Reversing Tolerance in Isotype Switch–Competent Anti-Insulin B Lymphocytes

Tolerance for B lymphocytes in the developing repertoire is maintained first by receptor editing and clonal deletion in the bone marrow (1–3). Not all self-reactive B cells are removed by central tolerance, however, as BCRs with monovalent or weak interactions with autoantigens may avoid elimination or revision (4, 5). Autoreactive B cells that escape central tolerance and mature in the periphery are a liability, and additional mechanisms of tolerance are necessary to guard against autoimmunity (6–9). B cells that continuously encounter self-antigens may be rendered anergic or functionally silent to immune stimuli in the periphery. Tolerant B cells exhibit decreased surface IgM expression, impaired Ca²⁺ mobilization, restricted competition for available survival factors and follicular niches, and impaired responses to both T cell help and B cell mitogens (7, 10). Such anergic B cells are recognized in both normal and autoimmune repertoires (11–13).

The importance of BCR signaling for maintaining peripheral tolerance is emphasized by the reversal of anergy upon removal of chronic cognate Ag (10, 11). Alterations in BCR signaling pathways and mediators such as phosphoinositide 3-kinase, protein kinase Cθ, and the negative regulator protein tyrosine phosphatase non–receptor type 22 have been shown to impact both the induction and maintenance of tolerance (14–16). Innate signaling via TLR and MyD88 reverses anergy in some autoreactive B cells, suggesting that environmental factors that lead to infection and inflammation may also alter tolerance (17, 18). B cells deficient in MyD88 demonstrate impaired IgM responses to bacterial Ags, indicating that innate signaling through TLR pathways is critical for early T cell–independent (TI) immune defense (19). TLR-4 stimulation by LPS unlocks alternate signaling pathways to ERK phosphorylation and NF-κB activation independent of conventional BCR-dependent signaling mediators (20) that may be impaired for anergic B cells. Adaptive interactions with T cells may also drive loss of B cell tolerance and promote somatic hypermutation and Ig class switch recombination (CSR) in germinal center (GC) reactions associated with ongoing autoimmune disorders (21, 22). The fact that most pathogenic autoantibodies are of the IgG isotype further implicates T cells as potential vectors for driving loss of B cell tolerance. Thus, the overall effectiveness of immune tolerance for B lymphocytes depends on the nature of BCR interaction with autoantigens, potential encounter with innate signals in the environment, and availability of epitopes that promote cognate T–B interactions.

Insulin is a protein hormone whose small size and low circulating concentration were previously thought to limit BCR interactions necessary for tolerance (23, 24). Studies using a conventional IgM-restricted anti-insulin BCR transgene revealed that anti-insulin B cells enter the mature repertoire but are anergic and fail to produce anti-insulin Abs following T cell–dependent (TD) immunization (25). Such functionally silenced B cells residing in the periphery retain cellular functions such as Ag presentation that enable them to promote autoimmunity in NOD mice (22, 26). Insulin autoantibodies associated with autoimmune disorders such as type 1 diabetes, as well as Abs that arise in response to insulin therapy and complicate disease management, are predominantly of the IgG isotype (27–32). How the ability to undergo CSR contributes to the maintenance or loss of tolerance for anti-insulin B cells is not known. To assess peripheral tolerance in anti-insulin B cells competent to undergo somatic hypermutation and CSR, we generated C57BL/6 (B6) mice that harbor an anti-insulin H chain site directed to its native locus (V_H125^SD). V_H125^SD pairs with endogenous L chains to generate a polyclonal B cell repertoire, where physiologically relevant CSR-competent anti-insulin B cells successfully compete and make up a small fraction of the repertoire. V_H125^SD B6 mice crossed with anti-insulin Vκ125 transgenic (Tg) mice generates a monoclonal B cell repertoire in which >98% of the B cells bind insulin (V_H125^SD/Vκ125Tg), a model used for in vitro experiments. These two models are used to assess the fate and function of anti-insulin B cells that are competent to undergo isotype switch in either a monoclonal or polyclonal repertoire.

In this study, we examine how peripheral tolerance is governed for autoreactive B lymphocytes that bind the relevant autoantigen, insulin. CSR-competent anti-insulin B cells enter mature compartments but are anergic, demonstrated by impaired proliferation to stimulation by a panel of B cell mitogens in vitro and total lack of IgG anti-insulin Ab production following TD immunization of V_H125^SD B6 mice. Reversal of anti-insulin B cell anergy is demonstrated by proliferation to insulin plus LPS in vitro and IgG2a Ab production following immunization of V_H125^SD B6 mice with insulin conjugated to a type 1 TI Ag (24, 33). This combined BCR/TLR costimulation effect in vivo is accompanied by entry of insulin-binding B cells into GCs, where anti-insulin L chains are not discarded but rather selected from the preimmune repertoire. These studies reveal a new pathway to drive loss of tolerance for CSR-competent anti-insulin B cells.

pIV_HL2neoR is a vector designed for targeted insertion of a rearranged V_H gene replacing J_H loci, and was a gift from Dr. Klaus Rajewsky (34). We first modified pIV_HL2neoR by cloning in two regions of short arm homology (SAH) at ClaI and NotI sites. This was necessary after initial efforts of homologous recombination were poor. Anti-insulin VDJ_H-125 from our original H chain plus L chain 125Tg mice (25) was subcloned into the pGEM-5Z vector and then cloned into modified pIV_HL2neoR-SAH at SalI and ClaI sites to generate our targeting vector, pIV_H-SAH-125-VDJ_H. All sequences were verified throughout the cloning process. pIV_H-SAH-125-VDJ_H was linearized through digestion with NotI and electroporated into 129/Ola embryonic stem cells. The electroporated cells were selected for neomycin resistance. Embryonic stem cell clone DNA was digested with HindIII, and Southern blot hybridization using a cDNA probe against an Xba1 enhancer located upstream of V_H125^SD detected the proper 6.1-kb fragment size of the construct in the targeted locus. Clones were confirmed by PCR using H chain primers 5′-CAG ATC CAG TTG GTG CAG TC-3′ (forward) and 5′-CCA GAC ATC GAA GTA CCC CT-3′ (reverse). Positive embryonic stem cell clones were injected into blastocysts and transplanted into pseudopregnant female mice. Tail DNA from founder chimeric pups and their progeny were screened for V_H125^SD. Southern blot was used to confirm the presence of the targeted allele.

V_H125^SD mice were backcrossed onto the B6 background. B6 mice and MD4 anti–hen egg lysozyme (HEL) mice (B6-Tg[IghelMD4]4Ccg/J) were purchased from The Jackson Laboratory (Bar Harbor, ME). EIIA-Cre B6 mice, provided by Dr. Richard Breyer (Vanderbilt University), were intercrossed with V_H125^SD mice to permanently remove the neomycin resistance cassette. Transgene transmission was monitored in subsequent generations of site-directed V_H125^SD B6 mice using PCR, and flow cytometry confirmed the presence of >95% transgenic IgM^a+ B cells. Male and female mice aged 7–14 wk were used in these studies and backcrossed >10 times to B6 mice. All animals were housed in specific pathogen-free conditions, and all studies were approved by the American Association for the Accreditation of Laboratory Animal Care–certified Vanderbilt Institute of Animal Care and Use Committee.

Abs reactive to murine B220 (RA3-6B2), IgM^a (DS-1), IgM^b (AF6-78), CD21 (7G6), CD23 (B3B4), IgD (11–26), GL7, or Fas/CD95 (Jo2) (BD Pharmingen), or IgM (μ-chain specific, Invitrogen) were used, along with biotinylated human insulin (35) or biotinylated anti-insulin mAb 123 (36) with streptavidin-conjugated fluorochrome to detect insulin-binding B cells, and 7-aminoactinomycin D for cell viability. Flow cytometry acquisition was performed after cell suspension in FACS buffer (1× PBS, 10% FBS, 1% sodium azide, 1% EDTA) using an LSR II (BD Biosciences), and sorting experiments were performed using a FACSAria I or II cell sorter in the Vanderbilt University Shared Resource Facility. Analysis was performed using FlowJo software (Tree Star).

B cells were purified from whole spleens using negative selection with anti-CD43 magnetic beads (MACS, Miltenyi Biotec). Average B cell purity across all experiments performed was: V_H125^SD/Vκ125Tg, 91.4%, n = 9; MD4 anti-HEL, 91%, n = 8; wild-type (WT), 92.5%, n = 3. Cells were plated at 2 × 10⁵ per well in complete RPMI (10% FBS, 1% HEPES, 1% l-glutamine, 1% gentamicin) and cultured for 72 h in a 37°C, 5% CO₂ incubator. Cells were pulsed with [³H]thymidine deoxyribose (NEN/PerkinElmer) on day 2 and harvested using a semiautomated cell harvester (Skatron) on day 3. [³H]thymidine deoxyribose incorporation was measured by scintillation counting, and results are expressed as the mean cpm ± SD for the indicated number of mice in each experimental group. For in vivo proliferation experiments, both unimmunized and insulin–Brucella abortus ring test Ag (BRT)-immunized mice were injected i.p. with BrdU per the manufacturer’s instructions (BD Pharmingen) 48 and 24 h before sacrifice. FITC-conjugated anti-BrdU (BD Pharmingen) was used to detect intracellular incorporation of BrdU.

For TD assays, preimmune sera were collected and mice were immunized with 40–50 μg bovine insulin in 1× PBS emulsified in either CFA s.c. at the base of the tail or in IFA i.p., and sera were collected 2–3 wk later. For TI assays, mice were immunized i.p. with human insulin–conjugated insulin-BRT (24), with BRT alone, or with physically mixed BRT and insulin, and sera were collected at 5–7 d following immunization. Anti-insulin Abs were measured by ELISA as follows: 96-well flat-bottom Nunc plates were coated with human insulin in borate-buffered saline overnight at 37°C. Plates were extensively washed with 1× PBS-Tween 20 (0.05%). Diluted sera (1:100) were added to the coated plates with or without 100× soluble human insulin to measure specific, inhibitable anti-insulin Abs. IgM and IgG anti-insulin Abs were detected using the following allotype-specific secondary Abs: IgM^a-biotin (DS-1), IgG1^a-biotin (10.9), or IgG2a^a-biotin (8.3) for transgenic anti-insulin B cells, or IgM^b-biotin (AF6-78), IgG1^b-biotin (B68-2), or IgG2c^b-biotin (5.7) (BD Pharmingen) for nontransgenic or WT B cells. Avidin–alkaline phosphatase (Sigma-Aldrich) or goat anti-mouse IgM–alkaline phosphatase or IgG–alkaline phosphatase Abs (SouthernBiotech) were added. After washing, p-nitrophenyl phosphate substrate (Sigma-Aldrich) was added to the plate and immediately read on a Microplate Autoreader (Bio-Tek Instruments) at an OD of 405 nm. Inhibitable (insulin-specific) binding was determined by the difference in binding in the presence or absence of excess competitive insulin.

Spleens were removed from unimmunized or insulin-BRT–immunized V_H125^SD B6 mice and soaked in 30% sucrose (w/v) overnight at 4°C and snap frozen in OCT medium on dry ice. Eight-micrometer sections were obtained from the Vanderbilt Translational Pathology Shared Resource. Sections were fixed with 4% paraformaldehyde in 0.1 M PBS and stained with the following Ab cocktails: IgM Alexa Fluor 488 (Life Technologies), IgD^a-biotin (AMS 9.1, BD Pharmingen) plus streptavidin Alexa Fluor 350 (Life Technologies), and CD3 PE (17A2, BD Pharmingen) to define follicular architecture, or GL7 FITC, IgD^a-biotin plus streptavidin Texas Red (Molecular Probes), and DAPI (Molecular Probes) to detect GCs. Images were obtained using an Olympus BX60 fluorescent microscope with MagnaFire software. Adobe Photoshop was used to adjust brightness and contrast and to overlay images.

Pancreata were dissected from unimmunized or insulin-BRT–immunized V_H125^SD B6 mice, or from 12- to 16-wk-old female NOD mice (as a positive control for insulitis assessment) and placed into neutral buffered formalin. After 4–6 h, tissues were transferred to 70% ethanol and incubated overnight. Tissues were subsequently paraffin embedded, 5-μm sections were cut, and slides were stained with H&E by the Vanderbilt Translational Pathology Shared Resource. Slides were blind scored for insulitis by light microscopy using the following scale: 0, no insulitis; 1, peri-insulitis, <25% islet infiltration; 2, 25–50% islet infiltration; 3, 50–75% islet infiltration; 4, >75% islet infiltration. Half of the pancreas was fixed with paraformaldehyde in 0.1 M PBS and soaked in 30% sucrose overnight at 4°C, and then snap frozen in OCT medium on dry ice. Eight-micrometer sections were stained for immunofluorescence with DAPI and biotinylated anti-IgG2a^a plus FITC-conjugated streptavidin.

Spleens were harvested from unimmunized V_H125^SD B6 mice or from mice 4 d following insulin-BRT immunization. Human insulin-binding B cells (B220⁺ live lymphocytes) were identified by FACS and further gated as IgM⁺ non-GC (GL7⁻Fas⁻), IgM⁺ GC (GL7⁺Fas⁺), or IgM⁻ GC populations. Cells were sorted directly into RNAqueous lysis buffer (Life Technologies), and RNA was isolated using the RNAqueous-micro total RNA isolation kit (Life Technologies) per the manufacturer’s instructions. Independent clones were derived from separate tubes of lysate. RNA was reverse transcribed into cDNA using SuperScript II RT (Invitrogen) and oligo(dT) primer (GE Healthcare) in a standard protocol. Resulting cDNA was used as a PCR template to amplify Igκ genes using a murine Vκ primer, 5′-ATT GTK MTS ACM CAR TCT CCA-3′, and murine Cκ primer, 5′-GGA TAC AGT TGG TGC AGC ATC-3′, where K = G or T, M = A or C, S = C or G, and R = A or G with a 44°C annealing temperature for 35 cycles. Appropriately sized PCR products were gel purified using the MinElute gel extraction kit (Qiagen) and Igκ was cloned as described previously (37, 38). Vκ and Jκ gene segment identities were assigned using Ig BLAST (http://www.ncbi.nlm.nih.gov/igblast/) and IMGT nomenclature with CDR boundaries defined by the KABAT V domain delineation system. Nucleotide mutation analyses were also performed using the Ig BLAST tool.

Previous μ-only BCR transgenic mouse models do not address how native cellular functions such as CSR impact the state of immune tolerance for anti-insulin B cells. To examine peripheral tolerance for anti-insulin B cells competent to undergo isotype switch, B6 mice that harbor anti-insulin VDJ_H-125 site directed to the IgH locus were developed as described in 2Materials and Methods. Fig. 1A summarizes the strategy for generating the targeting construct used to develop site-directed V_H125^SD B6 mice. Southern blot, PCR, and flow cytometry confirmed successful IgH locus targeting, as B220⁺ splenic B cells from V_H125^SD B6 mice coexpress IgM^a and IgD^a whereas those from conventional μ-only V_H125Tg B6 mice express only IgM^a as expected (Fig. 1B).

Flow cytometry with biotinylated insulin was used to investigate the contribution of different anti-insulin transgenes to insulin-binding B cells (Fig. 1C). In Fig. 1C (left), mice that possess only an anti-insulin L chain (Vκ125Tg) show low numbers of insulin-binding B cells in the spleen (0.12 ± 0.01%, n = 3), but these weakly binding cells are not specific, as they are not inhibited with excess competitive insulin (not shown). In Fig. 1C (center), V_H125^SD B6 mice have B cells in which the targeted transgene pairs with endogenous L chains to generate a small population of insulin-binding B cells (0.46 ± 0.05%, n = 9). In Fig. 1C (right), intercrosses that pair V_H125^SD with Vκ125Tg confer insulin-binding specificity to >98% of the B cells. Unlike other site-directed mouse models, such as the SW_HEL model (39), V_H gene revision or editing is not prevalent for IgM⁺ B cells in V_H125^SD, as >95% are [a] allotype and retain their insulin-binding potential. These data demonstrate a novel site-directed BCR transgenic model can be used to assess the fate and function of CSR-competent anti-insulin B lymphocytes within a monoclonal or polyclonal repertoire.

To determine whether CSR-competent anti-insulin B cells are anergic, B cells were purified from V_H125^SD/Vκ125Tg or WT B6 mice using MACS and tritiated thymidine incorporation was used to assess B cell proliferation in vitro to anti-IgM, anti-CD40, or LPS stimulation (Fig. 2). Anti-insulin B cells exhibit impaired proliferative responses to stimulation through BCR, CD40, and TLR-4, compared with B cells from WT B6 mice (Fig. 2A). Thus, anergy is maintained for CSR-competent IgM^a+IgD^a+ anti-insulin B cells in vitro, similar to that reported for anti-insulin B cells in conventional μ-only 125Tg mice (H plus L chain) (26).

Different costimulation pathways were assessed for their impact on proliferation in the context of BCR encounter with autoantigen. B cells were purified from V_H125^SD/Vκ125Tg B6 mice or naive anti-HEL MD4-Tg mice (10) and stimulated with anti-CD40 (Fig. 2B) or LPS (Fig. 2D) in the presence or absence of cognate Ag. Using B cells specific for HEL as naive control, stimulation with HEL plus anti-CD40 at low concentration (0.1 μg/ml) synergized to significantly augment proliferation (30-fold increase, Fig. 2C). In contrast, stimulation of anti-insulin B cells with insulin plus anti-CD40 did not enhance their proliferation, but instead blunted the response (0.52-fold, Fig. 2C). Synergy is defined in this study as the mathematical value representing the effect of B cell costimulation relative to the effect of the sum of each individual stimulus. Failure to demonstrate synergy between insulin and anti-CD40 was consistent across a wide range of Ag and Ab concentrations (data not shown). These data suggest that anergy is maintained for CSR-competent anti-insulin B cells costimulated with cognate Ag and anti-CD40, which mimics T cell help.

Using naive anti-HEL B cells as control, costimulation with HEL and LPS at low concentration (0.1 μg/ml) only marginally boosted proliferation, with no evidence of synergy (1.2-fold, Fig. 2E). In striking contrast, costimulation of anti-insulin B cells with insulin and LPS synergized to enhance B cell proliferation (10-fold, Fig. 2E). These data demonstrate that in vitro responses of anti-insulin B cells are altered by prior exposure of the monoclonal repertoire to Ag in vivo. Anti-insulin B cells are anergic, as they fail to proliferate to anti-CD40 with or without Ag. However, this anergy is readily reversed by simultaneous engagement of BCR and TLR-4 in vitro.

We sought to understand how the features of tolerance observed in a nearly monoclonal population of anti-insulin B cells apply to a polyclonal repertoire that contains relatively few anti-insulin B cells (V_H125^SD). Factors that govern anergy for autoreactive B lymphocytes in a polyclonal repertoire include competition for survival and entry into mature compartmental niches (6, 26, 40, 41). To assess whether CSR-competent anti-insulin B cells in the polyclonal repertoire of V_H125^SD B6 mice enter mature subsets, flow cytometry was used to identify B cells (B220⁺IgM^a+ live lymphocytes) that distributed into transitional 1 (T1, CD21⁻CD23⁻), follicular (FO, CD21^intCD23^high), and marginal zone (MZ, CD21^highCD23^low) compartments of the spleen (Fig. 3). In contrast to WT B6 mice, a small population of anti-insulin B cells is observed in V_H125^SD B6 mice (0.42 ± 0.05%, n = 9), and their specificity is confirmed by inhibition with unlabeled competitive insulin (Fig. 3A, left panels). Rare insulin-binding B cells in WT B6 mice (0.03 ± 0.01%, n = 3) are not inhibited by excess insulin (Fig. 3A, right panels). The ability of B cells to enter T1, FO, and MZ subsets was compared for non–insulin-binding (ins⁻) or insulin-binding B cells (ins⁺) in V_H125^SD B6 mice, or for WT B cells (Fig. 3B). As represented in Fig. 3C, ins⁺ B cells predominantly populate the FO subset in V_H125^SD B6 mice, whereas fewer ins⁺ B cells are found in other subsets (8.1 ± 3.2% T1, 78.9 ± 5.5% FO, 2.7 ± 1.9% MZ, n = 7). The reduced representation of ins⁺ B cells in the MZ contrasts ins⁻ B cells (12.25 ± 2.6% T1, 57.7 ± 3.5% FO, 16.5 ± 1.7% MZ) in V_H125^SD B6 mice and normal B cells (8.4 ± 1.7% T1, 78.3 ± 3.7% FO, 6.0 ± 0.9% MZ, n = 6) in WT B6 mice. These data suggest that anti-insulin B cells in the peripheral repertoire of V_H125^SD B6 mice are principally mature FO B cells.

To determine whether anti-insulin B cells are clonally ignorant or whether their BCRs have encountered endogenous rodent insulin in vivo, a second anti-insulin mAb (mAb123) was used. mAb123 recognizes a different insulin epitope and binds insulin-occupied BCRs (25, 35). Flow cytometry staining with mAb123 confirms that BCRs are occupied by endogenous insulin in V_H125^SD B6 mice (0.4 ± 0.1%, Fig. 3D). These data demonstrate that CSR-competent anti-insulin B cells in V_H125^SD B6 mice successfully compete in a polyclonal repertoire and populate FO compartments despite encounter with physiologic autoantigen during development and early peripheral maturation.

Prior studies in mice harboring IgM^a-restricted anti-insulin B cells demonstrated failure to respond to TD immunization (25). To dissect the functional status of CSR-competent IgM^a+IgD^a+ anti-insulin B cells in V_H125^SD B6 mice, Ab responses were assessed following two different immunization strategies, and serum IgM and IgG anti-insulin Abs were measured by ELISA (Fig. 4). For TD responses, mice were immunized with bovine insulin emulsified in CFA (insulin/CFA). Bovine insulin was chosen for TD immunization because the MHC class II of B6 mice (I-A^b) is genetically a strong responder to bovine but not to other insulins, notably human (24). Conventional WT B6 mice generated a strong IgG1^b anti-insulin response following insulin/CFA immunization, whereas CSR-competent anti-insulin B cells in V_H125^SD B6 mice failed to produce IgG1^a anti-insulin Abs (Fig. 4A). To ensure that the absence of response to insulin/CFA immunization in these mice is not attributed to lack of bovine insulin processing by anti-insulin BCRs, we used competitive inhibition in flow cytometry to confirm that bovine insulin at concentrations well below those used in immunization fully competes with human insulin for binding to anti-insulin B cells in V_H125^SD B6 mice (Supplemental Fig. 1). IgG1 was the predominant isotype observed following insulin/CFA immunization, a finding that agrees with previous work that examined the Ig response to insulin in B6 mice (24). To test whether TLR signaling provided by mycobacteria in CFA is necessary, mice were also immunized with bovine insulin emulsified in IFA (insulin/IFA). Whereas the IgG1^a anti-insulin response remained absent in V_H125^SD B6 mice, WT B6 mice generated IgG1^b anti-insulin Abs following insulin/IFA immunization (data not shown). This agrees with prior studies that demonstrated B cells deficient in both MyD88 and TRIF were able to mount Ab responses to multiple Ags administered in a variety of adjuvants, including CFA and IFA (42). These data validate that tolerance for anti-insulin B cells is maintained for TD immune responses to heterologous insulin in either the presence or absence of mycobacterial adjuvant.

To assess anti-insulin B cell competence to produce Abs in the absence of cognate T cell help, V_H125^SD B6 mice were immunized with human insulin conjugated to BRT or with BRT alone (24). Human insulin is used in BRT conjugates to avoid introduction of T cell epitopes. Prior studies showed that insulin-BRT behaves as a typical type 1 TI Ag, including response kinetics, expected Ab isotypes, and responses observed in both athymic and X-linked immunodeficient mice (33, 43). Following immunization with insulin-BRT, B cells from WT B6 mice and CSR-competent anti-insulin B cells from V_H125^SD B6 mice produced IgM^b and IgG2c^b or IgM^a and IgG2a^a anti-insulin Abs, respectively (Fig. 4B). Immunization with BRT alone, however, failed to induce any detectable IgG anti-insulin Abs in either V_H125^SD or WT B6 mice (Fig. 4B). Similarly, immunization with physically mixed BRT and insulin failed to induce any response (data not shown). Insulin-BRT immunization did not elicit IgG1 Abs, consistent with the expected isotypes associated with BRT conjugates (33, 43). These data suggest that the insulin-BRT conjugate promotes loss of tolerance for CSR-competent anti-insulin B cells in vivo through a combination of BCR and TLR signaling that drives production of IgM and IgG anti-insulin Abs.

To determine the pathological consequences of the breach in tolerance observed in V_H125^SD B6 mice following insulin-BRT immunization, pancreata were examined for insulitis and the presence of IgG anti-insulin Abs at day 12 of the response. A scoring system of 0–4 was used to assess insulitis by H&E staining in pancreata sections from unimmunized or insulin-BRT–immunized mice (outlined in 2Materials and Methods). NOD mouse pancreata sections served as a positive control for islet infiltration. No insulitis was detected in 60 islets examined in two unimmunized V_H125^SD B6 mice or in 82 islets examined in four immunized mice, such that 100% of islets examined received a score of 0. In contrast, 64 islets were scored in two 16- to 20-wk-old female, nondiabetic NOD mice, with 53% islets scoring 0, 14% islets scoring 1, 9% islets scoring 2, 8% islets scoring 3, and 16% islets scoring 4. Endogenous Ig deposition in the islets was assessed by direct immunofluorescence. Ig deposition was not detected in pancreata of either unimmunized or insulin-BRT–immunized mice (data not shown). Ig was observed by indirect immunofluorescence, confirming that Abs produced following insulin-BRT immunization are autoreactive. These data indicate that the breach in tolerance driven by insulin-BRT does not elicit a detectable organ-specific autoimmune attack.

To assess the fate of anti-insulin B cells undergoing an active breach of tolerance, clonal expansion and proliferation were measured by BrdU incorporation in vivo following insulin-BRT immunization. IgM⁺ insulin-binding B cells (IgM^a+ins⁺) in V_H125^SD B6 mice immunized with BRT alone (0.33 ± 0.07%, n = 7) were not increased compared with unimmunized mice (0.34 ± 0.08%, n = 6) (Fig. 5A). In contrast, IgM⁺ insulin-binding B cells were expanded in mice immunized with insulin-BRT (2.97 ± 1.28%, n = 9) (Fig. 5A). A small population of IgM⁻ insulin-binding B cells (IgM^a−ins⁺) that have likely undergone CSR were detected in V_H125^SD B6 mice immunized with insulin-BRT (0.79 ± 0.46%, n = 9) but not in unimmunized mice or in mice immunized with BRT alone. BrdU incorporation was used concomitantly to demonstrate that the increase in anti-insulin B cells represented Ag-specific expansion (Fig. 5B). Insulin-binding B cells in V_H125^SD B6 mice immunized with BRT alone did not incorporate BrdU and were similar in frequency to those in unimmunized mice. In contrast, both IgM⁺ and IgM⁻ insulin-binding B cell populations in mice immunized with insulin-BRT incorporated BrdU (39.8 ± 7.2% for IgM⁺ and 40.5 ± 14.1% for IgM⁻, n = 5). These findings correlate with the Ab responses observed in V_H125^SD B6 mice following immunization and show that the insulin-BRT conjugate drives clonal expansion and proliferation of anti-insulin B cells in vivo.

One hallmark of anergy is reduced surface IgM expression (6, 7, 10). Surface IgM for insulin-binding B cells (ins⁺) in unimmunized V_H125^SD B6 mice is reduced relative to non-insulin binders (ins⁻), expressed as a ratio of mean fluorescence intensity (MFI) of ins⁺ to ins⁻ (0.58 ± 0.12, n = 6) (Fig. 5C). To test whether insulin-BRT immunization promotes restoration of normal surface IgM expression, the MFI of surface IgM was compared for ins⁺ and ins⁻ B cells following immunization (Fig. 5C). Surface IgM for ins⁺ B cells was not restored in V_H125^SD B6 mice immunized with BRT alone (0.64 ± 0.08, n = 5). In contrast, surface IgM for ins⁺ B cells was completely restored following insulin-BRT immunization (1.04 ± 0.09, n = 6). Surface IgM restoration is consistent with reversal of anergy in CSR-competent anti-insulin B cells.

Ig isotype switch in B cells is principally recognized to occur in GC reactions that arise in the follicle following cognate interactions between CD4⁺ Th cells and Ag-specific B cells (44–46). Most rapid TI B cell responses are expected to occur in the extrafollicular spaces of the spleen (47); however, it has been demonstrated that some TI B cell responses can promote the unconventional formation of GCs in the absence of T cell help (48, 49). Accordingly, we assessed whether the observed breach in peripheral tolerance for anti-insulin B cells in the insulin-BRT response extends to the formation of GCs. Flow cytometry was used to assess expression of the GC markers, GL7 and Fas, on anti-insulin B cells in V_H125^SD B6 mice 5 d following immunization. IgM⁺ insulin-binding B cells (IgM^a+ins⁺) in unimmunized mice established the background for the GC phenotype in the spleen (0.01% GL7⁺Fas⁺, n = 5), and immunization with BRT alone did not increase this response (0.03% GL7⁺Fas⁺, n = 7) (Fig. 6A). In striking contrast, IgM⁺ insulin-binding B cells in mice immunized with insulin-BRT acquired the GC phenotype (24.4 ± 19.0% GL7⁺Fas⁺, n = 9) (Fig. 6A). An increased frequency of IgM⁻ insulin-binding B cells (IgM^a−ins⁺) acquired the GC phenotype following insulin-BRT immunization (34.2 ± 27.6% GL7⁺Fas⁺, n = 9), demonstrating that a portion of ins⁺ B cells participating in GC reactions have undergone CSR, a finding consistent with production of IgG anti-insulin Abs. Examination of these GC reactions at a later time point in the insulin-BRT response (day 12) yielded a sporadic population of ins⁺ GC B cells. Three of eight mice examined at day 12 had detectable ins⁺ GC B cells (5.22 ± 0.08%), whereas the other five mice did not have a detectable population of the same B cells (0.20 ± 0.13%). Rapid decline in insulin-specific GC B cells and Ab production at day 12 following insulin-BRT (data not shown) is consistent with previous work in other models describing the fate of Ag-specific GCs in the absence of cognate T cell help (48). These data support the concept that the insulin-BRT conjugate drives IgG anti-insulin Ab production largely through generation of Ag-specific GCs that are short-lived.

Notably, non–insulin-binding B cells (IgM^a+ins⁻) in V_H125^SD B6 mice did not upregulate GC markers following insulin-BRT immunization (0.39 ± 0.19% GL7⁺Fas⁺) when compared with unimmunized mice (0.36 ± 0.46% GL7⁺Fas⁺) (Fig. 6A, top row), suggesting that acquisition of the insulin-specific GC B cell phenotype is limited to cognate anti-insulin B cells and not merely a consequence of global B cell activation from BRT-mediated TLR stimulation. Flow cytometry staining confirmed that GL7⁺Fas⁺ GC B cells were IgD⁻ (Fig. 6B), consistent with induction of a GC B cell phenotype.

To confirm formation of anatomical GCs following insulin-BRT, immunofluorescence staining was performed on spleen sections from both unimmunized and immunized V_H125^SD B6 mice. IgM, IgD, and CD3 were used to define FO architecture, and GCs were identified as IgD⁻ and GL7⁺. Whereas some GL7⁺IgD⁻ GCs were observed in unimmunized mice and in mice immunized with BRT alone, they were small by microscopy and non–insulin-specific by flow cytometry (Fig. 6A, 6C). In contrast, large GL7⁺IgD⁻ GCs were readily detected in spleen sections from mice immunized with insulin-BRT (Fig. 6C). These data show that anti-insulin B cells acquire a GC phenotype and localize to GC structures in the spleen following immunization with insulin-BRT.

Anti-insulin B cells are detected in GCs following insulin-BRT immunization (Fig. 6). These GC B cells may reflect expansion of insulin-binding B cells that were present in the preimmune repertoire (Fig. 3D). Alternatively, the unique environment of GC reactions may select rare anti-insulin B cells that are clonally ignorant or below the level of detection. Accordingly, we investigated selection of anti-insulin Igκ from the preimmune repertoire by insulin-binding GC B cells. Spleens were harvested from unimmunized V_H125^SD B6 mice or from mice immunized with insulin-BRT. Non-GC (IgM⁺GL7⁻Fas⁻), IgM⁺ GC (IgM⁺GL7⁺Fas⁺), or IgM⁻ GC (IgM⁻GL7⁺Fas⁺) insulin-binding B cells were purified by flow cytometry sorting as identified in Figs. 5A and 6A. RNA was purified, and cDNA was used as template for Igκ amplification. The Ig BLAST tool was used to identify Vκ and Jκ gene segment usage (using IMGT nomenclature) as well as any nucleotide mutations (see 2Materials and Methods).

The Vκ125Tg in the previously published anti-insulin 125Tg (H plus L chain) mouse model is a Vκ4-74 (25, 50). This L chain is associated with anti-insulin B cells loaded with endogenous rodent insulin detected by mAb123 in V_H125^SD B6 mice (Fig. 3D and data not shown), confirming that this Vκ is autoreactive when combined with V_H125. Insulin-binding B cells sorted in unimmunized mice exclusively used Vκ4-74 (Fig. 7A). Vκ4-74 was also the dominant Vκ used by insulin-binding B cells following insulin-BRT immunization, including 11 of 12 non-GC (Fig. 7B), 8 of 9 IgM⁺ GC, and 5 of 6 IgM⁻ GC clones (Fig. 7C, 7D). Vκ4-74 clones were paired with all Jκ, forming different CDR3 amino acid junctions associated with previously recognized low/moderate (P-L) and high (P-P) affinity for rodent insulin, as deduced from sequences identified in studies of anti-insulin hybridomas (35). The 5′ degenerate primer sequence was removed, and Igκ sequences were deposited into GenBank (http://www.ncbi.nlm.nih.gov/genbank/) with the following accession numbers: unimmunized (KP790058–KP790071), non-GC (KP790072–KP790083), IgM⁺ GC (KP790084–KP790092), and IgM⁻ GC (KP790093-KP790098).

Most (five of nine) IgM⁺ GC anti-insulin B cell clones sorted from V_H125^SD B6 mice on day 5 following insulin-BRT retained nucleotide sequences in germline Igκ configuration with CDRs identical to those found in unimmunized mice (Fig. 7G). However, an increased frequency of both IgM⁺ and IgM⁻ GC anti-insulin B cell clones sorted from immunized mice possessed one or more nucleotide mutations relative to non-GC anti-insulin B cells from immunized mice or anti-insulin B cells isolated from unimmunized mice (Fig. 7E–H). Mutations were randomly yet equally distributed in framework regions and CDRs, although no amino acid changes were observed (data not shown). Taken together, these data suggest that anti-insulin B cells enter GCs from the preimmune repertoire following immunization with the insulin-BRT conjugate, and a proportion of B cells undergo somatic hypermutation in these nonconventional GC reactions.

Autoreactive B lymphocytes are recognized to enter the peripheral repertoire of normal individuals, and their presence is considered a liability for developing autoimmune disease. The association of IgG autoantibodies with pathological processes suggests that self-reactive B cells are recruited into immune responses. How this occurs is not clear, although a general assumption is that T cell–mediated help is required to reverse tolerance in such B cells. The present studies identify an alternative mechanism by which autoreactive B cells may escape peripheral tolerance without cognate T cell help. We describe the V_H125^SD B6 mouse, a new site-directed transgenic model that has a polyclonal B cell repertoire with a detectable population of IgM⁺IgD⁺ anti-insulin B cells that have the ability to undergo isotype switch. Anti-insulin B cells in V_H125^SD B6 mice are not eliminated but rather rendered clonally anergic in the periphery, unlike other models in which high-affinity Ag-specific B cells are eliminated because they fail to compete in the repertoire for residence in mature compartments (51–53). Tolerance is maintained in CSR-competent anti-insulin B cells following TD immunization, as immunized V_H125^SD B6 mice do not produce IgG anti-insulin Abs (Fig. 4A). Immunization of V_H125^SD B6 mice with insulin conjugated to a type 1 TI Ag (insulin-BRT), however, provides BCR/TLR costimulation that reverses anergy in anti-insulin B cells and promotes proliferation, differentiation into GCs, and IgG autoantibody production.

Insulin-BRT was previously used to uncover rare anti-insulin B cells in the repertoires of normal mice independently of T cell help (33). As originally observed, conjugation of insulin and BRT was necessary to drive Ab production (33). As such, immunization of V_H125^SD B6 mice with insulin-BRT elicits IgM and IgG2a anti-insulin Abs, whereas neither BRT alone (Fig. 4B) nor BRT physically mixed with insulin (data not shown) elicits Ab production. The kinetics of the insulin-BRT response (3–5 d) and Ab isotypes observed following immunization are in agreement with those seen in response to classic type 1 TI Ags. Whereas insulin-BRT responses have been shown to occur in the absence of T cells (33, 43), they are likely influenced by bystander effects or other non-specific T cell factors, as demonstrated for other TI Ags (54).

Other examples of B cell tolerance reversal independent of T cell help have been described. B cells specific for MHC class I in 3-83 transgenic mice develop normally in the bone marrow but undergo clonal deletion in the periphery when they encounter liver-expressed H-2K^b (55, 56). However, administration of a bacteriophage containing a 15-aa mimotope that is recognized by 3-83 B cells reverses tolerance and drives robust Ig production (55). Thus, tolerant B cells that are normally deleted in the periphery are actually rescued when self-antigen is presented to the B cell in a TI fashion. Similarly, anergy in anti-insulin B cells is reversed by BCR/TLR costimulation in the insulin-BRT response that is relatively TI.

Insulin-specific GCs are found in V_H125^SD B6 mice at day 5 of the insulin-BRT response (Fig. 6), consistent with previous studies that demonstrated B cells specific for the hapten nitrophenol (NP) can enter GCs without T cell help following immunization with the type 2 TI immunogen NP-Ficoll (48). NP-Ficoll immunization only elicited GCs in QM transgenic mice, which have ∼60% B cells specific for NP, but not in nontransgenic (WT) littermates. de Vinuesa et al. (48) used transfer experiments to confirm that TI GCs only arose when the NP-specific B cell precursor frequency was 1 in 1000 or higher. V_H125^SD B6 mice possess ∼0.4% anti-insulin B cells in the periphery (1 in 250), whereas the frequency in WT B6 is closer to 1 in 100,000 (Fig. 3). As such, we did not observe any detectable insulin-specific GCs in WT mice following insulin-BRT immunization (data not shown), despite IgM and IgG anti-insulin Ab production in the same mice (Fig. 4B). The increased frequency of anti-insulin B cell precursors in V_H125^SD B6 mice supports the generation of insulin-specific GCs by day 5 and could explain the increased IgM anti-insulin Ab production observed in V_H125^SD compared with WT B6 mice. Insulin-specific GCs are found in V_H125^SD B6 mice at day 12 but at a much lower frequency, as expected for GCs over time in the absence of T cell help. We are currently introducing V_H125^SD into NOD mice to determine how these unconventional GC reactions are controlled in an autoimmune strain of mice predisposed to type 1 diabetes.

Igκ genes from the preimmune repertoire do not undergo negative selection in GCs, as evidenced by exclusive anti-insulin Vκ4-74 usage by BCRs in anti-insulin B cells from unimmunized V_H125^SD B6 mice and in IgM⁺ and IgM⁻ insulin-binding GC B cells from mice following insulin-BRT immunization (Fig. 7). Prior studies show that Vκ4-74 pairs with several Jκ to form autoreactive insulin-binding BCRs when combined with V_H125 (35). Thus, insulin-binding B cells from the preimmune repertoire, and not obscure clonally ignorant B cells, are selected for GC entry during the insulin-BRT response.

The presence of nucleotide mutations in Vκ4-74 clones indicates that somatic hypermutation occurs in these TI GCs. Approximately half of the mutations observed in both IgM⁺ and IgM⁻ insulin-binding GC clones were found in CDRs, and although amino acid replacements were not observed, the findings suggest that Ag binding could be altered in these atypical GCs. Our data demonstrate that reversal of anergy in anti-insulin B cells provided by BCR/TLR costimulation in vivo extends to evasion of a tolerance program normally designed to eliminate such autoreactive clones in GCs. Although insulin-BRT may not mimic any natural molecule, loss of B cell tolerance via similar pathways may be possible in circumstances where the levels of autoantigen and innate stimulus are high, such as on the surface of APCs. This could occur in association with islet inflammation in type 1 diabetes where APCs are found in close contact with insulin-producing β cells (57) or following uptake of insulin injected through contaminated skin by epidermal and dermal APCs (58). Recently, TLR-7 signaling was implicated in the development of spontaneous GCs and autoantibodies in a systemic lupus erythematosus model (59), in addition to the previously recognized role of TLR-9 in autoimmune disease (60). Furthermore, BCR/TLR synergy has been shown to induce TI CSR through the noncanonical NF-κB pathway (61). Thus, there is growing evidence to support a unique role for TLRs in the generation of GCs and production of IgG autoantibodies.

We examined the capacity of various TLR mitogens, such as ssRNA and CpG DNA, to synergize with insulin in vitro to reverse anergy in anti-insulin B cells and found that LPS was the only TLR agonist that enhanced proliferation. BRT contains noncanonical LPS (62–64), and thus we expect that the proliferation induced by insulin-BRT in vivo (Fig. 5B) occurs in part by TLR-4 stimulation. These data implicate TLR-4 in the reversal of anergy in anti-insulin B cells, although other TLRs are clearly important for tolerance to different autoantigens. The fact that this phenotype occurred at very low stimulating concentrations (Fig. 2E) suggests that signaling thresholds for autoreactive B cells are quite low, and that loss of anergy may arise more quickly and easily than previously thought.

IgG anti-insulin Abs that arise following insulin-BRT immunization were confirmed to be autoreactive by indirect immunofluorescence, as they bound mouse insulin on pancreata sections from both unimmunized and immunized mice. These same autoantibodies were not detected endogenously on the tissue using direct immunofluorescence, suggesting that a lack of access of autoantibodies to their tissue targets may limit their autoimmune potential. For example, anti–H-2K^b Abs generated by phage-mimotope immunization were found on the surfaces of hepatocytes in 3-83 Tg mice (55), reflecting the ubiquitous surface expression of MHC class I in the target organ. Access to insulin epitopes may be more restricted in the islets. Experiments in V_H125^SD NOD mice will determine whether reversal of anergy in anti-insulin B cells driven by BCR/TLR costimulation accelerates immunopathology.

Type 1 diabetes in mice and humans is a multigenic autoimmune disorder in which detection of IgG insulin autoantibodies indicates that loss of tolerance in anti-insulin B cells is a critical event in disease pathogenesis. We show evidence that this breach in the B cell compartment surprisingly does not depend on T cell help. Common environmental triggers resulting from infection and inflammation that are prevalent in children could combine with endogenous insulin to break anti-insulin B cell tolerance. BCR/TLR costimulation of autoreactive B lymphocytes could represent an early, underappreciated event in the induction of autoimmunity.

We thank Dr. Klaus Rajewsky for providing the pIV_HL2neoR targeting vector, and we thank Dr. Richard Breyer for EIIA-Cre B6 mice. We thank the Vanderbilt Technologies for Advanced Genomics Sequencing Core, the Transgenic Mouse/Embryonic Stem Cell Shared Resource, the Vanderbilt Flow Cytometry Shared Resource, and the Translational Pathology Shared Resource. We thank the Division of Animal Care for assistance in the maintenance of the mice. We also graciously thank Dr. John Williams, Dr. Daniel Moore, and Dr. Damian Maseda for critical review of the manuscript.

The authors have no financial conflicts of interest.