FcγRIIb differentially regulates pre-immune and germinal center B cell tolerance in mouse and human

Several tolerance checkpoints exist throughout B cell development to control autoreactive B cells and prevent the generation of pathogenic autoantibodies. FcγRIIb is an Fc receptor that inhibits B cell activation and, if defective, is associated with autoimmune disease, yet its impact on specific B cell tolerance checkpoints is unknown. Here we show that reduced expression of FcγRIIb enhances the deletion and anergy of autoreactive immature B cells, but in contrast promotes autoreactive B cell expansion in the germinal center and serum autoantibody production, even in response to exogenous, non-self antigens. Our data thus show that FcγRIIb has opposing effects on pre-immune and post-immune tolerance checkpoints, and suggest that B cell tolerance requires the control of bystander germinal center B cells with low or no affinity for the immunizing antigen.

T hree major tolerance checkpoints during B-cell development control the generation of autoreactive B-cell clones. The first two checkpoints shape the pre-immune repertoire, while the third occurs following antigen-mediated activation 1 . First, the self-reactivity of immature B cells in the bone marrow (BM) is tested, with a large fraction of autoreactive clones being eliminated (a process called central tolerance) [2][3][4] . Autoreactive B cells that manage to leave the BM are subjected to a second checkpoint as they transit from immature to mature B cells in the blood and spleen, leading to a further decrease in the frequency of autoreactive clones among mature B cells 5,6 . Finally, following B-cell activation, autoreactive B cells may also be eliminated from the germinal center or corrected by somatic hypermutation, further reducing the chance of generating potentially harmful autoreactive memory B cells and autoantibody-producing plasma cells. The last two checkpoints, taking place outside the BM, contribute to peripheral tolerance 1,6,7 . At a cellular and molecular level, several mechanisms of B-cell tolerance have been described. Clonal deletion and anergy were first proposed as major mechanisms regulating B-cell tolerance 2,3,5,[8][9][10] . Receptor editing has also been shown to play a key role in central tolerance in the BM 4,11,12 . Autoreactive B cells can also be excluded from the B-cell follicles, with a lack of T cell help precipitating their subsequent apoptosis 13,14 . The nature and location of specific autoantigens seem to determine which tolerance mechanism is used, with deletion being preferentially induced by high-avidity antigens while low-avidity interactions tend to drive receptor editing and anergy [9][10][11] . Germinal center (GC) tolerance mechanisms are still only partially characterised. It has been suggested that autoantigens need to be present within the germinal center to properly control autoreactive clones, either by apoptosis 15,16 or by redemption of their B-cell receptor (BCR) via corrective somatic hypermutation 7,17,18 .
The strength of BCR signalling is thought to be important in all of these mechanisms, suggesting that inhibitory receptors regulating the BCR threshold of activation may be central actors in B-cell tolerance. The inhibitory receptor FcγRIIb negatively regulates signalling induced by the BCR 19,20 and is involved in the control of autoimmunity; however, its role in the regulation of specific B-cell tolerance checkpoints has not been determined.
This question is important, as reduced FcγRIIB expression or function has been associated with autoimmune disease in both humans and mice. In humans, a single-nucleotide polymorphism in FCGR2B leading to the replacement of an isoleucine by a threonine at position 232 (T232I) results in reduced inhibitory function 21,22 , and has been associated with susceptibility to SLE [23][24][25][26][27] , but protection against malaria 26,28 . Humanised mice reconstituted with cord blood cells bearing the 232T polymorphism display defective B-cell development and produce autoantibodies 29 . Naturally occurring variations have also been described in the promoter of human FCGR2B, and it has been suggested these too may predispose to SLE 30,31 . FcγRIIb appears to play a remarkably similar role in mouse and human biology. Thus, FcγRIIb-deficient mice present an exacerbated immune response 32,33 , are prone to inducible and spontaneous autoimmunity [34][35][36][37][38] with more autoreactive GC B-cell clones 39 , and are protected from malaria 26 . Transgenic mice overexpressing FcγRIIb on B cells display a reverse phenotype, with reduced immune responses and resistance to autoimmune disease 40 . Interestingly, natural variants common in wild mice from the Mus musculus genus have also been reported in the promoter region of Fcgr2b 41 , and have been shown to be associated with autoimmunity-prone laboratory strains 42,43 . By generating a knock-in (KI) mouse model bearing these wild promoter variants on the C57BL/6 background (FcγRIIb wild/H1 KI), we demonstrated that they control the upregulation of FcγRIIb on B cells upon activation. In the absence of FcγRIIb upregulation on GC B cells, we observed an increased number of total GC B cells following T-dependent immunisation, but, surprisingly, the number of those specific for the immunising antigen was normal 44 . This suggested that FcγRIIb upregulation may be necessary to prevent bystander B cells, that is those with no or negligible affinity for the immunising antigen, from persisting or expanding in the GC. Moreover, we observed a transient production of autoantibodies following exogenous antigenic challenge in our FcγRIIb wild/H1 KI mice, demonstrating that tight regulation of FcγRIIb expression is necessary to maintain B-cell tolerance 44 .
Here, we use an experimental system enabling us to monitor the frequency of autoreactive (or potentially autoreactive) B cells throughout B-cell differentiation and activation, in the context of different levels of FcγRIIb expression and autoantigen exposure. Our results show that a single inhibitory receptor can have contrasting effects on different tolerance checkpoints, limiting pre-immune while promoting post-immune tolerance. Our results suggest a novel mechanism of GC tolerance, in which FcγRIIb prevents the emergence of bystander autoreactive B cells. Moreover, they also suggest that, in this context, GC tolerance is dominant over pre-immune checkpoints for controlling the development of autoimmunity.

Reduced FcγRIIb expression increases central tolerance.
Absence or reduced expression of FcγRIIb on B cells has been shown to drive autoimmunity. We thus set out to determine its impact upon the different checkpoints of B-cell tolerance. We used mouse strains expressing reduced, normal or increased levels of FcγRIIb (FcγRIIb-deficient mice (KO), FcγRIIb WT mice (WT) and FcγRIIb-BFcR transgenic mice (BTG) that overexpress FcγRIIb on B cells only 40 ), crossed to the SW HEL mouse strain. In SW HEL mice, a transgene encoding an Ig heavy chain specific for the protein Hen Egg Lysozyme (HEL) has been knocked-in to the endogenous heavy chain locus, allowing class-switching and somatic hypermutation. These features, together with the fact that SW HEL mice can generate a polyclonal immune response against other antigens as only 10-30% of their B cells are specific for HEL, enables the physiological tracking of a GC B-cell response following immunisation 45 .
We first evaluated the role of FcγRIIb in central tolerance against a membrane-bound autoantigen using the membrane-HEL (mHEL) mouse strain 5 . Lethally irradiated wild-type or mHEL mice were reconstituted with bone marrow from SW HEL -FcγRIIb KO, SW HEL -FcγRIIb WT or SW HEL -FcγRIIb BTG donor mice (Fig. 1a). As previously described 45 , the frequency and absolute number of HEL-specific BM immature B cells was significantly reduced in mHEL compared with wild-type recipients irrespective of the donor used, consistent with clonal deletion (Fig. 1b and Supplementary Fig. 1a). Interestingly, deletion was enhanced in mHEL mice reconstituted with SW HEL -FcγRIIb KO BM cells and reduced in mHEL mice reconstituted with SW HEL -FcγRIIb BTG BM cells compared to WT controls (Fig. 1c). Thus, reduced FcγRIIb expression increases the deletion of autoreactive clones at the central tolerance checkpoint.
Reduced FcγRIIb expression enhances peripheral tolerance. After leaving the BM, immature B cells experience another round of selection at the transitional B-cell stage 6 . We thus determined whether FcγRIIb expression had an effect on the frequency and number of splenic HEL-specific immature and mature B cells ( Fig. 2a and Supplementary Fig. 1b, c). Absence of FcγRIIb expression was associated with more stringent counter-selection of HEL-specific B cells at all stages from transitional T1 to mature follicular and MZ B cells when compared with WT controls (Fig. 2b-g). Supporting this observation, B cell-specific overexpression of FcγRIIb was associated with increased frequency of HEL-specific B cells in mHEL mice reconstituted with SW HEL -FcγRIIb BTG BM cells compared with KO, and in some populations WT, groups, consistent with reduced tolerance at the transitional T2/T3 (Fig. 2d) and mature follicular (Fig. 2e, f) Bcell stages. Low but detectable levels of anti-HEL IgG1 were seen in sera of unimmunised mHEL mice reconstituted with BM from all SW HEL mice ( Supplementary Fig. 1d). This observation makes it possible that the effect of FcγRIIb on selection at this stage might be driven by co-cross-linking of the BCR and FcγRIIb with cognate antibody, although we cannot exclude an effect of FcγRIIb on tonic BCR signalling independently of antigen recognition and of FcγRIIb co-crosslinking 46,47 .
To confirm the role of FcγRIIb in clonal deletion at these central and peripheral tolerance checkpoints, we crossed the SW HEL -FcγRIIb KO-CD45.2 and the SW HEL -FcγRIIb BTG-CD45.1 mouse strains with mHEL mice (Supplementary Fig. 2). In this experimental model, in contrast to BM chimeras where HEL is only expressed by radiation-resistant recipient cells, all cells including hematopoietic cells express the membrane-bound HEL autoantigen. We observed a reduced frequency of HELspecific B cells in SW HEL -FcγRIIb-KO × mHEL compared with SW HEL -FcγRIIb-BTG × mHEL mice ( Supplementary Fig. 2b-  b Representative flow plots of immature B cells in the bone marrow (B220 + IgM + , left top panel). B cells originating from the donor bone marrow were identified based on FcγRIIb expression (detected with an anti-CD16/32 Ab) and expression of the congenic marker CD45.1 (right top panels). The frequency of HEL-specific cells for each condition is showed using the smoothing function to help the visualisation of rare events (bottom panels). c Quantification of the frequency of HEL-specific immature B cells in WT (left panel) and mHEL (right panel) recipient mice. mHEL recipients: n = 12-19 mice per group; wt recipients: n = 8-11 mice per group. The mean is represented and each dot corresponds to an individual mouse. Four pooled experiments are shown. The p-values were determined with the Kruskal-Wallis (KW) non-parametric test when comparing the three groups (KW: the pvalue is indicated in purple for the mHEL recipient. All KW tests were not significant for WT recipients) and with the Mann-Whitney non-parametric test when two conditions were compared (indicated by black stars). *p < 0.05; **p < 0.01; ***p < 0.001. Comparisons generating non-significant p-values are indicated with "ns". Source data are provided as a Source Data file FcγRIIb limits autoreactive B-cell anergy. We therefore tested whether FcγRIIb expression controls B-cell anergy 10,13,45 in addition to clonal deletion. We measured anergy by ex vivo crosslinking of the BCR and analysis of the phosphorylation of a downstream kinase, normally reduced in anergic B cells. The intensity of the phospho-Syk staining was equivalent between HEL-specific (HEL+) and non-HEL-specific B cells (HEL−) in WT recipients independent of FcγRIIb expression, that is, the ratio of the geometric mean expression of phospho-Syk in HEL+ to HEL− B cells was 1 (Fig. 3a, b). This ratio was less than 1 in   mHEL mice reconstituted with SW HEL -FcγRIIb-WT BM, consistent with increased anergy among the remaining HEL+ B cells that have avoided deletion (Fig. 3a, b). This ratio was further reduced to 0.5 for mice reconstituted with SW HEL -FcγRIIb-KO BM, demonstrating reduced phosphorylation of Syk upon BCR engagement on HEL-autoreactive B cells in absence of FcγRIIb, and conversely was around 1.5 for mHEL mice reconstituted with SW HEL -FcγRIIb-BTG BM (Fig. 3a, b). Moreover, autoreactive HEL+ B cells proliferated less than HEL− B cells in response to LPS as shown by reduced CFSE dilution (Fig. 3c, d), and this reduced proliferation was more marked in the absence of FcγRIIb, suggesting that, indeed, FcγRIIb expression controls autoreactive B-cell anergy (Fig. 3c, d). Consistent with this, the frequency of HEL+ plasmablasts (CD138 + B220 lo ) was higher in mHEL recipients reconstituted with SW HEL -FcγRIIb-BTG and SW HEL -FcγRIIb-WT than with SW HEL -FcγRIIb-KO BM (Fig. 3e, f). Hence, our results show that absence of FcγRIIb expression promotes increased anergy, with reduced signalling, proliferation and differentiation of HEL-specific autoreactive B cells.
FcγRIIb limits autoreactive GC response against mHEL. Thus FcγRIIb expression regulates pre-immune tolerance, with FcγRIIb-deficient mice showing enhanced deletion and anergy compared with their control counterparts. This is at odds with the observation that FcγRIIB defects are associated with autoimmunity in both mice and humans, raising the question of which tolerance checkpoint is defective in absence of FcγRIIB. We therefore assessed GC tolerance, generating chimeras as above and inducing GCs by immunising with HEL conjugated to sheep red blood cells (SRBC-HEL) (Fig. 4a). Following SRBC-HEL immunisation of mHEL recipients reconstituted with SW HEL -FcγRIIb-KO BM cells, the frequency and number of HEL-specific follicular B cells were reduced (Fig. 2, Supplementary Figs. 1 and 3, as expected, due to enhanced pre-immune tolerance) but despite this, HEL-specific GC B-cell number was increased compared with SW HEL -FcγRIIb-WT and SW HEL -FcγRIIb-BTG controls ( Fig. 4b-d). The ratio of the frequencies of GC versus follicular HEL+ B cells was significantly increased in mHEL mice reconstituted with SW HEL -FcγRIIb-KO BM cells compared with the two other experimental groups, emphasising the large expansion of HEL-autoreactive GC B cells in the absence of FcγRIIb (Fig. 4e). Despite this, the frequency and number of T FH and T FR cells were not significantly different ( Supplementary  Fig. 4). To determine if this increase in HEL-specific autoreactive GC B cells led to the development of HEL-specific plasma cells, we performed ELIspot, and observed more HEL-specific antibody-forming cells (AFCs) in the spleen of mHEL recipients reconstituted with SW HEL -FcγRIIb-KO BM compared with the the SW HEL -FcγRIIb-WT and -BTG groups (Fig. 4f). The size and intensity of the spots were similar between the different experimental conditions ( Supplementary Fig. 5), consistent with equivalent antibody output and in vitro survival. Thus, despite a mature B-cell repertoire shaped by enhanced clonal deletion and anergy at pre-immune tolerance checkpoints, mice deficient in FcγRIIb have a marked defect in the negative-selection of autoreactive GC B cells.
FcγRIIb limits autoreactive GC response against sHEL. To assess if the influence of FcγRIIb on B-cell tolerance was dependent on the strength of BCR signalling, we next used mL5 transgenic mice that express soluble HEL, which binds the BCR with less avidity than mHEL 2 , as an autoantigen (Fig. 5a). As expected from the previous studies 2 , the deletion of HEL-specific B cells was less pronounced in mL5 recipients compared with mHEL recipients (Figs. 1, 2 and 5). Nevertheless, deficient FcγRIIb expression still reduced the frequency of HEL-specific BM immature B cells in mice reconstituted with SW HEL -FcγRIIb-KO compared with SW HEL -FcγRIIb-WT or SW HEL -FcγRIIb-BTG BM (Fig. 5b). The frequency and number of HEL-specific autoreactive B cells in the marginal zone were also decreased in absence of FcγRIIb ( Fig. 5d) compared with SW HEL -FcγRIIb-BTG BM, whereas the frequency and number of autoreactive non-GC follicular B cells was not significantly affected by FcγRIIb expression (Fig. 5c).
Following immunisation, we observed an increase in the number of HEL-specific GC B cells (Fig. 5e) and in the ratio of HEL-specific GC to follicular B cells ( Fig. 5f) in sHEL recipient mice reconstituted with SW HEL -FcγRIIb-KO BM compared with mice reconstituted with SW HEL -FcγRIIb-BTG BM. Finally, sHEL recipients reconstituted with SW HEL -FcγRIIb-KO BM had an increased number of HEL-specific AFCs compared with the two other groups (Fig. 5g). Thus, whether autoantigen is soluble or membrane-bound, FcγRIIb deficiency enhances pre-immune and reduces GC tolerance.
FcγRIIb controls bystander autoreactive GC B cells. We previously generated a knock-in mouse model bearing naturally occurring polymorphisms of the Fcgr2b promoter common in wild mice and associated with spontaneous autoimmunity in inbred strains. These FcγRIIb wild/H1 KI mice have defective upregulation of FcγRIIb upon activation, leading to an increase in total GC B cell number, but no change in the number of antigenspecific GC B cells after a T-dependent immunisation 44 . This suggested that FcγRIIb expression might regulate the number of bystander GC B cells, that is, those displaying no or very low affinity for the immunising antigen. Moreover, the FcγRIIb wild/H1 KI mice also produced anti-dsDNA and anti-chromatin autoantibodies transiently following exogenous antigenic challenge 44 . Consistent with this, SRBC immunised FcγRIIb-deficient mice produced significantly more anti-dsDNA autoantibodies than BTG mice (Fig. 6a). These findings raised the possibility that FcγRIIb might contribute to GC tolerance by restricting the Fig. 2 Reduced FcγRIIb expression increases peripheral tolerance at the transitional B-cell stage. a Representative flow plots of immature and mature B cells in the spleen: immature and mature B cells were identified as B220 + CD93 + and B220 + CD93 − , respectively. Amongst the immature B cells, transitional T1 and T2/T3 were identified based on the expression of IgD and CD23 (T1 = IgD lo CD23 lo ; T2/T3: IgD hi CD23 hi ). Mature B cells were separated as follicular (CD23 + CD21 + ) and marginal zone (CD23 lo CD21 hi ) (left top three panels). On each subset, B cells originating from the donor bone marrow were identified based on the expression of FcγRIIb (CD16/32) and of the congenic marker CD45.1 (right top three panels). The frequency of HEL-specific cells was determined (bottom panels). b-g Quantification of the frequency of HEL-specific B cells in WT (left panel) and mHEL (right panel) recipient mice: total immature B cells (b), T1 transitional B cells (c), T2/T3 transitional B cells (d), total mature B cells (e), follicular B cells (f) and marginal zone B cells (g). mHEL recipients: n = 14-20 mice per group; wt recipients: n = 7-8 mice per group. The mean is represented and each dot corresponds to an individual mouse. Four pooled experiments are shown. The p-values were determined with the Kruskal-Wallis (KW) non-parametric test when comparing the three groups (KW: the p-value is indicated in purple for the mHEL recipient. All KW tests were not significant for the WT recipients) and with the Mann-Whitney non-parametric test when two conditions were compared (indicated by black stars). *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. Comparisons generating non-significant p-values were indicated with "ns". Source data are provided as a Source Data file We then asked whether, in the absence of FcγRIIb expression, autoreactive bystander B cells could expand in the GC in response to exogenous antigen and generate autoantibodies. We crossed the mHEL strain to the Rag2 −/− background (ensuring all B cells were of BM donor origin, allowing precise quantification of rare HELspecific GC B cells and serum HEL-specific antibodies), irradiated both Rag2 −/− and Rag2 −/− mHEL mice sub-lethally, and reconstituted them with SW HEL -FcγRIIb-KO, -WT or -BTG BM cells (Fig. 6b). As expected, enhanced pre-immune tolerance resulted in fewer HEL-specific B cells in mHEL recipient mice reconstituted with SW HEL -FcγRIIb-KO compared with SW HEL -FcγRIIb-WT and SW HEL -FcγRIIb-BTG BM cells (Fig. 6c). However, SRBC immunisation drove an increase in HEL-specific GC B cells (Fig. 6d) and in serum HEL-specific IgG1 (Fig. 6e) Fig. 6a). We had shown previously that these mice displayed subtly reduced FcγRIIb expression on pre-B, immature and GC B cells, suggesting these natural Fcgr2b polymorphisms may impact upon both pre-immune and GC tolerance 44 . The results followed the same trend as those generated using SW HEL -FcγRIIb KO chimeras. There was enhanced tolerance at the pre-immune checkpoints (reduced frequency of HEL-specific autoreactive B cells in the BM and spleen of SW HEL -FcγRIIb wild/H1 KI/mHEL mice compared with control ( Supplementary Fig. 6b, c) Fig. 6d, e). A non-significant trend towards increased serum HEL-specific autoantibodies in mice reconstituted with SW HEL -FcγRIIb wild/H1 KI BM was also observed ( Supplementary Fig. 6f). In summary, a naturally occurring genetic variant reducing the FcγRIIb increase that follows B-cell activation can drive the expansion of autoreactive B cells in the GC, despite enhanced pre-immune tolerance and even in response to an irrelevant exogenous antigen.
FcγRIIB limits VH4-34 gene usage in humans. We then asked whether FcγRIIB had a similar impact on tolerance checkpoints in humans. We used the National Institute of Health Research (NIHR) Cambridge BioResource to recruit 29 healthy volunteers: 19 volunteers homozygous for isoleucine at position 232 in FCGR2B (hereafter referred to as I232 individuals), the common variant associated with normal FcγRIIB function, and 10 volunteers homozygous for the SLE-associated single-nucleotide polymorphism in FCGR2B (hereafter referred to T232 individuals) encoding a receptor that has markedly reduced inhibitory function 21,22 . We flow-sorted peripheral blood CD19 + IgD + CD27 − naive/transitional B cells; CD19 + IgD − CD27 − CD38 mid/hi activated B cells; CD19 + IgD + CD27 + B cells that include marginal zone B cells, and perhaps B1-like and unswitched memory cells [48][49][50][51] ; CD19 + IgD − CD27 + CD38 low/mid memory B cells; and CD19 + IgD − CD27 + CD38 hi plasmablasts ( Supplementary Fig. 7). The BCR repertoire for each of these subsets was analysed by high-throughput sequencing using a method that can distinguish between isotype classes (Supplementary Fig. 8 and 52 ), generating an average of 20,573 BCR sequences per B-cell subset (Supplementary Table 1). No significant differences in the proportion of B cells within each B-cell subset, nor of isotype usage, was seen between genotypes ( Supplementary Figs. 9 and 10). The effect of FcγRIIB on tolerance was evaluated by examining the frequency of cells expressing the IgHV4-34 gene. This specific heavy chain has been shown to bind red blood cell antigens 53,54 as well as commensal bacteria 55 , and was significantly enriched in SLE patients (Fig. 7a) in agreement with previous studies 56,57 . Moreover, IgHV4-34 was enriched among BCRs without sequence evidence of switching or SHM (comprising predominantly naive B cells), as well as antigen-experienced BCRs as evidenced by isotype switching or SHM, suggesting a defect in both pre-and post-immune tolerance (Fig. 7a).
In I232 healthy volunteers, the frequency of IGHV4-34 gene usage was higher in CD19 + IgD + CD27 − than in CD19 + IgD − CD27 − CD38 mid/hi and CD19 + CD27 + IgD − populations (Fig. 7b), consistent with the impact of peripheral tolerance on this autoreactive population. Healthy T232 and I232 individuals had similar IGHV4-34 proportions in CD19 + IgD + CD27 − cells, suggesting the enhanced central tolerance seen in FcγRIIB-deficient mice may not occur in humans. IGHV4-34 was, however, enriched in CD19 + CD27 + IgD + and CD19 + IgD − CD27 − CD38 mid/hi populations to a level similar to that seen in SLE patients (Fig. 7b), and consistent with impaired post-immune tolerance associated with FcγRIIB dysfunction, though this difference was not seen in the CD19 + IgD − CD27 + CD38 low/mid population. These human data suggest that, even in a healthy heterogeneous population, reduced FcγRIIB function impacts upon post-immune, rather than central/ pre-immune tolerance, consistent with the differential impact of FcγRIIB on pre-and post-immune tolerance seen in the mouse, but with notable species-specific differences.

Discussion
Immune tolerance in B cells is governed by a number of checkpoints that share the common goal of preventing the emergence of autoreactive lymphocytes. Given this shared goal, it might seem likely that molecules that regulate tolerance would impact upon different tolerance checkpoints in a similar direction. One such molecule is FcγRIIb, an inhibitory receptor that antagonises BCR signalling. Its reduced expression or function enhances BCR signalling and also predisposes to autoimmunity in mouse and human. To investigate the role of FcγRIIb in the development of autoimmunity, we used mice either under-or overexpressing FcγRIIb on their B cells, crossed to the SW HEL mouse model to allow interrogation of specific tolerance checkpoints.
Three distinct mouse models were used in this study, as each alone had inherent limitations. The FcγRIIb-deficient mouse has a profound defect affecting not only B cells, but also other cells implicated in the immune response 35,37 . We therefore also used two additional models. In one FcγRIIb was over-expressed on all B cells by between two-and tenfold (but was expressed at normal levels on other cell types 40 ). The other is a more physiological model of FcγRIIb under-expression, in which a knock-in of a naturally occurring promoter variant results in a failure of activation-induced upregulation of FcγRIIb specifically on B cells 44 . This genetic model would be expected to give a less extreme phenotype than the deficient mice, as its defect is subtle, but would confirm the B-cell specificity of any effect. Taken together comparisons between these three models increase confidence in our findings, as they provide confirmation in two under-expression models-one an extreme knockout, and one a subtle B-cell specific physiological genetically modified mouseand a cell-specific overexpression model.

(top right three panels) and the frequency of HEL-specific cells was measured (bottom panels). c, d Quantification of the frequency and number of HEL-specific GC B cells in WT (left panels) and mHEL (right panels) recipient mice. e Quantification of the frequency of HEL-specific follicular B cells in WT (left panels) and mHEL (right panels) recipient mice. f Ratio of the frequency of HEL-specific GC B cells and total follicular B cells in WT (left panel) and mHEL recipient mice (right panel). g Quantification by ELISpot of
HEL-specific antibody-forming cell (AFC) numbers in the spleen of immunised WT (left panels) and mHEL (right panels) recipient mice. For all panels two experiments were pooled. For panels (c-f), wild-type recipients: n = 5-6 mice per group; mHEL recipients: n = 8-11 mice per group. For panel (g), wild-type recipients: n = 5 mice per group; mHEL recipients: n = 6-9 mice per group. The mean is represented and each dot corresponds to an individual mouse. In panel b, the FlowJo smoothing option was used to improve visibility due to low event rates. The p-values were determined with the Kruskal-Wallis (KW) non-parametric test when comparing the three groups (KW: the p-value is indicated in purple for the mHEL recipient. All KW tests were not significant for WT recipients with the exception of panel d, for which the KW p-value is indicated for the WT group as well) and with the Mann-Whitney non-parametric test when two conditions were compared (indicated by black stars). *p < 0.05; **p < 0.01; ***p < 0.001. Comparisons generating non-significant p-values were indicated with "ns". Source data are provided as a Source Data file In both major pre-immune tolerance checkpoints, at the BM immature B-cell stage and at the transitional B-cell stage in the spleen, no or low-FcγRIIb expression was associated with more stringent tolerance, with increased deletion and anergy of HELspecific autoreactive B cells. In contrast, such reduced FcγRIIb expression was associated with the increased production of autoreactive B cells in the GC, and in serum autoantibody, consistent with a failure of tolerance at the GC checkpoint.
FcγRIIB is expressed not only on B cells, but on other hematopoietic cells, such as macrophages and dendritic cells. Chimeras reconstituted with FcγRIIb KO BM will therefore contain non-B cells that are FcγRIIb deficient. The impact of FcγRIIb on tolerance,  Fig. 5 The impact of FcγRIIb on GC tolerance is also observed with a soluble autoantigen. a Wild type and sHEL recipients (both CD45.2) were irradiated and reconstituted with bone marrow from SW HEL -FcγRIIb KO-CD45.2, SW HEL -FcγRIIb WT-CD45.1 or SW HEL -FcγRIIb BTG-CD45.1 donor mice. Post reconstitution, chimera were immunised with SRBC-HEL intraperitoneally and analysed 8 days later. All populations were gated as in Figs. 1, 2 and 4

. b-e Quantification of the frequency of HEL-specific B cells in WT (left panels) and sHEL (right panels) recipient mice: BM immature B cells (b), splenic mature MZ B cells (c), splenic follicular B cells (d) and splenic GC B cells (e). f Ratio of HEL-specific GC B cells and total follicular B cells in WT (left panel) and sHEL recipient mice (right panel). g Quantification by ELISpot of HEL-specific antibody-forming cells (AFCs) in the spleen of immunised WT (left panels)
and sHEL (right panels) recipient mice. sHEL recipients: n = 4-6 mice per group; wt recipients: n = 2-3 mice per group. One experiment representative of at least two is shown. The mean is represented and each dot corresponds to an individual mouse. The p-values were determined with the Kruskal-Wallis (KW) non-parametric test when comparing the three groups (KW: the p-value is indicated in purple for the mHEL recipient. All KW tests were not significant for WT recipients with the exception of panel e (absolute number) for which KW p-value is indicated for the WT group as well) and with the Mann-Whitney non-parametric test when two conditions were compared (indicated by black stars). *p < 0.05; **p < 0.01. Comparisons generating nonsignificant p-values were indicated with "ns". Source data are provided as a Source Data file shown. RAG2KO recipients: n = 5-8 mice per group; RAG2KO-mHEL recipients: n = 5-10 mice per group. The mean is represented and each dot corresponds to an individual mouse. The p-values were determined with the Kruskal-Wallis (KW) non-parametric test when comparing the three groups (KW: the p-value is indicated in purple for the RAG2KO-mHEL recipient. All KW tests were not significant for RAG2KO recipients with the exception of panel c for which KW p-value is indicated for the RAG2KO group as well) and with the Mann-Whitney non-parametric test when two conditions were compared (indicated by black stars). *p < 0.05; **p < 0.01. Comparisons generating non-significant p-values were indicated with "ns". Source data are provided as a Source Data file however, is likely to be B cell intrinsic for several reasons. The first is that we observed a similar effect on tolerance in chimeras reconstituted with BM from SWHEL Fcgr2b wild/H1 KI mice, in which FcγRIIB expression is reduced only on B cells 44 . The second is that transgenic overexpression of FcγRIIb on B cells alone leads to the opposite phenotype, with very strong tolerance at the immature stage despite less stringent GC tolerance. Finally, we also performed our BM reconstitution experiments in Rag2 −/− recipient mice that were non-lethally irradiated. In consequence, cells other than B and T lymphocytes were mostly coming from the recipients that were WT for FcγRIIb expression.
The two pre-immune checkpoints were similarly influenced by FcγRIIb expression, suggesting a shared mechanism distinct from that in the GC. Co-ligation of FcγRIIb to the BCR by cognate antibody is known to inhibit BCR signalling. It is also clear, however, that in immature B cells FcγRIIb can inhibit the BCR independent of cross-linking by antibody 46 . As we observed detectable levels of anti-HEL IgG1 in relevant mice, it is plausible that they are involved in co-ligation of FcγRIIb to the HELspecific BCR. It is also possible that FcγRIIb may be having its impact through tonic inhibitory signals. Whether or not the impact of FcγRIIb on selection at the immature/transitional B-cell stage is dependent upon antibody ligation of FcγRIIb, therefore remains an open question.
Our results demonstrate that tolerance in the GC cannot be mediated solely by a deletion of autoreactive clones driven by BCR signal strength as occurs in the BM, as earlier studies might have implied 7,15-17 , as the impact of reduced FcγRIIb on BCR signalling should enhance such tolerance. By demonstrating FcγRIIb deficiency increased HEL-specific GC B cells in response to SRBC in mice that had never been exposed to HEL, we confirmed our earlier suspicion 44 that FcγRIIb controlled the expansion of bystander B cells in the GC. Such bystander cells, with no measurable affinity for the immunising antigen, have recently been implicated in contributing to ongoing GCs 58,59 , supporting our observation. It should be noted, however, that while low affinity cross-reactivity between the transgenic anti-HEL BCR and SRBC antigens cannot be measured, and would seem unlikely, the presence in the bystander population of cells with very low affinity for a given antigen cannot be excluded, and could contribute to subsequent measurable bystander responses. Nonetheless, no or low expression of FcγRIIb led to the presence of autoreactive cells into the GC that could differentiate into AFCs that produced isotype-switched autoantibody detectable in the serum in response to a distinct exogenous antigen. This could explain the appearance of autoantibodies sometimes noted after infection or inflammation in humans 60,61 , which is presumably impacted by genetic background. Most importantly, it points to a potential novel mechanism-governing tolerance and the development of autoimmunity in the GC-the control of bystander activation and expansion of autoreactive B cells in the GC could be a central mechanism for avoiding autoimmunity, and one in which inhibitory receptors such as FcγRIIb play an important role. A recent paper by Silver et al. has formally demonstrated, using a mouse transgenic system, that non-cognate B cells can enter the GC and diversify via somatic hypermutation even in a competitive setting, and can then generate new antigenic specificities and ongoing immune responses 58 . This mechanism was proposed to have perhaps arisen to fill holes in the B-cell repertoire, and thus improve defence against infection. Our data suggest it could also be responsible for generating autoreactive responses, and perhaps also for the GC-driven autoantigen spreading recently described by Carroll and colleagues 59 . Thus, the generation of autoimmunity by bystander activation of B cells is consistent with previous observations in the setting of both non-autoreactive and autoreactive immune responses. Further work will be required to determine the precise molecular mechanisms at play.
The results presented here might make the concepts of preimmune and post-immune tolerance a more logical way of subdividing tolerance than the more traditional central versus peripheral tolerance nomenclature. This constitutes an example of a regulatory receptor having opposing effects on pre-and postimmune B-cell tolerance. It could be envisaged that other molecules that regulate BCR signal strength may also have different impacts upon tolerance checkpoints, as, for example, was reported for a Ptpn22 risk mutant that differentially regulates autoreactive follicular and marginal zone B cells 62 . The net outcome of FcγRIIb deficiency in the B-cell compartment is autoimmunity, demonstrating that in this particular context post-immune tolerance is dominant over pre-immune tolerance.
This seems also to be the case in humans, as complementary findings were made in healthy volunteers bearing a loss-of- function variant in FCGR2B. The proportion of IGHV4-34 clones, known to be autoreactive, was significantly higher in the activated CD19 + IgD − CD27 − CD38 mid/hi and CD19 + IgD + CD27 + B-cell subsets in individuals homozygous for the FCGR2B T232 loss-offunction SLE-associated polymorphism. This is consistent with a relative impairment of post-immune tolerance in people with reduced FcγRIIB function, as was seen in mice. Mouse and human results were not fully concordant (there was, for example, no evidence in humans of increased pre-immune tolerance associated with reduced FcγRIIb function, at least as this is reflected in IGHV4-34 clone frequency). Nonetheless, human data support a predominant role for FcγRIIB in post-immune tolerance, as was seen in the various mouse models. This unexpected complexity of checkpoint control in B-cell tolerance points to a nuanced regulation of the balance between the risk of autoreactivity and the maintenance of repertoire breadth to optimise defence against infection. It may also have implications for the increasingly sophisticated attempts to manipulate the B-cell immune response following the success of B-cell depletion by rituximab in autoimmune disease. The possibility that a given inhibitory receptor may have opposing effects on the generation of autoreactivity at different checkpoints in different contexts may impact upon therapy targeted at inhibitory receptors, helping explain the mechanism of new therapeutic agents (e.g. anti-CD22 therapy epratuzumab, which generated unexpected dose-related complexity in clinical studies 63 ), but perhaps also creating opportunities for synergistic therapies targeting individual checkpoints.

Methods
Mice. The SW HEL mice were a kind gift of Robert Brink. They were crossed with the FcγRIIb-deficient mouse strain 35 , the FcγRIIb-BTG strain (itself crossed to the CD45.1 background) 40 and the FcγRIIb wild/H1 KI mouse strain 44 . FcγRIIb KO mice (C57BL/6 FcγRIIb −/− ) were provided by J. Ravetch and S. Bolland. This strain was originally generated on the 129/sv background and backcrossed for more than 15 generations on the C57BL/6 background. Despite this extensive backcrossing, we cannot exclude the persistence of 129/sv-derived polymorphisms in the Fcgr2b genetic region. Nevertheless, as we also used the FcγRIIb-BTG and the FcγRIIb wild/H1 KI strains that were both generated on pure C56BL/6 background, we are confident that the results observed are indeed due to the level of expression of FcγRIIb. Fcgr2b is on the same chromosome than the Ly5 locus, making it very difficult to obtain FcγRIIb KO mice bearing the congenic marker CD45.1. The mHEL (membrane-bound HEL transgenic) and ML5 (soluble HEL transgenic) mice have been described previously 2,5 . We crossed the mHEL strain with the Rag2 −/− (here called RAG2KO) strain to obtain the mHEL × RAG2KO strain, expressing one allele of the mHEL transgene. For all experiments, the number of animals included in each experiments was determined based on our previous expertise using mouse models with different expression level of FcγRIIb and by using the G*Power software for a priori power calculation.
To generate bone marrow chimeras, recipient mice were irradiated lethally (1000 Gy for WT, mHEL and mL5 recipients) or sub-lethally (500 Gy for RAG2KO and mHEL × RAG2KO recipients) and reconstituted by intravenous injection of fresh bone marrow cells. Reconstitution was assessed 6 weeks later and mice that were not reconstituted fully were excluded. Reconstituted mice were randomly numbered independently of their genotype or BM used for their reconstitution and analysed at least 8 weeks post irradiation. Mice were immunised with sheep red blood cells (SRBC) conjugated with HEL or SRBC alone by intra-peritoneal injection, as indicated 64 . All experiments were performed according to the regulations of the UK Home Office Scientific Procedures Act (1986) and permission was given by the UK Home Office to perform such procedures at our local animal facility.
Flow cytometry. Single cell suspensions of the spleen and bone marrow were prepared by mashing or flushing the organs, respectively 44 . For phosphoflow staining, splenocytes were stimulated at 37°C for 5 min with a F(ab')2 goat anti-mouse IgG (H + L) antibody (from Jackson immunosearch, 10 μg/ml). Cells were then fixed with ice-cold methanol on ice for 30 min before staining with the relevant antibodies. HELspecific B cells were detected by staining with either recombinant HEL (40 ng/ml, Sigma-Aldrich) followed by a rabbit anti-HEL biotinylated (AbCam) and streptavidin eFluor450 (eBioscience) or 1 μg/ml of biotinylated HEL (conjugation with the EZ-link Sulfo-NHS-biotin kit from Pierce following the manufacturer's instructions) and streptavidin eFluor450 (eBioscience). For all experiments, the gating of HEL-binding cells was set up based on control staining with a biotinylated rabbit anti-HEL and streptavidin eFluor450, or streptavidin eFluor450 alone when biotinylated HEL was used. For the analysis, the HEL+ gates were drawn on the condition minus HEL and applied to all other samples.
FACS analysis was performed on a BD LSR Fortessa (BD biosciences) and data were analysed with the Flowjo software (TreeStar, Ashland, OR). All antibodies used are described in Supplementary Table 2. ELISA and ELISPOT. Serum HEL-specific IgG1 were detected by ELISA using a modified version of the protocol previously described 44 . Briefly, plates were coated with 5 μg/ml of HEL in PBS overnight at 4°C. Sera were added to the saturated plates and were incubated for 2 h at 37°C. Pooled sera from hyperimmune mice (immunised with SRBC-HEL twice) were used as a standard. Plates were developed with the BD Opteia kit (BD Biosciences). Serum dsDNA-specific IgG1 was detected by ELISA as previously described 65 .
For detection of HEL-specific antibody-forming cells (AFCs) ELISPOT plates (Millipore) were coated with 5 μg/ml of HEL in PBS overnight at 4°C. Single-cell suspensions of the spleen were added to saturated ELISPOT plates in quadruplicate and incubated overnight at 37°C in 5% CO 2 in a humidified incubator with culture medium. AFCs were detected with goat anti-mouse IgG1 antibody conjugated to horseradish peroxidase (Southern Biotech). Plates were developed using 3-amino-9-ethylcarbazole tablets (Sigma-Aldrich). Plates were read using an AID ELISpot reader according to the manufacturer's instructions.
In vitro proliferation and differentiation. Splenocytes were labelled with 50 nM CFSE for 10 min before being washed twice in cold culture medium (RPMI, 10% FCS, penicillin/streptomycin, sodium pyruvate). Cells were then plated at 2 × 10 6 / ml in round bottom 96-well plates and stimulated for 96 h with 10 μg/ml of E. coli lipopolysaccharide (LPS) (Sigma-Aldrich). B-cell proliferation was measured by quantifying CFSE dilution by flow cytometry. B-cell differentiation was measured by quantifying the percentage of CD138 hi B220 lo plasmablasts by flow cytometry.
Statistical analysis of the mouse data. When three groups were compared, the pvalues were determined with the Prism software using the Kruskal-Wallis nonparametric test (significant p-values are indicated on the figures). When two groups were compared, the p-values were determined with the Prism software using the Mann-Whitney unpaired two-tailed non-parametric test (significant p-values are indicated on the figures by stars; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). Non-significant p-values are indicated by "n.s".
Cell sorting for the human repertoire analysis. Peripheral blood mononuclear cells (PBMCs) were obtained from healthy individuals homozygous for the FcγRIIB-T232 or FcγRIIB-I232 site, under appropriate ethics approval from the NIHR Cambridge Bioresource. Inclusion criteria for individuals were people aged between 44 and 77 years, with no serious co-morbidities, no direct family history of autoimmune disease, no use of immunosuppressants or steroids and no hospitalisation within the last 12 months. Individuals were age and sex matched (18 females and 11 males matched between genotypes). Flow sorting was performed using CD19-BV785, CD38-BV711, CD3-NC650, CD14-605NC, CD24-PerCP-Cy5.5, IgD-FITC, CD27-PE-Cy7 (all from BD Bioscience) and Aqua (for live-dead cell detection, Invitrogen), where flow protocol is outlined in Supplementary Fig. 7.
Inclusion criteria for the SLE cohort. Inclusion criteria for healthy individuals were people aged between 20 and 77 years, with no serious co-morbidities, no direct family history of autoimmune disease, no use of immunosuppressants or steroids, and no hospitalisation within the last 12 months.
The SLE cohort comprised patients attending or referred to the Addenbrooke's Hospital specialist vasculitis unit between July 2004 and June 2016 who met at least four American College of Rheumatology SLE criteria 66 , presenting with active disease (defined as a) new BILAG score of A or B in any system (including switch from C to A or D to B), b) physician assessment of active disease, and c) a requirement for an increase of immunosuppressive treatment (excluding isolated increase/commencement of prednisolone at a dose ≤25 mg). After treatment with an immunosuppressant, patients were followed up monthly. Disease monitoring was undertaken with serial British Isles Lupus Assessment Group disease scoring (BILAG, a major international SLE activity scoring system) 67 and serum antinuclear antibody status.
IgH sequencing and analysis. MiSeq libraries were prepared using Illumina protocols and sequenced using 300 bp paired-ended MiSeq (Illumina). Raw MiSeq reads were filtered for base quality (median Phred score > 32) using QUASR (http://sourceforge.net/projects/quasr/) 68 . MiSeq forward and reverse reads were merged together if they contained identical overlapping regions of >50 bp, or otherwise discarded. Universal barcoded regions were identified in reads and orientated to read from V-primer to constant region primer. The barcoded region within each primer was identified and checked for conserved bases (i.e. the T's in NNNNTNNNNTNNNNT). Primers and constant regions were trimmed from each sequence, and sequences were retained only if there was >80% sequence certainty between all sequences obtained with the same barcode, otherwise discarded. The constant region allele with highest sequence similarity was identified by 10-mer matching to the reference constant region genes from the IMGT database 69 , and sequences were trimmed to give only the region of the sequence corresponding to the variable (V-D-J) regions. Isotype usage information for each IgH was retained throughout the analysis hereafter. Sequences without complete reading frames and non-immunoglobulin sequences were removed and only reads with significant similarity to reference IgHV and J genes from the IMGT database were retained using BLAST 70 . IgH gene usages and sequence annotation was performed in IMGT V-QUEST, where repertoire differences were performed by custom scripts in python, and statistics were performed in R using Wilcoxon tests for significance (non-parametric test of differences between distributions).
Reporting Summary. Further information on experimental design is available in the Nature Research Reporting Summary linked to this article.

Data availability
A reporting summary for this article is available as a Supplementary