Unchecked collaboration between islet-reactive T and B lymphocytes drives type 1 diabetes (T1D). In the healthy setting, CD8 T regulatory cells (Tregs) terminate ongoing T–B interactions. We determined that specific CD8 Tregs from NOD mice lack suppressive function, representing a previously unreported regulatory cell deficit in this T1D-prone strain. NOD mice possess 11-fold fewer Ly-49+ CD8 Tregs than nonautoimmune mice, a deficiency that worsens as NOD mice age toward diabetes and leaves them unable to regulate CD4 T follicular helper cells. As IL-15 is required for Ly-49+ CD8 Treg development, we determined that NOD macrophages inadequately trans-present IL-15. Despite reduced IL-15 trans-presentation, NOD Ly-49+ CD8 Tregs can effectively transduce IL-15–mediated survival signals when they are provided. Following stimulation with an IL-15/IL-15Ra superagonist complex, Ly-49+ CD8 Tregs expanded robustly and became activated to suppress the Ag-specific Ab response. IL-15/IL-15Ra superagonist complex–activated CD8+CD122+ T cells also delayed diabetes transfer, indicating the presence of an underactivated CD8 T cell subset with regulatory capacity against late stage T1D. We identify a new cellular contribution to anti-islet autoimmunity and demonstrate the correction of this regulatory cell deficit. Infusion of IL-15–activated CD8 Tregs may serve as an innovative cellular therapy for the treatment of T1D.

Circulating islet autoantibodies remain the best clinical predictor of type 1 diabetes (T1D) in at-risk patients (1). Mechanistically, this clinical observation results from unchecked anti-islet immunity wherein islet-reactive B lymphocytes are inappropriately activated by islet-reactive T lymphocytes. Clinicians have attempted to halt this collaboration by nonselectively targeting the whole B or T cell compartment with anti-CD20, anti-CD3, or CTLA-4/Ig, but these approaches have not resulted in permanent islet protection (24). Fundamentally, the physiologic regulation of these cellular interactions remains incompletely understood. Identifying pathways that control T–B interactions holds promise to dampen progressive autoimmunity.

Regulation of the Ab response may be carried out by CD4 T regulatory cells (Tregs) (5, 6) and newly identified CD4 T follicular regulatory cells (7), although the effectiveness of general CD4 Tregs against the Ab response may be limited. In addition to these cells, several different types of CD8 based regulatory cell have been identified in T1D and have shown some potential to prevent islet destruction (810). In this study, we focus on a germinal center selective CD8 T cell, which plays an important role in limiting autoantibody production. Because the development of the autoantibody response heralds the future development of T1D, it is vital to determine whether and how CD8 Tregs may prevent the progression of anti-islet autoimmunity. Germinal center–targeting CD8 Tregs have been previously defined by expression of the activation marker CD44 and by expression of the IL-15/IL-2R β-chain CD122 (11). These CD8 Tregs can suppress experimental autoimmune encephalomyelitis (1215), collagen-induced arthritis (16), lupus (17), and prevent skin (18) and islet (19) allograft rejection in nonautoimmune mice. Mechanistically, these CD8 Tregs eliminate CD4 T follicular helper (TFH) cells that drive B cell–mediated immunity (17). Recently, the most potent population of TFH-targeting CD8 Tregs was reported to reside with the Ly-49–positive fraction of these CD44+CD122+ CD8 Tregs (20). These cells regulate the Ab response and quell further B cell–mediated immune activation that would otherwise promote epitope spreading. Therefore, understanding Ly-49+ CD8 Treg function in autoimmune T1D is a significant new opportunity in immune regulation that could be part of a comprehensive strategy to terminate this disease.

In the current study, we examined the role of germinal center–targeting CD8 Tregs in the NOD mouse. We discovered that wild-type NOD mice possess a pool of nonfunctional CD44+CD122+ CD8 Tregs. This functional deficiency may result from our observation that NOD mice possess a profoundly diminished pool of TFH-targeting Ly-49+ CD8 Tregs within their CD44+CD122+ CD8 Treg pool. We trace this deficiency to inadequate IL-15 trans-presentation by macrophages, a cell known to promote the development, maintenance, and activation of these CD8 Tregs (20). We demonstrate that NOD CD8 Treg function can be rescued by an IL-15 superagonist (21, 22), thereby restoring their ability to suppress the Ag-specific Ab response and delay diabetes progression. Overall, these studies further define the phenotype and function of CD8-based regulation of the germinal center reaction and Ab response in T1D and lay the foundation for a CD8 Treg–based cell therapy for its treatment.

C57BL6/J (B6), C57BL/6NTac-IL15tm1ImxN5 (B6.IL-15−/−), C57BL6/J.RAG1−/− (B6.RAG), NOD/ShiLtJ (NOD), and NOD/ShiLtJ.RAG1−/− (NOD.RAG) mice were purchased from The Jackson Laboratory (Bar Harbor, ME) or Taconic Biosciences (Rensselaer, NY). Mice were housed in a specific pathogen-free facility at Vanderbilt University. The Institutional Animal Care and Use Committee at Vanderbilt University approved all procedures.

Splenocytes were stained with the following fluorophore-conjugated Abs: B220, CD11b, CD11c, CD44, F4/80, ICOS, IL-15Ra, Ly-49E/F, and programmed death receptor 1 (PD-1) (eBioscience, San Diego, CA) or CD4, CD8a, CD122, Ly-49C/F/I/H, Ly-49F, and Ki67 (BD Biosciences, San Jose, CA). Samples were collected on a BD LSRFortessa flow cytometer and analyzed by FlowJo software (Tree Star, Ashland, OR). Gates were set on live lymphocytes using forward and side scatter and doublets were excluded.

B6 and NOD mice were injected i.p. with 100 μg of NP33/keyhole limpet hemocyanin (KLH)/CFA (20). Seven days later, splenic CD8+ T cells were purified magnetically (MACS) (Ly-2; Miltenyi, San Diego, CA) and then sorted fluorescently (FACS) to select for CD8 Treg (CD8+CD44+CD122+ or CD8+CD44+CD122+Ly-49E/F+) and non–CD8 Treg populations (CD8+CD44+CD122 or CD8+CD44+CD122+Ly-49E/F) (BD FACsAria III). FACS-sorted CD8 Tregs or non–CD8 Tregs (1 × 106) were i.v. injected into recipient B6.RAG or NOD.RAG mice. Recipients in both arms also received MACS purified splenic B cells (2 × 106) and CD4+CD25 T cells (1 × 106) from naive B6 or NOD donors. Mice were immediately injected i.p. with 100 μg of NP33/KLH/CFA, boosted with 50 μg of NP33/KLH/IFA 10 d later, and the anti-NP IgG response was measured via ELISA against NP8 on day 17 (NP8/BSA; Biosearch Technologies) (20).

Splenocytes were plated in cell culture media (DMEM plus 10% FCs, 1% penicillin/streptomycin, and 0.1% β-ME) along with 10 μg/ml of the TLR3 agonist poly(I:C) (InvivoGen, San Diego, CA). Forty-eight hours later, cells were stained with anti–IL-15Ra (clone: DNT15Ra; eBioscience) to assess IL-15 trans-presentation by APC populations via flow cytometry: macrophages (CD11b+F480+), plasmacytoid dendritic cells (B220+CD11c+), conventional dendritic cells (CD11b+CD11c+), and B cells (B220+). For isotype control staining, a rat-IgG1/PE conjugate was used (eBioscience).

To generate the IL-15/IL-15Ra superagonist complex (IL-15C), carrier-free IL-15 (eBioscience) was incubated with the high-affinity IL-15R α-chain Fc complex (R&D Systems, Minneapolis, MN) for 1 h at 37°C then snap frozen at −80°C until further use as previously described (21). For ex vivo signaling assays, whole splenocytes were exposed ex vivo to increasing concentrations of IL-15C (1, 10, 100, 1000, and 10,000 pM) for various periods of time (0, 5, 10, 15, 30, and 60 min) in cell culture media (23). Cells were then fixed with 1% paraformaldehyde and permeabilized with 100% ice-cold methanol, and p-STAT5 levels were assessed within Ly-49+ CD8 Tregs by staining with a primary anti–p-STAT5 (Y694) rabbit Ab, followed by a secondary anti-rabbit Fab2-Alexa Fluor 647 conjugate (Cell Signaling Technology, Danvers, MA). For in vivo signaling assays, mice were i.v. injected with 1 μg of IL-15C or saline as a control (21). One hour after IL-15C or saline injection, spleens were immediately fixed with 1% paraformaldehyde and permeabilized with methanol, and p-STAT5 signaling was assessed (24).

Mice were i.p. injected with 2 μg of IL-15C every day for 4 d. Splenocytes were counted using an Automated Cell Counter (Bio-Rad Laboratories, Hercules, CA) and relative expansion of Ly-49+ CD8 Tregs analyzed between strains on day 5. To test IL-15C–expanded Ly-49+ CD8 Treg suppressive function, NOD donor mice received a 1 μg i.p. injection of IL-15C every day for 7 d after initial immunization with 100 μg of NP33/KLH/CFA. The 1 × 106 Ly-49+ CD8 Treg or Ly-49 non-CD8 Treg populations were FACS sorted on day 7 and transferred into NOD.RAG mice along with naive 2 × 106 B and 1 × 106 CD4+CD25 T cells. Mice were immunized with 100 μg of NP33/KLH/CFA and boosted with 50 μg of NP33/KLH/IFA on day 10 and the high-affinity anti-NP IgG response was analyzed by ELISA on day 17, as above.

NOD.RAG recipients were divided into six groups and i.v. injected with either 1) saline control; 2) saline control; 3) 5 × 104 naive CD8+CD122 T cells from prediabetic NOD donors; 4) 5 × 104 naive CD8+CD122+ T cells from prediabetic NOD donors; 5) 5 × 104 CD8+CD122+Ly-49 CD8 T cells; or 6) 5 × 104 CD8+CD122+Ly-49+ CD8 T cells. Transferred CD8 T cells for groups 5 and 6 were sorted from prediabetic NOD mice undergoing a 7-d course of 1 μg/d in vivo stable IL-15C. Two days later, these six groups received 5 × 106 CD8-depleted diabetic splenocytes purified via MACS from hyperglycemic NOD mice (25). NOD.RAG mice in groups 2, 5, and 6 also received a single 1-μg injection of in vivo stable IL-15C to maintain CD8 Treg activation on the day of transfer. Glycemia was monitored by twice-weekly glucose checks (Accu-Chek; Roche Diagnostics, Indianapolis, IN). Two consecutive days of blood glucose readings >250 mg/dl confirmed disease onset. Splenic cell populations were assessed by flow cytometry within a week of disease onset.

Statistical analysis was performed with GraphPad Prism version 8 (La Jolla, CA) using the Student t test for comparison of two normally distributed conditions. One- or two-way ANOVA followed by Bonferroni posttest was used to compare multiple groups. The anti-NP response was analyzed via performing a semilogarithmic linear regression analysis, followed by y-intercept and slope curve comparison. Diabetes onset was graphed as Kaplan–Meier curve. Statistical comparisons with p values <0.05 were deemed significant. Experimental replicates and numbers of animals (n) are listed within each figure legend.

In nonautoimmune B6 mice, CD44+CD122+ CD8 Tregs suppress the Ab response and were recently described to be more potent protectors of islet allografts than their CD4 Treg counterparts (19, 20). To determine whether these CD8 Tregs were functional in NOD mice, we used a well-established in vivo CD8 Treg suppression assay developed by Cantor et al. (Fig. 1A) (11, 16, 20). Like B6 mice, NOD mice mount a robust T cell response to the foreign Ag KLH (26), which enabled us to extend this assay to the NOD model to test whether NOD CD8 Tregs could target the high-affinity Ab response as propagated by CD4 TFH, a cell population whose pathogenic role in T1D pathogenesis is just beginning to be elucidated (27, 28). KLH-activated CD8 Tregs (CD8+CD44+CD122+) or non–CD8 Treg (CD8+CD44+CD122) controls were transferred to immunodeficient recipients along with B lymphocytes and CD4+CD25 T cells from matched, Ag-naive B6 or NOD strains (CD4 Tregs [as defined by CD25 positivity] were not transferred to limit suppression to the transferred CD8 T cell populations). These animals were immunized and boosted with the original test stimulus (NP33/KLH), and the relative suppression of the high-affinity anti-NP IgG response was compared between mice receiving CD8 Tregs or non–CD8 Tregs. Whereas B6 CD44+CD122+ CD8 Tregs suppressed the generation of high-affinity anti-NP IgG Abs in recipient RAG mice as previously reported (Fig. 1B), NOD CD8 Tregs failed to suppress this Ag-specific Ab response (Fig. 1C).

FIGURE 1.

NOD mice possess nonfunctional, CD44+CD122+ CD8 Tregs. (A) In vivo CD8 Treg suppression assay. Briefly, naive B6 and NOD mice were injected i.p. with NP33/KLH/CFA. Seven days later, splenic CD8+ T cells were sorted fluorescently (FACS) to select for CD8 Treg (CD8+CD44+CD122+) and non–CD8 Treg populations (CD8+CD44+CD122). FACS-sorted CD8 Tregs or non–CD8 Tregs (1 × 105) were i.v. injected into recipient B6.RAG or NOD.RAG mice alongside matched MACS purified splenic B cells (2 × 106) and CD4+CD25 T cells (1 × 106) from naive B6 or NOD donors. Mice were immediately injected i.p. with 100 μg of NP33/KLH/CFA and boosted with 50 μg of NP33/KLH/IFA 10 d later, and the high-affinity anti-NP IgG response was measured via ELISA against NP8 on day 17. (B) CD122+ CD8 Tregs (black triangles) from KLH-activated B6 mice readily suppress the high-affinity anti-NP IgG response as compared with mice receiving CD122 non-CD8 Tregs (black squares) or no CD8 T cells (black circles) as a control. (C) In comparison, CD122+ CD8 Tregs (white triangles) from KLH-activated NOD mice fail to suppress the high-affinity anti–NP IgG response over mice receiving CD122 non-CD8 Tregs (white squares) or no CD8 T cells (white circles). Data represent one experiment that was repeated twice with similar results (n = 4 mice per B6.RAG recipient; n = 5 mice per NOD.RAG recipient). *p < 0.05, by semilogarithmic linear regression analysis, followed by y-intercept and slope curve comparison or two-way ANOVA followed by Bonferroni posttest. ns, nonsignificant.

FIGURE 1.

NOD mice possess nonfunctional, CD44+CD122+ CD8 Tregs. (A) In vivo CD8 Treg suppression assay. Briefly, naive B6 and NOD mice were injected i.p. with NP33/KLH/CFA. Seven days later, splenic CD8+ T cells were sorted fluorescently (FACS) to select for CD8 Treg (CD8+CD44+CD122+) and non–CD8 Treg populations (CD8+CD44+CD122). FACS-sorted CD8 Tregs or non–CD8 Tregs (1 × 105) were i.v. injected into recipient B6.RAG or NOD.RAG mice alongside matched MACS purified splenic B cells (2 × 106) and CD4+CD25 T cells (1 × 106) from naive B6 or NOD donors. Mice were immediately injected i.p. with 100 μg of NP33/KLH/CFA and boosted with 50 μg of NP33/KLH/IFA 10 d later, and the high-affinity anti-NP IgG response was measured via ELISA against NP8 on day 17. (B) CD122+ CD8 Tregs (black triangles) from KLH-activated B6 mice readily suppress the high-affinity anti-NP IgG response as compared with mice receiving CD122 non-CD8 Tregs (black squares) or no CD8 T cells (black circles) as a control. (C) In comparison, CD122+ CD8 Tregs (white triangles) from KLH-activated NOD mice fail to suppress the high-affinity anti–NP IgG response over mice receiving CD122 non-CD8 Tregs (white squares) or no CD8 T cells (white circles). Data represent one experiment that was repeated twice with similar results (n = 4 mice per B6.RAG recipient; n = 5 mice per NOD.RAG recipient). *p < 0.05, by semilogarithmic linear regression analysis, followed by y-intercept and slope curve comparison or two-way ANOVA followed by Bonferroni posttest. ns, nonsignificant.

Close modal

As the most specific population of TFH-targeting CD8 Tregs has been reported to reside within the Ly-49F fraction of classically defined CD8 Tregs (20), we explored whether NOD mice possess an altered pool of Ly-49F+ CD8 Tregs as compared with CD8 Treg sufficient B6 mice. We determined that NOD mice possess an extremely diminished pool of Ly-49E/F+ CD8 Tregs (Fig. 2A), possessing nearly 11-fold fewer splenic Ly-49+ CD8 Tregs at 8 wk of age as compared with the healthy controls. As allelic differences in the Ly-49 locus exist between B6 and NOD strains (29), we tested numerous anti–Ly-49F–detecting Ab clones in B6 and NOD prior to choosing this specific clone for analysis (Supplemental Fig. 1); specifically, the anti–Ly-49E/F clone CM4 bound unique populations of CD8 T cells and NK cells in both B6 and NOD mice. The mean fluorescence intensity (MFI) of Ly-49 staining in the small population of positive cells was similar or greater in NOD, as compared with B6, suggesting similar reactivity of the CM4 clone to Ly-49 in both strains.

FIGURE 2.

Diabetes-prone NOD mice are profoundly deficient in TFH-targeting Ly-49+ CD8 Tregs. (A) Wild-type, 8-wk-old NOD mice possess 11-fold fewer splenic TFH-targeting Ly-49+ CD8 Tregs (CD8a+CD122+Ly-49E/F+) as compared with nonautoimmune, age-matched B6 mice. (B) Whereas healthy B6 mice maintain a relatively robust and stable population of Ly-49+ CD8 Tregs as they age, NOD mice progressively lose this protective population as they age toward diabetes. Data represent one experiment that was repeated twice with similar results (n = 3 mice per B6 at each age; n = 3 mice per NOD at each age). (C) The observed Ly-49+ CD8 Treg deficiency in NOD mice extends to additional lymphoid compartments, including the bone marrow, inguinal lymph nodes, cervical lymph nodes, and pancreatic lymph nodes. Data represent one experiment from pooled immune compartments of B6 (n = 3) and NOD (n = 3) mice at 8 wk of age. (D) Ly-49+ CD8 Tregs in NOD mice divide at a rate three times lower than their B6 counterparts, as determined by their Ki67 positivity. Data are from two independent pooled experiments, for a total of n = 6 B6 mice and n = 6 NOD mice (8 wk). Although age-matched NOD mice possess similar percentages of Ly-49+, CD8 Tregs target TFH cells versus B6 mice (E). Data pooled from three independent experiments, for a total of n = 12 B6 mice and n = 11 NOD mice (8 wk); TFH cells in NOD mice demonstrate a 2.5-fold higher proliferative rate (F). This discrepancy may account for our observation that CD4 TFH cells outnumber Ly-49+ CD8 Tregs in NOD mice ∼87 to 1 (G). Data are pooled from two independent experiments, for a total of n = 6 B6 mice and n = 6 NOD mice (8 wk). *p < 0.05, **p < 0.01, ****p < 0.0001, by either two-way ANOVA followed by Bonferroni posttest or by Student t test. ns, nonsignificant.

FIGURE 2.

Diabetes-prone NOD mice are profoundly deficient in TFH-targeting Ly-49+ CD8 Tregs. (A) Wild-type, 8-wk-old NOD mice possess 11-fold fewer splenic TFH-targeting Ly-49+ CD8 Tregs (CD8a+CD122+Ly-49E/F+) as compared with nonautoimmune, age-matched B6 mice. (B) Whereas healthy B6 mice maintain a relatively robust and stable population of Ly-49+ CD8 Tregs as they age, NOD mice progressively lose this protective population as they age toward diabetes. Data represent one experiment that was repeated twice with similar results (n = 3 mice per B6 at each age; n = 3 mice per NOD at each age). (C) The observed Ly-49+ CD8 Treg deficiency in NOD mice extends to additional lymphoid compartments, including the bone marrow, inguinal lymph nodes, cervical lymph nodes, and pancreatic lymph nodes. Data represent one experiment from pooled immune compartments of B6 (n = 3) and NOD (n = 3) mice at 8 wk of age. (D) Ly-49+ CD8 Tregs in NOD mice divide at a rate three times lower than their B6 counterparts, as determined by their Ki67 positivity. Data are from two independent pooled experiments, for a total of n = 6 B6 mice and n = 6 NOD mice (8 wk). Although age-matched NOD mice possess similar percentages of Ly-49+, CD8 Tregs target TFH cells versus B6 mice (E). Data pooled from three independent experiments, for a total of n = 12 B6 mice and n = 11 NOD mice (8 wk); TFH cells in NOD mice demonstrate a 2.5-fold higher proliferative rate (F). This discrepancy may account for our observation that CD4 TFH cells outnumber Ly-49+ CD8 Tregs in NOD mice ∼87 to 1 (G). Data are pooled from two independent experiments, for a total of n = 6 B6 mice and n = 6 NOD mice (8 wk). *p < 0.05, **p < 0.01, ****p < 0.0001, by either two-way ANOVA followed by Bonferroni posttest or by Student t test. ns, nonsignificant.

Close modal

Whereas nonautoimmune B6 mice maintain a comparatively robust and stable population of splenic Ly-49+ CD8 Tregs throughout their lifetime, NOD mice progressively lose Ly-49+ CD8 Tregs as they age, as determined by both percentage and total cell number (Fig. 2B). Moreover, this deficiency extends to additional immune tissue compartments in NOD mice, including the bone marrow, inguinal lymph nodes, cervical lymph nodes, and pancreatic lymph nodes (Fig. 2C). This global Ly-49+ CD8 Treg deficiency in NOD mice may, in part, be due to the lower proliferative capacity of these cells as determined by their baseline Ki67 positivity. Nearly 5% of B6 CD8 Tregs are Ki67+, whereas only 1.5% of NOD Ly-49+ CD8 Tregs are positive for this marker of cellular division (Fig. 2D), suggesting a limitation in available stimulatory factors.

As Ly-49+ CD8 Tregs target the action of TFH cells, we explored whether splenic TFH cell populations differed between these two strains. B6 and NOD mice harbored similar percentages of ICOSHIPD-1+ splenic TFH cells at 8 wk of age (Fig. 2E). However, we observed that TFH cells in NOD mice divide at a 2.5-fold greater rate as determined by their Ki-67 positivity (Fig. 2F). Accordingly, regulation of these cells may be compromised as TFH cells outnumber Ly-49+ CD8 Tregs 87:1. In healthy, nonautoimmune B6 mice, the ratio of TFH cells to Ly-49+ CD8 Tregs was 6:1 (Fig. 2G).

The decrease in Ly-49+ CD8 Treg proliferation in NOD mice suggested a lack of endogenous stimulation. We therefore examined whether various populations of APCs failed to provide adequate survival signals for this regulatory cell type. Notably, work by Kim et al. (20) revealed that B6 mice deficient in IL-15, the requisite cytokine for CD8 T memory cells and NK cells, lack a splenic population of CD122+Ly-49+ CD8 Tregs, which we have also observed (Supplemental Fig. 2). These animals do retain a small population of CD122+Ly-49 CD8 T cells, suggesting that they can sustain a modest population of classical CD8 T memory cells despite the absence of IL-15. Thus, we next investigated whether IL-15 trans-presentation by Ly-49+ CD8 Treg supporting cells in NOD mice was dysfunctional.

Unlike traditional γ-chain–dependent cytokines, like IL-2, IL-15 is trans-presented by its high-affinity receptor (IL-15Ra) to neighboring cells (30). Thus, as a readout for IL-15 trans-presentation, we compared the relative expression of surface IL-15Ra on splenic resident APCs at baseline between B6 and NOD strains. Splenic resident macrophages (CD11b+F4/80+) in NOD mice expressed ∼2-fold less IL-15Ra at baseline (Fig. 3A). As CD8 Tregs are activated during episodes of ongoing immunity, we determined whether NOD macrophages increased IL-15 trans-presentation during stimulation. Whereas splenic B6 macrophages demonstrated ∼150% upregulation of surface IL-15Ra after 48 h of immune stimulation with poly(I:C), splenic NOD macrophages upregulated surface IL-15Ra only 75% over baseline levels (Fig. 3B).

FIGURE 3.

Splenic macrophages from NOD mice inadequately trans-present and upregulate IL-15. (A) Baseline IL-15 trans-presentation, as measured by relative surface IL-15R α expression (IL-15Ra), is significantly reduced on splenic resident NOD macrophages (MΦs). (B) Splenic resident macrophages from NOD mice fail to upregulate IL-15Ra expression to the same extent as B6 macrophages when stimulated with 10 μg of the TLR3 agonist poly(I:C). Isotype control staining for IL-15Ra on B6 and NOD MΦs is shown by the dotted line. Percent upregulation of IL-15Ra is calculated by first subtracting the isotype MFI from the individual mouse MFIs for each strain. Data represent one experiment, independently repeated three times with similar results. Data shown include n = 5 B6 mice and n = 5 NOD mice. *p < 0.05, **p < 0.01, by two-way ANOVA followed by Bonferroni posttest or Student t test.

FIGURE 3.

Splenic macrophages from NOD mice inadequately trans-present and upregulate IL-15. (A) Baseline IL-15 trans-presentation, as measured by relative surface IL-15R α expression (IL-15Ra), is significantly reduced on splenic resident NOD macrophages (MΦs). (B) Splenic resident macrophages from NOD mice fail to upregulate IL-15Ra expression to the same extent as B6 macrophages when stimulated with 10 μg of the TLR3 agonist poly(I:C). Isotype control staining for IL-15Ra on B6 and NOD MΦs is shown by the dotted line. Percent upregulation of IL-15Ra is calculated by first subtracting the isotype MFI from the individual mouse MFIs for each strain. Data represent one experiment, independently repeated three times with similar results. Data shown include n = 5 B6 mice and n = 5 NOD mice. *p < 0.05, **p < 0.01, by two-way ANOVA followed by Bonferroni posttest or Student t test.

Close modal

In addition to the reduced IL-15 availability in the NOD system, Ly-49+ CD8 Tregs in this T1D-prone setting could fail to thrive because of inadequate IL-15 signal transduction, which proceeds via the JAK3/STAT5 system (31). To evaluate IL-15 signaling dynamics in splenic Ly-49+ CD8 Tregs, we determined relative STAT5 phosphorylation in B6 and NOD Ly-49+ CD8 Tregs in response to IL-15/IL-15Ra superagonist (IL-15C) exposure (as described in the 2Materials and Methods section) (21). Importantly, NOD Ly-49+ CD8 Tregs phosphorylated STAT5 to the same extent as B6 Ly-49+ CD8 Tregs when exposed to increasing IL-15C concentrations ex vivo for 30 min (Fig. 4A). To further define IL-15–mediated signaling dynamics between B6 and NOD Ly-49+ CD8 Tregs, we explored STAT5 phosphorylation kinetics over time by exposing B6 and NOD splenocytes to 100-, 1000-, and 10,000-pM concentrations of IL-15C. IL-15C–mediated STAT5 phosphorylation in Ly-49+ CD8 Tregs did not substantially differ between strains across these concentrations (Fig. 4B). Additionally, to determine whether Ly-49+ CD8 Tregs signal differently in the whole-animal setting, we injected B6 and NOD mice with the maximal dose of IL-15C, which was calculated to achieve a 10,000-pM concentration at a whole-animal blood volume of 2 ml. Strikingly, both B6 and NOD Ly-49+ CD8 Tregs robustly upregulated STAT5 phosphorylation nearly 15-fold 60 min after injection (Fig. 4C). Despite similar overall p-STAT5 MFI levels between IL-15C–stimulated B6 and NOD Ly-49+ CD8 Tregs, we did observe a slight bimodality in the p-STAT5 response in stimulated NOD Ly-49+ CD8 Tregs. Thus, unlike B6 Ly-49+ CD8 Tregs, there may exist a subpopulation of NOD Ly-49+ CD8 Tregs that does not respond as robustly to extrinsic IL-15C stimulation, although nearly all cells increased their signaling over baseline.

FIGURE 4.

NOD Ly-49+ CD8 Tregs adequately transduce IL-15–mediated survival signals. (A) Within whole-plated splenocytes, B6 and NOD Ly-49+ CD8 Tregs phosphorylate STAT5 at Y694 to the same extent when exposed ex vivo to increasing concentrations of the IL-15C superagonist for 30 min. To account for the relatively small cell population analyzed, a minimum of 150 Ly-49+ CD8 Treg events were captured, and, from that, the MFI of p-STAT5Y694 was calculated. Data represent one experiment, repeated twice with similar results, that includes n = 3 B6 mice and n = 3 NOD mice. (B) B6 and NOD Ly-49+ CD8 Tregs phosphorylate STAT5 with similar time kinetics when exposed ex vivo to either 100, 1000, and 10000 pM of IL-15C. Data represent one experiment, repeated twice with similar results, that includes n = 3 B6 mice and n = 3 NOD mice. Of note, NOD CD8 Tregs demonstrate statistically lower p-STAT5 levels 30 and 60 min after stimulation with the maximal 10,000-pM concentration of IL-15C (right panel). (C) When exposed i.v. to 1 μg of IL-15C for 60 min (calculated to reach 10,000 pM in a 2-ml blood volume), both B6 and NOD Ly-49+ CD8 Tregs increase p-STAT5 levels 15 times over animals injected with saline as a control. Data represent one experiment that includes n = 3 B6 mice and n = 3 NOD mice. *p < 0.05, **p < 0.01, ***p < 0.005, by two-way ANOVA followed by Bonferroni posttest. ns, nonsignificant.

FIGURE 4.

NOD Ly-49+ CD8 Tregs adequately transduce IL-15–mediated survival signals. (A) Within whole-plated splenocytes, B6 and NOD Ly-49+ CD8 Tregs phosphorylate STAT5 at Y694 to the same extent when exposed ex vivo to increasing concentrations of the IL-15C superagonist for 30 min. To account for the relatively small cell population analyzed, a minimum of 150 Ly-49+ CD8 Treg events were captured, and, from that, the MFI of p-STAT5Y694 was calculated. Data represent one experiment, repeated twice with similar results, that includes n = 3 B6 mice and n = 3 NOD mice. (B) B6 and NOD Ly-49+ CD8 Tregs phosphorylate STAT5 with similar time kinetics when exposed ex vivo to either 100, 1000, and 10000 pM of IL-15C. Data represent one experiment, repeated twice with similar results, that includes n = 3 B6 mice and n = 3 NOD mice. Of note, NOD CD8 Tregs demonstrate statistically lower p-STAT5 levels 30 and 60 min after stimulation with the maximal 10,000-pM concentration of IL-15C (right panel). (C) When exposed i.v. to 1 μg of IL-15C for 60 min (calculated to reach 10,000 pM in a 2-ml blood volume), both B6 and NOD Ly-49+ CD8 Tregs increase p-STAT5 levels 15 times over animals injected with saline as a control. Data represent one experiment that includes n = 3 B6 mice and n = 3 NOD mice. *p < 0.05, **p < 0.01, ***p < 0.005, by two-way ANOVA followed by Bonferroni posttest. ns, nonsignificant.

Close modal

NOD Ly-49+ CD8 Tregs sufficiently transduce IL-15–mediated survival signals via STAT5 phosphorylation. Therefore, we explored whether systemically administered IL-15C would expand NOD Ly-49+ CD8 Tregs and, moreover, whether these expanded cells would be functionally activated. In vivo administration of IL-15C robustly expanded Ly-49+ CD8 Tregs, increasing total Ly-49+ CD8 Treg numbers 15-fold in B6 mice and 612-fold in NOD mice (Fig. 5A). The observed massive NOD CD8 Treg expansion was also associated with splenomegaly in NOD mice after IL-15C treatment; total B6 splenocytes increased 1.7-fold in absolute number after treatment (95.4 × 106 ± 32.2 versus 165.8 × 106 ± 26.9), whereas NOD splenocytes increased 3.8-fold (80.4 × 106 ± 16 versus 307.1 × 106 ± 70.9) (n = 6 for each strain, treated and untreated).

FIGURE 5.

In vivo administration of IL-15C robustly expands NOD Ly-49+ CD8 Tregs and partially rescues their suppressive function. (A) When injected with 2 μg of IL-15C for 4 d consecutively, B6 and NOD Ly-49+ CD8 Tregs expand robustly. Data are from two independent pooled experiments, for a total of n = 6 B6 mice and n = 6 NOD mice (8 wk). (B) IL-15C–activated Ly-49+ CD8 Tregs (white triangles) from NOD mice immunized with KLH partially suppress the high-affinity anti–NP IgG response as compared with mice receiving Ly-49 non-CD8 Tregs from the same donor mouse (white squares) or mice receiving no CD8 T cells as a control (white circles). Data represent one independent experiment, repeated two times with similar results, with a total of n = 3 NOD mice in each recipient group (8 wk). (C) In comparison, nonactivated Ly-49+ CD8 Tregs from CD8 Treg–sufficient B6 mice that did not receive concomitant IL-15C activation (black triangles) robustly suppress the high-affinity anti–NP IgG response as compared with mice receiving Ly-49 non-CD8 Tregs from the same donor mouse (black squares). Data represent one independent experiment, repeated two times with similar results, with a total of n = 4 B6 mice in each recipient group (8 wk). ****p < 0.0001, by semilogarithmic linear regression analysis followed by y-intercept and slope curve comparison or two-way ANOVA followed by Bonferroni posttest. ns, nonsignificant.

FIGURE 5.

In vivo administration of IL-15C robustly expands NOD Ly-49+ CD8 Tregs and partially rescues their suppressive function. (A) When injected with 2 μg of IL-15C for 4 d consecutively, B6 and NOD Ly-49+ CD8 Tregs expand robustly. Data are from two independent pooled experiments, for a total of n = 6 B6 mice and n = 6 NOD mice (8 wk). (B) IL-15C–activated Ly-49+ CD8 Tregs (white triangles) from NOD mice immunized with KLH partially suppress the high-affinity anti–NP IgG response as compared with mice receiving Ly-49 non-CD8 Tregs from the same donor mouse (white squares) or mice receiving no CD8 T cells as a control (white circles). Data represent one independent experiment, repeated two times with similar results, with a total of n = 3 NOD mice in each recipient group (8 wk). (C) In comparison, nonactivated Ly-49+ CD8 Tregs from CD8 Treg–sufficient B6 mice that did not receive concomitant IL-15C activation (black triangles) robustly suppress the high-affinity anti–NP IgG response as compared with mice receiving Ly-49 non-CD8 Tregs from the same donor mouse (black squares). Data represent one independent experiment, repeated two times with similar results, with a total of n = 4 B6 mice in each recipient group (8 wk). ****p < 0.0001, by semilogarithmic linear regression analysis followed by y-intercept and slope curve comparison or two-way ANOVA followed by Bonferroni posttest. ns, nonsignificant.

Close modal

To determine whether these IL-15C–expanded NOD Ly-49+ CD8 Tregs were functionally rescued, we used the CD8 Treg suppression assay described above (Fig. 1A). Donor NOD mice were immunized with NP33/KLH and CD8 Tregs were expanded via concomitant IL-15C administration (1 μg/d for 7 d). Recipient immunodeficient NOD.RAG mice then received either IL-15C–expanded Ly-49+ CD8 Tregs, IL-15C–expanded Ly-49 CD8 T cells, or no CD8 T cells as a control. These mice then received naive B cells and CD4+CD25 T cells, as well as the NP/KLH test stimulus. Whereas IL-15C–activated Ly-49 CD8 T cells did not suppress the Ag-specific, high-affinity anti-NP IgG Ab response over mice receiving no CD8 T cells, IL-15C–activated Ly-49+ CD8 Tregs suppressed the Ab response (Fig. 5B). However, equivalent numbers of transferred IL-15C–activated Ly-49+ CD8 Tregs in the NOD setting did not suppress the Ab response as completely as non–IL-15C–activated Ly-49+ CD8 Tregs in the B6 setting (Fig. 5C), perhaps because of limited IL-15 availability for trans-presentation in recipient NOD.RAG mice as compared with B6.RAG mice.

Finally, we assessed whether IL-15C–activated CD8 Tregs alter diabetes progression and whether this is enhanced in the Ab response–regulating Ly-49 compartment. Immunodeficient NOD recipients were injected with either 5 × 104 CD122 CD8 T cells, 5 × 104 CD122+ CD8 Tregs, 5 × 104 IL-15C–expanded CD122+ Ly-49+ CD8 Tregs, or 5 × 104 IL-15C–expanded CD122+ Ly-49 CD8 T cells from prediabetic NOD mice. Two additional groups also received saline injection controls. Two days later, all mice received 5 × 106 CD8-depleted diabetic splenocytes purified from hyperglycemic NOD mice. Mice receiving IL-15C–expanded CD122+ Ly-49+ CD8 Tregs, IL-15C–expanded CD122+ Ly-49 CD8 T cells, and one of the saline-injected control groups also received a single 1-μg injection of IL-15C on the day of cell transfer to maintain CD8 Treg activation and to assess the role of this superkine in the absence of a transferred cell population. Whereas immunodeficient mice receiving nonactivated CD122 CD8 T cells or classically defined CD122+ CD8 Tregs developed diabetes by 4 wk, on average (median survival time, 3.9 and 4.1 wk, respectively), progression to overt hyperglycemia took on average 10.4 wk in mice receiving as few as 5 × 104 IL-15C–activated CD122+ CD8 Tregs, regardless of their expression of Ly-49 (Fig. 6A). Mice receiving no CD8 T cells, whether in the absence or the presence of IL-15C, progressed to diabetes at a pace falling between these extremes (median survival time, 5.1 and 7.2 wk, respectively), indicating enhanced suppression by the transfer of activated CD8 Tregs. Mice receiving IL-15C–activated Ly-49+ CD8 Tregs possessed the greatest percentage of splenic Ly-49+ CD8 Tregs (CD122+Ly-49+; Fig. 6B), suggesting these cells maintain their Ly-49+ phenotype well after adoptive transfer. IL-15C–activated CD8 Tregs, both Ly-49 and Ly-49+, also demonstrated the smallest percentage of splenic CD4 TFH cells after recovery (ICOSHIPD-1+; Fig. 6C).

FIGURE 6.

IL-15C–activated Ly-49+ and Ly-49 CD122+ CD8 Tregs delay diabetes transfer. (A) A single injection of 5 × 104 Ly-49+ (black diamonds) or Ly-49 (white diamonds) CD122+ CD8 Tregs FACS-purified from prediabetic NOD mice receiving 1 μg of in vivo stable IL-15C for 7-d delay diabetes onset 2.6-fold longer than either 5 × 104 unactivated CD122+ CD8 T cells (black squares) or 5 × 104 CD122 CD8 T cells (white squares) from naive prediabetic NOD mice. NOD.RAG mice in all arms received 5 × 106 CD8-depleted splenocytes from hyperglycemic NOD mice 2 d prior to CD8 T cell infusion. On the day of CD8 T cell transfer, NOD.RAG mice receiving IL-15C–activated CD8 Tregs also received a 1-μg injection of in vivo stable IL-15C to maintain CD8 Treg activation. One week after diabetes onset, mice receiving IL-15C–activated Ly-49+ CD8 Tregs possessed the largest population of splenic Ly-49+ CD8 Tregs (B). Mice receiving either IL-15C–activated Ly-49+ or Ly-49 CD8 Tregs possessed the smallest populations of target CD4 TFH cells (C). (A) The N for each group is indicated below the survival curve. Groups were compared with naive CD122+ cell transfer by log-rank analysis: CD8-depleted control (p = 0.03), IL-15 alone (p < 0.001), naive CD122 (p = NS), Ly-49 (p < 0.0001), and Ly-49+ (p < 0.0001). Additional comparisons include Ly-49+ or Ly-49 versus CD8-depleted control (p = 0.01) and Ly-49+ or Ly-49 versus IL-15C alone (p < 0.01). (B and C) Naive CD122 (n = 5), naive CD122+ (n = 6), IL-15C–activated Ly-49 CD8 Tregs (n = 7), and IL-15C–activated Ly-49+CD8 Tregs (n = 6). Numbers of animals in (B) and (C) do not match numbers in (A), as some recipients died prior to analysis. **p < 0.01, ****p < 0.0001, significance determined by one-way ANOVA followed by Bonferroni posttest or by t test.

FIGURE 6.

IL-15C–activated Ly-49+ and Ly-49 CD122+ CD8 Tregs delay diabetes transfer. (A) A single injection of 5 × 104 Ly-49+ (black diamonds) or Ly-49 (white diamonds) CD122+ CD8 Tregs FACS-purified from prediabetic NOD mice receiving 1 μg of in vivo stable IL-15C for 7-d delay diabetes onset 2.6-fold longer than either 5 × 104 unactivated CD122+ CD8 T cells (black squares) or 5 × 104 CD122 CD8 T cells (white squares) from naive prediabetic NOD mice. NOD.RAG mice in all arms received 5 × 106 CD8-depleted splenocytes from hyperglycemic NOD mice 2 d prior to CD8 T cell infusion. On the day of CD8 T cell transfer, NOD.RAG mice receiving IL-15C–activated CD8 Tregs also received a 1-μg injection of in vivo stable IL-15C to maintain CD8 Treg activation. One week after diabetes onset, mice receiving IL-15C–activated Ly-49+ CD8 Tregs possessed the largest population of splenic Ly-49+ CD8 Tregs (B). Mice receiving either IL-15C–activated Ly-49+ or Ly-49 CD8 Tregs possessed the smallest populations of target CD4 TFH cells (C). (A) The N for each group is indicated below the survival curve. Groups were compared with naive CD122+ cell transfer by log-rank analysis: CD8-depleted control (p = 0.03), IL-15 alone (p < 0.001), naive CD122 (p = NS), Ly-49 (p < 0.0001), and Ly-49+ (p < 0.0001). Additional comparisons include Ly-49+ or Ly-49 versus CD8-depleted control (p = 0.01) and Ly-49+ or Ly-49 versus IL-15C alone (p < 0.01). (B and C) Naive CD122 (n = 5), naive CD122+ (n = 6), IL-15C–activated Ly-49 CD8 Tregs (n = 7), and IL-15C–activated Ly-49+CD8 Tregs (n = 6). Numbers of animals in (B) and (C) do not match numbers in (A), as some recipients died prior to analysis. **p < 0.01, ****p < 0.0001, significance determined by one-way ANOVA followed by Bonferroni posttest or by t test.

Close modal

In addition to the important regulatory functions ascribed to CD4 T cells and some B lymphocytes, the capacity for immune regulation by CD8 T cells continues to be revealed. Several such regulatory populations have been identified in studies of T1D, including roles for CD28 low (CD8+CD28) cells and foxp3-expressing CD8 T cells (810). The potential for CD8 cells to regulate the Ab response was first identified in CD8 T cell deficient animals, which had unexpectedly exaggerated Ab responses to immunization (32). Regulation of the developing autoantibody response is now thought to be a critical target in T1D pathogenesis, as the development of multiple autoantibodies defines a diagnosis of stage 1 T1D (33). A specific subpopulation of CD8 Tregs are emerging as a population of regulatory cells that check the germinal center response, prevent dangerous epitope spreading, and halt autoimmunity, making them potentially important players in diabetogenic autoimmunity (11). In this study, we have determined that T1D-prone NOD mice lack a functional population of CD44+CD122+ CD8 Tregs. Recently, the most potent population of TFH-targeting CD8 Tregs was identified within the Ly-49F–positive fraction of these classically defined CD44+CD122+ CD8 Tregs. Interestingly, the Ly-49F isoform of the NK cell family of Ly-49 inhibitory receptors is believed to interact with Qa-1 (20), which is required for CD8 Treg development and is linked to the diabetes risk locus Idd24, but may be dispensable for islet tolerance (34). This locus has further been connected to the prolonged immune response in NOD mice (35), which is an expected biological consequence of defective CD8 Treg function.

To define a cellular mechanism for CD8 Treg functional insufficiency, we determined that NOD mice are profoundly deficient in TFH-targeting Ly-49F+ CD8 Tregs. Genetic analysis has revealed that NOD mice possess the largest Ly-49 haplotype of any known mouse strain (29). The Ly-49 locus in NOD mice contains an overabundance of activating receptors whose function has been linked to diabetes progression. This locus resides in the diabetes susceptibility Idd6 region on chromosome 6 in NOD mice. NOD mice congenic for the B6 chromosomal region D6 Mit 254 to D6 Mit 14 (NOD.NK1.1 mice) have reduced diabetes incidence (36). Although the authors suggest improved NK/NKT cell performance as a mechanism of disease protection, introduction of the Ly-49 locus from B6 mice could also restore Ly-49+ CD8 Treg function in this NOD congenic strain, although this possibility has not been studied. Thus, although NOD mice possess an extremely polymorphic Ly-49 locus, our ability to rescue the Ab-suppressive function of these Ly-49+ CD8 Tregs suggests that Ly-49 does identify retention of functional CD8 T cells in NOD with the potential to regulate islet-Ab production.

In addition to the use of Ly-49 as a functional marker for CD8 Tregs, a recent report highlighted the potential role of the PD-1 in CD8 Treg–mediated suppression of the allograft response. These PD-1+CD44+CD122+ CD8 Tregs from B6 mice delayed rejection of BALB/C skin allografts via an IL-10–dependent mechanism (18). As TFH cells also express components of the PD-1/PD-1L cellular exhaustion pathway (17), CD8 Treg expression of PD-1 may allow direct TFH cell targeting. In our analysis of Ly-49+ CD8 Tregs in NOD mice, we detected no expression of PD-1 on CD44+CD122+ CD8 T cells (data not shown). This finding is corroborated by a report that wild-type NOD mice lack PD-1+CD122+ CD8 Tregs, which, in turn, permitted enhanced islet effector function by their PD-1-CD122+ CD8 T cell counterparts (37). Alternatively, the Cantor group recently published data demonstrating that the transcription factor Helios of the Ikaros family represents that master transcription factor of TFH-regulating CD8 Tregs (38). Preliminary work by members of our laboratory have since found that NOD mice also possess a diminished population of Helios-expressing CD122+ CD8 Tregs (data not shown). As our results demonstrate functional rescue of Ly-49+ CD8 Tregs by the IL-15C superagonist, future studies could investigate whether treatment with IL-15C restores PD-1 expression on, IL-10 secretion by, and/or Helios expression within CD122+ CD8 T cells in wild-type NOD mice.

As IL-15–deficient B6 mice lack Ly-49+ CD8 Tregs, we explored whether IL-15 insufficiency in the NOD system contributed to the deficiency of these cells. We determined that NOD macrophages inadequately trans-present the CD8 Treg–requisite cytokine IL-15. In 2010, Suwanai et al. (39) reported that NOD mice possess a defective IL-15 allele, which underlies this strain’s NK cell functional deficiency. Exogenous administration of low doses of IL-15C to the diabetes-protected BDC2.5/NOD mouse preferentially expanded NK cells, which broke islet cell tolerance and rapidly precipitated diabetes. The authors reported no expansion of CD44+CD122+ CD8 T cells, suggesting that, in contrast to the high doses of IL-15C used to expand CD8 Tregs in our study, low doses of IL-15C may favor the expansion of diabetes-promoting cells rather than their disease-protective regulatory cell counterparts. In fact, NOD mice demonstrate reduced disease incidence when they genetically lack IL-15 (40) or are treated with an anti–IL-15Rb–blocking Ab (41). Thus, although disruption of the IL-15 axis in these systems could interrupt residual Ly-49+CD8 Treg function, we hypothesize that the already profoundly diminished pool of IL-15 Ly-49+ CD8 Tregs would limit any addition deleterious effect from loss of IL-15. It was further reported, however, that in the absence of IL-15–dependent NK cells, administration of IL-15 to NOD mice prevented disease (42). Moreover, in the nonautoimmune B6 setting, coadministration of IL-15 with naturally occurring naive CD122+ CD8 Tregs prolonged foreign islet allograft survival (19). Thus, we hypothesize that activating CD8 Tregs with IL-15C specifically, and not their pathogenic IL-15–dependent cellular counterparts, affords disease protection in the NOD setting as we observed in our transfer study.

Although we determined that IL-15C activation rescued the Ab-suppressive function of Ly-49+ CD8 Tregs (see Fig. 5B), we found that the transfer of either IL-15C–activated Ly-49 or Ly-49+ CD8 T cells delayed diabetes progression (see Fig. 6A). Diabetes prevention may not solely rely on the Ab-suppressing action of Ly-49+ CD8 T cells. Ab production is an early aspect of T1D pathogenesis with T cells escaping their dependence on B lymphocytes by disease onset. Both IL-15C–activated Ly-49 or Ly-49+ CD8 T cells diminished the TFH cell response and diabetes progression; thus, the significant activation provided by the IL-15 superagonist may also stimulate regulation in the Ly-49 compartment, although these cells were not as effective in suppressing Ab production as the Ly-49+ cells. Ly-49+ CD8 T cells may be more relevant in prevention/early pathogenesis, leading to Ab maturation or in the absence of an activating stimulus, whereas IL-15C activation of CD122+ (either Ly-49+ or Ly-49) CD8 Tregs may be broadly valuable for halting progression of later disease.

Finally, patients with T1D have been reported to possess nonfunctional peripheral blood resident CD8 Tregs (43) as well as a unique TFH cell phenotype (28) that may result from this dysfunction. Specifically, CD8 Tregs from patients with T1D failed to eliminate GAD-reactive CD4 T cells via TCR-restricted interactions with the nonclassical MHC class Ib molecule HLA-E expressed by target cells. In related clinical studies, patients with recent onset T1D responding positively to anti-CD3 therapy (teplizumab) demonstrated an expanded pool of circulating central memory like CD8 T cells (3). In fact, CD8 T cells isolated from the peripheral blood of patients treated with teplizumab have restored regulatory function as compared with CD8+ T cells isolated from patients treated with a control IgG (9). These reprogrammed CD8 T cells upregulated the expression of the regulatory cell identifier GITR (44, 45), a marker that we also determined to be upregulated on CD8 Tregs from B6 and NOD mice (data not shown). Thus, anti-CD3 may reprogram CD8 T cells from an effector to an islet-protective regulatory phenotype. We have observed that NOD mice treated with a single 50-μg injection of anti-CD3 demonstrate a 10-fold expansion of Ly-49+ CD8 Tregs 7 d later (data not shown).

In conclusion, NOD mice lack a functional population of Ab-suppressive CD44+CD122+ CD8 Tregs, a dysfunction that may result from a severe deficiency of the Ly-49+ CD8 Tregs that regulate TFH cells in healthy animals. Despite reduced IL-15 availability in the NOD system, these Ly-49+ CD8 Tregs respond adequately to IL-15 and can be restored numerically and functionally with a novel IL-15C superagonist to prevent the high-affinity Ab response. IL-15C–activated CD8 T cells, both Ly-49+ or Ly-49, prevented diabetes transfer and reduced target CD4 TFH cell number. Overall, IL-15C may activate CD8 Tregs in patients with T1D, thereby offering new approaches for using CD8 Tregs as biomarkers for disease progression or as a novel cell-based therapy.

We thank the staff of the Vanderbilt Flow Cytometry Shared Resource for assistance with the experiments in this manuscript.

This work was supported by National Institutes of Health Grants R03-DK097410 (to D.J.M.), R21AI119224 (to D.J.M.), T32-GM007347 (through which the Vanderbilt Medical Scientist Training Program provided support for B.T.S.), and F31DK107321 (to C.S.W.) as well as a JDRF Career Development Award (to D.J.M.) and institutional funds provided by the Vanderbilt Department of Pediatrics (to D.J.M.). The Vanderbilt University Medical Center Flow Cytometry Shared Resource is supported by the Vanderbilt Ingram Cancer Center (P30-CA68485) and by the Vanderbilt Digestive Disease Research Center (DK058404).

The online version of this article contains supplemental material.

Abbreviations used in this article:

B6

C57BL6/J; B6.RAG, C57BL6/J.RAG1−/−; IL-15/IL-15Ra superagonist complex

IL-15C

IL-15/IL-15Ra superagonist complex

KLH

keyhole limpet hemocyanin

MFI

mean fluorescence intensity

NOD.RAG

NOD/ShiLtJ.RAG1−/−

PD-1

programmed death receptor 1

T1D

type 1 diabetes

TFH

T follicular helper

Treg

T regulatory cell.

1
Ziegler
,
A. G.
,
M.
Rewers
,
O.
Simell
,
T.
Simell
,
J.
Lempainen
,
A.
Steck
,
C.
Winkler
,
J.
Ilonen
,
R.
Veijola
,
M.
Knip
, et al
.
2013
.
Seroconversion to multiple islet autoantibodies and risk of progression to diabetes in children.
JAMA
309
:
2473
2479
.
2
Orban
,
T.
,
B.
Bundy
,
D. J.
Becker
,
L. A.
DiMeglio
,
S. E.
Gitelman
,
R.
Goland
,
P. A.
Gottlieb
,
C. J.
Greenbaum
,
J. B.
Marks
,
R.
Monzavi
, et al
Type 1 Diabetes TrialNet Abatacept Study Group
.
2011
.
Co-stimulation modulation with abatacept in patients with recent-onset type 1 diabetes: a randomised, double-blind, placebo-controlled trial.
Lancet
378
:
412
419
.
3
Herold
,
K. C.
,
S. E.
Gitelman
,
M. R.
Ehlers
,
P. A.
Gottlieb
,
C. J.
Greenbaum
,
W.
Hagopian
,
K. D.
Boyle
,
L.
Keyes-Elstein
,
S.
Aggarwal
,
D.
Phippard
, et al
AbATE Study Team
.
2013
.
Teplizumab (anti-CD3 mAb) treatment preserves C-peptide responses in patients with new-onset type 1 diabetes in a randomized controlled trial: metabolic and immunologic features at baseline identify a subgroup of responders.
Diabetes
62
:
3766
3774
.
4
Pescovitz
,
M. D.
,
C. J.
Greenbaum
,
H.
Krause-Steinrauf
,
D. J.
Becker
,
S. E.
Gitelman
,
R.
Goland
,
P. A.
Gottlieb
,
J. B.
Marks
,
P. F.
McGee
,
A. M.
Moran
, et al
Type 1 Diabetes TrialNet Anti-CD20 Study Group
.
2009
.
Rituximab, B-lymphocyte depletion, and preservation of beta-cell function.
N. Engl. J. Med.
361
:
2143
2152
.
5
Lim
,
H. W.
,
P.
Hillsamer
,
A. H.
Banham
,
C. H.
Kim
.
2005
.
Cutting edge: direct suppression of B cells by CD4+ CD25+ regulatory T cells.
J. Immunol.
175
:
4180
4183
.
6
Okamura
,
T.
,
S.
Sumitomo
,
K.
Morita
,
Y.
Iwasaki
,
M.
Inoue
,
S.
Nakachi
,
T.
Komai
,
H.
Shoda
,
J.
Miyazaki
,
K.
Fujio
,
K.
Yamamoto
.
2015
.
TGF-β3-expressing CD4+CD25(-)LAG3+ regulatory T cells control humoral immune responses.
Nat. Commun.
6
:
6329
.
7
Linterman
,
M. A.
,
W.
Pierson
,
S. K.
Lee
,
A.
Kallies
,
S.
Kawamoto
,
T. F.
Rayner
,
M.
Srivastava
,
D. P.
Divekar
,
L.
Beaton
,
J. J.
Hogan
, et al
.
2011
.
Foxp3+ follicular regulatory T cells control the germinal center response.
Nat. Med.
17
:
975
982
.
8
Yarde
,
D. N.
,
K.
Lorenzo-Arteaga
,
K. P.
Corley
,
M.
Cabrera
,
N. E.
Sarvetnick
.
2014
.
CD28 CD8+ T cells are significantly reduced and correlate with disease duration in juveniles with type 1 diabetes.
Hum. Immunol.
75
:
1069
1074
.
9
Ablamunits
,
V.
,
B.
Bisikirska
,
K. C.
Herold
.
2010
.
Acquisition of regulatory function by human CD8(+) T cells treated with anti-CD3 antibody requires TNF.
Eur. J. Immunol.
40
:
2891
2901
.
10
Wang
,
R.
,
G.
Han
,
L.
Song
,
J.
Wang
,
G.
Chen
,
R.
Xu
,
M.
Yu
,
J.
Qian
,
B.
Shen
,
Y.
Li
.
2009
.
CD8+ regulatory T cells are responsible for GAD-IgG gene-transferred tolerance induction in NOD mice.
Immunology
126
:
123
131
.
11
Kim
,
H.-J.
,
H.
Cantor
.
2011
.
Regulation of self-tolerance by Qa-1-restricted CD8(+) regulatory T cells.
Semin. Immunol.
23
:
446
452
.
12
Hu
,
D.
,
K.
Ikizawa
,
L.
Lu
,
M. E.
Sanchirico
,
M. L.
Shinohara
,
H.
Cantor
.
2004
.
Analysis of regulatory CD8 T cells in Qa-1-deficient mice.
Nat. Immunol.
5
:
516
523
.
13
Leavenworth
,
J. W.
,
C.
Schellack
,
H.-J.
Kim
,
L.
Lu
,
P.
Spee
,
H.
Cantor
.
2010
.
Analysis of the cellular mechanism underlying inhibition of EAE after treatment with anti-NKG2A F(ab’)2.
Proc. Natl. Acad. Sci. USA
107
:
2562
2567
.
14
Lu
,
L.
,
H.-J.
Kim
,
M. B. F.
Werneck
,
H.
Cantor
.
2008
.
Regulation of CD8+ regulatory T cells: interruption of the NKG2A-Qa-1 interaction allows robust suppressive activity and resolution of autoimmune disease.
Proc. Natl. Acad. Sci. USA
105
:
19420
19425
.
15
Yu
,
P.
,
R. N.
Bamford
,
T. A.
Waldmann
.
2014
.
IL-15-dependent CD8+ CD122+ T cells ameliorate experimental autoimmune encephalomyelitis by modulating IL-17 production by CD4+ T cells.
Eur. J. Immunol.
44
:
3330
3341
.
16
Leavenworth
,
J. W.
,
X.
Tang
,
H.-J.
Kim
,
X.
Wang
,
H.
Cantor
.
2013
.
Amelioration of arthritis through mobilization of peptide-specific CD8+ regulatory T cells.
J. Clin. Invest.
123
:
1382
1389
.
17
Kim
,
H.-J.
,
B.
Verbinnen
,
X.
Tang
,
L.
Lu
,
H.
Cantor
.
2010
.
Inhibition of follicular T-helper cells by CD8(+) regulatory T cells is essential for self tolerance.
Nature
467
:
328
332
.
18
Dai
,
H.
,
N.
Wan
,
S.
Zhang
,
Y.
Moore
,
F.
Wan
,
Z.
Dai
.
2010
.
Cutting edge: programmed death-1 defines CD8+CD122+ T cells as regulatory versus memory T cells.
J. Immunol.
185
:
803
807
.
19
Dai
,
Z.
,
S.
Zhang
,
Q.
Xie
,
S.
Wu
,
J.
Su
,
S.
Li
,
Y.
Xu
,
X. C.
Li
.
2014
.
Natural CD8+CD122+ T cells are more potent in suppression of allograft rejection than CD4+CD25+ regulatory T cells.
Am. J. Transplant.
14
:
39
48
.
20
Kim
,
H.-J.
,
X.
Wang
,
S.
Radfar
,
T. J.
Sproule
,
D. C.
Roopenian
,
H.
Cantor
.
2011
.
CD8+ T regulatory cells express the Ly49 Class I MHC receptor and are defective in autoimmune prone B6-Yaa mice.
Proc. Natl. Acad. Sci. USA
108
:
2010
2015
.
21
Rubinstein
,
M. P.
,
M.
Kovar
,
J. F.
Purton
,
J.-H.
Cho
,
O.
Boyman
,
C. D.
Surh
,
J.
Sprent
.
2006
.
Converting IL-15 to a superagonist by binding to soluble IL-15Ralpha.
Proc. Natl. Acad. Sci. USA
103
:
9166
9171
.
22
Stoklasek
,
T. A.
,
K. S.
Schluns
,
L.
Lefrançois
.
2006
.
Combined IL-15/IL-15Ralpha immunotherapy maximizes IL-15 activity in vivo.
J. Immunol.
177
:
6072
6080
.
23
Ring
,
A. M.
,
J.-X.
Lin
,
D.
Feng
,
S.
Mitra
,
M.
Rickert
,
G. R.
Bowman
,
V. S.
Pande
,
P.
Li
,
I.
Moraga
,
R.
Spolski
, et al
.
2012
.
Mechanistic and structural insight into the functional dichotomy between IL-2 and IL-15.
Nat. Immunol.
13
:
1187
1195
.
24
Krutzik
,
P. O.
,
J. M.
Crane
,
M. R.
Clutter
,
G. P.
Nolan
.
2008
.
High-content single-cell drug screening with phosphospecific flow cytometry.
Nat. Chem. Biol.
4
:
132
142
.
25
Christianson
,
S. W.
,
L. D.
Shultz
,
E. H.
Leiter
.
1993
.
Adoptive transfer of diabetes into immunodeficient NOD-scid/scid mice. Relative contributions of CD4+ and CD8+ T-cells from diabetic versus prediabetic NOD.NON-Thy-1a donors.
Diabetes
42
:
44
45
.
26
Serreze
,
D. V.
,
S. A.
Fleming
,
H. D.
Chapman
,
S. D.
Richard
,
E. H.
Leiter
,
R. M.
Tisch
.
1998
.
B lymphocytes are critical antigen-presenting cells for the initiation of T cell-mediated autoimmune diabetes in nonobese diabetic mice.
J. Immunol.
161
:
3912
3918
.
27
Martinov, T., J. Spanier, L. Swanson, and B. T. Fife. 2018. Programmed death-1 restrains the germinal center reaction in type 1 diabetes. J. Immunol. 200(Suppl. 1): 40.5.
28
Kenefeck
,
R.
,
C. J.
Wang
,
T.
Kapadi
,
L.
Wardzinski
,
K.
Attridge
,
L. E.
Clough
,
F.
Heuts
,
A.
Kogimtzis
,
S.
Patel
,
M.
Rosenthal
, et al
.
2015
.
Follicular helper T cell signature in type 1 diabetes.
J. Clin. Invest.
125
:
292
303
.
29
Belanger
,
S.
,
L.-H.
Tai
,
S. K.
Anderson
,
A. P.
Makrigiannis
.
2008
.
Ly49 cluster sequence analysis in a mouse model of diabetes: an expanded repertoire of activating receptors in the NOD genome.
Genes Immun.
9
:
509
521
.
30
Mortier
,
E.
,
T.
Woo
,
R.
Advincula
,
S.
Gozalo
,
A.
Ma
.
2008
.
IL-15Ralpha chaperones IL-15 to stable dendritic cell membrane complexes that activate NK cells via trans presentation.
J. Exp. Med.
205
:
1213
1225
.
31
Waldmann
,
T. A.
2006
.
The biology of interleukin-2 and interleukin-15: implications for cancer therapy and vaccine design.
Nat. Rev. Immunol.
6
:
595
601
.
32
Noble
,
A.
,
Z. S.
Zhao
,
H.
Cantor
.
1998
.
Suppression of immune responses by CD8 cells. II. Qa-1 on activated B cells stimulates CD8 cell suppression of T helper 2 responses.
J. Immunol.
160
:
566
571
.
33
Insel
,
R. A.
,
J. L.
Dunne
,
M. A.
Atkinson
,
J. L.
Chiang
,
D.
Dabelea
,
P. A.
Gottlieb
,
C. J.
Greenbaum
,
K. C.
Herold
,
J. P.
Krischer
,
Å.
Lernmark
, et al
.
2015
.
Staging presymptomatic type 1 diabetes: a scientific statement of JDRF, the Endocrine Society, and the American Diabetes Association.
Diabetes Care
38
:
1964
1974
.
34
Stocks
,
B. T.
,
C. S.
Wilson
,
A. F.
Marshall
,
L. A.
Brewer
,
D. J.
Moore
.
2017
.
Host expression of the CD8 Treg/NK cell restriction element Qa-1 is dispensable for transplant tolerance.
Sci. Rep.
7
:
11181
.
35
Sundström
,
M.
,
K.
Lejon
.
2010
.
The prolonged and enhanced immune response in the non-obese diabetic mouse is dependent on genes in the Idd1/24, Idd12 and Idd18 regions.
J. Autoimmun.
35
:
375
382
.
36
Carnaud
,
C.
,
J.
Gombert
,
O.
Donnars
,
H.
Garchon
,
A.
Herbelin
.
2001
.
Protection against diabetes and improved NK/NKT cell performance in NOD.NK1.1 mice congenic at the NK complex.
J. Immunol.
166
:
2404
2411
.
37
Arndt
,
B.
,
L.
Witkowski
,
J.
Ellwart
,
J.
Seissler
.
2015
.
CD8+ CD122+ PD-1- effector cells promote the development of diabetes in NOD mice.
J. Leukoc. Biol.
97
:
111
120
.
38
Kim
,
H.-J.
,
R. A.
Barnitz
,
T.
Kreslavsky
,
F. D.
Brown
,
H.
Moffett
,
M. E.
Lemieux
,
Y.
Kaygusuz
,
T.
Meissner
,
T. A. W.
Holderried
,
S.
Chan
, et al
.
2015
.
Stable inhibitory activity of regulatory T cells requires the transcription factor Helios.
Science
350
:
334
339
.
39
Suwanai
,
H.
,
M. A.
Wilcox
,
D.
Mathis
,
C.
Benoist
.
2010
.
A defective Il15 allele underlies the deficiency in natural killer cell activity in nonobese diabetic mice.
Proc. Natl. Acad. Sci. USA
107
:
9305
9310
.
40
Bobbala
,
D.
,
X.-L.
Chen
,
C.
Leblanc
,
M.
Mayhue
,
J.
Stankova
,
T.
Tanaka
,
Y.-G.
Chen
,
S.
Ilangumaran
,
S.
Ramanathan
.
2012
.
Interleukin-15 plays an essential role in the pathogenesis of autoimmune diabetes in the NOD mouse.
Diabetologia
55
:
3010
3020
.
41
Chen
,
J.
,
L.
Feigenbaum
,
P.
Awasthi
,
D. O.
Butcher
,
M. R.
Anver
,
Y. G.
Golubeva
,
R.
Bamford
,
X.
Zhang
,
M. B.
St Claire
,
C. J.
Thomas
, et al
.
2013
.
Insulin-dependent diabetes induced by pancreatic beta cell expression of IL-15 and IL-15Rα.
Proc. Natl. Acad. Sci. USA
110
:
13534
13539
.
42
Xia
,
J.
,
W.
Liu
,
B.
Hu
,
Z.
Tian
,
Y.
Yang
.
2010
.
IL-15 promotes regulatory T cell function and protects against diabetes development in NK-depleted NOD mice.
Clin. Immunol.
134
:
130
139
.
43
Jiang
,
H.
,
S. M.
Canfield
,
M. P.
Gallagher
,
H. H.
Jiang
,
Y.
Jiang
,
Z.
Zheng
,
L.
Chess
.
2010
.
HLA-E-restricted regulatory CD8(+) T cells are involved in development and control of human autoimmune type 1 diabetes.
J. Clin. Invest.
120
:
3641
3650
.
44
Sonawane
,
S. B.
,
J. I.
Kim
,
M. K.
Lee
,
S.-H.
Lee
,
P. E.
Duff
,
D. J.
Moore
,
M.-M.
Lian
,
S.
Deng
,
Y.
Choi
,
H.
Yeh
, et al
.
2009
.
GITR blockade facilitates Treg mediated allograft survival.
Transplantation
88
:
1169
1177
.
45
Kim
,
J. I.
,
S. B.
Sonawane
,
M. K.
Lee
,
S.-H.
Lee
,
P. E.
Duff
,
D. J.
Moore
,
M. R.
O’Connor
,
M.-M.
Lian
,
S.
Deng
,
Y.
Choi
, et al
.
2010
.
Blockade of GITR-GITRL interaction maintains Treg function to prolong allograft survival.
Eur. J. Immunol.
40
:
1369
1374
.

The authors have no financial conflicts of interest.

Supplementary data