Tfr deficiency leads to accumulation of anti-nuclear and tissue-specific antibodies (Abs) (Gonzalez-Figueroa et al., 2021), leading to development of auto-Ab-mediated autoimmunity in older mice (Fu et al., 2018). NucPrs that are often targeted in autoimmune diseases by auto-Abs include nucleosomal histones, SS-A/Ro, RNP-Sm (small nuclear ribonucleoproteins), Scl70, and Jo-1 (aminoacyl-tRNA synthetase, both nuclear and cytoplasmic) (Mahler and Fritzler, 2010; Tan, 1983). In this study we hypothesized that these NucPrs may contain peptides cognate to Tfr. Because the rise in Tfr is usually observed at the same time or slightly after the increase in Tfh/GC B cells (Aloulou et al., 2016; Turner et al., 2017a), we first examined whether acquisition of NucPrs by GC B cells promotes accumulation of Tfr (Figure 1A). Bovine nucleosomes, SSA-Ro, RNP-Sm, Scl70, and Jo1 (that have over 92% homology to murine NucPrs), were biotinylated and conjugated to streptavidin (SA) to generate SA-NucPrs. Alternatively, SA was conjugated to foreign antigens (Ags): ovalbumin-biotin or duck egg lysozyme (DEL)-biotin to generate SA-OVA or SA-DEL Ags respectively. C57BL/6 (B6) mice were subcutaneously immunized with SA-DEL to recruit both SA and DEL-specific B cells into the GC response. After formation of GCs (8 days later) mice were boosted with SA-NucPrs, SA, or SA-OVA to promote acquisition of these Ags by the SA-specific GC B cells. The Ag-draining inguinal lymph nodes (dLNs) were collected 3 days later for flow cytometry analysis. We found that Tfr in the dLNs of SA-NucPrs-boosted mice were significantly increased as compared to mice boosted with SA or SA-OVA Ags (Figure 1A–C). In contrast, the numbers and percentage of Tfh cells in the SA-NucPrs-boosted mice slightly decreased (Figure 1D and E). The observed accumulation of CXCR5^high PD1^high FOXP3⁺ Tfr was not accompanied by increase in the frequencies of CXCR5^int PD1^int FOXP3⁺ or CXCR5^low PD1^low FOXP3⁺ Tregs (Figure 1B and F–H). This data suggests that selected experimental scheme with the NucPrs promote rapid induction of Tfr in mice.

To assess gene expression and clonal repertoire of CXCR5^high PD1^high FOXP3⁺ Tfr induced by SA-NucPrs booster in mice, we performed 10× TCR immunorepertoire analysis of follicular (CXCR5^high PD1^high) T cells sorted from the dLNs of mice immunized with SA-DEL and boosted with SA or SA-NucPrs as described above (Figure 1A, Figure 2, Figure 2—figure supplement 1A). Graph-based clustering analysis of follicular T cells revealed 12 clusters of cells that express Bcl6 with clusters 9–12 enriched for expression of Foxp3 (Figure 2A–D, Figure 2—figure supplement 1A–E, Supplementary file 1). Based on that and uniform manifold approximation and projection (UMAP) dimension reduction of the data, we will call clusters 9–12 as Tfr-like cells and clusters 1–8 as Tfh-like cells. Clusters 9 and 10 have majority of Tfr-like cells, while the cells in cluster 12 have upregulated expression of Mki67, Ccna2, and other genes associated with cell proliferation (Figure 2A and B, Figure 2—figure supplement 1D, E, Supplementary file 1). As previously reported (Maceiras et al., 2017), we found that TCR clonal repertoire of Tfr-like cells is very diverse, with majority of cells present at 1 cell per clone in the 10× data. We therefore performed analysis of the more abundant clones (>1 cell per clone) in Figure 2E and F. Of note, the majority of TCR clones in clusters 9–12 overlapped with TCR clones from other follicular T cell clusters (Figure 2E). In clusters 9 and 10 at least half of the clones were found exclusively in other Tfr-like clusters. In contrast, in clusters 11 and 12 the majority of clones were also present in Tfh-like clusters (Figure 2E and F).

In parallel to the observed accumulation of CXCR5^high PD1^high FOXP3⁺ cells in mice boosted with SA-NucPrs by flow cytometry (Figure 1C and F), 10× analysis revealed increase of cells in cluster 10 (Figure 2A and G) that had the highest expression of Foxp3, Icos, Il2ra, Entpd1, and Ctla4 (Figure 2D). Cluster 10 was also enriched for cells expressing Il10 and granzyme B, while Nrn1 (encoding neuritin 1)- expressing cells were present in both clusters 9 and 10 and to some extent in other Tfr-like and Tfh-like cell clusters (Figure 2C, Figure 2—figure supplement 1F). In accord with the 10× data for cluster 10, flow cytometry analysis confirmed that CXCR5^high PD1^high FOXP3⁺ cells with the highest levels of ICOS were CD25^pos and had elevated levels of CTLA4, and granzyme B, as compared to Tfr with ICOS^int/low and Tfh cells (Figure 2H–K). Overall, the data suggests that the accumulating Tfr in SA-NucPr-boosted mice are likely to have more immunosuppressive Treg phenotype.

Importantly, we detected no increase in the Tfh-like cell TCR clones (from clusters 1–8) within Tfr-like clusters (9–12) in mice boosted with SA-NucPrs as compared to SA. Moreover, there was a trend for decreased frequency of Tfh-like cluster-associated clones among Tfr in mice boosted with SA-NucPr (Figure 2E and F). Therefore, based on the 10× TCR immunorepertoire analysis, we suggest that the observed increase in Tfr in the SA-NucPrs-boosted mice occurs due to accumulation of the cells with immunosuppressive phenotype that are not derived from abundant clones of Tfh cells.

Interestingly, single-cell analysis of Tfh cells (clusters 1–8) revealed a few genes with statistically different levels of expression in SA compared to SA-NucPrs-boosted mice (Figure 2—figure supplement 1G). For example, expression of proton-coupled peptide transporter Slc15a2 (that was detected in a few Tfh cells) was completely abolished in SA-NucPrs-boosted mice. In addition, the gene Apoe encoding apolipoprotein E (expressed in Tfh cluster 1) was downregulated in the Tfr-induced mice (Figure 2—figure supplement 1H).

To determine whether human Tfr have any subsets similar to the murine Tfr with immunosuppressive phenotype, we performed PCA of the publicly available integrated multimodal single-cell data from human LN samples (GSE195673). We performed graph-based clustering analysis of CD4 T cells and identified follicular T cell clusters using a modular score generated based on the expression of CXCR5, PDCD1, BCL6, ICOS, CTLA4, IL1R2, and CXCL13 (Figure 2—figure supplement 2A). Follicular T cell subsets were then reclustered and expression of FOXP3 and other genes associated with Treg’s immunosuppressive activity were analyzed (Figure 2—figure supplement 2A-C). Within eight of the detected follicular CD4 T cell clusters, FoxP3 expression was detected in clusters 4–7, with the highest expression in cluster 5 and the lowest in cluster 7. Elevated expression of IL2RA and ENTPD1, CTLA4, LAG3 was detected in clusters 5–7. Expression of TGFβ-1 and ICOS was the highest in cluster 7. Importantly, clusters 6 and 7 were also enriched in IL10-expressing cells. However, these clusters also expressed some Tfh-specific genes, such as IL21 and CD40L (especially in cluster 6) (Figure 2—figure supplement 2B, C). Therefore, based on the single-cell analysis, human LNs have follicular T cells that resemble some gene expression characteristics of the immunosuppressive murine Tfr subset (cluster 10), but have reduced or undetectable expression of FOXP3 and express some factors typical for Tfh cells.

To determine whether NucPrs-induced Tfr can in turn affect GC responses, we examined the total and SA-specific GC B cells in the SA-DEL-immunized mice boosted with SA-NucPrs or control Ags (Figure 1A). We found that in mice boosted with SA-NucPrs the frequency of GC B cells and the size of GCs were reduced compared to mice boosted with SA (Figure 3A and B, Figure 3—figure supplement 1A, B). Of note, while SA-boosting induced about threefold increase in the GC response as compared to no boosting control, no significant increase in GCs was detected in the SA-NucPrs-boosted mice (Figure 3—figure supplement 1C). Moreover, SA-specific GC B cells that were expected to preferentially reacquire SA-linked NucPrs underwent greater suppression than total GC B cells or DEL-specific B cells (Figure 3A, C and D; Figure 3—figure supplement 1D–J). Importantly, in mice boosted with SA-OVA no decrease in the total or SA-specific GC B cells was detected (Figure 3A–D). In parallel to the observed reduction in the numbers of SA-specific GC B cells, we detected reduced accumulation of the SA-specific class-switched memory B cells in mice boosted with SA-NucPrs (Figure 3E–G). We also found decreased formation of SA-specific plasmablasts (PB), including the PB recently derived from GC B cells (GCPB) (Figure 3H–L). Finally, in mice boosted with SA-NucPrs the titers of the SA-specific Abs were reduced as compared to mice boosted with SA (Figure 3—figure supplement 1K).

Importantly, our suggested immunization scheme (Figure 1A) promoted suppression of the NucPrs-acquiring GC B cells not only in the B6 mice but also in NZM2328 mice. NZM2328 mice spontaneously develop dsDNA auto-Abs and systemic lupus erythematosus-like disease after 5 months of age (Rudofsky and Lawrence, 1999; Waters et al., 2001; Wolf et al., 2018) and form anti-ribonucleoprotein (RNP) auto-Abs (Figure 4A). In this study we found that the ratio of GC B cells to Tfh cells (that is highly conserved in GCs; Baumjohann et al., 2013) is elevated over fivefold in 6- to 8-week-old female NZM2328 compared to B6 mice, suggesting dysregulated GC responses even in young NZM2328 mice (Figure 4B and C). Importantly, SA-DEL immunization followed by boosting with SA-NucPrs decreased the GC B/Tfh cell ratio in the NZM2328 mice, as well as in B6 mice (Figure 4B and C). Even more significant was the observed decrease in the SA-specific GC B cell/Tfh cell ratio in the SA-NucPrs-boosted NZM2328 (Figure 4D and E). The observed suppression of the GC responses was likely due to increased Tfr responses in the SA-NucPrs-boosted mice, although in different sets of NZM2328 mice Tfr levels were variable (Figure 4F). Therefore, even in pre-autoimmune state in mice where B cell responses are inflated, SA-DEL immunization followed by SA-NucPrs boosting provided a suppressive effect.

To verify whether observed suppression of the GC responses in the SA-NucPrs-boosted mice depended on Tfr, we repeated the immunization/boosting experiments (as in Figure 1A) in Foxp3^cre Bcl6^fl/fl mice that are deficient for Tfr (Wu et al., 2016) or in control Foxp3^cre Bcl6^+/+ mice. As expected, the frequency of CXCR5^high PD1^high FoxP3⁺ Tfr in the Foxp3^cre Bcl6^fl/fl mice was low, and no increase in Tfr was observed after boosting these mice with SA-NucPrs (as opposed to Foxp3^cre Bcl6^+/+ control mice) (Figure 5A–C). At the same time, Tfh responses in Foxp3^cre Bcl6^fl/fl mice were not significantly affected as compared to Foxp3^cre Bcl6^+/+ controls (Figure 5D and E). Importantly, while boosting Foxp3^cre Bcl6^+/+ mice with SA-NucPrs promoted significant suppression of the SA-specific and total GC B cells, both effects were completely abolished in the Tfr-deficient Foxp3^cre Bcl6^fl/fl mice (Figure 5F–I). This data suggests a direct role of NucPrs-induced Tfr in the negative regulation of the GC responses with predominant suppression of the NucPrs-acquiring GC B cells.

We then examined which immunization strategy promotes accumulation of the NucPrs-induced Tfr (Figure 6). Direct immunization of naïve mice with SA-NucPrs did not induce rapid expansion of Tfr that was observed in mice preimmunized with SA-DEL (Figure 6A–D, left panels). We then assessed whether anti-SA Abs (that should be induced after preimmunization of mice with SA-DEL) may facilitate SA-NucPrs uptake, presentation and efficient activation of Tfr by antigen-presenting cells. To address this, naïve mice were transferred with serum from SA-immunized mice and then immunized with SA-NucPrs. However, at 3 days following vaccination no effect on the Tfr frequency was detected as compared to mice that did not get serum transfer (Figure 6A, C and E, left panels). We next examined whether accumulation of SA-specific B cells induced by preimmunization of mice with SA-DEL was important for rapid increase in Tfr. To test this, we compared induction of the Tfr response by SA-NucPrs boosting in mice preimmunized with SA-DEL or with OVA. In contrast to the rapid expansion of Tfr and suppression of GCs in the SA-DEL-preimmunized mice, no increase in Tfr and/or suppression of GC B cells was detected in mice preimmunized with OVA (Figure 6B, D and F). Since boosting of mice with an ongoing SA-specific GC response with SA-NucPrs is expected to promote acquisition of these Ags by the SA-specific GC B cells, we suggest that directing a combination of NucPrs to a selected subset of GC B cells may be a quick and efficient method for triggering rapid induction of Tfr and suppression of GC B cell subset that present the acquired NucPrs’ peptides.

We then examined which of the selected NucPr Ags (nucleosomes, SSA-Ro, RNP-Sm, Scl70, and Jo1) promotes the observed accumulation of Tfr and suppression of GC B cells. To address this, SA-DEL-preimmunized mice were boosted with SA conjugated to each one of these Ags to assess Tfr frequency and SA-specific GC response (Figure 6—figure supplement 1A–D). While robust increase in Tfr was detected in mice boosted with combined SA-NucPrs (Figure 6—figure supplement 1B), boosting with separate SA-NucPr Ags did not promote significant increase in Tfr (Figure 6—figure supplement 1C). Based on this data we suggest that the observed increase in Tfr (Figure 6—figure supplement 1B) occurs due to additive accumulation of Tfr specific to various NucPrs Ags.

Despite the non-significant increase in Tfr after targeting separate NucPrs to SA-specific B cells, we detected decreased frequency of the SA-specific GCs B cells in mice boosted with all selected Ags except SA-Jo1 (Figure 6—figure supplement 1D). In addition, we have tested the frequency of the nucleosome-specific B cells in mice boosted with SA-nucleosomes conjugates. In accord with previous study, we found (based on the staining with fluorescent nucleosome tetramers) that about ~1% of GC B cells in SA-DEL-immunized mice boosted with SA were specific to nucleosomes (Gonzalez-Figueroa et al., 2021). However, boosting of mice with SA-nucleosomes led to significant drop in the frequency and numbers of the nucleosome-specific GC and memory cells (Figure 6—figure supplement 1E–H). Overall, this data suggests that nucleosomes and other NucPr Ags contain peptides cognate to Tfr and if acquired by GC B cells may promote their suppression by Tfr.

While antigen-specific targeting of NucPrs to GC B cells promotes rapid accumulation of Tfr in mice, whether this finding maybe relevant to the regulation in humans is unclear. We therefore assessed whether human Tfr may have a subset of cells cognate to the NucPrs. Based on the previous studies, cognate interactions between B and T cells cocultured ex vivo lead to increased formation of B-T cell conjugates detectable by flow cytometry analysis (Choudhuri et al., 2005). We therefore utilized a similar approach with Tfr (see Figure 7A) and B cells (CD19⁺ CD3^- CD27^-) sorted from the blood of healthy human subjects. To promote antigen uptake by human B cells we first coupled anti-human IgM-biotin Abs to SA, SA-DEL or SA-NucPrs to generate protein complexes αIgM (αIgM-SA or αIgM-SA-DEL) and αIgM-NucPrs (αIgM-SA-NucPr). Sorted B cells were then cultured in the presence of αIgM or αIgM-NucPrs to promote crosslinking of B cell receptors and internalization of the linked to αIgM protein complexes for degradation and HLA class II /peptide presentation. Tfr from the same patient were then cocultured with activated or control B cells for 36 hr and formation of Tfr-B cell conjugates was assessed by flow cytometry analysis (Figure 7B). In three independent experiments we detected an increase in the frequency of Tfr-B cell conjugates when Tfr were cocultured with activated B cells that have internalized NucPrs as compared to control Ags or B cells that did not acquire Ags (Figure 7B and C). Based on that, we suggest that a significant fraction of the circulating human Tfr may be specific to the peptides from the NucPrs selected in this study.

C57BL/6 (B6, WT) mice were purchased from The Jackson Laboratory. Bcl6^fl/fl (Hollister et al., 2013) and Foxp3-Yfp^cre mice (Rubtsov et al., 2008) were crossed in Alexander Dent’s Lab (Wu et al., 2016). NZM2328 mice (Jacob et al., 2003) were a gift to Michelle Kahlenberg from Chaim Jacob (USC). All mice were bred and maintained under specific pathogen-free conditions. Relevant mice were interbred to obtain Bcl6^fl/fl Foxp3-Yfp^cre mice. All the animal experiments were conducted in compliance with the protocols reviewed and approved by the Institutional Animal Care and Use Committee of the University of Michigan.

SA-DEL was generated as previously described (Turner et al., 2017b). Purified DEL was conjugated to biotin at a 1:2 molar ratio using Sulfo-NHS-LC-Biotin (Thermo Fisher Scientific) according to the manufacturer’s directions and incubated for 3 hr on ice. After dialyzing three times in PBS DEL-biotin incubated with SA (Sigma) at a 10:1 molar ratio for 30 min on ice, followed by removal of unbound DEL-bio by passage through a 30 kDa molecular weight cut-off desalting column (Bio-Rad). Nucleosome, RNP/Sm, Jo-1, Scl-70, and Ro (SSA) (AROTEC DIAGNOSTICS) were conjugated to biotin at a 1:50 molar ratio using Sulfo-NHS-LC-Biotin and incubated for 3 hr on ice. After dialyzing three times in PBS, nucleosome-biotin, RNP/Sm-biotin, Jo-1-biotin, Scl-70-biotin, and Ro (SSA)-biotin were incubated with SA at a 1:1 molar ratio for 30 min on ice. Antigen conjugation was verified by SDS gel electrophoresis. After incubation, antigens were aliquoted and kept at –20°C. OVA-bio (Thermo Fisher, NC0887816) were incubated with SA at a 4:1 molar ratio for 30 min on ice. All antigens were aliquoted and kept at –20°C.

To generate anti-IgM-NucPr, SA, anti-IgM-bio, nucleosome-bio, Jo-1-bio, Scl-70-bio, RO (SSA)-bio, and RNPsm-bio were incubated at a 1:2:0.5:0.5:0.5:0.5:0.5 molar ratio for 30 min on ice. To generate anti-IgM-DEL, SA, anti-IgM-bio, and DEL-bio were incubated at a 1:2:2 molar ratio for 30 min on ice. To generate anti-IgM-SA, SA and anti-IgM-bio were incubated at a 1:2 molar ratio for 30 min on ice.

In some experiments mice were subcutaneously (s.c.) immunized with 50 µg SA-DEL in Ribi, reimmunized with 10 µg SA, or with SA-OVA or SA-NucPrs (with 10 µg of SA in each conjugate prep) in Ribi at day 8 after immunization. In some experiments serum from mice immunized with 50 µg SA in Ribi at day 8 after immunization were collected and injected into unimmunized mice followed by immunization with 10 µg SA or SA-NucPrs (containing 10 µg SA). In some experiments naïve mice were immunized with 10 µg SA or SA-NucPrs (containing 10 µg SA) directly. Lymphoid cells from the draining lymph nodes and the spleens of immunized mice were analyzed at the indicated time points. Blood was collected into Microvette CB 300 tube via cardiac puncture when the mice are under deep anesthesia. About 0.5–1 ml blood was obtained from one mouse. Serum recovered after centrifugation at 10,000 × g 5 min 20°C.

Single-cell suspensions from draining lymph nodes and spleens were prepared and filtered through a 70 µm nylon cell strainer (BD). Red blood cells were lysed. Cells were washed in FACS buffer (2% FBS, 1 mM EDTA, 0.1% NaN₃ in PBS) and followed by surface staining for the indicated markers for 20 min at 4°C. NP-specific B cells or SA-specific B cells were detected with BCR-specific binding with NP-PE (Biosearch Technologies) or SA-PE (BioLegend). For intracellular staining, the FoxP3 intracellular staining kit (Thermo Fisher Scientific) was used according to the manufacturer’s instructions. Samples were then incubated with anti-FoxP3, CTLA4, granzyme B, or Ig (H+L). All samples were acquired on a BD FACSCanto flow cytometer. For cell sorting, enriched B cells and T cells were incubated with Abs in Sorting buffer (0.5% FBS and 2 mM EDTA in PBS) and were performed on a BD FACSAria III cell sorter. All data were analyzed with FlowJo (version 10.6.0) software.

SA-specific IgG were detected in serum from blood by ELISA. Nunc 96-well ELISA plates were coated with 50 µl of 2 µg/ml SA (Sigma) in borate saline buffer (100 mM boric acid, 0.9% NaCl, pH = 7.4) overnight at 4°C. Wells were blocked with 0.1% Bio-Rad Gelatin in PBS plus 0.05% Tween-20. Twofold diluted serum samples were loaded into the plate. ELISA plates were incubated for 1 hr at room temperature. Plates were washed with PBS containing 0.05% Tween-20. Bound Ab detected with 1.5 µg/ml IgG-HRP (Invitrogen). After washing, the color was developed with TMB (Thermo Fisher). The chromogenic reaction was stopped with 2N sulfuric acid and the plates were read with a Synergy HT microplate reader (Bio-Tek Incorporated) at 405 and 630 nm. All plates contained serial dilutions of the serum that was used to generate the calibration curve for quantitative comparison of the samples.

Anti-RNP was detected in female 52-week-old C57Bl/6 and 32- to 52-week-old NZM2328 mice (harvested at time of nephritis or at 52 weeks) via ELISA (Alpha Diagnostics, San Antonio, TX, USA) according to the manufacturer’s instructions.

Freshly isolated lymph nodes were fixed in 1% PFA for 1 hr at room temperature, washed with PBS three times, and stored at 4°C in 30% sucrose solution overnight. Fixed samples were then transferred to OCT (Tissue-Tek) and snap-frozen. Thirty µm thick sections were cut via cryostat (Leica). The sections were dried at room temperature for 3 hr. They were blocked using normal rabbit serum (Sigma-Aldrich) in PBS with 0.2% Triton X-100 and then stained with anti-mouse CD3 PE-CF594 (BD) and anti-mouse Bcl6 Alexa Fluor 488 (BD) overnight. After washing, slides were then mounted in fluoromount-G (Southern Biotech) and analyzed via confocal microscopy with Leica SP5 II (Leica Microsystems) using a 25× objective. Images were processed using Imaris and ImageJ. For quantitative analysis of GCs, Bcl6-positive areas in B cell follicles were identified, manually outlined in ImageJ, and their area was then calculated.

B6 mice were s.c. immunized with SA-DEL in Ribi and at day 8 were s.c. reimmunized with SA, SA-NucPr. Follicular T cells (CD4⁺PD1^hiCXCR5^hi) were sorted from mice at day 11 after immunization into PBS + 2% FBS. Single-cell suspensions were subjected to counting and viability checks on the LUNA Fx7 Automated Cell Counter (Logos Biosystems) and diluted to a concentration of 700–1000 cells/µl. Single-cell libraries were generated using the 10× Genomics Chromium Controller with Immune Profiling reagents following the manufacturer’s protocol (10× Genomics). Final library quality was assessed using the LabChip GX (PerkinElmer). Libraries were subjected to paired-end sequencing according to the manufacturer’s protocol (Illumina NovaSeq 6000). Bcl2fastq2 Conversion Software (Illumina) was used to generate de-multiplexed Fastq files and the CellRanger Pipeline (10× Genomics) was used to align reads and generate count matrices. Barcodes with mitochondrial reads >5% were removed. Preprocessing, clustering, and dimensionality reduction were performed using Loupe Browser and Loupe V(D)J browser (10× Genomics, Cell Ranger). Graph-based clustering of Tfr were visualized in 2D using UMAP algorithm.

For human scRNA-seq analysis, publicly available datasets were downloaded for lymph nodes cells from SARS-CoV2 mRNA-vaccinated patients (GSE195673) (Kim et al., 2022). Seurat v4 (Hao et al., 2021) was used for downstream analysis based on Seurat vignettes. Cells which had less than 200 or more than 8000 transcripts and mitochondrial genes representing greater than 15% of total transcripts were removed. All remaining cells were then clustered and projected into UMAP plots. The optimal number of PCs used for UMAP dimensionality reduction was determined using Jackstraw permutations which resulted in the first 15 PCAs being used. Identification of follicular T cells was accomplished by adding a module score in Seurat. Clusters in which the module score was significant enriched for were identified as being follicular. Follicular cells were then reclustered and differentially gene expression analysis between each cluster and all other cells was performed.

Human blood was collected with informed consent from healthy donors in accordance with a University of Michigan IRB approved protocol (HUM0007150). Fresh blood was subjected to centrifugation over a Ficoll-Paque Plus (GE Healthcare) density gradient, washed twice in PBS and resuspended cells in PBS. PBMC were stained with fluorophore-conjugated Abs. After washing, CD3^-CD19⁺CD27^- B cells and CD4⁺CD19^-CD45RA^loCXCR5⁺CD25^hiCD127^lo Tfr were sorted and resuspended in RPMI 1640 medium (GIBCO) with 10% FBS (Hyclone), 1% HEPES (Hyclone), 100 U/ml penicillin and 100 µg/ml streptomycin (GIBCO). 5×10⁴ B cells were first incubated with 5 µg/ml anti-IgM complexes and 1 µg/ml recombinant human CD40L (BioLegend) for 30 min and then 1×10⁴ Tfr were added. After incubation at 37 °C and 5% CO₂ for 36 hr, the cells were stained and analyzed by FACS. The ratio of CD3⁺CD19⁺ cells and CD3⁺ cells were calculated.

Statistical tests were performed as indicated using Prism 8 (GraphPad). No blinding or randomization was performed for animal experiments, and no animals or samples were excluded from analysis. All the statistical details of experiments and statistical analysis can be found in figure legends. Differences between groups not annotated by an asterisk did not reach statistical significance. Outliers were not excluded. Of note, outlier experimental data points in several figures did not affect statistical significance of the data.