Immunotherapy With Low-Dose IL-2/CD25 Prevents β-Cell Dysfunction and Dysglycemia in Prediabetic NOD Mice Free

Type 1 diabetes is an autoimmune disease that has no cure and affects over 9 million people (1). Experimental immunotherapies for type 1 diabetes have shifted away from broad immunosuppression and moved toward enhancing immune tolerance (2). Low-dose IL-2 therapy centers on induction and proliferation of regulatory T cells (Tregs), which are integral to promoting immune tolerance and preventing β-cell destruction (2,3). This therapy is being refined by the development of IL-2 analogs to promote selectivity for the high-affinity IL-2R and Tregs (4). One is an IL-2/CD25 fusion protein (5). This fusion protein exists predominantly as inactive, noncovalent transdimers that slowly dissociate into active monomers. IL-2/CD25 has a longer half-life than recombinant IL-2, and, on a per molecule basis, supports greater in vivo Treg expansion, suppressive activity, migration into the pancreas, and decreased numbers of T follicular helper cells (6). While efficacy trials for low-dose IL-2 are ongoing (NCT02411253, NCT0378263, and NCT01862120), data from preclinical studies indicate that low-dose mouse IL-2/CD25 (mIL-2/CD25) treatment prevents type 1 diabetes in nonobese diabetic (NOD) mice (5,6). This treatment limits β-cell destruction, but its effects on preserving β-cell function have not yet been assessed.

Impaired β-cell function is an early feature of type 1 diabetes pathogenesis (7–10). In stage 1 pre–type 1 diabetes, individuals display autoimmunity (two or more islet autoantibodies) but have no metabolic defects, while, in stage 2, individuals have β-cell dysfunction (11). This dysfunction is seen long before symptomatic type 1 diabetes in mouse (10) and human disease (8), and it is caused by immune cell infiltration (10,12–14). Data from recent clinical trials show that immunotherapy treatment can preserve β-cell function as determined by C-peptide in patients who have prediabetes and recent-onset diabetes (15–17). Intervening early to preserve β-cell function when greater β-cell mass remains could have long-term benefits (18–20). Therefore, it is imperative to understand how modulation of immune infiltrates with IL-2/CD25 not only prevents β-cell destruction but also β-cell dysfunction early in disease progression.

Here, we treated 12-week-old female NOD mice with mIL-2/CD25 for 5 weeks. These mice are in a state resembling stage 1 pre–type 1 diabetes (10). We evaluated how mIL-2/CD25 therapy affects β-cell function downstream of its known effects on infiltrating immune cells in the pancreatic islet during and after cessation of treatment. We measured glucose tolerance and plasma insulin levels to estimate β-cell function and to define the stage of pre–type 1 diabetes. To understand how β-cell function changes during the pathogenesis of type 1 diabetes and how it is affected by the treatment, we investigated β-cell physiology in detail. We 1) examined the insulin secretory capacity of islets by measuring dynamic hormone secretion and 2) recorded changes in intracellular free Ca²⁺ concentration ([Ca²⁺]_i) of individual β-cells in response to glucose in living pancreas slices. We found that treatment with low-dose mIL-2/CD25 limits β-cell dysfunction and prevents NOD mice from progressing to stage 2 pre–type 1 diabetes. Importantly, these effects were sustained for at least 5 weeks after cessation of treatment.

Female NOD-ShiLtJ (NOD) and NOD-scid IL2Rgamma^null (NSG) mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and housed in a pathogen-free animal facility. Mice were screened for diabetes biweekly by checking glucose levels using Contour Next EZ glucometer and test strips (Bayer, Barmen, Germany). Diabetes was defined by two consecutive readings of ≥250 mg/dL glucose. Diabetic mice were excluded from analyses. Mice received saline or mIL-2/CD25 (5 µg, intraperitoneally) (5,6) (Fig. 1A). Animal experiments were approved by the Institutional Animal Care and Use Committee, University of Miami, Miami, FL.

Mice were perfused with 4% (wt/vol) paraformaldehyde. The pancreas was excised and postfixed in 4% paraformaldehyde overnight. Tissue was cryoprotected (30% sucrose), then cut into sections (40 μm). After permeabilization (0.3% Triton X-100; Sigma-Aldrich, St. Louis, MO), sections were incubated in blocking buffer (BioGenex, Fremont, CA). Primary antibodies were diluted in blocking buffer: insulin (1:5; Agilent, Santa Clara, CA), glucagon (1:500; Sigma Aldrich, St. Louis, MO), and CD45 (1:1,000; BioLegend, San Diego, CA). Cell nuclei were stained with DAPI.

Confocal images (pinhole = airy 1) were acquired on a Leica SP5 confocal laser-scanning microscope using a ×40 magnification objective (numerical aperture 0.8; Leica, Wetzlar, Germany). Images were analyzed using ImageJ (National Institutes of Health, https://imagej.net/ij/index.html).

Intraperitoneal glucose and insulin tolerance tests (IPGTTs and ITTs) were performed after overnight fasting. Mice were injected with glucose solution (2 g/kg body weight) or insulin (0.75 units/kg insulin; Humulin; Eli Lily, Indianapolis, IN). Glucose was monitored at predetermined time points, and blood samples were collected in capillary tubes coated with EDTA (Sarstedt, Nümbrecht, Germany). Plasma insulin levels were measured by ELISA (Mercodia, Uppsala, Sweden).

Tissue slices were prepared from 8-, 12-, 17-, and 22-week-old NOD or NSG mice, as described previously (21) (Fig. 1B). Slices were incubated in HEPES-buffered solution (125 mmol/L NaCl, 5.9 mmol/L KCl, 2.56 mmol/L CaCl₂, 1 mmol/L MgCl₂, 25 mmol/L HEPES, 0.1% BSA [wt/vol], pH 7.4) with 3 mmol/L glucose and 100 µg/mL soybean trypsin inhibitor (Sigma-Aldrich, St. Louis, MO).

Five slices were placed in perifusion columns (PERI4 perifusion system; Biorep Technologies, Miami Lakes, FL). Tissue slices were perfused at 100 μL/min, and perifusate was collected every minute. Slices were washed for 90 min with 3 mmol/L glucose. Slices then were perfused with 3 mmol/L glucose (10 min), 16.7 mmol/L glucose (30 min), 3 mmol/L glucose (20 min), 60 mmol/L KCL (5 min), and 3 mmol/L glucose (10 min). Insulin and proinsulin were measured via ELISA (Mercodia, Uppsala, Sweden). After perifusion, slices were removed from chambers and placed in 500 μL acid ethanol. The solution was diluted 1:500, and insulin content of tissue slices was measured by ELISA.

Living tissue slices were incubated in 3 mmol/L glucose with 6 μmol/L Fluo-4 AM dye (1 h) (Thermo Fisher, Waltham, MA) (Fig. 1C). Bath applied stimuli were high glucose (16.7 mmol/L), kainate (100 μmol/L), and KCl (30 mmol/L, 3 mmol/L glucose in all solutions unless otherwise specified). Z-stacks of nine confocal planes were acquired every 5 s using an SP8 confocal laser-scanning microscope (Leica, Wetzlar, Germany).

We selected regions of interest around cells that responded to KCl and measured changes in mean [Ca²⁺]_i fluorescence intensity. Raw Ca²⁺ data were detrended and converted to percent change over baseline (% ΔF/F) using a MATLAB script (Mathworks, Natick, MA). Data were displayed as heat maps to include all recorded β-cells. β-cells were distinguished based on morphology, location, and lack of response to kainate (100 μmol/L). Non-β-cells were defined as cells responding to kainate or a drop in glucose concentration (Supplementary Fig. 1A) and were not included in analyses. Glucose-responsive β-cells were those with a mean fluorescence intensity greater than 1 SD of baseline. Glucose responses were analyzed for time to peak, height of peak, and response slope (Supplementary Fig. 1B). Independent, blinded observers quantified percentages of oscillating β-cells.

Statistical comparisons were performed using unpaired Student t test or one-way ANOVA with Tukey multiple comparison test. Statistical analysis of IPGTTs and ITTs was performed using two-way ANOVA followed by Šidák multiple comparison test. Simple linear regression models were fit to data comparing hormone secretion and content. Slope and y-intercept of linear regressions were compared, and a P value was calculated using hypothesis testing. Significance of differences between proportions were assessed using the χ² test. Data are shown as mean ± SD/SE/interquartile range as specified in each figure legend.

Further information and requests for resources, reagents, and data should be directed to and will be fulfilled by F.M.Q.

Short-term low-dose mIL-2/CD25 treatment (5–10 weeks) delays onset of type 1 diabetes in NOD mice by expanding Tregs and promoting peripheral tolerance (5,6). Notably, mice remained normoglycemic for extended times off therapy, after Tregs returned to homeostatic numbers. This finding raised the possibility that mIL-2/CD25 improves β-cell function.

To study how mIL-2/CD25 affects β-cell function before diabetes ensues, we treated 12-week-old NOD mice for 5 weeks with mIL-2/CD25 or saline (Fig. 1A). During the period that precedes the onset of hyperglycemia (pre–type 1 diabetes), NOD mice develop glucose intolerance, reproducing preclinical stages during the progression to type 1 diabetes (10) (Fig. 1A). There is, however, variability in diabetes onset, with a median at 18 weeks in female NOD mice (22,23). Therefore, to ensure that experiments were conducted in the prediabetic state and not in overt diabetes, we excluded mice that became hyperglycemic (>250 mg/dL, two consecutive readings) (Fig. 2A). At 17 weeks of age, islets in pancreatic slices from treated and control mice displayed insulitis but had a preserved cytoarchitecture with ∼20% α-cells and ∼80% β-cells (Fig. 1B–E), as reported for the prediabetic NOD mouse (24).

Data from clinical trials have shown that immunotherapies targeting type 1 diabetes (e.g., anti-CD3 mAb) can improve in vivo β-cell function in patients with pre–type 1 diabetes as measured by glucose tolerance tests (15). We therefore measured glucose tolerance as well as plasma insulin and proinsulin levels during IPGTT in mIL-2/CD25-treated and control NOD mice. Before treatment, at 12 weeks of age, all mice were glucose tolerant and displayed insulitis (Fig. 1C and D) (Supplementary Fig. 2), as described (10,25). By 17 weeks, 18% of control mice were overtly diabetic (Fig. 2A). By contrast, only one mouse from the mIL-2/CD25-treated group was diabetic by 17 weeks (3%), and two mice were diabetic by 22 weeks (9%), in line with our findings that this treatment prevents diabetes in the long term (5,6).

Prediabetic control mice progressed to stage 2 pre–type 1 diabetes (i.e., higher fed glycemia and larger glucose excursion in the IPGTT) (Fig. 2B–D) by 17 weeks, whereas mIL-2/CD25-treated mice had lower fed glycemia and remained glucose tolerant 5 weeks after cessation of therapy, at 22 weeks of age. While fasting and glucose-stimulated insulin levels during the IPGTT were not significantly different (Fig. 2E), 17-week-old mIL-2/CD25-treated mice had a threefold change in insulin levels in response to glucose, significantly greater than the twofold change in control mice (Fig. 2F).

To assess whether mIL-2/CD25 improves glucose tolerance and insulin secretion, we performed a paired assessment within the same mouse group at 12 and 17 weeks of age. Between 12 and 17 weeks of age, mIL-2/CD25-treated mice remained glucose tolerant and increased their insulin response (fold change) during the IPGTT (Supplementary Fig. 3A and B). By contrast, saline-treated mice became glucose intolerant, and their insulin response did not increase.

Elevations in plasma proinsulin:insulin or proinsulin:C-peptide ratios precede onset of type 1 diabetes and suggest β-cell endoplasmic reticulum (ER) stress (26–28). We measured this ratio during the peak of insulin secretion in the IPGTTs. The mIL-2/CD25 group had a significantly higher proinsulin:insulin ratio at 17 and 22 weeks (Fig. 2G).

We also found that mIL-2/CD25-treated mice had lower fasting insulin levels at 17 weeks (Supplementary Fig. 3C). We thus examined insulin sensitivity (ITTs). At 17 weeks, both groups had comparable drops to ∼25% of baseline glycemia, indicating similar insulin sensitivities (Supplementary Fig. 3D). Overall, our results indicate that mIL-2/CD25 therapy prevents progressive loss of glucose homeostasis and preserves β-cell function in the NOD mouse.

Our in vivo data showed that glucose homeostasis was maintained for at least 5 weeks after mIL-2/CD25 treatment, suggesting that the secretory capacity of β-cells was preserved. To measure dynamic hormone secretion in response to high glucose (16.7 mmol/L) and KCl (60 mmol/L), we conducted perifusion studies with living pancreatic slices from 17- and 22-week-old NOD mice. We found that glucose-induced insulin secretion was higher after mIL-2/CD25 treatment and continued to increase after cessation of treatment (Fig. 3A–C) (Supplementary Fig. 4A and B). When normalized to total insulin content in slices, insulin secretion was greater in the mIL-2/CD25-treated group (Supplementary Fig. 4A and B). Because of variability in slice insulin contents (Fig. 3B), we plotted data as insulin secretion (area under the curve) versus insulin content (Fig. 3C). Linear regressions showed that islets from both mIL-2/CD25 and saline-treated mice had a positive relationship between the insulin content and amount of insulin secreted (slope of linear regression). Linear regressions also showed that insulin secretion from mIL-2/CD25-treated mice was increased for any level of insulin content (y-intercept of linear regression) and continued to increase 5 weeks after stopping mIL-2/CD25 therapy (22-week-old mice).

We also measured proinsulin secretion (Fig. 3D–F) (Supplementary Fig. 4C and D). Proinsulin content, like insulin content, varied among slices (Fig. 3E). When plotting proinsulin secretion against proinsulin content, linear regression indicated no relationship between proinsulin content and proinsulin secretion; regression slopes for all groups were not significantly different from zero (Fig. 3F). Furthermore, there was no significant difference between y-intercepts of the three groups, indicating that the total amount of proinsulin secretion was not significantly different.

Next, we compared the ratio of proinsulin to insulin secreted from the slices in the perifusion experiments (Fig. 3G–I). The mIL-2/CD25 group had a significantly greater proinsulin:insulin ratio in their slices at both 17 and 22 weeks (Fig. 3H).

We plotted proinsulin secretion as a function of insulin secretion. The slope of the linear regression indicated that, at 17 weeks, islets from mIL-2/CD25 mice secreted more proinsulin per amount of insulin secretion than saline controls. However, by 22 weeks, there was no longer a relationship between insulin and proinsulin secretion (Fig. 3I). These data are in line with our in vivo results showing a greater fold change in insulin secretion in response to glucose and a higher plasma proinsulin:insulin ratio.

Measuring [Ca²⁺]_i responses in the β-cell population helps identify changes in stimulus secretion coupling (from glucose sensing to insulin exocytosis) and β-cell synchronicity that are related to functional status (29). Thus, it is important to understand how [Ca²⁺]_i dynamics change in pre–type 1 diabetes. A healthy Ca²⁺ response to glucose in a β-cell is triphasic—phase 0 is an initial dip in the [Ca²⁺]_i levels (caused by ER uptake of cytoplasmic Ca²⁺), followed by a sharp rise in [Ca²⁺]_i levels (phase 1), and a plateau with oscillations (phase 2) (29) (Fig. 4A). NSG mice, which have no immune infiltration, showed this healthy response (Fig. 4A). NSG mice, which do not become diabetic, had the highest proportion of glucose-responsive cells (75%) (Fig. 4B). The proportion of glucose-responsive cells decreased over time; 68% of β-cells responded in 8-week-old NOD mice, dropping to 67% by 12 weeks and 51% by 17 weeks. Eight-week-old female NOD mice, which are at an earlier stage of pre–type 1 diabetes, had [Ca²⁺]_i responses similar to NSG mice, though they had a lower proportion of cells showing oscillatory behavior (Fig. 4C). By 17 weeks, when NOD mice are in stage 2 pre–type 1 diabetes, [Ca²⁺]_i responses were impaired, and no cells had oscillatory behavior.

NSG, 12-week-old, and 8-week-old NOD mice had sharper, more robust increases in peak [Ca²⁺]_i in response to glucose than 17-week-old NOD mice (Fig. 4D and E). The latency of these responses also decreased over time (Fig. 4F), which is thought to be a compensatory mechanism for impaired insulin secretion (30).

Highly coordinated and synchronized [Ca²⁺]_i responses are another feature of a healthy islet. Β-cells are electrically coupled, and intact islets will respond to stimuli as a single unit, which is important for the pulsatile insulin secretion of healthy individuals (29). We quantified β-cell synchronicity as the average Pearson correlation coefficient of [Ca²⁺]_i responses in all β-cell pairs, as described previously (31). The closer the average correlation to one, the greater the degree of synchronicity. Our data show that responses to glucose are highly coordinated in NSG mice (healthy controls), and become less coordinated in NOD mice between 8 and 17 weeks of age (Fig. 5A and B). As an internal control, we measured the synchronicity of β-cells in response to KCl. As expected, all islets had a high degree of synchronicity when depolarized (Fig. 5C and D). The overall decline in the incidence, dynamics, and synchronicity of [Ca²⁺]_i responses between 8 and 17 weeks in NOD mice indicates a progressive loss of β-cell function from stage 1 to stage 2 pre–type 1 diabetes.

We next investigated the Ca²⁺ dynamics in β-cells in pancreas slices from NOD mice treated with mIL-2/CD25 or saline for 5 weeks. We also investigated these dynamics 5 weeks after mIL-2/CD25 therapy was discontinued, in 22-week-old NOD mice. A significantly higher proportion of cells responded to glucose in the mIL-2/CD25 group (Fig. 6A–D). β-Cells in islets from mIL-2/CD25-treated animals had sharper response slopes at 17 and 22 weeks (Fig. 6E–G), and higher peak [Ca²⁺]_i and response latencies at 22 weeks (Fig. 6E, F, and H). Our data show that β-cells become more synchronous after mIL-2/CD25 treatment and continue to improve 5 weeks after therapy is stopped (Fig. 7A and B). Responses to KCl depolarization were not affected at any stage or treatment (Figs. 5, 6, and 7C), indicating a specific defect in glucose responsiveness. In sum, these findings indicate that treatment with mIL-2/CD25 improved aspects of β-cell health, and that these improvements persist.

mIL-2/CD25 is a promising therapy for type 1 diabetes and has been proven to prevent diabetes in NOD mice more efficiently than equimolar recombinant IL-2 (5,6). It prevents diabetes by inducing proliferation and activating Tregs in secondary lymphoid organs and the pancreatic islet, which reduces inflammation and prevents further destruction of β-cells. Our study establishes that, secondary to changes in immune infiltration, mIL-2/CD25 therapy also restores β-cell function. While it has been shown that IL-2 may also increase β-cell proliferation in NOD mice (32,33), we demonstrate that, independent of cell number, individual β-cell and islet function improves with mIL-2/CD25 treatment. The treatment improved [Ca²⁺]_i responses to glucose, in vivo and in vitro insulin secretion, and glucose tolerance. Intervention with mIL-2/CD25 delayed or reversed the progressive loss of β-cell function. These effects on β-cells likely contribute to diabetes control in many NOD mice after cessation of mIL-2/CD25 therapy (5,6), as we now confirm at the level of individual β-cell physiology.

NOD mice mimic human type 1 diabetes in that there is a progressive decline in β-cell function before diagnosis (10). This prediabetic state has two stages: the first begins with autoimmunity, and the second begins with dysglycemia (11). Our study now provides a comprehensive and accurate picture of how β-cell function decays during the first two stages of pre–type 1 diabetes in NOD mice (Figs. 4 and 5). When treating type 1 diabetes with immunotherapy, it is important to start while there is still a significant pool of functional β-cells. While most trials are conducted in new-onset patients, prevention trials target patients in stages 1 and 2 (34). Our model with 12-week-old NOD mice mirrors these prevention trials in which autoantibody-positive patients are enrolled. At this age, immune cell infiltration is widespread, but β-cells are still functional and can be rescued from further dysfunction (35,36). We show that targeting autoimmunity at this early stage prevents not only overt diabetes (5,6) but also progression from stage 1 to stage 2 pre–type 1 diabetes.

In clinical and preclinical models, glucose tolerance is an indirect measure of β-cell function used in the staging of pre–type 1 diabetes to distinguish stage 1 from stage 2, where β-cell defects first occur (11). We show that mice in stage 1 pre–type 1 diabetes at 12 weeks of age progress to stage 2 by 17 weeks (Fig. 2B–D). Treatment with mIL-2/CD25 limited the progressive decline in β-cell function, even 5 weeks after treatment cessation. We also found that treated mice had healthier insulin secretory phenotypes. At 17 weeks, mIL-2/CD25-treated mice had a higher fold change insulin secretion to glucose stimulation compared with saline controls. This mirrors clinical data, where normoglycemic individuals have a higher fold change in insulin secretion than glucose-intolerant individuals, despite secreting less insulin overall (37,38). Our in vitro results also showed enhanced insulin secretion after treatment with mIL-2/CD25.

In addition to insulin, we measured proinsulin secretion, because it is often used as a surrogate marker for β-cell ER stress (26,27,39). Because cytokines from infiltrating immune cells are known to cause ER stress (26,40), we expected that reducing inflammation with mIL-2/CD25 would decrease proinsulin secretion. Given the general improvement of β-cell function, it was surprising that our in vivo results showed a higher plasma proinsulin:insulin ratio in mIL-2/CD25-treated mice (Fig. 2G). It is likely that there is still immune infiltration of some islets in mIL-2/CD25-treated mice that could induce ER stress (6). But this higher ratio may not necessarily indicate pathogenic ER stress; it could be a function of enhanced secretion from the β-cell. When β-cells increase secretory capacity, the proinsulin:insulin ratio increases in the absence of abnormal ER stress. It is thought that this is due to the rapid secretion of new insulin granules, preventing enzymes from cleaving C-peptide from proinsulin before the granule contents are released (41–43). By ameliorating the immune attack stress, mIL-2/CD25 therapy allows β-cells to rapidly secrete more insulin, with the side effect of increasing proinsulin:insulin ratios. Nevertheless, the increase in proinsulin:insulin ratio was not expected and should be addressed in future studies.

mIL-2/CD25 had beneficial effects on β-cell function as determined by [Ca²⁺]_i imaging. NSG mice, which do not have insulitis, had typical triphasic [Ca²⁺]_i responses (Fig. 4A). These [Ca²⁺]_i responses are diminished in the presence of immune infiltration and cytokine secretion (40,44,45). We showed that 8-week-old NOD female mice, which are early in the disease process but do have immune infiltration, had very few oscillating β-cells, indicating incipient β-cell dysfunction (46,47). The [Ca²⁺]_i responses in β-cells of 17-week-old NOD female mice were smaller, sluggish, and no longer oscillatory (Fig. 4). The [Ca²⁺]_i responses to glucose also began earlier, which is thought to help overcome diminished insulin secretion (30). When treated with mIL-2/CD25, β-cells in NOD mice regained aspects of normal [Ca²⁺]_i responses; that is, they were larger, sharper, and more synchronous. Increased synchronicity may prevent over secretion of insulin at baseline (29,47), in line with our in vivo results showing lower fasting insulin levels. Strikingly, 5 weeks after treatment cessation, β-cell function continued to improve.

Our study in a preclinical model of stage 1 type 1 diabetes provides further support for translating mIL-2/CD25 to clinical trials. While a major focus for type 1 diabetes therapeutics has rightfully been on preventing autoimmunity, effects on β-cell function have been relatively neglected (48). Recent clinical trials include end points such as glucose tolerance and C-peptide levels that estimate β-cell function (15–17). Targeting this aspect of diabetes pathogenesis is important because preserving β-cell function early in the disease process reduces long-term sequelae (18–20). Our study with mIL-2/CD25 lends mechanistic insight into how immunotherapies for type 1 diabetes exert their effects on β-cell function and provides a framework for future studies on type 1 diabetes therapeutics.

Acknowledgments. The authors thank Alicia Santos Savio (Department of Microbiology and Immunology, University of Miami Miller School of Medicine, Miami, FL) and Nicholas Borowsky (Medical Scientist Training Program, University of Miami Miller School of Medicine, Miami, FL) for technical assistance.

Funding. This research was supported by National Institutes of Health grants (National Institute of Diabetes and Digestive and Kidney Diseases F30DK126310 [F.M.Q.]; National Institute of Allergy and Infectious Diseases R01AI148675 [T.R.M.]; National Institute of Diabetes and Digestive and Kidney Diseases R01DK084321 [A.C.], R01DK111538 [A.C.], R01DK113093 [A.C.], and U01DK120456 [A.C.]; National Institute of Environmental Health Sciences R33ES025673 [A.C.]; and National Institute of Diabetes and Digestive and Kidney Diseases R01DK130328 [A.C.]); and Leona M. and Harry B. Helmsley Charitable Trust grants G-2018PG-T1D034 (A.C.) and G-1912-03552 (A.C.).

Duality of Interest. The University of Miami and T.R.M. have a patent pending on IL-2/CD25 fusion proteins (WO2016022671A1) that has been licensed exclusively to Bristol-Myers Squibb. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. F.M.Q., J.K.P., and J.P. performed experiments. F.M.Q., J.K.P., and J.P. analyzed data. F.M.Q., T.R.M., and A.C. designed the study and wrote the paper. All authors discussed the results and commented on the manuscript. A.C. 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.