Although memory CD4 T cells are critical for effective immunity to pathogens, the mechanisms underlying their generation are still poorly defined. We find that following murine influenza infection, most effector CD4 T cells undergo apoptosis unless they encounter cognate Ag at a defined stage near the peak of effector generation. Ag recognition at this memory checkpoint blocks default apoptosis and programs their transition to long-lived memory. Strikingly, we find that viral infection is not required, because memory formation can be restored by the addition of short-lived, Ag-pulsed APC at this checkpoint. The resulting memory CD4 T cells express an enhanced memory phenotype, have increased cytokine production, and provide protection against lethal influenza infection. Finally, we find that memory CD4 T cell formation following cold-adapted influenza vaccination is boosted when Ag is administered during this checkpoint. These findings imply that persistence of viral Ag presentation into the effector phase is the key factor that determines the efficiency of memory generation. We also suggest that administering Ag at this checkpoint may improve vaccine efficacy.
Although much progress has been made in defining the early activation events required for the generation of effector CD4 T cell subsets, the pathways that drive a cohort of effector T cells to successfully transition to a memory state remain poorly defined. It is unclear to what extent programming during initial cognate interaction of T cells with APC determines the fate of effector T cells and whether later signals affect memory generation.
Various models defining the role of Ag in effector and memory differentiation have been proposed. Some suggest that the initial interaction with Ag/APC is sufficient to program a cohort of T cells to become memory and further exposure to Ag and inflammation drives terminal differentiation of effector T cells (1–5). In contrast, other studies suggest that late Ag enhances the function, but not the number, of memory CD8 T cells (6, 7). It was shown that CD4 T cells require more Ag stimulation for effector and memory generation than do CD8 T cells, but most of these analyses were limited to the priming phase of the response (8–11). Other studies concluded that, although prolonged Ag stimulation can enhance effector CD4 T cell proliferation, it is deleterious to memory formation (12), and continuous Ag stimulation may drive CD4 T cells to a state of reduced responsiveness (13, 14). In vivo, responding T cells disengage from APC 24 h after initial interaction, engaging in few APC contacts during the last phase of priming (9, 15). Thus, it remains unclear how often responding CD4 T cells encounter Ag after the initial priming phase of the response and whether later Ag exposure impacts memory generation.
During a response to a live pathogen, it would be advantageous for the quality and quantity of the effector and memory response to be determined at the effector stage when the immune system could sense whether there is still a threat. In earlier studies, we found that in vitro–generated effector CD4 T cells are programmed to undergo activation-induced cell death but can be rescued by addition of IL-2 and TGF-β (16). In an in vivo model of influenza A virus (IAV) infection, we recently found that autocrine IL-2 production by effector CD4 T cells during a defined memory checkpoint (days 5–7) of the response was essential to promote survival and memory formation (17). These findings help to explain studies with lymphocytic choriomeningitis virus, in which IL-2 complexes added late in the response promoted memory formation (18). Because IL-2 production is typically induced by cognate Ag recognition, in this study we investigated whether the interaction of effector CD4 T cells with APC during this checkpoint is the key event that drives them to make IL-2, to survive, and to differentiate into long-lasting memory cells. A defined stage of effector CD4 T cell development, during which CD4 effector fate is determined by cognate Ag interaction, would suggest a new paradigm in which the formation of memory depends on a cohort of cells being selected by persisting Ag to become memory cells. If Ag presentation during the effector phase determines memory, it might explain why most vaccines lacking live organisms induce much less durable immunity than infection. In this study, we ask whether late recognition of Ag on APC is the signal that initiates the transition of CD4 effectors to memory cells at this memory checkpoint, and we define the key parameters that are involved.
In most previous in vivo studies, it was not possible to define the necessary timing and duration of the signals needed for the rescue of effectors from apoptosis and excessive contraction. Additionally, as T cells reach the effector stage, the roles that ongoing infection play in promoting memory have not been definitively examined in an in vivo model of infection. Defining these elements is critical for rational vaccine design.
To address these gaps in our understanding, we use a well-defined model of IAV infection to determine the role that Ag presentation and ongoing infection, during the effector phase, play in shaping memory CD4 T cell formation. IAV induces a highly protective memory CD4 T cell population that synergizes with B cells and CD8 T cells to provide protection from challenge with supralethal viral doses (19–22). Thus, the response epitomizes successful memory CD4 T cell generation in response to infection and, therefore, is well-suited to reveal the mechanisms involved in effective memory generation.
We find that effector CD4 T cells, induced by IAV infection, require cognate Ag recognition at 6 d postinfection (dpi) for continued expansion, survival, and all but a minor fraction of memory generation. In well-controlled adoptive-transfer models, we find that Ag/APC encounter at the effector stage (6 dpi) enhances the recovery of memory cells in secondary lymphoid organs and in the lung ≥10–100-fold. Notably, other infection-induced effects, such as inflammation, are not required for this increased memory generation. Effector T cells, exposed to Ag/APC for as little as 2 d, expressed higher levels of memory-associated molecules CD25 (IL-2Rα), B cell lymphoma 6 (Bcl-6), CD127 (IL-7Rα), and CXCR3. The memory cells generated by Ag encounter between 6 and 8 dpi had enhanced ability to make cytokines and provided better protection against a lethal dose of IAV than did those that were not exposed to checkpoint Ag. Moreover, in a cold-adapted (ca) vaccine model, we found very little Ag presentation during this late checkpoint; however, when additional Ag/APC were introduced at this time, memory CD4 T cell formation was enhanced. This suggests that low levels of Ag presentation from 6 to 8 dpi may limit the efficacy of vaccines that do not provide high levels of persisting Ag. These findings imply that whether pathogen infection persists into the effector stage determines effector fate by supplying late Ag/APC that are needed to program memory formation and that interventions to achieve this need not involve long-lived infection and its potentially deleterious effects.
Materials and Methods
Naive CD4 T cells were isolated from OT-II.Thy1.1+/−, OT-II.Nr4a1eGFP.Thy1.1+/−, OT-II.Bcl2l11+/−, or OT-II.Osb.eGFP mice bred at the University of Massachusetts Medical School (UMMS) breeding facility. Hosts were C57BL/6 (B6) male mice ordered from The Jackson Laboratory. Nr4a1eGFP (Nur77GFP) mice originally obtained from The Jackson Laboratory and bred at the UMMS breeding facility were also used. Mice used in experiments were 8–12 wk of age. The Institutional Animal Care and Use Committee of the UMMS approved all animal procedures.
Viral stocks, infections, and immunizations
For all influenza viral infections described, mice were lightly anesthetized with isoflurane (Piramal Healthcare) before intranasal infection with 50 μl of virus diluted in PBS. Influenza A/Puerto Rico/8/34-OVA323–339 (PR8-OVAII) and A/Puerto Rico/8/34 (PR8; H1N1) viruses were produced in the allantoic cavity of embryonated hen eggs from stock obtained from Dr. Peter Doherty of St. Jude Children’s Hospital. A sublethal dose of 0.3 LD50 was used. Protection experiments were performed using a lethal dose of 2 LD50. Attenuated ca.A/Alaska/72/CR9 (ca.Alaska) (H3N2) was originally supplied by S. Epstein (National Institutes of Health, Bethesda, MD) and then grown at the Trudeau Institute (23). Mice were immunized with 2500 TCID50 ca.Alaska, a dose shown to elicit T cell–mediated protection (23). Influenza A/Philippines/2/82/x-79 (H3N2) was also supplied by S. Epstein. Mice were infected with 100 PFU.
Naive CD4 T cell isolation and effector generation in primary hosts
Spleens and peripheral lymph nodes were harvested from TCR-transgenic or wild-type (WT) mice that were 6–10 wk old. Resting cells were enriched using a Percoll gradient. CD4 T cells were isolated using CD4 MACS beads (Miltenyi Biotec). Naive CD4 T cells were washed twice and resuspended in PBS, and a total of 3 × 105–5 × 105 cells was transferred by i.v. injection into hosts. Hosts were infected with PR8-OVAII on the same day.
Isolation of 6 dpi effector CD4 T cells
Spleen and lung-draining lymph nodes (DLN) were harvested from B6 mice on day 6 after PR8-OVAII infection. Cell suspensions were pooled, and donor cells were isolated by Thy1.1 or CD4 MACS isolation (Miltenyi Biotec). Cells were resuspended in PBS, and 1–4 × 106 effector cells were transferred by i.v. injection to hosts. All steps were conducted at room temperature (with the exception of one 15-min incubation at 4°C) to maintain effector phenotype. This minimal protocol ensures that effector cells are only out of mice for 2.5 h.
Bone marrow dendritic cell preparation
Bone marrow was harvested from B6 mice and washed with RPMI 1640 including 1% FBS. Cells were plated at 7–8 × 106 cells per milliliter in RPMI 1640 with 7.5% FBS including 10 ng/ml GM-CSF. After 7 d, cells were harvested, and CD11c+ cells were isolated by MACS. Purified cells were replated at 2 × 106 cells per milliliter, stimulated with polyinosinic-polycytidylic acid at 10 μg/ml for 1 d in culture, and used as dendritic cells (DC). DC were harvested and pulsed with 10 μM OVA323–339 (OVAII)- or NP311–325-peptide at 37°C for 1 h with shaking. Cells were resuspended in PBS, and 3–5 × 105 cells were injected i.v. per mouse.
PR8-infected splenic APC preparation and in vitro culture
Spleens from PR8-infected B6 mice were harvested at 6 dpi. Cell suspensions were pooled and washed with RPMI 1640 containing 1% FBS. Cells were depleted of Thy1.2+ cells using MACS beads. Cells were irradiated with 3000 rad. This APC population was then cocultured with isolated 6 dpi effectors at a 5:1 ratio of APC/T cell. OVAII- or NP311–325-peptide was added to culture at 0.5 μM. IL-7 was added to cultures at 0.1 ng/ml (a concentration that does not promote proliferation). All blocking Ab were used at 10 μg/ml.
APC for protection experiment
Spleen cells were harvested from uninfected B6 mice. Thy1.2+ cells were depleted using MACS beads. The Thy1.2-depleted fraction was plated at 3 × 106 cells/ml in RPMI 1640 containing 7.5% FBS, 10 ng/ml LPS, and 10 ng/ml dextran sulfate. After 2 d in culture, these activated APC-enriched cells were harvested and pulsed with 10 μM OVAII-peptide at 37°C with shaking for 1 h. APC were transferred to hosts with 6 dpi effectors at a 1:1 ratio.
B6 mice were infected with 0.3 LD50 PR8 or PR8-OVAII. Lungs were harvested at 6 dpi and fixed in 10% buffered formalin. Ten-micrometer sections were taken and stained with H&E. Lungs were scored as follows: 1, healthy-looking bronchioles with consolidation and mononuclear infiltrates making up <5% of the lung; 2, mild bronchiolitis with consolidation and mononuclear infiltrates making up >5% of the lung; 3, moderate bronchiolitis with consolidation and mononuclear infiltrates making up ≥15% of the lung; 4, moderate bronchiolitis with consolidation and mononuclear infiltrates making up ≥25% of the lung; and 5, severe bronchiolitis with consolidation and mononuclear infiltrates making up >50% of the lung. Scoring was done blind, four sections of each lung were scored, and the average is presented.
Viral titers of PR8- or PR8-OVAII–infected lungs were determined by quantification of viral RNA. Whole lungs were homogenized in TRIzol/Chloroform (Sigma-Aldrich), and RNA was extracted using the VWR E.Z.N.A kit and TURBO DNA-free Kit (Thermo Fisher Scientific). A total of 2.0 μg of RNA was reverse transcribed into cDNA using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). Quantitative PCR was performed to amplify the acidic polymerase (PA) gene using the Bio-Rad CFX96 Real-Time PCR Detection System, with 50 ng of cDNA per reaction. The following primers and probe were used: forward primer: 5′-CGGTCCAAATTCCTGCTGA-3′; reverse primer: 5′CATTGGGTTCCTTCCATCCA-3′; and probe: 5′-6-FAM-CCAAGTCATGAAGGAGAGGGAATACCGCT-3′. Data were analyzed using CFX Manager Software Version 20 (Bio-Rad). A standard curve generated using a PA-containing plasmid obtained from Dr. Rob Webster (St. Jude’s Children’s Research Hospital) was used to calculate the PA gene copy number per 50 ng of cDNA. This was used to calculate the total PA copy number per lung.
Flow cytometry and cytokine staining
For cytokine staining, total splenocytes were stimulated with PMA and ionomycin for 4 h at 37°C. Brefeldin A (10 μg/ml) was added after 2 h of stimulation. Following a surface stain, cells were fixed in 4% paraformaldehyde and permeabilized in 0.1% saponin for 30 min at 4°C. Cytokines were then stained for 30 min at 4°C. Bim, Bcl-2, Ki67, Bcl-6, and T-bet were stained using the eBioscience Foxp3 staining buffer kit, following the manufacturer’s recommendations. Bim Ab was stained with a fluorescent Goat anti-Rabbit Ab from Invitrogen. Host IAb-NP311–325–specific CD4 T cells were stained with the IAb-NP311–325–allophycocyanin tetramer obtained from the National Institutes of Health Tetramer Core Facility. All Ab were obtained from eBioscience, with the exception of anti-Bim (Cell Signaling) and anti-Bcl-6–PE (BD Biosciences). Gating strategy includes gating on single cells, lymphocytes, and live cells distinguished by Invitrogen LIVE/DEAD cell viability dye. Samples were run on an LSR II instrument (BD Biosciences), and analysis was done using FlowJo (TreeStar) software.
Groups of at least three mice were used for all experiments to ensure sufficient power. For analysis comparing more than two samples, a one-way ANOVA analysis was conducted with GraphPad Prism software. To compare two samples, an unpaired, two-tailed Student t test was conducted with GraphPad Prism software. All data were included unless found to be a significant outlier using the Grubbs’ test (Extreme Studentized Deviate method) available through GraphPad Prism software. Welch’s correction was applied when the SD were unequal.
Ag recognition at the effector phase of the CD4 T cell response is limited
To determine when responding CD4 T cells encounter Ag in vivo following IAV infection, we crossed OVA323–339 [OVAII]-specific TCR-transgenic mice (OT-II.Thy1.1+/−) with Nur77GFP mice that transiently express GFP following TCR stimulation (24, 25). To evaluate the feasibility of using Nur77GFP as an indicator of recent Ag-induced TCR stimulation in effector T cells, we isolated CD4 T cells from Nur77GFP mice and stimulated them in vitro. GFP expression was rapidly induced and remained high with continued TCR stimulation (Fig. 1A) but was significantly reduced within 24 h following removal of stimulation (Fig. 1B). Additionally, GFP was rapidly re-expressed following secondary exposure to Ag (Fig. 1C) and did not decrease with division (Fig. 1D).
To determine the kinetics of IAV Ag recognition in vivo, we transferred naive OT-II.Nur77GFP.Thy1.1+/− cells to B6 mice and infected them with a sublethal dose of PR8-OVAII (Fig. 1E, 1F). As expected, during priming (3 dpi) most cells were GFP+, indicating recent Ag exposure. However, by 5 dpi, only a fraction of effector CD4 T cells had recently encountered Ag, and by 9 dpi (the peak of the lung effector T cell response) very few cells expressed GFP (Fig. 1E, 1F). This was true of OT-II cells, as well as polyclonal host cells, following IAV infection of Nur77GFP mice (Fig. 1F). Thus, effector CD4 T cells only intermittently respond to cognate Ag in vivo, and the Ag recognition that does occur is mostly limited to just before the peak of T cell effector response corresponding to the memory checkpoint. These findings suggest that Ag recognition at the effector stage could act to select a limited number of effectors to become memory.
Ag recognition at the effector phase is required for memory formation
We next asked whether Ag recognition during the effector phase had any effect on memory generation. For this, we performed a sequential adoptive-transfer experiment as described in Fig. 2A. We first transferred naive OT-II.Thy1.1+/− cells to B6 mice and infected them with a sublethal dose of PR8-OVAII. At 6 dpi, donor OT-II.Thy1.1+/− effector cells were isolated from the secondary lymphoid organs of IAV-infected hosts. These 6 dpi effectors were fully activated (Supplemental Fig. 1). Donor cells were transferred into three groups of recipients, also infected 6 d previously with PR8-OVAII (Ag and virus), PR8 (virus without Ag), or no virus (Fig. 2A). We did not include lung effector T cells because they are more likely to have recently encountered Ag (Fig. 1). The kinetics of endogenous CD4 T cell responses to PR8-OVAII and PR8 viruses are the same (Supplemental Fig. 2A), indicating that, with respect to factors that govern the CD4 T cell response, these viruses are very similar. Additionally, viral titer kinetics and lung pathology are similar in PR8-OVAII– and PR8-infected mice (Supplemental Fig. 2B–D).
We enumerated donor cells in the lung, spleen, and DLN at 3, 7, and 14 d posttransfer (dpt). At 7 dpt, there were 60–200-fold more donor OT-II cells in the lung, 15–30-fold more in the spleen, and 80–400-fold more in the DLN of PR8-OVAII-infected hosts compared with PR8-infected or uninfected hosts, which were equally poor in supporting donor cell recovery (Fig. 2B, 2C). In PR8-OVAII hosts, donor OT-II numbers peaked at 3 dpt (9 dpi) and then contracted slowly over the subsequent 12 d (Fig. 2D, Supplemental Fig. 3), mimicking the endogenous CD4 T cell response (26). However, in PR8-infected and uninfected hosts, donor cells underwent a sharp, immediate contraction and by 14 dpt (20 dpi) were reduced to close to the limit of detection (Fig. 2D, Supplemental Fig. 3). A highly significant difference in memory recovery was still seen at 53 dpt (Fig. 2E). These results imply that re-exposure to Ag at or after 6 dpi is necessary to maximize the effector CD4 T cell response, prevent excessive contraction, and generate a long-lived memory population and that infection without Ag has little, if any, impact on memory formation.
Some investigators reported that late Ag promotes increased effector expansion but leads to exacerbated contraction, resulting in fewer or similar numbers of long-lived memory cells (6, 12). To determine whether the increased number of donor cells in the PR8-OVAII–infected hosts was the result of an extended expansion of short-lived effectors, we assayed the size and phenotype of donor cells at 2 and 8 dpt (Fig 2F, 2G). At 2 dpt, the donor cells were large in size (Fig. 2F), with high expression of effector markers ICOS and PD-1 (Fig. 2G), but by 8 dpt they were small (Fig. 2F) and had downregulated ICOS and PD-1 (Fig. 2G). Thus, by 8 dpt donor cells no longer had an effector phenotype and had mostly transitioned to resting cells.
To test whether the ability of Ag recognition to promote memory is transient or persists to later time points, we isolated donor OT-II effectors at 14 dpi instead of 6 dpi and transferred them to kinetically matched PR8-OVAII–infected, PR8-infected, or uninfected hosts. The presence of Ag in the hosts had little or no impact on the recovery of these 14 dpi donor cells (Fig. 2H), indicating that, at this time, CD4 T cells are no longer dependent on Ag recognition. Therefore, we postulate that the checkpoint is defined by a unique differentiation state of effectors during which they require cognate interaction for survival, but after which their fate has mostly been determined so that they no longer require TCR triggering.
A short duration of Ag presentation at the checkpoint is sufficient to restore memory formation
Given that re-encounter with Ag was required at 6 dpi but not at 14 dpi, we tested whether a short exposure of donor cells to Ag might be sufficient to induce memory formation. We transferred 6 dpi OT-II donor cells to PR8-infected mice and asked whether i.v. injected bone marrow–derived DC pulsed with OVAII-peptide (DC-OVAII) during this memory checkpoint would be sufficient to restore memory formation. We found that these DC present Ag for no longer than 2 d after transfer in vivo by tracking their ability to induce proliferation of naive OT-II cells (Supplemental Fig. 4A, 4B). The Ag signal provided by DC-OVAII did not appear to be excessive, because OT-II cell numbers in groups with DC-OVAII were similar to those seen in hosts infected with PR8-OVAII at 2 dpt (Supplemental Fig. 4C). Strikingly, the donor cells transferred to PR8-infected hosts that received DC-OVAII were recovered at similar levels as those transferred to PR8-OVAII–infected hosts out to 14 dpt (Fig. 3A, 3B). This indicates that encounter with cognate Ag for ≤48 h, starting at 6 dpi, was sufficient to prevent excessive contraction and promote memory formation.
Virus infection is not required at the memory checkpoint
To determine whether viral infection itself is important in promoting memory formation, other than providing Ag presentation, we tested whether adding DC-OVAII similarly increased memory formation in uninfected hosts. Strikingly, DC-OVAII strongly promoted donor recovery to a similar extent in PR8-infected and uninfected hosts (Fig. 3C). These data, combined with Fig. 1, in which there was no difference in memory formation following transfer of 6 dpi effectors into PR8-infected and uninfected hosts, suggest that aspects of infection other than providing Ag presentation, such as induction of lung inflammation, have no discernable impact at the memory checkpoint. The DC that we used were activated, so infection-induced viral-sensing pathways may be needed to activate in situ APC. The donor OT-II effectors exposed to Ag in PR8-infected or uninfected hosts downregulated effector molecules PD-1 and ICOS by 7 dpt (Fig. 3D), suggesting their loss of effector phenotype.
We further examined whether memory formation occurred normally when late Ag was provided by short-lived DC-OVAII. We directly compared OT-II memory generated when naive cells were transferred on day 0 and left in the same initial host or when 6 dpi OT-II effectors were isolated and transferred to kinetically matched PR8-OVAII–infected, PR8-infected with DC-OVAII, or uninfected hosts with DC-OVAII. To highlight the changes that distinguish memory cells from effectors, we included 6 dpi OT-II effectors for comparison.
One functionally important characteristic of memory cells is their ability to produce multiple cytokines upon restimulation (19). We found that the memory cells generated following transfer (to hosts with virally produced Ag or with short-lived Ag provided by DC-OVAII) regained the ability to produce multiple cytokines to a similar extent as those generated without transfer (Fig. 3E, 3F). Additionally, memory cells generated with and without transfer upregulated the critical memory marker CD127 that is necessary for their persistence (Fig. 3F, Supplemental Fig. 4D). Interestingly, when comparing memory cells generated in PR8-infected or uninfected hosts with late transfer of DC-OVAII, there was a decrease in CD127 expression in the uninfected hosts (Fig. 3F, Supplemental Fig. 4D). This suggests that, although systemic virus–induced inflammation may not be needed for memory cell numbers, function, or subset differentiation, virus-induced inflammation may be required for full CD127 upregulation. However, the equivalent recovery of memory cells argues that DC-OVAII exposure induced sufficient levels of CD127 for persistence.
We next examined CD4 memory subset differentiation. The tissue-resident memory population identified by CD69 expression (27, 28) in the lung was similar with and without transfer (Fig. 3G). IFN-γ production, an indicator of Th1 differentiation (29, 30), was also similar in all memory groups (Fig. 3G). CXCR5 was shown to mark a memory subset that is thought to be T follicular helper–like or T central memory–like (31, 32). CXCR5 expression was also similar among all memory groups (Fig. 3G, Supplemental Fig. 4E). Therefore, the limited Ag provided by DC-OVAII at 6 dpi is sufficient to generate canonical memory formation.
Effector CD4 T cells require Ag recognition at the checkpoint for continued proliferation
Some studies suggested that effector CD4 T cell division is programmed by initial Ag encounter (33, 34), whereas others suggested that CD4 T cells do not undergo such “autopilot” proliferation after 2 d of stimulation during priming (10); however, it remains unclear whether they acquire this ability later during infection. To determine whether division past 6 dpi depends on Ag recognition, we labeled isolated 6 dpi effectors with CFSE, transferred them to hosts with and without Ag, and assayed dilution of dye at 3 dpt. Only donor cells in hosts with Ag divided more than once (Fig. 4A). To determine whether this proliferation was an artifact of the transfer system, we used Ki67 staining to compare the proliferation of donor OT-II cells with that of endogenous IAb-NP311–325–specific host cells in PR8-OVAII–infected hosts. We found that there was a similar percentage of proliferating donor and hosts cells at 2 dpt; by 8 dpt, neither were undergoing division, a pattern seen in the lung, spleen, and DLN (Fig. 4B, 4C). Thus, division after 6 dpi is Ag dependent, short-lived, and followed by the transition to nondividing cells within a week. This also illustrates that the kinetics of proliferation of the transferred donor cells mimics that of the endogenous host CD4 T cell response to live IAV.
Ag recognition at the effector phase promotes survival of CD4 T cells
After viral clearance, most effector T cells undergo apoptosis leading to contraction, whereas a cohort survives to become memory. This suggests that avoiding apoptosis is a key step in the transition to memory. We propose that a cohort of effector CD4 T cells recognize Ag/APC, which drives them to make and respond to IL-2, which drives their survival and supports their transition to memory (17). We evaluated several components of this hypothesis.
To test whether Ag recognition at the checkpoint promoted enhanced survival of effector CD4 T cells, we transferred naive OT-II.Nur77GFP cells to hosts and infected them with PR8-OVAII. At 7 dpi, donors that had seen Ag during the first 1–2 d of the checkpoint were GFP+, whereas those that did not were GFP−. We analyzed donor CD4 T cells from the lung, spleen, and DLN directly ex vivo, gating on GFP+ and GFP− cells. To detect cell death directly ex vivo, we measured 7-aminoactinomycin D (7-AAD) staining (Fig. 5A). In each organ, 7-AAD staining was significantly greater in GFP− cells than in GFP+ cells, indicating that more effector cells that recognized Ag between 5 and 6 dpi survived than those that did not recently encounter Ag.
To further dissect the mechanisms involved in this survival, we developed an in vitro model to better control the signals that the effectors receive. We isolated 6 dpi OT-II effectors and cocultured them with T-depleted splenocytes isolated from PR8-infected mice, a physiologically relevant APC, with or without OVAII-peptide. To mimic the short-term Ag presentation that occurs in vivo, we irradiated the APC, ensuring that Ag presentation was restricted to the first 2 d of culture (Fig. 5B) (5). In this model, we found that Ag enhanced cell recovery (Fig. 5C) and promoted expression of CD127, which is strongly associated with memory cell survival (17) (Fig. 5D). After 2 d of culture, effector CD4 T cells exposed to Ag had decreased 7-AAD staining, indicating reduced cell death (Fig. 5E), as well as reduced levels of Bim, a proapoptotic protein known to mediate death during T cell contraction (35, 36) (Fig. 5F). Thus, instead of inducing widespread cell death, Ag presentation to in vivo–generated 6 dpi effectors drove effector survival.
Ex vivo recognition of Ag enhances survival via downregulation of Bim
We tested whether the reduced level of Bim seen in the Ag-exposed effectors was responsible for their increased survival. We cotransferred WT GFP Bcl2l11+/+ or Bcl2l11+/− [which express half the WT levels of Bim (37)] OT-II cells at a 1:1 ratio into B6.Thy1.1+/− mice and infected them with PR8-OVAII. We harvested total effector CD4 T cells at 6 dpi and stimulated them ex vivo with APC, with or without OVAII-peptide, and determined which donor cells preferentially survived after 14 d. When no Ag was present in vitro, the Bcl2l11+/− OT-II cells survived much better than did WT OT-II cells (Fig. 5G), implicating the high levels of Bim in the death/contraction of the 6 dpi effectors. In contrast, in the presence of Ag, the Bcl2l11+/− OT-II and WT OT-II cells survived comparably (Fig. 5G), consistent with the hypothesis that Ag counteracts apoptosis by causing Bim reduction. Indeed, in the absence of Ag, the Bcl2l11+/− OT-II cells expressed less Bim than WT, but with Ag, Bim levels were similar (Fig. 5H). This supports the hypothesis that Ag recognition by effectors at the checkpoint acts, in part, through reduction of Bim expression, which prevents short-term apoptosis and promotes survival.
The prosurvival effects of Ag recognition at the checkpoint are IL-2 dependent
Because our previous studies found that autocrine IL-2 was required for CD4 effector survival (17), we tested whether Ag stimulation of 6 dpi effectors ex vivo would promote IL-2 production and whether that IL-2 was necessary for enhanced survival. Indeed, the ex vivo effector CD4 T cells produced IL-2 only when cultured with Ag/APC (Fig. 5I). We cultured 6 dpi effectors with APC, Ag/APC, or Ag/APC plus Ab specific for CD25 (IL-2Rα) and CD122 (IL-2Rβ) to block IL-2 function. The exposure to Ag/APC enhanced donor cell recovery after 6 d, and blocking IL-2 signaling reduced that recovery (Fig. 5J). Notably, blocking IL-2 inhibited Ag/APC-induced proliferation only slightly (Fig. 5K), but it increased cell death dramatically, as measured by 7-AAD staining (Fig. 5L). Thus, in vitro Ag/APC stimulation of 6 dpi effectors induces IL-2 production that prevents apoptosis and enhances the survival necessary for memory formation. Partial effects seen on cell recovery (Fig. 5J) and cell proliferation (Fig. 5K) imply that factors beyond IL-2 also play a role in the effects of Ag seen at the memory checkpoint.
Ag recognition at the checkpoint promotes expression of a memory phenotype
Because Ag/APC exposure of 6 dpi effectors at the checkpoint promotes formation of a larger cohort of memory cells, as opposed to driving terminal differentiation, we asked whether it also promoted expression of known memory-associated markers. We transferred OT-II.Nur77GFP.Thy1.1+/− naive cells to hosts, infected them with PR8-OVAII, and harvested lung, spleen, and DLN at 7 dpi. We analyzed donor GFP+ cells (recent Ag exposure) versus GFP− cells (no recent Ag exposure). The GFP+ cells expressed higher levels of CD25 at 5–7 dpi in the lung compared with GFP− cells (Fig. 6A), consistent with the role of IL-2 postulated above. The 7 dpi GFP+ cells in lung, spleen, and DLN also expressed higher levels of Bcl-6, a transcription factor implicated in memory formation (38, 39) (Fig. 6B), whereas expression of T-bet, which is thought to promote terminal differentiation (40–42), was equivalent (Fig. 6C). We also cultured OT-II donors and polyclonal effector CD4 T cells with APC pulsed with OVAII or NP311–325 as respective cognate peptides. After 2 d, Ag/APC induced OT-II and NP311+ cells to upregulate CD25 and Bcl-6 compared with culture with APC alone (Fig. 6D, 6E). STAT3 signaling promotes the transcription of Bcl-6 and may promote memory formation (43, 44). We found exposure of 6 dpi CD4 effectors to Ag/APC significantly increased p-STAT3 after 4 h (Fig. 6F). The ability of Ag/APC to rapidly upregulate these memory signature proteins in vivo and in vitro provides further evidence that Ag recognition at the checkpoint drives effectors to express a comprehensive program that initiates and carries out their transition to memory cells.
Memory cells generated by cognate interactions at the checkpoint have enhanced phenotype, function, and protective ability
In the transfer model (Figs. 2–4), 6 dpi effector cells transferred to hosts without Ag underwent extensive contraction and were often at or below the limit of detection within 7 dpt. This low number of memory cells in hosts without Ag hampered our ability to determine the long-term phenotypic and functional differences between memory cells generated with or without Ag at the checkpoint. To increase the recovery of memory cells that develop without Ag at the checkpoint, we cultured in vivo–generated effector CD4 T cells with or without Ag for 2 d in vitro (as in Fig. 5), transferred equal numbers of each to uninfected mice, and allowed the cells to transition to memory for 7 d. In vivo, 3 d without Ag is sufficient for effector CD4 T cells to become virtually identical to memory (45). We then assayed cell recovery, phenotype, and cytokine production.
As expected, the donor cells that had been exposed to Ag/APC in vitro formed a significantly larger memory population after transfer to uninfected hosts even though their numbers were equivalent at the time of transfer, with 18-fold more in lung and 5-fold more in spleen (Fig. 7A). This indicates that the 2 d in vitro exposure to Ag was sufficient to confer significantly greater survival. Compared with APC without Ag, the donor effector cells exposed to Ag/APC in vitro expressed increased levels of CD127 and CXCR3, a memory marker needed for homing and protective function (46, 47) (Fig. 7B). Moreover, they secreted more IFN-γ and had a higher frequency of IFN-γ/TNF-α double producers after restimulation (Fig. 7C, 7D). These results indicate that even short-term Ag recognition at the 6 dpi checkpoint results in a much larger and a functionally superior memory population.
To evaluate whether the differences in memory formation with or without Ag at the checkpoint would lead to differences in protection against a lethal challenge of IAV, we transferred 6 dpi OT-II effectors to uninfected B6 mice along with OVAII-pulsed or unpulsed APC. To account for a potential host-naive CD4 T cell response, we included a group of mice that received OVAII-pulsed APC without transfer of 6 dpi effectors. We also included a group that received naive OT-II cells to control for the possibility that a similar number of naive donor OT-II cells could provide enhanced protection. Hosts were rested for 2–3 wk to ensure memory generation and then challenged with a lethal dose of PR8-OVAII.
Despite the fact that the only memory cells in the hosts were the donor 6 dpi effectors, the hosts that received effectors plus APC-OVAII were mostly protected against lethal infection (12/15), whereas those that received naive OT-II, OT-II 6 dpi effectors without Ag, or APC-OVAII alone were largely unprotected (Fig. 7E). Thus, providing effectors with only short-term in vivo Ag stimulation at the checkpoint drove the formation of protective memory cells. Because the hosts were not previously infected, we conclude that short-term Ag stimulation by activated APC, without any viral infection, is sufficient to promote the transition of 6 dpi effectors to become protective memory.
Ag during the checkpoint boosts memory in a ca vaccine model
Our findings establish a checkpoint that occurs at 6–8 dpi following IAV infection where Ag recognition drives functional memory CD4 T cell formation. Because many standard vaccinations likely do not induce the persistent levels of Ag that live virus does, we postulate that memory CD4 T cell formation following vaccination is normally constrained by a lack of Ag at the checkpoint. Therefore, the addition of Ag/APC at this time may enhance vaccine-induced memory. To test this premise, we immunized with a live attenuated, ca influenza A virus (ca.IAV). Replication of ca.IAV is limited to the upper respiratory tract, potentially limiting the duration of Ag presentation. The ca.IAV vaccines were shown to induce enhanced T cell responses compared with inactivated vaccines (48). To determine whether Ag persisted into the memory checkpoint following ca.IAV inoculation, we immunized Nur77GFP mice with ca.A/Alaska/6/77CR29 (ca.Alaska) and measured GFP expression in immunization-induced effector T cells. Our earlier studies showed that ca.Alaska induces a strong heterosubtypic response to PR8 and that NP311–325 is a dominant CD4 epitope shared between these two viruses (23). At 7 dpi, effector NP311–325-specific cells expressed no GFP after ca.Alaska immunization, indicating no recent Ag recognition (Fig. 8A), whereas in mice infected with PR8 or a non-ca H3N2 strain (A/Philippines/2/82/x-79), a cohort of NP311–325+ cells were GFP+, indicating recent Ag recognition in the live infections.
To determine whether the addition of Ag during the checkpoint could boost memory following ca.Alaska immunization, we added NP311–325-pulsed APC at 6 dpi to ca.Alaska-immunized mice and assayed memory CD4 T cell formation by enumerating NP311–325 tetramer+ cells at 33–44 dpi (Fig. 8B). We found significantly more NP311–325+ CD4 T cells in the lung and spleen, although there was no difference in the small number of donors found in the DLN (Fig. 8C). This finding suggests that the memory checkpoint exists for effectors generated by attenuated, as well as live, WT influenza infection. Additionally, it shows that the introduction of Ag/APC at the checkpoint can promote effector CD4 T cells induced by attenuated virus immunization to form more memory without the need for persisting live virus.
Cognate interactions during initial priming of naive CD4 T cells affect many aspects of the T cell response, including effector and, some suggest, memory differentiation (3–5). Our findings using a sequential transfer model show unequivocally that Ag recognition is again required at the effector phase to drive mature effector CD4 T cells to become memory. This implies that continuing Ag presentation indicative of pathogen persistence is required for optimum generation of CD4 memory. Our studies define a clear-cut checkpoint during the effector phase, lay out the timing of events that occur, define key pathways that are involved, and show that the memory cells generated are protective.
We show that primary effector CD4 T cells must engage in cognate interactions with Ag/APC at a checkpoint from 6 to 8 dpi. Although the IL-2–dependent checkpoint was 5–7 dpi for primed CD4 effectors in our earlier study (17), in this study we used primary cells, which seem to have a slightly delayed kinetics. This fate-determining effector:Ag/APC interaction induces a limited expansion of effectors, depends on IL-2 production, and leads to increased survival that is mediated, in part, by a downregulation of Bim. Importantly, although Ag persistence during the checkpoint normally depends on continued live pathogen, our transfer model revealed that a short-lived Ag-bearing APC population acting for as little as 2 d at the checkpoint is sufficient for strong memory generation that confers protection and is phenotypically and functionally similar to memory cells generated during viral infection.
We find that Ag recognition at the memory checkpoint initiates a program of memory-associated changes that results in a larger, long-lived memory population with increased CD127 and CXCR3 expression, as well as increased cytokine production. Our Nur77GFP experiments highlight that the early signaling events that occur following Ag recognition at the checkpoint include an upregulation of CD25, Bcl-6, and p-STAT3. CD25 expression is generally heterogeneous at the effector time point, and one recent study found that CD25hi effector T cells present late in the response preferentially form memory (49). Increased expression of IL-2R ensures that the cells that encountered late Ag effectively use the autocrine IL-2 required for memory formation (17). Bcl-6 was recently shown to promote the metabolic switch required for memory formation (38). Therefore, Ag at the memory checkpoint may serve to selectively upregulate Bcl-6 late in the response because cells destined to become memory must transition to a self-renewing, resting population. The regulatory effect of Bcl-6 in Th1 cells was shown to reflect the relative levels of Bcl-6 and T-bet (50). Because no significant increase in T-bet occurred following late Ag stimulation, even a modest increase in Bcl-6 expression may tip the balance in favor of a Bcl-6–mediated gene-expression program. The main known promoter of Bcl-6 expression is p-STAT3, which was also induced. Future work will determine whether TCR stimulation alone promotes Bcl-6 transcription or whether a STAT3-inducing cytokine produced by the responding T cell or the APC is responsible for increased Bcl-6 expression.
The effects of exposure to Ag/APC from 6 to 8 dpi are dramatic, leading to differences in memory recovery that sometimes exceed 100-fold, so there is little question that cognate Ag recognition is critical for transition of effectors to memory. This suggests that effectors generated by live IAV infection reach a stage of differentiation where they will die by default programmed death unless they are rescued by Ag/APC interaction. The donor OT-II CD4 T cells that we followed have a single, high-affinity receptor, so it could be argued that less vigorous infections or T cells with less avid receptors might generate effectors with a less-differentiated phenotype that do not reach the default death stage. However, we point out that the polyclonal response of NP311–325+ cells behaves similarly in our studies and that ca.Alaska IAV also produces effectors that form enhanced memory when they recognize Ag at the checkpoint. Additionally, it is important to consider that less-differentiated effectors will have divided fewer times and will be present in much lower numbers, so their potential contribution to the memory pool is limited. Many viral infections follow a similar time-course of virus accumulation and effector CD4 T cell generation as are seen in the IAV response, so we predict that the checkpoint will be broadly applicable to other acute viral responses.
Although the vast majority of studies that found no role for late Ag in memory T cell formation were CD8+ T cell studies (2, 51–53), some showed similar results for CD4+ T cells during Listeria monocytogenes and vesicular stomatitis virus (VSV) infection (3, 12). In the L. monocytogenes study, ablation of Ag after 48 h via antibiotic treatment had no effect on effector or memory cell numbers. However, a contradictory study demonstrated that truncating L. monocytogenes infection significantly reduced CD4+ memory formation (54). Additionally, because both studies used antibiotics to truncate infection, it cannot be ruled out that lingering Ag presentation may have been present. The VSV study is quite elegant in its use of Ab directed against a specific epitope in the context of MHC class II. In this study, the investigators found that blocking Ag 24 h postinfection had no impact on memory formation. However, this study also found differences in the Ag dependency of CD8+ T cells responding to VSV versus IAV, suggesting that memory requirements for these two pathogens may differ. Additionally, studies using in vitro–generated effector T cells found no role for late Ag (5, 55). However, we have found that in vitro–generated effectors do not mimic the kinetics of endogenous T cell expansion and contraction when transferred to infected mice (B.L. Bautista, unpublished observations). We propose that such in vitro–generated effectors, although useful for some studies, do not adequately mimic effectors generated in vivo by viral infection in terms of their requirements for memory T cell formation.
A requirement of Ag recognition at the effector stage to generate CD4 T cell memory makes teleological sense. First, it would ensure that a substantial memory population is only formed when Ag, indicating a continuing threat, persists. Generation of memory would be undesirable if the pathogen were rapidly cleared. Second, the additional round of Ag-dependent selection may help to select a memory pool with greater multifunctionality. Many recent studies demonstrated that memory CD4 T cells retain a significant level of the differentiation acquired during the effector phase (31, 32, 56, 57). An intriguing hypothesis is that, via the memory checkpoint described in this article, this late Ag interaction may be responsible for the selection and formation of more specialized subsets of effectors that become memory cells particularly tailored to combat the given pathogen upon re-encounter.
It is well established that live infections (and vaccines mimicking them) generate the best immunity (smallpox and others), whereas newer vaccines containing purified proteins with little or no adjuvant induce weak T cell immunity (48, 58). Our results suggest that one key reason why such vaccines may generate poor memory is because they do not induce sufficient persisting Ag presentation during the checkpoint. We tested this using ca.IAV because the attenuated virus initially has the properties of live vaccine; it replicates in the cooler upper respiratory tract but then fails to replicate in the warmer environment of the lower respiratory tract and is, therefore, short-lived. Indeed, by 7 dpi after ca.IAV there was no evidence of Ag presentation and when we introduced Ag/APC at 6 dpi after ca.IAV vaccination, it significantly improved memory CD4 T cell generation. This suggests that strategies to provide Ag/APC at a relevant checkpoint for each vaccine may often enhance memory CD4 T cell formation. Indeed, in another scenario, an early “boost” strategy efficiently promoted CD8 T cell memory (59). Importantly, because we find no need for live virus at the checkpoint, our findings also suggest that an optimal vaccine response could be achieved without the destructive inflammation caused by replicating live virus or systemic adjuvants. Although the exact timing and optimum approach for providing Ag at the checkpoint may need to be tailored to the specific vaccine, we anticipate that such an approach could be developed to improve vaccines in humans.
We thank Dr. Leslie Berg for advice on the manuscript and Dr. Liisa Selin for her advice on scoring lung pathology, as well as members of the Swain laboratory for helpful discussions throughout.
This work was supported by National Institutes of Health Grants P01 AI046539, P01 AI46530, and R01 AI118820 (all to S.L.S.).
The online version of this article contains supplemental material.
Abbreviations used in this article:
ca influenza A virus
bone marrow–derived DC pulsed with OVAII-peptide
lung-draining lymph node
influenza A virus
University of Massachusetts Medical School
vesicular stomatitis virus
The authors have no financial conflicts of interest.