Introduction

Glucagon-like peptide-1 (GLP-1), an incretin hormone secreted by enteroendocrine L cells of the gut, potentiates glucose-dependent insulin secretion in pancreatic beta cells. GLP-1 activates G-protein coupled receptors, leading to activation of adenylyl cyclase, which results in generation of cyclic AMP. This second messenger plays an important role in mediating the actions of GLP-1, although other signalling pathways are also involved [1]. Several studies have reported the anti-inflammatory action of cyclic AMP [2]. For example, the cytokine-activated janus kinase (JAK)-signal transducer and activator of transcription-1 (STAT-1) signalling pathway is inhibited by cyclic AMP in monocytes, in vascular endothelial cells and in hepatic stellate cells [3–5]. The effects of cyclic AMP are transient because it is degraded by phosphodiesterases. The isozymes of phosphodiesterase play important tissue-specific roles in the regulation of signalling pathways mediated by cyclic AMP. Anti-inflammatory actions of phosphodiesterase inhibitors in NOD mice and in BioBreeding rats have been reported [6, 7].

Chemokines are a family of ∼50 proteins, consisting of 70–130 amino acids that play critical roles in recruitment of leucocytes at the site of inflammation [8]. There are two major families, termed CC and CXC, depending on the arrangement of first two conserved cysteine residues near their N terminus. Chemokine (C-X-C motif) ligand (CXCL)10 is secreted by various cell types, including keratinocytes, intestinal epithelial cells, endothelial cells, monocytes and neutrophils [9]. Diverse biological effects of this highly inducible chemokine have been reported [10]. CXCL10 stimulates T cell adhesion to endothelial cells, suppresses food intake and inhibits angiogenesis [11]. Chemokine (C-X-C motif) receptor 3 (CXCR3), the receptor for CXCL10, is produced in several immune cell types including activated T cells and natural killer cells [10]. CXCL10 plays an important role in trafficking effector T cells to islets during insulitis in type 1 diabetes [12]. In NOD mice, CXCL10 production in islets is increased prior to initiation of insulitis [13]. CXCL10 neutralisation has been shown to suppress the incidence of diabetes in NOD mice [14]. Levels of CXCL10 are elevated in type 1 diabetic patients especially during early and subclinical stages [15]. Islets isolated from type 2 diabetic patients produce high levels of this chemokine [16]. A recent study with pancreatic tissues from four type 1 diabetic patients reported the presence of CXCL10 and infiltration of lymphocytes expressing the corresponding chemokine receptor CXCR3 in islets, regardless of enterovirus infection [17]. These studies suggest that suppression of CXCL10 production in islets could be an important therapeutic approach to reducing insulitis in type 1 diabetes. Recently, in human islets exposed to a combination of proinflammatory cytokines IL-1β, TNF-α and IFN-γ, we observed induction of several chemokines, including CXCL10 [18]. The objective of this study was to determine whether exendin-4-mediated generation of cyclic AMP in human islets leads to suppression of cytokine-induced CXCL10 expression.

Methods

Cell culture

Human islets isolated from cadaveric donors were provided by Islet Cell Resource Center Human Islet Basic Science Distribution Program (City of Hope, Duarte, CA, USA). This study was approved by Institutional Ethics Committee. None of the donors had a previous history of diabetes (Electronic supplementary material [ESM] Table 1). Islets with purity of 80% to 95% and viability of 75% to 99% were precultured for 24 h in Miami medium (CMRL 1066 medium [Mediatech, Hendon, VA, USA] supplemented with 0.5% [wt/vol.] human serum albumin and nicotinamide [10 mmol/l]) [18, 19]. Islets were exposed to IL-1β (50 U/ng), IFN-γ (20 U/ng) and TNF-α (100 U/ng) (Roche Applied Science, Indianapolis, IN, USA), and rolipram and cilostamide (Enzo Life Sciences, Plymouth Meeting, PA, USA) or exendin-4 (Sigma Chemical, St Louis, MO, USA). MIN6 cells, a mouse pancreatic beta cell line [20] obtained from J. Miyazaki (First Department of Surgery, Kyoto University, Japan), were cultured in DMEM containing 5.6 mmol/l glucose, 10% (vol./vol.) FBS, 100 μg/ml streptomycin, 100 U/ml penicillin and 50 μmol/l β-mercaptoethanol at 37°C in a humidified atmosphere with 5% CO2.

RNA isolation and RT-PCR analysis

Total RNA was isolated from treated human islets using an isolation kit (Versagene RNA; Fisher Scientific, Pittsburgh, PA, USA). The mRNA levels of CXCL9, CXCL10 and CXCL11 were measured by real-time quantitative RT-PCR using Taqman probes. The primers and probes (ESM Table 2) were designed by Primer Express (PE ABI, Foster City, CA, USA). Amplicons corresponding to the amplification sequence were synthesised and used as standards in the RT-PCR analysis. mRNA levels of chemokines were expressed in ag/pg of GAPDH.

Western blot analysis

Following treatment, human islets and MIN6 cells were washed with ice-cold PBS and cell lysates were prepared. Western blot analysis was performed by a procedure described previously [18] using CXCL10 (1:2000; R&D Systems, Minneapolis, MN, USA), phospho STAT-1, STAT-1, phospho cyclic AMP response element binding protein (CREB) or CREB antibodies (1:1000; Cell Signaling, Danvers, MA, USA). After incubation with secondary antibodies conjugated to alkaline phosphatase, signals were developed with CDP-Star reagent (New England Biolabs, Beverly, MA, USA). The membranes were then stripped for 45 min at 55°C in buffer containing 62.5 mmol/l Tris–HCl, pH 6.7, 2% (wt/vol.) SDS and 100 mmol/l β-mercaptoethanol, and reprobed with β-actin antibody (Sigma). The intensity of bands was measured using Fluor-S MultiImager and Quantity One (Bio-Rad, Hercules, CA, USA) and corrected for the levels of β-actin.

ELISA of CXCL10

CXCL10 released into the medium of cultured human islets and MIN6 cells was assayed using kits (DuoSet ELISA Development; R&D Systems) for human or mouse CXCL10 respectively. A 96-well microplate was coated overnight with capture antibody for CXCL10 that was blocked for 1 h at room temperature with 1% (wt/vol.) BSA in PBS. The coated plates were incubated for 2 h at room temperature with samples of medium or CXCL10 standards. After washing, the plates were incubated with biotinylated detection antibody for another 2 h at room temperature. This was followed by incubation with streptavidin conjugated to horseradish peroxidase and colour development with H2O2 and tetramethylbenzidine. The plates were read at 450 nm with wavelength correction at 570 nm. The concentration of CXCL10 was expressed as pg/ml of medium.

Human inflammatory cytokines and receptors PCR array

A PCR-based gene expression profiling specific to inflammatory pathway genes (PAHS-011A; SA Biosciences, Frederick, MD, USA) was carried out with total RNA isolated from treated human islets. RNA samples following DNase treatment were converted to cDNA [21]. Real-time PCR was performed with a sequence detector (ABI Prism 7700; Applied Biosystems, Foster City, CA, USA). The experimental cocktail was prepared by adding 1,278 μl of the RT2 qPCR SYBR Green master mix (SA Biosciences) and 1,173 μl H2O to 102 μl of the diluted cDNA; 25 μl of this cocktail was added to each well of the 96-well PCR array plate containing primers for the 84 inflammatory pathway genes (Table 1), five housekeeping control genes, and three RNA and PCR quality controls. The thermal cycling conditions were: 1 cycle of 10 min at 95°C, followed by 40 cycles of 15 s at 95°C and of 1 min at 60°C. After amplification, real-time data acquisition and analysis were performed through the Data Analysis Web Portal (SA Biosciences). Data analysis was based on the ΔΔCt method with normalisation of the raw data to GAPDH as described in the manufacturer’s manual.

Table 1 Functional gene grouping of inflammatory cytokines and receptors array
Full size table

ELISA of cyclic AMP

The cyclic AMP levels were determined using a cyclic AMP assay kit (Parameter; R&D Biosystems). In this assay based on competitive binding, cell lysates and cyclic AMP standards are added to a microplate coated with antibody to cyclic AMP. A horseradish peroxidase-labelled cyclic AMP conjugate is also added. Cyclic AMP present in the samples competes with the fixed amount of horseradish peroxidase-labelled cyclic AMP conjugate for sites on the monoclonal antibody. After 2 h incubation, unbound sample and excess conjugate are removed by a wash. A substrate solution is added to the wells to determine bound enzymatic activity. Addition of stop solution halts colour development and absorbance is then measured at 450 nm with correction at 540 nm.

Statistical analysis

This was performed by one-way ANOVA with Dunnett’s multiple comparison test.

Results

Suppression of IFN-γ-induced CXCR3-binding chemokines by exendin-4 and cyclic AMP in human islets

CXCL9, CXCL10 and CXCL11 are ligands for immune cells expressing CXCR3 and play critical roles in the initiation of insulitis in type 1 diabetes. The mRNA levels of these chemokines, as measured by real-time RT-PCR analysis using Taqman probes, increased by 230- to 265-fold in islets following treatment with IFN-γ when compared with untreated control. These induced levels were subsequently taken as 100% to determine the anti-inflammatory action of the GLP-1-cyclic AMP pathway. Exendin-4 in the presence of rolipram, an inhibitor of phosphodiesterase 4, which degrades cyclic AMP, suppressed IFN-γ-mediated induction of CXCL9 (24%), CXCL10 (30%) and CXCL11 (22%; Fig. 1). Exendin-4 decreased CXCL10 modestly (p < 0.05) in the absence of rolipram. In subsequent experiments, exendin-4 was used along with a phosphodiesterase inhibitor to increase the levels of cyclic AMP generated. Forskolin, which generates cyclic AMP, and dibutyryl cyclic AMP (DBC), decreased IFN-γ-stimulated mRNA levels of the three chemokines more effectively (p < 0.001; Fig. 1), namely by 45% to 70% and 33% to 60%, respectively.

Fig. 1
figure 1

Suppression of IFN-γ-induced expression of CXCR3-binding chemokines by exendin-4, rolipram, forskolin and DBC. Human islets (2,000 islet equivalents) were pre-incubated with 100 nmol/l exendin-4 (Ex), 15 μmol/l rolipram (R), 20 μmol/l forskolin (Fors) or 1 mmol/l DBC for 30 min, followed by exposure to 2 ng/ml IFN-γ for 6 h. Total RNA was isolated from islets for real-time quantitative RT-PCR analysis of CXCL10 (a), CXCL9 (b) and CXCL11 (c), and expressed in ag/pg of GAPDH. Results are mean±SE from six independent batches of human islets. IFN-γ-induced chemokine levels were taken as 100%. *p < 0.05; **p < 0.01 and ***p < 0.001 compared with IFN-γ control

Full size image

Decrease of IFN-γ-stimulated production of CXCL10 protein by exendin-4 and cyclic AMP in human islets

Among the CXCR3 ligands, CXCL10 is considered to play a more significant role in initiation of insulitis in autoimmune diabetes. Therefore we examined the regulation of this chemokine in more detail. Exendin-4 with rolipram decreased CXCL10 protein levels by 38% (p < 0.01; Fig. 2a), this decrease being reversed by SQ22536, an inhibitor of adenylyl cyclase, suggesting that cyclic AMP plays a role. Exendin-4 action was strongly (p < 0.001) mimicked by forskolin and DBC, which decreased CXCL10 protein by 40–45% (Fig. 2b). Secretion of CXCL10 into the medium of cultured islets in the presence of IFN-γ was also decreased by pre-incubation with exendin-4 (27%), forskolin (70%) and DBC (55%; Fig. 2c).

Fig. 2
figure 2

Decrease of IFN-γ-stimulated production of CXCL10 protein by exendin-4, rolipram, forskolin and DBC. Human islets (2,000 islet equivalents) were pre-incubated with 100 nmol/l exendin-4 (Ex), 15 μmol/l rolipram (R), 100 μmol/l SQ22536 (SQ), 20 μmol/l forskolin (Fors) or 1 mmol/l DBC for 30 min, followed by exposure to 2 ng/ml of IFN-γ, for 6 h, along with untreated control islets (Con). Treated islets were processed for western blot analysis of CXCL10 (a, b). The blots were reprobed for β-actin and band intensities determined by scanning. Band intensity is presented as ratio of CXCL10/β-actin. c Levels of secreted CXCL10 in the medium were determined by ELISA. The results are mean ± SE from five independent batches of human islets. IFN-γ-induced CXCL10 levels were taken as 100%. **p < 0.01 and ***p < 0.001 compared with IFN-γ control

Full size image

Decreased production of STAT-1 by cyclic AMP generation in human islets

Induction of CXCL10 by IFN-γ has been shown to proceed through activation of the transcription factors STAT-1 and nuclear factor kappa B (NF-kB) [22]. We used: (1) a pan JAK inhibitor that blocks activation of STAT-1 by phosphorylation; and (2) Bay 11-7085, which decreases IkB phosphorylation and nuclear localisation of NF-kB. Increases in levels of CXCL10 protein stimulated by IFN-γ were blocked by JAK inhibitor, but not by Bay 11-7085 as shown by ELISA of CXCL10 in culture medium (Fig. 3a) and by western blot analysis of islet lysates (Fig. 3b). This suggests that IFN-γ action in human islets is primarily through STAT-1. Exendin-4, forskolin and DBC decreased the levels of STAT-1 protein by 37% to 47% (p < 0.01; Fig. 3c, d).

Fig. 3
figure 3

Decrease of IFN-γ-stimulated production of STAT-1 by the GLP-1–cyclic AMP pathway. a, b Human islets (2,000 islet equivalents [IEQ]) were pre-incubated for 30 min in the presence of 1 μmol/l pan JAK inhibitor (JI) or 15 μmol/l Bay 11-7085 (Bay), an inhibitor of NF-kB, followed by exposure to 2 ng/ml of IFN-γ for 6 h, along with untreated control islets (Con). Medium was collected to determine levels of CXCL10 by ELISA (a). Islets were processed for western blot analysis and quantification of CXCL10 and β-actin (b). c, d Human islets (2,000 IEQ) were pre-incubated for 30 min with 100 nmol/l exendin-4 (Ex) and 15 μmol/l rolipram (R), and 20 μmol/l forskolin (Fors) or 1 mmol/l DBC, followed by exposure to 2 ng/ml IFN-γ for 6 h. The treated islets were processed for western blot analysis and quantification of STAT-1. The blots were reprobed for β-actin and band intensities determined by scanning. Results are mean ± SE from five independent batches of human islets. **p < 0.01 vs untreated control; † p < 0.01 compared with IFN-γ control

Full size image

Synergistic induction of CXCL10 by cytokines in human islets

Experiments described thus far examined the anti-inflammatory action of the GLP-1-cyclic AMP pathway in human islets exposed to IFN-γ. Following infiltration of islets with immune cells, insulitis is sustained through induction of CXCL10 by proinflammatory cytokines IL-1β, TNF-α and IFN-γ released by immune cells. Therefore human islets were exposed to these cytokines, alone or in combination. IL-1β and TNF-α did not induce CXCL10 expression even at a concentration of 5 ng/ml, whereas IFN-γ increased CXCL10 mRNA by 200- to 500-fold at 2 to 5 ng/ml (Fig. 4a). When a combination of all three cytokines was used, synergistic induction was observed as shown by the 700- to 1,000-fold increase in CXCL10 mRNA levels. Western blot analysis revealed similar synergy between cytokines in CXCL10 protein production in islets (Fig. 4b). The secretion of CXCL10 into the medium was measured by ELISA and the findings were consistent with the western blot analysis results (Fig. 4c).

Fig. 4
figure 4

Synergistic induction of CXCL10 by IL-1β, TNF-α and IFN-γ in human islets. Human islets (2,000 islet equivalents) were exposed to indicated doses of IFN-γ, IL-1β and TNF-α alone or a combination of all three (Cyt) for 6 h. a Total RNA was isolated from islets for RT-PCR analysis of CXCL10. b Islets were processed for western blot analysis and quantification of CXCL10, and blots were reprobed for β-actin. A representative blot is shown. Band intensities of CXCL10 were determined by scanning and corrected for levels of β-actin. c Levels of secreted CXCL10 in the medium were determined by ELISA. The results are mean ± SE from five independent batches of human islets. ***p < 0.001 compared with untreated control

Full size image

Suppression of cytokine-induced CXCL10 expression by cyclic AMP generation in human islets

Although inhibition of NF-kB did not decrease IFN-γ-induced CXCL10 production (Fig. 3a, b), CXCL10 induction by a combination of cytokines was decreased by Bay 11-7085 alone and more effectively when Bay 11-7085 was used in combination with JAK inhibitor (Fig. 5a). Because CXCL10 induction was significantly greater in the presence of a combination of cytokines, we used rolipram and cilostamide, inhibitors of phosphodiesterase 4 and 3 respectively, to enhance the anti-inflammatory action of exendin-4. Pancreatic beta cells are known to contain these two phosphodiesterase isozymes. Exendin-4 and phosphodiesterase inhibitors inhibited CXCL10 by 24% and 28% respectively; when used together, 50% inhibition was observed (Fig. 5b). Exendin-4-mediated suppression of CXCL10 was lost in the presence of an adenylyl cyclase inhibitor (SQ22536). Forskolin and DBC inhibited cytokine-induced CXCL10 expression by 61% and 28% respectively (Fig. 5c).

Fig. 5
figure 5

Suppression of cytokine-induced expression of CXCL10 by cyclic AMP generation in human islets. Human islets (2,000 islet equivalents) were pre-incubated for 30 min with 15 μmol/l Bay 11-7085 (Bay) and 1 μmol/l JAK inhibitor (JI) (a), with 100 nmol/l exendin-4 (Ex), 15 μmol/l rolipram (R), 10 μmol/l cilostamide (C) and 100 μmol/l SQ22536 (SQ) (b), and with 20 μmol/l forskolin (Fors) or 1 mmol/l DBC (c), followed by exposure to 2 ng/ml each of IL-1β, TNF-α and IFN-γ (Cyt) (a–c) as indicated for 6 h. Total RNA was isolated from islets for the RT-PCR analysis of CXCL10 and corrected for levels of GAPDH. Results are mean ± SE from five independent batches of human islets. **p < 0.01 and ***p < 0.001 compared with cytokine control

Full size image

Suppression of cytokine-induced expression of genes encoding chemokines by inhibitors of NF-kB and STAT-1 pathways, and by cyclic AMP generation in human islets

To determine whether the anti-inflammatory action of exendin-4 is seen with other inflammatory mediators, a pathway-specific gene expression profiling of 84 genes in the inflammatory pathway (Table 1) was carried out in islets under basal conditions and following stimulation with a combination of IL-1β, TNF-α and IFN-γ. Exendin-4 decreased the basal expression of several chemokines and cytokines in human islets by 20% to 45% even in the absence of phosphodiesterase inhibitors (Table 2). When islets were stimulated with a combination of cytokines, several chemokines in the CCL and CXCL family were induced (Table 3). In addition, the cytokines IL17C, IL1A, IL1B, IL8 and TNF-α (also known as TNF) were also induced. A combination of inhibitors of the NF-kB and JAK-STAT-1 pathways suppressed (17–96%) all the inflammatory mediators except CCL2, CCL7 and IL8. Exendin-4 in the presence of phosphodiesterase inhibitors reduced expression of CCL2, CCL3, CCL4, CCL7, CCL20, CXCL6, CXCL9, CXCL10, CXCL11, IL1B and IL8 by 20% to 30%.

Table 2 Suppression of chemokines and cytokines by exendin-4 in human islets
Full size table
Table 3 Suppression of cytokine-induced chemokines by cyclic AMP generation and by inhibitors of NF-kB and STAT-1 pathways
Full size table

Regulation of CXCL10 production by cytokines in MIN6 cells

To determine the regulation of CXCL10 production by cytokines specifically in beta cells, we used MIN6 cells, a mouse beta cell line. Among the cytokines tested, both IFN-γ and IL-1β were equally effective in increasing CXCL10 production (Fig. 6a, c), unlike in human islets (Fig. 4). A synergistic effect between the cytokines IL-1β, TNF-α and IFN-γ in the production of CXCL10 was observed, as in human islets. IFN-γ-mediated increase in CXCL10 protein was blocked by JAK inhibitor, which also decreased STAT-1 phosphorylation and STAT-1 production (Fig. 6b). These findings were further confirmed by ELISA of secreted CXCL10 in the culture medium. Although Bay 11-7085 did not decrease CXCL10 release into culture medium, it did enhance the inhibitory effects of JAK inhibitor (Fig. 6c).

Fig. 6
figure 6

Regulation of CXCL10 production by cytokines in MIN6 cells. MIN6 cells were exposed to 2 ng/ml of IL-1β, TNF-α and IFN-γ alone or in combination (Cyt) for 18 h (a) or pre-incubated in the presence of indicated concentrations of JAK inhibitor (JI) followed by exposure to 2 ng/ml of IFN-γ for 18 h (b). Treated MIN6 cells were processed for western blot analysis of CXCL10, phosphorylated and total forms of STAT-1 and β-actin. c MIN6 cells were incubated in the presence of cytokines and inhibitors as indicated for 18 h and levels of CXCL10 secreted into the culture medium were determined by ELISA. The results are mean±SE from four independent experiments. **p < 0.01 compared with MIN6 cells incubated in the presence of cytokines and 0.5 μmol/l of JAK inhibitor; ***p < 0.001 compared with untreated control; † p < 0.001 vs a combination of cytokines

Full size image

Suppression of IFN-γ-mediated CXCL10 production by cyclic AMP generation through decrease in the levels of STAT-1 in MIN6 cells

To determine whether the anti-inflammatory action of exendin-4 in human islets can be extended to beta cells, we examined the abundance of CXCL10 in MIN6 cells pre-incubated with exendin-4 and phosphodiesterase inhibitors. IFN-γ-mediated increase in CXCL10 protein level was significantly (p < 0.01) decreased by exendin-4 (21%), rolipram (24%) and cilostamide (23%), whereas a combination of these three resulted in an additive decrease (57%; p < 0.001) of CXCL10 protein (Fig. 7a). Inhibition of adenyl cyclase with SQ22536 partially reversed CXCL10 suppression by exendin-4 and phosphodiesterase inhibitors (Fig. 7b). Exendin-4 action on chemokine levels was mimicked by forskolin and DBC (Fig.7c). IFN-γ-stimulated increase in the levels of STAT-1 was significantly (45%; p < 0.01) inhibited by exendin-4 and phosphodiesterase inhibitors (Fig. 7d).

Fig. 7
figure 7

Suppression of IFN-γ-induced CXCL10 production by GLP-1-cyclic AMP pathway through decrease of STAT-1. MIN6 cells were exposed to IFN-γ-as indicated and to: (a) exendin-4 (Ex), rolipram (R) and cilostamide (C), alone and together; (b) to Ex+R+C and adenylyl cyclase inhibitor SQ22536 (SQ) as indicated; (c) to forskolin (Fors) and DBC; and (d) to Ex, R+C and Ex+R+C. Concentrations were as indicated in Fig. 4, exposure was for 18 h. Treated cells were processed for western blot analysis of CXCL10, STAT-1 or β-actin

Full size image

Sustained release of cyclic AMP by exendin-4 in the presence of phosphodiesterase inhibitors in MIN6 cells

Our results described thus far suggest increased production of cyclic AMP by exendin-4 in the presence of phosphodiesterase inhibitors. Therefore we measured the levels of this second messenger in exendin-4-treated MIN6 cells. The levels of cyclic AMP in MIN6 cells increased by twofold in the presence of exendin-4 and by 3.5-fold in the presence of exendin-4 and phosphodiesterase inhibitors, compared with untreated cells (Fig. 8a). CREB, an anti-apoptotic transcription factor that mediates the effects of exendin-4, is phosphorylated by cyclic AMP-activated protein kinase A (PKA) [19]. Therefore we determined the levels of phosphorylated CREB as a measure of increase in cyclic AMP levels. Exendin-4-mediated CREB phosphorylation increased by twofold at 15 min, but decreased gradually over the following 30 to 60 min, probably due to phosphodiesterase-mediated degradation of cyclic AMP (Fig. 8b). In the presence of rolipram and cilostamide, exendin-4-mediated CREB phosphorylation increased by 3.8-fold and was sustained over a 15 to 60 min period. Phosphodiesterase inhibitors also increased the level of phosphorylated CREB by 2.3-fold in the absence of exendin-4, suggesting that phosphodiesterase plays a role in degradation of cyclic AMP under basal conditions.

Fig. 8
figure 8

Sustained release of cyclic AMP (cAMP) by exendin-4 in the presence of phosphodiesterase inhibitors. MIN6 cells were pre-incubated in the presence of 10 μmol/l each of rolipram (R) and cilostamide (C), followed by incubation in the absence and presence of 100 nmol/l exendin-4 (Ex). Incubation was for 1 h (a) or for the indicated times (b). c, d MIN6 cells were also pre-incubated in the presence of 10 μmol/l H89, an inhibitor of protein kinase A, followed by exposure to exendin-4 and phosphodiesterase inhibitors as indicated. The cell lysates were processed for determination of cyclic AMP levels by ELISA (a) or for western blot analysis of PCREB and CREB (b, c) or CXCL10 (d). **p < 0.01 and ***p < 0.001 with respect to untreated control

Full size image

Protein kinase A is not involved in cyclic AMP-mediated suppression of CXCL10 levels in MIN6 cells

To delineate the role of PKA-dependent and -independent pathways of cyclic AMP action, we used the PKA inhibitor, H89. Inhibition of PKA decreased cyclic AMP-mediated CREB phosphorylation, as expected (Fig. 8c). However, suppression of IFN-γ-stimulated CXCL10 production by exendin-4 and phosphodiesterase inhibitors was not reversed by H89 (Fig. 8d). This observation suggests that the cyclic AMP-mediated decrease of CXCL10 is unlikely to proceed through PKA. Surprisingly, in human islets, CXCL10 suppression by cyclic AMP was further enhanced by H89 (results not shown), probably via inhibition of a feedback regulation by PKA.

Discussion

Chemokines play an important role in the initiation of insulitis in type 1 diabetes and in islet graft rejection [12, 13, 23, 24]. In this study, we demonstrate that exendin-4-mediated suppression of CXCL10 expression is enhanced by phosphodiesterase inhibitors in human islets and in a mouse beta cell line. Similar and stronger suppression of CXCL10 was observed in islets treated with: (1) forskolin, which generates cyclic AMP by activation of adenylyl cyclase; and (2) DBC, a non-degradable cyclic AMP analogue. These agents also decreased levels of STAT-1, a transcription factor through which IFN-γ induces CXCL10. PKA does not appear to play a role in mediating the action of cyclic AMP on CXCL10. The anti-inflammatory action of the GLP-1-cyclic AMP pathway was also seen with several other inflammatory mediators generated by human islets. These findings have potential implications for the treatment of diabetes and improvement of islet transplantation outcomes.

IFN-γ was found to be a strong inducer of CXCL10, whereas IL-1β and TNF-α did not increase expression of CXCL10 in human islets (Fig. 4). It is surprising that IL-1β and TNF-α, which are known to induce CXCL10 in other cell types [25], did not have a significant effect on this chemokine in human islets. However, there was a strong synergy when a combination of these three cytokines was used [18]. The synergy between cytokines could be due to the involvement of multiple transcription factors in the induction of CXCL10. Majumder et al. (1998) have demonstrated the role of STAT-1 and NF-kB in the induction of CXCL10 by IFN-γ and TNF-α [22]. A pathway-specific gene expression profiling analysis revealed induction of a panel of chemokines (CCL2, CCL3, CCL4, CCL5, CCL7, CCL8, CCL11, CCL20, CXCL1, CXCL2, CXCL3, CXCL5, CXCL6, CXCL9, CXCL10, CXCL11) and cytokines (IL17C, IL1A, IL1B, IL8 and TNF-α) in human islets following exposure to IL-1β, TNF-α and IFN-γ. The transcription factors NF-kB and STAT-1 seem to be involved in the induction of all these inflammatory mediators except CCL2, CCL7 and IL8. The anti-inflammatory action of exendin-4 and phosphodiesterase inhibitors was seen with CCL2, CCL3, CCL4, CCL7, CCL20, CXCL6, CXCL9, CXCL10, CXCL11, IL1B and IL8. Islet-generated chemokines could play a role in beta cell death in addition to causing insulitis. Impairment of beta cell function in human islets exposed to CXCL10 has been reported [16].

GLP-1 is known to increase proliferation and neogenesis of beta cells, and to prevent their death by apoptosis [1]. Inhibitors of dipeptidyl peptidase-IV, which is known to increase circulating levels of GLP-1, and degradation-resistant GLP-1 analogues are being used in the treatment of type 2 diabetes to preserve beta cell mass [26]. Anti-inflammatory action of GLP-1 has been suggested by Iwai et al. (2006), who reported the inhibition of lipopolysaccharide-induced IL-1β production in cultured rat astrocytes by GLP-1 through generation of cyclic AMP [27]. Exendin-4 has also been reported to decrease microglia-derived TNF-α and IL-1β in mouse brain [28]. Several in vivo studies have reported the beneficial effects of exendin-4 in NOD mice. For example, chronic administration of exendin-4 to NOD mice leads to modest and significant decrease in insulitis [29]. Continuous infusion of GLP-1 also delayed the onset of diabetes in NOD mice [30]. Other studies have used exendin-4 in combination with anti-inflammatory agents to demonstrate a more significant blocking of insulitis and diabetes onset in NOD mice [31–33]. Furthermore, a recent study has reported that increasing the levels of circulating GLP-1 by inhibiting dipeptidyl peptidase-IV results in prolonged islet graft survival and decreased insulitis in diabetic NOD mice [34]. The mechanism of anti-inflammatory action of exendin-4 could involve downregulation of STAT-1. STAT1 is an autoregulated gene induced by IFN-γ (Fig. 6). Exendin-4 is reported to decrease basal mRNA levels of JAK1 and STAT1 in human islets and in INS-1 cells, a rat insulinoma cell line [35]. We observed here a decrease in IFN-γ-stimulated STAT-1 protein levels pre-incubated with exendin-4 and phosphodiesterase inhibitors (Figs 3c, 7d).

Induction, triggered by viral infection, of CXCL10 expression in islets has been suggested to play a role in initiating insulitis in type 1 diabetes [12]. In a lymphocytic choriomeningitis virus-induced mouse model of autoimmune diabetes, increased expression of CXCR3-binding chemokines was reported as an important early event initiating islet infiltration with immune cells [36]. Cultured human islets have been shown to secrete CXCL10 following enterovirus infection [37]. In a recent report, Tanaka et al. examined sections of pancreas from autopsy material derived from a deceased patient with fulminant type 1 diabetes, an acute subtype of type 1 diabetes identified in the Japanese population, and detected enterovirus capsid protein infiltration of CXCR3-bearing T cells and macrophages [38]. They also observed IFN-γ and CXCL10 in all subtypes of islets. Another recent study has observed the presence of CXCL10 and infiltration of lymphocytes expressing the corresponding chemokine receptor CXCR3 in islets from pancreatic tissues of deceased recent-onset type 1 diabetic patients, regardless of enterovirus infection; this suggests a significant role for this chemokine in inducing insulitis [17]. Viral infection can also cause direct injury to beta cells [39]. Coxsackievirus-induced beta cell death in cultured human islets has been reported [40]. Double-stranded RNA, generated in virus-infected cells, induces the pro-apoptotic gene FAS and synergises with cytokines in inducing apoptosis in rodent beta cells [41].

The role of inflammatory mediators in type 2 diabetes is also being recognised [42, 43]. The presence of immune cells in islets from human type 2 diabetic patients and from animal models of type 2 diabetes has been reported [44]. Chronic infiltration of diabetic islets with immune cells could play a significant role in beta cell death in type 2 diabetes through release of cytokines and chemokines. Thus our findings of an anti-inflammatory action of exendin-4 have implications for the treatment of both types of diabetes.

Targeting beta cell/islet-generated chemokines could be a potential therapeutic approach along with conventional anti-inflammatory therapy directed against immune cells. Infiltration of islets in type 1 diabetes is preceded by changes in islet milieu [45]. Frigerio et al. have reported that beta cells secrete CXCL10 in response to inflammation [46]. The anti-inflammatory action of the GLP-1-cyclic AMP pathway through decrease of levels of STAT-1 was similar in human islets and MIN6 cells (Figs 3c, 7d). Although exendin-4 did have modest independent anti-inflammatory action in selected experiments (Table 2, Figs 1a, 5b, 7a), addition of phosphodiesterase inhibitors augmented exendin-4 action significantly. Beta cell-specific inhibition of phosphodiesterase could be a potential strategy to sustain the level of cyclic AMP generated by exendin-4. Tissue-specific expression of 11 families of phosphodiesterase isozymes has been reported [47]. Pancreatic beta cells predominantly express phosphodiesterase 3 and phosphodiesterase 4 [48]. Therefore we used cilostamide and rolipram, inhibitors of phosphodiesterase 3 and 4 respectively, to sustain exendin-4-mediated cyclic AMP levels (Fig. 8). The anti-inflammatory action of pentoxifylline, a general phosphodiesterase inhibitor, has also been reported in type 1 diabetic patients [49] and in animal models of type 1 diabetes [6, 50]. Theophylline, another phosphodiesterase inhibitor, decreases the incidence of diabetes in BioBreeding rats [7]. In the presence of phosphodiesterase inhibitors, exendin-4-mediated phosphorylation of CREB at serine 133, a regulatory step needed for CREB activation, was sustained over time (Fig. 8). Previously, we have reported that CREB is needed for the anti-apoptotic effects of exendin-4 [19]. Thus the present study suggests that a combination of exendin-4 and phosphodiesterase 3 and 4 inhibitors could decrease islet inflammation in addition to improving beta cell mass in patients with diabetes.