Interleukin-10⁺ Regulatory B Cells Arise Within Antigen-Experienced CD40⁺ B Cells to Maintain Tolerance to Islet Autoantigens

Although type 1 diabetes (T1D) has been classically described as a CD4⁺ T-cell–mediated disease, B cells play a crucial role in the autoimmune destruction of pancreatic islets (1–4). B cells can promote T1D by 1) presenting islet-derived peptides to autoreactive T cells (2,4–10), 2) producing autoantibodies (aAbs) against β-cell antigens (Ags) (7,11,12), and 3) secreting proinflammatory cytokines (2,3,13,14). B-cell depletion in nonobese diabetic (NOD) mice as well as neutralization of the B-cell–activating factor (BAFF) delay the onset of diabetes in prediabetic NOD mice and re-establishes normoglycemia in hyperglycemic NOD mice (3,15–18). Likewise, B-cell depletion in individuals newly diagnosed with T1D prevents decline of the islet mass (19). Importantly, preclinical and clinical studies suggest that re-emerging B cells exhibit an immunosuppressive phenotype that may directly contribute to the clinical efficacy of B-cell depletion therapies (3,15,20). Recent evidence supports the existence of regulatory B cells (Bregs), which, like regulatory T cells (Tregs), are able to negatively modulate T-cell–dependent autoimmune responses in an Ag-specific manner (14,21–28). To date, several populations of murine and human Bregs have been described based on the expression of various surface markers. Reported phenotypes for murine Bregs include CD19⁺CD1d^hiCD5⁺ (26,27,29), CD19⁺CD21^hiIgM^hiCD23^hi (30–32), and CD19⁺Tim-1⁺ (33). Human Breg populations have been identified based on expression of CD19⁺CD27⁻CD24^hiCD38^hi (immature Bregs) (34) or CD19⁺CD27⁺CD24^hi (memory B cells) (21) surface markers. Lineage-specific markers to exclusively identify Breg populations have yet to be described. The expression of the cytokine interleukin-10 (IL-10) appears to be the only known distinguishing feature and the central mediator that conveys the immunosuppressive functions of all Bregs subsets (24,28,35,36). Bregs have been previously shown to suppress autoimmune disease in mouse models of experimental autoimmune encephalomyelitis (EAE) (23,37) and arthritis (31), and impaired Breg responses are associated with several autoimmune diseases in humans, including multiple sclerosis (38) and lupus erythematosus (34). However, the role of Bregs in autoimmune T1D remains unknown. Moreover, the mechanisms regulating 1) the selective inhibition of antigen-specific immune responses and 2) the physiological signals controlling IL-10 production by Bregs require further elucidation.

Although reports indicate that ∼20% of NOD mice are naturally protected against the spontaneous development of diabetes (39), the mechanisms underlying this protection have yet to be identified (40). We previously showed that following depletion of the B-cell pool in NOD mice, the re-emerging B cells exhibit an immunosuppressive phenotype and inhibited diabetes onset (3). Furthermore, published data from other groups suggest a key role for Bregs in limiting autoimmune disease. On this basis, we hypothesized that 1) a population of naturally occurring, islet-infiltrating Bregs abrogates autoimmune responses and thus protects these 20% of NOD mice from the onset of autoimmune diabetes and 2) hyperglycemic NOD mice and individuals with T1D lack this population of Bregs.

Blood samples were obtained from healthy individuals and from individuals with T1D and their first-degree relatives who were free from diabetes but had detectable levels of the aAbs anti-GAD and anti–IA-2 in their peripheral blood (Table 1), in accordance with the Joslin Diabetes Center and Boston Children’s Hospital Institutional Review Boards.

Female NOD/ShiLtJ (NOD), NOD.CB17-Prkdcscid/J (NOD.Scid), NOR/LtJ, NOD.BDC2.5 (BDC2.5), and C57BL/6J (B6) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Human CD20 transgenic NOD mice [hCD20(^+/+) NOD] were generated and maintained as previously reported (15). All mice were used according to institutional guidelines, and animal protocols were approved by the Boston Children’s Hospital Institutional Animal Care and Use Committee.

Overt diabetes was defined as blood glucose levels above 250 mg/dL for 3 consecutive days. Blood glucose was measured using the OneTouch Ultra (LifeScan Inc., Milpitas, CA) blood glucose meter. Insulitis score was determined on 150 hematoxylin and eosin (H&E)-stained islets per group (n = 3 mice), as previously described (41).

Splenic CD4⁺ T cells (2 × 10⁵) were isolated from BDC2.5 TCR transgenic mice using CD4⁺ monoclonal (m)Ab-coated microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany) stimulated with 150 ng/mL BDC2.5 peptide and cocultured with splenic IL-10–producing B cells isolated from NOD mice using the Breg isolation kit (Miltenyi Biotec) in a 2:1 ratio, respectively. When indicated, dendritic cells (DCs), isolated using CD11c⁺ mAb-coated microbeads (Miltenyi Biotec), were added in a 2:1:1 ratio as described (42,43). To study the effect of IL-10 secreted by IL-10⁺ Bregs on T-effector cell activation and differentiation, 5 μg/mL of anti–IL-10 blocking Ab was added to the coculture system. Interferon-γ (IFN-γ) ELISA spot (ELISPOT) assays and flow cytometric analysis of cytokine production and activation marker expression were performed as described.

B cells were isolated from lymphocyte preparations of peripheral blood mononuclear cells (PBMCs) of healthy donors, T1D individuals, and their aAb⁺, diabetes-free relatives using CD19 mAb-coated microbeads (Miltenyi Biotec). B cells (2.5 × 10⁵) were cultured for 4 days in the presence of anti-human CD40 ligand (2 µg/mL; R&D Systems) and lipopolysaccharide (10 µg/mL; Sigma-Aldrich, St. Louis, MO) in RPMI-1640 (Gibco, Grand Island, NY) containing 10% human serum (Mediatech Inc., Herndon, VA).

Unless otherwise indicated, all data are shown as mean ± SEM. Statistical analysis was performed using the unpaired Student t test. A two-sided value of P ≤ 0.05 was considered statistically significant. The Kaplan-Meier curve with the Wilcoxon test was used to analyze the development of diabetes in mice. Statistical analysis was performed using GraphPad Prism software (GraphPad Software, Inc., La Jolla, CA).

To study B-cell infiltration patterns of pancreatic islets during the onset and progression of T1D in NOD mice, histological grading was performed on pancreatic cross-sections of 4- and 10-week-old normoglycemic (Nglc), hyperglycemic (Hglc) (average age, 19 weeks), and long-term normoglycemic (Lnglc) (average age, 30 weeks) female NOD, as well as 10-week-old nonautoimmune B6 mice.

Islets of 4-week-old Nglc NOD mice and B6 nonautoimmune control mice demonstrated a well-preserved islet architecture with abundant insulin staining devoid of lymphoid infiltrate as indicated by negative staining for the pan–B-cell marker B220 and the T-cell marker CD3 (Fig. 1A and B). Islets from 10-week-old Nglc and Lnglc NOD mice exhibited peri-insulitis, constituted by B cells and T cells (Fig. 1A and B). Insulin staining was preserved despite infiltration (Fig. 1B). Conversely, islets from Hglc NOD mice exhibited substantial lymphoid infiltrate, consisting of clustered B220⁺ B cells and scattered CD3⁺ T cells, while staining negative for insulin. Overall, islets from Lnglc mice showed a reduced, less organized B-cell–infiltration pattern when compared with Hglc NOD mice. Consistent with histopathological findings, islet-specific B220 mRNA expression was upregulated 13.4-fold in 10-week-old Nglc NOD mice (P ≤ 0.05), 23.7-fold in Hglc mice (P ≤ 0.05), and 20.5-fold in Lnglc NOD mice compared with 4-week-old mice (P ≤ 0.001), as determined by quantitative real-time PCR (Fig. 1C).

To address whether differences in the humoral immune response may convey protection against the autoimmune destruction of pancreatic islets in Lnglc NOD mice, we analyzed serum insulin aAb (IAA) (Fig. 1D) levels in different groups of NOD mice; however, no significant differences were observed. BAFF (Fig. 1E) was significantly (P ≤ 0.05) elevated in the serum of 10-week-old Nglc, Hglc, and Lnglc NOD when compared to 4-week-old mice.

Serial pancreatic tissue sections were stained for the pan–B-cell marker B220, the proliferation marker Ki-67, and TUNEL assay to assess a potential imbalance between proliferating and apoptotic B cells. No significant differences were evident in B-cell proliferation or apoptosis among Hglc and Lnglc NOD mice (Fig. 1F). Importantly, a significant (P ≤ 0.001) increase in apoptotic B cells was detected within islet-infiltrating B-cell populations of Lnglc, but not Hglc, compared with 10-week-old Nglc NOD mice (Fig. 1F). Because mice are not fully grown until 8 weeks of age, the pancreatic tissue of 4-week-old NOD mice showed Ki-67 positivity (Fig. 1B, panel 26).

Ectopic germinal center (GC)–like structures were quantified based on simultaneous expression of Bcl-6, Ki-67, and PNA-binding sites within morphologically distinct areas of proliferating B220⁺ B cells on serial tissue sections (Fig. 1H shows an examplary case from a 10-week-old NOD mouse). No ectopic GCs were identified in pancreatic islets of 4-week-old Nglc NOD or B6 control mice; however, 2.0% of islets of 10-week-old Nglc and Hglc NOD mice (P ≤ 0.001 vs. 4-week-old Nglc NOD, respectively) and 1.3% of islets of Lnglc mice (NS) contained extranodal GCs (Fig. 1G).

To characterize the islet-infiltrating B-cell repertoire in the different groups of NOD mice, B-cell receptor sequence libraries were generated (44,45) from individual islets, and the variable region of the heavy chain (V_H) was analyzed. A comparison of immunoglobulin (Ig)V_H sequences from the islet-infiltrating B-cell repertoire of NOD mice to germline segments using IMGT/V-QUEST software revealed both silent and amino acid replacement mutations throughout the variable regions of the B-cell receptor. Compared with the germline, low frequencies of nucleotide mutations per IgV_H transcript were found in 4- and 10-week-old Nglc NOD mice, indicating a primarily naïve B-cell phenotype (Fig. 2A and B). No significant increases in the number of nucleotide mutations were observed in B-cell receptor libraries of Hglc NOD mice; however, the islet-infiltrating B-cell repertoire of Lnglc mice showed significantly higher mutation rates compared with 10-week-old Nglc NOD mice (P ≤ 0.05) (Fig. 2A and B). Similarly, islet-infiltrating B cells from 4- and 10-week-old Nglc NOD mice acquired less than two amino acid replacement mutations per transcript within the IgV_H region (NS) (Fig. 2C and D). However, significant increases in amino acid replacement mutations were evident in Lnglc, but not Hglc, NOD mice compared with 4-week-old Nglc and with 10-week-old Nglc mice (P ≤ 0.05, respectively) (Fig. 2C and D).

The complementarity-determining region 3 (CDR3) within the variable region of the IgV_H plays an important role in determining Ag specificity by forming part of the epitope-binding site. Nucleotide and amino acid mutations within the CDR3 region were detected in B cells isolated from individual islets of all groups of NOD mice. No significant differences in frequencies of amino acid mutations within the CDR3 region were observed in 4-week-old Nglc, 10-week-old Nglc, Hglc, and Lnglc mice (Fig. 2E and F, NS). A semi-nested PCR approach was used to amplify IgV_H genes from islet-harbored B cells to study isotype expression among the B-cell infiltrate. The diversity of different isotypes expressed by the B-cell infiltrate was determined from sequence analysis of the isotype-specific constant regions and visualized by agarose gel separation of isotype-specific PCR products. Amplification of the IgV_H gene repertoire of B cells isolated from cross-sections of individual islets of 4-week-old Nglc NOD mice showed that 100% of the islets examined were infiltrated with B cells expressing IgM, whereas only 20% of the islets were infiltrated with detectable levels of B cells expressing IgG1, IgG2a/c, and IgG2b. No detectable levels of B cells expressing IgG3 were isolated from islets of 4-week-old Nglc NOD mice (Fig. 3A). Similarly, 100% of islets isolated from 10-week-old Nglc NOD mice were infiltrated by B cells expressing IgM and IgG2a/c, whereas IgG1 expressing B cells were found in 60% of islets and IgG2b⁺ expressing B cells were found in 20% of islets. No B cells expressing IgG3 were found infiltrating the islets of these mice (Fig. 3A). Taken together, the isotype profile of islet-infiltrating B cells suggests a naïve B-cell pool present in islets isolated from 4- and 10-week-old Nglc NOD mice. In contrast, most of the islet-infiltrating B cells from Hglc and Lnglc NOD mice had class switched from IgM to IgG isotypes (IgG1: 83 and 71%; IgG2a/c: 100 and 60%; IgG2b: 100 and 81%; IgG3: 83 and 57% in Hglc and Lnglc mice, respectively), demonstrating an Ag-experienced phenotype for most of the B cells (Fig. 3A). Representative agarose gel images visualizing isotype distribution within individual islets for each group are shown in Fig. 3B.

In addition to somatic hypermutation and class switching, IgV_H sequences within libraries constructed from pancreatic islets of different groups of NOD mice were analyzed for evidence of clonal expansion. Sequences were considered progeny of the same clone when their CDR3 region was identical and the IgV_H did not differ in more than one nucleotide replacement mutation. Analysis of IgV_H sequence libraries constructed from 4- and 10-week-old Nglc and Hglc NOD mice showed that the number of clones per 100 sequences analyzed corresponded with disease progression (Fig. 3C). Interestingly, fewer individual clones where found within libraries constructed from islet cross-sections of Lnglc compared with Hglc NOD mice (NS) (Fig. 3C). The CDR3 sequences of the individual clones within B-cell receptor libraries can be found in the Supplementary Table 1.

Next, we screened for clonal variants to determine whether diversification of variable regions in B cells had yielded a mature B-cell receptor repertoire. We defined IgV_H sequences as clonally related variants when they shared recombination events in the VDJ junction, as indicated by identical CDR3 regions, and differed in at least two nucleotide replacement mutations. Sequence analysis revealed significantly (P ≤ 0.05) higher percentages of clones with variants in 10-week-old Nglc and Lnglc compared with 4-week-old Nglc NOD mice (Fig. 3D). Intraclonal isotype switches were not observed in B-cell receptor libraries obtained from 4-week-old Nglc NOD mice; however, libraries generated from 10-week-old Nglc, Hglc, and, to an even greater extent, Lnglc demonstrated markedly enhanced (P ≤ 0.05) frequencies of intraclonal isotype switches (Fig. 3E).

CD40 ligation is known to play a key role in B-cell proliferation, induction of Ig class switches, somatic hypermutation, and generation of Bregs (24). Although the frequencies of splenic B220⁺ B cells expressing CD40 (Fig. 4A), CD80 (Fig. 4C), and MHC class II (Fig. 4E) isolated from B6, NOR, 10-week-old Nglc, Hglc, and Lnglc NOD mice were not statistically different, significantly higher percentages of splenic B cells expressing both CD40 and MHC class II were observed in 4-week-old Nglc NOD mice compared with the other groups (P ≤ 0.05) (Fig. 4A and E). Interestingly, the frequency of islet-infiltrating B cells expressing CD40 (Fig. 4B), but not CD80 and MHC class II (Fig. 4D and F), was significantly (P ≤ 0.05) increased in Lnglc compared with 10-week-old Nglc but not Hglc NOD mice. The percentages of B220⁺CD93⁺CD23⁺IgM^low anergic B cells (46) obtained from splenocytes were similar between the different groups (Fig. 4G). Frequencies of anergic B cells in islets of 10-week-old Nglc and Hglc NOD mice were comparable to those observed in the spleen; however, a significant (P ≤ 0.001) increase in the percentage of islet-infiltrating anergic B cells was observed in Lnglc NOD mice (Fig. 4H). Finally, the frequencies of naturally occurring IL-10⁺ B cells, suggested to be endowed with immunoregulatory properties (14,22,26), were then analyzed. Percentages of IL-10⁺ B cells recovered from the spleens of B6, 10-week-old NOR, 4-week-old Nglc, and 10-week-old Nglc, Hglc, or Lnglc NOD mice were similar among groups (NS) (Fig. 4I). The frequencies of IL-10⁺ B cells among the islet-infiltrating B-cell pool of 10-week-old Nglc and Hglc NOD mice were comparable to frequencies observed in the spleen; however, the percentages of islet-infiltrating IL-10⁺ B cells isolated from the pancreata of Lnglc NOD mice were significantly higher (P ≤ 0.05) (Fig. 4J). Importantly, frequencies of CD4⁺CD25⁺FoxP3⁺ Tregs were not significantly different among islet-infiltrating lymphocytes obtained from Hglc and Lnglc NOD mice (data not shown).

To assess the ability of total splenic B-cell populations to present BDC2.5 peptides and hence to stimulate autoreactive BDC2.5 transgenic CD4⁺ T cells, IFN-γ production by responder T cells was measured. Total splenic B cells isolated from 10-week-old Nglc NOD mice stimulated autoreactive CD4⁺ T cells more efficiently (P ≤ 0.05) than B-cell populations isolated from Hglc or Lnglc NOD mice, respectively (Fig. 5A). Next, we assessed whether total splenic B cells were endowed with immunoregulatory properties. Although the addition of B cells from the three experimental groups augmented T-cell activation compared with cocultures of CD4⁺ T cells with CD11c⁺ DC alone, no significant differences in IFN-γ production were evident in cocultures with B cells from 10-week-old Nglc, Hglc, and Lnglc NOD mice (Fig. 5B), suggesting that peripheral B-cell populations have stimulatory but not regulatory properties.

To evaluate the regulatory properties of peripheral B cells from 10-week-old Nglc, Hglc, and Lnglc NOD mice in vivo, B220⁺ B-cell and CD4⁺ T-cell populations were isolated from the different groups of NOD mice and adoptively transferred into NOD.Scid mice (n = 5, respectively). No significant differences in diabetes onset were observed among mice injected with B and T cells from Hglc mice, B cells from 10-week-old Nglc and T cells from Hglc mice, B cells from Lnglc and T cells from Hglc mice, or mice injected with T cells from Hglc mice alone (Fig. 5C). Interestingly, all NOD.Scid recipient mice that received CD4⁺ T cells from Lnglc mice remained normoglycemic after a 20-week follow-up (Fig. 5C).

To study whether IL-10–producing B-cell subsets exhibit deficits in Ag presentation, B220⁺ cells producing IL-10 versus IL-10⁻ B-cell bulk populations were isolated from the spleen of Nglc NOD mice and cocultured with BDC2.5 transgenic CD4⁺ T cells in the presence of BDC2.5 peptide. Upon isolation, IL-10–producing B220⁺ cells (IL-10⁺ B cells) were enriched by >12-fold compared to the flow-through B-cell subset (IL-10⁻ B cells) (Fig. 5D). Interestingly, compared IL-10⁻ B-cell populations, cocultures with IL-10⁺ B cells resulted in significantly reduced (P ≤ 0.05) levels of IFN-γ production (Fig. 5E), suggesting defects in the ability of IL-10⁺ B cells to present BDC2.5 self-peptide to BDC2.5-transgenic CD4⁺ T cells. Ab-mediated blockade of IL-10 did not alter B-cell Ag-presenting function as determined indirectly via assessment of IFN-γ production by cocultured T cells (Fig. 5F). To test whether secretion of the immunoregulatory cytokine IL-10 by Bregs may have a direct effect on activation and/or differentiation of T-effector cells, flow cytometric analysis was performed on CD4⁺ T cells that were cocultured in the presence of anti–IL-10 blocking or control Ab. Ab-mediated blockade of IL-10 did not significantly affect activation of CD44⁺CD62L⁻ T-effector cells or IL-2 and IL-17 production by CD4⁺ T cells in Breg cocultures but did result in a significant (P ≤ 0.01) reduction of IL-4 production by CD4⁺ T cells (Fig. 5G).

To study the immunoregulatory properties of naturally occurring IL-10–producing B cells ex vivo, splenic IL-10⁺ B cells were isolated from Nglc NOD mice and cocultured with CD11c⁺ DCs and BDC2.5 transgenic CD4⁺ T cells in the presence of BDC2.5 peptide. IFN-γ production by transgenic CD4⁺ T cells was significantly (P ≤ 0.05) decreased in cocultures with IL-10⁺B220⁺ B cells compared with IL-10⁻B220⁺ B cells (Fig. 5H). To test whether IL-10⁺ B-cell subsets are able to suppress T-cell–mediated destruction of pancreatic islets in vivo, splenic IL-10⁺ B cells and IL-10⁻ B-cell subsets were isolated from Nglc NOD mice and adoptively cotransferred with diabetogenic CD4⁺ T cells into NOD.Scid recipient mice. Consistent with our ex vivo findings, IL-10⁺B220⁺ cells suppressed T cell–mediated autoimmune destruction of pancreatic islets and maintained normoglycemia in NOD.Scid mice significantly (P ≤ 0.05) longer compared with transferred IL-10⁻ B cells (time free from diabetes: IL-10⁺ B cells: 15 weeks vs. IL-10⁻ B cells: 7 weeks post-adoptive transfer; n = 4 mice, respectively) (Fig. 5I).

To show that B cells are at least partially responsible for maintaining normoglycemia in Lnglc NOD mice, we attempted to deplete naïve B cells using an anti-mouse CD20-depleting mAb (47). However, consistent with published data (48), flow cytometric analysis revealed a downregulation of CD20 within the peripheral B-cell pool of Lnglc NOD mice even before injection with the anti-mouse CD20-depleting Ab (data not shown), thus rendering Lnglc NOD mice resistant to B-cell depletion. We therefore used hCD20(^+/+) NOD mice (15) in which B cells are engineered to stably express high levels of human CD20 and are hence susceptible to B-cell depletion with rituximab (anti-human CD20 mAb). Similar to wild-type NOD mice, approximately 10–15% of these animals fail to develop hyperglycemia. We successfully depleted the B-cell compartment of Lnglc hCD20(^+/+) NOD mice and adoptively transferred B cells isolated from Hglc NOD mice 1 week after depletion (Fig. 5J). Although 100% of untreated control mice (n = 5), mice depleted with rituximab that did not receive adoptively transferred B cells (n = 4), and mice that were not depleted but received B cells from Hglc NOD mice (n = 3) maintained normoglycemia for >10 weeks, five of seven B-cell–depleted mice that subsequently received B cells from Hglc NOD mice (n = 7) became hyperglycemic (P ≤ 0.05) (Fig. 5J).

To test whether our findings in the NOD mouse model of autoimmune diabetes extend to the human setting, we studied the phenotype of B cells in the peripheral blood of healthy volunteers, individuals with T1D, and relatives of T1D patients with detectable serum levels of aAb but without overt disease. Interestingly, frequencies of CD40⁺ B cells (Fig. 6A), but not CD80⁺ or MHC class II⁺ B cells (Fig. 6B and C), were significantly (P ≤ 0.05) higher in the peripheral blood of healthy individuals and relatives with serum aAb compared with T1D patients. On contrary, percentages of anergic IgD⁺IgM⁻CD27⁻ B cells (as previously described [49]) were markedly (P ≤ 0.05) lower in healthy individuals compared with individuals with T1D (Fig. 6D). Importantly, IL-10⁺ B cells were significantly (P ≤ 0.05) higher in the peripheral blood of healthy individuals compared with relatives with aAb and individuals with T1D (Fig. 6E).

To study whether unseparated peripheral B cell populations isolated from healthy individuals or those with T1D possess immunoregulatory functions ex vivo, PBMCs containing or depleted of B cells were cultured in the presence of the islet-specific peptides GAD or IA-2, and IFN-γ–production was quantified. Interestingly, compared with B-cell–depleted PBMC cocultures, stimulation of B-cell–containing PBMCs isolated from T1D patients with IA-2, but not with GAD, resulted in significantly (P ≤ 0.001) increased levels of IFN-γ production (Fig. 7B and C), suggesting that whole B-cell subsets have stimulatory but not immunoregulatory activity. T-cell activation in response to Pediacel served as a positive control (P ≤ 0.05) (Fig. 7A).

To assess whether individuals with T1D showed defects in Breg generation, peripheral B cells were isolated from healthy control subjects, individuals with T1D, and those with aAb but without T1D and subjected to an ex vivo Breg generation assay. Frequencies of IL-10⁺ B cells generated from healthy individuals were significantly (P ≤ 0.05) higher compared with individuals with T1D and aAb⁺ relatives (Fig. 7D).

To study the immunoregulatory properties of IL-10–producing B cells, IL-10⁺ B cells were isolated from PBMCs of individuals with T1D, expanded ex vivo, and cocultured with matched PBMCs in the presence of IA-2 peptide. Paralleling our findings in the NOD mouse model, IFN-γ production by T cells was significantly (P ≤ 0.05) decreased when IL-10⁺CD19⁺ B cells, but not IL-10⁻CD19⁺ B cells, were added to the coculture (Fig. 7E). Pediacel served as a positive control for T-cell activation.

Our current understanding of T1D pathogenesis suggests that B cells play a pivotal role in promoting disease onset and progression, yet little is known about their ability to inhibit the autoimmune destruction of pancreatic islets (1–3). Here, we show that a subset of IL-10–producing, immunoregulatory B cells plays a key role in maintaining normoglycemia in a group of NOD mice that were naturally protected from overt disease (i.e., hyperglycemia) and potentially healthy individuals. Conversely, Hglc NOD mice, individuals with established T1D, and individuals with aAbs but without disease demonstrated reduced frequencies of this population of Bregs.

Specifically, we demonstrated that the pancreatic islets of Lnglc NOD mice that are naturally protected from autoimmune diabetes for as-of-yet unknown reasons contained a limited B-cell infiltrate. Surprisingly, IgV_H transcriptome analysis revealed a highly mutated, class-switched B-cell receptor repertoire within islets of Hglc and, to an even greater extent, Lnglc NOD mice, suggestive of an Ag-experienced B-cell pool. Further examination of the VDJ junction within IgV_H transcripts revealed a progressive increase in clonal diversity within islet-infiltrating B cells from 4-week-old Nglc to Hglc NOD mice, suggesting that the B-cell repertoire continues to undergo affinity maturation (44,45) during the manifestation of diabetes. In contrast, we observed fewer individual clones within islets of Lnglc mice. In summary, we identified hallmark features of affinity maturation, namely somatic hypermutation, insertions and deletions, and isotype switching (44,45) in Hglc and, to an even greater extent, Lnglc NOD mice. We therefore propose that islets from Lnglc NOD mice are infiltrated by a pool of highly matured, selected B-cell clones with a unique specificity against islet-associated antigens, whereas lower-affinity clones were previously removed by negative selection. This hypothesis is supported by our findings of increased frequencies of apoptotic and anergic B cells within the pancreatic islets of Lnglc mice.

Although the age difference between Hglc (∼19 weeks) and Lnglc (∼30 weeks) mice may allow for B cells to acquire receptor diversity by incorporating stochastic mutations over time, data showing the accumulation of mutations specifically within the CDR3 region suggests that the pancreatic islets provide a distinct microenvironment in which B cells, upon Ag encounter, can undergo affinity maturation and clonal expansion. Importantly, islet-infiltrating B cells from Lnglc NOD mice proved to be highly mature and clonally expanded without promoting autoimmune destruction of insulin-producing β-cells, indicating that these B cells are functionally impaired and/or able to negatively regulate autoimmunity. Islets from Lnglc mice maintained high frequencies of both CD40⁺ and IL-10⁺ B cells compared with 10-week-old Nglc or Hglc NOD mice. Likewise, healthy individuals and aAb-positive relatives of T1D patients exhibit increased frequencies of CD40⁺ and IL-10⁺ B cells when compared to individuals with T1D. It has been previously shown that CD40 signaling induces IL-10 competence in B cells (50,51). In addition IL-10 has been described to exert anti-inflammatory effects, while enhancing survival, proliferation, differentiation, and isotype switching of B cells (52). Increased CD40 signaling in Lnglc NOD mice, healthy individuals, and relatives with serum aAb may possibly provide a strong stimulus, which could induce anergy or apoptosis in these hyperactivated B cells (14) or stimulate the generation of Bregs (50,51). Phenotypic analysis of B-cell populations in the periphery revealed differences between B-cell subsets in the blood of healthy individuals versus T1D patients, while no significant differences were observed between splenic B cells from normoglycemic versus hyperglycemic NOD mice. This may, at least in part, be due to the fact that the disease persists much longer in humans compared to mice, thus allowing sufficient numbers of B cells to migrate from the pancreatic microenvironment, where they are imprinted, to the periphery.

Our data further suggest that IL-10 plays a pivotal role in mediating immunosuppressive functions, as only IL-10⁺ B cells, but not whole B-cell populations or IL-10⁻ B-cell subsets, were able to suppress autoreactive T-cell activation ex vivo and prevent diabetes onset in vivo. Importantly, rituximab-mediated depletion of B cells followed by adoptive transfer of B cells from Hglc NOD mice converted Lnglc hCD20(⁺/⁺) NOD mice to hyperglycemia. Finally, the observed defect in generation of IL-10⁺ B cells in patients with T1D compared to healthy individuals further suggests that the tolerance mechanism we identified in the experimental NOD mouse model may, at least in part, translate to human disease.

Together, our findings provide novel insights on a population of naturally occurring, highly matured Breg that mediate protection against autoimmune destruction of pancreatic islets and possess the potential requirements for maintenance and expansion in vivo (Fig. 7F). Our data further suggest that in the setting of T1D, the inflammatory microenvironment renders this B-cell subset functionally impaired or apoptotic. This novel population of B cells may hold the potential to restore tolerance to aAg in individuals with T1D, which remains a major challenge for investigators in the field.

Funding. P.F. received a JDRF Career Development Award, an American Society of Nephrology Career Development Award, an American Diabetes Association (ADA) Mentor-based Fellowship grant, and is supported by an American Heart Association (AHA) Grant-In-Aid, Translational Research Program (TRP) grant from Boston Children's Hospital, and Harvard Stem Cell Institute grant (“Diabetes Program” DP-0123-12-00). P.F. also received Italian Ministry of Health grants RF-2010-2303119 and “Staminali” RF-FSR-2008-1213704. A.V. has been supported by a National Institutes of Health (NIH) Research Training grant to Boston Children's Hospital in Pediatric Nephrology (T32-DK-007726-28). M.B.N. is supported by an ADA Mentor-based Fellowship grant to P.F. R.B. is supported by an American Society of Transplantation/Genentech Clinical Science Fellowship grant. L.W. is supported by JDRF (4-2007-1059) and NIH (RC1DK087699 and DK088181).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.K., A.V., and P.F. planned the project. S.K., A.V., S.T., M.B.N., M.A.N., S.W., and R.B. carried out experimental work. S.K., A.V., S.T., M.B.N., M.A.N., S.W., R.B., F.D., T.S., R.A., M.A., M.H.S., L.W., C.H.W., K.C.O., and P.F. analyzed data. S.K. and P.F. wrote the manuscript. All authors discussed the results and commented on the manuscript. P.F. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.