Selective Expansion of Cross-Reactive Cd8⁺ Memory T Cells by Viral Variants

C57BL/10 mice at 5–6 wk of age were obtained from the animal department of the Netherlands Cancer Institute. Mice were handled at all times in accordance with institutional guidelines.

Peptides were produced using standard g-fluorenylmethoxycarbonyl (FMOC) chemistry. Soluble fluorochrome (PE or allophycocyanin [APC])-labeled MHC tetramers were produced as described previously 17,19 and stored frozen in Tris-buffered saline/16% glycerol/0.5% BSA.

Influenza A viruses A/NT/60/68 and A/HKx31 were provided by Dr. R. Gonsalves, National Institute for Medical Research, London, UK. Influenza virus B/Lee/40 was obtained from the American Type Culture Collection. Mice were killed at indicated time points after infection, and organs were removed for further analysis. Inflamed lung tissue and spleens were minced in single chamber mesh filters. The single cell suspensions obtained were treated with NH₄Cl solution to get rid of contaminating erythrocytes, before staining for flow cytometry purposes.

The influenza A virus (A/NT/60/68) NP–derived H-2D^b-restricted CTL epitope, ASNENMDAM, was introduced into EL4 tumor cells by retroviral insertion as a COOH-terminal fusion with the enhanced green fluorescent protein (eGFP) gene product (Wolkers, M.C., manuscript in preparation).

In all instances, mononuclear cells were stained with directly labeled mAbs or MHC tetramers. Analysis was performed on a FACSCalibur™ (Becton Dickinson) using CELLQuest™ software (Becton Dickinson). Before staining, propidium iodide (PI) was added to gate for PI-negative (living) lymphocytes.

Cytolytic activity of sorted CD8⁺ T cells derived from inflamed pulmonary tissue was determined in a standard 5-h ⁵¹Cr-release assay. EL4 target cells were preincubated with peptides for 1 h at 37°C. Percent specific lysis was calculated from the equation: [(experimental ⁵¹Cr release – spontaneous ⁵¹Cr release)/maximal ⁵¹Cr release – spontaneous ⁵¹Cr release)] × 100%.

Spleen-derived mononuclear cells were stained with anti-CD8 mAb in the presence of PE-labeled MHC tetramers containing increasing concentrations of MHC monomers. Cells were subsequently analyzed by flow cytometry.

Intracellular cytokine staining was performed as described 20. In brief, spleen cells were incubated with peptide (0.5 μM) for 5–6.5 h at 37°C in the presence of recombinant human (rh)IL-2 (50 U/ml) and Brefeldin A (0.1 μl/ml). After incubation, cells were surface stained with anti-CD8a–APC mAb (PharMingen), incubated in Cytofix/Cytoperm solution (PharMingen) for 20 min on ice, washed, and stained for intracellular cytokine with anti–IFN-γ–FITC (PharMingen) or FITC-labeled isotype control antibody (PharMingen). Analyses were performed on a FACSCalibur™ (Becton Dickinson) using CELLQuest™ software. Isotype control antibodies resulted only in background staining (data not shown).

Mice were injected subcutaneously with 100 μg peptide in IFA, 4–6 wk after a primary influenza A virus infection. On days 0, 1, and 2, 100 μg anti-CD40 mAb (FGK.45) was injected intravenously 21. On day 10 after peptide immunization, spleen cells were used for flow cytometry analysis.

To assess the effect of random antigen variation on the dynamics of T cell responses in vivo, we infected mice with pairs of influenza A viruses. These viruses expressed either the same NP_366–374 epitope or epitope variants. The specificity of the resulting T cell repertoire was assessed by two-parameter MHC tetramer staining, and association of MHC tetramers to NK receptors was ruled out through analysis of CD8b-expressing cells only. When mice are infected once or twice with either influenza virus strain A/NT/60/68 or A/PR/8/34, which differ in the sequence of the immunodominant NP CTL epitope at positions 7 and 8 (ASNENMDAM vs. ASNENMETM), the vast majority of the resulting NP-specific T cells selectively recognize the epitope of the strain encountered and not that of the opposite strain (Fig. 1 A, panels 1 and 3). This dominant role of the peptide side chains at positions 7 and 8 in ligand recognition by the majority of T cells is in accord with the prominent contribution of p8 and especially p7 to the TCR-exposed surface of this peptide–MHC complex 22.

This virus strain specificity of the NP-reactive CTLs contrasts sharply with the apparent lack of strain specificity during secondary responses against a variant strain. When mice that had recovered from a previous infection with influenza virus strain A/NT/60/68 are challenged with viral strain A/PR/8/34, most if not all NP^PR-reactive T cells are fully cross-reactive between the two NP variants (Fig. 1 A, panel 4). This phenomenon is reciprocal: in mice that have previously experienced infection with strain A/HKx31 (a reassortant strain with the NP gene from A/PR/8/34), the NP-reactive T cell population that emerges upon infection with A/NT/60/68 is fully cross-reactive between the two viral strains (Fig. 1 A, panel 2).

Although somewhat variable between individual mice, the extent of binding of the two MHC tetramers appears independent of the order in which the epitopes were encountered over a large series of experiments. To compare the affinity for primary and secondary antigen in a more direct manner, competition studies were performed. These experiments demonstrate that the binding of fluorochrome-labeled MHC tetramers can be inhibited by similar concentrations of ASNENMDAM- and ASNENMETM-containing monomers, indicating equal affinity of the cross-reactive TCRs for either antigen (Fig. 1 B). In addition, these data rule out the possibility of dual TCR expression 23,24 by the cross-reactive T cells, since both MHC monomers compete for binding of ASNENMDAM-containing MHC tetramers.

The above results indicate that the subsequent encounter of variants of a T cell epitope results in the expansion of a cross-reactive T cell population, as established by biochemical assays. In fact, this expansion inhibits the expansion of the largely strain-specific population observed during a regular primary response. However, T cell recognition of ligands with similar affinities can have drastically different functional outcomes, due to differences in off rates of TCR–ligand interactions 25,26,27,28,29. Therefore, we examined the functional behavior of this biochemically cross-reactive T cell population towards both the primary and the mutant epitope. The cross-reactive T cells that are observed during a recall influenza A virus infection lyse target cells pulsed with both the primary and the recall antigen directly ex vivo to the same extent over a wide range of E/T ratios, and over a range of peptide concentrations (Fig. 2 A).

As an independent test of functional cross-recognition, we measured the ability of these cells to initiate IFN-γ synthesis upon stimulation with the original or variant epitope. Since simultaneous staining of cells with MHC tetramers and intracellular IFN-γ staining is technically difficult, cells were stimulated in the presence of the original and variant peptide separately or simultaneously (Fig. 2 B). The proportion of IFN-γ–positive CD8⁺ T cells is similar in all three cases, indicating that the T cell populations that recognize the two epitopes are largely overlapping and therefore cross-reactive.

We conclude that the T cell population that emerges upon encounter of an antigenic variant is biochemically cross-reactive, and this cross-reactivity is fully reflected in their functional behavior.

To better understand the ontogeny of the cross-reactive T cell population, we examined the kinetics and composition of this T cell pool. The cross-reactive cytotoxic T cell population appears 2–3 d earlier at the site of infection (i.e., pulmonary tissue; data not shown) than the antigen-specific T cells during infection of naive mice 15,16,17, suggesting that these cells originate from a preexisting memory T cell population. Several studies have shown a narrowing of the antigen-specific polyclonal TCR repertoire during recall infection, due to the preferential outgrowth of a subpopulation of memory cells 30,31,32,33. In naive mice that are infected with influenza virus A/PR/8/34, the repertoire of NP-specific T cells involves a variety of BV elements (Fig. 3 A). The slight preferential usage of the BV8.3 element reported previously for C57BL animals infected with influenza virus A/PR8/34 was not observed in these experiments 34. In contrast, in A/NT/60/68-primed mice that are infected with influenza virus A/PR/8/34, the repertoire of A/PR/8/34-specific T cells is highly restricted (Fig. 3 A). This narrow T cell repertoire is likely to reflect the affinity maturation observed previously 33 compounded by the low number of cross-reactive T cells within the original memory population. In certain animals, the oligoclonal nature of the cross-reactive T cell population appears to be directly visible from double-tetramer analyses of the influenza-reactive CD8⁺ T cell population. In these mice, the expanded cross-reactive T cell population appears as two to three separate populations (an example is shown as Fig. 3 B), which may reflect slightly distinct affinities for the primary and secondary antigen.

Several reports have indicated that the requirements for activating memory T cells differ from those for activating naive T cells, making them more susceptible to low-affinity TCR triggering or cytokine-mediated stimulation 35,36,37,38,39. To address the possible contribution of aspecific or more broadly reactive T cells in the formation of the cross-reactive T cell population, influenza A/NT/60/68-primed mice were infected with the antigenically unrelated influenza B virus (B/Lee/40). Both at day 4 (data not shown) and day 8 after infection (Fig. 4, panel 1), no expansion of the influenza A/NT/60/68-specific memory T cell population is observed. This is in agreement with results by others showing that bystander activation during an acute viral infection is minimal 20,40,41. Furthermore, the influenza A virus–specific T cells in mice that underwent sequential infections with A/NT/60/68 and A/PR/8/34 do not bind peptide–MHC tetramers that contain an adenovirus E1A-derived CTL epitope (Fig. 4, panel 2). Thus, the cross-reactive T cell population that expands upon infection of primed mice does require a significant structural homology between two antigens that are encountered sequentially, and is specific for these two peptide–MHC complexes. Finally, selective outgrowth of cross-reactive T cells does not take place to an appreciable extent when the two related antigens are introduced simultaneously (Fig. 4, panel 3). This suggests that the selective advantage of cross-reactive cells relies on quantitative or qualitative traits of the T memory population, such as homing properties or the requirements for costimulatory signals.

To provide a first estimate of the extent of structural homology between two antigens that is required for the selective expansion of cross-reactive T cells, we searched the National Center for Biotechnology Information (NCBI) GenBank database for other natural variants of the H-2D^b–restricted influenza NP epitope. Thus far, 10 variants of this epitope have been identified, 7 of which contain mutations that affect TCR-exposed side chains (positions 6, 7, and 8) (reference 22, and Table). Because influenza A virus is not a common mouse pathogen, this set of mutants should be random with respect to T cell recognition. Synthetic peptides corresponding to these mutants were generated, and mice that had previously been exposed to influenza A/NT/60/68 or A/HKx31 were challenged by subcutaneous immunization. To circumvent the need of helper T cells for the induction of an effective CTL response, mice were treated with anti-CD40 mAb (FGK.45) after vaccination 21,42,43,44. These experiments establish that in most if not all cases where influenza A virus–primed mice are confronted with antigens that contain a single mutation within the T cell epitope, an influenza-specific T cell population emerges that is fully cross-reactive between the primary and secondary antigen (Table). This phenomenon is not only observed for conservative substitutions, but also for more drastic amino acid changes (e.g., Ala8→Asn). For variants that contain multiple alterations in TCR-exposed residues, cross-reactivity is observed for some sequences (e.g., ASNENMDAM→ASNENMETM), but not for others (e.g., ASNENMDAM→ASNENVEAM). Both the type of mutation and the contribution of the mutated residues to the TCR-exposed surface of the peptide are likely to be determining factors in this regard. For one of the mutant epitopes (ASNENVETM), the functional behavior of the cross-reactive T cells was also tested. In line with the biochemical data, intracellular IFN-γ staining of spleen cells that were stimulated with the primary antigen (ASNENMETM) or the peptide variant used for vaccination revealed an identical percentage of responding cells (data not shown).

Cross-protection or the lack thereof by an epitope-specific T cell population cannot readily be tested through the use of mutant or recombinant viral strains. Even in settings where B cell immunity is absent, such as in μ-MT mice, it is difficult to exclude a possible contribution of T cell responses against subdominant epitopes that are conserved between the primary and the recall strain. To circumvent these difficulties, we developed a model system in which only the T cell epitope under study is shared between the primary and secondary antigen encounter. The NP_366–374 CTL epitope of influenza virus strain A/NT/60/68 was introduced into the EL4 tumor cell line as a COOH-terminal fusion with the eGFP gene product. Although the parental cell line grows progressively when injected subcutaneously in C57BL/10 mice, the NP epitope–expressing tumor is rejected over a 2–3-wk time period. Furthermore, tumor rejection is paralleled by the appearance of an NP_366–374–specific T cell population and is markedly enhanced by simultaneous infection with influenza virus A/NT/60/68 (Wolkers, M.C., manuscript in preparation). Thus, the introduced NP epitope appears to function as a bona fide (neo)tumor antigen in this setting, as recognition of this antigen is correlated with tumor rejection.

To examine the in vivo effects of cross-reactive T cell populations, mice that had previously been exposed to influenza virus A/NT/60/68 (carrying the homologous NP epitope), A/HKx31 (reassortant of A/PR/8/34; carrying a variant NP epitope), or an unrelated influenza B virus were challenged with EL4-NP_366–374 tumor cells. After tumor cell injection, blood samples were taken from individual mice and the frequency of virus strain-specific and cross-reactive CD8⁺ T cells was measured. In mice that had been infected with the unrelated influenza B virus (Fig. 5, panel 1), tumor growth is comparable to that in uninfected mice (data not shown). As expected, mice that were previously exposed to influenza virus A/NT/60/68, which shares the CTL epitope with the EL4-NP_366–374 tumor, showed a strong reduction in tumor growth (Fig. 5, panel 1). This protection is accompanied by a massive and rapid increase in the number of A/NT/60/68-monospecific T cells, which were apparently reactivated from the memory T cell pool by the NP epitope–expressing tumor (Fig. 5, panel 2). Importantly, intermediate tumor growth was seen in mice that had previously been infected with variant virus A/HKx31 (Fig. 5, panel 1). In addition, the reduced tumor outgrowth in these mice is accompanied by the expansion of a large population of CD8⁺ T cells, which cross-react between the NP_366–374–expressing EL4 tumor and the viral strain used for priming (Fig. 5, panel 3). These results illustrate that the presence of a structurally related T cell antigen promotes the expansion of infrequent cross-reactive T cells even in the absence of any further noticeable homology between primary and secondary challenge, and this expansion is accompanied by a significant reduction in tumor outgrowth.

The ability of cytotoxic T cell populations to cross-react between different viral strains was first appreciated by Townsend and Skehel in 1984 18. They showed that repeated in vitro restimulation protocols of splenic lymphocytes derived from influenza A virus–primed mice could lead to the selection of NP-specific T cell lines that cross-reacted between different influenza A viral strains. At that time, it was unknown whether these T cell lines cross-reacted at the level of the (variable) immunodominant NP_366–374 CTL epitope, or whether shared subdominant epitopes were recognized. However, with the benefit of hindsight, these experiments can be said to have revealed for the first time the capacity of cytotoxic T cells to productively recognize viral variants.

In spite of the early recognition of T cell cross-reactivity in in vitro assays, the biological in vivo significance of this cross-reactivity for the immune system to cope with viral variants has since remained unclear. Our findings now document that prior antigen exposure dramatically affects the repertoire of T cells used in a subsequent response to antigenic variants in vivo. The propensity for cross-reactivity between two T cell antigens appears roughly proportional to the sequence similarity between the epitopes tested, and seems to be a common event for antigens that are closely related. These cells that expand in vivo are phenotypically indistinguishable from conventional effector T cell populations (CD44^high, CD62L^low; data not shown) and are functional ex vivo and in vivo. This process appears to be due to the selective expansion of cross-reactive T cells present in the memory T cell pool, and is not dependent on shared B or T helper epitopes between primary and recall antigen.

In recent years, several groups have studied the impact of epitope variants on antigen-specific T cell responses (for reviews, see references 6, 45, and 46). These variant epitopes were generally isolated as immune escape variants, or were identified in in vitro assays by their aberrant recognition by T cell clones. These selected antigen variants function as either partial agonists or antagonists of antigen-specific T cell responses not only in vitro but also in vivo. We sought to examine whether this type of T cell antagonism or partial agonism is a common phenomenon when a polyclonal T cell repertoire is confronted with antigenic variation. To this purpose, we studied the development of antigen-specific T cell repertoires after encounter of random natural variants of influenza A viruses. We conclude that during such encounters, the T cell repertoire generally reacts with the outgrowth of a T cell population for which the variant epitope is a full agonist. The TCRs encoded by these T cells bind with equal affinity to MHC molecules complexed with wild-type and variant epitopes. Furthermore, the functional capacity of this T cell population towards target cells expressing the original or the variant antigen is indistinguishable both in vitro and in vivo.

How do these data fit in with previous observations that mutations in T cell epitopes can lead to CTL escape by the virus? During chronic HIV and HBV infections 10,11, and also in a murine lymphocytic choriomeningitis virus (LCMV) model 12, the mechanism that we have identified apparently does not operate efficiently. In theory, this could be due to the type of amino acid changes in the T cell epitopes involved. However, examination of the type of mutations in the T cell epitopes in those cases and the mutations studied here does not point towards obvious differences (not shown). Alternatively, the impaired ability to react to emerging antigenic variants in those settings may be due to alterations at the T cell level. Specifically, repetitive antigen-specific T cell stimulation results in a narrowing of the reactive T cell repertoire 33. Indeed, the antigen-specific T cell expansions observed during chronic HIV infection are oligoclonal 47. It may be hypothesized that for such restricted antigen-specific T cell populations, a single antigenic variant could antagonize a substantial part of the antigen-specific T cell response, and it will be a challenge to test this notion in a direct manner.

Why do cross-reactive T cells dominate the response against antigenic variants over the largely strain-specific response observed normally during a primary T cell response? The cross-reactive T cells that we have identified appear to originate from the preexisting memory T cell pool, and naive and memory T cells differ both in quantitative and qualitative terms. Specifically, even though cross-reactive cells are infrequent in the original memory pool, these cells may still outnumber the strain-specific T cells that—due to their specificity—failed to get activated in the primary response. In addition, the activation requirements of memory T cells are less stringent than those of naive T cells 38,48,49. This is reflected in a lessened requirement both for costimulatory signals and for prolonged antigenic stimulation. Finally, the ability of memory T cells to enter peripheral tissues may promote the early encounter of viral antigens 50. Indirect evidence for functional T cell cross-reactivity during subsequent viral infections has been obtained previously. Experiments by Welsh and colleagues 51 have suggested that a measure of heterologous immunity may occur even for apparently unrelated viruses. Although in that case the antigenic determinants involved were not identified and cross-reactive cells were of low avidity, these results do suggest that the selective expansion of cross-reactive memory T cells could be a more general phenomenon.

We conclude that a polyclonal T cell repertoire responds to the encounter of random variants by the expansion of memory T cells that are best fit for the recognition of the variant antigen. These findings suggest that random mutations in T cell epitopes are unlikely to result in CTL escape, and that a polyclonal T cell memory population can provide protection not only against the index sequence, but also against a swarm of related sequences.

The authors would like to thank Mireille Toebes and Marjo van Puijenbroek for preparation of MHC tetramers and technical assistance, Helmut Kessels for peptide synthesis, Drs. G. Rimmelzwaan and A. Osterhaus (Erasmus University, Rotterdam, The Netherlands) for growing and titrating the various influenza strains, and members of the Kruisbeek and Schumacher labs for discussions.