figure b

Introduction

Insulin secretion is driven by the elevation of blood glucose, which activates metabolism–secretion coupling in the pancreatic beta cell. Glucose is transformed into pyruvate and metabolised in the mitochondria. The resulting elevation of cytosolic ATP closes the potassium channels, inducing membrane depolarisation and calcium influx that triggers insulin exocytosis. Multiple coupling factors have been proposed to amplify glucose-stimulated insulin secretion (GSIS), notably solute metabolites and fatty acids [1, 2]. Fatty acids can potentiate GSIS through either receptor-mediated free fatty acid receptor (FFAR)1 signalling or metabolically active derived molecules, such as long-chain acyl-CoA (Lc-CoA), diacylglycerol (DAG) and monoacylglycerol (MAG). Lc-CoA participates in the beta cell secretory response, which is primarily controlled by glucose levels [3,4,5]. Under glucose deprivation, Lc-CoA may represent an energy source through mitochondrial β-oxidation to maintain basal cellular activities [6]. Increasing glucose concentration inhibits β-oxidation and elevates the availability of cytosolic Lc-CoA [7], thereby modulating GSIS by targeting ATP-sensitive potassium (KATP) channels and protein kinase C (PKC) [8, 9] or entering the glycerolipid/NEFA cycle [10]. In turn, the glycerolipid/NEFA cycle may produce DAG and MAG species; among them 1-MAG, which was shown to modulate GSIS near the plasma membrane by facilitating insulin granule exocytosis [11].

The glycerolipid/NEFA cycle is composed of a lipogenic and a lipolytic arm, both activated during glucose stimulation in the beta cell [10, 12, 13]. In support for a role of this glucose-enhanced glycerolipid/NEFA cycling, inhibition of lipolysis or complete depletion of esterified lipid stores markedly impair GSIS [14,15,16]. Furthermore, in animals presenting with obesity and insulin resistance, thereby requiring higher insulin production, the lipogenesis and lipolysis turnover in islets is enhanced, while it becomes impaired with type 2 diabetes [10]. Lipogenesis involves the consecutive formation of lysophosphatidic acid (LPA), phosphatidic acid (PA), DAG and finally triacylglycerol (TAG) from glycolysis-derived glycerol triphosphate and Lc-CoA. The lipolytic arm of the cycle hydrolyses TAG to DAG and MAG, finally releasing NEFA and glycerol [10]. Enzymes involved in fatty acid synthesis have been extensively investigated, mainly acetyl-CoA carboxylase (encoded by ACACA) forming malonyl-CoA, then used by the fatty acid synthase (FASN) to produce de novo palmitate [4]. In contrast, enzymes of the glycerolipid/NEFA cycle in the beta cell catalysing lipogenesis and lipolysis reactions remain poorly characterised and consist of partially redundant and compartmentalised acyltransferases (glycerol-3-phosphate acyltransferase [GPAT], acyl-glycerol-3-phosphate acyltransferase [AGPAT], DAG-acyltransferase [DGAT]) and lipases (adipose triglyceride lipase [ATGL], hormone-sensitive lipase [HSL], MAG lipase [MAGL]/abhydrolase domain containing 6 [ABHD6]), respectively [10, 17]. The pathways through which the glycerolipid/NEFA cycle intermediates potentiate GSIS remain unclear, while they have been associated with DAG and MAG. Specifically, 1,2-DAG and 1-MAG promote insulin release, the former by activating PKC [18] and the latter by interacting with Munc13–1, a component of the exocytotic machinery [19].

The aim of the present work was to investigate the links between glucolipotoxic conditions that the beta cells may face in the context of type 2 diabetes and the putative contribution of the glycerolipid/NEFA cycle. For this purpose, we analysed the expression of glycerolipid/NEFA cycle-associated genes in human and rat islets and INS-1E beta cells, as well as the lipid storage vs mobilisation during glucose-stimulated secretion. More specifically, the chronic effects of the saturated and monounsaturated fatty acids palmitate and oleate were investigated at normal and high glucose concentrations. The study points to the glycerolipid/NEFA cycle as a consistent adaptive response to glucolipotoxic conditions.

Methods

Reagents

Polyornithine, fatty acid-free BSA, fatty acids, d-glucose and KCl were obtained from Sigma-Aldrich (St Louis, MO, USA). RPMI-1640 medium, Halt protease inhibitor and the fluorescent 4,4-Difluoro-1,3,5,7,8-Pentamethyl-4-Bora-3a,4a-Diaza-s-Indacene dye (Bodipy 493/503) were purchased at Thermo Fischer Scientific (Waltham, MA, USA). Inhibitors (R)-(+)-Etomoxir and Orlistat were obtained from Bio-Techne (Minneapolis, MN, USA) and Atglistatin from MedChem Express (Monmouth Junction, NJ, USA). Glycerol measuring kit was purchased from BioVision (Milpitas, CA, USA).

Culture and treatment of human and rat islets and INS-1E beta cells

Human islets were isolated at the Geneva University Hospital (Switzerland) from five pancreases of deceased multiorgan donors, who had provided written informed consent (ECIT [European Consortium for Islet Transplantation]). None of the donors were diagnosed with the metabolic syndrome or diabetes. Donors had an average BMI of 25.8 ± 2.6 kg/m2 and were aged 52.2 ± 8.2 years (means ± SD; electronic supplementary material [ESM] Table 1). Islets were maintained for a standard recovery period of time (1–4 days) in CMRL-1066 medium (Corning, New York, NY, USA) at 5.5 mmol/l glucose supplemented with 10% heat-inactivated FCS and used for experiments straight away without shipping (isolated on site). Islets were randomly hand-picked, washed and further cultured for 3 days in Falcon Petri dishes (Corning) in the presence of 10% FCS at either physiological 5.5 mmol/l glucose (G5.5, control) or high 25 mmol/l glucose (G25) supplemented or not with 0.4 mmol/l palmitate (C16:0), 0.4 mmol/l oleate (C18:1) or a mix of both fatty acids (0.2 mmol/l each) in the presence of 0.5% fatty acid-free BSA.

Clonal mycoplasma-free INS-1E beta cells (RRID: CVCL_0351) were cultured in RPMI-1640 GlutaMAX-1 medium (Thermo Fischer Scientific) at 11.1 mmol/l glucose supplemented with 10 mmol/l HEPES, 5% (vol./vol.) FCS, 100 U/ml penicillin, 100 μg/ml streptomycin, 1 mmol/l sodium pyruvate and 50 μmol/l β-mercaptoethanol. This medium is referred to as control medium. After 24 h seeding in 6-well Falcon culture dishes (Corning) or 96-well plates (Greiner Bio-One, Kremsmünster, Austria) coated with polyornithine (Sigma-Aldrich), INS-1E beta cells were exposed for 3 days to different glucose and fatty acids concentrations as described above for human islets. Stock solutions of fatty acids were adjusted to 8 mmol/l in 11% fatty acid-free BSA solution, without organic solvent, and stored at −20°C as previously described [20]. This preparation results in a concentration of unbound fatty acids in the range of 0.1–0.5 μmol/l [21]. Female 30-week-old Wistar rats weighing 200–250 g obtained from an in-house colony (University Medical Centre, Geneva, Switzerland) were housed 2–3 per cage with 12 h light–dark cycle and free access to a standard mouse chow diet and water. Rat islets were isolated by collagenase digestion f  [22], randomly hand-picked and cultured overnight free-floating in INS-1E complete RPMI-1640 medium before the 3 day culture at 11.1 mmol/l glucose used as control condition [23] and G25 without or with 0.4 mmol/l palmitate or oleate. Due to various culture conditions used in this study, the experimenters were not blinded to group assignments, while blinded for analysis and outcome.

RNA analyses

RNA was collected using peqGOLD Trifast (Peqlab, Germany)/chloroform extraction before RNA-Seq analysis (R/Bioconductor package EdgeR v. 3.4.2 and Cytoscape StringApp v 3.6.1.) for human islets and quantitative RT-PCR (qRT-PCR) on a StepOnePlus real-time PCR machine (Applied Biosystems, USA) for rat islets and INS-1E cells, as described in the ESM Methods. Accession numbers and fold changes with p values for RNA-Seq are provided in ESM Table 2. Primers used for qRT-PCR are listed in ESM Table 3.

Transmission electron microscopy

INS-1E beta cells treated for 3 days with different glucose concentrations and fatty acids were fixed with 2% glutaraldehyde in 0.1 mol/l pH 7.4 phosphate buffer for 1 h at room temperature. Cells were scraped, washed and pelleted in phosphate buffer. Samples were further dehydrated, epon-embedded and cut at the faculty electron microscopy core facility (PFMU, University of Geneva). Images were acquired on the Morgagni transmission electron microscope (FEI Company, Eindhoven, the Netherlands).

Insulin and luminescence-based secretion assay

INS-1E beta cells expressing a Gaussia luciferase in place of the insulin c-peptide [24] were used to assess the kinetics of secretion after chronic treatments. This luminescence-based assay gives similar results compared with measurement of immunoreactive insulin [25]. Cells were starved for 2 h in glucose-free RPMI-1640 medium (1% FCS), washed and incubated further for 40 min in KRB-HEPES at 2.5 mmol/l glucose. Native coelenterazine (5 μmol/l, Nanolight Technologies, Pinetop, AZ, USA) was added to the wells and after 20 min of signal stabilisation luminescence was monitored for 40 min using the Fluostar plate reader (BMG Labtech, Ortenberg Germany), first at basal 2.5 mmol/l glucose (10 min) and then at stimulatory 15 mmol/l glucose [26]. Blockade of lipolysis was achieved by incubating the cells with 200 μmol/l Orlistat during the starvation period (3 h) and during the assay. Secretion assay on rat islets (20 per tube), after the same starvation procedure used for INS-1E cells, was performed for 30 min at basal 2.5 and stimulatory 15 mmol/l glucose, as well as 30 mmol/l KCl, and insulin was measured by radioimmunoassay as previously described [27].

Intracellular lipid quantification and mobilisation

The storage of neutral lipids in INS-1E cells was assessed by loading the cells with 1 μg/ml of Bodipy dye as described previously [25] in the presence or not of the panlipase inhibitor Orlistat (200 μmol/l) or the ATGL specific inhibitor Atglistatin (20 μmol/l). In brief, Bodipy fluorescent signal (493/503 nm) was quantified using the ImageXpress XL plate reader and MetaXpress software (Molecular Devices, Sunnyvale, CA, USA) and expressed as fluorescent signal normalised to cell density. Mobilisation of TAG was measured through the release of glycerol by INS-1E cells and rat islets, quantified using a lipase-free commercial kit (#K622–100, BioVision).

Data treatment and statistical analysis

Statistical tests between each condition were performed using one-way ANOVA analysis followed by Tukey multiple comparison test. Results are presented as means ± SEM of at least three independent experiments, an independent experiment being representative of one cell passage. For secretion kinetics, a multiple unpaired t test was performed between the fatty acid-treated condition and the corresponding glucose BSA control using the false discovery rate approach. A p value lower than 0.05 was considered statistically significant. For RNA-Seq analysis, significant changes were considered when two or more independent islet batches (donors) exhibited down- or upregulation with a log2 fold change (log2 FC) threshold of 0.5 associated with at least one or more p<0.05.

Results

In human islets palmitate and oleate differentially regulate genes involved in lipid metabolism

The human islet transcriptome of lipid-associated genes was assessed by RNA-Seq following 3 day exposure to palmitate (C16:0) and oleate (C18:1) at standard (G5.5) or high (G25) glucose (Fig. 1a). We observed important variability among donors (Fig. 1) regarding the metabolic stress responses and, accordingly, only genes with significant and consistent changes among at least two donors were considered as relevant, as described in the Methods section. High glucose upregulated FASN, encoding fatty acid synthase, and this effect was maintained in the presence of oleate or palmitate (Fig. 1d–f). Conversely, high glucose, as well as oleate, downregulated ACSL6, encoding acyl-CoA synthetase, which catalyses the activation of NEFA to CoA esters. At both standard and high glucose concentrations, fatty acids induced expression of PLIN2, which encodes perilipin 2, a lipid droplet-associated protein. Palmitate promoted the upregulation of the desaturase-encoding gene SCD, at both G5.5 and G25 (Fig. 1b,e), an effect prevented by the addition of the unsaturated fatty acid oleate (Fig. 1g). At physiological glucose concentration (G5.5) oleate and palmitate upregulated AGPAT2, encoding the acyl-transferase that converts LPA into PA in the second step of TAG synthesis (Fig. 1b,c). This effect was abrogated when palmitate and/or oleate was combined with G25 (Fig. 1e–g). Oleate and high glucose increased expression of PNPLA2, encoding ATGL, but, concomitantly, robustly upregulated its inhibitor G0/G1 switch gene 2, encoded by G0S2 (Fig. 1c–f). Interestingly, both MAG hydrolase enzymes were upregulated; MGLL by high glucose on its own (Fig. 1d) and ABHD6 by palmitate and oleate at G25 (Fig. 1e,f), ABHD6 being associated with reduced insulin secretion [19]. Regarding fatty acid signalling, the NEFA (free fatty acid) receptors (FFARs) appeared to be essentially downregulated by high glucose (Fig. 1d). Of note, the combination of palmitate and oleate at G25 reverted, or attenuated, the effects of high glucose only (Fig. 1g). Short-chain fatty acids may regulate beta cell function either directly through their own FFAR2/3 [28] or as a consequence of intestinal L cell activation and the production of GLP-1, which activates the corresponding GLP-1 receptor of the beta cell [29]. Interestingly, GLP-1 receptor was downregulated by the glucotoxic condition (Fig. 1d).

Fig. 1
figure 1

Lipid metabolism-related transcriptomic regulation in human islets under metabolic stress conditions. (a) Functional interaction network of genes involved in lipid pathways. Node connections were established according to the STRING interaction knowledgebase (using the Cytoscape StringApp) with a confidence score >0.4. (bg) Effects of high 25 mmol/l glucose (G25) and 0.4 mmol/l oleate (C18:1) or palmitate (C16:0) on the transcriptional regulation of genes involved in lipid pathways. Human islets were exposed to (b) C16:0 at G5.5, (c) C18:1 at G5.5, (d) G25, (e) G25 + C16:0, (f) G25 + C18:1 and (g) G25 + C16:0 + C18:1 for 3 days before RNA-Seq analysis. Effects of culture conditions on transcript levels are compared with standard G5.5 medium and shown as upregulated (red), downregulated (blue), or unchanged (white). Missing values are represented in grey. Each disk is split into individual changes for the different donors. The colour code reflects the transcriptional changes in log2 fold changes (log2 FC) for that particular gene in individual donors. Significant changes were considered when two or more independent islet batches (donors) exhibited down- or upregulation with a log2 FC threshold of 0.5 associated with at least one or more p<0.05. Genes significantly regulated are highlighted in bold. *adjusted p<0.05, **adjusted p<0.01, ***adjusted p<0.001 between control G5.5 and the specific culture condition

In INS-1E beta cells fatty acids counteract part of the gene expression changes induced by high glucose

We further analysed mRNA levels of genes regulating lipid signalling and the glycerolipid/NEFA cycle (Fig. 2a) in INS-1E beta cells. Cells were cultured for 3 days in the presence of 0.4 mmol/l palmitate or oleate at different glucose concentrations. Because the EC50 of the glucose secretory response of INS-1E cells corresponds to its normal culture concentration [27], we selected 11.1 mmol/l glucose (G11.1) as the control standard condition; then 5.5 mmol/l as low basal glucose (G5.5) and 25 mmol/l as high glucose (G25). Consistent with human islets, expression of some of the key genes implicated in lipid metabolism was preserved by the different culture conditions (Fig. 2). The most salient difference with human islets was the absence of fatty acid-induced upregulation of Plin2 in INS-1E beta cells (Fig. 2j). Similar to the human islet response, the stearyl-CoA desaturase Scd1 was upregulated in INS-1E beta cells by palmitate at G5.5 (Fig. 2k, 3.2-fold, p<0.05), while it was not in the presence of the unsaturated fatty acid oleate. This effect was no longer observed in cells treated with higher glucose concentrations (G11.1 and G25).

Fig. 2
figure 2

Transcript levels of key enzymes of the glycerolipid/NEFA cycle and lipid-related genes in INS-1E beta cells. Cells were exposed to low 5.5 mmol/l (G5.5, blue), standard 11.1 mmol/l (G11.1, grey) and high 25 mmol/l (G25, red) glucose concentrations without (BSA) or with 0.4 mmol/l palmitate (C16:0) or oleate (C18:1) for 3 days. (a) Pathways of the lipogenic (green) and lipolytic (blue) arms of the glycerolipid/NEFA cycle. Transcript levels of glycerol-3-phosphate phosphatase (Pgp, b), glycerol-3-phosphate acyltransferase (Gpam, c), DAG O-acyltransferase 1 and 2 (Dgat1, d; Dgat2, e), ATGL (Pnpla2, f), HSL (Lipe, g), monoacylglycerol lipases (Mgll, h; Abhd6, i), perilipin 2 (Plin2, j) and stearoyl-CoA desaturase (Scd1, k). Results are presented as means ± SEM of 4 independent experiments (n = 2 in each experiment) and expressed as mRNA levels normalised to cyclophilin relative to G11.1 BSA condition; *p<0.05, **p<0.01, ***p<0.001 (one-way ANOVA, Tukey vs all conditions)

Expression of enzymes of the glycerolipid/NEFA cycle was more sensitive to treatments in INS-1E beta cells than in human islets. At G11.1 expression of the TAG forming Dgat2 was profoundly decreased by palmitate (−73%, p<0.01) or oleate (−75%, p<0.001) compared with control cells (Fig. 2e). Regarding TAG mobilisation, expression of genes encoding ATGL (Pnpla2, Fig. 2f) and the beta cell 89 kDa HSL isoform [30] was preserved (Lipe, Fig. 2g). The MAG lipase-encoding gene Mgll was upregulated 4.1-fold after chronic high glucose culture (Fig. 2h, p<0.01), an effect dampened by the presence of palmitate during the treatment (−30%, p<0.05). The major beta cell MAG lipase Abhd6 was high at basal glucose (G5.5) and low at higher glucose culture condition (Fig. 2i). As in human islets, G25 upregulated G0s2 (Fig. 3a).

Fig. 3
figure 3

Transcript levels of lipid-related genes in INS-1E beta cells and rat islets. Cells were exposed to low 5.5 mmol/l (G5.5), standard 11.1 mmol/l (G11.1) and high 25 mmol/l (G25) glucose concentrations without NEFA (BSA) or with 0.4 mmol/l palmitate (C16:0) or oleate (C18:1) for 3 days. Transcript levels in INS-1E cells of the G0/G1 switch gene 2 (G0s2, a), time course and endpoint of NEFA receptors 1 (Ffar1, b, c), 2 (Ffar2, d) and 3 (Ffar3, e). Results are means ± SEM of 4 independent experiments (n = 2 in each experiment) and expressed as mRNA levels normalised to cyclophilin relative to G11.1 BSA condition. Transcript levels in rat islets of Pnpla2 (f), Lipe (g), Mgll (h), Abhd6 (i), G0s2 (j) and Ffar1 (k). Results are means ± SEM (n = 4) expressed as mRNA levels (ΔΔCt) normalised to cyclophilin; *p<0.05, **p<0.01, ***p<0.001 (a, ck: one-way ANOVA, Tukey; b: multiple t test, Holm–Sidak)

The effects of glucose and fatty acids on the expression of Ffars in INS-1E cells were rather consistent with those observed in human islets, although mRNA from the long-chain fatty acid receptor Ffar4 (GPR120) was not detected in INS-1E cells. Expression levels of the medium and long-chain fatty acid receptor Ffar1 (GPR40) inversely correlated with glucose concentrations, showing upregulation at G5.5 and downregulation at G25 compared with control G11.1 (Fig. 3b,c). Interestingly, the time course for Ffar1 revealed that changes appeared rather late, pointing to the second transcriptional wave of the glucose response reported previously [31]. At G5.5 palmitate induced a twofold Ffar1 upregulation compared with G5.5 alone (Fig. 3c), while oleate did not. We also analysed expression of the receptors for short-chain fatty acids, produced mainly by the gut microbiota [32], which can also regulate insulin secretion [28, 29]. Similar to the Ffar1 pattern, Ffar2 (GPR43) expression levels exhibited an inverse glucose dependence, while at G5.5 palmitate induced a further 3.7-fold upregulation (Fig. 3d). As in human islets, Ffar3 (GPR41) expression was not significantly changed by any of the tested treatments (Fig. 3e).

In rat islets, expression of TAG mobilisation enzymes was not altered by high glucose and fatty acids, although G0s2 was upregulated by palmitate (Fig. 3f–j). Of note, compared with Abhd6, expression levels of the MAG lipase Mgll were much lower, consistent with beta cell disallowed genes [33]. Finally, fatty acids at G25 downregulated Ffar1 (Fig. 3k).

Oleate combined with high glucose sets the limit of cellular fat storage

Neutral lipids were quantified in INS-1E beta cells exposed to basal (G5.5), control (G11.1) and high glucose (G25) in the presence of palmitate, oleate or a mix of both. High glucose combined with oleate induced massive accumulation of lipid droplets (15.1-fold increase compared with G25 alone, p<0.001), while the effect of palmitate was marginal (Fig. 4a–c). Interestingly, palmitate induced similar accumulations of esterified lipids at all glucose concentrations, whereas oleate exhibited strong glucose dependence, resulting in a wide range of lipid storage levels. Lipid esterification in cells exposed to chronic high glucose and a mix of palmitate and oleate was 8.9-fold higher than G25 alone (Fig. 4c), showing an intermediate phenotype between palmitate and oleate-treated cells.

Fig. 4
figure 4

Neutral lipid accumulation in INS-1E beta cells under metabolic stress conditions. Cells were exposed to low 5.5 mmol/l (G5.5), standard 11.1 mmol/l (G11.1) and high 25 mmol/l (G25) glucose concentrations without (BSA) or with 0.4 mmol/l palmitate (C16:0) or oleate (C18:1) for 3 days. (a) Electron micrographs of cells cultured at G11.1 and G25 plus C16:0 and C18:1. Scale bar, 2 μm. (b) Representative fluorescence microscopy images of cells stained with Bodipy revealing neutral lipids after culture for 3 days at G25 plus C16:0 and C18:1 in the absence (DMSO) or presence of 200 μmol/l of the panlipase inhibitor Orlistat. Scale bar, 20 μm. (c) Quantitative analysis of intracellular neutral lipids (Bodipy signal) following the different culture conditions. Results are presented as means ± SEM of at least 5 independent experiments and expressed as Bodipy fluorescent signal normalised to cell density. ***p<0.001 and $p<0.001 vs, respectively, G11.1 BSA control and the DMSO control group of the corresponding culture condition (one-way ANOVA, Tukey)

We tested maximal fat storage capacity in INS-1E beta cells by concomitantly incubating cells with the different glucose concentrations and fatty acids plus the panlipase inhibitor Orlistat. However, Orlistat dramatically increased lipid storage in cells exposed to palmitate, oleate, or the mix of both, at any of the glucose culture conditions (Fig. 4c); without significant effects on cell density and morphology (ESM Fig. 1). Blockade of lipases by Orlistat uncovered an apparent maximal storage capacity, as indicated by a plateau phase of the lipid signal. This effect was not observed with Atglistatin used as a selective inhibitor of ATGL (ESM Fig. 2), indicating that alternative lipases, such as HSL [34], can hydrolyse TAG. Interestingly, Orlistat added at high glucose without exogenous fatty acids was not able to increase intracellular lipid stores, potentially because of its inhibitory action on the thioesterase domain of fatty acid synthase required for glucose-derived lipogenesis [35].

This set of data shows that the presence of oleate favours intracellular accumulation of neutral lipids in a glucose-dependent way and that the combination of 25 mmol/l glucose and 0.4 mmol/l oleate over 3 days is sufficient to promote nearly maximal fat storage capacity of the cells.

Mobilisation of intracellular fatty acids preserves part of the secretory response altered by glucotoxic conditions

We next examined to what extent storage of neutral lipids and their potential recruitment may affect acute glucose-stimulated secretion in INS-1E beta cells. After the 3 day exposure to different glucose concentrations and fatty acids resulting in various degrees of lipid accumulation (Fig. 4), secretion was tested in the absence or presence of the panlipase inhibitor Orlistat. Cells from the culture at low glucose (G5.5) exhibited weak responses to acute stimulatory 15 mmol/l glucose (Fig. 5a). The presence of palmitate in the culture medium partially preserved the secretory response, an effect not observed when only oleate was provided during the culture. Interestingly, only the second phase of secretion was restored in palmitate-treated cells. In INS-1E cells cultured at standard G11.1, the presence of oleate during the 3 days of pretreatment resulted in the potentiation of the secretory response, while palmitate alone had no effects (Fig. 5b). Cells from the G25 culture exhibited much blunted secretory responses. This glucotoxic effect was partially alleviated by the presence of oleate and the mix of palmitate and oleate in the culture media; not by palmitate alone (Fig. 5c). When cells were stimulated by KCl used as a calcium-raising agent, similar effects of oleate were observed (ESM Fig. 3). The inhibition of lipases during the 3 h starvation period prior to and during the secretion assay dramatically reduced glucose-stimulated secretion in all of the tested conditions, an effect particularly salient in cells cultured at standard G11.1 (Fig. 5d–f; ESM Fig. 4). This suggests a predominant role of fatty acid recruitment in the amplifying pathway of the secretory response. Overall, TAG mobilisation appears to be necessary for sustained glucose-stimulated secretion in INS-1E beta cells. This effect was particularly relevant for oleate-treated cells under glucotoxic culture conditions.

Fig. 5
figure 5

Contribution of endogenous lipid mobilisation to glucose-stimulated secretion in INS-1E beta cells. After the 3 day treatments with different glucose concentrations and fatty acids, secretion was tested following a 3 h starvation period in the absence (DMSO) or presence of the panlipase inhibitor Orlistat (200 μmol/l). Secretion was monitored first for 10 min at basal 2.5 mmol/l glucose and then at stimulatory 15 mmol/l glucose (Glc) for the remaining 30 min. Representative secretion kinetics in DMSO control (ac) and Orlistat-treated cells (d–f) previously incubated at different glucose concentrations (a, d: G5.5; b, e: G11.1; c, f: G25) and 0.4 mmol/l palmitate (C16:0), oleate (C18:1) or a mix of both (C16:0 + C18:1, 1:1) for 3 days, expressed relative to basal release at 2.5 mmol/l glucose. Values are means of triplicates ± SD from 1 out of 4 experiments; *p<0.05 for C16:0, †p<0.05 for C18:1, §p<0.05 for C16:0 + C18:1 vs BSA group of the corresponding glucose culture condition (Multiple t test, Benjamini, Krieger and Yekutieli)

Fasting-induced intracellular lipid mobilisation is proportional to fat storage

As part of a standard insulin secretion assay, cells are subjected to a glucose-starvation period prior to acute stimulation. This fasting procedure typically lasts for about 3 h, during which intracellular lipid stores are eventually mobilised, thereby producing lipid-derived coupling factors for the secretory response [19]. Because beta cells, as opposed to hepatocytes, do not exhibit glycerol kinase activity [36, 37], glycerol produced from the breakdown of intracellular TAG by the MAG lipases is released in the extracellular space as a dead-end product [38] and its quantification in vitro provides a readout of the lipolytic activity. Following the 3 days of culture at different glucose concentrations combined with fatty acids, the various culture media were replaced by fasting medium that was collected after 3 h for glycerol quantification (Fig. 6a). Glycerol release positively correlated with glucose concentrations during the 3 day culture (Fig. 6b). The presence of fatty acids during the culture significantly increased TAG recruitment, exhibiting a strong synergistic effect in cells from the G25 culture. Thus, in cells cultured at G25 with either palmitate or oleate, glycerol release was 2.8 times higher than in fatty acid-free G25 (Fig. 6b, p<0.01). Glycerol may not only derive from lipolysis but also from glycerol-3-phosphate phosphatase activity [39]. Accordingly, the mobilisation rate of neutral lipids was also assessed by Bodipy staining over the 3 h fasting period (Fig. 6c). In agreement with glycerol release, Bodipy assessment showed glucose dependence for the levels of neutral lipid being mobilised, at least in oleate-treated cells. Following both oleate and palmitate/oleate treatments, neutral lipid mobilisation was 5.7 times higher in cells cultured at G25 compared with the G5.5 culture (p = 0.054 and p<0.05, respectively). Once mobilised from TAG stores, fatty acids may undergo β-oxidation. In order to evaluate the contribution of this catabolic pathway, the inhibitor of β-oxidation Etomoxir was added during the 3 h fasting period. This treatment did not modify the profile of neutral lipid mobilisation in oleate-treated cells (Fig. 6d). Cells cultured with palmitate exhibited slightly reduced fat recruitment during the starvation period in the presence of Etomoxir, potentially indicating oxidative catabolism of stored palmitate in these conditions. Taken as a whole, this set of data points to a lipid mobilisation rate governed mostly by the levels of stored fat, i.e. active glycerolipid/NEFA cycle, rather than energetic demands during the fasting period.

Fig. 6
figure 6

Lipid mobilisation and insulin secretion in INS-1E beta cells and rat islets loaded with neutral lipids. (a) Cells were exposed to low 5.5 mmol/l (G5.5), standard 11.1 mmol/l (G11.1) and high 25 mmol/l (G25) glucose concentrations without (BSA) or with 0.4 mmol/l palmitate (C16:0) or oleate (C18:1) or both for 3 days. Then, cells were washed (T0) and incubated at low glucose (G5.5) during 3 h (T3). (b) Quantitative analysis of glycerol release in the medium at T3. (c, d) Overall mobilisation rates of neutral lipids during the 3 h at G5.5 in the absence (c) or presence (d) of 200 μmol/l of Etomoxir (+Eto) in order to block β-oxidation calculated from neutral lipid quantifications using Bodipy staining at T0 and T3. Results are presented as mean ± SEM of 3 independent experiments and expressed as nmol of glycerol per μg of protein (b) and absolute slope values of neutral lipid quantification between T0 and T3 (c, d) *p<0.05, **p<0.01, ***p<0.001 relative to BSA control and §p<0.05, §§p<0.01 vs G25 without NEFA. (e, f) Rat islets were exposed to the indicated culture conditions for 3 days before a 3 h fasting period followed by 30 min stimulation. (e) Glycerol release from rat islets during the fasting period and the subsequent acute 15 mmol/l glucose stimulation (Glc-stim). (f) Insulin secretion from rat islets stimulated with 15 mmol/l glucose (Glc-stim) or 30 mmol/l KCl at 2.5 mmol/l glucose (KCl-stim) in the absence or presence of the panlipase inhibitor Orlistat (Orl, 200 μmol/l). Results are means ± SEM, n = 3–4, *p<0.05 and **p<0.01 (one-way ANOVA, Tukey)

In rat islets, the glycerol release measured during the starvation period following the 3 day culture period was similar among groups (Fig. 6e). However, over the ensuing acute glucose response, islets cultured at standard G11.1 exhibited a near exhaustion of glycerol output, while the release was still robust in islets from the G25 plus fatty acids cultures. Exposure to oleate at G25 for 3 days conferred hypersecretory responses to both acute 15 mmol/l glucose and 30 mmol/l KCl stimulations (Fig. 6f), as well as higher insulin release under non-stimulatory conditions (ESM Fig. 5). Importantly, the potentiation of the secretory responses was fully prevented when lipases were inhibited by Orlistat. These oleate-specific effects are in line with those observed in INS-1E cells (Fig. 5).

Cells exposed to oleate resist starvation-induced intracellular lipid clearance

The dynamic of intracellular fat storage over a prolonged 12 h starvation period was evaluated through a time course of neutral lipid staining in INS-1E beta cells. After 3 days of exposure to different glucose concentrations and fatty acids, the various culture media were replaced by a fatty acid-free 5.5 mmol/l glucose medium for a 12 h period during which we monitored intracellular lipid droplets (Fig. 7a). Compared with cells cultured at standard G11.1, cells from the G25 culture without fatty acids contained more esterified fat, i.e. derived from de novo synthesis, not significantly changed over the 12 h period. In cells cultured at G25 with palmitate, as much as 83% (p<0.05) of neutral lipids were already mobilised after 3 h at G5.5, exhibiting a basal background up to 12 h of starvation. Similarly, in cells cultured at G25 with the mix of palmitate and oleate, lipid droplet staining was reduced by 73% (p<0.01) after 3 h. However, in cells initially cultured at G25 with oleate, only 28% (p<0.05) of the much higher neutral lipid contents were mobilised; and after 12 h the reduction of stored fat reached 50% (p<0.05), leaving half of the initial intracellular lipids.

Fig. 7
figure 7

Time course of INS-1E intracellular neutral lipid mobilisation over 12 h. INS-1E cells were exposed to low 5.5 mmol/l (G5.5), standard 11.1 mmol/l (G11.1) and high 25 mmol/l (G25) glucose concentrations without (BSA) or with 0.4 mmol/l palmitate (C16:0) or oleate (C18:1) or both for 3 days. Then, cells were washed and incubated at G5.5 over 12 h in the absence (b) or presence (c) of 200 μmol/l of Etomoxir in order to block β-oxidation. Neutral lipids were quantified using Bodipy staining at time (h) T0, T3, T6 and T12. Results are presented as means ± SEM of 3 independent experiments and expressed as Bodipy fluorescent signal normalised to cell density. *p<0.05, **p<0.01, ***p<0.001 (one-way ANOVA, Tukey)

Inhibition of β-oxidation by Etomoxir during the 12 h starvation period delayed the mobilisation of neutral lipids specifically in palmitate-treated cells, while the effect was marginal in cells exposed to oleate only (Fig. 6b). Indeed, after 3 h at G5.5, 50% and 59% of neutral lipids where mobilised in cells previously cultured at G25 with palmitate and the mix of fatty acids, respectively; as compared with 83% and 73% recruitments without Etomoxir, respectively. By contrast, in cells initially cultured with oleate at G25, the presence of Etomoxir had little effect on fat mobilisation with a 30% reduction at 3 h vs 28% in the absence of Etomoxir. This indicates that β-oxidation accounted for only a fraction of the rapid mobilisation of fatty acids. Moreover, storage of palmitate favoured subsequent TAG mobilisation towards β-oxidation, which was much less pronounced in oleate-loaded cells.

Discussion

In animal models, obesity and insulin resistance in the prediabetic state are associated with increased glycerolipid/NEFA cycling in islets, a possible compensatory mechanism enhancing insulin secretion [10]. Under nutrient-rich conditions, the glycerolipid/NEFA cycle is driven by fuel availability, notably glucose and fatty acids, and the expression of related genes. Here, accumulation of neutral lipids in INS-1E beta cells following fatty acid exposure was highly dependent on glucose concentration, probably explained by the provision of the glycerol-3-phosphate backbone derived from glycolysis and necessary for lipid esterification. As shown previously [25, 40], palmitate and oleate induced different degrees of lipid storage, the unsaturated fatty acid being much more potent in this regard. Palmitate esterification relies on the Scd1-encoded stearoyl-CoA desaturase, which converts palmitate-derived stearate to oleate for further esterification [41, 42]. In INS-1E beta cells, Scd1 expression was increased upon palmitate chronic treatment at basal glucose. Conversely, oleate exposure was associated with low Scd1 expression, whatever the glucose concentration in the media. Such low Scd1 expression did not prevent massive fat storage with oleate at high glucose, the latter favouring de novo palmitate synthesis. Consequently, the provision of both endogenous saturated and exogenous unsaturated fatty acids promotes lipid storage. Such a favourable balance was also illustrated by the combination of exogenous oleate and palmitate, showing important lipid accumulation. Inhibition of lipases by Orlistat during the various chronic treatments with fatty acids induced massive lipid storage among all culture conditions, probably reaching maximal capacity of the cells. This strong effect indicates that, with active lipases, both arms of the glycerolipid/NEFA cycle were constitutively running, i.e. both lipogenesis and lipolysis, even under high glucose conditions not requiring mitochondrial β-oxidation.

Our analyses on mRNA levels revealed that the key enzymes of the glycerolipid/NEFA cycle were normally expressed in both human and rat islets as well as INS-1E beta cells. Sensitivity to high glucose was more pronounced in human islets compared with rodent cells, i.e. rat islets and INS-1E cells. Chronic treatments with fatty acids at standard glucose had rather modest effects on such expression levels. However, glucolipotoxicity induced robust changes in the expression of some genes related to signalling (FFARs) and metabolism of fatty acids, in particular the MAG lipase MGLL and the G0/G1 switch gene 2 G0S2. Regarding the latter, upregulation of this endogenous ATGL inhibitor appears contradictory to an active glycerolipid/NEFA cycle, although in accordance with the lack of effect of Atglistatin used as a selective ATGL pharmacological inhibitor, as opposed to the panlipase inhibitor Orlistat. These observations indicate that ATGL is dispensable for the glycerolipid/NEFA cycle in beta cells, being partially redundant with HSL [34].

Genes encoding DAG-acyltransferases and MAG lipases were both sensitive to glucose concentrations in the culture media, glucose inducing an elevated capacity for lipid cycling. In contrast to observations gathered in the mouse beta cell line MIN6 [40], palmitate and oleate reduced glucose-induced expression of the DAG-acyltransferases DGAT1 and Dgat2 in human islets and rat INS-1E cells, respectively. Of note, in adipocytes DGAT1 and DGAT2 exhibit overlapping functions regarding TAG storage, while DGAT1 only may protect against lipotoxicity [43]. DAG-acyltransferases and MAG lipases are key enzymes in the glycerolipid/NEFA cycle that produce secretion coupling factors (1,2-DAG and 1-MAG) in the beta cell. DAG is formed either during lipogenesis from PA or during lipolysis from direct hydrolysis of TAG. However, different species of DAG may be produced according to the reaction implicated. The hydrolysis of PA by lipins generates 1,2-DAG, whereas the hydrolysis of TAG by ATGL forms 1,3-DAG or 2,3-DAG [44, 45]. Among these three species, only 1,2-DAG qualifies as a signalling molecule. It activates PKC, which further phosphorylates key proteins of the exocytotic machinery [18, 46]. The GSIS coupling activity of 1-MAG targets insulin granules by binding to the exocytotic protein Munc13-1 [19, 47]. The role of 1-MAG has been substantiated in beta cell-specific ATGL-KO mice, in which exogenous MAG supply restores GSIS [16]. Among the MAG lipases controlling intracellular levels of 1-MAG, Abhd6 appears to play a predominant role compared with Mgll [19, 33].

Our results highlight the importance of lipolysis intermediaries. Indeed, inhibition of lipases completely blunted the acute secretory responses, in particular when potentiated by chronic exposure to oleate at high glucose. These results are in line with previous studies conducted in rat [14] and mouse [48] islets. Chronic exposure to fatty acids during the culture period resulted in the rapid mobilisation of esterified lipids during the starvation period that precedes the acute secretory response, which exhibited lipase dependent potentiation upon glucose stimulation. This effect was also observed with KCl, showing that in beta cells loaded with lipids the calcium signal is necessary and sufficient to induce exaggerated insulin release. Because inhibition of β-oxidation did not prevent the recruitment of stored lipids, this catabolic pathway is an unlikely contributor of the enhanced secretory response. Past and present data point to intermediaries of the glycerolipid/NEFA cycle for the observed potentiation of glucose-stimulated secretion. Additionally, NEFA released by lipolysis may exit the cell and activate the NEFA receptors in a paracrine and/or autocrine manner [49].

In conclusion, the study highlights the central role of the glycerolipid/NEFA cycle in the adaptation of beta cells to chronic glucolipotoxic conditions (Fig. 8). The resultant stored lipids followed by their active recruitment during GSIS are instrumental for the preservation of the secretory response to acute glucose stimulation.

Fig. 8
figure 8

Effects of palmitate and oleate on INS-1E beta cells exposed to glucose. Schematic view of palmitate (C16:0, pink) and oleate (C18:1, blue) chronic effects on INS-1E beta cell expression of lipid-related genes, lipid cycling and GSIS under low (G5.5, a), standard (G11.1, b) and elevated (G25, c) glucose conditions. Arrows show up- or downregulated genes or increased or decreased pathways according to fatty acid colour code. Ψc, cellular potential; ABHD6, abhydrolase domain containing 6; AGPAT, acyl-glycerol-3-phosphate acyltransferase; ER, endoplasmic reticulum; G3P, glycerol triphosphate; G3PP, glycerol triphosphate phosphatase; GL, glycerolipid; Glc, glucose; LTCC, L-type calcium channel