Selective IL-2 Responsiveness of Regulatory T Cells Through Multiple Intrinsic Mechanisms Supports the Use of Low-Dose IL-2 Therapy in Type 1 Diabetes

Interleukin 2 (IL-2) influences tolerance and immunity by promoting regulatory T cell (Treg) development and homeostasis and T effector and memory responses (1). However, at limiting concentrations of IL-2, tolerance appears to be favored over immunity. In preclinical mouse studies, for example, the administration of a relatively low amount of IL-2 increased Tregs, lowered instances of autoimmune diabetes, and reduced the severity of experimental autoimmune encephalomyelitis (2–4). Additionally, low signaling associated with mutant IL-2 receptor subunit β (IL-2Rβ) readily supported Treg development and homeostasis, whereas T memory (T_M) responses remained substantially impaired (5,6).

There is much interest to boost Tregs cells in patients with type 1 diabetes (T1D) to enhance normal tolerogenic and immune suppressive mechanisms to control islet autoimmunity. Furthermore, single nucleotide polymorphisms in IL2RA represent a genetic risk for T1D (7), and impaired IL-2R signaling in Tregs has been observed in patients with T1D (8,9). Thus, the application of IL-2 represents not only a direct approach to enhance Tregs but also a means to target one of the underlying abnormalities associated with this disease. Recent clinical experiences suggest that the administration of IL-2 at doses lower than those previously used in attempts to boost immunity led to increased Tregs and improved clinical outcomes for patients with chronic graft-versus-host disease (GvHD), hepatitis C virus (HCV)–induced vasculitis, and alopecia areata (10–13). Prophylactic use of low-dose IL-2 was associated with lower incidences of GvHD (14). A short-term safety trial of low-dose IL-2 therapy was completed in T1D patients enrolled between 0.5 and 2 years from diagnosis and showed that this therapy was well tolerated and accompanied by a dose-dependent increase in Tregs (15). On the basis on these results, an efficacy trial has begun in patients with new-onset T1D (ClinicalTrial.gov identifier NCT01862120). The effectiveness of low-dose IL-2 therapy in these diverse groups of patients suggests that selective IL-2 responsiveness is a general property of human Tregs. However, little is known concerning the mechanisms for enhanced IL-2 responsiveness by Tregs.

In this report we quantify the extent that IL-2–dependent signaling and downstream gene activation are favored in Tregs compared with T_M cells and other lymphoid cell populations from healthy subjects and T1D patients. Our studydefines key mechanisms that explain the preferentialinduction of Tregs with low-dose IL-2 and provides additional rationale for this approach in T1D.

Peripheral blood samples from 32 healthy adult donors were purchased from the Continental Blood Bank, Miami, FL. Peripheral blood was obtained from eight patients with T1D, five males and three females (age range 10–52 years; mean ± SD 26.8 ± 14.0). The University of Miami Institutional Review Board approved the study, and all patients signed written informed consent. Frozen peripheral blood samples were obtained from two male and six female T1D patients (age range 22–49 years; mean 31.3 ± 8.4 years) undergoing a short course of low-dose IL-2 as described in the Dose-effect Relationship of Low-dose IL-2 in Type 1 Diabetes (DF-IL2) trial (Clinical Trials.gov identifier NCT01353833) (15). This study was approved by the Pitié-Salpêtrière Hospital Ethical Committee, and written informed consent was obtained from all subjects preceding the start of the study.

Blood was processed the day after collection according to U.S. Food and Drug Administration guidelines for clinical testing. Heparinized blood was diluted in PBS (1:3 final ratio), layered on Ficoll-Paque Plus (GE Healthcare, Little Chalfont, U.K.) and centrifuged at 400g for 30 min at room temperature, without braking. The interphase cells were collected and washed once with PBS.

The following monoclonal antibodies, obtained from BD Biosciences (San Jose, CA), or BioLegend or eBiosciences (both San Diego, CA), were used with the clone names designated in parenthesis: FOXP3 (259D), CD56 (B159), CD8 (RPA-T8), CD3 (SK7 or OKT3), CD4 (RPA-T4), CD45RA (H100), CD122 (9A2-CD122), IL2Rα (M-A251or BC96), CD45RO (UCHL1), CD127 (HCD127), CD132 (TUGh4), and phosphorylated STAT5 (pSTAT5) (pY694). Cells were stained in FACS buffer (PBS, 2% heat inactivated FBS, 1 mmol/L EDTA, 0.1% sodium azide). With the exception of pSTAT5 analysis (see below), FOXP3 staining was performed after fixing and permeabilizing using the FOXP3/Transcription Factor Staining Buffer Set (eBioscience) according to the manufacturer’s instructions. Typically, 100,000 events were collected using BD LSR II or BD LSRFortessa flow cytometers. Data were analyzed using BD FACSDiva 6.0 software, where viable cells were gated based on forward versus side light scatter profiles and doublets were excluded based on forward light scatter area versus scatter width. A viability dye was sometimes added to exclude dead cells.

Human CD4⁺ T cells were enriched by negative selection with the MACS CD4⁺ T Cell Isolation Kit II (Miltenyi Biotec, Auburn, CA). The purified CD4⁺ T cells (2 × 10⁶ cell/well in 24-well plates) were cultured (37°C in 5% CO₂) in 1 mL complete medium (RPMI 1640 medium supplemented with 10% human AB serum, sodium pyruvate [1 mmol/L], penicillin [50 units/mL]/streptomycin [50 μg/mL], and glutamine [2 mmol/L]) overnight in the absence or presence of human IL-2. These CD4⁺ T cells were stained and sorted using a BD FACSAria into Tregs (CD4⁺ IL2Rα^hi CD127⁻) and T_M (CD4⁺ IL2Rα⁺ CD127⁺ CD45RA⁻), where dead DAPI⁺ cells were excluded. Sorted cells were typically 98% pure. After washing, some cells were placed in QIAzol (0.5 mL) (Qiagen, Venlo, the Netherlands) and stored at −70°C until used for RNA isolation. Other sorted Tregs were evaluated for FOXP3 expression and viability by counterstaining using LIVE/DEAD fixable yellow (Life Technologies). To generate T cell blasts, peripheral blood mononuclear cells (PBMCs; 1 × 10⁶ cells/well in 24-well flat-bottom plates) were cultured in 1 mL complete medium and stimulated with LEAF purified anti-human CD3 (5 μg/mL; OKT3, BioLegend) for 72 h.

PBMCs or T blasts (0.5–1 × 10⁶/well in 24-well plates) were cultured in 1 mL complete medium for 0.5 and 4 h, respectively. After this “rest” culture, IL-2 was added for 15 min at 37°C. The cells were fixed by addition of 100 μL paraformaldehyde (16% solution, EM Grade; Electron Microscopy Sciences, Hatfield, PA) to the cultures for 10 min at 37°C. The cells were harvested, transferred to a culture tube (12 × 75 mm), and pelleted by centrifugation. The cells were resuspended and permeabilized by the addition of 0.5 mL ice-cold methanol. After incubation for 30 min on ice, the cells were washed twice in PBS containing 0.5% BSA and 0.02% sodium azide and stained for 60 min at room temperature in the dark with monoclonal antibodies to pSTAT5, relevant surface molecules, and FOXP3. The cells were washed twice with PBS containing 0.5% BSA and analyzed by flow cytometry. In some experiments, DMSO (0.02%) vehicle control or calyculin A (50 nmol/L; Cell Signaling, Danvers, MA) were added during the rest culture and remained present after the addition of IL-2.

Total RNA was isolated using the miRNeasy Micro Kit (Qiagen). The quality and quantity of the RNA was assessed and verified to be undegraded using an Agilent 2100 BioAnalyzer and a NanoDrop. One round of linear probe amplification using 100 ng RNA and labeling was performed using Ambion WT Expression Kit Gene (Life Technologies, Carlsbad, CA), and gene expression was assessed using Affymetrix Human Gene ST 1.0 arrays at Expression Analysis (Durham, NC). Image analysis was performed using Affymetrix Command Console Software. The robust multiarray averaging method was used to normalize the data. Differentially expressed genes were identified, and gene enrichment analysis was performed through software at GeneSifter (Seattle, WA). Hierarchical clustering and heat map were obtained using the GENE-E software from the Broad Institute, Massachusetts Institute of Technology (http:www.broadinstitute.org/cancer/software/GENE-E/). Gene array data have been submitted to the Gene Expression Omnibus (http:/www.ncbi.nlm.nih.gov/geo/) under accession number GSE49817.

cDNA was prepared with the High-Capacity cDNA Reverse Transcription Kit using random hexamer primers (Life Technologies). The primer pairs are listed in Supplementary Table 1. Real-time PCR was done in triplicate using Power SYBR Green PCR Master Mix (Life Technologies), primers (0.3 μmol/L), and cDNA (2.5 ng/μL). PCR conditions were 95°C for 10 min for 1 cycle, followed by 40 cycles of 95°C for 15 s and 60°C for 60 s. Results were adjusted based on the amplification of 18S RNA as an endogenous control using a commercially available primer set (Life Technologies).

Graphical representations of the data and statistical analyses were performed using GraphPad Prism 5.0 software. Data are shown as means ± SEM. To determine the half-maximal effective concentration (EC₅₀) for the pSTAT5 dose-response studies, values for the medium control were subtracted and the maximal response by each individual was normalized to 100%.

A key question is the extent to which low-dose IL-2 therapy selectively stimulates human Tregs with no effect on other IL-2R–bearing cells. We therefore assessed IL-2–dependent tyrosine phosphorylation of STAT5 (pSTAT5) 15 min after stimulation, where pSTAT5 responses were maximal. Multiple lymphoid subpopulations were examined, including CD45RA⁺ naive (n)Treg and CD45RO⁺ memory (m)Treg cells (16) in the PBMCs of healthy individuals. The gating of these and other cells after pSTAT5 staining is shown in Fig. 1A. FACS histograms (Fig. 1B) and dose-response curves for each individual (Fig. 1C) demonstrated that mTregs and nTregs were more responsive to low levels of IL-2 than the other cell populations. Although some variation was noted, the individual T-cell populations closely approximated each other, and this is evident by the average pSTAT5 responses (Fig. 1D). With respect to natural killer (NK) cells (Fig. 1A and C), the major CD56^lo population showed poor pSTAT5 activation, whereas CD56^hi NK cells, which express the high-affinity IL-2R (17), exhibited higher activation (Fig. 1A–D). Approximately 3 vs. 100 pmol/L IL-2 was required to support 50% pSTAT5 activation for mTreg versus CD4⁺ T_M cells, respectively (Fig. 1D). Nonlinear regression analysis of these data revealed that the response by mTregs was more sensitive to IL-2 compared with all other cell populations (P < 0.001) (Fig. 1E). Compared with mTregs, the EC₅₀ values for pSTAT5 responses by nTreg, CD56^hi NK, CD4⁺ T_M, CD8⁺ T_M, and CD4⁺ T_naive cells were 1.9-, 6.9-, 10.5-, 14.0-, and 19.3-fold higher.

Because pSTAT5 was measured using PBMCs, the shifts in dose-response curves by non-Tregs might be related to their inability to compete for IL-2 with Tregs. This seems unlikely, however, because the dose-response for pSTAT5 activation by CD4⁺ Treg and T_M cells was nearly identical when unfractionated PBMCs or highly purified cells were examined (Fig. 2).

Tregs expressed the highest levels of IL-2Rα and γ chain (γc), whereas NK cells expressed the highest levels of IL-2Rβ (Fig. 3A and B). Plotting the mean fluorescent intensity (MFI) versus the log-EC₅₀ (pmol/L) for pSTAT5 activation by mTregs, nTregs, CD4⁺, and CD8⁺ T_M cells, and CD56^hi NK cells revealed a relationship between the levels of expression of IL-2Rα and γc and the IL-2–dependent activation of pSTAT5 (Fig. 3C). Higher levels of these IL-2R subunits correlate to higher pSTAT5 responses with lower amounts of IL-2. On the basis of the ligand-binding properties of the IL-2R (18), higher IL-2Rα on Tregs may preferentially capture IL-2 to promote signaling. The trend for increased γc in Tregs, although not statistically significant, may increase recruitment of JAK3 to the IL-2R, also favoring STAT5 activation (19).

In vitro–activated CD4⁺ CD45RO⁺ T blasts expressed the highest levels of all IL-2R subunits (Fig. 3D and Supplementary Fig. 1A), but their IL-2–dependent pSTAT5 dose-response required higher levels of IL-2 than Tregs (Fig. 3E and Supplementary Fig. 1B). This observation couples with the observation that CD56^hi NK cells expressed lower levels of IL-2Rα than T_M cells yet were more responsive to IL-2, suggesting that other undefined cell-intrinsic properties distinctively regulate IL-2R signaling.

Protein phosphatase (PP)2A has been implicated in promoting IL-2R signaling in the YT cell line through its ability to dephosphorylate serine and threonine residues in IL-2Rβ, JAK3, and STAT5 (20). Therefore, we tested whether the PP1/PP2A inhibitor calyculin A affected IL-2–dependent pSTAT5 activation in various populations of primary human T cells. Pretreatment of human PBMCs with calyculin A inhibited IL-2–dependent pSTAT5 activation in mTregs, nTregs, and CD4⁺ T_naive and T_M cells, with greatest inhibition at lower levels of IL-2 (Fig. 4A). The pSTAT5 responses by CD4⁺ T_naive and T_M showed more inhibition than Tregs. However, if we considered the dose of IL-2 that supports 50% pSTAT5 activation (i.e., 3 pmol/L for mTregs and 100 pmol/L for CD4⁺ T_M cells) (Fig. 1D), the level of pSTAT5 inhibition (75–85%) was similar.

We also assessed the role of PP1/PP2A on the pSTAT5 responses by Tregs and CD4⁺ T_M cells at concentrations of IL-2 that support less than maximal responses (i.e., 1–10 pmol/L for Tregs and 30–100 pmol/L for CD4⁺ T_M cells) because these levels dictate the cellular sensitivity in response to IL-2. Nonlinear regression analysis was performed for pSTAT5 dose-response data in the absence or presence of calyculin A. The pSTAT5 responses measured in the DMSO vehicle control were higher than typically seen by cells cultured only in media containing IL-2, and these responses were inhibited by calyculin A (Fig. 4B). Calyculin A shifted the dose-responses curves such that more IL-2 was required for pSTAT5 activation. The EC₅₀ values in the DMSO control compared with those in calyculin A revealed a dose-response shift of 0.6 to 6.0 pmol/L for mTregs, 1.6 to 24 pmol/L for nTregs, and 23 to 87 pmol/L for CD4⁺ T_M cells. These shifts correspond to a requirement for 10-, 15-, and 3.8-fold more IL-2, respectively, to achieve the same level of IL-2–dependent pSTAT5 activation in the presence of calyculin A. Thus, compared with CD4⁺ T_M cells, blockade of PP1/PP2A with calyculin A in Tregs had a greater effect on the sensitivity to IL-2. These data are consistent with PP1/PP2A activity as another determinant in the high sensitivity of Tregs to IL-2.

To further investigate this, quantitative (q)PCR was used to measure the levels of the catalytic subunits of PP1 and PP2A in CD4⁺ Treg, T_M, and T_naive cells, and the levels were similar in these cell types (Fig. 4C). However, Tregs expressed an approximately twofold lower level of SET, an inhibitor of PP2A (21), as evidenced by the higher ΔCT value determined for Tregs (Fig. 4C). Thus, Tregs may express increased PP2A activity due to lower levels of SET.

To broadly define the outcome of IL-2R signaling for human Tregs, genome-wide profiling of expressed mRNAs was performed for purified Tregs (Supplementary Fig. 2) cultured for 24 h in the absence or presence of IL-2, and 388 mRNAs were differentially expressed by at least 1.5-fold, with 84.0% increased in the presence of IL-2 (Fig. 5A). Several of the most highly differentially expressed genes represent molecules involved in binding (IL2RA, PTGER2, IL1R1) or regulation (CISH, SOCS2) of responses to cytokines (Fig. 5A), suggesting that an important role of IL-2 signaling in Tregs is to regulate responses to cytokines involved in growth, homeostasis, and inflammation. Consistent with this notion, gene enrichment analysis (z score ≥4) of the differentially expressed genes identified pathways and processes in Tregs regulated by IL-2 (Supplementary Table 2). Some of these include the JAK-STAT signaling pathway, immune system processes, lymphocyte growth and death, metabolism, cell migration, and inflammation. Representative genes regulating some of these processes are shown in Fig. 5B. Several of these gene enrichment groups have also been identified in the mouse (5,22) and represent activities known to be controlled by IL-2.

Hierarchical clustering of the differentially expressed genes as they relate to individual subjects revealed two main clades (1 and 2) within the media and IL-2–stimulated cells (Fig. 5C). Two clusters of genes were also visually evident for individual samples in clade 1 that may represent normal subjects with low versus high responses to IL-2 as related to down- (box A) and up- (box B) regulation of genes expression (Fig. 5D). Comparison of all IL-2–dependent genes revealed that 32% were expressed at significantly different levels between the individuals within clades 1 and 2 and that 95% of these genes were expressed at lower levels in clade 1 (Supplementary Table 3). These findings suggest that IL-2–dependent regulation of these target genes is heterogeneous between individuals, which might affect the outcome of low-dose IL-2 therapy.

To assess the relative levels of IL-2–dependent activation in Treg versus T_M cells by flow cytometry, PBMCs were cultured with various amounts of IL-2. Nearly maximal levels of FOXP3 and IL-2Rα were noted for Tregs after culture with low IL-2 (0.3 units/mL) (Fig. 6A). Using a new group of normal subjects, RNA from purified CD4⁺ Tregs and T_M cells (Supplementary Fig. 2) was used for real-time qPCR analysis of 12 selected target genes from the gene array profiling (Fig. 6B). The analyses of Tregs in this independent sample confirmed the initial gene array findings. The maximal fold IL-2–dependent gene upregulation was always associated with Tregs. When considering this value, ≥50% of the maximal responses occurred at 1 (8 targets) and 10 units/mL (4 targets) for Tregs, but 8 of 11 targets required at least 100 units/mL for T_M cells to achieve such an IL-2–dependent response (Table 1 and Fig. 6B). The level of most mRNAs (average ΔCT_media) was similar between CD4⁺ Tregs and T_M cells, but some varied substantially; for example, AHR, FLT3LG, and FURIN were relatively high in CD4⁺ T_M cells, whereas FOXP3 and HLADR were relatively high in Tregs. Thus, Treg and T_M cells show overlapping IL-2–dependent responses, but the activation of such genes in T_M cells generally requires at least 10–100-fold greater levels of IL-2. Compared with the response by CD4⁺ T_M cells, this induction of mRNAs by IL-2 in Tregs was often even more selective than pSTAT5 activation.

IL-2–dependent pSTAT5 responses were assessed using fresh PMBCs from eight T1D patients with a long duration of disease (15.0 ± 7.4 years). Dose-response curves of these responses (Fig. 7A) revealed that mTregs were more responsive to IL-2 than CD4⁺ T_M cells. Similar to normal control subjects (Fig. 1C), some variation was noted in the magnitude of the pSTAT5 responses at high concentrations of IL-2 (Fig. 7A, left). Nevertheless, normal control subjects and T1D patients showed similar and not statistically different EC₅₀ values for mTregs and CD4⁺ T_M cells (Fig. 7B). This result did not change even when we removed the one outlier for the T1D group.

Identical IL-2 dose-response experiments were performed on frozen/thawed PBMCs for eight additional T1D patients with a shorter duration of disease (1.1 ± 0.8 years) who were treated with low-dose IL-2 (0.33–3 million IU [MIU]) in the DF-IL2 trial (15). Each patient received a daily subcutaneous injection of IL-2 for 5 consecutive days. IL-2–dependent pSTAT5 was assessed on a sample of PBMCs collected immediately before the start of low-dose IL-2 and on another sample collected 4–6 days later, at the end of the treatment course. The EC₅₀ values for mTregs, nTregs, and CD4⁺ T_M cells were indistinguishable for each population before and after the start of low-dose IL-2 (Fig. 7C). Responses by mTregs and nTregs showed lower EC₅₀ values than CD4⁺ T_M cells, and the EC₅₀ values from these frozen/thawed T cell populations were essentially identical to those measured when using fresh PBMCs from normal control subjects (Fig. 1E) or T1D patients (Fig. 7C).

Two genes identified that are responsive to low levels of IL-2 in vitro are FOXP3 and IL-2Rα (Fig. 6). Therefore, the expression of FOXP3 and IL-2R subunits in Tregs and CD4⁺ T_M cells was assessed for T1D patients undergoing low-dose IL-2 treatment. Initially we confirmed that mTregs and nTregs expressed substantially higher levels of IL-2Rα (Fig. 7D) and FOXP3 (data not shown) than CD4⁺ T_M cells. mTregs also expressed a slight (∼10%) increase in cell surface γc. When these molecules were examined 4–6 days after the start of low-dose IL-2 treatment, FOXP3 and IL-2Rα were consistently increased in mTregs and nTregs, but not in CD4⁺ T_M cells, for all T1D patients (Fig. 7E), which paralleled the in vitro responses (Fig. 6A). Low-dose IL-2 also mediated a small but selective increase in γc in mTregs. Collectively, these data indicate that Tregs from T1D patients retained selective responsiveness to IL-2 as assessed by pSTAT5 activation in vitro and by FOXP3 and IL-2Rα expression in vivo.

There is much interest in developing protocols that boost Treg number and function in T1D to suppress unwanted immune responses. Our study provides an initial mechanistic underpinning for the selectivity of low-dose IL-2 therapy and its capacity to broadly enhance IL-2–dependent genes in human Tregs from healthy control subjects and patients with T1D. IL-2–dependent signaling and downstream gene activation readily occur at ∼10–15- and 100-fold lower levels of IL-2, respectively, than in T_M cells. This 100-fold difference for gene activation provides a considerable therapeutic window where low levels of IL-2 may target Tregs in the absence of effects on T_M cells. The higher amount of IL-2 needed for gene activation compared with STAT5 signaling by T_M cells may be related to a requirement for sustained signaling over time, which amplifies the difference in pSTAT5 activation. Another possibility may be related to increased complexity of IL-2R signaling in T_M cells because they are more dependent on the IL-2–induced PI3K/mTOR pathway, which is muted in Tregs due to high PTEN levels (23). The potential to broadly and selectively regulate the IL-2 gene program in Tregs represents an attractive feature of low-dose IL-2 therapy and likely accounts for the beneficial effects on Tregs and clinical outcomes reported in patients that have received low-dose IL-2 therapy (10–12). However, our gene profiling studies raise the possibility that there may be individuals that are low and high responders with respect to a subset of genes that are IL-2 dependent in Tregs. Additional assessment of this point is required to confirm and extend this finding because it may lead to a means to predict patients who may optimally benefit from this therapy and also help to personalize IL-2 doses.

Because IL-2R signaling has been reported to be impaired in patients with T1D (8,9), an important consideration for the application of low-dose IL-2 therapy to these patients is whether the window of selective responsiveness to IL-2 by Tregs is maintained. By performing extensive dose-response studies and determining the EC₅₀ values for these responses, our analysis of IL-2–dependent pSTAT5 activation showed that the window of selective responsiveness by Tregs compared with CD4⁺ T_M cells was similar for normal control subjects and 16 of 16 T1D patients. We conclude from these results that there is not a fundamental impairment in IL-2R signaling in Tregs in T1D patients and that these patients are good candidates for low-dose IL-2 therapy. On the basis of the similar EC₅₀ values, one low dose of IL-2 is likely to be effective to selectively boost Tregs in most T1D patients. We noted some individual variation in these responses by Tregs and other cells populations from normal control subjects and T1D patients, but this was primarily in the maximal pSTAT5 response at high concentrations of IL-2. Past work indicating impaired IL-2R signaling in T1D primarily noted this latter type of variation (8). In our data set, where the number of patients and control subjects was more limited and not rigorously aged matched, we saw a similar trend for a lower percent of pSTAT5⁺ Tregs from T1D patients at high IL-2 levels, but this was statistically nonsignificant compared with the responses by Tregs from normal subjects (data not shown). This variation in maximal IL-2–induced pSTAT5 in Tregs might reflect differences imposed by disease or polymorphisms in IL2RA or PTPN2 (7,8), which may potentially be more active in a subset of Tregs. Overall, our findings and the lack of activation of T effector responses in patients with GvHD, HCV vasculitis, and T1D (10–12,15) suggest that the mechanism responsible for Treg-selective responsiveness to low-dose IL-2 is robust and maintained under various conditions of immune activation.

One determinant of selective IL-2–dependent activation of human Tregs in response to low levels of IL-2 is relatively high expression of surface IL-2Rs compared with other lymphoid cells. This finding is consistent with past studies using cloned mouse-activated T cells that showed that greater pSTAT5 activation occurred at lower levels of IL-2 for those cells that expressed higher amounts of IL-2Rα and IL-2Rβ (24). For human Tregs, their increased levels of IL-2Rα, and perhaps γc, support their relatively strong response to low levels of IL-2. Our data also make plain that other factors besides IL-2R levels contribute to their responsiveness to low-dose IL-2. On the one hand, activated T cell blasts expressed substantially greater levels of IL-2Rα and β subunits than Tregs and yet were less responsive at low doses of IL-2 than Tregs. On the other hand, CD56^hi NK cells expressed very low levels of IL-2Rα yet showed good pSTAT5 responses to lower amounts of IL-2, albeit responses that were not as robust as those by Tregs. In addition, the Tregs in T1D patients who were treated with low-dose IL-2 expressed twofold higher levels of IL-2Rα but did not exhibit enhanced IL-2–dependent pSTAT5 activation.

Several mechanisms, besides high IL-2R levels, may promote high IL-2R signaling sensitivity in Tregs. One mechanism may be at the level of SOCS proteins (25), which attenuate IL-2 signaling and STAT5 activation in activated T cell blasts (26–29). Tregs express low levels of SOCS3 (30), which can be further decreased through degradation by SOCS2, to enhance IL-2R signaling (31). Notably, SOCS2 is a highly IL-2–dependent gene in Tregs (Fig. 5B), which may enhance IL-2R signaling in a positive feedback loop. Tregs express high levels of FOXP3-dependent microRNA miR-155 that lowers SOCS1 and enhances IL-2R signaling (32). Another mechanism may be related to the levels of serine/threonine phosphorylation associated with IL-2Rβ, JAK3, and STAT5, where this modification downregulates IL-2R signaling (20). Indeed, we showed that inhibition of the serine/threonine phosphatase activity with calyculin A, which targets PP1 and PP2A, highly inhibited IL-2–dependent pSTAT5 at low doses of IL-2 in Tregs. Moreover, we found that Tregs contain diminished levels of SET, an inhibitor of PP2A, suggesting that heightened activity of PP2A may normally contribute to their increased IL-2 signaling sensitivity.

Our findings indicate that Treg-specific low-dose IL-2 therapy must identify a dose of IL-2 that does not activate CD4⁺ T_M and CD56^hi NK cells, because these two cell populations are next in line with respect to responsiveness to IL-2. Impaired β-cell function and increased CD56^hi NK cells were noted in T1D patients who received 54 MIU of IL-2 during a 4-week period in conjunction with rapamycin (33), which was given for 3 months. It is not clear whether increases in T effector, NK cell, and/or the well-established rapamycin-mediated β-cell toxicity (34) were responsible for decreasing β-cell function when assessed 3 months after the start of therapy. The benefit for patients with GvHD, HCV vasculitis, or alopecia areata who have undergone low-dose IL-2 therapy suggests that reactivation of T effector cells may not easily occur (10–13). More recently, much lower doses of IL-2 were solely administered to T1D patients, and a dose of IL-2 was identified that was safe and boosted Treg without increasing in CD56^hi NK cells (15). The initial bolus dose of IL-2 at 1 week was approximately threefold lower than the dose in the IL-2/rapamycin trial, and the dosing that the efficacy trial plans to administer is a 7.5-fold lower amount of IL-2 over 1 month.

In conclusion, there are likely a set of permissive conditions that converge to selectively support the activation and function of Tregs to a limiting amount of IL-2 during low-dose IL-2 therapy. Three conditions defined here for human Tregs include enhanced pSTAT5 activation, decreased negative signaling, most likely mediated by PP2A, and integration of proximal IL-2R signals to amplify IL-2–dependent mRNA expression. These conditions ensure that human Tregs readily, selectively, and effectively respond to low concentrations of IL-2. We show here that FOXP3 gene activation and protein expression in Tregs is supported by low levels of IL-2. Upregulation of FOXP3 by low levels of IL-2 is expected to reinforce the human Treg gene program and improve suppressive function. Human Tregs also express high levels of the high-affinity IL-2R compared with other T cells. Therefore, human Tregs in vivo are also expected to outcompete T effector cells for limiting IL-2. In summary, our study defines mechanisms of action and provides additional rationale for low-dose IL-2 therapy in patients to enhance tolerance over self-reactivity.

Acknowledgments. The authors thank the Flow Cytometry Cores of the Diabetes Research Institute and the Sylvester Comprehensive Cancer Center at the University of Miami, and Oliver Umland, of the University of Miami, for expert help with flow cytometry.

Funding. This work was supported by the Diabetes Research Institute Foundationhttp://dx.doi.org/10.13039/100001078, Hollywood, FL, the Peacock Foundationhttp://dx.doi.org/10.13039/100001792, Inc., Miami, FL, the Anton E.B. Schefer Foundation, and French state funds managed by the ANR within the Investissements d'Avenir programme (ANR-11-IDEX-0004-02).

Duality of Interest. M.R. and D.K. are inventors on a patent application related to the therapeutic use of low-dose IL-2, which belongs to their respective academic institutions and has been licensed to ILTOO Pharma. M.R. and D.K. hold shares in ILTOO Pharma. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. A.Y., I.S., and F.V. performed experiments. A.Y., I.S., F.V., A.P., and T.R.M. analyzed the data. M.R. and D.K. designed the clinical trial and provided patients’ samples. D.K. and A.P. edited the manuscript. A.P. and T.R.M. designed the study. T.R.M. wrote the manuscript. T.R.M. 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.