Nfatc1 ^caaA (c.n.Nfatc1) (36), Nfatc1 ^fl/fl (Nfat2 ^fl/fl) (34, 37) and/or Prdm1 ^fl/fl (Prdm1^floxlflox ) (38) mice were crossed to FIC (Foxp3-IRES-Cre) (39) in order to generate Nfatc1 ^caaA.FIC, Nfatc1 ^fl/fl.FIC, Prdm1 ^fl/fl.FIC and Nfatc1 ^fl/fl .Prdm1 ^fl/fl.FIC mice. Further breeding with reporter mice Prdm1 ^+/gfp (Blimp ^gfp/+) (40) or R26R3-YFP (41) generated Nfatc1 ^caaA.Prdm1 ^+/gfp.FIC, Prdm1 ^fl/gfp.FIC, Nfatc1 ^caaA.Prdm1 ^fl/gfp.FIC and R26R-YFP.FIC, respectively. Nfatc1/Egfp (42) and DEREG (43) have been described. All mice, male or female, were bred and maintained on a C57BL/6J background at the ZEMM, University of Würzburg.

Mice were immunized i.p. with 125 µg NP-KLH (4-Hydroxy-3-Nitrophenylacetyl hapten conjugated to Keyhole Limpet Hemocyanin (24–32) (Biosearch) emulsified (1:1) in ImJect® Alum (ThermoScientific) and boosted on day 7.

RNA was extracted using RNeasy Micro Kit (QIAGEN) followed by cDNA synthesis with the iScript II Kit (BioRad). Real-time qRT-PCR was carried out with an ABI Prism 7700 detection system using following primers: Nfatc1 GATCCGAAGCTCGTATGGAC plus AGTCTCTTTCCCCGACATCA, Nfatc1 P1 CGGGAGCGGAGAAACTTTGC plus CAGGGTCGAGGTGACACTAG, Nfatc1 P2 AGGACCCGGAGTTCGACTTC plus CAGGGTCGAGGTGACACTAG, Foxp3 GGCCCTTCTCCAGGACAGA plus GCTGATCATGGCTGGGTTGT, Bcl6 GATACAGCTGTCAGCCGGG plus AGTTTCTAGGAAAGGCCGGA, Prdm1 TAGACTTCACCGATGAGGGG plus GTATGCTGCCAACAACAGCA, Cxcr5 TCCTGTAGGGGAATCTCCGT plus ACTAACCCTGGACATGGGC, Hprt AGCCTAAGATGAGCGCAAGT plus TTACTAGGCAGATGGCCACA.

Flow cytometry staining was performed with the following antibodies: anti-B220-Percp (RA3-6B2, Biolegend), anti-B220-PE (RA3-6B2, Invitrogen), anti-CD11b-PE (M1/70, Biolegend), anti-CD11c-APC (N418, eBioscience), anti-CD21- Percp-cy5.5 (7E9, Biolegend), anti-CD23-BV510 (B3B4, Biolegend), anti-CD25-PE (PC61, Biolegend), anti-CD25-APC/cy7 (PC61,Biolegend), anti-CD3-APC (145-2C1, Biolegend), anti-CD3-BV421 (145-2C1, Biolegend), anti-CD4-BV510 (RM4-5, Biolegend), anti-CD4-PacificBlue (GK1.5, Biolegend), anti-CD4-Percp (RM4-5, Biolegend), anti-CD44-FITC (IM7, eBioscience), anti-CD44-PE/Cy7 (IM7, Biolegend), anti-CD62L-PE (MEL-14, Biolegend), anti-CD8-BV510 (53-6.7, Biolegend), anti-CD8-APC/Cy7 (53-6.7, Biolegend), anti-CXCR5-BV421 (L138D7, Biolegend), anti-Fas-PE (Jo2, BD Pharmigen^TM), anti-GITR-PE (DTA1, Biolegend), GL-7-FITC (GL-7, BD Pharmigen^TM), GL-7-PE (GL-7, Biolegend), anti-ICOS-PE/Cy7 (C398.4A, Biolegend), anti-IgD-APC (11-26c.2a, Biolegend), anti-IgM-PE/Cy7 (RMM-1, Biolegend), anti-Ly6G-BV510 (1A8, Biolegend), anti-PD1-APC (J43, eBioscience), anti-ST2-PE (DIH9, Biolegend), anti-Klrg1-PE/cy7 (2F1/KLRG1, Biolegend), Fc -receptors were blocked with anti-CD16/anti-CD32 (clone 93, Invitrogen). Intracellular Foxp3 staining (anti-Foxp3-PE (FJK-16s, Invitrogen), anti-Foxp3-FITC (FJK-16s, eBioscience) was performed with the eBioscience^TM Foxp3/Transcription Factor Staining Buffer Set (ThermoFisher). Life-/ dead discrimination was done with the Zombie-Aqua Fixable viability kit (Biolegend). For concurrent analyses of intracellular YFP and Foxp3 the following requirements were necessary (44): after dead cell staining and surface staining, the cells were pre-fixed with FA 1% (from methanol-free FA 16%, ThermoFisher) for 30 minutes at RT. The samples were then washed with 1X permeabilization Buffer (eBioscience™ Permeabilization Buffer 10X, ThermoFisher) and subsequent incubation with antibodies against intracellular proteins was performed overnight at 4°C. Samples were acquired at a FACS Canto II and analyzed with the FlowJo software (Tree star).

Liver was perfused through the vena cava with 10 ml ice-cold PBS. During perfusion, the hepatic portal vein was cut to allow outflow of the blood from the liver. Afterwards, liver was gently meshed through a 100 µm metal cell strainer into a 50 ml falcon. The pellet was washed twice with RPMI, centrifugated at 500 x g for 10 min (4°C). Separation of lymphocytes, hepatocytes and RBCs was done with 40 % / 80 % Percoll gradient centrifugation for 20 min at 2000 x g (4°C, no brakes). Hepatocytes float on the top of gradient, while the pellet contains RBCs. The middle phase contains lymphocytes, which were collected in a fresh 50 ml falcon, filled with RPMI. After another centrifugation step (1800 rpm, 5 min, 4°C) lymphocyte fraction was ready.

For the lung, thorax as well as abdominal cavity was exposed. Lung was perfused by opening the inferior vena cava. 10 ml of ice-cold PBS was flushed through the right ventricle of the heart until the lung turned colorless. Afterwards, lung was minced and transferred in a 50 ml falcon containing 10 ml of digestion buffer (1mg/ml Collagenase D, 20 µg/ml DNAse I, 5 mg/ml BSA, RPMI) to be incubated on a rotating shaker (37°C) for 20 min. Next, lung suspensions were filtered via a 100 µm filter into a new 50 ml tube with RPMI and centrifuged for 5 min with 300 x g (room temperature). Isolation on leukocytes was executed as described for the liver.

To isolate lymphocytes from the fat tissue, abdominal fat pads were carefully excised, and fat was cut into small pieces. For digestion, pieces were transferred into a 50 ml falcon containing 10 ml of digestion buffer (1 mg/ml Collagenase D, 20 µg/ml DNAse I, 20 mg/ml BSA, DMEM), incubated on a rotating shaker (37°C) for 40 – 45 min. To stop the digestion, 0.5 M EDTA-PBS was added; incubation for 2 min. Fat was centrifuged for 5 min with 300 x g (room temperature); the pellet contained the lymphocytes for further analyses.

The antibody titers for various isotypes in the sera of NP-KLH immunized mice were quantified relative to a sample pooled from sera of NP-KLH immunized mice. Nunc MaxiSorp™ plates (ThermoFisher) were coated with 1µg/ml of NP-(14)-BSA or NP-(2)-BSA (Biosearch). An initial dilution of 1:500 of the serum was prepared followed by a series of 1:5 dilutions. For the quantification of anti-ds-DNA-antibodies, high-binding half-area plates (Corning) were coated with Poly-L-lysin (Sigma-Aldrich), followed by 25 ng/ml calf thymus dsDNA (Sigma-Aldrich). An initial dilution of 1:10 of the sera was prepared, following a series of 1:2 dilutions. The following isotype-specific detection antibodies were used: donkey anti-mouse IgG-HRP (JacksonImmunoResearch), goat anti-mouse IgG1-HRP, anti-mouse IgG2b-HRP, anti-mouse IgG2c-HRP, anti-mouse IgG3-HRP, anti-mouse IgM-HRP (all from SouthernBiotech). Total quantification of IgG, was performed using a mouse-IgG-Kit (Roche) according to the manufacturer’s instructions.

mLN were extracted and fixed for 72 h in 4 % Formalin at RT. The formalin was changed every 24 h. The tissues were embedded in paraffin. Sections of 3 µm thickness were generated. Heat-induced antigen retrieval of tissue sections was performed in 20 mM citric acid buffer (pH 6.0). The following primary antibodies/conjugates were used: rabbit anti-CD3 (A0452 Dako), PNA-biotin (B-1075, VECTOR) and rat anti-Foxp3 (FJK-16s, ThermoFisher). For detection the following secondary antibodies were used: donkey anti-rat Alexa Fluor^TM 488, donkey anti-rabbit Alexa Fluor^TM 555, streptavidin Alexa Fluor^TM 647 all from ThermoFisher. The tissues were blocked with normal rat serum (NRS) (STEMCELL), in order to allow a third staining step with the rat anti-B220-AlexaFluor594 (RA3-6B2, Biolegend) antibody and Hoechst (Sigma-Aldrich). Tissues were mounted in Mowiol supplemented with DABCO (ROTH). Images were acquired at the Evos FL Auto 2 fluorescence microscope (ThermoFisher) and evaluated using the software Fiji (ImageJ) (45).

CD4⁺ cells from spleen and mLN of NP-KLH immunized mice 10 days i.p. were enriched using CD4 (L3T4) MicroBeads (Miltenyi) according to the manufacturer’s instructions. B220⁻CD4⁺BlimpGFP⁺CXCR5⁺GITR^hi T_FR cells were enriched via fluorescence-activated cell sorting (BD FACSAria II, VIM Würzburg) according to the sort strategy in Figure S8A . Cells were sorted in RLT-Buffer (QIAGEN) supplemented with 2-Mercaptoethanol (ROTH) and kept on dry ice or at -80°C until further processing.

RNA was purified with the RNeasy Plus Micro Kit according to the manufacturer’s protocol (QIAGEN). RNA was quantified with a Qubit 2.0 fluorometer (Invitrogen) and the quality was assessed on a Bioanalyzer 2100 (Agilent) using a RNA 6000 Pico chip (Agilent). Samples with an RNA integrity number (RIN) of > 8 were used for library preparation. Barcoded mRNA-seq cDNA libraries were prepared from 10ng of total RNA using NEBNext® Poly(A) mRNA Magnetic Isolation Module and NEBNext® Ultra™ II RNA Library Prep Kit for Illumina® according to the manual with a final amplification of 15 PCR cycles. Quantity was assessed using Invitrogen’s Qubit HS assay kit and library size was determined using Agilent’s 2100 Bioanalyzer HS DNA assay. Barcoded RNA-Seq libraries were onboard clustered using HiSeq® Rapid SR Cluster Kit v2 using 8pM and 59bps were sequenced on the Illumina HiSeq2500 using HiSeq® Rapid SBS Kit v2 (59 Cycle). The raw output data of the HiSeq was preprocessed according to the Illumina standard protocol. Sequence reads were trimmed for adapter sequences and further processed using Qiagen’s software CLC Genomics Workbench (v12 with CLC’s default settings for RNA-Seq analysis). Reads were aligned to GRCm38 genome. Heatmaps were generated using the online tool Morpheus https://software.broadinstitute.org/morpheus.

HA-Nc1-RSD, NFATc1/A-ER, NFATc1/C-ER, Blimp-1-Flag, Blimp-1-HA, Foxp3 and Nfatc1 P1 as well as Cxcr5 promoter /HS2 luciferase-reporter constructs have been described (27, 30, 46–48). Further HS2 mutations were introduced by overlapping oligos applied as primers in PCR ( Figure S2B ). The retroviral vector pBcl-6-Flag was constructed by cloning the complete cDNA of murine Bcl-6 into pEYZ/MCS-F (47) thereby directly fusing the Flag-peptide to the C-terminal end of Bcl-6.

EL-4 and HEK 293T cells were cultured in complete RPMI or DMEM medium containing 5 % and 10 % FCS, respectively (34). They were transiently transfected with Nfatc1 P1 or different Cxcr5 promoter luciferase-reporter constructs alone or in combination with plasmids encoding for NFATc1/A, NFATc1/C, Blimp-1 and Foxp3 using conventional calcium phosphate or PEI (Sigma Aldrich) for HEK 293T cells and standard DEAE Dextran for EL-4 cells. 36 h post transfection, luciferase activity was measured from the cells that were left untreated or treated with TPA (50 ng/ml), ionomycin (0.5 µM) o/n and relative light units were corrected for the transfection efficacy relative to total protein concentrations. Normalized mean values of at least 3 independent experiments are depicted in relative light units as fold activation over empty vector control.

HEK 293T cells were transiently transfected alone or in combination with expression plasmids coding for ER-tagged NFATc1/A and NFATc1/C or Flag-tagged Blimp-1 and its mutants ( Figure 2F ) (30, 47). Cells were lysed in IP lysis buffer (Thermo Scientific) and CoIP was performed as described earlier (30), using anti-ER (Santa Cruz) and anti-Flag (M2, Sigma) Abs. For CoIP of Blimp-1 and NFATc1 from primary T cells, 1x10⁷ tTreg and Tconv cells were activated and expanded with anti-CD3/CD28 beads (Invitrogen) for 7 days. CoIP was performed with the Nuclear Complex Co-IP Kit (Active motif) using anti-Blimp-1 (C14A4, cell signaling) and anti-NFATc1 (7A6, BD Pharmingen) Abs.

The transiently transfected HEK 293T cells were incubated for 24 – 48 h and stimulated with TPA (50 ng/ml), ionomycin (0.5 µM) and CaCl₂ (2 mM) for 4 h, when ER-tagged proteins were expressed, additionally with 4-hydroxytamoxifen (Tm, 200 nM; Sigma-Aldrich). Either whole cellular extracts were prepared using the ProteoJET Kit (Thermo Scientific) and EMSAs performed with radioactively labeled probes as described before (47), or nuclear extracts prepared using the Nuclear Extract Kit (Active Motif) or NE-PER^TM Nuclear and Cytoplasmic Extraction reagents (ThermoFisher). In that case, EMSAs were performed using the Gelshift^TM Chemiluminescent EMSA Kit (Active Motif) according to the standard protocol. Oligonucleotides were synthesized by Eurofins. DNA-probes (sense strang) were biotinylated, but competitors were left non-biotinylated. DNA-probes:

Nfatc1-P1-tan_s (5’ GGAAGCGCTTTTCCAAATTTCCACAGCG),

Nfatc1-P1-tan_a (5’ CGCTGTGGAAATTTGGAAAAGCGCTTCC),

Myc-PRE_s (5’ CGCGTACAGAAAGGGAAAGGACTAG),

Myc-PRE_a (5’ CTAGTCCTTTCCCTTTCTGTACGCG),

Cxcr5-pro-B_s (5’ AAAGAAAAGAAAAGAAAAGAAGGGGGAAAACACA),

Cxcr5-pro-B1_a (5’ TGTGTTTTCCCCCTTCTTTTCTTTTCTTTTCTTT),

Cxcr5-pro-N1_s (5’ GAAAAGACTCAGTGGAAAAAAAAAAAAAAAG),

Cxcr5-pro-N1_a (5’ CTTTTTTTTTTTTTTTCCACTGAGTCTTTTC),

Cxcr5-HS2-B_s (5’ GGGCAGCTGTGAGTGAAAGGTATG),

Cxcr5-HS2-B_a (5’ CATACCTTTCACTCACAGCTGCCC),

Cxcr5-HS2-N1_s (5’ GGAGCTGAGGAAACGCAGGTGC),

Cxcr5-HS2-N1_a (5’ GCACCTGCGTTTCCTCAGCTCC),

Cxcr5-HS2-N2_s (5’ GCCCCCTTCTTTTCCACTCAGAAAA),

Cxcr5-HS2-N2_a (5’ TTTTCTGAGTGGAAAAGAAGGGGGC),

Cxcr5-HS2-N3_s (5’ TAGGAGGCCATTTCCTCAGTTTCAG),

Cxcr5-HS2-N3_a (5’ CTGAAACTGAGGAAATGGCCTCCTA),

Competitors: Myc-PRE_s (5’ CGCGTACAGAAAGGGAAAGGACTAG),

Myc-PRE_a (5’ CTAGTCCTTTCCCTTTCTGTACGCG),

Il2-Pubd_s (5’ CAAAGAGGAAAATTTGTTTCATACAG),

Il2-Pubd_a (5’ CTGTATGAAACAAATTTTCCTCTTTG).

Supershifts were performed with anti-NFATc1 (7A6, SCBT), anti-ER (MC-20, SCBT) and anti-Blimp-1 (6D3, Biolegend). The assays were blotted onto a Roti®-Nylon plus membrane (ROTH). DNA-protein complexes were cross-linked onto the nylon membrane with an UV Stratalinker^TM 1800 (Stratagene) using the auto cross-link function.

ChIP was performed as before (27). In brief, ChIP-IT Express kit (Active Motif) was used according to the manufactures’ instructions, except enzymatic shearing followed by additional 25 min sonication. Following precipitating, 5 µg anti-Blimp-1 (C14A4, cell signaling) was used. Quantification of DNA-binding was carried out by real-time PCR using the following primers: Cxcr5 distal CTAGTATTCTTAGGGTTCTTCC plus GGGCACTTGATCAACCTGTG, Cxcr5 middle GGCTCGCCTGGGACTGAG plus GGGGCTAAGAAAAGAGTACTC, Cxcr5 proximal ACTGACTCTGTGGGGGGAG plus CTTGCCTCTCGACTCATCTC.

All results are shown as median with interquartile range. The statistical significance of the differences between the groups was determined via a Mann-Whitney and unpaired t tests. Results were calculated with the software Prism 5 (GraphPad). Differences for p-values > 0.05 were considered not significant, but p-values ≤ 0.05 as significant and indicated in figures as p ≤ 0.05 (*), p ≤ 0.01 (**), p ≤ 0.001 (***), p ≤ 0.0001 (****).

We had shown before that NFATc1 is highly expressed in T-follicular (T_FOLL) cells. After immunization, only in T_FR and not in T_FH cells NFATc1 expression is necessary for CXCR5 expression (27). Therefore, we rationalized that NFATc1 has to overcome a T_FR-specific repressor and Blimp-1 was the prime candidate (18–20). To verify Blimp-1 expression in CD4⁺CXCR5^hiPD-1^hiCD25^hi T_FR cells after immunization with NP-KLH, we made use of Prdm ^gfp mice (40) and indeed found a substantial amount of GFP, i.e. Blimp-1 expression in T_FR, but not in T_FH cells ( Figure 1A ). Comparable cells from immunized Nfatc1/egfp (42), also revealed clear NFATc1 expression in T_FR, although to a lesser extent than in T_FH cells, but still distinctly higher than in unstimulated CD4⁺CXCR5^—PD-1^— Tconv ( Figure 1A ). To evaluate if NFATc1 and Blimp-1 can simultaneously be present in nuclei of Tregs, we isolated Tregs from DEREG mice, which express GFP under the control of Foxp3 (43), stimulated them in vitro in the presence of IL-2 for 24 h and stained them for confocal microscopy. Reassuringly, Blimp-1 did not exclude NFATc1 from the nucleus and vice versa ( Figure 1B ). To this end, conditional Nfatc1 ^fl/fl and Prdm1 ^fl/fl mice were crossed to Foxp3-IRES-Cre (FIC) mice (34, 37, 39, 49), creating mice with Tregs ablated for NFATc1, Blimp-1 or both. All mice were and stayed healthy for at least 6 months. Upon immunization, we defined PD-1⁺CXCR5⁺ T_FOLL cells ( Figure 1C ). T_FR cells were underrepresented in the T_FOLL population of Nfatc1 ^fl/fl.FIC mice as described before (27). On the contrary, ablation of Blimp-1 in Tregs (Prdm1 ^fl/fl.FIC mice) caused a substantial shift towards T_FR on the expense of T_FH cells ( Figure 1D ). When NFATc1 was deleted additionally to Blimp-1 in Tregs, the T_FR/T_FH ratio appeared fairly normalized as if NFATc1 could no longer trigger overshooting CXCR5 expression on T_FR cells, which was unrestrained in the sole absence of Blimp-1.

As we found before, NFATc1 binds to a consensus site in the proximal Cxcr5 promoter, which transmits transactivation (27). Since Blimp-1 and NFATc1 demonstrated a reciprocal influence on CXCR5⁺ T_FR cells, we were wondering if Blimp-1 is equally able to bind to the promoter. ChIP assays with all splenic CD4⁺ T cells (Tconv and Treg) from NP-KLH-immunized WT mice suggested that Blimp-1 engages at the same proximal region as NFATc1 (27) ( Figure 2A ). Although a complete core consensus sequence, (AGn)GAAAG (50, 51), is not present in the proximal promoter, electromobility shift assays (EMSA) with nuclear extracts from expression vector-transfected HEK 293T cells revealed binding to a stretch of repetitive A and G nucleotides ( Figures 2B, C ). The binding pattern was comparable to the known Blimp-1 response element in the Myc promoter. The consensus sequences for Blimp-1 and NFAT are very similar, but NFATc1 did not recognize the Blimp-1-recruiting oligonucleotide and Blimp-1 not the formerly defined NFATc1 site. Thus, we termed the sites Cxcr5-pro-B and -N1.

Still, it appeared that the presence of Blimp-1 enhanced the binding of NFATc1 at Cxcr5-pro-N1 (6^th lane compared to 2^nd of the lower right gel). Therefore, we tested if NFATc1 and Blimp-1 would be able to interact. Indeed, we found that both NFATc1/αA and NFATc1/βC co-immunoprecipitated (CoIPs) Blimp-1 from extracts of transiently transfected HEK 293T cells ( Figure 2D ). Using primary T cells, we could verify a direct interaction between NFATc1 and Blimp-1 in nuclei of tTregs, but not in Blimp-1⁻ Tconv cells ( Figure 2E ). In additional Co-IPs, we compared full length Blimp-1 with a naturally occurring version (Δ7) devoid of the first two, DNA-binding Zn-fingers (47) and several deletion constructs. We included N-terminally truncated Blimp-1 starting C-terminal to the proline-rich (Pro) region (52) and the mirroring C-terminally deleted one with or without the Pro domain, furthermore a short C-term consisting of the Zn and the acidic region only ( Figure 2F right). Those partial Blimp-1 proteins revealed NFATc1 interaction to rely on the Pro domain of Blimp-1 ( Figure 2F ). This suggests that NFATc1 is masking one HDAC2-interacting domain of Blimp-1 (52), thus constraining Blimp-1-mediated repression. Consistently, in HEK 293T cells NFATc1 not only transactivated the Cxcr5 promoter (1127 bp) if Cxcr5-pro-N1 was intact, but was supported by the presence of Blimp-1 ( Figure 2G ). Exogenous Blimp-1 even restored the activity of NFATc1 irrespective of the Cxcr5-pro-N1 mutation. This hints to the recruitment of NFATc1 via its response element and via protein interaction with Blimp-1. Interestingly, overexpression of Blimp-1 was not sufficient to repress CXCR5-controlled luciferase expression ( Figure 2G ).

The luciferase reporter construct also contained an enhancer derived from the first Cxcr5 intron and originally found as a DNase I hypersensitive site. We had already shown the enhancer quality of this ‘HS2’ (27) and recently a consensus sequence-encompassing Blimp-1 response element has been described within (19). A snapshot of the Cxcr5 locus with own and uploaded ChIPseq data (37, 48, 53) documented not only Blimp-1 binding (in plasma blasts and CD8⁺ T cells), but also NFATc1 / NFATc2 binding (in CD8⁺ T cells) to HS2, which seemed even more potent than to the promoter ( Figures 3A , S1A ). Blimp-1 also attaches strongly to an upstream region, but here we focused on the interplay of NFATc1 and Blimp-1, which appeared prominent at the HS2 enhancer. Neighboring NFAT and Blimp-1 binding occurs also in other gene loci like Pdcd1, Il2, Il2ra, Il10, Ctla4, and Nfatc1; whereas other loci like Myc, Tigit, or Dnmt3a exhibit distant NFAT and Blimp-1 ChIPseq peaks ( Figures S1B–J ).

Since the Blimp-1 site within Cxcr5 HS2 had been verified (19), we sought to evaluate putative NFAT response elements. We found five candidates by a computational search, of which three are conserved between mice and men ( Figures 3B and S2A ). When those three were subjected to an EMSA, the middle one, named Cxcr5-HS2-N2, responded with a considerable shift comparable to the one at the Nfatc1 P1 tandem site and displaceable by the ‘cold’ NFAT response element Pubd from the Il2 promoter (31, 54) ( Figure 3C ). Luciferase reporter assays in transiently transfected EL-4 cells with an expression plasmid for NFATc1 and constructs containing the proximal Cxcr5 promoter [329 bp (27)] and the HS2 enhancer, wild typic and / or NFAT binding site-mutated, demonstrated that both elements take part in NFATc1-mediated transactivation ( Figures 3D and S2B ). As at the promoter, NFATc1 as well as Blimp-1 recognized solely their respective response elements from the HS2 ( Figure 3E ). Further Cxcr5-luciferase reporter assays with HS2-mutated in N2, B or N2+B revealed that eradication of the consensus Blimp-1 site within the HS2 enhancer unleashed expression ( Figures 3F and S2B ). However, this was only observable if pro-N1 and/or HS2-N2 were intact. In fact, all activity was lost upon mutation of HS2-B in combination of both, promoter- and enhancer-derived NFAT sites. Altogether, these experiments demonstrated that NFATc1 and Blimp-1 cooperate to ensure an adequate regulation of Cxcr5 expression and that, while NFATc1 clearly transactivates Cxcr5 via promoter and HS2-located response elements, Blimp-1 plays an ambivalent role as a repressor simultaneously supporting the recruitment of NFATc1.

Although Blimp-1 is dominantly expressed in T_FR and not in T_FH cells, NFATc1 appeared to be far more pronounced in T_FH than in T_FR cells ( Figure 1A ). To investigate the intrinsic role of Blimp-1 and NFATc1 for T_FR cells in vivo, we first determined how exclusive the Cre activity of FIC mice is for Tregs, especially since other Foxp3-driven Cre lines are rather leaky (55). We created R26R-YFP.FIC mice by crossing FIC to the R26R-YFP reporter mouse deleting the STOP cassette and setting free YFP expression in Foxp3⁺ cells (39, 41). Immune cells in thymus, LNs and spleen of nine-week-old R26R-YFP.FIC were analyzed thoroughly. CD4CD8 double-negative (data not shown) as well as double-positive thymocytes did not express YFP, but CD4⁺Foxp3⁺ Tregs started to be positive (20 %; Figure S3A ). Thymic and peripheral mDCs, pDCs, eosinophils, neutrophils, inflammatory and resident monocytes, however, were devoid of any YFP⁺ cells ( Figure S3B ). Similar, developmental and subtype stages of B cells were negative for YFP expression in LNs and spleen ( Figure S3C ). This notion extended to GC-B cells ( Figure S4A ). All CD8⁺ T cells proved to be negative as well ( Figure S4B ).

The majority of CD4⁺CD25⁺ as well as CD4⁺CD25⁻ Tregs were YFP⁺ in mLN, the combined peripheral LNs and the spleen ( Figure S4C ). However, some CD4⁺CD25⁻Foxp3⁻ Tconv also elicited YFP expression. This pattern was reflected by follicular CD4⁺ICOS⁺CXCR5⁺ T cells. While most T_FR cells documented FIC-mediated Cre activity, this was also true for a substantial number - up to a quarter in peripheral LNs - of CD4⁺CD25⁻Foxp3⁻ T_FH cells ( Figure S4D ). We reasoned that this could be due to ex-Tregs / ex-T_FR cells (12) and evaluated YFP in non-T_FH ( Figure S4E ). Indeed, while only a small fraction of the abundant naive ICOS⁻ Tconv was YFP⁺, this was enriched in the few activated ICOS⁺ and especially in sole CXCR5⁺ CD4⁺ T cells. Thus, for the analyses of FIC mice it had to be taken into account that, while different from other Foxp3-Cre lines B and CD8⁺ T cells stayed untouched, gene-edited Tconv and here especially T_FH cells – possibly resembling ex-Tregs/T_FR cells - contribute to the measured effects.

The role of Blimp-1 as a T_FR-restraining factor has been implicated early (9) and our Treg-specific Blimp-1 ablation demonstrated boosted numbers of CXCR5⁺ T_FR cells, which we had measured relative to T_FH cells within the T_FOLL population after immunization ( Figure 1D ). Additionally, we wondered about the frequency of T_FR cells within the Treg population and applied a different gating strategy ( Figure 4A ). Supporting the former data, immunization of Prdm1 ^fl/fl.FIC mice caused an enhanced differentiation towards B220⁻CD4⁺CD44⁺Foxp3⁺CXCR5⁺PD1⁺ T_FR cells ( Figures 4B, C ). The CD4⁺CD25⁺Foxp3⁺ T_FR population was already positively affected without immunization in steady state ( Figure S5A ). CXCR5 expression per Treg was not significantly enriched compared to WT, but higher than on NFATc1-deficient Tregs ( Figure 4D ).

Reciprocally, the frequency of T_FH cells was determined independently from T_FR cells within the activated CD44⁺CD4⁺ Tconv population ( Figure 4A ). While loss of NFATc1 and a consequential decrease in CXCR5⁺ T_FR cells led to more T_FH cells indicating a shortfall in GC control (27), the excess of Blimp-1-deficient T_FR cells did not cause a gain in suppression, i.e. the expected reduced frequency of T_FH cells ( Figures 4E, F ). It rather appeared as if the T_FH population would expand in the presence of Blimp-1-ablated Tregs. Nevertheless, the robust surplus of Blimp-1-deficient T_FR cells still shifted the balance of T_FR / T_FH in favor of T_FR cells ( Figure 4G ), which would have suggested a better control with the former gating strategy. Now we strongly suggest that Blimp-1 is certainly restraining the number of Foxp3⁺ cells within GCs, but is necessary for their functional competence.

T_FR cells limit the magnitude of the GC reaction by direct and indirect repression of B cells, i.e. the number of GC-B cells and the quantity and quality of secreted immunoglobulins (10, 56–58). Thus, we evaluated the number of B220⁺GL-7⁺Fas⁺ GC-B cells ( Figure 5A ) and found a similar picture as for T_FH cells. In line with less T_FR cells due to NFATc1 ablation, more GC-B cells could be detected. This was also true in Prdm1 ^fl/fl.FIC and in Nfatc1 ^fl/fl.Prdm1 ^fl/fl.FIC mice ( Figures 5B, C ) albeit their enhanced frequencies of CXCR5⁺Foxp3⁺ T_FR cells ( Figures 4B, C ). It was reflected in significantly higher titers of antigen-specific IgM as well as total IgG or IgG subclasses and even of overall IgG ( Figures 5D, E and Figure S6A ). The affinity of antibodies did not drop, but rather raised, while anti-dsDNA auto-antibodies did not occur to a major extent ( Figures 5F, G ). It is noteworthy to mention that NFATc1-Blimp-1 double-deficiency in Tregs resulted in the most prominent rise in antibody titers as if in the absence of Blimp-1 the loss of NFATc1 additionally affected the function of T_FR cells.

To assess if elevated GC responses and Ab production was due to a limited ability of Treg cells to migrate into B-cell areas in the absence of NFATc1, we evaluated the localization of the differentially ablated Tregs within the follicle as well as within the GC itself. As expected, NFATc1-deficient Tregs were less abundant in both areas, whereas Blimp-1-deficient Tregs were prominently detectable in follicles and GCs ( Figures 5H–J ). With regard to homing, double-deficient Tregs of immunized Nfatc1 ^fl/fl.Prdm1 ^fl/fl.FIC, however, behaved like WT Tregs. This was in line with normalized numbers of T_FR cells due to NFATc1 ablation in absence of the repressor Blimp-1 and reduced function due to loss of Blimp-1 cumulating in a severe failure of GC control.

NFATc1 was well expressed in T_FR cells, although not as pronounced as in T_FH cells ( Figure 1A ). We had demonstrated before that cTregs only express the P2-originating long isoforms of NFATc1 (30), but were wondering whether T_FR cells – as one type of effector Tregs (eTreg) – express the inducible short isoform NFATc1/αA, which could contribute to the heightened level of NFATc1. We sorted WT T_FR cells by means of Blimp-GFP expression ( Figure 6A ) and verified cell type-specific Prdm1, Bcl6 and Foxp3 RNA expression ( Figure S6B ). Nfatc1 RNA was clearly present in CD4⁺CXCR5^hiPD1^hi T_FH and CD4⁺CXCR5^hiPD1^hiGFP⁺ T_FR cells although again more prominent in T_FH than in T_FR cells ( Figure 6B , compare to Figure 1A ). Remarkably, in both T_FOLL types, P1 transcripts dominated which lead to NFATc1/αA expression (32).

Accordingly, we exogenously expressed NFATc1/αA Treg-specifically in a constitutive active form [caNFATc1/αA; originally c.n.NFATc1 (36)]. For unexplored reasons, we received only minor numbers of offspring, but mice elicited normal distribution of CD4⁺ and CD8⁺ thymocytes and thymic Tregs ( Figures S6C, D ). The frequency of peripheral Tregs was clearly enriched ( Figure S6D ). Intriguingly, both CD4⁺ and CD8⁺ T cells of these mice displayed an activated (CD44⁺CD62L^–) phenotype in peripheral lymphoid organs, indicating that caNFATc1/αA-expressing Treg cells – irrespective of their elevated number – could be less suppressive ( Figures S6E, F ).

Upon immunization, Nfatc1 ^caaA.FIC Tregs exhibited a robust heightened CXCR5 expression per cell ( Figure 6C ). This was not followed by gain in T_FR frequencies, although the relative number of Tregs within the CD4⁺ population was still increased ( Figures 6D, E ). The frequency of T_FH cells was fairly stable ( Figure 6F ). In line with elevated CXCR5 expression per cell, Foxp3⁺ T cells altered their homing behavior and now preferentially migrated into the GC itself ( Figures 6G, H ). Thus, WT NFATc1 expression is sufficient for differentiation of CXCR5⁺ T_FR cells, but the level of CXCR5, hinging on the level of NFATc1 or even NFATc1/αA, determines the sub-localization of T_FR cells.

Although immunized Nfatc1 ^caaA.FIC mice showed only a moderately reduced GC-B-cell frequency ( Figure 7A ), antigen-specific IgM and IgG were significantly reduced ( Figures 7B–D ; Figure S6G ), suggesting that NFAT controls the quality and not the quantity of humoral immune responses. The affinity might have increased ( Figure 7E ) and the amount of anti-dsDNA-specific antibodies decreased ( Figure 7F ), although those effects did not reach significance.

Finally, to reveal possible functional NFATc1-mediated changes, we determined the transcriptome of T_FR cells. For the most distinct effect, we decided to compare NFATc1-deficient with NFATc1/αA-overexpressing T_FR cells and further explored, whether caNFATc1/αA-mediated effects depended on the presence of Blimp-1. Frequencies of T_FR, T_FH and GC-B cells presented like before as T_FR cells from Nfatc1 ^caaA.Prdm1 ^fl/fl.FIC mice were enriched the most, but could not control the number of T_FH and GC-B cells due to Blimp-1 deficiency ( Figures S7A–C ).

In parallel, we characterized the Treg compartment of these mice, when still young. We determined their frequencies in spleen, combined peripheral LN and mLN as well as in non-lymphoid liver, lung and visceral fat tissue (VAT) ( Figure S7D ). Mostly, overexpression of NFATc1/αA in absence of Blimp-1 increased the relative numbers of Tregs. However, differentiation towards a PD-1⁺Klrg1⁺ eTreg or ST2⁺Klrg1⁺ tissue Treg (tisTreg) phenotype was diminished in the absence of Blimp-1 and/or upon overexpression of NFATc1/αA ( Figures S7E, F ). NFATc1-ablated Tregs were normal in frequency with a tendency to more eTreg and tisTregs.

To isolate T_FR cells, we once again took advantage of Prdm1 ^gfp mice (40), crossed to the Nfatc1 ^fl/fl.FIC, Nfatc1 ^caaA.FIC and Nfatc1 ^caaA.Prdm1 ^fl/fl.FIC mouse lines and sorted CD4⁺ B220⁻GFP⁺CXCR5⁺GITR^hi cells after immunization ( Figure S8A ). The NGS data were filtered for genes that showed an expression value of more than five in at least one of the groups and differed more than twice in their expression between Nfatc1 ^fl/fl .Prdm1 ^+/gfp.FIC and Nfatc1 ^caaA .Prdm1 ^+/gfp.FIC or Nfatc1 ^caaA .Prdm1 ^+/gfp.FIC and Nfatc1 ^caaA .Prdm1 ^fl/gfp.FIC ( Figure S8B ; a selection shown in Figure 7G ). Exogenous expression of caNFATc1/αA reduced Jak2, Bcl2, Klrg1 and Il4 expression, while Il6st, Il1rn, Il10, Tigit and Ctla4 were enriched. All of these genes were rather low expressed in the additionally Blimp-1-deficient T_FR cells indicating that Blimp-1 regulates effector function in T_FR cells. Besides, the removal of Blimp-1 on the background of caNFATc1/αA overexpression changed the expression of multiple chemokine receptors. Without Blimp-1, expression levels of Ccr7, S1pr1, Ccr4, Ccr10, Ccr6, Cx3cr1 were more abundant, while those of Ccr3, Ccr8, and Ccr9 were reduced. Of note, since CXCR5 had to be part of the gating strategy, no dramatically altered Cxcr5 mRNA expression could be expected. Nevertheless, it became clear that Blimp-1 is highly involved in migration of T_FR cells, but also supports their effector phenotype. Similar to Cxcr5, two of the differentially regulated genes, i.e. Il10 and Ctla4, exhibited overlapping NFAT and Blimp-1 binding in ChIPseq data ( Figure 7H and Figures S1G, H ).

Alarmingly, Foxp3 mRNA expression was distinctly less in caNFATc1/αA⁺ compared to NFATc1⁻ T_FR cells and even less upon Blimp-1 deletion ( Figure 7I ). Measured as MFI in flow cytometry, this hold true on protein level for T_FR cells of Nfatc1 ^caaA.Prdm1 ^fl/fl.FIC mice ( Figure 7J ) and for the whole Treg population in steady state Nfatc1 ^caaA.FIC mice ( Figure S7G ). This pointed to the high risk of NFATc1/αA expression in Tregs, not present in WT cTregs (30). In T_FR cells, this could have been triggered by overexpression of the constitutive active version beyond the normal – although pronounced – endogenous level of NFATc1/αA ( Figure 6B ). Since NFATc1/αA levels are distinctly higher in T_FH cells, we determined whether Blimp-1 – or Foxp3 – could counteract NFATc1 upregulation and keep the level below that of T_FH cells. Indeed, both transcriptional regulators limited NFATc1-mediated P1 transactivation in a luciferase assay ( Figure 7K ). Thus, expression and function of NFATc1/αA and Blimp-1 are interconnected in T_FR cells ensuring the specific follicular effector Treg phenotype, while constraining the conversion to T_FH cells.