Introduction

Insulin deficiency and insulin resistance contribute to the development of type 2 diabetes. Pancreatic beta cell dysfunction is present early in the pathogenesis of type 2 diabetes and worsens over time [1], at least in part as a result of loss of functional beta cell mass [2]. By increasing insulin requirements, insulin resistance can precipitate the onset of hyperglycaemia. Both beta cell failure and insulin resistance are determined by genetic and environmental factors, the latter causing the rapidly increasing prevalence of type 2 diabetes worldwide. Sedentary lifestyle and energy-dense diets rich in refined sugars and saturated fats are known risk factors for type 2 diabetes, but the mechanisms involved are only partially understood.

Glucotoxicity and lipotoxicity are terms coined for the deleterious effects of high circulating concentrations of glucose and lipids, both at the level of insulin sensitivity and in beta cells. Chronically elevated glucose concentrations affect beta cell function and survival through increased generation of reactive oxygen species and mitochondrial dysfunction, endoplasmic reticulum (ER) stress, and JNK signalling [3–5]. NEFA, and in particular saturated NEFA, impair beta cell function and cause apoptosis through ceramide synthesis, JNK activation and oxidative and ER stress [5–9]. Saturated NEFA cause ER stress by depleting ER calcium, which secondarily affects protein folding in the organelle [10]. Human islets exposed in vitro to palmitate activate the three branches of the ER stress response under the control of PRKR-like endoplasmic reticulum kinase (PERK), activating transcription factor 6 (ATF6) and inositol-requiring 1 (IRE1) [10]. Markers of oxidative [11] and ER stress [8, 12, 13] are present in islets of type 2 diabetic patients, suggesting these stress responses may indeed play a role in the development and progression of the disease.

Mild inflammation by activation of the innate immune system has been suggested to play a role in the pathogenesis of type 2 diabetes, and it is present years before disease onset [14]. In the adipose tissue of obese individuals, infiltrating macrophages produce cytokines, modulate adipocyte adipokine secretion and contribute to the development of insulin resistance [15]. The spill over of inflammatory cytokines into the circulation has been associated with the development of type 2 diabetes. While the single increase in circulating IL-6 does not enhance disease risk, the combined elevation of IL-6 and IL-1β is associated with a threefold increased risk for type 2 diabetes [16]. A new role for inflammation in type 2 diabetes has been recently proposed, namely as a trigger of beta cell dysfunction and death [17]. Macrophage infiltration was observed in islets from type 2 diabetic individuals compared with non-diabetic (ND) individuals, with 20% of the type 2 diabetic islets containing >3 CD68+ cells per islet compared with 5% of the ND islets [18, 19]; the role of these macrophages is unknown. It has been proposed that beta cells contribute to the inflammatory process in type 2 diabetes by their own secretion of IL-1β in response to high glucose. In a series of studies, it was suggested that glucotoxicity in human beta cells is mediated by Fas (TNF receptor superfamily, member 6) receptor upregulation and nuclear factor κB (NF-κB) activation, culminating in beta cell dysfunction and apoptosis [20, 21], though this was not confirmed by others [22, 23]. Altogether, the cause and role of islet inflammation in type 2 diabetes remains controversial and poorly understood.

We have used human pancreatic beta cells and islets from ND and type 2 diabetic individuals to answer the following questions: (1) is an inflammatory process present in type 2 diabetic beta cells, as evaluated by array analysis of laser-capture microdissected beta cells from type 2 diabetic and normoglycaemic individuals; (2) do NEFA and/or high glucose contribute to this process; (3) by which signalling events and cellular response mechanisms do saturated NEFA induce an inflammatory response in human islets; and (4) does NEFA-induced inflammation contribute to lipotoxicity? The results demonstrate the presence of mild inflammation in islets from type 2 diabetic individuals, a process that is mimicked by in vitro exposure of ND human islets to palmitate. The saturated-NEFA-induced inflammation could be a consequence of NEFA-induced ER stress and does not directly contribute to human islet lipotoxicity.

Methods

Laser-capture microdissection and microarray analysis

Pancreatic samples from ten ND multiorgan donors (age 60 ± 2 years, four women/six men, BMI 30.6 ± 1.6 kg/m2) and ten type 2 diabetic donors (age 67 ± 7 years, three women/seven men, BMI 30.9 ± 6.2 kg/m2, known duration of diabetes 5.3 ± 2.3 years, insulin/oral therapy 33%/66%) were obtained with approval of the Ethics Committee in Pisa, Italy. The causes of death were similar. The available clinical characteristics of the donors are provided in Electronic supplementary material (ESM) Table 1; no data were available on levels of circulating lipids or inflammatory markers. Laser-capture microdissection (LCM) was performed as previously detailed [24, 25]. Briefly, frozen pancreatic sections were fixed, dehydrated and air-dried. LCM was performed using PixCell II Laser Capture Microdissection System (Arcturus Engineering, Mountain View, CA, USA) by melting thermoplastic films mounted on transparent LCM caps (Arcturus) on beta cells, identified by their intrinsic autofluorescence, in islets with no signs of amyloid deposits. The microdissected beta cells were incubated in guanidine isothiocyanate extraction buffer and RNA was extracted by modification of the RNA micro-isolation protocol [24]. The total RNA underwent two amplification rounds using the RiboAmp HS RNA Amplification Kit (Arcturus) and was biotinylated using the BioArray High Yield RNA Transcript Labeling Kit (Enzo Diagnostics, Farmingdale, NY, USA). RNA products were fragmented and hybridised to GeneChip Human X3P Array (Affymetrix, Santa Clara, CA, USA). Array data were normalised and analysed using DNA-Chip Analyzer (dChip) software (available from http://biosun1.harvard.edu/complab/dchip/) that assesses the standard errors for the expression indices and calculates confidence intervals for fold changes. Lower confidence bound (used at a cut-off value of 1.2), a conservative estimate of the fold change, and the p value were used to assess differentially expressed genes. The complete array data will be reported elsewhere (L. Marselli, D. C. Sgroi, J. Thorne, S. Dahiya, A. Sharma, S. Bonner-Weir, P. Marchetti and G.C. Weir, unpublished results).

Pancreatic islet preparation and culture

With the approval of the Ethics Committee in Pisa, Italy, human islets were isolated from 12 ND donors (age 58 ± 6 years, seven women/five men, BMI 24.1 ± 0.7 kg/m2) by collagenase digestion and density gradient purification [9], cultured in M199 medium (5.5 mmol/l glucose) and shipped to Brussels, Belgium, within 1–5 days of isolation. After overnight recovery in Ham’s F-10 containing 6.1 mmol/l glucose, 10% (vol./vol.) FCS, 2 mmol/l GlutaMAX (Gibco, Invitrogen, Merelbeke, Belgium), 50 μmol/l 3-isobutyl-1-methylxanthine, 1% (wt/vol.) charcoal-absorbed BSA (Boehringer, Indianapolis, IN, USA), 50 U/ml penicillin and 50 μg/ml streptomycin, islets were exposed to NEFA in the same medium without FCS and a glucose concentration of 6.1 or 28 mmol/l [10]. The percentage of beta cells, examined by insulin immunofluorescence (see below), in the islet preparations was 50 ± 3% (range 35% to 70%, n = 12).

Oleate and palmitate (sodium salt, Sigma, Bornem, Belgium) were dissolved in 90% (vol./vol.) ethanol and diluted 1:100 to a final concentration of 0.5 mmol/l, corresponding to a NEFA/BSA ratio of 3.4 [10, 26]. The control condition contained 1% charcoal-absorbed BSA and a similar ethanol dilution. IL-1 receptor antagonist (IL-1ra; R&D Systems, Abingdon, UK) was used alone, or in combination with palmitate, at a concentration of 30 ng/ml. In combination with recombinant human IL-1β (50 U/ml) and IFN-γ (1,000 U/ml; PeproTech, Paris, France), 300 ng/ml IL-1ra was used (>100-fold concentration compared with the added IL-1β) and added 30 min prior to cytokines [27]. Recombinant murine TNF-α (Innogenetics, Gent, Belgium) was used at a concentration of 1000 U/ml. Cyclopiazonic acid (CPA; Sigma) was used at 50 μmol/l [10] and salubrinal (ChemBridge, San Diego, CA, USA) at 75 μmol/l [28].

Human ductal cell lines

Pancreatic ductal adenocarcinoma, Capan-2, and epithelioid carcinoma, Panc-1, cell lines [29] were a kind gift of L. Bouwens (Vrije Universiteit Brussel, Brussels, Belgium). Capan-2 cells were cultured in Advanced-RPMI medium (Gibco, Paisley, UK) containing 10% (vol./vol.) FCS, 2 mmol/l GlutaMAX and antibiotics (as above). Panc-1 cells were cultured in DMEM (without pyruvate; Gibco) containing 10% (vol./vol.) serum, MEM non-essential amino acids (Sigma) and antibiotics. For the NEFA treatment, medium contained 1% (vol./vol.) FCS and 1% (wt/vol.) BSA.

Assessment of cell death

The percentage of islet cell death was determined by at least two observers (one unaware of sample identity) using inverted fluorescence microscopy following staining with propidium iodide (5 μg/ml; Sigma) and Hoechst 33342 (10 μg/ml; Sigma) as described [10, 30].

Real-time PCR

Poly(A)+ RNA was isolated from human islets and reverse transcribed. The real-time PCR amplification was done using iQ SyBR Green Supermix on iCycler MyiQ Single Color (BIO-RAD, Hercules, CA, USA) as described [10]. The gene expression level was normalised to the housekeeping gene β-actin, expression of which is not modified by the treatments and comparable with that of ornithine decarboxylase antizyme 1 (OAZ1; ESM Fig. 1). Primer sequences are given in ESM Tables 2 and 3.

Immunofluorescence

Dispersed human islet cells were stained with anti-p65 (Santa Cruz Biotechnology, Santa Cruz, CA, USA) and anti-insulin antibodies (Sigma) as described in the ESM. To determine the islet cell type(s) producing chemokine (C-C motif) ligand (CCL)2 pancreatic specimens were stained by double immunofluorescence using anti-CCL2 (Abcam, Cambridge Sciences Park, Cambridge, UK), anti-insulin (Dako, Glostrup, Denmark), and anti-glucagon (R&D Systems) antibodies and analysed with a Leica TCS SP5 laser scanning confocal microscope (see ESM).

Insulin RIA

Insulin accumulated in the culture medium was measured as described [30] and normalised to β-actin mRNA expression.

Chemokine/cytokine ELISA

Human chemokine (C-X-C motif) ligand 1 (CXCL1) and IL-6 protein were measured in the culture medium by ELISA following the manufacturer’s instructions (Quantikine, R&D Systems, Minneapolis, MN, USA) and normalised to β-actin mRNA expression.

Statistical analysis

Data are means±SEM. Real-time PCR data were log transformed to obtain normal distribution and compared by ANOVA followed by paired t test with the Bonferroni correction for multiple comparisons. Microarray data were analysed by Mann–Whitney test. Linear regression was done using the least squares method. Multiple regression models were built to determine whether the association between a dependent and an independent variable remained significant after adjusting for other independent variables. A p value < 0.05 was considered significant.

Results

Microarray profiling of human beta cells reveals increased cytokine and chemokine expression in type 2 diabetes

Human beta cells obtained from type 2 diabetic donors by laser-capture microdissection had increased mRNA expression of chemokines and cytokines compared with ND controls. The expression of CCL2 (also known as MCP-1) and CCL13 (also known as MCP-4) was threefold higher in type 2 diabetic beta cells (p < 0.05, Fig. 1), while CCL11 (eotaxin), CXCL1 (Gro-α) and IL6 expression tended to be higher (Fig. 1).

Fig. 1
figure 1

Expression of pro-inflammatory chemokines and cytokines in human beta cells from type 2 diabetic patients and ND controls. Microarray analysis of LCM human beta cells from ten ND and ten type 2 diabetic (T2D) donors. Individual mRNA levels (expressed as model based expression indices, MBEI) and means ± SEM are shown in black/white and grey symbols, respectively. a CCL11, p = 0.063; b CCL13, p = 0.001; c CXCL1, p = 0.089; d IL6, p = 0.48; e CCL2, p = 0.015; f IL8, p = 0.063; g IL1=β, p = 0.105; h IFNG, p = 0.91

Full size image

Array data on IL1-β (also known as IL1B) and IL8 from our LCM samples has already been reported and confirmed by real-time PCR [20]. These previously published data were obtained for nine ND and ten type 2 diabetic patients; one of the ten controls with no type 2 diabetes history was excluded because of some high glycaemia readings during the antemortem period. Values were obtained from an initial analysis in which filtering criteria were used and the signals and statistical results of 30,990 probe sets were retrieved. The published results for IL1-β were 28 ± 3.9 vs 58 ± 14.2 (p = 0.032) and for IL8 22.2 ± 3.8 vs 50.7 ± 12.9 (p = 0.034) for ND vs type 2 diabetic patients, respectively [20]; for all ten ND, IL1-β was 27 ± 3.7 (p = 0.057 vs type 2 diabetes) and IL8 22.2 ± 3.8 (p = 0.059).

For our present study of ten ND controls and ten type 2 diabetic patients, analysis was performed to obtain the expression signal and the statistical results of all the probe sets present in the chip, returning a list of 61,295 probes, and no individuals were excluded. With this approach, data for IL1-β were 28.3 ± 5.7 vs 53.2 ± 12.2 and for IL8 22.7 ± 4.3 vs 47.3 ± 11.4 (Fig. 1). Interferon-γ (IFNG) expression was similar in type 2 diabetes and controls. The cytokine and chemokine expression in type 2 diabetic beta cells varied considerably between individuals, suggesting that an inflammatory response was present in islets from some but not all donors. The presence of CCL2 protein was detected in insulin-positive beta cells in pancreatic tissue sections from ND and type 2 diabetic donors using confocal microscopy, whereas little or no CCL2 was detected in glucagon-positive alpha cells or exocrine cells (Fig. 2).

Fig. 2
figure 2

CCL2 co-localises with insulin in human islets. Confocal microscopy analysis of insulin- (a, g) and CCL2-stained (b, h) sections of pancreas from ND donors (a–c) and type 2 diabetic patients (g–i) shows predominant presence of CCL2 protein in insulin-positive cells. In the merged images (c, i), the superposition of insulin (green) and CCL2 (red) results in a yellow colour, while Hoechst 33342 stains nuclei blue. In contrast, in glucagon- (d, j) and CCL2-stained (e, k) sections of pancreas from ND donors (d–f) and type 2 diabetic patients (j–l) there is little or no CCL2 protein in glucagon-positive cells. In the merged images (f, l), there is no yellow colour seen when glucagon (green) and CCL2 (red) are superposed. Images are representative of 85 islets examined in pancreases from three ND donors and 60 islets in pancreas samples from two type 2 diabetic patients. The lower panels (m–o) show negative control sections probed with anti-rabbit Alexa 594 in the absence of primary antibody: m Alexa 594; n Hoechst; o merged

Full size image

Palmitate, but not oleate or high glucose, induces expression of chemokines and cytokines in human islets

We next examined whether metabolic stress plays a role in islet inflammation. The mild inflammatory response observed in type 2 diabetic beta cells was mimicked in vitro by ND human islet exposure to palmitate. Palmitate increased the mRNA expression of CXCL1, IL6, CCL2, IL8, IL1-β and TNF-α (also known as TNF) by two- to tenfold after 24 h and 48 h (Fig. 3). In contrast, oleate did not induce any of the chemokines and cytokines (Fig. 3). The mRNA induction by palmitate was paralleled by protein secretion by the human islets. Palmitate induced secretion of IL-6 and CXCL1 by two- to threefold after 24 and 48 h (Fig. 4). The induction of cytokines by palmitate was paralleled by the upregulation of nuclear factor κB inhibitor, α (IκBα; Fig. 3), a direct target of the transcription factor NF-κB. At variance with a previous report describing IL-1β induction in human islets cultured for 44 h at 33 mmol/l glucose [21], 28 mmol/l glucose neither induced chemokine/cytokine mRNA/protein production nor IκBα expression (Figs 3 and 4), while it did increase insulin release to the culture medium (Fig. 4). This is in line with our previous report on high-glucose-treated human islets [23]. Moreover, high glucose did not potentiate the effects of palmitate (Figs 3 and 4). We have previously shown that neither NEFA nor high glucose induces chemokines or cytokines in rodent insulin-secreting cells [7], suggesting that the observed pro-inflammatory effects of palmitate are present in human but not in rodent islets of Langerhans.

Fig. 3
figure 3

Palmitate, but not oleate or high glucose, induces chemokine and cytokine mRNA expression in human islets. Human islets from ND donors were cultured for 24 h (a–g) or 48 h (h–n) in the absence (Ct) or presence of oleate (Ol) or palmitate (Pal) at 6 or 28 mmol/l glucose (G). Expression of mRNA was measured by real-time PCR, corrected by β-actin mRNA and expressed as fold induction compared with control (Ct, 6 mmol/l glucose). a, h CXCL1; b, i IL6; c, j CCL2; d, k IL8; e, l IL1-β; f, m TNF-α; g, n IκBα. The results are means ± SEM of six to nine independent experiments. *p < 0.05; **p < 0.01; ***p < 0.001 vs Ct at glucose 6 mmol/l

Full size image
Fig. 4
figure 4

Palmitate, but not oleate or high glucose, induces IL-6 and CXCL1 protein secretion by human islets. Human islets were cultured for 24 h (a, c, e) or 48 h (b, d, f) in the absence (Ct) or presence of oleate (Ol) or palmitate (Pal) at 6 or 28 mmol/l glucose (G). The chemokine and cytokine secreted to the culture medium were measured by ELISA, and values are expressed in pg/l normalised to β-actin mRNA expression. Insulin released to the medium was measured by RIA and normalised to β-actin mRNA. a, b CXCL1; c, d IL-6; e, f insulin secretion. The results are means ± SEM of four to six independent experiments. *p < 0.05; **p < 0.01 vs Ct at G 6 mmol/l

Full size image

Human islet chemokine and cytokine expression is positively correlated with IL-1β and TNF-α

The expression of IL8, IL6, CXCL1 and CCL2 in the cultured human islet samples was positively correlated with the expression of IL1-β and TNF-α (Fig. 5). To examine whether an independent association exists between IL1-β, TNF-α and IL6 mRNA (considered as independent variables) and downstream chemokines and cytokines, CXCL1, CCL2 and IL8 mRNA (considered as dependent variables), multiple regression analysis was performed. IL1-β and TNF-α but not IL6 were significantly associated with CXCL1 and CCL2 (p < 0.01), while all three cytokines were independently correlated with IL8 expression (p < 0.01).

Fig. 5
figure 5

Expression of chemokine and cytokine mRNA is positively correlated with IL1-β and TNF-α mRNA levels in cultured human islets. Linear correlation between log-transformed chemokine and cytokine mRNA and IL1-β (a–e) and TNF-α (f–i) mRNA expression in human islets cultured under different glucose and NEFA conditions for 24 h and 48 h; n = 135 samples. Correlation of IL1-β with: a CXCL1, R = 0.85, p < 0.001; b IL6, R = 0.75, p < 0.001; c CCL2, R = 0.87, p < 0.001; d IL8 R = 0.94, p < 0.001; e TNF-α, R = 0.85, p < 0.001. Correlation of TNF-α with: f CXCL1, R = 0.81, p < 0.001; g IL6, R = 0.81, p < 0.001; h CCL2, R = 0.78, p < 0.001; i IL8 R = 0.91, p < 0.001

Full size image

IL-1β mediates the chemokine/cytokine induction by palmitate

To establish whether IL-1β is upstream and causal in the inflammatory response induced by palmitate in human islets, we blocked IL-1β signalling using IL-1ra as previously described [27]. As a positive control for IL-1ra activity, we used it in combination with the cytokines IL-1β and IFN-γ. IL-1β and IFN-γ induced a much greater expression of the mRNA of cytokines and chemokines (three- to tenfold, Fig. 6) than palmitate. IL-1ra significantly inhibited the cytokine-induced expression of IL6 at the mRNA level and IL-6 production at the protein level (by seven- and ninefold, Fig. 6) and also decreased expression of the other chemokines and cytokines (Fig. 6). Interestingly, IL-1ra prevented palmitate-induced IL6 mRNA expression and IL-6 secretion, as well as the induction of CXCL1, IL8 and IκBα (also known as NFKBIA; Fig. 6). IL-1ra also decreased basal expression of IL6, CXCL1, CCL2, IL8 and TNF-α, pointing to a basal IL-1β-induced chemokine production by human islets, as previously reported [23]. The palmitate-induced TNF-α expression was also prevented by IL-1ra, suggesting that it is downstream of IL-1β signalling. While IL-1ra reduced cytokine-induced human islet cell death, blocking IL-1β signalling did not protect against palmitate (Fig. 6), suggesting that IL1-β induction by palmitate regulates the expression of downstream cytokines and chemokines, but does not play a major role in lipotoxic cell death.

Fig. 6
figure 6

IL-1ra prevents palmitate-induced chemokine/cytokine mRNA expression and protein production in human islets but not islet cell death. For mRNA and protein analysis, human islets were cultured for 48 h in the presence of palmitate (Pal), alone or in combination with IL-1ra. The cytokines (Cyt) IL-1β (50 U/ml) plus IFN-γ (1,000 U/ml), alone or in combination with IL-1ra, were used as positive controls. Chemokine, cytokine and IκBα mRNA expression was analysed by real-time PCR and IL-6 protein by ELISA. mRNA data are expressed as fold induction of control (Ct) and IL-6 secretion as pg/l normalised to β-actin. Results are means±SEM of eight independent experiments. For the evaluation of human islet cell death, islets were cultured for 3 days. a IL6 mRNA expression; b IL-6 protein secretion; c CXCL1 mRNA expression; d CCL2 mRNA expression; e IL8 mRNA expression; f IL1-β mRNA expression; g TNF-α mRNA expression; h IκBα mRNA expression; i human islet cell death. The results are means±SEM of four independent experiments. *p < 0.05, **p < 0.01; ***p < 0.001 for treated vs control; † p < 0.05, †† p < 0.01; ††† p < 0.001 for cytokine vs cytokine + IL-1ra; ‡ p < 0.05, ‡‡ p < 0.01; ‡‡‡ p < 0.001 for palmitate vs palmitate + IL-1ra

Full size image

Cell source of the chemokines and cytokines produced by human islets

Because the purification of human beta cells from islet preparations is technically challenging, we evaluated the cell type in which palmitate induces NF-κB activation by staining dispersed human islet cells for insulin and p65, the major NF-κB component in beta cells. Palmitate, and the combination of IL-1β plus IFN-γ plus TNF-α used as positive control, induced NF-κB activation in beta and non-beta cells, indicated by nuclear translocation of p65 after 2 h (ESM Fig. 2). Palmitate also activated NF-κB in the clonal human ductal cells Panc-1 and Capan-2 (ESM Fig. 3), and it induced chemokines and cytokines in ductal cells, while oleate did not (ESM Fig. 3). This suggests that both human beta cells and exocrine ductal cells exhibit pro-inflammatory responses to saturated NEFA.

Role of ER stress in the induction of chemokines and cytokines by palmitate

ER stress is present in islets from type 2 diabetic individuals [31], and can be elicited in human islets in vitro by saturated NEFA [10]. To examine whether ER stress signalling contributes to the observed induction of chemokines and cytokines, we exposed human islets to the synthetic ER stressor CPA. This ER stressor induced cytokine expression and NF-κB activation to a similar extent as palmitate (ESM Fig. 4). To evaluate the role of the PERK/eukaryotic translation initiation factor 2, subunit 1 α (eIF2α) pathway on chemokine/cytokine expression we used salubrinal, a selective inhibitor of eIF2α dephosphorylation [28, 30]. Salubrinal did not lead to chemokine/cytokine or IκBα expression in human islets (ESM Fig. 5), suggesting that other branches of the ER stress response mediate the pro-inflammatory response.

Discussion

Systemic low-grade inflammation precedes the onset of type 2 diabetes by years [14], but its role in the pathogenesis of the disease is debated. In genetic studies, a polymorphism in CCL2 has been associated with type 2 diabetes risk [32]. In a meta-analysis of type 2 diabetes genome-wide association scans, however, gene variants influencing circulating levels of IL-1ra, IL-18, the IL-6 receptor, macrophage migration inhibitory factor and C-reactive protein did not predispose to type 2 diabetes [33], suggesting that inflammation might be secondary to rather than causal for diabetes. Human and animal studies point to a role for inflammation in enlarged adipose tissues, increased secretion of cytokines and chemokines by adipocytes and fat-infiltrating macrophages, and insulin resistance [14, 15]. Whether inflammation is also present in islets and whether this contributes to the development of type 2 diabetes are important questions because of the central role of beta cell dysfunction in the disease pathogenesis [22]. Moreover, IL-1ra therapy is under evaluation for the treatment of type 2 diabetes [34, 35], based on studies suggesting that high-glucose-induced IL-1β secretion by beta cells causes beta cell dysfunction and death [21].

In the present study, we observed a gene expression signature of mild inflammation in human type 2 diabetic beta cells obtained by LCM. There was increased expression of the mRNA for: the β-chemokines CCL2 and CCL13, implicated in monocyte chemotaxis [15]; α-chemokines CXCL1 and IL-8, potent chemoattractants and activators of neutrophils [15]; and the pro-inflammatory cytokines IL-1β and IL-6. The chemokine and cytokine mRNA expression was heterogeneous in the type 2 diabetic group, suggesting that mild inflammation is present in some but not all type 2 diabetic individuals. Mild macrophage infiltration has been observed in islets from type 2 diabetic individuals in two recent studies, with 20% of the islets containing >3 CD68+ cells per islet compared with 5% of the ND islets [18, 19]. This islet inflammation could be mediated by CCL2 and CCL13 produced in human beta cells. Islet inflammation was also present in animal models for type 2 diabetes, namely the high-fat-diet-fed and db/db mice [18, 36], the Cohen diabetic rat [36] and the Goto–Kakizaki rat [37, 38]. The role of these infiltrating macrophages is unknown. Some propose them to be beneficial to islet function by promoting angiogenesis [39]; others suggest that activated CD68+ cells could cause beta cell dysfunction and death [36, 37]. The beta cell inflammatory response was presently reproduced at the RNA and protein level in ND human islets by their in vitro exposure to the saturated NEFA palmitate, and was dependent on the canonical mediator of early inflammation IL-1β. While confocal microscopy suggests that CCL2 protein production in ND and type 2 diabetic pancreas is largely restricted to beta cells, palmitate induced a pro-inflammatory response in both beta and islet non-beta and ductal cells.

At variance with previous reports that elevated glucose concentrations increase IL1-β expression in human islets cultured either in suspension or on an extracellular matrix [20, 21], glucose did not induce an inflammatory response in human islets in this or a previous study by our group [23]. Rather, the chemokine/cytokine expression and NF-κB activation was induced by the saturated NEFA palmitate, but not by the unsaturated oleate. Saturated and unsaturated NEFA have different functional effects in beta cells [40–42]. Saturated NEFA are pro-inflammatory in several cell types [43, 44] and also cause ER stress [10]. Because palmitate (but not oleate or high glucose) induces ER stress in human islets [10] and there is a crosstalk between ER stress and NF-κB pathways [45–47], we explored this mechanism further. We have previously shown that palmitate activates PERK, and also leads to ATF6 and IRE1 activation through depletion of ER Ca2+ stores [10]. The crosstalk between ER stress and NF-κB activation can occur at different levels. Ca2+ efflux from the ER activates NF-κB through generation of reactive oxygen species [46] and release of a diffusible, still unknown, NF-κB activator [48]. IRE1 and TNF receptor-associated factor 2 (TRAF-2) can activate NF-κB [45], and stalled protein translation caused by eIF2α phosphorylation decreases IκBα protein availability, consequently activating NF-κB [49]. CPA, a canonical ER stressor, upregulates chemokines and cytokines and activates NF-κB in human islets with an effect comparable with that of palmitate (present data). The CPA-induced p65 nuclear translocation in beta and non-beta cells occurred after short (2 h) exposure, in line with previously reported kinetics showing maximal NF-κB activation by CPA and thapsigargin after 1 h due to ER Ca2+ depletion [50]. To examine involvement of the PERK pathway in NF-κB activation, we used salubrinal as a selective inhibitor of eIF2α dephosphorylation [28, 30]. Salubrinal did not induce chemokines/cytokines or IκBα (ESM Fig. 5), suggesting that ER Ca2+ depletion or ER stress-induced IRE1 signalling but not inhibition of protein translation mediates palmitate-induced NF-κB activation in human islets. ER stress is not the only mechanism by which palmitate could induce mild inflammation in human islets. Toll-like receptors (TLRs) may also recognise saturated NEFA [51]. These TLRs have been suggested to mediate some of the NEFA-induced islet inflammation in a recent study [52].

Blocking IL-1β signalling with IL-1ra prevented the palmitate-induced expression of chemokines and cytokines and NF-κB activation, showing that the observed inflammatory response is largely dependent on IL-1β production. The magnitude of IL-1β induction is modest, and probably insufficient to cause human islet cell damage. In fact, low concentrations of IL-1β have been shown to stimulate human islet function [53]. IL-1ra did not prevent palmitate-induced human islet cell death in vitro, suggesting that the mild IL-1β induction is not directly deleterious to the human islets and does not mediate lipotoxicity. Administration of IL-1ra in vivo to high-fat-fed mice improved beta cell function but also fully protected against changes in insulin sensitivity [54]. IL-1ra reduced adipose tissue inflammation, altered adipokine secretion and decreased circulating lipid levels, which may have contributed to the preservation of insulin secretion and glucose tolerance in the high-fat-fed mice. We propose, therefore, that the induction of IL-1β and downstream chemokines and cytokines by saturated NEFA results in mild islet inflammation in type 2 diabetes, but that this does not directly contribute to beta cell dysfunction and apoptosis. ER stress [31] and other signalling pathways [55] contribute to lipotoxicity in human islets in type 2 diabetes.

In conclusion, we demonstrate here that there is a mild inflammatory process in islets from some but not all type 2 diabetic patients. Palmitate, but not oleate or high glucose, induces in vitro chemokine and cytokine expression in ND human islets, a process largely dependent on IL-1β production. ER stress could be one of the mechanisms involved in saturated-fatty-acid-induced human islet inflammation. The induction of IL-1β and downstream chemokines and cytokines by saturated NEFA does not directly contribute to lipotoxicity.

The mild inflammation in islets from type 2 diabetic patients could be the result of in vivo exposure to saturated NEFA. Whether this islet inflammation causes loss of functional beta cell mass and thereby contributes to the development and progression of type 2 diabetes remains to be elucidated.