot morphology should reflect the kinetics of IFN-γ production. Using computer-assisted image analysis, we studied the spot size distribution and found that on a logarithmic scale this size distribution was closely approximated by a normal distribution function. Similar to the mean channel number in FACS analysis, we used the mean spot size (the hatched line) to characterize this integrated parameter of the secretory response for the entire T cell population.

When the mean spot size was studied after 24 h of assay duration as the function of the signal strength (Fig. 4,A), it was found to increase with the peptide concentration and to reach a plateau (however, the spot size distribution was Gaussian for all peptide concentrations similar to Fig. 3, only the mean spot size differed, data not shown), confirming the notion that cytokine production is indeed Ag dose dependent; also, the IFN-γ concentration in the culture supernatants was found to be peptide dose dependent at all time points tested (data not shown). Thus, the net amount of cytokine produced by individual cells within 24 h was proportional to the signal strength. Accordingly, at 24 h, 0.4 μg/ml peptide not only activated all T cells to secrete cytokine, but also induced the maximal production rate. Spot size measurements at different time points (addressing the kinetics of cytokine production) vs the peptide concentration showed (Fig. 4,B) that at the highly stimulatory peptide concentrations (≥0.025 μg/ml), the T cells started to produce IFN-γ within 2 h, while following weak stimulation (0.0016 μg/ml), spots became detectable only after 6 h. At maximally stimulating doses, the T cells continued to produce at a constant rate (up to 72 h) that also appeared to be the maximal rate (in Fig. 4 B the linear function between mean spot size and time for doses was ≥0.4 μg/ml). However, below 0.4 μg/ml, the average spot sizes increased more slowly (suggesting a lower production rate) and reached a plateau within 24 h. Thus, at weakly stimulating peptide doses, cytokine production was not only slowed down, but also started with a delay and ceased earlier, involving both a lower production rate and a shorter production period.

When the kinetics of spot formation was studied in greater detail for different spot size categories at a submaximally stimulating peptide dose, spots in each size category were found to reach a plateau within 24 h (data not shown). In contrast, at the maximally stimulating peptide dose the transition from smaller to larger spot categories continued throughout the 72-h test period, also suggesting ongoing IFN-γ production in these cultures.

To validate the assumption that the spot size reflects the amount of cytokine produced per cell, and to reproduce these observations by an independent read out, we also measured IFN-γ by ELISA in the culture supernatants. A close to linear correlation was seen between the net amount of IFN-γ measured in the supernatants at different peptide concentrations and the mean spot sizes (Fig. 5), suggesting that the spot size and the total amount of cytokine produced by individual cells were directly proportional. Based on these data, we conclude that the net per-cell output of cytokine and the onset and the duration of IFN-γ production are each functions of the signal strength. The observed heterogeneity of spot sizes, even at maximally stimulating peptide doses, appears to reflect both different productivity of the individual T cells (resulting from their different activation thresholds) and the asynchronous onset of cytokine production. Therefore, considerable biologic complexity is concealed behind the close to perfect sigmoidal dose-response curve of a clonal population of T cells. Although such dose-response curves apparently do not fully characterize the clone’s affinity (as genetically defined by the TCR specificity expressed), they are suited to characterize the overall responsiveness of the cell population. In vivo, T cell populations, not individual cells, mediate effector functions, and the frequency of cytokine-producing cells along with the per-cell cytokine production rate will define the magnitude of effect. This overall Ag responsiveness seems to be well reflected in the ELISPOT dose-response curve. Is functional avidity a constant feature of a memory cell?

In Fig. 6,A, the dose-response curve is shown for NBHAC25 cells 5 days after Ag stimulation. Ninety-five percent of the 300 T cells plated produced IFN-γ spots at HA peptide concentrations of 0.22 μg/ml, reaching 100% at higher peptide doses. The 5% activation threshold was at 0.0032 μg/ml. Fifty percent maximal activation was calculated to be at 0.040 μg/ml peptide. When the T cell clone was kept in culture for an additional 20 days (without restimulating with Ag), and was retested on day 25, 100% of the T cells were eventually induced to produce cytokine, but the dose-response curve was shifted to the left. The 5, 50, and 95% activating peptide doses were calculated to be at 0.0001, 0.0028, and 0.016 μg/ml, respectively, which are ∼20-fold lower than when the clone was tested at the 5-day time point. The dose-response curves measured at the 10- and 15-day time points gave intermediate results with an incremental shift to the left with time (Fig. 6,A). To ascertain that these changes in peptide responsiveness are intrinsic to the T cells, we restimulated the NBHAC25 clone with Ag after the 25-day time point, and retested it at days 5 and 10. The dose-response curves seen overlapped with the measurements initially made at day 5 and 10, respectively; this time-dependent shift was reproduced in two additional experiments (data not shown). Therefore, the time-dependent changes in dose-response curves seem to be cyclic changes in the activation thresholds for IFN-γ production that are intrinsic to the clone. Parallel to these ELISPOT measurements, we also performed proliferation assays. Reproducing the previously published data on the proliferative potential of the clone (30), we found that higher peptide doses were required to induce a proliferative response at the 5-day time point than at the 25-day time point (open symbols in Fig. 6 A). At both time points shown (and at all the other time points tested), the induction of IFN-γ production required approximately one log lower concentration of peptide, suggesting different activation thresholds for the two T cell functions.

The significant increase in functional avidity of cloned T cells with time might suggest that T cells require time before they can reset their cytokine synthesis machinery. If this were the case, as opposed to a general decrease in the T cell’s signaling functions, one might expect decreased per cell cytokine production rates in the early memory cells. Aided by the ability to study independently the frequencies and per-cell cytokine output of T cells vs the signal strength, we addressed this question. Fig. 6 B shows the dose-response graphs of average spot sizes vs the peptide concentration for the NBHAC25 clone tested at 5, 15, and 25 days after the last stimulation with Ag; the testing of the T cells with the different Ag history was done in parallel. These data show that the day 5 cells require a stronger signal to reach the same average spot size than cells of day 25, and cells from day 15 gave intermediate results. While there was a parallel shift to the left on the peptide concentration axis, the cells reach the same plateau level, but at different peptide concentrations. Therefore, independent of the Ag history, the T cells had comparable maximal IFN-γ productivity. These data suggest that the cyclic changes in Ag responsiveness operate at the level of the signaling machinery and do not result from a decreased ability of these cells to perform these functions

Using a highly defined system, involving clonal APC and T cells, we provide here the direct visualization of the signal strength-dependent cytokine secretion of individual cloned T cells. Although the dose response of the T cell clone, at the cell population level, had the appearance of a classic affinity/functional avidity curve, the individual T cells within the clone seemed to display a broad range of activation thresholds. Thus, at submaximally stimulating peptide doses, only a fraction of all T cells became eventually activated; this fraction of cytokine-producing cells increased as a function of the signal strength asymptotically until the 100% mark was reached. The peptide dose that stimulated 5 vs 95% of the T cells spanned over a log, suggesting that the activation thresholds for individual T cells within the clone were orders of magnitudes apart, even though all of them expressed the same TCR specificity. This intraclonal heterogeneity in Ag responsiveness could result from the different numbers of TCRs and of the various accessory molecules expressed by the individual T cells. By flow cytometry, this clone, like clones in general, shows a close to Gaussian distribution for the cell surface density of each of these molecules, whereby their expression patterns are not correlated (e.g., TCR^high clones are not necessarily CD44^high or ICAM^high). Accordingly, individual T cells that happen to express a higher copy number of TCRs and of costimulatory molecules should conceivably have a lower activation threshold than those individual cells that express low copy numbers of these molecules. Therefore, the range of peptide concentration between the minimally and the maximally stimulating dose, as well as the slope of the dose-response curve, seems to reflect the level of microheterogeneity in intraclonal activation thresholds. It proved to be a constant feature of the clone: the slopes of the dose-response curves stayed constant, irrespective of the time point at which the clone was last restimulated by Ag (Fig. 6 A). The sigmoidal shape of the dose-response curves suggests that the activation thresholds for the individual cells within the clone show normal distribution (the Gaussian distribution being an integral of the sigmoidal curve). Despite this considerable interindividual variation in activation thresholds within a single clone, different T cell clones can have fundamentally different dose-response characteristics, with their own Gaussian spectrum of interindividual variations. Thus, in our previous work we found that myelin basic protein (MBP)-specific T cell clones isolated from MBP-expressing wild-type mice required a 10,000 times higher peptide dose for becoming stimulated than did clones isolated from MBP^−/− congenic mice (37). With the same immunodominant MBP peptide and the same APC involved, this fundamentally different Ag dose responsiveness could only be attributed to different affinities of the clones, i.e., the deletion of high-affinity clones in the MBP-expressing host. Therefore, the sigmoidal graphs of dose-dependent frequency measurements of cytokine-producing cells permit us to characterize clonal (or bulk) T cell populations with respect to average functional avidity and activation threshold distribution.

So far, it was not possible to directly visualize the signal strength-dependent production of cytokine by individual cells. Therefore, the very nature of the T cell activation threshold has remained unresolved. It has been suggested that individual T cells “count” the number of TCR ligations by MHC-peptide complexes on the APC, and that the T cell becomes activated as soon as a critical number of TCR ligations have occurred (a number that has been approximated in the order of several thousands) (16, 39). Accordingly, at low peptide concentrations (MHC-peptide ligand density), the individual T cell may simply require more time for “counting” before the threshold number of TCR ligations occurs. Therefore, is there a defined minimum number of MHC-peptide complexes on the APC above which TCR ligation does result in induction of IFN-γ, or do weak signals merely lead to delayed cytokine production?

Studying the spot size distributions vs peptide concentration by ELISPOT, we showed that per-cell cytokine output is a function of the signal strength, confirming previous observations (32, 35). However, our detailed analysis of the kinetics of spot formation suggests that not only the rate of production was affected, but also the time required for the beginning of secretion and duration of the production period itself. The plots for the average spot size as the function of the peptide concentration showed a similar sigmoidal curve, as did the dose-response frequency plots (Fig. 2,A vs 4A) with a Gaussian distribution of spot sizes at each concentration (Fig. 3), comparable to the activation threshold curve derived from the dose-response frequency plots. Most likely, this outcome reflects the normal distribution of “avidities” for individual T cells within the clone (if cell “A” within the clone has a lower avidity than cell “B,” “B” will produce more cytokine and a bigger spot at a given peptide concentration). However, we cannot rule out that the potency to produce (a certain maximal amount of) cytokine is also normally distributed (if cell “A” produces half the amount of cytokine at concentration “x” than does cell “B,” it might still produce half the amount of cytokine at concentration “y”, presuming both cells have the same avidity), also contributing to the Gaussian shape of the curves. In fact, this is likely, as at the maximally stimulating doses we still see a normal distribution with a similar range, though shifted to higher values.

This parallel shift of close to normal distribution toward higher cytokine output implies that the dose responses for the individual cells should follow (sigmoidal) curves similar to the average spot size graphs. Indeed, our analysis of spot size distributions clearly demonstrates that activation of individual T cells is not a quantum process with a simple on/off threshold, but follows a sigmoidal/asymptotic behavior. Therefore, it is impossible to determine an absolute cutoff representing the activation threshold for individual T cells. However, an arbitrary cutoff can be defined at a certain spot size, similar to the 5% frequency threshold.

Although the ELISPOT system can detect per-cell cytokine production over a wide range (>2 logs of spot sizes and densities, Fig. 3), we found no evidence for even minimal cytokine production below a certain range of Ag doses during an up to 72 h incubation period (Figs. 2 and 4). Therefore, although a discrete lower threshold for the T cell activation may not exist, it seems that when the T cell “counts” MHC-peptide ligations, these TCR engagements have to occur at a minimal frequency rate, to generate a stimulatory signal, consistent with the kinetic proofreading model (39). Confirming this notion, subthreshold triggering seems to lead to the formation of inactive phosphorylation intermediates of the signal-transducing TCR-ζ-chain (17).

It was striking that the Ag responsiveness of a T cell clone (at the population and single-cell level) was not a constant. T cells 5 days after the last Ag stimulation required ∼20-fold higher peptide concentrations to start secreting IFN-γ and to reach 50% cell activation than they did 25 days after the last Ag contact (Fig. 6). Although it has been previously described that the activation threshold for a given T cell can vary when it interacts with a different type of APC (that expresses different levels of MHC and costimulatory molecules), in our experimental setting the APCs were clonal and constant. This T cell clone’s different activation threshold and functional avidity for IFN-γ production was found to be cyclic (a function of the time that elapsed since the last restimulation by Ag). The cyclic Ag responsiveness of this clone, and of three others studied, was previously noted using proliferation assays (30), which we reproduced here (Fig. 6,A). It is unlikely that these changes in activation thresholds reflect different maturation stages of the clone. Rather, the changes are likely to result from activation-state-dependent variations in cell surface molecule expression. Thus, in a previous report (30), the proliferative behavior of this clone (and three others) was correlated with the cell surface expression of 24 molecules. The level of TCR and CD4 expression was comparable early and late after restimulation (which we confirmed, data not shown). Therefore, changes in the avidity of TCR-MHC-peptide binding are unlikely to account for the cyclic changes of the activation threshold. A consistent increase in the expression of CD26, LFA-1, and very late Ag-1 was noted on late time points after restimulation (30). Being molecules involved in costimulation, the up-regulation of these molecules might explain the lowered activation thresholds. Additional molecular correlates seem possible, and even likely. Because these changes did not affect the T cells’ cytokine expression machinery (the maximal per-cell cytokine production was unimpaired, Fig. 6 B), it is likely that this tuning of activation threshold occurs at the level of the signal transduction apparatus. It was suggested that the increased proliferative potential of the late vs early memory cell helps in maintaining the memory T cell pool when the Ag becomes limiting in the body (30). Although such changes in the proliferative responsiveness of foreign Ag-reactive T cells can be well reconciled with the regulation of clonal sizes and with the maintenance of an effective immune system, such tunable activation thresholds for T cell effector functions are more difficult to accommodate in the theoretical framework of T cell biology, particularly with respect to self/nonself discrimination.

It is at present thought that all the T cells that are positively selected in the thymus must have a minimal affinity for self-MHC and an associated self-peptide, and that this low level of TCR stimulation (by the positively selecting MHC-self-peptide complex) has to be maintained also in the immune periphery for the naive T cell to survive (3, 4). Therefore, all naive T cells in the immune periphery are weakly autoreactive. When naive T cells become activated, they are thought to become more responsive to Ag (24); that is, they seem to decrease their activation threshold. The reason for this might be in part that the memory T cell up-regulates accessory molecule expression. Indeed, it requires either immunization with an autoantigen or an infection with a cross-reactive microorganism to convert “ignorant” naive autoreactive T cells into autoaggressive effector cells; after becoming primed, these autoreactive memory cells start to become stimulated by the same APC whose constitutive level of autoantigen presentation was ignored by the naive T cell (9). The data presented here suggest that this lowering of the T cell activation threshold from that of the naive T cell to that of effector/memory T cell is not a discrete quantum leap, but one that continues after the priming event. Accordingly, many low-affinity autoreactive T cell clones that are engaged by a peripheral immunization or cross-reactive infection will further lower their activation threshold, while they make the transition to the late memory cell stage. Only after these ∼2 wk will they be stimulated by the level of autoantigen generally presented (Fig. 7).

This behavior of memory cells might explain why experimentally induced autoimmune diseases typically develop only 14–20 days after the immunization with the autoantigen, although the cytokine commitment and the full clonal size of the induced T cells is reached within 4–7 days (40). This might also shed light on the poorly understood relapsing course of most autoimmune diseases. Once the peripherally primed autoreactive T cells become stimulated by the endogenous autoantigen in the target organ (which induces them to secrete cytokines and to cause the first episode of the disease), they may raise their activation threshold and stop producing cytokine. Only after the renewed gradual lowering of their activation threshold can they again perceive the level of endogenous autoantigen presentation as stimulatory and can they reengage in the cytokine secretion that causes a relapse. Each new cycle could lead to a new relapse (Fig. 7). The time course of relapses, up to weeks apart, are consistent with the cyclic regulation of the T cell activation threshold, as seen in Fig. 6.

In summary, we show here that while dose-response curves do not fully characterize a T cell clone’s affinity as genetically defined by the TCR specificity expressed, they are suited to characterize the overall Ag responsiveness of the cell population. Also this Ag responsiveness is not solely defined by the TCR specificity expressed, but codefined by the clone’s Ag history. Cyclic changes in T cells’ Ag responsiveness might contribute to the maintenance of immune memory and might explain the intermittent course of T cell-mediated autoimmune diseases.

We thank Richard Trezza and Tameem Ansari for expert technical and Earl Sigmund for editorial assistance.