Induction of Indoleamine 2,3-Dioxygenase by Interferon-γ in Human Islets

Human islets were prepared by collagenase digestion by the Islet Cell Resource Center at the University of Colorado at Denver and Health Sciences Center using the Edmonton protocol (cold ischemia time between 4 and 9 h). All donors were brain dead, heart-beating individuals from the state of Colorado who died in motor vehicle accidents. None had a previous history of diabetes or inflammatory diseases. Islet purity (75–80%) and viability (76–96%) were determined by dithizone and Syto13/ethidium bromide staining according to the Standard Operation Procedure Manual, Clinical Islet Laboratory, SMRI, (Edmonton Alberta, Canada). The islets were precultured for 12 to 24 h in Miami medium (CMRL 1066 supplemented medium 99-603-CV; Mediatech, Herndon, VA) containing 0.5% human serum albumin and then further cultured in the presence and absence of cytokines, namely 10 μg/ml interleukin (IL)-1β (500 units/ml), 25 μg/ml IFN-γ (500 units/ml), 25 μg/ml tumor necrosis factor-α (TNF-α) (2,500 units/ml), and a cocktail mixture of 2 μg/ml IL-1β (100 units/ml), 10 μg/ml IFN-γ (200 units/ml), and 10 μg/ml TNF-α (1,000 units/ml) (Roche Applied Science, Indianapolis, IN) in 5% CO₂ and 37°C for 24 h. Islet preparations (2,000 islet equivalents) from two of the donors when transplanted under the kidney capsule of streptozotocin-induced diabetic Rag2−/− B6 mice normalized hyperglycemia, indicating long-term survival and functionality.

Total RNA was extracted from groups of 8,000–10,000 human islets from four unrelated donors (two men and two women, age 26–46.6 years) by TRIzol reagent (Invitrogen, Carlsbad, CA) and purified by RNeasy columns (Qiagen, Valencia, CA), and RNA quality was verified by capillary electrrophoresis (Agilent-2100 Bioanalyzer; Agilent, Palo Alto, CA). Biotin-labeled cRNA was synthesized from total RNA (6 μg) according to standard Affymetrix protocol (Affymetrix, Santa Clara, CA), and 15 μg was hybridized to human genome HG U133 Plus 2.0 chip. The basic experiment included untreated islet tissue (n = 4) and exposure to IL-1β (n = 3), TNF-α (n = 4), IFN-γ (n = 3), and cytokine cocktail (n = 3). Data were analyzed using the GeneSpring Software (Silicon Genetics, Palo Alto, CA). The dataset was normalized as follows: 1) Scanner intensity values below 0.01 were set to 0.01. 2) Each measurement was divided by the 50th percentile of all measurements in that particular sample. 3) Specific samples (cytokine treated; n = 13) were normalized to the median of the corresponding control sample (n = 4). Cross gene error model was based on replicates, and the average of base/proportion ration was calculated as 68.15. For statistical analysis, the data were categorized on the basis of the treatment parameters, and a parametric test was performed that assumes equal variances. A one-way ANOVA for multiple groups was chosen in Gene Spring software to initially filter the data. A multiple correction test in the form of Bonferroni Step Down (Holm) was chosen because, during testing a set of genes for statistical significance across various groups, some of the genes may be falsely considered as statistically significant. The purpose of a multiple testing correction was to keep the overall error rate/false positives to less than the P value cutoff (0.01) (20). This restriction tested 54,675 genes and estimated that ∼0.01 genes could be expected to pass the restriction by chance. Tukey’s test was used for post hoc analysis.

cDNA was prepared from total RNA (1 μg) using iScript cDNA synthesis kit (Bio-Rad Laboratories, Hercules, CA). Samples equivalent to 0.1 μg of the original RNA sample were used as templates for amplification in the linear phase in a 5′ nuclease assay–based system using FAM dye–labeled Taqman MGB probes (Applied Biosystems, Foster City, CA) on a 96-well ABI 7000 PCR system instrument. Hypoxanthine phosphoribosyl transferase (HPRT) gene was selected for sample normalization based on preliminary experiments with the ABI control plate (part no. 430 9199). The cycle threshold values (C_T) were measured in triplicate, and the samples were re-normalized to the C_T values from islet samples without cytokine exposure (control group/calibrator), and data were quantified using 2^−ΔΔC_T method (ABI User Bulletin #2).

All chemicals were purchased from Sigma-Aldrich (St. Louis, MO) unless specified otherwise. Groups of 1,000 human islets were cultured for 24 h in 1 ml medium with cytokines, recovered by centrifugation for 5 min at 800 × g and sonicated in 150 μl PBS containing a protease inhibitor cocktail (Set 2; Calbiochem, EMD Biosciences, San Diego, CA). The sonicate was centrifuged for 10 min at 10,000 × g, and the supernatant was assayed in triplicate by incubating a 40-μl sample with an equal volume of 100 mmol/l potassium phosphate buffer, pH 6.5, containing 40 mmol/l ascorbic acid (neutralized to pH 7.0), 100 μmol/l methylene blue, 200 μg/ml catalase, and 400 μmol/l l-Trp for 30 min at 37°C. The assay was terminated by the addition of 16 μl 30% (w/v) trichloroacetic acid (TCA) and further incubated at 60°C for 15 min to hydrolyze N-formylkynurenine to KYN. The mixture was then centrifuged at 12,000 rpm for 15 min, and KYN was quantified by mixing equal volume of supernatant with 2% (w/v) Ehrlich’s reagent in glacial acetic acid in 96-well micro-titer plate and reading the absorbance at 480 nm using l-KYN as standard. Protein in the islet samples was quantified by Bio-Rad Protein assay at 595 nm. For the detection of l-KYN in the islet culture supernatants, proteins were precipitated with 5% (w/v) TCA and centrifuged at 12,000 rpm for 15 min, and determination of KYN in the supernatant with Ehrlich’s reagent was as described above. IL-4 (10 μg/ml; 500–2,000 units/ml) and 1-α-methyl Trp (1-MT; 40 μmol/l) were added to the incubation media as indicated.

Groups of 1,000–1,200 islets incubated for 24 h in Miami medium in the presence of cytokines were harvested and sonicated in PBS as above, and 50-μg protein samples were electrophoresed on 10% SDS–PAGE gels. COS7 cells (0.6 × 10⁶ cells/60-mm³ petri dish) transfected with human-IDO plasmid (3 μg) or empty vector cells were used as positive and negative controls, respectively. Proteins were transferred electrophoretically onto polyvinylidine fluoride membranes by semidry method and blocked for 1 h with 5% (w/v) nonfat dry milk in Tris-buffered saline and 0.1% Tween and then incubated overnight with anti-human mouse IDO antibody (1:500; Chemicon, Temecula, CA), phospho-STAT_1α p91, and STAT_1α p91 (1:500; Zymed, San Francisco, CA). Immunoreactive proteins were visualized with ECL Plus Western blotting detection reagent (Amersham BioSciences, Buckinghamshire, U.K.) after incubation for 1 h with anti-mouse horseradish peroxidase–conjugated secondary antibody (Jackson Immunolabs, West Grove, PA).

Islets were fixed in 4% paraformaldehyde in PBS (Invitrogen) for 1 h, immobilized in molten 10% porcine skin gelatin blocks (37°C), and embedded in optimal cutting temperature compound. Immunofluorescent staining on islet tissue was performed on 7-μm sections that were stained with antibodies raised against pancreatic duodenal homeobox 1 (PDX1) and IDO. Antigen retrieval was performed in a water bath for 30 min in a buffer containing 10 mmol/l Tris and 1 mmol/l EDTA (pH 9.0) at 97°C. The sections blocked for 1 h with 5% normal goat serum in PBS. The tissues were then reacted with mouse monoclonal anti-human IDO antibody (1:20; Chemicon) and goat polyclonal anti-human PDX1 antibody (1:2,000; gift from Chris Wright, School of Medicine, Vanderbilt, TN) overnight at room temperature in a humid chamber. Secondary antibodies anti-goat (labeled with Cy3) and anti-mouse (labeled with Cy2) were purchased from Jackson Immunolabs and were used at a concentration of 1:200. The nuclei were stained with Hoechst 33258 (Molecular Probes, Eugene, OR). Images were acquired by Intelligent Imaging System software from an Olympus 1X81 inverted motorized microscope equipped with Olympus DSU (spinning disk confocal) and Hamamatsu ORCA IIER monochromatic CCD camera.

Human islets exposed for 24 h to IFN-γ, either alone or in combination with other cytokines, exhibited profound changes in a number of genes, many of which have been previously documented in either rodent islets or pancreatic β-cells purified by flow cytometry (21,22). Affymetrics HG U133 Plus 2.0 gene chips that enable concurrent analysis of 54,675 genes were used to analyze global gene expression profile of human islet preparations exposed to cytokines. In all, 975 transcripts appeared to be upregulated and 523 downregulated (more than threefold) by IFN-γ alone, and 1,197 transcripts appeared to be upregulated and 660 downregulated (more than threefold) by IFN-γ + IL-1β and TNF-α. The effect on IDO mRNA was the most remarkable (>139-fold, P = 7.99E-06) (Fig. 1) along with another enzyme, tryptophanyl t-RNA synthase (WARS) (>17-fold) (P = 4.89E-06) (Table 1). The effect of IFN-γ on these mRNAs was further enhanced by combination with TNF-α and IL-1β (>235-fold, P = 3.23E-06) and WARS (>19-fold, P = 2.57E-05), although TNF-α and IL-1β alone had modest effects. Two related enzymes of Trp metabolism (KYN monooxygenase), the mitochondrial WARS2 mRNAs were unaffected, as were major secretory products in islets such as insulin and glucagons (Fig. 1; data not shown). Several genes from the JAK-STAT pathway, which is thought to mediate the effects of IFN receptor binding, were also upregulated with IFN-γ pretreatment, including JAK2 (4.5-fold, P = 3.75E-04), STAT1α (4.2-fold, P = 2.07E-04), and IFN-γ regulatory factor-1 (IRF-1) (8.3-fold, P = 1.73E-04). Other transcripts showing dramatic responses to IFN-γ included the chemokines (C-X-C motif) ligand 9 (CXCL9) (106.4-fold, P = 1.43E-06), ligand 10 (CXCL10) (84.7-fold, P = 7.65E-07), ligand 11 (CXCL11) (80.6-fold, P = 5.07E-07), and ligand CCL5 or RANTES (31-fold, P = 2.95E-04) (Table 1). The genes most upregulated by IFN-γ are shown in Table 1. The remaining genes modified by IFN-γ and other cytokines will be described in a separate publication (S.A.S., B. Kutlu, R.W., A. Valentine, H.W.D., A.W., R.G.G., J.C.H., unpublished data) and will also be accessible at http://genespeed.uchsc.edu. The induction of IDO and WARS by IFN-γ and the further potentiation by the addition of TNF-α and IL-1β was confirmed by quantitative real-time PCR (Q-RTPCR) using the 5′ nuclease Taqman assay (Fig. 2). The induction of IDO and WARS transcription occurred within 6 h, peaked between 24 and 48 h, and was sustained for up to 86 h (Fig. 3). Consistent with the microarray data, no significant changes were seen in the expression of KYN under these conditions (Figs. 2 and 3). No difference in β-cell apoptosis between control and cytokine-treated islet preparations was noted by terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling at the end of 24 h (data not shown), although previous studies have suggested that chronic exposure to these cytokine combinations may induce functional changes in human islets after 2–6 days (23) and apoptotic death of the β-cells after 6–9 days (24).

Western blot analysis of human islets after exposure to IFN-γ alone or in combination with TNF-α and IL-1β showed the induction of a single 45-kDa IDO immunoreactive component, a size consistent with that of the primary translation product (45,331 Da). IL-1β and TNF-α alone had little effect and did not appear to potentiate the response to IFN-γ, although synergism with IL-1β and TNF-α has been reported previously (25). Basal IDO protein expression in COS7 was undetectable but reached a level of expression that was 10- to 20-fold higher than treated islets after transfection with a cDNA construct under a cytomegalovirus promoter (Fig. 4).

IFN-γ signaling occurs through the JAK-STAT pathway, leading to the phosphorylation of STAT1α and its translocation to the nucleus, where it initiates transcription (26,27). Western blot analysis detected STAT1α protein (91 kDa) in the control sample and induction of expression with cytokine treatment, most noticeably with IFN-γ–treated tissue (Fig. 4). Blotting with an antibody specific to the transcriptionally active phospho-form of STAT_1α indicated marginal induction by TNF-α and marked increase in response to IFN-γ alone or in combination with IL-1β and TNF-α (Fig. 4).

IDO enzyme activity in intact islet cells was quantified by the determination of the appearance of KYN in the islet culture media after 24 h of exposure to cytokines (Figs. 5 and 6). Treatment with IFN-γ but not IL-1β or TNF-α increased KYN production by the islets (Fig. 5), an effect that was partly suppressed by co-incubation with IL-4. The IDO competitive substrate inhibitor 1-MT blocked the KYN production induced by IFN-γ, without having any effect on the basal level of KYN in the medium. Determination of IDO catalytic activity in islet homogenates after treatment for 24 h with cytokines demonstrated the induction of IDO by IFN-γ but not with other cytokines. IL-4 pretreatment did not significantly induce IDO enzyme activity but countered the effect of IFN-γ. 1-MT inhibited the activity induced by IFN-γ (Fig. 6), as expected. Homogenates of nontransfected COS7 cells showed negligible basal IDO enzyme activity but marked increase upon transfection with a plasmid construct encoding human IDO under the cytomegalovirus promoter. The observed changes in IDO protein expression and enzyme activity suggested that IDO mRNA was translated into enzymatically active protein, although perhaps with a lesser efficiency than may have been indicated by the fold changes in IDO transcripts after cytokine incubation.

Immunohistochemical analysis performed on human islets under basal conditions demonstrated the presence of strong IDO immunoreactivity in a few scattered islet cells (Fig. 7). These included β-cells, as demonstrated by their immunoreactivity for transcription factor PDX1, but also other endocrine cells that showed a similar nuclear morphology and chromogranin immunostaining but lacked PDX-1 immunoreactivity. An increase in the number of IDO immunopositive cells was observed in response to IFN-γ treatment that included both PDX1 immunoreactive β-cells and other cells (Fig. 7). The cytokines IL-1β and TNF-α did not produce an obvious change in the number or intensity of staining, and in combination with IFN-γ, they appeared to decrease the number of immunopositive cells. In this respect, the response appeared to parallel the effects of these agents on the protein expression level (Fig. 4) more than the mRNA response (Fig. 1). As in the case of the basal immunostaining, it was not possible to ascribe the increased staining with IFN-γ treatment to a specific cell type, although it was clear from the distribution of immunoreactivity that it included PDX1-positive β-cells and other endocrine and nonendocrine cells.

Changes in Trp metabolism are likely to play a role in the ability of IFN-γ to induce a tolerizing phenotype in splenic CD8 dendritic cells and of CTLA4-Ig to correct the defective tolerogenesis young type 1A diabetic NOD mice (28). Further support for a tolerogenic role for IDO in islets in vivo has come from the observation that adenovirus-mediated IDO expression in transplanted NOD islets prolonged their survival in NOD/SCID recipients that received diabetogenic splenocytes (29). These effects are in part attributed to local depletion and perturbation of the function of activated B and T lymphocytes by local depletion of free Trp (19,30). It is not clear whether such in vitro observations have a counterpart in the interaction of antigen-presenting cells and effector and regulatory populations of T-cells in the insulitic lesion in the pancreas and whether the endocrine cells can participate in the process. IDO has not previously been described as a significant gene transcript in the pancreatic islet, as evidenced by the low frequency of appearance in pancreas cDNA libraries and its representation in Unigene EST gene profiling analyses (www.ncbi.nlm.nih.gov/Unigene/ESTProfileViewer) (1 clone from 14,299 expressed sequence tags in human islets). Nevertheless, IDO transcripts have been documented in recent microarray profiling experiments on islets incubated in vitro with cytokine, especially in response to IFN-γ (31).

In this study, we demonstrated that IFN-γ alone or in combination with IL-β and TNF-α induces >100-fold the level of IDO mRNA in human islets obtained from heart-beating cadaveric donors. These changes were accompanied by a concomitant increase in IDO protein in human islets and IDO enzymatic activity measured both in intact cells and in islet homogenates. These responses were largely attributable to an IFN-γ effect and not to a nonspecific effect on islet viability. IFN-γ under these conditions does not induce apoptosis, although this can be achieved with cocktail of IL-1β, TNF-α, and IFN-γ after several days of exposure in culture (21,24). Such a cocktail further increased IDO mRNA relative to IFN-γ alone but paradoxically appeared to reduce the level of active enzyme. This may in part be attributable to the production of reactive nitrogen species via IL-1β signaling through nuclear factor-κB to upregulate inducible nitric oxide synthase (32), which in concert with SOD produces peroxynitrite-mediated nitrosylation of tyrosine residues and inactivation of STAT1 and its phosphorylation (33). The inverse correlation between IDO induction and the apoptotic reaction to cytokines suggests that IDO is not deleterious and may even play a protective role.

Although Trp depletion is the favored mechanistic explanation for the tolerogenic effect of IDO, the Trp downstream catabolites KYN and picolinate can directly inhibit T-cell proliferation (34,35), suggesting that other elements of the KYN pathway can also significantly influence cellular function. The effects of chronic β-cell exposure to KYN are unknown, although its metabolites 3-OH-KYN and 3-hydroxyanthranilate can acutely block leucine-stimulated insulin secretion by rat islets (36). Another metabolite of Trp, quinolinate, is a potent agonist of NMDA (N-methyl-d-aspartate) receptors, which during central nervous system infection and AIDS-dementia complex has been implicated in neurotoxicity induced by activated macrophages (37). Both pancreatic β-cells (38) and T-cells (39) express glutamate receptors that may be relevant in this regard. Compared with TDO, IDO has a high K_m for molecular oxygen with little activity expressed <100 μmol/l (40). It has been suggested that endogenous electron donors for IDO are superoxides generated from tetrahydrobiopterin, flavoenzymes, xanthine oxidase, gluthathione reductase, or reduced flavin or pyridine nucleotides (10) and that the enzyme may play a protective biological role as an antioxidant (41) and by suppression of the cell surface expression of major histocompatibility complex (MHC) class I molecules (42).

IDO is only one of many gene transcripts that are regulated by IFN-γ in islets. Several genes from the JAK-STAT signaling pathway showed increased expression, including JAK2, IRF-1, and STAT1α (Table 1). Downstream genes, notably the chemokines CCL5/RANTES, CXCL9, CXCL10, and CXCL11, were upregulated >25-fold (Table 1). The extravasation and recruitment of lymphocytes to the islet in response to these signals would likely exacerbate insulitis and result in further exposure of the tissue to inflammatory cytokines (43). Increased expression of Trp-tRNA synthase (WARS) in response to IFN-γ might compensate for a reduced availability of Trp for protein synthesis resulting from IDO activation. An anti-inflammatory role for WARS is also conceivable because the vascular endothelial WARS is proteolytically processed to an anti-angiogenic factor (44,45)

Elevated circulating levels of CXCL9, CXCL10, and CXCL11 are reported in type 1 diabetes in human serum (40) and suggest that the response to IFN-γ seen here in vitro may model a physiologically relevant response. Further information is needed concerning the spatio-temporal pattern of expression of IFN-γ–responsive genes and their modifiers such as IL-4 and IL-4 receptors. The available immunohistochemical data suggest a complex pattern of IDO expression in human islets (Fig. 7) that encompass a variety of cell types, including a heterogeneous pattern of expression in the major endocrine cell types, including PDX1-positive β-cells. Vascular endothelium, macrophages, and dendritic cells are other potential sites of IDO induction by cytokines based on studies in the other tissues, although these cell types could not account for the overall pattern of islet expression that we observe. Such complexity is perhaps not unexpected given that within the pancreas, islets that are heavily infiltrated with mononuclear cells can occur immediately adjacent to the others that are morphologically normal, both in long-standing human type 1 diabetic subjects and at any stage of disease development in NOD mouse. It attests to the importance of local factors and the dynamic instability of the inflammatory reaction. Within this scenario, we postulate that in the short-term, IDO activation may protect islets from cytotoxic damage through depletion of superoxides and maintenance of redox potential and perhaps a reduction in MHC class 1 surface expression. The release of Trp metabolites such as 3-OH-KYN, 3-hydroxyanthranilate, quinolinate, and picolinic acid could also provide bystander inhibition of immune cell function. However, in the longer term, the metabolites of Trp may adversely affect the endocrine cells, particularly the β-cell that has a low antioxidant capacity but generates reactive oxygen species through mitochondrial electron transfer as part of the sensing of metabolizable secretagogues. In addition, further downstream metabolites derived from NAD such as ADP ribose, NAADH, and cyclic ADP ribose could lead to intracellular Ca²⁺ build up to toxic levels and ultimately lead to β-cell attrition.

S.A.S. was a Juvenile Diabetes Research Foundation postdoctoral fellow and has received Juvenile Diabetes Research Foundation Grant 02-05-60294) and the provision of pancreatic islets and core resources from the Islet Cell Resource Center at the University of Colorado at Denver and Health Sciences Center (National Institutes of Health Grant 5-U42-RR-016599) and the Diabetes and Endocrinology Research Center cores (National Institutes of Health Grant P30-DK-57516).

We thank Dr. Chris Wright (School of Medicine, Vanderbilt, TN) for the gift of PDX1 antibody. We thank Travis Still and Iyabo Osifeso for their assistance with histology and Tony Valentine and Joshua Beilke for the isolation of islets.