A T-dependent B cell response involves germinal centers (GCs), in which antigen-activated B cells proliferate, hypermutate, and undergo affinity-based selection by follicular T helper (TFH) cells.16,17 While the GC reaction is a hallmark of active immune responses to foreign antigens following vaccination or infections, GCs also spontaneously form in mice genetically predisposed to autoimmunity18 and even in normal mice.19 In the latter case, microbiota, food components and other environmental materials are plausible source of antigen, while bona fide autoantigen cannot be ruled out. Therefore, we decided to look for BDA-reactive B cells and T cells in GCs that spontaneously develop in healthy, unmanipulated animals.

Spontaneous GC development

Extending previous findings,19 we found that, in specific pathogen-free (SPF) B6 mice of 8–10 weeks of age, ~0.2% of total B220+ B cells were Fas+GL7+ GC B cells, and ~0.2% of total splenocytes were plasma cells (SPPC; Fig. 1a). These frequencies of splenic GCs and SPPCs are consistent with what are generally reported in literature as the background or control condition of unmanipulated animals. At these frequencies, GCs could be seen histologically as sporadic clusters of Ephrin B1-expressing cells20,21 in some but not all B-cell follicles (Supplementary information, Fig. S2). These spontaneous GCs were accompanied by appearance of CD4+CD44+CXCR5hiPD-1hiFoxP3− TFH and CD4+CD44+CXCR5hiPD-1hiFoxP3+ TFR cells (Fig. 1b). We further examined MD4 mice that carry transgenic IgMa B cell receptor (BCR) recognizing hen egg lysozyme. Spontaneous GCs, SPPCs, and bone marrow PCs (BMPC) frequencies were actually comparable between co-housed MD4 and B6 littermates (Fig. 1c, d), and so were frequencies of spontaneous TFH and TFR cells (Fig. 1e). Importantly, however, because essentially none of spontaneous GC, SPPC, and BMPCs in MD4-transgenic animals expressed IgMa MD4 BCR or antibodies (Fig. 1d), these cells must carry BCRs generated by V(D)J recombination at the endogenous immunoglobulin (Ig) locus. Considering the stringent allelic exclusion exerted by transgenic IgMa MD4 BCR, these results support the possibility that spontaneous GCs in a non-autoimmune background may include autoantigen-driven components.

Fig. 1: Spontaneous splenic GC responses.

a, b Spontaneous GCs in WT B6 mice of 2–3 months of age. a Representative contour plots and summary statistics of GC and SPPC frequencies in B220+ B cells and total splenocytes, respectively. b Spontaneous TFH cells in CD4+CD44hiFoxp3− and TFR cells in CD4+CD44hiFoxp3+ cells. c–e Spontaneous GCs in B6 and MD4 mice of 2–3 months of age. c Representative contour plots and summary statistics of GC, SPPC and BMPC frequencies in B220+ B cells, total splenocytes and total BM cells, respectively. For MD4 mice, IgMa+ cells in GC or plasma cells as identified by surface or intracellular staining are also shown. d Summary statistics of total GCs, SPPCs and BMPCs in B6 and MD4 mice and IgMa+ GCs, SPPCs and BMPCs in MD4 mice. e Summary statistics of TFH and TFR frequencies in CD4+CD44hiFoxp3− and CD4+CD44hiFoxp3+ cells. Each symbol represents one mouse, and lines denote mean values. Data were pooled from three independent experiments. P values by Mann–Whitney tests.

Spontaneous GCs depend on T cell help and are constrained by TFR cells

To determine whether spontaneous GCs depend on T-cell help just like conventional GCs induced following immunization or infection, we examined wild-type (WT) B6 mice and mice deficient in SLAM-associated protein (SAP) that is essential for cognate T–B interactions and GC formation.22,23 As shown in Supplementary information, Fig. S3a, 3- to 4-month-old WT mice had ~0.5% of total B cells as GCs, while SAP-deficient mice had essentially none. SPPCs were also significantly reduced in the absence of SAP (Supplementary information, Fig. S3b). Moreover, CD40L blockade essentially abrogated spontaneous GCs in 4 days, while SPPCs was also slightly reduced as a consequence in the same time frame (Supplementary information, Fig. S3c, d). Therefore, spontaneous GCs depend on conventional T-cell help.

Regulatory T (Treg) cells enforce dominant tolerance extrathymically, in part by suppressing activation of self-reactive T cells.24,25,26 Follicular T regulatory (TFR) cells are the Treg subset that regulates the GC response.27,28,29 Similar to TFH cells, TFR cells exhibit a CXCR5+PD-1+ surface phenotype, are localized in the B-cell follicle, and express both the Treg-defining Foxp3 and the TFH-defining Bcl6 transcription factor.30,31,32

To determine whether spontaneous GCs are constrained by TFR cells, we used Treg reporter FOXP3-GFP-hCre (called “Foxp3-cre”) mice.33 As shown in Fig. 2a, under the SPF condition, ~10% CD44hiGFP+ Treg cells in these mice co-expressed CXCR5 and PD-1, exhibiting a TFR phenotype. In Foxp3-cre;Bcl6fl/fl mice, in which the Bcl6 gene is specifically ablated in FoxP3-expressing Treg cells, the FoxP3+CXCR5hiPD-1hiTFR frequency was markedly reduced. By immunohistochemical staining, B-cell follicles in Foxp3-cre mice contained GFP+ TFR cells, whereas much fewer were seen in follicles of Foxp3-cre;Bcl6fl/fl mice (Fig. 2b). For simplicity, we hereafter refer to Foxp3-cre;Bcl6fl/fl mice as TFR-insufficient mice. At 2–3 months of age, TFR-insufficient mice showed a 2- to 3-fold increase in spontaneous GCs, SPPCs, and BMPCs, as compared to their WT counterparts (Fig. 2c–e). As mice aged to 10–12 months, WT mice harbored twice to three times as many spontaneous GCs and SPPCs and four times as many BMPCs as their 2- to 3-month-old counterparts, while TFR-insufficient mice continued to show a two-fold increase over the WT in terms of GCs and SPPCs but not BMPCs, which amounted to ~0.4% of total BM cells in both groups of mice (Fig. 2f–h). This latter observation is consistent with the notion of BMPC survival niche of a finite size.34 In addition, receptor sequencing by 10× Genomics showed that, in TFR-insufficient mice, a significantly higher fraction of spontaneous GCs was isotype-switched, and this increase was particularly pronounced in the SPPC compartment (Fig. 2i). With or without intact TFR cells, most of spontaneous GC B cells harbor somatic mutations (Fig. 2j).

Fig. 2: Spontaneous GCs are constrained by TFR cells.

a, b Validation of TFR insufficiency. a Representative contour plots and summary statistics of CD4+CD44hiCXCR5+PD-1+Foxp3+ TFR cells in CD4+CD44hiFoxp3+ cells in 2- to 3-month-old mice of indicated genotypes. b Representative images (left) and summary statistics (right) of follicular TFR distribution in mice of indicated genotypes. Scale bar, 50 μm. Each symbol represents one mouse (a) or follicle (b), and lines denote mean values. Data were pooled from three independent experiments. P values by two-tailed unpaired t-tests. c–h Representative contour plots and summary statistics of GC frequencies in B220+ B cells (c, f), SPPCs in total splenocytes (d, g) and BMPCs in total bone marrow cells (e, h) in mice of 2 to 3 months (c–e) or 10 to 12 months (f–h) of age. Each symbol represents one mouse, and lines denote mean values. Data were pooled from two (f–h) or three (e) or five (c, d) independent experiments. P values by two-tailed unpaired t-tests. i, j Isotype distribution (i) and somatic hypermutation (SHM) fraction (j) of GCs and SPPCs by single-cell BCR sequencing. Shown are proportions of indicated components in GCs (two pie charts on the top) and SPPCs (two on the bottom) in mice of indicated genotypes. Total numbers of cells sequenced are shown at the center. Cells were pooled from 8 TFR-sufficient and 10 TFR-insufficient mice of 2–3 months of age, respectively. P values by Fisher’s exact tests.

Therefore, similar to conventional GCs induced by immunization or infection, spontaneously formed GCs are dependent on T cell help, constrained by TFR cells, and involve somatic hypermutation and class-switching.

Spontaneous GCs harbor specificities against B cell-derived autoantigen

To explore whether spontaneous GCs contain specificities against BDAs, we set up Nojima culture35,36 to clone and convert single GC B cells to antibody-producing cells (Fig. 3a). The culture supernatant was taken at day 9 and 10 from each well and used to test reactivities to surface molecules of B cells by flow cytometry. MD4 B cells were used as the test target in an effort to minimize potential interference by polyclonal BCR idiotypes.

Fig. 3: Spontaneous GCs recognize B cell-surface autoantigen.

a The protocol for cloning single spontaneous GC B cells from 2–3-month-old TFR-sufficient and -insufficient mice to produce antibody-secreting cells (ASCs) in vitro. b–e Histogram profiles of representative ASC supernatants that were derived from TFR-sufficient (b, d) or -insufficient (c, e) GCs and did (b, c) or did not (d, e) stain the B cell surface, measured with splenocytes from MD4 mice as target cells. f Frequencies of BDA-reactive ASC supernatants. Total numbers of clones analyzed are shown at the centers of pie charts. P value by the χ2 test.

We cloned 236 spontaneous GC B cells from the WT and 108 from TFR-insufficient mice. After IgG concentrations in those culture supernatants were normalized to 2 μg/mL, they were directly tested as the primary staining reagent, with an anti-mouse IgG antibody as the secondary. As exemplified in Fig. 3b, c, many supernatants of GC clones stained components on the B cell surface but showed far less or not at all staining on CD4+ or B220−CD4− splenocytes, while other clones did not react to B cells (Fig. 3d, e). Such B cell-reactive clones accounted for ~19% and ~29% in spontaneous GCs of WT and TFR-insufficient mice, respectively (Fig. 3f; see Supplementary information, Fig. S4 for staining histograms of all tested antibody supernatants). These data indicate that at least 20% of spontaneous GCs recognize B cell-derived surface components and that TFR cells restrain these cells.

FcγRIIB is an autoantigen for spontaneous GCs

In an effort to positively identify surface BDAs, we cloned Ig heavy and light chains from 13 single GC B cells that produced B cell surface-binding antibodies (Fig. 4a; Supplementary information, Table S1). When these IgH/L pairs were expressed in 293 F cells as recombinant IgG1 antibodies, they all demonstrated an ability to recognize antigen on the B cell surface. Interestingly, we found that 3 antibodies (UNW10, K3H12 and W2H5) isolated from different mice were reactive to FcγRIIB, the low-affinity IgG Fc receptor that binds to IgG immune complexes and transmit inhibitory signals to B cells.37 As exemplified in Fig. 4b and quantified in Fig. 4c, those 3 antibodies (20 μg/mL) can positively stain WT but not FcγRIIB-deficient B220+IgD+ naïve B cells, and the staining was abrogated by pre-incubation of the cells with the 2.4G2 anti-FcγRII/III antibody. On the other hand, K2A1 and W3A11, two additional surface BDA-reactive antibodies expressed in the same IgG1 format and tested at the same 20 μg/mL concentration, positively stained both WT and FcγRIIB-deficient B220+IgD+ naïve B cells to a similar extent, and the staining was not affected by 2.4G2 pre-treatment. Recombinant 4-hydroxy-3-nitrophenyl (NP) hapten-specific B1-8hi antibody of the same IgG1 isotype did not show any staining of the B cell surface above the background level achieved with non-specific mouse IgG1 (Fig. 4b, c). Moreover, when mouse FcγRIIB was retrovirally expressed in WT or FcγRIIB-deficient B cells (Supplementary information, Fig. S5), staining of such cells with UNW10, K3H12 and W2H5 was markedly higher than staining of vector-transduced cells (Fig. 4d). These data indicate that these 3 antibodies uniquely recognize FcγRIIB as antigen.

Fig. 4: FcγRIIB is an autoantigen for spontaneous GCs.

a The protocol for producing recombinant antibodies. b Histogram profiles of staining B220+IgD+ B cells of indicated genotypes with 3 spontaneous GC-derived and B1-8hi BCR-derived recombinant antibodies, with or without 2.4G2 blockade. c Mean fluorescence intensities (MFI) of staining as in a. d MFIs of staining B cells that were transduced with FcγRIIB or vector control with the 3 spontaneous GC-derived recombinant antibodies or mouse IgG1 isotype control. e Direct ELISA showing binding of indicated recombinant F(ab)’2 antibodies to the FcγRIIB extracellular domain. One of two independent experiments with similar results are shown.

To completely rule out the remote possibility that the observed differential binding was because, by unknown mechanisms, the FcγRIIB receptor could preferentially recognize Fc portions of UNW10, K3H12 and W2H53 antibodies but not K2A1, W3A11 or B1-8hi antibodies, we made recombinant F(ab)’2 versions of UNW10, K3H12 and W2H5 and control B1-8hi antibodies. By direct ELISA, UNW10, K3H12 and W2H5 but not B1-8hi F(ab)’2 were found to bind to the extracellular domain of the FcγRIIB receptor (Fig. 4e).

FcγRIIB is constitutively expressed by naïve follicular B cells and is upregulated on follicular dendritic cells after GC formation,38 making FcγRIIB an abundant follicular autoantigen. Among UNW10, K3H12 and W2H5 antibodies, except for K3H12 harboring 2 somatic mutations that cause an amino acid replacement in the IgL complementarity-determining region 3 (CDR3), there is no somatic mutation in UNW10 and only a synonymous mutation in W2H5 (Supplementary information, Table S1). We can thus rule out the possibility that the FcγRIIB reactivity was acquired through somatic hypermutation in a GC response against unknown environmental antigen. We conclude that FcγRIIB is a BDA, and B cells specific to this BDA can readily develop into GCs spontaneously in non-autoimmune mice.

Spontaneous TFH and TFR cells preferentially recognize B cell autoantigen

Given the proof of concept provided by the FcγRIIB as a BDA and the estimated > 20% prevalence of BDA specificities in spontaneous GCs, we suspect spontaneous TFH cells would have to recognize B cell-derived autoantigen so as to provide cognate help. As TFR cells clearly constrain spontaneous GCs, spontaneous TFR cells may also preferentially recognize BDAs.

To test BDA-directed specificities, we sequenced T cell receptor (TCR) α and TCRβ chains from sorted single cells by 10× Genomics (see Supplementary information, Fig. S6 for the sorting strategy using Foxp3-IRES-GFP mice). As shown in Fig. 5a, in young adult mice of 2–3 months of age, ~25% of spontaneous TFH (FoxP3-GFP−CD4+CD44+CXCR5+PD-1+) cells were clonally expanded (identical pairs of TCRα and TCRβ DNA sequences; 1667 of 6680 cells), in contrast to ~0.7% in non-TFH (FoxP3-GFP−CD4+CD44+CXCR5−PD-1−)-activated TH cells (22 of 3082). On the other hand, TFR (FoxP3-GFP+CD4+CD44+CXCR5+PD-1+) cells and other Treg (FoxP3-GFP+CD4+CD44+CXCR5−PD-1−) cells contained ~11% and ~8% expanded clones, respectively (512 of 4629 TFR cells vs 332 of 3964 Treg cells). These data indicate more frequent clonal expansion of spontaneous TFH and TFR cells as compared to their respective non-follicular counterparts, suggestive of active and ongoing antigen stimulation that these cells experience in the follicle.

Fig. 5: Spontaneous TFH and TFR cells preferentially recognize B-cell autoantigen.

a Clonal characteristics of TH (FoxP3-GFP–CD4+CD44+CXCR5–PD-1–), TFH (FoxP3-GFP–CD4+CD44+CXCR5+PD-1+), Treg (FoxP3-GFP+CD4+CD44+CXCR5–PD-1–), and TFR (FoxP3-GFP+CD4+CD44+CXCR5+PD-1+) cells. The gray color indicates unique TCR clones without observed repeats, while the red color of increasing shades indicates increasing clonal size (in cell number) of the same TCR. Total numbers of cells sequenced are given in the pie chart. Data were pooled from 5 independent experiments or meta-mice sequenced separately, each of which contained a pool of 8 Foxp3-IRES-GFP mice of 2–3 months of age. b Four TCRs (10 × 1, 10 × 2, Tfh13 and Tfh14) analyzed in detail below, with their CDR3 sequences and their observed counts in indicated cell populations. c Conjugate indices of the four TCRs, with WT B cells (blue), class II MHC-deficient B cells (red) or WT DCs (dark green) as the APC partner. Raw data of conjugate frequencies are provided in Supplementary information, Fig. S6. The horizontal dashed line indicates index 1, at which a TCR confers a conjugation efficiency equal to that of the OT-II TCR. One of two independent experiments with similar results are shown. Bars indicates the standard deviation (STD) of triplicated samples. P values by two-tailed unpaired t-tests. d TCR signal indices of the four TCRs, as assayed in NFAT-GFP reporter hybridoma stimulated with WT B cells (blue), class II MHC-deficient B cells (red) or WT DCs (dark green). Raw data of NFAT-GFP+ frequencies are provided in Supplementary information, Fig S7. One of two independent experiments with similar results are shown. Bars indicates STD of triplicated co-culture wells. P values by two-tailed unpaired t-tests. e The protocol for constructing TCR libraries and testing antigen specificity preference of receptor repertoires (more details in Materials and methods). f Antigen specificity index (ASI) of TCR libraries from indicated T cell populations, as assayed in NFAT-GFP reporter hybridoma stimulated with WT or class II MHC-deficient B cells (blue) or DCs (red). One of two independent experiments with similar results are shown. Bars indicate STD of 4–6 co-culture wells. P values by two-tailed unpaired t-tests.

As shown in Fig. 5b and Supplementary information, Table S2, we further identified αβ TCRs with identical amino acid sequences that appeared in TFH and TFR cells (e.g., 10 × 1) or in TFH, TFR and Treg cells (e.g., 10 × 2). 10 × 1 and 10 × 2 share the same TCRα CDR3 amino acid sequence, and they differ by only one amino acid in TCRβ CDR3 sequences. 10 × 2 TCR was observed in different meta-mice or individual mice that were barcoded and analyzed in multiple separate experiments (see Materials and methods for details). Importantly, 10 × 2 manifested itself in different DNA sequences in different mice or meta-mice (Supplementary information, Table S2) and even between TFH and Treg cells from the same mice (i.e., the 5-4 mouse, Supplementary information, Table S2), indicating a strong selection driven by an abundant common autoantigen. Tfh13 and Tfh14 were the most and the second most abundant clones in spontaneous TFH cells, respectively, and they were not found in TFR or Treg cells (Fig. 5b). Tfh13 was shared by 5-1 mouse and a meta-mouse (meta-3; Supplementary information, Table S2), again indicating selection by common autoantigen.

By retroviral transduction into polyclonal B6 T cells, we tested reactivities of those four TCRs toward B cells. We reason that if these TCRs recognize epitopes derived from B cell autoantigen, such epitopes should be naturally presented by B cells themselves. Therefore, we examined direct conjugate formation between TCR-transduced T cells and WT B cells or B cells deficient in class II major histocompatibility complex (MHC) expression. As compared to control OT-II TCR that was expressed in B6 T cells, 10 × 1, 10 × 2, Tfh13 and Tfh14 TCRs afforded a significant increase in conjugation with WT B cells; such advantage over OT-II TCR was abrogated when class II MHC-deficient B cells were used (Fig. 5c; Supplementary information, S7a, b). When splenic dendritic cells (DCs) were used as the antigen-presenting cells (APCs), those 4 TCRs did not show any advantage over OT-II TCR in conjugation (Fig. 5c; Supplementary information, S7c). These data support that those 4 TCRs recognize BDAs.

To corroborate these findings, we reconstituted mhCD4-NFAT-GFP T cell hybridoma39 (see Materials and methods for details) with those four TCRs or control OT-II TCR. While four test TCRs were expressed at comparable or slightly lower levels than the control OT-II TCR (Supplementary information, Fig. S8a), they all gave rise to more GFP+ cells than OT-II, 24 h after hybridoma cells were co-cultivated with WT B cells but not with class II MHC-deficient B cells (Fig. 5d; see Supplementary information, Fig. S8b–d for cytometry profiles and raw quantitation). When DCs were used as APCs, only 10 × 2 showed a detectable but weaker response while the other three did not detectably respond (Fig. 5d; Supplementary information, Fig. S8b, e). The fact that these TCR-reconstituted hybridoma cells responded more strongly to B cells was not because B cells were generally better APCs, as OT-II TCR-reconstituted hybridoma cells responded much stronger to DCs than B cells when exogenous OVA323-339 peptide was used as the antigen (Supplementary information, Fig. S8f). Therefore, all 4 TCRs isolated from spontaneous TFH or TFR cells preferentially react to B cell-derived but not DC-derived autoantigen.

Encouraged by these results, we systemically tested whether spontaneous TFH and TFR cell populations are selectively enriched with TCR specificities against BDAs. We took advantage of the 1D2 mouse strain that has been created through nuclear transfer of the nucleus from a T cell of unknown specificity40 (a gift from Dr. Shohei Hori). As a result, the 1D2 strain has a recombined Tcrb locus expressing β chain of the 1D2 TCR. The β chain-fixed 1D2 strain was further bred with the FoxP3-hCD2 knock-in reporter strain to obtain Tcrb1D2/1D2Foxp3hCD2/y or Tcrb1D2/1D2Foxp3hCD2/hCD2 mice (Fig. 5e). From these mice, we isolated spontaneous TFH cells, TFR cells, conventional non-TFH activated T cells, and Treg cells (Supplementary information, Fig. S9a) to construct their respective α chain libraries. These α chain libraries were then separately transduced into an NFAT-GFP hybridoma subline that was engineered to stably express 1D2 TCR β chain (Fig. 5e). While such 1D2 β chain-modified hybridoma does not express TCR on the cell surface, introduction of a TCRα chain led to receptor complementation and surface expression of TCR (Supplementary information, Fig. S9b). After 24 h culture of these TCR-reconstituted hybridoma cells together with WT or class II MHC-deficient B cells or DCs, GFP+ frequencies were enumerated in each group (Supplementary information, Fig. S9c–f). By dividing the frequency of GFP+ hybridoma cells after stimulation with WT APCs by the GFP+ frequency achieved with MHC-deficient cells, we define the antigen specificity index (ASI), which reflects the extent to which a polyclonal TCR repertoire can react to autoantigen presented by APCs. As shown in Fig. 5f, TCRs from non-TFH activated T cells or non-TFR Treg cells exhibited an ASI close to 1, no matter whether B cells or DCs were the APC. In striking contrast, TCRs from spontaneous TFH cells exhibited an ASI close to 15 in response to B cells, and ASI of TCRs from spontaneous TFR cells was ~4 in response to B cells. In response to autologous DCs, spontaneous TFH but not TFR TCRs showed a marginally detectable antigen-specific response (Supplementary information, Fig. S9e, f; Fig. 5f). Notably, among the four types of T cells analyzed (TFH, TH, TFR, and Treg cells), the ASI or the strength of their TCR repertoires reacting to B cells matched the relative extent of clonal expansion these cells exhibited (Fig. 5a). Taken together, these data indicate that spontaneous TFH cells and TFR cells are selectively enriched with TCR specificities for BDAs.

In summary, in normal mice with an unmanipulated receptor repertoire, the immune system does not silence all autoreactive clones, particularly not BDA-reactive ones. Instead, follicular homeostasis is maintained in association with ongoing T-dependent GC responses to BDAs.

BDA-specific GCs persist and expand following infection

Given the constitutive nature of spontaneous GCs, GCs induced by immunization or infection would likely be intermingled with BDA-reactive GC cells. To test this possibility, we treated the S1PR2-CreERT2;Rosa-Ai14 reporter strain with tamoxifen to label spontaneous GCs before NP-Keyhole Limpet Hemocyanin (KLH) immunization (Fig. 6a). By flow cytometry, tdTomato labeled only spontaneous GCs (80%–90%) already existed before immunization and did not label any NP-specific GC B cells induced after immunization (Fig. 6b, c), demonstrating the specificity of this approach. As shown in Fig. 6d, while sporadic but tightly packed tdTomato+EFNB1+spontaneous GC clusters were seen in non-immunized mice, large GCs were seen to contain tdTomato+ cells 8 days after NP-KLH immunization (29 out of 32 GCs from 8 mice examined in 2 independent experiments). Analysis of single GC B cells using the protocol in Fig. 3a revealed that 3 of 14 tdTomato+ cells that did not bind to NP were reactive to surface BDAs (Fig. 6e). Therefore, immunization-induced foreign antigen-reactive GCs merge with pre-existing spontaneous GCs. Using the same approach as described in Fig. 5e, we further constructed α chain libraries for TFH cells, TFR cells, conventional non-TFH activated T cells, and Treg cells isolated from NP-KLH-immunized Tcrb1D2/1D2Foxp3hCD2/y or Tcrb1D2/1D2Foxp3hCD2/hCD2 mice. As shown in Fig. 6f and Supplementary information, Fig. S10, TCR repertoires of polyclonal TFH cells and TFR cells following immunization still contained specific reactivities toward B cells. Therefore, BDA-reactive B cells, TFH and TFR cells are a regular component in GC responses even after immunization. Interestingly, we notice that ASI of TFH cells isolated from immunized mice is much lower than that of spontaneous TFH cells, ~3 in the former vs ~15 in the latter, consistent with newly generated immunogen-specific TFH cells being dominant in the total TFH population. On the other hand, ASI of TFR cells isolated from immunized mice is quite similar to the ASI of spontaneous TFR cells, ~2 in the former vs ~3 in the latter, indicating that the TFR repertoire is less affected by immunization.

Fig. 6: GC merging and increase in BDA-specific antibodies during virus infection.

a The protocol for labeling spontaneous GCs prior to immunization. b Representative flow-cytometry profiles, showing GCs in B220+ B cells, NP-binding and non-binding GC cells and their respective tdTomato+ fractions in PBS control or NP-KLH-immunized mice. c Frequencies of tdTomato+ cells in NP non-binding (NP−) or NP-specific (NP+) GC B cells in NP-KLH-immunized mice. Each symbol represents one mouse, and lines denote mean values. Data were pooled from three independent experiments, with all mice of 3–4 months of age. The P value was by a two-tailed unpaired t-test. d The representative distribution of tdTomato+ cells in GCs of PBS control or NP-KLH-immunized mice. Scale bar, 50 μm. e Histogram profiles of representative ASC supernatants derived from single NP–tdTomato+ GC B cells as in Fig. 3a, tested with splenocytes from MD4 mice as target cells. f ASI of TCR libraries from indicated T cell populations of NP-KLH-immunized mice, as assayed in NFAT-GFP reporter hybridoma stimulated with WT or class II MHC-deficient B cells (blue) or DCs (red). g The protocol for labeling spontaneous GCs prior to LCMV infection. h Representative flow-cytometry profiles, showing GCs in B220+ B cells, SPPC in total splenocytes and their respective tdTomato+ fractions in control or LCMV-infected mice 14 days post infection. i Total numbers of tdTomato+ GCs and SPPCs in the spleen of control or LCMV-infected mice 7 and 14 days after infection (8 and 15 days after labeling). Each symbol represents one mouse, and lines denote mean values. Data were pooled from two independent experiments, with all mice of 3–4 months of age. P values by Mann–Whitney tests. j Histogram profiles of staining of B220+IgD+ B cells or CD4+ T cells from FcγRIIB-deficient mice. Primary reagents were sera from control or infected (4-month old) mice, collected 14 days after infection. Mouse IgG1 (0.5 mg/mL) was used as the negative control. k MFIs of staining as in i. Each symbol represents one mouse serum sample, and lines denote mean values. P values by Mann–Whitney tests.

We conducted similar experiments in mice acutely infected with Lymphocytic Choriomeningitis Virus (LCMV) (Fig. 6g, h). First, between the first and the second week after tamoxifen-induced labeling, the number of tdTomato+ GC cells was reduced by ~5 folds, whereas the number of tdTomato+ SPPCs did not change in control mice (comparing day 8 and 15 post labeling in Fig. 6i). These data suggest that spontaneous GCs rapidly turn over in a few days without significant output to SPPCs, consistent with homeostatic maintenance by the regulatory triad. In infected mice, the number of tdTomato+ GCs was comparable to control mice, either on day 8 or 15 after labeling. In striking contrast, the number of tdTomato+ SPPCs in infected mice outstripped that in control mice by 10 folds 7 days after infection (8 days after labeling), and the difference was further increased by another factor of ~10 over the second week after labeling. Moreover, this is despite the fact that tdTomato+ GCs in infected mice went through similar decay over the same period as in control mice. These data indicate that, following viral infection, mixed-in spontaneous GC components become highly productive in giving rise to SPPCs, and/or their progeny SPPCs are massively expanded. Consistent with accumulation of SPPCs derived from spontaneous GCs, which prominently contain BDA reactivities, we observed a markedly increased serum IgG response to B-cell but not T-cell surface autoantigen, tested with FcγRIIB-deficient naïve B and T cells as the target (Fig. 6j, k). The reason to use FcγRIIB-deficient cells as the target was to ensure no interference from Fc binding by FcγRIIB. These results suggest that flares of such auto-antibody production following episodes of viral infections could increase the possibility of developing overt autoimmunity.