Follicular regulatory T cells control humoral autoimmunity via NFAT2-regulated CXCR5 expression

Because we had observed that NFAT2 has especially high expression in human T_FH cells (Rasheed et al., 2006), we addressed the expression of NFAT transcription factors in their murine counterparts. Microarray experiments revealed enhanced RNA expression of NFAT1 in CD4⁺ICOS^hiCXCR5^hi T_FH compared with activated CD4⁺ICOS^hiCXCR5^– T cells and/or CD4⁺ICOS^–CXCR5^– naive conventional T (T_conv) cells, but the family member most clearly up-regulated was NFAT2 (Fig. 1 A). Alternate definition of T_FH cells by CD4⁺PD-1^hiCXCR5^hi in comparison to CD4⁺PD-1^–CXCR5^– T_conv cells revealed the same dominance of NFAT2 by qRT-PCR, which was primarily caused by high Nfat2 P1 promoter activity (Fig. 1 B). More importantly, the P1 promoter was also more effective in CD4⁺PD-1^hiCXCR5^hi T_FH cells than in activated CD4⁺ICOS^hi T_conv cells (Fig. 1 B, lower graphs). Making use of BAC-transgenic Nfatc1/egfp reporter mice (Bhattacharyya et al., 2011; Hock et al., 2013), T_FH cells (defined as CD4⁺PD-1^hiCXCR5^hi or CD4⁺ICOS^hiCXCR5^hi) from KLH- or SRBC-immunized mice exhibited the highest green fluorescence (Fig. 1 C and not depicted). This was reflected by the strong nuclear appearance of NFAT2, similar to that found in activated T_conv cells (Fig. 1 D). Corresponding to the robust Nfat2 P1 activity, NFAT2/α isoforms were dominantly expressed and activated in T_FH like in T_conv cells. NFAT1 and NFAT4 were also found to be activated in CD4⁺PD-1^hiCXCR5^hi cells, but to a lesser extent (Fig. 1 E). Prominent NFAT2 and especially NFAT2/α expression was verified in human T_FH cells in chronically inflamed tonsils (Fig. 1, F and G), conclusively confirming NFAT2 as robustly expressed and activated in human T_FH cells (Rasheed et al., 2006).

Next, we tested the consequence of T cell–specific NFAT2 ablation for the GC reaction Nfat2^fl/fl x Cd4cre mice are healthy and show a fairly normal composition of lymphoid compartments (Vaeth et al., 2012). Nonetheless, we detected modest but considerably increased frequencies of T_FH cells in spleen and mesenteric LNs (mLNs) from NP-KLH–immunized Nfat2^fl/fl x Cd4cre mice compared with Cd4cre littermates being significant over the complete distribution of the datasets (Fig. 2, A and B). By PNA staining of histological sections, however, we observed an even more clearly pronounced GC response (Fig. 2 C and not depicted). Flow cytometry reinforced a boosted frequency of B220⁺GL-7⁺Fas⁺ GC-B cells in NP-KLH– or SRBCs-immunized Nfat2^fl/fl x Cd4cre mice (Fig. 2 D and not depicted). Consecutively, the titer of the NP-specific Abs IgM and class-switched IgG1 and IgG3 was increased in Nfat2^fl/fl x Cd4cre mice upon NP-KLH immunization (Fig. 2 E), revealing an overall amplification of the GC reaction upon loss of NFAT2 in T cells.

To unravel the increase in GC reaction in Nfat2^fl/fl x Cd4cre mice at the molecular level, we performed global transcription analyses of FACS sorted ICOS^hiCXCR5^hi T_FH cells in comparison to activated (ICOS^hiCXCR5^–) and naive (ICOS^–CXCR5^–) T_conv cells. In line with highest expression of NFAT2 in T_FH cells, NFAT2-deficient T_FH cells showed prevalence in the number of differently regulated transcripts (Fig. 3, A and B). We analyzed affected transcripts and found that T_reg-specific or –associated genes (Hill et al., 2007) were a self-contained group being reduced in Nfat2^fl/fl x Cd4cre mice (Fig. 3 C). Performing quantitative RT-PCR from NFAT2-sufficient and NFAT2-deficient T_FH in comparison to T_conv cells again demonstrated a clear transcriptional decline of two tested key molecules for T_FR, namely Foxp3 and Blimp-1, whereas Bcl-6 was slightly increased, if quantities changed at all (Fig. 3 D). The additional loss of NFAT1 in follicular T cells had no major impact, suggesting NFAT2 as the prominent NFAT family member in T_FR cells (Fig. 3 D).

Because microarray analyses of CD4⁺ICOS^hiCXCR5^hi T_FH cells revealed a reduction in the T_reg cell signature (Hill et al., 2007) upon NFAT2 deficiency, we tested the frequency of Foxp3⁺ T_reg cells in the population of T_conv versus T_FH cells upon immunization. As NFAT2-sufficient and -deficient T_conv cells displayed the same number of Foxp3⁺ T_reg cells, significantly fewer Foxp3⁺ T_FR cells were detected among PD-1⁺CXCR5⁺ T_FH cells in mice lacking NFAT2 (Fig. 4 A). This was true irrespective of the kind and strength of immunization, shown by immunization with NP-KLH or SRBCs as well as by Ova in Nfat2^fl/fl x Cd4cre x OT-II mice tested by flow cytometry and, in individual follicles, by immunohistochemistry (Fig. 4, A–D; and not depicted). Confocal microscopy discovered reduced numbers of Foxp3⁺ T_FR cells within PNA-positive GC-areas in Nfat2^fl/fl x Cd4cre mice (Fig. 4 E). For comparison, we addressed the impact of NFAT1 on T_FR cells. The numbers of T_FR cells in B cell follicles of immunized Nfat1^−/− mice were found unaltered, whereas additional NFAT1 deficiency in Nfat2^fl/fl x Cd4cre mice might further impair the frequency of T_FR cells (Fig. 4, F and G). In summary, this defines NFAT2 as an important transcription factor for one specific subtype of nT_reg cells, namely T_FR cells, which are distributed mainly between B cells in GCs (unpublished data).

Earlier we demonstrated that the generation of thymus-derived nT_reg cells, to which T_FR belong (Linterman et al., 2011), does not depend on NFAT1 and NFAT2 in terms of Foxp3 expression (Vaeth et al., 2012). This was not consistent with the observed decrease of Foxp3⁺ cells within NFAT2-deficient CXCR5^hi T_FH cell population. We therefore asked if the loss of Foxp3⁺ cells was T_reg-intrinsic by crossing Nfat2^fl/fl mice to Foxp3-IRES-Cre (FIC; Wing et al., 2008) mice and found, again, a profound reduction of CXCR5^hi T_FR cells upon immunization (Fig. 5 A). In line with our former findings, no alteration of nT_reg cell frequencies within the CD4⁺CXCR5^–PD-1^– T_conv population could be observed (Fig. 5 A). Immunohistochemistry confirmed fewer Foxp3⁺ T_FR cells in reactive B cell follicles (Fig. 5 B). Comparable to the eradication of NFAT2 in all T cells, T_reg-specific NFAT2 deletion led to an augmented GC reaction (Fig. 5 C and not depicted) paralleled by an increase in antigen-specific IgM and IgG after immunization (Fig. 5 D). To exclude secondary effects caused by the influence of other cells, e.g., via IL-2 production, we adoptively co-transferred WT and NFAT2-deficient CD4⁺ T cells at a ratio of 1:1 into Rag1^−/− mice, followed by immunization (Fig. 5 E). WT T cells were distinguished by CD90.1 from CD90.2⁺ Nfat2^−/− T cells. Again, we found a decrease in T_FR cells specifically within the CD90.2⁺ NFAT2-deficient T cell population (Fig. 5 E). The phenotype achieved by T cell–specific loss of NFAT2 for the GC reaction was similar to a total depletion of T_reg cells. This was performed by injection of diphtheria toxin into immunized Foxp3^DTR-eGFP mice (Lahl et al., 2007). Clearly reduced numbers of T_FR cells were seconded by an expansion of GCs upon depletion of T_reg cells (Fig. 5 F and not depicted). Overall, these data account for an NFAT2-dependent and T_FR cell–intrinsic effect to limit humoral immune reaction.

As Foxp3 expression is independent on NFAT factors in thymus-derived nT_reg cells (Vaeth et al., 2012), we reasoned that NFAT2-deficient nT_reg cells might be unable to home to GCs. Indeed, when evaluating the frequency of Foxp3⁺ cells in naive CD4⁺ICOS^–CXCR5^– or activated CD4⁺ICOS^hiCXCR5^– T_conv and CD4⁺ICOS^hiCXCR5^hi T_FH subsets from WT and Nfat2^fl/fl x Cd4cre mice, the decrease in T_FR cells was paralleled by an increase in nT_reg cells within the CD4⁺ICOS^hiCXCR5^– (T_act) population (Fig. 6 A). This was a particular feature of NFAT2 and not affected by NFAT1 deficiency in Foxp3⁺ T_reg cells of an activated CD4⁺CD44^hi phenotype (Fig. 6 B). Because CXCR5 is the essential homing receptor for T_FH and T_FR cells, we evaluated whether its expression differs in WT and NFAT2-deficient T cells. Therefore, we defined follicular T cells as CD4⁺BTLA^hi and found that NFAT2 selectively regulates CXCR5 expression in T_FR but not in T_FH cells (Fig. 6 C). In line with a T_FR cell–specific effect, transwell experiments suggested an impairment (although highly variable and therefore not significant) of NFAT2-deficient Foxp3⁺ nT_reg cells, but not T_conv cells from immunized littermate mice in migration toward a CXCL13 gradient (Fig. 6 D). Altogether, these data indicate that Foxp3⁺ nT_reg cells specifically depend on NFAT2 expression for the up-regulation of CXCR5.

Although CXCR5 expression was clearly dependent on the presence of NFAT2, this could have been just one among many decisive reasons for a dysregulated GC reaction. In line with the notion that the observed defect laid within the nT_reg compartment, adoptive transfer of in vitro anti-CD3/CD28-activated congenic WT nT_reg cells into immunized WT or Nfat2^fl/fl x Cd4cre recipient mice complemented the reduced number of T_FR cells in NFAT2-deficient mice and rescued the GC phenotype (Fig. 7 A). Therefore, Nfat2^−/− nT_reg cells were lentivirally transduced for CXCR5-GFP expression and adoptively transferred together with B cells and naive T_conv cells into Rag1^−/− mice, which were immunized 2 wk thereafter (Fig. 7 B). Pre-activated NFAT2-deficient (in parallel with NFAT2-sufficient) nT_reg cells were successfully manipulated to express CXCR5 at their surface (Fig. 7 C). In fact, this made up for their NFAT2 deficiency during the provoked GC-reactions and restored GC-B cell numbers to normal levels (Fig. 7 D).

Computational analyses of the TATA-less Cxcr5 promoter region revealed three conserved noncoding sequences (CNS) between men, dog, and mice with putative NFAT responsive elements (Fig. S1, A and B). Furthermore, two DNase hypersensitive sites (HS1 and HS2) could serve as 3′ enhancers (Fig. S1 C). Reporter gene assays in human Jurkat T or murine EL-4 thymoma cells using a −1,127-bp fragment of the Cxcr5 promoter region containing most putative NFAT-responsive elements demonstrated inducibility and sensitivity to CsA (Fig. 8, A and B). Here, cells were activated by TPA plus ionomycin (T/I), which mimics antigen-receptor engagement. The presence of HS2 slightly enhanced, but did not influence the quality of activity (Fig. 8 A). Within the Cxcr5 promoter three conserved regions (named as proximal, middle, and distal regions; Fig. S1, A and B) were defined, each containing putative NFAT-responsive elements (Fig. S1, B and D). To address if NFAT can bind to these sites, we performed electrophoretic mobility shift assays (EMSAs) using Cxcr5-N1 to Cxcr5-N4 as probes. Total protein extracts were prepared from transfections with an expression plasmid for the DNA-binding domain of NFAT2 (HA-Nc1-RSD; Klein-Hessling et al., 2008)). The most proximal site (Cxcr5-N1) showed a similar bandshift as the Pubd from the Il2 promoter, but could not respond to Blimp-1, which recognizes a similar core consensus sequence, namely GAAAG (Kuo and Calame, 2004). Cxcr5-N1 therefore assigns for a potent and specific NFAT-responsive element (Fig. 8 C). Using nuclear extracts from stimulated primary T_reg cells corroborated the strong binding of NFAT to Cxcr5-N1. Both NFAT1 and NFAT2 could be recruited to this site as demonstrated by anti-NFAT1– and anti-NFAT2–mediated supershifts (Fig. 8 D). Chromatin immunoprecipitation (ChIP) assays using T cells from WT and Nfat2^fl/fl x Cd4cre mice immunized with KLH verified the proximal part of the Cxcr5 promoter to be bound by NFAT2 in vivo (Fig. 8 E). To test the function of the Cxcr5-N1 site, we deleted distal parts of the Cxcr5 promoter and found that the proximal part alone (−329 bp, containing Cxcr5-N1) could mediate T/I induction in Jurkat cells (Fig. 8 F). Next, we mutated the two G-nucleotides of the Cxcr5-N1 motif essential for NFAT-binding within the 1,127-bp-long promoter fragment (NFAT-mutated). When nonlymphoid human embryonic kidney (HEK 293T) cells were transfected, we found that exogenous NFAT2 transactivated the Cxcr5 promoter robustly, but only when Cxcr5-N1 was intact (Fig. 8 G and Fig. S1 D). In line with this, a significantly reduced induction was witnessed in Jurkat cells, when this single NFAT-responsive element was mutated (Fig. 8 H). The same tendency was observed for the 329-bp fragment, as shown for EL-4 cells (Fig. 8 I). Overall, this demonstrated the important role of Cxcr5-N1 for NFAT2-mediated transactivation of Cxcr5.

Patients suffering from SLE commonly exhibit increased frequencies of T_FH and GC-B cells and high titers of pathogenic auto-Abs (Simpson et al., 2010), often correlated with impaired numbers and function of T_reg cells (Valencia et al., 2007; Bonelli et al., 2008, 2010). Clinical hallmarks of SLE are antinuclear antigen (ANA) and anti–double-stranded DNA (dsDNA) auto-Abs. Accordingly, immunization with chromatin isolated from syngeneic-activated lymphocytes leads to a lupuslike disease in mice (Li et al., 2004; Qiao et al., 2005), which we performed with WT and Nfat2^fl/fl x Cd4cre littermate mice (Fig. 9 A). Equivalent with all other immunizations, we observed a modestly increased number in CD4⁺CXCR5^hiPD-1^hi T_FH cells (unpublished data), as well as GL-7^hiFAS^hi GC-B cells and, importantly, a sharp drop in the frequency of T_FR cells in spleen and mLN (Fig. 9 B and not depicted). Consequently, titers of pathogenic anti-dsDNA auto-Abs were found to be increased upon T cell–specific loss of NFAT2 (Fig. 9 C). PNA-immunohistochemistry of chromatin-immunized mice exhibited enlarged splenic B cell follicles in Nfat2^fl/fl x Cd4cre mice (unpublished data). Indicative of a lupuslike disease, PAS-staining detected increased intracapillary, mesangial hypercellularity and thickened contours of the mesangial matrix in kidney sections of NFAT2-deficient mice (Fig. 9 D). Furthermore, detection of pathogenic immunoglobulin deposition in the kidneys was indicative of an aggravated lupuslike disease in chromatin-immunized Nfat2^fl/fl x Cd4cre mice (Fig. 9 E). All these data demonstrate that NFAT2 is essential for T_FR cells to control humoral autoimmunity.

Earlier transcriptional profiling of CD4⁺ICOS^hiCXCR5^hi T_FH cells from chronically inflamed tonsils identified Nfat2 as one of the most prominently up-regulated genes (Rasheed et al., 2006). When we now analyzed murine T_FH cells we found a similar prevalence of NFAT2 among transcription factors in general and NFAT members especially. Immunization of mice, in which Nfat2 was specifically ablated in T cells, resulted in an unexpected phenotype, i.e., an enhanced GC reaction. This could be tracked down to an impairment of Foxp3⁺ T_reg cells to up-regulate CXCR5, leading to reduced numbers of Foxp3⁺ T_FR cells within GCs. NFAT2-dependent CXCR5 up-regulation and consequential homing to GCs was T_FR cell–intrinsic as well as specific. Furthermore, rescue experiments verified that the inability to induce CXCR5 in nT_reg cells accounted for the observed phenotype. Therefore, absence of NFAT2 was sufficient to exclude T_FR cells from GCs. The outstanding importance for precise homing of CXCR5^hiFoxp3⁺ T_FR cells was emphasized in a murine model of a lupuslike disease with a boosted severity in the absence of NFAT2.

One possible interpretation of the enhanced GC reaction upon Nfat2 ablation in T cells might have taken into account that NFAT proteins are decisive transactivators of Il2 and that less IL-2 favors GC formation (Ballesteros-Tato et al., 2012; Johnston et al., 2012). However, (a) single deficiency of NFAT2 does not dramatically reduce the levels of IL-2 (Vaeth et al., 2012), (b) we confirmed the increase of T_FH and GC-B cells after T_reg-specific deletion of Nfat2, (c) demonstrated the reduction of CXCR5 on T_reg cells as cell-intrinsic by co-transferred WT and Nfat2^−/− CD4⁺ T cells, and (d) were able to rescue the overshooting GC reaction in Nfat2^fl/fl x Cd4cre mice by transfer of CXCR5-sufficient WT nT_regcells.

Still, the phenotype of T cell–specific Nfat2 ablation, which is dominated by Nfat2^−/− T_reg cells, was unexpected, as we had revealed that overall frequency and distribution of nT_reg cells remains unaltered in mice that were NFAT2-deficient upon thymic CD4 expression (Vaeth et al., 2012). However, the observed augmented GC formation correlated with an exclusive decrease in T_FR cells and could be reversed by an adoptive transfer of WT and NFAT2-deficient, but CXCR5-sufficient, nT_reg cells. Consistently, nT_reg cells were selectively affected in up-regulating the chemokine receptor CXCR5. CXCR5 empowers T cells to migrate toward a CXCL13 gradient and accumulate site-specifically within GCs. This is reminiscent of expression of the Th1-defining transcription factor T-bet in nT_reg cells, leading to CXCR3 induction and migration to sites of Th1 responses, or of STAT3 induction for adapting CCR6 expression like Th17 cells (Chaudhry et al., 2009; Josefowicz et al., 2012; Koch et al., 2009). Therefore, the specific high expression of NFAT2 in GC-B, T_FH, and, concomitantly, in T_FR cells is another example of transcription factor cooption allowing T_reg cells to suppress distinct immune responses (Josefowicz et al., 2012).

So far, Bcl-6, which is obligatory for GC-B and T_FH cells, had been designated as the coopting transcription factor in T_FR cells, as it is mainly responsible for CXCR5 expression (Chung et al., 2011; Linterman et al., 2011; Josefowicz et al., 2012). However, ectopic expression of Bcl-6 is insufficient to drive CXCR5 expression (Ma et al., 2012). Strikingly, transferred naive CD4⁺ T cells from Bcl6^−/− mice exhibit the same CXCR5 induction as WT cells, although T_FH cells depend on Bcl-6 for a stable lineage commitment (Liu et al., 2012). Because Bcl-6 acts as a repressor, it instead intervenes with the transcriptional fate of other CD4⁺ T cell lineages; e.g., it antagonizes GATA3 expression and Th2 responses (Kusam et al., 2003) and Tbx21 and Rorc expression (Yu et al., 2009), or binds and inhibits T-bet protein (Oestreich et al., 2012). Therefore, follicular T cell specificity is provided indirectly by Bcl-6.

Currently, little is known about the transcriptional control of Cxcr5 in T_FH and T_FR cells. In mature B cells, constitutive expression is cooperatively regulated by Oct-2, Bob1 (OBF-1), and NF-κB, which bind to consensus sequences within its proximal promoter (Wolf et al., 1998). In addition, a putative NFAT-responsive element (Pu.box) was recognized. Indeed, in retinoic acid–treated HL-60 cells, the human homologous element binds NFAT4 (Wang and Yen, 2004). However, we could not detect any binding of NFAT proteins to the murine Cxcr5-N3. Another putative site, Cxcr5-N2, is essential for NF-κB responsiveness in B cells (Wolf et al., 1998), in line with a high degree of overlap between NFAT and NF-κB sites (McCaffrey et al., 1994). Importantly, the most proximal and thus far unrecognized Cxcr5-N1 proofed to be a true and decisive NFAT response element. Therefore, NFAT2 is likely to contribute to CXCR5 expression in GC-B and T_FH cells.

The fact that only T_FR, but not T_FH cells depend on NFAT2 to migrate to GCs attributes to the special transcriptional landscape of nT_reg cells. Effector T_reg (eT_reg) cells, which have differentiated in parallel with their respective T_conv cells highly express Blimp-1 in addition to Foxp3 (Cretney et al., 2013). From PCs it is known that Blimp-1 represses CXCR5, which is a requisite to egress from GCs (Shaffer et al., 2002). Similarly, Blimp-1 deletion (Linterman et al., 2011) led to an up-regulation of CXCR5, whereas overexpression of Blimp-1 repressed CXCR5 (Oestreich et al., 2012). One possibility also supported by our preliminary data could be that NFAT2 specifically overcomes the Blimp-1–mediated repression in T_FR cells. In such a scenario, this molecular checkpoint would be controlled by an inhibiting (i.e., Blimp-1) and activating (i.e., NFAT2) factor, which ensures an adequate expression of CXCR5, and thereby an appropriate GC response. A mutual balancing of Blimp-1 and NFAT2 might even culminate in synergy. At least this has been described for eT_reg cells, where Blimp-1 in conjunction with IRF4 acts as a positive regulator at the Il10 intron1 enhancer (Cretney et al., 2011).

T_reg cells can directly suppress B cells in GCs and periphery (Lim et al., 2005; Vaeth et al., 2011; Sage et al., 2013), where T_reg-derived IL-10 and the GITR are functionally involved (Alexander et al., 2011). In addition, we found that direct cell–cell contact enables the transfer of cAMP (cyclic adenosine monophosphate) from T_reg into B cells via gap-junction intracellular communication, resulting in the induction of ICER (inducible cAMP early repressor; Bopp et al., 2007; Vaeth et al., 2011; Bodor et al., 2012). ICER is a transcriptional repressor, which inhibits lymphocyte activation not only by interfering with Nfat2 expression directly, but also the suppression of NFAT-mediated transcription (Vaeth et al., 2011; Bodor et al., 2012). Because GC-B cells require high levels of NFAT2 for GC formation and the production of class-switched, high-affinity Abs (Bhattacharyya et al., 2011), the inhibition of NFAT in target cells appears as an additional mechanism in how T_FR cells regulate GC reaction.

In summary, NFAT2 is instrumental for both an adequate and nonpathological humoral immune response. Whereas NFAT2 is of importance for GC-B and T_FH cells (Rasheed et al., 2006; Bhattacharyya et al., 2011), it is concomitantly needed to limit the GC reaction by T_FR cells. NFAT2 regulates the specific homing of T_FR cells to GCs via up-regulating CXCR5, essentially to prevent humoral autoimmune diseases like SLE.

Nfat2^fl/fl mice were generated in A. Rao’s laboratory (Harvard Medical School, Boston, MA). Phenotypes, when crossed to Cd4cre and FIC, have been explored earlier and proof of NFAT2 deficiency had previously been provided on genetic and protein levels for all T cells including nT_reg cells (Vaeth et al., 2012). Nfat1^−/−, Nfat4^−/−, B6-Tg (Cd4cre) 1Cwi/Cwilbcm (EMMA, Italy), Foxp3-IRES-cre (FIC), OT-II, Rag1^−/−, Nfatc1/egfp, and Foxp3^DTR-eGFP (DEREG) mice have been described (Barnden et al., 1998; Ranger et al., 1998; Lee et al., 2001; Jasinski et al., 2006, Lahl et al., 2007; Wing et al., 2008; Hock et al., 2013). Animals were used at 6 to 16 wk and maintained in accordance with institutional guidelines for animal welfare. Jurkat, EL-4, and HEK 293T cells were cultured in complete RPMI or DMEM medium containing 10% FCS (Nayak et al., 2009; Vaeth et al., 2012).

Sex-matched littermate mice were immunized (approval by the government of Lower Franconia; 55.2–2531.01-80/10) intraperitoneally (i.p.) with 100 µg NP-(27)-KLH (BioCat) emulsified in Imject Alum (Thermo Fisher Scientific) or 10⁹ SRBCs (Dunn Labortechnik) and analyzed on d 7. In some experiments, mice were boosted with 50 µg NP-(27)-KLH on day 7 and analyzed on d 10–12. OT-II mice were immunized i.p. with 100 µg OVA_323-339 emulsified in Imject Alum. Depletion of Foxp3⁺ T_reg cells in Foxp3^DTR-eGFP (DEREG) mice was performed by daily injection of 1 µg diphtheria toxin (DT) in PBS i.p. on day 1–3 after KLH immunization. For adoptive cell transfer experiments, 12 × 10⁶ negatively isolated WT B cells (Miltenyi Biotec) were transferred along with 6 × 10⁶ WT negatively isolated (Invitrogen) CD90.1⁺CD4⁺ T cells and 6 × 10⁶ CD90.1⁺CD4⁺ Nfat2^fl/fl x Cd4cre T cells i.p. into Rag1^−/− mice. After 28 d mice were immunized with 100 µg NP-(27)-KLH. For rescue experiments, CD90.1⁺CD4⁺CD25⁺ WT nT_reg cells were isolated and activated with anti-CD3/CD28 beads (Invitrogen). 7 d after immunization, 5 × 10⁶ nT_reg cells were injected retroorbitally into CD90.2⁺ WT and Nfat2^fl/fl x Cd4cre littermates.

Induction of a lupuslike disease by chromatin immunization was performed as reported by others (Li et al., 2004; Qiao et al., 2005). In brief, chromatin was isolated from syngenic BL/6 splenocytes stimulated for 2 d in complete RPMI with T/I. After washing, nuclei were extracted, washed, and lysed in buffer containing 300 mM NaCl, 25 mM EDTA, 50 mM Tris-HCl, pH 8.0, 0.2% SDS, and 0.5 mg/ml Proteinase K (Fermentas) overnight at 50°C. Chromatin was precipitated with EtOH, dried, and digested by incubation with S1 Nuclease (Fermentas) for 60 min at 30°C in the respective buffer. The reaction was stopped by addition of 40 mM EDTA for 15 min at 70°C and the chromatin was finally precipitated with EtOH, dried, and resolved in PBS at 1 µg/µl concentration. Mice were immunized subcutaneously (s.c.) with 100 µg chromatin in 100 µl CFA (Difco) and the immune reaction was boosted with additional s.c. injection of 100 µg chromatin in IFA (Difco) in the 1st, 3rd, and 23rd week.

FACS staining was performed with following Abs: FITC-conjugated CD4 (GK1.5), CD44 (IM7), CD90.1 (OX-7), CXCR4 (2B11), GL-7 (Ly-77); PE-conjugated CD4 (RM4-5), CD19 (1D3), CD25 (PC61), CD95/Fas (DX2), CD272/BTLA (8F4), CD279/PD-1 (RMP1-30), CXCR5 (2G8); biotin-conjugated B220 (RA3-6B2), CD4 (GK1.5) CD90.2 (53–2.1), CXCR5 (SPRCL5), ICOS (7E.17G9); eFluor 450-conjugated CD4 (RM4-5), GL-7 (Ly-77); and secondary streptavidin-APC, streptavidin-PE, or streptavidin-PE-Cy5.5 mAbs (all BD or eBioscience). Intracellular Foxp3 (FJK-16s, FITC-, and APC-conjugated) staining was performed using the Foxp3 staining kit (eBioscience). Samples were analyzed on a FACSCanto II (BD) with FlowJo software (Tree Star). FACS sorting was performed with the FACS Aria cell sorter (BD).

For confocal microscopy, FACS-sorted cells were harvested on slides using cytospin. Cells were fixed and permeabilized with the Foxp3 staining kit (eBioscience). For tissue samples, formalin-fixed paraffin-embedded specimens were prepared (Vaeth et al., 2012). Immunohistochemical reactions were prepared using heat-induced antigen retrieval performed under pressure in citrate-buffer (pH 6.0; Sigma-Aldrich). The following primary antibodies were used: rabbit anti-NFAT1 (Sigma-Aldrich), mouse anti-NFAT2 (7A6; BD), rabbit anti-NFAT2/α (IG-457; immunoGlobe), mouse anti-NFAT4 (F-1; Santa Cruz Biotechnology, Inc.), mouse anti-CD3ε (PS1; Leica), PNA-biotin (Vector Laboratories), rabbit anti-Ig (Dako), and rat anti-Foxp3 (FJK-16s, eBioscience). Secondary staining was performed using the following Abs: anti–rabbit Alexa Fluor 647, anti–mouse Alexa Fluor 488, and anti–rat Alexa Fluor 555 (all from Molecular Probes); anti–IgM-Cy5 (Jackson ImmunoResearch Laboratories); and streptavidin-Alexa Fluor 488 (Invitrogen). Slides were mounted with Fluoromount-G (SouthernBiotech) containing DAPI. Images were taken with a confocal microscope (TCS SP2 equipment, objective lens; HeX PL APO, 40×/1.25-0.75; Leica) and LCS software (Leica). For statistics, >30 cells from at least 2 independent experiments were counted and mean fluorescence intensity per cell (MFI) was calculated. For immunohistochemistry, anti–rabbit HRP (Dianova), anti–rat HRP (Abcam), and Neutravidin HRP (Dianova) were used as secondary Abs. Hematoxylin and eosin staining and PAS staining were performed as previously described (Vaeth et al., 2012).

RNA was extracted using RNeasy Micro kit (QIAGEN) followed by cDNA synthesis with the iScript II kit (Bio-Rad Laboratories). Real-time qRT-PCR was performed with an ABI Prism 7700 detection system using following primers: Nfat2, 5′-GATCCGAAGCTCGTATGGAC-3′ plus 5′-AGTCTCTTTCCCCGACATCA-3′; Nfat1, 5′-TCATAGGAGCCCGACTGATTG-3′ plus 5′-CCATTCCCATCTGCAGCAT-3′; Nfat4, 5′-GCCTCCATTGACAGAGCAACT-3′ plus 5′-CACATCCCACAGCCCAGTG-3′; Nfat2 P1, 5′-CGGGAGCGGAGAAACTTTGC-3′ plus 5′-CAGGGTCGAGGTGACACTAG-3′; Nfat2 P2, 5′-AGGACCCGGAGTTCGACTTC-3′ plus 5′-CAGGGTCGAGGTGACACTAG-3′; Foxp3, 5′-GGCCCTTCTCCAGGACAGA-3′ plus 5′-GCTGATCATGGCTGGGTTGT-3′; Bcl6, 5′-GATACAGCTGTCAGCCGGG-3′ plus 5′-AGTTTCTAGGAAAGGCCGGA-3′; Prdm1, 5′-TAGACTTCACCGATGAGGGG-3′ plus 5′-GTATGCTGCCAACAACAGCA-3′; Cxcr5, 5′-TCCTGTAGGGGAATCTCCGT-3′ plus 5′-ACTAACCCTGGACATGGGC-3′; Hprt, 5′-AGCCTAAGATGAGCGCAAGT-3′ plus 5′-TTACTAGGCAGATGGCCACA-3′.

CD4⁺ T cells from day 7 KLH-immunized WT and Nfat2^fl/fl x Cd4cre mice were negatively enriched (Miltenyi Biotec) followed by FACS-sorting of CXCR5^–ICOS^– T_naiv, CXCR5^–ICOS^hi T_act, and CXCR5^hiICOS^hi T_FH cells to >95% purity. RNA was extracted using the RNeasy Mini kit (QIAGEN). Biotin-labeled aRNA was prepared with the GeneChip 3′ IVT Express kit and hybridized to GeneChip mouse genome 430 2.0 arrays (Affymetrix) according to the manufacturer’s protocol. The trimmed mean signals of the probe arrays were scaled to a target value of 500 and expression values determined using the Affymetrix GeneChip Operating Software (GCOS). Data were analyzed using GeneSpring GX 12.0 software (Agilent Technologies) for significant changes in gene expression between mutant and control samples using the t test or 2-way ANOVA, and resulting P values were corrected for multiple testing using the Benjamini-Hochberg method. The original microarray data can be found in the ArrayExpress database under the accession no. E-MEXP-3820 (https://www.ebi.ac.uk/arrayexpress/experiments/E-MEXP-3820/).

ChIP-IT Express kit (Active Motif) was used according to the manufacturer’s instructions, except enzymatic shearing followed by additional 25-min sonication. After precipitating, the following Abs (5 µg) were used: anti-NFAT2 (7A6; BD) and anti–acetyl-histone H3 (Active Motif). Quantification of DNA binding was performed by (real-time) PCR using the following primers: Cxcr5 distal, 5′-CTAGTATTCTTAGGGTTCTTCC-3′ plus 5′-GGGCACTTGATCAACCTGTG-3′; Cxcr5 middle, 5′-GGCTCGCCTGGGACTGAG-3′ plus 5′-GGGGCTAAGAAAAGAGTACTC-3′; Cxcr5 proximal, 5′-ACTGACTCTGTGGGGGGAG-3′ plus 5′-CTTGCCTCTCGACTCATCTC-3′.

5 × 10⁵ CD4⁺ T cells from day 7 KLH-immunized WT and Nfat2^fl/fl x Cd4cre mice were set in 100 µl complete RPMI without FCS. The cells were placed in the upper transwell (Costar; polycarbonate, pore size 5 µm, 6.5 diam) with the lower transwell containing 600 µl complete RPMI (without FCS) with various concentrations of CXCL13 (R&D Systems) or PBS as control. After 5 h of incubation, cells from the lower well were harvested, stained for flow cytometry (CD4 plus Foxp3), and collected in 150 µl FACS buffer. Transwell cell numbers and T cell phenotype were assessed using a FACS Canto II cytometer (BD; 100 µl/min flow speed for 30 s).

For determination of NP-specific Ab levels in sera from 10 d NP-(27)-KLH–immunized mice (Bhattacharyya et al., 2011), samples were applied in threefold serial dilutions starting from a 1:20 initial dilution, and their concentration was quantified against a reference serum. For detection of anti-dsDNA Abs in sera of chromatin-immunized mice, blood samples were taken 8 wk after first immunization. Total Ig or IgG anti-dsDNA serum titers were analyzed with a quantitative mouse anti-dsDNA Ig (total A+G+M) ELISA kit (Alpha Diagnostic International) according to the manufacturer’s instructions.

Nuclear extracts from murine T_conv and Foxp3⁺ T_reg cells as well as whole cellular extracts from transiently transfected HEK 293T cells were prepared using the ProteoJET kit (Thermo Fisher Scientific). EMSA was performed, as previously described (Schmidt et al., 2008), with the following probes: Cxcr5-N1, 5′-GAAAAGACTCAGTGGAAAAAAAAAAAAAAAG-3′; Cxcr5-N1mut, 5′-GAAAAGACTCAGTAAAAAAAAAAAAAAAAAG-3′; Cxcr5-N2, 5′-GGGGGGAGTTTCCCTTTTTCTTAAA-3′; Cxcr5-N3, 5′-GGGATGGTTGGTCACCCTAGTGATAAGGAAAGTG-3′; Cxcr5-N4, 5′-GGACTGAGGGGTTTCCCAGGACAGGGGACTT-3′; IL2-Pubd, 5′-CCCCAAAGAGGAAAATTTGTTT-3′. The mutations are underlined. For super-shifting anti-NFAT2 (7A6; BD) and anti-NFAT1 (Cell Signaling Technology) Abs were used.

HA-Nc1-RSD, HA-NFAT2/C (HA-NFATc1/C), and Blimp-1-Flag have been described (Klein-Hessling et al., 2003; Schmidt et al., 2008; Nayak et al., 2009). The Cxcr5 promoter (Wolf et al., 1998) elements of murine Cxcr5 (exact sequences in Fig. S1) were cloned into NheI–NcoI matching Cxcr5-ATG with luciferase ATG and HS1 or HS2 into the SalI-site of pGL3-Basic (Invitrogen).

HEK 293T cells, Jurkat T, or EL-4 cells were transiently transfected with different Cxcr5 promoter luciferase-reporter constructs alone or in combination with a plasmid encoding for NFAT2/C using calcium phosphate, Superfect reagent (Invitrogen), and DEAE dextran, respectively. 36 h after transfection, luciferase activity was measured from the cells that were left untreated, treated with T/I or T/I plus CsA overnight and relative light units were corrected for the transfection efficacy due to total protein concentrations. Normalized mean values of at least three independent experiments are depicted in relative light units as fold activation over empty vector control.

Primary CD4⁺CD25⁺ nT_reg cells from WT and Nfat2^fl/fl x Cd4cre mice were activated with anti-CD3/CD28 beads (Invitrogen) for 3 d in presence of 50 U/ml IL-2. Beads were removed and 5 × 10⁶ nT_reg cells were collected and resuspended in 1 ml fresh RPMI medium containing 8 µg/ml polybrene (Santa Cruz Biotechnology) and 50 U/ml lL-2 (PeproTech). After 30 min, cells were centrifuged and the supernatant was mostly removed. Cell pellets were resuspended in 0.25 ml/2.5 × 10⁶ IFU ready-made lentiviral solution (gfp or Cxcr5-gfp [pLenti-EF1a(mCXCR5)-Rsv(GFP-Puro)]; Amsbio) and transduction was performed by spin infection (1,800 rpm for 90 min at 32°C). Then cells were diluted in 2 ml RPMI medium containing 50 U/ml lL-2 (PeproTech) and incubated for additional 4 h at 37°C. In the meantime, CD4⁺CD25^–CD62L^hi T cells and B cells were isolated from WT mice using the naive T cell isolation kit and the untouched B cell isolation kit, respectively (Miltenyi Biotec). nT_reg cells were washed twice and 4 × 10⁵ transduced WT or NFAT-deficient nT_regcells along with 5 × 10⁶ B cells and 2 × 10⁶ naive T cells were adoptively transferred i.p. into Rag1^−/− mice. After 14 d, mice were immunized with 100 µg NP-KLH and analyzed on day 21.

Results were compared with Prism software (GraphPad) using two-tailed paired or unpaired Student’s t test and two-way ANOVA. Differences with p-values of <0.05 are considered significant: *, P < 0.05; **, P < 0.005; ***, P < 0.001.

Fig. S1 shows a scheme of the Cxcr5 promoter region.

We are indebted to Anjana Rao for providing Nfat2 floxed mice as well as Laurie H. Glimcher for Nfat1^–/– and Nfat1^–/– x Nfat4^–/–, Kajsa Wing and Shimon Sakaguchi for Foxp3-IRES-cre⁺, and Tim Sparwasser for Foxp3^DTR-eGFP mice. Mice were kept at the Center for Molecular Medicine – ZEMM of the University of Wuerzburg. Appreciation is expressed to Miriam Eckstein, Andra Eisenmann, Ralf Kielenbeck, Doris Michel, Andrea Peters, and Ilona Pietrowski for excellent technical support.

This work was made possible by funding from the Fritz Thyssen Stiftung/Az.10.13.2.215 (M.V. + F.B.-S.), the German Research Foundation DFG: Priority Program SPP1365 (M.V. + F.B.-S.), Collaborative Research Centre/Transregio CRC/TR52/A3 (F.B.-S.), CRC/TR52/B8 (M.L. & G.M.), and CRC/TR52/C5 (E.S.), and Clinical Research Unit CRU216 (I.B.) as well as from the Federal Ministry for Education and Research: IZKF/A-167 in Würzburg, Germany (L.D. + F.B.-S.) and the Wilhelm Sander-Stiftung/2012.047.1 (M.V. + F.B.-S.).

The authors have no competing financial interests.