Id2 reinforces TH1 differentiation and inhibits E2A to repress TFH differentiation.

The differentiation of helper T cells into effector subsets is critical to host protection. Transcription factors of the E-protein and Id families are important arbiters of T cell development, but their role in the differentiation of the TH1 and TFH subsets of helper T cells is not well understood. Here, TH1 cells showed more robust Id2 expression than that of TFH cells, and depletion of Id2 via RNA-mediated interference increased the frequency of TFH cells. Furthermore, TH1 differentiation was blocked by Id2 deficiency, which led to E-protein-dependent accumulation of effector cells with mixed characteristics during viral infection and severely impaired the generation of TH1 cells following infection with Toxoplasma gondii. The TFH cell-defining transcriptional repressor Bcl6 bound the Id2 locus, which provides a mechanism for the bimodal Id2 expression and reciprocal development of TH1 cells and TFH cells.

Introduction

Recognition of a pathogen by the immune system initiates a multi-step transcriptional program which directs CD4+ T cell differentiation into distinct T helper populations (TH) that coordinate eradication of infection. TH1 effector cells secrete inflammatory cytokines and activate immune cells[1]. Follicular helper (TFH) cells secrete cytokines and upregulate ligands that induce B cells to form germinal centers, class switch, and generate high-affinity antibodies[2]. Differentiation of CD4+ T cells is directed by cytokine-induced activation of STAT proteins and lineage determining transcription factors such as T-bet and Bcl6[3]. Following activation, TH1 cells receive signals that initiate T-bet expression and induce migration from the lymphoid tissues to infected or inflamed areas of the body[1]. In contrast, to properly differentiate, TFH cells must upregulate Bcl6 and the chemokine receptor CXCR5 allowing movement from the T cell zone into the B cell follicle[2]. The differentiation of TH1 and TFH cells is interconnected through antagonistic interplay between T-bet and Bcl6, and Bcl6 and Blimp-1[4-8].

E protein transcription factors and their natural repressors, the inhibitor of DNA binding (Id) proteins, play a crucial role in the differentiation of numerous lymphocyte populations such as B cells, innate lymphoid cells, natural killer cells, invariant NKT cells, αβ γδ and CD8+ effector and memory T cells[9-17]. Recently, studies have highlighted the role played by Id2, Id3 and E proteins in mature CD4+ T cells, particularly in the differentiation and maintenance of regulatory T (Treg) cells and TH17 cells[18-21]. Deletion of E proteins leads to an increase in Treg cell populations; however, deletion of Id2 and Id3 cripples the differentiation and localization of Foxp3+ Treg cells[18,20]. Additionally, Id2-deficient CD4+ T cells were unable to mount a robust TH17 response in a mouse model of experimental autoimmune encephalomyelitis[21]. Ectopically expressed basic helix-loop-helix (bHLH) transcription factor achaete-scute homolog 2 (Ascl2) binds E-box sites to drive upregulation of CXCR5 in vitro, resulting in augmented accumulation of CD4+ T cells in the B cell follicle in vivo[22]. However, Ascl2 did not induce Bcl6, raising the question of how E protein activity and the induction of CXCR5 and Bcl6 are interrelated. Furthermore, our own studies have shown differential Id2 and Id3 mRNA expression in TH1 and TFH cells[23]. Thus, we further explored the biology of Id2 and Id3 in the differentiation of TH1 and TFH during infection.

Results

Id2 and Id3 define TH1 and TFH effector CD4+ T cell subsets

We determined the abundance of Id2 and Id3 in CD4+ T cell subsets by using reporter mice in which cDNA encoding for yellow fluorescent protein (YFP) or green fluorescent protein (GFP) were inserted in the Id2[14] or Id3[24] gene, respectively. We crossed both lines to SMARTA TCR transgenic mice (LCMV-specific, gp66-77 IAb-restricted) to generate Id2YFP/+ and Id3GFP/+ SMARTA CD4+ T cells, which were transferred into B6 hosts that were then infected with LCMV-Armstrong. TH1 and TFH differentiation was assessed in Id2-YFPlo and Id2-YFPhi subsets following infection. In parallel, we infected Id2YFP/+ or Id3GFP/+ mice with LCMV to follow polyclonal CD4+ T cell differentiation. We observed that Id2-YFPlo cells were almost exclusively TFH cells (CXCR5+SLAMlo or CXCR5+PD-1lo) and GC TFH (CXCR5+PD-1+), while the vast majority of Id2-YFPhi cells displayed a TH1 phenotype (SLAM+CXCR5− or CXCR5−PD-1−) (Fig. 1a, Supplementary Fig. 1a). Similarly, we found that Id3 expression was highly polarized: Id3-GFPlo cells enriched in the TH1 population and Id3-GFPhi cells enriched in TFH and GC TFH populations (Fig. 1b, Supplementary Fig. 1b).

Figure 1

Differential Id2 and Id3 expression defines TH1 and TFH cell subsets

Id2YFP/+ (a) or Id3GFP/+ (b) SMARTA CD4+ T cells were transferred into B6 mice and analyzed 7 days after LCMV infection. TH1 (SLAM+CXCR5− or CXCR5−PD-1−), TFH (SLAMloCXCR5+ or CXCR5+PD-1−) or GC TFH (CXCR5+PD-1+) cell development in the indicated SMARTA CD4+ T cell populations was analyzed by flow cytometry and quantified. (c) Id2YFP/+Id3GFP/+ SMARTA CD4+ T cells were transferred into B6 mice and analyzed 0, 4 or 7 days after LCMV infection. Flow cytometric analysis of Id2-YFP and Id3-GFP expression (bottom plots) in SLAM+CXCR5− (black lines) or CXCR5+SLAMlo (dashed lines) populations (top plots). (d) Id2YFP/+ reporter mice were immunized subcutaneously with phycoerythrin (PE) emulsified in TiterMax Gold Adjuvant and analyzed by histology 13 days later. Draining lymph node sections were stained with IgD (blue) and CD4 (red). Id2-YFP reporter expression is shown in green. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 (two-tailed unpaired Student’s t test). Data are representative of three experiments (a–c), each with n = 3 mice per group, or are representative of two experiments (d), each with n = 2 mice per group (mean ± s.e.m.).

We also generated Id2YFP/+Id3GFP/+ dual reporter mice and analyzed coordinate Id2 and Id3 expression after LCMV infection. Naive CD4+ T cells expressed intermediate and high amounts of Id2 and Id3, respectively. After infection, TH1 cells showed a substantial increase in Id2-YFP expression compared to naive cells (Fig. 1c). In contrast, TH1 cells lost Id3-GFP expression, and TFH cells maintained high expression of Id3-GFP (Fig. 1c). Compared to TFH cells, GC TFH cells had equivalent expression of Id3-GFP reporter (data not shown). Histology revealed that the majority of Id2-YFP expressing CD4+ T cells were excluded from the B cell follicle and germinal center (Fig. 1d). Our results show that Id2 and Id3 have contrasting expression patterns in TH1 and TFH cells following acute LCMV infection.

Impaired Id2 expression enhances TFH differentiation

To determine if differential Id2 expression in CD4+ T cells influences CD4+ T cell differentiation in vivo, we expressed an shRNAmir specific for Id2 (shId2) or a control shRNAmir (shCtrl) in SMARTA CD4+ T cells, transferred cells into B6 mice, and analyzed T cell differentiation after LCMV infection. Expression of shId2 in SMARTA CD4+ T cells reduced Id2 mRNA expression (Supplementary Fig. 2a). Among shId2+ cells, there was an increased frequency of TFH cells and decreased frequency of TH1 cells compared to shCtrl (Fig. 2a–c). The bias was predominantly attributed to GC TFH cells, identified as CXCR5+Bcl6+ cells (Fig. 2d–f) or CXCR5+PSGL-1lo (Supplementary Fig. 2c–e). We then examined the differentiation of shId2+ cells earlier after infection (Supplementary Fig. 2f) and observed an increased frequency of CXCR5+Bcl6+ TFH and a decreased TH1 (CXCR5− Bcl6−) population among shId2+ cells (Fig. 2g–i). Analysis of early TFH cells (CXCR5+CD25−)[25,26] also revealed an increased proportion among shId2+ cells (Supplementary Fig. 2g–i). Thus, impaired Id2 expression appears to favor TFH differentiation.

Figure 2

Id2 knockdown results in increased TFH differentiation

SMARTA CD4+ T cells transduced with the indicated shRNAmir-RV were transferred into B6 mice and analyzed 6 (a–f) or 3 days (g–i) after LCMV infection. (a–c) TFH (CXCR5+SLAMlo) and TH1 (SLAM+CXCR5−) differentiation was analyzed by flow cytometry (a) and quantified as a fraction of SMARTA CD4+ T cells (b) or total splenocytes (c). (d–f) GC TFH (CXCR5+Bcl6+) differentiation was analyzed by flow cytometry (d) and quantified as a fraction of SMARTA CD4+ T cells (e) or total splenocytes (f). (g–i) TFH (CXCR5+Bcl6+) and TH1 (CXCR5−Bcl6−) cell development was analyzed by flow cytometry (g) and quantified as a fraction of SMARTA CD4+ T cells (h) or total splenocytes (i). (j–m) OT-II CD4+ T cells transduced with the indicated shRNAmir-RV were transferred into Bcl6fl/fl CD4-Cre+ mice and analyzed 11 days after footpad immunization with NP-OVA in alum. (j) Quantitation of OT-II CD4+ T cells. GC B cells (Fas+PNA+) (k) and plasma cells (CD138+IgD−) (l) were analyzed by flow cytometry and quantified as a fraction of B cells. (m) NP-specific IgG titers. The dotted line indicates NP-specific titers in naive mice. *p<0.05, **p<0.001, ***p<0.0001 (two-tailed unpaired Student’s t test). Data are pooled from four (a–f) five (g–i) or two (j–m) independent experiments with n=6-14 mice per group (mean ± s.e.m.).

To examine B cell help by shId2+ CD4+ T cells we transferred shId2+ or shCtrl+ OT-II CD4+ cells into Bcl6fl/fl CD4-Cre+ mice[27], followed by immunization with NP-OVA in alum. After immunization, more shId2+ T cells had differentiated into CXCR5+PD-1+ or CXCR5+Bcl6+ GC TFH than control cells (Supplementary Fig. 2j,k). Increased frequencies of GC B cells (Fig. 2k) and plasma cells (Fig. 2l) were observed in shId2+ recipients compared to control, with no differences in the NP-specific IgG titers (Fig. 2m), despite slightly reduced accumulation of shId2+ CD4+ T cells (Fig. 2j). Thus, Id2 expression appears to negatively impact TFH cell differentiation.

Id2 is necessary for generation of TH1 cells during infection

We next asked how the total absence of Id2 affects CD4+ T cell differentiation. We crossed Id2-floxed (Id2fl/fl) mice[28] to SMARTA TCR- and CD4-Cre transgenic lines to generate Id2fl/flCD4-Cre+ SMARTA (Id2−/−) mice in which Id2 is deleted in αβ thymocytes. Naive Id2−/− SMARTA cells were CD44low and indistinguishable from Id2+/+ cells (data not shown). We transferred naive Id2+/+ CD4-Cre+ SMARTA (Id2+/+) or Id2−/− cells into B6 hosts and followed their differentiation after infection. Id2−/− cells did not form a distinct TH1 population (Fig. 3a,b, Supplementary Fig. 3a). Loss of TH1 differentiation by Id2−/− cells was accompanied by decreased granzyme B, T-bet and IFNγ expression, and increased TCF-1 expression (Supplementary Fig. 3b,c). Notably, a prominent SLAMmidCXCR5mid population emerged among Id2−/− effector CD4+ T cells, which was not observed in Id2+/+ cells (Fig. 3a,b). This phenotype was also apparent in a polyclonal CD4+ T cell response (Supplementary Fig. 3d). These results show that Id2 is required for differentiation of TH1 cells.

Figure 3

Id2 is necessary for the generation of TH1 CD4+ helper cells during infection

(a–d,g) Id2+/+ CD4-Cre+ (Id2+/+) or Id2fl/fl CD4-Cre+ (Id2−/−) SMARTA CD4+ T cells were analyzed 4 (a,c,d,g) or 7 (b) days after LCMV infection. (a,b) SLAMhiCXCR5− (green), SLAMmidCXCR5mid (purple) and SLAMloCXCR5+ (blue) cells were analyzed (left panels) and quantified as a frequency of SMARTA CD4+ T cells (middle panels) and total cell numbers (right panels). PSGL-1 and Ly6C (c) and IL-2Rα or Bcl6 (d) expression on the indicated subsets, histogram color (d) corresponds to the gates drawn in (a). (e,f) Id2+/+ or Id2−/− cells were transferred into Bcl6fl/fl CD4-Cre+ mice and analyzed 8 (e) days after LCMV infection. (e) Analysis and quantification of GC B cells (Fas+PNA+) or plasma cells (CD138+IgD−) as a normalized frequency of B cells. (f) LCMV-specific IgG titers at the indicated time points. (g) viSNE analysis of total splenic CD4+ T cells (left panel) or SMARTA CD4+ T cells (right panels), with visualization of SLAM and Bcl6 expression by color gradient. (h) Id2+/+ CD4-Cre+ or Id2fl/fl CD4-Cre+ mice were infected with T. gondii and analyzed 7 days after infection. Analysis and quantification of IFNγ, T-bet and Foxp3 expression by lamina propria CD4+ T cells. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 (two-tailed unpaired Student’s t test). Data are representative of two (f,h) or three (a–d,g) experiments, each with n= 5–10 mice per group, or are pooled from two (e) independent experiments each with n= 10 mice per group (mean ± s.e.m.)

To understand the dysregulation of Id2−/− TH1 cells, we further characterized the phenotypes of the Id2−/− populations after LCMV infection. Id2+/+ and Id2−/− TH1 cells maintained high expression of PSGL-1 and IL-2Rα, which both need to be downregulated for proper TFH differentiation[25,29-31] (Fig. 3c,d). Analysis of the SLAMmidCXCR5mid population revealed that Id2−/− cells shared a partial phenotype with TH1 cells, including high expression of PSGL-1, Ly6C and IL-2Rα, and were Bcl6low when compared to Id2+/+ or Id2−/− TFH cells (Fig. 3c,d). Bcl6 expression was normal in shId2+ early TH1 and TFH (Supplementary Fig. 3e) suggesting that partial loss of Id2 does not impact TH1 cell phenotype. In contrast, the complete absence of Id2 impacted CD4+ T cells throughout differentiation, permanently disrupting TH1 cells.

Id2−/− cells were then assessed for their ability to support B cell responses. Eight days after LCMV infection of Bcl6fl/fl CD4-Cre+ mice that had received Id2+/+ or Id2−/− cells, we observed an increased frequency of Id2−/− TFH cells, but their numbers were similar (Supplementary Fig. 3f). Plasma cell frequencies and anti-LCMV IgG titers in the serum were unchanged (Fig. 3e,f). However, GC B cell development was impaired in mice that received Id2−/− cells (Fig. 3e), suggesting that Id2−/− TFH cells may have impaired function.

To further characterize the differentiation status of Id2−/− CD4+ T cells outside the limitations imposed by 2-parameter flow cytometric plots, and because many of the Id2−/− cells could not be unambiguously assigned to either TH1 or TFH based on SLAM or CXCR5 expression, we employed viSNE multi-parameter clustering, where the overall position of each cell reflects similarity to neighboring cells or dissimilarity to non-neighboring cells based on expression of CD4, CD45.1, SLAM, CXCR5, Bcl6, TCF-1, PD-1 and T-bet[32]. We observed two geographically distinct populations in total CD4+ or specifically in SMARTA CD4+ T cells that uniquely expressed the TH1 and TFH markers SLAM and Bcl6, respectively (Fig. 3g). Id2+/+ and Id2−/− TFH populations are similar in location and appearance (Fig. 3g). However, Id2−/− TH1 cells were located outside of the TH1 multiparameter gate defined by Id2+/+ TH1 cells (Fig. 3g), further suggesting that Id2 is required for proper TH1 differentiation.

To examine the defect of TH1 differentiation in Id2-deficient cells, we employed a model of Toxoplasma gondii (T. gondii) infection for which the role of IFNγ-mediated TH1 responses for long-term resistance and control of infection is well established[33]. Id2fl/fl CD4-Cre+ mice infected with T. gondii harbored CD4+ T cells with markedly reduced expression of both IFNγ and T-bet in the lamina propria of the small intestine compared to Id2+/+ CD4-Cre+ controls (Fig. 3h). No significant alteration in the frequencies of Foxp3+ Treg cells could be detected (Fig. 3h). Thus in two distinct infections, we observe a severe loss of TH1 effector T cells.

Id3 restrains TFH differentiation

A second inhibitor of E protein activity, Id3, is chiefly expressed by TFH and GC TFH cells following infection and has been implicated as an inhibitor of TFH differentiation. However, the role for Id3 in the generation of TFH cells was not examined in the context of infection[22]. Thus, we generated Id3-floxed[34] CD4-Cre+ SMARTA transgenic mice (Id3−/−), transferred cells into B6 mice, and infected with LCMV. In response to infection, Id3−/− cells displayed an increased propensity to become both TFH and GC TFH when compared to Id3+/+ cells (Supplementary Fig. 4a). Further, an increase in frequency of GC TFH was also observed in polyclonal Id3-deficient CD4+ T cells (Supplementary Fig. 4b). We next investigated whether constitutive expression of Id3 could inhibit TFH differentiation; Id3-RV+ or GFP-RV+ LCMV NP-specific (NIP) CD4+ T cells[35] were transferred into B6 mice, which were infected with LCMV. The acquisition of both TFH (Supplementary Fig. 4c) and early TFH (Supplementary Fig. 4d) characteristics were abrogated when Id3 was overexpressed. This is consistent with the observation that Id3 inhibited TFH differentiation following protein immunization[22].

Loss of Id2 regulates expression of key TH genes

To understand how Id2 impacts TH1 and TFH cell differentiation, we studied the global transcriptional changes in CD4+ T cells resulting from Id2 deficiency. Id2+/+ and Id2−/− (encompassing both SLAMhiCXCR5− and SLAMmidCXCR5mid populations) TH1 and Id2+/+ and Id2−/− CXCR5+SLAMlo TFH cells were used for comparative gene expression profiling (Fig. 4a). Differential expression of TH1- and TFH-associated genes was confirmed for Id2+/+ TH1 and TFH cells (Fig. 4b). We then compared gene expression profiles of Id2+/+ and Id2−/− TH1 cells and observed downregulation of TH1-associated genes (Gzmb, Slamf1 and Cxcr6) in the context of Id2 deficiency, while genes associated with the TFH program (Cxcr5, Il6ra and Tcf7) were upregulated (Fig. 4c). However, Bcl6, Ascl2, Pdcd1 and Icos, all highly expressed by TFH cells and important in TFH differentiation, were not elevated in the Id2−/− TH1 population when compared to Id2+/+ TH1 cells, indicating that E proteins control the expression of only a portion of TFH signature genes (Fig. 4d).

Figure 4

Increased E2A binding in the absence of Id2 regulates expression of key T helper genes

Id2+/+ CD4-Cre+ (Id2+/+) or Id2fl/fl CD4-Cre+ (Id2−/−) SMARTA CD4+ T cells were transferred into B6 hosts and infected with LCMV. (a) Microarray analysis was performed on the indicated TH1 and TFH populations sorted 7 days after LCMV infection. (b and c) Comparisons between the indicated populations are shown as expression-by-expression plots with FC ≥1.4, CV ≤ 0.10 and expression ≥40. (d) Expression of putative E2A-target genes identified by ChIP-seq between Id2+/+ TH1 (purple) and Id2−/− TH1 (grey). White bars indicate genes not significantly differentially regulated. (e-h) Volcano plots for the indicated data sets. (e) Differentially regulated Id2+/+ TH1 genes from Id2+/+ TH1 vs Id2+/+ TFH mean-class expression (MCE) plot (b). (f) Differentially regulated Id2+/+ TH1 genes from Id2+/+ TH1 vs Id2−/− TH1 MCE plot (c). (g) Differentially regulated Id2+/+ TFH genes from Id2+/+ TFH vs Id2+/+ TH1 MCE plot (b). (h) Differentially regulated Id2−/− TH1 genes from Id2−/− TH1 vs Id2+/+ TH1 MCE plot (c). (i) Comparisons between Id2+/+ and Id2−/− TFH are shown as expression-by-expression plots with FC ≥1.4, CV ≤0.10 and expression ≥40. (j) Frequency of differentially regulated genes (FC ≥1.4) between Id2+/+ TH1 and Id2−/− TH1 populations, which are also E2A targets as indicated by E2A ChIP-seq. Black bar indicates the background frequency of E2A targets. Data are representative of 2 independent experiments each with n= 5 mice per group.

We next characterized the impact of Id2 deficiency on the expression of TH1-associated genes. A TH1 gene set was selected as all genes upregulated ≥1.4 fold in Id2+/+ TH1 cells compared to Id2+/+ TFH cells (Fig. 4b, green). Id2−/− TH1 cells had reduced expression for ~78% of the TH1-associated genes (Fig. 4e). Additionally, of the 144 genes most substantially downregulated in Id2−/− TH1 cells compared to Id2+/+ TH1 cells (Fig. 4c, purple), ~79% were expressed at higher levels in Id2+/+ TH1 compared to Id2+/+ TFH cells (Fig. 4f). Thus, deletion of Id2 impaired acquisition of the TH1 program.

A TFH gene set was defined as all genes expressed ≥1.4 fold in Id2+/+ TFH cells compared to Id2+/+ TH1 cells (Fig. 4b, blue). Id2−/− TH1 cells inappropriately upregulate ~69% of the TFH-associated genes (Fig. 4g). Analysis of genes most upregulated in Id2−/− TH1 cells compared to Id2+/+ TH1 cells (Fig. 4c, grey) revealed that ~65% were preferentially expressed in Id2+/+ TFH cells (Fig. 4h). These analyses indicate a substantial bias towards the TFH gene expression program in Id2−/− TH1 cells. When gene expression of Id2+/+ and Id2−/− TFH cells was contrasted, only 140 genes showed significant differential expression, indicating that established TFH cells, which express lower levels of Id2, were moderately impacted by Id2 deficiency (Fig. 4i).

The absence of proper TH1 development in Id2−/− cells suggests that unchecked E2A activity impaired TH1 differentiation. Analysis of genes differentially expressed between Id2+/+ TH1 and TFH cells revealed a larger number of E2A-bound genes were upregulated in Id2−/− TH1 cells compared to Id2+/+ TH1 cells (Fig. 4d), consistent with Id2 inhibition of E2A. We compared changes in gene expression with a list of E2A-target genes[16] and found that 62% of the genes upregulated in Id2−/− TH1 cells are targets of E2A (Fig. 4j). These results suggest that Id2 is important for the maintenance of TH1 gene-expression program and that its absence results in acquisition of a partial TFH program.

E proteins drive CXCR5 expression

Our microarray results suggest that Id2 and E proteins act together to control CD4+ T cell differentiation in part by regulating CXCR5 and expression of TH1 effector molecules such as SLAM. We hypothesized that diminished E2A levels could rescue the defect observed in Id2-deficient cells. We expressed an shRNA targeting Tcf3 (encoding E2A) or a control shRNA in Id2+/+ and Id2−/− SMARTA CD4+ T cells (Supplementary Fig. 5a). Cells were adoptively transferred into B6 mice infected the previous day, and differentiation of the transferred cells was analyzed. As expected, Id2−/− cells expressing control shRNA could not correctly differentiate into TH1 cells (Fig. 5a). However, Id2−/− TH1 cells were rescued by shTcf3 expression and defects in SLAM and CXCR5 expression were both corrected (Fig. 5a). Thus, the defective TH1 differentiation we observe in the absence of Id2 is the result of increased E protein activity.

Figure 5

E proteins drive CXCR5 expression and inhibit TH1 formation

(a) Id2+/+ CD4-Cre+ (Id2+/+) or Id2fl/fl CD4-Cre+ (Id2−/−) SMARTA CD4+ T cells were transduced with the indicated shRNAmir-RV, transferred into B6 mice and analyzed 7 days after LCMV infection. TH1 cells were analyzed by flow cytometry and quantified as a fraction of transduced SMARTA CD4+ T cells. (b) CXCR5 expression in GFP+ CD4+ T cells and pooled quantitation. (c) Quantitation of CXCR5 expression by RV+ SMARTA TFH (CXCR5+Bcl6+) cells 3 days after LCMV. (d–k) RV transduced NIP CD4+ T cells were transferred into B6 mice and analyzed 3 (d–g) or 6 days (h–k) after LCMV infection. (d–g) TFH (CXCR5+Bcl6+) and TH1 (CXCR5−Bcl6−) differentiation was analyzed by flow cytometry (d) and quantified as a fraction of NIP CD4+ T cells (f) or total splenocytes (g). (e) Quantitation of CXCR5 expression in NIP TFH (CXCR5+Bcl6+) cells. (h–k) TH1 (PSGL-1+CXCR5−), TFH (CXCR5+PSGL-1+) and GC TFH (CXCR5+PSGL-1−) cell development was analyzed by flow cytometry (h) and quantified as a fraction of NIP CD4+ T cells (j) or total splenocytes (k). (I) Quantitation of CXCR5 expression in NIP TFH (CXCR5+SLAMlo) cells. (l) CXCR5 expression by GFP+Ame+ CD4+ T cells and quantitation. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001(two-tailed unpaired (a–k) or paired (l) Student’s t test). Data are representative of two (a,c), three (d,e) or four (h,i) independent experiments or pooled from two (f,g,j,k), four (l) or nine (b) individual experiments each with n= 3-8 mice per group (mean ± s.e.m.).

The E-box binding bHLH transcription factor Ascl2 has been shown to drive robust TFH differentiation by inducing CXCR5 when overexpressed in CD4+ T cells[22]. E proteins Tcf3 and Tcf12 (HEB) are both highly expressed in TFH cells early after LCMV infection (Supplementary Fig. 5b)[26]. In contrast, Ascl2 is essentially undetectable in either TFH or TH1 cells at the same time point (Supplementary Fig. 5b). Retroviruses overexpressing the Tcf3 isoforms E12, E47 or Ascl2 all induced CXCR5 expression by CD4+ T cells in vitro (Fig. 5b). Ectopic expression of E47 led to enhanced expression of CXCR5 by both early TH1 cells and early TFH cells when compared to their GFP-RV+ counterparts in vivo (Fig. 5c, Supplementary Fig. 5c).

Given that Id2 inhibits the transcriptional activity of E proteins and E proteins induce CXCR5 expression, we investigated whether Id2 inhibited TFH differentiation by preventing expression of CXCR5. Id2-RV+ or GFP-RV+ NIP CD4+ T cells were transferred into B6 mice, which were infected with LCMV. Id2-RV+ NIP CD4+ T cells displayed decreased differentiation into early TFH cells (Fig. 5d,f,g) and impaired CXCR5 expression (Fig. 5e, Supplementary Fig. 5e). Overexpression of Id2 did not impact expression of T-bet or accumulation of Id2-RV+ NIP CD4+ T cells (Supplementary Fig. 5d,f). Similar to what was previously observed, Id2-RV+ NIP CD4+ T cells poorly differentiated into TFH cells 6 days after LCMV infection (Fig. 5h,j,k, Supplementary Fig. 5g–j), with impaired CXCR5 expression by TFH cells (Fig. 5i, Supplementary Fig. 5k). Additionally, accumulation of Id2-RV+ GC TFH cells was impaired (Fig. 5h,j,k). Next, we constitutively co-expressed E proteins, Ascl2 and Id2 in CD4+ T cells. E proteins were expressed from a GFP reporter RV while Id2 was expressed from an Ametrine reporter RV. As expected, the E proteins E12, E47 and Ascl2 drove strong CXCR5 expression when CD4+ T cells were co-transduced with an empty Ametrine-RV (Fig. 5l). When Id2-Ame-RV was introduced into E12- or E47-RV+ CD4+ T cells, there was a reduction in CXCR5 expression (Fig 5l) by GFP+Ametrine+ cells. Surprisingly, Id2 was not able to block the Ascl2-driven induction of CXCR5 (Fig. 5l). These data indicate that Id2 prevents E proteins from inducing CXCR5 expression.

Bcl6 inhibits Id2 expression

Our data show that the Id2-E protein axis modulates TH1 and TFH differentiation, and that Id2 inhibits expression of Cxcr5. The transcriptional repressor Bcl6 is essential for TFH differentiation and is important for CXCR5 expression by TFH cells in vivo[8,25,36], but it does not directly regulate Cxcr5[22,37]. We therefore asked whether Bcl6 induces CXCR5 expression by inhibiting Id2 transcription. Bcl6 ChIP-Seq of primary tonsillar GC TFH revealed recruitment of BCL6 to the ID2 locus (Fig. 6a)[37], which was confirmed by ChIP-qPCR (Fig. 6b). To test whether Bcl6 represses Id2 expression, we transduced Bcl6fl/fl CD4-Cre+ SMARTA CD4+ T cells (Bcl6−/−) with GFP-RV or Bcl6-RV, transferred RV+ cells into B6 mice, infected with LCMV, and assessed expression of Id2 (Supplementary Fig. 6a, Fig. 6c). Re-introduction of Bcl6 into Bcl6−/− cells led to significant repression of Id2 in IL-2Rαhi Th1 cells (Fig. 6c). Recent work demonstrated that separate domains of Bcl6 control TFH differentiation, and mutation of lysine K379Q significantly hinders the activity of Bcl6[27,35]. Introduction of the Bcl6 K379Q mutant into Bcl6−/− cells failed to repress Id2 compared to WT Bcl6 (Fig. 6c). Thus, Bcl6 directly represses Id2 in CD4+ T cells.

Figure 6

Bcl6 inhibits Id2 expression

(a) BCL6, H3K27Ac, H3K4me1 and H3K4me3 density tracks of ID2 in human GC TFH cells. Arrows above the BCL6 tracks indicate primers used in ChIP-qPCR analysis. Sequence conservation with mouse is shown at the bottom. (b) PD-1hi GC TFH cells were isolated from human tonsil cells, chromatin was prepared and ChIP was performed for BCL6 at the indicated loci. (c) Bcl6fl/fl CD4-Cre+ SMARTA CD4+ T cells transduced with the indicated vectors were transferred into B6 mice. IL-2Rα+ SMARTA CD4+ T cells were sorted 3 days after LCMV infection and Id2 expression was tested by qRT-PCR. (d) WT (Bcl6+/+), Bcl6fl/WT CD4-Cre+ (Bcl6+/−), Bcl6fl/fl CD4-Cre+ (Bcl6−/−) SMARTA CD4+ T cells were transferred into B6 mice. IL-2Rα+ or IL-2Rα− SMARTA cells were sorted and Id2 expression was tested by qRT-PCR. *p<0.05, **p<0.01, ***p<0.0001 (two-tailed unpaired Student’s t test). Data are pooled from two (c,d) or four (b) independent experiments, with n= 5 mice per group (b) (mean ± s.e.m.). For c and d, 3 mice were pooled to create a data point.

We also asked how Bcl6 gene copy number affected Id2 expression. Bcl6−/−, Bcl6+/− or WT CD4-Cre+ SMARTA CD4+ T cells were transferred into B6 mice and IL-2Rαhi (TH1) and IL-2Rαlo (TFH) cells were sorted following LCMV infection (Supplementary Fig. 6b). As expected, WT TFH cells had reduced expression of Id2 compared to TH1 cells (Fig. 6d). Id2 expression was significantly increased in IL-2Rαlo Bcl6+/− cells in comparison to WT cells (Fig. 6d). Further, complete loss of Bcl6 resulted in a significant upregulation of Id2 expression in IL-2Rαlo cells compared to IL-2Rαlo Bcl6+/− and WT cells (Fig. 6d). Thus, Bcl6 inhibits Id2 and Bcl6 haploinsufficiency results in inappropriate Id2 expression.

Discussion

E and Id proteins are pivotal regulators of lymphocyte development and function. Here, we investigated an unexplored role for Id2 in the differentiation TH cells in response to acute viral infection and show that Id2 controls the balance of TH1/TFH by inhibiting E protein activity. Id2 and Id3 are preferentially expressed in TH1 and TFH cells, respectively. Reduction of Id2 levels in CD4+ T cells results in a higher proportion of TFH cells. Complete ablation of Id2 hampered TH1 generation, resulting in an abnormal effector population exhibiting mixed traits of TH1 and TFH lineages. Further, Bcl6 specifically inhibits Id2 to ensure E protein activity, driving a portion of the TFH program, thus establishing Id2 as a critical enforcer of proper TH differentiation.

Reduction of Id2 levels shifts the balance of TH1/TFH cells indicating that partial expression of Id2 can inhibit enough E protein expression to maintain both helper populations while still biasing cells towards the TFH lineage. Strikingly, CD4+ T cells that completely lack Id2 lose the ability to form a TH1 effector cell population, while maintaining an intact TFH population. Id2-deficient effector cells exhibit weak expression of TH1-associated genes and show simultaneous upregulation of a large portion of the TFH gene program (Cxcr5, Il6ra, Lef1 and Tcf7, but not Bcl6, Icos and Pdcd1). Id2-deficient cells may be unable to commit to one lineage for a number of reasons. While they adopt aspects of the TFH transcriptional program, Id2-deficient cells also upregulate and maintain high levels of Id3, Foxo1 and Il2ra, which may explain this dichotomy. Foxo1 specifically inhibits TFH development[38,39]. Within the first two cell divisions, expression of IL-2Rα is a key factor in driving the TH1 lineage decision[25]. Their expression in the absence of Id2 may counterbalance the TFH gene program. Thus, Id2 and E proteins are powerful regulators of key TFH-associated genes and many TH1-associated genes, but the unusual phenotype of Id2-deficient effector CD4+ T cells demonstrates that Id2 and E proteins control gene sets that do not per se result in polarized differentiation of TH1 and TFH cells.

Previous work demonstrated a role for Id3 in regulating the TFH gene expression program[20,22,24,40]. Our work further shows that specific deletion of Id3 in TH cells promotes TFH/GC TFH formation but does not impact TH1 differentiation following infection. Both differential expression patterns and unique binding partners are plausible explanations as to how Id2 and Id3 may control distinct TH subsets. They may inhibit DNA binding of different E proteins with differing affinities and have differentially regulated binding activity and protein stability. In support of this, we observe that Id2 inhibits E47 but not Ascl2 induction of Cxcr5. Loss of Id2 dramatically impairs TH1 differentiation, but Id3-deficiency does not phenocopy Id2-deficiency, instead enhancing TFH and GC TFH differentiation. These observations support the hypothesis that inhibition of E proteins alters the TH1/TFH balance: Id2 is important in the upregulation of TH1 genes while Id3 restrains TFH differentiation.

Ascl2 has been suggested to act upstream of Bcl6 to regulate early TFH differentiation[22]. However, Ascl2 is generally not detectable in naive CD4+ T cells[41-43](Immgen.org), nor early TFH cells[26]. Instead, Ascl2 expression is reported for fully differentiated GC TFH cells in mice and humans[22,43]. The high expression of E2A and HEB early after LCMV infection suggests that these E proteins, not Ascl2, direct early TFH differentiation. Interestingly, Ascl2-induced expression of CXCR5 was not dampened by co-expression of Id2. GC TFH have the highest levels of CXCR5 expression and Ascl2 may be important for amplifying expression at later stages of TFH differentiation into GC TFH cells.

There are multiple plausible models for the coordination of genes regulating the earliest stages of TFH differentiation in vivo. Bcl6 function is critical for TFH differentiation[2,27,35], and can be detected as early as the second cell division in vivo[25]. Tcf7 and Lef1, both of which are known E protein targets[44], are epistatic to Bcl6 and promote TFH differentiation by enhancing expression of Bcl6, IL-6R and ICOS and repressing Prdm1[26,45,46]. Notably, we observed increased expression of Tcf7, Lef1, and Il6ra in the absence of Id2, supporting the idea that E proteins such as E2A and HEB normally promote TCF-1 and LEF-1 expression. In this regard, Id2 and E proteins act upstream of Bcl6. However, our data also demonstrate that Bcl6 directly represses Id2 expression. Altogether these data suggest that positive feedback mechanisms involving TCF-1, LEF-1, Bcl6, IL-6R, ICOS, and E proteins support TFH differentiation under low Id2 conditions, and that a feedforward loop can potentially be generated by starting at any of several genes in that gene network.

The relationship between E/Id proteins, Bcl6 and CXCR5 expression is of particular interest. Ectopic expression of Bcl6 in human CD4+ T cells results in CXCR5 expression[47]. Coordinated expression of Bcl6 and CXCR5 in early TFH cells is observed in multiple in vivo models[25,36,48,49]. Yet Bcl6 does not bind to Cxcr5[37], and thus must regulate its expression indirectly. One mechanism involves repression of Cxcr5 by Blimp-1[7]. But naive T cells do not express Blimp-1, indicating that this mechanism primarily regulates later CXCR5 expression. Here, we provide a new mechanism whereby Bcl6 inhibits Id2 expression, which yields enhanced E protein activity to drive Cxcr5 expression.

Our data uniquely position Id2, Bcl6 and E proteins in a regulatory triad that controls the balance of TH1 and TFH differentiation. Through inhibition of E proteins, high expression of Id2 in the TH1 population enforces proper development of the TH1 lineage. Early expression of Bcl6 in TFH cells ensures repression of Id2, allowing E proteins to drive TFH differentiation. Thus, dichotomous expression of Id2 is critical in ensuring the reciprocal development of TH1 and TFH cell differentiation.

ONLINE METHODS

Mice

CD4-Cre+ mice were from Jackson Laboratory. Mouse strains described below were bred and housed in specific pathogen–free conditions in accordance with the Institutional Animal Care and Use Guidelines of the University of California San Diego or the La Jolla Institute. Id2-YFP[14], Id3-GFP[24], Id2fl/fl28, Id3fl/fl34, SMARTA TCR transgenic (specific for LCMV gp66-77)[50], Bcl6fl/fl51 CD45.1 congenic, NIP TCR transgenic (specific for LCMV NP311-325)[35] and OT-II TCR transgenic (specific for OVA 323-339) mice were on a fully B6 background. Recipient C57BL/6J mice were either bred at UCSD or received from The Jackson Laboratory. Both male and female mice were used throughout the study, with sex and age matched T cell donors and recipients.

LCMV Infection and protein immunizations

Recipient mice were infected by intraperitoneal injection of 2 × 105 or 5 × 105 plaque-forming units of LCMV-Armstrong for days 6/7 or 3 analysis, respectively. In adoptive transfer experiments of naive CD4+ T cells, LCMV infection was performed one day after cell transfer. A total of 20 μg of 4-Hydroxy-3-nitrophenylacetyl-OVA (NP16-OVA; Biosearch Technologies) was prepared in 5% alum (Aluminum potassium sulfate, Sigma) in a total volume of 20 μL and injected into each footpad.

Toxoplasma gondii infection and lymphocyte isolation from small intestine

The ME49 strain of T. gondii was maintained in CBA/CaJ mice by intraperitoneal injection of 20 cysts and cysts were obtained from brain homogenates after 5–6 weeks. Mice were infected with 40 cysts of ME49 by gavage. Small intestine was harvested on day 7-post infection to analyze TH1 response. To isolate lamina propria lymphocytes, small intestine were cut and washed with plain RPMI-1640, and epithelial cells were removed by incubation with 5 mM EDTA and 1mM DTT for 20 min at 37 °C, followed by enzyme digestion with 0.16 U/ml liberase TL (Roche) for 30 min at 37 °C. Lymphocytes were enriched by centrifugation with 47% percoll.

Flow cytometry and histology

Single-cell suspensions of spleen or draining popliteal lymph nodes were prepared by standard gentle mechanical disruption. Surface staining for flow cytometry was done with monoclonal antibodies against CD4 (RM4-5, 1:400), CD8 (53-6.7, 1:400), CD45.1 (A20, 1:400), CD25 (PC61.5, 1:400), B220 (RA3-6B2, 1:400), IgD (11-26, 1:400), PD-1 (J43, 1:400) and CD138 (eBiosciences, 1:500); PSGL-1 (2PH1, 1:800), CD138 (281-2, 1:500), Fas (Jo2, 1:400) (from BD Biosciences); SLAM (TC15-12F12.2, 1:400), CD25 (PC61, 1:400), CD4 (GK1.5, 1:400), Ly6C (AK1.4, 1:800) (BioLegend) and PNA (Vector Laboratories). Stains were done for 30 min at 4 °C in PBS supplemented with 0.5% bovine serum albumin and 0.1% sodium azide, unless specified otherwise. PE-labeled IAb/GP66–77 tetramer was supplied by the National Institute of Health (NIH) tetramer core facility. Single-cell suspensions were stained with tetramer at 37 °C for 2 h. CXCR5 staining was done as described[52], using purified anti-CXCR5 (2G8; BD Pharmingen) for 1 h, followed by biotinylated anti-rat IgG (Jackson Immunoresearch), and then by PE-, PE-Cy7- or APC-labeled streptavidin (eBioscience) at 4 °C in PBS supplemented with 0.5% bovine serum albumin, 2% fetal calf serum, and 2% normal mouse serum. Intracellular staining was performed with an Alexa 647- or PE-conjugated monoclonal antibody to Bcl6 (clone K112-91; BD Pharmingen, 1:20), TCF1 (clone C63D9; Cell Signaling, 1:200), IFN-γ (clone XMG1.2; eBioscience, 1:200), T-bet (clone 4B10; eBioscience, 1:200), FoxP3 (clone FJK-16s; eBioscience, 1:200) and the Foxp3 ICS kit buffers and protocol (eBioscience). Stained cells were analyzed using LSRII, LSRFortessa or LSRFortessa X-20 (BD) and FlowJo software (TreeStar). All sorting was done on a FACS Aria (BD Biosciences). For RT-PCR analyses, early TH1 (IL-2Rα+) and TFH (IL-2Rα−) amongst total or RV+ SMARTA CD4+ T cells were sorted 3 days after infection with LCMV. Histology was performed as previously described[53].

ELISA

Nunc MaxiSorp plates (Thermo Fisher Scientific) were coated overnight at 4 °C with 1 μg/mL NP23-BSA (Biosearch Technologies) or with a 1:60 dilution of LCMV lysate (prepared from LCMV-infected BHK cells) in PBS. Plates were blocked with PBS + 0.2% Tween-20 + 1% BSA for 90 min at 25°C. After washing, mouse serum was added in a dilution series in PBS + 0.2% Tween-20 + 1% BSA and incubated for 90 min. After washing, horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG (Thermo Fisher Scientific) was added at 1:5,000 in PBS + 0.2% Tween-20 + 1% BSA for 90 min at 25°C. Colorimetric detection was performed using a TMB substrate kit (Thermo Fisher Scientific). Color development was stopped after approximately 10 min with 2 N H2SO4, and absorption was measured at 450 nm.

Retroviral vectors, transductions and cell transfers

shRNAmir and pMIG, Bcl6 MIG and middle domain mutant Bcl6 (K379Q) vectors were described previously[27,52]. E12, E47, Ascl2, Id2 and Id3 were cloned into the pMIG or pMIA vectors, which contain an IRES-GFP or IRES-mAmetrine, respectively. Virions were produced by transfection of the PLAT-E cell line, as described previously[8]. Culture supernatants were collected 24 and 48 h after transfection, filtered through a 0.45 μm syringe filter and stored at 4°C until transduction. CD4+ T cells were isolated from whole splenocytes by negative selection (Stemcell Technologies) and resuspended in D-10 (DMEM + 10% fetal calf serum, supplemented with 2 mM Glutamax (Gibco) and 100 U/mL Penicillin/Streptomycin (Gibco)) with 2 ng/mL recombinant human IL-7 (Peprotech) and 50 μM β-mercaptoethanol (BME). 2 × 106 cells were stimulated in 24-well plates pre-coated with 8 μg/mL anti-CD3 (17A2; BioXcell) and anti-CD28 (37.51; BioXcell). At 24 and 36 h after stimulation, cells were transduced by adding RV supernatants supplemented with 50 μM BME and 8 μg/mL polybrene (Millipore), followed by centrifugation for 90 min at 524 x g at 37 °C. Following each transduction, the RV-containing medium was replaced with D-10 + 50 μM BME + 10 ng/mL human IL-2. After 72 h of in vitro stimulation, CD4+ T cells were transferred into six-well plates in D-10 + 50 μM BME + 10 ng/mL human IL-2, followed by incubation for 2.5 days. One day before transfer, the culture medium was replaced with D-10 + 50 μM BME + 2 ng/mL human IL-7. Transduced cells were sorted based on reporter expression (FACSAria; BD Biosciences). Transfer of sorted cells into recipient mice was performed by intravenous injection via the retro-orbital sinus. Transferred cells were allowed to rest in host mice for 3-4 days before infection or immunization. 2 × 104 or 4 × 105 transduced CD4+ T cells were transferred into each mouse for day 6 or 3 analysis, respectively. For protein immunization, 1 × 105 transduced CD4+ T cells were transferred into each mouse. (Fig. 5a) DNA fragments encoding shRNAs targeting mouse Tcf3 or Cd8a were subcloned into a custom retroviral vector containing the miR30 backbone plus the murine PGK promoter and dsRED as a reporter. 1 × 106 naive Id2+/+CD4-Cre+ and Id2fl/flCD4-Cre+ SMARTA CD4+ T cells were stimulated in 24-well plates pre-coated with anti-CD3 and anti-CD28 for 18 h. Following stimulation, cells were transduced by adding RV supernatants supplemented with 100 U/ml human IL-2 and 8 μg/mL polybrene, followed by centrifugation for 90 min at 2000 × g at 37 °C. Following transduction, cells were incubated for 3 h at 37 °C. 5×104 cells were transferred into day −1 LCMV infected hosts and remaining cells were cultured in vitro with D-10 + 50 U/mL human IL-2 in a parallel time course to assess for knockdown efficiency.

Microarray and ChIP-seq

Id2+/+ CD4-Cre+ and Id2fl/fl CD4-Cre+ CD4+ T cells (pooled from 5 mice) were isolated via flow cytometry on day 7 of LCMV infection (FacsARIA, BD). For microarray analysis, RNA was extracted with TRIzol reagent, amplified and hybridized to the Affymetrix Mouse Gene 1.0 ST Array. Data were normalized and analyzed with the GenePattern software suite. E2A Bio-Chip was performed as previously described on total thymocytes from Tcfe2aBio/Bio Rosa26BirA/BirA mice[16,54].

Quantitative RT-PCR and ChIP-qPCR

Total RNA from the sorted cells was extracted and reverse-transcribed, and quantitative PCR was performed using SYBR Select MasterMix (Thermo Fisher Scientific). Results were normalized to the expression of Gapdh transcripts. Primary GC TFH were isolated from human tonsil by staining with biotin-conjugated PD-1 (J105, eBioscience) followed by isolation using Streptavidin microbeads (Miltenyi). GC TFH were crosslinked with 1% formaldehyde and then quenched with 125 mM glycine. Cells were lysed in RIPA buffer (150 mM NaCl, 1% NP-40, 0.5% Na-Deoxycholate, 0.1% SDS, 50 mM Tris and 5 mM EDTA) supplemented with protease inhibitors (Thermo Fisher Scientific) and 0.5 mM PMSF followed by sonication and isolation of chromatin. Protein G Dynabeads (Life Technologies) were conjugated to antibodies specific to Bcl6 (N-3 and C-19, Santa Cruz). Normal rabbit IgG (Santa Cruz) was used as a control. Chromatin was immunoprecipitated using the conjugated beads, eluted, and reverse crosslinked using 0.3 M NaCl at 65 °C overnight. qPCR was performed on isolated DNA and sample values were given as a percentage of input. Primers are listed in Supplementary Fig. 6c.

Statistical Methods

Statistical tests were performed using Prism 6.0 (GraphPad). Significance was determined by unpaired Student’s t-test with a 95% confidence interval.