Affinity Thresholds for Naive CD8⁺ CTL Activation by Peptides and Engineered Influenza A Viruses

Virus-specific CD8⁺ CTLs recognize non–self-peptides complexed with self-MHC class (pMHC) I glycoproteins. This requirement for pMHCI recognition focuses CTL effectors onto virus-infected cells and is mediated via clonally expressed TCR αβ heterodimers. Naive CTL clonal expansion and differentiation depends on the integration of signals subsequent to TCR/pMHCI ligation, costimulation mediated via other interactions (CD28/B7) between cell surface molecules (1), and the availability of secreted helper factors and inflammatory cytokines (2). The homeostatic cytokines IL-7 and IL-15 (3) are thought to be particularly important, as is the upregulation of antiapoptotic molecules such as bcl-2 (4). Efficient recruitment into and survival through an immune response requires that responding T cells achieve a necessary level of “fitness” by crossing various signaling thresholds (3, 5). The primary measure of such molecular “fitness” is considered to be the strength of TCR/pMHCI binding (6), which determines the differential selection of naive precursors into an immune response (7–9), the spectrum of subsequent clonal expansion and extent of differentiation to effector/memory status (6, 9). What constitutes optimal, or even sufficient, TCR/pMHCI affinity/avidity is, however, a matter of debate, with different conclusions being reached from a variety of in vivo and in vitro approaches using altered peptide ligands (APLs) (4, 9–14). APLs of the agonist OVA_257–264 peptide (amino acid sequence, SIINFEKL [N]) that vary in potency for the OT-I TCR were used to define the affinity threshold for thymic positive and negative T cell selection (15). Interestingly, APLs that demonstrate the lowest affinity for the OT-I TCR were still able to support positive selection (16). Particularly informative (9) have been recent studies with engineered variants of Listeria monocytogenes that express some of the OVA_257–264 APLs to examine the role of TCR affinity in activating pathogen-specific T cell responses. In this case, presentation of both intermediate and high-affinity TCR ligands was able to initiate efficient CTL activation. However, high-affinity interactions were necessary to sustain continued effector CTL expansion during infection (9). What was not addressed is whether those APLs able to support T cell-positive selection were also capable of supporting T cell activation during infection. Therefore, further work is needed to gain a better understanding of how low-affinity interactions can impact pathogen-specific T cell responses.

The present analysis uses mutant peptides to determine how altering the K^b OVA_257–264 (N) epitope modifies the avidity of cognate TCR/pMHCI interactions. The normal N, EIINFEKL (E1), and SIIGFEKL (G4) variants were used in vitro as free peptide to stimulate cultured, transgenic OT-I T cells, with or without added costimulation. Peptide-pulsed, bone marrow-derived dendritic cells (BMDCs), lipopeptide vaccine preparations (17), and influenza A viruses engineered to express N, E1, or G4 in the influenza A neuraminidase gene segment (NA) stalk (18) were then used to prime B6 mice that had been given naive, congenic OT-I T cells.

The findings following in vitro and in vivo stimulation were very different and are considered in the context of TCR/pMHCI avidity/affinity measurements and pMHCI structural constraints. Overall, this analysis supports the idea that CD8⁺ T cell activation depends on TCR/pMHCI interactions reaching a minimal affinity threshold and that changes in pMHCI stability as a result of small alterations in peptide sequence represent a potential mechanism for viral escape. The experiments also indicate that, whereas in vitro studies of antigenicity provide important insights, they give only a partial reflection of what happens in an immune response.

C57BL/6J (B6, H-2^b, and Ly5.2⁺), congenic Ly5.1⁺OT-I (OT-I), and RAG1je-deficient (Rag^−/−) mice were bred and housed under specific pathogen-free conditions at the Department of Microbiology and Immunology, University of Melbourne. All experiments followed guidelines stipulated by the University of Melbourne Animal Ethics Experimentation Committee.

Pooled lymph nodes from naive OT-I mice were disrupted through a 70-mm sieve, and an aliquot was stained for CD8α expression to determine the percentage of OT-I CTL precursors. Unlabeled or CFSE-labeled OT-I cells (2 × 10⁴) were stimulated with 1 μM N, E1, G4, or Q4 peptide (Auspep) with or without 5 mg/ml anti-CD28 (clone 37.51; BioLegend) and 10 U/ml IL-2 (Apollo) or ex vivo-purified dendritic cells (DCs) activated with LPS (19) and pulsed with 1 μM N, E1, or G4 peptide. Stimulation was carried out in round-bottom 96-well plates for 5 h to 4 d in 200 ml RPMI 1640 medium supplemented with 2 mM l-glutamine, 1 mM sodium pyruvate, 100 mM nonessential amino acids, 5 mM HEPES buffer, 55 μM 2-ME, 100 U/ml penicillin, 100 mg/ml streptomycin, and 10% FCS. For analysis of cytokine production at 1–4 d, OT-I cells were restimulated with 1 ng/ml PMA and 1 mM ionomycin in the presence of 5 mg/ml brefeldin A (all Sigma-Aldrich) for 4 h prior to intracellular cytokine staining.

B6 mice were anesthetized by inhalation of vaporized methoxyfluorane and infected with 10⁴ PFU influenza A/Hong Kong/x31 virus (x31; H3N2) or vaccinated with 25 nmol lipopeptide intranasally (i.n.). For analysis of recall responses, mice were primed i.p. with 1.5 × 10⁷ PFU A/Puerto Rico/8/34 virus (PR8; H1N1) 6 wk prior to challenge with 10⁴ PFU x31 virus. Lipopeptide constructs (20) contained the CD4⁺ Th OT-II epitope (ISQAVHAAHAEINEAGR), conjugated to a Pam2Cys moiety and the OT-I ligands N, E1 or G4, termed lipopeptide containing N, E1, or G4, respectively.

Influenza A viruses were engineered to express G4 and E1 from the NA stalk by reverse genetics (21, 22), as described for N (18). Briefly, plasmids containing the mutant epitopes were generated by PCR using primers encoding the E1 or G4 variants and part of the x31 or PR8 virus NA sequence (primer sequence available on request) and NA-specific primers. PCR products were digested with BsmBI or BsaI and ligated into the pHW2000 vector. Recombinant viruses (x31-N, -E1, -G4, -Q4 or PR8-N, -E1, and -G4) were rescued following transfection of cocultured 293T and Madin–Darby canine kidney cells (21). Viruses were grown in 10-d embryonated hen’s eggs, and both lung and stock virus titers were determined as PFU using Madin–Darby canine kidney monolayers (23).

Spleen, bronchoalveolar lavage, and mediastinal lymph node (MLN) samples were recovered from infected B6 mice at the peak of the virus-specific CD8⁺ T cell response (days 9 and 10). The bronchoalveolar lavage populations were depleted of adherent cells by incubation on plastic for 1 h at 37°C, and the lymph nodes were disrupted between the grooves of two forceps, whereas the spleen samples were forced through a 70-μm sieve, and the erythrocytes lysed prior to further analysis.

For analysis of surface phenotype, cells (1–2 × 10⁶) were stained with anti-CD8a (clone 53-5.7) with or without anti-CD45.1 (clone A20, for detection of OT-I cells), anti-CD62L (clone MEL-14), or anti-CD44 (clone IM7) Abs (all BD Pharmingen or BioLegend) in PBS/0.1% BSA/0.02% sodium azide. For analysis of tetramer binding, samples were stained with the K^bN, K^bG4, or K^bQ4 tetramers (monomers obtained from ImmunoID) for 1 h at room temperature prior to Ab staining. Determination of tetramer dissociation was conducted as previously described (24), using the anti–H-2K^b/D^b Ab (clone 28-6-6; BD Pharmingen) to absorb dissociated tetramer. Data were collected on FACSCalibur, LSRII, or FACSCanto II (all BD Biosciences) and analyzed with FlowJo software (Tree Star).

Cells (1–2 × 10⁶) obtained from infected mice were cultured in 96-well round-bottom plates for 5 h at 37°C in RPMI 1640 medium supplemented as above in the presence of 1 μM peptide (N, E1, G4, or Q4), 1/1000 dilution of GolgiPlug (BD Biosciences), and 10 U/ml IL-2. Samples were then stained with anti-CD8α and anti-CD45.1 (for detection of OT-I cells), fixed, and permeabilized with Cytofix/Cytoperm kit (BD Biosciences) and stained with anti–IFN-γ (clone XMG1.2), anti–TNF-α (clone MP6-XT22), and anti–IL-2 (clone JES5-5H4) Abs (all Abs BD Pharmingen or BioLegend) to detect cytokine production. Samples were collected and analyzed as above. For measurement of affinity by peptide titration, cells were incubated with decreasing doses of peptide and analyzed as the %IFN-γ⁺ cells relative to that of 1 μM peptide.

Pooled lymph nodes from OT-I mice (10⁷ cells/ml in PBS/0.1% BSA) were labeled with 5 mM CFSE (Sigma-Aldrich) for 10 min at 37°C and washed in media containing 10% FCS. CFSE-labeled cells were then adjusted to the desired concentration for in vitro stimulation or adoptive transfer to B6 recipients via tail vein inoculation. B6 mice that had received OT-I cells were infected 1 d later with x31-N, x31-E1, x31-G4, or x31-Q4.

Naive K^bG4-specific T cells were depleted from B6 spleen and lymph nodes using the method developed by Moon et al. (25) and adapted by La Gruta et al. (7) for identifying naive, epitope-specific, CD8⁺ T cells. Briefly, pooled spleen and lymph node suspensions were stained with 1 μl K^bG4 tetramer-PE for 1 h in 0.1 ml spent 2.4G2 supernatant/1% normal mouse serum/1% normal rat serum and 0.1 ml PBS/2 mM EDTA/0.5% BSA at room temperature. Cells were washed and stained with anti–PE-conjugated magnetic microbeads (Miltenyi Biotec) for 20 min at 4°C and passed over a magnetized LS column (Miltenyi Biotec). Unbound (K^bG4 tetramer-negative) cells were collected and adoptively transferred to Rag^−/− mice. Approximately one B6 mouse was used to reconstitute two Rag^−/− mice, which were then left for 4wk prior to the adoptive transfer of OT-I cells and subsequent infection.

Peptide-pulsed B6 BMDCs were used for direct, in vivo stimulation with the variant peptides. The BMDCs were harvested from naive B6 mice and expanded in vitro in complete DMEM supplemented with 20 ng/ml GM-CSF (a gift from Prof. L. Brown, University of Melbourne) at 2 × 10⁵ BMDC/ml for 6 d at 37°C in 10% CO₂. They were then activated with 5 μg/ml LPS (a gift from Prof. L. Brown), peptide pulsed, and transferred (3–4 × 10⁵ in 200 μl) to B6 recipients that had also been injected i.v. with 10⁶ naive OT-I cells. BMDC activation was assessed by flow cytometry to demonstrate increased IA^b and CD86 expression (data not shown).

Stabilization of K^b on RMA-S cells was carried out as described previously (26). Briefly, after overnight incubation at 27°C, 1 × 10⁵ RMA-S cells were cultured with graded concentrations of OVA APLs for 1 h and then transferred to 37°C for 2 h. H2-K^b cell surface expression was detected using the K^b-specific mAb, Y3, conjugated to FITC. Cells were analyzed using a FACSCanto II (BD Biosciences), and the mean fluorescence intensity was determined using FlowJo software (Tree Star). The RMA-S cells were washed three times in PBS to remove excess peptide and used to stimulate day 10 effector OT-I CTL, generated after adoptive transfer and subsequent infection as described above. After 5 h, samples were stained with anti-CD8α and anti-CD45.1 (for detection of OT-I cells), fixed, and permeabilized with Cytofix/Cytoperm kit (BD Biosciences) and stained with anti–IFN-γ (clone XMG1.2; BD Pharmingen) to detect intracellular cytokine production. Samples were analyzed by flow cytometry as described above.

Statistical analyses were conducted using the unpaired two-tailed Student t test, with significance denoted as *p < 0.05, **p < 0.01, and ***p < 0.001.

The H-2K^b and β₂-microglobulin molecules were expressed in Escherichia coli as inclusion bodies, refolded with each of the N, E1, and G4 peptides and then purified (27). The H-2K^bN complex crystals were obtained (2 mg/ml at 20°C by the hanging-drop vapor diffusion technique. Crystals were grown in a reservoir containing 0.2 M sodium cacodylate at pH 6.5, 15% polyethylene glycol 8000 (w/v), and 0.2 M calcium acetate. The variant H-2K^b complexes were crystallized in variations of this condition, using 0.1 M sodium malonate at pH 6 as the buffer. All the crystals belong to space group P2₁ and the unit cell dimensions were consistent with two molecules per asymmetric unit. The crystals were flash frozen to 100 K before data collection in-house with a Rigaku RU-200 rotating-anode X-ray generator or at the Australian synchrotron using the MX-1 beam line. The data were processed and scaled with the XDS program (28). The crystal structures were solved using the molecular replacement method using Phaser (29) from the CCP4 suite of programs (30). The search probe used to solve the structure was the structure of mouse MHC class I H-2K^b [1VAC (31)] minus the peptide. The progress of refinement was monitored by the R_free value with neither a σ nor a low-resolution cutoff being applied to the data. The protocol used includes several cycles of refinement with REFMAC (30), followed by manual model rebuilding with coot program (32). Translation, liberation, and screw–rotation displacement refinement was used to model anisotropic displacements of defined domains during the refinement process. The coordinates of the H-2K^b in complex with N, E1, and G4 have been deposited in the Protein Data Bank (http://www.pdb.org/pdb/home/home.do) under accession numbers 3P9L, 3PAB, and 3P9M, respectively, and will be released for public access at the time of publication.

Circular Dichroism Spectra were measured on a Jasco 815 spectropolarimeter using a thermostatically controlled cuvette. Far-UV spectra were collected from 190 to 250 nm. The UV minimum was determined as 218 nm for H2K^bN, 220 nm for H2K^bG4, and 219nm for H2K^bE1. The measurements for the thermal melting experiments were made at the minimum for each peptide–MHC complex at intervals of 0.1°C at a rate of 1°C/min from 20 to 70°C. The Jasco Spectra Manager software was used to view and smooth the traces, and then, the GraphPad Prism software was used to plot temperature versus percent unfolded. The midpoint of thermal denaturation (Tm) for each protein was determined as the point at which 50% unfolding was achieved. Measurements were done in duplicate at two concentrations (H-2K^bN, 5 and 2.5 mM; H-2K^bG4, 10 and 5 mM; and H-2K^bE1, 6 and 3 mM) in a solution of 10 mM Tris (pH 8) and 150 mM NaCl.

Peptide variants of the wild-type (wt) OVA peptide N, including E1 and SIIGFEKL G4, provide a convenient model for probing the nature of the TCR/pMHCI interaction (33). Although the affinity of the K^bN-specific OT-I TCR for these H2K^b + mutant peptide epitopes falls within the general range of TCR/pMHCI interactions (K_d values: 20–22 μM for E1, 10 μM for G4, 5–6 μM for N (34–37)), such variant pMHCI ligands induce a spectrum of response profiles when used in vitro to activate OT-I cells. This provides a model for assessing how hierarchies of TCR/pMHCI affinity translate to diverse functional outcomes following different modes of T cell activation. For example, although K^bE1 has been described as a weak agonist (16, 35) capable of triggering limited lytic activity in OT-I T cells (16, 38), K^bG4 induces poor activation and survival (11, 36).

The consequences of costimulation for the phenotypic and functional changes induced by in vitro culture of OT-I T cells with 1 μM of the N, E1, or G4 peptide ligands were analyzed over a 4-d interval following the addition of exogenous IL-2 and/or anti-CD28 Ab. As expected, activation of naive OT-I T cells with cognate N peptide resulted in extensive proliferation (Fig. 1A, top panel) together with the rapid upregulation of CD44 (Fig. 1B, top panel) and loss of CD62L (Fig. 1C, top panel). Interestingly, stimulation of OT-I T cells with concurrent CD28 cross-linking appeared to limit the extent of expansion (Fig. 1A, Supplemental Fig. 1). In response to G4 stimulation, almost all cells became CD44^hi within 48 h, regardless of the secondary signals (Fig. 1B, middle panel), although significant accumulation was observed only if IL-2 was present in the culture (Fig. 1A, middle panel). Irrespective of the signals provided by IL-2 and CD28 ligation, stimulation with the low-affinity E1 peptide failed to induce significant OT-I proliferation, with delayed CD44 upregulation and little change in the CD62L phenotype (Fig. 1B, 1C, bottom panels). Therefore, qualitative changes in cell surface phenotype are associated with the quality of TCR/costimulatory signals received during initial T cell activation.

We next analyzed how these variant peptides influence the acquisition of cytokine production. Naive OT-I T cells were stimulated for 5 h in vitro in the presence of brefeldin A, and then, profiles of IFN-γ, TNF-α, and IL-2 production were determined by the flow cytometric intracellular cytokine staining (ICS) assay (Fig. 2A–C). Much to our surprise, we found that naive OT-I T cells stained strongly for cytoplasmic TNF-α protein within 5 h of stimulation with the cognate N peptide (Fig. 2A, top panel). This suggests that, at least from the aspect of TNF-α production, naive T lymphocytes are “poised” for immediate effector function prior to division (39). Furthermore, this very rapid cytokine induction depends on a high-affinity TCR/pMHCI interaction, as <4% of cells showed evidence of 5-h TNF-α production after stimulation with the G4 peptide (Fig. 2A, middle panel), whereas E1 induced no detectable early response (Fig. 2A, bottom panel).

How persistent is this profile of very early TNF-α production? From the 24-h time point in Fig. 2, naive OT-I cells that had been activated in vitro as described above were restimulated with PMA plus ionomyocin for 5 h prior to reading the cytokine staining profile. This additional step bypassed any effect of TCR downmodulation following the initial peptide stimulation. Strikingly, there was a rapid continuing drop in TNF-α production from 48 h in the N-stimulated OT-I cultures (Fig. 2A, top panel). The downmodulation effect was not a consequence of selective cell death as loss of TNF-α production correlated with increased proliferation (Fig. 1A, top panel). Interestingly, sustained TNF-α production was observed in the presence of costimulation (Fig. 2A, top panel), and there was evidence of IFN-γ synthesis within 24 h of exposure to the N peptide, with a subsequent, gradual increase over the culture period that was dependent on CD28 signaling (Fig. 2B, top panel). Conversely, IFN-γ expression in the G4- and E1-stimulated cultures was only observed in the presence of IL-2 (Fig. 2B, middle and bottom panels, respectively) with there being little dependence on CD28 when it came either IFN-γ production or maintaining T cell numbers (Fig. 1A, middle and lower panels). Overall, these results suggest that acquisition of effector phenotype is differentially regulated depending on the varying quality of the TCR/pMHC interaction that, in turn, establishes differential requirements for costimulatory signals.

To examine the impact of TCR/pMHCI avidity on the coexpression of IFN-γ and TNF-α upon activation, naive OT-I cells were stimulated with either the N or G4 peptides in the presence of anti-CD28, and the extent of coexpression of IFN-γ and TNF-α was determined by ICS (Fig. 2C, 2D). As expected (from Fig. 1A), the naive OT-I CTLs were only making TNF-α at 5 h after N or G4 stimulation (Fig. 2C, 2D), whereas IFN-γ production was observed by 24 h in a subset of the TNF-α producers for both the N and G4 cultures (Fig. 2C, 2D). This acquisition of IFN-γ was, however, sustained only in the N-stimulated population and soon fell off for those activated by the G4 peptide (Fig. 2C, 2D). The apparent emergence of TNF-α–only producers at days 2–4 within the G4-stimulated cultures likely reflects the survival of undivided naive OT-I (Supplemental Fig. 1, middle panel). It thus seems that the acquisition and maintenance of cytokine polyfunctionality requires an optimal, high-avidity TCR/pMHCI interaction.

The in vitro analysis shown in Figs. 1 and 2 established that low-affinity pMHCI ligands induce the activation of OT-I cells, although less optimally than the wt peptide. Are, however, these mutant epitopes immunogenic in normal mice? As described for N (18), recombinant influenza A viruses were engineered to express either the E1 or the G4 variant within the viral NA. Naive B6 mice were then infected separately with 10⁴ PFU of the x31-N, x31-E1, or x31-G4 viruses, and lung virus titers were measured (days 3, 6, and 9) by plaque assay (Fig. 3A). Both x31-E1 and x31-G4 had grown to a higher titer than x31-N by day 3 (p < 0.01). Within an additional 3 d, however, the lung titers for x31-N and x31-E1 were comparable and greater than those for x31-G4 (p < 0.01). All viruses were cleared by day 9 (Fig. 3A).

The K^bN-, K^bE1-, and K^bG4-specific CD8⁺ CTL populations were measured by tetramer staining on day 10 postinfection. All viruses induced tetramer⁺CD8⁺ CTL populations that were readily detected in the spleen (Fig. 3B), demonstrating that there is indeed a TCR repertoire specific for each of these variant epitopes. Furthermore, K^bG4 tetramer⁺ CTL expansion induced by infection with x31-G4 was significantly greater in magnitude than the endogenous K^bN-specific response (Fig. 3C; p < 0.05), possibly contributing to the faster rate of virus clearance (Fig. 3A).

Despite these quantitative differences, the K^bN-, K^bE1-, and K^bG4-specific CTL sets were qualitatively similar in terms of multiple cytokine production profiles (Fig. 3D, 3E). In addition, all three CD8⁺ CTL populations showed a similar capacity to produce IFN-γ in response to decreasing doses of the homologous peptide (Fig. 3F), suggesting that the spectrum of “functional avidity” was broadly comparable for these different TCR/pMHCI interactions.

Graded numbers of OT-I cells were adoptively transferred into naive B6 recipients, which were then infected separately with 10⁴ PFU of the x31-N, x31-E1, or x31-G4 viruses. The various OT-I populations were then analyzed on day 10 to determine the extent of in vivo expansion and functional activation (Fig. 4). As expected, x31-N induced substantial cell dose-related proliferation of the OT-I sets, but somewhat surprisingly after the findings from the in vitro experiments (Figs. 1, 2) and from the endogenous response profiles (Fig. 3), exposure to x31-E1 or x31-G4 did not cause any significant increase in OT-I T cell numbers (Fig. 4A). This was further confirmed from the CFSE dilution profiles assessed for labeled OT-I CTLs at day 3 or 10 postinfection (Fig. 4B). Furthermore, the lack of obvious x31-E1 or x31-G4 virus-induced OT-I expansion correlated with the maintenance of CD62L expression (Fig. 4C) and the absence of any increase in IFN-γ production following ex vivo restimulation with the wt N peptide (Fig. 4D). Thus, we see completely different in vitro (peptide stimulation) and in vivo (virus infection) proliferation and functional activation profiles for OT-I T cells exposed to these mutant epitopes, although the control responses to the wt N peptide were equivalent for both situations (Figs. 1, 2, 4).

One possibility for the absence of any immediate recall of IFN-γ production (Fig. 4D) for the OT-I sets recovered from the x31-E1– and x31-G4–primed mice is that the T cells may have been rendered anergic as a consequence of suboptimal activation with the variant peptides and were thus refractory to secondary stimulation (40). To test this in vivo, B6 mice that had been given OT-I T cells were primed i.p. with the PR8-N, PR8-E1, or PR8-G4 virus, then challenged i.n. 6 wk later with the serologically different (H3N2 versus H1N1) x31-N virus (Supplemental Fig. 2) or, following PR8-N, with the x31-E1 or x31-G4 variants (Supplemental Fig. 3). Respiratory exposure to x31-N induced a secondary OT-I response in the PR8-N–immune mice and also triggered what looked to be a primary OT-I response (Fig. 3) in those that were first exposed to PR8-E1 and PR8-E4 (Supplemental Fig. 2). Reversing the prime/challenge protocol indicated that x31-G4 induced some activation of PR8-N–primed memory T cells, although the effect was minimal (Supplemental Fig. 3). In summary, although the mutant peptides can stimulate an OT-I response in vitro (Figs. 1, 2), there are no indications that in vivo exposure induces either anergy or promotes an effective cross-reactive recall response.

Considering that the nonresponsiveness to G4 in vivo is not due to the nature of infection, we next wondered whether the sizeable endogenous response to K^bG4 impedes OT-I cells responding to G4, because CD8⁺ T cell responses to the same epitope can impede one another, a phenomenon known as immunodomination (41, 42). To examine this possibility, we reconstituted RAG^−/− mice that lack mature T cells, with either whole B6 lymphocytes preparations or with lymphocyte preparations from which the K^bG4-specific cells had been depleted (6, 25). Mice were reconstituted for 4 wk prior to adoptive transfer of OT-I cells and subsequent infection with x31-N or -G4 i.n. (Fig. 5A). Following x31-G4 infection, a robust K^bG4-specific response was generated in mice reconstituted with whole lymphocytes, whereas mice that received depleted lymphocytes showed a significantly diminished K^bG4-specific response (Fig. 5B). Depletion of K^bG4-specific CD8⁺ T cells did not result in activation of naive OT-I cells in response to x31-G4, because OT-I cells did not accumulate (Fig. 5C) and did not undergo greater division than naive controls (Fig. 5D). Thus, the endogenous K^bG4-specific CD8⁺ T cell response does not impact upon the OT-I response to G4, and the inability of OT-I cells to respond to G4 is likely to be T cell intrinsic.

The difference between the findings for in vitro (Figs. 1, 2) and in vivo (Figs. 3, 4, Supplemental Figs. 2, 3) activation suggests that the levels of pMHCI epitope encountered by T cells exposed to peptide-pulsed versus in vivo-infected APCs may be very different. Naive B6 mice were thus primed with peptide-pulsed, LPS-activated BMDCs (Fig. 6A, 6B), and the K^bN- and K^bG4-specific responses were measured by tetramer staining. Although vaccination with G4-peptide–pulsed BMDCs induced an endogenous K^bG4-specific response (Fig. 6A), it failed to trigger naive OT-I cells (Fig. 6B, 6C). This is in contrast the OT-I response in mice that received activated BMDCs pulsed with the wt N peptide (Fig. 6B, 6C). This is not due to an inability of OT-I cells to respond to G4 presented by DCs, because N- and G4-pulsed LPS-activated ex vivo isolated DCs were able to induce proliferation by naive OT-I cells in vitro (Fig. 6D).

Priming naive mice with peptide bound (20) to a Pam2Cys lipid moiety (lipopeptide) activates DCs to induce potent wt CTL response (17, 43, 44) via TLR-2 ligation (17). Naive B6 mice were given OT-I cells, then vaccinated with then N-, E1-, or G4-lipopeptide constructs (Fig. 6E–H). As expected, the N-lipopeptide induced strong proliferation of the OT-I CTL (Fig. 6E). Interestingly, administration of both the E1- and G4-lipopeptides was also associated with limited division of the OT-I CTL sets within the draining lymph node (Fig. 6E, 6F) and caused some decrease (at least for E1) in cell surface CD62L expression (Fig. 6H), although this level of activation and proliferation did not lead to increased localization to the lung (Fig. 6G). Still, the lipopeptide formulations were more effective than the peptide-pulsed BMDCs at replicating the in vitro peptide cross-stimulation profile (Figs. 1, 2) in the in vivo situation.

Infection with recombinant L. monocytogenes expressing the SIIQFEKL (Q4) mutant activates naive OT-I T cells in vivo (9) to levels greater than those seen for the E1 and G4 peptides used in this study (Figs. 3–6), although to a lesser extent than that observed for the wt N. Would we see the same thing if the Q4 response is driven via infection with a modified influenza A virus? Mice that received adoptively transferred, naive OT-I CD8⁺ T cells (Fig 3) were thus infected i.n. with the engineered x31-Q4 virus. As expected from the L. monocytogenes study, x31-Q4 infection does activate OT-I cells, with the extent of early (to day 3) CTL division in the MLN being remarkably similar for mice exposed to x31-N or x31-Q4 (Fig. 7A). However, as observed previously by Zhen and colleagues following bacterial challenge (13), the extent of clonal expansion in the longer term was clearly diminished, with the magnitude of the OT-I response in the spleen on day 7 being significantly lower for the mice given x31-Q4 rather than x31-N (Fig. 7B; p < 0.01). Also, although no difference was observed for the cytokine production profile following high-dose N peptide restimulation in vitro (Fig. 7C), significantly more of the x31-Q4–primed OT-I cells remained CD62L^hi (Fig 7D; p < 0.01). A single experiment (data not shown) confirmed the observation made with the Listeria model that the OT-I T cells exit the lymph node more rapidly in the x31-Q4–primed mice, leading do an overall decrease in the extent of clonal expansion.

Do these response profiles (Figs. 3–7) correlate with the spectrum of “functional avidity” measured by IFN-γ production for in vivo-primed (day 10 x31-N) OT-I cells following in vitro stimulation with graded doses of the N, G4, or Q4 peptides (Fig. 7E)? Such effector OT-I CTLs showed 50% maximal IFN-γ responses at the 10⁻¹⁰, 10⁻⁹, and 10⁻⁷ M peptide concentration for the N, Q4, and G4 peptides, respectively (Fig. 7E). Stimulation of OT-I CTL with G4 had not yet reached saturation (Fig. 7E). Thus, the 50% maximal IFN-γ response at 10⁻⁷ M for the G4 peptide is likely underestimated. Furthermore, the spectrum of in vitro persistence for OT-I cells stimulated with graded doses of the N, Q4, and G4 peptides shows an equivalent hierarchy (Fig. 7F). Naive OT-I T cells failed to survive for 3 d following exposure to 10⁻⁷ M G4, whereas comparable Q4- and N-peptide–stimulated CTLs remained viable after stimulation with much lower doses (10⁻⁸ and 10⁻⁹ M, respectively). This in vitro evidence that the avidity of the TCR/pMHCI interaction for G4 falls below the minimal level required for OT-I CTL survival presumably explains the differential response profiles (Figs. 3–7) for G4 and Q4 following in vivo stimulation with these engineered influenza A viruses.

The crystal structures of the K^bN and K^bG4 complexes were solved to 2Å resolution, whereas the K^bE1 complex was solved to 2.4Å resolution with the final refinement statistics summarized in Table I. In agreement with the previously published structure (31), the N peptide lies in an extended conformation within the K^b cleft, anchored by the phenylalanine at p5 and leucine at p8 (Fig. 8A, Supplemental Fig. 4). The E1 and G4 peptides are also in the same extended conformation as N (root mean square deviation of 0.21 or 0.13Å compared with N, respectively (Fig. 8A), suggesting that these variants exhibit no major changes in the overall peptide binding mode. The p1-Ser residue in the N peptide acts as a secondary anchor, showing a high number of main chain hydrogen bonds with the MHC, with the side-chain hydroxyl group making two hydrogen bonds with p63-Glu in the MHC (Supplemental Fig. 4C, 4D). Interestingly, there was loss of two direct bonds with the Glu⁶³ in the MHC in the K^bE1 complex (Fig. 8B). Furthermore, the E1 side chain extended out of the MHC groove, becoming more solvent exposed in a manner that would likely impact OT-I TCR binding. Overall, the structural analysis indicates that substitution of p1 S to E within the N peptide may result in decreased peptide stability as a result of the loss of contacts around the secondary anchor residue, which could in turn impact TCR recognition and lead to lower avidity interactions.

The K^bG4 complex also showed a highly similar structure to the K^bN, with a root mean square deviation of 0.21Å within the MHC α1–α2 domains (Fig. 8B). Although most hydrogen bonds with the peptide main chain were conserved, there was loss of one direct hydrogen bond to the Arg¹⁵⁵ (Fig. 8B). This MHCα2 residue has been described as a key MHC contact residue for all TCR/pMHC ternary complexes solved to date (45–47). Again, mutation of the p4N to G results in loss of an essential solvent exposed TCR contact. Taken together, this substitution may impact on TCR recognition via direct and indirect conformational changes.

An implication of the above results is that E1 and G4 may destabilize the peptide within the MHC groove because of the loss of specific hydrogen bonds between the variant peptides and the MHC (Fig. 8). To determine whether this was indeed the case, the thermostability of each complex was determined. The K^bN complex shows high thermostability with a Tm point of 57.9°C, whereas the K^bE1 and K^bG4 complexes showed lower Tms of 50.2 and 50.3°C, respectively (Table II), indicative of less stable pMHCI complexes. To determine whether this impacted cell surface stability of pMHC complexes, APL stabilization of K^b was determined by RMA-S assay using graded doses of peptide (Fig. 9). The Q4 and G4 peptides were able to stabilize K^b cell surface expression to a similar degree as the WT N peptide (Fig. 9A). Thus, for G4, the thermodynamic stability does not reflect ability to stabilize cell surface pMHC expression. In contrast, a greater concentration of the E1 peptide was required to obtain 50% maximal K^b stability when compared with the WT N peptide (Fig. 9A). Despite demonstrating a similar K^b stability profile to WT N peptide, the G4 peptide demonstrated a diminished capacity to stimulate effector OT-I CTL (Fig. 9B). Moreover, although the E1 peptide could stabilize pMHC cell surface expression, albeit less efficiently than WT N peptide, there was a lack of OT-I activation (Fig. 9B). Thus, combined with structural analysis, these data suggest that the inability of these variant peptides to activate naive OT-I CTLs in vivo likely reflects abrogation of optimal TCR recognition, rather than decreased pMHC stability postinfection.

Stimulating naive OT-I CTLs in vitro demonstrated that, although low-affinity pMHC/TCR interactions can trigger clonal expansion, both survival and the acquisition of effector function are highly dependent on the quality of the inductive signals. Exposure to the agonist N peptide induced robust proliferation with rapid and sustained acquisition of polyfunctional cytokine production. The integration of multiple signals is clearly central to this induction of full effector CTL differentiation, because the absence of costimulation gave rise to OT-I cells that, despite robust proliferation and survival in culture, were functionally limited. Previous studies using the mutant G4 peptide as a stimulus have demonstrated that, whereas OT-I cells may be partially activated by K^bG4, they do not persist (11).

Furthermore, although stimulation with the mutant G4 and E1 peptides induced CD44 expression, the level of surface CD62L staining remained high after exposure to the low-affinity E1 peptide. Also, in contrast to the wt N peptide, exposure to E1 in the presence of concurrent CD28 costimulation failed to rescue either sustained OT-I proliferation or to initiate the sustained acquisition of multiple cytokine expression. The findings from this in vitro analysis are thus in accord with the notion that only high-avidity TCR/pMHCI interactions are capable of triggering CTLs in a way that is both optimal and sustainable.

Our data support earlier observations (39) that strong TCR signals can induce the production of TNF-α within 5 h of exposing naive CD8⁺ T cells to a potent inductive stimulus, in this case, the cognate N peptide. This same functional capacity is not triggered by suboptimal (G4 or E1) OT-I TCR/pMHCI interactions, perhaps reflecting decreased proliferation as a consequence of diminished autocrine cytokine production (48, 49). Interestingly, our data suggest that added exogenous IL-2 is relatively dispensable for the acquisition and maintenance of cytokine synthesis by CD8⁺ T cells that are induced in culture via a high (N), although not a low (G4 and E1), avidity TCR/pMHCI interaction. This agrees with a recent publication reporting that greater exposure to IL-2 during the course of CD8⁺ T cell activation did not increase cytokine production. However, this same study also demonstrated that the expression of perforin and granzyme B was greatly enhanced by IL-2 (50). The obvious implication is that the various CTL effector functions may be regulated via different signaling pathways.

A striking finding was that although CD28 ligation was key to robust, sustained effector function following stimulation with the cognate N peptide, we also observed more limited initial proliferation and maintenance of a CD62L^hi phenotype when this pathway was engaged. The maintenance of cell surface CD62L expression may be important to ensure that fully activated CTLs that produce multiple cytokines do not leave the lymph nodes and traffic to the site of infection too early. This limitation of OT-I cycling with CD28 engagement is, however, at odds with previous reports suggesting that costimulation promotes the proliferative capacity of CD8⁺ T cells (51). It is also possible, of course, that the effect observed in this study reflects the high concentrations of peptide and anti-CD28 mAb used in this in vitro system, leading to a situation that may never be seen under the physiological conditions that apply in the responding lymph node.

The OVA_257–264 APLs Q4, G4, and E1 exhibit a spectrum of ligand potency for the OT-I TCR that correlates with different outcomes for immature OT-I T cell development. Although both the agonist OVA_257–264 and weak agonist Q4 result in negative selection (15), the low-affinity E1 and G4 support positive selection of OT-I TCR transgenic T cells (15, 16). Interestingly, despite being able to support positive selection, infection of mice with recombinant influenza A viruses expressing either the G4 or E1 variant peptides failed to induce the activation of adoptively transferred naive OT-I cells. Our data indicate that OT-I recognition of the G4 and E1 peptides falls below a threshold required for inducing efficient T cell activation (52). The analysis of Zehn et al. (9) found that other p4 variants can activate OT-I T cells, and in particular, V4 was shown to induce a very small OT-I response with the expected consequences for diminished CTL expansion and effector function. Although V4 is slightly more immunogenic for OT-I T cells, the in vitro situation (13) was generally similar to that found for G4 in our study. Given these peptides represent the lower spectrum of APL potency for the OT-I TCR (15), these two ligands likely fall the minimal threshold for TCR-mediated OT-I CD8⁺ T cell activation under in vivo conditions. Importantly, a potential implication of these data are that self-ligands that support positive selection of a mature T cell repertoire are in themselves unlikely to be able to initiate activation of these selected T cells in the periphery as they fall below the affinity threshold for mature T cell activation.

The lack of OT-I responsiveness was not due to diminished antigenicity, because these viruses induced robust, endogenous CTL responses. Furthermore, competition for Ag was not obviously a factor, because the removal of K^bG4-specific naive precursors did not restore the OT-I response. Also, the OT-I cells rendered anergic by exposure to the low-avidity K^bG4 epitope, because they could be efficiently recalled when the mice were challenged 6 wk later with the x31-N virus. In addition, memory OT-I cells were not triggered by infection with x31-G4, suggesting that the heightened sensitivity characteristic of CD8⁺ T cell memory does not overcome the OT-I TCR/pMHCI signaling threshold required for effective CTL activation. This lack of recall was not a consequence of some immunosuppressive effect resulting from influenza A virus infection, because the x31-Q4 virus triggers robust OT-I responses. In fact, even in the context of peptide-pulsed BMDCs, K^bG4 could not induce an OT-I response, suggesting that, whereas the OT-I TCR can to some extent be engaged by high concentrations of G4 peptide encountered in vitro, these conditions are not replicated in the in vivo situation.

An interesting observation was that administration of the E1 and G4 as lipopeptides did induce some OT-I proliferation, although not resulting in full effector CTL differentiation. An implication is that targeting of TLR2 (44), either on the DC or on the T cell, somehow lowers the affinity threshold enabling activation of OT-I CTL. A recent report has suggested that TLR2 signaling can enhance CTL activation and effector function in response to peptide activation (53). It is tempting to speculate that in our model, TLR2 targeting on OT-I lowers the threshold of activation and promotes activation in response to low-affinity E1 and G1 APLs. In support of this, we have previously reported (43) that lipopeptide vaccination resulted in a broader responding T cell repertoire due to recruitment of CTL with suboptimal TCR characteristics. Our interpretation of these data is that TLR2 signaling enabled activation and recruitment of low-affinity T cell clones, although this still remains to be tested.

What may influence this threshold is the level of pMHCI stability (54, 55). Earlier experiments have suggested that the reduced activation of naive CD8⁺ T cells by altered peptide ligands reflects that formed pMHCI complexes are less robustly presented at the cell surface (56). The decreased thermodynamic stability of both the K^bE1 and K^bG4 complexes (when compared with K^bN) may reflect the loss of some hydrogen bonds that potentially impact the density and persistence of epitope presentation postinfection. This may be a factor in the inability of K^bE1 and K^bG4 to activate naive OT-I CTLs following in vivo priming. However, in agreement with others (38), we found the extent of N, G4, and E1 stabilization of K^b on RMA-S cells to be largely equivalent. As such, diminished thermostability does not necessarily translate to loss of stability at the cell surface. More importantly, we demonstrated that despite an ability to stabilize K^b cell surface expression, both the G1 and E1 APLs’ failed to effectively induce effector OT-I CTL activation. This, together with the fact that infection with the x31-G4 and x31-E recombinant influenza A viruses does induce robust, endogenous CTL responses to the K^bE1 and K^bG4 epitopes, demonstrates that the major factor in the failure of these pMHC complexes to activate OT-I CTL appears to be at the level of the OT-I TCR/pMHCI interaction.

In fact, the peptide substitutions analyzed in this study are associated with relatively minor alterations in pMHCI topology, reflected in the addition or loss of prominent side-chain residues for the E1 and G4 peptides, respectively. In particular, P4-Asn has been previously described as a TCR contact residue (57), so substitution with a Gly at this location may be expected to result in compromised OT-I TCR ligation and subsequent poor activation. Interestingly, the structural analysis indicates that the Asn to Gly change at position 4 resulted in a small but significant modification of the side-chain contacts made to Arg¹⁵⁵ within the MHCα2 helix. Position 155 is invariably contacted in all the TCR-pMHCI ternary structures solved to date and has been proposed as a “gate keeper” residue that is important for efficient TCR docking (45, 46, 58). Thus, it is tempting to speculate that any alteration in Arg¹⁵⁵ conformation is sufficient to compromise efficient OT-I TCR binding onto the pMHCI complex. In the case of the E1 peptide, the introduction of a prominent Glu side chain that extends out from the pMHCI cleft likely interferes with OT-I TCR ligation, leading to a low-avidity K^bE1/OT-I TCR interaction (34, 35) and consequent diminished CTL activation.

The present analysis thus establishes that small changes in peptide sequence that cause limited structural alteration can lead to low-avidity TCR/pMHCI ligation and inefficient T cell activation. The substitution of key contact TCR contact residues and consequent modification in atomic contacts are sufficient to render an antigenic peptide invisible to that particular CTL clonotype, providing a ready mechanism for viral escape from T cell immunity. This correlation of various parameters of T cell responsiveness with diminished TCR/pMHCI avidity and structural compromise has important implications for CD8⁺ T cell-mediated immunity. Only through increasing our understanding of these mechanisms can we improve the design of CTL-based immunotherapies and produce vaccines that promote effective, long-lived cross-protective CD8⁺ T cell memory and recall.

We thank Dr. Nicole La Gruta and Prof. Lorena Brown (University of Melbourne) for reagents and helpful discussion. We thank the staff at the Australian Synchrotron for assistance with data collection.

The authors have no financial conflicts of interest.