Bach2 Controls T Follicular Helper Cells by Direct Repression of Bcl-6

An effective humoral immune response requires cognate interaction of Ag-specific B and T cells (1). In the germinal center (GC) reaction, B cells proliferate and undergo affinity maturation under the control of a specialized T cell subset of T follicular helper (Tfh) cells (2). Consequently, Tfh cells are the prerequisite for the generation of high affinity memory B cells and long-lived plasma cells. Therefore, manipulation of the Tfh response is of particular clinical interest to either promote the generation of protective Abs during vaccination or to eliminate harmful Abs in autoimmune diseases or allergy (3).

Tfh cells provide B cell help by expression of CD40L and IL-21. They are characterized by high expression of the chemokine receptor CXCR5, the coinhibitory receptor PD-1, and the lineage-defining transcription factor Bcl-6. The development of Tfh cells from naive CD4⁺ T cells is a multistep process that requires interaction with different cell types and only takes place in the complex anatomical structure of secondary lymphoid organs (4). Although our knowledge about molecules regulating Tfh cells has substantially expanded in past years, many details are still ill defined. The specific signals maintaining the Tfh cell phenotype in later phases of the GC reaction are poorly understood. Furthermore, little is known how the Tfh cell master transcription factor Bcl-6 itself is regulated.

Much about the molecular regulation of Tfh cells can be learned from B cells, because the transcriptional networks defining the GC B cell and Tfh cell phenotype share many similarities (5). The prime example is Bcl-6, which has a key role for both the development of GC B cells and Tfh cells. Bcl-6 inhibits the expression of Blimp-1, which is the transcription factor driving alternative fate differentiation of B and T cells to plasma cells and non-Tfh effector cells, respectively (5). Another important transcription factor to maintain the phenotype of GC B cells is Bach2 (6). Bach2 is highly expressed in GC B cells and cooperates with Bcl-6 to suppress expression of Blimp-1, thereby preventing terminal differentiation of GC B cells into plasma cells (7–9). For many years, Bach2 was considered to be a B cell–specific transcription factor (10). Only recently, Bach2 was also demonstrated to be expressed in T cells to promote the development of regulatory T cells and prevent differentiation of Th2 CD4⁺ T cells (11–13). However, whether Bach2 has a function also for Tfh cells is still unknown.

In this study, we show that Bach2 is highly differentially expressed in Tfh versus non-Tfh effector cells. However, in striking contrast to GC B cells, Tfh cells are characterized by low expression of Bach2, indicating that the function of this transcription factor in T cells is completely different. Ectopic overexpression of Bach2 in already-differentiated Tfh cells results in rapid loss of their phenotype, identifying Bach2 as a key molecule to maintain Tfh cells in later phases of the GC reaction. Bach2 overexpression in Tfh cells inhibits a specific set of genes, including IL-21, TIGIT, and Bcl-6. Furthermore, we can demonstrate that Bach2 directly represses Bcl-6 by binding in the promotor region and that inhibition of Bcl-6 promoter activity resulted from displacing an activating complex of Irf-4 and Batf.

TCR-transgenic OT-II (14) and Smarta (15) mice were additionally crossed to B6PL mice (Jax stock 000406; Thy-1.1⁺) to track cells in adoptive transfer experiments. For the tamoxifen-inducible retroviral overexpression system, both strains were crossed to CreER^T2 mice (16). For CRISPR/Cas9-mediated knockout (KO), Smarta mice were crossed to Cas9-GFP transgenic mice (17). For analysis of Il21 expression, we used Smarta CreER^T2 mice crossed to Il21-IRES-FP635 reporter mice (18, 19). Nitrophenol (NP)–specific BCR knock-in B1-8i mice (20) were crossed to κ-L chain KO mice (21) to ensure NP-specificity of all B cells and Ly-5.1 mice (Jax stock 002014; CD45.1⁺) for tracking of transferred cells. As recipient mice for adoptive transfer experiments, C57BL/6 or CD28 KO mice (22) were used. All mice were bred under specific pathogen-free conditions in the animal facility of the Federal Institute for Risk Assessment, Berlin. Female mice were used for experiments at an age of 8–16 wk. Animal handling and experiments were conducted according to the German animal protection laws and approved by the responsible governmental authority (Landesamt für Gesundheit und Soziales).

For constitutive overexpression, cDNAs for Bach2, Tigit, Irf4, or Bcl6, followed by an IRES site and either eGFP or mAmetrine were cloned into the retroviral expression vector pMSCV (Takara Bio). For inducible overexpression, Bach2 was cloned behind a loxP-flanked dsRed stop cassette (23). For KO of genes, guide RNAs were selected using CrispRGold version 1.1 (https://crisprgold.mdc-berlin.de). Corresponding oligonucleotides (Supplemental Table I) were cloned into the BbsI site of a pMSCV-based guide RNA vector (17). Viral particles were generated in HEK293T cells by calcium phosphate transfection using the packaging plasmids pECO and pCpG. In vitro stimulated OT-II or Smarta T cells (OVA_323–339 or LCMV GP_61–80 peptide for 24 h) were infected with retroviral supernatants by centrifugation for 90 min at 400 × g, 32°C, and 8 μg/ml polybrene. Cells were cultured for additional 40 h and sorted for fluorescent protein–expressing cells on a FACSAria II flow sorter (BD Biosciences).

To determine the efficiency of CRISPR/Cas9-mediated KO, some cells were cultured for additional 24 h and sorted for Cas9-GFP and guide RNA–expressing T cells to prepare genomic DNA. The genomic region around the guide RNA binding site was amplified by PCR using the primers listed in Supplemental Table II. PCR products were cloned into pJet1.2 (Thermo Fisher Scientific) and subjected to Sanger sequencing. Sequences from at least 11 individual clones were aligned to the original DNA sequence.

A total of 2.5 × 10⁵ retrovirally infected transgenic T cells were adoptively transferred by i.v. injection into C57BL/6 mice. For some experiments, 1 × 10⁶ B1-8i B cells were cotransferred. Recipient mice were injected s.c. at the tailbase with cognate Ag (50 μg NP-conjugated OVA for OT-II T cells or 20 μg NP and GP_61–80 coupled to mouse serum albumin for Smarta T cells) in complete Freund’s adjuvant (Sigma-Aldrich). To induce nuclear translocation of CreER^T2 recombinase, mice were injected i.p. with 1.5 mg tamoxifen (Sigma-Aldrich) dissolved in sunflower oil. For determination of cell proliferation, 1 mg bromodeoxyuridine (BrdU; Sigma-Aldrich) was injected i.p. 12 and 24 h after tamoxifen administration. To block lymphocyte egress, 30 μg FTY720 (Sigma-Aldrich) was administered i.p. Draining inguinal lymph nodes were isolated at indicated time points and mashed through 70-μm sieves to prepare single-cell suspensions for flow cytometric analysis. Cells were counted with a Guava EasyCyte capillary flow cytometer and ViaCount solution (Merck Millipore). Transgenic T cells were identified in flow cytometry as CD4⁺B220⁻CD8⁻Thy-1.1⁺. Transgenic B cells were defined as CD19⁺CD3⁻CD8⁻CD45.2⁻CD45.1⁺.

Sorting of transgenic murine Tfh and non-Tfh cells was performed as described (24). In brief, transgenic T cells from inguinal lymph nodes were first enriched by magnetic sorting using Thy-1.1 MicroBeads (Miltenyi Biotec), followed by sorting on an BD FACSAria II flow sorter (BD Biosciences) for live B220⁻CD8⁻CD4⁺Thy-1.1⁺CD44^high and either CXCR5/PD-1 double-positive or negative cells. Human tonsils from patients undergoing routine tonsillectomy were obtained after informed consent. Mononuclear cells were prepared by mechanical disruption of tissue and Ficoll density centrifugation. Tfh and non-Tfh cells were sorted as live CD3⁺CD4⁺CD45RO⁺CD45RA⁻ and ICOS/CXCR5 double positive or double negative, respectively.

Single-cell suspensions from lymph nodes were stained with different combinations of mAbs (Supplemental Table III) conjugated to Biotin, FITC, PE, PerCP, PE-Cy7, Alexa Fluor 647, Alexa Fluor 700, APC-Cy7, Pacific Blue, Brilliant Violet 421, Brilliant Violet 605, Brilliant Violet 711, or Brilliant Violet 785. Streptavidin/PE-Cy7 was used as secondary reagent for biotinylated Abs. Fc-receptors were blocked with 2.4G2 (anti-CD16/32). Abs were bought from commercial suppliers or were purified from hybridoma supernatants and coupled to fluorophores by standard procedures. To detect early apoptotic cells, a FITC-conjugated inhibitor of caspase-3 (CaspGLOW; eBioscience) was used according to the manufacturer’s instructions. Intracellular staining for transcription factors and apoptotic proteins was performed using the Foxp3 Staining Buffer Set from eBioscience. DNA-incorporated BrdU was labeled with the APC BrdU Flow Kit (BD Biosciences) according to the manufacturer’s instructions. To discriminate dead cells, either DAPI was added to live cells immediately before analysis or cells were stained on ice for 25 min with 84 nM Alexa 700 succinimidyl ester (Thermo Fisher Scientific) prior to fixation. Cells were analyzed on a BD LSRFortessa (BD Biosciences). Analysis gates were set on live cells defined by scatter characteristics and exclusion of DAPI- or Alexa 700-positive cells. Data were analyzed with FlowJo Software (Tree Star).

RNA was isolated using the RNeasy Micro Kit (Qiagen) or Roche High Pure RNA Isolation Kit, and quality was checked on an Agilent 2100 Bioanalyzer. For quantitative RT-PCR, cDNA was synthesized with the Applied Biosystems High Capacity cDNA Kit. Gene expression was analyzed on an Applied Biosystems 7500 Real Time PCR System using TaqMan gene expression assays for murine Bach2 (Mm00464379_m1), human Bach2 (Hs00222364_m1), and Hprt for standardization (Mm01545399_m1 and Hs02800695_m1).

For gene expression analysis by next-generation sequencing, either the TruSeq Stranded Total RNA Library Prep Kit (Illumina) or the SMART-Seq v4 Ultra Low Input RNA Kit (Clontech Laboratories) and Nextera XT Library Preparation Kit (Illumina) were used according to the manufacturer’s instructions. Paired-end sequencing with 2 × 75 nt was performed on a NextSeq500 (Illumina), and raw sequence reads were mapped to the mouse GRCm38/mm10 genome with TopHat2 (25) in very-sensitive settings for Bowtie2 (26). Gene expression was quantified either by HTSeq (27) for total RNA or featureCounts (28) for mRNA and analyzed using DESeq2 (29). Differential expression between TFH and non-TFH or empty control vector and Bach2 was regarded as significant when the adjusted p value was ≤0.05 and the fold change was ≥1.3.

EL4 T cells (American Type Culture Collection, TIB-39; tested for mycoplasma) were transfected by nucleofection (Amaxa Cell Line Nucleofector Kit L; Lonza Biosciences) with the pGL3 basic luciferase plasmid containing the Bcl-6 promotor (bp +51 to −947), internal control pRL-TK Renilla plasmid (both from Promega) and the Bach2, Batf, and Irf4 coding sequence in pcDNA3.1 (Thermo Fisher Scientific). Luciferase activity was measured on a SpectraMax i3× microplate reader (Molecular Devices) after 24 h using the dual luciferase assay system (Promega). Data were normalized to the activity of Renilla luciferase.

Smarta T cells were stimulated with GP_61–80 peptide under Tfh-polarizing conditions in the presence of IL-21, IL-6 (BioLegend), α–IL-12 (17.8), α–IL-4 (11B11), α–IFN-γ (AN18.14), and α–TGF-β (1D11) as described (30) and retrovirally infected with either FLAG-tagged or control Irf-4 in combination with a Bach2 expression or empty control vector. Chromatin immunoprecipitation (ChIP) was performed with sorted double-transduced cells as described previously (23). Briefly, 1 × 10⁶ cells were cross-linked with 1% formaldehyde, followed by chromatin shearing by sonication and immunoprecipitation with an anti-FLAG Ab (clone M2; Sigma-Aldrich). Enrichment at the Bcl-6 promoter was measured by SYBR Green–based RT-PCR using the primer pair 5′-GAGGGAGAGTGCGCTTTGC-3′ and 5′-GGAGCCGAGTTTATGGG-TAGGA-3′. Binding was determined by calculating the percent of input relative to the nonflagged control.

Experimental groups consisted of four to seven individual mice per group (randomly assigned). Sample size was predetermined using G*Power 3 (31). Data are presented either showing the mean and results from single animals or the mean with error bars (SEM). Flow cytometry plots are always shown as concatenated data from all animals in the group. No data exclusion criteria were used. Data were analyzed using GraphPad Prism 7. For samples with normal distribution according to Shapiro–Wilk, differences between groups were calculated using a two-tailed Student t test or one-way ANOVA with Bonferroni multiple comparison test. Otherwise, data were analyzed by Mann–Whitney U test.

To identify novel factors that might be important to maintain the phenotype of already-differentiated Tfh cells, we analyzed the transcriptomes of Ag-specific Tfh and non-Tfh effector cells. To this end, TCR-transgenic cells expressing the congenic marker Thy-1.1 were transferred into syngenic Thy-1.2⁺ mice, which were immunized s.c. with cognate Ag. On day 8, the maximum of the GC response in this system, TCR-transgenic, OVA-specific, Thy-1.1⁺ OT-II T cells were reisolated from draining lymph nodes and sorted into Tfh and non-Tfh cells according to expression of CXCR5 and PD-1. Next-generation sequencing identified more than 980 genes with a highly significant (p < 0.01) and more than 1.3-fold differential expression between CXCR5/PD-1 double-positive versus double-negative cells.

Because of their key function for cell subset differentiation, we focused our analysis on the gene family of transcription factors (Fig. 1A). Among the 29 differentially expressed transcription factors, Bach2 was of special interest because of its already-reported major role for B cells during the GC reaction. Bach2 reads were almost four times lower in Tfh compared with non-Tfh effector cells (Fig. 1B). Similar results were obtained for Tfh cells isolated from human tonsil; Bach2 expression in CXCR5/ICOS double-positive Tfh cells was even 12 times lower than in double-negative cells (Fig. 1C). It has been previously reported that the KO of Bach2 resulted in a preferential differentiation of CD4⁺ T cells toward a Th2 phenotype (13). Therefore, we analyzed Bach2 expression in different T helper subsets. Whereas Bach2 expression in Th1 and Th17 cells was more than two times higher than in naive T cells, a downregulation by more than 85% was observed in Th2 and Tfh cells, indicating that Th subsets differentially rely on Bach2 and that Bach2 expression seems to be critical for Th2 and Tfh cells (Fig. 1D).

This low expression of Bach2 in Tfh cells was a highly unexpected finding, because in B cells, Bach2 is known to directly repress the transcription factor Blimp-1, which itself antagonizes Bcl-6 (5). Consequently, high Bach2 expression levels in GC B cells prevent Blimp-1 expression and ensure Bcl-6 expression (32). In contrast, Tfh cells seem to downregulate Blimp-1 independent of Bach2, indicating that Bach2 in Tfh cells functions in a way completely different from the regulatory network in B cells.

To functionally analyze the role of Bach2 for Tfh cells, we ectopically overexpressed Bach2 in already-differentiated Tfh cells. Thy-1.1⁺ OT-II T cells expressing tamoxifen-inducible Cre recombinase were transfected with a retroviral vector containing Bach2 cDNA behind a loxP-flanked dsRed stop cassette. T cells were transferred into recipient mice, and tamoxifen was applied 4 d after immunization when Tfh cells had already developed in vivo (Fig. 2A). Analysis of Ag-specific CXCR5⁺PD-1⁺ Tfh cells in draining lymph node revealed that their frequency and absolute number were decreased by almost 70% only 3 d after tamoxifen-induced Bach2 overexpression (Fig. 2B). The functional consequences of the Bach2-mediated loss of Tfh cells for the GC reaction were analyzed by cotransferring B cells specific for NP and immunization of recipient mice with cognate Ag. Within 6 d of Bach2 overexpression in Ag-specific T cells, the number of Ag-specific PNA⁺GL7⁺ GC B cells was reduced by 80% (Fig. 2C).

Because Bach2 has been previously described to be important for the homeostasis of regulatory T cells (11, 12), we tested whether Bach2 overexpression might increase the frequency of regulatory or follicular regulatory T cells in our system. However, intracellular staining for Foxp3 in transgenic T cells transfected with the Bach2 overexpression vector did not reveal any differences compared with the control group (Supplemental Fig. 1).

To confirm the Bach2-mediated regulation of Tfh cells by a reciprocal approach, we tested whether the KO of Bach2 would be able to increase the numbers of Tfh cells. We used T cells from Cas9 and Smarta TCR–transgenic mice and retroviral delivery of a Bach2-specific guide RNA (Fig. 2D). This system was demonstrated to be highly efficient, with KO frequencies up to 90% (17). Draining lymph nodes from recipients of transgenic T cells were analyzed by flow cytometry on day 6. In line with the reduced Tfh frequencies upon Bach2 overexpression, CRISPR/Cas9-mediated KO of Bach2 conversely resulted in increased frequencies of transgenic Tfh cells in vivo (Fig. 2E).

There are several possible explanations for the reduction of Tfh cells upon Bach2 overexpression: 1) decreased survival of Tfh cells, 2) decreased proliferation, 3) migration out of the draining lymph node, or 4) loss of the Tfh cell phenotype. To address the effect of Bach2 overexpression on apoptosis, we determined the frequency of Smarta TCR–transgenic T cells positive for active caspase-3, a central protease involved in the apoptosis pathway. No difference was observed between the Bach2 overexpression and control group 36 h after induction of Bach2 overexpression (Fig. 3A). In particular, the frequencies of active Caspase-3–positive Tfh cells were almost identical. As a second readout for cell survival, the expression ratio of the antiapoptotic factor Bcl-2 and the proapoptotic factor Bim (Bcl2L11) was determined by flow cytometry. Again, no differences were observed between the two groups, neither for total transgenic T cells nor for Tfh cells (Fig. 3B). Therefore, the reduced number of Tfh cells is not the result of enhanced apoptosis.

The proliferation of Ag-specific T cells was assessed by incorporation of the thymidine-analogue BrdU. BrdU application was started together with tamoxifen-induced Bach2 overexpression, and mice were analyzed 36 h later by flow cytometry. BrdU incorporation was the same between all transgenic T cells or Tfh cells overexpressing Bach2 and control cells (Fig. 3C), ruling out any proliferative effects as explanation for the reduced numbers of Tfh cells.

To test whether Bach2 overexpression directs Tfh cells to selectively migrate out of the draining lymph node, we treated recipient mice with the sphingosine-1-phosphate receptor antagonist FTY720 parallel to tamoxifen-induced Bach2 overexpression. This compound effectively blocks lymphocyte egress from lymph nodes, which was confirmed by the presence of only very low numbers of transgenic T cells in the peripheral blood and their accumulation in the draining lymph node. At the same time, the frequency of transgenic T cells present in the lymph node remained the same after Bach2 overexpression compared with the control group (Fig. 3D). Importantly, FTY720 treatment was not able to prevent the Bach2 overexpression-induced loss of CXCR5⁺/PD-1⁺ Tfh cells (Fig. 3E).

Taken together, these experiments show that neither increased apoptosis, reduced proliferation, nor egress out of the lymph node cause the loss of Tfh cells upon Bach2 overexpression. Instead, it can be concluded that low levels of Bach2 are required to prevent dedifferentiation of Tfh cells into non-Tfh effector cells.

To identify downstream targets of Bach2, which directly interfere with the stability of the Tfh cell phenotype, we performed global transcriptome analysis. Based on expression of CXCR5 and PD-1, Smarta TCR–transgenic Tfh and non-Tfh cells were sorted ex vivo 18 h after tamoxifen-induced Bach2 overexpression. At that time, transgenic Bach2 protein levels could be expected to be already high. However, the Tfh phenotype, as defined by expression of CXCR5 and PD-1, had not changed yet, enabling us to define early events of the Tfh phenotype loss and to find genes directly regulated by Bach2. The transcriptome of Tfh cells transduced with the Bach2 overexpression vector was compared with cells transfected with control vector by RNA sequencing. Because Bach2 is known as a transcriptional repressor, we focused on genes that were downregulated in Tfh cells by Bach2 overexpression. To further narrow down the number of 110 genes meeting this first criterion, we compared them with genes upregulated in Tfh versus non-Tfh cells, meaning they are likely to promote the Tfh cell phenotype. This finally resulted in a list of only 16 genes that were both Tfh signature genes and downregulated by Bach2 (Fig. 4A). The top three hits of this list were Bcl-6, the coinhibitory cell surface receptor TIGIT, and IL-21, the main effector cytokine of Tfh cells. Further genes suppressed by Bach2 included heme oxygenase 1 (Hmox1), which is able to degrade heme. As heme can inactivate Bach2 (33), the Bach2-mediated inhibition of Hmox1 appears to create a negative feedback loop.

Expression of the three top hits in the list (Bcl-6, TIGIT, and IL-21) was further analyzed by flow cytometry (Fig. 4B). This analysis did not only confirm the highly differential expression of all three molecules between Tfh and non-Tfh cells, but also demonstrated their suppression by Bach2.

To functionally test whether suppression of one of our candidate genes would result in a similar loss of the Tfh cell phenotype as observed with Bach2 overexpression, we performed CRISPR/Cas-mediated KO of the top genes from the list shown in Fig. 4A. T cells from Smarta Cas9 transgenic mice were retrovirally infected with appropriate guide RNAs (see Fig. 2D). Successful deletion of genes was verified by sequencing of the targeted region (Supplemental Table II). Transduced cells were transferred into recipient mice, and the frequency of Ag-specific Tfh cells was determined on day 6 after immunization. Only the KO of Bcl-6 resulted in a strong reduction in Tfh cells, whereas KO of the remaining genes did not have any effect (Fig. 5A).

To confirm this result, we additionally tested whether overexpression of Bcl-6 together with Bach2 could rescue the Bach-induced loss of the Tfh cell phenotype. In this test, we also included TIGIT, which was among the most strongly Bach2-downregulated targets and is a molecule with a long-known differential expression but still undefined function for Tfh cells (34, 35). The Bach2 overexpression plasmid was cotransfected with either Bcl-6 or TIGIT overexpression plasmids into Smarta T cells. Both plasmids contained different fluorescence reporters (GFP and mAmetrine, respectively), so that single- and double-transfected cells could be easily discriminated. Transduced cells were transferred into recipient mice, which were analyzed 6 d after immunization for Ag-specific Tfh cells. As expected, single overexpression of Bach2 resulted in an almost complete loss of Tfh cells (Fig. 5B). With the constitutive overexpression vector, this effect was even more pronounced than with the inducible vector (compare Fig. 2B). As shown before (see Fig. 4B), Bach2 overexpression also resulted in a substantial downregulation of TIGIT on the cell surface (Fig. 5B). Simultaneous overexpression of TIGIT restored the expression almost to the same level observed in the empty vector control. However, TIGIT overexpression was not able to rescue the Tfh cell phenotype (Fig. 5B). This clearly shows that TIGIT is not the Bach2 downstream target responsible for the observed loss of the Tfh cell phenotype. However, this does not rule out that the Bach2-induced downregulation of TIGIT still might have a role for specific effector functions of Tfh cells.

For overexpression of Bcl-6, the results of the rescue assay were very different. As expected, overexpression of Bcl-6 alone resulted in a substantial increase in Tfh cells, with almost 70% of all transgenic Smarta T cells displaying a CXCR5⁺PD-1⁺ Tfh cell phenotype. Vice versa, Bach2 overexpression reduced Tfh cell frequencies almost 10-fold compared with the empty vector control. Importantly, simultaneous Bach2 and Bcl-6 overexpression resulted in a highly significant rescue of the Tfh cell population (Fig. 5C). The partial recovery of Tfh cell frequencies goes in line with the incomplete recovery of Bcl-6 expression levels in the Bach2/Bcl-6 double-overexpression group. The rescue of the Tfh cell population by additional Bcl-6 overexpression shows that suppression of Bcl-6 by Bach2 is the mechanism for how Bach2 regulates Tfh cells.

To determine whether Bach2 directly binds in the Bcl-6 promotor, we used publicly available Bach2 ChIP sequencing data from in vitro–stimulated, wild-type, and Bach2-deficient CD4⁺ T cells (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE65084). These data showed a strong enrichment for Bach2 binding within a region of 1 kb immediately in front of the transcriptional start (Fig. 6A). Bach2 is known to bind to AP-1 motifs (10), and bioinformatic analysis revealed that this region contains several potential Bach2 binding sites (Fig. 6B).

Bach transcription factors typically form heterodimers with members of the small MAF transcription factor family (MafF, MafG, and MafK) (36). However, recently, it has been shown that Bach2 can also form a dimer with another AP-1 family transcription factor, Batf, which binds to palindromic AP-1 motives in the IL-4 promotor (37). The Bach2/Batf dimer displaced an activating complex of Batf, Irf4, and JunD, which resulted in a shutdown of gene expression. Interestingly, it has been shown that Batf binds to the Bcl6 gene in the same 1-kb region upstream of transcription start as Bach2 to regulate gene expression (38). At the same time, it has been demonstrated that Irf-4 promotes Bcl-6 expression and the Tfh cell differentiation program (39, 40). Therefore, competitive gene regulation similar to as demonstrated for the Il4 gene is conceivable for the suppression of Bcl-6 by Bach2. Indeed, analysis of ChIP sequencing data (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE39756) revealed that both Batf and Irf4 bind to the same region in the Bcl-6 promotor (Fig. 6C). To investigate whether Bach2 expression resulted in a replacement of Irf-4 at this site, Irf-4–FLAG-expressing T cells were additionally transduced with a Bach2 expression vector and cultured under Tfh cell–promoting conditions; in the presence of IL-21, IL-6 and Abs against IL-12, IL-4, TGF-β, and IFN-γ as previously described (30). Analysis by ChIP and RT-PCR confirmed that Irf-4 binding to the Batf/Bach2 binding site in the Bcl-6 gene was abrogated by Bach2 overexpression, indicating that Bach2 replaces Irf-4 to form an inhibiting complex with Batf instead (Fig. 6D).

To finally demonstrate that Bach2 not only binds to, but also functionally suppresses Bcl-6 transcription, a luciferase assay with the Bcl-6 promotor was performed in EL4 T cells. Cells were transfected with expression plasmids for Batf and Irf4 and either an empty control vector or a Bach2 expression vector to test whether Bach2 can replace Irf4 in the activating Batf/Irf-4 complex. Indeed, Bach2 cotransfection resulted in a significant reduction of luciferase activity (Fig. 6E). Importantly, this displacement effect was only visible when EL4 T cells were additionally transfected with Batf and Irf4. These data suggest that Bach2 functions as a suppressor of Bcl-6 expression by displacing an activating Batf/Irf4 complex in the Bcl-6 promotor.

Bach2 has been known as an important transcriptional regulator of B cells for more than 20 y. However, only recently it was discovered that Bach2 is also expressed in T cells (11–13). These studies demonstrated that Bach2 has a critical role for regulatory T cells. Bach2 KO mice had substantially reduced Foxp3⁺ regulatory T cells, which resulted in an autoinflammatory phenotype (11, 12). Moreover, a second study showed increased IL-4 production by effector T cells, indicating a specific function of Bach2 for Th2 cells (13). However, other Th subsets have not been analyzed so far.

Our data now indicate that Bach2 has much more complex and specific roles for different Th cell subsets. The exceptionally high expression of Bach2 in Th1 and Th17 cells indicates that these two Th subsets rely on high Bach2 expression similarly to regulatory T cells. To the contrary, Tfh cells show very low expression levels of Bach2 and functionally depend on these low levels to maintain their phenotype. This is in stark contrast to GC B cells, which share Bcl-6 with Tfh cells as lineage-defining transcription factor and require high Bach2 levels to maintain their phenotype by preventing upregulation of the antagonizing transcription factor Blimp-1 (32).

How can the different function of Bach2 for the transcriptional network in Tfh versus GC B cells be explained? Our data show that, contrary to the regulation in GC B cells, Bach2 in Tfh cells does not regulate Bcl-6 indirectly via repression of Blimp-1. Instead, Bach2 is downregulated to prevent direct inhibition of Bcl-6 by binding to its promotor region. The mechanism responsible for the suppressive effect of Bach2 appears to be the displacement of an activating Batf/Irf4 complex at the same promoter element. This model also provides an explanation why Bach2 is able to exert different functions in GC B cells compared with Tfh cells. In B cells, Irf4 induces Bcl-6 expression not only in a complex with Batf, but is also able to form activating complexes with PU.1, a factor that is not abundant in T cells (41). The PU.1/Irf4 complexes bind to distinct motifs in the promotor that are not susceptible to competitive displacement by Bach2. Consequently, B cells are in contrast to Tfh cells, able to coexpress Bach2 and Bcl-6.

Considering that Bcl-6 is the lineage-defining and functionally central transcription factor of Tfh cells, surprisingly little is known about its direct regulation. Although several transcription factors have been identified to promote early Tfh cell generation, only Tcf-1 and Bob1 have been shown to directly regulate Bcl-6 by binding to its promotor (42–44). In contrast, only STAT5 and Foxo1 were demonstrated to suppress Bcl-6 by direct binding to regulatory elements (45, 46). With Bach2, we have identified a novel potent negative regulator of Bcl-6 that is critical for Tfh cell maintenance.

A KO of Bach2 does not only result in autoinflammatory symptoms in mice, but genetic polymorphisms within the Bach2 locus were associated with several autoimmune and allergic diseases in humans (47–54). Moreover, a recent study demonstrated that haploinsufficiency of Bach2 resulted in a multiorgan autoimmune disease (55). Because autoimmune diseases are generally associated with increased frequencies of Tfh cells (3), it is tempting to speculate that reduced Bach2 levels might be related to this disease phenotype.

Our investigation demonstrates that overexpression of Bach2 as a single factor is sufficient to completely reverse the phenotype of established Tfh cells, resulting in the breakdown of the GC reaction. This provides a mechanism for how a physiological Tfh cell response may be limited on a transcriptional level. Further investigations will be required to determine the external and internal cues directing upregulation of Bach2 in a physiological immune reaction. In the context of dysregulated Tfh cell responses, the underlying cause of most Ab-driven autoimmune diseases, this is an interesting pathway to be targeted in therapeutic strategies. Affected patients require therapeutic strategies not only to prevent the de novo generation of Tfh cells, but rather to eliminate already existing autoreactive Tfh cells.

We thank Joseph E. Craft (Yale University School of Medicine, New Haven, CT) for generously providing IL-21 reporter mice and Klaus Rajewsky and Van Trung Chu (Max Delbrück Center for Molecular Medicine, Berlin) for providing Cas9-transgenic mice and for valuable advice on the CRISPR/Cas9 system. We further thank Gitta Anne Heinz, Katrin Lehmann, Heidi Schliemann, and the team of the animal facility for support.

The authors have no financial conflicts of interest.