Initial TCR affinity for peptide Ag is known to impact the generation of memory; however, its contributions later, when effectors must again recognize Ag at 5–8 d postinfection to become memory, is unclear. We examined whether the effector TCR affinity for peptide at this “effector checkpoint” dictates the extent of memory and degree of protection against rechallenge. We made an influenza A virus nucleoprotein (NP)-specific TCR transgenic mouse strain, FluNP, and generated NP-peptide variants that are presented by MHC class II to bind to the FluNP TCR over a broad range of avidity. To evaluate the impact of avidity in vivo, we primed naive donor FluNP in influenza A virus–infected host mice, purified donor effectors at the checkpoint, and cotransferred them with the range of peptides pulsed on activated APCs into second uninfected hosts. Higher-avidity peptides yielded higher numbers of FluNP memory cells in spleen and most dramatically in lung and draining lymph nodes and induced better protection against lethal influenza infection. Avidity determined memory cell number, not cytokine profile, and already impacted donor cell number within several days of transfer. We previously found that autocrine IL-2 production at the checkpoint prevents default effector apoptosis and supports memory formation. Here, we find that peptide avidity determines the level of IL-2 produced by these effectors and that IL-2Rα expression by the APCs enhances memory formation, suggesting that transpresentation of IL-2 by APCs further amplifies IL-2 availability. Secondary memory generation was also avidity dependent. We propose that this regulatory pathway selects CD4 effectors of highest affinity to progress to memory.
T cells recognize conserved viral epitopes, and hence T cell memory provides broad, heterologous immunity crucial to protection against rapidly mutating viruses, including influenza. Resting CD4 T cells respond to infection by generating Ag-specific effectors that protect against infection via multiple synergizing mechanisms (1), including help to CD8 and B cells, production of IFN-γ, and perforin-mediated cytotoxicity. After viral clearance, a cohort of CD4 effectors becomes memory cells that can protect against future infections. However, the signals and mechanisms required for the transition of CD4 effectors to memory are only partially defined.
Several studies indicate that longer duration of Ag stimulation during priming results in increased naive CD4 T cell response, leading to greater proliferation and function of effector cells (2–4). Both the amount of peptide Ag loaded onto MHC class II (MHC-II) receptors on an APC, which determines the density of peptide–MHC-II (pMHC-II) complexes on the APC, and the affinity of the pMHC-II interaction with TCR during priming determine the extent of T cell response (5). Although strong pMHC-II interactions with TCR favor Th1 over Th2 development (6–8), their role in T follicular helper cell (Tfh) versus Th1 differentiation is less clear, with conflicting reports (9–12). There are also conflicting reports that either strong or weak initial pMHC–TCR interactions better support memory formation and recall responses (11, 13–20). Thus, whether the extent of CD4 effector cell transition to memory is dependent on their affinity for Ag is unclear.
Mice infected with influenza present a wide diversity of Ag epitopes to T cells, with high viral titers and Ag presentation, soon after infection that remain high until infection is cleared (21, 22). Infection produces very strong CD4 T cell memory, suggesting that persistent high levels of Ag may explain the high levels of memory generated. Here, we specifically analyze the impact of TCR avidity for peptide on APCs at the effector phase, when both viral levels and CD4 effectors have peaked and when effectors that fail to recognize Ag undergo apoptosis (23, 24). We ask if high- versus low-affinity Ag drives generation of more CD4 memory cells and if they provide superior protection. Previously, we showed that to become memory cells, CD4 effectors generated in situ by influenza A virus (IAV) infection needed to recognize Ag during the effector phase checkpoint, spanning from 5 to 8 d postinfection (dpi) (23, 24). During this cognate interaction, the effectors must produce autocrine IL-2, which prevents their apoptosis, enabling their transition to memory (23–25). Moreover, we also showed that the addition of short-lived, Ag-pulsed APCs at this checkpoint boosted memory formation (23) but that viral infection, apart from its role in cognate Ag presentation, was not required between 5 and 8 dpi to promote memory (23).
To study the impact of peptide avidity at this crucial juncture, we developed a TCR transgenic (Tg) mouse (FluNP) specific for NP311-325, an immunodominant, highly conserved, IAV nucleoprotein (NP) epitope in C57BL/6 (B6) mice (22). We made a truncation and single amino acid substitutions to generate a library of NP311 peptides with a spectrum of functional avidities for the FluNP TCR. We used an adoptive transfer model to generate in vivo effectors from naive CD4 by IAV infection in a first host, isolated 6-dpi donor effectors, and then cotransferred them to a second host (23). Here we used activated APCs, pulsed with peptides from a panel spanning high to low avidity for the FluNP TCR, as the only source of Ag.
We find that APCs with higher avidity peptides at 6 d of the CD4 effector response promote a far larger CD4 memory population in the spleen, lung, and draining lymph node (dLN) and that this leads to better protection upon influenza rechallenge. The peptide affinity and dose used to pulse APCs determines the level of IL-2 produced by the 6-d effectors, and the levels of IL-2 correlate with the prevention of default effector apoptosis, survival, and development of memory (24). CD25, the high-affinity IL-2 receptor (IL-2Rα), is not expressed on 6-d CD4 effectors but is highly upregulated on APCs in IAV-infected 4–8-dpi mice. Holding the peptide Ag constant, APCs expressing CD25 generated higher numbers of CD4 T cell memory than CD25-deficient APCs. We propose that IL-2Rα expression on APCs acts in concert with IL-2Rβ/γ on the effector CD4 cells to enhance available IL-2 and signaling. We suggest that at the effector phase, the level of IL-2 production, determined by peptide affinity, and the efficacy of the response to autocrine IL-2 enhanced by IL-2 transpresentation are the dominant pathways that regulate the size of the memory CD4 population. We discuss the implications of this requirement for high peptide affinity at the effector stage for vaccine design that promotes protective CD4 memory with a focus on kinetics, Ag breadth, and adjuvants that activate APCs.
Materials and Methods
We use 8–12-wk-old B6 mice as hosts in all experiments. Naive CD4 T cells are isolated from B6.FluNP strains, including B6.FluNP.Thy1.1+/− and B6.FluNP.Nr4a1EGFP.Thy1.1+/−. The B6.Nr4a1 EGFP developed by Kris Hogquist and Steve Jameson (26) were from The Jackson Laboratory. Bone marrow-derived dendritic cells (BMDCs) were derived from B6 or B6.129S4-Il2ratm1Dw/J (CD25KO) mice obtained from The Jackson Laboratory and bred at the University of Massachusetts Chan Medical School (UMMS) breeding facility. Mice used in experiments were 8–12 wk of age. The institutional animal care and use committee of UMMS approved all animal procedures.
B6.FluNP TCR Tg mice were generated in collaboration with Eric Huseby’s Laboratory at UMMS. Briefly, B6 mice were infected with a sublethal dose of PR8 (0.3 LD50), and, at 21 dpi, 2 × 107 spleen and lung dLN cells were isolated and stimulated in vitro with irradiated spleen cells loaded with 100 μg/ml NP311-325 peptide. After 5 d, responding T cells were fused with BW5147 to generate T cell hybridomas (27). T cell hybridomas with reactivity to NP311-325 peptide, presented by lung APCs from IAV (A/PR8/34)-infected mice, were expanded. TCR Vβ-chains were identified by staining with a set of Vβ-specific Abs (BD Biosciences), and the TCRα-chains were identified by PCR analysis using a panel of TCR Vα primers that collectively amplify all TCR Vα gene families. We chose a hybridoma with Vα4.2 and Vβ2.1. A TCR Tg plasmid was made using cloned, rearranged cDNAs for 22.B6 TCR Vα4.2 and Vβ2.1. Vα4.2 (Arden) included the V region TRAV6-5, CDR3(CALRSSGSWQLIF) and J region(IMTG): TRAJ22. Vβ2 included TCRV (Arden):TRAV1, CDR3 (CTCSAEVGGDTQYF), and J region (IMTG): TRAJ2-5. Cloned products were fused with full-length TCR Cα and Cβ sequences (28). All the TCR genes were sequenced, and error-free full-length cDNAs were subcloned into the human CD2 promoter transgene cassette for T cell–specific expression (29). B6.FluNP were established by injecting B6 oocytes with the TCR-Tg plasmid.
Virus stocks and infections
Mice were anesthetized with either isoflurane (Piramal Healthcare) or ketamine/xylazine (at a dose of 25/2.5 mg/kg by i.p. injection) before intranasal infection with 50 μl of influenza virus diluted in PBS corresponding to a 0.2 to 0.3 (sublethal) LD50 for response, 2 LD50 for weight loss, and 4 LD50 for survival. Influenza virus A/Puerto Rico/8/34 (PR8, H1N1), originally from St. Jude Children’s Hospital, was from our stocks grown and maintained at the Trudeau Institute. The virus was also characterized by its ability to infect eggs, and we found that 2 LD50 corresponded to a 50% egg infective dose of ∼10,000. Our standard dose was sublethal 0.3 LD50, corresponding to 25 PFU.
Nucleoprotein peptide generation
We modified the NP311-325 peptide to produce peptides of shorter lengths by deletions of amino acids on both ends to determine the best length. We used single alanine substitutions to determine the peptide-I-Ab binding frame (P1 = Y), then selective amino acid side-chain modifications, at known peptide–TCR contact positions to generate peptides likely to have lower affinities. Peptide-I-Ab IC50 was determined with surface plasmon resonance (SPR) using a BIAcore 3000 (Cytiva). SPR analysis was performed using a BIAcore 3000 instrument. Briefly, biotinylated peptide exchanged MHCs were immobilized on a streptavidin chip. For the TCR, the FluNP TCR sequence was cloned into the pCDH lentiviral expression vector with a P2A site separating the α- and β-chains. This construct was used to generate stable lines in 293S GnTI cells. TCR was then purified from supernatant with a nickel-nitriloacetic acid column and subsequent size exclusion. Recombinant FluNP TCR was then passed over at increasing concentrations. A5 was used as a negative control, and the signal from this flow cell was subtracted from those of experimental flow cells. The resulting data points were plotted and fitted to hyperbolas to derive KD values.
BMDC generation and peptide/APC preparation
APCs were generated as described previously (23, 30). Bone marrow was harvested from B6 or CD25KO mice, washed with RPMI 1640 and 1% FBS, and cells were plated at 107 cells/ml in RPMI 1640 with 10% FBS and 10 ng/ml GM-CSF. After 7 d, CD11c+ BMDCs were isolated via MACS and activated with polyinosinic:polycytidylic acid [poly(I:C)] at 10 μg/ml overnight in culture or were used as APCs. APCs were pulsed with a standard concentration of 100 μM or dilutions thereof of each of a chosen panel of NP peptides. For in vivo experiments, pulsing was at 37°C for 1 h with shaking, APCs were washed three times and then resuspended in PBS, and 1 × 106 cells per mouse were injected i.v.
Sequential transfer model
We closely followed the model we developed previously (23). Spleens and peripheral lymph nodes were collected from B6.FluNP.Thy1.1+/− mice. Naive CD4 cells were isolated via negative selection with CD4 MACS (Miltenyi Biotec) and washed three times, then resuspended in PBS. Naive CD4 cells (0.5–1 × 106) were transferred via i.v. injection into B6 first hosts. First hosts were infected with a sublethal dose of PR8 the same day. Donor CD4 effectors were reisolated from the first hosts at 6 dpi. Single-cell suspensions were prepared from pooled spleens and dLNs, and donor FluNP cells were isolated via Thy1.1 positive selection by MACS (Miltenyi Biotec). Cells were resuspended in PBS, and 1.5 × 106 donor FluNP cells per mouse were injected i.v. into second B6 hosts, along with Ag/APC (peptide pulsed). To maintain effector phenotype, all steps were conducted at room temperature, except for one 15-min incubation at 4°C. This minimal protocol, without sorting, ensures that effector cells are only out of mice for a maximum of 2.5 h. IL-2 complex (IL-2–anti-IL-2 Ab) treatment was administered daily, at 5–7 dpi, as previously described (24).
In vitro culture of naive and 6-dpi effector FluNP CD4 T cells
To assess the functional avidity of peptides, we compared their ability to induce responses of FluNP naive CD4 T cells, isolated as described above (sequential transfer model). Following isolation of naive or 6-d FluNP effectors and generation of Ag/APCs, cells in complete RPMI 1640 media were plated at a CD4 T/APC ratio of 5:1. Following 2 d in culture, plates were centrifuged, supernatant was removed for cytokine protein analysis via ELISA, and cell pellets were stained for FACS analysis.
Supernatants were collected from in vitro culture. Plates (Nunc) were coated overnight with capture Ab (ELISAmax, BioLegend). The following day, plates were blocked according to the manufacturer’s protocol, and supernatants were added neat or diluted 1:10, 1:100, or 1:1000 and left at 4°C overnight. The following day, plates were washed, and detection was carried out per the manufacturer’s protocol.
Flow cytometry and cytokine staining
Cells were harvested, passed through a 70-μm cell strainer, and stained in FACS buffer (0.5% BSA, 0.01% sodium azide; Sigma-Aldrich) in PBS. Cells were blocked with anti-FcR (2.4G2) and stained with amine reactive viability dyes (Invitrogen) to exclude dead cells. Surface proteins were stained with fluorochrome-conjugated Abs at 4°C. Abs used included the following: CD4 (GK1.5 and RM), CD44 (IM7), CD90.1 (OX-7 and HIS51), CD11c, CD25, CD62L, CD69, CD80, CD86, CD122, CD132, CD185 (CXCR5, SPRCL5), NKG2A/C/E, and MHCII (I-Ab). For CD127 staining, anti-CD127-biotin was included in the surface stain mixture, cells were washed three times, and a streptavidin fluorochrome-conjugated secondary Ab was used in the second step. For cytokine staining, total splenocytes were stimulated with 10 μM of NP311-325 for 6 h at 37°C. Brefeldin A (10 μg/ml) was added after 1 h of stimulation. Following surface staining, cells were fixed in 2% paraformaldehyde for 20 min and permeabilized with 0.1% saponin buffer (1% FBS, 0.1% NaN3, and 0.1% saponin in PBS; Sigma-Aldrich) for 15 min. Subsequent staining for cytokines was carried out using the following Abs: anti-IFN-γ (XMG1.2), anti-TNF-α (MP6-XT22), anti-IL-2, and anti-IL-17. For transcription factor staining, cells were first surface stained, then fixed and permeabilized, using a Foxp3 fix/perm kit (eBioscience) overnight per the manufacturer’s protocol and stained with the following Abs: anti-BCL-6 (K112-91), anti-Foxp3 and anti-T-bet at 4°C for 1 h. Abs were obtained from BD Biosciences, BioLegend, and eBioscience. Stained cells were acquired on a BD LSRII flow cytometer and analyzed using FlowJo analysis software.
Groups of three to five mice were used in all experiments, and exact conditions were repeated to obtain sufficient statistical power. All experiments shown were repeated two or three or more times. For statistical analysis, an unpaired, two-tailed independent t test was used. All analyses were performed using GraphPad Prism software.
Characterization of FluNP TCR transgenic CD4 T cell response
We generated a B6 TCR Tg mouse specific for an immunodominant NP core protein epitope, NP311-325, from the internal NP of influenza PR8/34 (H1N1) presented by I-Ab (MHC-II). This epitope is conserved among all dominant outbreak strains of IAV in people (22) and in the IAV strains we used. The mouse was created by selecting T cell hybridomas specific for NP311-325 from IAV-infected mice (27–29). We call the mouse “FluNP.”
We generated B6.FluNP.Thy1.1/Thy1.2 mice so that we could readily detect donor FluNP cells by flow cytometry after transfer to second hosts by their Thy1.1 expression. To evaluate whether IAV induces a comparable response of the donor FluNP cells and polyclonal host CD4 T cells, we transferred naive FluNP CD4 T cells into hosts and infected them with a sublethal dose of PR8/34, a mouse-adapted IAV. We compared the kinetics of donor FluNP TCR Tg and endogenous host (CD4+CD44hi) T cell responses 4, 6, 8, 12, 21, and 63 dpi in the lung (Fig. 1A), mediastinal dLN, and spleen (Supplemental Fig. 1A). IAV infection induced a similar pattern of expansion and contraction of donor FluNP and host CD4 T cells in the lung (Fig. 1A), spleen, and dLN (Supplemental Fig. 1A). We assessed the subsets of CD4 effectors generated from both donor FluNP and host naive CD4 T cells at 8 dpi by their cytokine production and phenotypic markers. We enumerated Th1 (T-bet, IFN-γ, TNF-α, IL-2), triple cytokine producers (IFN-γ, TNF-α, IL-2), Th17 (IL-17), and regulatory T cells (Tregs) (Foxp3). We also used NKG2A/C/E expression to detect cytotoxic CD4 (ThCTL), which we found only in the lung (31), and CXCR5 and Bcl-6 coexpression markers to detect Tfh in the spleen (32) (Fig. 1B, Supplemental Fig. 1B–1I). Both donor and host responses in the lung were dominated by Th1 phenotype cells (33–35), with high expression of T-bet, IFN-γ, and TNF-α (Fig. 1B, left; and Supplemental Fig. 1B, 1E), and little expression of IL-17 or FoxP3 (Fig. 1B, left; and Supplemental Fig. 1D), markers of Th17 and Tregs. ThCTL in lungs was found in both at similar proportions (Fig. 1B, left; and Supplemental Fig. 1C). In the spleen, donor and host subset patterns were also similar (Fig. 1B, right; and Supplemental Fig. 1F–1I). Thus, the overall effector responses of donor monoclonal FluNP and host polyclonal CD4 to IAV were comparable.
We examined memory populations at 21 dpi, assessing phenotypically distinct memory subsets: central memory T cells (CD127+ CD44+ CD62L+), effector memory T cells (CD127+ CD44+ CD62L−), and resident memory T cells (CD44+ CD69+). The overall patterns in donor and host memory were again similar (Fig. 1C, Supplemental Fig. 1J–1M), and both donor and host memory cells expressed high levels of the canonical CD4 memory marker CD127 (∼75%) in all tissues (Supplemental Fig 1L). Together, these comparisons of donor FluNP and host polyclonal response support the conclusion that FluNP CD4 T cells respond equivalently to host polyclonal ones and that this model is suitable for investigating the impact of peptide avidity on memory generation.
Generation of a panel of nucleoprotein peptides with a broad range of avidity for FluNP TCR
To study the role of peptide avidity at the checkpoint in promoting memory generation, we first generated a series of NP peptides with single amino acid substitutions that have a broad range of abilities to stimulate the naive FluNP response when pulsed on B6-derived activated APCs. We sought substitutions that altered TCR interaction while maintaining tight peptide–MHC interaction. First, we identified the I-Ab binding frame by scanning partially overlapping 11-residue peptides that cover the full NP311-325 peptide for I-Ab binding, using a fluorescent peptide competition binding assay and purified recombinant I-Ab carrying a cleavable linker peptide (36). I-Ab binding was substantially reduced for NP313-323 and lost completely for NP314-324 (Fig. 1D), suggesting that Y313 occupied the key P1 position in the I-Ab binding site (37, 38). We confirmed this using alanine-scanning mutagenesis, revealing Y313 as the only position where I-Ab binding was substantially affected (Fig. 1D). To identify peptides that modulate TCR interaction, we introduced other substitutions in addition to alanine at the predicted TCR contact positions P2 (Ser→Gln), P5 (Arg→Lys), and P7 (Glu→Gln). Some of these substitutions caused moderate reductions in peptide-MHC binding, up to 3.4 for Q2 (Fig. 1D). To evaluate the effect of these substitutions on FluNP TCR interaction independent of peptide-MHC effects, we measured pMHC-TCR binding using an SPR (BIAcore) assay with streptavidin-immobilized biotinylated I-Ab-peptide complexes and recombinant soluble FluNP TCR (Fig. 1E). FluNP TCR bound to I-Ab carrying the parent NP311-325 peptide or truncated NP311-322 (NPT) with high affinity (KD ∼3 μM) (Fig. 1E). The other substitutions at predicted TCR contact position caused reductions in pMHC-TCR affinity ranging from ∼20-fold (for A7) to >150-fold (for Q7). The A5 substitution abrogated detectable binding.
We ranked the substituted NP peptides by their ability to stimulate an FluNP response (functional avidity). We loaded BMDCs activated with poly(I:C) with peptides over a broad dose range and evaluated how well they stimulated naive FluNP cells in vitro. After 2 d, we assessed induction of CD69 (Fig. 1F), CD25 (IL-2Rα) (Fig. 1G), and Nurr77 (Fig. 1H). By all three assays of naive CD4 T cell activation, FluNP CD4 T cells responded in a dose- and affinity-dependent manner. Thus, we could confidently rank the relative avidity of the peptide–MHC-II complex on APCs for the FluNP TCR on the CD4 T cells. We classified the NPT and NP311-325 as high avidity, A7 and K5 as medium (mid) avidity, and Q2 and Q7 as low avidity, and we used A5 and unpulsed BMDCs as negative controls. In each assay, the high peptides (NP311 and NPT) induced peak responses at doses 100-fold lower than the middle peptides (A7, K5), and the low peptides (Q2, Q7) required 10 times the dose of the two middle peptides (Fig. 1F–1H).
From here on, we used a high dose of 10−4 M to pulse APCs to minimize the contribution of peptide density and maximize the contribution of pMHC-TCR affinity. At this concentration, all six peptides (NPT, NP311-325, A7, K5, Q2, Q7) stimulate a measurable FluNP naive CD4 T cell response.
Peptide avidity at the effector phase determines the number of memory cells
We evaluated the impact of peptide avidity on in vivo memory generation using a sequential adoptive transfer model developed previously (23). Naive FluNP cells were transferred to first hosts, then infected with IAV to generate 6-d effectors in vivo. These were purified by flow cytometry on the basis of their expression of Thy1.1 and were cotransferred to uninfected second hosts with groups of peptide-pulsed APCs as the only source of Ag (23) (Fig. 2C). The APCs are short-lived and present Ag for only 48–72 h (23), defining the discrete checkpoint for effectors to recognize Ag that we have established is necessary for memory generation. To compare the number of FluNP memory cells generated from high-avidity peptide/APCs with those generated by IAV infection, we transferred 6-dpi FluNP effectors into either 6-dpi PR8 infection-matched hosts or uninfected hosts along with high-affinity peptide-pulsed NPT/APCs (Fig. 2A). At 21 dpi, NPT/APC-stimulated FluNP effectors produced as many memory cells in spleen and lung and only slightly fewer in dLNs as compared with PR8 infection (Fig. 2B). Thus, a high-avidity peptide/APC generates an equivalent number of memory cells from effectors as does IAV infection, both here and in an equivalent model using OT-II Tg CD4 T cells (23).
We investigated if modulating avidity over a broad range would lead to a corresponding influence on donor memory numbers in the second hosts in the same model (Fig. 2C). The impact of peptide avidity was striking, with numbers of recovered memory cells clearly dependent on the rank of the peptides. The greatest differences in memory cell numbers were seen in the lung, followed by the dLN, and last the spleen (Fig. 2D, 2E, and Supplemental Fig. 2A). The significance of these changes between groups of peptides is illustrated by the fold change in memory cell number between each peptide-pulsed APC compared with unpulsed APC (Fig. 2F, left), low peptides (Fig. 2F, right), and middle peptides (Fig. 2F, bottom). In the lung, there were 280-fold more donor FluNP cells in the NPT-pulsed group than in the unpulsed group, 29- and 36-fold more than in the low groups, and 6- and 8-fold more donor memory cells than in the mid groups. A nearly identical pattern was seen in the dLN. In the spleen, there were 16-fold more FluNP cells in the NPT group than in the unpulsed group and 5–8-fold more than in the mid groups, with both findings being highly significant (Fig. 2E, 2F). These data indicate that the strength of pMHC-TCR interaction at the effector checkpoint determines the size of the memory population in both secondary lymphoid organs and the lung, but that the effect is most dramatic in the lung and dLN.
We analyzed the donor-derived memory cells for surface phenotype. They uniformly express high CD127, the IL-7Rα (Supplemental Fig. 2B–E), which supports CD4 memory generation in the secondary lymphoid tissues (39), in all stimulated groups in the spleen and dLN (Fig. 2C–2E), whereas in the lung, many donor cells do not express CD127 as expected (35). We assessed cytokine production ex vivo by the memory cells stimulated with different peptides on APCs by intracellular cytokine staining (Supplemental Fig. 2F–2H). Overall, all memory cells produced equivalent levels of IL-2, TNF-α, and IFN-γ. The high-affinity peptide resulted in only a slightly higher fraction of IFN-γ-capable memory. Overall, these results suggest that even the memory cells that develop at lower avidity are likely to be functional if they encounter sufficient Ag on restimulation.
Peptide avidity at the effector checkpoint promotes effector cell survival
To probe the mechanisms by which peptide avidity has such a dramatic impact on memory cell recovery, we asked if affinity impacts FluNP effector cell proliferation, survival, or both (Fig. 3A). To evaluate proliferation, we stained the 6-d donor effectors with CellTrace Violet (CTV) before transfer and examined their division in vivo 3 d after cotransfer with Ag/APCs. As indicated by dilution of CTV, the spleen FluNP cells in each peptide group divided multiple times, whereas the unpulsed group barely proliferated (Fig. 3B). Thus, proliferation of transferred cells required Ag recognition, but peptide avidity did not impact the pace or numbers of division in any organ (Supplemental Fig. 3A–3C). These results suggest that the dramatic differences in memory cell numbers (Fig. 2) are not due to a greater rate or number of cell divisions.
We determined survival of FluNP effectors by measuring expression of caspase 3/7 and viability staining at 3 d post-transfer (dpt) (Fig. 3C, Supplemental Fig 3D–3F). In the spleen, we found a much greater fraction of live caspase 3/7-negative cells in the NPT (high) and A7 (mid) groups compared with the Q2 (low) group (Fig. 3C, Supplemental Fig. 3D), indicating there is greater apoptosis in the low-avidity group compared with both mid and high groups. Most donor cells recovered in the unpulsed APC group were proapoptotic by 3 dpt (Fig. 3C), consistent with a strict requirement for Ag recognition to prevent effector apoptosis (23). We did not see a clear difference in cell survival with different affinity peptides on APCs in dLN or lung, although all peptides led to higher survival than APCs with no peptide (Supplemental Fig 3E, 3F). Perhaps the response kinetics are different in each organ, or maybe the spleen is a major source of memory generation after which developing memory cells migrate to the lung and dLN. Therefore, we also evaluated donor cell recovery as a more integrated measure of net survival. We found a clear-cut pattern with recovery corresponding to higher peptide avidity in all sites (Fig. 3D) that was of similar magnitude at 3 dpt (late effector) as it was at 15 dpt (memory) (Fig. 2E). We conclude that the impact of peptide avidity in regulating memory cell recovery is realized within a few days of effector cell Ag encounter, and that it is due primarily to the fraction of effectors that survive rather than to their extent of division.
Peptide avidity at the effector checkpoint induces more effector IL-2 production
Effector phase autocrine IL-2 signaling is required for CD4 T cell memory development, and autocrine IL-2 acts by downregulating proapoptotic Bim, which promotes the survival of effector CD4 survival (23, 24). We thought this likely requires IL-2Rα expression on the effectors (24). Therefore, we analyzed the expression of IL-2R subunits CD132 (IL-2Rγ), CD122 (IL-2Rβ), and CD25 (IL-2Rα) on donor FluNP effectors before transfer and 3 d after in vivo stimulation with the various peptides/APCs. Before restimulation, 15% of FluNP effectors express CD132, 40% express low levels of CD122, but <10% express CD25 (Fig. 3E). Following stimulation with high-affinity peptides/APCs in vivo, three-fourths of FluNP cells in the lung at 3 dpt strongly express CD132, and almost all express CD122 (>90%) (Fig. 3E), regardless of the avidity of peptide used to stimulate them. Again, few FluNP effectors in the spleen and lung express CD25 (Supplemental Fig. 3G), and only half are CD25+ in the dLN (Fig. 3E, Supplemental Fig. 3G). In the in vivo setting, we found no difference in IL-2R subunit expression on 3-dpt donor cells following high, mid, and low TCR signaling in any tissue, and even the unpulsed group 3-dpt effectors expressed a similar pattern (Supplemental Fig. 3G). These findings suggest that avidity does not act by upregulating IL-2R chain expression and leave unresolved the mechanism by which autocrine IL-2 effectively signals the effectors.
We examined the impact of peptide avidity on the level of IL-2 produced by 6-d effectors. We restimulated 6-d FluNP effectors ex vivo with the panel of NP peptide-pulsed APCs introduced in Fig. 1, and we quantified IL-2-producing cells by intracellular cytokine staining (Fig. 3F) and levels of secreted IL-2 by ELISA (Fig. 3G, 3H). The fraction of IL-2-positive 6-d FluNP effectors corresponded closely with increasing peptide avidity, mirroring the impact on memory. Thus, induction of IL-2 secretion by FluNP effector cells is highly dependent on the avidity of Ag recognition. Because the CD4 T cell response is also dependent on the density of peptide Ag/MHC, we varied the peptide concentration used to pulse the APCs (Fig. 3G, Supplemental Fig. 3J). We found that peptide dose also strongly influenced IL-2 production by the 6-d effectors. With higher-avidity peptides, higher levels of IL-2 were seen, and they increased with concentration so that they were detectable even at a very low dose (10−7 M), whereas with mid and low peptides, much higher concentrations were required. We integrated the area under the curve as a reflection of overall IL-2 that could be (Fig. 3H). This confirmed that IL-2 production mirrored the rank of peptides determined by measures of affinity and functional avidity (Fig. 1). Because peptide avidity determines the level of autocrine IL-2 produced, and because IL-2 availability determines their survival (24), this is likely the key pathway that translates the peptide avidity into the number of memory cells generated following influenza infection. IFN-γ and TNF-α production were also dependent on peptide avidity and on dose (Supplemental Fig. 3H, 3I), indicating a general avidity dependence for the elicitation of cytokine production at the effector stage.
IL-2 complex addition can overcome lower peptide avidity for effector TCR
In the in vivo response to live influenza, autocrine but not paracrine IL-2 produced by cotransferred wild-type (WT) CD4 T cells supports CD4 effectors to efficiently transition to memory (24). However, when high levels of additional IL-2 are provided as a complex (IL-2c) that has extended bioavailability at 5, 6, and 7 dpi, IL-2-deficient CD4 T cells responding against influenza can develop into memory cells (40). We reasoned that if, as we suggested above, the impact of peptide avidity acts by enhancing IL-2 production during cognate effector–Ag/APC interaction, treating mice in which CD4 effectors responding to low-avidity peptides with IL-2c should enhance memory. To test this, we transferred 6-dpi FluNP effectors along with high-avidity NPT/APC or with low-avidity Q2/APC to second hosts. One group of second hosts that received Q2/APC was also treated with IL-2c for 3 d after cell transfer. We assessed donor cell memory generation at 21 d (Fig. 4A–4D). As before, CD4 memory generated in response to low-affinity Q2-APC was dramatically lower in all sites than that generated in mice receiving NPT-APCs. Importantly, the addition of IL-2c enhanced memory generation to the low-avidity peptide significantly so that the level was nearly comparable to the high-avidity response.
CD25 expression on APCs during the effector phase promotes effector transition to memory
When effector cells recognize peptide Ag presented by APCs, they make IL-2 within several hours, which traditionally was predicted to bind to the tripartite IL-2R on the same cell because the IL-2 needs to be autocrine (24). However, we see in Fig. 3E that the 6-d FluNP effectors do not express CD25. IL-15 shares two-thirds of its receptor (IL-2Rβ/γ) with IL-2 and is known to be transpresented by IL-15Rα on APCs to IL-2Rβ/γ on T cells (41–43). Several previous studies found that IL-2 can also be transpresented and speculated that this impacts responses by limiting (44) or augmenting IL-2 availability (45). We asked if IL-2 transpresentation might play a role during the cognate interaction of CD4 effectors with peptides/APCs. We first analyzed whether IAV infection generates CD25-expressing APCs. In uninfected mice, we found no CD25 expression on MHC-II+ cells, but at 6 dpi, a cohort of dLN CD11c+, MHC-II+ cells clearly express CD25 (Fig. 4E). At 4–8 dpi, the time when autocrine IL-2 signals are needed for CD4 T cell memory (40), substantial populations of CD11c+, MHC-II+ cells express CD25 in the lung, dLN, and spleen (Fig. 4F).
To analyze whether CD25 expression on APCs plays a role in CD4 effector survival to memory, we generated 6-d FluNP effectors and cotransferred them with WT or CD25−/− activated APCs (BMDCs) pulsed with high and mid NP peptides (Fig. 4G). Both APCs expressed high levels of CD11c, MHC-II, CD80, and CD86, but only WT BMDCs expressed CD25 (Supplemental Fig. 4A). We found that 6-d effectors stimulated in vitro with WT versus CD25−/− peptides/APCs produced equivalent amounts of IL-2 and IFN-γ (Supplemental Fig 4B, 4C), indicating that both activate FluNP similarly. However, when we assessed in vivo generation of memory in the transfer model with WT or CD25−/− APCs (Fig. 4G–4J), significantly fewer donor memory cells developed in the spleen, dLN, and lung of second hosts when NPT high- and middle-avidity peptides were presented by CD25−/− BMDCs (Fig. 4H–4J), with more dramatic differences in the dLN and lung and lesser fold differences in the spleen. The fraction of donor memory cells that produced IL-2, IFN-γ, and TNF-α was equivalent (Supplemental Fig 4D–4F). This provides clear evidence that transpresentation of IL-2 by APCs during cognate interaction serves to amplify the IL-2 signal to the CD4 effectors, and this leads to generation of more memory cells.
Peptide avidity at the effector checkpoint determines protection from influenza challenge
We asked if second hosts of FluNP effectors that receive higher-avidity peptides/APCs are better protected from rechallenge with IAV. We generated memory from FluNP effectors cotransferred with high or low Ag/APC (Fig. 5A). After 21 dpi, we challenged the mice with a sublethal dose of PR8 and analyzed their weight loss (Fig. 5B). The high-avidity group recovered weight significantly faster than the low-avidity group, and both recovered more rapidly than the unpulsed group (Fig. 5B). When we challenged hosts with a higher, lethal dose of PR8 and analyzed survival, more animals in the two high-avidity groups (>80%) survived than their two low-signaling counterparts (50–60%), and these were better protected than unpulsed APCs (20%) (Fig. 5C). Thus, higher-avidity peptide at the effector phase promoted a more protective memory FluNP population, and the degree of protection increased with the size of the memory population. This supports the concept that peptide avidity determined functioning memory.
Peptide avidity during Ag recognition by secondary effectors determines their memory generation
Memory CD4 T cells have less stringent requirements than naive cells for Ag dose and costimulatory interactions (46), and they become more protective secondary effectors during reinfection with influenza (47). Influenza-specific memory T cells accumulate with age in humans due to multiple exposures and their longevity through both influenza infections and vaccinations (48–50) and thus may dominate responses. We asked whether avidity of TCR for peptide determines the level of memory generated from secondary effector cells. We generated primary memory cells from 6-d FluNP effectors with high peptide/APC in the second host, whose polyclonal CD4 T cells remained otherwise naive. After 21 d, we infected these second hosts and isolated 6-d donor secondary effectors. We cotransferred the donor secondary 6-d FluNP effectors along with high-, mid-, and low-avidity peptide-pulsed APCs or unpulsed APCs (Fig 6A) into third hosts and enumerated secondary memory FluNP after 21 d in the spleen, dLN, and lung (Fig. 6B). In each organ, memory generation was dependent on Ag recognition as expected (23, 24), with the high peptide NPT producing 14.4-fold more memory in spleen, 40-fold more in dLN, and 34-fold more in lung than APCs with no peptide Ag. Thus, like naive cells, CD4 memory cells need to recognize Ag again as secondary effectors during the defined kinetic window we call the “effector checkpoint” to form optimal secondary memory and the avidity of TCR for peptide Ag again determines the amount of secondary memory cells that are formed.
We examined the impact of peptide avidity for the FluNP TCR, delivered by peptide-pulsed APCs at the CD4 effector checkpoint (23, 24), on memory formation. Our earlier studies established that peptide-pulsed poly(I:C) activated APCs added at the effector checkpoint, generated memory equivalent to that primed by infection (23), and showed that peptide-pulsed APCs used as here efficiently presented Ag to transferred 6-d effectors in the spleen, resulting in a strong systemic response like that of infection (32). We recovered the greatest number of CD4 memory cells after 6-d effectors were stimulated with the highest-avidity peptide, with 200-fold more in the lung compared with no peptide, and 20–50-fold more compared with low-avidity peptide. Higher avidity drove higher levels of autocrine IL-2 production, which promoted greater effector survival and donor cell recovery in the late effector phase, laying out the mechanisms likely to be responsible.
The importance of autocrine IL-2 in memory formation was emphasized by the fact that optimum levels of memory required that activated APCs express IL-2Rα during the cognate interaction with CD4 effectors that express IL-2Rβ/γ but not CD25, the IL-2Rα. Higher avidity also led to increased protection from lethal rechallenge. Secondary CD4 effector cells derived from memory cells also required Ag recognition to form secondary memory and were also favored by high-avidity peptide. To increase CD4 T cell memory, we suggest that vaccine strategies must supply a second round of high-avidity Ag and pathogen recognition signals from infection during the effector stage of CD4 T cell response.
Previous studies varied Ag avidity at the initiation of the immune response and found that higher levels and avidity could lead to greater effector and memory cell numbers (52–55). Extending the period of Ag presentation also promoted enhanced CD4 and CD8 memory formation (2–4, 23, 24, 56, 57). The pathways responsible were not determined, and we know of no studies by others that restricted Ag to the effector checkpoint 6–8 dpi, which we showed is strictly required for memory generation (23, 24, 40) and is when the TCR avidity for peptide/APCs has the striking impact shown here. Our study fills this gap in knowledge, by analyzing the effect of a broad range of affinities during the effector checkpoint on memory formation. Our results indicate that at the effector checkpoint, 6 dpi after initial infection, CD4 effector TCR avidity for peptide/MHC-II on APCs during cognate interaction determines the magnitude of memory by regulating the level of effector production of autocrine IL-2. In turn, this determines how many effectors survive over the next few days and progress to memory. Autocrine IL-2 at the effector stage is also required in polyclonal responses (24), so we suggest that this is a universal mechanism to select for CD4 memory cells that are of the highest affinity for the pathogen Ag epitopes they recognize.
In our studies, the impact of peptide avidity on memory was consistently more dramatic in the lung and dLN than in the spleen. One likely explanation for this is that higher-avidity peptide also promotes effector migration from the spleen to the lung, so the impact is less apparent in the spleen. We and others have shown that checkpoint Ag recognition and strong Th1 skewing promote greater expression of CXCR3 by effector cells, which promotes their trafficking to tissue sites (23, 35, 57, 58). The results are particularly striking here because in the second host, there is no infection or inflammation in the lung to attract effectors. Thus, we suggest that peptide avidity at the effector checkpoint also regulates migration to the tissues. A previous study indicated that high affinity of responding CD8 effectors for Ag was associated with prolonged proliferation, delayed contraction, and migration (59), suggesting that a parallel mechanism may occur in CD8 memory formation.
We found previously that Ag presentation to effector CD4 T cells at the effector checkpoint induces autocrine IL-2 production, which prevents their default apoptosis and thus supports formation of memory cells (24). IL-2 from coadministered WT TcR Tg cells could not rescue IL-2-deficient cell memory, indicating that paracrine IL-2 is not effective (23, 24). Here, we show that higher peptide avidity at the effector checkpoint does not drive greater division of 6-d effectors but does proportionally increase their level of IL-2 production and leads to a dramatic increase in short-term effector survival and recovery, such that the number of effectors at 9 d (Fig. 3D) is proportional to the size of the memory population at 21 d (Fig. 2E). In each organ, there were more than 100-fold higher donor effector cells recovered with the highest peptide versus no peptide, with very few donor cells in the unpulsed groups (spleen ∼103, dLN ∼500, and lung ∼50). At this time point, even the low-avidity peptide results in 20–40-fold more effector cells than no peptide (Fig. 3D). The increase in IL-2 production by effectors at 6 d, which is determined by peptide avidity and dose of peptide (Fig. 3), clearly links the effect of peptide avidity to IL-2 rescue of effectors and thus increased memory formation. We found that IL-2 complexes added at 6–8 dpi improved the responses to a lower-affinity peptide epitope (Fig. 4A–4D), producing almost equal numbers of memory CD4 T cells as the high-avidity ones, further supporting that the availability of IL-2 is the key factor that determines the efficiency by which CD4 effector cells transition to memory.
However, our analyses indicate that CD25 was either transiently expressed or not expressed on most CD4 effectors, raising the possibility that optimum autocrine IL-2-mediated survival of effectors might be regulated by a mechanism in addition to IL-2 binding to the tripartite IL-2R complex on the CD4 T cell. Another survival cytokine, IL-15, which shares IL-2Rβ/γ with IL-2, is presented in trans when bound to IL-15Rα on APCs while signaling through IL-2Rβ/γ on the T cell (40–42), and previous studies reported that activated APCs can express CD25 in both mice and humans (60, 61). We found that the activated BMDCs we used as APCs, here and in previous transfer models (Fig. 4E), as well as a substantial subset of activated APCs in mice infected with influenza 4–8 d earlier (Fig. 4F), express high levels of CD25. We tested CD25-deficient APCs for their ability to promote memory formation at the checkpoint from CD4 effectors and found that they generated fewer memory cells than WT APCs in vivo (Fig. 4). This suggests that Ag/APCs transpresent autocrine IL-2 to the interacting CD4 T cells and thus increase IL-2 availability.
We speculate that this autocrine IL-2 transpresentation may play a particularly important role at the checkpoint because most CD4 effectors do not express CD25, and that it helps by driving greater effector survival and better memory formation. We suggest that influenza infection also provides pathogen recognition signals that continue to activate APCs until virus is cleared, enhancing CD25 expression, as well as MHC-II and costimulatory ligand expression, so that during cognate interaction, APCs efficiently present both peptide Ag and CD4 effector-produced IL-2. All the results together illustrate the dominant role of IL-2 availability in supporting memory generation from effectors at the checkpoint.
We suggest that this set of mechanisms evolved to require that memory develops best only when infection is still ongoing at the effector stage and can therefore provide high-avidity peptide Ag at high doses and high levels of pathogen recognition signals to promote optimum effector transition to memory, especially in the lung, which is the site of infection. Thus, effective vaccine strategies likely need to provide such high-dose, high-avidity Ag and pathogen recognition signals during the T cell effector phase, something nonreplicating or nonpathogen Ags are unlikely to do.
In adult humans with a long history of exposure to influenza viruses, many responses likely stem from existing memory cells (48–51). We find that efficient generation of secondary memory also requires effector checkpoint Ag recognition and is increased by high-avidity interactions (Fig. 6). This contrasts with observations that memory responses in general are less dependent on Ag dose and costimulation (48). We suggest that effector functions of memory cells are more easily achieved, but that forming new secondary memory is again under stringent regulation to avoid unnecessary memory cells to nonpathogens and to select only those with high affinity for persistent Ag only when infection is ongoing.
The impressive impact of Ag avidity for TCR on the size of the memory pool from both primary (Fig. 2) and secondary (Fig. 6) CD4 effectors strongly suggests that both a high dose of available peptides and CD4 effectors bearing TCR with high affinity for some of those peptides are strictly required for robust memory generation. This implies that immunization with a wide range of viral proteins, including those with known immunodominant CD4 epitopes, will be more effective than single proteins in inducing memory. Because of the heterogeneity of human HLA molecules, only a fraction of total potential viral epitopes will be immunodominant in a given individual, which further argues for vaccines expressing a wide breadth of proteins. A broader repertoire of high-affinity memory CD4 T cells should lower the likelihood that escape variants, which arise by random mutations and selection, will have the opportunity to develop and escape so that they can be passed on, because this would require multiple mutations. Many of the immunodominant T cell epitopes, such as the FluNP NP311 used here, are in core proteins of viruses, not in the viral surface proteins that B cells recognize; thus, it follows that vaccines should include core as well as surface protein epitopes to elicit both T and B cell immune memory, including heterosubtypic determinants not likely to be selected by Ab to external surface proteins. An advantage of vaccines is that they can supply these signals in a less dangerous context than infection. Additional studies are needed to support these latter implications in detail, but we suggest that the evolutionary advantage of this requirement for a high-avidity interaction at the effector stage is to generate memory cells with higher affinity for the infecting virus while not allowing memory when virus does not persist at high levels into the effector stage or when there is no replicating infectious entity.
The authors have no financial conflicts of interest.
This work was supported by the National Institutes of Health grants R01AI118820, R21AI153120, and R21AI146532 (S.L.S.); U19 AI109858 (L.J.S. and E.S.H.); T32AI007349-30 (M.C.J. and O.A.K.-U.); T32 AI132152 (O.A.K.-U.); R21 AI146647 (K.K.M.); and R21 AI117457 (T.M.S.).
The online version of this article contains supplemental material.
Susan L. Swain is a Distinguished Fellow of AAI.
bone marrow-derived dendritic cell
draining lymph node
influenza A virus
peptide–MHC class II
surface plasmon resonance
T follicular helper cell
regulatory T cell
University of Massachusetts Chan Medical School