Natural Variation in Interleukin-2 Sensitivity Influences Regulatory T-Cell Frequency and Function in Individuals With Long-standing Type 1 Diabetes

Mechanisms leading to type 1 diabetes (T1D) depend on a complex combination of genetics (1–3) and environmental factors resulting in the breakdown of peripheral tolerance. We and others have reported that suppression of autologous conventional CD4⁺ T cells (Tconv) by CD4⁺CD25^hiFOXP3⁺ regulatory T cells (Tregs) in individuals with newly diagnosed T1D (NDT1D) and long-standing T1D (LST1D) is reduced compared with age-matched control subjects (4–8). Although the precise reason for reduced suppressive activity has not yet been fully elucidated, several intrinsic defects in Tregs have been observed in (at least a subgroup of) individuals with T1D, including decreased interleukin (IL)-2 signaling, increased Treg apoptosis, and decreased stability of Treg FOXP3 expression (5,6,9,10). However, it is highly significant that, to date, all studies examining functional aspects of Treg biology have observed a large degree of overlap between individuals with and without T1D, with only a subgroup of patients with T1D displaying the immune phenotype associated with reduced Treg function. Furthermore, Hughson et al. (11) reported in a longitudinal analysis of Treg functions during the first year of T1D diagnosis not only that there was great heterogeneity in patient immunophenotypes but also that the time of sampling and the state of progression of the disease may also affect Treg function.

IL-2 plays a key role in the generation and maintenance of peripheral fitness and function of Tregs in mice and humans (12–16). Observations that polymorphisms in genes in the IL-2–signaling pathway associate with T1D (1–3) thus suggest that these genetic variants may alter T1D risk via effects on Treg numbers or function. In support of this, we and others have performed candidate gene-to-phenotype studies and reported that multiple T1D-associated polymorphisms in the IL-2 receptor α-chain (IL-2RA/CD25) and protein tyrosine phosphatase 2 (PTPN2) genes conferred decreased IL-2 signaling in CD4⁺CD25^hi Tregs (9,17–19). We further observed that the presence of the main T1D IL2RA susceptibility allele also associated with lower levels of FOXP3 expression in Tregs and a reduction in their ability to suppress proliferation of autologous Tconv (17).

Owing to their constitutively high expression of CD25 (20,21), Tregs display enhanced sensitivity to IL-2 compared with Tconv and require lower IL-2 levels to support their development, homeostasis, and function (17,21,22). This key characteristic underlies the use of low-dose IL-2 therapy to enhance Treg frequency and function. IL-2 administration in mouse models of autoimmunity has shown therapeutic effects (23,24) and has also shown clinical efficacy in humans with chronic graft-versus-host disease (25,26), hepatitis C virus–induced vasculitis (27), and alopecia areata (28). Therefore, there is a strong rationale for investigating IL-2 immunotherapy in human T1D (29–31). Nevertheless, the immune system of a patient with T1D is relatively normal (32–35) compared with lymphopenic patients (e.g., chronic graft-versus-host disease) or patients with other severe inflammatory conditions (e.g., hepatitis C virus–induced vasculitis). Furthermore, a recent phase I trial of IL-2/rapamycin in T1D was terminated because there was a partial decline in β-cell function (36). The doses and frequency of IL-2 dosing may have been too high in this trial, and inadvertent Tconv activation could have accelerated β-cell damage. These considerations highlight an urgent need to determine dose and frequency of dosing of IL-2 in T1D (31), including the identification of baseline characteristics of a patient’s immune system that could predict the level of response to IL-2.

Despite the interest in boosting Treg function in T1D by IL-2 administration, detailed studies directly linking IL-2 signaling with T1D-associated Treg immune phenotypes are lacking. To address this, we examined IL-2 sensitivity in CD4⁺ T-cell subsets in 70 individuals with LST1D and assessed the effect of low IL-2 sensitivity on Treg frequency and function. This study reveals extensive interindividual variation in IL-2 responsiveness in Tregs that was stable within an individual and influenced by T1D-associated gene polymorphisms. In individuals with low IL-2 signaling, Tregs, especially of the antigen-experienced subsets, were reduced in frequency. Furthermore, Tregs from these individuals were less able to maintain expression of FOXP3 under limiting concentrations of IL-2 and displayed reduced suppressor function. Our results indicate that stratification of trial participants by Treg frequency and IL-2 signaling capacity could be a useful strategy in the optimization of IL-2 therapy in T1D.

Blood samples were obtained from 18 control subjects without diabetes and 70 individuals with LST1D (>3 years postdiagnosis, <40 years of age) at two time points >3 months apart. Large blood samples from two donors were used as internal biological controls in batch analyses of IL-2 sensitivity. Peripheral blood mononuclear cells (PBMCs) were isolated from heparinized blood samples by density gradient centrifugation (Lymphoprep; Axis-Shield PoC AS, Oslo, Norway) and cryopreserved in FBS (Gibco) with 10% DMSO or used immediately for functional studies. In addition, fresh blood samples were obtained from 22 age- and sex-matched adults with LST1D, 20 control subjects without T1D, 17 individuals with NDT1D (<2 years postdiagnosis), and 15 autoantibody-negative unaffected siblings (UAS). Details of study participants are reported in Supplementary Table 1. Ethical approval for this study was granted by the local ethics committee, and informed consent was obtained.

Single nucleotide polymorphisms (SNPs) in the genes PTPN2 (rs45450798 and rs478582) and IL2RA (rs12722495 and rs2104286) were genotyped using TaqMan 5′ nuclease assays (Applied Biosystems), according to the manufacturer’s protocol.

Antibodies used in this study are detailed in Supplementary Table 2.

Phosphorylated (p)STAT5a analyses for cryopreserved PBMC samples were performed in a batch manner where each batch consisted of duplicate samples from eight individuals, including six with LST1D and two biological control subjects (Supplementary Fig. 1). The assay was performed using a violet fluorescent cell barcoding kit (BD Biosciences, San Diego, CA). Briefly, for the initial IL-2 sensitivity screen, cryopreserved PBMC samples were thawed and rested for 10 min at 37°C in X-VIVO-15 media with 1% human pooled AB⁺ sera (Sigma-Aldrich, Dorset, U.K.). PBMCs were then stimulated with 0.1, 0.25, or 10 IU/mL human (h)IL-2 (Proleukin; Novartis) for 30 min at 37°C, fixed with BD Lyse/Fix buffer for 10 min, washed in PBS, and permeabilized with prechilled (−20°C) BD Perm Buffer III for 30 min on ice. Cells were spun and resuspended in prechilled 50% BD Perm Buffer III (diluted with PBS) and incubated with the barcoding dye mixture at 4°C for 30 min. After extensive washes with barcoding wash buffer, samples stimulated with the same IL-2 concentration were combined into a single FACS tube and stained with anti–CD4-fluorescein isothiocyanate, anti–CD25-phosphatidylethanolamine (PE), anti–CD45RA-PerCP-Cy5.5, and anti–pSTAT5a-AlexaFluor 647 for 1 h in the dark at 20°C.

Analyses of pSTAT5a in cryopreserved PBMCs from selected groups of high (n = 12) and low (n = 12) IL-2 responders were stimulated with 0.2, 0.4 or 10 IU/mL hIL-2 for 30 min at 37°C, barcoded and stained with anti–CD4-allophycocyanin (APC)-eFluor 780, anti–CD25-PE, anti–CD45RA-PE-Cy7, anti–FOXP3-AlexaFluor 488, and anti–pSTAT5a-AlexaFluor 647 using a combination of the BD violet fluorescent cell barcoding kit and the BD Pharmingen Transcription Factor PhosphoPlus Buffer set (BD Biosciences). Data acquisition was performed on a BD FACSCanto II (BD Biosciences). Flow cytometry data were analyzed using FlowJo Software (Tree Star Inc., Ashland, OR). The data set per IL-2 concentration for each cell subset was normalized across the batch using the data from the two biological control subjects as detailed in Supplementary Fig. 1 to account for day-to-day variation. pSTAT5a analyses for fresh whole-blood samples were performed as previously described (17).

Fresh PBMCs were stained on ice with anti–CD4-qDot605, anti–CD14-AlexaFluor 488, anti–CD19-Pacific Blue, anti–CD25-PE (M-A251), and anti-CD127-PerCP-Cy5.5. Single lymphocytes were identified based on forward and side scatter parameters and populations isolated for functional analyses using a BD FACSAria II flow cytometer and FACSDiva software (BD Biosciences).

CD4⁺CD14⁻CD19⁻CD25^hiCD127^lo Tregs from fresh blood were stained with anti–FOXP3-AlexaFluor 647 and anti–Ki67-fluorescein isothiocyanate immediately after sorting and after 48 h cultured with or without limiting concentrations (0.1 or 1 IU/mL) of hIL-2 using the FOXP3/transcription factor staining buffer set (eBioscience).

Suppression assays were performed in V-bottom 96-well plates using fresh PBMCs by coculturing 500 sorted CD4⁺CD25^int-loCD127⁺ Tconv per well in the presence or absence of CD4⁺CD25^hiCD127^lo Tregs at various ratios with or without 1 × 10³ CD19⁺CD4⁻ B cells as a source of accessory cells in X-VIVO-15 media with 10% human sera. Samples were stimulated with phytohemagglutinin (4 µg/mL; Alere) (APC-dependent assay) or Human T-Activator anti–CD3/CD28 beads (Life Technologies) at a bead-to-Tconv ratio of 1:1 (APC-independent assay) and incubated at 37°C in 5% CO₂ for 6 days. Proliferation was assessed by the addition of 0.5 μCi/well [³H]thymidine (PerkinElmer, Waltham, MA) for the final 20 h of coculture. All conditions were run in quintuplicate and proliferation readings in counts per minute (cpm) averaged. Samples with proliferation <3,000 cpm were excluded. In cultures containing stimulated Tregs alone, in the absence of Tconv, proliferation was similar to the background of the assay (<500 cpm; mean, 166). Percentage suppression was calculated as previously described (17).

The normality of data sets was tested using the D’Agostino and Pearson omnibus normality test, and the unpaired Student t test, ANOVA, or Mann-Whitney test was used as appropriate. Correlations were assessed using linear regression. One-tailed tests were performed if there was prior evidence of association; otherwise, values from two-tailed tests were reported (GraphPad Software, Inc., La Jolla, CA). Case subject– and control subject–matched data were analyzed using bootstrap analysis (https://github.com/nicholasjcooper/misc/blob/master/YangBootStrap.R) accounting for a mixed design of pairs and trios, using the “boot” package in R software (www.r-project.org). Sample size and power calculations were calculated using Stata software (www.stata.com) and are detailed in Supplementary Fig. 6.

To assess responsiveness to IL-2, we measured pSTAT5a in cryopreserved PBMCs from 70 individuals with LST1D after brief in vitro exposure to IL-2 (Fig. 1). Barcoding (37) and normalization methods were used to reduce intra- and interstaining variability of the IL-2 sensitivity assay (Supplementary Fig. 1). An example of pSTAT5a staining and the gating strategy used to identify CD4⁺ T-cell subsets is shown in Fig. 2. Because fixation precluded the use of CD127 as a surface marker, Tregs were identified by CD4^loCD25⁺ staining, as previously described (17) (Fig. 2D). In addition, a more stringent definition was applied to identify Tregs by gating on the top 2% of CD25-staining CD4⁺ T cells (CD25^hi Tregs) (5,6,17,38). As previously observed, sensitivity to IL-2 in this assay was lowest for naïve Tconv (nTconv), with memory Tconv (mTconv), CD4^loCD25⁺CD45RA⁺ Tregs, CD4^loCD25⁺CD45RA⁻ Tregs, and CD25^hi Tregs showing successively higher sensitivities (Fig. 2F), which correlated with their respective CD25 expression levels (17,21).

To investigate the stability of IL-2 responsiveness and identify individuals who show reproducibly high or low IL-2 responsiveness, we obtained two independent blood draws from each subject, separated by a minimum of 3 months. We observed considerable interdonor variation in IL-2 responsiveness in all CD4⁺ T-cell subsets (Fig. 3). Notably, we observed a strong correlation between the two blood draws in all CD4⁺ T-cell subsets (r² = 0.34–0.87). Similarly, the frequency of T-cell subsets, including Tregs, was highly correlated between the two blood draws (r² = 0.68–0.92; Supplementary Fig. 2).

Long et al. (18) reported an association between a T1D-associated variant (rs1893217) in PTPN2 and reduced IL-2 signaling in CD4⁺ T cells in individuals without diabetes. We observed a similar association in all Treg subsets between two independent PTPN2 risk alleles of SNPs rs45450798 (r² = 1 with rs1893217) and rs478582 (r² = 0.159 with rs45450798) and reduced IL-2 signaling in our cohort of individuals with T1D (P = 4.4 × 10⁻³-0.02; Fig. 4).

Expression of FOXP3 is currently the most reliable marker to identify bona fide Tregs by flow cytometry. However to date, costaining of FOXP3 and pSTAT5a in cryopreserved PBMCs has been problematic. We therefore used novel staining reagents optimized for this purpose and were able to reliably identify and delineate three different populations of CD25⁺FOXP3⁺ T cells, as described by Sakaguchi and colleagues (39): resting (rTreg, FOXP3⁺CD45RA⁺, Fraction (Fr.) I), activated (aTreg, FOXP3^hiCD45RA⁻, Fr. II), and memory (mTreg, FOXP3⁺CD45RA⁻, Fr. III) Tregs (Fig. 5A and B). To confirm that the intrinsic differences in IL-2 signaling between individuals were maintained when this more definitive method for identifying Tregs was used, we selected cryopreserved PBMCs from subgroups of 24 individuals with LST1D with extremes of IL-2 signaling, identified from the IL-2 responses from their CD4^loCD25⁺CD45RA⁻ and CD25^hi Tregs (Supplementary Fig. 3). We observed that low IL-2 responders maintained reduced number of pSTAT5a⁺ cells in total FOXP3⁺ Tregs compared with high IL-2 responders (P = 3.6 × 10⁻⁵; Fig. 5C), with the greatest difference observed in the aTreg subset (P = 4 × 10⁻⁴-0.016; Fig. 5D–F). Kinetic analysis of IL-2–induced pSTAT5 induction in selected individuals demonstrated that the difference between high and low responders was maintained at several time points after stimulation (Supplementary Fig. 4A and B). No difference in IL-2 responsiveness was observed between the two groups at the higher concentration of IL-2 (10 IU/mL), indicating that optimal/saturating IL-2 concentration could “recover” deficient response observed in individuals with low IL-2 responsiveness (Supplementary Fig. 4C). Expression of CD25 was also reduced in all Treg subsets from low IL-2 responders, most notably in aTregs (P = 1 × 10⁻³; Supplementary Fig. 5). Furthermore, we observed a reduced frequency of FOXP3⁺ Tregs in individuals with low IL-2 responsiveness (P = 1.8 × 10⁻³; Fig. 6A). The greatest difference in Treg frequency between the two subgroups of individuals with LST1D was observed in aTregs (P = 5 × 10⁻⁴; Fig. 6B). A similar difference was observed in mTregs (P = 0.011) but not in antigen-inexperienced rTregs (P = 0.49; Fig. 6C and D).

To examine the relationship between IL-2 sensitivity and Treg fitness (FOXP3 maintenance and proliferation), we recalled the same subgroups of individuals with extremes of IL-2 signaling to examine fresh PBMCs from a third blood draw. Consistent with the previous data in cryopreserved PBMCs (Fig. 6), we observed a reduced frequency of Tregs in low IL-2 responders compared with high IL-2 responders (P = 7.1 × 10⁻³–0.03; Fig. 7A and B). Tregs were more proliferative (5–16% Ki-67⁺) in vivo compared with Tconv (0.8–3% Ki-67⁺), in agreement with previous reports (40). However, no difference was found for the steady-state proliferation of immediately postsorted CD25^hiCD127^lo Tregs between high and low IL-2 responders (P = 0.64; Fig. 7C and D). Previous studies have shown that defects in IL-2 signaling contribute to diminished maintenance of FOXP3 expression in Tregs in a subgroup of individuals with T1D (10). Here, we also observed that Tregs from high IL-2 responders cultured with a limiting concentration of IL-2 (1 IU/mL) were better at maintaining FOXP3 expression compared with low IL-2 responders (P = 0.017; Fig. 7E). Furthermore, the level of FOXP3 maintenance was positively correlated with the level of IL-2 signaling in FOXP3⁺ Tregs (r² = 0.22, P = 0.04; Fig. 7F).

To determine if in vitro Treg suppressive capacity differs between the two subgroups of individuals with extremes of IL-2 signaling, we used an in vitro coculture suppression assay. We observed reduced levels of suppression of Tconv proliferation in low IL-2 responders compared with high IL-2 responders under APC-dependent (P = 0.026) and APC-independent (P = 0.036) conditions (Fig. 7G). In cultures without Tregs, low IL-2 responders were observed to have increased Tconv proliferation compared with high IL-2 responders (P = 0.079 for APC-dependent and P = 0.027 for APC-independent assays; Fig. 7H).

Given the extensive interindividual variation observed in our LST1D cohort, we wanted to test whether IL-2 responsiveness in CD4⁺CD25⁺ T cells from individuals with T1D was different from control subjects without diabetes in the light of two previous studies (9,10). We compared IL-2 signaling in cryopreserved PBMCs from 18 individuals without diabetes recruited contemporaneously with the 70 individuals with LST1D. However, no difference in IL-2 sensitivity was observed in any CD4⁺ T-cell population between these two groups (P = 0.05–0.45; Fig. 8A–D). In addition, we examined IL-2 sensitivity using fresh whole-blood samples from 17 individuals with NDT1D and 22 with LST1D and compared these with 15 matched UAS and 20 control subjects without diabetes, respectively. Samples from individuals with T1D were run in parallel with a control subject without diabetes in an attempt to minimize interday variation. However, despite the paired nature of the study design, and consistent with the studies with cryopreserved PBMCs, we observed a similar distribution of IL-2 responsiveness in individuals with and without T1D, with no significant difference in IL-2 responsiveness observed between groups in any of the CD4⁺ T-cell subsets analyzed (P = 0.13–0.73; Fig. 8E–G).

A major motivation for our research is to develop stratified or precision medicine for the treatment of T1D. This goal is based on detailed and reproducible knowledge of the mechanisms of disease and identification of accurately measured phenotypes that not only measure the effects of potential immunotherapies but also might indicate at baseline (before drug administration) which patients might respond more or less than others. To this end, here we have established robust methods and procedures for measuring IL-2 signaling in T cells, including pSTAT5a measurement, in relation to Treg function. We discovered that Tregs from patients with low IL-2 responsiveness were less able to maintain the expression of FOXP3 under limiting concentrations of IL-2 and displayed reduced suppressor function with lower overall frequencies of Tregs in the circulation compared with individuals with higher IL-2 responsiveness. These results suggest that T1D patients with lower IL-2 responsiveness might benefit more, in terms of safely enhancing Treg function, from treatment with physiological, or ultra-low, doses of IL-2.

Assessing cell phenotype and function using fresh blood samples poses several technical challenges, especially when large sample sizes are involved. When fresh blood is used, only a relatively few samples can be collected and tested on the same day, leading to unavoidable day-to-day variation inherent when measuring the levels of intracellular proteins such as pSTAT5a. Here, we opted to assess IL-2 responsiveness in individuals with LST1D using cryopreserved PBMCs from two independent blood draws. To reduce day-to-day variation, we performed batch analyses and exploited barcoding (37) and normalization methods to minimize intra- and interassay variability. In accordance with the study by Long et al. (10), we demonstrated that IL-2 responsiveness is a stable phenotype of CD4⁺ T cells within an individual with highly correlated IL-2 signaling between the two blood draws. Extensive interindividual variation in IL-2 responsiveness was observed, not only in cohorts with T1D but also in control subjects without diabetes. We and, more recently, Yu et al. (21) failed to replicate the finding by others that individuals with T1D have reduced IL-2 responsiveness compared with control subjects (9,10). We observed that control subjects present a similar distribution of IL-2 responsiveness compared with individuals with T1D, with a gradient of response, with some control subjects displaying reduced IL-2 signaling. However, we acknowledge that larger sample sizes are required to rigorously address what may be a subtle phenotype. The level of heterogeneity of this assay between studies precludes us from accurately estimating a combined effect size for the sensitivity to IL-2 signaling. Given the wide range of effect sizes, we estimate that sample sizes exceeding 300 individuals would be needed to reveal differences of <5% in IL-2 signaling between case subjects and control subjects (Supplementary Fig. 6). It is not surprising that some control subjects also present reduced IL-2 responsiveness, particularly because the degree of IL-2 responsiveness is influenced by polymorphisms in several genes in the IL-2–signaling pathway, such as T1D-associated variants in IL2RA and PTPN2, where individuals without diabetes carrying the risk alleles had reduced IL-2 signaling in Tregs (17,18). Here, in a cohort of individuals with LST1D, we replicated the association of T1D-associated PTPN2 variant(s) with IL-2 signaling that was initially observed in control subjects without diabetes (18), further supporting the robustness of our sample quality control and methods.

Definitive identification of bona fide Tregs by flow cytometry is problematic, especially because activated CD4⁺ T cells share many phenotypic characteristics with Tregs. For the initial assessment of IL-2 sensitivity in cryopreserved PBMC samples from individuals with LST1D, Tregs were identified from high levels of CD25 expression, as previously described (5,6,17,38). During the course of the study, we were able to robustly dual-stain for FOXP3 and pSTAT5a in cryopreserved PBMCs to allow for more definitive gating of Tregs, thus enabling the confirmation of extremes of IL-2 signaling in FOXP3⁺ Tregs in selected individuals from the screening study.

A study by Long et al. (10) observed a reduced IL-2 responsiveness in CD4⁺CD25⁺ T cells from individuals with T1D and also a reduced ability to maintain the expression of FOXP3, although a direct association of these two phenotypes was not shown. In the current study, we established a direct link between IL-2 sensitivity and expression of FOXP3. One of the pleiotropic roles of IL-2 is its requirement for the maintenance of FOXP3 expression in Tregs (41) to sustain suppressive function because downregulation of FOXP3 has been associated with a loss of suppressor function (42). Here we observed that Tregs from individuals with T1D with reduced IL-2 responsiveness indeed are inferior at suppressing proliferation by autologous Tconv compared with Tregs from individuals with high IL-2 responsiveness. Interestingly, in cultures without Tregs, proliferation of Tconv was increased in individuals with reduced IL-2 responsiveness upon T-cell receptor stimulation. We, therefore, cannot rule out that Tconv of low IL-2 responders may also have enhanced resistance to Treg suppression, a phenotype that we and others had previously observed in individuals with T1D (7,43), or that cells in these individuals have an intrinsic proliferative advantage owing to deficient regulation. Recent results suggest that elevated production of IL-21 by Tconv, including T-follicular helper cell, could be part of the intricate balance between Treg and Tconv activities (33,34,44,45).

Intriguingly, we consistently observed a reduced frequency of Tregs in individuals with reduced IL-2 responsiveness using fresh and cryopreserved PBMCs, due to a reduction in numbers of Tregs mainly of the antigen-experienced activated/memory phenotype. There was no difference between the in vivo steady-state proliferation of Tregs in individuals with extremes of IL-2 responsiveness. Ghosh and colleagues (5,6) observed a higher level of apoptosis in Tregs using fresh blood samples from individuals with T1D compared with control individuals, which may be partially mediated by IL-2 deprivation resulting in lower expression of antiapoptotic genes. We therefore compared apoptosis in isolated Tregs cultured with low-dose IL-2 and Treg expression of antiapoptotic Bcl-2 in cryopreserved PBMCs by flow cytometry in selected individuals with extremes of IL-2 responsiveness. However, we observed no significant difference between the two groups in any of these analyses (data not shown). The apparent disparity between these results and those obtained by Ghosh and colleagues is likely due to methodological differences, including use of fresh blood versus isolated/cultured or cryopreserved Tregs. Further studies are required to investigate the link between IL-2 responsiveness, Treg frequency, and Treg survival.

IL-2–dependent STAT5a phosphorylation requires more than 10-fold lower concentrations of IL-2 in Tregs compared with Tconv subsets due to expressing higher levels of IL-2RA and common γ-chain and also, as recently observed, an increased activity of endogenous serine/threonine phosphatase 1/2A (17,21). Well-tolerated ultra low–dose IL-2 has been shown to not only expand the frequency but also augment suppressor function of Tregs, albeit with significant interindividual variation, in healthy individuals (46). Most recently, Yu et al. (21) demonstrated a consistent increase in the expression of FOXP3 and CD25 in Tregs, but not mTconv, for all individuals with LST1D upon treatment with low-dose IL-2. Interestingly, their study also showed a heterogeneous IL-2–dependent gene expression profile in Tregs in healthy subjects, suggesting differential regulation of IL-2–dependent genes in an individual might affect the outcome of low-dose IL-2 therapy. In most, if not all, of these aforementioned IL-2 immunotherapy studies, the ability to enhance the frequency of Tregs in response to IL-2 treatment is heterogeneous between individuals (26,29,30). As observed in our study and by others, because of the heterogeneity of the responses observed in individuals with autoimmune diseases, it may be advantageous to characterize IL-2 responsiveness and/or IL-2–dependent gene expression profiles and stratify individuals to be targeted for low-dose IL-2 immunotherapy for T1D and other immune-related diseases.

Acknowledgments. The authors thank all subjects for their participation. The authors thank members of the National Institute for Health Research (NIHR) Cambridge BioResource Scientific Advisory Board and management committee for their support and access to NIHR Cambridge BioResource volunteers and their data and samples. Documents describing access arrangements and contact details are available at http://www.cambridgebioresource.org.uk/. The authors thank the recruitment teams of the NIHR Cambridge BioResource and the NIHR Biomedical Research Centre at Guy's and St Thomas' National Health Service Foundation Trust and King's College London for assistance with volunteer recruitment and K. Beer, T. Cook, L. Eckhardt, S. Gillman, D. Goyme, S. Mbale, and J. Rice for blood sample collection. The authors thank M. Woodburn and T. Attwood for their contribution to sample management, N. Walker and H. Schuilenburg for data management, H. Stevens for providing clinical resources support, and L. Bell, G. Coleman, S. Dawson, J. Denesha, S. Duley, and M. Maisuria-Armer for preparation of blood and DNA samples at JDRF/Wellcome Trust Diabetes and Inflammation Laboratory. The authors thank E. O’Donnell and J. Vidal at Becton, Dickinson and Company for technical support and providing reagents. The authors also thank Thomas Hayday at Department of Immunobiology, Faculty of Life Sciences and Medicine, King's College London for performing flow cytometry cell sorting.

Funding. This work was supported by the JDRF UK Centre for Diabetes Genes, Autoimmunity and Prevention (D-GAP, 4-2007-1003), the Wellcome Trust (WT061858/091157), the NIHR Biomedical Research Centre based at Guy's and St Thomas' NHS Foundation Trust and King's College London, and the NIHR Cambridge Biomedical Research Centre. The Cambridge Institute for Medical Research is in receipt of a Wellcome Trust Strategic Award (100140).

Duality of Interest. C.S.B. and G.-J.G. are employees of Becton, Dickinson and Company. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. J.H.M.Y., J.A.T., L.S.W., and T.I.M.T. participated in the design and interpretation of the experiments and results. J.H.M.Y., A.J.C., R.C.F., J.L.R., P.C., D.J.S., and T.I.M.T. participated in the acquisition and analysis of data. N.J.C. and C.W. participated in statistical analyses. C.S.B. and G.-J.G. devised, designed, and optimized reagents for dual FOXP3 and pSTAT5a staining. J.H.M.Y., J.A.T., and T.I.M.T. wrote the manuscript. T.I.M.T. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.