BIM Deficiency Protects NOD Mice From Diabetes by Diverting Thymocytes to Regulatory T Cells

The immune system has evolved to allow robust responses against a diverse range of pathogens while preserving self-tolerance to prevent immunopathology. Developing self-reactive thymocytes that express T-cell receptors (TCRs) that bind with high affinity to MHC/self-peptide complexes are usually deleted by apoptosis in the thymus, a process termed “deletion” or “recessive tolerance” (1). A complementary mechanism of immune tolerance, termed “dominant tolerance,” involves the generation of FoxP3-expressing regulatory T cells (Tregs) capable of suppressing the activation of autoreactive, conventional T cells that fail to undergo negative selection in the thymus (e.g., because the cognate self-antigen is not expressed and presented in the thymus) (2). The importance of Tregs and thymic deletion in preventing autoimmunity is illustrated in immunodysregulation polyendocrinopathy enteropathy X-linked syndrome and in autoimmune polyglandular syndrome type 1, respectively, both of which result in widespread autoimmune disease, including type 1 diabetes (3,4).

Tregs have considerable therapeutic potential in autoimmune diseases such as type 1 diabetes. Tregs that express TCR specific for self-antigen can home, along with autoreactive effector T cells, to antigen-expressing target peripheral tissues, such as pancreatic islets and their draining lymph nodes. At these sites, they can suppress not only autoreactive effector cells with TCR specificities for the same antigen they recognize but also others, a phenomenon called bystander suppression.

The signaling events that regulate and presumably distinguish between thymic deletion and Treg development remain incompletely defined. Apoptosis is critical for deletion of self-reactive T cells in both the thymus and the peripheral lymphoid tissues and for thymic Treg development. BIM is a proapoptotic BH3-only protein required for deletion of autoreactive thymocytes (5). BIM-deficient mice not only have defects in T-cell deletion (6) but also have increased numbers of FoxP3-expressing Tregs (7). How BIM controls the development of Tregs in BIM-deficient mice and the functional significance of increased Tregs in BIM-deficient mice have not been studied.

Although it has been demonstrated that high-affinity TCR interaction with self-antigen is required to trigger both thymic deletion and Treg differentiation, these studies involved transgenic expression of antigen or peptide administration or used TCR transgenic T cells present at high clonal frequencies (4,8–10). However, other data suggest that thymic Treg development uses a limited antigenic niche (11,12), implying that models using ubiquitous antigen presentation or TCR transgenic T cells at high clonal frequencies may not reflect normal events. Therefore, it is important to study autoreactive T-cell deletion and Treg differentiation in a disease with a polyclonal T-cell repertoire and natural self-antigens expressed at physiological levels. To this end, we studied the impact of removing a critical mediator of thymocyte apoptosis, BIM, on thymic deletion and Treg differentiation in the NOD mouse, an animal model of type 1 diabetes (13). In this model, mice express natural self-antigens and have a polyclonal T-cell repertoire. The data show that reducing thymic deletion by disabling thymic apoptosis ameliorates autoimmune diabetes by increasing Tregs.

BIM-deficient C57BL/6 mice (6) were backcrossed onto the NOD/Lt genetic background for >10 generations. NOD.BDC2.5 mice were provided by D. Mathis and C. Benoist (Harvard University). NOD.Rag1^−/− and NOD mice expressing GFP under control of the FoxP3 promoter were obtained from The Jackson Laboratory. C57BL/6 and NOD/Lt mice were obtained from the Walter and Eliza Hall Institute animal breeding facility (Kew, VIC, Australia). Diabetes monitoring and insulitis scoring were performed as described previously (14). Mice with two consecutive blood glucose readings >15 mmol/L were considered diabetic. All animal studies were conducted under specific pathogen-free conditions at St. Vincent’s Institute (Melbourne, VIC, Australia) following the guidelines of the institutional animal ethics committee.

The Bim gene (Bcl2l11) lies within the Idd13 diabetes susceptibility locus. To confirm the genetic background of NODBim^−/− mice and define the congenic interval encompassing the Bcl2l11 null allele, DNA was isolated from tail biopsy specimens from 10th generation backcrossed mice and genotyped by the Australian Genome Research Facility using the mouse 5K targeted genotyping array run on the Affymetrix GeneChip Scanner 3000 7G MegAllele system. To refine the 129/Sv-derived congenic interval encompassing the Bcl2l11 null allele, additional polymorphic markers (nucleotide repeats and single nucleotide polymorphisms [SNPs]) were identified using the Mouse Genome Informatics and National Center for Biotechnology Information databases. DNA was genotyped by standard PCR and gel electrophoresis methods (selected nucleotide repeats) and by The Centre for Applied Genomics (The Hospital for Sick Children, Toronto, ON, Canada) using the Illumina medium-density linkage panel (1,449 SNPs), which contained SNPs that further refined the congenic interval. NODBim^−/− mice were of NOD genotype across the whole genome except for a region on chromosome 2 encompassing the Bcl2l11 locus. The congenic interval is 129-derived between, but not including, D2mit277 (Chr2:123,255,831) and D2mit338 (Chr2:130,655,046), and no B6 alleles were detected within this interval. The 129-congenic interval is unlikely to contribute to diabetes protection because 129 mice are predicted to harbor a susceptibility, not a resistance, allele in this region of chromosome 2 (15).

Antibodies used were anti-CD4-PerCp-Cy5.5 (RM4-5), anti-CD3 (145-2C11), anti-CD5 (clone 53-7.3), and anti-CD24 (clone M1/69) all conjugated to fluorescein isothiocyanate (FITC); anti-CD44-AlexaFluor 700 (1M7), anti-CD69-allophycocyanin (H1.2F3), and anti-62L-allophycocyanin-Cy7 (MEL-14) (all from BD Biosciences, San Jose, CA); anti-CD11c (N418), anti-B220 (RA3-6B2), anti-CD11b (M1/70), and anti-F4/80 (BM8) all conjugated to eFluor 450 (all from BioLegend, San Diego, CA); and anti-CD8a-Pacific Orange (5H10) (Invitrogen, Carlsbad, CA). FITC-conjugated antibodies specific for individual TCR Vβ chains were purchased as a kit from BD Biosciences. Anti-FoxP3 (FJK-16s), anti-Nur77 (12.14) conjugated to phycoerythrin, anti-Helios-FITC (22F6), and anti-GITR-allophycocyanin (DTA-1) were purchased from eBioscience (San Diego, CA). Analysis was performed on a BD LSRFortessa (BD Biosciences, Franklin Lakes, NJ) using FlowJo (Tree Star, Ashland, OR) software. Intracellular staining was performed according to the manufacturer’s specifications using the BD Cytofix/Cytoperm Plus kit (BD Biosciences, San Jose, CA).

The phycoerythrin-conjugated I-A (g7) tetramers, insulin B_9–23 (HLVERLYLVCGGEG) (16) and control CLIP (AMKRHGLDNYRGYSL), were provided by the National Institutes of Health Tetramer Core Facility. Tetramer staining and magnetic bead–based cell enrichment were performed as described previously (17).

Whole thymus or sorted thymocyte subsets were homogenized in lysis buffer containing 50 mmol/L Tris-HCl (pH 8), 150 mmol/L NaCl, 0.5% Triton X-100, and protease inhibitor cocktail (Sigma). Proteins (8 μg/lane) were resolved by SDS-PAGE and transferred using standard procedures to nitrocellulose. Monoclonal mouse antibodies to nuclear factor-κB (NF-κB) p65, phosphorylated IκB-α, and β-actin (Santa Cruz Biotechnology, Dallas, TX), and horseradish peroxidase–conjugated anti-mouse Ig antibodies were used.

For carboxyfluorescein succinimidyl ester (CFSE) dilution assay, T cells were labeled with CFSE as previously described (18). For Treg suppression assays, sorted CD25⁻GITR⁻CD4⁺ or CD4⁺FoxP3⁻CD25⁻ were used as responder T cells, and CD4⁺GITR⁺ or CD4⁺FoxP3⁺ cells were used as Tregs. CFSE-labeled responder T cells (5 × 10⁴) were cultured for 72 h with γ-irradiated (4,000 rad) splenocytes (5 × 10⁴) and anti-CD3 monoclonal antibody 1 μg/mL (145-2C11; Bio X Cell) in the presence or absence of the indicated ratio of Tregs.

Data are presented as mean ± SD. Data analysis was performed using GraphPad Prism software (GraphPad Software, San Diego, CA), and the Mann-Whitney test was used to assess statistical significance. The log-rank test was used to perform survival curve analysis.

NODBim^−/− mice had higher percentages of CD4⁺CD8⁻ and CD8⁺CD4⁻ thymocytes and a reduction in CD4⁺CD8⁺ thymocytes compared with control NOD mice (Fig. 1A, B, E, and G). The total splenocyte number was significantly increased in NODBim^−/− mice, whereas the proportions of CD4⁺ T cells, CD8⁺ T cells, and B cells were not statistically different from control NOD mice (Fig. 1A, B, E, and G). These findings are similar to Bim^−/− mice on other genetic backgrounds (5,6,19) and suggest impaired thymocyte deletion.

In NOD mice, T cells bearing the TCR Vβ3 chain represent only 0.4% of the CD4⁺ T cells because of clonal deletion of most Vβ3⁺ immature thymocytes by the mammary tumor virus-3 (Mtv-3) superantigen presented by class II MHC (20). T cells bearing the TCR Vβ3 chain were significantly increased in the spleen and thymus of NODBim^−/− mice (Fig. 1C and F and Supplementary Fig. 1A), indicating that loss of BIM causes a defect in superantigen-mediated deletion within a polyclonal T-cell repertoire on the NOD background.

We then assayed the numbers of insulin B_9–23–specific CD4⁺ T cells, a population of T cells known to play a central role in the pathogenesis of type 1 diabetes (21). Proinsulin 2 is expressed in the mouse thymus (22,23), and NOD mice deficient in proinsulin 2 develop accelerated diabetes, suggesting that high-affinity insulin-specific T cells are tolerized in the normal NOD thymus (24). Using the method of staining with tetramer followed by enrichment using magnetic beads (17), we detected rare insulin B_9–23–specific CD4⁺ T cells in the peripheral lymphoid organs and thymus of NOD mice. The small number of insulin B_9–23–specific CD4⁺ T cells (16) in the thymus of control NOD mice suggests that deletion of thymocytes with this specificity is stringent but incomplete. This population was significantly increased (P < 0.05) in NODBim^−/− mice (Fig. 1D and H). These data demonstrate that BIM deficiency impairs deletion of autoreactive T cells in autoimmune disease–prone NOD mice.

We analyzed the effect of impaired deletion on the development of autoimmunity in NODBim^−/− mice. NODBim^−/− mice developed glomerulonephritis as shown by mesangial deposits of immune complexes (Supplementary Fig. 1B) consistent with the systemic lupus erythematosus–like disease observed in BIM-deficient mice on other genetic backgrounds. By contrast, insulitis was significantly reduced in NODBim^−/− mice (Fig. 2A and B). Remarkably, NODBim^−/− mice were completely protected from diabetes, and even heterozygous NODBim^+/− mice were partially protected from this disease (Fig. 2C). This was surprising because we had expected that the defective deletion caused by BIM deficiency would accelerate diabetes.

We next investigated whether protection from diabetes in NODBim^−/− mice was due to protection of islets from death by apoptosis mediated by effector T cells. We transferred splenocytes from recently diabetic NOD mice into γ-irradiated NODBim^−/− and control NOD mice (Fig. 2D). Both NOD and NODBim^−/− recipients developed diabetes at similar rates. The transfer of activated diabetogenic T cells bypasses T-cell priming and activation in the pancreatic lymph node (PLN). To test whether these were affected by loss of BIM, we transferred CFSE-labeled IGRP-specific naive T cells from NOD 8.3 mice into NOD and NODBim^−/− mice. We and others have previously used CD8⁺ T cells from NOD 8.3 to assess the priming of autoreactive T cells in the PLN (18,25). Transferred CFSE-labeled naive IGRP-specific transgenic cells proliferated similarly in PLNs of NOD and NODBim^−/− mice (Supplementary Fig. 1C and D), indicating that protection does not derive from impaired priming of autoreactive cells in NODBim^−/− mice. These experiments collectively indicate that the protection observed in NODBim^−/− mice was not due to β-cell resistance to autoimmune destruction.

It has been reported that BIM deficiency can protect from autoimmunity by affecting the function of effector T cells in a cell-intrinsic manner (26). We detected insulin B_9–23–specific CD4⁺ T cells in the thymus and IGRP_206–214-specific and insulin B_15–23–specific CD8⁺ T cells within the insulitis of NODBim^−/− mice (Supplementary Fig. 1E), demonstrating that the protection from diabetes seen in NODBim^−/− mice is not due to the absence of autoreactive T cells. Moreover, the effector T cells in NODBim^−/− mice were functional in the NOD.BDC2.5 model (as discussed later; see Fig. 5C). These experiments demonstrate that the effector T cells in NODBim^−/− mice are functional.

Thymocytes expressing TCRs that recognize self-peptide/MHC complexes with sufficient high affinity can also differentiate into Tregs (27). We examined whether the self-reactive T cells that escape deletion in NODBim^−/− mice preferentially differentiate into Tregs. The percentages and absolute numbers of CD4⁺FoxP3⁺ T cells in the spleen, thymus, PLNs, and inguinal lymph nodes were substantially increased in NODBim^−/− mice compared with control NOD mice (Fig. 3A and E). The proportion of Tregs was also increased in the islet-infiltrating cells of NODBim^−/− mice compared with NOD mice (Supplementary Fig. 1F). Because NODBim^−/− had less insulitis, we could not compare the absolute number of islet-infiltrating Tregs between NOD and NODBim^−/− mice. More than 90% of the CD4⁺FoxP3⁺ T cells in both NOD and NODBim^−/− mice also expressed the cell surface marker GITR (Fig. 3B and Supplementary Fig. 2C). As shown previously in C57BL/6 mice, there were increased CD4⁺FoxP3⁺CD25⁻ T cells in NODBim^−/− mice compared with wild-type NOD mice (7) (Fig. 3C).

The increase in Tregs was most prominent in the thymus of NODBim^−/− mice, suggesting that BIM deficiency promotes an increase in natural Tregs. To further confirm that the excess FoxP3⁺ cells in NODBim^−/− mice are derived from the thymus, we performed expression analysis for the natural Treg-specific marker Helios (28). Indeed, >85% of the Treg population in NODBim^−/− mice expressed Helios (Fig. 3D).

Next, we investigated whether thymocytes bearing TCR Vβ3 chains and self-reactive insulin B_9–23–specific CD4⁺ T cells that are normally deleted after encountering self-antigen in NOD mice differentiate preferentially into Tregs in NODBim^−/− mice. We examined GFP⁺ (FoxP3-expressing) insulin B_9–23–specific CD4⁺ T cells and CD4⁺ T cells bearing TCR Vβ3 chains in transgenic NOD and NODBim^−/− mice that express GFP under control of the FoxP3 promoter (NODBim^−/−FoxP3-GFP mice) (29). First, we found that FoxP3 (GFP)–expressing T cells bearing TCR Vβ3 and insulin B_9–23–specific CD4⁺ thymocytes were increased in NODBim^−/−FoxP3-GFP mice compared with NODFoxP3-GFP mice (Fig. 4A and B). Second, because the majority of FoxP3⁺ T cells also express the cell surface marker GITR, we also used GITR as a marker of Tregs. We found that GITR-expressing Tregs were enriched in the fraction of T cells bearing TCR Vβ3 chains, suggesting that autoreactive T cells are enriched within the Treg population in NODBim^−/− mice (Supplementary Fig. 2A). Incomplete deletion of insulin B_9–23–specific CD4⁺ thymocytes in NOD mice yielded both GITR⁺ and GITR⁻ insulin-specific T cells (Supplementary Fig. 2B). Of note, insulin B_9–23–specific CD4⁺ T cells were significantly increased (P < 0.05) in NODBim^−/− mice, and these insulin B_9–23–specific CD4⁺ T cells were enriched for GITR-expressing Tregs (Supplementary Fig. 2B). These results indicate that impaired deletion in NODBim^−/− mice diverts most autoreactive T cells into Tregs.

To examine whether the protection from diabetes in NODBim^−/− mice could be due to increased antigen-specific Tregs, we tested the function of the Tregs. Sorted CD4⁺GITR⁺ T cells from NODBim^−/− mice or CD4⁺FoxP3⁺ T cells from NODBim^−/−FoxP3-GFP mice substantially diminished the proliferation of conventional CD4⁺ T cells in response to stimulation with CD3 agonist antibody in a standard in vitro T-cell suppression assay; their activity was comparable to that of Tregs from control NOD mice or NODFoxP3-GFP mice (Fig. 4C and Supplementary Fig. 2D). We did not observe a significant difference in the suppressive function between CD25^hi and CD25^lo FoxP3⁺CD4⁺ T cells from NODBim^−/−FoxP3-GFP mice (Supplementary Fig. 2D). Furthermore, when sorted CD4⁺GITR⁺ Tregs from NOD or NODBim^−/− mice were cotransferred with diabetogenic splenocytes, both delayed diabetes in NOD.Rag1^−/− mice with similar efficacy. These experiments confirm that Tregs in NODBim^−/− mice are functional (Fig. 4D).

To examine the function of effector T cells and Tregs in NODBim^−/− mice, we used NOD.BDC2.5 mice. Even though sorted BDC2.5 TCR transgenic T cells are highly diabetogenic when transferred into NOD.SCID or NOD.Rag^−/− recipients, NOD.BDC2.5 mice spontaneously develop diabetes only at a very low incidence. However, when NOD.BDC2.5 mice are crossed to mice with the Rag mutation and/or FoxP3 Scurfy mutation to eliminate Tregs, 100% of the resulting compound mutant animals develop diabetes rapidly, indicating that Tregs protect NOD.BDC2.5 mice from autoimmune diabetes (30). BIM-deficient NOD.BDC2.5 mice were completely protected from diabetes, and the percentages of CD4⁺FoxP3⁺ T cells were increased in these animals compared with NOD.BDC2.5 control mice (Fig. 5A and B). As expected, total spleen cells from NOD.BDC2.5 or BIM-deficient NOD.BDC2.5 mice did not transfer diabetes efficiently when injected into NOD.Rag1^−/− mice: Only one of the six recipients receiving NOD.BDC2.5 spleen cells and none of the six recipients receiving BIM-deficient NOD.BDC2.5 spleen cells became diabetic. On the other hand, 5 × 10⁵ sorted CD4⁺GITR⁻ T cells with high BDC2.5 TCR clonotype expression from NOD.BDC2.5 or BIM-deficient NOD.BDC2.5 mice transferred diabetes in all recipients tested (Fig. 5C). These experiments demonstrate that both Tregs and effector T cells in NODBim^−/− mice are functional, consistent with the notion that protection from diabetes in NODBim^−/− mice is due to increased antigen-specific Tregs.

Thymic deletion of developing T cells requires high-avidity TCR signaling. Of note, during their development in the thymus, Tregs are selected by stronger TCR signals compared with conventional CD4⁺ T cells (27,31). BIM deficiency may raise the threshold of TCR signaling that developing T cells in the thymus can survive, enabling the maturation of highly self-reactive T cells that would not normally be produced. Consistent with this hypothesis, we detected abnormally increased percentages of double-positive (DP) thymocytes expressing CD69 (an indicator of recent TCR stimulation) in NODBim^−/− mice (Fig. 6A), indicating that more DP thymocytes survived after TCR ligation in NODBim^−/− mice than in NOD mice. Thymocytes receiving strong TCR signals have been shown to rapidly upregulate CD25 and GITR, and these cells are enriched for Treg precursors (32). We found that the Treg precursors (GITR and CD25 expressing DP thymocytes) were markedly increased in NODBim^−/− mice (Fig. 6A), indicating that the antigen-experienced surviving thymocytes in NODBim^−/− mice might be diverted to develop into Tregs.

To deduce the TCR signal strength received by NODBim^−/− thymocytes, we measured the expression of Nur77 by flow cytometry. The Nr4a family protein Nur77 is required for thymic deletion as well as for FoxP3 expression and Treg differentiation in thymocytes following TCR stimulation (33). Nur77 expression levels are directly proportional to the strength of TCR stimulation (31). Thymocytes from NODBim^−/− mice had increased expression of Nur77 compared with their counterparts from control NOD mice (Fig. 6B and C and Supplementary Fig. 3A). This is consistent with the notion that BIM deficiency allows escape of thymocytes bearing high-affinity self-antigen–specific TCRs. It has been reported that increased TCR signaling is a requirement for natural killer T cells, a population that has been shown to be defective in NOD mice. We did not see increased numbers of natural killer T cells in NODBim^−/− mice compared with NOD mice (Supplementary Fig. 3C).

Following TCR stimulation, activation of the NF-κB pathway is critical for induction of FoxP3 expression and Treg differentiation (34,35). Thymocytes from BIM-deficient mice displayed augmented levels of phosphorylated IκB-α compared with their counterparts from NOD mice (Fig. 6D and Supplementary Fig. 3B), consistent with the notion that the cells surviving deletion are diverted to Tregs through NF-κB activation.

We studied the impact of BIM deficiency on thymic deletion, Treg development and function, and spontaneous autoimmunity in NOD mice. BIM deficiency increased autoreactive T cells in NOD mice by impairing thymocyte deletion. Unexpectedly, these mice were completely protected from diabetes, although the insulitis observed indicated that some autoimmunity remained. Thymocytes from NODBim^−/− mice that expressed an autoreactive TCR survived strong TCR signaling and showed preferential differentiation into Tregs. These cells expressed high levels of Nur77, a transcriptional regulator that plays a key role in Treg development by translating strong TCR signals induced by self-antigens into the FoxP3 developmental program. They also had increased activation of NF-κB, a transcriptional program that is essential for Treg development.

Deficiency of another proapoptotic BH3-only protein, PUMA, in addition to loss of BIM, was reported to further impair deletion (19). In contrast to Bim^−/− mice, Bim^−/−Puma^−/− mice on a C57BL/6 background spontaneously develop organ-specific autoimmune pathology, with leukocyte infiltration in the salivary gland, liver, lung, and pancreas and serum autoantibodies to these tissues. However, despite the almost complete block observed in thymocyte deletion, the autoimmunity in C57BL/6.Bim^−/−Puma^−/− mice is not fatal. The present data, in conjunction with the observation that Bim^−/−Puma^−/− mice also have more Tregs than Bim^−/− mice (7), suggest that increased Tregs substantially restrain autoimmune disease in these settings. The current study shows that this increase in Tregs involves the diversion of thymocytes with autoreactive TCRs and reveals for the first time to our knowledge that this process can prevent autoimmune disease.

We found that diabetes protection is complete when Treg number is further increased by BIM deficiency in NOD.BDC2.5 mice and that the intermediate Treg increase and diabetes protection in NODBim^+/− mice suggest that diabetes protection in NOD mice is due to expansion of Tregs. Additionally, it is noteworthy that the phenotype of NODBim^−/− is similar to other mouse models in which protection from diabetes is predominantly due to Tregs, such as in the NOD.BDC2.5 mice and in NOD mice deficient in tumor necrosis factor receptor 1 (30,36). In these models, like ours, some insulitis develops, indicating the initiation of autoimmune responses. However, Tregs prevent the progression of autoimmune attack by acting in the draining PLNs or directly in the islets, thereby protecting from diabetes (37,38).

FoxP3⁺-expressing Tregs in NODBim^−/− mice expressed lower levels of CD25 than those from control NOD mice. This has also been observed in BIM-deficient mice on a C57BL/6 background (7). Developing Treg precursors in the thymus upregulate CD25 and GITR as early as the DP (CD4⁺CD8⁺) stage (31,39). It was hypothesized that DP precursors that upregulate CD25 have a survival advantage due to increased affinity for interleukin 2. Transforming growth factor-β might also be protective to Tregs experiencing strong TCR signals from self-MHC and, thus, may be another key signal in the discrimination between clonal deletion and clonal diversion (40). BIM-deficient thymocytes are intrinsically excellent survivors, even in the absence of upregulation of CD25, and these cells are diverted to become Tregs after they encounter self-antigens in the thymus. Therefore, thymocytes that encounter high-affinity antigens are deleted unless they are saved either by interleukin 2 and transforming growth factor-β or, in this case, by inhibition of apoptosis due to loss of BIM. Consistent with this notion, thymic FoxP3⁺ Tregs in transgenic mice that express constitutively active IκB-α with a resultant increased NF-κB activity (41) did not express CD25 in the thymus because of the prosurvival function of NF-κB. Potentially, loss of BIM could also promote Treg survival in the periphery and be a strategy to enhance Treg therapy.

Immunological tolerance is maintained in normal individuals by a delicate balance between effector T cells and Tregs. In NOD mice, a combination of several defects in tolerance mechanisms, including failure to eliminate autoreactive T cells, results in diabetes. Several studies have proposed that this is due to either resistance to clonal deletion, perhaps as a result of a partial defect in BIM, or increased competition for positively selecting niches in the thymus (42–48). The present study shows that complete BIM deficiency in NOD mice substantially impairs deletion of diabetogenic autoreactive, insulin-specific T cells but that the health of the host is preserved because autoreactive T cells that survive deletion differentiate into antigen-specific Tregs and thereby maintain immune tolerance. BIM promotes Tregs from thymic generation to peripheral survival of Tregs. Thus, modulating apoptosis may be one of the ways to increase generation and survival of antigen-specific Tregs to prevent autoimmune disease and transplant rejection.

Acknowledgments. The authors thank L. Mackin (St. Vincent's Institute, Melbourne, Australia) for help with genetic analysis; A. Irvin, S. Fynch, L. Elkerbout, N. Sanders, D. Novembre-Cycon, R. Branch, and A. Gomes (all from St. Vincent's Institute, Melbourne, Australia) for help with mice and cytometry; D. Mathis (Department of Microbiology and Immunobiology, Harvard Medical School, Boston, MA) for the NOD.BDC2.5 mice; P. Bouillet (The Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia) for Bim^−/− mice; O. Kanagawa (Akashi City Hospital, Akashi, Japan) for the BDC2.5 clonotypic antibody; P. Santamaria (Julia McFarlane Diabetes Research Centre, Snyder Institute for Chronic Diseases, Faculty of Medicine, University of Calgary, Calgary, Alberta, Canada) for the NOD.8.3 mice; J. Lin (University of Melbourne, Melbourne, Australia) for help with MHC class I tetramer production; and the National Institutes of Health Tetramer Core Facility for MHC class II tetramers.

Funding. This work was supported by grants and fellowships from the National Health and Medical Research Council of Australia, JDRF, the Leukemia and Lymphoma Society of America, and the Operational Infrastructure Support Scheme of the Government of Victoria.

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

Author Contributions. B.K. contributed to designing and performing the experiments and to the writing of the manuscript. J.C., G.J., P.T., T.C., C.S., E.N.G., K.L.G., J.A.W., and Y.Z. performed experiments. T.C.B., D.G., and A.S. contributed to the writing of the manuscript. J.A. contributed to designing and performing the experiments. H.E.T. and T.W.H.K. supervised the study and contributed to the writing of the manuscript. B.K. and T.W.H.K. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.