β-Cell–Specific IL-2 Therapy Increases Islet Foxp3⁺Treg and Suppresses Type 1 Diabetes in NOD Mice

NOD/LtJ, NOD.CB17-Prkdcscid/J (NOD.scid), NOD.Cα^−/−, NOD.BDC (39), NOD.BDC.Foxp3^GFP (40), and NOD.8.3 (41) mice were bred and maintained under specific pathogen-free conditions in an American Association for Laboratory Accreditation–accredited animal facility. NOD mice were diagnosed as diabetic after three consecutive blood glucose readings >250 mg/dL. All procedures were approved by the University of North Carolina Animal Use and Care Committee.

Full-length cDNA encoding murine Il2 (of the NOD genotype) and EGFP were subcloned into a dsAAV plasmid (37) containing the murine preproinsulin II promoter (mIP). dsAAV vector packaging was carried out as previously described (42). Briefly, HEK 293T cells were cotransfected via polyethyleneimine with adeno helper plasmid (pXX6–80), AAV8 capsid plasmid, and dsAAVmIP-IL2/enhanced green fluorescent protein (EGFP) plasmid. Nuclear fractions were harvested 72 h posttransfection and packaged vector purified with a Cesium Chloride (RPI Corp.) gradient. Vector containing fractions and titers were determined by Southern dot blot. NOD mice were vaccinated intraperitoneally with dsAAV8mIP-IL2 or dsAAV8mIP-EGFP.

The anti–IL-2 antibody set (JES6-1 and JES6-5; eBioscience) was used at a concentration of 2 μg/mL on a high binding ELISA plate (Costar) to measure IL-2 levels in serum and islet cultures. Interferon-γ (IFN-γ) was measured with the mouse IFN-γ ELISA set (BD optEIA; BD Biosciences) as per the manufacturer’s recommendations.

Pancreases were perfused with 2 mg/mL collagenase P (Roche) and digested for 30 min at 37°C. Islets were purified via Ficoll gradient, handpicked, and counted. For flow cytometry analysis, lymphocytes were collected from freshly isolated islets after culturing in RMPI complete medium for a minimum of 4 h at 37°C and filtered through a 70-micron nylon mesh.

NOD mice were treated intraperitoneally with 200 mg/kg of streptozotocin (STZ; Sigma-Aldrich), resuspended in fresh sodium citrate buffer (pH 4.0), and 72 h later vaccinated intraperitoneally with AAV8mIP-IL2 or left untreated. Blood glucose levels and serum IL-2 were assessed over 10 days post–rAAV vector vaccination.

Cells from various tissues were stained with previously titrated fluorochrome-conjugated monoclonal antibodies specific for: CD3 (145-2C11), CD4 (GK1.5), CD8 (Ly-2), CD25 (PC61.5), CD62L (MEL-14), glucocorticoid-induced tumor necrosis factor receptor–related protein (GITR, DTA-1), inducible costimulator (ICOS, 7E.17G9), cytotoxic T lymphocyte–associated antigen-4 (CTLA-4, UC10-4F10-11), and Foxp3 (FJK-16s). Fc receptors were blocked with rat anti-mouse CD16/32 (2.4G2; BD Biosciences) prior to staining. Intracellular KI67 (B56; BD Biosciences) and Bcl-2 (3F11; BD Biosciences) staining was done using the Fix/Perm and Perm/Wash reagents (eBioscience) according to the manufacturer's recommendations.

Intracellular cytokine staining was performed on single-cell suspensions as previously described (43). Briefly, lymphocytes were stimulated with 500 ng/mL phorbol myristic acid (Sigma-Aldrich) and 1,000 ng/mL ionomycin (Sigma-Aldrich) in complete RPMI medium for 5 h at 37°C and then treated with 10 μg/mL brefeldin A (Sigma-Aldrich) for 4 h. Cells were stained for surface molecules, fixed, and permeabilized with Cytofix/Cytoperm (BD Biosciences) and stained for intracellular IFN-γ (XMG1.2) or IL-10 (JES5-16E3). For pSTAT5 staining, tissues were immediately processed in PBS containing 2% paraformaldehyde. Single-cell suspensions were incubated on ice for 20 min, centrifuged, and resuspended in ice-cold methanol for 30 min. Cells were then counted, washed twice in 1% BSA in PBS, and stained with anti-pSTAT5 antibody (47/Stat5 [pY694]; BD Biosciences) for 1 h on ice. Data were acquired on a Cyan flow cytometer (DakoCytomation) and analyzed using Summit software (DakoCytomation).

Islets were pooled from NOD.BDC.Foxp3^GFP mice vaccinated with 2.5 × 10¹⁰ vector particles (VPs) of AAV8mIP-IL2 or AAV8mIP-EGFP, and single-cell suspensions stained with CD4- and CD3-specific antibodies in PBS supplemented with 2.5% BSA and 2 mmol/L EDTA. CD4⁺CD3⁺GFP⁺ Tregs were sorted on a MoFlo sorter (DakoCytomation). For T cell responders, naive CD4⁺CD62L⁺ splenocytes were isolated from NOD.BDC mice using the CD4⁺CD62L⁺ T cell isolation kit II (Miltenyi Biotec) and labeled with Cell Trace Violet (Invitrogen) as per the manufacturer’s recommendations. Cell purity in naive populations was >95% by flow cytometry. Splenocytes from NOD.Cα^−/− mice pulsed with 1 μg/mL sBDC mimetic peptide were used as antigen-presenting cells (APCs). Naive BDC CD4⁺ T cells (2.5 × 10⁴), APCs (2.5 × 10⁴), and varying ratios of Foxp3⁺Tregs were cocultured in complete RPMI medium for 72 h and proliferation of BDC CD4⁺ T cells assessed by dilution of Cell Trace Violet via flow cytometry.

To target transgene expression to β-cells, dsAAV8mIP-IL2 encoding IL-2 driven by mIP was engineered. The serotype capsid 8 protein used to package the AAV vector has β-cell–specific tropism (35,36). IL-2 was detected in cultures of islets prepared from NOD.scid mice injected intraperitoneally with dsAAV8mIP-IL2, confirming ectopic IL-2 expression by β-cells in vivo (Fig. 1A). No IL-2 was observed in supernatants from single-cell suspensions of liver or heart (data not shown), which are major tissues transduced in vivo by AAV8 vectors (44). Persistent low levels of IL-2 were detected in the serum of 10–12-week-old NOD female mice treated with dsAAV8mIP-IL2 (Fig. 1B). Reduction of β-cell mass by STZ injection prior to dsAAV8mIP-IL2 vaccination significantly decreased serum IL-2 levels (Fig. 1C), further indicating ectopic IL-2 expression by β-cells in vivo. No difference in the frequency of conventional CD4⁺ and CD8⁺ T cells, B cells (CD19⁺B220⁺), and NK cells (DX5⁺) was detected in the blood of dsAAV8mIP-IL2–vaccinated animals compared with control groups (Fig. 1D).

The effect of increased local IL-2 on islet and PLN Foxp3⁺Tregs was examined in 12-week-old NOD female mice 4 weeks after dsAAV8mIP-IL2 treatment. The frequency and number of islet Foxp3⁺Tregs were increased up to two-fold in NOD mice treated with dsAAV8mIP-IL2 compared with animals injected with dsAAV8mIP-EGFP or left untreated (Fig. 2A and B). Notably, no significant effect on PLN Foxp3⁺Tregs was detected after treatment (Fig. 2A and B), suggesting that dsAAV8mIP-IL2 administration preferentially affected islet Foxp3⁺Tregs.

Phenotypic analyses demonstrated that the frequency of islet CD62L^HIFoxp3⁺Tregs was significantly increased in dsAAV8mIP-IL2–treated versus control animals (Fig. 2C). In addition, the level of CD25 expression by islet Foxp3⁺Tregs, based on mean fluorescence intensity, was elevated in a dsAAV8mIP-IL2 dose-dependent manner (Fig. 2A and D). In contrast, the frequency of islet Foxp3⁺Treg-expressing CTLA-4, GITR, and ICOS and the levels of expression of the respective molecules were similar between the dsAAV8mIP-IL2 and control groups (Supplementary Fig. 1). Unlike the islets, the phenotype of PLN Foxp3⁺Tregs was unaffected by dsAAV8mIP-IL2 vaccination (Fig. 2).

In contrast to islet Foxp3⁺Tregs, no change in the number of conventional IFN-γ⁺CD4⁺ and CD8⁺ Teffs was detected in the islets of dsAAV8mIP-IL2–treated NOD mice (Fig. 2E). However, the ratio between the pool of IFN-γ⁺ Teffs and Foxp3⁺Tregs in the islets but not PLNs was decreased in dsAAV8mIP-IL2 versus untreated and dsAAV8mIP-EGFP cohorts (Fig. 2F).

To determine if the increased islet Foxp3⁺Treg pool correlated with elevated IL-2R signaling, levels of pSTAT5 were measured. Up to 50% of islet Foxp3⁺Tregs were pSTAT5⁺, reflecting an ∼10-fold increase in dsAAV8mIP-IL2 versus control groups (Fig. 3A). Only a modest two-fold increase in the frequency of pSTAT5⁺ cells was detected in conventional Foxp3⁻CD4⁺ T cells in the islets of NOD mice vaccinated with 2.5 × 10¹⁰ VPs dsAAV8mIP-IL2 (6.4 ± 1.1%) relative to control animals (3.6 ± 0.6%) (Fig. 3B). In contrast, levels of pSTAT5 detected in Foxp3⁺Tregs or Foxp3⁻CD4⁺ T cells in the PLNs were unaffected by β-cell–specific IL-2 (Fig. 3). Together, these findings demonstrate that β-cell–specific IL-2 expression results in an increased pool of phenotypically distinct islet but not PLN Foxp3⁺Tregs marked by elevated IL-2R signaling.

Enhanced survival and/or expansion of Foxp3⁺Tregs may explain the increased islet Foxp3⁺Treg pool in dsAAV8mIP-IL2–treated animals. To determine if AAV8mIP-IL2 vaccination increased islet Foxp3⁺Treg survival, expression of the antiapoptotic protein Bcl-2 was measured. The level of Bcl-2-expression following dsAAV8mIP-IL2 treatment was increased in islet Foxp3⁺Tregs (Fig. 4A) and a trend toward an elevated frequency of Bcl-2⁺ Foxp3⁺Tregs observed (Fig. 4B). In contrast, Bcl-2 expression was unaffected in PLN Foxp3⁺Tregs (Fig. 4A and B) or in islet IFN-γ⁺CD4⁺ and CD8⁺ Teffs (Fig. 4C).

Proliferation of islet and PLN Foxp3⁺Tregs after dsAAV8mIP-IL2 administration was analyzed by KI67 staining. The frequency of KI67⁺Foxp3⁺CD25⁺ T cells was significantly increased in the islets by 8 days post–dsAAV8mIP-IL2 treatment (Fig. 4D). In contrast, the frequency of KI67⁺Foxp3⁺CD25⁺ T cells in the PLNs remained unchanged over time (Fig. 4D), as was the frequency of KI67-staining islet IFN-γ⁺CD4⁺ and CD8⁺ Teffs (Fig. 4E).

Next, the influence of β-cell–specific IL-2 on islet Foxp3⁺Treg suppressor activity was assessed. NOD.BDC mice transgenic for a Foxp3–internal ribosome entry site–GFP reporter cassette (NOD.BDC.Foxp3^GFP) were vaccinated with dsAAV8mIP-IL2 or dsAAV8mIP-EGFP and GFP⁺Foxp3⁺Tregs sorted from the islets 4 weeks later. The capacity of the respective Foxp3⁺Tregs to suppress Cell Trace Violet–labeled naive BDC CD4⁺ T cells following stimulation by peptide-pulsed APC was then measured. Islet Foxp3⁺Tregs from dsAAV8mIP-IL2–treated NOD.BDC.Foxp3^GFP mice exhibited significantly increased suppressor activity compared with islet Foxp3⁺Tregs isolated from dsAAV8mIP-EGFP donors at all ratios tested (Fig. 4F). Together, these results demonstrate that β-cell–specific IL-2 enhances the fitness of islet Foxp3⁺Tregs characterized by increased Bcl-2 expression, proliferation, and suppressor function.

Since β-cell–specific IL-2 expression increased islet Foxp3⁺Tregs, the efficacy of AAV8mIP-IL2 to suppress ongoing β-cell autoimmunity and prevent overt diabetes was examined. NOD female mice 12 or 16 weeks of age, representing late preclinical stages of type 1 diabetes, received a single intraperitoneal injection of dsAAV8mIP-IL2, dsAAV8mIP-EGFP, or were left untreated, and diabetes was monitored. All NOD female mice treated with 2.5 × 10¹⁰ VPs of dsAAV8mIP-IL2 at either 12 or 16 weeks of age remained diabetes-free (Fig. 5A and B). Similarly, the frequency of diabetes was significantly reduced in NOD female mice treated at 12 weeks of age with 1 × 10¹⁰ VPs of dsAAV8mIP-IL2 (2 of 8) relative to animals left untreated (18 of 22) or receiving 2.5 × 10¹⁰ VPs dsAAV8mIP-EGFP (10 of 10) (Fig. 5A). The latter indicated that diabetes protection was not attributed to a nonspecific effect of dsAAV vector transduction of β-cells.

Diabetes-free NOD mice at 35 weeks of age exhibited a higher frequency of islet Foxp3⁺Tregs in dsAAV8mIP-IL2 versus control groups, while no marked difference in the PLNs was observed (Fig. 5C). Furthermore, islet Foxp3⁺Tregs of dsAAV8mIP-IL2–treated NOD mice were characterized by an increased frequency of CD62L^HI Foxp3⁺Tregs (Fig. 5D). Additionally, the ratio of IFN-γ⁺ Teffs to Foxp3⁺Tregs was decreased in the islets of dsAAV8mIP-IL2–treated NOD mice compared with normoglycemic untreated controls (Fig. 5E). No difference, however, was detected in the IFN-γ⁺ Teffs to Foxp3⁺Tregs ratio for PLN of dsAAV8mIP-IL2–vaccinated animals (Fig. 5E). These results demonstrate that β-cell–specific IL-2 expression effectively suppresses ongoing autoimmunity long-term by enhancing the islet Foxp3⁺Treg pool.

IL-2 is essential for the homeostasis of Foxp3⁺Tregs, and dysregulation of the IL-2–IL-2R axis has been linked to aberrant Foxp3⁺Treg survival and/or function in NOD mice and type 1 diabetic patients (9–12). Systemic administration of low-dose rIL-2 or IL-2–antibody complexes is effective in preventing ongoing β-cell autoimmunity in NOD mice (16,25–29). Nevertheless, achieving a protective versus proinflammatory response via systemic rIL-2 delivery can be problematic. For instance, systemic low-dose rIL-2 plus rapamycin resulted only in a transient increase in FOXP3⁺Tregs in type 1 diabetic subjects, and β-cell function declined, suggesting exacerbated autoimmunity (30). Our results demonstrate that dsAAV vector–mediated targeting of IL-2 expression to β-cells is an effective strategy to manipulate the islet Foxp3⁺Treg pool and diabetogenic response, while avoiding complications associated with systemic delivery of rIL-2.

A key finding made in this study was the differential effect of β-cell–specific IL-2 on islet-infiltrating Foxp3⁺Tregs and Teffs (Fig. 2). Elevated local levels of IL-2 resulted in marked expansion of islet Foxp3⁺Tregs but not Teffs. This observation is consistent with studies demonstrating that Foxp3⁺Tregs are more sensitive to IL-2 compared with naive T cells and Teffs, partly due to constitutive expression of CD25 by Foxp3⁺Tregs (45). Indeed, the level and frequency of IL-2R signaling as measured by pSTAT5 was significantly increased in islet Foxp3⁺Tregs versus Teffs in dsAAV8mIP-IL2–treated NOD mice (Fig. 3). No effect of β-cell–specific IL-2 expression on PLN-resident Foxp3⁺Tregs was detected (Figs. 2 and 3), presumably due to β-cell secretion of low levels of IL-2 into the bloodstream and not the lymphatics. The increased number of islet Foxp3⁺Tregs correlated with enhanced proliferation early after dsAAV8mIP-IL2 treatment (Fig. 4D). Interestingly, islet Foxp3⁺Tregs division was reduced compared with untreated animals 4 weeks post–dsAAV8mIP-IL2 vaccination (Fig. 4D). This result suggests that homeostasis is achieved overtime within the Foxp3⁺Treg pool despite increased local levels of IL-2 and elevated IL-2R signaling (Fig. 3). Accordingly, long-term maintenance of islet Foxp3⁺Tregs in dsAAV8mIP-IL2–treated animals (Fig. 5C and D) is likely due to enhanced survival, reflected by increased expression of anti-apoptotic Bcl-2 (Fig. 4A) and Bcl-xL (Supplementary Fig. 2) as opposed to persistent levels of elevated Foxp3⁺Treg proliferation in the islets.

In addition to increasing the size of the pool, β-cell–specific IL-2 also induced qualitative changes in islet Foxp3⁺Tregs. For instance, up to a four-fold increase in suppressor activity was detected in vitro by BDC peptide–specific islet Foxp3⁺Tregs from dsAAV8mIP-IL2–treated NOD.BDC.Foxp3^GFP mice (Fig. 4F). The latter may be explained by the approximate two-fold increase in the frequency of CD62L^HI Foxp3⁺Tregs in the islets of dsAAV8mIP-IL2–treated NOD mice (Fig. 2C). We and others (6,19,46,47) have reported that CD62L^HI versus CD62L^LO Foxp3⁺Tregs exhibit increased suppressor activity and that elevated systemic levels of IL-2 promote expansion of CD62L^HI Foxp3⁺Tregs in vivo. An increase in the general fitness of islet Foxp3⁺Tregs, marked by elevated levels of CD25 and anti-apoptotic molecules for instance, may also contribute to more robust suppressor function. Enhanced islet Foxp3⁺Treg suppression, however, did not correlate with elevated levels of IL-10 or transforming growth factor-β1 expression (data not shown).

Importantly, increases in islet Foxp3⁺Tregs by β-cell–expressed IL-2 were sufficient to suppress ongoing β-cell autoimmunity at late preclinical stages and prevent the onset of diabetes long-term (Fig. 5A and B). Initial efforts to induce remission in recent-onset diabetic NOD mice by dsAAV8mIP-IL2 vaccination have proven to be unsuccessful (Supplementary Fig. 3). The latter may be due to an insufficient β-cell pool that can be transduced, thereby limiting local levels of IL-2 (Supplementary Fig. 3). It is noteworthy that the reduced ratio of Teffs to Foxp3⁺Tregs and increased frequency of CD62L^HI Foxp3⁺Tregs in the islets observed 4 weeks post–dsAAV8mIP-IL2 vaccination (Fig. 2) were also evident in NOD mice monitored for diabetes long-term (e.g., up to 23 weeks after dsAAV vaccination) (Fig. 5). This result is particularly striking in view of the progressive increase in the Teffs to Foxp3⁺Tregs ratio, as well as the declining frequency of CD62L^HI Foxp3⁺Tregs apparent in the islets of 35-week-old control NOD female mice (Fig. 5D and E), typical of ongoing β-cell autoimmunity. These findings demonstrate that once established, the islet Foxp3⁺Treg pool is maintained long-term by persistent β-cell–expressed IL-2. Indeed, similar numbers of islet Foxp3⁺Tregs were detected in NOD mice treated 23 (Supplementary Fig. 4) versus 4 weeks (Fig. 2B) post–dsAAVmIP-IL2 vaccination. Whether expression of β-cell–specific IL-2 is required to maintain islet Foxp3⁺Tregs needs to be determined. Interestingly, dsAAV8mIP-IL2 failed to prevent diabetes in transgenic NOD.8.3 mice expressing the H2K^d-restricted T-cell receptor specific for islet-specific glucose-6-phosphatase catalytic subunit-related protein (Supplementary Fig. 5), suggesting that Foxp3⁺Tregs are required for disease prevention; for instance, CD4⁺ Foxp3⁺Tregs make up only <1% of the islet T cell pool following dsAAV8mIP-IL2 vaccination.

This study further highlights the efficacy of dsAAV vector–mediated gene transfer to selectively modify the immunogenicity of β-cells in vivo. Transduction of β-cells in vivo with a dsAAV vector results in rapid transgene expression that persists long-term (Fig. 1B); the latter indirectly demonstrating that β-cell viability and function are unaffected by dsAAV vector transduction. The use of an insulin promoter restricts transgene expression to β-cells, so that high local cytokine levels can be obtained within the islets and complications associated with systemic cytokine delivery minimized (30). A similar approach has been taken to target ectopic IL-4 expression to β-cells (35). Recent technological advances can also be exploited to refine β-cell specificity and reduce immunogenicity of AAV capsid proteins, thereby enhancing application of dsAAV vector gene transfer for the treatment of type 1 diabetes in the clinic (48).

Our observations are consistent with a model in which elevated local IL-2 has direct effects on islet Foxp3⁺Tregs, only minimal (if any) on islet Teffs, and no effect on non–T cell islet residents such as NK cells and B cells (Supplementary Fig. 6). An increased pool of islet Foxp3⁺Tregs characterized by a phenotypic shift toward CD62L^HI Foxp3⁺Tregs and increased fitness and suppressor function effectively blocks ongoing β-cell destruction by Teffs. Once established, the islet Foxp3⁺Treg pool is stable and long-term tolerance maintained. These results provide strong rationale to further pursue β-cell–specific targeting of IL-2 and other cytokines as a strategy to manipulate islet-resident Tregs and block the diabetogenic response.

This work was supported by a grant from the National Institutes of Health (5R01-DK-081585). M.C.J. and A.L.G. were supported by a National Institutes of Health training grant (5T32-AI-07273).

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

M.C.J., A.L.G., and S.C.N. researched data, contributed to discussion, and wrote and edited the manuscript. C.L., R.J.S., B.W., and R.T. contributed to discussion and wrote and edited the manuscript. R.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.

The authors thank Dr. Charles Kroger for critical reading of the manuscript.