Characterisation of anti-insulin B cells in VH125.hCD20/NOD mice
VH125/NOD transgenic mice expressing an increased frequency of anti-insulin B cells  were crossed with hCD20/NOD mice  to generate VH125.hCD20/NOD mice, facilitating the depletion of B cells using the 2H7 mAb (depletes B cells expressing human CD20). Expression of single or double transgenes did not affect T and B cell development (ESM Fig. 1a,b). However, we noted an increase in splenic marginal zone B cells when the VH125 transgene was present, as previously reported  (ESM Fig. 1c).
We examined anti-insulin B cells using FITC-conjugated insulin in four transgenic mouse strains (hCD20/NOD, VH125 and VH125.hCD20/NOD mice and 125Tg mice [transgenic for both heavy (VH) and light (VL) chains of the 125 insulin-specific BCR] as a positive control ). We demonstrated an increased percentage of anti-insulin B cells in the spleen of both VH125 and VH125.hCD20/NOD mice compared with hCD20.NOD mice (Fig. 1a,b). Anti-insulin B cells populating all splenic compartments in the VH125.hCD20/NOD mice (Fig. 1c–e), were comparable with anti-insulin B cells that escape central tolerance in the spleen in VH125 mice . Anti-insulin B cells were proportionally enriched in the T2 population (CD21hiCD23hi) (Fig. 1d), which plays a role in antigen-specific positive selection as these cells proliferate upon BCR engagement . The absolute number of anti-insulin B cells was significantly higher in both follicular and marginal zones (Fig. 1e), sites where BCR signalling is fundamental for B cell maturity .
Anti-insulin B cells can successfully present antigen to insulin-specific CD8+ T cells
To assess functionality, we investigated whether anti-insulin B cells could present insulin to and activate insulin-specific CD8+ T cells from the monoclonal G9Cα−/− CD8+TCR transgenic mouse. To obtain sufficient anti-insulin B cells, we used a two-step enrichment approach of insulin–FITC detection and anti-FITC microbeads (Fig. 1f). We obtained approximately 20 times more anti-insulin B cells (Fig. 1f) compared with the multi-step biotinylated antibody technique used by Smith et al  whereby enrichment was approximately 8%. Insulin-positive and insulin-negative fractions were then cultured with G9Cα−/− CD8+ T cells for 24 h. Significant CD69 upregulation (p = 0.026) and increased IFN-γ (p = 0.114, not statistically significant) production from CD8+ T cells was observed, when cultured with insulin-positive B cells compared with insulin-negative B cells (Fig. 1g,h). Insulin-positive B cells cultured with G9Cα−/− CD8+ T cells expressed increased CD86 (geometric mean fluorescent intensity [GMFI], p = 0.0571) when compared with insulin-negative B cells (Fig. 1i). We also observed a non-significant increase in the GMFI of CD80 and MHC I (Fig. 1i). Thus, anti-insulin B cells can successfully present to and activate IS-CD8+ T cells, corroborating previous studies demonstrating that anti-insulin B cells, despite being tolerant to antigen, are effective antigen-presenting cells (APCs) .
Selective recruitment of anti-insulin B cells to pancreatic islets
We studied anti-insulin B cells in PLNs and pancreatic islets from VH125.hCD20/NOD mice of different ages (Fig. 2). To demonstrate specific binding to insulin, we performed a competitive binding assay . B cells from spleen, PLNs and pancreatic islets were stained with insulin–FITC in the presence of human insulin (Fig. 2a). Successful inhibition of insulin binding on B cells was found in all tissues (ESM Fig. 2b); there was less inhibition in the PLNs (50%), in keeping with insulin-binding B cells in the NOD PLNs having different binding specificities . We ensured that insulin BCR staining was not due to increased insulin receptors on B cells (ESM Fig. 2c), as anti-insulin receptor staining (CD220) did not compete with insulin–FITC detection.
Next, we examined mice with a fully mature B cell repertoire developing early insulitis (aged 6–8 weeks old) and long-established insulitis (18–20 weeks old). We demonstrated a significantly increased frequency of anti-insulin B cells in pancreatic islets (p < 0.05) and a non-significant increase in the PLNs (p = 0.109); this appeared to coincide with disease progression (Fig. 2b). We compared the number of anti-insulin B cells in the PLNs and islets with numbers of B cells that were not specific for insulin and showed selective recruitment of anti-insulin B cells to the islets (spleen vs islets, p < 0.05), whereas the absolute number of anti-insulin B cells was unchanged in the spleen (Fig. 2c). These data suggested that anti-insulin B cells are recruited to the target tissue during beta cell destruction. This was also shown recently  and supports the notion that anti-insulin B cells contribute to type 1 diabetes .
Established pancreatic islet B cells are heterogeneous early-differentiated plasma-cell populations
To investigate total islet-infiltrating B cell population in our transgenic model, we assessed islet B cells from VH125.hCD20/NOD and NOD mice aged 12–20 weeks (Fig. 3). B cells from VH125.hCD20/NOD mice were IgM+IgD−, whereas in wild-type NOD islets, B cells were IgMlowIgD+ cells (Fig. 3a). During established insulitis, B cells become CD5 negative [12, 14] and upregulate CD138 . We observed few CD11b+ or CD5+ cells in both NOD and VH125.hCD20/NOD mice, confirming that few B1a B cells were present during insulitis (Fig. 3a).
We defined four subpopulations based on CD138 and IgD/IgM (for gating controls see ESM Fig. 3) expression (IgM on VH125.hCD20/NOD B cells, as the VH125 transgene is IgM and VH125.hCD20/NOD mice do not express IgD). The percentage of splenic B cells (Fig. 3b) expressing CD138 was less than in islet B cells (Fig. 3c) in both NOD and VH125.hCD20/NOD mice. We detected more CD138+ B cells in NOD mouse islets compared with VH125.hCD20/NOD mouse islets, although the difference was not statistically significant (p = 0.09) (Fig. 3c,d). We observed few plasma cells, defined by high expression of CD138 and loss of IgD/M (Fig. 3c, red gate). CD138hiIgD/Mlo cells expressed Blimp-1, a key transcription factor for plasma-cell differentiation , displayed increased CD43 expression  and expressed MHC II molecules (Fig. 3c, red histograms). Although proportions were small, significantly fewer CD138hiIgD/Mlo B cells were found in VH125.hCD20/NOD islets, compared with NOD islets (Fig. 3d, p < 0.05). CD138intIgD/M+ cells had a similar profile to the CD138− population (Fig. 3c, orange gate vs blue gate).
Single-cell analysis, using imaging flow cytometry, was performed in NOD mice to confirm our observations (Fig. 3e–i). The CD138int cells (Fig. 3f, orange/grey gate) were a heterogeneous population that still expressed B220+ but had a reduced (or had lost) CD19 and IgD expression (Fig. 3g). We demonstrated that CD138hiIgDlo cells (Fig. 3f, red gate,) expressed Blimp-1, although even in this small population, heterogeneity was visible, as not all cells had lost B220 (Fig. 3g). CD138int and CD138hi populations displayed significantly increased cell size (measured by area), compared with CD138− populations (Fig. 3i). We showed a larger expanse of cytoplasm synonymous with plasma cells, by bright-field imaging. These findings indicated that some B cells, upon entering pancreatic islets, progress into the plasma-cell pathway, although most exist as heterogeneous early-differentiated plasma-cell populations.
Anti-insulin B cells in pancreatic tissue are enriched in the CD138int subset
We next questioned whether islet anti-insulin B cells were present in the compartments shown in Fig. 3. Islet anti-insulin B cells may already have endogenous insulin bound to their BCRs. To ensure that we were detecting all anti-insulin B cells, islets from groups of mice were pooled and incubated with insulin–FITC at 37°C. We observed an increased frequency of anti-insulin B cells compared with islets stained at 4°C (Fig. 4a). Two insulin-specific populations were identified: insulin+CD19+ and insulin+CD19− B cells (Fig. 4a). We detected increased proportions of insulin+CD19+ B cells in VH125.hCD20/NOD mice compared with NOD mice, with variability in both strains (Fig. 4b,c). We also demonstrated increased insulin+CD19+ cells in pancreatic islets, compared with spleen, in both strains (Fig. 4b,c), supporting earlier observations (Fig. 2). Competitive binding experiments confirmed that both insulin+CD19+ and insulin+CD19− populations were insulin specific (ESM Fig. 2a,b). To study expression of CD138 and IgD/IgM, we used the gating strategy shown in Fig. 3. Insulin+CD19+ and insulin+CD19− cells expressed intermediate CD138 in VH125.hCD20/NOD and NOD strains (Fig. 4d–f). We found a higher proportion of plasma cells (CD138hiIgM/Dlo) among the insulin+CD19− population. Differences between the mice were highlighted by a higher proportion of the insulin+CD19+ cells in the CD138− fraction (Fig. 4d, blue gate, p < 0.001) and a lower proportion in the CD138int fraction (Fig. 4d, orange gate, p < 0.01) in the VH125.hCD20/NOD mice. The converse relationship was found in NOD mice (Fig. 4d; blue gate, p < 0.05; orange gate, p < 0.001).
To evaluate cell morphology, we used single-cell imaging. Spleens and pooled islets from VH125.hCD20/NOD mice (ESM Fig. 4) were analysed alongside pooled NOD mouse islets. Gating on live CD3−CD11c−CD11b− revealed anti-insulin B cell populations divided by their CD19 expression (Fig. 4g–j). We confirmed that anti-insulin B cells were enriched in the CD138 subset. Insulin+CD19− cells also displayed loss of IgD and little expression of Blimp-1 (Fig. 4k), supporting the observation that very few cells in the CD138hiIgDlo population bound insulin. We demonstrated that CD138 staining intensity was significantly increased in the insulin+CD19− B cell population, compared with the CD19+ population (Fig. 4l). However, some insulin+CD19+ B cells displayed increased CD138 expression, when compared with the intensity of the CD138− B cell subset (Fig. 4m, dotted line), but insulin+CD19− B cells displayed significantly greater expression (Fig. 4l). Finally, cell size analysis revealed that both of the insulin-positive populations were larger compared with CD19+ insulin-negative B cells, indicating that these cells were activated and blasting (Fig. 4n).
Anti-insulin B cells are recruited to pancreatic islets after anti-CD20 treatment
B cell depletion therapy has been successful in delaying the onset of diabetes [2, 3, 30]. However, the effect of global B cell depletion on anti-insulin B cells has not been studied. We confirmed that expression of VH125 had no effect on hCD20 expression (ESM Fig. 5a) and that hCD20-expressing B cells were present in the tissues examined (ESM Fig. 5b). We treated groups of 6- to 8-week-old VH125.hCD20/NOD mice with anti-CD20 mAb, and analysed spleen and PLN tissue for anti-insulin B cells after treatment. Expression of hCD20 was similar in anti-insulin and non-insulin-binding B cells, with successful targeting of both populations in spleen (ESM Fig. 5c) and PLNs at 24 h after treatment (ESM Fig. 5d). In agreement with previous studies [31, 32], expression of murine CD20 was parallel with the expression of human CD20 (ESM Fig. 5e). These results suggest that autoreactive B cells in the periphery are successfully depleted and not spared by anti-CD20 treatment.
As anti-insulin B cells are altered upon entry into islets, we investigated whether these cells would be targeted. We confirmed that IgM+ B cells were targeted by treatment (ESM Fig. 6a), which we determined was a result of hCD20 expression (ESM Fig. 6b). CD19+insulin+ B cells were successfully depleted, although cell numbers were low, making it difficult to determine statistical significance (ESM Fig. 6c,e). However, we observed hCD20 expression on CD19+insulin+ B cells (ESM Fig. 6d), indicating that anti-insulin B cells are targeted in pancreatic islets. Again, we found that hCD20 expression was parallel with murine CD20 on islet B cells (ESM Fig. 6b,d). Conversely, hCD20 was not expressed on insulin+CD19− B cells, indicating that these cells would be spared by anti-CD20 treatment (ESM Fig. 6d). During the 12 week observation period, 48% of the control anti-CD20 antibody-treated mice became diabetic compared with 15% of the 2H7-treated mice.
To investigate the repopulation of anti-insulin B cells, we examined CD19+insulin+ and CD19+insulin− cells at 8 and 12 weeks after depletion (Fig. 5). In hCD20/NOD mice, B cells repopulate peripheral tissues by 12 weeks post treatment [2, 19]. We found repopulation dynamics were similar for anti-insulin B cells in spleen (Fig. 5a–g) and PLNs in VH125.hCD20/NOD mice (Fig. 5b–h). However, we observed that anti-insulin B cells repopulated earlier than non-insulin B cells in pancreatic islets (Fig. 5i).
To analyse anti-insulin B cells further, we pooled pancreases to gain more cells for analysis (Fig. 6). Considering that anti-insulin B cells are enriched in the CD138int fractions (Fig. 4), we analysed IgM and CD138 subpopulations after 12 weeks of treatment. Islet B cells expressing IgM were reduced in 2H7-treated mice (Fig. 6a,b), although interestingly an increase in the CD138intIgMlo fraction was observed (Fig. 6b). In line with this, we observed an increase in the percentage of insulin+CD19− cells (Fig. 6c,d), which were significantly enriched in the CD138intIgMlo fraction (Fig. 6e–g) after anti-CD20 treatment.