|Home | About | Journals | Submit | Contact Us | Français|
Interleukin(IL)-2 and inflammation regulate effector and memory cytolytic T-lymphocyte (CTL) generation during infection. We demonstrate a complex interplay between IL-2 and inflammatory signals during CTL differentiation. IL-2 stimulation induced the transcription factor eomesodermin (Eomes), upregulated perforin (Prf1) transcription, and repressed re-expression of memory CTL markers Bcl6 and IL-7Rα. Binding of Eomes and STAT5 to Prf1 cis-regulatory regions correlated with transcriptional initiation (increased recruitment of RNA polymerase II to the Prf1 promoter). Inflammation (CpG, IL-12) enhanced expression of IL-2Rα and the transcription factor T-bet, but countered late Eomes and perforin induction while preventing IL-7Rα repression by IL-2. After infection of mice with lymphocytic choriomeningitis virus, IL-2Rα-deficient effector CD8+ T cells expressed more Bcl6 but less perforin and granzyme B, formed fewer KLRG-1+ and T-bet-expressing CTL, and killed poorly. Thus, inflammation influences both effector and memory CTL differentiation, whereas persistent IL-2 stimulation promotes effector at the expense of memory CTL development.
Naive CD8+ T cells differentiate into effector and memory cytolytic T-lymphocytes (CTL) upon antigen stimulation in the context of infection and inflammation. During this process, the differentiating cells induce the expression of effector proteins such as the cytokine IFNγ, the pore-forming protein perforin, and a family of serine esterases known collectively as granzymes (Cruz-Guilloty et al., 2009; Harty et al., 2000). Perforin and granzymes are essential for cytolytic activity of CTL (Pipkin and Lieberman, 2007). IFNγ, perforin, and granzymes are each induced at the transcriptional level after activation, but distinct regulatory mechanisms appear to be involved—most, if not all, antigen-specific CD8+ T cells express IFNγ and granzyme B during the course of an infection, but only a fraction of these express perforin and IFNγ expression does not necessarily correlate with cytolytic activity (Harrington et al., 2008; Johnson et al., 2003; Peixoto et al., 2007; Zaiss et al., 2008). The expression of all three classes of effector genes in activated cells has been correlated with memory CTL development (Bannard et al., 2009; Harrington et al., 2008; Joshi et al., 2007; Opferman et al., 1999; Sarkar et al., 2008). However, little is known about the signals that regulate transcription of these different classes of effector genes in activated CD8+ T cells, what mechanisms are involved, and how those signals might regulate effector or memory CTL differentiation.
The factors and mechanisms that drive the differential development of effector versus memory CTL during clonal expansion are not completely understood (Badovinac and Harty, 2007; Kaech and Wherry, 2007; Williams and Bevan, 2007). A single brief T cell receptor (TCR) stimulus (signal 1) combined with costimulation (signal 2) can induce an extended period of proliferation, acquisition of effector functions, and ultimately, memory CTL formation (Kaech and Ahmed, 2001; Mercado et al., 2000; van Stipdonk et al., 2001). The duration of TCR stimulation mainly affects the magnitude of effector CD8+ T cell accumulation (Prlic et al., 2006), whereas altered TCR signaling in the context of mutant TCRs affects the balance of effector and memory CTL development (Teixeiro et al., 2009).
IL-2 signals are sometimes considered part of signal 2 (Valenzuela et al., 2002). However, the role of IL-2 signaling in CD8+ T cell differentiation has been difficult to discern in vivo because results from infection of IL-2-deficient mice have differed. This variability may reflect autoimmunity secondary to defective regulatory T cell development in IL-2-deficient mice (Bachmann and Oxenius, 2007; Malek, 2008). More recent studies that avoided these caveats have shown that IL-2 is essential for normal accumulation of effector CD8+ T cells (D’Souza et al., 2002) and for programming the ability of memory CTL to reexpand upon secondary infection in vivo (Bachmann et al., 2007; Williams et al., 2006). In addition, IL-2Rβ, an essential signaling subunit of the IL-2R complex, and STAT5, a transcription factor activated by IL-2R stimulation, are required for normal expression of perforin, granzyme B, and IFNγ in activated CD8+ T cells (Imada et al., 1998; Malek et al., 2001). Although both IL-2 and IL-15 signal through IL-2Rβ, each cytokine has different effects on CTL differentiation; stimulation of IL-2Rβ on CD8+ T cells in cell culture with IL-2, as opposed to IL-15, favors effector rather than memory CTL generation (Carrio et al., 2004; Manjunath et al., 2001), suggesting that how IL-2Rβ is activated affects gene expression.
An inflammatory signal (signal 3) provided by cytokines such as type I interferons and/or IL-12 is essential for normal effector and memory CTL generation. In different settings, signal 3 has been shown to be crucial for inducing CTL effector functions (Curtsinger et al., 2003; Mescher et al., 2006), for driving antigen-activated CD8+ T cells toward a short-lived effector cell fate (Joshi et al., 2007), and for programming contraction of the effector cell population (Badovinac et al., 2004). At the same time, type I interferons and IL-12 have also been shown to be required for memory CTL development (Xiao et al., 2009). Thus, a number of extracellular signals regulate effector and memory CTL development in vivo, but it is still unclear how these signals act individually and in combination to regulate gene expression programs in activated CD8+ T cells and control their differentiation.
In this study, we used a simple cell-culture system to investigate how the strength of IL-2R signaling regulated perforin gene (Prf1) expression after TCR activation of naive CD8+ T cells and how inflammatory signals (CpG and IL-12) influenced this process. We also investigated the requirement of IL-2 signaling for CTL responses in vivo in a setting that avoids autoimmunity; mixed bone marrow chimeric mice reconstituted with both wild-type and Il2ra-deficient cells were infected with lymphocytic choriomeningitis virus (LCMV). We found that persistent IL-2 stimulation induced Prf1 transcription and promoted effector CTL differentiation, whereas inflammatory signals had distinct effects. Our data indicate that inflammation and IL-2 signals interface in a complex way and that their relative strengths are likely to regulate the development of both effector and memory CTL.
To begin investigating how the “strength” of IL-2R signaling regulated the effector functions of CD8+ T cells, naive CD8+ T cells were cultured for 2 days with a strong TCR stimulus in conjunction with CD28 costimulation, then removed from stimulation and recultured in low (10 U/ml) or high (100 U/ml) IL-2 concentrations. As an indirect measure of IL-2R signal intensity, we assessed cell-surface expression of IL-2Rα, which is regulated in part through a well-characterized positive feedback loop in which STAT5 directly targets the Il2ra gene (Nakajima et al., 1997). All cells displayed uniformly high expression of IL-2Rα until day 4, as judged by flow cytometry; after day 4, cells cultured in high IL-2 maintained high IL-2Rα expression, whereas cells cultured in lower IL-2 concentrations showed decreased IL-2Rα expression (Figure S1A available online). Evaluating IL-2 signaling more quantitatively as the content of tyrosine-phosphorylated STAT5 (Figures S1B and S1C), we found that phospho-STAT5 amounts were high in cell populations cultured in either high or low IL-2 until day 5, after which they were sustained in high IL-2 but declined abruptly in cells cultured in low IL-2 (Figure S1B). At the single-cell level, flow cytometric analysis showed that 94% of the cells cultured in high IL-2, but only 15% of the cells cultured in low IL-2, maintained high amounts of phospho-STAT5 on day 6 (Figure S1C). The mean fluorescence intensities (MFIs) of phospho-STAT5 staining in each population suggested that there was an approximately 5- to 10-fold difference in IL-2 “strength” when comparing 10 versus 100 U/ml IL-2 in our cultures.
The two essential effectors of the cytolytic program, perforin and granzyme B, displayed distinct expression kinetics in activated CD8+ T cells under high or low IL-2 conditions (Figures 1A and 1B). Naive CD8+ T cells expressed perforin mRNA but very little protein, and neither was induced by 2 days of TCR stimulation. After removal from TCR stimulation, perforin mRNA expression decreased until day 4, regardless of IL-2 concentration, then continued to decrease in cells cultured in low IL-2, but increased strongly, beginning at day 5, in cells cultured with high IL-2 (Figure 1A). Similarly, perforin protein was strongly induced after day 4, when cells were cultured in high IL-2 (Figure 1B). In contrast, granzyme B mRNA was undetectable in naive T cells but was strongly induced by TCR stimulation. Although granzyme B expression was maintained in both low and high IL-2, granzyme B mRNA and protein were both more highly expressed in cells cultured in high IL-2 (Figure 1A and 1B). The concentration of IL-2 in culture did not affect the ability of CD8+ T cells to express IFNγ and TNF upon brief restimulation (Figure 1C), consistent with previous studies (Bachmann et al., 2007; Williams et al., 2006), although IL-2Rβ was likely required initially (Malek et al., 2001).
Strong IL-2 signals were necessary to induce cytolytic function. Activated CD8+ T cells cultured until day 6 in high IL-2 rapidly killed antigen-pulsed target cells in a short-term (30 min–2 hr) assay (Cruz-Guilloty et al., 2009), without prior restimulation (Figure 1D and Figure S1D), whereas those cultured in low IL-2 did not (Figure 1D). Nevertheless, cells cultured in low IL-2 were fully functional in all other respects—they accumulated exponentially for at least 6 days, with kinetics equivalent to those of cells cultured in high IL-2 (Figure 1E), and displayed efficient cytolytic activity after a brief restimulation (Figure 1F). We attribute the de novo increase in cytolytic activity to the strong induction of perforin in response to restimulation (Figure 1G).
The strength of IL-2 stimulation regulated the expression of additional genes in CD8+ T cells. As expected, initial stimulation of naive P14 TCR transgenic CD8+ T cells, primed either by TCR crosslinking or by coculture with splenic APCs loaded with GP33 peptide, induced surface markers characteristic of antigen-stimulated T cells (CD44hi, CD25hi, CD69hi, CD127−). After day 4, however, cells cultured in low IL-2 quickly converted to a central memory-like phenotype (CD25lo, CD122hi, CD127 hi, CD62Lhi) and were not cytolytic, whereas the majority of those cultured in high IL-2 retained a characteristic effector phenotype (CD25 hi, CD122hi, CD127−, CD62Llo or −) and killed efficiently (Figures S2A–S2C and data not shown).
We examined IL-7Rα (CD127) regulation in more detail because effector cells that reinduce IL-7Rα at the peak of the response to some acute infections are enriched for cells that form long-lived memory CTL (Kaech et al., 2003). TCR stimulation completely downregulated surface IL-7Rα expression, and high IL-2 prevented IL-7Rα re-expression. However, IL-7Rα was reinduced after day 4 in a dose-dependent fashion if the IL-2 concentration was reduced (Figures 2A and 2B). IL-7Rα re-expression did not correlate with cell division, because all cells completely diluted CFSE (>7 divisions) by day 4 (Figure 2C) and accumulated similarly until day 5, by which time IL-7Rα had already been reinduced (Figure 2B; also see Figure 2D). At the mRNA level, IL-7Rα mRNA was repressed and perforin mRNA was reciprocally induced at the same high concentrations of IL-2 (Figure 2D). Consistent with this observation, surface expression of IL-2Rα and IL-7Rα chains was mutually exclusive: even at low IL-2 concentrations (1–10 U/ml), IL-2Rα-positive cells tended to be IL-7Rα negative (Figure 2D). Enforced expression of IL-2Rα by retroviral transduction confirmed that low IL-2 could repress IL-7Rα in cells that expressed higher IL-2Rα (Figures S2D and S2E). Thus, expression of the high-affinity IL-2Rαβγ receptor is sufficient to repress IL-7Rα re-expression, even in cells cultured in low concentrations of IL-2.
Distinct populations of effectorCD8+T cells are present early after acute viral infection that have different potential to become memory CTL, and they can be distinguished by their ability to produce IL-2 and to proliferate upon secondary stimulation ex vivo (Joshi et al., 2007; Sarkar et al., 2008). These two attributes were controlled by different strengths of IL-2 stimulation in culture. IL-2 production was not affected by different IL-2 concentrations through day 4 (Figure 3A), but by day 6, cells cultured in low IL-2 produced much more IL-2 upon restimulation than cells cultured in high IL-2 (Figure 3A); they also proliferated more strongly in response to low levels of stimulation (Figure S2F).
Studies of gene-disrupted mice have shown that Blimp-1 and Bcl6 are required in vivo for the development of effector and memory CTL, respectively (Ichii et al., 2002; Rutishauser et al., 2009). We therefore examined whether IL-2 regulated expression of these two transcription factors. Blimp-1 mRNA was not expressed in naive cells and was only induced upon removal from the TCR stimulus and culture in IL-2 (Figure 3B). By day 6, Blimp-1 mRNA expression was not maintained in low IL-2, but Blimp-1 remained expressed in high IL-2. Conversely, Bcl6 mRNA was expressed in naive CD8+ T cells, downregulated during TCR stimulation, and re-expressed after day 4 in low, but not high, IL-2, and this correlated inversely with expression of IL-2Rα mRNA (Figure 3B). Thus, the inverse expression of Blimp-1 and Bcl6 typical of effector cells in vivo was regulated by the degree of IL-2 stimulation.
The T-box transcription factors T-bet and Eomes are both required for normal CTL differentiation (Intlekofer et al., 2008; Intlekofer et al., 2005; Joshi et al., 2007). We asked whether IL-2 stimulation regulated expression of T-bet and Eomes upon CD8+ T cell activation. Expression of T-bet and Eomes mRNA and protein was low or undetectable in naive CD8+ T cells. Two days of TCR stimulation strongly induced T-bet (Figure 3C), and its expression was maintained in an IL-2-independent manner through day 6 (Figure 3C). In contrast, Eomes was induced by day 4 and was upregulated through day 6 in a manner strongly dependent on IL-2 concentration in culture (Figure 3C). By switching cells from low to high IL-2 and vice versa on day 4 of culture, we confirmed that Eomes and perforin expression were both regulated by IL-2, whereas T-bet expression was not (Figure 3D and data not shown).
The kinetics and IL-2 dependence of Eomes expression was closely paralleled by perforin expression (Figures 3C and and2D),2D), leading us to ask whether Eomes could induce perforin without strong IL-2 stimulation. Under these conditions, endogenous Eomes was not highly expressed and STAT5 phosphorylation was low (Figures 3C and 3D and Figures S1B and S1C). We activated purified naive CD8+ T cells from wild-type and T-bet-deficient B6 mice for 2 days, transduced them with a retrovirus expressing a hyperactive form of Eomes (Eomes-VP16) linked to IRES-GFP, then cultured the cells in low IL-2. Cells infected with Eomes-VP16, but not the control GFP retrovirus, strongly expressed perforin mRNA on day 6 (Figure 3E). Eomes-VP16 had no effect on the expression of granzyme B mRNA (Figure 3C). Control experiments showed that the effect of Eomes could not be attributed to a large increase in IL-2Rα or IL-2Rβ expression (Figure S3). Thus, Eomes most likely acts directly at Prf1 after induction by IL-2.
We tested this hypothesis by chromatin immunoprecipitation (ChIP) (Figure 4). Endogenous Eomes and STAT5 proteins both bound to Prf1 on day 6 in cells cultured in high IL-2, although their binding patterns were distinct. STAT5 primarily bound DNase I hypersensitive (DHS) site 4 at –1 kb and within intron 3 downstream of DHS7, comparable to its binding to the TSS of Il2ra, a known STAT5 target gene (Nakajima et al., 1997) (Figure 4B). Eomes bound to the Prf1 transcription start site (TSS) and to a lesser degree DHS4 (Cruz-Guilloty et al., 2009), comparable to its binding to the TSSs of Il2rb and Ifng, genes that are direct targets of Eomes (Intlekofer et al., 2005). Eomes also bound modestly near DHS7 (Figure 4B). Compared to cells cultured in high IL-2, cells cultured in low IL-2 showed considerably less binding of STAT5 and Eomes to Prf1, consistent with their lower phospho-STAT5 content and lower Eomes expression (Figures 4A and 4B). However, the amount of binding was still somewhat greater than that observed at the TSS of ll4 and Hprt, genes that are not expressed in CD8+ T cells or that are expressed, but not regulated, by IL-2, respectively. Therefore, even low binding of phospho-STAT5 and Eomes to the Prf1 gene may be relevant.
TCR stimulation of naive CD4 T cells leads them to differentiate into distinct subsets with respect to expression of cytokine genes, a process that involves differential chromatin remodeling of specific cytokine gene loci (Ansel et al., 2006; Avni et al., 2002; Fields et al., 2002; Grogan et al., 2001). Because activated CD8+ T cells differentiated in high IL-2 express at least 20 times more perforin mRNA and exhibit a differentially remodeled DNase I hypersensitivity (DHS) site pattern in the Prf1 locus compared to differentiated Th1 cells (Pipkin et al., 2007), we asked whether differential perforin expression by CD8+ T cells in response to low versus high IL-2 was accompanied by differences in the DHS site pattern surrounding Prf1 (Figure S4) (Pipkin and Lichtenheld, 2006). Unexpectedly, we found that there was no difference in the pattern of DHS sites across ~200 kb of the Prf1 locus regardless of whether the cells had been differentiated in low or high IL-2 (Figure S4 and data not shown). Thus, IL-2 signals did not appear to “open” IL-2 specific cis-regulatory regions in the Prf1 gene but, rather, acted on a previously opened locus.
We asked whether IL-2 stimulation regulated recruitment of RNA polymerase II (Pol II) (Figure 4C). Relative to cells differentiated in low IL-2, cells differentiated in high IL-2 exhibited dramatically increased recruitment of Pol II to the Prf1 TSS at day 6, along with a small increase in Pol II density across the gene body (Figure 4C). Conversely, Pol II was recruited efficiently to the Il7ra TSS in CD8+ T cells cultured in low IL-2, but not in cells cultured in high IL-2 (Figure 4D). In both high and low IL-2, the level of Pol II binding at the Prf1 TSS was greater than that observed at the TSS of the muscle-specific gene MyoD1, which is not expressed in T cells (Figure 4C). Pol II binding at the TSS of the housekeeping gene Hprt (not regulated by IL-2) was comparable to the level of Pol II binding at the Prf1 TSS in high IL-2 (data not shown). Thus, the strength of IL-2 stimulation determined whether Pol II was recruited to the Prf1 and Il7ra genes in activated CD8+ T cells.
To examine the effect of IL-2 upon effector CD8+ T cell differentiation, we generated mixed bone marrow chimeras by transferring a mixture of congenically distinct wild-type and IL-2Rα-deficient bone marrow cells into lethally irradiated B6 mice and infecting them with LCMV (Bachmann et al., 2007; Williams et al., 2006). Eight days after infection, CD44hi, antigen-experienced Il2ra+/+, and Il2ra−/− CD8+ T cells were sorted and their cytotoxic activity was analyzed in a standard chromium release assay using GP33 peptide-loaded targets. Despite equal representation of GP33-specific cells among wild-type and Il2ra−/− effectors based on tetramer staining (data not shown), Il2ra−/− CD8+ T cells were severely defective in their capacity to lyse targets: 10-fold more effectors were needed for comparable target cell lysis (Figure 5A). This impairment correlated with decreased expression of perforin and granzyme B mRNA in sorted Il2ra−/− CD44hi cells compared to wild-type cells (Figure 5B). In addition, on day 6 after infection, when responding wild-type CD8+ T cells expressed maximal IL-2Rα, fewer total Il2ra−/− CD8+ T cells expressed granzyme B, T-bet, or KLRG-1, compared to wild-type cells, whereas proportionally more knockout cells expressed IL-7Rα and CD62L (Figure 5C and data not shown) (Williams et al., 2006). Furthermore, Il2ra−/− CD8+ T cells contained a much larger proportion of effector cells that coproduced IFNγ and IL-2 upon restimulation and expressed much more Bcl6 mRNA (Figures 5D and 5E). Thus, CD8+ T cells that could not signal via high affinity IL-2Rs differentiated inefficiently into effector CTL, even in an environment containing normal cells and inflammatory signals.
To determine the in vivo fate of in vitro-activated CD8+ T cells, we transferred primed P14 TCR-transgenic CD8+ T cells that had been cultured in high or low IL-2 to naive B6 mice (mixed at a 1:1 ratio, Figure 6). Engraftment of cells cultured in high and low IL-2 18 hr after transfer was equivalent in both the spleen (Figure 6A) and the lung (data not shown), and this equal representation was maintained when the mice were bled 10 days later (data not shown). When the same mice were analyzed 35 days after transfer (41 days after initial TCR activation in vitro), the P14 cells that had been cultured in low IL-2 maintained their frequencies, whereas those that had been cultured in high IL-2 had decreased in frequency and absolute number in all tissues analyzed (Figures 6A and 6B). Although cells from the high IL-2 cultures maintained some CD62Llo cells both 10 and 35 days after transfer, cells from the low IL-2 cultures were essentially all CD62Lhi (Figure 6C). Thus, the differential phenotypic programming observed in vitro was preserved upon in vivo transfer, indicating that T cells that had received higher and more prolonged IL-2 signals were impaired for memory differentiation.
As a functional read-out for the fitness of the memory P14 cells derived from culture in varying concentrations of IL-2,we infected recipient mice with LCMV 35 days after transfer. Five days after infection, both populations of P14 T cells robustly expanded in the spleen, lymph nodes ,and liver; despite the fact that the P14 cells differentiated in low IL-2 were represented at an increased frequency prior to rechallenge, they did not exhibit an advantage upon re-expansion (Figures 6D and 6E). Thus, with CD8+ T cells from this in vitro system, alterations in IL-2 exposure altered their differentiation to memory but did not subsequently influence their ability to respond to secondary stimulation.
Given that both IL-2 and inflammation regulate CTL development, we examined how inflammatory stimuli interfaced with IL-2 signals to control gene expression in activated CD8+ T cells. We primed naive P14 CD8+ T cells with APCs and GP33 peptide, with or without inflammation (CpG), followed by culture in high or low IL-2 with or without IL-12 (Figure 7). The kinetics of perforin and granzyme B mRNA expression in cells primed with peptide and APC were similar to those previously observed after priming with plate-bound αCD3 and αCD28 (Figure 7A, compare with Figure 1A). However, compared to cells stimulated with GP33 peptide alone, cells stimulated in the presence of CpG followed by culture in IL-12 and low IL-2 upregulated T-bet and IL-2Rα mRNA, as well as CD25 surface expression (Figures 7A and 7B). IL-2Rα upregulation occurred prior to the initial cell division (Figure 7C) in a manner that was largely resistant to treatment with blocking antibodies to IL-2 (Figure 7D; Figure S5), suggesting that IL-2Rα expression was induced prior to, and independently of, feedback through IL-2/STAT5. In addition, inflammation increased the peak amount of IL-2Rα mRNA and surface protein expression in cells cultured in low IL-2 and sustained CD25 expression beyond the precipitous drop otherwise seen at day 3 in the absence of inflammation (Figures 7A and 7B). Although inflammation itself enhanced IL-2Rα expression, high IL-2 still upregulated IL-2Rα expression further in cells primed with inflammation (Figure 7A). Together, these data showed that by increasing the early expression of IL-2Rα, inflammation enhanced the duration and responsiveness of activated CD8+ T cells to IL-2 signals.
Nevertheless, some of the effects of high IL-2, namely upregulation of perforin and Eomes, were different when inflammatory signals were present (Figure 7A). Inflammatory signals during priming impaired the ability of high IL-2 to induce Eomes and perforin at later times (Figure 7A; Figure S5). In addition, they impaired the ability of high IL-2 to repress IL-7Rα mRNA expression. Thus, inflammatory stimuli prolonged IL-2 responsiveness but, paradoxically, also interfered with the ability of IL-2 to regulate certain genes.
Here, we tested the hypothesis that different “strengths” of IL-2 stimulation induce different transcriptional responses that alter CD8+ T cell differentiation. We showed that increasing IL-2R signal strength promoted effector CTL differentiation in a simple cell-culture system and that IL-2R signals were required for normal gene expression and accumulation of effector CTL during viral infection. Moreover, we showed that inflammatory stimuli (CpG/IL-12) potentiated early IL-2 responsiveness but simultaneously altered the transcriptional programs induced by IL-2. Our findings are consistent with those of an accompanying study (Kalia et al., 2010) reporting that viral infection induces some effector cells to express more IL-2Rα for a longer duration than others and that these cells are both more responsive to IL-2 and more prone to differentiate into effector rather than memory CTL.
CD8+ T cell differentiation upon infection is complex and integrates multiple signals including strength of TCR stimulus and costimulation, IL-2R signals, and inflammation (Williams and Bevan, 2007). We found that IL-2 regulated perforin and granzyme B expression independently of inflammation (CpG and IL-12), a result that seems to contradict previous studies showing that inflammatory signals were obligatory for inducing cytolytic function (Curtsinger et al., 2003). The difference could lie in the strength of initial TCR signals: Curstsinger et al. primed naive cells with MHC I-peptide and B7-coated microspheres to mimic APC, and under these conditions, initial IL-2Rα induction was low and transient and strongly dependent on inflammation for sustained expression in the presence of low IL-2 concentrations (Curtsinger et al., 2003; Valenzuela et al., 2002). In contrast, we used plate-bound anti-CD3 and anti-CD28 or live APC with high concentrations of peptide, conditions that maintained IL-2Rα expression without a requirement for inflammatory signals. Our studies revealed that IL-2 enhanced perforin transcription, whereas CpG and IL-12 did not. We hypothesize that in conditions of weak TCR stimulation, inflammation enables the IL-2 responsiveness that is necessary for expression of perforin and granzyme B; however, inflammation is less critical under conditions of strong TCR stimulation.
Increased inflammation promotes “short-lived” effector CTL development, programs clonal contraction, and induces CTL effector functions (Badovinac et al., 2004; Curtsinger et al., 2003; Joshi et al., 2007). Somewhat counterintuitively, inflammatory signals also appear to be required for memory CTL development (Xiao et al., 2009). In the simplified cell-culture setting, our results showed that inflammation could regulate both effector and memory CTL development in the context of different strengths of IL-2 signals. On the one hand, inflammatory signals increased T-bet and IL-2Rα expression even in low IL-2 conditions, whereas without inflammation, strong IL-2R signals increased Eomes, perforin, and granzyme B expression. Thus, both inflammation and IL-2 promoted aspects of effector CTL differentiation. On the other hand, however, as best seen in our high IL-2 conditions, inflammatory signals attenuated the late expression of perforin and Eomes and increased IL-7Rα expression, counteracting certain effects of persistent IL-2 signals. Thus, the generation of effector and memory CTL during infection in vivo is likely to be determined by the relative balance of TCR, costimulatory, inflammatory, IL-2, and IL-15 signals encountered by individual responding cells and possibly additional unidentified signals as well.
Analysis of CD8+ T cell differentiation after activation in vitro and the CD8+ T cell response to LCMV in the absence of IL-2R signals in vivo supports two conclusions regarding the roles of IL-2. First, IL-2Rα-deficient CD8+ T cells in vivo resembled cells cultured in low concentrations of IL-2: both types of cells showed reduced perforin and granzyme B expression, as well as premature re-expression of IL-7Rα, CD62L, and Bcl6. Thus, strength of IL-2 stimulation directly regulated expression of genes characteristic of both effector and memory CTL. Second, IL-2Rα-deficient CD8+ T cells developed proportionally fewer KLRG-1-and T-bet-expressing effector cells relative to wild-type CD8+ T cells near the peak of the LCMV response. This was unlikely to be due to a direct effect of IL-2 upon T-bet expression because IL-2 did not regulate T-bet in CD8+ T cells primed in culture. Therefore, IL-2 appeared to drive selectively the accumulation of primed effector CD8+ T cells that had already induced T-bet. These two conclusions are consistent with those of an accompanying study, demonstrating that CD25hi effector CD8+ T cells (i.e., more IL-2-responsive) generated during LCMV infection exhibited a more effector-like gene expression profile, proliferated more extensively at the tail-end of the effector phase, and contributed inefficiently to the formation of memory CTL relative to CD25lo effector cells (Kalia et al., 2010).
Our results help to clarify the regulation of T-bet and Eomes during CD8+ T cell activation. As expected, T-bet was induced upon TCR stimulation; however, we found that Eomes was induced in response to IL-2 several days after removal from the TCR stimulus. The induction of T-bet by TCR signals, and Eomes by IL-2 stimulation, might partially explain the sequential upregulation of T-bet and Eomes during infection (Intlekofer et al., 2005). In addition, we found that the presence of CpG at priming was sufficient to repress later Eomes induction by IL-2 (data not shown). Previous studies have shown that T-bet and Eomes are inversely regulated in activated CD8+ T cells by IL-12 (Takemoto et al., 2006). However, CpG stimuli did not enhance T-bet expression, unless exogenous IL-12 was also provided (Joshi et al., 2007; Szabo et al., 2002). Thus, T-bet-independent pathways are likely to prevent Eomes upregulation by effector cells.
Both T-bet and Eomes can positively regulate IL-2Rβ expression in CTL (Intlekofer et al., 2008; Intlekofer et al., 2005; Joshi et al., 2007). Here, we showed that IL-2 induced Eomes. In conjunction with the previous studies, our results are consistent with the hypothesis that T-bet and Eomes operate in feedback loops. For example, T-bet could initiate elevated IL-2Rα expression to enable later IL-2-induced Eomes upregulation that feeds back to increase Il2rb even further; Eomes together with phospho-STAT5 would then feed forward to activate late effector genes such as Prf1. Our data suggest that inflammation could be important to initiate this process by enhancing T-bet and IL-2Rα expression. However, we also showed that strong inflammation inhibited Eomes upregulation by IL-2 and, thus, could result in effector cells that fail to upregulate Eomes and IL-2Rβ expression efficiently. Thus, the paradoxical positive and negative effects of inflammation on the development of both effector and memory CTL in vivo could relate in part to positive and negative effects of inflammation on IL-2-regulated gene expression.
Finally, our data confirm previous studies implicating IL-2 as an important regulator of Prf1 transcription (Zhang et al., 1999) and provide additional mechanistic insights as to how IL-2 mediates Prf1 gene activation. Under the conditions we examined, IL-2 did not induce global chromatin remodeling of previously reported DHS sites (Pipkin et al., 2007) but acted on a previously “opened” locus. Our data suggest that the main mechanism of gene activation by IL-2 is increased transcription initiation via increased RNA Pol II recruitment to the TSS, as opposed to stimulated elongation of preinitiated Pol II complexes. In addition, RNA Pol II loading at the promoter was coordinated with binding of endogenous STAT5 at an enhancer located at −1 kb, previously shown to be controlled by STAT5 in transient reporter assays (Zhang et al., 1999). Moreover, Eomes bound to the −1 kb enhancer and the TSS concurrently with STAT5 and RNA Pol II, suggesting that STAT5 and Eomes might participate in recruiting RNA Pol II. At the same time, RNA Pol II was recruited away from the Il7ra transcription start site under persistent IL-2 signaling, suggesting that IL-2-mediated regulation of RNA Pol II recruitment to and away from accessible genes might be a general mechanism for controlling the differentiation of activated CD8+ T cells.
CD8+ T cells were isolated from 4- to 12-week-old Tcrα −/− × P14 TCR transgenic (Taconic), C57BL6/J or Tbx21−/− mice (Jackson Laboratory). P14 CD8+ T cells were uniformly naive based on staining with antibodies recognizing CD25, CD44, CD62L, and CD122. All mice were maintained in specific pathogen-free barrier facilities and used according to protocols approved by the Immune Disease Institute and Harvard Medical School animal care and use committees. For bone marrow chimeric mice and LCMV infection, 6- to 8-week-old B6.SJL-PtprcaPep3b/BoyJ (CD45.1) mice were purchased from the Jackson Laboratory. B6.SJL × B6 (CD45.1/CD45.2, heterozygote) and B6.129S4-Il2ratm1Dw/J (Il2ra−/−, CD45.2) mice were bred in our facility at the University of Washington (Seattle, WA) under specific pathogen-free conditions. Bone marrow preparations from the femur and tibia of donor mice were incubated with anti-CD3-, CD4- and CD8-biotin, followed by anti-biotin magnetic beads, and applied to LS columns for depletion of T cells (Miltenyi). 2–5 × 106 total T cell-depleted bone marrow cells, containing a roughly 1:1 mixture of wild-type and IL-2Rα-deficient bone marrow, were injected intravenously into lethally irradiated B6.SJL hosts (1000 Rad). Mice were infected with 2 × 105 PFU of LCMV-Armstrong intraperitoneally 10–12 weeks posttransplant.
CD8+ T cells were purified (>95% purity) by negative selection (Invitrogen) from RBC-lysed single-cell suspensions from pooled spleen and lymph node cells. For stimulation, purified CD8+ T cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% heat-inactivated fetal calf serum, 2 mM L-glutamine, penicillin-streptomycin, nonessential amino acids, sodium pyruvate, vitamins, 10 mM HEPES, and 50 µM 2-mercaptoethanol as 4 × 105 cells/cm2 in T25 flasks coated with anti-CD3 (clone 2C11) and anti-CD28 (clone 37.51) (1 µg/ml) by precoating with 300 µg/ml goat anti-hamster IgG. After 48 hr, cells were removed from the TCR signal and recultured at a concentration of 5 × 105 cells/ml in media supplemented with the indicated concentration of recombinant human IL-2 (rhIL-2). For activation of P14 CD8+ T cells with APCs, combined spleen and lymph nodes were used to generate single-cell suspensions after RBC lysis. Washed cells resuspended in complete in T cell media as 4 × 106 cells/ml and GP33 peptide was added to 1 µM final concentration and were untreated or supplemented with 3 ng/ml CpG (ODN1826, Invivogen, Inc.). After 48 hr, CD8+ T cells were purified by negative selection for direct analysis, or cocultures were counted and recultured as 5 × 105 cells/ml in rhIL-2-containing media either with or without a single dose of IL-12 to 5 ng/ml. Purified naive P14 CD8+ T cells were also activated with mitomycin C-treated APCs from B6 Tcra−/− spleens and gave similar results (data not shown). Every 24 hr, cells were counted and readjusted to 5 × 105 cells/ml with fresh media containing 10 or 100 U/ml rhIL-2. Viral supernatants were generated by transfection of Phoenix packaging cells and concentration by overnight centrifugation at 6000 × g. At ~42 hr after TCR priming of 106 CD8+ T cells in 1 ml per well in 12-well plates, the culture media was replaced with complete media supplemented with 8 µg/ml polybrene containing concentrated virus. The plates were centrifuged at 700 × g for 1 hr at room temperature and then incubated at 37°C for 5 hr. Retroviral constructs for Eomes-VP16 and the MIG control empty vector were a gift from Dr. Steve L. Reiner (Intlekofer et al., 2005).
For flow cytometric killing analysis, EL-4 thymoma target cells were loaded with 0 or 1 µM GP33 peptide for 2 hr before a 2 hr coincubation with P14 CD8+ T cells in 96-well round-bottom plates (Cruz-Guilloty et al., 2009). After coincubation, cells were stained with AnnexinV-FITC and anti-CD8+-APC. For standard killing assays, bulk, polyclonal CD44hi wild-type, and Il2ra−/− CD8+ effector T cells from spleens of mixed BMC mice, infected 8 days prior with LCMV, were FACS-sorted based on congenic marker expression. EL-4 cells were labeled with 51Cr (Perkin Elmer) for 1 hr at 37°C, washed extensively, and then incubated with effector cells (adjusted for proportion of antigen-specific cells based on staining an aliquot of each population with Db-GP33 tetramer) with either 10 µM GP33 peptide or no peptide for 4 to 5 hr at 37°C. Percent specific lysis from duplicate or triplicate wells was determined relative to the spontaneous (targets alone) and total release (2% Triton X-100 detergent) controls, as described previously (Tyznik et al., 2004).
Chromatin immunoprecipitation (ChIP) was performed as described (Cruz-Guilloty et al., 2009). Briefly, formaldehyde-fixed chromatin was isolated from 2–4 × 107 CD8+ T cells for each immunoprecipitation and was sonciated to yield 0.5–1 kb chromatin fragments. Immunoprecipitation was performed by adding optimized antibody amounts (see Supplemental Experimental Procedures), followed by overnight incubation at 4°C; protein A-sepharose beads were added for the last 3 hr of the incubation. After bead washes, chromatin was treated with RNase A for 1 hr at 37°C, followed by addition of Proteinase K and overnight incubation at 65°C to reverse crosslinking, and DNA was purified with QIAquick columns (QIAGEN). For real-time PCR detection of immunoprecipitated targets using the SYBR Green PCR kit, a standard curve was generated for each sample based on amplification of serial dilutions of input DNA. ChIP DNA PCR reactions were performed in duplicates. Agarose gel analysis and melt curves were analyzed to ensure amplification of specific target sequences.
Additional Experimental Procedures can be found in Supplemental Experimental Procedures available with this article online.
We thank Dr. S. Reiner for generously providing the Eomes-VP16 retroviral construct. This work was funded by NIH grants AI44432 and AI707088 (to A.R.) and AI19335 (to M.J.B.); University of Miami Developmental Center for AIDS Research 5P30AI073961 (to M.G.L.); and the National Cancer Institute F32CA126247 (to M.E.P). F.C.G. was a predoctoral fellow of the Ryan Foundation and was supported by a Ford Foundation Predoctoral Fellowship.
The Supplemental Information include five figures, Supplemental Experimental Procedures, and one table and can be found with this article online at doi:10.1016/j.immuni.2009.11.012.