The Development and Function of Memory Regulatory T Cells after Acute Viral Infections

Naturally occurring CD4⁺CD25⁺ regulatory T cells (Tregs) are a cell lineage devoted to the maintenance of homeostasis in the immune system. The natural Treg lineage develops in the thymus and exits to the periphery, where these cells represent 5–10% and 2–4% of the CD4⁺ T cell population in healthy mice and humans, respectively (1, 2). They are phenotypically characterized by the expression of the IL-2R α-chain (CD25), CTLA-4, glucocorticoid-induced TNFR (GITR), and high levels of folate receptor 4 (FR4) (3–5). The most definitive signature of natural Tregs, however, is the expression of the transcription factor Foxp3, which is required for Treg development and function (6–8).

Functionally, Tregs are defined by their ability to suppress the activation of multiple cell types, including CD4⁺ and CD8⁺ T cells, B cells, NK cells, and dendritic cells (9–13). Suppression can be achieved by the production of immunosuppressive cytokines, by direct cell-to-cell contact, or by modulating APCs (14). The suppressive ability of natural Tregs is crucial in maintaining dominant tolerance—the active, trans-acting suppression of the immune system for the prevention of autoimmune diseases. This critical role of natural Tregs has been exemplified by studies wherein functional deletion of Tregs leads to aggressive and fatal autoimmune diseases (15–17). Dysregulation of Tregs is associated with other autoimmune diseases, including, but not limited to, multiple sclerosis, type 1 diabetes, and rheumatoid arthritis (18–20). Studies have also demonstrated a critical role for the natural Treg in downmodulating antitumor immunity, suppressing allergic diseases such as asthma, and achieving transplant tolerance (21–23).

Accumulating evidence has also suggested an important role for Tregs in the control of immune responses to pathogens (24, 25). The overwhelming majority of these studies have focused on chronic infections, in which the presence of Tregs controls effector immune responses, affecting disease outcome in a variety of ways. By downmodulating the immune response, Tregs can function to reduce immune-mediated disorders, as demonstrated in models of pulmonary inflammation caused by Pneumocystis pneumonia, inflammatory eye lesions with HSV, and liver disease in Schistosoma mansoni infection (26–28). In addition to reducing immunopathological conditions, the presence of Tregs can lead to pathogen persistence. However, the presence of Tregs can also be beneficial to the host by maintaining protective immunity. Specifically, in models of Leishmania infection, Leishmania-specific Tregs found at the site of infection prevent pathogen clearance, which is critical for the development of concomitant immunity (29, 30).

In contrast to chronic infections, how Tregs behave and function during acute infections remains largely undefined. In this article, we show that pathogen-specific Tregs can be activated and expand upon acute infections with vaccinia virus (VV) or influenza virus in vivo. Similar to non-Treg CD4⁺ T cells, activated Tregs undergo contraction to form a memory pool after resolution of the infection. These memory Tregs can rapidly expand, secrete high levels of IL-10, and suppress the collateral tissue damage and inflammation elicited by recall expansion of non-Treg memory CD4⁺ T cells. These results suggest that development of a memory Treg population following an acute infection may help prevent excessive immunopathological effects during a recall response. Furthermore, we provide evidence that the ability of memory Tregs to suppress memory non-Treg CD4⁺ T cells during a recall response is dependent on IL-10.

B10.D2 and BALB/c mice were purchased from The Jackson Laboratory (Bar Harbor, ME). The 6.5 HA-TCR transgenic mice on BALB/c background that express a TCR recognizing an I-E^d–restricted hemagglutinin (HA) epitope (¹¹⁰SFERFEIFPKE¹²⁰) were kindly provided by Dr. H. von Boehmer (Harvard University, Boston, MA). These mice were backcrossed onto the Thy1.1⁺ B10.D2 background (31). Mice expressing enhanced GFP under control of the endogenous Foxp3 promoter (Foxp3-GFP knockin mice on BALB/c background) were purchased from The Jackson Laboratory. These mice were intercrossed with 6.5 HA-TCR transgenic mice to generate 6.5 HA-TCR Foxp3-GFP mice. All mice used in these studies were between 8 and 12 wk of age. Experimental procedures were performed in accordance with protocols approved by the Animal Care and Use Committee of the Duke University Medical Center.

Naive clonotypic HA-specific Treg and non-Treg CD4⁺ T cells (Thy1.1⁺) were prepared from 6.5 HA-TCR transgenic mice, as described (32). CD25⁺ T cells were enriched by MACS-positive selection by staining with PE-conjugated anti-CD25, followed by anti-PE MicroBeads (Miltenyi Biotec). CD25⁺ cells were selected using a MACS column (Miltenyi Biotec) according to the manufacturer’s instructions. Following this procedure, unbound cells (CD25⁻ enriched) were sorted, gating on 6.5⁺CD4⁺CD25⁻ cells (non-Treg), and CD25⁺-enriched cells were sorted, gating on 6.5⁺CD4⁺CD25⁺ cells (Treg), with a FACSVantage (BD Biosciences) high-speed cell sorter. The purity of FACS-sorted populations of cells was >98%. After sorting, 1 × 10⁵ cells (Treg or non-Treg) were transferred into recipient B10.D2 mice (Thy1.2⁺) i.v. in 200 μl HBSS.

Recombinant VV encoding HA (VV-HA) and recombinant E1-deleted adenoviruses encoding HA (Ad-HAs) were previously described (32). Mice were infected with 2 × 10⁶ PFU VV-HA i.p. or 5 × 10⁵ PFU intranasally, or 2 × 10⁹ PFU Ad-HA i.v. Influenza A/PR/8/34 (H1N1) was obtained from American Type Culture Collection (Manassas, VA). The viral titer was determined by a plaque assay, as previously described by Lin et al. (33). Immediately prior to intranasal administration of the influenza, mice were anesthetized with 3% isoflurane. Then 40 μl 4000 tissue culture-infective dose, at 50%, of influenza in sterile PBS was instilled intranasally.

Abs for staining were PE-, FITC-, and APC-conjugated anti-CD4; PE- and PerCP-conjugated anti-Thy1.1; APC-conjugated anti–IFN-γ; biotinylated anti-CD62L; PE-conjugated anti-CD44; PE-conjugated anti–IL-10; streptavidin-PeCy5; and PE-conjugated anti-CD25 (all from BD Biosciences). FITC-conjugated anti-Foxp3 was purchased from eBioscience. Anti–TCR-HA Abs (6.5) were purified and conjugated in our laboratory. Collection of flow cytometry data was carried out using a FACSCanto (BD Biosciences), and events were analyzed using FACSDiva software (BD Biosciences).

For intracellular Foxp3 staining, 1 × 10⁶ cells were surface stained, and then fixed/permeabilized using the Cytofix/Cytoperm Kit (BD Biosciences) according to the manufacturer’s instruction. Following fixation, cells were stained with FITC-conjugated anti-Foxp3. For intracellular IFN-γ and IL-10 staining, cells were cultured for 6 h in the presence of 5 μg/ml GolgiPlug (brefeldin A; BD Biosciences) and 20 μg/ml I-E^d–HA peptide.

For histopathological study, tissues were harvested and prepared for paraffin sections and cryosections. Paraffin sections (5 μm) were stained with H&E according to standard procedures. Random sections were examined for histopathological changes in a blinded fashion, using the criteria we described previously for adenoviral hepatitis, with minor modifications (34).

For immunohistochemistry, frozen sections (5 μm) were acetone fixed and then stained with FITC-conjugated anti-CD4, followed by Alexa Fluor 488-conjugated anti-FITC and Alexa Fluor 546-conjugated H36 (anti-HA mAb, kindly provided by Dr. W. Gerhard, Wistar Institute, Philadelphia, PA). With use of a Zeiss fluorescence microscope, the number of CD4⁺ cells in each section was quantified by counting CD4⁺ cells found in ten randomly selected fields of view.

For in vivo blocking of IL-10R, TGF-β, and CTLA-4, mice were injected with blocking Abs 6 h before infection with VV-HA intranasally, as well as 2 d following Ag rechallenge. The following Abs were used: 0.5 mg anti–IL-10R (1B1.3A) mAb, which was purified from hybridoma cell culture supernatant (American Type Culture Collection); 0.5 mg anti–TGF-β (1D11) Ab (R&D Systems); and 0.25 mg anti–CTLA-4 (UC10-4F10-11) Ab (BioXCell, Lebanon, NH), as well as proper IgG controls (rat, mouse, and hamster, respectively), all from Jackson ImmunoResearch Laboratories.

Results were expressed as mean ± SD. Differences between groups were examined for statistical significance, using the Student t test.

To study the behavior of pathogen-specific Tregs in response to an acute infection in vivo, we used a model of influenza HA-specific Tregs in response to infection with VV-HA. HA-specific Tregs were derived from 6.5 HA-TCR transgenic mice (Thy1.1⁺) that express a TCR recognizing an I-E^d–restricted HA epitope (¹¹⁰SFERFEIFPKE¹²⁰) (31, 35–37). We first characterized the activation and expansion of HA-specific Tregs following VV-HA infection. HA-specific Tregs (6.5⁺CD4⁺CD25⁺) were purified from naive 6.5 HA-TCR transgenic mice by FACS sorting. The sorted HA-specific Treg population was consistently >85% Foxp3⁺ (data not shown). The 1 × 10⁵ Thy1.1⁺, purified HA-specific Tregs or non-Tregs (6.5⁺CD4⁺CD25⁻) were transferred into congenic B10.D2 mice (Thy1.2⁺). Mice were subsequently infected with VV-HA (2 × 10⁶ PFU) i.p. or left uninfected as a control. At 7 d post infection, splenocytes were analyzed for clonal expansion of the HA-specific Treg (Thy1.1⁺CD4⁺Foxp3⁺) or non-Treg (Thy1.1⁺CD4⁺Foxp3⁻) population. Of total lymphocytes, 0.47% were clonotypic Tregs (Fig. 1A, 1B), which represented ∼50-fold expansion compared with that of the mock-infected control mice (Fig. 1A). This expansion was less than that of the clonotypic non-Tregs, which represented 2.9% of the total lymphocytes (∼290-fold expansion) (Fig. 1A, 1B). Despite expansion, very few clonotypic Tregs produced IFN-γ, compared with the 24% of the non-Treg population that were IFN-γ⁺ (Fig. 1B). In contrast, no Foxp3⁺ cells were detected in the non-Treg population (Fig. 1B). Similar degrees of expansion for both Treg and non-Treg populations were also found in other lymphoid tissues, including mesenteric and peripheral lymph nodes (data not shown).

We next analyzed the phenotype of clonotypic Tregs after VV-HA infection. Similar to HA-specific non-Tregs_, HA-specific Tregs downregulated CD62L, compared with naive controls (Fig. 1C). However, unlike non-Treg CD4⁺ T cells, which upregulated CD44 upon infection (Fig. 1C), HA-specific Tregs exhibited high levels of CD44 expression even before VV-HA infection, and no obvious changes were observed post infection (Fig. 1C). This finding is consistent with the idea that Tregs are Ag experienced, as their selection in the thymus requires intermediate- to high-affinity TCR:peptide ligand interactions (35, 38). Collectively, these results show that pathogen-specific Tregs can be activated and expand following an acute viral infection. To further confirm that Foxp3⁺ Tregs are capable of expanding upon VV infection in vivo, we intercrossed HA-TCR transgenic mice with Foxp3-GFP knockin mice, and similar results were obtained when 6.5⁺CD4⁺Foxp3-GFP⁺ Tregs were used for adoptive transfer experiments followed by infection with VV-HA (Fig. 1D).

We next examined the fate of activated Tregs following the peak of clonal expansion at day 7. The majority of activated pathogen-specific Tregs underwent contraction after resolution of the infection, and only a small fraction (∼5%) survived the contraction to form a stable memory population by day 28, which persisted through day 50 (Fig. 2A). This situation was similar to the formation of non-Treg memory CD4⁺ T cells in vivo (Fig. 2A). On the basis of homing characteristics and effector functions, at least two subsets of memory T cells have been described for both CD4⁺ and CD8⁺ T cells (39): central memory T cells (T_CM) that express CD62L and effector memory T cells (T_EM) that lack CD62L. Indeed, non-Treg memory CD4⁺ T cells displayed both the CD62L^lo (T_EM) and CD62L^hi (T_CM) populations. However, the memory Tregs were almost entirely CD62L^lo, corresponding to a T_EM phenotype (Fig. 2B).

One central hallmark of memory T cells is rapid recall expansion upon a secondary Ag challenge (40). To determine whether the memory Tregs also possessed this salient memory quality, mice that had received clonotypic Tregs, and were infected with VV-HA 50 d before, were rechallenged with Ad-HA i.v. At 4 d after rechallenge, memory Tregs rapidly expanded (∼10-fold expansion compared with that of nonboosted mice; Fig. 2C, 2D). This fold expansion was similar to that of the non-Treg memory CD4⁺ T cells (Fig. 2C, 2D). Thus, these pathogen-specific memory Tregs can also rapidly expand upon a secondary challenge. Taken together, these results suggest that a stable population of Tregs with memory characteristics can develop after resolution of the infection.

We next investigated the role of pathogen-specific Tregs during primary and recall responses to an acute infection. We cotransferred Thy1.1⁺, HA-specific Tregs (1 × 10⁵) with non-Tregs (1 × 10⁵)_, or HA-specific non-Tregs only (1 × 10⁵), into congenic B10.D2 mice (Thy1.2⁺), followed by infection with VV-HA. At 7 d post infection, we found that although the initial transfer was at a 1:1 ratio (non-Treg:Treg), clonotypic Tregs represented only ∼10% of the donor (Thy1.1⁺) population in mice that had received both Tregs and non-Tregs (data not shown). This is most likely due to the differences in expansion observed between clonotypic Tregs and non-Tregs in response to VV-HA infection (Fig. 1). We observed no significant difference in the expansion of non-Treg CD4⁺ T cells, with or without the presence of clonotypic Tregs (Fig. 3A, 3B). In addition, no significant difference was noted in the percentage of IFN-γ⁺ cells among clonotypic non-Tregs, with or without HA-specific Tregs (data not shown). Collectively, these results indicate that the presence of activated, pathogen-specific Tregs does not significantly affect the expansion and effector differentiation of non-Treg CD4⁺ T cells during the primary response to VV-HA.

We then evaluated whether HA-specific memory Tregs could suppress the recall response of HA-specific non-Treg memory CD4⁺ T cells. At 50 d post infection with VV-HA, the above mice that had received either Tregs with non-Tregs, or non-Tregs only, were rechallenged with Ad-HA. At 4 d after boost infection, mice were analyzed for the expansion of non-Treg CD4⁺ T cells. Massive recall expansion of non-Treg memory CD4⁺ T cells was detected in the absence of memory Tregs (Fig. 3C, 3D). In contrast, the extent of recall expansion of non-Treg CD4⁺ T cells was significantly (p < 0.001) reduced in the presence of memory Tregs (Fig. 3C, 3D). In addition, a reduction was noted in the percentage of IFN-γ⁺ cells among non-Treg CD4⁺ T cells in the presence of memory Tregs (Fig. 3C). The differences in expansion were not due to differential contraction of non-Tregs in the presence of Tregs, as the percentage of non-Treg memory CD4⁺ T cells was similar between both groups prior to Ag rechallenge (Fig. 3D). Thus, HA-specific memory Tregs suppress the expansion and effector function of HA-specific non-Treg memory CD4⁺ T cells during an Ag recall response.

We next tested whether the observed function of memory Tregs could be confirmed in a different model of acute viral infection. In this experiment, we used a model of influenza viral infection that induces lung disease. Again, we cotransferred Thy1.1⁺, HA-specific Tregs (1 × 10⁵) with non-Tregs (1 × 10⁵)_, or HA-specific non-Tregs only (1 × 10⁵), into congenic B10.D2 mice (Thy1.2⁺), followed by infection with influenza virus (4000 tissue culture-infective dose, at 50%) intranasally. At 7 d later, we evaluated the spleen (data not shown) and draining hilar lymph nodes for clonal expansion of HA-specific non-Tregs and Tregs. We observed no significant difference in the primary expansion of non-Treg CD4⁺ T cells with or without the presence of clonotypic Tregs (Fig. 4A, 4B). In addition, no significant difference was observed in the percentage of IFN-γ⁺ cells among clonotypic non-Tregs, with or without HA-specific Tregs (data not shown). Thus, similar to infection with VV-HA, the presence of activated, pathogen-specific Tregs does not affect the expansion or effector differentiation of non-Treg CD4⁺ T cells during the primary response to influenza virus.

We then evaluated whether HA-specific memory Tregs formed following influenza priming would suppress activation of HA-specific memory non-Tregs after Ag rechallenge with VV-HA. At 50 d post infection with influenza, the above mice that had received either Tregs with non-Tregs, or non-Tregs only, were rechallenged with VV-HA (5 × 10⁵ PFU) intranasally. At 5 d after boost infection, mice were analyzed for the expansion of non-Treg CD4⁺ T cells. The recall expansion of non-Treg memory cells was significantly (p < 0.001) reduced in the presence of memory Tregs (Fig. 4C, 4D). We also observed a significant reduction in IFN-γ⁺ cells among clonotypic non-Tregs in the presence of memory Tregs (Fig. 4C). Taken together, the results from two different models of acute viral infection support our conclusion that memory Tregs suppress the expansion and function of non-Treg CD4⁺ T cell expansion during a recall response.

What, then, is the biological significance of memory Treg-mediated suppression on non-Treg CD4⁺ T cells during a recall response in vivo? Because of the vigorous expansion of non-Treg CD4⁺ T cells that occurred during a recall response, we hypothesized that memory Treg-mediated suppression may prevent the collateral damage that could result from such a robust response. To test this hypothesis, we examined immunopathological conditions in mice with or without memory Tregs. For the first model, we selected liver tissues for histopathological examination because studies have shown that the liver is the primary target organ for adenoviral vectors (41). Again, we cotransferred HA-specific Tregs with HA-specific non-Tregs, or HA-specific non-Tregs alone, into B10.D2 mice, followed by VV-HA infection. At 50 d post infection, mice were rechallenged with Ad-HA. At 4 d after rechallenge, liver tissues were evaluated for tissue effects. We found widespread severe immunopathological changes and tissue damage characterized by hepatocellular degeneration, focal necrosis, and periportal infiltration in mice that received only non-Treg CD4⁺ T cells (Fig. 5A, 5B). Immunohistochemical staining revealed that the portal inflammation mainly consisted of infiltrating CD4⁺ T cells (Fig. 5A, 5C). By contrast, the overall immunopathological condition was diminished, with a significant (p < 0.001) reduction of periportal infiltrating CD4⁺ T cells in the presence of memory Tregs (Fig. 5). This reduction was accompanied by a significant accumulation of memory Tregs in the liver upon rechallenge (Fig. 5D). These results suggest that memory Treg-mediated suppression prevents overwhelming immunopathological changes during a robust recall response. However, the expression of HA in the liver was similar in mice, with or without memory Tregs, at day 4 post rechallenge (Fig. 5A), and it became undetectable by day 10 post rechallenge (data not shown), suggesting that the presence of memory Tregs did not appear to affect the clearance of HA in the liver.

We then assessed immunopathological changes in the second model of infection with influenza virus. Again, we cotransferred HA-specific Tregs with HA-specific non-Tregs, or HA-specific non-Tregs alone, into congenic B10.D2 mice, followed by influenza infection. At 50 d post infection, mice were rechallenged with VV-HA. At 5 d after rechallenge, lung tissues were evaluated for histopathological effects. We found severe immunopathological changes and tissue damage characterized by perivascular and peribronchial infiltration mainly consisting of infiltrating CD4⁺ T cells in mice that received non-Tregs alone (Fig. 6A). By contrast, the overall immunopathological condition was diminished, with a reduction of perivascular and peribronchial infiltration in the presence of memory Tregs (Fig. 6B). Collectively, these results suggest that memory Treg-mediated suppression prevents overwhelming immunopathological changes during a robust recall response.

We next investigated what is responsible for memory Treg-mediated suppression on non-Treg CD4⁺ T cells during the recall response. Because CTLA-4, IL-10, and TGF-β have been implicated in Treg-mediated suppression in vivo (14), we examined the effects of CTLA-4, IL-10, and TGF-β blockade on the recall response. We transferred HA-specific non-Tregs with Tregs into congenic B10.D2 mice, which then received influenza virus intranasally. At 50 d following infection, mice were treated with anti–CTLA-4, anti–IL-10 receptor (IL-10R), anti–TGF-β, or a control IgG, followed by a secondary challenge with VV-HA. At 5 d later, mice were analyzed for the expansion of non-Treg CD4⁺ T cells. We found that only IL-10R blockade led to significantly (p < 0.01) enhanced expansion of non-Treg CD4⁺ T cells (Fig. 7A, 7B) and exacerbation of lung disease (Fig. 7C) in the presence of memory Tregs, but not in the absence of memory Tregs, suggesting IL-10 is responsible for memory Treg-mediated suppression on non-Treg CD4⁺ T cells. We further found that although quiescent memory Tregs produced little IL-10, they produced much higher levels of IL-10 upon Ag rechallenge (Fig. 7D, 7E). Taken together, these data indicate that memory Treg-mediated suppression on non-Treg CD4⁺ T cells and immunopathological effects during the recall response are mediated by IL-10.

In this study, we have examined the kinetics and function of pathogen-specific Tregs in two different models of acute viral infection. Our work reveals a previously unknown memory Treg population that develops following acute viral infections in vivo. We show that memory Tregs suppress the expansion and effector function of non-Treg memory CD4⁺ T cells upon secondary challenge. These memory Tregs play a critical role in controlling excessive immunopathological changes that occur during the recall expansion of non-Treg memory CD4⁺ T cells upon secondary challenge. Furthermore, we demonstrate that these memory Tregs secrete high levels of IL-10 upon Ag rechallenge and that the function of memory Tregs is mediated by an IL-10–dependent mechanism.

Our observation that Ag-specific Tregs expanded in response to acute VV or influenza viral infections is consistent with previous observations in models of HSV and Leishmania infections (29, 30, 42, 43), but stands in contrast to a model of Listeria infection (44). In this model, infection with Listeria monocytogenes led to expansion of L. monocytogenes-specific CD4⁺Foxp3⁻ T cells, but L. monocytogenes-specific CD4⁺Foxp3⁺ Tregs. It is not entirely clear what contributes to the discrepancy, but it could be related to different pathogens, routes of administration, and experimental conditions.

Upon viral infections, non-Treg CD4⁺ and CD8⁺ T cells expand and then contract, with only a small fraction (5–10%) surviving to form the stable memory population (39). Following infection with VV-HA, we observed that pathogen-specific Tregs behave with similar kinetics to non-Treg CD4⁺ T cells, albeit with a lower initial expansion. The memory Tregs that survive the contraction phase can also respond rapidly to secondary Ag rechallenge. However, unlike non-Treg memory CD4⁺ T cells that consist of both CD62L^lo, effector memory (T_EM), and CD62L^hi, central memory (T_CM), populations, these memory Tregs display almost exclusively a CD62L^lo, T_EM phenotype. The expression of CD62L on T_CM allows them to home to lymph nodes efficiently and differentiate into effector cells upon Ag restimulation. In contrast, owing to lack of lymph node homing receptors, CD62L, T_EM migrate into nonlymphoid tissues and display immediate effector functions for the rapid control of invading pathogens. The T_EM phenotype of the memory Tregs may allow them to migrate quickly into nonlymphoid tissues, such as liver and lung, to exert immediate control of collateral tissue damage (39). Indeed, we found rapid accumulation of memory Tregs in the liver upon Ag rechallenge (Fig. 5D).

Our observation that memory Tregs suppress the expansion and effector function of non-Treg memory CD4⁺ T cells and play a critical role in controlling excessive immunopathological effects during the recall response, further supports the growing body of evidence that Tregs are essential in preventing immunopathological disorders in various models of bacterial, viral, parasitic, and fungal infections (26–28, 45). Inhibition of immunopathological changes during a recall response may be of particular importance because of the rapid and robust expansion of memory non-Tregs. Indeed, we observed much greater immunopathological effects in the liver or lungs during a recall response than during a primary response (A.M. Sanchez and Y. Yang, unpublished observations).

Multiple mechanisms have been proposed for Treg-mediated suppression in vivo, including the production of immunosuppressive cytokines such as IL-10 and TGF-β, direct cell–cell contact, and inhibition of APC function (14). In this article, we have provided in vivo evidence that IL-10R blockade significantly reduces the ability of memory Tregs to suppress the recall expansion of non-Treg CD4⁺ T cells and the attendant immunopathological changes, and that memory Tregs produced high levels of IL-10 upon rechallenge, suggesting that IL-10 is responsible for memory Treg-mediated suppression. It remains to be defined how IL-10 suppresses non-Treg CD4⁺ T cell responses and whether other mechanisms of Treg-mediated suppression are also used. Thus, it would be important to address these questions in the future.

In this study, no obvious Treg-mediated suppression on the expansion of non-Treg CD4⁺ T cells was detected during the primary response to VV or influenza in vivo. Our results are in line with the observations in models of infection with Pseudomonas and Trypanosoma (46, 47), but are in contrast to the model of HSV infections (42, 43), as well as the model of peptide IFA immunization (48). The reasons for the discrepancy are not entirely clear but may be due to different experimental models and conditions. In the studies with HSV infection, the polyclonal population of Tregs was deleted, as opposed to addition of Ag-specific Tregs in our study, which could affect the overall suppressive capability of Tregs. In addition, different infections and immunizations could elicit different inflammatory milieu and/or innate immune response, both of which have been suggested to play a role in overcoming Treg-mediated suppression (32, 49, 50).

We observed a difference in the suppressive ability of naive or activated Tregs during the primary response versus memory Tregs during the recall response. This difference is not likely due to changes in Treg frequency between the primary and recall responses, as the percentage of Tregs among total Ag-specific T cells remains similar during the initial priming and the recall response. Similarly, this difference is not likely a result of different viruses used between the priming and recall response because VV has been used in both settings in our models. Rather, our data suggest that the increased suppressive potential of memory Tregs is due to the ability to produce markedly elevated levels of IL-10 upon rechallenge. Thus, a qualitative difference exists between memory and naive Tregs in the ability to suppress non-Tregs, and this qualitative difference is reflected in their ability to rapidly produce high levels of IL-10 in response to Ag stimulation. The underlying mechanism leading to this qualitative gain by memory Tregs requires further investigation.

In conclusion, we have shown that pathogen-specific Tregs expand and contract following acute viral infections, with a small population developing into the memory Treg pool. These memory Tregs suppress memory non-Tregs upon a secondary challenge. The suppression by memory Tregs is critical to the control of immunopathological changes that occur during the recall response as a result of robust non-Treg expansion. Finally, we demonstrate that suppression by memory Tregs results, at least in part, from their ability to produce high levels of IL-10 during the recall response. Our work reveals a previously unknown memory Treg population that develops following acute viral infection in vivo and may help in the design of effective strategies to circumvent excessive immunopathological changes during a recall response.

We thank Dr. H. von Boehmer for providing 6.5 HA-TCR transgenic mice.

The authors have no financial conflicts of interest.