Screening with the flow cytometric IFN-γ assay has led to the identification of a new immunogenic peptide (SSYRRVPGI) from the influenza PB1 polymerase (PB1703–711) and a mimotope (ISPLMVAYM) from the PB2 polymerase (PB2198–206). CD8+ T cells specific for KbPB1703 make both IFN-γ and TNF-α following stimulation with both peptides. The CD8+ KbPB1703+ population kills PB2198-pulsed targets, but cell lines stimulated with PB2198 neither bind the KbPB1703 tetramer nor become CTL. This CD8+KbPB1703+ population is prominent in the primary response to an H3N2 virus, although it is much less obvious following secondary challenge of H1N1-primed mice. Even so, we can now account for >40% of the CD8+ T cells in a primary influenza pneumonia and >85% of those present after H3N2 → H1N1 challenge. Profiles of IFN-γ and TNF-α staining following in vitro stimulation have been traced for the four most prominent influenza peptides through primary and secondary responses into long-term memory. The DbNP366 epitope that is immunodominant after the H3N2 → H1N1 challenge shows the lowest frequencies of CD8+ IFN-γ+TNF-α+ cells for >6 wk, and the intensity of IFN-γ staining is also low for the first 3 wk. By 11 wk, however, the IFN-γ/TNF-α profiles look to be similar for all four epitopes. At least by the criterion of cytokine production, there is considerable epitope-related functional diversity in the influenza virus-specific CD8+ T cell response. The results for the KbPB1703 epitope and the PB2198 mimotope also provide a cautionary tale for those using the cytokine staining approach to identity antigenic peptides.
The concordance in virus-specific CD8+ T cell numbers (1, 2) measured by flow cytometric analysis with tetrameric complexes of MHC class I gp plus peptide (tetramers), or by staining for intracellular IFN-γ subsequent to short-term stimulation with peptide to cause cytokine production (PepC)4 assay, has greatly facilitated the dissection of cell-mediated immunity of H-2b mice infected with the influenza A viruses (3). Quantitative analysis has focused on the primary and recall CD8+ T cell responses in the regional mediastinal lymph nodes (MLN), the spleen, and the pneumonic lung sampled by bronchoalveolar lavage (BAL). The secondary challenge experiments rely on the fact that the surface hemagglutinin (H) and neuraminidase (N) glycoproteins of the A/Aichi/68/HKx31 (H3N2), and A/PR/8/34 (PR8, H1N1) influenza A viruses are sufficiently different to avoid any cross-neutralization with antibody (4), while these two viruses share the internal components that provide allof the peptides recognized to date by influenza-specific, H-2b-restricted CD8+ T cells (5, 6, 7). Productive infection is essentially limited to the lung, reflecting that a protease required to cleave the influenza virus H molecule is restricted in distribution to the respiratory epithelium (8). The influenza mouse model thus provides an optimal experimental system for the quantitative analysis of CD8+ T cell responsiveness in a localized infection (9, 10).
The primary CD8+ T cell response following intranasal (i.n.) challenge with the HKx31 virus is largely specific for two peptides derived from the viral nucleoprotein (NP366–374) and the PA polymerase (PA224–233), respectively (7). Both are presented by H-2Db. The CD8+DbNP366+ and CD8+DbPA224+ sets detected by tetramer staining comprise <5% of the CD8+ T cells in spleen and >30% of those recovered by BAL from the infected lung. The secondary response (HKx31 → PR8) is dominated by the CD8+DbNP366+ population, although the CD8+DbPA224+ T cells also increase in prevalence.
The present experiments identify another epitope that is prominently recognized in the primary response to the HKx31 influenza A virus. These same T cells also make both IFN-γ and TNF-α (11) following stimulation with a totally different peptide derived from another influenza virus protein, although this mimotope does not seem to stimulate an independent immune response. Any assignment of antigenicity based on screening with the PepC assay must clearly be confirmed by other, more functional criteria.
Patterns of IFN-γ and TNF-α staining have also been analyzed for a spectrum of immunogenic peptides in both the acute response and in long-term memory. The cytokine expression profiles for influenza-specific CD8+ T cells are clearly a function of the particular epitope, the level of local antigen stimulation, and the time that has elapsed since primary or secondary challenge. These observations raise intriguing questions about possible selective loss, or further differentiation with time, in Ag-specific CD8+ memory T cell populations.
Materials and Methods
Mice and tissue sampling
Female C57BL/6J (B6, H-2b) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Mice that were at least 8 wk of age were anesthetized by i.p. injection with avertin (2,2,2-tribromoethanol) and infected i.n. with 106.8 50% egg infectious doses of the HKx31 influenza A virus (3, 12). “Memory” mice for secondary challenge experiments were injected i.p. at least 6 wk before with 108.5 50% egg infectious doses of the PR8 influenza virus. The PR8 and HKx31 influenza A viruses (4) have different surface H and neuraminidase (N) glycoproteins (H1N1 and H3N2, respectively), but share internal components (NP, NS1, NS2, M, PA, PB1, PB2). At the time of sampling, the mice were anesthetized and exsanguinated from the axillary artery. Lymphocytes were obtained from the pneumonic lung (12) by BAL, and adherent cells were removed by incubating on plastic for 1 h at 37°C. Spleen and MLN samples were disrupted and enriched for CD8+ T cells by incubation with mAbs to CD4 (GK1.5) and MHC class II glycoprotein (M5/114.15.2), followed by magnetic depletion subsequent to binding anti-rat and anti-mouse Ig-coated beads (Dynal, Oslo, Norway).
PepC intracellular cytokine assay
The PepC assay utilized lymphocytes that were cultured for 5 h in 96-well round-bottom plates (Costar, Corning, NY) at a concentration of 5–8 × 105 cells/well in 200 μl of RPMI 1640 medium containing 10% FCS, 10 u/ml human recombinant IL-2, and 5 μg/ml brefeldin A (Epicentre Technologies, Madison, WI) in the presence or absence of viral peptide (7). They were then washed and stained with anti-mouse CD8α-FITC Ab (BD Pharmingen, San Diego, CA). Nonspecific Fc binding was blocked using anti-mouse CD16/32 (BD Pharmingen). The cells were fixed in 1% formaldehyde in PBS for 20 min, then permeabilized in PBS/0.5% saponin for 10 min before staining (30 min) with conjugated mAbs to IFN-γ (PE-XMG 1.2) and/or TNF-α (APC-MP6-XT22). The mAbs to IFN-γ and TNF-α were mixed. The specificity of the staining reaction was checked initially by blocking with excess, purified cytokine and by the use of isotype control Ab. The data were acquired on a Becton Dickinson FACScan or FACSCalibur flow cytometer, then analyzed using CellQuest software (Becton Dickinson Immunocytometry Systems, San Jose, CA). In each assay, the percent CD8+IFN-γ+ or CD8+TNF-α+ without peptide (<0.2%) was subtracted from the percent CD8+IFN-γ+ or CD8+TNF-α+ with peptide to give the percent specific CD8+ T cells staining for the particular cytokine. Kinetic analysis of epitope-specific CD8+ T cell populations recovered directly ex vivo utilized 1 μM viral peptide.
The influenza A virus peptides were synthesized at the Center for Biotechnology, St. Jude Children’s Research Hospital, using a model 433A peptide synthesizer (Applied Biosystems, Berkely, CA) and purified by HPLC. The previously identified (5, 6, 7) peptides/epitopes were: DbNP366, H-2Db + NP366–374; DbPA224, H-2Db + PA224–233; KbNS2114, H-2Kb + NS2114–121, KbM1128, and H-2Kb + M1128–135. In addition, a new epitope (KbPB1703, H-2Kb + PB1703–711) and a mimotope (PB2198–206) were identified by screening (at 10 μM) immune spleen cells with a panel of overlapping 15-mer peptides spanning the PB1 and PB2 proteins of the PR8 virus (13) using the PepC assay for IFN-γ. The positives were resynthesized as shorter sequences (9–10 aa). The restriction element for the new PB1 peptide was identified using the RMAS peptide stabilization assay. The RMAS cells (14) were first incubated for 16 h at 26°C in complete medium, then in 96-well plates in the presence of 2-fold dilutions of peptide (100 μM to 760 pM) for 30 min at room temperature followed by 3 h at 37°C. The cells were then stained with mAbs to H-2Db (28.14.85) or H-2Kb (AF188.8.131.52), followed by FITC-conjugated rabbit anti-mouse IgG (Dako, Copenhagen, Denmark), and analyzed by flow cytometry.
Tetramer staining of peptide-specific CD8+ T cells
Virus-specific CD8+ T cells were identified using tetrameric complexes (1, 3, 7) of the H-2Db glycoprotein and peptides derived from the NP ASNENMETM (5) and polymerase 2 protein (PA) SSLENFRAYV, or H2Kb + SSYRRPVGI from PB1. These are referred to in the text as the DbNP366, DbPA224, and KbPB1703 tetramers. Recombinant H-2Db or H-2Kb molecules with a birA biotinylation motif substituted for the carboxyl-terminal transmembrane domain were refolded with human β2-microglobulin plus the appropriate viral peptide, biotinylated with birA and complexed at a 4:1 molar ratio with neutravidin-PE (Molecular Probes, Eugene, OR). Lymphocytes were stained for 60 min at room temperature with the tetrameric complexes in PBS/BSA/azide, followed by staining with anti-CD8α-FITC, washed twice, and analyzed by flow cytometry.
Cell lines and CTL assays
Polyclonal cell lines were generated from mice infected acutely with the HKx31 virus. Irradiated (3000 rad) HKx31-infected naive spleen cells were washed twice and incubated (106/ml) with responder lymphocytes (1.5 × 106/ml) for 5–7 days in complete medium at 37°C with 5% CO2. The cell lines were then generated by multiple cycles of restimulation with peptide-pulsed (1 μM) irradiated spleen cells every 5–7 days in medium incorporating 10 U/ml IL-2.
The EL4 (H-2b) target cells were labeled with Na51Cr for 1 h, pulsed with viral peptides or infected with the HKx31 influenza A virus for 60–90 min, washed twice, and plated at 5000 targets/well. They were then incubated with the effector populations for 5 h before harvesting supernatants for gamma counting. Two-fold lymphocyte dilutions were assayed in triplicate, while untreated and Triton X-100-disrupted controls were measured in quadruplicate. The percent specific lysis was calculated as 100 × (51Cr release from targets with effectors − 51Cr release from targets alone)/(51Cr release from targets with Triton X-100). The level of 51Cr release from targets incubated in the absence of T cells did not exceed 15% of the total Triton X-100-mediated 51Cr release.
A PB1 epitope and a cross-reactive PB2 mimotope
Overlapping 15-mer and 10-mer peptides were made (15) for the PR8 influenza PB1 and PB2 proteins, screened at 10 μM in the PepC assay for IFN-γ synthesis, and then tested for up-regulation of MHC class I molecules in TAP-2-deficient RMAS cells (Fig. 1). A PB1 epitope was identified as a H-2Kb-restricted 9-aa (SSYRRPVGI; KbPB1703–711) sequence which loosely conforms to the defined H-2Kb motif (16) XXYXF/YXX(X)L/M/I/V. A second stimulatory peptide was identified in PB2 (ISPLMVAYM; PB2198–206) by the PepC assay. However, it was not possible to assign this PB2198 peptide to either H2Kb or H2Db. No specific cytolysis above background was detected when PB2198-pulsed L-929 cells transfected with H-2Db or H-2Kb molecules were used as targets (data not shown), and only high concentrations of peptide enhanced the expression of both MHC class I glycoproteins in the RMAS assay (Fig. 1).
Freshly isolated BAL populations from mice acutely infected with the HKx31 virus were assayed for CTL activity on EL4 cells pulsed with graded concentrations of the PB1703 and PB2198 peptides. Both peptides sensitized the EL4 targets for lysis, although the effect was clearly greater with PB1703 (Fig. 2). Cell lines were then generated by six cycles of restimulation with the PB1703 or PB2198 peptides and assayed for CTL activity. The PB1703 lines were lytic for EL4 cells pulsed with PB1703–711, PB2198–206, and the truncated PB1703–710 (Fig. 3,A), although PB1703–710 showed little capacity to up-regulate H-2Kb expression in the RMAS assay (Fig. 1). The PB2198 lines, however, showed minimal evidence of CTL activity (Fig. 1,A). In addition, the PB2198 lines were less uniformly CD8+ (Fig. 3,B), although this phenotype only emerged after three cycles of in vitro culture (data not shown), and did not stain with the KbPB1703 tetramer (Fig. 3 C).
We then asked whether the PB1703-specific and PB2198-specific CD8+ T cells that could be detected in freshly isolated BAL populations by the PepC assay for IFN-γ (Fig. 4, C and D) were indeed distinct sets of T cells. The BAL cells were stained with the KbPB1703 tetramer, sorted into CD8+KbPB1703+ and CD8+KbPB1703− sets, and then stimulated with the two epitopes together (Fig. 4, J and O) or separately (Fig. 4, H, M, I and N). It was very obvious that the PB2198 peptide was indeed stimulating the majority of the CD8+KbPB1703+ T cells (Fig. 4 IN). The only significant response in the CD8+KbPB1703− population was caused by the control DbNP366 peptide (Fig. 4,G). It thus seems reasonable to assume that PB2198 (although completely unrelated in sequence) is a mimotope that can stimulate an IFN-γ response by highly activated PB1703-specific CD8+ T cells (Fig. 4) and is recognized by PB1703-specific CTL (Figs. 2 and 3,A), but is not capable of driving either the clonal expansion of cells that bind the KbPB1703− tetramer or effector CTL generation following in vitro culture (Fig. 3, A and C).
Magnitude of epitope-specific primary and secondary CD8+ T cell responses
Immunologically naive (primary) and PR8-primed mice (secondary) were challenged i.n. with the HKx31 virus, then analyzed by the PepC/IFN-γ assay for the number of CD8+ T cells responding to five different H-2b-restricted epitopes (Fig. 5). In looking at Fig. 5, it is important to realize that the y-axes differ for the various panels. We chose to express the results as numbers rather than percent values, as this was the only way that the response profiles for the different epitopes and anatomical sites could be adequately illustrated. Following primary challenge, the KbPB1703 peptide was found to rank third (after DbNP366 and DbPA224) in magnitude for the response in the lymphoid tissue (Fig. 5, A and B) and was also prominent in the BAL (Fig. 5,C). The recall response was, as shown previously by concordant patterns of staining with both tetramers and the PepC/IFN-γ assay (3, 7), dominated by the expansion of the DbNP366-specific set (Fig. 5, D–F). Although the secondary DbNP366-specific response was increased (Fig. 5, D–F) 10- to 15-fold above that found in the primary (Fig. 5, A–C), the difference in magnitude for the subdominant epitopes tended to be more in the range of 2- to 4-fold (Fig. 5, A–C and G–I). Adding the percent values (data not shown) used to calculate the data presented in Fig. 5 also established that influenza-specific T cells account for >40% of the CD8+ T cells that localize to the lungs of mice with a primary influenza virus pneumonia, with that value increasing to >85% after secondary challenge.
Epitope-related patterns of IFN-γ and TNF-α production
The next question was whether the PB1703 epitope and the PB2198 mimotope would trigger the same spectrum of TNF-α and IFN-γ staining (PepC assay) in the CD8+ T cell population recovered by BAL from the site of virus-induced pathology. The TNF-α/IFN-γ profiles were essentially similar following exposure to the PB1703 and PB2198 peptides, but the spectrum for the NP366 peptide was different (Fig. 6). Approximately 40% of the IFN-γ-producing, DbNP366-specific CD8+ T cells were not producing TNF-α. In addition, a small fraction (<5%) of DbPA224-specific CD8+ T cells produced low levels of IL-2 (data not shown) while no DbNP366-specific CD8+ T cells appeared to secrete this cytokine.
The experiment was then repeated at several time points during the primary and recall response (Fig. 7) to the four principle influenza CD8+ T cell epitopes (see Fig. 5). More CD8+IFN-γ+/TNF-α+ than CD8+IFN-γ+TNF-α− spleen cells were found for the KbNS2114, DbPA224−, and KbPB1703-specific sets on day 8 after secondary challenge (Fig. 7). The exception was the splenic DbNP366-specific population, although the majority of the CD8+DbNP366-specific T cells in the BAL were IFN-γ+/TNF-α+ at the same time point (Fig. 7). The difference is likely to reflect that the levels of DbNP366 encountered in the lung, the site where the virus is replicating, are much higher than those in the spleen. Previous experiments have shown very clearly that potent CTL effector function is most obvious in the BAL population. By day 21 after the H3N2 → H1N1 challenge, the majority of both the BAL and the splenic CD8+DbNP366-specific T cells were IFN-γ+/TNF-α+.
A more detailed kinetic analysis with the two major epitopes (DbNP366 and DbPA224) showed that >75% of the CD8+DbPA224+ T cells were IFN-γ+/TNF-α+ by day 12 after primary infection, and retained that profile as both memory T cells (Fig. 8,A, day 30, and day 0 B) and in the long-term following secondary challenge (Fig. 8, A–D). This was also true for the secondary, but not the primary, response in the BAL to DbNP366 (Fig. 8, C and D). Long-term memory following both primary and secondary challenge was characterized by a slow shift of the CD8+DbNP366+ T cells toward an IFN-γ+/TNF-α+ phenotype.
Analysis of the intensity of cytokine staining indicated that the IFN-γ profiles were essentially similar for the KbNS2114−, DbPA224-, KbPB1703-, and DbNP366-specific sets on days 8 and 30 of the primary response (Fig. 9,A). However, the massive expansion (Fig. 5, D–F) of the CD8+DbNP366-specific T cell population that occurs during the secondary response was characterized by a much lower level of IFN-γ staining (Fig. 9,C, days 8 and 21). This effect was no longer apparent for the memory T cells recovered on day 78 after secondary challenge (Fig. 9,C, day 78) and was not obvious for TNF-α production in either the primary or secondary response. Clearly, the patterns of cytokine response can vary for different epitopes, at least for the first 1–2 mo, after a primary or secondary encounter with an influenza A virus (Figs. 8 and 9).
These experiments establish that different peptides from the same virus induce varied spectra of functional activity. At one extreme we have a mimotope (ISPLMVAYM, PB2198) derived from the viral polymerase 3 protein that stimulates T cells generated during the course of an in vivo response to the SSYRRPVGI polymerase 1 epitope (KbPB1703) to make both IFN-γ and TNF-α. However, PB2198 neither promotes clonal expansion of the CD8+KbPB1703+ set nor generates CTL effectors specific for targets pulsed with either the PB1703 or PB2198 peptides. At the other end of the spectrum, the prominent DbNP366-specific response is characterized by levels of IFN-γ and TNF-α production that are low relative to those induced by KbPB1703 and DbPA224. Although DbNP366 drives lymphocyte differentiation and proliferation very efficiently in vivo, there is no evidence that PB2198 is ever expressed on the surface of either virus-infected targets or Ag-presenting stimulator cells. What useful generalizations be drawn from these findings?
The results with the PB2198 mimotope provide several lessons. The first is the very practical point that, although the PepC/IFN-γ assay allows the rapid identification of potentially immunogenic peptides, this can only be regarded as a screening procedure. Any such assignment of antigenicity must be confirmed by other functional criteria. The second is that there is no a priori mechanism for identifying possible cross-reactivities: the relationship between ISPLMVAYM and SSYRRPVGI is defined only by a spectrum of the TCRs that are responding to the KbPB1703 epitope. The third is that the profiles of cytokine production identified subsequent to stimulation with PB2198 must be those that are established by the initial stimulation with KbPB1703. A low-affinity-avidity interaction (17) between KbPB2198 and the KbPB1703-specific TCRs will then cause IFN-γ and TNF-α production, but cannot drive clonal expansion or CTL generation. The implication is that different functions are set at different thresholds. The triggering of a particular response is presumably determined by the avidity of the TCR-epitope interaction that in turn dictates the magnitude of signal (18, 19, 20).
It is intriguing that the lowest level of concordance for IFN-γ and TNF-α production is detected throughout for the DbNP366-specific population that is so prominent in the secondary H3N2 → H1N1 response (3, 7). Perhaps DbNP366 is selecting for the development of memory T cells with a much broader range of TCR affinities. We know, for example, that both the acute and memory CD8+ T cell responses to DbPA224 utilize predominantly the Vβ7.1 TCR chain (21). Although Vβ8.3 tends to be expressed on 30–60% of the DbNP366-specific T cells (22), the pattern is much less consistent than that for DbPA224 and Vβ7.1. Although the impression to date is that the percent of Vβ8.3+ T cells in the DbNP366-specific set does not change with time, it would be necessary to look sequentially within individuals to see whether there is a progressive focusing of TCR clonotypes as this T cell population evolves toward the “memory” IFN-γ+TNF-α+ phenotype (23).
Another possibility is that the generation of high levels of TNF-α following a primary encounter with Ag is correlated with the development of less responsive or diminished (24, 25, 26) memory T cell populations. The primary response to DbPA224 is initially higher than that to DbNP366 though, by day 12 after i.n. challenge, there are larger numbers of DbNP366-specific T cells in the spleen. Proportionally more of these DbPA224-specific T cells are producing TNF-α. The nature of both the primary and secondary CD8+ T cell response to these two epitopes should be compared for mice that are not capable of making, or responding to TNF-α. We also need to re-examine the situation for H-2kxb F1 mice (21), where the CD8+DbNP366-specific response is not immunodominant following the secondary H3N2 → H1N1 challenge. Will decreasing the size of the DbNP366-specific population modify the profile of cytokine expression, or is this a function of the TCR-epitope interaction?
The results presented here are generally in accord with the observation of Slifka and Whitton (11) that lymphocytic choriomeningitis virus-specific, CD8+IFN-γ+TNF-α+ double-positive cells are more prominent in the established memory T cell pool than in activated CD8+ T cell populations recovered during the acute Ag-driven phase of the response. The results from the influenza model add the insight that the extent of this effect differs for particular peptides. The secondary CD8+ DbPA224+ response shows the characteristic IFN-γ+TNF-α+ memory profile identified by Slifka and Whitton (11). This is also true for the CD8+DbNP366+ set in the BAL, but not for the comparable population detected in the spleen following the H3N2 → H1N1 challenge. The difference probably reflects that the higher level of antigenic stimulation in the lung (12) is driving the emergence of the IFN-γ+TNF-α+ phenotype.
Perhaps the CD8+DbNP366+IFN-γ+TNF-α+ set that predominates with time is the “true” memory population (27). Both tetramer staining and the PepC assay demonstrate that CD8+DbNP366-specific T cells are present in the memory T cell pool at an order of magnitude higher (2, 3) than the frequencies determined by earlier (28) limiting dilution analysis. This difference may reflect that only 10% of the CD8+DbNP366+ set survives the 12–15 cycles of cell division required to read-out as positive in the limiting dilution analysis CTL assay (27), although all can apparently cycle to some extent after secondary in vivo stimulation (29). The differences in cytokine profiles presented here suggest that the spectrum of functional diversity may be much greater for the massive DbNP366+ population generated following the H3N2 → H1N1 challenge than for the more modest secondary response to, for example, DbPA224. It is obviously important to clarify whether the characterization of an epitope as “dominant” or “subdominant” (30) based on the comparison of T cell numbers (21) is indeed reflective of functional efficacy in the in vivo situation.
We thank Gabriela Byers for technical assistance and Vicki Henderson for help with this manuscript.
This work was supported by Public Health Service Grants AI29579, AI38359, and CA21765, and by the American Lebanese Syrian Associated Charities. G.T.B. is a C. J. Martin Fellow of the Australian National Health and Medical Research Council (Reg. Key. 977 309).
Abbreviations used in this paper: PepC, peptide/cytokine; BAL, bronchoalveolar lavage; H, hemagglutinin; HKx31, the A/Aichi/68/HKx31 H3N2 influenza A virus; M1, viral matrix protein; N, neuraminidase; NP, nucleoprotein; NS2, viral nuclear export protein; PA, PB1 and PB2, the three influenza virus polymerase proteins; PR8, the A;/PR/8/34 H1N1 influenza A virus; i.n., intranasal; MLN, mediastinal lymph node.