E2-2 Dependent Plasmacytoid Dendritic Cells Control Autoimmune Diabetes

Autoimmune diabetes is a consequence of immune-cell infiltration and destruction of pancreatic β-cells in the islets of Langerhans. We analyzed the cellular composition of the insulitic lesions in the autoimmune-prone non-obese diabetic (NOD) mouse and observed a peak in recruitment of plasmacytoid dendritic cells (pDCs) to NOD islets around 8–9 weeks of age. This peak coincides with increased spontaneous expression of type-1-IFN response genes and CpG1585 induced production of IFN-α from NOD islets. The transcription factor E2-2 is specifically required for the maturation of pDCs, and we show that knocking out E2-2 conditionally in CD11c+ cells leads to a reduced recruitment of pDCs to pancreatic islets and reduced CpG1585 induced production of IFN-α during insulitis. As a consequence, insulitis has a less aggressive expression profile of the Th1 cytokine IFN-γ and a markedly reduced diabetes incidence. Collectively, these observations demonstrate a disease-promoting role of E2-2 dependent pDCs in the pancreas during autoimmune diabetes in the NOD mouse.


Introduction
Type-1-diabetes (T1D) is an immune-mediated disease caused by insufficient insulin production from the pancreas. T1D as well as the spontaneous autoimmune diabetes in the non-obesediabetic (NOD) mouse is characterized by immune cell infiltration of the pancreatic islets of Langerhans and a subsequent T-cell mediated destruction of the insulin-producing β-cells. In NOD mice, the inflammation that precedes overt diabetes is characterized by the presence of myeloid cells such as macrophages as well as dendritic cells (DCs) followed by recruitment of immune cells of lymphoid origin such as T-and B-cells [1][2][3][4][5]. Recently, several publications have suggested distinct and important roles for plasmacytoid DCs (pDCs) as well as type-1-IFN (IFN-I) signaling in initiation and progression of human T1D [6,7] and diabetes of the NOD mouse model [8][9][10]. In contrast, pDCs have also been reported to play a regulatory role in T1D [11][12][13] and during progressive insulitis in animal models of diabetes [14][15][16][17][18]. This dual role of pDCs in autoimmune diabetes may be explained by the diverging abilities of activated pDCs to either stimulate or inhibit immune reactions by presenting antigen and producing IFN-I or by producing tolerogenic enzymes and cytokines, respectively (reviewed in [19,20]).
In this study we have performed a detailed analysis of the cellular composition of infiltrating immune cells during progression of autoimmune diabetes. We describe that pDCs display unique kinetics of recruitment into the islets of Langerhans suggesting that this cell type plays a role in the pathogenesis. Analysis of conditional E2-2 knockout NOD mice which are defective in maturation of pDCs support this notion since pDC-deficient NOD mice display a significantly reduced expression profile of the Th1 cytokine IFN-γ during advanced insulitis and consequently a reduction in diabetes incidence.

IFN-α-secreting pDCs peak in the pancreatic islets of NOD mice at 8-9 weeks
We isolated leukocytes from the islets of both NOD and control B6 mice at different ages and analyzed them using flow cytometry. The accumulation of recruited CD45 + leukocytes appeared in the NOD islets between 4-6 weeks of age ( Fig 1A). From this time point we observed a gradual increase in CD45 + cells including T-cells (CD4 + and CD8 + ), B-cells, DCs, macrophages and NK-cells peaking at 12-14 weeks of age in NOD islets (Fig 1A and S1A Fig). In B6 islets no such accumulation was detected ( Fig 1A and S1A Fig). The predominating cell types were of T-or Bcell origin making up more than 70% of the total CD45 + cells from 6 weeks of age. The decrease in CD45 + cells observed in islets at older ages (>23 weeks of age) was most likely due to progressive β-cell destruction resulting in reduced immune cell recruitment [21]. Together, this data concurs with previous reports analyzing NOD pancreases [1,2,22].
Remarkably, pDCs accumulated with a distinct peak around 8-9 weeks and then disappeared from the islets at later time points (Fig 1A and 1B). The accumulation of pDCs is accompanied by an increased IFN-I signaling confirmed by the increased expression of IRF7 (interferon response factor 7) and ISG15 (interferon stimulated gene 15) after 8-9 weeks compared to 3 week expression levels ( Fig 1C) as well as increased production of IFN-α after CpG 1585 stimulation ( Fig 1D). Analysis of the IFN-I response genes in islets from B6 mice shows equal or even slightly reduced expression of both IRF7 and ISG15 from 3 weeks to >8 weeks ( Fig 1C). Collectively these findings support the notion that pDCs and IFN-I signaling increase in islets of NOD mice after 8 weeks of age.
doi:10.1371/journal.pone.0144090.g001 mice and to fl/fl cre − littermate controls (Fig 2A and 2B). Impaired pDC development was consistent at all ages in spleen and peripheral LNs (S2A and S2B Fig). Other lymphoid and myeloid subsets including T-cells, B-cells, DCs, granulocytes and macrophages appeared to be unaffected (Fig 2A). In bone marrow, pDC levels are reduced to 60% of levels in NODwt in both heterozygous +/fl cre + and homozygous fl/fl cre + (S2A and S2E Fig). In the spleen, the effect of homozygous floxing of the E2-2 locus is more pronounced as the pDC frequency is reduced to 35% and 50% of levels in NODwt for fl/fl cre + and +/fl cre + , respectively (S2A and S2E Fig). Both observations are consistent with our initial findings [23]. Because of inefficient recombination in precursors of pDCs, which only express CD11c intermediately, some pDCs will remain in the +/fl cre + and fl/fl cre + mice [26].

Impaired recruitment of pDCs to pancreatic islets alters cytokine profile of insulitis and prevents diabetes development
We next investigated if the reduction of pDCs in the NOD.E2-2 CKO mice would affect the cellular composition of insulitis as well as the cytokine profile of the infiltrated pancreatic islets. We could not detect any effect on overall insulitis score analyzed at 8-14 weeks of age (S2C Fig) or the total number of infiltrating CD45 + leukocytes ( Fig 3A) when comparing fl/fl cre − to fl/fl cre + . The percentage of pDCs recruited to islets was reduced at all ages in fl/fl cre + mice compared to fl/fl cre − (Fig 3B).
Correspondingly, the number of pDCs within pancreatic islets was clearly reduced in the fl/fl cre + mice older than 8 weeks of age but not at 3 weeks of age (Fig 3A), likely reflecting the low numbers of pDCs present at the early time point. In contrast, we did not observe significant differences in the number of islet-infiltrating DCs, T-or B-cells, regulatory T-cells or in the CD4/ CD8 ratio of islet infiltrating T-cells at any age ( Fig 3A and S2D Fig). The cellular composition of pancreatic lymph nodes (PaLN) also remained unchanged except for the markedly reduced pDC population in the fl/fl cre + mice at all ages (S2A Fig) and a slight increase in DCs most likely resulting from an effect on the common precursor of pDCs and DCs [26].
As expected, analysis of the inflammatory cytokine profile from fl/fl cre + mice and littermate fl/fl cre − controls revealed that CpG 1585 induced IFN-α secretion was markedly reduced from islets isolated from the fl/fl cre + mice (Fig 3C). A similar tendency was also observed in PaLN node cells after stimulation and in spontaneous IFN-α production from pancreatic islets ( Fig 3C). Autoimmune diabetes of the NOD mouse is a Th1 mediated disease and IFN-γ is the main pro-inflammatory cytokine. We observed a marked reduction in the expression of IFN-γ from infiltrated pancreatic islets of fl/fl cre + mice at 14 and >22 weeks of age ( Fig 3D). This reduction suggests a less aggressive Th1 phenotype of the infiltrating T-cells in the scenario where pDCs are not present in the islets.
Finally, we analyzed the potential effect of the reduction of pDCs and IFN-γ expression on diabetes development. The extend of pDC reduction may control the progression to overt diabetes and therefore the diabetes incidence was determined in both homozygous fl/fl cre + and heterozygous +/fl cre + mice and littermate control cre − mice. As illustrated in Fig 3E, a significant drop in diabetes incidence from >75% for control to <10% for fl/fl cre + mice was observed. The effect on diabetes incidence was also observed in +/fl cre + mice but was less pronounced in accordance with the lower reduction in peripheral pDCs in +/fl mice compared to fl/fl cre + .

Discussion
Several studies have investigated the role of pDCs in the development of autoimmune diabetes and both disease promoting and inhibiting roles of this cell subset have been suggested. Diana et al. recently proposed a model in which pDCs via IFN-α interplay with neutrophils and natural antibodies to promote disease early in the development of murine T1D [8]. In that study, a peak in recruitment of pDCs to pancreatic islets is reported at 4 weeks. The significant discrepancy between these finding and the findings presented in this study could be caused by difference in the ages of mice used for analysis. Thus, while we here describe a peak in accumulation of pDCs at 8-9 weeks of age, Diana et al. do not analyze islets from mice older than 6 weeks. In our hands, pDCs and neutrophils only constitute small fractions of the total islet leukocytes at 3-4 weeks of age (Fig 1A, 1B and S1A Fig). Given the small number of CD45 + cells isolated from islets at this early stage we could, therefore, not confirm the early recruitment of large numbers of pDCs or neutrophils reported by Diana et al. at the same age [8]. One explanation for this discrepancy could be technical differences, such as the method of islet isolation. In the current study we handpicked islets directly after collagenase digestion minimizing the risk of contaminating islet derived leukocytes with myeloid cells from the exocrine pancreas. Alternatively, the observed discrepancy in kinetics of pDC recruitment to islets could be a result of differences between the NOD colonies. However, we find this unlikely since multiple studies performed in independent laboratories have failed to identify notable accumulation of inflammatory leukocytes in pancreatic islets of young NOD mice (analyzed in e.g. [1,4]).
Despite this apparent absence of pDC accumulation in the pancreatic islets at 3-4 weeks, we did observe expression of interferon response genes already at this time point (Fig 1D). These observations are in line with other recent studies, which identify an early IFN-I transcriptional signature in NOD islets [4,27], but suggest that the cellular source of IFN-α is likely to be islet resident cells at this time point.
The peak in pDC recruitment to NOD islets at 8-9 weeks of age (Fig 1A and 1B) coincide with increased CpG 1585 induced secretion of IFN-α ( Fig 1D) and increased spontaneous expression of the interferon response genes IRF7 and ISG15 (Fig 1C). The study by Diana et al. reports a peak in islet IFN-α specifically at 3 weeks of age [8]. This discrepancy could be caused by a difference in the ages of mice used for analysis, since analysis of IFN-α by Diana et al. covers islets of NOD mice at 2 to 6 weeks of age, while the present study covers IFN-α production from islets at 4 weeks and > 8 weeks of age. In addition, the 3 week peak in IFN-α observed by Diana et al. is not overlapping with the observed peak recruitment of pDCs at 4 weeks [8], suggesting that the cellular contributors to IFN-α at 3-4 weeks of age could include additional cell types.
The contribution of pDCs in T1D pathogenesis remains highly controversial and both tolerogenic and inducing roles have been linked this cell type. The present study supports an inducing role of pDCs in diabetes pathogenesis by linking reduced IFN-γ expression and lower diabetes incidence (Fig 3D and 3E) to a genetic defect in pDC maturation. This link is supported by studies of NOD mice and material from human T1D patients. Analysis of leukocytes from human T1D patients shows that mature pDCs can drive the differentiation of CD4 + T cells towards Th1 phenotype with increased production of IFN-γ in vitro in a process which is likely dependent on IFN-I [6]. Additional studies show that pDCs in vitro enhance auto-antigen presentation in the presence of auto-antibodies in a more efficient manner than other DC and antigen presenting cell subsets from peripheral blood of T1D patients [7]. In NOD mice, depletion of pDCs cause reduced potential of pancreatic CD8 + Tcells to produce IFN-γ in response to re-stimulation with IGRP 206-214 peptide [8]. Collectively, these studies show a pDC sensitive Th1-polarization of T-cell responses causing destruction of insulin producing β-cells. Thereby pDCs control the progression from insulitis to overt diabetes. Despite the observed preventive effect on diabetes incidence, the pDC deficiency in the NOD.E2-2 CKO mice did not result in any detectable alterations in insulitis score analyzed at 8-14 weeks (S2C Fig). It is likely that analysis at time points of more advanced insulitis will show detectable differences in insulitis and β-cell mass and these studies are ongoing.
The analysis of the regulatory role of pDCs in progressive NOD insulitis demonstrates that pDCs also can provide inhibitory signals by secreting IDO [16,17]. In the wild type NOD, the production of IDO at progressive insulitis is, however, not sufficient to control the ongoing destruction of β-cells and prevent diabetes. Moreover, a recent finding suggests that IDO expression and catalytic function is defective in pDCs from NOD mice [9]. Another set of evidence supporting a protective and regulatory role of pDCs has been obtained by analyzing induced models of diabetes [14,15,18]. The mechanistic evidence from these studies suggests that pDCs via TGF-β production, CXCR3 up-regulation and interplay with NKT cells and Tregs act in several ways to dampen tissue destruction.
Based on the SNP analysis of Chr 18 we cannot rule out that the alleles of Idd21.1 and Idd21.2, [28,29] located in the same chromosomal region as the E2-2, could be of non-NOD origin. However, since we use littermate controls in this study we can rule out that the differences observed between fl/fl cre − control and fl/fl cre + results from a carryover of resistance conferred by non-NOD alleles in this region of chromosome 18 or elsewhere in the genome. In addition, the integration of the Cre-recombinase transgene has previously been reported not to affect the incidence or insulitis progression after backcrossing to NOD [30].
In conclusion, this study complements previous conflicting findings of reduced diabetes incidence upon pDC depletion using antibodies [8,10] or by blocking IFN-I signaling [31] but no effect on diabetes incidence or insulitis in NOD.IFNAR -/- [27]. However, it is still unclear whether the disease-promoting role of pDCs observed here is a consequence of cytokine mediated or antigen presentation mediated activation of diabetogenic T-cells. Future studies including analysis of the particular subset of cells responding to pDC stimulation are necessary to fully understand the molecular mechanisms by which pDCs activate and enhance the diabetogenic response in NOD islets.

Ethics statement
Animal experiments were performed in strict accordance with the recommendations for the use of laboratory animals from the Swedish board of agriculture. The ethics committees of Umeå University and Lund University approved all animal experimental procedures (Permit number: A120-12 and M130-13). All efforts were made to minimize suffering.

Mice
C57BL/6 (B6) and NOD mice were purchased from Taconic (Ejby, Denmark). B6.E2-2 fl/fl -CD11c.cre [23] were backcrossed to NOD genetic background for >10 generations to generate NOD.E2-2 fl/fl -CD11c.cre. SNP-based scans were used to ensure that all chromosomes, except chromosome 18 carrying the E2-2 locus were of NOD origin. By in depth SNP analysis of chromosome 18 we have limited the region flanking the E2-2 locus to <19 cM corresponding to <24 Mb (S1 Table). All animals were bred and maintained in a specific pathogen-free environment at the animal facilities at Umeå University or Lund University and all strains were maintained by continuous inbreeding. Mice were sacrificed by cervical dislocation. Only female mice were used for experiments.

Diabetes and insulitis scoring
Female mice were monitored for diabetes incidence until 30 weeks of age by weekly measurements of urine glucose levels. Mice were considered diabetic when urine glucose was >13 mmol/l for two consecutive weeks. Preparation of pancreas sections and staining were performed as described previously [32]. Insulitis was scored in >50 islets per mouse as no infiltration (score 0), peri-insulitis (score 1), <50% of islet infiltrated (score 2) or >50% of islet infiltrated (score 3).

Real-time PCR analysis
Total RNA was extracted and DNase-treated using RNeasy micro kit (Qiagen, Venlo, Netherlands). cDNA was generated using High Capacity kit (Applied Biosystems, Waltham, MA, USA). qPCR was performed using MyIQ machine (BioRad, Hercules, CA, USA) and Taqman primer/probe sets recognizing IFN-γ and interferon response genes IRF7 and ISG15 (Applied Biosystems). Expression levels are shown relative to expression of GAPDH and β-actin (Applied Biosystems).

Supporting Information
(TIF) S1 Table. Microsatellite validation of B6.E22flox-CD11c.cre backcross to NOD background. Control (B6) and 2 NOD.E22flox-CD11c.cre mice at generation N5 were assayed using 82 probes to the indicated SNPs. These mice were further backcrossed for 2 generations to NOD background followed by inbreeding. Yellow indicate B6 loci and orange indicates NOD loci. Numbers in the box are the size of the expected PCR fragments. (TIF)