Increased spontaneous autoantibody IgG3 responses in Fcgr2b-cKO mice. We first characterized the spontaneous autoantibody production in mice with a B cell–specific FcγRIIB conditional KO (Fcgr2b-cKO). As reported previously, we observed increased anti-dsDNA IgG, which was present in mice at around 4–5 months of age (Figure 1A and Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/JCI157250DS1). Interestingly, when analyzing the IgG subclasses, we observed a significant increase only in IgG3 anti-dsDNA (Figure 1B). Using a flow cytometric assay we developed to assess anti–nuclear antibody–positive (ANA+) PCs (31), we observed an increased frequency of ANA+IgG+ PCs in the spleens of Fcgr2b-cKO mice (Figure 1, C and D). In line with the serum data, the increase in ANA+IgG+ PCs was only significant in the IgG3 subclass (Figure 1E). The frequency of ANA+ PCs within other isotypes or in the BM was not increased (Supplemental Figure 1, B–F). The increase in ANA+IgG3+ PCs in the spleen suggested an extrafollicular B cell response.

Increased spontaneous autoantibody IgG3 responses in Fcgr2b-cKO mice. Female control (Ctr) and Fcgr2b-cKO mice were bred and maintained until experiments at 10–12 months of age, at which point (auto)antibodies in serum and PCs in spleen were characterized. ANA reactivity of PCs was established using flow cytometry. (A and B) dsDNA ELISA for total IgG and IgG subclasses in serum from Fcgr2b-cKO mice. (C) Representative example of ANA staining in IgG and IgM PCs in spleen. (D and E) Frequency of ANA+ IgG+ PCs in spleen, total IgG (D), and by IgG subclass (E). (F) Representative example of IgM and IgG3 staining in total PCs. IgG3+ cells are indicated in green. (G) Frequency of IgG+ PCs in spleen separated by subclass. (H) Frequency of IgG3+ B cells in control and Fcgr2b-cKO mice. (I) Representative example of staining strategy for B-1 and B-2 cells in spleen. (J) Percentage of splenic IgG3+ B cells with a B-1 or B-2 phenotype, respectively, gated as in I. (K) Representative example of staining for B220 and CD5 in total IgG3+ B cells compared with B-1, FO, and MZ B cells in spleen. Data are shown as the median, with each symbol representing an individual mouse (A, B, D, E, G, H, and J) (n = 12–17 per group pooled from 2–3 independent experiments). *P < 0.05 and ***P < 0.001, by Mann-Whitney U test.

Since we previously showed that increases in ANA+ IgG PCs in patients with SLE and lupus-prone mice occur through aberrant IgG PC differentiation rather than as a result of an antigen-specific tolerance defect (31), we also analyzed tolerance checkpoints for ANA+ B cells and PCs in Fcgr2b-cKO mice. B cell–intrinsic FcγRIIB deficiency did not affect the percentage of ANA+ mature naive B cells or more immature B cell subsets in the BM or spleen (Supplemental Figure 2, A–C and Supplemental Figure 3). In contrast, we detected a specific increase in the frequency of splenic IgG+ PCs and serum levels of total IgG (Supplemental Figure 1, G–I, M, and N). Again, the most prominent increase was observed in the IgG3 subclass, both in PCs and in serum titers (Figure 1, F and G, and Supplemental Figure 1, J–L and O–R).

Together, these results indicate that spontaneous (auto)antibody production occurred through enhanced differentiation or survival of IgG3 PCs.

Fcgr2b-cKO mice spontaneously displayed an increased frequency of resting IgG3+ B cells (Figure 1H and Supplemental Figure 4, A–E). IgG3 is usually derived from B-1 cells or MZ B cells, and increased numbers of peritoneal B-1 cells have been reported in complete Fcgr2b–/– mice (24). Most IgG3+ B cells in Fcgr2b-cKO mice had a B-2 (CD19+B220hi) phenotype and were CD5– (Figure 1, I–K). In addition, the frequencies of B-1a or B-1b cells in the spleen and peritoneum were unaffected in Fcgr2b-cKO mice (Supplemental Figure 4, F–M), whereas MZ B cell frequencies were increased in Fcgr2b-cKO mice (Supplemental Figure 4, N and O).

Increased extrafollicular PC responses. IgG3 has been associated primarily with extrafollicular PC responses derived from MZ or B-1 cells (32–34). B-1 cells produce natural antibodies and can even produce these in the absence of antigen stimulation; and MZ B cells can be directly activated by antigens with repeated epitopes and TLRs and act as a first line of defense against blood-borne pathogens (24, 35, 36). Both of these cell types have been associated with extrafollicular humoral responses that are independent of cognate T cell help. Therefore, we analyzed the extrafollicular response to immunization with prototypical T-independent and T-dependent antigens, (4-hydroxy-3-nitrophenyl)acetyl–Ficoll (NP-Ficoll), and NP-CGG in alum, respectively. We observed a large increase in NP-IgG, but not NP-IgM, serum titers in Fcgr2b-cKO mice on day 7 following immunization with NP-Ficoll (Figure 2A). The anti-NP response of all IgG subclasses was significantly increased, with a marked increase in IgG3 (Figure 2B). In line with serum antibody levels, NP-specific IgG+ PCs in the spleen were increased, whereas no increase in IgM PCs was observed (Figure 2, C–E). The greatest increase in NP-specific PCs was present in the IgG3 subclass (Figure 2, F and G). Besides an increase in NP+IgG3+ PCs, there was an increased frequency of NP+IgG3+ B cells in the spleens of Fcgr2b-cKO mice (Figure 2, H and I).

Increased extrafollicular responses following immunization. (A–F) Female control and Fcgr2b-cKO mice were immunized with NP-Ficoll. Serum and splenocytes were obtained 7 days later. (A and B) Levels of NP-specific antibodies, separated by isotype and subclass. (C) Representative example of PC staining following NP-Ficoll immunization (concatenated data on 4000 cells from 4 different mice per group). (D and E) Frequency of NP-specific PCs in spleen by isotype, as a percentage of B cells. (F) Representative examples of intracellular IgG3 and NP staining in splenic PCs. (G) Frequency of NP-specific PCs in spleen as a percentage of B cells, separated by IgG subclass. (H) Representative example of surface NP gating on IgD– B cells (left) and IgG1 and IgG3 staining in IgD–NP+ B cells (middle and right; ~10,000 cells were concatenated from 4 mice per group). (I) Frequency of IgG3+NP+ B cells among total B cells. Data are shown as the median, with each symbol representing an individual mouse (n = 8–9 mice per group; data were pooled from 2–3 independent experiments). **P < 0.01 and ***P < 0.00, by Mann-Whitney U test.

Whereas total levels of NP-IgG and NP-IgM were not increased in Fcgr2b-cKO mice following a T-dependent response to NP-CGG (Supplemental Figure 5A), we observed an increase in NP-IgG3 serum levels and PCs (Supplemental Figure 5, B and C). Likewise, NP+IgG3+ B cells were increased (Supplemental Figure 5, D and E). In line with an extrafollicular origin of NP-specific IgG3 following NP-CGG immunization, low-affinity NP-25 IgG3 levels peaked early (around day 14), and high-affinity NP-2 IgG3 levels were detected only at low levels at all time points up to day 42 (Supplemental Figure 5, F and G).

Together, these data suggest that B cell–specific FcγRIIB deficiency led to greatly enhanced extrafollicular responses to immunization, with a particularly strong increase in IgG3 responses that may have been of MZ origin.

Increased activation of MZ B cells in the absence of FcγRIIB. Since MZ or B-1 cells could be the origin of increased extrafollicular responses in Fcgr2b-cKO mice, we analyzed the expression of FcγRIIB in several B cell subsets in mice (Figure 3, A and B). We found that the expression of FcγRIIB was highest in MZ B cells and IgG3+ B cells compared with that in B-1 cells and FO B cells. Following immunization with NP-Ficoll and NP-chicken gamma globulin (NP-CGG), the we observed the highest expression of FcγRIIB in NP+IgG3+ B cells compared with expression in naive B cells or IgG1+NP+ B cells (Supplemental Figure 6, A–C). These results suggest that MZ B cells and IgG3+ B cells may be most susceptible to inhibition by FcγRIIB; stated otherwise: MZ B cells may be most activated in Fcgr2b-cKO mice.

Phenotype and activation of MZ B cells in Fcgr2b-cKO mice and humans. (A) Representative example of staining for FcγRIIB in different B cell subsets and IgG3+ B cells in the spleen. Panel on the right shows staining in Fcgr2b-cKO cells. (B) FcγRIIB staining intensity in different B cell subsets from control mice. (C) Schematic of the experimental approach used to investigate the effect of FcγRIIB on B cell activation. Using intact anti-IgM, both the BCR and FcγRIIB are engaged (top); using Fab′2 anti-IgM (middle) or Fcgr2b-cKO cells (bottom), only the BCR is engaged. (D–G) Sorted FO and MZ B cells were stimulated with 3 μg/mL anti-IgM for 20 hours. Activation was measured by flow cytometry. Representative examples (D) and summary (E–G) of the expression of CD80, CD86, and IAd. (H and I) Representative examples and summary of expression of CD32 analyzed by flow cytometry using different human B cell subsets. (J) qPCR for FCGR2B in sorted human B cell subsets. Relative expression was normalized to polr2a, after which ΔΔCt was calculated compared with naive B cells. (K–M) Sorted human naive and MZ-like B cells were stimulated with 3 μg/mL intact anti-IgM or equimolar concentrations (2 μg/mL) of Fab′2 anti-IgM for 20 hours or were left untreated as controls. Upregulation of HLA-DR, CD80, and CD86 was measured by flow cytometry. The percentage of inhibition was calculated for each donor (the median percentage of inhibition is indicated). Data are shown as the median, with each symbol representing an individual mouse or human (n = 6 per group for B; n = 5–7 per group for D–F; n = 6–10 per group for I–M; data were pooled from 2–5 independent experiments; except for the data in B, which were from 1 experiment). *P < 0.05, **P < 0.01, and ***P < 0.001, by 1-way ANOVA with Bonferroni’s post hoc test (B, I, and J) or 2-way ANOVA with Bonferroni’s post hoc test (E–G and K–M). mem, memory B cells.

Since we observed increased expression of FcγRIIB in MZ B cells, we asked whether there was a differential inhibitory effect of FcγRIIB on FO B cells and MZ B cells. Sorted B cell subsets were activated through BCR crosslinking in the presence (intact IgM crosslinking antibodies) or absence of FcγRIIB engagement (Fcgr2b-cKO cells or Fab′2 IgM crosslinking antibodies) (Figure 3C). We first determined optimal concentrations by Fab′2 IgM crosslinking antibodies for upregulation of CD80, CD86, and MHC class II (IAd) in FO B cells (Supplemental Figure 6, D–F) and observed increased upregulation of CD80 and CD86 in MZ B cells compared with FO B cells (Supplemental Figure 6, G–I). We then analyzed the effect of FcγRIIB engagement on BCR-mediated activation of FO and MZ B cells. We found that FcγRIIB deficiency increased the expression of CD80, CD86, and MHC class II in MZ B cells, whereas only CD86 was significantly upregulated in FO B cells in the absence of FcγRIIB (Figure 3, D–G). The absence of a strong effect on FO B cells was surprising, so we confirmed the observation by comparing equimolar concentrations of intact versus Fab′2 anti-IgM antibodies in control mice (Supplemental Figure 6, J–L).

To address these findings in humans, we analyzed the expression and inhibitory function of FcγRIIB in various B cell subsets from healthy donors. IgM+CD27+ B cells have been proposed as the human equivalent of MZ B cells in mice (MZ-like B cells) (37). We isolated these cells by FACS and confirmed their MZ-like phenotype by high SOX7 and low HOPX expression, relative to conventional naive and IgG+ memory B cells (37) (Supplemental Figure 7, A and B). Next, we analyzed the expression of FcγRIIB in MZ-like B cells, naive B cells, and IgG and IgA memory B cells. As in mice, the human MZ-like compartment had the highest expression of FcγRIIB on both the protein and RNA level (Figure 3, H–J). MZ-like B cells were also more strongly activated than were naive B cells by BCR crosslinking using Fab′2 anti-IgM across a range of concentrations (Supplemental Figure 7, C–E). We next analyzed the effect of FcγRIIB-BCR crosslinking in MZ-like B cells compared with naive B cells and observed a stronger inhibitory effect of FcγRIIB engagement on MZ-like B cells than on naive B cells (Figure 3, K–M, and Supplemental Figure 7I). For example, although CD86 upregulation by BCR engagement was similar between naive B cells and MZ B cells (Supplemental Figure 7H), the inhibitory effect of FcγRIIB on MZ-like B cells returned CD86 expression close to baseline levels (Figure 3M). Upregulation of HLA-DR and CD80 upon BCR crosslinking was also stronger in MZ-like B cells (Supplemental Figure 7, F and G), and FcγRIIB-mediated inhibition was most pronounced in this subset (Figure 3, K and L).

MZ-like B cells have been reported to be more prone to activation; we now show that they also exerted stronger inhibition of activation through FcγRIIB, an effect that was observed in both mice and humans.

FcγRIIB inhibits MZ B cell activation through inhibition of Erk phosphorylation and calcium flux. We next analyzed the phosphorylation of signaling molecules downstream of the BCR and FcγRIIB (Figure 4A). We chose to study Syk, SHIP1, and Erk1/2 signaling. Syk is an early B cell signaling molecule that is not inhibited by FcyRIIB or SHIP1, the phosphatase activated by FcyRIIB engagement. Erk1/2 is downstream of FcyRIIB/SHIP1 signaling and is important for B cell activation and PC differentiation (38). Consistent with the increased upregulation of activation markers, we found increased phosphorylation of Syk and Erk1/2 in murine MZ B cells compared with FO B cells stimulated with Fab′2 anti-IgM (Figure 4, C and D). There was strong inhibition of Erk1/2 phosphorylation in MZ B cells in the presence of Fcgr2b engagement with intact anti-IgM antibodies, but no inhibition of Syk phosphorylation (Figure 4, B–D). A significant decrease in Erk1/2 phosphorylation was only observed in MZ B cells and not in FO B cells (Figure 4D). Furthermore, MZ B cells exhibited increased SHIP1 phosphorylation following combined BCR-FcγRIIB crosslinking compared with FO B cells (Figure 4E). Similar results for Erk1/2 phosphorylation were obtained when we studied the absence of FcγRIIb engagement in Fcgr2b-cKO mice compared with control mice (Figure 4, F and G). Again, this difference was only observed in MZ B cells.

The effect of FcγRIIB on B cell signaling in MZ B cells. (A) Simplified schematic diagram of key signaling molecules downstream of the BCR and FcγRIIB. (B–D) Representative examples and summary of FO and MZ B cells from control mice; cells were stimulated with 3 μg/mL intact anti-IgM (αIgM) or equimolar concentrations (2 μg/mL) of Fab′2 anti-IgM for 10 minutes, followed by Phosphoflow analysis of the phosphorylation of signaling molecules. (E) SHIP1 phosphorylation was analyzed by capillary Western blotting. (F and G) Representative examples and summary of p-Erk expression in FO and MZ B cells from control and Fcgr2b-cKO mice; cells were stimulated with intact anti-IgM as described in B–D. (H–J) PBMCs from healthy donors were stimulated with 15 μg/mL intact anti-IgM or equimolar concentrations (10 μg/mL) of Fab′2 anti-IgM for 60 minutes, after which the phosphorylation of signaling molecules was analyzed by Phosphoflow. (H) Representative example of Syk and Erk phosphorylation in MZ-like B cells. (I and J) Comparison of intact versus Fab′2 anti-IgM in naive and MZ-like B cells. The median percentage of inhibition calculated per donor is indicated. (K–M) Calcium flux following 75 μg/mL intact anti-IgM or equimolar concentrations (50 μg/mL) of Fab′2 anti-IgM. Peak calcium flux and the AUC were calculated using FlowJo. (K) Representative examples of calcium flux in naive and MZ-like B cells. (L and M) Comparison of the peak and AUC of calcium flux following intact versus Fab′2 anti-IgM in naive and MZ-like B cells. inh, inhibition. Data are shown as the median, with each symbol representing an individual mouse (n = 4–7 mice per group for B–D; n = 8 mice per group for G; n = 6 mice per group for H–J; n = 5 mice per group for L and M; data for each were pooled from 2–3 independent experiments). *P < 0.05, **P < 0.01, and ***P < 0.001, by 2-way ANOVA with Bonferroni’s post hoc test.

In human MZ-like B cells, we likewise observed increased phosphorylation of Erk1/2 following BCR triggering in MZ-like cells compared with naive B cells (Supplemental Figure 8, A and B). FcγRIIB engagement strongly inhibited p-Erk in MZ-like B cells. Decreased Erk phosphorylation was also observed in naive B cells (Figure 4, H–J). MZ-like B cells compared with naive B cells had increased calcium flux upon BCR triggering (Supplemental Figure 9), and the calcium flux, particularly the peak flux, was more strongly inhibited by FcγRIIB in MZ-like B cells than in naive B cells (Figure 4, K–M).

These results point to a strong inhibitory effect of FcγRIIB on the activation of MZ B cells through decreased phosphorylation of Erk1/2 and decreased calcium flux.

IgG3 responses in Fcgr2b-cKO mice are derived from MZ B cells. Because our results suggest that MZ B cells are a likely source for the increased IgG3 response in Fcgr2b-cKO mice, we generated double-KO (dKO) mice that lacked Notch2 and FcγRIIB in B cells, resulting in large reductions in MZ B cell numbers (39) and B cell–specific FcγRIIB deficiency. These mice exhibited deficiency of MZ B cells (~80% reduction), whereas FO B cells and B-1 cell numbers were maintained (Figure 5, A and B, and Supplemental Figure 9, A–C). Similar to what has been previously described (32), Notch2-cKO mice had decreased NP-specific IgM and IgG serum levels and PC numbers, but there was no effect of FcγRIIB on the IgM response in either the presence or absence of Notch2 (Supplemental Figure 9, D and F). More important, the increased IgG response to NP-Ficoll that we observed in Fcgr2b-cKO was completely reversed in the dKO mice lacking both Notch2 and FcγRIIB in B cells (Supplemental Figure 9, E and G). Furthermore, the increase in NP-IgG3 serum titers and NP+IgG3+ B cells and PCs generated by FcγRIIB deficiency was absent in dKO mice (Figure 5, C–F). IgG3+ B-2 B cells, but not IgG3+ B-1 B cells, were significantly reduced in MZ-deficient dKO mice (Supplemental Figure 9, H and I). Finally, we analyzed spontaneous autoantibody production triggered by deficiency of FcγRIIB in MZ-deficient mice. The increased IgG and IgG3 anti-dsDNA observed in FcγRIIB-cKO mice was abolished in dKO mice lacking both FcγRIIB and MZ B cells (Figure 5, G–I). Together, these results show that MZ B cells were responsible for enhanced extrafollicular (auto)antibody responses in Fcgr2b-cKO mice.

Combined FcγRIIB and MZ deficiency reverses the enhanced response to antigen challenge and the increase in autoantibody production. (A–F) Female control, Notch2-cKO, Fcgr2b-cKO, and Notch2 Fcgr2b–dKO mice were immunized with NP-Ficoll. Serum and splenocytes were obtained 7 days later. (A and B) Representative examples of MZ B cell frequencies in Fcgr2b-cKO and Notch2 Fcgr2b–dKO mice. (C) Representative examples of intracellular IgG3 and NP staining in splenic PCs. (D) Frequency of NP-specific IgG3+ PCs in spleen, as a percentage of B cells. (E) Levels of NP-specific IgG3 in serum. (F) Frequency of NP-specific IgG3+ B cells in spleen. (G–I) Female control, Notch2-cKO, Fcgr2b-cKO, and Notch2 Fcgr2b–dKO mice were bred and maintained until 6–7 months of age, after which dsDNA antibodies in serum were characterized. dsDNA ELISAs for total IgM, IgG, and IgG subclasses were performed. Data are shown as the median, with each symbol representing an individual mouse (n = 5 mice per group for A–F; n = 13–17 mice per group for G–I). *P < 0.05, **P < 0.01, and ***P < 0.001, by 2-way ANOVA with Bonferroni’s post hoc test.

Diminished expression of FcγRIIB in MZ-like B cells from patients with SLE. Previous studies have shown diminished expression of FcγRIIB on CD27+ B cells from patients with SLE (9, 10). However, these cell populations were not studied in further detail, and CD27+ cells comprise both switched memory B cells as well as MZ-like B cells. We therefore wanted to assess the expression of FcγRIIB in MZ-like B cells from patients with SLE. We performed high-dimensional spectral flow cytometry, which allowed a detailed identification of more than 10 B cell subsets, in combination with staining for CD32B/C, on PBMCs from patients with SLE (n = 15) and healthy donors (n = 10). The patients’ characteristics are shown in Supplemental Table 2.

In line with previous studies, patients with SLE showed decreased expression of CD32B/C on CD27+ B cells, but not on CD27– B cells (Figure 6A). Next, B cell subsets were clustered using FlowSOM and visualized on a uniform manifold approximation and projection (UMAP) projection (Figure 6B). MZ B cells were identified in cluster 08, characterized as CD27+, IgMhi, IgDlo, CD21hi, CD24hi, and CD38lo B cells (Supplemental Figure 10). Likewise, other B cell subsets were identified according to published characteristics (40). Interestingly, we observed a significant decrease in CD32B/C expression in MZ-like B cells (Figure 6, C–E). Although some other cell populations (activated naive B cells, DN1 B cells, resting switched memory B cells, and plasmablasts) also exhibited some decrease in expression of CD32B/C, this difference was no longer significant after correction for multiple testing (Figure 6F). Similar results were obtained using conventional gating (data not shown).

Reduced FcγRIIB expression in MZ B cells from patients with SLE. High-dimensional spectral flow cytometry was used to identify multiple B cell subsets within PBMCs from patients with SLE (n = 15) and healthy donors (HD) (n = 10). (A) Expression of CD32B/C in CD27– and CD27+ B cells. (B) Live B cells were clustered using FlowSOM and are shown in a UMAP plot. Live B cells from healthy donors and patients with SLE were concatenated from each donor (n = 250,000 cells per group). (C) Expression of CD32B/C among all B cell clusters in the UMAP plot. (D) Representative example of CD32B/C expression in MZ-like B cells from healthy donors and patients with SLE, including the FMO control. (E) Summary of CD32B/C expression in MZ B cells. (F) Summary of CD32B/C expression in cells in all FlowSOM clusters. (G) Expression of CD32B/C in MZP cells identified by manual gating as live CD19+CD27–IgD+CD38loCD21+CD24hi. (H) Representative examples of CD80 expression and gating in MZ B cells from healthy donors and patients with SLE. MZ B cells from healthy donors and patients with SLE were concatenated from each donor (n = 25,000 cells per group). (I) Summary of the percentage of CD80+ MZ B cells from healthy donors and patients with SLE. (J) Correlation of the percentage of CD80+ cells and CD32B/C MFI in MZ B cells. Data are shown as the median, with each symbol representing an individual donor (n = 10 for healthy donors; n = 15 for patients with SLE). **P < 0.01 (A) and other P values were determined by Mann-Whitney U test (E–G and I) or Spearman’s rank test (J). Act. Nv, activated naive; Rest. Nv, resting naive; Atyp. UnswM, atypical unswitched memory; Act. SwM, activated switched memory; Rest. SwM, resting switched memory; PB, plasmablast.

Interestingly, decreased expression of CD32B/C was already observed in the MZ precursor (MZP) population, gated as CD27–CD38loIgD+CD24hiCD45RBhi, further characterized in a previous report as IgMhiCD21+CD1cint (37) (Figure 6G and Supplemental Figure 11).

MZ-like B cells also exhibited significantly increased expression of the activation marker CD80 in patients with SLE compared with healthy donors (Figure 6, H and I). The increased expression of CD80 was not seen in other B cell populations (Supplemental Figure 12). No difference in other activation markers (CD40, CD69, CD86, HLA-DR) was observed (data not shown). CD80 expression in MZ-like B cells was inversely correlated to levels of CD32B/C (Figure 6J), suggesting that the lower expression of FcγRIIB was associated with higher activation of MZ B cells in SLE, as we demonstrated in the murine studies

In summary, our results show increased extrafollicular responses in B cell–intrinsic FcγRIIB-deficient mice characterized by a major increase in MZ-derived IgG3 responses. MZ B cells are highly sensitive to inhibition through FcγRIIB, which can be explained by the high expression of FcγRIIB in MZ B cells compared with expression levels in other B cell subsets. High FcγRIIB expression resulted in strong inhibitory signaling through FcγRIIB in MZ B cells, an effect that we observed in both mice and humans. Finally, patients with SLE exhibited a marked decrease in FcγRIIB expression in B cells, most strongly in MZ-like B cells.