TET enzymes control antibody production and shape the mutational landscape in germinal centre B cells

Upon activation by antigen, B cells form germinal centres where they clonally expand and introduce affinity‐enhancing mutations into their B‐cell receptor genes. Somatic mutagenesis and class switch recombination (CSR) in germinal centre B cells are initiated by the activation‐induced cytidine deaminase (AID). Upon germinal centre exit, B cells differentiate into antibody‐secreting plasma cells. Germinal centre maintenance and terminal fate choice require transcriptional reprogramming that associates with a substantial reconfiguration of DNA methylation patterns. Here we examine the role of ten‐eleven‐translocation (TET) proteins, enzymes that facilitate DNA demethylation and promote a permissive chromatin state by oxidizing 5‐methylcytosine, in antibody‐mediated immunity. Using a conditional gene ablation strategy, we show that TET2 and TET3 guide the transition of germinal centre B cells to antibody‐secreting plasma cells. Optimal AID expression requires TET function, and TET2 and TET3 double‐deficient germinal centre B cells show defects in CSR. However, TET2/TET3 double‐deficiency does not prevent the generation and selection of high‐affinity germinal centre B cells. Rather, combined TET2 and TET3 loss‐of‐function in germinal centre B cells favours C‐to‐T and G‐to‐A transition mutagenesis, a finding that may be of significance for understanding the aetiology of B‐cell lymphomas evolving in conditions of reduced TET function.


Introduction
Epigenetic regulation at the level of DNA is largely mediated by covalent addition of a methyl moiety at the 5th carbon of cytosines via DNA methyltransferases (DNMT) [1]. 5-methylcytosine (5mC), enriched in the context of CpG dinucleotides, associates with a local repression of gene expression [2]. Embryonic development is accompanied by a vast turnover of DNA methylation, guiding developmental transitions and ultimately stabilizing cell identity. It has been assumed that the DNA methylation patterns that are established during embryonic development are self-perpetuating and not substantially altered in adult tissues. Yet, mounting evidence suggests that context-dependent changes in DNA methylation may be functional during postnatal lineage priming, commitment and cell function [3][4][5][6][7][8].
Germinal centres are transient structures in secondary lymphoid tissues that are initiated by B cells upon engagement of their cell surface B-cell antigen receptor (BCR) by antigen, such as during pathogen infection or vaccination [25,26]. During the GC reaction, the enzyme activation-induced cytidine deaminase (AID) is vital for the induction of somatic hypermutation (SHM) and class-switch recombination (CSR) at the BCR gene loci. Clonally expanding GC B cells successively reprogramme their transcriptome, eventually differentiating into memory B cells or antibody-secreting cells, that is plasma cells that produce large amounts of high-affinity antibodies.
The transcriptional signatures that separate B-cell progenitors from mature na€ ıve B cells, and mature na€ ıve B cells from functionally distinct progeny are established by ordered and stepwise reconfiguration of the DNA methylome [27][28][29][30][31]. In humans, an estimated 30% of the entire DNA methylome is modified during B-cell development, affecting several million CpG sites [30,32]. B-cell-specific deletion of DNMT1 in mice resulted in a complete block in early B-cell development [33]. Using conditional mouse models, we and others have demonstrated that a combined lack of TET2 and TET3 from early B-cell development on prevents focal DNA hypomethylation at enhancers that are enriched for consensus binding motifs of key B-lineage transcription factors, and consequently corrupts the gene expression programme associated with maturation transitions [34,35]. The transcription factors PU.1 and EBF1 were shown to interact with and possibly recruit TET2 to target regulatory elements [35,36], suggesting a sequence-specific mechanism of TET-mediated DNA demethylation. Double-deficiency of TET2 and TET3 partially blocked pro-B to pre-Bcell transition, impaired BCR light chain rearrangement and resulted in reduced numbers of peripheral mature B cells. Of note, these mature B cells were nonresponsive to T-cell-dependent immunization [34].
Preventing DNA methylation has been shown to terminate GCs [33]. In contrast, much remains to be learned about the role of TET enzymes in the GC response and for humoral immunity. To this aim, we generated a mouse model of acute joint ablation of TET2 and TET3 in established GC B cells, and report a vital role for TET function in antibody-mediated immunity.

Results and Discussion
TET2 and TET3 are expressed throughout B-cell development and terminal differentiation Human and mouse B cells express TET2 and TET3, whereas TET1 is barely detectable throughout development and upon activation [37][38][39][40]. To analyse the relative mRNA expression of TET2 and TET3 in cell populations spanning the entire B-cell lineage, we FACS-sorted developing and mature na€ ıve B-cell subsets from unchallenged wild-type mice and antigen-experienced B cells from mice that were immunized with sheep red blood cells (RBC), and performed quantitative real-time PCR (qRT-PCR).
During B-cell development expression of both, TET2 (Fig. 1A) and TET3 (Fig. 1B) is progressively increasing, peaking in splenic transitional 1 B cells. Such a pattern indicates that the two proteins may have partially overlapping functions in developmental transitions or the selection of immature B cells, potentially controlling the appearance of autoimmune or nonfunctional mature B cells. However, a robust understanding of TET-function in peripheral immature and mature na€ ıve B-cell subsets could only be achieved using conditional Cre transgenes with selective activity in the cell type of interest. As compared to mature na€ ıve follicular (FO) B cells, TET2 and TET3 are substantially down-  in antigen-experienced GC B cells and plasma cells, a result in agreement with a recent report in human GC B cells [37] (compare Fig. 1A and B; FO vs. GC vs. PC). GC B cells cyclically migrate between the GC dark zone (DZ), where they undergo clonal expansion and SHM, and the GC light zone (LZ) where cells expressing a high-affinity BCR are positively selected. Whereas TET3 mRNA is not differentially expressed between the DZ centroblasts (CB) and the LZ centrocytes (CC), TET2 reaches its lowest level in centrocytes. Altogether, these results indicate that TET2 and TET3 might serve both, overlapping and unique functions in antibody-mediated immunity.
Dependent on the mouse strain and animal facility, young mice lacking only TET2 or TET3 display no or only moderate B-cell phenotypes at steady state [34,41,42]. Hence, we reasoned that only combined deletion of TET2 and TET3 may unravel significant TET-dependent effects in B cells upon activation.
Combined loss of TET2 and TET3 impairs plasmablast differentiation in vitro B-cell transit through the GC is accompanied by extensive DNA demethylation, focal methylation gains and an overall increased heterogeneity in DNA methylation patterns [30,31,43]. DNMT1 hypomorphic mice present with a diminished GC size and abundance upon challenge [28]. This suggests that maintenance of a specific DNA methylation pattern during the GC reaction is critical. Enforced DNA demethylation upon in vitro treatment of activated B cells with 5-azacytidine augmented the appearance of plasmablasts in a division-dependent manner [31]. Conversely, inhibition of DNA demethylation might impair plasma cell generation.
Addressing the involvement of TET proteins in this process, we generated Cg1-Cre;Tet2 F/F ;Tet3 F/F mice in which physiologic germ-line Cg1 transcription drives expression of the Cre-recombinase [44]. Using this system, joint Cre-mediated deletion of both Tet genes is expected in a majority of GC B cells upon IgG1-priming. Of note, acute GC B cell-specific Tet deletion circumvents indirect effects caused by extended TETdeficiency during B-cell development.
First, we used a co-culture system that allows the generation and exponential growth of in vitro induced GC (iGC) B cells [45]. In this system, mature na€ ıve B cells are cultured on feeder cells that stably express CD40 ligand and secrete BAFF thus mimicking T cell help. Dependent on the cytokine provided, that is exclusive exposure to IL-4 for 8 days or initial exposure for 4 days to IL-4 followed by IL-21 for another 4 days, this culture allowed us to determine the dependency of iGC B cells on TET-proteins for proliferation, CSR and plasmablast generation.
After 4 days of iGC culture, acute Tet deletion is complete as indicated by qRT-PCR analysis ( Fig. 2A). Within the limited duration of the 8 days culture system, double-deficiency of TET2 and TET3 did not alter cell growth, as indicated by an identical increase in cellularity between control and Cg1-Cre;Tet2 F/F ; Tet3 F/F iGC B cell cultures (Fig. 2B). This is consistent with a comparable fraction of apoptotic cells (Fig. 2C). To confirm in an independent culture system that TET-deficiency does not impact the proliferation of activated B cells, na€ ıve B cells were labelled with a proliferation-tracking dye and stimulated with aCD40/IL-4/IL-21 or LPS/IL-4/IL-5. No alterations in proliferation between the genotypes were observed (Fig. 2D) despite the highly efficient and division-independent deletion of Tet2 and Tet3 after 3 days in culture (Fig. 2E). In TET-proficient B cells, both TET mRNAs were down-regulated in a cell division cycledependent manner, albeit with different kinetics. Whereas TET2 was initially down-regulated and moderately up-regulated in division cycles 5-6, down-regulation of TET3 was only apparent once the cells had divided > 4 times. From these results a picture emerges where GC B cells down-regulate TET proteins to prevent premature terminal differentiation, and up-regulation of TET2 is required for optimal plasmablast differentiation. This is in line with Dominguez et al. [46] who propose that low TET2 levels, such as in patients with clonal hematopoiesis of indeterminate potential that present with TET2 loss-of-function mutations [47], prevent terminal differentiation hence facilitating B lymphomagenesis. The cytokine milieu in the iGC culture system favours CSR from IgM to IgG1 or IgE. CSR to these two isotypes was significantly though moderately impaired in Cg1-Cre;Tet2 F/F ;Tet3 F/F cells (Fig. 2F). Strikingly, IL-21-driven differentiation into CD138 + plasmablasts, antibody-secreting precursors of longlived plasma cells, was strongly diminished (Fig. 2G). Accordingly, the amount of IgG1 and IgE secreted into the medium was significantly reduced in TET2/ TET3 double-deficient iGC B cell cultures (Fig. 2H). The dependence of B cells on TET activity for CSR to IgG1 and plasmablast differentiation could be recapitulated using an independent culture system (Fig. 2I,J). Hence, our data suggest that TET function is essential for proper plasmacytic differentiation. TET2 might serve a dominant role, as it was shown to cause the demethylation of intronic CpGs in the Prdm1 locus encoding BLIMP-1 [46], a key transcriptional repressor for plasmacytic differentiation.

A B
We note that we have not yet assessed a causal relationship between TET activity and the methylation changes during plasmacytic differentiation. While the distribution of 5hmC has been determined in various B-cell subsets [30,31], technical limitations prevent so far a direct identification of the genomic regions where TET2 and TET3 exert their functions.
Targeted hypomethylation at cis-regulatory sites might be initiated by the binding of transcription factors such as EBF1 [35], that subsequently recruit TET proteins allowing for 5hmC deposition. Focally demethylated regions were found enriched for binding motifs for NF-jB and AP-1 or IRF4 and Oct-2 transcription factors during early and late phases of plasmablast differentiation respectively [31]. However, how transcription factor-assisted recruitment of TET enzymes and consequential oxidation of 5mC result in targeted DNA demethylation is still a matter of debate. 5hmC-mediated DNA demethylation may occur in an enzyme-dependent active, or a replicationcoupled passive manner [40]. Given that plasmablast differentiation is tightly coupled to cell division [43], it is likely that passive dilution of oxidized cytosine species accounts for a substantial fraction of the observed global demethylation. In accordance, a recent report has described a progressive replication-dependent loss of 5hmC in activated B cells in the absence of TET function [40]. Nevertheless, as TET2/TET3 double-deficiency does not impair cell growth in culture (Fig. 2B,D) despite preventing plasmacytic differentiation ( Fig. 2G,H,J), our results support the idea that TET function during plasmablast generation does not depend on proliferation. This is consistent with a report by Caron et al. [43] concluding that DNA demethylation during plasmacytic differentiation in vitro is associated with the deposition of 5hmC at regions undergoing targeted demethylation, independent of proliferation. Next, we turned our attention to the humoral immune response in vivo.

TET-deficiency restricts IgG1 production in vivo
We assessed the levels of antibodies circulating in the peripheral blood of TET2/TET3 double-deficient mice at steady-state. Of note, Cg1-Cre-mediated recombination mostly involves IgG1, whereas IgM + B cells largely retain the loxP-flanked gene [44]. In accordance, steady-state levels of IgM in peripheral blood were unchanged in Cg1-Cre;Tet2 F/F ;Tet3 F/F mice (Fig. 3A). However, TET activity was required for optimal IgG1 titres, as indicated by significantly reduced basal IgG1 levels (Fig. 3B). Of note, neither TET2-nor TET3-deficiency resulted in alterations in steady-state IgG1 levels, indicating compensatory potential during steady-state humoral immunity (Fig. 3B).
To determine whether GC B cell-specific TET2/ TET3 double-deficiency manifests defects in the immunization, and ELISA was performed to determine the abundance of serum antibody isotypes. Whereas IgG1 titers increased upon immunization in Cg1-Cre mice, Cg1-Cre;Tet2 F/F ;Tet3 F/F mice displayed only a marginal increase in circulating IgG1 from day 0 to day 11 (Fig. 3B). These in vivo data mirror the defects in antibody production observed in vitro (Fig. 2H), and indicate that complete TET loss-of-function impairs humoral immunity in a GC B-cell-autonomous manner.

Loss of TET-function impairs GC maintenance and CSR upon immunization
To explore the role of TET proteins in the GC response, we used an alum-adsorbed hapten-carrier conjugate, 4hydroxy-3-nitrophenylacetyl (NP) conjugated to chicken gamma globulin (NP-CGG). NP-specific GC B cells that form in response to alum-adjuvanted immunogens predominately switch to IgG1 [48,49]. Joint Tet2 and Tet3 deletion in Cg1-Cre;Tet2 F/F ;Tet3 F/F GC B cells was efficient by day 10 post immunization ( Fig. 4A). Fourteen days post immunization, flow cytometry analysis revealed a substantially decreased overall and NPspecific GC response in Cg1-Cre;Tet2 F/F ;Tet3 F/F mice, whereas the number and fraction of na€ ıve B cells remained unaltered (Fig. 4B,C). Of note, within the fraction of GC B cells NP-specific IgG1 + B cells were further reduced in Cg1-Cre;Tet2 F/F ;Tet3 F/F mice (Fig. 4C, bar graph). A reduced number of GC was also evident when performing immunohistochemistry for PNA and Ki67 on splenic sections of immunized mice (Fig. 4D). Despite the differential expression of TET2 between dark and light zone GC B cells (Fig. 1A), compartmentalization was not phenotypically altered in Cc1-cre;Tet2 F/F ;Tet3 F/F mice (Fig. 4E). Within the total GC B-cell population we observed a striking deficit for CSR from IgM to IgG1 (Fig. 4F), in accordance with the impaired CSR observed in vitro (Fig. 2F). Overall, these results indicate that TET proteins positively control the magnitude of the GC reaction, and facilitate CSR to IgG1.
Ki67 mRNA levels are unaltered in FACS-sorted Cg1-Cre;Tet2 F/F ;Tet3 F/F GC B cells, suggesting that these cells proliferate normally (Fig. 2G). A recent report showed that the bZIP transcription factor BATF recruits TET proteins to TET-responsive regulatory elements in the AID gene locus, promoting 5hmC deposition and demethylation, sustaining chromatin accessibility and bolstering AID expression [50]. Consistent with this report and the observed CSR defect in Cg1-Cre;Tet2 F/F ;Tet3 F/F GC B cells, AID mRNA levels were found decreased by 50% in GC B cells (Fig. 4G) [50]. Despite the fact that optimal AID expression is required for efficient CSR [50], transcriptional regulation of AID may not be the only mechanism by which TET proteins mediate AID function. Recently, AIDmediated recruitment of TET2 to the FANCA locus and associated demethylation has been described in human lymphoma cell lines [39]. Hence, it appears possible that AID recruits TET enzymes to further gene loci during somatic hypermutation. We wondered whether the reduced AID levels would impair affinity maturation in Cg1-Cre;Tet2 F/F ;Tet3 F/F GC B cells.
Cg1-Cre;Tet2 F/F ;Tet3 F/F mice show compromised serum a-NP IgG1 levels Following immunization with NP-CGG, we determined the abundance and affinity of serum In contrast, immunized Cg1-Cre;Tet2 F/F ;Tet3 F/F mice had substantially less circulating NP-specific IgG1 in the peripheral blood as compared to Cg1-Cre control mice (Fig. 5B). Despite a two-fold reduction in NPspecific IgG1, the levels of both total (NP 18 -BSA) and high-affinity (NP 1.7 -BSA) a-NP IgG1 were decreased similarly, resulting in no significant skewing of the affinity ratio in Cg1-Cre;Tet2 F/F ;Tet3 F/F mice (Fig. 5B).
Within the limits of this experimental approach, our results indicate that TET activity compromises a-NP IgG1 quantity but not quality. The lack of phenotype in affinity maturation appeared surprising given the reduced levels of AID in Cg1-Cre;Tet2 F/F ;Tet3 F/F GC B cells (Fig. 4G), and motivated us to explore SHM in more detail.

Loss of TET function does not impair affinity maturation but alters the mutation spectrum
Antibody affinity increases throughout affinity maturation, ultimately leading to the generation of highly protective neutralizing antibodies. The NP-CGG model is amenable to SHM analysis as humoral immune responses against carrier-coupled NP are dominated by B cells carrying the V H 186.2 gene rearrangement [51,52].
We FACS-sorted B220 + CD19 + CD138 À CD95 hi CD38 lo/À IgG1 + cells from the spleens of three Cg1-Cre control and three Cg1-Cre;Tet2 F/F ;Tet3 F/F mice 14 days post immunization. After RNA isolation, a region of 315 nucleotides of the V H 186.2-constant region g1 gene encompassing CDR1 and CDR2 was amplified, sequenced and mutations in 30 unique GC B-cell clones per genotype were analysed by comparison with the germ-line V H 186.2 gene sequence.
The well-characterized tryptophan to leucine switch at position 33 of V H 186.2 (W33L) [53], conferring a 10-fold increase in the BCR affinity for NP, was  Fig. 3. TET-deficiency restricts the levels of circulating IgG1 in vivo. ELISA for IgM (A) and IgG1 (B) was performed on sera from unchallenged (steady state; n = 16 for Cg1-Cre, n = 4 for Tet2 F/F ;Tet3 F/F , n = 8 for Cg1-Cre;Tet2 F/F , n = 6 for Cg1-Cre; Tet3 F/F and n = 15 for Cg1-Cre;Tet2 F/F ;Tet3 F/F mice) and sheep RBC-immunized mice (days 5 and 11; n = 3 for Cg1-Cre and n = 3 for Cg1-Cre;Tet2 F/F ;Tet3 F/F ). Data are shown as mean AE SD. *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.0001 vs Cg1-Cre;Tet2 F/F ;Tet3 F/F mice (One-way ANOVA and Tukey post-hoc test for multiple comparisons; unpaired Student's t-test when comparing two groups). comparably frequent in Cg1-Cre control (55%) and Cg1-Cre;Tet2 F/F ;Tet3 F/F (54%) GC B-cell clones (Fig. 6A). These data are in accordance with the ELISA results in Fig. 5. Furthermore, NP-specific GC B cells from Cg1-Cre;Tet2 F/F ;Tet3 F/F mice exhibited similar fractions of amino acid replacement and silent mutations across the entire length of V H 186.2 (Fig. 6B). Hence, SHM appears phenotypically unaltered in GC B cells deficient for TET2 and TET3. Our current analysis is limited to the hapten NP, a simple antigen where a few specific mutations suffice for efficient affinity maturation. It will be interesting in the future to test the impact of TET activity on the formation of broadly neutralizing antibodies upon infection with pathogens carrying complex antigens, such as HIV or influenza that are characterized by extraordinary frequencies of affinity-enhancing mutations.
Methylated CpG dinucleotides are a mutation hotspot in human cancers, a phenomenon that has been linked to the propensity of 5mC to deaminate. Hydroxymethylation of cytosines has been reported to enhance [54] or reduce [55] the likelihood of mutations. Hence, the modification state of cytosine likely influences the mutability of DNA, however, the relationship between mutability and cytosine modifications remains poorly understood. We found that Cg1-Cre;Tet2 F/F ;Tet3 F/F GC B cells carried slightly less somatic mutations per V H 186.2 sequence (Fig. 6C), though this difference did not reach statistical significance. Strikingly, however, V H 186.2 mutations in TET-deficient GC B cells exhibited C-to-T and G-to-A transition mutation biases (Fig. 6D). In GC B cells, mutagenesis is largely a consequence of AID activity [56]. In accordance with our results, B-cellspecific Tet2-deficiency promoted age-dependent transformation in a fraction of mice, an outcome that strictly required AID and involved a bias towards Gto-A and C-to-T mutations in the established tumours [38]. This suggests that TET proteins may at least partially exert their tumour suppressor activity in GC B cell-derived lymphomas by orchestrating the mutagenic potential of AID. One possible scenario is that cytidine deaminases (AID and APOBEC1-3) have substantially lower activity on 5hmC compared to 5mC [57,58]. In addition, deamination of 5mC produces thymine, whereas 5hmC deaminates to 5-hydroxymethyluracil, an atypical base that might be repaired more efficiently [55]. Alternatively, 5hmC generated by TET enzymes may promote chromosome stability by supporting DNA damage repair [59]. Up to 10% of human DLBCL exhibit TET2 loss-of-function mutations, and these mutations usually occur mutually exclusive with mutations in regulators of TET function [46,60]. Finally, reduced TET function may actively impact on the aetiology of GC B-cell-derived tumours by shaping the mutational landscape or increasing chromosomal aberrations.

Concluding remarks
The major goal of this study was to explore whether TET activity affects the magnitude or quality of humoral immune responses. Our data emphasize an essential role of TET function in GC expansion, plasmacytic differentiation and antibody production. The observed relative reduction in CSR correlated with decreased AID levels.   In our model system, affinity maturation did not depend on TET function and was not limited by reduced AID levels. Strikingly, however, TET proteins appear to influence the mutagenic potential of AID as TET deficient GC B cells showed mutational skewing towards transition mutations, a function that may be of relevance for the aetiology of GC B-cell-derived lymphomas and potentially further tumour entities.
To determine cell proliferation in the presence of mitogens in vitro, B cells were labelled with 10 lM Cell Proliferation Dye eFluor 450 (Thermo Fisher Scientific, 65-0842-90) as per manufacturer's instructions and placed in culture.

Flow cytometry analysis of iGC B cells
To collect iGC B cells from 10 cm dishes, 9/10 of the medium was gently removed, and the cells were incubated for 5 min at 37°C in 3 mL harvest buffer (PBS/0.5% BSA/ 2 mM EDTA). Cells were harvested by addition of B-cell medium and vigorous pipetting. For CSR analysis, 5 9 10 5 iGC B cells were washed twice with 1 mL PBS at room temperature and 460 g for 4 min. Subsequently, the cells were resuspended in 250 lL of 37°C 0.05% trypsin (Sigma, T4174) and surface digested for 4 min at 37°C. To stop trypsin digestion, 3 mL of staining buffer (described above for flow cytometry) were added, the cells were pelleted at 1800 r.p.m. for 4 min and the supernatant was removed. The cells were transferred to a U-bottom 96 well plate and dead cells were labelled with the Fixable Viability Dye eFluor 780 (Thermo Fisher Scientific, 65-0865-14) as per manufacturer's instructions. Subsequently, cells were fixed and processed using the active Caspase 3 staining kit (BD Biosciences, 550914) as per manufacturer's instructions. Before staining with the active Caspase 3 antibody, the iGC B cells were incubated with 1 lgÁmL À1 of aCD16/32 Fc-Block (BioLegend, 101310) in 25 lL perm/wash buffer for 10 min at 4°C. Subsequently another 25 lL of aIgE-bio (BD Biosciences, 553419, 1 : 400) diluted in perm/wash buffer was added, the cells were further incubated for 15 min at 4°C and washed with perm/wash buffer at 568 g for 2 min. Next, the cells were incubated with 25 lL of antibody mix containing aIgM-PeCy7 (Thermo Fisher Scientific, 25-5790-81, 1 : 500), aIgG1-FITC (BD Biosciences, 562026, 1 : 300), aCD19-BV605 (BioLegend, 115540, 1 : 300) and Streptavidin-APC (BioLegend, 405207, 1 : 800) for 15 min at 4°C. The cells were washed with perm/wash buffer at 2000 r.p.m. for 2 min and stained for active caspase 3 as per manufacturer's instructions.
The processing of cells for plasmablast flow cytometry analysis was performed at room temperature in the dark. Data were acquired on an LSRII cytometer (BD Biosciences) and analysed using FLOWJO software (Tree Star). Nonsinglet events were excluded from analyses using FSC-H/FSC-W and SSC-H/SSC-W characteristics.
Cell sorting was carried out on a FACS Aria III (BD Biosciences). Nonsinglet events were excluded from analyses based on characteristics of FSC-H/FSC-W and SSC-H/ SSC-W.

Quantitative real-time-PCR
RNA from snap-frozen cell pellets of FACS-sorted B cells was isolated using the Quick-RNA Micro Prep Kit (Zymo Research, Irvine, CA, USA, R1050) and DNase digestion as per manufacturer's instructions. RNA from snap-frozen in vitro-cultivated B cells was isolated using Trizol reagent (Thermo Fisher Scientific, 15596026) as per manufacturer's instructions. DNA was removed using the RQ1 RNAsefree DNase (Promega, Madison, WI, USA, M610A), and RNA was retrieved using GlycoBlue Coprecipitant (Thermo Fisher Scientific, AM9515) as per manufacturer's instructions. First-strand cDNA was generated from 100 ng of total RNA using the iScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA, 170-8891) and cDNA was amplified using the AceQ qPCR SYBR Green Master Mix (Vazyme biotech, Nanjing, China, Q111-02) as per manufacturer's instructions. The qRT-PCR was run on a StepO-nePlus Real-time PCR system (Applied Biosystems, Foster City, CA, USA) and melt curve analysis was performed for every run. The expression of individual mRNAs was normalized to HPRT with the following formula: fold induction = 2 ðÀDCtÞ , where DC t ¼ C t ðtargetÞ À C t ðHPRTÞ . The following primers were used: HPRT F: ELISA For NP-specific serum IgG1 and IgM titres, ELISA was performed as described in [62]. For total immunoglobulin levels in serum or secreted immunoglobulin in iGC culture medium, 50 lgÁmL À1 capture antibody (Southern Biotech, 1010-01) was coated overnight onto 96-well enzyme-linked immunosorbent assay plates at 4°C (Sigma, CLS3590). Plates were washed three times with wash buffer (PBS containing 0.05% TWEEN 20), blocked with 100 lL per well 1% BSA in PBS for 4 h at room temperature and washed three more times with wash buffer. Subsequently, wells were incubated over night at 4°C with 100 lL per well of mouse serum serially diluted 1 : 4 in blocking buffer (range 1 : 800 to 1 : 160 000). Plates were washed three times with wash buffer and incubated with 100 lL per well of HRPconjugated a-mouse IgG1 (Southern Biotech, 1070-05, 1 : 5000 in 1% BSA in PBS) or HRP-conjugated a-mouse IgM (Southern Biotech, 1020-05, 1 : 5000 in 1% BSA in PBS) for 4 h at room temperature. For detection, 100 lL of ABTS substrate solution per well [200 lL ABTS (Stock: 15 mgÁmL À1 in a.d.), 10 mL citrate-phosphate buffer (574 mg citric acid monohydrate in 50 mL a.d.) and 10 lL H 2 O 2 ] was incubated for 20 min. Absorbance was measured at 405 nm using a microplate reader (Tecan Sunrise, M€ annedorf, Switzerland).

Immunohistochemistry
IHC was performed on 7 lm sections of murine spleens as described in Ref. [62]. Imaging was performed on a Zeiss Axioplan 2 microscope using a 109 objective (Zeiss, Oberkochen, Gemany) and the Axiocam 305 colour camera (Zeiss). Pictures were generally processed with the ZEISS ZEN BLUE 2.5 LITE software (Zeiss).

VH186.2 SHM analysis
Somatic hypermutation analysis was performed according to Ref. [63] for three Cg1-Cre and three Cg1-Cre;Tet2 F/F ; Tet3 F/F mice. Briefly, IgG1 + GC B cells were FACS-sorted (CD19 + B220 + CD138 À Fas hi CD38 lo/À IgG1 + ), pelleted and total RNA was isolated using the Quick-RNA Micro Prep Kit (Zymo Research, R1050) including the DNase digestion step as per manufacturer's instructions. First-strand cDNA was generated from 100 ng of total RNA using the Promega GoScript Reverse transcription system (Promega, A5000) and a gene-specific primer for the IgG1 locus: Cg1 5 0 -CATGGAGTTAGTTTGGGCAG-3 0 . Subsequently, two rounds of semi-nested PCRs were performed using the Herculase II Fusion DNA polymerase kit (Agilent, Santa Clara, CA, USA, 600675) as per manufacturer's instructions. Following each round of PCR, the gene product was run on an agarose gel and purified using the peqGold Gel extraction kit (VWR, Radnor, PA, USA, 732-2777) as per manufacturer's instructions. For the first PCR, the following primers were used: V186.2 leader 5 0 -AGCTGTAT-CATGCTCTTCTTGGCA-3 0 , Cg1 5 0 -CATGGAGTTAGT TTGGGCAG-3 0 . For the second PCR, the following primers were used: V186.2 nested 5 0 -CATGCTCTTCTTGG-CAGCAACAG-3 0 , Cg1 5 0 -CATGGAGTTAGTTTGGG CAG-3 0 . For cloning of VH186.2 segments, the pJet1.2 blunt vector was used and cleaved with EcoRV (New England BioLabs, Ipswich, MA, USA, R0195). The purified PCR products were ligated into the vector using T4 ligase (Promega, M1794) as per manufacturer's instructions. Using a standard transformation protocol, DH5a bacteria were transformed with the bulk of ligated plasmids, spread onto LB Agar/ampicillin plates and incubated over night at 37°C. For each of the six mice, >10 colonies were transferred into a PCR Master Mix (Vazyme Biotech, P211), and the very same colonies were selected on new LB agar plates. For the colony PCR, the following primers were used: pJet1.2 F 5 0 -CGACTCACTATAGGGAGAGCGGC-3 0 , pJet1.2 R 5 0 -AAGAACATCGATTTTCCATGGCAG-3 0 . Cycle conditions were: 1 cycle: 95°C for 1 min; 30 cycles: 95°C for 15 s, 56°C for 15 s, 72°C for 45 s; 1 cycle: 72°C for 3 min. The PCR products were loaded onto an agarose gel, and colonies with correct insertions were grown in 2 mL LB medium over night. Plasmid DNA was extracted using the Monarch Plasmid Miniprep Kit (New England BioLabs, T1010L) as per manufacturer's instructions. Subsequently, Sanger sequencing was performed using the pJet1.2 F primer detailed above. Sequences were aligned to the V H 186.2 germ line sequence using VBASE2 (http://www.vbase2.org/). Mutation data were only used if the entire sequence was intact, and only unique sequences were included in the analysis. According to these criteria, two sequences from Cg1-Cre;Tet2 F/F ; Tet3 F/F GC B cells had to be excluded.

Quantification and statistical analysis
Results are always shown as mean and standard deviation (SD). Graphs were plotted and statistical analysis was performed with GRAPHPAD PRISM 7 software (San Diego, CA, USA) using Student's t test when comparing two groups, One-way ANOVA and Tukey post hoc test when comparing multiple groups and Mann-Whitney test for SHM analysis. The number of biological repetitions (n) is stated in each figure legend, and every experiment was performed at least twice. Differences between groups were considered statistically significant when P < 0.05. In figures, asterisks stand for: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.