Autoreactive B lymphocytes that are not culled by central tolerance in the bone marrow frequently enter the peripheral repertoire in a state of functional impairment, termed anergy. These cells are recognized as a liability for autoimmunity, but their contribution to disease is not well understood. Insulin-specific 125Tg B cells support T cell–mediated type 1 diabetes in NOD mice, despite being anergic to B cell mitogens and T cell–dependent immunization. Using this model, the potential of anergic, autoreactive B cells to present Ag and activate T cells was investigated. The data show that 1) insulin is captured and rapidly internalized by 125Tg BCRs, 2) these Ag-exposed B cells are competent to activate both experienced and naive CD4+ T cells, 3) anergic 125Tg B cells are more efficient than naive B cells at activating T cells when Ag is limiting, and 4) 125Tg B cells are competent to generate low-affinity insulin B chain epitopes necessary for activation of diabetogenic anti-insulin BDC12-4.1 T cells, indicating the pathological relevance of anergic B cells in type 1 diabetes. Thus, phenotypically tolerant B cells that are retained in the repertoire may promote autoimmunity by driving activation and expansion of autoaggressive T cells via Ag presentation.
Autoreactive B lymphocytes in the developing repertoire are subject to central tolerance in the bone marrow that includes receptor editing and clonal deletion. However, a number of B cells escape central tolerance and enter the mature repertoire in a functionally silent, or anergic, state (1–3). Anergy is considered the principal mechanism that keeps peripheral B cell autoreactivity in check, as anergic cells fail to proliferate or produce Ab in T cell–dependent responses (4). However, their role in autoantigen presentation is not clear.
Transgenic mice (125Tg), in which B cells express anti-insulin BCRs, have enabled the study of tolerance in B cells that recognize a physiologically relevant hormone Ag that is a critical target in type 1 diabetes (T1D) (5, 6). 125Tg BCRs bind rodent insulin with a Kd of <10−7 mol/l (7), and most BCRs in 125Tg mice are occupied by insulin in vivo (8). Proliferative responses to anti-IgM, LPS, or CD40 are significantly impaired in vitro (9), and 125Tg B cells fail to produce insulin-specific Ab responses following immunization in vivo (8). Nevertheless, these B cells are able to support the development of diabetes in NOD mice (5, 9). T1D in both mice and humans results from T cell–mediated destruction of insulin-producing β cells in pancreatic islets. B cells are necessary for T1D pathogenesis, and several studies indirectly support their role in Ag presentation, in a capacity that is not redundant with other APCs (10–16). However, the functional status of B cells in the polyclonal populations used in these studies is not clear.
To directly address the function of tolerant B cells in Ag presentation, we used 125Tg B cells and insulin, or insulin conjugated to peptide mimotopes, to probe for Ag-specific responses from functionally distinct CD4+ T cell populations. Although the anergic state of anti-insulin B cells was confirmed in studies of calcium mobilization, these B cells nevertheless capture and rapidly internalize insulin for processing and presentation. Anergic 125Tg B cells are competent to activate disease-relevant T cells from NOD mice, including anti-insulin T cells, which require APCs to process and present a critical low-affinity B chain epitope (17). We find that tolerant 125Tg B cells are also competent to present specific epitopes to nonautoreactive, naive T cells, which have not been previously primed. When compared with naive B cells, anergic B cells prove efficient for activating T cells when transiently exposed to Ag, indicating that they may be particularly effective when Ag is present intermittently or at low levels. Because B cells exhibiting a similar functional state are present in normal repertoires, these findings indicate that anergic B cells are an ever-present liability for activating autoaggressive T cells. In contrast to the common assertion that autoimmunity arises from a breach in immune tolerance, we find that the pernicious actions of anergic B cells are a consequence of their tolerant state and not its loss.
Materials and Methods
Calcium transients were measured using MACS LS column (Miltenyi Biotec)–purified B cells loaded with the ratiometric dye fura 2-AM (Molecular Probes) and a FlexStation II scanning fluorometer (Molecular Devices). The FlexStation II fluorometer was used to measure calcium fluxes following the addition of ligands at 5 s (insulin 5 μg/ml, hen egg lysozyme [HEL; Sigma-Aldrich] 5 μg/ml, or ionomycin 1 μg/ml) and of 2 mM calcium to the calcium-free buffer at 20 s. Measurements (340/380 nm excitation ratios) were obtained at 5-s intervals.
Cells for Ag presentation
Lymphocytes were purified by negative selection using MACS LS columns (Miltenyi Biotec) following the manufacturer’s instructions. CD4+ T cells were selected using a CD4+ T cell sort kit II (Miltenyi Biotec). For B lymphocyte purification, biotinylated anti-CD43 (1 μl/106 cells), anti-CD11c (0.25 μl/106 cells), and anti-CD11b (0.25 μl/106 cells) were used, followed by streptavidin beads (Miltenyi Biotec) prior to negative selection. Cell purity was confirmed by flow cytometry (>85% for T cells, >90% for B cells).
Insulin internalization assays
For cell labeling, insulin (Eli Lilly) or HEL (Sigma-Aldrich) was modified using sulfo-m-maleimidobenzoyl-N-hydoxysuccinimide ester (sulfo-MBS; Sigma-Aldrich), reacted with N-hydroxysuccinimide ester–biotin (Sigma-Aldrich), or directly conjugated with Alexa 488 succinimide ester (Molecular Probes). These conjugation reactions for insulin target the lysine at B30 for both Alexa 488 and biotin, and these labeled insulins have been interchangeably used for flow cytometry. Biotin-insulin or Alexa 488-insulin (50 ng/ml final) were incubated on ice with 106/100 μl cells in 1.5-ml microfuge tubes, then washed. Pellets were resuspended in 0.5 ml culture medium and returned to ice or transferred to prewarmed tubes and incubated at 37°C for the time indicated. Reactions were stopped by the addition of 1 ml cold FACS buffer containing EDTA and azide, and cells were stained for flow cytometry. For confocal microscopy, total splenocytes were collected from 125Tg/NOD mice and treated as above with Alexa 488-insulin. Following incubation with Ag, cells were fixed with 1 ml PBS containing 4% formaldehyde, then stained for B220 using B220-biotin and Texas Red–avidin. Cells were then washed with 1% FBS in PBS and mounted to glass slides (Cytospin 4; Thermo Shandon, Waltham, MA; 1 min at 750 rpm) using fluorescent mounting medium (Dako, Carpinteria, CA). Visualization was performed at 488 and 561 nm by confocal microscopy using an LSM 710 (Zeiss) equipped with a ×63/1.4 numerical aperture plan apochromat objective lens to image 200-μm fields. At least two fields were collected from each sample. Z-stacks of 6–11 slices were prepared at 0.83- to 0.86-μm intervals. Slices with distinct B220 staining were selected for further analysis. Images were prepared for publication using Zen2011 software. Contrast, brightness, and image magnification were adjusted equally across all images to enable optimal visualization. Images were then exported to Photoshop (Adobe) for organization without further manipulation.
Pork insulin (Eli Lilly) or HEL (Sigma-Aldrich) was modified using the heterobifunctional reagent MBS (Sigma-Aldrich) in dimethyl formamide (Sigma-Aldrich) (18). The MBS-modified Ags were isolated and reacted with OVA323–329 peptide (KISQAVHC; Sigma-Genosys), or BDC2.5 1040-63 peptide (RTRPLWVRMEC; Schafer-N) (19) with N-terminal cysteines to form stable thioether bonds. The conjugates were purified by dialysis and chromatography under sterile conditions and protein concentration was determined by spectrophotometry.
Ag presentation assays
CD4+ T cells (105) were plated in 96-well flat-bottom plates (Costar) with 105 B cells (irradiated, 6 Gy) in 200 μl culture medium. Tritiated [3H]thymidine (PerkinElmer), which was added on day 3 (1 μCi/well), was measured on day 4 as cpm using a Beckman Coulter LS 6500. Anti-insulin 125Tg, control 281Tg B cells, and anti-HEL B cells on a C57BL/6 background were used for assays in which responding transgenic T cells were also C57BL/6 (OTII). For T cells from the NOD background (BDC2.5 and BDC12-4.1), 125Tg/NOD and 281Tg/NOD B cells were used. For Ag-pulsing assays, B cells were incubated with Ag at 37°C for 30 min, then washed extensively, prior to coculture with T cells. CFSE labeling was performed as previously described (20). CFSE-labeled T cells (5 × 105) were plated with B cells (nonirradiated, 5 × 105) in 24-well plates (Costar) with 1 ml culture medium for 4 d. The number of cells in each peak was determined using cell enumeration for each well, multiplied by relevant subset percentages, as determined by flow cytometry. Mitotic events were calculated as previously described (20–22). Briefly, the formula [N × (2n − 1)]/2n was used to determine mitotic events for each generational peak, or n, and in which N is the total number of cells under each peak. This formula allows determination of the original number of precursors for each peak (N ×1/2n) as well as extrapolation of the number of mitotic events for the precursors from each peak by incorporating (2n − 1) into the formula. For example, if 8 cells (N) were present in the second divided peak (n), they would have arisen from 2 precursors (8 × 1/22 = 2). Each precursor would have undergone one mitotic event to give rise to the first generation of daughter cells, and each of those daughter cells would have undergone a second mitotic event, expressed as 22 − 1 = 3 mitotic events. In this example, [8 × (22 − 1)]/22 = 6 mitotic events required to give rise to the 8 cells found in the second generation.
Splenocytes were stained with fluorochrome-conjugated Abs to B220, IgMa, CD4, CD69, and 7-aminoactinomycin D (BD Pharmingen). Biotinylated insulin followed by streptavidin-fluorochrome or insulin-Alexa 488 was used to identify insulin-specific B cells. Data were collected on an LSRII flow cytometer (Becton Dickinson Biosciences) and analyzed using FlowJo software (Tree Star).
ELISPOT was performed as previously described, using an IFN-γ ELISPOT kit (Becton Dickinson Biosciences, San Diego, CA) following the manufacturer’s instructions (23). Briefly, purified BDC12-4.1/NOD T cells (105) and 125Tg/NOD or 281Tg/NOD B cells (105) were cocultured for 48 h at 37°C in anti–IFN-γ-coated 96-well plates (Millipore). IFN-γ–producing T cells were enumerated using a CTL ImmunoSpot analyzer and CTL ImmunoSpot software, version 3.2 (Cellular Technology).
Cell supernatants from CFSE-labeled cells in Ag presentation experiments were analyzed using a mouse cytokine/chemokine multiplex panel (Millipore, catalog no. MPXMCYTO-70K) following the manufacturer's instructions. The plate was read on a Luminex 100 instrument, using Exponent 3.2 software. Raw fluorescent values were analyzed using Milliplex Analyst 3.4 software. A five-parameter log curve fit was used.
125Tg and VH281Tg mice were developed as previously described, harboring nontargeted VH and Vk Tgs on C57BL/6 (B6) or NOD backgrounds (6, 9). BDC2.5/NOD mice, OTII/C57BL/6 mice, and MB4/C57BL/6 mice were purchased from The Jackson Laboratory as breeding pairs, then bred and housed in the Vanderbilt University specific pathogen-free facility. BDC12-4.1/NOD mice were developed as previously described, bred and housed at the University of Colorado (23). All studies were approved by the Institutional Animal Care and Use Committee of Vanderbilt University, fully accredited by the American Association for the Accreditation of Laboratory Animal Care.
B lymphocytes that express an anti-insulin BCR are refractory to insulin stimulation
Insulin-specific 125Tg B cells have impaired proliferative responses to B cell mitogens (9). To confirm that uncoupling of the BCR from Ag-specific signals, the hallmark of B cell tolerance, is present in 125Tg B cells, we assessed calcium mobilization by purified B cells in response to their cognate Ag. Because the essential physiologic requirement for insulin in vivo precludes the availability of naive anti-insulin B cells, anti-HEL MD4 transgenic B cells were used as controls for an Ag-specific response in naive B cells. Mobilization of Ca2+ by anti-insulin and anti-HEL B cells was measured using dual-wavelength ratiometry on fura 2-AM–loaded B cells (Fig. 1A). Calcium flux was initially measured in the absence of extracellular calcium, then following the addition of 2 mM Ca2+, to assess mobilization from intracellular stores and channels in the cell membrane, respectively. The data show that encounter of HEL by naive anti-HEL control B cells triggers rapid release of Ca2+ from intracellular stores. This is followed by a rapid influx through store-operated channels in the membrane upon addition of extracellular Ca2+. In contrast, anti-insulin B cells fail to demonstrate any Ca2+ mobilization following encounter with ligand. This failure to respond to cognate Ag is highly consistent across a wide range of insulin concentrations (1–500 μg/ml, data not shown). Slight elevation of basal Ca2+ levels is sometimes observed in anti-insulin B cells, but this finding is not consistent. As expected, anti-HEL control B cells do not respond to insulin. Although unable to signal in response to Ag, anti-insulin B cells are competent to flux calcium as demonstrated by the response to the calcium ionophore, ionomycin (Fig. 1B). These data and prior functional studies indicate that exposure of anti-insulin B cells to their Ag in vivo is sufficient to uncouple Ag-specific BCR signaling and render anti-insulin B cells tolerant (8, 9).
B cell receptors on tolerant anti-insulin B cells efficiently capture and internalize Ag
To investigate the potential for Ag internalization by 125Tg B cells, labeled insulins were used to track the fate of BCR-associated Ag. Purified B cells were chilled on ice and then incubated with Ag (50 ng/ml). After washing in cold buffer, cells were either incubated at 37°C to allow internalization of the Ag or returned to ice (controls). To provide differential staining, insulin was conjugated to either biotin or Alexa 488. Insulin-Alexa 488 permits identification of internalized Ag by flow cytometry. Biotinylated insulin, however, only allows detection of Ag remaining on the cell surface and is revealed by the addition of fluorochrome-conjugated streptavidin following incubations. Cells incubated on ice do not internalize Ag, and they served as controls. As shown in Fig. 1C, 125Tg B cells loaded with biotinylated insulin and incubated on ice, followed by staining with avidin-FITC, bind insulin in proportion to surface IgM, as do those loaded with insulin-Alexa 488 (Fig. 1E). After incubation at 37°C, however, loss of fluorescence (FITC) indicates insulin is no longer present on the surface of cells incubated with biotinylated insulin (Fig. 1D). IgM on the surface remains unchanged, consistent with BCR turnover known to continuously replace surface IgM. In contrast, insulin-Alexa 488 allows visualization of the fluorescent signal even after Ag internalization (Fig. 1F), indicating that the Ag has not simply dissociated from the surface, but remains diffusely cell associated. Confocal microscopy was used to confirm internalization of insulin-Alexa 488 by anti-insulin B cells. Splenocytes from 125Tg mice, incubated on ice with insulin-Alexa 488 and then washed, were returned to ice or placed at 37°C to allow for Ag internalization. Cells were counterstained with B220 to discriminate the B cell perimeter. Cell imaging shows that most of the Ag remains on the surface of cells maintained in the cold, whereas Ag is readily visualized internal to the perimeter after warming the cells for 5 min (Fig. 1G and 1H, respectively). Similar Ag internalization was also observed by confocal microscopy after 10 and 20 min at 37°C (Supplemental Fig. 1). These findings indicate that tolerant anti-insulin B cells efficiently capture and internalize their ligand in vitro under conditions that reflect monovalent BCR encounter with Ag in vivo. Findings do not differ between 125Tg B cells on C57BL/6 versus NOD backgrounds for either calcium flux or Ag internalization.
Naive T cells are activated by anergic B cells
Internalized Ag must be processed for presentation to T cells. To test the ability of tolerant anti-insulin B cells to perform this function, we conjugated insulin to a peptide recognized by OTII CD4+ T cells, OVA323–329, for internalization by the insulin-specific BCR of 125Tg B cells. All cells used in these studies were from mice fully backcrossed onto C57BL/6 backgrounds. Purified OTII CD4+ T cells were cocultured with irradiated, purified B cells, with or without insulin-OVA323–329, and T cell proliferation was measured using tritiated thymidine. Fig. 2A shows robust proliferative responses of OTII CD4+ T cells activated by 125Tg B cells in the presence of 1 μg/ml conjugate Ag (black bars). This is not due to nonspecific cellular activation, as indicated by poor proliferation in Ag-free cocultures. 281Tg noninsulin-binding B cells (dark gray bars) do not stimulate T cells to proliferate above background of T cells and Ag alone (light gray bars), indicating that Ag-specific uptake by tolerant 125Tg B cells, not fluid phase pinocytosis, drives activation of naive T cells (p < 0.01; 125Tg versus 281Tg or versus T cells alone, with continuous Ag). These data indicate that the Ag processing and presentation function of tolerant 125Tg B cells is intact.
Ag-pulsed anergic B cells are sufficient for T cell activation
B cells occupy a unique Ag-presenting niche, as they are able to internalize and concentrate specific Ag. We therefore examined the capability of anergic B cells to process and present limited amounts of physiologic Ag. Purified B cells were pulsed with insulin-OVA323–329 conjugate at 37°C for 30 min, then washed to mimic Ag acquisition under temporal conditions. These Ag-loaded B cells were irradiated and cocultured with OTII CD4+ T cells. No free Ag was added to the cultures. Data presented in Fig. 2B show that OTII CD4+ T cells are induced to proliferate by Ag-pulsed 125Tg B cells (black bars). In the absence of Ag exposure, 125Tg B cells do not induce T cell proliferation, and Ag-pulsed control 281Tg B cells induce only low levels of proliferation, which may be due to limited Ag uptake via pinocytosis (dark gray bars; p = 0.01; pulsed 125Tg versus pulsed 281Tg). These data indicate that tolerant B cells specific for a physiologic Ag, insulin, can present Ag and activate naive cognate T cells in a setting wherein Ag is present only in small amounts or for limited periods of time.
Tolerant 125Tg B cells present Ag more efficiently than do naive B cells under Ag-limiting conditions
We next compared the efficiency of Ag presentation by anergic 125Tg B cells with that of naive B cells. Because their Ag is present physiologically, mature, naive, anti-insulin B cells are not available for study. We therefore used transgenic B cells that recognize HEL (MD4) as naive comparators (C57BL/6 background). Conjugated HEL-OVA323–329 Ag was used to test Ag-presenting capacity of naive anti-HEL B cells to OTII CD4+ T cells, in continuous and pulsed conditions, in parallel assays with anergic 125Tg B cells and insulin-OVA323–329. As shown in Fig. 2C, OTII CD4+ T cell proliferation in response to continuous Ag is not significantly different when activated by anergic 125Tg B cells (black bars) or naive anti-HEL B cells (gray bars). However, when B cells are pulsed with Ag, then washed prior to coculture with T cells, naive anti-HEL B cells are unable to stimulate T cell proliferation as well as anergic anti-insulin 125Tg B cells can (p = 0.01). This finding for anti-insulin B cells is consistent with previously published studies indicating that naive anti-HEL B cells internalize Ag less efficiently than do anergic ones (24). Thus, under conditions of limited or intermittent Ag availability, tolerant B cells may have an advantage in activating their cognate T cells compared with their naive counterparts.
Tolerant 125Tg B cells drive anti-insulin BDC12-4.1 T cells to produce IFN-γ
To determine the pathogenic relevance of Ag processing and presentation by tolerant anti-insulin B cells, we next used the NOD mouse model of T1D. In studies published to date, 125Tg B cells are the only Ig transgenic B cells that fully support diabetes development in NOD mice (5). To directly test their ability to present autoantigen in this setting, we used BDC12-4.1/NOD CD4+ T cells, which recognize a key insulin epitope that mediates diabetes development in NOD mice (23). Importantly, BDC12-4.1 CD4+ T cells were recently shown to recognize a truncated peptide from insulin B:9-23, which binds IAg7 in register 3 with low affinity (17). Therefore, Ag handling by APCs is critical to the activation of BDC12-4.1 T cells, as processing must generate this low-affinity epitope to be effective. We used ELISPOT assays to determine whether anti-insulin 125Tg/NOD B cells could drive BDC12-4.1/Rag−/−/NOD T cells to produce IFN-γ, as this response is known to report splenocyte APC and peptide stimulation (23). As shown in Fig. 3, coculture of 125Tg/NOD B cells (black bars) with BDC12-4.1/Rag−/−/NOD CD4+ T cells, in the presence of intact insulin, induced increased numbers of T cells to produce IFN-γ, compared with those cocultured with Ag and 281Tg/NOD B cells (gray bars, p < 0.01). These data directly demonstrate that tolerant anti-insulin B cells are capable of driving effector function of pathogenic T cells. Furthermore, the findings indicate that processing of insulin by tolerant 125Tg B cells permits presentation of low-affinity pathogenic epitopes from insulin.
Cognate interaction with tolerant 125Tg/NOD B cells induces proliferation and Th1-predominant cytokine production by islet-specific BDC2.5/NOD T cells
To investigate the ability of tolerant B cells to fully activate diabetogenic T cells, we used another well-characterized TCR transgenic model of T1D, BDC2.5/NOD CD4+ T cells that recognize chromogranin A (25), with 125Tg/NOD or 281Tg/NOD B cells. For these experiments, insulin was conjugated to a BDC2.5 peptide mimotope, 1040-63, known to activate these T cells (19). The data show that BDC2.5/NOD T cells proliferate better when cocultured with Ag and 125Tg/NOD B cells (Fig. 4A, black bars) than with noninsulin-binding controls (dark gray bars). In pulsed conditions, T cells are induced to proliferate by Ag-pulsed 125Tg/NOD B cells (Fig. 4B, black bars), but not Ag-pulsed 281Tg/NOD control B cells (dark gray bars; p = 0.01). To understand the extent of this cellular activation, additional experiments were performed using CFSE-labeled BDC2.5/NOD CD4+ T cells. This approach allows analysis of all cells that proliferate during the full 4-d course of the culture, as well as confirmation of cellular activation, and cytokine analysis of supernatants. Fig. 5A shows flow cytometric analysis of CD4+ gated, CFSE-labeled BDC2.5/NOD T cells after coculture with 125Tg/NOD B cells. T cells that have undergone division are represented by CFSE+ peaks of decreasing fluorescence (x-axis), and they are also activated, indicated by increased expression of CD69 (y-axis). The histogram overlay depicts CFSE+ peaks from CD4+ gated T cells after coculture with 125Tg/NOD B cells (blue), 281Tg/NOD control B cells (red), or alone (black) in the presence of Ag. Enumeration of CD4+ T cells in each CFSE+ generational peak confirms the robust proliferative responses of BDC2.5/NOD T cells cocultured with 125Tg/NOD B cells compared with control 281Tg/NOD B cells (Fig. 5B, p < 0.05 for the first generation, and p < 0.01 for generations two through five). Total mitotic events were calculated for each peak and summed, as shown in Fig. 5C (7.89 ± 1.48 × 105 total mitotic events for cells cocultured with 125Tg, versus 2.19 ± 1.48 × 105 for those with 281Tg, p < 0.01). To determine the functional outcome of interactions between tolerant B cells and BDC2.5/NOD CD4+ T cells, cytokine analysis was performed by Luminex assays on cellular supernatants from the above studies. As shown in Fig. 5D, IFN-γ predominates the cytokine profile from T cells cocultured with 125Tg/NOD (4514 ± 594 pg/ml) versus 281Tg/NOD (762 ± 54 pg/ml, p < 0.01). Other cytokines shown to be differentially produced are: IL-10 (821 ± 58 versus 340 ± 25 pg/ml, p < 0.01), IL-17 (278 ± 43 pg/ml versus 193 ± 23 pg/ml, p < 0.05), TNF-α (100 ± 10 versus 53 ± 7 pg/ml, p < 0.01), and IL-4, which was found only in extremely small amounts (7.5 ± 2.2 versus 3.2 ± 0.0 pg/ml, p < 0.05). Other cytokines tested that were not statistically different between groups included IL-2, IL-6, and IL-12 (data not shown).
Tolerance, or anergy, in self-reactive B cells is expected to protect against autoimmunity, and a breach in tolerance is generally considered necessary for initiation of autoimmune disorders. However, the data presented in this study indicate that the anergic state does not prevent anti-insulin B cells from presenting Ag for both conventional and autoimmune T cell epitopes. Rather, BCR internalization results in effective delivery of Ag to endosomal compartments and loading of MHC class II, as evidenced by the ability of anti-insulin B cells to activate Ag-specific T cells. Using insulin conjugated to known T cell epitopes, anti-insulin B cells are shown to be effective APCs for both naive CD4+ T cells that recognize OVA peptide (OTII) and for potentially experienced autoreactive (BDC2.5) T cells that recognize a peptide mimotope (1040-63) for chromogranin A. Anti-insulin B cells that present this mimotope to BDC2.5 T cells support a Th1 effector response in this model. In addition to conjugated epitopes, anti-insulin B cells are also competent to generate and present epitopes from intact insulin and drive IFN-γ production from diabetogenic anti-insulin (BDC12.4-1) T cells. These in vitro studies are consistent with the finding that anti-insulin transgenic BCRs provide NOD B cells with an essential component that promotes T cell–mediated T1D in vivo, despite their functionally anergic state (5, 9).
The interaction of 125Tg B cells with anti-insulin T cells allows new insight into the question of how the critical epitopes recognized by these T cells may be presented in vivo. BDC12-4.1 T cells recognize distinct residues from the B:9-23 peptide of insulin, register 12-20, that binds IAg7 with a low affinity. This low-affinity interaction is proposed to permit T cells to escape deletion in the thymus (26, 27). Some studies suggest that this type of low-affinity epitope may only be generated within the islets, where insulin concentrations are sufficiently high to load weakly binding peptides onto diabetogenic MHC class II, IAg7. Our findings identify an additional mechanism for generating low-affinity epitopes that is mediated by the efficient capture and processing of autoantigen by specific B cells in which the continuous delivery of insulin to endosomal compartments aid MHC class II loading of low-affinity epitopes. As B cells heavily populate inflamed islets, and many are insulin-specific (28, 29), this may, in fact, be a major means by which pathogenic T cells are activated in islets.
Fully functional, Ag-specific B cells are recognized to be highly efficient APCs because of their ability to capture and concentrate Ags for processing and presentation to T cells (30, 31). The present study on anti-insulin B cells, as well as recent data on tolerant B cells in anti-HEL/sHEL or anti-ssDNA models, clearly demonstrates that facilitated BCR internalization is a component of the tolerant state (24, 32, 33). Tolerance in the HEL model is characterized by developmental arrest and rapid turnover of B cells at the transitional stage (34). In contrast, anti-insulin B cells are not blocked in development and enter mature B cell subsets (5, 8, 9). Thus, receptor internalization remains intact in tolerant B cells from these two different models, independent of the maturational subset in which they reside. Similarly, internalization of BCRs is also intact in anergic anti-ssDNA and in ArsA1 B cells (32). However, the anergic state of B cells in these models includes a reversible block in late endosomal entry of BCR and TLR necessary for effective Ag presentation (32). This level of control is not evident in either anti-HEL or anti-insulin models (this study and Refs. 35, 36). Thus, the diverse mechanisms by which B cell anergy is maintained for different self-molecules also includes differences in how tolerogenic signaling regulates Ag processing and presentation.
The question of how these various models relate to anergy in the physiologic setting is still open. Anergic, autoreactive-prone B cells have been shown to be part of the normal repertoire, in both mice and humans, and may be increased in some disease states (1–3). B cell anergy can be reversed in certain circumstances (9, 37, 38), raising the question of whether interaction with cognate T cells could deliver signals that would reverse B cell anergy and lead, in turn, to more efficient T cell activation by those B cells. In fact, the underlying assumption regarding anergic B cell contributions to disease is that this reversal is a necessary component. Elegant studies in the anti-HEL model have shown that T cell help can effectively rescue those cells from Ag-induced anergy, and that this rescue is more effective in NOD than in B6 B cells (38). Anti-insulin B cells used in the present studies, however, retain their anergic properties in most circumstances, with the exception of in vitro stimulation with anti-CD40 together with IL-4 (9). Coculture with cognate T cells in the assays presented in Fig. 5, which did not require B cell irradiation, failed to expand B cell numbers, indicating lack of proliferation even in the setting of direct interaction with cognate helper T cells (Supplemental Fig. 2A). We have also performed adoptive transfer studies using CFSE-labeled anti-insulin B cells and find that they can successfully populate inflamed islets with little evidence of proliferation even as late as 5 d after transfer (Supplemental Fig. 2B). Furthermore, as we have previously shown, even T cell–dependent immunization using insulin-CFA fails to induce them to produce insulin Abs in vivo in either C57BL/6 or NOD mice (8). Nevertheless, they are able to support the development of T1D in the NOD model (9). Therefore, it does not seem likely that the Ag-presenting function of 125Tg B cells depends on their loss of anergy.
A related question concerns the anergic state of B cells that promote T1D in the endogenous BCR repertoire of wild-type NOD mice, as well as in humans. Because anti-insulin Abs are the hallmark, and are predictive, of T1D, then by definition there are nonanergic anti-insulin B cells present. It is possible that these Abs may arise in the wake of somatic hypermutation from nonanergic cells that originally recognized insulin with lower affinity. This idea is supported by the fact that the 125Tg was originally created from an anti-insulin Ab that emerged with two CDR changes postimmunization from a germline H chain (281Tg) that does not detect insulin at high enough affinity to be detected by flow cytometry. However, Ab production may be a separate issue from Ag presentation in terms of disease promotion. Although the findings presented in this study do not prove that anergic cells in the wild-type NOD repertoire promote disease development in T1D, they support the idea that anergy need not necessarily be reversed to contribute to autoimmune disease via Ag presentation.
These findings contribute to a body of emerging data on the importance of B cells in T1D in both NOD mice and in humans (10, 15, 16, 20, 39–42). Furthermore, our data emphasize the important autoantigen-presenting function of B lymphocytes that have not emerged from their tolerant state. We propose that the presence of such B cells expands the pool of autoreactive T cells prior to the appearance of autoantibodies. Therefore, the limited success of T1D interventions in subjects with autoantibodies to insulin, GAD65, or other β cell Ags may reflect autoreactive T cell expansion occurring in response to anergic B cells, before overt loss of tolerance in the B cell compartment gives rise to autoantibodies. Understanding the regulation of these events in tolerant anti-insulin B cells may provide new targets for preventing the expansion of autoaggressive T cells in T1D.
We acknowledge Dr. George Eisenbarth (University of Colorado, Denver, CO) for support and encouragement throughout the development of this project. We thank Chrys Hulbert and Hunter Houston at Vanderbilt University Medical Center for technical assistance, and Maki Nakayama, University of Colorado, Denver, for her role in developing the BDC12-4.1/NOD mouse model. The Vanderbilt Hormone Assay and Analytical Services Core performed the Luminex cytokine analysis. Flow Cytometry experiments were performed in the Vanderbilt Medical Center Flow Cytometry Shared Resource. Confocal experiments were performed in part through the use of the Vanderbilt University Medical Center Cell Imaging Shared Resource.
This work was supported by National Institutes of Health Grants R01 DK 084246, R01 AI 051448, K08 DK 070924, and R21 DK 084568, as well as by Juvenile Diabetes Research Foundation Grants APC 4-2007-1056 and 1-2008-108. The Vanderbilt University Medical Center Flow Cytometry Shared Resource is supported by the Vanderbilt Ingram Cancer Center (National Institutes of Health Grant P30 CA68485) and the Vanderbilt Digestive Disease Research Center (National Institutes of Health Grant DK058404). The Vanderbilt University Medical Center Cell Imaging Shared Resource is supported by National Institutes of Health Grants CA68485, DK20593, DK58404, HD15052, DK59637, and EY08126.
The online version of this article contains supplemental material.
The authors have no financial conflicts of interest.