A Two-Step Process of Effector Programming Governs CD4+ T Cell Fate Determination Induced by Antigenic Activation in the Steady State.

Various processes induce and maintain immune tolerance, but effector T cells still arise under minimal perturbations of homeostasis through unclear mechanisms. We report that, contrary to the model postulating primarily tolerogenic mechanisms initiated under homeostatic conditions, effector programming is an integral part of T cell fate determination induced by antigenic activation in the steady state. This effector programming depends on a two-step process starting with induction of effector precursors that express Hopx and are imprinted with multiple instructions for their subsequent terminal effector differentiation. Such molecular circuits advancing specific terminal effector differentiation upon re-stimulation include programmed expression of interferon-γ, whose production then promotes expression of T-bet in the precursors. We further show that effector programming coincides with regulatory conversion among T cells sharing the same antigen specificity. However, conventional type 2 dendritic cells (cDC2) and T cell functions of mammalian target of rapamycin complex 1 (mTORC1) increase effector precursor induction while decreasing the proportion of T cells that can become peripheral Foxp3+ regulatory T (pTreg) cells.

INTRODUCTION

The steady-state conditions have been considered as synonymous with an induction and maintenance of immune tolerance (Iberg and Hawiger, 2020a; Iberg et al., 2017; Steinman et al., 2003). However, T cells with effector and memory phenotype and functions can paradoxically also arise under minimal perturbations of homeostasis (Huang et al., 2003; Kawabe et al., 2017; Long et al., 2007; Vokali et al., 2020). It remains unclear how autoimmune and other effector responses can be initiated despite the presence of mechanisms that induce and maintain T cell tolerance in the steady state (ElTanbouly et al., 2020; Iberg and Hawiger, 2020a; Iberg et al., 2017; Vokali et al., 2020). The current model postulates that in the steady state, dendritic cells (DCs) with tolerogenic functions induce peripheral Foxp3+ regulatory T (pTreg) cells, whereas pro-inflammatory conditions skew toward priming of effector T cells (effectors) (Iberg et al., 2017; Iwasaki and Medzhitov, 2015; Merad et al., 2013; Pulendran, 2015; Segura and Amigorena, 2013; Steinman, 2012; Yatim et al., 2017; Zelenay and Reis e Sousa, 2013). Nevertheless, after activation in the steady state, substantial numbers of the initially activated antigen-specific CD4+ T cells, including among the monoclonal populations of T cell receptor transgenic (TCR tg) cells, fail to convert into pTreg cells (Iberg et al., 2017; Jones and Hawiger, 2017). Such Foxp3neg T cells do not induce the expression of the established key regulators of T cell fate, and the functional polarization of these T cells remains unclear (Iberg et al., 2017; Jones et al., 2016). We now found that after their antigenic activation in the steady state, CD4+ T cells undergo a two-step process of fate determination that begins with an early induction of effector precursors followed by their programmed terminal differentiation. This initial process is independent of an expression of well-established master regulators of T cell fate. Instead, the effector precursors are characterized by the early induced expression of Homeodomain only protein (Hopx).

Expression of Hopx in naive CD4+ T cells can be induced de novo during differentiation, such as during a conversion of pTreg cells (Hawiger et al., 2010; Jones et al., 2015). Hopx does not govern the initial conversion of pTreg cells, but instead Hopx controls the maintenance of mature Foxp3+ pTreg cells during their suppressor functions under the pro-inflammatory conditions (Hawiger et al., 2010; Jones and Hawiger, 2017; Jones et al., 2015). Hopx is also expressed in some Foxp3neg differentiated lymphocytes, including CD4+ effector T cells (such as cytotoxic CD4+ cells) (Albrecht et al., 2010; Cano-Gamez et al., 2020; De Simone et al., 2019; Serroukh et al., 2018). In addition to lymphocytes, Hopx is also present in various non-hematopoietic tissues and organs, including brain, heart, intestines, stem cells, and various tumors (Mariotto et al., 2016). Contrasting with such semi-ubiquitous expression of Hopx, its genetic deletion leads to only a few specific abnormalities (Jones and Hawiger, 2017; Mariotto et al., 2016). Instead, Hopx has recently emerged as a crucial marker of the specific developmental/differentiation potentials of progenitor populations in various non-hematopoietic tissues both in human and mouse systems (Mariotto et al., 2016). Our current results now uncovered the key roles of Hopx in indicating specific differentiation potentials induced early in CD4+ T cells following their activation in the steady state. We found that Hopxhi effector precursors become imprinted with transcriptional and epigenetic (defined by methylation and acetylation of the histone H3 associated with active genes and enhancers) instructions for their subsequent terminal differentiation that may result in autoimmune functions. These molecular circuits that advance such robust terminal differentiation even under minimally skewing conditions include a programmed expression of interferon-γ (IFN-γ), whose production then promotes expression of the crucial master regulator T-bet in effector precursors. Our findings further reveal that effector programming coincides with a process of regulatory conversion among T cells sharing the same antigen specificity, consistent with the maintenance of tolerance. Functions of mammalian target of rapamycin (mTOR) complex 1 (mTORC1) and antigenic activation by conventional type 2 DCs (cDC2) increase the induction of effector precursors while correspondingly reducing numbers of T cells that can become pTreg cells. Overall, a two-step process of fate determination extends the plasticity of the responses that are initiated in the immune system in the steady state.

RESULTS

Induction of Regulatory and Effector-Related T Cells in the Steady State

To examine T cells that fail to efficiently convert into Foxp3+ pTreg cells, we delivered a self-antigen, myelin oligodendrocyte glycoprotein (MOG), to induce a specific pTreg cell conversion by tolerogenic DEC-205+ DCs, using well-established methods of targeted antigen delivery in vivo with recombinant chimeric antibodies (Hawiger et al., 2001; Iberg and Hawiger, 2019, 2020b). Because Foxp3+ pTreg cells express Hopx (Hawiger et al., 2010), we used HopxFoxp3 double reporter mice (Jones et al., 2015) for this analysis. The results revealed that some T cells that remained Foxp3neg also expressed Hopx (Figure S1A). We further found that expression of Hopx was induced early (before Foxp3 expression) following activation in vivo in about half of the T cells (Figure S1B). We also found a similar early induction of Hopx expression in other antigen-specific T cells, as well as T cells activated by their corresponding cognate antigens that were specifically delivered to DCs by targeting through either DEC-205 or CD11c present on DCs (Figures S1C and S1D). Such Foxp3neg T cells with either “high” (induced) or “low” (not induced) expression of Hopx (hereafter referred to as Hopxhi and Hopxlo T cells) showed comparable patterns of proliferation, and also expression of multiple markers (Figure S1E), including Ly6c and CD5 that were previously identified to indicate specific differentiation potentials in T cells (Fulton et al., 2015; Henderson et al., 2015; Mandl et al., 2013; Martin et al., 2013; Persaud et al., 2014).

Because expression of Hopx was previously found in multiple various types of differentiated T cells, including Treg cells and effectors (Albrecht et al., 2010; De Simone et al., 2019; Hawiger et al., 2010; Jones et al., 2015), we hypothesized that an early induction of Hopx expression in naive T cells activated in the steady state may uniquely distinguish varying differentiation potentials of such T cells. To test this, we isolated Hopxhi and Hopxlo T cells that we initially induced in vivo and transferred these Foxp3neg T cells into new recipients. We then re-stimulated such T cells with the same antigen presented in vivo by DEC-205+ DCs (Figure 1A). These conditions have been established before to promote a robust conversion of antigen-specific naive T cells into Foxp3+CD25+ pTreg cells (Iberg et al., 2017; Jones et al., 2016), and consistently, we observed an almost 60% conversion of pTreg cells among Hopxlo T cells. In contrast, significantly fewer (20%) Hopxhi T cells converted into pTreg cells (Figure 1B). To independently confirm such varying Treg cell differentiation potentials of Hopxhi and Hopxlo T cells, we tested Hopxhi and Hopxlo T cells for their ability to convert into Foxp3+CD25+ induced Treg (iTreg) cells in vitro. Given that most Hopxhi and Hopxlo T cells did not downregulate CD62L expression following their initial induction in vivo, we used for these and all other subsequent experiments T cells that were CD62Lhi (as shown in Figure S1F) to avoid skewing of the results by T cells that might have already differentiated. Such Hopxhi and Hopxlo (CD62Lhi) T cells had a similar expression of antigen-specific TCR alpha and beta chains and were also similarly activated in comparison with naive T cells as indicated by increased expression of CD69 and Nur77, as well as a downregulation of VISTA, whose expression is associated with T cell quiescence (ElTanbouly et al., 2020) (Figures S1G–S1I). However, the conversion to iTreg cells was impaired among multiple tested populations of Hopxhi T cells that were specifically induced in vivo (Figures S1J–S1L), despite a comparable survival of the Hopxhi, Hopxlo, and naive T cells (Figure S1M). In mature Foxp3+ pTreg cells, Hopx controls the maintenance of these T cells during their suppressor functions under pro-inflammatory conditions. However, Hopx does not govern the conversion of pTreg cells (Hawiger et al., 2010; Jones and Hawiger, 2017; Jones et al., 2015). In accordance with these previous results, we observed that Hopx and Hopx−Foxp3neg T cells pre-activated by cognate antigen in the steady state did not differ in their specific ability to convert into iTreg cells (Figures S1N and S1O). Overall, we conclude that despite being dispensable for the induction of Treg cells, an early induction of Hopx expression in T cells activated in the steady state serves as a selective marker of their decreased potential to convert into Treg cells de novo.

Figure 1.

Dendritic Cells Induce Regulatory and Effector-Related T Cells in the Steady State

(A and B) Induction of Hopxhi and Hopxlo T cells that differ in their pTreg cell conversion potential in vivo.

(A) General experimental outline for examining Hopxhi and Hopxlo T cells. Plots (representative of multiple independent experiments) show Hopx (GFP) and Foxp3 (RFP) expression, and the corresponding regions indicate gating for automated cell sorting in pooled CD4+ T cells combined from peripheral lymph nodes and spleens obtained from MOG-specific TCR tg (2D2) HopxFoxp3 mice 5 days after treatment with αDEC-MOG chimeric antibody. Arrows indicate adoptive transfers of Hopxhi and Hopxlo T cells into WT recipient mice that were then treated with either αDEC-MOG or PBS as indicated.

(B) Representative plots show Foxp3 (RFP) expression and anti-CD25 staining intensity in the transferred CD4+ T cells among splenocytes from the recipients as in (A) analyzed by flow cytometry after 21 days. Numbers in quadrants show corresponding percentages. Graph shows mean ± standard deviation (SD) percentages of Foxp3+CD25+ pTreg cells among transferred CD4+ T cells in the indicated groups of recipients; ***p < 0.001 determined by unpaired t test, n = 4 per group from two independent experiments.

(C) Expression of master regulators. The graphs show expression of Tbx21, Gata3, Rorc, Sfpi1, Bcl6, Irf4, Batf, Stat3, and Nr4a1 determined by RNA-sequencing (RNA-seq) in multiple independent replicates from the Hopxhi, Hopxlo, and naive 2D2 T cells obtained as in Figure S1F and analyzed as described in Star Methods. None of those genes have significantly different expression between Hopxhi and Hopxlo T cells.

(D) Gene set enrichment analysis (GSEA). Individual genes whose expression is significantly different between Hopxhi and Hopxlo T cells and that were identified by the GSEA are shown as present in the indicated numbers of publicly available gene sets either positively (containing genes with upregulated expression in effector/memory T cells) or negatively (containing genes with downregulated expression in effector/memory T cells) associated with effector or memory T cells. The vertical axes show the specific ratio of expression of such genes in Hopxhi T cells and Hopxlo T cells, and red and blue dots denote individual genes with either upregulated or downregulated, respectively, expression in Hopxhi T cells in comparison with Hopxlo T cells. A shaded contour represents the distribution of projections of the dots onto the horizontal plane.

(E) Immune gene expression. Heatmap shows expression of additionally identified genes whose expression is significantly different between Hopxhi and Hopxlo T cells and that are associated with specific immune cell functions (see text and Table 1). Multiple independent replicates are shown.

Under homeostatic steady-state conditions, endogenous T cells are constantly exposed to stimulation by agonistic peptides originating from self, as well as foreign antigens, possibly resulting in their activation and initial differentiation. To test the expression of Hopx among such polyclonal endogenous CD4+ T cells, we analyzed CD4+Foxp3neg T cells isolated from unmanipulated TCR wild-type (WT; non-TCR tg) HopxGFPFoxp3RFP mice. Consistent with a previously suggested expression of Hopx in some effector CD4+ T cells (Albrecht et al., 2010; Cano-Gamez et al., 2020; Serroukh et al., 2018), we found Hopx+ cells among the CD4+Foxp3negCD62Llo population. However, we also detected a small population of Hopx+ cells among the CD4+Foxp3negCD62Lhi endogenous T cells (Figure S2A). We propose that this Hopx+CD4+Foxp3negCD62Lhi population arises as a result of a constant activation of polyclonal T cells under homeostatic conditions, and these T cells may correspond to Hopxhi T cells that we identified above. However, it is currently not possible to distinguish polyclonal Hopxlo and other T cells that lack expression of Hopx. Also, the timing of activation and induction of Hopx expression in specific T cells among the polyclonal population is rather unpredictable, further constraining an in-depth analysis of these mechanisms in the polyclonal repertoire. Therefore, to explain the observed differences in the fate of Hopxhi and Hopxlo T cells, we performed a transcriptome analysis of Hopxhi, Hopxlo, and naive T cells obtained from TCR tg animals using RNA sequencing (RNA-seq). Principal-component analysis (PCA) revealed numerous differences in the transcriptomes of these T cells, and expression of multiple individual genes differed significantly between the Hopxhi, Hopxlo, and naive T cells (Figures S2B and S2C). However, the genes whose expression was specifically altered in either Hopxhi or Hopxlo T cells in comparison with naive T cells had only a small overlap with the genes in the multiple clusters and modules that were previously identified in naive and some specifically activated T cells (ElTanbouly et al., 2020) (Figure S2D). Also, the expression of multiple master regulators of T cell differentiation (Wang et al., 2015), including Tbx21, Gata3, Rorc, Sfpi1, Bcl6, Irf4, Batf, Stat3, and Nr4a1, represented here as biological “averages” of their corresponding expression among huge numbers of individual cells, did not differ between such Hopxhi and Hopxlo T cells that are characterized by distinct and non-overlapping expression of Hopx (Figure 1C).

To characterize the relevant differences in gene expression between Hopxhi and Hopxlo T cells, we performed gene set enrichment analysis (GSEA) (Subramanian et al., 2005) using a comprehensive group of publicly available gene sets related to functional subsets of CD4+ T cells (Figure S3). This analysis revealed multiple enrichments of specific gene sets containing genes positively associated with effector differentiation (gene expression upregulated in effector CD4+ T cells) among the genes whose expression was upregulated in Hopxhi T cells in comparison with Hopxlo T cells. The specific genes identified by GSEA included genes known to govern biological processes, including development, metabolism, and proliferation (Figures S4A and S4B; Table 1). All of these specific genes were also individually present in multiple other gene sets positively associated with effector/memory T cells (Figure 1D). The expression of Il2ra was downregulated in Hopxhi T cells in comparison with Hopxlo T cells (Figure S4B), likely consistent with the known requirement of Il2ra for Treg cell differentiation and functions (Fontenot et al., 2005) and the observed decreased conversion of Hopxhi T cells into pTreg cells. In addition to the genes identified by GSEA, a separate literature-based analysis of the genes whose expression differed between Hopxhi and Hopxlo T cells revealed in Hopxhi T cells several other genes specifically associated with the differentiation of effector and memory immune cells (Figure 1E; Table 1). We also performed an analysis of microRNAs (miRNAs) whose expression significantly differed between Hopxhi and Hopxlo T cells. We found an increased expression in Hopxhi T cells of Mir21a, Mir183, Mir31, and Mir148a, all previously proposed to either promote T helper (Th) 1/17 effector differentiation or limit Treg cell differentiation. In contrast, expressions of Mir150 and Mir26a, both positively associated with Treg cell differentiation, were decreased in Hopxhi T cells (Figure S4C; Table 1). Overall, we conclude that despite the initial activation under generally tolerogenic steady-state conditions, Hopxhi T cells resist the conversion into Treg cells and acquire a unique transcriptional signature that includes multiple genes related to Th1/17 cells, while lacking the major master regulators of effector differentiation.

Table 1.

Genes Related to Effector Differentiation and Functions Specifically Identified in Effector Precursors by Transcriptional and Epigenetic Analyses

Gene	Function
Ace	regulates Th1/Th17 cell cytokine production and migration (Platten et al., 2009; Uzawa et al., 2014)
Ahr	regulates both Treg and Th17 cell differentiation in autoimmune disease (Quintana et al., 2008)
Alcam	regulates CD4+ T cell migration across blood-brain barrier endothelium during EAE (Cayrol et al., 2008)
Aspm	regulates cyclin activity by modulating its phosphorylation and localization into the nucleus (Capecchi and Pozner, 2015)
Bhlhe40	promotes effector T cell differentiation by positively regulating GM-CSF and inhibiting the production of IL-10 (Lin et al., 2014; Martínez-Llordella et al., 2013)
Ccr6	regulates inflammatory T cell priming and lymphoid egress in EAE and arthritis (Liston et al., 2009; Razawy et al., 2020)
Cd38	associated with expression of Th2 effector cytokine IL-13 (Scalzo-Inguanti and Plebanski, 2011)
Cd70	important for T-bet and IFN-γ expression by both Th1 and Th17 effector cells (Dhaeze et al., 2019)
Cdk1	regulates Cdk2/cyclin A2 activity (Diril et al., 2012)
Cdr2	correlates with attenuated hypoxia-inducible factor responses (Balamurugan et al., 2009)
Cxcr3	required for optimal differentiation of IFN-γ-secreting Th1 effector cells (Groom et al., 2012)
Cxcr5	mediates co-localization of effector Tfh cells and B cells in lymphoid organs during arthritis pathogenesis (Moschovakis et al., 2017)
Dusp14	regulates T cell proliferation and cytokine production (Yang et al., 2014)
Eea1	regulates the polarization of T cell and NK cell cytotoxic granules (Chiang et al., 2017)
Egr4	interaction with nuclear mediator NF-κB regulates the transcription of inflammatory cytokines (Wieland et al., 2005)
Fgf2	synergistically exacerbates tissue inflammation by regulating the production of cytokines and chemokines (Shao et al., 2017)
Foxm1	increases Ube2c transcription in glioma cells (Guo et al., 2017)
Il12rb2	modulates the differentiation of IFN-γ+ Th1 effector cells (Bettelli and Kuchroo, 2005; Martínez-Barricarte et al., 2018)
Il13	mediates the expansion of Th2 cells (Deepak and Acharya, 2012; Scalzo-Inguanti and Plebanski, 2011)
Il1r2	negatively regulates IL-1 signaling (Peters et al., 2013)
Il9r	mediates Th17-driven inflammatory disease (Li et al., 2010; Nowak et al., 2009)
Kif4	suppresses the activity of DNA repair enzyme poly(ADP-ribose) polymerase-1 (Midorikawa et al., 2006)
Kpna2	regulates the expression of glycolytic genes downstream of effector-inducing mTORC1 (Chen et al., 2016)
Map3k8	activates mTORC1, promotes Th1/Th17differentiation, and suppresses iTreg differentiation (Acuff et al., 2015; Li et al., 2016)
Mir148a	controls Th1 cell survival by regulating Bim expression (Haftmann et al., 2015)
Mir150	cooperates with miR-150 to repress the expression of the Th17-promoting factor mTOR (Warth et al., 2015)
Mir183	enhances Th17 cell cytokine production and autoimmunity, and represses expression of the transcription factor Foxo1 (Ichiyama et al., 2016)
Mir21a	promotes Th17 differentiation by targeting and depleting SMAD-7, a negative regulator of transforming growth factor β (TGF-β) signaling (He et al., 2017; Murugaiyan et al., 2015)
Mir26a	downregulates Th17 and upregulates Treg cell function by targeting IL-6 (Zhang et al., 2015b)
Mir31	inhibits the induction of pTreg cells by suppressing Gprc5a (Zhang et al., 2015a)
Nkg7	upregulated by Th1-polarized T cells (Lund et al., 2005)
Nrp2	controls migration of T cell precursors (Mendes-da-Cruz et al., 2014)
Pdcd1	associated with Th cells in proliferative lupus nephritis (Caielli et al., 2019)
Plxnb2	required for normal germinal center humoral responses (Yan et al., 2017)
Ptger2	promotes differentiation and functions of Th1/Th17 cells (Boniface et al., 2009; Yao et al., 2009)
Ptpn3	regulates TCR signaling (Han et al., 2000; Sozio et al., 2004)
Runx2	required for memory T cell persistence (Olesin et al., 2018)
Rgs16	regulates T cell activation and specific chemokine receptor-mediated migration (Lippert et al., 2003)
S1004	expression identifies memory CD4+ T cells (Weatherly et al., 2015)
Shcbp1	regulates Th1/Th17 cell-mediated autoimmunity (Buckley et al., 2014)
Syk	activated in pathogenic CD4+ T effectors (Chauhan et al., 2016; Krishnan et al., 2003)
Tjp2	interacts with SNX27 at the T cell immunological synapse (Tello-Lafoz et al., 2017) and regulates proliferation, cell size, and apoptosis (González-Mariscal et al., 2017)
Tnfsf4	promotes T cell survival, effector T cell phenotype, T cell memory, and reduction of regulatory function (Webb et al., 2016)
Ube2c	promotes phosphatidylinositol 3-kinase (PI3K)-Akt-mTOR pathway (Guo et al., 2017)
Uhrf1	regulates iNKT apoptosis and Tbet+ iNKT1 effector frequencies (Cui et al., 2016)

Terminal Differentiation of Effector Precursors

Based on the above results, we hypothesized that Hopxhi T cells might be specialized precursors of effector T cells. Consistent with this hypothesis, we found that after re-stimulation under non-skewing (Th0) conditions in vitro, Hopxhi T cells robustly induced the expression of T-bet, a crucial master regulator of Th1 differentiation. This induction of T-bet expression by Hopxhi T cells was highly specific, with only very few Hopxlo or naïve T cells inducing T-bet expression under these conditions (Figure 2A). A similar induction of T-bet expression was also observed among all other tested populations of Hopxhi T cells that were either of different antigen specificity or were induced in vivo by different subsets of cDCs (Figures S5A and S5B). In contrast, no cells in the Hopxhi, Hopxlo, or naive T cell populations induced expression of Foxp3 under these Th0 conditions. This is further consistent with a specific effector skewing of Hopxhi T cells (Figure S5C).

Figure 2.

Effector Precursors Undergo Robust Terminal Differentiation In Vitro and In Vivo

(A) Induction of T-bet expression. Representative histograms show anti-T-bet intracellular staining intensity analyzed by flow cytometry in Hopxhi, Hopxlo, and naive 2D2 T cells that were obtained as in Figure S1F and then re-stimulated in vitro under Th0 conditions for either 1 or 3 days as indicated. Graphs show mean ± SD median fluorescence intensity (MFI) in the indicated groups, n = 6–20 replicates from three independent experiments using pooled material from multiple mice per group.

(B) General experimental outline of analyzing the differentiation and functions in vivo of the Hopxhi, Hopxlo, and naive 2D2 T cells that were obtained as in Figure S1F and then adoptively transferred into multiple independent groups of Rag− recipient mice also treated with pertussis toxin (PT).

(C) Induction of IFN-γ expression. Representative plots show anti-CD4 and intracellular anti-IFN-γ staining intensity in the transferred CD4+CD3+ T cells among splenocytes analyzed by flow cytometry after 12 days and then additional re-stimulation in vitro. Graphs show mean ± SD percentages of IFN-γ+ cells among transferred T cells in the indicated groups, n = 5–6 individual mice per group from two independent experiments.

(D–F) Terminal effector differentiation and acquisition of autoimmune functions of Hopxhi, Hopxlo, and naive T cells that were transferred as in (B).

(D) Graphs show mean ± SD percentages of T-bet+ cells among transferred T cells in the indicated groups after 7, 11, or 14 days as indicated.

(E) Graphs show mean ± SD percentages of T-bet+RORγt+ cells among transferred T cells in the indicated groups after 7, 11, or 14 days as indicated. (D and E) n = 3–7 individual mice per group from three independent experiments.

(F) Graphs show mean ± SEM EAE disease scores in the indicated groups and at the indicated days; n = 25 individual mice per group from five independent experiments.

*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, determined by Welch’s t test (A and C) or two-way ANOVA with Fisher’s least significant difference (LSD) (D–F). ns, not significant.

An activation of T cells in the steady state is known to prevent subsequent immunogenic (autoimmune) activation (Iberg et al., 2017). Therefore, to clarify the impact of terminal differentiation of Hopxhi T cells in vivo on autoimmune responses, we examined the responses of self-reactive, neural antigen (MOG)-specific Hopxhi, Hopxlo, and naive T cells that we transferred into separate groups of Rag1−mice that we also treated with pertussis toxin (PT) (Figure 2B). Neither Hopxhi nor Hopxlo T cells induced expression of Foxp3 (Figure S5D). However, under these mild PT-induced pro-inflammatory conditions, about 30% of Hopxhi T cells expressed T-bet and produced IFN-γ, a crucial cytokine of Th1 effectors (Lighvani et al., 2001; Zhu, 2018). In contrast, fewer than 10% of Hopxlo or naive T cells started production of IFN-γ or T-bet expression in most animals within 11–12 days (Figures 2C and 2D; Figure S5E). About 10% of the Hopxhi T cells also became double expressors of T-bet and RORγt, while such T-bet+RORγt+ cells were almost completely absent from the Hopxlo and naive populations (Figure 2E; Figure S5E). The double expression of T-bet and RORγt by Hopxhi T cells could be consistent with a formation of “pathogenic” Th17 cells (Awasthi et al., 2009; Hirota et al., 2011; Lee et al., 2012; Wu et al., 2020). Overall, Hopxhi T cells are effector precursors prone to an enhanced terminal effector differentiation in vivo.

To elucidate the relevant immune functions of the self-reactive Hopxhi effector precursors, we followed the development of experimental autoimmune encephalomyelitis (EAE) in Rag1−mice transferred with either Hopxhi, Hopxlo, or naive T cells and also treated with PT. This experimental system precludes any activation of other encephalitogenic T cells, therefore allowing for a direct assessment of the impact on a disease process of the effector programming of transferred T cells. Mice that received Hopxlo or naive T cells developed only a baseline disease consistent with a limited induction of effectors from these T cells. In contrast, and in accordance with the robust terminal effector differentiation of Hopxhi cells, mice that received these cells developed more severe symptoms of EAE (Figure 2F). Overall, we conclude that Hopxhi effector precursors readily complete their terminal effector differentiation both in vitro and in vivo, and the self-reactive effector precursors can acquire enhanced autoimmune functions.

Molecular Instructions Governing the Differentiation of Effector Precursors

Because the specific transcriptional signature of effector precursors lacks master regulators known to directly govern effector differentiation, we hypothesized that the genes encoding such relevant regulators of T cell fate would be marked epigenetically for expression. To identify such epigenetic modifications associated with transcriptional activation in Hopxhi, Hopxlo, and naive T cells, we performed chromatin immunoprecipitation followed by sequencing (ChIP-seq) and examined trimethylation of the lysine 4 of the histone H3 (H3K4me3), as well as acetylation of the lysine 27 of the histone H3 (H3K27Ac), which are typically associated with active genes and enhancers. This analysis revealed in Hopxhi effector precursors specific H3K4me3 and H3K27Ac enrichments at multiple genes, including Ahr, Bhlhe40, Ccr6, Cxcr3, Cxcr5, Ifng, Map3k8, Ptger2, and Shcbp1 (Figure 3A). These genes encode crucial cytokines, receptors, transcription factors, and other mediators important for Th1 and Th17 effector mechanisms (Table 1), consistent with the subsequent terminal differentiation of effector precursors. However, we found no specific H3K4me3 or H3K27Ac enrichments in Hopxhi effector precursors at Tbx21 and Rorc genes that encode crucial master regulators T-bet and RORγt, respectively. We also found either increased or similar H3K4me3 and H3K27Ac enrichments at multiple genes in the Hopxlo T cells relative to the Hopxhi T cells (Figure 3B). Most such genes have no known specific functions in governing T cell differentiation. Therefore, these data stress that epigenetic programming in Hopxhi T cells is overall specific to the genes relevant for T cell differentiation, and especially effector functions.

Figure 3.

Epigenetic Modifications of T Cell Effector Fate Regulators in Effector Precursors

(A and B) Hopxhi, Hopxlo, and naive 2D2 T cells were obtained as in Figure S1F and analyzed by ChIP-seq. Genomic tracks were calculated from multiple independent experiments and show H3K4me3 and H3K27Ac enrichments for indicated genes or established enhancer regions (in case of Ifng) at the given positions in the genome.

(A) Genomic tracks shown for identified relevant regulators of T cell fate.

(B) Genomic tracks shown for select genes with no known functions in T cell differentiation.

Although the genes identified above are marked by epigenetic alterations associated with transcriptional activation, most of these genes are not yet expressed in the Hopxhi effector precursors (Figure 1; Figure S4). This suggests these cells may be poised for a process of terminal differentiation, dependent upon the rapidly upregulated expression of the crucial genes following T cell re-stimulation. To identify such genes, we examined by RNA-seq the transcriptomes of the Hopxhi, Hopxlo, and naive T cells that were re-stimulated under non-skewing conditions in vitro. As already shown in Figure 2A, only Hopxhi, but not Hopxlo or naive T, cells induce expression of T-bet under these conditions. In addition to induction of gene expression of Tbx21 that encodes T-bet, we found in Hopxhi T cells specifically increased expression of several other genes relevant for effector functions (Figure 4A; Table 1), and some of these genes were also identified in Figure 3 as epigenetically marked for expression. Among all such T cell-relevant genes whose expression in Hopxhi and Hopxlo T cells was found to be significantly different after the re-stimulation, but not in the absence of re-stimulation, expression of Ifng that encodes IFN-γ was most upregulated (Figure 4A). This rapid induction of Ifng expression in Hopxhi T cells is consistent with the specific epigenetic modifications identified in Figure 3 in the established Ifng enhancer region (Shnyreva et al., 2004) and also in the promoter of Ifngas1, which is involved in positive regulation of Ifng expression (Collier et al., 2012; Gomez et al., 2013; Petermann et al., 2019). Such induced Ifng expression also corresponds with the robust induction of IFN-γ production observed in Hopxhi T cells in vivo (Figure 2C), and the specific production of IFN-γ by these cells was maintained even under Treg-skewing conditions in vitro (Figure 4B).

Figure 4.

Programmed Expression of Interferon-g in Effector Precursors Mediates an Expression of T-bet

(A and B) Robust induction of Ifng expression and IFN-γ production upon re-stimulation.

(A) Hopxhi, Hopxlo, and naive 2D2 T cells were obtained as in Figure S1F and then additionally re-stimulated in vitro under Th0 conditions for 12 h as indicated (see Star Methods). RNA-seq heatmaps show expression of T cell-relevant genes whose expression in Hopxhi and Hopxlo T cells was found to be significantly different after the re-stimulation, but not in the absence of re-stimulation. Multiple independent replicates of Hopxhi, Hopxlo, and naive T cell groups are shown.

(B) Hopxhi, Hopxlo, and naive 2D2 T cells were obtained as in Figure S1F and were then cultured in vitro under Treg cell-skewing conditions for 4 days. The concentration of IFN-γ was measured in the supernatants at the end of the cultures. Results show mean ± SD MFI in the indicated groups; n = 12–20 replicates from three independent experiments using pooled material from multiple mice per group. ***p < 0.001, determined by Welch’s t test.

(C) IFN-γ induces T-bet expression. Representative histograms show anti-T-bet intracellular staining intensity analyzed by flow cytometry in Hopxhi, Hopxlo, and naive 2D2 T cells that were obtained as in Figure S1F and then re-stimulated under Th0 conditions for 3 days in the presence of either anti-IFN-γ or isotype control antibody as indicated. Graphs show mean ± SD MFI in the indicated groups, n = 10–14 replicates from three independent experiments using pooled material from multiple mice per group. ***p < 0.001, determined by Welch’s t test.

IFN-γ can induce expression of T-bet (Lighvani et al., 2001; Schulz et al., 2009). We therefore hypothesized that either autocrine or paracrine production of IFN-γ by Hopxhi T cells is responsible for the subsequent induction of T-bet expression in these cells. In agreement with this hypothesis, we found that blocking of IFN-γ by a specific neutralizing antibody almost completely inhibited the expression of T-bet in Hopxhi T cells after their re-stimulation (Figure 4C). Collectively, these results show that the terminal differentiation of effector precursors depends on specific instructions, such as those promoting production of IFN-γ that subsequently induces expression of T-bet.

Regulation of Effector Precursor Induction

Crucial immunomodulatory pathways, such as B and T lymphocyte associated (BTLA) or programmed death-ligand 1 (PD-L1), can promote induction of pTreg cells (Francisco et al., 2009; Jones et al., 2016; Wang et al., 2008). Because we found that effector pre-determination of Hopxhi T cells is accompanied by the formation of Hopxlo T cells that can subsequently differentiate into pTreg cells (Figure 1), we hypothesized that specific immunomodulatory pathways may also reciprocally regulate formation of effector precursors. In agreement with the role of BTLA in promoting the conversion of Foxp3+ pTreg cells as we previously established (Jones et al., 2016), the percentage of developing Foxp3+ pTreg cells appears to be lower in the Btla−recipients. However, neither genetic ablation of BTLA nor antibody blocking of PD-L1 affected the formation of Hopxhi (Foxp3neg) and Hopxlo T cells (Figures S6A and S6B).

A de novo induction of pTreg cells is adversely affected by the activation of the mTOR (Haxhinasto et al., 2008; Sauer et al., 2008). Furthermore, mTORC1 dependent on Raptor can promote T cell priming and effector differentiation (Delgoffe et al., 2009; Geltink et al., 2018; Yang et al., 2013; Zeng and Chi, 2017). We found that despite a similar initial activation by tolerogenic DCs of Rptorfl/fl CD4cre (Rptor−) and Rptor T cells, an early conversion of pTreg cells from the adoptively transferred Rptor−T cells was increased (Figures S6C and S6D). However, mTOR has been proposed not to directly suppress Foxp3 expression in differentiating Treg cells (Harada et al., 2010; Ouyang et al., 2010). Therefore, we hypothesized that mTORC1 may instead skew the balance governing the formation of effector precursors and potential pTreg cells. In agreement with this hypothesis, we found an about 3-fold specific decrease in Hopxhi effector precursor induction in vivo from Rptor−HopxnegFoxp3neg T cells (Figure 5A) but similar total numbers of the remaining Rptor−and Rptor T cells (Figure S6E). The small numbers of Hopxhi T cells that did develop in the absence of mTORC1 could still readily induce T-bet expression upon re-stimulation under Th0 conditions, whereas, as expected, both Rptor−and Rptor Hopxlo T cells failed to induce expression of T-bet (Figure 5B). Therefore, functions of mTORC1 increase the induction of effector precursors while correspondingly reducing numbers of T cells that can potentially become pTreg cells, independently of the process of subsequent terminal effector differentiation of such T cells.

Figure 5.

cDC2 and mTORC1 Increase Induction of Effector Precursors

(A) Impact of mTORC1 on the formation of effector precursors. Hopx(GFP)negFoxp3(RFP)negCD25neg 2D2 T cells were isolated from Rptor− and Rptor 2D2 HopxFoxp3 mice and were adoptively transferred into congenically labeled recipient mice that were then treated with αDEC-MOG. Representative plots show Hopx(GFP) and Foxp3(RFP) expression in the transferred CD4+ T cells among splenocytes analyzed by flow cytometry after 3 and 5 days as indicated. Numbers in quadrants show corresponding percentages. Graphs show mean ± SD percentages of Hopxhi cells among transferred CD4+ T cells in indicated groups of recipients; n = 11–12 individual mice per group from five independent experiments.

(B) Induction of T-bet expression in the absence of mTORC1. Representative histograms show anti-T-bet intracellular staining intensity analyzed by flow cytometry in Rptor− and Rptor Hopxhi, Hopxlo, and naive 2D2 T cells that were obtained as outlined in Figure S1F from the Rptor− and Rptor 2D2 HopxFoxp3 mice pre-treated with either αDEC-MOG or PBS. T cells were then re-stimulated in vitro under Th0 conditions for 3 days as indicated. Graphs show mean ± SD MFI in the indicated groups; n = 4–15 replicates using pooled material from multiple mice per group from four independent experiments.

(C) Enhanced formation of effector precursors after antigenic stimulation mediated by cDC2. Hopx(GFP)negFoxp3(RFP)negCD25neg 2D2 T cells were adoptively transferred into congenically labeled recipient mice that were then treated with either αDEC-MOG or αDCIR2-MOG. Representative plots show Hopx(GFP) and Foxp3(RFP) expression in the transferred CD4+ T cells among splenocytes analyzed by flow cytometry after 3 and 5 days as indicated. Numbers in quadrants show corresponding percentages. Graphs show mean ± SD percentages of Hopxhi cells among transferred CD4+ T cells in the indicated groups of recipients; n = 3–6 individual mice per group from three independent experiments.

(D) T-bet expression. Representative histograms show anti-T-bet intracellular staining intensity analyzed by flow cytometry in Hopxhi, Hopxlo, and naive 2D2 T cells obtained as outlined in Figure S1F from 2D2 HopxFoxp3 mice that were pre-treated with αDCIR2-MOG and then re-stimulated in vitro under Th0 conditions for 3 days. Graphs show mean ± SD MFI in the indicated groups; n = 3–15 replicates using pooled material from multiple mice per group. Results represent one of two independent experiments.

*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, determined by two-way ANOVA with Sidak’s multiple comparisons (A and C) or Welch’s t test (B and D). ns, not significant.

Two major subsets of cDC, cDC1 and cDC2, are well established in eliciting diverse immune responses. Whereas cDC1 can efficiently uptake and present self-antigens important for induction of pTreg cells and tolerance in the steady state, cDC2 are generally poor inducers of pTreg cells (Iberg and Hawiger, 2020a; Iberg et al., 2017). In contrast, antigen presentation by cDC2 can lead to priming of multiple types of effector T cells, especially under pro-inflammatory conditions, but the roles of cDC2 in the steady state remain incompletely understood (Durai and Murphy, 2016). To examine T cell responses elicited by the antigenic stimulation specifically mediated by cDC2, we used a well-established experimental system based on chimeric antibodies specific for DCIR2 present on cDC2 to deliver antigens to these cDCs (Dudziak et al., 2007) in addition to targeting antigens through DEC-205 present on cDC1. We observed about a 50% increase in the formation of Hopxhi effector precursors among T cells following their responses to the antigen presented specifically by cDC2 compared with the induction of effector precursors mediated by cDC1 (Figure 5C). Such effector precursors induced expression of T-bet that was dependent on the produced IFN-γ (Figure 5D; Figure S6F). Overall, we conclude that cDC2 mediate the induction of effector precursors more robustly than cDC1.

DISCUSSION

The current framework underlying the functions of the immune system postulates that pro-inflammatory signals induce immunity, whereas homeostatic conditions facilitate tolerance (Iberg et al., 2017; Iwasaki and Medzhitov, 2015; Merad et al., 2013; Pulendran, 2015; Segura and Amigorena, 2013; Steinman, 2012; Yatim et al., 2017; Zelenay and Reis e Sousa, 2013). Previous results also revealed a greater extent of plasticity of T cell differentiation under nominally pro-inflammatory conditions, such as the presence of commensal microorganisms, manifested through mechanisms responsible for the induction of pTreg cells and tolerance to prevent excessive immune responses (Coombes et al., 2007; Esterházy et al., 2019; Manicassamy et al., 2010; Naik et al., 2012; Sun et al., 2007). In contrast, specific mechanisms of tolerance are induced in both CD4+ and CD8+ T cells following activation of these cells in the steady state as reviewed in Iberg and Hawiger (2020a), Iberg et al. (2017), and Steinman et al. (2003). However, a formation of T cells with transient effector characteristics has also been observed under the steady-state conditions, but the specific mechanisms governing the formation of effector T cells with potentially autoimmune functions have remained unclear (Huang et al., 2003; Kawabe et al., 2017; Long et al., 2007; Vokali et al., 2020).

Our results now revealed that effector programming is an integral part of fate determination induced in the steady state. Effector precursors are distinguished by a specific epigenetic and transcriptional profile that includes instructions for a subsequent terminal differentiation and acquisition of effector and autoimmune functions under minimal perturbations of homeostasis. Consistent with the maintenance of tolerance, effector programming can coincide with regulatory conversion among T cells sharing the same antigen specificity. However, our results suggest that these processes are mechanistically separate. The transcriptional profile of effector precursors lacks expression of established master regulators, such as T-bet or RORγt, but instead includes an induced expression of Hopx. Hopx expression has been found before in Foxp3+ and some Foxp3neg lymphocytes, including CD4+ effector T cells, although the absence of Hopx leads to only a few specific abnormalities (Albrecht et al., 2010; Cano-Gamez et al., 2020; De Simone et al., 2019; Hawiger et al., 2010; Jones et al., 2015; Mariotto et al., 2016; Serroukh et al., 2018). In mature Foxp3+ pTreg cells, Hopx controls the maintenance of these T cells during their suppressor functions under pro-inflammatory conditions. However, Hopx does not govern the conversion of pTreg cells (Hawiger et al., 2010; Jones and Hawiger, 2017; Jones et al., 2015). An early induction of Hopx expression has recently also emerged as a specific marker of the developmental and differentiation potentials of multiple non-hematopoietic progenitor populations in brain, heart, intestines, and skin (as reviewed in Mariotto et al., 2016). Our results now uncovered that expression of Hopx crucially indicates the specifically altered differentiation potentials in CD4+ T cells.

A pre-activation of CD4+ and CD8+ T cells was proposed before to facilitate their differentiation by acquisition of early expression of CD5 and other molecules (De Simone et al., 2019; Fulton et al., 2015; Henderson et al., 2015; Mandl et al., 2013; Persaud et al., 2014; Sood et al., 2019). CD4+ Hopxhi effector precursors appear to be induced independently of the specific expression of established regulators of T cell responses and instead they express multiple other genes related to Th1 and Th17 cells. The molecular circuits governing the key transition of precursors into effectors include the programmed expression of specific factors encoded by genes such as Ifng that are rapidly induced upon re-stimulation, even under non-skewing or tolerizing conditions. The production of IFN-γ then induces expression of T-bet, a crucial master regulator of T cell differentiation. Under minimally skewing conditions, the majority of Hopxhi effector precursors predominantly skew toward Th1 phenotype. However, such effectors can also become T-bet+RORγt+ cells in vivo, and Hopxhi effector precursors also induce expression of genes specific for other types of Th cells, such as Il13. This may indicate a broader differentiation plasticity of the effector precursors, possibly in the presence of additional signals. Overall, effector programming appears to be an integral part of fate determination induced by antigenic activation in the steady state. However, it also coincides with a regulatory conversion among T cells sharing the same antigen specificity, consistent with the maintenance of tolerance. Although tolerance is generally maintained by multiple mechanisms, including thymically derived tTreg cells (Josefowicz et al., 2012), specific mechanisms controlling the relative proportions of antigen-specific effector precursors and pTreg cells would be crucial for regulation of tolerance.

Antigen presentation by DCs is critical for T cell responses against foreign antigens, as well as self-antigens, and the decisions that determine specific T cell fates are at least in part influenced by the type of DCs that mediate T cell activation (Baptista et al., 2019; Durai and Murphy, 2016; Eisenbarth, 2019; Esterházy et al., 2019; Iberg and Hawiger, 2020a; Iberg et al., 2017). cDCs are the major population among both human and murine DCs and can be further divided into the cDC1 and cDC2 subsets based on their development and expression of specific markers (Collin and Bigley, 2018; Durai and Murphy, 2016; Guilliams et al., 2016). Although designations of cDC subsets do not strictly overlap with their distinct immune functions, the specific subsets are characterized by a degree of functional specialization, attributed to various factors, including different localization within the local architecture of immune organs, differences in the efficiencies of processing and presentation of antigens to T cells, and the presence of various immunomodulatory mechanisms (Baptista et al., 2019; Dudziak et al., 2007; Eisenbarth, 2019; Esterházy et al., 2019; Iberg et al., 2017). The recently characterized BTLA-HVEM-CD5 axis initiated by BTLA expressed in cDC1 helps to govern tolerogenic outcomes of interactions between cDCs and T cells (Bourque and Hawiger, 2019; Henderson et al., 2015; Jones et al., 2016). The BTLA signals that are mediated through HVEM in naive CD4+ T cells increase expression of CD5 that promotes efficient conversion of pTreg cells in concert with other immunomodulatory mechanisms (Bourque and Hawiger, 2018; Henderson et al., 2015; Iberg and Hawiger, 2020a; Jones et al., 2016). However, we found that immunomodulatory pathways, such as those mediated by BTLA or PD-L1, do not affect the induction of Hopxhi T cells. Therefore, such immunomodulatory pathways might rather govern the separate process of subsequent conversion of pTreg cells from Hopxlo T cells.

In contrast, induction of Hopxhi effector precursors is enhanced by antigenic stimulation mediated specifically by cDC2. This is consistent with the robust priming of effector T cells mediated by cDC2 (Baptista et al., 2019; Dudziak et al., 2007; Durai and Murphy, 2016; Eisenbarth, 2019). Our results now clarify the roles of cDC2 in the steady state and reveal that under such conditions, cDC2 do not prime effector T cells directly but instead induce effector precursors more efficiently than cDC1. The subsequent transition of Hopxhi effector precursors into effectors occurs during an additional re-stimulation even under minimally skewing conditions. This is in contrast with a pTreg cell conversion process that requires the continuous presence of specific signals that can be provided by tolerogenic DCs. Hopxhi effector precursors resist such Treg cell-promoting signals, in accord with their effector skewing under Th0 conditions. In contrast, and consistent with their programmed expression of IL-12 receptor, under various subsequent immune conditions these cells may possibly remain susceptible to the specific signals such as IL-12, in agreement with the roles of IL-12 to control effector T cell differentiation (van Panhuys et al., 2014). Overall, various fate-modifying signals might further enable specific differentiation programs of the effector precursors.

The functions of mTOR help to sense activation of T cells, and such mechanisms were previously proposed to inhibit Treg cell induction and promote effector differentiation (Delgoffe et al., 2009; Delgoffe and Powell, 2015; Geltink et al., 2018; Haxhinasto et al., 2008; Huang and Perl, 2018; O’Sullivan and Pearce, 2015; Sauer et al., 2008; Zeng and Chi, 2017). These previous results are therefore consistent with our current findings that the functions of mTORC1 promote effector programming. However, our results now reveal that although such roles of mTORC1 promote the induction of Hopxhi effector precursors, mTORC1 is not essential for a terminal differentiation of these precursors. Therefore, by elucidating a two-step process of fate determination, our findings also help to clarify crucial roles of mTORC1 in governing priming of effectors under conditions that are relevant for autoimmune and possibly some other immune responses. Further, by dissociating the processes of the initial T cell fate commitment and subsequent terminal differentiation, our current findings also resolve an outstanding issue of how mTORC1 maintains the balance between the effector and regulatory T cell differentiation. Previous observations suggested an inhibition of Foxp3 expression by mTOR, but the corresponding mechanisms have remained unclear because other results disconnected the direct functions of mTOR and an inhibition of Foxp3 expression in T cells (Harada et al., 2010; Ouyang et al., 2010). By elucidating that mTORC1 specifically reduces the number of T cells that can potentially convert into pTreg cells, our results now explain how mTORC1 limits pTreg cell conversion without directly inhibiting Foxp3 expression in T cells. Future investigation may further clarify the detailed molecular mechanisms of mTORC1 relevant for these processes.

Overall, our results help to clarify the functions of cDCs initiating effector priming under minimal perturbations of homeostasis. We propose that current findings are in agreement with the emerging concepts of “homeostatic maturation” of cDCs and also build on the roles of DCs originally envisioned by Ralph Steinman to function as “Nature’s adjuvant” inherently able to initiate certain immune and, particularly, autoimmune responses (Ardouin et al., 2016; Baratin et al., 2015; Jiang et al., 2007; Steinman, 1996, 2007; Vander Lugt et al., 2017). However, the capacity of such cDCs to initiate effector T cell differentiation in the steady state has remained unclear in light of the seminal functions of the pathways triggered by pro-inflammatory signals that crucially promote effector responses (Iwasaki and Medzhitov, 2015; Merad et al., 2013; Pulendran, 2015; Sancho and Reis e Sousa, 2012; Segura and Amigorena, 2013; Yatim et al., 2017; Zelenay and Reis e Sousa, 2013). Based on our current findings, we propose that the effector programming in the steady state can trigger certain immune responses, including those directed against self, that may possibly be then further advanced by the pro-immunogenic mechanisms mediated by the recognition of specific molecular patterns and other signals. We also would like to speculate that in accord with the key roles of CD4+ T cells to promote CD8+ T cell responses during immunosurveillance (Alspach et al., 2019; Ferris et al., 2020), the programming of effectors under minimal perturbations of homeostasis might help to initiate T cell responses against tumors. Eventually, the regulation of the two-step process of effector programming and differentiation might be harnessed for future immunotherapies skewing the balance between immunity and tolerance.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Daniel Hawiger (daniel.hawiger@health.slu.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

The accession numbers for the RNA-seq and ChIP-seq data reported in this paper are GEO: GSE120277, GSE141724.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Mice

All mouse strains were bred on C57BL/6 background for at least 12 generations. 6–9 week old sex- and age-matched male or female littermates were used for experiments. Littermates of the same sex were randomly assigned to experimental groups. 2D2 TCR tg mice (Bettelli et al., 2003) or OTII TCR tg (Barnden et al., 1998) were crossed onto Hopx−/− (Shin et al., 2002) or Hopx (Takeda et al., 2013) and Foxp3 (Wan and Flavell, 2005) reporter mice to derive 2D2 Hopx−/−Foxp3, 2D2 Hopx Foxp3 or OTII Hopx Foxp3 mice respectively. Additionally, mice were crossed with Rptor (Peterson et al., 2011) and CD4-cre (Lee et al., 2001) mice to generate 2D2 Rptor CD4-cre Hopx Foxp3 mice. TCR tg were maintained by crossing with non-carriers within the individual specific genotypes. Btla−/− (Sedy et al., 2005), Rag1−/− (Mombaerts et al., 1992) and C57BL/6 congenically mismatched recipient mice for adoptive transfer experiments were all purchased from Jackson Laboratories. All mice were maintained in our facility under specific pathogen free conditions and used in accordance with the guidelines of the Saint Louis University Institutional Animal Care and Use Committee.

METHOD DETAILS

In vivo targeting of antigens to DC

Antigen-specific T cells were activated in vivo by cognate antigens (MOG35–55 or OVA323–339) that were delivered by chimeric antibodies specific to DEC-205, CD11c, or DCIR2 (Dudziak et al., 2007; Hawiger et al., 2004, 2010; Jones et al., 2016). Antibodies were produced and mice were specifically treated with corresponding MOG or OVA delivering chimeric antibodies as previously established (Hawiger et al., 2004, 2010) and further described in (Jones et al., 2016). Briefly, antibodies were expressed in A293 cells (ATCC) by transient transfection of plasmids using calcium-phosphate. Cells were grown in serum-free DMEM supplemented with Nutridoma SP (Roche) and antibodies were purified on protein-G Sepharose beads (GE Healthcare). Chimeric antibodies were injected in PBS intraperitoneally at 12.5 μg/mouse (MOG-delivering antibodies) or 250 ng/mouse (OVA-delivering antibodies) as established previously (Hawiger et al., 2004, 2010; Jones et al., 2016).

Experimental autoimmune encephalomyelitis (EAE)

Rag1− recipients of 2D2 T cells were injected intraperitoneally with 200 ng per mouse of Pertussis toxin (List Biological Laboratories Inc.) reconstituted in PBS on days 1 and 3 after T cell transfer. Clinical disease score of EAE was graded on a scale of 0–4: 0 – no clinical signs; 1 – flaccid tail; 2 – hind limb weakness; abnormal gait; 3 – complete hind limb paralysis; 4 – complete hind limb paralysis and forelimb weakness or paralysis. Mice were scored daily in a blinded fashion.

In vitro T cell stimulation

Treg cell induction in vitro was performed by culturing 0.2 × 106 sorted T cells/well in 96 well plates pre-coated with anti-CD3 (145–2C11) (1 μg/mL) using Click’s media (Millipore-Sigma) containing 10% fetal bovine serum (FBS) (Gemini); penicillin-streptomycin, L-glutamine, HEPES, β-mercaptoethanol, sodium pyruvate (all from GIBCO); recombinant mouse IL-2 (200 units/ml), human TGF-β1 (0.25 ng/ml, or as indicated in figure legends of individual figures), and anti-CD28 (37.51) (1.5 μg/mL) (Biolegend).

Th0 cultures were prepared by culturing 0.2 × 106 sorted T cells/well with 0.1 × 106 antigen presenting cells (APCs) in the presence of anti-CD3 (145–2C11) (1 μg/ml) (Biolegend) using Click’s media (Millipore-Sigma) containing 10% FBS (Gemini) and penicillin-streptomycin, L-glutamine, β-mercaptoethanol, HEPES, and sodium pyruvate (all from GIBCO) for 1–3 days. The APCs used in these cultures were isolated from pooled lymph node and spleen cells depleted of T cells (CD4+, CD8+, CD3+) and CD49b+ NK cells using biotinylated antibodies (Biolegend) and magnetic microbead selection (Miltenyi).

Flow Cytometry

Lymphocytes from peripheral lymph nodes and spleens were isolated and analyzed separately unless otherwise indicated. Similar results were obtained from spleens and peripheral lymph nodes from corresponding experimental groups. For surface marker staining, cells were first analyzed with Zombie Aqua Live/Dead viability dye according to manufacturer’s protocol (BioLegend), pre-incubated with Fc-block (anti-CD16/32, clone 2.4G2, produced in-house from corresponding hybridomas obtained from ATCC), and then stained with fluorochrome-conjugated antibodies for 25 minutes at 4°C. To conduct intracellular staining for transcription factors, cell populations as described in the manuscript were sorted then fixed and permeabilized according to manufacturer’s protocol (eBioscience). Alternatively, sorted T cell populations were cultured in vitro as described in the manuscript and then processed as above. Additionally, ex vivo isolated T cells obtained as described in the manuscript from individual groups of Rag1− recipients were fixed and stained for transcription factors as above. To conduct intracellular staining for cytokines, ex vivo isolated T cells obtained as described in the manuscript from individual groups of Rag1− recipients were stimulated for approximately 4 hours with phorbol 12-myristate 13-acetate (PMA) (Millipore-Sigma) (100ng/ml), and Ionomycin (Millipore-Sigma) (0.25 μM). BD GolgiStop Protein Transport Inhibitor (containing monensin) was added for the last 2 hours, and cells were fixed and permeabilized with Cytofix-Cytoperm buffers according to manufacturer’s protocol (BD). Flow cytometry acquisition was performed using BD LSR II or BD LSRFortessa instruments, and data was analyzed with FlowJo software (FlowJo, LLC).

Antibodies used for flow cytometry and in vitro culture experiments

Antibodies and reagents used for experiments include anti-CD11c (N418), anti-CD11b (M1/70), anti-CD8a (53–6.7), anti-B220 (RA3–6B2), anti-CD49B (DX5), anti-CD25 (PL61), Streptavidin (405232), anti-CD4 (GK1.5), anti-CD45.2 (104), anti-CD62L (Mel-14), Ly-6C (AL-21), anti-IFN-γ (XMG1.2), anti-CD3 (145–2C11), Rat IgG1 Isotype control (RTK2071), anti-CD28 (37.51), anti-CD5 (53–7.3), anti-VISTA (MIH63), anti-T-bet (4B10), (BioLegend), anti-CD69 (H1.2F3), anti-RORgt (q31–378), anti-TCR V beta 11 (RR3–15), anti-Nur77 (12.14), (BD), anti-TCR V alpha 3.2 (RR3–16), anti-Foxp3 (FJK-16 s), Fc-block (anti-CD16/32) (2.4G2), (ATCC).

Cytokine measurement

Cytokine concentrations were measured using a BD Cytometric Bead Array (CBA) mouse cytokine kit according to manufacturer’s protocols.

IFN-γ blocking

IFN-γ was blocked in vitro by using 10 ug/ml of anti-IFN-γ (XMG1.2) (BioXcell) or purified Rat IgG1, k Isotype Control Antibody (RTK2071) (Biolegend).

T cell isolation and sorting

Lymph nodes and spleens were pooled, and CD4+ T cells were enriched by depletion. Non-CD4 T cells were stained with anti-CD8a (53–6.7), B220 (RA3–6B2), CD11c (N418), CD11b (M1/70), and CD49b(DX5) biotinylated antibodies (Biolegend) followed by magnetic microbead selection (Miltenyi). Cells were sorted on FACSAria III or FACSAria Fusion instruments (BD). Th0 cultures were harvested after 12 hours and stained with fluorochrome-conjugated anti-CD4 and Zobmie Auqua Live/Dead viability dye and sorted as described above.

In vivo cell proliferation assay

T cell proliferation in vivo was assessed by using eBioscience™ Cell Proliferation Dye eFluor™ 450 according to manufacturer’s protocol. Briefly, sorted T cells were washed in PBS and stained in 10 μM dye in PBS at concentration 10 × 107 cells/mL at room temperature for 20 minutes. Next, the staining was quenched by adding FBS, and the cells were washed and transferred to recipient animals.

T cell adoptive transfers

Sorted cells were washed twice with PBS and counted. Cells were then reconstituted in PBS and injected into tail veins. 5 × 106 T cells were transferred into C57BL/6 recipients, and 3 × 106 T cells were transferred into Rag1− recipients.

PD-L1 blocking

αPD-L1 (10F.9G2) (150 μg/mouse) antibody (BioXCell) was injected in PBS i.p. 4 and 26 hours after chimeric antibodies were administered

RNA library construction and sequencing

Total mRNA and miRNA were isolated by using TRIzol reagent (Invitrogen) and mirVana miRNA isolation Kit (Invitrogen) to isolate both large and small RNA from Hopxhi, Hopxlo, and naive 2D2 CD4+ T cells obtained as described in (Figure S1F). In some experiments individual T cell groups were additionally re-stimulated in vitro and CD4+ T cells were then re-sorted. The material for every replicate within each of the Hopxhi, Hopxlo, and naive T cell groups was obtained from an independent experiment. Within each such an independent experiment Hopxhi and Hopxlo T cells were pooled from multiple (5–6) mice 5 days after treatment with αDEC-MOG; the naive T cells were pooled from multiple (4–5) mice that were treated with PBS. Ribosomal RNA was depleted from total RNA using the Eukaryotic RiboMinus Core Module v2 (Life Technologies, ThermoFisher), and libraries were constructed using the Ion Total RNA-seq v2 kit (Life Technologies, ThermoFisher) according to the manufacturer’s protocol. Sequencing was performed on an Ion Torrent Proton with a mean read length of ~140 nucleotides, and reads were aligned to the mm10 mouse genome assembly using the TMAP (Torrent Mapping Program) aligner (Homer, 2011) map4 algorithm. Reads required a minimum seed length of 20 nucleotides, and soft-clipping at both 5′ and 3′ ends was permitted to accommodate spliced reads. Analysis methods that count reads operate under the assumption that reads in each sample are of the same length. Unlike some other platforms (for example Illumina), Ion Torrent Proton platform generates reads which are variable in length. To help ensure that differences in read lengths do not impact the quantification of the expression of each gene, the number of nucleotides within each read that aligns to gene’s exons were counted using the genomecoveragebed program included in the freely available BEDtools (Quinlan and Hall, 2010). As most recently described in (Basta et al., 2020; Kreienkamp et al., 2018), total exon coverage was calculated for each sample and divided by the average total exon coverage across all replicates to obtain normalization factors. The summed coverage for each gene in each replicate was then multiplied by these normalization factors to obtain expression values for each gene as total normalized nucleotide coverage.

Small RNA-seq libraries were constructed using the Ion Total RNA-seq v2 kit (Life Technologies, Thermofisher) according to the manufacturer’s protocol for small RNA libraries. Sequencing was performed on an Ion Torrent Proton with a mean read length of ~55 nucleotides, and reads were aligned to the Mus musculus (house mouse) genome assembly GRCm38 (mm10) from Genome Reference Consortium using the TMAP aligner (Homer, 2011) as described above. The nucleotide coverage for miRNAs was calculated and normalized to total coverage as above. Expression values are given as total normalized nucleotide coverage per miRNA.

Analysis of gene expression

Expression values were identified for 23219 genes in three individual replicates for each experimental group (Hopxhi, Hopxlo, and naive T cells). Within each group, the mean expression values were calculated for every gene that had at least one replicate with the expression value above 5. Expression values below 5 were considered as “undetected” and excluded from averaging. If all three replicates were undetected, then the mean expression value of that gene within the given group was set as equal to 1. The genes specifically expressed (expression values greater than 103 in all replicates) in both groups within each comparison were analyzed by t test as described in “Quantification and statistical analysis” below. The genes whose mean expression values differed between two compared groups by at least 2-fold and that were either specifically expressed in both groups within each comparison with a p value less than 0.05 or that were classified as specifically expressed in one group, but not in the other group within each comparison were then used for subsequent Gene Set Enrichment Analysis (GSEA) as well as for additional literature-based analysis. As required for GSEA analysis, the official mouse gene symbols were converted first to the ENSEMBL format and then to HUGO format using the BioMart tool (http://useast.ensembl.org/biomart/martview/c0a53d4785d930f83bf91ca0b07f7f18) in conjunction with the ENSEMBL_mouse_gene.chip dictionary available in the GSEA web page. In rare cases when a gene was absent in these dictionaries, its official mouse gene symbol was converted to uppercase and used directly as a HUGO format unless multiple alternative translation variants were present. After the name conversion, 1525 genes were pooled in the comparison between Hopxhi and naive T cells, 813 genes were pooled in the comparison between Hopxlo and naive T cells, and 372 genes were pooled in the comparison between Hopxhi and Hopxlo T cells. Each of those three gene lists was submitted independently to GSEA Preranked with the following metric: GSEA were performed using a customized database of published gene sets containing the genes against which we tested. This specific database was constructed from the c7.all.v6.1 database (https://www.gsea-msigdb.org/gsea/downloads.jsp) by selecting those gene sets whose titles contained at least one of the “positive” keywords (CD4, TREG, TEFF, _TH_, TFH, ROR, FOXP3, T-BET, GATA3) and none of the “negative” keywords (CD8, DN_, DP, DC, DENDRITIC, MACROPHAGE, BCELL, MAST, THYMUS, MEDULLA, CORTEX, MYELOID, KO). The resulting database contained 994 gene sets and is denoted herein as “CD4+ T cell-related gene sets.”

For RNA-seq analysis of T cells following their re-stimulation in vitro, expression values were identified for 3 replicates of each Hopxhi T cells, Hopxlo T cells, and naive T cells after excluding snoRNA exons from all samples. The expression values less than 10^(1.5) ≈31.6 were considered undetected and excluded from averaging as described above. Genes were then analyzed as described in the section above. Additionally, the analysis included genes that had expression values higher than 150 in all three replicates of Hopxhi T cells, but whose expression was undetected in all replicates of Hopxlo and naive T cells. All identified genes whose expression had not been significantly different between Hopxhi and Hopxlo T cells before the in vitro re-stimulation, were selected and then further analyzed based on the literature search.

Analysis of miRNA expression

Expression values were identified for 2220 miRNAs in three individual replicates for each experimental group (Hopxhi, Hopxlo, and naive T cells). Within each group, the mean expression values were calculated for every miRNA that had at least one replicate with the expression value above 5. Expression values below 5 were considered as “undetected” and excluded from averaging. If all three replicates were undetected, then the mean expression value of that miRNA within the given group was set as equal to 1. The miRNAs specifically expressed (expression values greater than 200 in all replicates) were additionally tested by t test, as described in “Quantification and statistical analysis” below, followed by the positive false discovery rate correction (Storey and Tibshirani, 2003). The miRNAs whose mean expression values differed between Hopxhi and Hopxlo T cells by at least 2-fold and that were either specifically expressed in both groups with a confidence value of at least 0.05, or that were classified as specifically expressed in only one group, but not in the other group were identified. In total, 48 miRNAs were identified and for normalization their expression values were divided by the mean expression values of the corresponding miRNAs in the naive group.

Chromatin Isolation

After sorting, cells were adjusted to 2×106 cells/ml in 3%–5% FCS in PBS. Cells were fixed with 1% formaldehyde for 12 minutes at room temperature. Fixation was stopped by adding glycine to a final concentration of 0.25M. Cells were collected by centrifugation at 1000 g for 5 minutes, washed twice with cold PBS, and frozen as a pellet on dry ice and stored at −80°C. Chromatin was prepared as previously described (Dorsett et al., 2014; Swain et al., 2016) with slight modifications. To isolate chromatin, 10×106 cells were suspended in 300 mL sonication buffer [0.4% SDS, 10mM Tris pH 7.6] with protease inhibitors (Roche cOmplete, EDTA-free Protease Inhibitor Cocktail) and incubated on ice for 20 minutes. The cell suspension was sonicated, and chromatin was sheared to a range of 200–1000bp using a Diagenode Biorupter. After sonication, 900 μL of adjustment buffer [10mM Tris pH 7.6, 1.33 mM EDTA pH 8, 0.133% Na deoxycholate, 1.33% Triton] plus protease inhibitors was added and chromatin was incubated on ice for an additional 20 minutes. Chromatin was clarified by centrifugation at 16,000 g for 10 minutes at 4°C. The supernatant was collected and glycerol was added to a final concentration of 5% prior to snap freezing on dry ice and storing at −80°C.

ChIP-seq

Multiple independent biological replicates were performed for each ChIP-seq experiment. All ChIP antibodies were purchased from Cell Signaling Technology: H3K4me3 (Rabbit mAb9751), H3K27Ac (Rabbit mAb8173). To perform ChIP, 600 mL of chromatin (~5×106 cells) was used. Chromatin was precleared by incubating with 15ul Protein A Dynabeads (Invitrogen) that were pre-washed with RIPA Buffer [10mM Tris pH 7.6, 1mM EDTA pH 8, 0.1% SDS, 0.1% Na deoxycholate, 1% Triton] for 2 hours at 4°C. Beads were discarded and 5ul of antibody was incubated with the chromatin overnight at 4°C. The chromatin-antibody mix was then incubated with 20ul of Protein A Dynabeads (prewashed with 0.5% BSA in PBS) at 4°C for 4 hours. Beads were washed 5 times for 5 minutes each at room temperature with cold wash buffer [50mM HEPES pH 7.5, 1mM EDTA, 1% NP-40, 0.7% Na Deoxycholate, 500mM LiCl] and 4 times with cold TE. Bound chromatin was eluted by suspending beads in 200 μL Elution Buffer [50mM Tris HCl pH 8, 10mM EDTA, 1% SDS] and incubating at 65°C for 40 minutes with mixing. After discarding the beads, 200 μL of 0.2 μg/μl RNase A (QIAGEN, 158924) in TE was added to the supernatant and incubated for 1 hour at 37°C. To reverse the crosslinks, Proteinase K (Roche, 03115828001) was added to a final concentration of 0.5mg/mL and incubated at 65°C overnight (up to 15 hours). Immunoprecipitated DNA was purified using MinElute columns (QIAGEN) and the concentration measured using a Qubit Fluorometer. A nanogram of the DNA was sheared to an average size of 150 to 200 nt by sonication and used to make ChIP-seq libraries using the Ion Plus Fragment Library Kit (Life Technologies) without enzymatic shearing.

ChIP-seq processing

ChIP-seq libraries were sequenced on an Ion Torrent Proton to approximately 2X genome coverage and input chromatin libraries were sequenced to approximately 8x genome coverage. Reads were aligned to the mm10 mouse genome assembly using the TMAP (Homer, 2011) aligner map4 algorithm without soft-clipping. Reads flagged as PCR duplicates were removed using Samtools (Li et al., 2009). Genome coverage was calculated for ChIP and input samples using BedTools (Quinlan and Hall, 2010). ChIP sample coverage was normalized to input coverage in R (R Development Core Team, 2015) using a previously described script (Dorsett and Misulovin, 2017) to generate ChIP enrichment values every 50 bp in sgr files. Prior to averaging independent biological replicates, they were assessed for reproducibility both visually in IGB (Freese et al., 2016) and by genome-wide Pearson correlation coefficients. To allow for quantitative comparisons between groups, average enrichment values from naive, Hopxhi and Hopxlo groups for each antibody were quantile normalized to each other using the normalize.quantiles command in R, from the preprocessCore package. Quantile normalized average enrichment values were used for all downstream analysis.

QUANTIFICATION AND STATISTICAL ANALYSIS

Mice of particular genotypes were randomly assigned to individual experimental groups. All experimental groups and individual mice were included in statistical analysis. The data was collected from 2–5 independent experiments of each kind. P values were calculated using unpaired two-tailed t test (Welch’s unequal variances t test), paired two-tailed t test (Student’s t test) (for comparisons of gene expression in Hopxhi and Hopxlo groups whose corresponding replicates were both obtained from the same cohorts of mice) and two-way ANOVA. Additional specific statistical analysis and corrections are listed in respective figure legends and elsewhere in methods. Statistical analysis was performed using MATLAB (The MathWorks) and GraphPad (Prism). No new methods were used to determine whether the data met assumptions of the statistical approach.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
anti-CD11C	BioLegend	N418, RRID:AB_313773
anti-CD11b	BioLegend	M1/70 RRID: AB_312787
anti-CD8a	BioLegend	53–6.7 RRID: AB_312743
anti-B220	BioLegend	RA3–6B2 RRID: AB_312989
anti-CD49B	BioLegend	DX5 RRID: AB_313411
anti-CD25	BioLegend	PL61 RRID: AB_893288
anti-CD4	BioLegend	GK1.5 RRID: AB_312689 RRID: AB_312707
anti-CD45.2	BioLegend	104 RRID: AB_389211
anti-CD62L	BioLegend	Mel-14 RRID: AB_493719
anti-CD69	BD	H1.2F3
anti-IFNγ	BioLegend	XMG1.2 RRID: AB_315396 RRID: AB_315404
anti-CD3	BioLegend	145–2C11 RRID: AB_312669 RRID: AB_1877073
anti-Foxp3	eBioscience	FJK-16s RRID: AB_467575
anti-CD28	BioLegend	37.51 RRID: AB_312877
Fc-block (anti-CD16/32)	ATCC	2.4G2 (HB-197)
Streptavidin	BioLegend	N/A
anti-T-bet	BioLegend	4B10 RRID: AB_1595466
anti-RORγt	BD	q31–378 RRID: AB_2651150
anti-Ly-6C	BD	AL-21 RRID: AB_2737949
anti-CD5	BioLegend	53–7.3 RRID: AB_312733
anti-VISTA	BioLegend	MIH63 RRID: AB_2728190
anti-TCR V alpha 3.2	eBioscience	RR3–16
anti-TCR V beta 11	BD	RR3–15 RRID: AB_394704
anti-Nur77	BD	12.14 RRID: AB_395232
anti-PD-L1	BioXCell	10F.9G2 RRID: AB_10949073
Purified Rat IgG1, kappa isotype Ctrl antibody	BioLegend	RTK2071 RRID: AB_326508
Anti-Tri-Methyl-Histone H3 (Lys4) (C42D8)	Cell Signaling Technology	Rabbit mAb9751
Anti-Acetyl-Histone H3 (Lys27) (D5E4) XP®	Cell Signaling Technology	Rabbit mAb8173
Experimental Models: Organisms/Strains
Mouse: C57BL/6J	Jackson Laboratory	Stock# 000664
Mouse: B6.SJL-Ptprca Pepcb/BoyJ	Jackson Laboratory	Stock# 002014
Mouse: Foxp3RFP reporter	Jackson Laboratory	(Wan and Flavell, 2005)
Mouse: HopxGFP reporter	available at Jackson Laboratory on a mixed background	(Takeda et al., 2013)
Mouse: Hopx−/−	available at Jackson Laboratory on a mixed background	(Shin et al., 2002)
Mouse: 2D2 TCR tg	Jackson Laboratory	(Bettelli et al., 2003)
Mouse: OTII TCR tg	Jackson Laboratory	(Barnden et al., 1998)
Mouse: Rptorfl/fl	Jackson Laboratory	(Peterson et al., 2011)
Mouse: CD4-cre	Jackson Laboratory	(Lee et al., 2001)
Mouse: Rag1−/−	Jackson Laboratory	(Mombaerts et al., 1992)
Cell line: A293	ATCC	N/A
Chemicals, Peptides and Recombinant Proteins
Nutridoma SP	Millipore-Sigma	11011375001
Protein-G sepharose beads	GE Healthcare	17061801
Zombie Aqua Live/Dead viability dye	BioLegend	423102
Phorbol 12-myristate 13-acetate	Millipore-Sigma	P8139
Ionomycin	Millipore-Sigma	I0634
Pertussis toxin	List Biological Laboratories Inc.	180
Penicillin-streptomycin	Gibco	15140-122
Sodium Pyruvate	Gibco	11360-070
L-glutamine	Gibco	25030-081
HEPES	Gibco	15630-080
β-mercaptoethanol	Gibco	21985-023
Recombinant mouse IL-2	Biolegend	575406
Recombinant human TGF-β1	Biolegend	580704
TRIzol reagent	Invitrogen	15596026
Critical Commercial Assays
Transcription Factor Staining Buffer Set	Invitrogen	00-5523-00
Fixation/Permeabilization Solution Kit	BD	554715
Mouse Th1/Th2/Th17 Cytokine Kit	BD Biosciences	560485
Streptavidin magnetic microbead	Miltenyi	130-048-101
mirVana miRNA isolation Kit	Invitrogen	AM1560
Eukaryotic RiboMinus Core Module v2	Life Technologies	N/A
Cell Proliferation Dye eFluor™ 450	Invitrogen	65-0842
Ion Total RNA-seq v2 kit	Life Technologies	N/A
cOmplete™, EDTA-free Protease Inhibitor Cocktail	Roche	N/A
Software and Algorithms
TMAP (Torrent Mapping Program) aligner map4 algorithm	N/A	https://github.com/iontorrent/TMAP/blob/master/doc/tmap-book.pdf
R statistical computing	R Software	www.r-project.org
MATLAB®	The MathWorks	https://www.mathworks.com/downloads/
Gene Set Enrichment Analysis	Broad Institute, Inc.	ImmuneSigDB
BioMart	ENSEMBL	http://useast.ensembl.org
GraphPad Prism	Graphpad Software	https://www.graphpad.com
GSEA	UC San Diego, Broad Institute	https://www.gsea-msigdb.org/gsea/downloads.jsp
Bedtools	Quinlan and Hall, 2010	http://bedtools.readthedocs.io/en/latest/
Flowjo 9	FLOWJO, LLC	www.flowjo.com
Deposited Data
RNA-seq and ChIP-seq data	This paper	NCBI GEO: GSE120277 and GSE141724