IRF4 Regulates the Ratio of T-Bet to Eomesodermin in CD8+ T Cells Responding to Persistent LCMV Infection

CD8+ T cell exhaustion commonly occurs in chronic infections and cancers. During T cell exhaustion there is a progressive and hierarchical loss of effector cytokine production, up-regulation of inhibitory co-stimulatory molecules, and eventual deletion of antigen specific cells by apoptosis. A key factor that regulates T cell exhaustion is persistent TCR stimulation. Loss of this interaction results in restoration of CD8+ T cell effector functions in previously exhausted CD8+ T cells. TCR stimulation is also important for the differentiation of Eomeshi anti-viral CD8+ effector T cells from T-bethi precursors, both of which are required for optimal viral control. However, the molecular mechanisms regulating the differentiation of these two cell subsets and the relative ratios required for viral clearance have not been described. We show that TCR signal strength regulates the relative expression of T-bet and Eomes in antigen-specific CD8+ T cells by modulating levels of IRF4. Reduced IRF4 expression results in skewing of this ratio in the favor of Eomes, leading to lower proportions and numbers of T-bet+ Eomes- precursors and poor control of LCMV-clone 13 infection. Manipulation of this ratio in the favor of T-bet restores the differentiation of T-bet+ Eomes- precursors and the protective balance of T-bet to Eomes required for efficient viral control. These data highlight a critical role for IRF4 in regulating protective anti-viral CD8+ T cell responses by ensuring a balanced ratio of T-bet to Eomes, leading to the ultimate control of this chronic viral infection.


Introduction
Acute virus infections are characterized by the formation of robust CD8 + T cell effector responses followed by the generation of immunological memory. Both CD8 + effector T cells as well as CD8 + memory cells produce a variety of cytokines and cytotoxic molecules, and have high proliferative capacity [1]. In contrast, during chronic viral infections, high viral loads cause CD8 + T cell exhaustion that is characterized by hierarchal loss of effector functions and eventual deletion of antigen-specific cells [2][3][4]. The remaining virus-specific CD8 + T cells lose the ability to make IFNγ, TNFα, and IL-2, and up-regulate high levels of inhibitory receptors such as PD-1 and LAG-3. Eventually the cells become completely dysfunctional and are deleted by apoptosis [2]. T cell exhaustion was initially thought to be a viral immune evasion mechanism, but recent studies have indicated that it serves to protect the host from T cell-mediated immunopathology [5,6].
Many factors regulate T cell exhaustion. The expression of the immuno-suppressive cytokine IL-10 and inhibitory co-receptors like PD-1 enhance T cell exhaustion, whereas help from CD4 + T cells aids in the restoration of CD8 + T cell function [7][8][9][10]. Persistent T cell signaling due to high viral loads and increased MHC-I presentation is detrimental as well as beneficial during chronic infection. Increased antigen presentation results in reduced numbers and impaired function of anti-viral CD8 + T cells; however, loss of this interaction also leads to poor viral control [4]. Antigen is also required for the long-term maintenance of virus-specific cells during chronic infections, as these cells do not undergo homeostatic proliferation in response to IL-7 and IL-15; instead, they require viral antigen [11,12].
In the presence of a persistent infection, exhausted CD8 + T cells were found as two distinct subsets, one subset expressing high levels of the transcription factor, T-bet, and the other subset expressing high levels of the related transcription factor, Eomesodermin (Eomes). Further, Paley, et.al. showed that, in response to antigenic stimulation, T-bet hi cells give rise to Eomes hi cells. T-bet hi cells were found to be less exhausted and exhibited higher proliferative rates in response to antigen, while the Eomes hi cells were more exhausted, but exhibited greater cytotoxic activity than the T-bet hi cells. Both subsets were essential for viral control, as loss of either transcription factor resulted in viral persistence [13]. These data indicated that optimal expression of T-bet and Eomes, and the presence of both CD8 + T cell subsets, were essential for efficient viral control. However, the pathway that regulates the differentiation of antigen-specific T cells into these subsets, and whether a specific ratio of the two subsets is critical for viral control, are not known.
IRF4 is a pleiotropic transcription factor that regulates a myriad of functions in a wide variety of cell populations [14]. Recently, we and others have shown that levels of IRF4 in CD8 + T cells are regulated by the strength of TCR signaling. Thus, robust expression of IRF4 is important for maximal CD8 + T cell expansion in response to acute viral infections [15][16][17][18][19]. When IRF4 levels are reduced by a haplo-deficiency of the Irf4 gene, the magnitude of the CD8 + T cell response is dramatically impaired. The decreased numbers of virus-specific T cells are accounted for by a reduction in terminal effector cells (SLEC; KLRG1 hi CD127 lo ) without a significant effect on the numbers of memory precursor effector cells (MPEC, KLRG1 lo CD127 hi ) [15]. These studies also highlighted a role for IRF4 in the expression of key transcription factors T-bet and Eomes, important for differentiation and maintenance of SLEC and MPEC populations, respectively, during acute infections [15][16][17][18][19][20][21].
Here we show that TCR signal strength maintains an optimal balance of T-bet to Eomes, and that this process is regulated by the levels of IRF4 expressed. Reduced expression of IRF4 skews this ratio in the favor of Eomes during infection with LCMV-clone 13, resulting in reduced differentiation of T-bet + Eomesprecursors and impaired viral control. Reducing Eomes expression in Irf4 heterozygous mice re-establishes the protective balance of T-bet to Eomes, restores differentiation of T-bet + Eomesprecursors, and ultimately rescues defective viral clearance. These data indicate a critical role for IRF4 in regulating T cell exhaustion by balancing the relative expression of T-bet and Eomes during chronic infection. Overall, these findings demonstrate that reduced differentiation of the T-bet + Eomes -CD8 + T cell population impairs viral clearance, whereas a partial reduction in Eomes expression can restore viral control during persistent LCMV-clone 13 infections.

Results
TCR signal strength via IRF4 regulates the ratio of T-bet and Eomesodermin in activated CD8 + T cells To understand the role of TCR signal strength in the differentiation of CD8 + T cell subsets, we utilized P14 TCR transgenic TCRα -/cells (hereafter referred to as P14 WT), that respond to the lymphocytic choriomeningitis virus (LCMV) GP 33-41 epitope (GP33) bound to H2-D b ; additionally, P14 WT cells respond to a lower affinity 'F6L' peptide variant of GP [33][34][35][36][37][38][39][40][41] , containing a single amino acid substitution from phenylalanine to leucine at the sixth position of the peptide. This alteration leads to a 5-fold decrease in the affinity of the P14 TCR for the F6L-H2-D b complex relative to GP 33-41 -H2-D b , and as a consequence, a 100-1000-fold reduction in functional avidity [22]. P14 WT T cells were stimulated in vitro and then examined for T-bet (S1A Fig) and Eomes (previously published in [15]) expression by intracellular staining. At 24hr post-stimulation, P14 WT T cells stimulated with the high affinity GP33 epitope, expressed a higher T-bet to Eomes ratio relative to cells stimulated with the lower affinity F6L epitope ( Fig 1A). A similar trend was also observed when P14 cells were stimulated with decreasing doses of the GP33 peptide. These data indicate that variations in TCR signal strength, such as those achieved by varying the affinity of the peptide-MHC for the TCR or by varying the dose of stimulating peptide, affect the relative expression levels of these key transcription factors.
The transcription factor IRF4 is upregulated in CD8 + T cells in a graded manner in response to variations in TCR signal strength. Furthermore, IRF4 is known to be a negative regulator of Eomes and a positive regulator of T-bet expression in CD8 + T cells [15][16][17][18]23,24]. To determine if alterations in IRF4 expression affect the relative up-regulation of T-bet and Eomes, we compared P14 WT T cells to P14 TCRα -/cells lacking one or two functional alleles of Irf4 (Irf4 +/fl X CD4-Cre and Irf4 fl/fl X CD4-Cre; referred to as P14 Irf4 +/fl and P14 Irf4 fl/fl , respectively). In response to stimulation with the GP33 peptide, P14 WT cells expressed the highest T-bet to Eomes ratio, while P14 Irf4 +/fl cells showed a reduced T-bet to Eomes ratio, and P14 Irf4 fl/fl cells exhibited the lowest T-bet to Eomes ratio ( Fig 1A). This pattern was observed at all time points examined (Fig 1B and 1C). Furthermore, no alterations in the ratio of T-bet to Eomes expression was observed when P14 Irf4 fl/fl cells were stimulated with varying doses of GP33 peptide (Fig 1D-1F). These data indicate that the strength of TCR signaling via IRF4 regulates the relative expression of T-bet and Eomes in CD8 + T cells.

Levels of IRF4 regulate CD8 + T cell differentiation into T-bet + Eomesand T-bet -Eomes + subsets in response to LCMV Cl13 infection
The data described thus far indicated a critical role of IRF4 in maintaining the balance of T-bet and Eomes in-vitro. To determine whether the levels of IRF4 regulate the relative expression of T-bet and Eomes in-vivo, WT, Irf4 +/fl X CD4-Cre and Irf4 fl/fl X CD4-Cre mice (henceforth referred to as Irf4 +/fl and Irf4 fl/fl , respectively) were infected with an exhausting dose of LCMVclone 13. Due to the defects in regulatory T cell populations in Irf4 fl/fl mice, CD4 + and CD8 + T cell populations are activated in Irf4 fl/fl mice even in the absence of infection [23,25]. As a result, we focused on understanding how modest reductions, rather than a complete loss of IRF4 expression, impacted the protective response to persistent virus infection.
In response to LCMV-clone 13 infection, anti-viral CD8 + T cells differentiate into T-bet + Eomesprecursors that give rise to T-bet -Eomes + progeny in response to antigenic stimulation [13]. Examination of these subsets at D8 p.i. indicated an IRF4 dose-dependent defect in the differentiation of these populations. Irf4 +/fl mice exhibited a significant decrease in the numbers and proportions of the T-bet + Eomesprecursors with a concomitant increase in the differentiation to T-bet + Eomes + cells (Fig 2D and S2C Fig). These double positive cells could potentially be an intermediate population that differentiates into T-bet -Eomes + cells later during infection. As the viral titers are not different between WT and Irf4 +/fl at D12 p.i. (Fig 2B), these data indicate an inherent defect in the differentiation of CD8 + T cells expressing lower levels of IRF4. The differentiation of the Irf4 fl/fl cells into T-bet + Eomesand T-bet + Eomes + populations was severely compromised, likely accounting for the increased viral titers in the Irf4 fl/fl mice. These data indicate that IRF4 regulates the relative expression of T-bet and Eomes and therefore may impact the differentiation of T-bet + Eomesprecursors into T-bet -Eomes + progeny early during persistent LCMV infection. Further, high levels of IRF4 are required for robust expansion of virus-specific T-bet + Eomeseffector CD8 + T cells and viral control at early times post-infection.
WT levels of IRF4 are required to maintain the balance of T-bet to Eomesodermin expression during persistent infection In activated T cells, the expression of IRF4 is regulated by the strength of TCR signaling via the mTOR pathway; furthermore, mTOR signaling is suppressed during chronic infection by signaling through the inhibitory receptor PD-1 [15][16][17]23,26]. To test the importance of IRF4 expression levels at later timepoints during the persistent infection, we examined CD8 + T cell responses to LCMV-clone 13 at D21-24 p.i. At this time point, we observed reduced proportions of virus-specific T cells in the spleens of Irf4 +/fl mice compared to WT mice; however, absolute numbers were not significantly different (Fig 3A and 3D). As expected, Irf4 fl/fl mice had significant reductions in both the proportions and numbers of both subsets of virus-specific CD8 + T cells.
Examination of T-bet and Eomes expression in virus-specific CD8 + T cells at D21-24 p.i. with LCMV-clone 13 indicated that for both of the LCMV epitopes examined, virus-specific Irf4 +/fl CD8 + T cells exhibited an altered T-bet to Eomes ratio relative to WT cells ( Fig 3C and 3E). This skewed ratio resulted in reduced differentiation of T-bet + Eomesand increased differentiation of T-bet -Eomes + populations in Irf4 +/fl mice relative to WT mice ( Fig 3C and 3E and S3A and S3C Fig). High viral titers have been implicated in the excessive proliferation and eventual conversion of the T-bet hi CD8 + T cell population into Eomes hi CD8 + T cell population in patients with chronic HCV infection [13,27]. However, at day 26 p.i., we observed no difference in serum virus titers between Irf4 +/fl and WT mice, indicating that the altered CD8 + T cell populations were not simply a reflection of differences in viral load ( Fig 3B). These data indicate that a modest reduction in IRF4 expression leads to impaired generation of T-bet + Eomesvirus-specific CD8 + T cells at later timepoints of the persistent infection.

Intrinsic role of IRF4 in regulating the balance of T-bet to Eomesodermin expression in CD8 + T cells responding to LCMV-clone 13 infection
To address whether differences in IRF4 expression were regulating the balance of T-bet to Eomesodermin by acting in virus-specific CD8 + T cells, we performed adoptive transfer experiments. P14 WT or P14 Irf4 +/fl CD8 + T cells were transferred into congenic hosts, and recipients were infected with LCMV-clone 13. Beginning at D10 p.i., we observed that a proportion of recipients receiving WT P14 cells were succumbing to the infection, a response not seen in recipients that received P14 Irf4 +/fl cells ( Fig 4A). Analysis of viral titers in the sera of the surviving mice at D15 p.i. indicated that 75% of recipients receiving WT P14 cells had viral titers <7.5x10 4 , whereas for recipients receiving P14 Irf4 +/fl cells, this value was 2.9x10 5 ; however, the means of the two datasets were not significantly different. Based on previous studies showing that fatality from LCMV-cl13 infection results from excessive T cell responses leading to immunopathology [6], these data suggest a more robust effector response of WT compared to Irf4 +/fl P14 cells to the LCMV-clone 13 infection, consistent with previous observations [19].
To examine the P14 cells directly, surviving mice were analyzed at D22 p.i.. At this timepoint, we recovered fewer P14 Irf4 +/fl cells from recipient mice compared to WT P14 cells ( Fig  4C). Examination of these cells for T-bet and Eomes expression showed that WT P14 cells had increased levels of T-bet and reduced levels of Eomes relative to P14 Irf4 +/fl cells, resulting in higher T-bet to Eomes ratio in the WT P14 cells (Fig 4D). Overall, these data indicate that the magnitude of IRF4 expression in virus-specific CD8 + T cells regulates their expansion, as well as the relative expression of T-bet and Eomes in response to LCMV-clone 13.
WT levels of IRF4 are essential for optimal control of persistent virus infection While modest reductions in the levels of IRF4 were inconsequential for early control of virus replication through d26 p.i., long-term studies revealed that Irf4 +/fl mice had a defect in controlling the persistent virus infection. Relative to WT mice, a lower proportion of Irf4 +/fl mice controlled LCMV-clone 13 infection in the kidney, liver and serum when examined more than 100 days p.i. (Fig 5A-5C). In addition, enumeration of viral titers in sera over time indicated that the kinetics of viral control were slower for Irf4 +/fl mice than WT mice ( Fig 5D). As expected, Irf4 fl/fl mice exhibited the highest viral titers and a complete impairment in viral control (Fig 5A-5D).
IRF4 is also important for the differentiation of CD4 + Tfh cells that regulate plasma cell differentiation. As previous studies have shown that during LCMV infection reduced antibody responses lead to virus persistence [28,29], we considered whether the slower kinetics of virus control in Irf4 +/fl mice could be accounted for by reduced anti-LCMV antibody titers. Analysis of LCMV-specific IgG antibody titers at D40 p.i. indicated that Irf4 fl/fl mice were impaired in antibody production, consistent with the requirement for IRF4 in Tfh differentiation ( Fig 5E). However, we observed no difference in antibody titers between WT and Irf4 +/fl mice, supporting a CD8 + T cell-intrinsic defect in viral control in Irf4 +/fl mice. We then analyzed virus-specific antibody titers at D112-114 p.i., a timepoint at which the majority of WT mice were controlling the infection. When antibody data were segregated so that WT and Irf4 +/fl mice were compared based on whether they had detectable virus in their sera (Fig 5C), we found slightly reduced antibody titers in Irf4 +/fl mice that were controlling the virus relative to WT mice in this group (Fig 5F). However, no difference in antibody titers between the two genotypes of mice were seen for those that still had persistent virus replication in their sera. These data indicate that high levels of IRF4 expression are essential for optimal long-term control of the persistent LCMV-clone 13 infection, and that differences in viral control are not correlated with differences in anti-viral antibody titers.
Examination of virus-specific CD8 + T cells at D112-114 p.i. showed no significant differences in their numbers of virus-specific CD8 + T cells in Irf4 +/fl mice compared to WT controls (Fig 6A and 6D). Nonetheless, consistent with our observation from D8 and D21-24 p.i., virusspecific Irf4 +/fl CD8 + T cells at D112-114 p.i. exhibited reduced T-bet to Eomes ratio relative to WT cells (S4A and S4C Fig). This skewed ratio resulted in higher proportions and numbers of T-bet + Eomesvirus-specific CD8 + T cells in WT mice relative to Irf4 +/fl mice (Fig 6B and 6E). When the data from the WT mice were segregated according to viral titers, we found that WT mice that had not controlled LCMV-clone 13 infection at this time point, appeared as outliers and had proportions and numbers of T-bet + Eomesvirus-specific CD8 + T cells comparable to Irf4 +/fl mice (Fig 6B and 6E). These findings are consistent with the conclusion that persistent antigen is essential for conversion of T-bet + Eomescells into T-bet -Eomes + progeny [13].
Control of persistent LCMV-clone 13 infection has been associated with high numbers of T-bet + virus-specific CD8 + T cells that lack expression of the exhaustion marker, PD-1 [13]. When analyzed at D112-114 p.i., we observed increased proportions and numbers of T-bet hi PD-1 lo cells in WT mice relative to Irf4 +/fl and Irf4 fl/fl mice infected with LCMV-clone 13 ( Fig  6C and 6F). However, as with the analysis of T-bet + Eomescells, the T-bet hi PD-1 lo population in WT mice that had not cleared the virus by this timepoint appeared more similar to that of the Irf4 +/fl cells. These data suggest that the proportions and numbers of T-bet hi PD-1 lo CD8 + T cells may be more closely associated with viral control than with IRF4 expression levels. Similar findings are evident from analysis of Eomes + PD-1 hi CD8 + T cell proportions and numbers at this timepoint (S4B and S4D Fig).  experiments; significant differences determined by Ordinary one-way ANOVA using Tukey's multiple comparison test. $ denotes statistically significant difference between WT and Irf4 +/fl samples when analyzing only mice with undetectable serum viral titers (bold outlined symbols). Significant differences between outlined samples were determined by unpaired t test with Welch's correction. doi:10.1371/journal.pone.0144826.g006

IRF4 Maintains the Balance of T-Bet to Eomes
Reducing Eomes expression improves viral control in LCMV-clone 13-infected Irf4 +/fl mice The data presented above indicated that LCMV-clone 13-infected Irf4 +/fl mice have a lower Tbet to Eomes ratio than WT mice in their virus-specific CD8 + T cells at all time points investigated post-infection. Since optimal expression of both of these factors is essential for viral control, we considered whether manipulation of this ratio in favor of T-bet would restore protective CD8 + T cell responses in Irf4 +/fl mice, leading to improved viral control. To address this, Irf4 +/fl mice were crossed to Eomes +/fl mice to generate Irf4 +/fl Eomes +/fl CD4-Cre mice (henceforth referred to as Irf4 +/fl Eomes +/fl mice). These compound heterozygotes were infected with LCMV-clone 13, along with WT, Irf4 +/fl , and Eomes +/fl mice and analyzed at D77-82 p.i.
We first observed higher proportions of GP276-specific T cells in the spleens of WT mice relative to the other three genotypes. Analysis of absolute cell numbers indicated that WT mice had more GP276-specific CD8 + T cells than did Eomes +/fl or Irf4 +/fl Eomes +/fl mice; in addition, Irf4 +/fl Eomes +/fl mice had significantly fewer cells than did Irf4 +/fl mice (Fig 7A). In contrast, the proportions of GP33-specific CD8 + T cells were similar between all four genotypes with only modest differences in the absolute numbers of these virus-specific CD8 + T cell populations ( Fig 7C).
Analysis of the levels of T-bet and Eomes in virus-specific CD8 + T cells between D77-82 p.i. indicated that the Irf4 +/fl and the compound heterozygotes had lower T-bet expression relative to WT, as would be expected due to the role of IRF4 in positively regulating T-bet expression (Fig 7B and 7D) [15][16][17]30]. The expression of Eomes was higher in the Irf4 +/fl cells relative to Irf4 +/fl Eomes +/fl cells, indicating that the heterozygous deficiency in Eomes was able to reduce Eomes protein expression in Irf4 +/fl cells (Fig 7B and 7D). In addition, Irf4 +/fl Eomes +/fl cells showed normalized ratios of T-bet to Eomes expression levels that were comparable to those seen in WT cells. Furthermore, the re-establishment of the WT T-bet to Eomes ratio was also accompanied by increased proportions of T-bet + Eomes -GP33-specific cells in Irf4 +/fl Eomes +/ fl mice compared to those found in Irf4 +/fl mice (Fig 7D).
To test the functional consequences of the restored T-bet to Eomes ratio and increased differentiation of the T-bet + Eomesvirus-specific CD8 + T cell population in Irf4 +/fl Eomes +/fl mice, we examined the viral titers in these mice between days 77-82 p.i. This analysis showed a significantly higher titer of virus in the kidneys of Irf4 +/fl mice compared to the other three genotypes analyzed (Fig 8A). Additionally, in contrast to the Irf4 +/fl mice, virus was efficiently cleared from livers and sera of WT, Eomes +/fl , and Irf4 +/fl Eomes +/fl mice at this timepoint ( Fig  8B and 8C). Finally, we observed no difference in the kinetics of virus clearance when comparing WT to Irf4 +/fl Eomes +/fl mice (Fig 8D). These data demonstrate that reducing Eomes expression in Irf4 +/fl mice restores a protective balance of T-bet to Eomes levels, leading to improved protective T cell responses to this persistent virus infection To determine whether these mice showed differences in T cell exhaustion, we examined expression of 2B4, CD160, LAG-3, and PD-1 on virus-specific cells. At this timepoint, Irf4 fl/fl cells showed a marked reduction in 2B4 expression relative to all other genotypes, and Irf4 +/fl cells had higher PD-1 expression relative to WT cells (S5B Fig and S6B Fig). While WT mice exhibited higher proportions of T-bet hi PD-1 lo and lower proportions of Eomes hi PD-1 hi populations relative to Irf4 +/fl mice, we did not observe any differences between Irf4 +/fl and Irf4 +/fl Eomes +/fl cells (S5C and S5D Fig and S6C and S6D Fig). Overall, these data indicate that differences in viral control are unlikely to be due to differential expression of exhaustion markers.
We next examined cytokine production by virus-specific CD8 + T cells from WT, Irf4 +/fl , Irf4 fl/fl , Eomes +/fl , and Irf4 +/fl Eomes +/fl mice at D22 p.i. Ex-vivo stimulation of splenocytes from infected with GP276 or GP33 peptide showed few cells capable of producing IFNγ, and little evidence of multi-functional T cells in any of the mice (S5E and S5F Fig and S6E and S6F Fig). We also failed to observe a difference in Granzyme B expression between any of the genotypes, with the exception of Irf4 fl/fl cells (S5H Fig and S6H Fig), arguing against a difference in cytotoxic activity between WT, Irf4 +/fl , Eomes +/fl , and Irf4 +/fl Eomes +/fl mice at this timepoint.
To determine whether Irf4 +/fl Eomes +/fl produced higher levels of anti-viral antibodies compared to Irf4 +/fl mice, we examined anti-LCMV IgG titers at D40 and D78-82 p.i. with LCMVclone 13. No differences in anti-viral antibody titers were observed in this analysis (Fig 8E), consistent with the comparable proportions of IL-21-producing GP61-specific CD4 + T cells in these mice (S6I Fig).

IRF4 Maintains the Balance of T-Bet to Eomes
Discussion CD8 + T cell exhaustion is commonly observed during persistent infections and cancers [31]. Although a number of cell extrinsic factors such as the presence of the immune-suppressive cytokine IL-10, absence of help from CD4 + T cells, and high viral loads have been implicated in inducing exhaustion, a clear understanding of the CD8 + T cell intrinsic molecular mechanisms is still lacking [4,32]. The role of several transcription factors, such as Blimp-1, T-bet and Eomes, have been examined during chronic infection [13,33,34]. Blimp-1 is highly expressed in exhausted CD8 + T cells and its expression correlates with greater expression of exhaustion markers. Yet, Blimp-1 knockouts have a defect in controlling LCMV-clone 13 infections relative to WT mice, whereas mice haplo-deficient for Blimp-1 are more proficient at controlling the infection. These data indicate that varying levels of Blimp-1 regulate distinct transcriptional modules in exhausted CD8 + T cells, and that levels of Blimp-1 that are too high or too low are detrimental to the immune response [33]. Similarly, the expression of T-bet and Eomes segregate into T-bet hi and Eomes hi populations during persistent infections. Even though both proteins are T-box transcription factors and function redundantly early in LCMV Armstrong infection, loss of either protein results in impaired control of LCMV-clone 13 infections likely due to the different gene networks they regulate during LCMV-clone 13 infections [13,32]. Together these data demonstrate that efficient control of LCMV-clone 13 requires an optimal level of each of these transcription factors. Interestingly, mice lacking Blimp-1, T-bet, or Eomes have no defect in clearing acute LCMV Armstrong infections, yet are impaired in controlling LCMV-clone 13 [21,35,36]. Here we find that IRF4 joins this group of transcription factors, as IRF4 Maintains the Balance of T-Bet to Eomes reduced expression of IRF4 leads to persistence of LCMV-clone 13 infection, but does not affect clearance of LCMV Armstrong [15].
Despite the fact that the function of Tbet hi and Eomes hi subsets is well defined, the factors regulating the differentiation of anti-viral CD8 + T cells into these subsets is not known. We explored the role of TCR signaling via IRF4 in the relative expression of T-bet and Eomes and the differentiation of T-bet + Eomesand T-bet -Eomes + subsets during chronic LCMV infection. We find that the balance of these two transcription factors is dependent on the strength of TCR signaling and reduction in the affinity or the dose of the stimulating ligand resulted in lowering of this ratio. Molecularly, the relative levels of T-bet and Eomes were dependent on the magnitude of IRF4 expression, and reduced IRF4 expression, such as those achieved by haplo-deficiency of Irf4, resulted in skewing of this ratio in the favor of Eomes, both in-vitro and in-vivo. Previous studies have demonstrated that IRF4, together with its binding partner, BATF, directly bind to the both the T-bet and Eomes loci in effector CD8 + T cells, indicating a direct role of IRF4/BATF in regulating the expression of these two genes [17,30].
The most severe consequence of the altered T-bet to Eomes ratio was the reduced differentiation of T-bet + Eomesprecursors, and long-term viral persistence in Irf4 +/fl mice. IRF4 regulates multiple pathways such as proliferation, metabolism, and expression of effector cytokines in anti-viral CD8 + T cells [16,17]. Thus, it is likely that defects in any of these pathways could contribute to impaired viral clearance in Irf4 +/fl mice rather than the lower T-bet to Eomes ratio. To directly test the importance of Tbet to Eomes ratio, we reduced Eomes expression in Irf4 +/fl mice by generating Irf4 +/fl Eomes +/fl mice. These compound heterozygotes had lower Eomes expression relative to Irf4 +/fl mice, exhibited no alteration in the T-bet to Eomes ratio relative to WT cells, and showed enhanced differentiation of T-bet + Eomescells relative to Irf4 +/fl mice. Consequently, Irf4 +/fl Eomes +/fl mice were indistinguishable from WT mice in terms of viral clearance from multiple organs. These data, to the best of our knowledge, are the first to test the importance of relative expression of T-bet and Eomes in LCMV-clone 13 viral control.
Another consequence of reduced T-bet to Eomes ratio was the altered differentiation of Tbet hi and Eomes hi populations in Irf4 +/fl mice. Using T-bet and Eomes knockout mice, Paley et. al. showed that both T-bet hi and Eomes hi subsets are important for viral control [37]. However, it is not known if reduced differentiation of one population over the other affects viral clearance. Our studies suggest that reduced differentiation of T-bet + Eomespopulation in Irf4 +/fl mice is detrimental to efficient viral control, and that increasing this population in Irf4 +/fl Eomes +/fl mice restores viral control. Consistent with this observation, lower proportions of CD8 + T-bet + Eomescells are observed in lung transplant recipients with relapsed CMV infection relative to controllers [27]. A recent study showed that loss of FoxO1 results in higher levels of T-bet and lower levels of Eomes, thus biasing CD8 + T cell differentiation to the T-bet hi subset. Interestingly, the FoxO1 deficient mice were also defective in viral control [26]. Together these data underscore the importance of previous observations that both subsets of CD8 + T cells, the T-bet hi and the Eomes hi populations, play unique and essential functions during anti-viral immune responses to LCMV-clone 13 [13]. Further, these data support our conclusion that balanced differentiation of T-bet + Eomesand T-bet -Eomes + subsets are important for viral control.
Recently, we and others have shown that the levels of IRF4 regulate the magnitude of CD8 + T cell expansion during acute infections [15][16][17][18][19]. Similar to that seen with acute infections, we find here that the levels of IRF4 also regulate the magnitude of the CD8 + T cell response to LCMV-clone 13 at D8 post infection. Furthermore, the increased numbers of virus-specific cells in WT mice relative to Irf4 +/fl mice also resulted in a significant increase in the numbers of Tbet + Eomescells at this timepoint. Somewhat surprisingly, total virus-specific CD8 + T cell numbers were not different between WT and Irf4 +/fl mice at D22 p.i., nor were they different at later timepoints examined. Therefore, we speculate that this early difference in the magnitude of the CD8 + T cell response may be responsible for the ultimate ability of WT, but not Irf4 +/fl mice to clear LCMV clone 13.

Mice
Mice were bred and housed in specific pathogen-free conditions at the University of Massachusetts Medical School (UMMS) in accordance with institutional animal care and use committee guidelines. Irf4 fl/fl and Eomes fl/fl mice have been described previously [38,39]. CD4-Cre + transgenic mice were a gift from Joonsoo Kang (UMMS). Irf4 fl/fl mice were crossed to CD4-Cre + transgenic mice at UMMS to generate Irf4 fl/fl CD4-Cre + mice. P14 TCR transgenic TCRα -/used for in-vitro studies were purchased from Taconic Farms (Germantown, New York) and crossed to Irf4 fl/fl CD4-Cre + mice at UMMS. P14 TCR transgenic mice used for in-vivo studies were a gift from Susan Kaech (Yale), and were crossed to Irf4 fl/fl CD4-cre + mice to generate P14 Irf4 +/fl CD4-cre + mice. Eomes fl/fl mice were purchased from The Jackson Labs (Maine) and crossed to Irf4 +/fl CD4-Cre + at UMMS. Irf4 +/+ and Irf4 +/+ CD4-Cre + mice were used as WT controls.

Virus and Infections
LCMV-clone 13 stocks were propagated in baby hamster kidney 21 cells at UMMS [40] and were generously provided by Dr. Raymond M. Welsh and amplified by us. Adult male mice (6-9 weeks old) were infected with an exhausting dose, 2X10 6  monomers were prepared at UMMS; LCMV-specific H2D b -GP276 and H2D b -GP33 monomer were obtained from the NIH Tetramer Core Facility (Atlanta, GA). Single cell suspensions from spleens were prepared, RBC lysed and Fc receptors were blocked using supernatant from 2.4G2 hybridomas. Cells were stained with H2D b -GP276 or H2D b -GP33 tetramers prior to staining with cell-surface antibodies. For cytokine staining, cells were stimulated ex-vivo with 1μg/ml GP 276-286 or GP 33-41 peptide for 5 hours at 37°C. Intracellular cytokine staining was performed using BD Cytofix/Cytoperm™ Fixation/Permeabilization Solution kit. Intracellular transcription factor staining was performed using eBioscience Foxp3/transcription factor staining kit, as per manufacturer's instructions, unless specified. All samples were analyzed on an LSRII flow cytometer (BD Biosciences), and data were analyzed using FlowJo (Tree Star).

Cell culture
Lymph node cells from P14 WT, P14 Irf4 +/fl or P14 Irf4 fl/fl mice were mixed with equal numbers of WT CD45.1 splenocytes and stimulated with GP 33-41 (GP33) or F6L peptides for 24, 48, and 72 h. Cells were harvested and analyzed for T-bet and Eomes expression by intracellular staining.

Viral Titers
Sera, livers and kidneys were harvested from infected mice at the indicated time points postinfection (p.i.). Organs were homogenized in one ml of complete RPMI media and stored at -80°C. LCMV-clone 13 virus titers were determined by plaque assays as previously described [40].

LCMV-Specific Antibody Titers
To quantify LCMV-specific antibody titers, high-binding 96 well flat bottom ELISA plates (Corning) were coated overnight at room temperature with cell lysate from LCMV-clone 13 infected BHK21 cells. Plates were washed, blocked and three-fold serial dilutions of serum samples were plated. Plates were washed again and incubated with horseradish peroxidase labeled goat anti-mouse IgG detection antibody (Bethyl labs) and developed using 3,3',5,5'-Tetramethylbenzidine substrate. The reaction was quenched 0.18M H 2 SO 4 ELISA stop solution (Bethyl Labs) and the optical density (OD) was measured at 450nm using an Emax Endpoint ELISA microplate reader (Molecular Devices). LCMV-specific antibody titers were determined by end-point titer method and two times the mean OD of uninfected control sera was used as the cut-off.

Statistical Analysis
All data are represented as mean ± Standard Error of Mean (SEM). Statistical significance is indicated by ns p>0.05, Ã p 0.05, ÃÃ p 0.01, ÃÃÃ p 0.001, and ÃÃÃÃ p 0.0001. For outlined D112-114 samples, statistical significance is indicated by $p 0.05 and $ $p 0.01. Statistical analysis was performed using unpaired t test with Welch's correction, ordinary one-way ANOVA using Tukey's multiple comparison test or Log-rank (Mantel-Cox) test as indicated in the figure legend. experiments; significant differences determined by Ordinary one-way ANOVA using Tukey's multiple comparison test. Symbols with bold outlines represent mice whose serum viral titers were below the limit of detection at D112-114 p.i.. $ denotes statistically significant difference between WT and Irf4 +/fl samples when analyzing only mice with undetectable serum viral titers (bold outlined symbols). Significant differences between bold outlined samples were determined by unpaired t test with Welch's correction.