Cell Tissue Res (2002) 310:169–175 DOI 10.1007/s00441-002-0628-6

REGULAR ARTICLE

Cheng Luo · Markku Kallajoki · Rene Gross Mika Mulari · Tamara Teros · Laura Ylinen Marjaana Mäkinen · Jukka Laine · Olli Simell

Cellular distribution and contribution of cyclooxygenase (COX)-2 to diabetogenesis in NOD mouse Received: 22 July 2002 / Accepted: 31 July 2002 / Published online: 27 August 2002 © Springer-Verlag 2002

Abstract Unlike most other mammalian cells, β-cells of Langerhans constitutively express cyclooxygenase (COX)-2 rather than COX-1. COX-2 is also constitutively expressed in type 1 diabetes (T1D) patients’ periphery blood monocytes and macrophage. To understand the role of COX-2 in the β-cell, we investigated COX-2 expression in β-cells and islet infiltrates of NOD and BALB/c mice using fluorescence immunohistochemistry and cytochemical confocal microscopy and Western blotting. Immunostaining showed that COX-2 is expressed in islet-infiltrating macrophages, and that the expression of insulin and COX-2 disappeared concomitantly from the β-cells when NOD mice progressed toward overt diabetes. Also cultured INS-1E cells coexpressed

insulin and COX-2 but clearly in different subcellular compartments. Treatment with celecoxib increased insulin release from these cells in a dose-dependent manner in glucose concentrations ranging from 5 to 17 mM. Excessive COX-2 expression by the islet-infiltrating macrophages may contribute to the β-cell death during insulitis. The effects of celecoxib on INS-1E cells suggest that PGE2 and other downstream products of COX-2 may contribute to the regulation of insulin release from the βcells.

The project was supported by grants from the Academy of Finland, the Diabetes Research Foundation, Finland, the Sigrid Jusélius Foundation, the Novo Nordisk Foundation, and the Juvenile Diabetes Research Foundation International (grant nos. 197058 and 4-1999-731 to O.S.)

Introduction

C. Luo (✉) · T. Teros · L. Ylinen · M. Mäkinen · O. Simell Juvenile Diabetes Research Foundation (FDRF) Center for Prevention of Type 1 Diabetes in Finland, University of Turku, Turku, Finland e-mail: [email protected] Tel.: +358-2-3338001, Fax: +358-2-3337000 C. Luo · T. Teros · L. Ylinen · M. Mäkinen · O. Simell MediCity Research Laboratory, University of Turku, Turku, Finland C. Luo · T. Teros · L. Ylinen · M. Mäkinen · O. Simell Department of Pediatrics, University of Turku, Turku, Finland M. Kallajoki · J. Laine Department of Pathology, University of Turku, Turku, Finland M. Mulari Department of Anatomy, University of Turku, Turku, Finland R. Gross UMR 5094 du Centre National de la Recherche Scientifique (CNRS), Université Montpellier I, Montpellier, France C. Luo MediCity Research Laboratory, University of Turku, Tykistökatu 6 A, 20520, Turku, Finland

Keywords COX-2 · Macrophage · Confocal microscopy · INS-1 cells · Non obese diabetic (NOD) mouse · Mouse (BALB/c)

Cyclooxygenase (COX) is a key enzyme in the conversion of a polyunsaturated fatty acid, arachidonic acid, to prostaglandin (PG) H2, which is then converted into various prostanoids (PGs, prostacyclins and thromboxanes). Its two isozymes, COX-1 and COX-2, are encoded by two separate genes in different chromosomes (Hla and Neilson 1992), and differ in tissue distribution, promoter structures and non-steroid anti-inflammatory drug-binding sites (Smith and Dewitt 1996). COX-1 is constitutively expressed in most tissues, where it synthesizes small physiological amounts of prostaglandins. COX-2, on the other hand, is mainly expressed in activated macrophages and other inflammatory cells and becomes strongly upregulated after exposure to growth factors or inflammatory stimuli. COX-2 expression is commonly elevated in malignant cells, suggesting that the enzyme may contribute to carcinogenesis (Ristimäki et al. 1997; Molina et al. 1999; Luo et al. 2001). Recent studies suggest that COX-2 rather than COX-1 is predominantly expressed in the islets of Langerhans, a situation opposite to most other mammalian cell types (Robertson 1999). Interestingly, CD14+ monocytes of the patients with type 1 diabetes (T1D), and of healthy sub-

170

jects with diabetes-associated autoantibodies and increased risk of developing T1D, also constitutively express COX-2 (Litherland et al. 1999). COX-2 gene expression in β-cells of isolated murine islets increases modestly with treatment of interleukin (IL)-1β, and diminishes when the islets are exposed to dexamethasone (Sorli et al. 1998; Tran et al. 1999). PGE2 downregulates T-cell-dependent immune responses including protection against immunity to self antigens (Takano et al. 1998). It also mediates at least part of the immunosuppressive effects of some tumours (Walker et al. 1992). Treatment of pancreatic β-cells or islets with exogenous PGE2 or with PGE2 inducers decreases insulin secretion in vitro. PGE2 also controls Tcell-receptor-dependent release of interferon (IFN)-γ from islet-reactive CD8+ T cells in NOD mouse, and inhibits IFN-γ release from cloned NOD CD8+ T cells and polyclonal cytotoxic T lymphocytes. These effects are mediated by the main PGE2 receptors EP2 and EP4 (Ganapathy et al. 2000). Excessive INF-γ production may harm β-cells and enhance progression to T1D, but it may also conversely promote Th2 response that is believed to slow down progression towards overt T1D. The protective effect of oral insulin administration on T1D development is associated with increased production of PGE2 in the islets (Hancock et al. 1995), known to be involved in the differentiation of the T-helper subsets, where it particularly promotes expansion of the Th2 subset (Hilkens et al. 1995). Expression of COX-2- and COX-2-dependent production of PGE2 may regulate insulin production and secretion in β-cells. Furthermore, they may induce and regulate T-cell differentiation and cytokine production via macrophages particularly during exposure to exogenous disease trigger(s). To understand these phenomena, we studied whether COX-2 and insulin are coexpressed in the islets of Langerhans of BALB/c mice, and of NOD mice during progression toward overt T1D. We also studied COX-2 expression in vitro in INS-1E cells, and examined how exposure of these cells to a selective COX-2 inhibitor, celecoxib, modifies their insulin secretion.

Materials and methods Animals Six female NOD and six female BALB/c mice at the age of 1 week together with their four dams, and six female NOD and six female BALB/c mice at the age of 7 weeks, were purchased from Bomholtgard (Ry, Denmark). The mice were maintained in pathogen-free conditions in a 12 h-12 h light-dark cycle with free access to water and normal laboratory chow ad libitum. The NOD and BALB/c young were killed with CO2 1 week later. The older NOD and BALB/c mice were killed at 14 or 17 weeks of age. For Western blotting samples, three 2-, 17-, and 23-week-old NOD mice, and three 2- and 30-week-old BALB/c mice, were ordered from the same place. The distal end of the vas deferens from a 6-weekold male BALB/c mouse was used as a positive control for immunohistochemistry and Western blotting. The use of the animals was permitted by the local animal authorities (permission nos. 834/98 and 1050/00).

Antibodies For primary antibodies rabbit anti-mouse insulin polyclonal antibody was purchased from Santa Cruz Biotechnology (Santa Cruz, CA), COX-2 monoclonal antibody (Clone 33) from Transduction Laboratory (Lexington, KY), affinity purified rabbit anti-mouse COX-2 polyclonal antibody from Cayman (Ann Arbor, MI), and macrophage marker CD68 anti-human monoclonal antibody from DAKO Immunochemicals (Glostrup, Denmark). CD3 (TCR-ε) monoclonal antibody was purchased from Novocastra (Newcastle, UK). For secondary antibodies biotin-conjugated goat anti-rabbit IgG was used for polyclonal insulin and COX-2 antibodies and biotin-conjugated goat anti-mouse IgG for monoclonal COX-2 antibody. Horseradish peroxidase (HRP)-streptavidin was purchased from Zymed (Burlingame, CA). The fluorescein isothiocyanate (FITC)-conjugated polyclonal swine anti-rabbit IgG was obtained from DAKO. Alexa Fluor 568-conjugated goat anti-mouse IgG and fluorescent DNA stain ToPro-3 (a monomeric thiazole orange derivative, TP3) were obtained from Molecular Probes (Eugene, OR). Western analysis For immunodetection, the tissue samples of three mice from each group were directly transferred in protein extraction buffer (30 mM TRIS-HCl, pH 7.8, 1% Tween 20, 100 µM phenylmethylsulphonyl fluoride) in a wt/vol ratio of approximately 200 mg/ml, and gently homogenized on ice with Ultra-Turrax T25 (IKA Labortechnik, Staufen, Germany). The protein concentration of the supernatant was determined using a Bio-Rad assay (Hercules, CA) with bovine serum albumin (BSA) as standard. A total of 100 µg protein was resolved in 10% sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE). The SDS-PAGE-resolved proteins were transferred onto polyvinylidene difluoride (PVDF) membrane (Bio-Rad) using a Bio-Rad protein transfer apparatus. The immunodetection of COX-2 was carried out using 1:2,500 dilution of COX-2 monoclonal antibody. The secondary antibody was biotin-goat anti-mouse IgG (1:5,000). The complex was detected with HRP-streptavidin (1:5,000). Peroxidase activity was visualized using ECL reagents (Amersham-Pharmacia, Bucks., UK), and exposed on highly sensitive BioMax MS-1 film (Kodak, New York, NY). Cell culture INS-1E cells were a generous gift from Prof. C.B. Wollheim (Geneva, Switzerland). The cells were grown at 37°C in humidified 95% air and 5% CO2 atmosphere and passaged every week using 0.05% trypsin-ethylenediaminetetraacetic acid (EDTA). The RPMI-1640 culture medium (Life Technologies, Scotland, UK) was supplemented with 10% fetal calf serum (Autogen Bioclear, Calne, Wilts., UK), 100 U/ml penicillin, 100 µg/ml streptomycin, 2 mM L-glutamine, 10 mM hydroxyethylpiperazine ethanesulphonic acid (HEPES), 1 mM sodium pyruvate and 50 µmol/l βmercaptoethanol (Beffy et al. 2001). Insulin assay INS-1E cells were seeded in six-well plates (approximately 1×105 cells/well) and cultured for 5 days. The cells were then washed and preincubated at 37°C for 1 h in Krebs Ringer’s bicarbonate (KRB) buffer (pH 7.4) containing 0.2% BSA and no glucose. After removal of preincubation medium, the cells were again washed once and incubated at 37°C for 1 h in the same buffer containing 5 mM, 11 mM and 17-glucose and celecoxib (provided by Searle, Pharmacia, Chicago, IL) in a concentration of 1 µM, 10 µM or 20 µM (similar to high serum concentrations in human in vivo). At the end of the incubation period, medium was collected, centrifuged to avoid floating cells and frozen for insulin deter-

171 mination. One millilitre of cold acid ethanol mixture (absolute ethanol: H2O: concentrated HCl, 150:47:3, v/v) was then added on the cells to the wells to extract and homogenize the contents, and subsequently to measure the cellular insulin. Insulin was measured using radioimmunoassay with rat insulin standard (Novo Nordisk, Copenhagen, Denmark). The ratio of insulin secreted per hour/insulin remaining in the cells was used to avoid possible differences in proliferation of the INS-1E cells. The sensitivity of the insulin assay was 0.1 ng/ml. Immunofluorescence histochemistry and cytochemistry with confocal microscopy Deparaffinized, antigenically retrieved tissues and immunofluorescence were studied as described (Luo et al. 2002), except that CD3 antigen retrieval was carried out in 10 mM TRIS-HCl, 1 mM EDTA, pH 9.0, in a microwave oven for 10 min. Consecutive sections were used for staining with two separate antibodies in order to detect the two antigens in approximately the same zone. For immunohistochemistry, polyclonal anti-mouse insulin antibody (1:200 dilution) alone in 0.2% gelatine in phosphate-buffered saline (PBS), or with monoclonal anti-human COX-2 antibody (1:200), or with monoclonal CD68 antibody (1:100 dilution) was incubated on tissue sections for 30 min at 37°C. Slides were washed first with 0.2% gelatine PBS for 2×10 min, and then with PBS for 2×10 min. The FITC-conjugated swine anti-rabbit IgG in 1:100 dilution with 0.2% gelatine PBS for polyclonal COX-2 antibody, Alexa Fluor 568-conjugated goat anti-mouse IgG in 1:100 dilution for monoclonal COX-2, and CD68 and CD3 antibodies were incubated for 30 min at 37°C, then washed with 0.2% gelatine in PBS for 2×10 min, and finally with PBS for 2×10 min. Slides were air dried and mounted with 90% glycerol in PBS. The FITC, or Alexa Fluor 568-labelled, specimens were scanned using Leica TCS SP confocal laser scanning microscopy (Leica Microsystems, Heidelberg, Germany). The 488-nm and 568-nm excitation laser lines of the Omnichrome ArKr-laser (Melles Griot, Carlsbad, CA) were used for FITC and Alexa 568 excitation, respectively, or both fluorescences were simultaneously utilized. For the fluorescence cytochemistry the INS-1E cells were collected at the concentration of 2.5×106/ml. One millilitre freshly made 4% paraformaldehyde fixative was added to 1 ml cell suspension, and incubated at RT for 30 min, washed twice with PBS (pH 7.4), then exposed to cytospin at 300 rpm for 3 min to let cells attach to the slides. The cells were then permeabilized by adding a few drops of 0.1% Triton X-100 in PBS, incubated at 4°C for 4 min, and used for fluorescence immunohistochemistry as described. Image analysis and statistics Immunohistochemical images were photographed with a Leica DM RB E research microscope using a Leica DC 100 digital camera. The data were directly transmitted to a computer with a Leica DC Viewer version 3.2 and saved as tiff files. Confocal images were recorded as tiff files with TCS NT software, version 1.6.587 (Leica Microsystems, Heidelberg, Germany). The image of exposed film of Western blot was captured by a Kodak EDAS 120 camera with an inverted light source, and transferred to computer using KDS 1D 2.0 software from Kodak, and saved in tiff format. The relative optical density (ROD) was measured and analysed with the Analysis Image System (AIS) version 4.0 software (AIS, Ontario, Canada). All images were edited using CorelDraw 9.0 (Corel, Ottawa, Canada) and Adobe PhotoShop 6.01 without adding artefacts and without loss of original resolution.

Results COX-2 and insulin are concomitantly expressed in the β-cells during progression toward T1D In all BALB/c mice and young NOD mice without insulitis (age 2.5 weeks), confocal immunofluorescence microscopy with insulin and COX-2 antibodies showed that the two antigens were expressed solely in the β-cells in the islets (data not shown). As insulitis in NOD mice usually develops by the age of 4–8 weeks, all our 14week-old NOD mice had mostly severely inflamed islets with only few remaining insulin-containing β-cells (Fig. 1A, B). These β-cells all also coexpressed COX-2 just like the β-cells in healthy islets, but COX-2 alone was strongly expressed also by non-insulin-containing cells, likely to be islet-infiltrating macrophages that had emerged to the islets, or islet dendritic cells (Fig. 1C). Within the β-cells the nuclear membrane and its inside constantly showed the strongest COX-2 expression. By 14–17 weeks of age, when a large proportion of the islets of Langerhans in the NOD mice had become infiltrated with lymphocytes and monocytes (Fig. 1A1–A3), numerous COX-2-expressing cells also stained with antiCD68 (Fig. 1B1–B3, C1–C3), suggesting that they were macrophages. The morphology of these cells was also typical for macrophages and they were mainly located around the capillaries and at the periphery of the islet. The remaining β-cells only expressed insulin (green), but never CD68 (Fig. 1C2). The vast majority of the infiltrating mononuclear cells expressed neither CD68 nor COX-2, as indicated by the nuclear staining (Fig. 1A2, B3, C3).

COX-2 is present in the BALB/c and NOD mice in pancreases at all ages, but its origins differ Western blotting showed that the rate of COX-2 protein expression was similar in 4-week-old BALB/c and NOD mice pancreases, i.e. before the onset of insulitis in NOD mice (Fig. 2A). COX-2 expression in the pancreas of adult BALB/c mice was almost two-fold higher than that in 19- and 25-week-old non-diabetic or in diabetic NOD mice, probably because a marked proportion of the βcells had been destroyed during advancement of insulitis. The fact that COX-2 expression continued in the adult NOD pancreas despite almost complete loss of insulin and COX-2 containing β-cells was probably due to the large number of COX-2-expressing activated macrophages that had entered the islets or were located in its immediate vicinity (Fig. 2B).

The expression and function of COX-2 in INS-1E cells Cultured insulin-producing INS-1E cells coexpressed insulin and COX-2 just like the BALB/c and NOD

172

Fig. 1A–C Representative images of distribution of COX-2 in inflamed islets of NOD mice. A An islet from a 14-week-old NOD mouse with advanced insulitis. The autoimmune attack has led to destruction of almost all β-cells. B COX-2 expression in remaining β-cells and in apparent macrophages or dendritic cells. C The overlay picture demonstrates that remaining β-cells express insulin and COX-2 (yellow), but that the apparent dendritic cells and macrophages express only COX-2 (red). A1–A3 CD3-expressing lymphocytes and CD68-expressing macrophages in the islets of Langerhans of NOD mice with advanced insulitis. A1 Islet-infiltrating CD3+ T cells detected using TCR-ε monoclonal antibody (red). A2 An islet with moderate infiltrate of lymphocytes. The nuclei of the lymphocytes were stained by ToPro-3 (blue). COX-

2-positive cells (red) in the periphery of the islets are probably macrophages. Insulin (green) and COX-2 (red) in β-cells is overlaid (yellowish). A3 Haematoxylin and eosin staining of an inflamed islet from a 17-week-old diabetic NOD mouse. B1 From a 14-week-old NOD mouse stained for insulin (green) and COX-2 (red). C1 The next section from the same islet as in B1, stained for insulin (green) and CD68 (red). B2 A section from a 14-week-old NOD mouse islet stained for insulin and COX-2. C2 Consecutive section from the same islet as in B2 stained for insulin (green) and CD68 (red). B3 A section from a 17-week-old diabetic NOD mouse islet, stained for COX-2 (red). C3 Consecutive section from the same islet as in B3 stained for CD68 (red). Scale bars 10 µM

173

Fig. 2A, B Comparison of COX-2 protein expression in vas deferens (VD, positive control) and pancreatic tissue of BALB/c and NOD mice using Western blotting and 10% SDS-PAGE. A The sample is from 4-week-old animals (lane 1 vas deferens, lane 2 pancreases of 4-week-old BALB/c mice, lane 3 pancreases of 4week-old NOD mice). B The samples are from 19-week-old or older mice (lane 1 vas deferens, lane 2 pancreas of 32-week-old BALB/c mice, lane 3 pancreas of 25-week-old non-diabetic NOD mice, lane 4 pancreas of 25-week-old diabetic NOD mice, lane 5 pancreases of 19-week-old non-diabetic NOD mice, lane 6 pancreases of 19-week-old diabetic NOD mice, ROD relative optical density)

Fig. 3A–C Representative images showing dual-channel confocal microscopy pictures of double-stained INS-1E cells. A Insulin staining (green). B COX-2 staining (red). C The dual-channel overlay picture shows that the INS-1 cells express insulin and

mouse β-cells. In the INS-1 cells, double immunostaining and confocal microscopy localized insulin and COX-2 mainly to different subcellular compartments (Fig. 3). As expected, insulin was found in the excretory granules mainly at the periphery of the cell, while COX2 was primarily located at the nuclear membrane, and during COX-2 overexpression also around the nucleus and endoplasmic reticulum with traces in cytosol (Fig. 3B, C). To further analyse the role of COX-2 and its metabolically active end products in INS-1E cells, we studied the effects of different concentrations of celecoxib, a selective COX-2 inhibitor, on insulin secretion in different glucose concentrations (Fig. 4). In KRB buffer containing 5 mM glucose, the INS-1E cells secreted insulin to the medium in a manner that depended on the celecoxib dose (P for the difference between 0 µM and 1 µM, 10 µM and 20 µM concentrations of celecoxib = 0.0023, 0.011 and 0.0012, respectively). However, in 11 mM glucose 10 µM and 20 µM celecoxib released similar amounts of insulin from INS-1E cells. This dampening effect of high medium glucose concentration on celecoxib-induced insulin release is probably explained by the strong effect of glucose alone on insulin release at a high glucose concentration, because the amount of insulin released increased by 15% when medium glucose concentration increased from 5 mM to 11 mM, but by 54% when glucose concentration increased from 11 mM to 17 mM. The addition of even 1 µM celecoxib to the assay system thus modified insulin release markedly, but celecoxib was unable to greatly alter glucose-dependent basal secretion rates.

Discussion Our study visually shows that β-cells in islets of Langerhans belong to the few cell types in mice that constitutively express COX-2. A few previous studies have suggested constitutive COX-2 expression in the islets, but

COX-2 mainly in different intracellular compartments. The yellow colour indicates some intracellular colocalization. Scale bars 10 µM

174

Fig. 4A–C The effect of various concentrations of COX-2 inhibitor celecoxib on glucose-induced insulin secretion by the INS-1 cells in culture. The columns present means ± SEM of four independent experiments performed in three to five replicates. The insulin secretion is expressed as the ratio of insulin secreted to the KRB buffer over the amount of insulin remaining in the cells after 1 h exposure. For further details of the assay system, see “Insulin assay” in “Materials and methods.” A Effects of increasing concentration of celecoxib in 5 mM glucose. The insulin-release-stimulating effect of celecoxib was dose dependent (P for difference between 0 µM celecoxib and 1 µM, 10 µM or 20 µM celecoxib = 0.0023, 00.011, 00,0012, respectively). B Effects of increasing concentration of celecoxib in 11 mM glucose. The stimulatory effect of celecoxib was again dose dependent (P=0.209, 0.011, 0,0006). C Effects of increasing concentration of celecoxib in 17 mM glucose. Insulin release values do not differ statistically between the different celecoxib concentrations (P>0.05)

our data provide direct evidence that of the islet endocrine cells, COX-2 is expressed only in the β-cells, and in tight linkage with the expression of insulin. Whether the two molecules share common regulatory pathways, or whether their metabolism is otherwise interdependent in the β-cells remains unknown so far. Interestingly, insulin and insulin-like growth factor I (IGF-1) amplify IL-1β-induced nitric oxide (NO) and COX-2-dependent prostaglandin biosynthesis in glomerular mesangial cells (Guan et al. 1998), which also express COX-2 and inducible nitric oxide synthase (iNOS) constitutively. Another interesting molecule, nephrin, previously documented to be expressed in the glomeruli, has now been found also only in β-cells in the islets (Palmen et al. 2001). COX-2 was not detected in the zone of islet-infiltrating lymphocytes or inside the peri-islet area, which is consistent with independent regulation of the trafficking of antigen-presenting cell (APC): macrophages and dendritic cells and T cells that participate in the development of islet inflammation (Lee et al. 1988; Fox and Danska 1998). The role of the macrophages in the inflamed islets is probably not restricted to secretion of cytotoxic and inflammation-inducing mediators only, because they also produce substances that participate in tissue reorganization including enzymes such as hyaluronidase, elastase, and collagenase, anti-proteases, regulatory growth factors and others. However, the presence of COX-2 in the β-cells and infiltrating macrophages strongly suggests that the enzyme has a disease-accelerating role in the genetically diabetes-prone NOD mouse, possibly through regulation of the subsets of T lymphocytes recruited to the islets, or via production of cytokines that favour progression of cell destruction. The fact that Western blotting analysis and immunostaining showed similar COX-2 expression patterns in the β-cells of young BALB/c and NOD mice, and the expression of insulin and COX-2 declined strictly in parallel during progression of insulitis, suggests that COX-2 has important cellular or paracrine functions in the βcells. However, the relative contributions of COX-2 in βcells, infiltrated macrophages and dendritic cells at different stages of diabetes development remain to be explored. Clearly, in young NOD mice a larger proportion of COX-2 is in the β-cells, but during progression of insulitis and development of diabetes an increasing proportion of COX-2, and probably its metabolically active end products, is produced by the activated islet-infiltrating macrophages. Our morphological data show that COX-2 and insulin are expressed tightly in parallel in the β-cells, even though insulin and COX-2 genes have totally different regulatory mechanisms. The mechanisms that link insulin and COX-2 expression together in the β-cells remain unknown, but the fact that the link apparently exists suggests an important role for COX-2 in β-cell physiology. PGE2 is a potent inhibitor for insulin secretion in vitro, but may promote insulin secretion during fasting or low plasma glucose concentrations. An inducible cyclooxy-

175

genase, COX-2, might under such circumstances respond better than the constitutive isoform COX-1. Given the results above, we further questioned whether the COX-2 has any functional correlation with the secretion of insulin. At 17 mM glucose the COX-2 inhibitor had no clear effect on insulin release from insulin-producing INS-1E cells, probably because insulin release had reached a plateau between 15 mM and 20 mM glucose concentration. However, in 5 mM and 11 mM glucose concentrations, COX-2 inhibitor enhanced insulin secretion in a dose-dependent manner. A previous study showed that IL1α or IL-1β stimulated COX-2 and iNOS expression in rat islets and insulin-producing RINm5F cells in vitro. However, the situation in vivo might be much more complicated, as PGE2 may also promote insulin secretion in vivo during physiological conditions. One possibility is that PGE2 releases β-endorphin, which then influences insulin secretion through sympathetic regulation. It is also possible that celecoxib to some extent directly rather than via COX2 and PGE2 inhibits insulin secretion. Confocal microscopy confirmed that insulin and COX-2 are coexpressed in the β-cells but mainly in different cellular compartments, previously known to be their prominent expression sites in the cell. COX-2 is also expressed by dendritic cells residing in the healthy islets, and by islet-infiltrating macrophages. A regulatory network triggered by COX-2 end products may be important for maintenance of insulin secretion under physiological circumstances. However, at present the relative importance of COX-2, its products, iNOS and its products, and different cytokines and other factors for insulin secretion remains speculative both during normal physiological conditions and during progression of insulitis toward clinical diabetes. Acknowledgements We thank Dr. Pierre Maechler, Division of Clinical Biochemistry and Department of Internal Medicine, University Medical Center, Geneva, Switzerland, for advice on the culture of INS-1E cells.

References Beffy P, Lajoix AD, Masiello P, Dietz S, Peraldi-Roux S, Chardes T, Ribes G, Gross R (2001) A constitutive nitric oxide synthase modulates insulin secretion in the INS-1 cell line. Mol Cell Endocrinol 183:41–48 Fox CJ, Danska JS (1998) Independent genetic regulation of T-cell and antigen-presenting cell participation in autoimmune islet inflammation. Diabetes 47:331–338 Ganapathy V, Gurlo T, Jarstadmarken HO, von Grafenstein H (2000) Regulation of TCR-induced IFN-gamma release from islet-reactive non-obese diabetic CD8(+) T cells by prostaglandin E(2) receptor signaling. Int Immunol 12:851–860

Guan Z, Buckman SY, Baier LD, Morrison AR (1998) IGF-I and insulin amplify IL-1 beta-induced nitric oxide and prostaglandin biosynthesis. Am J Physiol 274:F673–F679 Hancock WW, Polanski M, Zhang J, Blogg N, Weiner HL (1995) Suppression of insulitis in non-obese diabetic (NOD) mice by oral insulin administration is associated with selective expression of interleukin-4 and -10, transforming growth factor-beta, and prostaglandin-E. Am J Pathol 147:1193–1199 Hilkens CM, Vermeulen H, van Neerven RJ, Snijdewint FG, Wierenga EA, Kapsenberg ML (1995) Differential modulation of T helper type 1 (Th1) and T helper type 2 (Th2) cytokine secretion by prostaglandin E2 critically depends on interleukin-2. Eur J Immunol 25:59–63 Hla T, Neilson K (1992) Human cyclooxygenase-2 cDNA. Proc Natl Acad Sci U S A 89:7384–738 Lee KU, Amano K, Yoon JW (1988) Evidence for initial involvement of macrophage in development of insulitis in NOD mice. Diabetes 37:989–991 Litherland SA, Xie XT, Hutson AD, Wasserfall C, Whittaker DS, She JX, Hofig A, Dennis MA, Fuller K, Cook R, Schatz D, Moldawer LL, Clare-Salzler MJ (1999) Aberrant prostaglandin synthase 2 expression defines an antigen-presenting cell defect for insulin-dependent diabetes mellitus. J Clin Invest 104: 515–523 Luo C, Strauss L, Ristimaki A, Streng T, Santti R (2001) Constant expression of cyclooxygenase-2 gene in prostate and the lower urinary tract of estrogen-treated male rats. Z Naturforsch [C] 56:455–463 Luo C, Laine JVO, Ylinen L, Teros T, Mäkinen M, Ristimäki A, Simell O (2002) Expression of cyclooxygenase-2 in intestinal goblet cells of prediabetic NOD mice. Acta Physiol Scand 174:265–274 Molina MA, Sitja-Arnau M, Lemoine MG, Frazier ML, Sinicrope FA (1999) Increased cyclooxygenase-2 expression in human pancreatic carcinomas and cell lines: growth inhibition by nonsteroidal anti-inflammatory drugs. Cancer Res 59:4356–4362 Palmen T, Ahola H, Palgi J, Aaltonen P, Luimula P, Wang S, Jaakkola I, Knip M, Otonkoski T, Holthofer H (2001) Nephrin is expressed in the pancreatic beta cells. Diabetologia 44: 1274–1280 Ristimäki A, Honkanen N, Jankala H, Sipponen P, Harkonen M (1997) Expression of cyclooxygenase-2 in human gastric carcinoma. Cancer Res 57:1276–1280 Robertson RP (1998) Dominance of cyclooxygenase-2 in the regulation of pancreatic islet prostaglandin synthesis. Diabetes 47: 1379–1383 Smith WL, Dewitt DL (1996) Prostaglandin endoperoxide H synthases-1 and -2. Adv Immunol 62:167–215 Sorli CH, Zhang HJ, Armstrong MB, Rajotte RV, Maclouf J, Robertson RP (1998) Basal expression of cyclooxygenase-2 and nuclear factor-interleukin 6 are dominant and coordinately regulated by interleukin 1 in the pancreatic islet. Proc Natl Acad Sci U S A 95:1788–1793 Takano M, Nishimura H, Kimura Y, Washizu J, Mokuno Y, Nimura Y, Yoshikai Y (1998) Prostaglandin E2 protects against liver injury after Escherichia coli infection but hampers the resolution of the infection in mice. J Immunol 161:3019–3025 Tran PO, Gleason CE, Poitout V, Robertson RP (1999) Prostaglandin E(2) mediates inhibition of insulin secretion by interleukin1beta. J Biol Chem 274:31245–31248 Walker DA, Dillon M, Levitt G, Cervera A, Shaw D, Pritchard J (1992) Multiple exostosis (osteochondroma) and Wilms’ tumour, a possible association. Med Pediatr Oncol 20:360–361