Interferon regulatory factor 8 integrates T-cell receptor and -signaling pathways and drives effector differentiation of CD8 T cells

Fumi Miyagawaa, Hong Zhanga, Atshushi Terunumaa, Keiko Ozatob, Yutaka Tagayac,1,2,3, and Stephen I. Katza,1,3

aDermatology Branch, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD 20892; bLaboratory of Molecular Growth Regulation, National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD 20892; and cMetabolism Branch, National Cancer Institute, National Institutes of Health, Bethesda, MD 20892

Edited* by Thomas A. Waldmann, National Cancer Institute, National Institutes of Health, Bethesda, MD, and approved June 19, 2012 (received for review January 26, 2012)

We recently demonstrated that differentiation of cytotoxic T cells differentiation (27). Nevertheless, the role of IRF8 in CD8 T requires cooperation between T-cell receptor (TCR)/costimulation cells remains elusive; IRF8KO mice manifest impaired CD8 and γc-. Here we demonstrate that the transcription fac- T-cell responses against certain viruses (22). Thus, we embarked tor IFN regulatory factor 8 (IRF8) is expressed in CD8 T cells by the on a detailed assessment of the role of IRF8 in CD8 T cells. Here, we demonstrate the critical need for the convergence of combination of these two signals. More importantly, depletion of γ IRF8 in these cells abrogated the differentiation of naive CD8 T cells c-Jak3 and TCR/costimulation-signaling pathways in the tran- scription of IRF8 and demonstrate that removal of IRF8 cripples into effector cells in an experimental graft-vs.-host disease mouse fi − model. We also show that IRF8 seems to not operate upstream of CD8 effector T cells. Associated with these ndings, IRF8 OT-I cells failed to induce GvHD in our model (28). We also show other critical factors such as T-bet and eomesodermin, which have that either IRF8 action is independent of that of T-bet and been implicated in effector maturation. Collectively, our work Eomes (9–11) or IRF8 may operate downstream of these factors, shows that IRF8 integrates the TCR/costimulation and γc-cyto- – but not by way of them, and therefore propose that IRF8 be kine signaling pathways and mediates the transition of naive added to the roster of critical regulators for the differentiation of CD8 T cells to effector cells, thus identifying IRF8 as one of the naive CD8 T cells into effector cells. molecular regulators of CD8 T-cell differentiation. Results γ common c-cytokine | T-cell receptor signaling | CD8 effector Suboptimal Antigen Stimulation and a γc-Cytokine Cooperatively differentiation | transcription factors | Jak3 Generate Functional CD8 T Cells. We previously developed an ex- perimental model of GvHD (28) and reported that cooperation D8 T cells are essential for the adaptive immune response of γc-cytokine(s) and Ag/TCR-signaling events is indispensable Cagainst various intracellular pathogens and tumors. A typical for disease development. Briefly, a membrane-bound form of CD8 T-cell response consists of three phases: clonal expansion of chicken OVA was expressed in mice under the K14 promoter. antigen (Ag)-specific cells and acquisition of effector functions, Adoptive transfer of syngeneic OT-I cells (CD8+Vα2+Vβ5+) contraction of the effector cells through apoptosis, and genera- that recognize OVA in these mice causes GvHD-like pathogenic tion of long-lived memory cells (1–4). The acquisition of effector changes (28). In a strain with lower OVA copy number (K14- functions is critical for the control of intracellular pathogens and mOVAlow mice), GvHD did not occur spontaneously after tumors and is accompanied by the production of cytotoxic mol- adoptive transfer of OT-I cells but occurred when cells were ecules, perforin and granzyme, as well as two main cytokines, injected with IL-15 (and other selective γc-cytokines) (13) (Fig. IFN-γ and (TNF) (5–7). S1A). Furthermore, deletion of IL-15 by crossing K14-mOVAhigh The transition from naive to effector CD8 T cells requires mice (mice with high OVA copy number) with IL-15KO mice marked changes in gene expression (8) mediated by transcription abrogated disease (13). These observations led us to postulate factors (5). The T-box transcription factors, T-bet (9, 10) and that TCR and γc-cytokine signals cooperate physiologically to eomesodermin (Eomes) (11), are the best-described regulators enable full maturation of effector T cells. To test this hypothesis, of this process in CD8 T cells (11, 12). serially diluted OVA–peptide was cultured with OT-I cells in the We previously reported that transgenic mice expressing oval- high presence or absence of IL-15. Amounts greater than 10 pg/mL of bumin (OVA) protein by the keratin 14 promoter (K14-mOVA the peptide induced a maximal proliferative response by OT-I mice) develop graft-vs.-host disease (GvHD) after the transfer of cells, and 0.3 pg/mL was a suboptimal dose of peptide that in- OT-I CD8 cells, whereas similar transgenic mice expressing much A B low duced OT-I proliferation (Fig. 1 and Fig. S1 ). However, at lower OVA copy numbers (K14-mOVA mice) did not (13, 14). doses between 0.03 and 3 pg/mL of peptide, the cellular pro- We also demonstrated that the onset of GvHD in this model γ liferative response was greatly enhanced by the presence of IL- is dependent only on CD8 T cells. The injection of c-cytokines 15, and combinatorial treatment of 0.3 pg/mL of peptide and (15), especially IL-15, could create GvHD in K14-mOVAlow mice upon OT-I cell transfer, whereas K14-mOVAhigh mice on an IL-15KO background failed to develop GvHD upon OT-I cell transfer (13). These data suggested a critical requirement Author contributions: F.M., K.O., Y.T., and S.I.K. designed research; F.M., H.Z., and A.T. γ performed research; F.M., H.Z., and A.T. analyzed data; and F.M., K.O., Y.T., and S.I.K. for c-cytokines in the effector differentiation of CD8 T cells wrote the paper. in vivo. We then hypothesized that concurrent signaling events fl consisting of the γc-pathway and the T-cell receptor (TCR)/ The authors declare no con ict of interest. costimulation pathway are critical for the transition of naive *This Direct Submission article had a prearranged editor. CD8 T cells into effector cells. 1Y.T. and S.I.K. contributed equally to this work. In this study, we sought a molecular integrator of these two 2Present address: Division of Basic Science and Vaccine Research, Institutes of Human pathways in the effector differentiation of CD8 T cells and Virology, University of Maryland School of Medicine, Baltimore, MD 21201. identified IFN regulatory factor 8 (IRF8). IRF8 belongs to the 3To whom correspondence may be addressed. E-mail: [email protected] or yta- IFN regulatory factor family (16–18) and critically controls the [email protected]. lineage commitment between the myeloid and B cells (19–26). In This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. addition, IRF8 controls a silencing program for Th17 cell 1073/pnas.1201453109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1201453109 PNAS Early Edition | 1of6 Downloaded by guest on September 30, 2021 Fig. 1. Increased expression of IRF8 correlates with full effector function of OT-I cells and disease activity in K14-mOVA mice. (A) Synergistic effects of IL-15 and Ag peptide on the growth response of OT-I cells. Purified OT-I cells were cultured at 1.25 × 105 cells/mL (4-mL culture) in the presence of the same number of irradiated BMDC (peptide-pulsed or nonpulsed). Cells were counted on day 5 and divided by 5 × 105 to convert to a fold increase. Data are pooled from three independent experiments (error bars, SEM). (B) Synergistic effects of peptide and IL-15 in IFN-γ. Stimulation: peptide, 0.3 pg/mL of SIINFEKL; IL-15, 15 nM; peptide + IL-15, 0.3 pg/mL of SIINFEKL + 15 nM of IL-15. OT-I cells (2.5 × 104) were cultured with peptide-pulsed or nonpulsed BMDCs (2.5 × 104)inthe presence or absence of IL-15. Three days later, production of IFN-γ in the supernatants was determined by an ELISA (error bars, SEM). *P < 0.05 between groups. Data are representative of two independent experiments with duplicates in each experiment. (C) Venn diagram depicting the overlap and distinction of gene expression between OT-I cells stimulated with three different treatments (peptide, IL-15, and peptide + IL-15 as in B) for 6 or 12 h (adjusted P < 0.05; fold change > 1.5). Gene expression in each of the stimulated OT-I cells was measured by DNA microarray analysis to a single reference (untreated naive OT-I cells). (D) Induction of IRF8mRNA requires cooperation of cytokine and Ag-peptide stimulation: peptide, 0.3 pg/mL of SIINFEKL; IL-15, 15 nM; peptide + IL-15, 0.3 pg/mL of SIINFEKL + 15 nM of IL-15. OT-I cells were cultured as for A but harvested for RNA extraction 6 or 12 h later. Data represent the microarray assessment of the IRF8mRNA expression. An asterisk indicates adjusted P < 0.05 vs. untreated OT-I cells.

15 nM of IL-15 induced maximum proliferation (Fig. 1A and Fig. number of genes induced by the peptide alone dropped to 311 S1B). These combination treatments also induced IFN-γ pro- genes (Fig. 1C), suggesting the rather transient nature of Ag/ duction (Fig. 1B), indicating that OT-I cells gained effector costimulatory signaling under these conditions. Cytokine signaling function. These results suggest that partial-to-low T-cell seemed more persistent (1,064 genes remain induced), but the responses by suboptimal challenges of Ag can develop into a full combination of two pathways induced 921 distinct genes (Fig. 1C), response by the concomitant presence of certain γc-cytokines, an making the majority of the genes in this pool unique only to the observation that mirrors our previous in vivo results (13). Thus, combinatorial signal. Among the 921 genes, 144 genes were per- we conclude that γc-signaling, but not that of a particular γc- sistently detected in the combination group at both 6- and 12-h cytokine, is required to affect these changes. We also observed time points, whereas 748 genes were newly induced at 12 h. Be- that similar cooperation between a γc-cytokine and the TCR/ cause neither cytokine alone nor peptide alone led to the robust costimulation pathways operates in polyclonal CD8 T cells. proliferation of, and production of IFN-γ by, OT-I cells (Fig. 1 A and B), we reasoned that a critical transcription factor(s) that Transcription Factor IRF8 Potentially Integrates γc-Cytokine and TCR/ causes the differentiation of effector CD8 T cells via de novo Costimulation Signals. To determine the intracellular events by protein synthesis may be found in this gene pool of 144 genes which γc-cytokine signaling cooperates with TCR/costimulation, (Table S1). The results indicate the initiation of a robust tran- we conducted microarray analyses using cDNAs prepared from scription network in CD8 T cells upon . As a first step to OT-I cells stimulated by (1) Ag alone (peptide 0.3 pg/mL) (2), delineate such a network, we identified transcription factors in the cytokine alone (IL-15 15nM), or (3) a combination of both (Fig. pool of induced genes (Table S1). We thus identified IRF8 and 1A, arrows). Costimulation by CD28 was always present because c-jun and chose IRF8 for subsequent study (Fig. 1D and Table S1). we used bone marrow dendritic cells (BMDC) as antigen pre- Other key transcription factors known to control CD8 T-cell senting cells (APC). This dose of peptide, although suboptimal effector functions, such as T-bet and Eomes, were strongly and as assessed by the proliferative response of OT-I cells (Fig. 1A), persistently induced by IL-15 alone, suggesting that IRF8 and T- caused the induction of 2,825 transcripts at 6 h of stimulation bet/Eomes may belong to different categories (Fig. S1C). We do (Fig. 1C), and IL-15 stimulation induced 2,889 transcripts. When not know, however, if a strong Ag stimulation alone would induce these two signals were combined, induction of 2,088 genes was these two factors (Discussion). The microarray results were vali- observed, but the combination led to the induction of only 137 dated by real-time PCR (Fig. S1D, Table S2) and Western blot distinct transcripts. The picture dramatically changed at 12 h; the analysis (Fig. S1E).

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1201453109 Miyagawa et al. Downloaded by guest on September 30, 2021 To determine whether our observation could be extrapolated to polyclonal CD8 T cells, we treated CD8 T cells with sub- optimal doses of αCD3mAb and IL-15, and they expressed higher levels of IRF8mRNA than either treatment alone (Fig. S1F). OT-I cells stimulated with suboptimal doses of αCD3mAb+IL-15 (Fig. S1G) responded similarly in experiments depicted in Fig. S1 D and F. Thus, we hypothesized that the induction of IRF8 could be linked to the acquisition of effector functions by CD8 T cells.

Reasonable Correlation of IRF8 Expression with Disease Activity. We next asked whether IRF8 expression correlates with the activation status of OT-I cells and the clinical course of GvHD in K14-mOVA mice. To this end, 2 d after OT-I cells were injected into congenic (CD45.2) K14-mOVAlow,K14-mOVAhigh, and IL-15 treated K14- mOVAlow mice, we sorted the transferred OT-I cells by CD45.1 expression (Fig. S1A). Cells from K14-mOVAhigh mice expressed the highest levels of IRF8 followed by those sorted from K14- mOVAlow micetreatedwithIL-15andthosesortedfromK14- mOVAlow mice (Fig. S1H). Isolation of injected OT-I cells at earlier time points yielded too few cells for assessment although they may express higher IRF8 expression than the 48hr-OT-I cells. None- theless, this result suggests that the up-regulation of IRF8 correlates with effector function and disease activity in these K14-mOVA mice.

IRF8 Is an Activation Molecule in CD8 T Cells. IRF8 is the only IRF family member whose expression is confined to immune cells. However, its expression in CD8 T cells has not been extensively studied. Our results (Fig. S2 A and B) demonstrate the activa- tion-associated expression of IRF8 in OT-I cells. Similar results IMMUNOLOGY were obtained with polyclonal CD8 T cells (Fig. S2C), confirm- ing that IRF8 is expressed in CD8 T cells upon activation.

Effector Maturation of CD8 T Cells Is Dependent on the Integration of Jak-STAT and TCR/Costimulation-Signaling Pathways via IRF8. Our studies suggested that concurrent signaling through γc-cytokine and TCR/costimulation pathways may be critical in the differ- entiation of functional effector CD8 T cells via IRF8. This finding prompted us to determine whether blocking either of Fig. 2. Specific inhibition of γc-signaling compromises the induction of IRF8. these pathways individually would abrogate these outcomes. To (A) IRF8mRNA levels in OT-I cells cultured in the presence of CP-690,550. Two fi million cells/2 mL were stimulated with plate-bound αCD3mAb (10 μg/mL) and test this hypothesis, we rst used a Jak3 inhibitor, CP-690,550 α μ (hereafter referred to as CP) (29), to block endogenous Jak3- CD28mAb (5 g/mL) in the presence of various concentrations of CP or DMSO STAT signaling upon T-cell activation. Cells stimulated by high (vehicle) for 6 h. The IRF8mRNA level was determined by real-time PCR. Fold doses of Ag without exogenous cytokine with various concen- changes in reference to naive cells are shown. Data are pooled from two in- dependent experiments with triplicates in each experiment (error bars, SEM). trations of CP were tested for IRF8mRNA expression. CP IRF8 A (B) Proliferation of OT-I cells was not affected by CP. CFSE-labeled OT-I cells blocked the induction of mRNA in these cells (Fig. 2 ). were cultured as above with or without CD3/CD28 stimulation for 2 d. Cells Moreover, this blockade was seen with polyclonal CD8 T cells were gated on CD8Vα2 cells. Data are representative of two independent (Fig. S2D). Notably, CP did not inhibit the proliferation of OT-I – B experiments. (C E) Expressions of activation markers on cultured OT-I cells in cells (Fig. 2 ) nor did it inhibit the induction of the expression of the presence of CP. OT-I cells were cultured as in A for 2 d and stained for Vα2, CD44 (Fig. 2C and Fig. S2E), excluding nonspecific toxic effects Vβ5, CD44, CD25, and CD69. Cells were gated by Vα2 and Vβ5. Data shown are of this inhibitor. Interestingly, the induction of CD25 and CD69 representative of three independent experiments. The mean fluorescence was inhibited by CP (Fig. 2 D and E and Fig. S2E) in agreement intensity (MFI) of Fig. S2E was determined by flow cytometry. (F) IFN-γ pro- with the previous suggestion that Jak3-STAT5 regulates the ex- duction from OT-1 cells in the presence of CP. OT-I cells (1 × 105) were cultured pression of these genes (30, 31). Moreover, the CP treatment in 96-well plate as in A, and production of IFN-γ was determined by ELISA at reduced the amount of IFN-γ production by OT-I cells (Fig. 2F). the 48-h time point. Data are representative of two independent experiments These results suggest that the γc-signaling is intrinsic to the im- with duplicates in each experiment (error bars, SEM). munologic activation of CD8 T cells for the full generation of functional effector cells and that IRF8 could be critical in the integration of the TCR/costimulation and Jak-STAT–signaling had lymphadenopathy and splenomegaly and manifested in- creased percentages of Gr-1+CD11b+ granulocytes (Fig. S4). pathways. Likewise, we then used a Zap70 inhibitor, piceatannol. − Piceatannol inhibited IRF8 expression (Fig. 3A) and pro- We assessed the acquisition of effector functions of IRF8 OT-I liferation (Fig. 3B) and IFN-γ production (Fig. 3C) in OT-I cells cells in vivo using our K14-mOVAhigh mice (13, 28). Effector in a dose-dependent manner in agreement with our proposed maturation of OT-I cells was examined by the production of model for the requirements for IRF8 (Fig. S3). IFN-γ and by the manifestation of cytotoxicity against skin. Disease activity correlates well with the activation status of − Deletion of IRF8 from OT-I Cells Attenuates GvHD Responses in Vivo. injected OT-I cells (13). IRF8 OT-I cells failed to cause GvHD To determine if IRF8 is directly involved in the effector function in K14-mOVAhigh mice as judged by the lack of apparent weight of OT-I cells, we generated OT-I mice on an IRF8KO back- loss and pathological skin lesions (Fig. 4 A and B). Histology of − ground. Significant numbers of Vα2+Vβ5+OT-I cells were the ears showed that IRF8 OT-I cells caused no pathological present in all lymphoid organs of OT-I/IRF8KO mice, indicating changes (Fig. 4C). These data strongly indicate that IRF8 criti- that IRF8 is dispensable for the development of CD8 T cells cally controls processes that lead to effector differentiation of (Fig. S4). Like the IRF8KO mice (22), OT-I/IRF8KO mice also OT-I cells in vivo.

Miyagawa et al. PNAS Early Edition | 3of6 Downloaded by guest on September 30, 2021 Fig. 3. Inhibition of the TCR signaling abrogates the induction of IRF8. (A) IRF8mRNA levels in OT-I cells cultured in the pres- ence of piceatannol. OT-I cells were cultured as in Fig. 2A but with piceatannol or DMSO for 6 h. Data are pooled from two independent experiments with triplicates in each experiment (error bars, SEM). (B) Dose-dependent inhibition of pro- liferation of OT-I cells by piceatannol. CFSE-labeled OT-I cells were cultured as in Fig. 2B but with piceatannol. Data are representative of two independent experiments. (C) IFN-γ production from OT-1 cells is impaired in the presence of piceatannol. OT-I cells were cultured as in Fig. 2F but with piceatannol or DMSO. Data are representative of two in- dependent experiments with duplicates in each experiment (error bars, SEM).

Relationship Between IRF8 and T-bet/Eomes. T-bet and Eomes have of T-bet (Fig. S5 A and B, Table S2)andEomes(Fig. S5 A–D, Table beenshowntobecriticalinthefunctional activation and effector S2) did not differ much in the presence or absence of IRF8, sug- differentiation of CD8 T cells (9–11). Because the role of IRF8 de- gesting that either IRF8 operates completely independently of these fined in this study somewhat resembles those of T-bet/Eomes, we T-box factors or at least IRF8 is not situated upstream of T-box asked whether or not IRF8 operates dependently on these T-box factors. Furthermore, the Jak3 inhibitor, but not the Zap 70 inhibitor, factors. As described earlier, microarray analysis suggested that the blocked T-bet in a dose-dependent manner (Fig. S5 C and D). transcriptional requirement for the T-box factors (γc-signal alone) is different from that for IRF8 (both γc-cytokine and TCR/cos- IRF8−OT-I Cells Have Full Proliferative Capability but Produce Less timulation). Real-time PCR experiments revealed that the induction IFN-γ. Thus far our studies have defined a unique role of IRF8 in the in vivo activation of naive CD8 T cells. We next de- termined which compartment of T-cell activation might be controlled by IRF8 in vivo. First, we injected OT-I and − IRF8 OT-I cells into K14mOVAhigh mice and monitored their proliferation. Carboxyfluorescein diacetate succinimidyl ester (CFSE) experiments showed that OT-I cells proliferated simi- larly in the presence or absence of IRF8 in response to OVA (Fig. S6A), suggesting that IRF8 does not directly control the clonal expansion of the cells. Consistently, IRF8KO mice do not manifest any deficit in their T-cell compartments. These pieces of data strongly suggest that γc-cytokine and TCR/costimulation pathways are integral components enabling the functional mat- uration of CD8 T cells and that IRF8 functions as the molecular bridge integrating these two distinct pathways. − The total number of OT-I and IRF8 OT-I cells in secondary lymphoid organs did not differ between K14-mOVAhigh mice − injected with OT-I or with IRF8 OT-I cells (Fig. S6B). However, when assessing activation markers, injected OT-I cells exhibited an activated phenotype (CD25highCD44highCD62Llow) whereas − − in IRF8 OT-I cells there was less up-regulation of CD25 and Fig. 4. IRF8 OT-I cells induce less severe GvHD than do OT-I cells after C adoptive transfer into K14-mOVAhigh mice. (A) Kinetic change of body weight greater expression of CD62L (Fig. S6 ), suggesting that the lack − of OT-I–injected K14-mOVAhigh mice. One million naive OT-I or IRF8 OT-I cells of IRF8 selectively compromised the activation state of OT-I were injected into K14-mOVAhigh mice. The mice were weighed daily, and skin cells. Thus, IRF8 seems to be involved in the phenotypic matu- lesions were monitored for 14 consecutive days. Pooled results from two (of ration of CD8 T cells, but is dissociated from the clonal expan- six) independent experiments are shown (n = 9 each; error bars, SEM). **P < sion. This finding is reminiscent of the effect of the Jak3 inhibitor 0.001 and *P < 0.05. (B) Development of skin lesions in K14-mOVAhigh mice. CP that only impaired activation marker expression of OT-I cells Photos were taken on day 14 after injection of cells. (C) Histology of the ears of without affecting their proliferative response to the Ag (Fig. 2 B– K14-mOVAhigh mice on day 14 after injection of cells. Magnification: 20×. E). Next, we determined the relevance of IRF8 in the effector

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1201453109 Miyagawa et al. Downloaded by guest on September 30, 2021 function of CD8 T cells. IFN-γ serum levels and intracellular 3,000 genes in agreement with previous observations that the staining were significantly lower in the K14-mOVAhigh injected outcome of T-cell activation (i.e., proliferation, cytokine pro- − with IRF8 OT-I cells (Fig. 5A and Fig. S7A). In contrast, TNF-α, duction, cytokinesis, cytotoxicity, etc.) is hierarchically ordered another effector cytokine of CD8 T cells, was produced equally depending on the intensity of the TCR signaling (36, 37). − in both OT-I and IRF8 OT-I cells (Fig. S7B), suggesting that the Moreover, the CP experiment implied the inherent need for γc- transcriptional regulation of IFN-γ and TNF-α is distinct. These cytokine stimulation even at the point of T-cell activation in results strongly suggest that IRF8 is critically and selectively in- response to high doses of Ag. However, the combination of γc- volved in the acquisition of effector functions by CD8 T cells. cytokine and low-dose pMHC/TCR seems to qualitatively change the nature of the T-cell activation, as opposed to quan- IRF8 Controls Cytotoxicity. We next determined whether or not titatively amplifying the signaling, because the kinetics and rep- IRF8 deletion impairs the Ag-dependent cytotoxicity of OT-I cells. ertoire of the gene activation dramatically changed (more To address this question, we used an in vivo cytotoxic T lymphocyte lingering effect and newer genes differentially expressed) when − high γ (CTL) assay. IRF8 OT-I cells in K14-mOVA mice showed c-cytokine and pMHC/TCR stimulation were combined. modest, but significantly lower, cytotoxicity than OT-I cells under In vivo, T-cell activation occurs in a special milieu that is full γ the same conditions (Fig. 5B). The reduction in IFN-γ production of environmental factors including select c-cytokines (i.e., IL-7 and of the killing ability seen at day 5 is likely to be relevant to the and -15). It is very likely that some T cells encounter relatively high low levels of Ag and would still differentiate into functional ef- lack of development of GvHD in K14-mOVA mice because γ their kinetics paralleled that of the clinical symptoms of K14- fector cells with the c-cytokines provided by the microenviron- mOVAhigh mice injected with OT-I cells (Fig. 4A). It should be ment. It is possible that CD8 T cells stimulated by high doses of noted that Perforin and Granzyme mRNAs were also detected in Ag and those by concurrent cytokine/low-dose Ag may differ- entiate into different types of effector/memory T cells. the CD8 effector T cells, and like IRF8, both are inhibited by the γ γ Jak3 and Zap 70 inhibitors (Fig. S5 C and D, Table S2). Collec- Which c-cytokine may be involved in this function? Some c- tively, these results suggest that IRF8 is one of the critical factors in cytokines such as IL-2 or IL-4 are produced as a consequence of, and only after, T-cell activation (15). However, two γc-cytokines, the development of effector CD8 T cells from naive precursors. IL-7 and IL-15, are produced by nonlymphoid cells (15, 38). Furthermore, we and others have demonstrated that IL-15 can be Discussion α i γ surface-born via IL-15R by dendritic cells (DCs) activated Here we demonstrate ( ) that c-signaling concurrent or cooper- through TLR (the IL-15 transpresentation paradigm) (39, 40). ating with TCR/costimulation is indispensable for the differenti- ii Knowing that TLR signaling could facilitate the capacity of Ag to ation of effector CD8 T cells and ( ) that these two signaling present to T cells by the same DCs, it is plausible that DCs can IMMUNOLOGY cascades converge on IRF8, making this protein one of the critical concurrently present a γc-cytokine and Ag to neighboring CD8 T signal integrator molecules in this process. Many studies have cells. Our in vivo and in vitro data clearly support this proposal. fi shown that TCR stimulation, per se, is insuf cient for successful Here we have demonstrated that IRF8 is an important in- activation/maturation of T cells. Costimulatory molecules in- tegrator of TCR/costimulation and γc-signaling. Moreover, IRF8 cluding CD28, CD40L, and OX40 have been integrated into the seems to contribute to the phenotypic maturation and function- − equation to account for missing signals (32). However, our study ality of CD8 effector T cells because IRF8 OT-I cells demon- suggests that γc-cytokine signaling constitutes another critical el- strate a less diverse array of surface Ags upon activation and failed ement (i.e., the γc-Jak axis) in addition to the combination of TCR to induce GvHD when transferred into K14-mOVAhigh mice, al- and costimulatory signals. Due to the dramatic depletion of major − − − − − though IRF8 OT-I cells maintained a similar capacity of clonal lymphoid cells in γc / and Jak3 / mice (33–35), it has been dif- expansion as OT-I cells. ficult to precisely define the contribution of γc-cytokine signaling Naturally, differentiation of naive CD8 T cells into effector upon Ag-induced T-cell activation in the periphery. and/or memory cells is transcriptionally regulated. Recent stud- Our in vitro experiments showed that the γc-cytokine effect ies implicated the involvement of two related factors belonging becomes visible only at suboptimal doses of the Ag. However, to the T-box transcription factor family, T-bet and Eomes (5, 7), the microarray analysis showed that even this nonproliferative in the differentiation of CD8 T cells. In addition, several other TCR stimulation (peptide/MHC I and CD28) induced nearly factors have also been identified in this process, including Stat4,

Fig. 5. IRF8 mediates IFN-γ induction and killing activity in vivo. (A) Serum IFN-γ concentrations on day 5 after adoptive transfer. One million OT-I or IRF8−OT-I cells were transferred into WT or K14-mOVAhigh mice. Five days later, the serum IFN-γ level was measured by an ELISA (n = 5/group; error bars, SEM). Data − were pooled from two independent experiments. (B) In vivo killing activity of OT-I cells or IRF8 OT-I cells (5 d after they were adoptively transferred into K14- mOVAhigh mice) was assessed by flow cytometry. CFSE intensity of cells in the mice was determined 6 h after injection of a 1:1 mixture of SIINFEKL-pulsed splenocytes labeled with a high concentration of CFSE and unpulsed splenocytes labeled with a low concentration of CFSE. Numbers above the bracketed lines indicate the percentage of CFSEhigh cells. Data are representative of two independent experiments.

Miyagawa et al. PNAS Early Edition | 5of6 Downloaded by guest on September 30, 2021 Stat5, Notch2, Blimp-1, and Bcl6 (7). Furthermore, Runx3 and autoimmunity and chronic viral infection and may present an Notch1 control Eomes expression (7). Thus, the formation of opportunity for unique clinical intervention for these diseases. a network of transcription factors appears to support the dif- ferentiation of effector/memory T cells. Our studies suggest that Materials and Methods IRF8 is an important and nonredundant component of this Mice. K14-mOVA mice (13, 28) and IRF8KO mice (22) have been described network. We also determined whether IRF8 controls CD8 T-cell previously. OT-I/IRF8KO mice were generated by backcrossing OT-I mice with differentiation via these transcription factors. Real-time PCR IRF8KO mice. These mice were housed in a clean conventional facility and analyses demonstrated that the expression of the transcripts bred and used in accordance with protocols approved by the Animal Care encoding T-bet, Eomes, and Runx3 was not affected by the de- and Use Committee of the National Cancer Institute (41). letion of IRF8 from OT-I cells, suggesting that IRF8 does not induce these factors. Further analysis will determine if T-bet and Microarray Gene Profiling Analysis. OT-I cells were treated with OVA–peptide Eomes control IRF8 expression. alone, IL-15 alone, or OVA–peptide + IL-15 for 6 or 12 h, and total RNA was The concurrent response of γc-cytokine and TCR/co- extracted. Treated OT-I cells were then compared with naive OT-I cells that stimulation to low doses of Ag may present another intriguing were cultured for 6 or 12 h without stimuli using Mouse Genome 430 scenario for in vivo CD8 T-cell differentiation. Activation of 2.0 array (Affymetrix). The details are described in SI Materials and Methods. CD8 T cells in autoimmunity and chronic viral infection may be The protocols for the purification of cells and adoptive transfer, cell culture dependent on a lesser amount of circulating Ag than that in and cell growth assessment, ELISA, real-time PCR, antibodies, flow cytometry, acute immunologic activation. Such chronic disease conditions CFSE labeling, intracellular cytokine staining, in vivo CTL assay, and histo- may also likely cause continuous production of γc-cytokines, in logical analysis are described in detail in SI Materials and Methods. particular those of IL-7 and -15, by environmental cells. As- Statistical Analysis. Data were compared using a Student’s t test. Values of suming that the IRF8-mediated pathway would represent acti- < fi vation of CD8 T cells associated with a low dose of Ag in P 0.05 were considered a signi cant difference. combination with select γc-cytokines, IRF8 may be the key player during chronic activation of CD8 T cells whereas T-box ACKNOWLEDGMENTS. We thank J. Linton for technical assistance; P. Melzer, D. Edelman, A. Player, S. Davis, and Y. Wang for help with the microarray factors could primarily control acute CD8 T-cell activation. As experiments; and N. Voong for assistance with FACS sorting. This work was a corollary, depletion of IRF8 should have a more profound supported by intramural research funds of the Center for Cancer Research, effect in blocking pathological activation of CD8 T cells in National Cancer Institute/National Institutes of Health.

1. Dutton RW, Bradley LM, Swain SL (1998) T cell memory. Annu Rev Immunol 16: 23. Tamura T, Nagamura-Inoue T, Shmeltzer Z, Kuwata T, Ozato K (2000) ICSBP directs 201–223. bipotential myeloid progenitor cells to differentiate into mature macrophages. Im- 2. Murali-Krishna K, et al. (1998) Counting antigen-specific CD8 T cells: A reevaluation of munity 13:155–165. bystander activation during viral infection. Immunity 8:177–187. 24. Lu R, Medina KL, Lancki DW, Singh H (2003) IRF-4,8 orchestrate the pre-B-to-B tran- 3. Homann D, Teyton L, Oldstone MB (2001) Differential regulation of antiviral T-cell sition in lymphocyte development. Genes Dev 17:1703–1708. immunity results in stable CD8+ but declining CD4+ T-cell memory. Nat Med 7:913–919. 25. Lee CH, et al. (2006) Regulation of the germinal center gene program by interferon 4. Kaech SM, Wherry EJ, Ahmed R (2002) Effector and memory T-cell differentiation: (IFN) regulatory factor 8/IFN consensus sequence-binding protein. J Exp Med 203: – Implications for vaccine development. Nat Rev Immunol 2:251 262. 63–72. 5. Glimcher LH, Townsend MJ, Sullivan BM, Lord GM (2004) Recent developments in the 26. Wang H, et al. (2008) IRF8 regulates B-cell lineage specification, commitment, and transcriptional regulation of cytolytic effector cells. Nat Rev Immunol 4:900–911. differentiation. Blood 112:4028–4038. 6. Williams MA, Bevan MJ (2007) Effector and memory CTL differentiation. Annu Rev 27. Ouyang X, et al. (2011) Transcription factor IRF8 directs a silencing programme for Immunol 25:171–192. TH17 cell differentiation. Nat Commun 2:314. 7. Rutishauser RL, Kaech SM (2010) Generating diversity: Transcriptional regulation of 28. Shibaki A, Sato A, Vogel JC, Miyagawa F, Katz SI (2004) Induction of GVHD-like skin effector and memory CD8 T-cell differentiation. Immunol Rev 235:219–233. disease by passively transferred CD8(+) T-cell receptor transgenic T cells into keratin 8. Kaech SM, Hemby S, Kersh E, Ahmed R (2002) Molecular and functional profiling of 14-ovalbumin transgenic mice. J Invest Dermatol 123:109–115. memory CD8 T cell differentiation. Cell 111:837–851. 29. Changelian PS, et al. (2003) Prevention of organ allograft rejection by a specific Janus 9. Sullivan BM, Juedes A, Szabo SJ, von Herrath M, Glimcher LH (2003) Antigen-driven – effector CD8 T cell function regulated by T-bet. Proc Natl Acad Sci USA 100: kinase 3 inhibitor. Science 302:875 878. 15818–15823. 30. Ascherman DP, Migone TS, Friedmann MC, Leonard WJ (1997) Interleukin-2 (IL-2)- 10. Juedes AE, Rodrigo E, Togher L, Glimcher LH, von Herrath MG (2004) T-bet controls mediated induction of the IL-2 receptor alpha chain gene. Critical role of two func- autoaggressive CD8 lymphocyte responses in type 1 diabetes. J Exp Med 199:1153–1162. tionally redundant tyrosine residues in the IL-2 receptor beta chain cytoplasmic do- 11. Pearce EL, et al. (2003) Control of effector CD8+ T cell function by the transcription main and suggestion that these residues mediate more than Stat5 activation. J Biol factor Eomesodermin. Science 302:1041–1043. Chem 272:8704–8709. 12. Intlekofer AM, et al. (2005) Effector and memory CD8+ T cell fate coupled by T-bet 31. Welte T, et al. (1999) The PTK-STAT signaling pathway has essential roles in T-cell and eomesodermin. Nat Immunol 6:1236–1244. activation in response to antigen stimulation. Cold Spring Harb Symp Quant Biol 64: 13. Miyagawa F, et al. (2008) IL-15 serves as a costimulator in determining the activity of 291–302. autoreactive CD8 T cells in an experimental mouse model of graft-versus-host-like 32. Sharpe AH (2009) Mechanisms of costimulation. Immunol Rev 229:5–11. disease. J Immunol 181:1109–1119. 33. Cao X, et al. (1995) Defective lymphoid development in mice lacking expression of the 14. Miyagawa F, Gutermuth J, Zhang H, Katz SI (2010) The use of mouse models to better common cytokine receptor gamma chain. Immunity 2:223–238. understand mechanisms of autoimmunity and tolerance. J Autoimmun 35:192–198. 34. Nosaka T, et al. (1995) Defective lymphoid development in mice lacking Jak3. Science 15. Rochman Y, Spolski R, Leonard WJ (2009) New insights into the regulation of T cells 270:800–802. – by gamma(c) family cytokines. Nat Rev Immunol 9:480 490. 35. Thomis DC, Gurniak CB, Tivol E, Sharpe AH, Berg LJ (1995) Defects in B lymphocyte 16. Tamura T, Yanai H, Savitsky D, Taniguchi T (2008) The IRF family transcription factors maturation and T lymphocyte activation in mice lacking Jak3. Science 270:794–797. – in immunity and oncogenesis. Annu Rev Immunol 26:535 584. 36. Valitutti S, Müller S, Dessing M, Lanzavecchia A (1996) Different responses are elicited 17. Driggers PH, et al. (1990) An -regulated protein that binds the in cytotoxic T lymphocytes by different levels of T cell receptor occupancy. J Exp Med interferon-inducible enhancer element of major histocompatibility complex class I 183:1917–1921. genes. Proc Natl Acad Sci USA 87:3743–3747. 37. Auphan-Anezin N, Verdeil G, Schmitt-Verhulst AM (2003) Distinct thresholds for CD8 18. Nelson N, et al. (1996) Expression of IFN regulatory factor family proteins in lym- T cell activation lead to functional heterogeneity: CD8 T cell priming can occur in- phocytes. Induction of Stat-1 and IFN consensus sequence binding protein expression dependently of cell division. J Immunol 170:2442–2448. by T cell activation. J Immunol 156:3711–3720. 38. Waldmann TA, Tagaya Y (1999) The multifaceted regulation of interleukin-15 ex- 19. Schiavoni G, et al. (2002) ICSBP is essential for the development of mouse type I in- terferon-producing cells and for the generation and activation of CD8alpha(+) den- pression and the role of this cytokine in NK cell differentiation and host response to – dritic cells. J Exp Med 196:1415–1425. intracellular pathogens. Annu Rev Immunol 17:19 49. 20. Aliberti J, et al. (2003) Essential role for ICSBP in the in vivo development of murine 39. Dubois S, Mariner J, Waldmann TA, Tagaya Y (2002) IL-15Ralpha recycles and presents CD8alpha + dendritic cells. Blood 101:305–310. IL-15 in trans to neighboring cells. Immunity 17:537–547. 21. Tsujimura H, Tamura T, Ozato K (2003) Cutting edge: IFN consensus sequence binding 40. Lodolce JP, Burkett PR, Boone DL, Chien M, Ma A (2001) T cell-independent in- protein/IFN regulatory factor 8 drives the development of type I IFN-producing terleukin 15Ralpha signals are required for bystander proliferation. J Exp Med 194: plasmacytoid dendritic cells. J Immunol 170:1131–1135. 1187–1194. 22. Holtschke T, et al. (1996) Immunodeficiency and chronic myelogenous leukemia-like 41. Committee on Care and Use of Laboratory Animals (1985) Guide for the Care and Use syndrome in mice with a targeted mutation of the ICSBP gene. Cell 87:307–317. of Laboratory Animals (Natl Inst Health, Bethesda), DHHS Publ No (NIH) 85–23.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1201453109 Miyagawa et al. Downloaded by guest on September 30, 2021