Insulin Treatment in Patients With Type 1 Diabetes Induces Upregulation of Regulatory T-Cell Markers in Peripheral Blood Mononuclear Cells Stimulated With Insulin In Vitro

Our study included 33 patients with newly diagnosed type 1 diabetes. In 12 of 33, the blood sample was obtained at diagnosis before starting exogenous insulin therapy (group 0), whereas 21 patients had been treated with insulin for 1–21 days (median 8 days, group 1). Table 1 shows sex distribution, age, blood glucose concentration at the time of sampling, and HbA_1c (A1C) values 5–6 months after the diagnosis in the patients in both groups. During the same period, we also studied 10 (7 males) nondiabetic children without acute infections or autoimmune diseases as control subjects. The median age in the control group was 6.7 years (range 1.2–15.8).

In addition, 12 recently diagnosed type 1 diabetic patients (4 male) and 4 healthy children (3 male) were included in the analyses of CD4+CD25+ cells and Foxp3-expressing cells by means of flow cytometry. In this series, the median age of the patients was 4.6 years (range 1.1–12.8) in group 0 (n = 7) and 6.6 years (3.6–13.2) in group 1 (n = 5). The median age of the control children included in the flow cytometric analyses was 7.5 years (3.5–10.1). The study protocol was approved by the ethical committees of the participating hospitals (Hospital for Children and Adolescents, University of Helsinki, Helsinki, Finland, and Jorvi Hospital, Espoo, Finland).

PBMCs were isolated from heparinized blood by Ficoll-Paque density gradient centrifugation (Amersham Pharmacia Biotech, Uppsala, Sweden) and were suspended in RPMI-1640 containing 5% human AB+ serum (Finnish Red Cross Blood Transfusion Service, Helsinki, Finland) and 25 μg/ml gentamicin (Life Technologies, Paisley, Scotland). PBMCs were cultured in U-bottomed 96-well cell culture plates, 2 × 10⁵ cells (200 μl) per well in triplicate wells with or without antigens. Twenty micrograms per milliliter tetanus toxoid (National Public Health Institute, Helsinki, Finland), 300 μg/ml bovine insulin (purified from bovine pancreas, cell culture tested; Sigma, St. Louis, MO), and 300 μg/ml human insulin (yeast recombinant protein; Boehringer Mannheim, Mannheim, Germany) were used as antigens. After incubation for 72 h, the supernatants and the cells were collected and frozen. The cell pellets were frozen in the lysis buffer of the RNA kit (see methods in quantitative rt-pcr).

RNA was extracted from frozen cell pellets with the GenElute Mammalian Total RNA Miniprep kit (Sigma). Reverse transcription was performed in a total volume of 20 μl by using TaqMan Reverse Transcription Reagents (Applied Biosystems, Foster City, CA). Before adding multiscribe reverse transcriptase enzyme, the reaction mixture was treated with 0.01 units/μl DNase (Boehringer Mannheim). Real-time quantitative PCR was performed using TaqMan PDAR (predeveloped assay reagents) or assays-on-demand primers/probes and the ABI Prism 7700 Sequence Detection System (Applied Biosystems) in triplicate wells. PDAR primers/probes for IFN-γ (cat. no. 4327052), IL-5 (4327039), IL-4 (4327038), and TGF-β (4327054) and assays-on-demand for Foxp3 (Hs00203958_m1), CTLA-4 (Hs00175480_m), and ICOS (Hs00359999_m1) were used.

Ribosomal 18S RNA was used as an endogenous control (PDAR 4310893E and assays-on-demand Hs99999901_s1). On every plate, the expression of the same marker mRNA was measured in a calibrator sample that was prepared by stimulating PBMCs from a healthy volunteer with phytohemagglutinin (Sigma) for 48 h. The reaction mixture contained TaqMan Universal PCR Master Mix, 1 × primers/probes (both from Applied Biosystems), and 1.8 μl of template cDNA in 25 μl total volume. Thermal cycling conditions were 2 min at 50°C, 10 min at 95°C, and 50 cycles of 15 s at 95°C and 1 min at 60°C.

The quantities of the cytokines were analyzed by a comparative threshold cycle (C_t) method (as recommended by Applied Biosystems). ΔC_t stands for the difference between C_t of the marker gene and C_t of the 18S gene, whereas ΔΔC_t is the difference between the ΔC_t of the analyzed sample and ΔC_t of the calibrator. Calculation of 2^−ΔΔCt then gives a relative amount of the analyzed sample compared with the calibrator, both normalized to an endogenous control (18S). The relative amount (2^−ΔΔCt) of IFN-γ and CTLA-4 was multiplied by 1,000, the relative amount of Foxp3 and ICOS by 100, and the relative amounts of IL-5, IL-4, and TGF-β by 10 to get whole numbers for plots. The relative amount of each marker detected in the wells cultured without antigen was subtracted from the relative amount of respective marker detected in the antigen-stimulated wells.

Secreted IFN-γ, IL-5, IL-4, IL-2, and IL-10 were measured with a flow cytometry–based cytometric bead array kit (cat. no. 550749; BD Pharmingen, San Diego, CA) according to the manufacturer’s instructions. The amount of each secreted cytokine in the wells cultured without antigen was subtracted from the amount of respective cytokine expressed in the antigen-stimulated wells.

PBMCs were cultured with or without the antigens for 5–6 days in 24-well cell culture plate, 2 × 10⁶ cells (2 ml) per well, in order to detect antigen-induced Foxp3 expression in flow cytometric analysis. The staining protocol was adapted from the article by Roncador et al. (18). Briefly, the cells were fixed with 1% paraformaldehyde and 0.05% Tween-20 PBS overnight at 4°C. Cells were treated twice for 30 min with 100 units/ml DNase (Roche Diagnostics, Mannheim, Germany). Before staining, the cells were blocked with 1% goat IgG (Caltag, Burlingame, CA) in staining buffer that contained 3% FCS, 0.5% Tween 20, and 0.05% Na-azide in PBS. Mouse anti-human Foxp3 antibody (clone 150D/E4 cell culture supernatant was a kind gift from Dr. Alison H. Banham, University of Oxford, Oxford, U.K.) was diluted (1:50) in blocking buffer and incubated for 1 h at room temperature. Cell culture supernatant with irrelevant mouse monoclonal antibody was used as a control antibody. Alexa fluor 488–labeled goat anti-mouse IgG (Molecular Probes, Leiden, the Netherlands) was used as a secondary antibody in 1:500 dilution. Before cell surface staining, the cells were blocked with 1% mouse IgG (Caltag) for 20 min. CD4-PerCP (Becton Dickinson, San Jose, CA) and CD25-PE (Miltenyi Biotech, Bergisch Gladbach, Germany) were used for surface staining at room temperature for 20 min. Cells were analyzed by FACSCalibur and CellQuest Pro software (Becton Dickinson).

PBMCs isolated by density gradient centrifugation were cultured in U-bottomed 96-well cell culture plate, 1 × 10⁵ cells (200 μl) per well, in quadruplicates with the same antigen concentrations as described in cell culture for rt-pcr and cytometric bead array. After incubation for 5 days, 1 μCi of tritiated thymidine (Amersham, Buckinghamshire, U.K.) was added to each well and the cells were harvested 16–18 h later to measure the incorporation of radioactivity. The proliferative response has been presented as a stimulation index calculated by dividing the median counts per minute measured from the antigen-stimulated wells by median counts per minute detected in the wells cultured without antigen.

IAAs were analyzed with a specific radioligand assay as previously described (19). The assay had a disease sensitivity of 58% and a disease specificity of 98% in the 2005 Diabetes Autoantibody Standardization Program workshop.

The nonparametric Mann-Whitney U test was used for comparisons between the groups. The Spearman rank correlation test was applied to analyze correlations between different parameters (SPSS 10.0 for Windows; SPSS, Chicago, IL). P values <0.05 were considered significant.

No differences were seen between the study groups in any immunological parameters of the unstimulated PBMCs (data not shown).

The insulin-induced expression of Foxp3-, CTLA-4–, and ICOS-specific mRNAs in PBMCs was higher in newly diagnosed children who had received insulin treatment than in children with type 1 diabetes studied before starting insulin therapy. Both bovine and human insulin induced higher expression of Foxp3 (P = 0.040 and P = 0.055, respectively, and median [means ± SD] relative amounts were 10.6 [14 ± 15] and 2.8 [6.0 ± 7.8] for bovine insulin and 4.6 [8.1 ± 11] and 0.1 [2.0 ± 3.6] for human insulin), CTLA-4 (P = 0.035; 8.1 [10 ± 10] and 0.4 [3.6 ± 5.2] for bovine insulin and P = 0.009; 2.7 [6.1 ± 10] and 0 [0.6 ± 1.1] for human insulin), and ICOS (P = 0.025; 9.1 [16 ± 25] and 6.9 [6.3 ± 3.8] for bovine insulin and P = 0.018; 6.6 [6.7 ± 2.4] and 3.6 [4.0 ± 2.0] for human insulin) in diabetic children receiving insulin therapy compared with the diabetic children before starting insulin therapy (Figs. 1, 2, and 3).

When compared with healthy children, the children who had received exogenous insulin therapy showed higher expression of Foxp3 mRNA (P = 0.002; median [mean ± SD] relative amount 4.0 [3.2 ± 2.2] for control subjects), CTLA-4 mRNA (P = 0.002; median relative amount 0.2 [1.2 ± 1.6] for control subjects), and ICOS mRNA (P = 0.005; median relative amount was 5.0 [4.9 ± 2.1] for control subjects) than the unaffected children after bovine insulin stimulation of PBMCs.

The bovine insulin–induced expression of Foxp3 mRNA correlated with the expression of bovine insulin–induced CTLA-4 and ICOS mRNAs (r = 0.78, P < 0.001 for CTLA-4; r = 0.69, P < 0.001 for ICOS). The human insulin–induced expression of Foxp3 mRNA correlated with the human insulin–induced expression of CTLA-4 and ICOS mRNAs (r = 0.71, P < 0.001 for CTLA-4; r = 0.55, P = 0.003 for ICOS). The bovine insulin–and human insulin–induced expression levels of Foxp3, CTLA-4, and ICOS mRNAs in PBMCs correlated as well (r = 0.44, P = 0.011 for Foxp3; r = 0.42, P = 0.016 for CTLA-4; and r = 0.45, P = 0.017 for ICOS).

The human insulin–induced Foxp3 mRNA expression in insulin-treated patients correlated inversely with A1C values measured 5–6 months after diagnosis (r = −0.66, P = 0.038). There was a direct correlation between bovine insulin–induced Foxp3 mRNA expression studied after treatment and A1C values of the patients determined 5–6 months after diagnosis (r = 0.74, P = 0.010). No correlation was observed between Foxp3 mRNA and the blood glucose concentrations at the time of sampling or the age of all subjects. The duration of insulin treatment in patients in group 1 did not correlate with the mRNA expression of any regulatory T-cell markers (data not shown).

Insulin stimulation of PBMCs in vitro induced an increase in the number of CD4+CD25^high cells detected with flow cytometric analysis. The proportion of bovine insulin–induced CD4+CD25^high cells among CD4+ cells varied between 0.1 and 4.7% and that of human insulin–induced cells between 0 and 0.4% in children with newly diagnosed type 1 diabetes before starting insulin treatment. Respectively, the proportion of bovine insulin–induced CD4+CD25^high cells among CD4+ cells varied between 1.2 and 9.0% and that of human insulin–induced cells between 0.2 and 1.7% in insulin-treated children. The proportion of bovine insulin–induced CD4+CD25^high cells among CD4+ cells varied between 0.8 and 2.6% and that of human insulin–induced cells between 0 and 1.7% in the control children. The cells cultured without antigen showed very few CD4+CD25^high cells among CD4+ cells in all children (median 0.1%, mean 0.3%).

The expression of Foxp3 in CD4+CD25^high cells was studied by flow cytometry in three patients with newly diagnosed type 1 diabetes. Foxp3 expressing cells among CD4+CD25^high cells varied between 42.7 and 57.9% after in vitro bovine insulin stimulation (Fig. 4A and B). Also, human insulin induced CD4+CD25^high cells and approximately half of these cells (52.0–62.5%) expressed Foxp3 as well (Fig. 4C), although the absolute number of CD4+CD25^high cells induced by human insulin was lower than that induced by bovine insulin.

No differences in cytokine activation were observed between the patients studied before or after starting insulin treatment. Both groups of children with newly diagnosed type 1 diabetes, tested either before or after starting exogenous insulin therapy, had higher expression of IL-4 mRNA than unaffected children in response to stimulation with bovine insulin (P = 0.007 for group 0 and P = 0.002 for group 1; median [mean ± SD] relative amounts were 5.9 [5.0 ± 3.1] for group 0, 6.1 [21 ± 38] for group 1, and 0 [0.9 ± 1.8] for control subjects) and with human insulin (P = 0.033 for group 0 and P = 0.057 for group 1; medians were 1.3 [2.2 ± 1.7], 2.3 [4.5 ± 5.6], and 0 [0.6 ± 0.9]) (Fig. 5).

Also, the secretion of IL-4 by insulin-stimulated PBMCs was higher in insulin-treated patients than in healthy children. The recent-onset patients who had received exogenous insulin had higher levels of IL-4 than the unaffected children in response to bovine insulin (P = 0.014; median [mean ± SD] levels were 0.8 [1.0 ± 1.0] and 0 [0.1 ± 0.2] pg/ml). The levels of IFN-γ–, IL-5–, or TGF-β–specific mRNAs or secreted IFN-γ, IL-5, IL-2, or IL-10 did not differ between the groups after insulin stimulations.

The groups did not differ in the quantities of tetanus toxoid–induced cytokine or regulatory T-cell marker–specific mRNAs (Table 2). The children with newly diagnosed type 1 diabetes tested before insulin therapy had lower levels of secreted IFN-γ in response to tetanus toxoid than the nondiabetic children (P = 0.033; medians [means ± SD] were 53.4 [135 ± 160] and 642 [1,012 ± 1,069] pg/ml). No differences could be observed between the groups in response to tetanus toxoid in IL-4, IL-5, IL-2, and IL-10 secretion.

The two groups of patients with type 1 diabetes and the unaffected children did not differ in their T-cell proliferation response after stimulation with bovine or human insulin or with tetanus toxoid.

The proliferation response to bovine insulin correlated with the quantities of bovine insulin–induced IFN-γ or IL-5 at mRNA (r = 0.54, P = 0.002 for IFN-γ and r = 0.62, P < 0.001 for IL-5) and at protein level (r = 0.61, P = 0.001 for IFN-γ and r = 0.65, P < 0.001 for IL-5) in the total series.

The bovine insulin–induced stimulation indexes correlated with the bovine insulin–induced expression of Foxp3 and CTLA-4 mRNAs (r = 0.51, P = 0.003; r = 0.43, P = 0.016) (Fig. 6A shows the correlation between stimulation index and Foxp3 mRNA expression.) In contrast, the human insulin–induced expression of IFN-γ, IL-5, Foxp3, or CTLA-4 did not correlate with the proliferation response induced by human insulin (Fig. 6B shows the correlation between stimulation index and Foxp3 mRNA expression), but there was an inverse correlation between the quantity of human insulin–induced TGF-β mRNA and the proliferation response induced by human insulin (r = −0.41, P = 0.029). The proliferative response to bovine and human insulin correlated with each other (r = 0.52, P = 0.002).

The levels of IAAs correlated with the quantities of IL-4 and TGF-β mRNAs in response to human insulin in patients with newly diagnosed type 1 diabetes (r = 0.46, P = 0.027, n = 23 for IL-4 and r = 0.60, P = 0.003, n = 23 for TGF-β). The proliferative response induced by human insulin, as well as the age of the patients, inversely correlated with the levels of IAAs (r = −0.42, P = 0.044, n = 24 for proliferation and r = −0.45, P = 0.008, n = 34 for age).

The exogenous insulin treatment in children with newly diagnosed type 1 diabetes resulted in increased expression of regulatory T-cell–related markers Foxp3, CTLA-4, and ICOS after in vitro stimulation of PBMCs with insulin. No effect of insulin treatment was seen on the characteristics of immune responses to an unrelated antigen, tetanus toxoid. We had a unique possibility to study children with newly diagnosed type 1 diabetes before the start of insulin treatment. The expression of regulatory T-cell markers Foxp3, CTLA-4, and ICOS mRNAs in response to in vitro insulin stimulation was lower in patients with type 1 diabetes who had not received insulin therapy. This suggests that exogenous insulin expands or activates the population of insulin-specific regulatory T-cells in the periphery. The phenomenon of autoantigen-induced activation of regulatory T-cells in the periphery has been reported in experimental animal studies (20,21), but to our knowledge this is the first report indicating that it may occur also in humans.

Naturally arising Foxp3-expressing CD4+CD25^high regulatory T-cells are produced in the thymus (11,12). These endogenous regulatory T-cells specifically express Foxp3, which encodes a transcription factor responsible for the development and function of regulatory T-cells (13–15). Transduction of Foxp3 suppresses IL-2 production but upregulates the expression of regulatory T-cell–associated molecules, such as CD25 and CTLA-4. The X-linked immunodeficiency syndrome IPEX is caused by mutations in Foxp3 and is associated with autoimmune disease in multiple endocrine organs (type 1 diabetes and thyroiditis), inflammatory bowel disease, and allergies (22). Thus, natural regulatory T-cells play a key role in the control of immune responses to self and nonself in humans. The T-cell receptor repertoire of natural regulatory T-cells is skewed toward recognizing complexes of self-peptide and MHC expressed in thymus and periphery (23). Insulin is expressed as a self-peptide in thymus (24), and thus insulin-specific regulatory T-cells should be present in humans. It has been shown that regulatory T-cells undergo antigen-specific proliferation in vivo after antigenic stimulation (25). Based on these observations, we hypothesized that treatment of diabetic children with a disease-specific autoantigen, human insulin, induces the expansion and/or activation of autoantigen-specific regulatory T-cells in vivo. To detect this, we measured the mRNA levels of regulatory T-cell–associated molecules after in vitro stimulation of PBMCs with insulin. The blood volume received from the children, especially at diagnosis of type 1 diabetes before the start of insulin therapy, limited the number of PBMCs available for our studies. For this reason, analyses at single-cell level were not possible. The upregulation of the CD4+CD25^high cell population expressing Foxp3 at protein level was confirmed by flow cytometry in a subgroup of the children.

As a control antigen, we used tetanus toxoid, an antigen unrelated to type 1 diabetes. The increase in the expression of regulatory T-cell–related markers was seen after insulin treatment only when PBMCs were stimulated with human or bovine insulin and not after tetanus toxoid stimulation, which supports the interpretation that in vivo insulin stimulation specifically results in an expansion of insulin-reactive Foxp3-expressing regulatory T-cells. An increased expression of CTLA-4 and ICOS in in vitro insulin-stimulated PBMCs was also seen in the patients on insulin therapy. CTLA-4 is constitutively expressed by regulatory T-cells but is also expressed after activation in other T-cells. CTLA-4 blockade prevents regulatory T-cell activation (26), and CTLA-4 may be an important molecule to signal activation in regulatory T-cells together with the T-cell receptor. CTLA-4 may also directly mediate suppression by interacting with CD80/86 on responder T-cells.

Similarly to CTLA-4, ICOS is a member of the family of CD28 proteins (27). It is expressed in both CD4 and CD8 T-cells after their activation. ICOS regulates T-cell differentiation and cytokine production and provides critical signals for immunoglobulin production and class switching. In a mouse model of autoimmune diabetes, it has been shown that regulatory T-cells are present in the insulitis during the pre-diabetic stage and that their regulatory activity is dependent on the presence of ICOS (28). The expression of these different regulatory T-cell markers, Foxp3, CTLA-4, and ICOS, strongly correlated in our assay.

Since human insulin is expressed in thymus as a self-antigen, naturally arising regulatory T-cells in humans are primarily specific to human insulin. We also studied the response to bovine insulin for several reasons. Bovine insulin is encountered in the diet, and the regulatory mechanisms specific to bovine insulin may thus be induced in the intestinal immune system (i.e., peripherally [29]). Bovine insulin differs from human insulin by three amino acids, which is sufficient to make it more immunogenic in humans as observed when patients with type 1 diabetes were treated with bovine insulin (30). The increase in Foxp3 and IL-4 expression was more pronounced after bovine insulin stimulation than with human insulin. Due to the limited difference of only three amino acids between these two insulin molecules, the immune response cross-reacts (31,32). In our study, a strong correlation was seen between human insulin–and bovine insulin–induced regulatory T-cell markers. However, we found differences in the immune response to human and bovine insulin. Insulin from different species may induce functionally different immune responses at the cellular level, as suggested by studies in animal models (33,34). In vitro stimulation of PBMCs with insulin activates both effector T-cells and regulatory T-cells present in the culture wells. The Foxp3 activation induced by bovine insulin did not suppress, but rather correlated with, the proliferation of the effector cells in vitro and showed a direct correlation with A1C after diagnosis. Bovine insulin seemed to induce regulatory T-cells de novo as a consequence of immune activation. These kind of regulatory T-cells may not be as suppressive as the thymus-derived regulatory T-cells (11).

Instead, in vitro human insulin–induced Foxp3 response showed an inverse correlation with subsequent A1C levels, suggesting a link to the recovery of β-cell function after the diagnosis. However, we could not find any correlation between the duration of insulin treatment and the expression of Foxp3 mRNA or other regulatory T-cell markers after in vitro insulin stimulation (r = 0.07, P = 0.763 for Foxp3 mRNA after bovine insulin stimulation), which may indicate that the activation of regulatory T-cells occurs already soon after the start of insulin therapy. On the other hand, the number of individuals studied here was relatively small, and thus the results should be interpreted cautiously. In nonobese diabetic mice, it has been shown that an expansion of CD4+CD25^high regulatory T-cells in the pancreatic islets (35) and in vitro (21) inhibits the presentation of overt diabetes. As exogenous insulin treatment does induce a regulatory T-cell response to insulin, it might contribute to the remission period. A recently published work (36) reported that in vivo challenge with food allergens induced CD4+CD25^high regulatory T-cells in children who had become tolerant to the allergen.

The insulin-induced activation of a regulatory T-cell population together with insulin-specific IL-4 increment may explain, at least partly, the limited capacity of T-cell assays to detect insulin-specific responses in patients with type 1 diabetes (37–40). We could not see any difference in T-cell proliferation in response to insulin between the patients and the unaffected children, and similar findings have been reported by other study groups (41–43). An inverse correlation has been reported between IAAs and insulin-induced T-cell proliferation (4,42) and, indeed, was also seen in our study. Notably, the IAA levels of the patients correlated with insulin-induced IL-4 and TGF-β mRNAs, cytokines that inhibit cell proliferation.

Our finding of IL-4 upregulation in response to insulin in children with type 1 diabetes, although seemingly in conflict with the concept of type 1 diabetes as a Th1-mediated disease, is in agreement with the observation of IL-13–secreting insulin-specific T-cell lines derived from the pancreatic lymph nodes of patients with type 1 diabetes (44). Although the highest increase in insulin-induced IL-4 expression was seen in recent-onset patients receiving insulin therapy, the increased insulin-specific IL-4 expression at the level of mRNA was also observed in children with type 1 diabetes before starting insulin treatment. This suggests that an insulin-specific immune response of Th2 type is also associated with the autoimmune response to endogenous insulin in patients with type 1 diabetes and that exogenous insulin further enhances such reactivity. This observation is in accordance with previous publications reporting IL-4 and IL-10 secretion in response to preproinsulin epitopes in individuals with type 1 diabetes–associated autoantibodies (45) and increased mitogen-induced secretion of IL-4 by insulin from PBMCs (46). Also, the occurrence of IgG4-subclass IAAs in nondiabetic autoantibody-positive children (47) is in agreement with our finding of insulin-induced IL-4 activation, a cytokine inducing IgG4 production. In addition, Achenbach et al. (48) observed that autoantibodies of IgG1 subclass represent low predictive value for type 1 diabetes, and the risk increases when the response to β-cell autoantigens spreads to include IgG2 and IgG4 subclasses. It is also noteworthy that IL-4 activates the regulatory T-cell population (49).

The present observations suggest that insulin treatment induces activation of insulin-specific regulatory T-cells in patients with recent-onset type 1 diabetes. These results in humans, indicating that the administration of a thymus-expressed self-antigen activates regulatory T-cells, support the concept of the induction and expansion of regulatory T-cells by their antigen in the periphery. This effect of insulin may interfere with the insulin-specific T-cell assays performed in patients with type 1 diabetes. Furthermore, the enhancement of regulatory T-cell responses to insulin induced by insulin treatment may play a role in the induction of the remission phase. The transient nature of the remission suggests defects in the mechanisms of T-cell regulation as reported in patients with type 1 diabetes (50).

This work was supported by the Academy of Finland, the Juvenile Diabetes Research Foundation International, the Sigrid Juselius Foundation, and the Finnish Cultural Foundation.

We thank Dr. Alison H. Banham (University of Oxford, U.K.) for providing us the Foxp3 antibody. We are also grateful to Kristiina Luopajärvi, MD, for compiling the data on blood glucose concentrations and A1C values in the patients. We thank Harry Lybeck, Annika Cederlöf, Sinikka Tsupari, and Riitta Päkkilä for technical assistance.