Autoantigen-Specific B-Cell Depletion Overcomes Failed Immune Tolerance in Type 1 Diabetes

The anti-insulin VH125Tg [Cg-Tg(Igh-6/Igh-V125)2Jwt/JwtJ] and non–insulin-binding VH281Tg [Tg(Igh-6/Igh-V281)3Jwt/JwtJ] mice used in this study harbor a nontargeted heavy chain transgene on C57BL/6 or NOD backgrounds, as described previously (5,15). Sera from 125Tg/NOD mice [NOD.Cg-Tg(IGk-V125)1Jwt/JwtJ mice intercrossed with VH125Tg mice] served as positive controls in enzyme-linked immunosorbent assay (ELISA) to detect anti-insulin Ab (15). Age ranges are indicated in figure legends. All data are derived from lines backcrossed >20 generations to C57BL/6 or NOD that are hemizygous for all transgenes indicated. All mouse lines are available from The Jackson Laboratory. All mice were housed under sterile, specific pathogen-free housing conditions, and all studies were approved by the Institutional Animal Care and Use Committee of Vanderbilt University.

BM was eluted from long bones, and spleens were macerated with Hanks’ balanced salt solution (HBSS; Invitrogen) plus 10% fetal bovine serum (FBS; HyClone). Red blood cells were lysed using Tris-NH₄Cl. Freshly isolated pancreata were digested with 3 mL of 1 mg/mL collagenase P diluted in HBSS for 30 min, then tissue was disrupted using an 18-gauge needle. HBSS plus 10% FBS was immediately added to inhibit collagenase activity. Cells were directly analyzed by flow cytometry. Alternatively, 10 × 10⁶ cells/mL splenocytes was cultured in complete culture medium (Dulbecco’s modified Eagle’s medium, 10% FBS, l-glutamine, HEPES, minimum essential medium sodium pyruvate, nonessential amino acids, gentamycin, 2 × 10⁻⁵ M β-mercaptoethanol [Invitrogen]) for 18 h in a 37°C CO₂ incubator and stimulated with human insulin (Sigma-Aldrich) or anti-IgM (Jackson ImmunoResearch Laboratories).

Ab reagents reactive with B220 (6B2), IgM^a (DS-1), IgM^b (AF6-78), CD19 (1D3), CD21 (7G6), CD23 (B3B4), CD86 (GL1), 7-aminoactinomycin D, DAPI (BD Biosciences), or IgM (μ chain specific; Invitrogen) were used for flow cytometry. Biotinylated human insulin (16) was detected with fluorochrome-labeled streptavidin (BD Biosciences). To restrict analysis to antigen-specific cells: insulin-specific B cells = % insulin-binding B cells, no inhibition (Fig. 1A, top) − % insulin-binding B cells, 10× unlabeled insulin competitor (Fig. 1A, bottom); this was further confirmed by a linear relationship of insulin binding and IgM^a. BCR occupancy with endogenous insulin or BCR preloaded with 50 ng/mL human insulin on ice was detected using a second anti-insulin antibody, mAb123, which was biotinylated. mAb123 binds a distinct insulin epitope from mAb125 (from which VH125Tg is derived) and detects anti-insulin 125Tg BCR occupancy with endogenous insulin (15), but not insulin bound to the hormone receptor (17). Flow cytometry analysis was performed using an LSRII (BD Biosciences) and FlowJo software (Tree Star).

In short-term mAb administration experiments, 100 μg of anti-insulin mAb123 or IgG1 isotype control antibody (ATCC #HB-123 or #CRL-2395 hybridoma supernatant purified by the Vanderbilt Antibody and Protein Resource, respectively) in 100 μL 1× PBS (pH 7.4) and injected i.p. into VH125Tg/NOD mice weekly for 3 weeks. After 2–14 d, BM and spleens were harvested, and flow cytometry analysis was performed as above. For long-term administration of mAb in disease studies, 100 μg mAb123 was injected i.p. into WT/NOD or VH125Tg/NOD mice every other week starting at 3 weeks of age and continued throughout the disease study. Blood glucose was monitored weekly in untreated or mAb123-treated mice beginning at 10 weeks of age; diabetes was diagnosed following two consecutive >200 mg/dL blood glucose readings.

Insulin autoantibodies are commonly detected in WT/NOD mice prior to type 1 diabetes onset (18). We therefore employed flow cytometry to identify insulin-binding B cells in the spleen. Live B220⁺IgM^a+ lymphocytes and insulin-specific B cells were detected as in research design and methods. Insulin-specific B cells are difficult to reliably track in WT/NOD mice (Fig. 1A); therefore, to understand how B cells with this critical specificity are regulated during development, the VH125Tg BCR-transgenic model was employed, in which anti-insulin VH125 pairs with endogenous light chains in NOD mice to form a small population of B cells (1–2%) that recognize insulin (5). An increased frequency of anti-insulin B cells exists in the spleen of autoimmune VH125Tg/NOD mice compared with nonautoimmune VH125Tg/C57BL/6 mice (Fig. 1A).

The transitional 1 (T1) B-cell stage in the spleen is a major tolerance checkpoint that censors autoreactive specificities in the developing repertoire (10–12). Flow cytometry was therefore used to identify whether insulin-binding B cells may be culled at this checkpoint. T1 (CD21^lowCD23^low), follicular (CD21^midCD23^high), and marginal zone (CD21^highCD23^mid) B-cell subsets were identified within live B220⁺IgM^a+ lymphocytes, and insulin-specific B cells were detected as in Fig. 1A in VH125Tg/C57BL/6 and VH125Tg/NOD mice. Insulin-specific B cells are present in the T1 (0.65 ± 0.19), as well as mature follicular (0.78 ± 0.17) and marginal zone (1.15 ± 0.20) B-cell compartments of VH125Tg/NOD mice (Fig. 1B). No significant decrease in the percentage of insulin-binding B cells is observed between the T1 to follicular or marginal zone B-cell compartments within VH125Tg/NOD mice; rather, a significantly increased frequency of insulin-binding B cells is found among marginal zone B cells (P < 0.0001), a site associated with autoreactive B-cell accumulation (19). In contrast, VH125Tg/C57BL/6 mice show a markedly reduced (or absent) percentage of insulin-binding B cells among the T1 (0.10 ± 0.14), follicular (0.04 ± 0.13), and marginal zone (−0.02 ± 0.12) spleen B-cell compartments. These data show that NOD anti-insulin B cells fail to be effectively censored at the T1 stage and compete effectively within a polyclonal repertoire for entry into mature B-cell subsets.

Anti-insulin B cells maintain functions necessary to promote β-cell destruction in VH125Tg/NOD mice (5). To identify whether insulin autoantigen specifically evokes costimulatory molecule upregulation in anti-insulin B cells, VH125Tg/NOD splenocytes were cultured in the presence or absence of insulin in vitro, and CD86 expression was measured using flow cytometry. The change in CD86 expression (mean fluorescence intensity [MFI]) was significantly increased approximately twofold (P < 0.01) in insulin-binding B cells exposed to insulin in vitro compared with those cultured in the absence of insulin. CD86 expression remained unchanged in non–insulin-binding B cells present in the same cultures, indicating CD86 upregulation was driven by specific insulin recognition through the BCR (Fig. 2A). To further investigate whether the potential to upregulate CD86 was different between anti-insulin and non–insulin-binding B cells, VH125Tg/NOD splenocytes were stimulated in vitro with anti-IgM to provide uniform BCR stimulation. Insulin-binding B cells and non–insulin-binding B cells showed comparable CD86 upregulation (Fig. 2B). CD86 expression was found to be significantly increased two- to threefold in insulin-binding B cells (P < 0.001), compared with non–insulin-binding B cells identified in freshly isolated pancreata (Fig. 2C). These data suggest that endogenous levels of autoantigen induce upregulation of CD86 in the disease setting. Similar outcomes were obtained in studies examining CD40 upregulation (not shown).

To identify whether anti-insulin B cells in VH125Tg/NOD mice are functionally silent for autoantibody production, sera were tested for the presence of IgM^a insulin autoantibody using ELISA (see research design and methods). No insulin autoantibody was detected in the serum of VH125Tg/NOD mice compared with positive control 125Tg/NOD mice, which possess a small amount of insulin autoantibody (Fig. 2C). These data indicate that anti-insulin B cells that escape into the periphery of VH125Tg/NOD mice remain competent to upregulate costimulatory molecules in response to their cognate antigen. They do not produce detectable insulin autoantibody, suggesting that tolerance breakdown in VH125Tg/NOD mice includes the retention of T-cell stimulation capability necessary for APC function that is independent of autoantibody production.

Flaws in tolerance revealed in VH125Tg/NOD mice suggest that BCR signaling following endogenous insulin encounter does not sufficiently evoke a mechanism to cull these cells (Fig. 1). To identify whether these cells are clonally ignorant, due to low concentrations of circulating insulin, mAb123-biotin was used to identify whether T1, follicular, or marginal zone BCR were occupied by insulin using flow cytometry. T1 (1.18 ± 0.14), follicular (1.22 ± 0.30), and marginal zone (1.58 ± 0.21) B-cell populations are clearly stained with mAb123-biotin (Fig. 3). The percentages detected fall within the range detected with biotinylated insulin (Fig. 1). These data suggest that BCR recognition of insulin is occurring, despite the low concentration of autoantigen physiologically present.

The BCR of anti-insulin B cells in VH125Tg/NOD are occupied by insulin. The macroself model demonstrates the feasibility of BCR-mediated deletion of B cells in the periphery upon strong antigen cross-linking of the BCR (20). We therefore hypothesized that mAb123 administered in vivo would bind specifically to insulin-occupied B cells and enhance BCR signaling to reinforce tolerance as a strategy to specifically remove anti-insulin B cells. VH125Tg/NOD mice were injected weekly for 3 weeks with either 100 μg of mAb123 or 100 μg IgG₁ isotype control mAb. BM and spleens were harvested 2, 7, or 14 d following antibody treatment, and the percentage of insulin-binding B cells in the BM and spleen was assessed using flow cytometry (Fig. 1). mAb123 does not recognize insulin when it is bound to the insulin hormone receptor (17). The expected normoglycemia was observed, and no significant change in blood glucose levels was observed between the mAb123 or isotype control antibody-injected cohorts of mice (n = 11; P = 0.84) as calculated by a two-tailed t test.

The percentage of insulin-binding B cells was significantly reduced (2, 7, and 14 d, respectively) in immature B cells harvested from mice injected with mAb123 (0.33 ± 0.05, 0.45 ± 0.21, and 0.45 ± 0.20) compared with isotype control (0.72 ± 0.06, 0.93 ± 0.07, and 0.82 ± 0.07), as well as in T1 spleen B cells (mAb123: 0.11 ± 0.09, 0.42 ± 0.27, and 0.36 ± 0.17; isotype control: 0.56 ± 0.15, 1.05 ± 0.41, and 1.13 ± 0.34, respectively). However, complete depletion was not achieved in either of these developing compartments (Fig. 4A). The frequency of insulin-binding B cells was further reduced in follicular B cells isolated from mice treated with mAb123 (0.10 ± 0.05, 0.12 ± 0.05, and 0.12 ± 0.04) compared with isotype control antibody-treated mice (0.97 ± 0.17, 1.44 ± 0.11, and 1.27 ± 0.11), as well as marginal zone B cells (mAb123: 0.07 ± 0.01, 0.10 ± 0.11, and 0.09 ± 0.03; isotype control, 1.33 ± 0.19, 1.58 ± 0.42, and 2.46 ± 0.33, respectively; Fig. 4A). Elimination of anti-insulin B cells was also observed in the lymph nodes, blood, and peritoneal cavity (not shown). The percentages of total B cells present within B-cell developmental subsets were unchanged (not shown), suggesting that B cells were not depleted nonspecifically.

Tolerance induction in developing B cells that encounter insulin has been shown to result in BCR downregulation (21). To test whether IgM levels were reduced in insulin-binding B cells that had not yet undergone depletion, IgM^a MFI was compared among non–insulin-binding and insulin-binding B cells present in VH125Tg/NOD mice treated with isotype control or mAb123 antibodies. As shown in Fig. 4B, the IgM^a MFI was significantly reduced in insulin-binding immature B cells harvested from mAb123-treated mice (650 ± 133) when compared with non–insulin-binding B cells from the same mice (1,312 ± 79, P < 0.001) or with insulin-binding B cells present in isotype control antibody-treated mice (1,340 ± 54; P < 0.001), but was unchanged in non–insulin-binding B cells harvested from mice treated with isotype control (1,342 ± 92) or mAb123 antibodies (1,312 ± 79; P = 0.64). mAb123 bound to insulin occupying the BCR does not block detection of IgM^a (Fig. 3) and thus does not account for the reduced IgM^a observed. These findings are consistent with the hypothesis that mAb123 induces BCR downregulation in immature B cells in an antigen-specific manner.

Anti-insulin BCRs that are loaded with endogenous insulin are not 100% occupied, as they are further stained by fluorescently labeled insulin in vitro (not shown). To eliminate the possibility that mAb123 was capturing endogenous insulin and promoting 100% BCR occupancy, and thus blocking detection of anti-insulin B cells by biotinylated insulin, spleens from isotype control or mAb123-treated mice were placed at 37°C for 0–3 h to permit BCR turnover. As a positive control, spleens were harvested from untreated mice and incubated with excess human insulin to establish 100% BCR occupancy and were also placed at 37°C for 0–3 h. Whereas detection of insulin-binding B cells using biotinylated human insulin reappeared in the control spleens loaded with unlabeled insulin following incubation at 37°C, no insulin-binding B cells were observed in the spleens harvested from mAb123-treated mice, even after 3 h (Fig. 4C). As expected, insulin-binding B cells were present in isotype control Ab-treated mice at all time points (not shown). These data suggest that anti-insulin B cells are being specifically depleted from the repertoire, rather than being masked by mAb123 binding. Thus, mAb123 selectively enhances removal of anti-insulin B cells from the polyclonal repertoire. These data suggest that this strategy is highly effective at removing anti-insulin B cells from the mature repertoire, but less effective for developing B cells.

Anti-insulin B cells are critical for type 1 diabetes (5,8,22,23). This specificity is depleted from the repertoire in vivo by mAb123 treatment (Fig. 4); therefore, we tested whether this strategy could prevent disease in NOD mice. mAb123 was administered to VH125Tg/NOD or WT/NOD mice every other week beginning at 3 weeks of age, and diabetes outcomes were determined (see RESEARCH DESIGN AND METHODS). Significant disease protection is observed for mAb123-treated versus untreated WT/NOD mice (Fig. 5; P = 0.046). In contrast, disease protection is not observed in mAb123-treated VH125Tg/NOD mice (compared with untreated). These data suggest that mAb123 protects against disease in a polyclonal repertoire in which the frequency of anti-insulin B-cell formation in the developing repertoire is very low, but the same regimen is not sufficient to prevent disease if BM production of anti-insulin B cells is increased.

The findings presented in this study reveal how B-cell tolerance for insulin can be restored to eliminate this critical APC specificity to prevent type 1 diabetes. Tracking of islet antigen-specific B cells shows that multiple checkpoints fail to remove anti-insulin B cells and provides a realistic picture of how immune tolerance is lost in a multigenic autoimmune disease. Anti-insulin B cells escape into the peripheral repertoire, where they are functional to upregulate CD86 when stimulated with insulin (Figs. 1 and 2), highlighting their dangerous potential to activate autoreactive T cells. Developing and mature anti-insulin BCR are occupied by endogenous insulin, despite low levels of circulating autoantigen. Targeting autoantigen-occupied BCR specifically eliminates this pernicious specificity, but is unexpectedly less effective at targeting the developing repertoire (Fig. 4). This new approach protects against type 1 diabetes in WT/NOD mice; however, increased anti-insulin B-cell production in the BM of VH125Tg/NOD mice overwhelms this protective effect (Fig. 5). Together, these data suggest that BM production of anti-insulin B cells combines with tolerance flaws to create liabilities for the pathogenesis of type 1 diabetes, which can be overcome with effective specific targeting and elimination of autoantigen-specific B cells.

Breaches in tolerance represented by autoantibodies to insulin and other islet autoantigens are reliable predictors of disease progression (18,24,25). Maternal transmission of autoantibody promotes disease (26), and antibodies serve as cofactors in experimental type 1 diabetes when islet expression of the antigenic target is engineered (27). They are, however, neither necessary nor sufficient, as disease is observed in IgM isotype-restricted VH125Tg/NOD and 125Tg/NOD mice in the absence of anti-insulin autoantibody (5), and transfer of serum containing islet autoantibodies is insufficient to drive type 1 diabetes (7). These studies employ the VH125Tg/NOD model to probe how islet-reactive B cells escape tolerance to acquire APC function to drive type 1 diabetes, independent of peripheral tolerance blockade of plasma cell differentiation and autoantibody production.

CD21 and CD23 acquisition occurs normally for anergic anti-insulin B cells in a restricted repertoire (6), in contrast with other described states of anergy (28). These studies demonstrate that in a polyclonal repertoire, maturation and effective competition for follicular niches by anti-insulin B cells remain intact (Fig. 1). An increased proportion of insulin-binding B cells mature into marginal zone B cells (Fig. 1), which are superior to follicular B cells in activating naive CD4 T cells (29). Anti-insulin B cells retain the ability to upregulate CD86 following insulin engagement (Fig. 2), further highlighting the dangerous potential of these cells. The critical role of B cells as APC in driving type 1 diabetes has been well documented (7–9). These findings suggest that anergy is not as stringently enforced among anti-insulin B cells as for other autoantigens (6,30–32). Anti-insulin B cells are thus poised to activate anti-insulin T cells, a pathogenic function validated by the fact that they fully support the development of type 1 diabetes in NOD mice (5,6).

Global B-cell depletion therapy (rituximab) shows promise for the treatment of type 1 diabetes (33), highlighting a pivotal role for B cells in disease. Unwanted side effects resulting from immune suppression (34) could be avoided through targeted silencing or elimination of autoreactive B cells. Developing B cells in the BM and T1 stage in the spleen undergo cell death following BCR stimulation (35). Given the enhanced BCR sensitivity of immature B cells (36–39), developing B cells were expected to be the most sensitive B-cell subset to this type of approach. Surprisingly, the opposite was observed, whereby mature anti-insulin B cells were more readily culled, and the immature BM and T1 spleen B cells were somewhat more resistant to elimination (Fig. 4A). Insulin-binding B cells are continuously generated in the BM, making the immature and T1 B-cell compartments more dynamic than the longer-lived mature B-cell compartment. We propose that the feasibility of therapeutic depletion using this strategy is enhanced among B cells that are further from the source of the leak in the BM. The surviving insulin-binding immature B cells show markedly decreased surface BCR expression (Fig. 4B), consistent with tolerance induction for insulin (21). We propose that anti-insulin B cells may be programmed for receptor editing or deletion when mAb123 intensifies insulin-occupied BCR stimulation in vivo. Antibody-dependent cell-mediated cytotoxicity impacts clearance of B cells in anti-CD20 treated mice (40); thus, it is likely also a factor in this treatment scenario. Future studies utilizing mAb123 F(ab’)₂ fragments will be necessary to uncover the role that Fc receptor recognition is playing in anti-insulin B-cell depletion and whether BCR engagement alone can invoke depletion.

These results suggest that reinforcing B-cell tolerance for insulin is sufficient to protect against disease (Fig. 5). The frequency of insulin-binding B cells in WT/NOD mice is very low, whereas this population is clearly increased in VH125Tg/NOD mice (Fig. 1). Insulin-binding B cells were detected in many of the mAb123-treated VH125Tg/NOD mice following disease development and study removal (not shown). We hypothesize that a crucial threshold exists, beyond which the effect of this mAb123 dosage is overwhelmed by continuous BM production. Exquisite depletion efficiency is thus a key requirement for successful autoreactive B cell–directed mAb candidates. It is also possible that VH125 skews the repertoire toward other islet specificities not targeted by mAb123.

These data add to the growing body of evidence to suggest that insulin is a critical autoantigen in type 1 diabetes (4,5) and thus a clear target for immune tolerance restoration. The human B-cell repertoire for insulin may be restricted (41), emphasizing the potential feasibility of this approach in either prevention of disease or protection against future attack of β-cells in islet grafts. These findings provide proof of concept for how anti-insulin B cells might be therapeutically removed to prevent autoimmune disease. This concept could be broadly applied to additional autoimmune diseases, provided the targeted autoantigen is present on the surface of the B cells to be depleted or that highly selective anti-idiotype reagents could be generated. A panel of mAb specific for multiple autoantigens may show the best potential for therapeutic intervention by removing the complete array of autoreactive B cells from the repertoire. The success of intravenous γ globulin in treating many autoimmune disorders, although variable, may use the same fundamental principal, as heterogeneous preparations contain multiple specificities.

On the basis of our findings, we propose the following model: tolerance failure seeds anti-insulin B cells into the mature repertoire, where they subsequently escape peripheral tolerance and upregulate CD86 following insulin stimulation to provide critical APC functions for autoaggressive T cells. This breach in tolerance is only partial, as anti-insulin B cells remain impaired for antibody production. This outcome can be circumvented by autoantigen-specific targeting to selectively deplete self-reactive B cells to prevent disease. These studies identify multiple checkpoints at which tolerance fails to circumvent autoreactivity in type 1 diabetes-prone mice and provide proof of concept for the development of a new strategic approach for the elimination of escaped autoreactive B cells.

This work was supported by National Institutes of Health grants 5T32-HL-069765, 5T32-AR-059039, R01-AI-051448, and R01-DK-08246, as well as Juvenile Diabetes Research Foundation Grant 1-2008-108.

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

R.A.H. designed and performed the experiments and wrote the manuscript. P.L.K. and J.W.T. provided experimental design and discussion input, as well as critical manuscript review. P.L.K. also assisted in the disease study experiment. J.W.T. 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.

Parts of this work were presented as a short talk in Workshop 2A at the Keystone Symposia B Cells: New Insights into Normal versus Dysregulated Function, Whistler, British Columbia, Canada, 12–17 April 2011.

The authors thank Amita Rachakonda, Brett Case, and Allison Sullivan for technical support (Vanderbilt University), as well as the Vanderbilt Medical Center Flow Cytometry Shared Resource (supported by the Vanderbilt Ingram Cancer Center [P30 CA68485] and the Vanderbilt Digestive Disease Research Center [DK058404]), the Vanderbilt Antibody and Protein Resource, the Vanderbilt Hormone Assay & Analytical Services Core, and the Vanderbilt Diabetes Research and Training Center. The authors also thank Eugene Oltz (Washington University, St. Louis, MO), Daniel Moore, Mark Boothby, Kristen Hoek, and Patrick Collins (Vanderbilt University) for critical manuscript review.