Contrasting Roles of Islet Resident Immunoregulatory Macrophages and Dendritic Cells in Experimental Autoimmune Type 1 Diabetes

The innate immune system critically shapes diabetogenic adaptive immunity during type 1 diabetes (T1D) pathogenesis. While the role of tissue-infiltrating monocyte-derived macrophages in T1D is well established, the role of their tissue-resident counterparts remains undefined. We now demonstrate that islet resident macrophages (IRMs) from non-autoimmune mice have an immunoregulatory phenotype and powerfully induce FoxP3+ Tregs in vitro. The immunoregulatory phenotype and function of IRMs is compromised by TLR4 activation in vitro. Moreover, as T1D approaches in NOD mice, the immunoregulatory phenotype of IRMs is diminished as is their relative abundance compared to immunostimulatory DCs. Our findings suggest that maintenance of IRM abundance and their immunoregulatory phenotype may constitute a novel therapeutic strategy to prevent and/or cure T1D.


Introduction
Autoimmune type 1 diabetes (T1D) arises from T-cell mediated destruction of insulin-producing β-cells. However, alterations in innate immune cells profoundly influence diabetogenic T cell immunity and T1D pathogenesis [1]. In non-obese diabetic (NOD) mice, T1D pathogenesis is associated with elevated inflammatory cytokine secretion by monocyte-derived macrophages [2] and the emergence of an inflammatory Batf3+ intra-islet dendritic cell (DC) subset [3]. In humans, monocytes and myeloid DCs from individuals with T1D have altered TLR responses that give rise to heightened NF-κB activation and pro-inflammatory cytokine secretion [4].
Circulating monocyte-derived, tissue-infiltrating macrophages play a pro-inflammatory role in T1D; however, the role of their tissue-resident counterparts remains undefined. Tissue-resident macrophages (TRMs) are an Ly6C-lineage that arises from yolk-sac derived progenitors [5,6], distinguishing them from Ly6C+ monocyte-derived macrophages that are replenished by bone marrow-derived circulating progenitors [7]. Recently, we identified a CD169+TIM-4+ TRM subset in heart and skin that migrates upon activation to draining lymph nodes, produces TGFβ, induces FoxP3+ Tregs, and promotes allograft survival [8]. This TRM population joins kindred immunoregulatory Ly6C-subsets of bronchoalveolar TRMs that suppress asthmatic lung inflammation [9] and adipose TRMs that control local inflammation and insulin resistance [10]. Thus, many TRM subsets are immunoregulatory in the steady state; however, the influence of inflammation on the immunoregulatory phenotype and function of TRMs has not been intensively studied.
A CD169+ (MOMA-1+) population has been identified in pre-diabetic NOD mice by immunohistochemistry [11], suggesting that a population of TRMs might be present in the pancreatic islets of T1D-prone mice. However, whether these cells: i) are TRMs, ii) possess immunoregulatory phenotype and function in the steady state, or iii) exhibit phenotypic or functional plasticity, remains untested. To this end, we have identified a CD169+ islet-dwelling TRM subset and analyzed its immunoregulatory phenotype and function in quiescence, inflammation and autoimmune T1D.

Materials & Methods Mice
C57BL/6J (B6), female NOD/LtJ (NOD) and female NOR/LtJ (NOR) mice were obtained from the Jackson Laboratories (Bar Harbor, ME). Founder bm12 (B6(C)-H2-Ab1 bm12 /HkEgJ), B6. g7 (B6.NOD D17Mit21-D17Mit10 /LtJ), and ABM TCR [12] transgenic mice bearing the FoxP3/GFP reporter mice were maintained in our colony. Founder BDC2.5 TCR transgenic mice on the B6.g7 background [13] were obtained from Drs. Mathis and Benoist (Harvard Medical School, Boston, MA) and maintained as heterozygotes in our colony by paired mating with wild type B6.g7 mice. Glucose tolerance was defined as AUC 100mg/dl <6,000 and blood glucose <150mg/ dl at 120 minutes after i.p. injection of glucose (3g/kg). All mice were certified to be specific pathogen-free conditions. All animal experiments were performed under a protocol reviewed and approved by the Institutional Animal Care and Use Committee at Beth Israel Deaconess Medical Center (Boston, MA) and in accordance with the U.S. Department of Health and Human Services' Guide for the Care and Use of Laboratory Animals.

Immunofluorescence microscopy
Pancreases were snap-frozen; stained with rat anti-mouse CD169, goat anti-mouse TIM-4 (R&D Systems, Minneapolis, MN) and a triple antibody cocktail containing rabbit anti-human glucagon, somatostatin, and pancreatic polypeptide. Images were acquired on an LSM 510 Meta using a plan-apochromat 1.3 numerical aperture 40x objective with oil (Carl Zeiss, Thornwood, NY).

Isolation of islet resident leukocytes
Pancreatic islets were purified as described [14], gravity washed, and incubated at 37°C in HBSS containing 0.25% trypsin, 2mM EDTA, and 0.1mg/ml DNAse with agitation (120 RPM) for 10 minutes. Digestion was stopped by adding FBS to a 20% final concentration and placing samples on ice. Single cell suspensions were then stained with the indicated viability dye and antibodies. Absolute numbers were determined using CountBright absolute counting beads (Life Technologies, Carlsbad, CA).

Quantitation of IL-6 in supernatants
IL-6 was measured in the supernatants of MLR cultures using a cytokine bead array. The concentration of the experimental samples was interpolated using a 4-parameter logistic non-linear regression model.

Real-time PCR
Cells were FACS-sorted or resuspended into cell lysis buffer. cDNA was synthesized from column-purified RNA and pre-amplified for 10 cycles using the PreAmp (Life Technologies) target-specific pre-amplification system. Real-time PCR was performed using TaqMan probe/ primer sets from Life Technologies (Table 1). Data were normalized to GAPDH using the ΔCT method.

Statistics
Statistical analyses were performed with Prism software (GraphPad, La Jolla, CA). Means of two groups were compared by an unpaired t test. Means of three or more groups were compared using a one-way ANOVA with Bonferroni's post-test. All error bars are +/-S.D. Values of p < 0.05 were deemed statistically significant.

Islet resident macrophages powerfully induce FoxP3+ Tregs in vitro but their immunoregulatory function is compromised by TLR4 stimulation
To test whether IRMs from insulitis and T1D-free mice are functionally immunoregulatory, we performed mixed lymphocyte reactions (MLRs) with IRM, IRDC or splenic DC stimulator cells from bm12 mice (I-A bm12 ) and FoxP3/GFP-CD4+ T cells from ABM TCR transgenic mice. ABM mice express a clonotypic anti-I-A bm12 transgenic TCR and a FoxP3/GFP reporter [12]. After 4 days in the MLR, CD4+ ABM T cells stimulated by bm12 IRM, IRDC or splenic DC stimulator cells were collected and analyzed for proliferation by CellTrace Violet dilution and Treg conversion by de novo FoxP3/GFP reporter expression. At the 1:20 stimulator:T cell ratio used in our experiments, we observed inter-experimental variability in the frequency of non-activated T cells that failed to upregulate CD25. As proliferation and Treg conversion cannot be determined for this non-activated CD25-population, which likely failed to encounter antigen, we performed proliferation and Treg conversion analyses using an activated (CD25+) CD4+ TCR transgenic T cell gate. Whereas IRMs induce less proliferation as compared to IRDCs or splenic DCs (Fig 3A), IRMs induce a much higher frequency of FoxP3/GFP expression (Fig 3B and 3C; p<0.001). Strikingly, stimulation of the MLR by LPS, a TLR4 agonist, reduces the frequency of FoxP3/GFP+ T cells (20.9% for untreated IRMs vs. 9.3% for LPS-stimulated IRMs, p<0.01; Fig 3C). This decrease in FoxP3+ Treg frequency correlates with an increase in the Treg inhibitory cytokine IL-6 in the supernatant (Fig 3D; p<0.001).
IRMs from T1D NOD mice undergo phenotypic plasticity and express less immunoregulatory CD39 as compared to IRMs from NOR mice We next tested the hypothesis that IRMs, a cell population whose phenotype is altered by TLR4 stimulation, might manifest changes in phenotype associated with age that herald T1D onset. Interestingly, expression of CD39 is reduced on IRMs in diabetic (p<0.01) and non-diabetic NOD (p<0.05) mice as compared to control NOR mice at 15 weeks of age (Fig 6). In contrast, expression of CD39 is similar between NOD and NOR mice at earlier time-points ( The balance of islet resident APCs tilts from islet resident macrophages to islet resident DCs in T1D NOD relative to control mice We next tested the hypothesis that the balance of intra-islet IRMs to DCs is altered during T1D pathogenesis. To test this, we enumerated intra-islet APC populations in 5-, 9-, and 15-weekold NOD mice and age-matched NOR control mice. NOD mice exhibit mild peri-islet insulitis at 5 weeks, moderate peri-islet insulitis at 9 weeks, and invasive insulitis at 15 weeks. Fifteenweek-old NOD mice were stratified into non-diabetic, diabetic, and glucose intolerant cohorts based on the results of an i.p. glucose tolerance test (IPGTT) one day before analysis.
The ratio of immunoinflammatory IRDCs to IRMs is virtually identical in young 5-week-old NOD and NOR mice (Fig 7A, p = N.S.). However, the ratio of IRDCs to IRMs is dramatically increased in 9-week-old (p<0.01) and 15-week-old (p<0.001) NOD mice relative to age-matched NOR mice. In order to determine whether the increased frequency of IRDCs (CD45+Ly6C-CD11c+CD16/32-F4/80-) relative to IRMs (CD45+Ly6C-CD11c+CD16/32+F4/80+) is due to a decrease in IRMs and/or an increase in IRDCs, we quantified the absolute number of IRMs and IRDCs per islet in each group. At 9 and 15 weeks of age, IRMs, IRDCs, and tissue-infiltrating (CD11b+Ly6C hi ) macrophages are all increased in absolute abundance in NOD as compared to NOR mice (Fig 7B). In contrast, all three populations are indistinguishable in absolute number between young 5-week-old NOD and NOR mice (Fig 7B).

Discussion
Immunoregulatory TRM subsets are present in skin, heart [8], lung [9], and adipose tissue [10]. We now report the characterization of an IRM population with immunoregulatory phenotype and function. As T1D approaches, the immunoregulatory phenotype of IRMs is diminished as is their relative abundance as compared to immunoinflammatory DCs. Both of these changes are likely relevant to the onset of overt T1D.
Relative to IRDCs, IRMs more abundantly express immunoregulatory CD39, CD73, and galectin-9. The role of other T cell checkpoints, including PD1/PDL1 [25] and ICOSL [26], in IRM biology is not certain and warrants further study as these pathways play an important role in regulating autoimmune T1D pathogenesis. Compared to DCs, IRMs also more powerfully induce FoxP3+ Tregs in the MLR. However, TLR4 activation reduces FoxP3+ Treg induction by IRMs, possibly related to the downregulation of immunoregulatory CD39 transcripts and the upregulation of pro-inflammatory CCL2, IL-6, and IL-1β transcripts. Corroborating our in vitro findings, there is an age-related loss of CD39 expression by IRMs in 15-week-old NOD mice relative to diabetes-resistant controls. However, IRMs from NOD, NOR, and B6.g7 mice are similar in their ability to induce FoxP3+ Tregs in vitro, suggesting that IRMs from NOD mice are not inherently defective in their ability to induce Tregs.
It is still possible that IRM plasticity compromises other immunoregulatory functions such as attenuating local inflammation [9,10]. Loss of CD39 function, and the catabolism of proinflammatory ATP into anti-inflammatory adenosine as a consequence, is associated with other T cell dependent autoimmune diseases [27]. TRMs from type 2 diabetic mice undergo proinflammatory phenotypic plasticity [28]. Thus, IRM plasticity provides new potential mechanisms by which TLR activation may incite T1D [29,30] and warrants closer examination.
The expansion index for the gated live activated CD4+Vβ4+CD25+ population was determined using FlowJo analysis software and is noted in the upper-left corner. (B) FoxP3 expression is shown for the gated live activated CD4+Vβ4+CD25+ population. (C) Percent of gated live CD4+Vβ4+CD25+ transgenic T cells that acquire FoxP3 expression after culture with IRMs from the indicated strain are shown from independent experiments (n = 3-4 per strain). A One Way ANOVA indicated no significant difference among groups with (p = 0.95) or without peptide (p = 0.64).
doi:10.1371/journal.pone.0150792.g005 The MFI of CD39 expression on IRMs from NOD and NOR mice (n = 5 per group pooled from 2 independent experiments) was analyzed by flow cytometry using the CD45+Ly6C-CD11c+F4/80+CD16/32+ gate. Since each time-point was acquired separately, inter-graph comparison of MFIs is not possible. NOD groups were compared to the NOR control group using an unpaired t test (*, p<0.05, **, p<0.01). The absolute number of IRMs increases but its proportion relative to inflammatory IRDCs is diminished in non-diabetic and diabetic NOD mice. This suggests that a shift in the balance of intra-islet APCs from one that favors IRMs to one that favors inflammatory IRDCs heralds T1D. As the NOD and NOR differ at defined loci on chromosomes 2, 4 11, and 12, it seems probable that one or more of these genetic intervals drives changes in the balance of intra-islet APCs and, as a result, disease outcome.
doi:10.1371/journal.pone.0150792.g007 in the spleen [35], the impact of TRMs on immunoregulatory B cell development remains uncertain and warrants further study.
We now demonstrate the existence of a functionally immunoregulatory population of islet resident macrophages in non-autoimmune and T1D NOD mice. As T1D approaches, the immunoregulatory phenotype of IRMs is diminished as is their relative abundance as compared to immunoinflammatory DCs. These results suggest that the maintenance of IRM abundance and their immunoregulatory phenotype constitutes a potential therapeutic strategy to prevent and/or cure T1D.