Following the Fate of One Insulin-Reactive CD4 T cell: Conversion Into Teffs and Tregs in the Periphery Controls Diabetes in NOD Mice

BDC12-4.1 TCR Tg mice were previously described (11). Heterozygous Foxp3^wt/mut C57BL/6 mice I-A^b.RAG⁺ females (strain 4088) were purchased from The Jackson Laboratory (Bar Harbor, ME). A minimum of two crosses with NOD.BDC12-4.1.RAG^KO background was needed to generate I-A^g7/g7.BDC12-4.1.Foxp3^mut.RAG^KO animals. In all experiments, Tg littermates served as genetic controls. Mice were maintained in La Jolla Institute for Allergy and Immunology animal facility or at the University of Colorado Health Sciences Center for Laboratory Animal Care in Denver under specific pathogen-free conditions according to each institution’s animal care and use committee guidelines for the use and care of laboratory animals.

Genotyping for the BDC12-4.1 Tg TCRα and TCRβ, RAG, and I-A typing were performed as described previously (11). The presence of the Foxp3^mut allele could be detected by PCR (forward primer, TCA GGC CTC AAT GGA CAA AA; reverse primer, CAT CGG ATA AGG GTG GCA TA) with a band size near 1,000 base pairs (bp). Females were classified as either heterozygous Foxp3^wt/mut or homozygous Foxp3^mut/mut by microsatellite typing with 6-carboxyfluorescein (FAM)-conjugated forward primer, ATG AGA AGAAGG AAG ATC AGCG, and an unlabeled reverse primer, ACC TGG GAA GGA ACT ATT GC. The microsatellites were analyzed with a real-time PCR ABI 3100 Avant Analyzer (Foster City, CA). The mutant peaks appear at 344 bp and the wt peaks are at 342 bp.

Mice were followed for diabetes by weekly testing of tail blood using either the Freestyle glucometer and test strips from Abbott Laboratories (Abbott Park, IL) or the ReliOn Ultima system (Alameda, CA). Mice whose glucose values exceeded 250 mg/dL were tested the next day and were considered diabetic if the second reading also exceeded 250 mg/dL.

Cells were washed in fluorescence-activated cell sorter (FACS) staining buffer (0.1% sodium azide, 1 mmol/L EDTA, and 2% FCS in PBS) and incubated with 10 μl/mL Fc block (2.4G2) (BD-Pharmingen, San Diego, CA) for 10 min. After the blocking step, cells were stained for CD4, CD8α, CD44, CD62L, CD69, CD25 and CD127, killer cell lectin-like receptor subfamily G (KLRG)-1, CD103, V β 2 (Vβ2) (BDC12-4.1), or V β 4 (Vβ4) (2H6) (eBioscience or BD-Pharmingen, San Diego, CA). For intracellular Foxp3, Helios, or Ki-67 detection, cells were fixed with Foxp3 Fix/Perm buffer and stained with the corresponding monoclonal antibodies (eBioscience). For cytokine detection, splenocytes were activated in vitro for 4 h with anti-CD3/28 (10/5 μg/mL, respectively) or phorbol myristic acid/ionomycin (50/1000 ng/ml, respectively) in the presence of brefeldin A (BfA, 10 μg/mL) (Sigma-Aldrich, St. Louis, MO). Cells were costained for Foxp3, interferon-γ (IFN-γ), and interleukin-10 (IL-10) after fixation with the Foxp3 Fix/Perm kit. All antibody incubations were performed at 4°C for 30 min (isotype controls were included). Cells were immediately acquired on a FACSCalibur or LSRII flow cytometer (BD-Pharmingen) and analyzed using FlowJo software (Treestar).

Pancreas sections (6-μm) were cut by cryostat and dried overnight. For positive control, spleen sections were included. Upon rehydration with TBS-PI (Tris-buffered saline containing protease inhibitor, mini EDTA-free tablets; Roche) and fixation (0.4% paraformaldehyde, 15 min at room temperature [RT]), sections were blocked in 5% BSA and 2% goat serum TBS-PI (30 min RT). To block endogenous biotin activity, the avidin/biotin blocking kit from Vector was used according to the manufacturer’s instructions. As primary antibodies, anti–CD4-FITC (clone RM4-5, 1/50 dilution; BD-Pharmingen), guinea pig anti-swine insulin (1/100 dilution; Dako), and biotinylated anti-Foxp3 (clone FJK-16S, 1/50 dilution; eBioscience) were used. Primary antibody incubations were carried for 1.5–2 h at RT in a damp chamber. For detecting CD4, the Alexa Fluor 488 signal amplification kit for fluorescein dyes (Invitrogen) was used according to the manufacturer’s instructions. For detecting Foxp3 and insulin, Alexa Fluor 647–coupled streptavidin (1/1,000 dilution; Invitrogen) and Alexa Fluor 594–labeled polyclonal goat anti–guinea pig IgG antibody (1/1,000 dilution; Molecular Probes) were added overnight at 4°C or for 1.5 h at RT, respectively. Finally, nuclei were stained with Hoechst 33258 dye (1/200 dilution, 20 min; Invitrogen). After washing, sections were mounted with SlowFade Gold mounting reagent (Invitrogen), and coverslips were sealed with nail polish. Images were recorded using FluoView FV10i confocal microscope (Olympus), and analysis was performed with Fluoview ASW2.1 (Olympus) and Image Pro Analyzer 7.0 (Media Cybernetics) software packages.

Purified, FACS-sorted CD4SP T cells from the thymi of BDC12-4.1.RAG^KO mice were cultured in 96-well plates at 0.25 × 10⁶ cell/well in the presence of 0.5 × 10⁶/well of T cell–depleted splenocytes (TDS) as antigen-presenting cells and insulin B:9-23 peptide sequence 1 (PHLVEALYLVCGERG) or 2 (SHLVEALYLVCGERG) at 50 μg/mL. TDS consisted of spleens magnetically depleted of CD90.2⁺ T cells with the use of αCD90.2 antibody-coated microbeads (Miltenyi Biotech Inc., Auburn, CA) from NOD.Ins2^KO or Ins1^KO donor mice. Supernatants were collected 3 days after culture for IL-4, transforming growth factor-β1 (TGF-β1), IFN-γ, and IL-10 detection. Cytokines present in the supernatants were quantified by a sandwich enzyme-linked immunosorbent assay (ELISA) using standard commercially available kits (Pharmingen OptEIA mouse; BD-Pharmingen).

Spleens were extracted under aseptic conditions and homogenized by grinding between two glass microscope slides or passing through cell strainers. Erythrocytes (red blood cells) were lysed using red blood cell lysis buffer (R7757) (Sigma-Aldrich) or ACK buffer. Splenocytes were washed in sterile PBS three times. Cells were resuspended in the appropriate volume of sterile PBS before transfer. Varying amounts of primary, unsorted splenocytes (1.5 × 10⁵ to 6 × 10⁶ cells) were injected intraperitoneally in 200 μL into 6–8-week-old I-A^g7.C57BL/6.RAG^KO or NOD.RAG^KO recipients.

Bone marrow (BM) was harvested under aseptic conditions from the tibias and femurs of 4–5-week-old donors in MACS buffer (PBS, 0.5% BSA, and 2 mmol/L EDTA) and washed twice. Mature B and T cells were depleted by negative selection with CD4, CD8, and CD45RB/B220 magnetic microbeads and the autoMACS Magnetic Cell Sorter (Miltenyi Biotech Inc.) according to the manufacturer’s instructions. Ten million BM cells (in 200 μL) were injected retro-orbitally into sublethally X ray–irradiated (200 rads), 6–8-week-old I-A^g7.C57BL/6.RAG^KO or NOD.RAG^KO recipients.

The statistical significance of the difference between means was determined using the two-tailed Student t test. All statistical analyses, including Kaplan-Meier survival curve analyses, were performed using GraphPad Prism version 4.01 for Windows (GraphPad Software, San Diego, CA). P values were as follows: *P < 0.05, **P < 0.01, and ***P < 0.001.

Earlier studies had shown that BDC12-4.1 mice only develop T1D on the RAG^KO background with ∼40% penetrance by 40 weeks of age, whereas RAG⁺ mice develop insulitis but no overt diabetes (11). On the RAG^KO background, no other TCRα (TCRα) chains were expressed, and only CD4 single-positive (SP) cells were able to mature in the thymus (Supplementary Fig. 1). To investigate whether Treg activity would contribute to diabetes resistance in BDC12-4.1.RAG^KO mice, we performed detailed phenotypic analysis. A significant proportion (60–70% in the spleen and 20% in the PDLNs) of the remaining CD4⁺Vβ2⁺ cells displayed marked activation, exhibiting a typical Teff phenotype (CD44^hiCD62L^low) by 7–10 weeks of age (Fig. 1A and B), whereas CD69 levels were more pronounced in the PDLNs (data not shown). Interestingly, when we searched for the presence of Tregs in BDC12-4.1.RAG^KO mice, on average, ∼3% of CD4⁺Vβ2⁺CD127^lo cells in the spleen and ∼5% in the PDLNs were CD25⁺Foxp3⁺ (Fig. 1D–F). Foxp3⁺ Tregs were also detected with immunofluorescence staining directly in the pancreatic infiltrate of BDC12-4.1.RAG^KO mice (Fig. 1G and H). Thus, BDC12-4.1 T cells can develop into Teffs as well as aTregs likely after exiting the thymus in a naïve state (see next section).

Next, we monitored the Teff and aTreg frequencies over time and found that fewer CD44^hiCD62L^low cells accumulated in the PDLNs compared with the spleen, correlating with the slightly increased aTreg frequency in that site (Fig. 1C and F). In the spleen, the number and activation status of Teffs consistently increased with age, whereas aTreg frequency remained relatively stable. In contrast, in the PDLNs, Teffs and aTregs both underwent cyclical increases and decreases.

Foxp3⁺ Treg cells can be generated in the thymus as a result of agonist recognition during thymic selection. To define the precise origin of Foxp3⁺ cells in BDC12-4.1.RAG^KO mice, we analyzed thymocytes by flow cytometry. The percentage of viable CD4/CD8 double-positive (DP) thymocytes was high, but they did not exhibit the typical characteristics of cognate antigen recognition; no CD4/CD8 coreceptor downregulation and marginal CD69 downregulation were observed (Supplementary Fig. 2). However, TCRβ levels were somewhat reduced, and a 90% reduction in mature CD4 SP cells was seen in RAG^KO compared with RAG⁺ mice. This suggests that some degree of negative selection took place, most probably at the transition from the DP to SP stage (Supplementary Fig. 2). Had insulin-specific negative selection been complete, no CD4SP cells should be detected (25).

Interestingly, no Foxp3⁺ nTregs could be detected within the CD4SP population on the RAG^KO background (Fig. 2), arguing strongly that the Foxp3⁺ Tregs detected in the periphery were in fact induced BDC12-4.1 aTregs. Our findings also suggest that the CD25⁺Foxp3⁺ cells detected in the thymus in RAG⁺ mice were the result of TCR rearrangement (pairing between the Tg Vβ2 chain with non-Tg Vα chains). These observations highlight the significance of de novo aTreg induction in enforcing tolerance. Autoreactive T cells convert into aTregs to keep Teffs of the same or other specificities under their direction.

In contrast to humans, mice express two insulin genes, Ins1 and Ins2, and the B:9-23 peptides, to which BDC12-4.1 Tg cells react, differ in their amino acid sequence at position 9 (S for INS2 and P for INS1). ELIspot analysis showed that BDC12-4.1 Tg cells produce IFN-γ upon stimulation with both B:9-23 sequences, whereas they are not specific for B24-C36 or B16:A (Supplementary Research Design and Methods and Fig. 3). We hypothesized that one of the mechanisms involved in acquiring Foxp3 expression in the periphery would include insulin recognition. Therefore, we sorted CD4SP cells from the thymus of BDC12-4.1.RAG^KO mice and placed them in culture with TDS from Ins2^KO mice loaded with either B:9-23 peptide (INS1 or 2). After 48 h stimulation with either B:9-23 INS1 (1) or INS2 (2), upregulation of CD44 in conjunction with CD62 L downregulation occurred (Fig. 3A). Most importantly, B:9-23 stimulation also induced generation of CD25⁺Foxp3⁺ aTreg (Fig. 3B), suggesting that insulin recognition can induce BDC12-4.1 T cells to acquire both effector and regulatory properties. It is striking that a T cell with a fixed TCR can convert into a Treg upon antigen encounter in vitro in the absence of exogenously added TGF-β1. ELISA analysis of the cell culture supernatants revealed generation of IL-10, IFN-γ, and IL-4 and low amounts of TGF-β1 (Fig. 3C). This TGF-β1 expression could explain the Foxp3 induction and aTreg formation in the in vitro cultures. Altogether, activation of naïve, thymically derived CD4⁺ BDC12-4.1 TCR Tg cells (RAG^KO background) in vitro resulted in the generation of effector and Foxp3-expressing T cells and the production of cytokines with proinflammatory and immunoregulatory properties.

In order to address how the Foxp3-expressing CD4⁺ T cells in BDC12-4.1.RAG^KO TCR Tg mice might exert their Treg function, we determined their cytokine profile. As shown in Fig. 4A, polyclonal activation of ex vivo splenocytes showed that non-Treg cells (CD25⁻Foxp3⁻) produced predominantly IFN-γ, whereas Treg cells (CD25⁺Foxp3⁺) coproduced IFN-γ and IL-10. Interestingly, the IFN-γ/IL-10 coproduction signature was specific for the Treg cells derived from BDC12-4.1.RAG^KO and not BDC12-4.1.RAG⁺ mice (Fig. 4B). Thus, BDC12-4.1 mice generate Foxp3⁺ aTregs that coproduce IFN-γ and IL-10 in the periphery.

Since BDC12-4.1.RAG^KO mice exhibit a strong degree of lymphopenia (11), we wondered whether a proportion of Foxp3⁺ cells were the result of homeostatic expansion. Since homeostatically expanded Foxp3⁺ cells express the CD103⁺/KLRG-1⁺ signature (26), we performed flow cytometry staining analysis for both markers after gating on the CD25⁺Foxp3⁺ (Treg) population. Indeed, as shown in Fig. 5A, >20% of the Tregs coexpressed CD103/KLRG-1 and therefore were likely derived from homeostatic expansion. We also found that Teff and Treg populations stained marginally positive for Ki-67 (Fig. 5B). In control BDC12-4.1.RAG⁺ animals, Treg renewal was more pronounced. To confirm the non–thymically derived lineage of Foxp3⁺ cells in BDC12-4.1.RAG^KO mice, we performed additional staining for the nTreg-specific marker, Helios (27). Indeed, <5% of the Treg population in BDC12-4.1.RAG^KO mice expressed Helios (Fig. 5B and C). In summary, Foxp3⁺ aTregs develop in the periphery of BCDC12-4.1.RAG^KO mice as a result of insulin recognition and homeostatic expansion.

We found a strong correlation between the age and glucose intolerance in prediabetic BDC12-4.1.RAG^KO mice (Supplementary Fig. 4). This suggested that despite the induction of aTregs, Teffs accumulated and eliminated more and more β-cells. However, only a proportion of mice eventually turn diabetic. In order to address whether Foxp3⁺ cells in the periphery were responsible for this T1D protection in 60% of BDC12-4.1.RAG^KO mice, we back-crossed Foxp3 mutant (scurfy) C57BL/6 mice to Tg BDC12-4.1.RAG^KO NOD mice for two generations (see figure legend for details). As depicted in Fig. 6A and B, spontaneous autoimmune diabetes was significantly accelerated with 100% penetrance when the scurfy Foxp3 mutation was introduced into BDC12-4.1.(I-A^g7/g7) RAG^KO mice. At least one copy of I-A^g7 was needed for diabetes development, and one copy of the I-A^b gene somewhat delayed, but did not prevent, diabetes (Fig. 6C). However, despite the presence of a single I-A^b allele, all Foxp3-mutant (mut) BDC12-4.1.RAG^KO mice still progressed to diabetes (Fig. 6C).

The above results conclusively confirmed our initial hypothesis that in BDC12-4.1.NOD.RAG^KO mice, pathogenic as well as regulatory cells coemerge, and the generation of Tregs was crucial to curb diabetes development. A set of adoptive transfer experiments was performed to confirm pathogenic and regulatory roles of BDC12-4.1 CD4⁺ T cells. We reasoned that if the scurfy mutation truly eliminated any regulation by BDC12-4.1 T cells, then adoptive transfer of total splenocytes or BM transplantation (BMT) from BDC12-4.1.RAG^KO.Foxp3^mut mice into NOD.RAG^KO or I-A^g7.C57BL/6.RAG^KO (g7/B6.RAG^KO) recipients would cause diabetes much faster than transfer of cells from BDC12-4.1.Foxp3^wt donors. As shown in Fig. 7A, BMT from BDC12-4.1.RAG^KO.Foxp3^mut donors into both NOD and g7/B6.RAG^KO recipients led to more diabetes development when the donors had the Foxp3 scurfy mutation (Foxp3^mut) than when the donors did not (Foxp3^wt). Similar observations were made when splenocytes were transferred (Fig. 7B). Interestingly, in this set of experiments, only when NOD.RAG^KO, and not I-A^g7/57.RAG^KO, were used as hosts did diabetes occur, underlining the importance of other insulin-dependent diabetes susceptibility genes for influencing disease progression. Taken together, these results provide strong evidence that the aTreg population observed in BDC12-4.1.RAG^KO TCR Tg mice provide active peripheral tolerance and control pathogenic cells of the same specificity.

Here we show that the heterogeneous and much-reduced T1D development in mice uniformly expressing a TCR specific for a critical insulin B chain epitope (BDC12-4.1.RAG^KO TCR Tg mice) was due to de novo generation of insulin-specific Tregs in the periphery. A significant proportion of insulin B:9-23–specific T cells escaped negative selection and then recognized insulin in the periphery, which led to their differentiation into Teffs and aTregs. This is the first time where a T cell with a fixed (autoreactive) TCR can follow both Teff and aTreg fates when exposed to similar differentiation niches in vivo.

In accordance, in vitro CD4SP–sorted thymocytes that responded to B:9-23 stimulation produced IL-10 and TGF-β1 and a fraction became Foxp3⁺. With increasing age, the number of Teffs increased in the spleen, whereas the number of Tregs remained relatively constant, which explains the development of T1D in a fraction of the BDC12-4.1.RAG^KO mice. Accordingly, mice became more intolerant to glucose as they aged. Insulin-specific Teffs and Tregs underwent cyclical changes in the PDLNs, which could be a sign of a more active, iterative immunological process in that location; indeed, earlier studies had described similar cyclical alterations in islet-reactive CD8 responses in NOD mice. Alternatively, insulin-specific Teffs could have migrated to the pancreas, causing the decrease in cell counts in the PDLNs. In agreement, histological examination showed the presence of few Foxp3⁺ but far more Foxp3⁻CD4⁺ cells in the pancreas, especially in islets that were massively infiltrated.

In the mouse, Foxp3 is the most specific marker for CD4⁺ Treg cells (28), and accumulating evidence suggests that Foxp3⁺ Treg can be generated from peripheral naïve or already differentiated T cells in vivo under specific conditions (29). Such aTregs have been described to develop at lower levels of antigen availability under tolerogenic priming conditions (for example at mucosal surfaces or in the presence of TGF-β1) and are favored by homeostatic expansion (23). Usually, most Foxp3⁺ Treg originate in the thymus as a result of the regular T-cell selection process (25). Thymic differentiation of naturally occurring Tregs is regulated by TCR affinity, CD28 costimulation, common γ-c cytokine (IL-2) recognition, and TGF-β1 production (30,31). Previous studies with TCR Tg mouse models, where agonist peptides were expressed in the thymus, showed that self-peptide recognition not only induces negative selection but also nTreg differentiation, which suggests that the TCR repertoire of Treg cells overlaps somewhat with that of pathogenic T cells (23,32).

Since insulin is expressed in the thymus, we hypothesized that BDC12-4.1 Foxp3⁺ nTregs may also develop. However, only on the RAG⁺ background could thymic nTregs be observed, whereas on the RAG^KO they were completely absent. We have evidence to suggest that in the BDC12-4.1.RAG⁺ background, the presence of additional TCR α chains enabled thymic Treg commitment. However, since nTregs did not develop in the thymus of RAG^KO mice, either insulin B:9-23 expression is not adequate to drive their differentiation, medullary epithelial cells lack some important costimulatory signals, or the affinity/avidity of the trimolecular TCR/peptide/I-A^g7 complex is too low in NOD mice. In the future, it would be of interest to analyze thymi from BDC12-4.1 mice on diabetes-resistant genetic backgrounds, i.e., on BALB/c.

In the periphery, where the environment is largely lymphopenic, T cells start to proliferate. Perhaps these environmental cues together with stronger insulin recognition and presentation in the PDLNs permit autoreactive T-cell activation and differentiation into Teffs and aTregs. Despite the immediate effects of the immune system on the function and viability of insulin-producing β-cells, other factors, such as androgens, could have contributed to their decline (33). This is seen by the persistent gender discrepancy in several genetic scenarios (i.e., RAG^KO, I-A^g7 homozygosity, etc.). However, in order to address how the insulin-specific aTreg modulated disease progression, we examined their cytokine production profile. aTregs in BDC12-4.1.RAG^KO mice coproduce IFN-γ and IL-10, whereas Teffs produced solely IFN-γ. It is not clear whether some IFN-γ can be important for Treg function as we and others had shown previously (34,35). To corroborate this notion, T-bet expression (the master regulator of IFN-γ expression) in Treg cells is essential for their suppressive activity in the scurfy (Foxp3-deficient) model (36). The CD103/KLRG-1 coexpression signature (26) together with the absence of high amounts of folate receptor-4 and Helios, both distinctive markers for nTregs (37) (data not shown), strongly supports the adaptive origin.

The BDC12-4.1 TCR Tg mouse is the first single-Tg, RAG-deficient animal that exhibits spontaneous in vivo development of aTregs and Teffs with identical TCR sequence. Perhaps insulin is a unique autoantigen, and presentation in the context of I-A^g7 provides a permissive signal to instruct the acquisition of aTreg and Teff phenotype in cells bearing identical TCRs. It is important to note that the heterogeneous disease observed in BDC12-4.1.RAG^KO animals, where the TCR repertoire is restricted to one insulin-specific Tg TCR, is usually not observed in other Tg or retrogenic CD4⁺ TCR models for T1D. For example, BDC2.5 mice recognizing the chromogranin autoantigen on immunodeficient backgrounds develop fulminant diabetes in 100% of the animals, suggesting that no aTreg development takes place (38,39). Of evidence, when BDC2.5.RAG^KO mice are placed onto the scurfy background, no alteration in disease kinetics is seen.

The present findings support an intriguing TCR Tg model, where autoreactive cells escaping negative selection convert in the periphery into Teffs and aTregs, where their balance in the periphery determines the degree of islet destruction and consequently diabetes development. We conclude that it might be useful in further studies to assess therapies that might “tip the balance” toward antigen-specific Tregs rather than Teffs, and monitoring their frequency and the ratio could be an important biomarker to predict disease progression after certain types of therapeutic intervention.

This work was supported by a U01 National Institute of Allergy and Infectious Diseases prevention center grant to M.v.H., and the National Institutes of Health (DK-32083), Autoimmunity Prevention Center (AI050864), Diabetes Endocrine Research Center (P30-DK-57516), Juvenile Diabetes Research Foundation (4-2007-1056 and 11-2005-15), Children's Diabetes Foundation, and Brehm Coalition support to G.E.

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

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

G.F. and J.J. designed and conducted experiments, analyzed the data, and wrote the manuscript. A.D., T.J., and Y.C. performed experiments. M.N. performed experiments, provided important guidance, contributed to the discussion of the results, and critically improved the manuscript. P.P. performed immunofluorescence staining. F.L., G.S., M.C., and H.C. provided important guidance, contributed to the discussion of the results, and critically improved the manuscript. G.E. and M.v.H. supervised the study and wrote the manuscript. G.E. and M.v.H. 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.

The authors thank Malina Mc’Clure for excellent mouse husbandry, Natalie Amirian for assisting with histological preparation and staining, and Priscilla Colby for administrative assistance (all from La Jolla Insitute for Allergy and Immunology).