Functional Specialization of Islet Dendritic Cell Subsets

Abstract

Dendritic cells (DC) play important roles in both tolerance and immunity to β-cells in type 1 diabetes. How and why DC can have diverse and opposing functions in islets remains elusive. To answer these questions, islet DC subsets and their specialized functions were characterized. Under both homeostatic and inflammatory conditions, there were two main tissue resident DC subsets in islets, defined as CD11b^lo/−CD103⁺CX3CR1⁻ (CD103⁺DC), the majority of which were derived from Flt3 dependent pre-DC; and CD11b⁺CD103⁻CX3CR1⁺ (CD11b⁺DC), the majority of which were derived from monocytes. CD103⁺DC were the major migratory DC and cross-presented islet derived antigen in the pancreatic draining lymph node (LN), although this DC subset displayed limited phagocytic activity. CD11b⁺DC were numerically the predominant subset (60–80%), but poorly migrated to the draining LN. Although CD11b⁺ DC had greater phagocytic activity, they poorly presented antigen to T cells. CD11b⁺DC increased in numbers and percentage during T cell mediated insulitis, suggesting that this subset might be involved in the pathogenesis of diabetes. These data elucidate the phenotype and function of homeostatic and inflammatory islet DC, suggesting differential roles in islet immunity.

Introduction

Islets contain resident DC that play important roles in type 1 diabetes and allogeneic islet transplant rejection. Both protective and pathogenic roles for DC in type 1 diabetes have been reported (1–3). DC are a heterogeneous population that consists of different subsets with distinct developmental, phenotypic and functional properties. Several groups have shown that NOD mice have an imbalance in the numbers of specific splenic DC subsets, with a relative decrease in the number of CD8⁺DC and an increase in myeloid DC (4). Characterization of islet DC subsets in NOD has not been reported.

Non-lymphoid tissue DC capture antigen and migrate from the periphery to the draining LN, where they initiate immune responses, and deliver cues that influence T cell effector function. In the periphery, tissue resident DC phagocytose apoptotic cells generated during normal tissue turnover, migrate to draining LNs, and present antigens derived from the apoptotic cells (5, 6). This presentation of self-antigens under steady-state conditions is thought to lead to deletion or anergy of self-reactive T cells, thereby maintaining T cell tolerance. Distinct DC subsets have been identified in non-lymphoid tissues and can be categorized based on the surface markers CD103, CD207, CD11b and CX3CR1 (7–10). CD103⁺DC are derived exclusively from pre-DC under the control of fms-like tyrosine kinase 3 (Flt3) ligand and its receptor Flt3, whereas CD103⁻DC are a heterogeneous population dependent on both Flt3 and macrophage colony-stimulating factor receptor (MCSF-R) (7–9). The two DC subsets have different functions in the intestine and lung (8, 9, 11, 12). Lamina propria CX3CR1⁺CD103⁻DC sample intestinal antigens by projecting dendrites through the epithelial cell layer and into the lumen, may serve as a first line of defense by phagocytosing and killing bacteria (13–16). CD103⁺DC from gut-associated mesenteric LNs produce endogenous TGFβ and retinoic acid (RA), and are capable of differentiating naïve T cells into Foxp3⁺ Tregs independent of exogenous TGFβ (17–21). Lung-migrating CD103⁺DC are the major contributors in cross-presentation for CD8⁺T cells under tolerogenic conditions (22) and for activation following poxvirus infection (23). Dermis-derived CD103⁻DC but not CD103⁺DC constitutively produce RA and induce adaptive Treg (24). Thus, the heterogeneity of DC highlights their functional versatility in shaping local tissue immunity and their collaboration in orchestrating immune responses. These subsets are also found in islets (7), but their functional roles have not been tested.

Studies have shown that antigen presenting cells in islets present β-cell-derived peptides bound to their class II MHC molecules (25). T cell mediated inflammation induces islet DC maturation, which leads to further processing of captured antigens (26). However, in those reports separate DC subsets were not characterized, and whether they had distinct functions was not explored. In this study, we characterized two major DC subsets in islets, as well as their origins and specialized functions during both the steady-state and inflammation. CD103⁺DC were the major migratory DC subset and responsible for cross-presenting antigens to CD8⁺ T cells. CD11b⁺DC were the major phagocytic cells, whose number was significantly increased during the islet inflammation. Our studies uncovered islet DC heterogeneity, which contribute to an understanding of the mechanisms that balance islet inflammation and tolerance.

Materials and Methods

Mice

BALB/c, C57BL/6, MIP-GFP, RIPmOVA, CD11c-cre-GFP, CCR7^−/−, Plt, C57BL/6.OT1 and C57BL/6.OT2 mice were from The Jackson Laboratories (Bar Harbor, ME). CX3CR1^GFP/GFP on the C57BL/6 and BALB/c backgrounds were from Dr. D. Littman (Skirball Institute, New York, NY). CX3CR1^GFP/GFPC57BL/6 mice were crossed with C57BL/6 mice to produce CX3CR1^GFP/+. Flt3^−/− C57BL/6 mice (27) were provided by Dr. M. Merad (Mount Sinai, New York, NY). LysM-Cre × Rosa26-stopfloxEGFP mice were provided by Dr. G. Randolph (Mount Sinai). All mice were housed in a pathogen-free animal facility. All experimental protocols were approved by the Institutional Animal Care and Utilization Committee of University of Maryland Medical Center.

Cell preparations

Mice were euthanized, the common bile duct was exposed and injected with 3 ml cold Hanks’ buffer containing 1.5 mg/ml of collagenase-P (Roche Diagnostics, Indianapolis, IN), and the pancreas was excised and digestion allowed to continue at 37°C for 15 minutes. The digested pancreas was disrupted by trituration and the suspension washed twice with RPMI1640 containing 10% fetal bovine serum (FBS). Pancreatic islet separation was performed by centrifugation on a discontinuous Ficoll (Sigma, St. Louis, MO) gradient of 11%, 21%, 23%, and 25%. Islets were picked from the interface between the first and second layers. Islets or LNs were incubated for 30 min in 10% FBS in Hanks balanced salt solution (HBSS) containing 0.2 mg/ml collagenase type IV (working activity of 770 U/mg, Sigma), then homogenized and passed trough a 19G syringe to obtain a single cell suspension.

Flow cytometry

Pancreatic islet cells were resuspended in phosphate buffer solution (PBS) containing 1% bovine serum albumin (BSA), 2 µg/ml Fc-blocking buffer (eBioscience Inc., San Diego, CA) and 2 mM EDTA, and stained with antibodies at 4°C. Flow cytometric analyses were performed on an LSRII flow cytometer (eBioscience) with FlowJo software (Tree Star Inc., Ashland, OR). Dead cells were excluded by DAPI (Invitrogen, Carlsbad, CA) staining. Fluorochrome- or biotin-conjugated mAbs specific to mouse B220 (RA3-6B2), CD8α (53–6.7), MHC class II (MHCII) (M5/114.15.2), IA^g7 (IA^k, 10–3.6), CD103 (2E7), CD11b (M1/70), CD11c (N418), CD45 (30F11), Gr-1 (RB6-8C5), Langerin (eBioL31), CD40 (1C10), CD80 (16-10A1), CD86 (GL1), CCR7 (4B12), H-2Kb-OVA_257–264 (SIINFEKL) peptide bound to H-2Kb, and CD4 (L3T4), the corresponding isotype controls, and the secondary reagents (allophycocyanin-conjugated streptavidin) were purchased either from BD Biosciences-Pharmingen (San Jose, CA), eBioscience or Biolegend (San Diego, CA). Anti-F4/80 (A3-1) mAb was purchased from AbD Serotec, Oxford, UK.

Islet whole mount staining

Hand-picked islets isolated from CD11c-cre-GFP mice were incubated with anti-CD31 (eBioscience), followed by Cy3-conjugated goat anti-rat IgG (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA), and then mounted with Vectashield (with DAPI) (Vector Laboratories, Inc., Berlingame, CA). Images were acquired with a Leica DMRA2 fluorescent microscope (Leica Microsystems, Wetzlar GmbH, Germany), and Openlab (Improvision, Inc., MA).

Induction of islet inflammation

Male mice (8–10 weeks old) received intraperitoneal injections of streptozotocin (STZ, Sigma-Aldrich, St. Louis, MO), at a dose of 40 mg/kg daily for 5 consecutive days to induce T cell dependent inflammation and autoimmune diabetes (28).

T cell stimulation in vitro

OT1 CD8⁺T cells were isolated with the Dynal CD8 Negative Isolation Kit (Invitrogen). OT2 CD4⁺T cells were isolated with the Dynal Mouse Negative Isolation Kit (Invitrogen). CD103⁺CD11b^low/−DC, CD103⁻CD11b⁺DC and CD103⁺CD11b⁺DC from islets of RIPmOVA or littermate C57BL/6 mice were sorted by flow cytometry. Islet DC were gated on DAPI⁻CD45⁺MHCII⁺CD11c⁺ and CD103 vs CD11b. For T cell proliferation, isolated OT1 cells or OT2 T cells were labeled with carboxyfluorescein succinimidyl diester (CFSE, Invitrogen). OT1 or OT2 cells were co-cultured with the indicated number and subset of DC, or dispersed islet or LN cells in 200 µl complete medium with or without 1 ng/ml IL-2 (eBioscience) in 96-well round-bottomed plates. To get ovalbumin (OVA) pulsed DC, sorted DC subsets were incubated with 1 µg/ml OVA (Sigma) in 1 ml 10% FBS/RPMI1640 at 37°C for 1 hour and then washed three times. For cytokine production, pancreatic islet or LN DC subsets were sorted by flow cytometry from C57BL/6 mice. 1×10⁵ OT2 cells were co-cultured with 500 of the indicated DC subset pulsed with or without OVA_323–339 (American Peptide Company,inc. Sunnyvale, CA) in 200 µl complete medium with or without 2 ng/ml TGFβ (eBioscience) in 96-well round bottomed plates. After 5 days, cells were analyzed by flow cytometry. Intracellular staining for Foxp3 expression was performed following the manufacturer’s protocol (eBioscience). Intracellular cytokine (IL-4, IL-17 and IFNγ) staining was performed after restimulation with PMA (100 ng/ml, sigma) and ionomycin (500 ng/ml, eBioscience) in the presence of monesin (1/1000 dilution, eBioscience) for 4 hours.

Cell transfers

Naïve CD4⁺T cells were isolated with CD4⁺CD62L⁺T cell Isolation Kit II (Miltenyi Biotec, Auburn, CA) from OT2 LN and spleen cells, and expanded with anti-CD3 (eBioscience)-coated plate and 1 µg/ml anti-CD28 (eBioscience) for 3 days. 2 × 10⁶ OT2 CD4⁺T cells/mouse were transferred to RIPmOVA or littermate C57BL/6 recipients via i.p. injection. CD8⁺T cells were enriched from OT1 LN and spleen cells using the CD8α microbeads (Miltenyi Biotec), and expanded with anti-CD3-coated plate and 1 µg/ml anti-CD28 for 3 days. 3.5 × 10⁶ CD8⁺ cells/mouse were transferred into RIPmOVA or littermate C57BL/6 recipients via i.v. injection.

Quantitative real-time PCR (qRT-PCR)

The procedures for RNA isolation, cDNA synthetsis and quantification by qRT-PCR were described previously (29). qPCR used Oligo(dT) primers on the LightCycler 2.0 (Roche). Relative expression was calculated as 2^{cycle threshold [Ct] control – Ct gene} using cyclophilin A as an endogenous control. Primer sequences for CCR7, forward: 5’-CACGCTGAGATGCTCACTGG-3’ and reverse: 5’-CCATCTGGGCCACTTGGA-3’.

OVA uptake assay

200 µg alexa-488 labeled OVA (Invitrogen) were intravenously injected into C57BL/6 mice and islet single cell suspensions were assayed by flow cytometry 1.5 and 16 hours later. Islet single cell suspensions were incubated with 2 µg alexa-488 labeled OVA in 2 ml 10% FBS/RPMI1640 at 37°C with or without 0.5% sodium azide (NaN₃), or on ice for 1 hour, cells were washed three times, and assayed at 4°C by surface marker staining.

Statistical Analysis

The differences were assessed using the unpaired Student’s t test and expressed as the mean ± SEM. A p-value of <0.05 was taken to be statistically significant.

Results

Phenotypes and origins of islet DC during steady-state

Whole mount immunofluorescent staining of islets showed that a CD31⁺ blood vessel network was contained within normal naive islets. Most CD11c-GFP⁺ DC were in close proximity to the surface of intra-islet blood vessels (Fig. 1A), suggesting that DC were actively probing the vessel and the surrounding area (25, 30). Next we analyzed the islet DC subsets. As shown in Fig. 1B, there were two major tissue resident DC (CD11c⁺MHCII⁺) subsets in pancreatic islets during the steady-state (7), which we refer to as CD11b⁺DC (CD11b⁺CD103⁻CX3CR1⁺F4/80⁺) and CD103⁺DC (CD11b^lo/−CD103⁺CX3CR1⁻F4/80⁻). Neither subset expressed CD80 or CD40. Both expressed low levels of CD86 and high levels of MHC class II, showing their semi-mature phenotypes (Fig. 1B). In pre-diabetic NOD, there also were two major islet DC subsets, CD11b⁺DC and CD103⁺DC (Fig. 1C). The number of both CD11b⁺DC and CD103⁺DC was significantly increased in NOD mice, compared to diabetes resistant NOR mice (Fig. 1C, 1E). Although the percentage and co-stimulatory molecule expression of these two subsets were similar, the CD103 expression level was lower in NOD mice than in NOR mice (Figs. 1C, 1D).

FIGURE 1. Islet DC at steady state.

(A) Whole mount immunofluorescent staining of islet DC. 200 × or 400 × magnification. Blood vessel (CD31, red) and DC (CD11c-GFP, green). (B) Pancreatic islet single cell suspensions from normal C57BL/6 mice were gated on FSC-A vs. FSC-W, DAPI⁻CD45⁺MHCII⁺CD11c⁺, and CD103 vs. CD11b. Histograms show receptor expression profile of CD11b⁺ DC (red line) and CD103⁺ DC (blue line). (C)–(E) Identification of islet DC of pre-diabetic NOR or NOD (12 weeks). (C) and (D) Islet cells were gated on FSC-A vs. FSC-W and DAPI⁻CD45⁺MHCII⁺CD11c⁺, and CD103 vs. CD11b. (D) Histograms show CD80 and CD86 expression on islet CD11b⁺CD103⁻ and CD103⁺CD11b^low/−DC. Solid line: NOD; dashed line: NOR. (E) CD11b⁺CD103⁻ and CD103⁺CD11b^low/− DC gated on FSC-A vs. FSC-W and DAPI⁻CD45⁺IA^g7+CD11c⁺CD11b vs. CD103. Graphs show percentage of CD103⁺DC and CD11b⁺ DC in IA^g7+CD11c⁺ islet DC. Each data point corresponds to one mouse, n=4. Mean ± SEM.

Next we determined the origin of these DC subsets. The Flt3 receptor has been reported as a key molecule for the development of CD103⁺DC in most non-lymphoid organs (7, 11). As shown in Figs. 2A and 2B, both the number and percentage of islet CD103⁺DC were dramatically decreased in Flt3^−/− mice, whereas the number of CD11b⁺DC were not affected (7), demonstrating that the development of islet CD103⁺DC but not CD11b⁺DC was dependent on Flt3 and derived exclusively from pre-DC.

FIGURE 2. Islet DC origin at steady state.

(A) and (B) Flt3 is required for CD103⁺ DC homeostasis. (A) Dot plots show CD103 and CD11b expression among DAPI⁻CD45⁺MHCII⁺CD11c⁺ islet DC in wild type C57BL/6 (B6) or Flt3^−/−. (B) The graph shows the percentage (left) and absolute numbers (right) of CD103⁺CD11b^−/low and CD103⁻CD11b⁺ DC among DAPI⁻CD45⁺MHCII⁺CD11c⁺ islet DC in Flt3^−/− and control wild type C57BL/6. Each data point corresponds to one mouse. n=5–13. Mean ± SEM. (C) and (D) Tracing the origin of islet DC in LysM-Cre × Rosa26-stopfloxEGFP mice. (C) Dot plots show the percentage of GFP⁺ cells among CD115⁺Gr1^hi blood monocyte (upper panel) and islet DC subsets (middle panel). High side scatter (SSC) and lower levels of CD115 expression were used to discern blood neutrophils (31). Islet cells were gated on FSC-A vs FSC-W and CD45⁺DAPI⁻. (D) The percentage of GFP in islet DC subsets was normalized to the percentage of GFP⁺Gr1^hi monocytes in the same mouse. Graph showing the ratio of percentage of GFP⁺ islet DC and percentage of blood GFP⁺Gr1^hi monocyte. Each data point corresponds to one mouse, n=5. Mean ± SEM.

The possibility that islet steady-state DC were derived from monocytes was tested by using LysM-Cre × Rosa26-floxstopfloxEGFP mice. In these reporter mice, Cre activity removes a stop cassette upstream of the floxed reporter and results in irreversible GFP expression in LysM⁺ cells, including monocytes and their progeny. Although all monocytes express LysM, not all monocytes are GFP⁺ in these mice (31). Therefore, it has been reasoned that DC expressing GFP at levels and percentages comparable to blood monocytes are thus derived primarily from monocytes (31, 32). We found that CD11b⁺DC had similar GFP expression levels in comparison to blood monocytes, whereas CD103⁺DC expressed much lower levels of GFP (Figs. 2C, 2D), indicating that the majority of CD11b⁺DC but not CD103⁺DC were derived from monocytes under steady-state conditions.

CD103⁺DC are the major migratory DC

Tissue draining LN contain tissue migratory and resident DC (33). Migration of DC from tissue to the draining LNs is dependent on CCR7 (34, 35). Islet CD103⁺DC but not CD11b⁺DC expressed CCR7 (Fig. 3A), suggesting only the CD103⁺ subset had the potential to migrate to draining LN; and the distribution of islet DC subsets was not altered by CCR7 or its ligand deficiency (Fig. 3B). In the pancreatic draining LN, there were three DC populations, which were CD11b⁻CD103⁺, CD11b⁺CD103⁻ and a third population of CD11b⁺CD103⁺DC. Neither CD11b⁻CD103⁺ nor CD11b⁺CD103⁺ pancreatic LN DC expressed CX3CR1, which was expressed on the islet CD11b⁺ DC subset (Figs. 1B, 3C), suggesting that both LN CD11b⁻CD103⁺DC and the third population CD11b⁺CD103⁺DC were derived from islet CD103⁺DC (11). Significantly, both CD103⁺DC subsets were significantly reduced in the pancreatic LN of CCR7^−/− mice in the steady-state, whereas the pancreatic LN CD11b⁺CD103⁻DC remained unaltered in these mice (Figs. 3C, 3D). Together these data strongly suggested that islet CD103⁺DC activity migrated from islets to LN and were the major migratory DC subset.

FIGURE 3. CD103⁺DC are the major migratory DC.

(A) CCR7 expression on islet CD103⁺CD11b^−/low and CD103⁻CD11b⁺ DC. Left panel, mRNA expression profile, pooled islet DC sorted from 10 mice. Right panel, flow cytometry histogram. (B) Distribution of islet DC subsets by flow cytometric analysis in wild type and CCR7^−/− mice. Cells were gated on DAPI⁻CD45⁺MHCII⁺CD11c⁺ CD11b vs. CD103. n=2–3, Mean ± SEM. (C) and (D) LN Cells were gated on DAPI⁻CD8⁻MHCII⁺CD11c⁺ CD11b vs. CD103. (C) Dot plots showing pancreatic LN DC subsets from wild type and CCR7^−/−; and histogram showing CX3CR1 expression from B6.CX3CR1^GFP/+ mice, grey solid line: B6.CX3CR1^GFP/+ islet CD11b⁺DC; black solid line: B6.CX3CR1^GFP/+ pancreatic LN CD103⁺CD11b⁻DC; dashed line: B6.CX3CR1^GFP/+ pancreatic LN CD103⁺CD11b⁺DC; grey filled: B6 pancreatic LN CD103⁺CD11b⁻DC. (D) Absolute numbers and percentages of CD11b⁺CD103⁻, CD103⁺CD11b^low/− and CD11b⁺CD103⁺DC in pancreatic LN of C57BL/6 and CCR7^−/−. n=4–5, Mean ± SEM.

Steady-state antigen presentation: migratory CD103⁺DC present β cell-derived antigen while CD11b⁺DC have poor antigen presentation ability

It has been demonstrated that OT1 cells, recognizing an MHC class I-OVA peptide complex, can accumulate and proliferate in the pancreatic draining LN after adoptive transfer into RIPmOVA mice (36, 37), in which OVA is expressed on the membrane of β-cells (37), and DC are responsible for the cross-presentation of OVA (36). To recapitulate this cross-presentation ex vivo, naïve OT1 cells were co-cultured with islet cells or pancreatic LN cells. Few OT1 cells proliferated when co-cultured with islet cells either from C57BL/6 mice or from RIPmOVA mice. However, OT1 cells proliferated significantly when co-cultured with pancreatic LN cells from RIPmOVA mice, which was not observed when co-cultured with pancreatic LN cells from C57BL/6 mice (Fig. 4A), demonstrating that resident islet DC were not able to stimulate OT1 cells in islets, but some DC sampled islet-derived antigen, migrated to draining LN, cross-presented the antigen, and stimulated CD8⁺ T cells. Since CD11b⁺ DC have limited capability for migrating to LN, we reasoned that CD103⁺DC might deliver the islet derived antigen to LN. To confirm which DC subsets cross-present OVA peptide, we examined the H-2Kb-OVA_257–264 complex on the DC subsets of RIPmOVA mice by using a specific antibody that recognizes the peptide-MHC class I complex. Low levels of peptide-MHC class I complexes were detected on both of the two CD103⁺DC subsets from pancreatic LN, while the complexes were not detectable on other pancreatic LN DC subsets or on islet DC subsets (Fig. 4B and data not shown). Therefore, under steady-state, migratory CD103⁺DC were responsible for cross-presenting islet-derived peptides in the pancreatic LN and had the potential to cross-prime CD8⁺ T cells. To directly examine the ability of the migratory CD103⁺DC in stimulating CD8⁺ T cells, naïve OT1 cells were co-cultured with LN CD103⁺DC. Because the absolute number of the migratory DC in pancreatic LN was very low, we took advantage of the ability of FLT3L to expand these cells in vivo. The B16-FLT3L cell line produces FLT3L that drives in vivo expansion of FLT3L dependent DC subsets, and has no influence on the expression of costimulatory ligands or MHC class I and II on any DC subset (38). B16-FLT3L cells (1 × 10⁶) were injected i.v. into C57BL/6 and RIPmOVA mice and the DC harvested after 18 days. FLT3L-expanded LN CD103⁺DC from RIPmOVA but not from C57BL/6 mice stimulated OT1 cell to proliferate (Fig. 4C), demonstrating that migratory CD103⁺DC were capable of cross-priming CD8⁺ T cells. CD103⁺CD11b⁺DC showed greater ability for T cell stimulation than CD103⁺CD11b⁻DC (Fig. 4C).

FIGURE 4. Cross-presentation of islet-derived antigen by islet DC.

(A) CFSE labeled 1×10⁵ OT1 cells were co-cultured with 2.5×10⁵ dispersed islet cells or 2×10⁵ pancreatic LN cells from RIPmOVA or C57BL/6 for 3 days, and proliferation by CFSE dilution assayed. Cells were gated on CD8α⁺Vα2⁺. (B) Analysis of H-2Kb-OVA_257–264 complexes on DC subsets. Islet DC gated on CD45⁺MHCII⁺CD11c⁺ and CD11b vs. CD103. Pancreatic LN DC gated on CD8α⁻MHCII⁺CD11c⁺ and CD11b vs. CD103. Solid line: anti-H-2Kb-OVA_257–264; dashed line: isotype. (C) Pancreatic LN CD103+DC were isolated from C57BL/6 or RIPmOVA mice, which were injected i.v. with B16-FLT3L cells (day 18). DC gated on FSC-A vs. FSC-W, CD8α⁻MHCII⁺CD11c⁺CD103⁺ and CD11b. 6000 DC subset cells co-cultured with 5×10⁴ CFSE labeled OT1 cells, and OT1 proliferation analyzed after 5 days.

To investigate whether DC subsets present antigen to CD4⁺ T cells, native OT2 cells expressing a transgenic T cell receptor specific for OVA peptide presented by MHC class II, were labeled with CFSE and co-cultured with dispersed islet cells, pancreatic LN cells, or different DC subsets isolated from islets. Islet cells from RIPmOVA mice stimulated only about 6% of OT2 cells to proliferate, while inflamed islet cells stimulated around 16% of OT2 cells (Figs. 5A, 5B). Isolated DC subsets from uninflamed islets of RIPmOVA mice also poorly stimulated OT2 cells (Fig. S1B). Pancreatic LN cells from RIPmOVA mice did not show significant differences in stimulating OT2 cells compared to pancreatic LN cells from B6 mice (Fig. S1A). Together these data suggest that islet and pancreatic draining LN DC poorly primed CD4⁺ T cells during the steady-state. To amplify the T cell response, OT2 cells were stimulated with islet DC subsets pulsed either with OVA protein or MHC class II restricted OVA_323–339 peptide. After 5 days of co-culture, OVA protein pulsed CD103⁺DC stimulated OT2 cells, while OVA pulsed CD11b⁺DC still did not (Fig. 5C). In contrast only 500 CD11b⁺DC pulsed with OVA peptide stimulated 50% of OT2 cells to proliferate (Figs. 5C, S1B), thus the inability of CD11b⁺ DCs to stimulate CD4⁺ T cells was not a result of overall impairment in antigen presentation, but was likely due to limited ability to process antigen. Overall, islet CD11b⁺DC had poor abilities to stimulate either CD8⁺ or CD4⁺ antigen specific T cells under the steady-state condition.

FIGURE 5. Direct-presentation of islet-derived antigen by islet DC.

(A) and (B) CFSE labeled 5×10⁴ OT2 cells were co-cultured with dispersed islet cells from RIPmOVA or C57BL/6 with or without MLDS treatment (day 3 post initial STZ) for 3 days, and proliferation by CFSE dilution assayed. Cells were gated on CD4⁺Vα2⁺. (A) 5×10⁴ OT2 cells were co-cultured with 2.5×10⁵ dispersed islet cells. (B) 5×10⁴ OT2 cells were co-cultured with 10⁴–2.5×10⁵ dispersed islet cells. Duplex. Mean ± SEM. (C) CD103⁺CD11b^−/low and CD103⁻CD11b⁺DC sorted from islets of C57BL/6 mice. Islet cells were gated on FSC-A vs. FSC-W, DAPI⁻CD45⁺MHCII⁺CD11c⁺ and CD11b vs. CD103. 6000 DC subsets pulsed without or with OVA, or 500 DC pulsed with OVA_323–339 peptide co-cultured with CFSE labeled 5×10⁴ OT2 cells and OT2 proliferation analyzed after 5 days.

The local microenvironment conditions DC and influences CD4 helper cell differentiation. To explore the influence of the local microenvironment of the islets on DC function, OT2 cells were co-cultured with peptide-pulsed DC subsets. After 5 days neither islet nor pancreatic LN DC subsets induced CD4⁺ T cells to express the Treg lineage marker Foxp3 in the absence of exogenous TGFβ added to the culture (Fig. S2A). Flow cytometric analysis showed that peptide-pulsed islet CD103⁺DC and CD11b⁺DC were efficient in inducing OT2 cells to produce IFNγ (10%), but inefficient for IL-4 (<0.5%) or IL-17 (<0.5%) production, indicating that neither subset under homeostatic conditions was capable of inducing Th2 or Th17 differentiation (Fig. S2B). Overall then, the islet DC poorly processed and presented antigen to CD4⁺ T cells, and the few T cells that were induced were of the Th1 lineage.

CD103⁺DC have limited ability for antigen uptake

Tissue resident DC constitutively phagocytose apoptotic cells, and process and present peptides to T cells. To examine the ability of islet DC to sample and process islet-derived antigens, MIP-GFP mice were used. In this strain, GFP is expressed in the cytoplasm of β cells and can be taken up and processed by antigen presentation cells, so that the phagocytosis of islet-derived antigens can be monitored by assaying the levels of green fluorescence in DC (26). During homeostatic conditions, about 40% of islet CD11b⁺DC from MIP-GFP mice were GFP positive, whereas less the 10% of CD103⁺DC were GFP positive (Fig. 6A, 6B). To further assess the antigen uptake and presenting ability of islet DC during inflammatory conditions, we induced islet inflammation by injecting MIP-GFP mice with multiple low dose streptozotocin (MLDS), which induces T cell dependent insulitis (28). As shown in Fig. S3, the number and percentage of CD11b⁺DC increased with the progression of MLDS-induced inflammation. By day 3, MLDS treatment also caused slightly up-regulation of CD86 and down-regulation of MHC class II expression on CD11b⁺DC, but not on CD103⁺DC (Fig. 6D), suggesting that resident islet CD11b⁺DC acquired a more mature phenotype during islet inflammation. At this time, a marked decrease in both the percentage of GFP⁺DC and the mean GFP fluorescence intensity were also observed in the CD11b⁺ DC subset, while CD103⁺DC subset remained GFP low (Figs. 6A, 6B). This suggested that internalized GFP antigen in CD11b⁺DC had been processed and proteolytically destroyed during inflammation-induced maturation rather than recruitment of GFP⁻DC from the circulation (26), while the ingested antigen remained largely intact and unprocessed under homeostatic conditions. No GFP⁺ migratory DC (CD8α⁻CD11c⁺MHCII⁺) were found in either normal or MLDS treated draining pancreatic LN (Fig. 6C). These data suggested that the low level of GFP signal in CD103⁺DC may have been due to the limited capacity of these cells to sample islet-derived cell-bound antigens. Alternatively those CD103⁺DC that acquired tissue antigens may have migrated rapidly to the draining pancreatic LN and processed the ingested antigens during the migration, commensurate with ability of the migrated CD103⁺DC to cross-present islet-derived antigen in the pancreatic LN.

FIGURE 6. Capacity of DC in sampling antigen.

(A–D) Islet and pancreatic LN DC from MIP-GFP mice examined by flow cytometry. CTRL, untreated mice; MLDS, 3 days after initial MLDS treatment. (A) GFP fluorescence in islet CD11b⁺CD103⁻ and CD103⁺CD11b^low/−DC. Cells gated on FSC-W CD45⁺MHCII⁺CD11c⁺ and CD11b vs. CD103 (B) Percentage of GFP⁺ islet CD11b⁺CD103⁻ or CD103⁺CD11b^low/−DC. Each data point corresponds to one mouse, n=7–10. Mean ± SEM. (C) GFP fluorescence in pancreatic LN DC. Cells gated on FSC-W and CD8α^−/low. 3–4 pancreatic LNs were pooled. (D) CD86 and MHCII expression on islet CD11b⁺CD103⁻ and CD103⁺CD11b^low/−DC, 0 and 3 days after initial MLDS treatment. Solid line: MLDS treated; dashed line: untreated control. (E–G) Islet CD11b⁺CD103⁻ and CD103⁺CD11b^low/− DC gated on CD45⁺ FSC-W CD45⁺MHCII⁺CD11c⁺ and CD11b vs. CD103. (E) Islet DC uptake of OVA-alexa488 from blood. DC were examined 1.5 or 16 h after i.v. injection of OVA-alexa488. (F) and (G) Islet DC uptake OVA-alexa 488 in vitro. Islet cells were incubated OVA-alexa488 at 37 °C with or without NaN3, or on ice for 1 hour. (G) Graph showing percentage of alexa488⁺ cells in islet DC. Each data point corresponds to one experiment, n=3. Mean ± SEM.

We (Fig. 1A) and others (25) found that most islet DC were primarily located next to blood vessels and in contact with them, and thus may have the capacity to sample antigen in the circulation. To address the ability of different islet DC subsets to take up soluble antigen, OVA-alexa488 was intravenously injected and DC were examined for the presence of alexa488 signal 1.5 and 16 h later. As shown in Fig. 6E, about 25% of CD11b⁺DC and relatively fewer CD103⁺DC in both islets and pancreatic LN had green fluorescence, showing that although both DC subsets were able to sample soluble antigen from blood, CD11b⁺DC were more efficient. To rule out the possibility that this was caused by restricted access to sampling antigen from the blood by CD103⁺DC, islet single cell suspensions were incubated with OVA-alexa488 in vitro. Nearly 30% of CD11b⁺DC took up OVA, but less than 10% of CD103⁺DC did so. Uptake was an active process since it was inhibited by sodium azide and by incubation on ice (Figs. 6F, 6G). These data directly demonstrated that, as for cell-bound antigens, CD11b⁺DC were more efficient than CD103⁺DC in taking up soluble antigens.

Phenotype of islet DC during inflammation

To characterize islet DC during antigen-specific inflammation, we induced T cell mediated islet inflammation in RIPmOVA mice by the adoptive transfer of activated OT1 or OT2 cells. The phenotype of islet DC was examined 5 days after cell transfer. Both CD11c⁺MHCII⁺ DC and large numbers of CD11c⁺MHCII⁻ cells infiltrated into islets. A small percentage of these latter cells were CD11b⁺ or CD103⁺. Since these CD11c⁺ cells did not express surface MHC class II, these CD11c⁺ cells were either macrophages and/or immature DC (Figs. 7A, 7D). Similar to MLDS induced islet inflammation (Figure S3), CD4⁺ and CD8⁺ T cell-induced inflammation increased the number and percentage of the CD11b⁺DC subset (Figs. 7A, B, D and E). Under inflammation induced by either OT1 or OT2, CD11b⁺DC slightly up-regulated CD86, CD40 and CD11b, and down-regulated surface MHC class II expression (Figs. 7C, 7F). In contrast, the phenotype of CD103⁺DC was unaffected by the antigen specific CD8⁺ or CD4⁺ T cells (Figs. 7C, 7F), indicating that CD11b⁺DC but not CD103⁺DC became more mature in inflamed islets.

FIGURE 7. Identification of islet DC subsets during T cell mediated inflammation.

OT1 CD8⁺ T cells (A–C) or OT2 CD4⁺ T cells (D–F) were adoptively transferred i.v. into C57BL/6 (B6) and RIPmOVA. (A and D) Expression of CD103 and CD11b DC among DAPI⁻CD45⁺MHCII⁺CD11c⁺ islet DC. (B and E) Percentages and numbers of CD11b⁺CD103⁻ and CD103⁺CD11b^low/−DC among DAPI⁻CD45⁺MHCII⁺CD11c⁺ islet DC. Each data point corresponds to one mouse, n=5–6. Mean ± SEM. (C and F) Histograms comparing the expression of CD86, CD40, MHC II and CD11b of islet DC subsets between C57BL/6 (dashed line), and RIPmOVA (solid line) recipients. Grey filled, normal C57BL/6.

Discussion

DC can be divided into subsets according to their development, phenotype, anatomic location and function. The DC pool in islets consisted mainly of two subsets: CD103⁺DC and CD11b⁺DC. The relative percentage of CD103⁺DC was less than 20% of total islet DC, and was much lower than the CD11b⁺DC subset. We also noticed some individual variation in CD103⁺ DC, ranging from several percent to 20%. (7). Similar to origins of DC subsets in other non-lymphoid tissues (7– 10), the development of islet CD103⁺DC was dependent on Flt3 (7), while CD11b⁺DC were mainly derived from monocytes.

DC are considered to be present as semi-mature cells with high phagocytic activity within peripheral tissues under homeostatic conditions (14, 16). We showed here that islet CD11b⁺DC effectively took up islet antigen presumably derived from dead cells (26), and also sampled soluble antigens from blood. In contrast, CD103⁺DC were in a more mature stage compared to CD11b⁺DC under steady-state conditions, and expressed higher levels of co-stimulatory molecules, lower levels of MHC class II and sampled less antigen. (25). Although CD11b⁺DC more efficiently took up antigens, the internalized antigens remained largely unprocessed under homeostatic conditions. In line with this, the CD11b⁺ DC poorly activated antigen specific CD4⁺ or CD8⁺ T cells under homeostatic conditions. It has been shown that lamina propria CX3CR1⁺CD103⁻DC also have poor T cell stimulatory capacity, although these DC effectively took up OVA in vivo (14, 16). Thus, it was likely that under steady-state, extracellular antigens captured by the CD11b⁺ DC were poorly processed in the endosome/phagosome of DC and their degradation required phagosome maturation. Under islet inflammation, these resident CD11b⁺ DC underwent maturation, as well as their subcellular compartments, which allows antigens stored in the endosome/phagosome to access contents of lysosomes and to be fully processed (39–41). Although CD11b⁺DC processed most of the internalized antigens following exposure to inflammatory stimuli, their costimulatory molecule expression did not markedly increase, suggesting other signaling was also involved, such as through Toll-like-receptors, which have been shown to regulate antigen presentation (39–42). We and others showed that CD11b⁺DC also have a limited ability to migrate to the draining LN (8, 11, 16, 43) where T cells are primed. Thus, CD11b⁺DC populations may be more likely to modulate immune responses directly in the tissues by clearance of enteropathogens and/or dead cells by phagocytosis (8, 11, 16, 43), and thus may be more related to macrophages rather than DC according to their function.

In contrast, although the CD103⁺DC had a limited capacity for sampling tissue or soluble antigen under the steady-state, they were able to migrate to the pancreatic draining LN, stimulate CD4⁺ T cells and cross-present tissue antigen to CD8⁺ T cells. It remains unknown what mechanisms account for the different abilities of CD103⁺ and CD103⁻DC in antigen presentation. The capability of splenic CD8⁺DC to cross-present antigen seems dependent on the expression of specialized machinery for antigen processing (44). Peripheral CD103⁺ DC are developmentally related to CD8α⁺ DC (45). It will be interesting to investigate whether CD103⁺DC also share similar specialized machinery with splenic CD8⁺ DC. The cross-presentation of self-antigen by DC in pancreatic LN has been demonstrated to induce CD8⁺ T cell tolerance. (37, 46, 47). In vitro, we found that a few thousand LN CD103⁺ DC (DC/T cell: 1/16) from naïve RIPmOVA hardly induced CD8⁺ T cells to proliferate (data not shown). Due to limited numbers of migratory DC in pancreatic LN, we were unable to increase the DC/T cell ratio without using FLT3L to expand DC number. Flt3L expanded LN DC stimulated CD8⁺ T cells in vitro, showing that these migratory DC had potential for cross-priming. Further, total LN cells, which contained similar or even fewer numbers of DC, effectively induced CD8⁺ T cell proliferation, suggesting that signals from other LN cells promoted this cross-priming process. One study reported a subpopulation of DC in pancreatic LN, which was named merocytic DC, based on their capacity to present islet antigens to both CD8⁺ and CD4⁺ T cells. These DC were CD11c⁺CD11b^−/lowCD8α⁻PDCA-1⁻DC, and whose number was increased in NOD mice. Katz et al. showed that purified islet antigen loaded merocytic DC from diabetic NOD mice were able to break peripheral T cell tolerance to β-cells and induce rapid onset type 1 diabetes in young NOD mouse (38, 48). This suggests that merocytic DC may be a subpopulation of CD103⁺DC, since they are phenotypically and functionally similar.

Calderon et al. showed that 56% of DC from pancreatic LN of insulin promoter hen egg lysozyme (IP-HEL) and 90% of DC from membrane HEL (mHEL) mice contained peptide-MHC class II complex when examined by a specific antibody recognizing the complex (25). In these HEL transgenic mice, high concentrations of HEL are expressed in β-cells of the islet and low concentrations of HEL are found in the general circulation, comparable to those of insulin (49, 50). Thus, LN DC most likely sampled soluble HEL antigen and/or antigen transported by migratory DC, since over 50% of LN DC were HEL⁺ while only 15% of LN DC were migratory. Calderon et al. showed that dispersed islets from IP-HEL mice were slightly less effective in stimulating the T cell hybridoma 3A9 compared to inflamed islets. Since T cell hybridomas may be much more sensitive to antigen and easily activated compared to primary T cells, this could account for their observations, which are in contrast to ours where we observed that uninflamed islets very poorly presented antigen to and stimulated CD4⁺ T cells. In RIPmOVA mice, in which OVA is membrane bound, we observed that dispersed islets much less effective in stimulating primary OT2 cells, compared to inflamed islets. The limited capacity of these DC for OT2 cell stimulation was likely due to very few OVA peptide-MHC class II complexes presented during homeostasis, since ingested antigen was mostly unprocessed in this major phagocytic and non-migratory DC subset. These data are consistent with previous reports that OT2 mice expressing mOVA in islets remained diabetes free for at least 8 months (51, 52).

The tissue microenvironment shapes resident DC and regulates their functions. Unlike in skin, gut and lung, neither DC subset in islets induced Treg differentiation. In contrast, both islet DC subsets possessed the potential to induce Th1 differentiation under non-polarizing conditions, indicating that islet resident DC have less tolerogenic potential compared to mucosal DC, but instead have the propensity to induce Th1 responses that have been thought to be involved in autoimmune type 1 diabetes (53). This suggests that tolerance is maintained in islets by the poor processing and presentation ability of the CD11b⁺ DC under the steady-state.

During the progression of insulitis induced by autoreactive T cells, the balance between the activities and the relative proportions of these two islet DC changed. CD11b⁺DC became more frequent and matured. Studies have been shown that T cell infiltration only leads to a mild increase in the percentage of islet DC expressing Ki67 (26). This small increase of DC proliferation could not account for the much greater increase in the total DC numbers in the islets (26). Furthermore, the increase in islet DC after MLDS treatment depends on bone marrow function (25). Thus, it seemed most likely that the increase in islet CD11b⁺DC was mainly due to recruitment from the circulation rather than local expansion of DC. Compared to CD11b⁺DC, CD103⁺DC remained unchanged. Therefore, CD103⁺ and CD11b⁺DC had different functional responses. The balance between the activities of these subsets may be an important aspect of immune regulation in islets. Overall, the subsets differed in development, phenotype, uptake and processing of antigens and the ability to present MHC class I or II restricted antigens. These differences are important in mechanisms of how islet DC orchestrate islet immunity.

Abbreviation

DC

Dendritic cell

Flt3

fms-like tyrosine kinase 3

IP-HEL

Insulin promoter hen egg lysozyme

LP

Lamina propria

LN

Lymph node

MCSF-R

Macrophage colony-stimulating factor receptor

MLDS

Multiple low dose streptozotocin

RA

Retinoic acid