Introduction

Type 1 diabetes results from an autoimmune destruction of pancreatic beta cells, caused by a complex interplay between beta cells and the immune system. During the early stages of insulitis, activated macrophages and T lymphocytes are attracted to the islets of Langerhans. These cells release oxygen free radicals, nitric oxide and cytokines such as IL-1β, IFN-γ and TNF-α in the vicinity of the islet cells, contributing to beta cell dysfunction and apoptosis [1].

Cytokine-induced beta cell apoptosis depends on de novo gene expression [1]. To determine the nature of the genes implicated in this process, we previously performed a series of microarray studies in cytokine-treated beta cells (reviewed in [2]). We observed that nuclear factor-κB (NF-κB) is a key regulatory transcription factor with a pro-apoptotic role in beta cells [3–5]. NF-κB also contributes to the induction of the cytokine IL-15 and the chemokines monocyte chemoattractant protein-1 (MCP-1, now known as CCL2), interferon-inducible protein-10 and intercellular adhesion molecule-1 in the beta cell [3]. These chemokines contribute to the recruitment and activation of mononuclear cells and to the development of insulitis in early type 1 diabetes and during islet graft destruction [6]. The microarray analyses of IL-1β plus IFN-γ-exposed beta cells also revealed changes in the expression of numerous genes involved in lipid metabolism, among them the transcription factor peroxisome proliferator-activated receptor (Ppar)-δ and its target genes scavenger receptor B (CD36), lipoprotein lipase (Lpl), acyl CoA synthetase and adipocyte differentiation-related protein (Adrp, also known as adipophilin or Adfp) [7].

PPAR-δ (also called PPAR-β) belongs to the PPAR nuclear receptor family, which regulates transcription by binding to peroxisome proliferator response elements (PPREs). PPAR-α and PPAR-γ control cellular lipid metabolism, and PPAR-γ also affects the activity of the proinflammatory transcription factors STAT and NF-κB in macrophages [8]. The role of PPAR-δ is less well understood. In keeping with its ubiquitous expression, PPAR-δ has been implicated in placentation, cellular proliferation and differentiation, lipid metabolism and inflammation [9]. Synthetic PPAR-δ agonists promote cholesterol uptake [10] and efflux [11] in macrophages via regulation of CD36 and the ATP-binding cassette A1. In muscle, PPAR-δ overexpression or its activation by synthetic ligands induces lipid utilisation [12–14], while in the liver PPAR-δ promotes the synthesis of NEFA [15].

In addition to these effects on lipid metabolism, it was recently shown that PPAR-δ has a role in inflammation in atherogenesis [16]. Thus, deletion of PPAR-δ in foam cells was atheroprotective through the increased availability of inflammatory repressors and the downregulation of the pro-inflammatory genes Mcp-1, IL-1β and matrix metalloproteinase 9. While unliganded PPAR-δ sequesters the transcriptional repressor protein B cell lymphoma-6 (BCL-6) and has mainly proinflammatory effects, following ligand binding BCL-6 is released from PPAR-δ and inhibits inflammatory signals. Hence, deletion of PPAR-δ exerts anti-inflammatory effects, as does PPAR-δ activation by synthetic ligands [16]. PPAR-δ was therefore suggested to control an inflammatory switch through its ligand-dependent association with and dissociation from the transcriptional repressor BCL-6 [16].

PPAR-δ is the predominant PPAR isoform in islets [17], but little is known about its role in beta cells. Based on our previous observations that cytokines induce both Ppar-δ and the expression of proinflammatory chemokines in beta cells [7], we sought to determine whether the PPAR-δ network is present and active in beta cells and whether PPAR-δ activation modulates the inflammatory response of beta cells exposed to cytokines via its association with the transcriptional repressor BCL-6.

Materials and methods

Culture of primary beta cells purified by fluorescence-activated cell sorting and of insulin-producing INS-1E cells

Male Wistar rats (Charles River Laboratories Belgium, Brussels, Belgium) were housed and utilised in accordance with the guidelines of the Belgian Regulations for Animal Care and the Principles of Laboratory Animal Care (NIH publication no. 85-23, revised 1985; http://grants1.nih.gov/grants/olaw/references/phspol.htm accessed on 4 April 2005). Rat islets were isolated by collagenase digestion followed by hand picking under a stereomicroscope. For beta cell isolation, islets were dispersed and beta cells were purified by autofluorescence-activated cell sorting (FACS, FACStar, Becton-Dickinson and Co, Sunnyvale, CA, USA) ([18] and references therein). The preparations used in the present experiments contained 93±1% beta cells. Purified beta cells were pre-cultured for 16 h in HAM’s F-10 medium containing 10 mmol/l glucose, 2 mmol/l glutamine, 50 μmol/l 3-isobutyl-1-methylxanthine, 5% heat-inactivated fetal bovine serum (FBS), 0.5% BSA, 50 U/ml penicillin and 50 μg/ml streptomycin and then changed to serum-free medium with 1% BSA [18]. The well-differentiated INS-1E beta cell line, a kind gift from C. Wollheim (Centre Medical Universitaire, Geneva, Switzerland), was cultured in RPMI-1640 (with Glutamax-1) supplemented with 10 mmol/l HEPES, 5% FBS, 100 U/ml penicillin, 100 μg/ml streptomycin, 1 mmol/l sodium pyruvate and 50 μmol/l 2-mercaptoethanol [19]. FACS-purified beta cells and INS-1E cells were exposed to the cytokines IL-1β (50 U/ml for beta cells, 10 U/ml for INS-1E cells; kindly provided by C. W. Reynolds, National Cancer Institute, Bethesda, MD, USA), IFN-γ (500 U/ml, R&D Systems, Oxon, UK) and TNF-α (1000 U/ml, Innogenetics, Gent, Belgium). Cytokine concentrations were selected based on our previous time course and dose–response studies [1, 7].

Determination of mRNA expression

Poly(A)+ RNA was isolated from purified beta cells and INS-1E cells using oligo(dT)25-coated polystyrene Dynabeads (Dynal, Oslo, Norway), and reverse transcribed as previously described [20]. Standards for each gene were amplified by PCR using appropriate primers. Real-time PCR products were quantified using SYBR Green [21], and their concentration was calculated as copies/μl by comparison with the standard curve [22]. The mRNAs analysed were: Ppar-δ, Adrp, CD36, acyl CoA synthetase, inducible nitric oxide synthase (iNOS), Mcp-1 and the inhibitor κB-α (IκB-α). The expression levels of the genes of interest were corrected for that of the house-keeping gene glyceraldehyde-3-phosphate dehydrogenase (Gapdh) [21, 22]. The specific real-time PCR primers and their respective PCR product lengths are available as electronic supplementary material (ESM). For Bcl-6, a conventional PCR was performed as described [23] using forward primer 5′-CTGAGGGAAGGCAACATCAT-3′ and reverse primer 5′-CGGCTGTTCAGGAACTCTTC-3′ yielding a PCR product of 975 bp. The ethidium bromide-stained agarose gels were photographed under UV transillumination using a Kodak Digital Science DC 290 camera (Kodak, Rochester, NY, USA).

Luciferase reporter assays

To determine whether PPAR-δ is functional in beta cells, primary beta and INS-1E cells were transfected with a PPAR-δ reporter construct containing three copies of the acyl CoA oxidase (AoX) PPRE cloned upstream of the gene encoding the thymidine kinase luciferase reporter [24]. In parallel, cells were transfected with the parental plasmid (without PPRE). To evaluate the effect of PPAR-δ activation on cytokine-induced MCP-1 and NF-κB activation, we used, respectively, the pMCP-1-514(enh) luc construct containing two NF-κB binding sites [25] and the NF-κB reporter construct (BD biosciences Clontech, Palo Alto, CA, USA). In all conditions, cells were co-transfected with the internal control plasmid pRL-CMV (Promega, Madison, WI USA). Transfections were performed by 5 h lipofection with lipofectAMINE 2000 (Invitrogen, Paisley, UK), as previously described [26], using 0.5 μl lipofectamine, 135 ng test plasmid and 15 ng pRL-CMV for primary beta cells, and 250 and 25 ng of the respective plasmids for INS-1E cells. The cells were subsequently exposed for 6–24 h to GW501516 or cytokines. In a second series of experiments, 250 ng haemagglutinin-tagged BCL-6 expression or control plasmids (a kind gift from A. L. Dent, Indiana University School of Medicine, IN, USA) were co-transfected with the Mcp-1 promoter or NF-κB reporter and pRL-CMV plasmids. Luciferase activities were assayed with the Dual-Luciferase Reporter Assay System (Promega). Test values were corrected for the luciferase activity value of the internal control pRL-CMV. The different conditions tested did not affect pRL-CMV activity.

Rat islet insulin secretion

Rat islets were exposed to GW501516 (1 μmol/l) for 24 h, after which their acute glucose-induced and palmitate-induced insulin release was measured as previously described [27]. Briefly, triplicate groups of ten islets were incubated for 90 min in Krebs–Ringer bicarbonate buffer supplemented with 0.5% BSA and the secretagogues indicated in Table 1. Insulin release was corrected per islet insulin content and expressed as fold-increase compared to basal insulin release at 1.7 mmol/l glucose.

Table 1 Rat islet insulin secretion
Full size table

Determination of cell viability

To determine the effect of GW501516 on cytokine-induced beta cell apoptosis, the percentage of viable, apoptotic and necrotic cells was determined after 48 h of cytokine exposure. The treatment with GW501516 was initiated 12 h before the addition of IL-1β (50 U/ml) and IFN-γ (500 U/ml). To assess cell viability, FACS-purified beta cells were incubated for 15 min with the DNA-binding dyes propidium iodide (10 μg/ml; Sigma, Bornem, Belgium) and Hoechst 33342 (10 μg/ml; Sigma) [5, 7, 28]. In each experimental condition a minimum of 500 cells was counted by at least two observers, one of them blinded for the sample identity. There was >90% agreement between observers in the different assessments of viability, confirming the reproducibility of the method.

Statistical analysis

Data are presented as means±SE. Comparisons were performed by two-sided paired Student’s t-test or ANOVA followed by paired t-test with the Bonferroni correction for multiple comparisons. A p-value <0.05 was considered statistically significant.

Results

Cytokines induce Ppar-δ mRNA expression in pancreatic beta cells

INS-1E cell exposure to TNF-α or IL-1β, alone or in combination with IFN-γ, increased Ppar-δ expression (Fig. 1a). The induction occurred early (at 6 h) and was most pronounced (fivefold increase) for the combination of IL-1β and IFN-γ. Ppar-δ expression was also induced in FACS-purified primary rat beta cells exposed to IL-1β and IFN-γ for 6 h (2.2-fold, Fig. 1b), while mRNA levels returned to control levels after 24 h in both cell types. These data are in good agreement with our previous microarray analysis observations [7].

Fig. 1
figure 1

Cytokines induce Ppar-δ expression in beta cells. INS-1E (a) or FACS-purified primary beta cells (b) were cultured for 6–24 h under control conditions (empty bars), or with TNF-α (dotted bars), IL-1β (grey bars) or IL-1β + IFN-γ (black bars). Ppar-δ mRNA expression was analysed by real-time PCR and the data were normalised for the expression of the housekeeping gene Gapdh. Data represent means±SE of three to eight independent experiments. *p<0.05, **p<0.01, ***p<0.001 vs control

Full size image

Cytokines induce PPAR-δ target gene expression in beta cells

CD36 mRNA expression increased twofold in INS-1E and primary beta cells treated for 6 or 24 h with IL-1β alone or in combination with IFN-γ (Fig. 2a,b). INS-1E Lpl expression levels increased three- to sixfold following exposure to IL-1β or IL-1β + IFN-γ (Fig. 2c,d). In beta cells, Lpl induction was observed with IL-1β but not with IL-1β + IFN-γ. IL-1β and/or IFN-γ also increased acyl CoA synthetase mRNA levels at 6 and 24 h by two- to sixfold in INS-1E cells, and by 1.5- to twofold in primary beta cells (Fig. 2e,f). Finally, we observed increased Adrp mRNA expression in INS-1E cells exposed to TNF-α or IL-1β + IFN-γ, and in primary beta cells treated with IL-1β alone or in combination with IFN-γ (Fig. 2g,h). There was some variability in the timing and magnitude of PPAR-δ target gene induction in INS-1E and primary beta cells but as a whole these data suggest that the Ppar-δ mRNA induction at 6 h is followed by upregulation of downstream PPAR-δ target genes at 6 and 24 h.

Fig. 2
figure 2

Cytokines induce PPAR-δ target gene expression in beta cells. INS-1E (a, c, e, g) or FACS-purified primary beta cells (b, d, f, h) were cultured for 6–24 h under control conditions (empty bars), or with TNF-α (dotted bars), IL-1β (grey bars) or IL-1β+IFN-γ (black bars). CD36, Lpl, acyl CoA synthetase and Adrp mRNA expression was analysed by real-time PCR and the data were normalised for the expression of the housekeeping gene Gapdh. Data represent means±SE of three to eight independent experiments. *p<0.05, **p<0.01, ***p<0.001 vs control

Full size image

Cytokines activate the PPAR-δ nuclear receptor in beta cells

Following transfection with the AoX-PPRE luciferase reporter, primary beta cells were exposed for 12–24 h to IL-1β + IFN-γ or to the synthetic PPAR-δ agonist GW501516 (0.1–1 μmol/l), used as a positive control. Cytokines activated PPAR-δ, as indicated by the fourfold increase in luciferase activity after 12 h, which fell to about twofold after 24 h (Fig. 3a,b). GW501516 treatment led to a twofold increase in luciferase activity after 12 and 24 h at a concentration of 1 μmol/l (Fig. 3). This concentration of GW501516, previously shown to activate PPAR-δ highly selectively but not the other PPAR isoforms [11], was therefore selected for further experiments. Similar results were obtained in INS-1E cells (data not shown). No effect was observed when the parental thymidine kinase luciferase plasmid lacking the PPRE was used (Fig. 3), or with the PPAR-α agonist gemfibrozil (300 μmol/l, data not shown), confirming the specificity of the results.

Fig. 3
figure 3

Cytokines and GW501516 activate the PPAR-δ nuclear receptor in beta cells. Primary beta cells were co-transfected with the PPRE-containing acyl CoA oxidase luciferase reporter construct and the internal control pRL-CMV encoding Renilla luciferase (black bars). As negative controls, cells were transfected with the parental thymidine kinase luciferase plasmid lacking the PPRE (empty bars). After transfection, the cells were exposed for 12 h (a) and 24 h (b) to IL-1β + IFN-γ or GW501516 (GW 0.1–1 μmol/l) and assayed for firefly and Renilla luciferase activities. The results were normalised for Renilla luciferase activity and are means±SE of four or five independent experiments. *p<0.05 vs control

Full size image

PPAR-δ activation does not affect insulin secretion by beta cells

Glucose-induced and/or NEFA-induced insulin release from islets was measured following a 24-h exposure to 1 μmol/l GW501516. Insulin release to the medium by GW501516-treated islets was not different from that by control islets over the 24-h culture period (93±13% of control). Glucose (16.7 mmol/l) or glucose (16.7 mmol/l) plus palmitate (0.5 mmol/l) induced a four- to sevenfold increase in insulin release as compared to 1.7 mmol/l glucose, and this was not modified by GW501516 (Table 1). GW501516 also failed to affect insulin release at 5.6 mmol/l glucose (Table 1). Insulin release at 1.7 mmol/l glucose was similar in control and GW501516-treated islets (data not shown).

PPAR-δ activation does not prevent Mcp-1 expression or NF-κB activation in beta cells

IL-1β + IFN-γ treatment led to a clear induction of Mcp-1 mRNA levels in INS-1E cells (Fig. 4). The PPAR-δ agonist GW501516 did not affect Mcp-1 expression per se and did not prevent the cytokine effects (Fig. 4). Similarly, basal Mcp-1 promoter activity and activation of the reporter construct by TNF-α, IL-1β or IL-1β + IFN-γ were not modified by GW501516 (data not shown).

Fig. 4
figure 4

PPAR-δ activation does not modify cytokine-induced MCP-1 expression in beta cells. Mcp-1 mRNA expression was analysed by real-time PCR and normalised for the expression of the housekeeping gene Gapdh. INS-1E cells were cultured for 10 h in the presence or absence of GW501516 (1 μmol/l) before IL-1β+IFN-γ exposure in the continuous presence of GW501516 or control condition for 6 h (n=3). *p<0.05 vs control

Full size image

In INS-1E cells transfected with an NF-κB reporter construct, GW501516 did not affect basal luciferase activity or the tenfold cytokine-induced increase in luciferase activity (data not shown). In line with these observations, GW501516 did not inhibit the cytokine-mediated mRNA induction of the NF-κB-regulated genes iNOS and IκB-α (data not shown). We also evaluated the effect of PPAR-δ activation on cytokine-induced primary beta cell apoptosis. The combination of IL-1β and IFN-γ induced apoptosis in 19±2% of the beta cells after 48 h, and this was not affected by GW501516 (21±3% apoptosis, n=3). Apoptosis in control beta cells was 3±1% (n=3), and this was not changed by GW501516 (4±1%).

Pancreatic beta cells do not express the transcriptional repressor BCL-6, but transient BCL-6 expression inhibits the pro-inflammatory response in beta cells

PPAR-δ inhibits MCP-1 expression in macrophages by virtue of its ligand-dependent dissociation from the transcriptional repressor BCL-6 [16]. Because GW501516 did not decrease Mcp-1 induction (Fig. 4), we examined whether pancreatic islets, FACS-purified primary beta cells or INS-1E cells express BCL-6. Bcl-6 was not detected in any of these cells, whereas a clear expression was observed in splenic tissue, used as positive control (Fig. 5).

Fig. 5
figure 5

Beta cells do not express the transcriptional repressor BCL-6. Rat islets, FACS-purified primary beta cells and INS-1E cells were cultured in the presence or absence of IL-1β+IFN-γ for 24 h. Bcl-6 expression was assessed by PCR and Gapdh was used as a housekeeping gene. Rat spleen was used as a positive control. One representative experiment for three similar experiments is shown

Full size image

To test whether the lack of constitutive BCL-6 expression could explain the defective beta cell response to the PPAR-δ agonist, we transiently transfected INS-1E cells with a BCL-6 expression plasmid or an empty control vector, both containing a haemagglutinin tag. After 24 to 48 h, haemagglutinin immunocytochemistry confirmed nuclear staining in the cells transfected with the BCL-6 construct (data not shown). Following co-transfection with the Mcp-1 promoter reporter (Fig. 6a), BCL-6 reduced basal and cytokine-induced Mcp-1 promoter activity. This inhibitory effect appeared to be mediated through inhibition of NF-κB, because BCL-6 effectively prevented cytokine-mediated NF-κB activation in a co-transfection experiment with the NF-κB reporter (Fig. 6b). Basal NF-κB activity was decreased to 40% of control, and cytokine-induced activation to 25%.

Fig. 6
figure 6

BCL-6 decreases cytokine-induced MCP-1 promoter and NF-κB activity in beta cells. a MCP-1 activation was examined using the Mcp-1 promoter luciferase reporter construct transfected in parallel with the BCL-6 expression construct (black bars) or the empty control vector (empty bars) in INS-1E cells exposed to cytokines for 14 h. The results were normalised for Renilla luciferase activity and are means±SE of five independent experiments. b NF-κB activation was examined using a luciferase reporter construct transfected in parallel to the BCL-6 expression construct (black bars) or the empty control vector (white bars) in INS-1E cells exposed to cytokines for 6 h. The results were normalised for Renilla luciferase activity and are means±SE of four independent experiments. *p<0.05, **p<0.01 vs control; +p<0.05, ++p<0.01 vs empty vector

Full size image

Discussion

In previous microarray studies of INS-1E cells exposed to IL-1β + IFN-γ, we observed the induction of a large set of genes involved in lipid metabolism, among them the transcription factor Ppar-δ and several of its target genes [7]. In the present study, we confirm by real-time PCR that the PPAR-δ gene network is induced by cytokines in INS-1E cells, and we extend these findings to primary rat beta cells. The upregulation of Ppar-δ mRNA is functional, as suggested by the PPRE reporter activation by cytokines and the induction of the PPAR-δ target genes CD36, Lpl, acyl CoA synthetase and Adrp. Because PPAR-δ was recently shown to control an inflammatory switch via its interaction with the transcriptional repressor BCL-6 [16], we examined whether PPAR-δ induction and activation regulates the cytokine-mediated pro-inflammatory response in pancreatic beta cells. At variance with immune cells, the synthetic PPAR-δ agonist GW501516 did not affect the expression and promoter activity of the chemokine MCP-1, probably because the transcriptional repressor BCL-6 is not expressed in beta cells. Neither did PPAR-δ activation inhibit cytokine-mediated NF-κB activation, nor did it protect beta cells against apoptotic cell death. BCL-6 expression by transient transfection inhibited the beta cell pro-inflammatory responses.

Of the three isoforms of the PPAR family, PPAR-α, -γ and -δ, PPAR-δ is the predominant form in pancreatic islets [17]. Its expression levels increase when beta cells are exposed to cytokines, while no change was observed in PPAR-α or -γ expression [7]. The role of PPAR-δ in beta cells is unknown, but PPAR-α is induced and promotes NEFA oxidation in beta cells at low glucose levels, i.e. during fasting and starvation, and decreases glucose-stimulated insulin secretion [29], while PPAR-γ overexpression in islets also impairs glucose-stimulated insulin secretion [30, 31]. On the other hand, and at variance with the mention of increased islet insulin release in another report [14], PPAR-δ activation by its synthetic ligand GW501516 did not affect glucose- and NEFA-stimulated insulin secretion (present data).

In immune cells, PPAR-δ controls the inflammatory responses through its association (pro-inflammatory) and disassociation (anti-inflammatory) with the transcriptional repressor BCL-6 [16]. As in the case of MCP-1, BCL-6 can act directly or indirectly (via binding to other factors such as IRF and STAT) to repress chemokine expression [32]. We have previously demonstrated the importance of two NF-κB binding sites in the enhancer region of the Mcp-1 promoter, while the proximal promoter region with putative NF-κB or AP-1 sites, a γ-activated sequence and an IFN response-inhibitory sequence contributed little to the regulation of MCP-1 expression in pancreatic beta cells [25]. BCL-6 can also negatively regulate NF-κB [33] by repressing its transcription and inhibiting its nuclear binding activity. Independent of BCL-6, PPARs may also exert anti-inflammatory effects via cross-talk with NF-κB. Recent data suggest that NF-κB activation may lead to PPAR-δ inhibition in cardiomyocytes, through physical interaction between the p65 subunit of NF-κB and PPAR-δ [34]. Conversely, synthetic ligand-mediated PPAR-δ activation can inhibit NF-κB activation [35]. These regulatory mechanisms vary between tissues, as illustrated by the differences in PPAR-δ-mediated repression of MCP-1 expression in liver and aorta [36].

In pancreatic beta cells, cytokines activate both NF-κB and PPAR-δ, but PPAR-δ agonists did not decrease basal or cytokine-stimulated NF-κB activity, suggesting limited cross-talk between both pathways. Perhaps more importantly, and in contrast to immune cells, we found that beta cells do not express the transcriptional repressor BCL-6, and we demonstrated that transient BCL-6 expression inhibits these pro-inflammatory responses. The lack of BCL-6 expression, and PPAR-δ-mediated repression of inflammatory responses, may render the beta cell particularly prone to propagation of inflammation through uninhibited NF-κB activation and MCP-1 expression. Stable transfection of beta cells with BCL-6 in vitro may restrict their chemokine release in the acute inflammatory reaction following islet transplantation and in the case of recurrent autoimmune disease and could be a means to prevent the development and amplification of insulitis.

In conclusion, PPAR-δ activation does not regulate the pro-inflammatory and pro-apoptotic response of beta cells to cytokines during insulitis in type 1 diabetes because beta cells lack BCL-6. Additional experiments, using in vivo models of prolonged PPAR-δ activation and BCL-6 expression, are required to clarify the role of PPAR-δ and BCL-6 in beta cell physiology and to establish their therapeutic potential in type 1 diabetes.