Cellular Mechanisms of CCL22-Mediated Attenuation of Autoimmune Diabetes

Chemokines play a central role in the immune system by controlling leukocyte migration and coordinating immune responses through the differential expression of chemokine receptors. Originally described as a monocyte-derived chemokine, CCL22 is also produced by NK, B, and T cells during inflammatory processes (1, 2). It acts via the receptor CCR4, which is highly expressed by T regulatory cells (Tregs), and appears to promote Th2 immune responses (3). Accordingly, expression of CCL22 in several pathologies, including malignancies (4–7) and chronic infections (8, 9), is associated with the accumulation of Tregs. We recently reported the successful harnessing of CCL22 to protect β cells from autoimmune diabetes in NOD mice (10), as well as in islet allografts (11). CCL22 protects from spontaneous diabetes development when expressed in pancreatic β cells in prediabetic NOD mice, as well as from recurrent autoimmunity when expressed by syngeneic islets transplants in diabetic NOD recipients. In both models, CCL22 islet expression is associated with Treg accumulation, and depletion of Tregs abrogates the effectiveness of CCL22, indicating that Tregs are required for CCL22-mediated protection. To advance toward clinical translation of this approach, it is necessary to have a better understanding of the mechanisms underlying CCL22’s ability to attenuate autoimmunity.

Besides their role in controlling leukocyte trafficking, chemokines may also play a role in the activation and function of immune cells (12–16). Notably, it has been reported that CCR4 and its ligands are able to modulate innate immune responses (17, 18). A potential role for CCL22 in Treg function has been suggested by the observation that in breast tumors producing CCL22, infiltrating Tregs are characterized by higher activation status and proliferation rate (5). Another study showed that CCL22 promoted survival of human Tregs in vitro (19). These observations suggest that CCL22 may affect cell activation or function in addition to chemotaxis. In most reports of CCL22-producing cancers, immunosuppression in the tumor microenvironment is attributed to the activity of recruited Tregs. However, the receptor for CCL22 is expressed by other cells in addition to Tregs, including effector T cells, NK and NKT cells, and dendritic cells (DCs). It is conceivable that these cell populations may also be attracted by CCL22 and contribute to immune modulation. We sought to clarify the mechanisms by which CCL22 attenuates autoimmunity and protects from diabetes. To this end, we assessed whether different cell populations, attracted and perhaps activated by CCL22, collaborate to create a tolerogenic milieu, inhibiting autoreactive T cells and thereby protecting β cells.

Animals were housed in the Child and Family Research Institute Animal Unit and cared for according to the guidelines of the Canadian Council on Animal Care and regulations of the University of British Columbia. NOD/ShiLtJ and NOD.scid mice were purchased from The Jackson Laboratory and bred in our facility. CCR4^−/− mice were purchased from The Jackson Laboratory as frozen embryos. CD1d-deficient NOD mice were a gift of Dr. David Serreze at The Jackson Laboratory. Blood glucose was monitored in wild-type and CD1d-deficient NOD females once per week from 8 wk of age. The time of diabetes onset was determined as two consecutive blood glucose measures >20 mM. Islet transplantation was performed in NOD mouse recipients within 10 d of diabetes onset. To maintain normoglycemia in diabetic mice prior to transplant, mice received an s.c. insulin implant (LinShin Canada), which was removed the day before islet transplantation. In some experiments, mice were housed individually and given 5 mg/ml 1-methyl-dl-tryptophan (Sigma-Aldrich) in the drinking water to inhibit IDO with an average consumption of 3–4 ml/day.

For syngeneic islet transplantation in wild-type or CD1d-deficient NOD mice, pancreatic islets were isolated from NOD.scid donors by collagenase (Sigma-Aldrich) injection in the pancreatic duct, digestion, and purification by filtration through a 70-μm filter as described previously (20). Islets were hand-picked and cultured in complete RPMI consisting of RPMI 1640 (Life Technologies), 1% GlutaMAX (Life Technologies), 100 U/ml penicillin (Sigma-Aldrich), 100 μg/ml streptomycin (Sigma-Aldrich), and 10% FBS (Life Technologies). Freshly isolated islets were transduced in overnight culture with adenoviruses expressing a mouse CCL22 cDNA (Ad-CCL22) or LacZ (Ad-LacZ) as previously described (10). Multiplicity of infection (MOI) used for adenoviral transduction (MOI of 10) was estimated assuming that each islet contains an average of 1000 cells. On the following day, 500 islets were transplanted into the renal subcapsular space of diabetic recipients under isoflurane anesthesia. The time of graft failure was determined by measurement of blood glucose >20 mM on 2 consecutive days.

Islet grafts were excised 10 d after transplantation, fixed in 4% paraformaldehyde overnight, and embedded in paraffin. Tissue sections (5 μm) were deparaffinized in xylene, rehydrated in graded ethanol, and washed in PBS. Ag retrieval was performed (Dako) prior to blocking. Sections were incubated for 1 h with primary Abs at room temperature, followed by incubation with secondary Abs. The primary Abs used were guinea pig anti-insulin (Dako), rat anti-mouse Ki67 (Dako), goat anti-mouse IDO (Santa Cruz Biotechnology), rat anti-mouse Siglec-H (Imgenex), and rabbit anti-mouse Foxp3 (Novus Biologicals). The secondary Abs used for immunofluorescence included goat anti-guinea pig Alexa Fluor 488, donkey anti-rat Alexa Fluor 488, and goat anti-rat Alexa Fluor 594 (Molecular Probes), as well as CFL594 donkey anti-goat (Santa Cruz Biotechnology). Immunohistochemistry was performed using donkey anti-goat HRP (Santa Cruz Biotechnology), donkey anti-rabbit HRP (GE Healthcare), or biotinylated anti-rat (Vector Laboratories) followed by streptavidin/HRP (Dako). We applied 3,3′-diaminobenzidine substrate (BioGenex) and counterstained sections with hematoxylin (Invitrogen). IDO staining was performed with anti-goat alkaline phosphatase (Sigma-Aldrich) and New Fuschin Substrate System (Dako). All Abs were diluted in PBS with 1% BSA. Immunostained slides were washed, dehydrated, and mounted with Cytoseal (Thermo Scientific). Images were captured using an Olympus Bx61 microscope and InVivo (Mediacybernetics) or DP Controller software (Olympus).

Harvested grafts, lymph nodes (LNs), and spleens were strained through a 40-μm mesh, washed, and resuspended in PBS. Spleen suspensions were further treated with 0.155 M NH₄Cl (Sigma-Aldrich) plus 10 mM KHCO₃ (Thermo Fisher Scientific) plus 10 μM Na₂EDTA (Thermo Fisher Scientific) for 2 min to lyse RBCs and then washed with PBS. For some in vitro experiments, we further isolated immune populations with EasySep kits from StemCell Technologies for downstream applications such as tissue culture or flow cytometry. Using the kits for CD4 pre-enrichment and CD25 positive selection, we obtained a purity of 90% for the CD4⁺CD25^high population (Tregs) and 93% for CD4⁺CD25⁻ cells (CD4 conventional [CD4conv] cells). DCs were isolated with a CD11c positive selection kit with an average purity of 85%. Alternatively, cell subsets were identified by staining for specific markers and sorted from tissues using a FACSAria cytometer (BD Biosciences).

For immunostaining, cells were resuspended in PBS with 1 mM EDTA (Thermo Fisher Scientific) and 1% FBS (FACS buffer). The Abs used in flow cytometry studies were purchased from eBioscience, BioLegend, and BD Biosciences. CD4 subsets were first gated as CD3⁺CD4⁺CD8⁻ cells; Tregs and CD4conv cells were further identified as Foxp3⁺ and Foxp3⁻ cells, respectively. Lymphoid DCs were identified as CD11c⁺CD8a⁺ cells, myeloid DCs were identified as CD11c⁺CD11b⁺ cells, and plasmacytoid DCs (pDCs) were defined as CD11c⁺PDCA-1⁺ cells. NK cells were identified as CD49b⁺CD3⁻ cells, whereas cells expressing both CD49b and CD3 were defined as non–invariant NKT (niNKT) cells. To detect invariant NKT (iNKT) cells, we used CD3 Ab and CD1d tetramer loaded with α-galactosylceramide obtained from the National Institutes of Health Tetramer Facility. BDC2.5 tetramers were obtained from the National Institutes of Health Tetramer Facility to identify islet-specific CD4⁺ cells. 7-Aminoactinomycin D was used to detect nonviable cells. Cells were incubated with Abs or tetramers for 30 min at 4°C and washed with FACS buffer. For intracellular staining, we used the Cytofix/Cytoperm kit from BD Biosciences to fix and permeabilize the cells. Fluorescent counting beads (Invitrogen) were used to determine the absolute number of cells in samples. Data were acquired with FACSCalibur and LSR II (BD Biosciences) cytometers and CellQuest software. Data were then analyzed using FlowJo 7.6 software.

To study the impact of CCL22 on cell activation and function, splenocytes or isolated cell populations were resuspended in RPMI 1640 with 1% GlutaMAX, 100 U/ml penicillin, 100 μg/ml streptomycin, and 0.1% BSA. Cells were seeded at a density of 0.5–1 × 10⁶ cells/ml in 96-well plates (BD Biosciences) and treated with CCL22 for different periods of time. Recombinant CCL22 was purchased from R&D Systems and reconstituted in PBS with 0.1% BSA prior to addition to cells. In some experiments, media were supplemented with 100 U/ml recombinant human IL-2 (Roche) and CD3/CD28 beads (Invitrogen) at a ratio of 1:1 with cells.

For coculture experiments, Tregs (CD4⁺CD25^high) and CD4conv (CD4⁺CD25⁻) cells were isolated and pretreated with 500 pg/ml CCL22. After 48 h, DCs were freshly isolated and added to the culture at a ratio of 1:2 (DC/CD4 subsets). In some experiments, we added 10 μg/ml CTLA-4 blocking Ab or an isotype control (BioLegend). We also used 1000 U/ml IFN-γ (Roche) to stimulate DCs. Cells were harvested 2 d later for analysis by flow cytometry or real-time PCR. For suppression assays, CD4conv cells and Tregs were purified by cell sorting. Responder CD4conv cells were labeled with cell proliferation dye eFluor 450 (eBioscience) and cocultured with Tregs at ratios ranging from 1:1 to 1:16 (Tregs/CD4conv cells). Activator CD3/CD28 beads were added at ratio 1:7 with CD4conv cells. Cell proliferation was determined 4 d later by flow cytometry.

Chemotaxis assays were performed using 5-μm-pore Transwell plates (Millipore). The media used in these experiments was RPMI 1640 with 1% GlutaMAX, 100 U/ml penicillin, 100 μg/ml streptomycin, and 0.1% BSA. Chemotactic stimulus or medium alone was added to the bottom chamber and the upper chamber was seeded with 10⁵ cells. As chemotactic stimuli, we used recombinant CCL22 at various concentrations, as well as 100 islets transduced overnight with Ad-CCL22. Cells were left to migrate at 37°C for for 1, 2, or 4 h, at which time the upper chambers were aspirated and EDTA (25 mM) was added for 5 min to detach migrated cells from the membrane. The migratory fraction was recovered and resuspended in FACS buffer for cell identification and quantification. Migration index was calculated as the number of cells migrating toward the chemotactic stimulus divided by the number of cells migrating toward medium only.

Total RNA was directly extracted using the RNeasy micro kit from Qiagen. In some cases, tissues or cultured cells were placed in RNAlater stabilization reagent (Qiagen) until RNA isolation could be performed. cDNA was prepared using the SuperScript III reverse transcriptase (Invitrogen) and random primers (Invitrogen). The Applied Biosystems 7500 system was used to monitor real-time amplification of cDNA with the EvaGreen MasterMix (Applied Biological Materials). The primers used for real-time PCR were designed with OligoPerfect designer from Invitrogen. Relative quantification of gene expression was performed using the housekeeping genes rplp0 and 36b4 as controls. The comparative Ct (cycle threshold) method was used for relative quantification of mRNAs of interest. Gene expression was also calculated as fold difference compared with control samples using the 2^−ΔΔCt method.

A mouse cytokine multiplex bead array kit for Luminex was purchased from Millipore to measure concentrations of IL-2, IL-4, IL-10, IL-12, IL-17, IFN, and TNF-α. Cytokine production in supernatants was determined according to the manufacturer’s instructions and analyzed using a Bio-Plex 200 Luminex machine (Bio-Rad).

Data in graphs are represented as means ± SEM. Survival curves were generated using Kaplan–Meier life-table analysis and compared using the log-rank test. A Student t test was used to compare between two groups and one-way ANOVA to compare more than two groups, with Dunnett posttest for comparison between groups. All statistical analyses were performed with GraphPad software, and differences were considered significant when p < 0.05.

CCR4 is present on a variety of immune cells both in humans and mice. In the T cell compartment, CCR4 is found on CD4⁺ T cells upon activation (21, 22), particularly in the Th2 subset, and on memory CD8⁺ cells (23). Additionally, DCs (24, 25), NK cells (26), and NKT cells (23) have all been shown to express CCR4, among other chemokine receptors. To assess the ability of CCR4-expressing cells to respond to CCL22, we first examined CCR4 expression in mouse DC and T cell subsets, as well as NK and NKT cells with or without activation by Con A. The lectin Con A is widely used to activate T cells in vitro (27) and has been shown to stimulate monocytes and NKT cells as well (28, 29). Our results confirm that CCR4 is present on a large proportion of Tregs under steady-state conditions and even more after treatment with Con A (Supplemental Fig. 1A). We observed constitutive expression of the CCL22 receptor in ∼20% of DCs and 45% of NKT cells. As expected, conventional CD4⁺ and CD8⁺ T cells (Tconv cells) expressed CCR4 only upon activation. The converse expression profile was seen on NK cells, which lose CCR4 expression after stimulation with Con A. Analysis of expression levels in CCR4⁺ cells at rest revealed that all populations examined express lower amounts of CCR4 than Tregs, except for NKT cells, which displayed significantly higher levels of the CCL22 receptor (Supplemental Fig. 1B). This interesting finding suggests that NKT cells may be even more sensitive to CCL22 gradients than are Tregs.

The ability of these immune cells to migrate toward CCL22 was explored in chemotaxis assays. We examined cell recruitment by 100–1000 pg/ml recombinant CCL22 and observed significant migration of Tregs, iNKT cells, and pDCs (Fig. 1A), regardless of the concentration used. Interestingly, pDC recruitment was comparable to that of Tregs and iNKT cells, despite lower expression of CCR4. To determine the migration kinetics of these three populations, we performed a time course experiment. Migration of pDCs became detectable after 2 h and continued for the rest of the experiment, whereas Tregs and iNKT cells were already maximally recruited at the 1 h time point (Supplemental Fig. 1C), thus preceding and potentially contributing to pDC recruitment.

We next examined the ability of CCL22-expressing islets to attract immune cells in vitro and found that they selectively recruited Tregs, iNKT cells and pDCs (Fig. 1B). Of note, nontransduced control islets also recruited several cell subsets, particularly NKT cells, which were recruited both by CCL22-expressing and untreated islets. One possible explanation is that isolated islets produced a number of cytokines and chemokines (30, 31) that likely participate in cell recruitment in our assay and perhaps synergize with CCL22 (32). To investigate whether cell migration to CCL22 is indeed mediated by CCR4, we performed chemotaxis assays with CCR4^−/− splenocytes. Recruitment of Tregs, iNKT cells, and pDCs by recombinant CCL22 (Supplemental Fig. 1D) and Tregs to CCL22-expressing islets (Supplemental Fig. 1E) was significantly reduced when these cells lacked CCR4. However, CCL22-transduced islets maintained the ability to recruit, although to a lesser degree, CCR4-deficient pDCs, niNKT cells, and iNKT cells, suggesting that additional factors secreted by islets may contribute to their recruitment.

We next sought to characterize the immune infiltrate in CCL22-expressing islet grafts at different times after transplantation in diabetic NOD recipients. We observed a progressive infiltration by conventional CD8⁺ (CD8conv) and CD4⁺ (CD4conv) T cells that was comparable in LacZ- and CCL22-expressing transplants (Fig. 1C). Consistent with our previous findings, recruitment of Tregs was more evident in CCL22-expressing grafts, with a significant increase in Treg numbers 10 d posttransplantation. Interestingly, iNKT cells markedly infiltrated CCL22-expressing grafts 5 d posttransplantation, preceding the recruitment of Tregs, and this increased number of iNKT cells in the graft persisted over time. Both NK and niNKT cells were retained in CCL22-expressing grafts 10 d posttransplantation, contrary to LacZ-expressing grafts, in which these cell populations declined with time. Infiltration by myeloid and lymphoid DCs was similar in both groups, but the number of pDCs was higher in CCL22-expressing grafts 10 d posttransplantation.

These findings provide in vivo confirmation of our observations in chemotaxis assays showing the selective recruitment of Tregs, iNKT cells, and pDCs by CCL22. The kinetics of graft infiltration demonstrate that iNKT cells are the first population to be recruited by CCL22-expressing grafts, and that they remain within the graft thereafter.

Similar numbers of effector T cell populations in CCL22- and LacZ-expressing islet grafts suggest that these cells do not undergo apoptosis but might rather become anergic. To examine the possibility that CCL22 induces effector T cell anergy, we examined the ability of Tconv cells from islet grafts to proliferate and express cytokines. Histological analysis 10 d posttransplantation revealed lower expression of Ki67 in the immune infiltrate of CCL22-expressing islet grafts compared with control grafts (Fig. 2A). We examined proliferation of CD4conv and CD8conv cells by flow cytometry and found that cells of both subsets extracted from CCL22-expressing islet grafts, but not from draining LNs, displayed significantly lower proliferation rates (Fig. 2B).

We next sought to characterize effector function of Tconv cells in our model. Diabetogenic T cells use several mechanisms to destroy β cells, including secretion of cytokines such as IFN-γ and TNF-α, as well as cytolysis mediated by perforin and granzymes (33, 34). We used cell surface mobilization of CD107a as an indicator of degranulation, which is associated with cytotoxic activity in T cells (35). Lower frequencies of CD107a⁺ cells were found in Tconv cell subsets present in the CCL22-expressing islet grafts (Fig. 2C), suggesting decreased cytotoxicity in these cell populations. To evaluate cytokine gene expression in Tconv populations, CD4conv and CD8conv cells were sorted from islet grafts 10 d posttransplantation for quantitative PCR analysis. In CCL22-expressing islet grafts, both CD4conv and CD8conv cells exhibited significantly lower levels of IFN-γ, consistent with a 10-fold decrease of IFN-γ expression in the whole graft 10 d posttransplantation (Fig. 2D–F). IL-17 and TNF-α also tended to be downregulated in CD4conv and CD8conv cells, respectively, whereas IL-10 was significantly upregulated in CD4conv cells from CCL22-expressing grafts. These data show that proliferation and effector functions are inhibited both in CD4conv and CD8conv cells in the presence of CCL22, compatible with an anergic state (36, 37).

We next characterized the phenotypes of Tregs and CD4conv cells in CCL22-transduced islet grafts 10 d posttransplantation. We found higher proportions of Tregs expressing activation markers CTLA-4, ICOS, and CD62L in CCL22-expressing than in LacZ-expressing islet grafts (Fig. 3A). Moreover, Tregs displayed higher levels of CCR4 in CCL22-transduced transplants (Fig. 3B). In contrast, expression of these markers in CD4conv cells were comparable in LacZ- and CCL22-expressing islet transplants. In the renal LN draining the islet graft, Treg populations did not differ between transplant groups, pointing to a localized effect of CCL22 at the site of expression. The frequency of BDC2.5-specific CD4⁺ T cell subsets was similar in both groups of transplant recipients (Fig. 3C), suggesting that CCL22 did not preferentially recruit islet-specific Tregs. In contrast to a report showing Treg proliferation in CCL22-expressing tumors (5), the frequency of Ki67⁺ Tregs was similar in CCL22- and LacZ-transduced syngeneic islet grafts (Fig. 3D), in accordance with what we recently reported in CCL22-expressing islet allografts (11). This result indicates that Treg proliferation does not account for the increased numbers of Tregs observed in CCL22-expressing islet grafts.

Because it is possible that CCL22 simply recruits Tregs with both higher CCR4 expression and higher levels of cell-surface activation markers, we next incubated whole splenocytes or isolated CD4 cell subsets with recombinant CCL22 and analyzed the presence of activation markers by flow cytometry after 24 and 48 h. Notably, expression of CD86 and MHC class II on DCs (Supplemental Fig. 3A), as well as CD44 and CD69 on Tconv cells (data not shown), remained unchanged following treatment with CCL22, suggesting that CCL22 does not impact the activation status of these cell populations. In contrast, we observed selective activation of Tregs as evidenced by upregulation of CTLA-4, ICOS, CD62L and CCR4 after 48 h treatment with CCL22 (Fig. 4A, 4B). We then compared the effect of CCL22 treatment with TCR activation on the expression of CTLA-4, ICOS, and CCR4. TCR stimulation enhanced expression of CTLA-4 and ICOS on CD4conv cells, with no additional effect by CCL22 (Fig. 4C). CCL22 modestly increased CTLA-4 expression on Tregs in the presence of TCR stimulation, suggesting that CCL22 can induce maximal expression of CTLA-4 in Tregs. ICOS expression on Tregs was upregulated by CCL22 alone, although to a lower level than CD3/CD28 stimulation, and CCL22 had no additive effects on ICOS expression in the presence of TCR stimulation. CCL22 increased expression of its own receptor, CCR4, in Tregs, which was not impacted by CD3/CD28 stimulation (Fig. 4D). Of note, similar results were obtained with IL-2 supplementation (data not shown), suggesting that these findings were unlikely to be due to selective death of cells expressing lower surface levels of these markers. Collectively, these data indicate that CCL22 is able to promote an activated phenotype on Tregs, whether in the presence or absence of other stimulatory signals.

Because ICOS is involved in Treg survival (38) and is upregulated by CCL22 treatment of Tregs, we next determined whether CCL22 treatment affected Treg viability. In the presence of the prosurvival factor IL-2, viability was comparable between control and CCL22-treated Tregs during a 24-h period. As expected, in the absence of IL-2, Treg viability declined, but CCL22 treatment restored Treg viability (Fig. 4E). This finding indicates that CCL22 promotes Treg survival, consistent with a previous study on human Tregs (19).

We sought to determine the effect of CCL22 on Treg suppression of CD4conv cell proliferation in vitro. CCL22-treated Tregs showed similar suppressive capacity to untreated Tregs at all cell ratios tested (Supplemental Fig. 2A), indicating that CCL22 does not enhance Treg inhibition of CD4conv cell proliferation. Consistent with our in vivo data, we found no direct impact of CCL22 on the proliferation rate of Tregs (Supplemental Fig. 2B). We next examined gene expression of several Treg effector molecules after incubation with CCL22. Although the differences were modest, we observed downregulation of IL-17 and upregulation of IL-10 mRNA following 24 h exposure to CCL22 (Supplemental Fig. 2C). IL-10 protein levels tended to be higher in the supernatant of CCL22-treated Tregs after 24 h (Supplemental Fig. 2D), whereas IL-2, IL-4, IL-17, and IFN-γ remained undetectable (data not shown).

Tregs have been shown to condition DCs to express the potent regulatory enzyme IDO in a CTLA-4–dependent manner (39). Because CCL22 enhances the expression of CTLA-4 on Tregs, we tested the ability of CCL22-treated Tregs to modulate DC function in coculture experiments. We found that IDO expression was markedly induced in DCs cultured for 48 h with CCL22-treated Tregs (Fig. 5A). IDO upregulation in DCs was abrogated by CTLA-4 blockade, indicating that CCL22-treated Tregs promote expression of IDO in DCs via CTLA-4. DC activation status was modestly downregulated by CCL22-treated Tregs, as shown by lower levels of CD86 and MHC class II molecules (Supplemental Fig. 3A). DCs cultured with CCL22-treated Tregs produced less TNF-α and tended to secrete less IFN-γ and IL-12 (Supplemental Fig. 3B). These findings indicate that CCL22 potentiates the ability of Tregs to modulate DC function.

Immunostaining of islet grafts 10 d posttransplantation revealed a trend toward increased numbers of IDO-producing cells in recipients of CCL22- compared with LacZ-expressing islet grafts, surrounding islets and forming small clusters (Fig. 5B). Real-time PCR analysis of islet grafts and draining LNs showed that IDO mRNA levels were increased by 7-fold in CCL22-expressing grafts 10 d posttransplantation compared with LacZ-expressing grafts and tended to be higher in the draining LN of CCL22-expressing graft recipients as well (Fig. 5C). Because DCs are a major source of IDO (40), we examined IDO gene expression in CD11c⁺ cells isolated from the islet graft. The population of DCs found within CCL22-islet transplants was characterized by significantly higher levels of IDO, whereas expression of TNF-α and IFN-γ genes was comparable to that in DCs from LacZ-islet grafts (Fig. 5D). Our findings suggest that CCL22 expression in islets produces a local tolerogenic milieu by the interplay of Tregs with DCs. Inhibition of IDO by administration of the IDO inhibitor 1-methyl-d-tryptophan in the drinking water of a small cohort of recipients of CCL22-expressing islets tended to shorten graft survival (Supplemental Fig. 3C), suggesting that CCL22-mediated protection from recurrent autoimmunity is at least partly dependent on IDO induction by Tregs.

Enrichment of NKT cells is known to prevent autoimmune diabetes in NOD mice (41, 42), and the significant recruitment of iNKT cells by CCL22 points to their potential contribution to the protective effect of CCL22 in our model. To determine the role of NKT cells in CCL22-induced protection of syngeneic islet grafts, we used NOD mice lacking CD1d as transplant recipients. Thymic expression of CD1d is required for NKT cell development, and thus CD1d-deficient mice lack these cells (43). In CD1d^−/− recipients, islet grafts transduced with either Ad-LacZ or Ad-CCL22 had a mean survival of 11.1 ± 1.9 and 13.1 ± 3.2 d, respectively, comparable to that of LacZ-expressing grafts in wild-type recipients (Fig. 6A). Notably, CCL22-expressing grafts have shortened survival (p < 0.05) in the absence of NKT cells, suggesting that NKT cells might be important for CCL22 protection from recurrence of autoimmune diabetes.

To determine the potential impact of NKT cells on graft infiltration by Tregs and pDCs, we compared islet grafts from wild-type and CD1d-deficient recipients by histology. Overall, graft infiltration by Tregs and pDCs was lower in CD1d^−/− mice compared with wild-type recipients (Figs. 6B). Islet expression of CCL22 significantly promoted recruitment of Tregs (p < 0.05), but not pDCs, in CD1d^−/− recipients. Interestingly, we also found fewer cells expressing IDO in CD1d-deficient mice, and CCL22 did not enhance their numbers. In vitro migration assays showed that Tregs were similarly recruited by CCL22 regardless of the presence of NKT cells; however, pDCs no longer migrated toward recombinant CCL22 in the absence of NKT cells (Supplemental Fig. 4A). Differences in leukocyte subsets between CD1d^−/− mice and wild-type littermates could potentially confound this observation, but we normalized the migratory fraction to its original population to avoid this bias. We also tested the ability of DCs from CD1d^−/− mice to produce IDO in vitro and found similar upregulation of IDO levels in wild-type and CD1d^−/− DCs upon stimulation with IFN-γ or coculture with CCL22-treated Tregs (Supplemental Fig. 4B). Taken together, these data suggest that NKT cells promote pDC recruitment by CCL22, likely by stimulation of secretion of other chemokines, as previously described (44, 45). Hence, iNKT cells may contribute to CCL22 protection from autoimmune diabetes by helping to bring pDCs to the site of inflammation, where Tregs may condition them to produce IDO.

The chemokine CCL22 has been shown to modulate the antitumor immune response and thus promote tumor growth. We previously reported the harnessing of CCL22 to protect β cells from spontaneous and recurrent autoimmunity in NOD mice, using a gene delivery approach to engineer the expression of CCL22 in islets. In this study, we sought to understand the mechanism underlying the prevention of autoimmune diabetes by CCL22. Our data demonstrate that CCL22 prevents autoimmune diabetes not only by recruiting but also by activating Tregs, and by recruiting other immune cell subsets including iNKT cells and pDCs, which likely play a role in attenuating local autoimmunity in the islet.

CCR4 is known as a receptor involved in Treg trafficking, although its expression is not exclusive to Tregs. We report in the present study that a large proportion of NKT cells express CCR4 constitutively, and, more importantly, that CCR4 levels are higher on NKT cells, especially iNKT cells, than on Tregs, suggesting a greater sensitivity to CCL22 gradients. In accordance with this idea, we demonstrated that iNKT cells are the first leukocyte subset to be recruited specifically by CCL22-expressing islet grafts. Influx of Tregs as well as pDCs was observed later at 10 d posttransplantation. Intriguingly, all subsets of DCs express relatively low levels of CCR4, but only pDCs are found in increased numbers in CCL22-expressing grafts. Our data suggest that the influx of pDCs sequentially follows that of iNKT cells and that NKT cells are likely important for their recruitment by CCL22. Whereas migration of Tregs to CCL22-expressing islets appears to be entirely CCR4-dependent, recruitment of iNKT cells and pDCs by CCL22-expressing islets is only partly mediated by CCR4, indicating that other islet-derived chemokines may also be involved. Both NKT cells and pDCs are capable of inducing immune tolerance by influencing a variety of immune cells, including conventional T cells, Tregs, DCs, and NK cells, either by cell-to-cell contact or via cytokines (46, 47).

Several studies in the oncology field have demonstrated the capacity of CCL22 to suppress the immune response, a property that has been ascribed to Tregs recruited by CCL22. Besides showing that Tregs are not the only regulatory population recruited by CCL22, we now also provide evidence that CCL22 selectively activates Tregs and promotes their regulatory function. Indeed, our data show that CCL22-conditioned Tregs are characterized by modestly increased expression of CTLA-4, ICOS, CD62L, and the CCL22 receptor, CCR4. Notably, a comparable phenotype is observed on Tregs that infiltrate CCL22-expressing islet transplants. Whether these CCL22-induced changes in Treg gene expression are of significance for in vivo Treg function will require further study. Gobert et al. (5) reported a similar finding in the context of human breast tumors, where most Tregs expressed CTLA-4, ICOS, and CCR4. Expression of either CTLA-4, ICOS, or CD62L on Tregs endows them with superior suppressive activity and ability to protect from autoimmunity, particularly in the context of autoimmune diabetes (38, 48–50). Interestingly, we also found that CCL22 promoted Treg viability without inducing production of the prosurvival factor IL-2.

Importantly, Tregs treated with CCL22 enhance the production of IDO by DCs. In CCL22-expressing islet grafts, IDO expression was significantly higher, particularly within the DC compartment, and the number of IDO-expressing cells also tended to be higher. Our data further indicate that CCL22 promotes Treg interaction with DCs via CTLA-4 to induce production of IDO, an enzyme that inhibits T cell proliferation (51, 52). The activity of IDO is associated with tolerance induction in many settings, including transplantation and autoimmune diabetes (53, 54). Onodera et al. (55) described the importance of the CCL22/CTLA-4 axis in the induction of IDO under physiologic conditions. They showed that DCs in mesenteric LNs secreted large amounts of CCL22 and thereby attracted Tregs, which induced IDO expression in these DCs in a CTLA-4–dependent manner. Our findings suggest that we recreated this tolerogenic axis by expressing CCL22 in islet transplants. Moreover, our preliminary studies in a small cohort of NOD recipients of CCL22-expressing syngeneic islet grafts that received the IDO inhibitor 1-methyl-d-tryptophan suggest that CCL22-mediated protection from recurrent autoimmunity may be at least partly mediated by IDO.

A protective role has been attributed to both iNKT and niNKT cells in the context of autoimmune diabetes, as demonstrated by transgenic overexpression or adoptive transfer of these cells (42, 56). The shorter survival of CCL22-expressing islet grafts in CD1d-deficient mice suggested that NKT cells might be required for CCL22-mediated protection from recurrent autoimmunity. The pronounced and early recruitment of iNKT cells by CCL22-expressing grafts suggests that this particular subset of NKT cells might be important for the protective effect of CCL22.

That both Tregs and iNKT cells mediate CCL22 protection suggests that they cooperate to prevent immune attack of β cells. Although each of these regulatory subsets is capable of suppressing immune responses individually, such interactions have been implicated in the prevention of autoimmune diabetes (45, 57), rejection of heart transplants (58), and graft-versus-host disease (59) in rodent models. Several studies demonstrated that pDCs are able to prevent diabetes, notably by producing IDO and interacting with NKT cells (60, 61). In our model of recurrent autoimmune diabetes, elevated levels of IDO in CCL22-expressing grafts are concomitant with the local accumulation of pDCs, suggesting that pDCs participate in the protective effect of CCL22 by producing IDO. Our data support the hypothesis that one of the roles of iNKT cells is to promote pDC migration to CCL22-expressing islet grafts, where pDCs and other DC subsets may produce IDO under the influence of Tregs.

Finally, CCL22 expression in islet grafts did not prevent infiltration by effector T cells; moreover, their frequency and number did not decline over time. However, we observed reduced effector function of infiltrating conventional CD4⁺ and CD8⁺ cells, characterized by lower proliferation, cytotoxicity, and IFN-γ expression. Thus, CCL22 modulation of T cell responses seems to occur by induction of cell anergy and tolerance at the site of inflammation, leading to reduced immune destruction of β cells. Suppression of effector T cells likely results from the accumulation of regulatory subsets, in particular iNKT cells and Tregs, and upregulation of IDO in CCL22-expressing islet grafts.

In conclusion, the present study provides significant insight into the immunomodulatory mechanisms of CCL22, a chemokine that is often expressed by human cancers to evade immune destruction. We show, to our knowledge for the first time, that CCL22 expression in vivo recruits iNKT cells, which may collaborate with Tregs in mediating CCL22’s protective effect. Our data suggest that iNKT cells and Tregs cooperate to induce tolerance toward β cells through modulation of DCs, more specifically pDCs. This particular cellular interplay has recently been recognized, again in a model of protection from autoimmune diabetes (62). The mechanisms of CCL22-mediated protection that we reveal in the present study are likely at play not only in spontaneous autoimmune diabetes (10) and recurrent autoimmunity, but also in islet allograft rejection (11). The other ligand for CCR4, CCL17, has also been described in human tumors that attract Tregs (4–6); hence, it is possible that CCL17 will similarly offer protection from anti-islet immune responses. Because the suppressive function and response to CCR4 ligands of iNKT cells and Tregs are similar between humans and mice, we anticipate that CCL22 therapy will be efficient in humans as well. It is noteworthy, in this respect, that an even higher proportion of human than of mouse Tregs express CCR4 (63). Both iNKT cells and Tregs are known to protect from autoimmunity (64, 65) and induce tolerance to allografts (59, 66, 67). Therefore, CCL22 therapy could be beneficial in other autoimmune disorders and organ transplantation.

We thank members of the Verchere, Levings, Tan, and van den Elzen laboratories for insightful comments and assistance. We also thank the Diabetes Research Program at the Child and Family Research Institute for core support.

The authors have no financial conflicts of interest.