Regulatory T cells limit unconventional memory to preserve the capacity to mount protective CD8 memory responses to pathogens

Significance

T cell-mediated immune memory provides the host with the capacity to rapidly and efficiently respond to pathogen reencounter, often without symptoms or disease. In addition to conventional memory T cell responses that are pathogen-specific, there also exists a subset of unconventional memory T cells that share many of the characteristics of conventional memory T cells without the initiating requirement of cognate antigen recognition. Here, we demonstrate a mechanism whereby regulatory T cells limit the expansion of these unconventional memory T cells to preserve a host’s capacity to mount a memory CD8 T cell response that remains protective upon pathogen reencounter.

Abstract

Immunological memory exists so that following infection an expanded population of pathogen-specific lymphocytes can rapidly and efficiently control infection in the case of reexposure. However, in the case of CD8+ T lymphocytes, a population of unconventional CD44+CD122+ virtual memory T cells (T_VM) has been described that possesses many, though not all, features of “true memory” T cells, without the requirement of first encountering cognate antigen. Here, we demonstrate a role for regulatory T cell-mediated restraint of T_VM at least in part through limiting IL-15 trans-presentation by CD11b+ dendritic cells. Further, we show that keeping T_VM in check ensures development of functional, antigen-specific “true” memory phenotype CD8+ T cells that can assist in pathogen control upon reexposure.

In response to pathogen exposure, naïve T cells recognizing their cognate antigen in the context of costimulatory signals and appropriate cytokines undergo proliferation and differentiation. These expanded T cells give rise to a pool of effector T cells that can assist in pathogen clearance and a population of long-lived memory T cells that can rapidly respond to pathogen in the case of reexposure. These pathogen-specific memory T cells are an asset to the host, as they possess enhanced functional characteristics such as increased quality and speed of response in addition to metabolic and migratory reprogramming that enables these memory T cells to outperform their naïve precursors (1–3).

In addition to this canonical pathway, it has become evident that unconventional T cells with a memory phenotype can develop without exposure to foreign antigens, but rather through homeostatic proliferation driven by lymphopenia (4–8). Recently, it was shown that antigen-specific, nonself-reactive CD8+ T cells from unimmunized/unmanipulated hosts are detectable in germ-free and specific pathogen-free mice. Notably, these cells share characteristics with both conventional, “true” memory T cells and memory T cells generated through homeostatic proliferation. Shared traits include increased expression of CD44, rapid proliferation after T cell receptor (TCR) stimulation, and rapid production of IFN-γ following exposure to proinflammatory cytokines (9, 10). In naïve mice these cells, now termed virtual memory (T_VM) CD8+ T cells, have been further phenotyped and found to express high levels of CD44, CD122, CXCR3, and Ly6C and low levels of CD49d.

A distinct population, comprising ∼10 to 20% of total CD8+ T cells expressing these markers, has been documented in naïve mice, suggesting that T_VM likely play a role in the immune system. Indeed, CD8+ T_VM have been shown to confer bystander-mediated protection; they provide potent protective immunity against pathogens including Listeria monocytogenes and influenza virus, suggesting that T_VM are distinct from naïve T cells and can participate in protective immunity (11–13). While both conventional memory and T_VM are poised to proliferate following TCR stimulation, T_VM cells are less efficient at rapid and robust IFN-γ production under these conditions (11), highlighting the distinct characteristics of T_VM versus true memory T cells and further demonstrating their potential unique role in immunity. Furthermore, recent studies have identified a T_VM-like population in the human liver (14), underscoring the need to advance our understanding of T_VM generation and function, as these cells could potentially be harnessed for therapeutic interventions such as vaccination.

CD8+ T_VM cells arise extrathymically and their development is dependent on homeostatic rather than antigenic environmental cues (10, 12), with a demonstrated requirement for IL-15 presented in trans by CD8α+ dendritic cells (DCs) (12). However, it remains unclear how development of CD8+ T_VM cells is regulated such that the frequency of this population remains fairly stable in adult hosts, despite some age-related increases (15). Given the critical importance of coordinating and regulating a developing true memory CD8+ T cell response, we sought to investigate the potential role regulatory T cells (Tregs) might play in the regulation of T_VM development.

Tregs play a central role in the prevention of autoimmunity through their ability to suppress autoreactive cells and inflammation (16, 17). However, to date there has been no investigation of the role that Treg-mediated regulation may play in the development of the T_VM cell pool. Given that T_VM have the potential to respond to TCR signals with robust proliferation, respond to inflammatory cytokines with rapid IFN-γ production, and express CXCR3, a chemokine receptor that can immediately allow access to tissues, we hypothesized that these cells would necessarily be subject to immunomodulatory restraint. Here, we demonstrate the mechanistic role that Tregs play in the restraint of T_VM. Further, we demonstrate that restriction of the T_VM pool allows for the development of functional, antigen-specific true memory cells that can protect the host from secondary challenge.

Results

Tregs Limit Expansion of the Virtual Memory CD8+ T Cell Pool.

To test the hypothesis that CD8+ T_VM cells are subject to Treg-mediated restraint, we transiently depleted Tregs using the Foxp3^DTR mouse model and subsequently measured the frequency of T_VM in the blood and spleen. Surprisingly, only 4 d after Treg ablation, the frequency of T_VM cells in the blood more than doubled, and by 6 d postdepletion, a time at which there are not yet any overt signs of autoimmunity or weight loss, ∼35% of blood CD8+ T cells had a virtual memory phenotype (Fig. 1A), suggesting that Tregs play an important role in the regulation of T_VM. Similarly, in the spleen, there was also a significant increase in the frequency of T_VM after removal of Tregs (Fig. 1B), as well as a significant increase in the total number of splenic T_VM by day 6 after Treg depletion (Fig. 1C). Following Treg depletion, T_VM maintain their characteristic phenotype of CD49d^lo CD8+ T cells that coexpress CD122 and CD44 (Fig. 1D). Furthermore, Treg ablation results in a decrease in the fraction of CD8+ T cells that are naïve, although there is no difference in the number of naïve CD8+ T cells within the spleen (SI Appendix, Fig. S1). Finally, the increase in T_VM following transient Treg depletion is stable, as the significant increase in T_VM frequency and number is still detectable 28 d after transient Treg ablation (Fig. 1E). Altogether, our data demonstrate that at steady state Tregs play a role in limiting the expansion of unconventional CD8+ T cells such as T_VM.

Fig. 1.

Tregs restrain the virtual memory T cell pool. Foxp3^DTR or C57BL/6J (B6) mice were injected i.p. with DT for two consecutive days to deplete Tregs. For some experiments, controls were PBS-treated Foxp3^DTR mice. (A–E) On days 0, 2, 4, 6, and 28 after Treg depletion, cohorts of mice were bled or killed for collection of spleens and the frequency and number of T_VM cells was assessed by flow cytometry. T_VM were gated on single live lymphocytes that were CD3+CD8+ and further identified based on low CD49d expression and CD122 and CD44 coexpression. Results are representative of three independent experiments. Statistical significance was determined by ANOVA with Tukey’s multiple comparisons test. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. ns, not statistically significant.

Tregs Aid in the Maintenance of a Stable T_VM Population by Limiting T_VM Expansion.

Given our finding that removing Treg-mediated restraint unleashes a dramatic and significant increase in the frequency and number of T_VM cells (Fig. 1), we next sought to determine if the increased T_VM population remains stable upon repopulation of the Treg compartment. A previous study of T_VM cells in both neonatal and adult mice found that T_VM frequency peaks at about 30% of CD8+ T cells at 3 wk of age, followed by a decline to ∼20% in adult mice, which then remains relatively stable throughout life (10). The timing of this T_VM expansion has been attributed to lymphopenia within neonates. Interestingly, this timing corresponds to the development during ontogeny of Foxp3+ Tregs, which are delayed in development compared with conventional CD4 T cells in the thymus and do not begin to appear in appreciable quantities until about 3 wk of age (18). Thus, we hypothesized that transient removal of Tregs in adult mice would result in expansion of T_VM cells, and that while Treg recovery might somewhat diminish their numbers, as in the neonate-to-adult transition (30 to 20% of CD8+ T cells), the end result would be a net gain in T_VM frequency. To test this hypothesis, we transiently depleted Tregs using Foxp3^DTR mice and tracked T_VM frequency in the blood over time. We found that T_VM frequency remained elevated out to day 28 postdepletion, despite a rapid recovery in Tregs (Fig. 1E and SI Appendix, Fig. S2A). Furthermore, T_VM cells taken from mice previously depleted of Tregs, and thus enriched for T_VM, retain functional properties similar to those from unmanipulated mice in terms of IFN-γ production following exposure to IL-2, IL-12, and IL-18 (SI Appendix, Fig. S2B). Next, we examined clonal diversity by cataloging the TCR Vβ usage within the T_VM pool. While there were some minor but statistically significant differences in TCR Vβ usage between T_VM and naïve CD44^lo CD8+ T cells, there were no statistically significant differences in usage between T_VM from WT, unmanipulated mice and mice that had a higher frequency of T_VM due to previous Treg ablation (SI Appendix, Fig. S2C). Altogether, transient depletion of Tregs results in a long-term increase in the frequency of T_VM cells that retain both the phenotypic and functional characteristics of virtual memory cells.

To determine if the increase in T_VM that occurs upon removal of Treg-mediated restraint is due to conversion of naïve, CD44^lo CD8+ T cells into T_VM or expansion of existing T_VM, we performed an adoptive transfer experiment. Briefly, congenically marked CD8+CD44^lo naïve cells or CD44+CD122+ T_VM cells were transferred into Foxp3^DTR hosts that had been transiently depleted of Tregs 1 d before transfer. Five days later, proliferation of each of these transferred populations was assessed by CellTrace Violet (CTV) dilution, as well as virtual memory phenotype. We observed that T_VM cells retained their T_VM phenotype and underwent several rounds of division upon transfer into Treg-depleted mice whereas naïve CD44^lo CD8+ T cells did not proliferate or convert into T_VM (Fig. 2). In addition, the transferred T_VM proliferated only when infused into Treg-depleted hosts, indicating that the proliferation observed was not due to homeostatic proliferation of memory phenotype CD8+ T cells. Finally, when Tregs are transiently depleted in a serial manner, the frequency of T_VM does not increase further, indicating that there is a limited potential for expansion of T_VM (SI Appendix, Fig. S2D).

Fig. 2.

Tregs limit expansion of T_VM. CD44-CD8+ naïve T cells or CD122+CD44+ T_VM were FACS-sorted from Ly5.1 mice, labeled with CTV, and adoptively transferred into Foxp3^DTR mice depleted of Tregs (Treg−) or PBS-treated Foxp3^DTR mice (Treg+) 1 d before transfer. Five days later, the transferred cells located in the spleen were examined for proliferation and phenotypic identity by flow cytometry.

Tregs Restrain IL-15 Transpresentation by DCs.

IL-15 is a homeostatic cytokine that is constitutively expressed by many cell types and is known to play a role in the survival and function of T cells. After naïve, CD44^loCD5+ CD8+ T cells exit the thymus and enter the periphery, they receive IL-15 signals within the context of the IL-15Rα on CD8α+ DCs to become a T_VM, and additional IL-15 can subsequently lead to the expansion of the T_VM pool (19). Thus, given the importance of IL-15 in the generation and expansion of T_VM, we next investigated whether Tregs could be restraining the T_VM population by inhibiting trans-presentation of IL-15 by DCs.

Using transient Treg depletion in the Foxp3^DTR mouse model, we tested the hypothesis that Tregs directly restrain IL-15 trans-presentation by DCs. We assessed membranous IL-15 expression on DCs using ex vivo flow cytometric staining as previously described (20). Upon ablation of Tregs, there was no significant increase in the amount of IL-15 expressed by CD8α+ DCs (Fig. 3A and SI Appendix, Fig. S3A), although there was a modest increase in the frequency but not the total number of CD8α + DCs trans-presenting IL-15 (Fig. 3B). Interestingly, however, removal of Tregs resulted in a dramatic and significant increase in the amount of IL-15 presented by CD11b+ DCs (Fig. 3C and SI Appendix, Fig. S3A) as well as the frequency and number of CD11b+ DC that trans-present IL-15 (Fig. 3D). Removal of Tregs had no effect on the amount of IL-15Rα expressed by either CD8α+ or CD11b+ DCs (SI Appendix, Fig. S3B). Altogether, these data suggest that Tregs limit homeostatic IL-15 trans-presentation by CD11b+ DCs.

Fig. 3.

Loss of Tregs results in enhanced IL-15 trans-presentation by CD11b+ DCs. (A–D) Foxp3^DTR mice were injected with DT to deplete Tregs, and IL-15 expression was measured on the surface of CD8α⁺ DCs (A and B) or CD11b⁺ DCs (C and D). The percent or number of IL-15+ cells (B and D) was quantified using flow cytometry. (E) Direct ex vivo pSTAT5 staining was assessed by flow cytometry of splenic cells from Foxp3^DTR mice treated with DT (Treg−) or PBS (Treg+) to deplete Tregs for 3 d. (F) Foxp3^DTR mice were injected with DT to deplete Tregs (Treg−) and additionally treated with αCD122 antibody or isotype control before flow cytometric assessment of T_VM frequency in the spleen. (G) Foxp3^DTR mice were injected with DT to deplete Tregs (Treg−) and additionally treated with αCD4 antibody or isotype control before flow cytometric assessment of CD11b+ DC expression of IL-15 and number of T_VM in the spleen. (H) Mice from CC strains 008, 011, 027, 012, 071, 002, 040, 003, 005, 044, and 004 (from left to right on the x axis) were assessed for frequency of CD4+Foxp3+ Tregs in the spleen as a percentage of live lymphocytes by flow cytometry. Statistical significance was determined by unpaired t tests, ANOVA with Tukey’s multiple comparisons test, or linear regression. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. ns, not statistically significant.

To link this enhanced IL-15 trans-presentation with increased expansion of T_VM, we stained for pSTAT5 expression within CD8+ T cells after transient Treg ablation. Two days after depletion of Tregs, we noted a significant increase in the frequency of CD8+ T cells that have phosphorylated STAT5 (Fig. 3E), suggesting that they have received enhanced signals through the IL-15Rβ/γc as a result of Treg removal. To more directly demonstrate that this increase in IL-15 availability has an effect on CD8+ T cells, we treated mice with anti-CD122. This was done in the context of Treg ablation to limit the ability of T_VM to respond to the enhanced levels IL-15 presented by CD11b+ DCs upon removal of Tregs. We used TMβ-1, a murine antibody equivalent to the humanized Mikβ-1 (21), which has previously been shown to block IL-15 trans-presentation in humans (22–25). Foxp3^DTR mice treated with DT and anti-CD122 antibody had significantly fewer T_VM compared with mice depleted of Tregs and treated with an isotype control antibody (Fig. 3F). In addition, we found that treatment of Treg-sufficient mice with anti-CD122 resulted in a reduction in the frequency of T_VM compared with mice treated with an isotype control (Fig. 3F). A previous study using a mutated TMβ-1 that does not participate in antibody-dependent cytotoxicity (ADCC) found that in vivo treatment of mice with this mutated TMβ-1 achieved a similar reduction in the frequency of CD122+ CD8+ T cells compared with WT TMβ-1 (26), thereby suggesting that in vivo treatment with TMβ-1 results in a blockade of IL-15 trans-presentation rather than ADCC-mediated elimination of CD122+ cells. Thus, it appears that sustained IL-15 signaling may be required for the maintenance of T_VM.

Given that the removal of Tregs led to both expansion and a corresponding increase in IL-15 production by CD11b+ DCs (Fig. 3 C and D) (27), we next sought to determine the mechanism whereby Tregs restrict IL-15 trans-presentation by this DC subset. It was previously reported that Treg depletion leads to expansion of activated CD4+ T cells, which in turn help DCs to mature through CD40-CD40L signaling and other pathways (27). Thus, we hypothesized that removal of Treg-mediated restraint of CD4 T cells could contribute to enhanced IL-15 trans-presentation by CD11b+ DC. To test this hypothesis, Foxp3^DTR mice were treated with anti-CD4 antibody or isotype control in conjunction with DT-mediated ablation of Tregs. Compared with isotype control-treated mice depleted of Tregs, mice lacking CD4+ T cells had significantly reduced IL-15 expression by CD11b+ DC, as well as fewer CD11b+ DC making IL-15 (Fig. 3G). Further, there was no difference in IL-15 production by CD11b+ DC between Treg-replete mice and mice lacking CD4 T cells and Tregs, or in the number of T_VM (Fig. 3G). Thus, CD4+ T cells are critical for enhanced IL-15 trans-presentation by CD11b+ DC upon Treg ablation. Altogether, our data point to a role for Tregs in restricting IL-15 trans-presentation by CD11b+ DCs to limit expansion of T_VM cells.

While the complete removal of Tregs results in a dramatic increase in the frequency of T_VM (Fig. 1 A and B), the dramatic loss of this dominant immunomodulatory cell type likely has pleiotropic effects that influence more than IL-15 trans-presentation. Thus, we next turned to the Collaborative Cross (CC), a resource consisting of a set of recombinant inbred strains of mice that have distinct and unique genotypes. This afforded us the advantage of heterogeneity within genotype and thus immunophenotype. We have previously shown that the CC better mimics the natural diversity observed in the human population in terms of variability in the frequency of lymphocyte types and immune activation (28). Indeed, upon examination of 11 CC strains, we found a large degree of variability in the frequency of Tregs within the spleen, despite all of the mice being overtly healthy (Fig. 3H). Further, we observed an inverse correlation between the frequency of Tregs and the frequency of DCs expressing IL-15 (Fig. 3H), suggesting a dose–response effect consistent with the reduced availability of Tregs resulting in enhanced IL-15 trans-presentation. Finally, the frequency of Tregs was found to be inversely correlated with the frequency of splenic T_VM (Fig. 3H), and although there is only a trend toward significance, it is also likely that Tregs are not the sole contributing factor to the size of the T_VM population.

Tregs Restrain CD8+ T_VM Formation in Part Through CTLA-4–Mediated Inhibition of DCs.

Given our dramatic finding of increased T_VM cell frequency and IL-15 trans-presentation upon ablation of Tregs (Figs. 1 and 3), we next investigated the mechanism of this Treg-mediated restraint. Tregs are known to suppress T cell responses through a variety of well-studied mechanisms (29). One such mechanism, CTLA-4, has been shown to be a critical mediator of Treg suppression. Conditional deletion of CTLA-4 in Tregs reduces their suppressive capacity and leads to an inability to effectively down-regulate CD80 and CD86 on antigen presenting cells (APCs) via transendocytosis (30, 31). Given the dominant role of CTLA-4 in APC-mediated suppression governed by Tregs and the dependence on DCs for generation of T_VM, we investigated the importance of CTLA-4 expression by Tregs on T_VM formation using a Foxp3-driven conditional KO (cKO) mouse model. Similar to Foxp3^DTR mice transiently depleted of Tregs, in which there was a more than threefold expansion of T_VM within the spleen, CTLA-4^flox/flox × Foxp3^Cre mice have a significantly higher frequency and number of T_VM than Foxp3^Cre controls at steady state, with ∼50% of CD8+ T cells within the spleen displaying a T_VM phenotype (Fig. 4A). Additionally, we tested the necessity of Treg production of IL-10, a soluble mediator of suppression, and found that in contrast to CTLA-4 expression, Treg production of IL-10 was dispensable for the regulation of T_VM (SI Appendix, Fig. S4). These results suggest that Tregs likely restrain T_VM development, at least in part, through a contact-dependent CTLA-4–mediated inhibitory mechanism.

Fig. 4.

CTLA-4 expression by Tregs is critical for controlling the frequency of T_VM by limiting CD11b+ DC expression of IL-15. (A) CTLA-4^flox/flox × Foxp3^Cre (CTLA-4 cKO) or CTLA-4^WT/WT × Foxp3^Cre (WT) mice were assessed for the frequency and number of T_VM within the spleen by flow cytometry. IL-15 expression was measured on the surface of CD11b⁺ DCs from CTLA-4 cKO or WT mice (B). MFI and percent or number of IL-15+ cells was quantified using flow cytometry. Statistical significance was determined by unpaired t test. *P ≤ 0.05, **P ≤ 0.01. ns, not statistically significant.

Through its increased affinity for B7 molecules expressed by APCs, CTLA-4 effectively outcompetes T cell-expressed CD28 for binding, thus curtailing a developing T cell response (32). Furthermore, CTLA-4 has been demonstrated to enhance Treg adhesion to APCs via an LFA-1–dependent mechanism (33, 34). As such, we hypothesized that CTLA-4+ Tregs might be restraining IL-15 trans-presentation by DCs, and thereby regulating T_VM development, by restricting access of CD8+ T_VM precursors or T_VM, respectively, to IL-15. CTLA-4 itself appears to play a role in the blocking function of Tregs, as conventional T cell clustering around DCs in the presence of Tregs is enhanced by in vitro or in vivo treatment with a CTLA-4 blocking antibody (35). In addition, recent studies have shown that Tregs form prolonged adhesive contact with DCs independent of antigen or MHC recognition, which significantly suppressed the interaction of the same DCs with conventional T cells (36). Similarly, Tregs exhibit strong intrinsic adhesiveness to DCs that reduces the ability of Treg-contacted DCs to engage other antigen-specific cells (37). Therefore, we experimentally determined if Tregs affect IL-15 trans-presentation by DCs in a CTLA-4–dependent fashion using flow cytometric assessement of IL-15 on CD11c+ DCs as above (Fig. 3 A–D). In CTLA-4 cKO, there was no significant difference in the amount of IL-15 expressed on CD8α+ DCs, or in the frequency or number of CD8α+ DCs presenting IL-15 (SI Appendix, Fig. S5). Similar to a complete ablation of Tregs, however, CD11b+ DC in CTLA-4 cKO mice displayed a significant increase in the amount of IL-15 presented by CD11b+ DCs as well as the frequency and number of CD11b+ DCs that trans-present IL-15 (Fig. 4B). Altogether, these data suggest that Tregs limit homeostatic IL-15 trans-presentation by CD11b+ DCs through a CTLA-4–dependent mechanism, which is likely playing a role in regulation of the T_VM population.

An Expanded Population of T_VM Does Not Adversely Affect Outcome to Viral Infection, Despite a Deficiency in Pathogen-Specific Effector CD8+ T Cell Generation.

Given our finding that a transient unleashing of Treg-mediated restraint results in a stable increase in cells with T_VM phenotypic and functional properties (SI Appendix, Fig. S2), we next investigated the consequence of this expanded population of T_VM cells on the T cell response to primary infection. To do this, we first enriched mice for T_VM by transiently depleting Tregs, followed by a ≥30-d rest period to ensure that Treg frequencies returned to steady state in the presence of the previously noted T_VM enrichment in frequency (SI Appendix, Fig. S2A). We then infected these T_VM-high mice with West Nile virus (WNV) via s.c. injection in the rear footpad. The interaction between CD8+ T cells and Tregs in this infection model has been shown to play an especially critical role in modulating the delicate balance between a developing anti-WNV response and the immune-mediated collateral tissue damage that can occur after exposure (38, 39). Infected mice had a comparable frequency of Tregs in the spleen (SI Appendix, Fig. S6), and maintained T_VM enrichment, with more than double the frequency of T_VM in the T_VM-high mice compared with WT unmanipulated controls (Fig. 5A). Contrastingly, when we examined both the spleen and the brain for WNV-specific CD8+ T cells 12 d postinfection, we found a significant and dramatic reduction in the frequency of effector WNV-specific CD8+ T cells within both tissue compartments in the T_VM-high groups (Fig. 5 B and C). Similarly, the memory CD8+ T cell response directed at WNV was impaired in these T_VM-high mice, as there were fewer WNV-specific CD8+ T cells 4 wk postinfection (Fig. 5D). However, these differences cannot be attributed to variability in viral pathogenesis as both weight loss and viral titers were comparable between WT and T_VM-high mice following infection (Fig. 5E and SI Appendix, Fig. S7). Indeed, this result is consistent with previous findings that T_VM proliferate rapidly in response to antigen (9) and can participate in protective immunity against pathogens such as L. monocytogenes and influenza virus (11–13); while an expanded T_VM pool appeared to have consequences to expansion of WNV-specific CD8+ T cell effectors, it is likely that the T_VM still contributed sufficiently to lead to control, thereby resulting in an equivalent viral load.

Fig. 5.

Unrestrained expansion of T_VM compromises the frequency of anti-viral effector CD8+ T cells without affecting clinical outcome. Foxp3^DTR mice were injected with PBS (WT) or DT (T_VM-high) at least 4 wk before s.c. infection with WNV in the rear footpad. The frequency and number of splenic T_VM (A) was assessed by flow cytometry immediately before WNV infection, and the frequency and number of WNV-specific CD8+ T cells in the spleen (B) and brain (C) was assessed by NS4b tetramer staining at day 12 postinfection. The frequency and number of memory WNV-specific CD8+ T cells in the spleen was assessed by tetramer staining at day 35 postinfection (D). (E) Spleen viral load at day 4 postinfection and brain viral load at day 12 postinfection was determined by real-time qRT-PCR. Statistical significance was determined by unpaired t test. *P ≤ 0.05, **P ≤ 0.01. ns, not statistically significant.

β1 Integrin-Expressing Tregs Exert Superior Restriction of the T_VM Pool.

Our experiments using mice with high T_VM frequencies due to previous Treg depletion demonstrated that unregulated expansion of T_VM carries a cost in terms of the pathogen-specific immune response. However, while the frequency of Tregs had returned to baseline before infection, the use of transient depletion of Tregs to increase the size of the T_VM population could result in other long-term changes impacting the antiviral immune response. Thus, we sought an alternate model in which to study the regulation of T_VM population size and its resultant impact on antipathogen immunity. To do so, we investigated potential contributing mechanisms by which Tregs could be restraining T_VM expansion through inhibition of IL-15 trans-presentation.

As discussed above, one molecular mechanism by which Tregs are able to suppress T cell activation is through decreasing the stability of the in vivo T cell–DC synapse (40, 41). One method by which this is accomplished is through the increased ability of Tregs themselves to form stable conjugates with DCs to outcompete conventional T cells for access to DCs. Specifically, leukocyte function-associated antigen-1 (LFA-1, or a_Lb₂) expression by Tregs is required for Tregs to form stable aggregates with DCs (42, 43). This stabilized interaction with DCs allows Tregs to block conventional naïve T cell access to the DC, as well as down-regulate the expression of CD80/86 on DCs in a CTLA-4–dependent manner (43). β1-integrin, a component of the heterodimeric integrin very late antigen-4 (VLA-4), along with α4-integrin, has been reported to be necessary for T cell access to sites of inflammation (44). However, integrins are well known to facilitate cell–cell adhesion in addition to their critical roles in migration. For instance, LFA-1 forms the peripheral ring of the immunological synapse that forms between APCs and T cells during T cell activation (45, 46). VLA-4 has also been shown to colocalize with LFA-1 within the antigen-dependent immune synapse, suggesting that it could participate in stabilization of the synapse between T cells and DCs by promoting cell–cell adhesion (47, 48). In C57BL/6 mice, we have found that β1-integrin is constitutively expressed on ∼40% of Tregs (Fig. 6A). Further, β1-integrin+ Tregs express significantly greater levels of CTLA-4 and CXCR3 (Fig. 6 B and C), suggesting this population may have increased suppressive activity. Taken together, we hypothesized that β1-integrin expression by Tregs could be playing a role in Treg–DC interactions and could thus affect IL-15 trans-presentation to T_VM.

Fig. 6.

β1 integrin-expressing Tregs exert superior control over T_VM expansion. (A) Representative staining of β1 integrin expression by CD4+Foxp3+ Tregs. Flow cytometric determination of the percent of either β1 integrin+ or β1 integrin− Tregs expressing CTLA-4 (B) or CXCR3 (C). (D) The frequency and number of T_VM within the spleens of Itgb1^WT/WT × Foxp3^Cre (WT) or Itgb1^flox/flox × Foxp3^Cre (cKO) mice is shown along with representative flow cytometry staining. (E) IL-15 expression was measured on the surface of CD11b⁺ DCs from β1 cKO or WT mice. MFI and percent or number of IL-15+ cells was quantified using flow cytometry. Statistical significance was determined by t test or ANOVA with Tukey’s multiple comparisons test. *P ≤ 0.05, **P ≤ 0.01, ****P ≤ 0.0001. ns, not statistically significant.

To address this hypothesis and determine whether β1-integrin expression by Tregs plays a role in the regulation of DCs and T_VM expansion, we utilized a FoxP3-driven β1-integrin cKO mouse model. First, we confirmed that use of Itgb1^flox/flox × Foxp3^Cre mice results in a nearly complete abrogation of β1-integrin expression in Foxp3+ Tregs (SI Appendix, Fig. S8A). Importantly, through the β1-integrin cKO mouse model, we discovered that a lack of β1-integrin on Tregs correlated with a significant increase in the frequency of CD8+ T_VM cells at steady state (Fig. 6D). Thus, it appears that β1-integrin may contribute to the ability of Tregs to regulate expansion of T_VM.

It has previously been demonstrated that T_VM are likely derived from CD8+ T cells that were selected with increased affinity for self ligands in the thymus, as indicated by their increased expression of CD5 compared with naïve conventional CD8+ T cells (14). To determine if β1-integrin deficiency in Tregs had an effect on thymic selection, thereby leading to an increased frequency of CD5^hi cells that could then yield more T_VM, we examined expression of CD5 on peripheral T_VM and naïve CD44^lo CD8+ T cells in β1-integrin cKO or WT mice. While CD5 is expressed at significantly higher levels on T_VM compared with naïve CD8+ T cells, as previously reported (14), there is no difference in CD5 expression levels between T_VM or naïve CD8+ T cells in WT versus β1-integrin cKO mice (SI Appendix, Fig. S8B). Thus, it appears that β1-integrin deficiency in Tregs does not alter thymic output.

In Tregs, IL-15 signaling has been shown to increase the expression of CD25 and Foxp3 (49, 50). As such, we hypothesized that Tregs might consume IL-15 in a competitive fashion, as they do IL-2, thereby limiting the availability of IL-15 for trans-presentation to naïve CD8+ T cells and perhaps the subsequent expansion of T_VM. To investigate this possibility, we examined CD122 expression by β1-integrin–positive versus –negative Tregs compared with T_VM by flow cytometric staining of cells taken directly ex vivo from WT mice. We found that T_VM express far more CD122 on a per-cell basis as compared to Tregs, regardless of β1-integrin expression (SI Appendix, Fig. S8C). Unlike the well-characterized inhibitory mechanism of high-affinity CD25 mediated competion for IL-2 by Tregs, our data suggest that Tregs are not selectively outcompeting T_VM for IL-15 presented by DCs based on increased availability of the IL-15R.

To establish the mechanism whereby β1-integrin expression by Tregs limits T_VM generation, we next examined the ability of DCs to trans-present IL-15 in β1-integrin cKO mice by IL-15 staining on DCs as described above (Fig. 3). Absence of β1-integrin on Tregs did not alter the ability of CD8α+ DCs to present IL-15 (SI Appendix, Fig. S8D). However, there was a modest increase in the amount of IL-15 presented by CD11b+ DCs, although this did not affect the frequency or number of CD11b+ DCs expressing IL-15 (Fig. 6E). Thus, it appears that β1-integrin expression by Tregs plays a role in limiting IL-15 trans-presentation by CD11b+ DCs, as well as restraint of T_VM expansion.

Tregs Restrain the T_VM Population to Preserve the Capacity to Generate Protective True Memory T Cell Responses upon Pathogen Reexposure.

Next, we sought to extend our investigation of the effects of dysregulated expansion of T_VM cells to the immune response (Fig. 5) upon pathogen challenge or reexposure. Given our results using the model of T_VM expansion via transient Foxp3^DTR-mediated Treg ablation (Fig. 5), we hypothesized that dysregulated expansion of T_VM during the development of a pathogen-specific primary response might permanently alter a host’s ability to mount effective pathogen-specific CD8+ T cell recall responses. To test this we used the β1-integrin cKO mouse model in conjunction with WNV infection, as it allowed us to eliminate the additional manipulation that Treg depletion in our enriched T_VM model required. To ensure that the β1-integrin cKO mouse model was suitable to answer this question, we first extensively phenotyped these mice to rule out the presence of any other potential immune defects. β1-integrin cKO mice display no overt signs of disease and body weight at steady state was identical to that of age-matched WT mice (SI Appendix, Fig. S9A). Further, there was no difference in the frequency or number of Tregs in the cKO mice compared with WT controls. However, cKO mice had a significant reduction in the frequency and number of Tregs expressing CTLA-4, as well as CD73 (SI Appendix, Fig. S9B). The reduced frequency of Tregs expressing CTLA-4 is in line with our observation that β1+ Tregs express more CTLA-4 (Fig. 6B). This reduced expression of Treg suppressive markers, however, does not appear to alter conventional T cell frequency or activation, as there is no difference in CD3+ T cell frequency or number, or CD4+ or CD8+ T cell frequency or number (SI Appendix, Fig. S9 C–E). In addition, there is no difference in Ki67 expression on CD4+ or CD8+ T cells, and while there is an increased frequency and number of CD44+ CD4+ and CD8+ T cells (SI Appendix, Fig. S9 D and E), it is likely that this is due to an enhanced expansion of CD8+ T_VM cells. Thus, because the immune phenotypic differences between β1 cKO and WT mice were largely restricted to differences associated with the mechanisms underlying T_VM expansion, as well as T_VM frequency, we proceeded with this model to further investigate the effects of expanded T_VM on the pathogen-specific memory recall response.

β1-integrin cKOs or WT control mice were infected with WNV to first examine the primary WNV-specific CD8+ T cell response using tetramer staining. The frequency and number of WNV-specific CD8+ T cells at the peak of the primary response to WNV was comparable between cKO and WT mice. Additionally, when we examined the true, antigen-specific memory response at day 35 following WNV infection, we detected a similar frequency and number of WNV-specific CD8+ T cells within the spleen (SI Appendix, Fig. S10 A and B). In line with these observations, there was no difference in weight loss or survival following primary WNV infection in cKO mice, or in viral burden within the brain (SI Appendix, Fig. S10C). In summary, β1 cKO mice, which have a more modest yet significant increase in the size of the T_VM population (Fig. 6D), do not have an impaired effector CD8+ T cell response or atypical clinical outcome to primary WNV infection.

Next, we sought to assess the impact of a dysregulated T_VM population on the protective efficacy of a recall CD8+ T cell response to WNV. Although a protective memory response to WNV relies on both B and T cells, adoptive transfer of CD8+ T cells into Rag2^−/− mice has been shown to provide a significant protective effect (51). Therefore, to determine if the expansion of the T_VM population results in an altered CD8+ T cell memory recall response, we combined WNV infection with adoptive transfer and our constitutively T_VM-enriched FoxP3 driven β1-integrin cKO mouse model (Fig. 7A). Use of the β1-integrin cKO mouse model allowed us to eliminate the additional manipulation of Treg depletion that our enriched T_VM model requires. Furthermore, leveraging adoptive T cell transfers in this setting allowed us to determine the effects of a moderately dysregulated T_VM compartment on the virus-specific CD8+ T cell recall response and disease outcome without the confounding element of antibody-mediated viral clearance upon secondary challenge. Finally, because there was no difference in the frequency or number of WNV-specific effector or memory CD8+ T cells in β1 cKO mice (SI Appendix, Fig. S10 A and B), adoptive transfer of total CD8+ T cells from β1 cKO or WT mice at a memory time point results in a similar number of WNV-specific cells transferred, along with an enhanced frequency of T_VM coming from the β1 cKO compared with WT donors.

Fig. 7.

Dysregulation of T_VM dramatically impairs the recalled protective response to WNV infection. (A) Schematic of experimental design for Fig. 7. β1 cKO or WT mice were infected s.c. with WNV. At ≥30 d postinfection, 5 × 10⁶ total CD8+ T cells were isolated and adoptively transferred into naïve Ly5.1 congenically labeled recipient mice. Recipient mice were subsequently infected with 1,000 pfu WNV ∼24 h following CD8+ T cell infusion. (B) Weight loss and survival were monitored daily. (C) Brain WNV viral load at day 8 or 9 postinfection of recipients of β1 cKO and WT CD8+ T cells was quantified by real-time qRT-PCR. (D–I) Using separate cohorts of mice, spleens and brains were harvested from recipient groups at day 8 or 9 postinfection and analyzed by flow cytometry. (D) Frequency and number of splenic donor-derived CD8+ T cells. (E) Frequency and number of splenic donor-derived NS4b-specific CD8+ T cells. (F) Frequency and number of splenic endogenous NS4b-specific CD8+ T cells. (G) Frequency of splenic donor-derived CD8+ T cells that are KLRG-1+ effector phenotype cells. (H) Frequency of splenic donor-derived or endogenous CD8+ T cells that express IFN-γ after coculture with heat-inactivated WNV. (I) Frequency of splenic donor-derived CD8+ T cells or donor-derived NS4b tetramer+ CD8+ T cells that express PD-1. (J) β1 cKO or WT mice were infected s.c. with WNV. At ≥30 d postinfection, total CD8+ T cells were isolated and 5 × 10⁶ cells of each type were adoptively transferred into naïve Ly5.1 congenically labeled recipient mice. Recipient mice were subsequently infected with 1,000 pfu WNV ∼24 h following CD8+ T cell infusion. Survival was monitored daily. Statistical significance was determined by unpaired t test or log-rank (Mantel–Cox) analysis. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001. ns, not statistically significant.

β1 cKO or WT control mice were infected with WNV for a minimum of 30 d to generate memory CD8+ T cells for subsequent adoptive transfer. Naïve congenic Ly5.1-expressing mice were then i.v. infused with 5 × 10⁶ CD8+ T cells from either β1 cKO or WT controls 1 d before infection with WNV (Fig. 7A). Mice were then monitored for weight loss and survival, and the CD8+ T cell compartment was probed for recall responses to WNV infection. Approximately 8 to 9 d postinfection, β1 cKO-derived memory CD8+ T cell-recipient mice began losing weight at an accelerated rate that continued until mice met the recommended weight loss killing criteria of ≥20% of initial body mass culminating in a significantly increased mortality rate (Fig. 7B). In accordance with this increase in morbidity, there was a significantly increased viral load in the brain of mice that had received CD8+ T cells from β1 cKO donors compared with WT donors, as assessed by qRT-PCR (Fig. 7C). Therefore, while a moderately dysregulated pool of T_VM created by β1 integrin deficiency in Tregs did not affect the primary response or outcome to WNV infection (SI Appendix, Fig. S10), there appeared to be adverse consequences on the protective recall response.

To determine the specific effects on the recall response, we analyzed the CD8+ T cell compartment in recipient mice and quantified the frequency of WNV-specific CD8+ T cells in the transferred and endogenous T cell fractions after infection. For these experiments it was necessary to collect tissues from recipient mice by the previously determined weight loss inflection point, day 9 postinfection (Fig. 7B), to ensure our analyses would be sufficiently powered. We found that donor CD8+ T cells sourced from β1 cKO mice, which have an increased frequency of T_VM (Fig. 6D), significantly expanded compared with WT recipients (Fig. 7D), but that the expansion of recalled WNV-specific CD8+ T cells was significantly decreased in the spleen after challenge (Fig. 7E). Furthermore, the emerging endogenous WNV-specific primary CD8+ T cell response was also significantly limited (Fig. 7F), indicating that the introduction of a competing T_VM population has consequences for both developing effector and recall responses in this model. It has previously been demonstrated that T_VM are more likely to generate memory over effector phenotype cells upon antigen reexposure compared with true memory cells (11), so we next examined KLRG-1 expression on donor-derived CD8+ T cells. We found that CD8+ T cells from β1 cKO mice, that have increased T_VM, had a diminished effector, KLRG-1+ phenotype compared with CD8+ T cells from WT mice (Fig. 7G), which is consistent with this previous report and suggestive of being less equipped to impart functional, recalled protection. In line with this, fewer CD8+ T cells from β1 cKO compared with WT mice produced IFN-γ upon restimulation with heat-inactivated WNV (Fig. 7H), demonstrating a compromised recall response. Further, there were fewer endogenous CD8+ T cells producing IFN-γ in mice that received β1 cKO CD8+ T cells, again indicating that the emerging primary effector response was affected (Fig. 7H). Finally, we examined PD-1 expression, an inhibitory receptor up-regulated at activation and during exhaustion, and found that more donor-derived CD8+ T cells and WNV-specific donor-derived CD8+ T cells expressed PD-1 when comparing β1 cKO to WT cells (Fig. 7I). Considering that IL-15 has been demonstrated to increase PD-1 expression on CD8+ T cells (52), it is possible that this elevated PD-1 expression by β1 cKO CD8+ T cells is due to increased exposure of these cells to IL-15 trans-presented by CD11b+ DC before adoptive transfer.

Because WNV is a neurotropic infection that generates immunity within the CNS as well as in the lymphoid organs, we assessed the recall response within the brain. Similar to the spleen, in the brain there was also an increase in the frequency and number of donor CD8+ T cells sourced from cKO mice (SI Appendix, Fig. S11A), as well as a decrease in the fraction of donor CD8+ T cells that were specific for WNV (SI Appendix, Fig. S11B). Notably, there was no difference in the frequency of WNV-specific cells amoung endogenous CD8+ T cells within the brain (SI Appendix, Fig. S11C).

Finally, we sought to determine whether the CD8+ T cells from immunized β1 cKO mice might be immunosuppressive in addition to being ineffective at preventing death following WNV infection. To address this question, we performed an additional adoptive transfer experiment. CD8+ T cells were isolated from β1 cKO or WT mice that had been infected with WNV for at least 30 d. CD8+ T cells from β1 cKO, WT mice, or β1 cKO + WT mice were then adoptively transferred into naïve Ly5.1 congenically labeled recipient mice, which were subsequently infected with WNV to track survival following infection. Interestingly, we found that cotransfer of an equal number of CD8+ T cells from immunized WT and β1 cKO mice also resulted in increased mortality of recipient WNV-infected mice (Fig. 7J). This finding supports the notion that a failure to control expansion of the T_VM pool can result in suppression of the protective response upon infection. Altogether, our data point to a role for Tregs in limiting IL-15 trans-presentation and the size of the T_VM population to preserve the potential to develop adequate and protective memory T cells directed at microbial antigens upon secondary pathogen exposure.

Discussion

We demonstrate a role for Tregs in restraining expansion of the T_VM population, at least in part through the limitation of IL-15 trans-presentation by CD11b+ DC, and suggest that this restraint is critical to generate protective pathogen-specific memory CD8+ T cell responses upon pathogen reexposure. Further, our data point to a mechanistic role for both CTLA-4 and β1 integrin expressed by Tregs in restricting CD11b+ DC trans-presentation of IL-15. IL-15 is known to play a role in the survival and function of T cells, including Tregs, as it increases the expression of CD25 and Foxp3 (49, 50). We propose that under homeostatic or steady-state conditions, β1 integrin and/or CTLA-4+ Tregs form stable interactions with DCs, and that formation of this strong synapse-like interaction with Tregs leads to DCs with a reduced availability of IL-15 to present to CD8+ T cells. Further, Tregs maintain CD4+ T cell activation in a more quiescent state, thus limiting the ability of CD4+ T cells to aid in CD11b+ DC maturation (27) and production of cytokines such as IL-15 (Fig. 3G). However, upon disruption of Treg function either through a reduction/ablation of their numbers or through selective deletion of CTLA-4 or β1 integrin, the newly unleashed CD11b+ DCs are able to trans-present IL-15 without restraint, and we hypothesize that this at least in part leads to the increased expansion of T_VM. However, further studies are required to definitively demonstrate that restraint of IL-15 trans-presentation is the central mechanism whereby Tregs restrict T_VM expansion.

It has been proposed that CD8+CD122+ cells have regulatory properties similar to Foxp3+ Tregs (53, 54) and thus can assist in maintaining immune homeostasis. While we demonstrated that a large fraction of CD44^hiCD122+ CD8+ T cells that expanded as a result of Treg ablation produce IFN-γ after exposure to proinflammatory cytokines, suggesting that these cells cannot be exclusively suppressive, it remains possible that a portion of these unconventional memory phenotype cells possess regulatory properties within our model. In support of this idea, when we cotransferred CD8+ T cells from WT and β1 cKO immunized mice into recipients that were subsequently infected with WNV, we found that mice suffered from enhanced mortality, similar to mice that received CD8+ T cells from β1 cKO immunized mice alone (Fig. 7J). This suggests that the enlarged pool of T_VM cells from β1 cKO mice suppressed both the cotransferred WNV-specific CD8+ T cells from WT mice, as well as the endogenous WNV-specific response generated in the recipient mice. However, if some of the CD122+ CD8+ T cells present after Foxp3+ Treg depletion do indeed possess regulatory properties, then these cells could assist in providing functional suppressive coverage in the absence of canonical Foxp3+ CD4 Tregs, possibly through an IL-15–dependent mechanism as well. Although future studies are required to determine if unconventional memory CD8+ T cells that expand following Foxp3+ Treg depletion possess suppressive characteristics, it is clear that they cannot substitute for Foxp3+ Tregs, as depletion of Tregs using the Foxp3^DTR mice results in fatal and catastrophic autoimmunity (27).

Importantly, it appears that an unchecked T_VM population has long-lasting consequences to a host’s response to pathogens. The canonical pathway of T cell activation and memory development suggests that upon TCR-dependent recognition of cognate antigen in the presence of costimulatory signals and appropriate cytokine signals, naïve T cells expand to become effector or memory T cells, carrying out the immediate functions of pathogen clearance and becoming future guardians against reinfection, respectively. However, unconventional memory T cells, including T_VM, share many, though not all, characteristics of true memory T cells (19). Likewise, our data contribute further evidence to previous reports demonstrating that T_VM can play a beneficial role in immunity by participating in protection against pathogens (11–13). Specifically, we found that an enhanced T_VM population as a result of previous Treg depletion resulted in no difference in viral load or clinical symptoms, despite a significant deficit in virus-specific CD8+ effector T cells (Fig. 5). This suggests that while a dysregulated T_VM population has consequences for the generation of virus-specific effectors, or perhaps that previous depletion of Tregs had long-lasting effects, it is possible that these T_VM also contributed to viral clearance through bystander protective immunity (14).

In addition to this previously reported protective role for T_VM following microbial infection, we report here data to demonstrate that restraint of the T_VM population is critical to preserve the capacity to generate protective pathogen-specific memory CD8+ T cell responses upon pathogen reexposure. Specifically, we demonstrate that dysregulated T_VM expansion compromises both a burgeoning and recalled pathogen-specific response, leading to increased morbidity and mortality in a WNV model of infection (Fig. 7), with associated defects in the number and phenotype of pathogen-specific memory CD8+ T cells. While the significant increase in frequency and number of donor-derived CD8+ T cells from β1 cKO memory mice (Fig. 7D and SI Appendix, Fig. S11A) is consistent with the previously reported ability of T_VM, which are enriched in β1 cKO mice, to rapidly proliferate following antigen exposure (9), it is possible that this occurred at the expense of expansion (Fig. 7E and SI Appendix, Fig. S11B) and function (Fig. 7 H and I) of WNV-specific memory cells. Further, the reduced production of virus-specific IFN-γ in addition to increased expression of PD-1 by CD8+ T cells from β1 cKO memory mice compared with WT memory mice is consistent with the concomitant increase in brain viral load observed in mice that received β1 cKO memory CD8+ T cells (Fig. 7C).

The direct mechanism for the exacerabation of disease (Fig. 7 B and J) remains unclear but our data provide insight into the competing pressures at play between host and pathogen. Through the use of both human models of infection and murine IL-15 KO models, several groups have evaluated the importance of IL-15 in mediating effective CD44+CD8+ T cell responses in the context of viral infections (55). Specifically, in the context of influenza and HIV, IL-15 is critical in the development of a protective response (56, 57) and directly leads to up-regulation of cytolytic molecules such as granzyme B (58, 59). Recent work is beginning to unconver the critical role that IL-15 may play in WNV infection specifically, as WNV has been shown to actively antagonize STAT5 signaling in DCs (60), with at least one promising proposed vaccination strategy relying on IL-15 encoding plasmid coadminstration (61, 62). Our data suggest that CD8+ T cell access to IL-15 in the course of a viral infection, and in particular WNV infection, is a defining component of host protection and that introducing additional competition in the form of T_VM for this resource in an environment of virally imposed scarcity tips the balance further in favor of WNV. Future experiments will be required to directly determine if the exacerbated disease we observed (Fig. 7B) is due to limited IL-15 availability for virus-specific effector and memory T cells as the result of an expanded pool of T_VM.

We believe that our findings have critical implications for Treg-directed therapeutics, such as CTLA-4 and VLA-4 blockades, as there is potential for secondary long-term effects on the compositional dynamics of the host CD8+ T cell population. In particular, it is possible that such perturbations could result in decreased T cell responsiveness to vaccination and subsequent pathogen exposures. In sum, this study points to a critical role for Tregs in regulating both IL-15 trans-presentation as well as unconventional memory T cell development, with downstream implications for preserving the ability to recall effective memory T cells upon future pathogen encounter.

Materials and Methods

Mice, Injections, and Infections.

Foxp3^DTR, Foxp3^GFP, Itgb1^flox/flox × Foxp3^Cre mice, provided by Alexander Rudensky, Memorial Sloan Kettering Cancer Center, New York, CTLA-4^flox/flox × Foxp3^Cre mice (63), and IL-10^flox/flox × Foxp3^Cre mice (64) were bred under specific pathogen-free conditions onsite in the Fred Hutchinson Cancer Research Center animal facility. Spleens from 8- to 10-wk-old male CC mice were obtained from the Systems Genetics Core Facility at the University of North Carolina at Chapel Hill (65). To ablate Tregs, Foxp3^DTR mice were treated i.p. with 30 μg/kg body weight of diphtheria toxin (DT), followed by a second injection of 10 μg/kg DT on the following day. In some experiments, mice were treated with anti-CD4 (clone GK1.5), anti-CD122 (clone TMb1), or isotype control i.p. at a dose of 100 μg per mouse on two consecutive days. WNV TX-2002-HC (WN-TX) was provided by Michael Gale and Jason Netland, University of Washington, Seattle, WA, and mice were infected s.c. in the rear footpad with 100 to 1,000 pfu. Additional details are described in the SI Appendix, Supplemental Materials and Methods.

All animal experiments were approved by the University of Washington Institutional Animal Care and Use Committee and the Fred Hutchinson Cancer Research Center. The Office of Laboratory Animal Welfare of the NIH has approved Fred Hutchinson Cancer Research Center (no. A3226-01) and the University of Washington (no. A3464-01), and this study was carried out in strict compliance with the Public Health Service Policy on Humane Care and Use of Laboratory Animals.

Flow Cytometry.

Following preparation of single-cell suspensions, cells were plated at 1 × 10⁶ cells per well and stained for surface markers for 15 min on ice. For WNV tetramer staining, cells were stained with the WNV NS4b-H2D^b tetramer (generated by the Immune Monitoring Lab, Fred Hutchinson Cancer Research Center) along with cell-surface markers. Cells were subsequently fixed, permeabilized (Foxp3 Fixation/Permeabilization Concentrate and Diluent; Ebioscience) and stained intracellularly with antibodies for 30 min on ice. Flow cytometry was performed on a BD LSRII machine using BD FACSDiva software. Analysis was performed using FlowJo software.

Statistical Analysis.

All statistical analyses were performed using Prism software (GraphPad Software). Statistical significance was determined using unpaired t tests, ANOVA with Tukey’s multiple comparisons test, or linear regression. P values < 0.05 were considered significant. Survival curve analyses were performed using the log-rank (Mantel–Cox) test.