Pancreatic β-Cell Lipotoxicity Induced by Overexpression of Hormone-Sensitive Lipase

A pBluescript vector containing the rat adipocyte HSL cDNA (nucleotides 610–3,225) downstream of the scavenger receptor gene promoter was used as starting plasmid (kind gift from Drs. Karen Reue and Jean-Louis Escary, VA Lipid Research Laboratory, Los Angeles, CA). After excision of the scavenger receptor gene promoter, the rat insulin II gene promoter, obtained from a pBR322/RIP vector (kind gift from Dr. Henrik Semb, Göteborg University, Göteborg, Sweden), was subcloned into the vector, resulting in a 4.5-kb fragment, harboring the RIP-rHSL construct with a human growth hormone tail (Fig. 1A). This fragment was injected into the male pronucleus of fertilized eggs from superovulated (C57BL/6J × CBA/J) females that had been mated to males with the same genetic background. Genotyping was performed by PCR from tail biopsies (14), using primers specific for the rat insulin promotor (5′-GAAACTGCAGCTTCAGCCCCTCTGGC-3′) and HSL (5′-CAGTGCCTGTTCCCGAACACCTGCAAAGAC-3′), respectively.

Animals were housed in temperature-controlled facilities on a 12-h light-dark cycle. The mice had free access to food and water and were fed either a high-fat diet with 58% fat by energy or a low-fat diet with 11% fat by energy (Research Diets, New Brunswick, NJ). Blood samples were taken from the orbital capillary plexus, and plasma insulin, glucagon, and leptin were determined radioimmunochemically (Linco Research). Plasma glucose was determined with the glucose oxidase method (Sigma) and free fatty acids (FFAs) were measured photometrically (NEFA C; Wako Chemicals). The study was approved by the Animal Ethics Committee at Lund University.

Islets were isolated by collagenase digestion as described (15) and then lysed in SDS-PAGE sample buffer. Other tissues were homogenized in 10 volumes of homogenization buffer (0.25 mol/l sucrose, 1 mmol/l dithioerythritol, 1 mmol/l EDTA, 10 μg/ml leupeptin, 10 μg/ml antipain, 1 μg/ml pepstatin A) using a polytron. Total protein was measured (BCA protein, Pierce), and 100 μg of protein were fractionated by SDS-PAGE and electroblotted onto nitrocellulose membranes (Hybond-c extra, Amersham Pharmacia Biotech). Primary antibodies used were affinity-purified rabbit anti-rat HSL rabbit anti-mouse GLUT-2 (Chemicon), affinity-purified rabbit anti-human UCP2 (antibody 2–605) (16), rabbit anti-rat liver CPT-I (17), rabbit anti-rat acyl-CoA oxidase (a kind gift from Dr. Stefan Alexson, Karolinska Institute, Stockholm, Sweden), and rabbit anti-rat PDX-1/IPF1 (kind gift from Drs. Palle Serup and Ole D. Madsen, Novo Nordisk, Bagsvaerd, Denmark). The secondary antibody was a horseradish peroxidase-conjugated anti-rabbit IgG prepared in goat (Amersham Pharmacia Biotech) or donkey (Pierce). The blots were developed by enhanced chemiluminescene (SuperSignal, Pierce), and detection was performed using a CCD camera (LAS 1000; Fuji).

Total RNA was isolated from islets (18), electrophoresed, and transferred to nylon membranes (Hybond-N+). The blots were probed with a ³²P-labeled mouse UCP2 cDNA using ExpressHyb (Clontech). Membranes were analyzed and quantified by digital imaging (Fujix Bas 2000; Fuji).

Pooled total RNA isolated from islets was DNase-treated, and three independent reverse transcription reactions were performed using Superscript II reverse transcriptase (Invitrogen) following the manufacturer’s instructions. The mRNA expression levels were determined using real-time fluorescent detection and the following primer combinations: 5′-GTACCACTACGGAGTTCACGCAT-f-3′ and 5′-CGCCGAAAGAAGCCCTTAC-r-3′ for peroxisome proliferator-activated receptor (PPAR)-α, 5′-CAGATAATTGGCACCTACGCC-f-3′ and 5′-AAGATGAGTTCCGTGGCCC-r-3′ for acylCoA oxidase, 5′-TCTGGATGGCTATGGTCAAG-f-3′ and 5′-TACCACTGCAATTTGGGTCC-f-3′ for CPT-I, 5′-AGCTGCTAGTGGAGCACCAAG-f-3′ and 5′-TCGCCATTTCTGCGAGC-r-3′ for medium-chain acyl dehydrogenase (MCAD), and 5′-CCTACAGCGCCAGATGAGCT-f-3′ and 5′-GAGTCGTAGAGGCCAATGCG-r-3′ for UCP2. The mRNA expression levels were determined twice in each first-strand synthesis reaction and normalized to the expression levels of 18S rRNA as described by the vendor (Applied Biosystems).

Glucose tolerance tests were performed in anesthetized mice after intravenous administration of d-glucose (1 g/kg) or arginine (0.25 g/kg). Blood samples were taken at 0, 1, 5, 20, and 50 min.

Insulin secretion from isolated islets was measured as described (15). Total insulin content in islets was measured after extraction of insulin with acidic ethanol (0.2 mol/l HCl in 75% ethanol) (Linco).

For determination of ATP levels, batches of 10 islets were incubated in RPMI-1640 medium containing 16.7 mmol/l glucose at 37°C for 5 min. The ATP concentration was measured using a luciferase-based assay (ATP monitoring kit; Thermo Labsystems).

Acylglycerols were extracted from isolated islets (19). After hydrolyzation, the amount of released glycerol was measured using a bioluminescence assay (20).

Homogenates, prepared as for Western blot analysis, were assayed for diglyceride and triglyceride lipase activity, as described (21) using mono-oleoyl-2-O-monooleylglycerol, a diacylglycerol analog, and triolein, respectively, as substrates. One unit of enzyme activity corresponds to 1 μmol of fatty acid released per minute at 37°C.

Lipolysis was measured, in groups of 50 islets incubated with 11 mmol/l glucose and 10 μmol/l forskolin for 30 min at 37°C, as release of glycerol using a bioluminescence assay (20).

Glucose oxidation was determined as production of ¹⁴CO₂ from [U-¹⁴C]glucose (22), and palmitate oxidation was determined as production of ¹⁴CO₂ from [1-¹⁴C]palmitate (23).

Pancreatic tissue was prepared and used for indirect immunofluorescence (24) using affinity-purified rabbit anti-rat HSL, anti-insulin and anti-glucagon (Euro-Diagnostica), and rabbit anti-mouse GLUT2 (Chemicon) as primary antibodies and a secondary antibody coupled to fluorescein isothiocyanate (Euro-Diagnostica).

All results are expressed as means ± SE. Differences between individual data were compared with unpaired, two-tailed t test, where the critical probability to determine significance was P < 0.05.

A vector containing the rat insulin gene promoter and a rat HSL cDNA (Fig. 1A) was microinjected into zygotes. Seven male and nine female founder mice were obtained. After breeding the male founders to C57BL/6J female mice, six generated offspring. After analysis of the HSL expression in islets by Western blot, one founder line showing ∼50-fold overexpression compared with matched littermates (Fig. 1B) was chosen for the studies described in this paper. Following the complete phenotypic characterization of this line, an additional line with the same degree of overexpression was subjected to a limited characterization in order to confirm critical aspects of the phenotype (see last paragraph of results).

To confirm that the overexpression was confined to β-cells, the expression of HSL was analyzed by Western blot in several tissues, i.e., white and brown adipose tissue, skeletal muscle, liver, spleen, kidney, stomach, intestine, lung, heart, brain, and genitals. No differences in the expression pattern, for tissues other than islets, were observed between transgenic and wild-type mice (selected tissues are shown in Fig. 1B).

Diglyceride lipase activity was found to be elevated in islets of transgenic mice (Fig. 1C). Upon stimulation of isolated transgenic islets with the cAMP-elevating agent forskolin, triglyceride lipase activity (Fig. 1C) and the release of glycerol (an end product of lipolysis) were increased. Transgenic islets stimulated with forskolin secreted 88 ± 36 pmol glycerol · islet⁻¹ · 30 min⁻¹, while there was no detectable glycerol in medium samples from incubations with wild-type islets.

Body weight and plasma levels of glucagon, leptin, and free fatty acid (FFA) were not different between transgenic and wild-type mice fed a low-fat diet (Table 1 and data not shown). Insulin levels were significantly elevated in young transgenic male mice (7 weeks) and were accompanied by reduced blood glucose. On the contrary, in older male transgenic mice (11–20 weeks), insulin levels were lower (transgenic 121 ± 14 pmol/l [n = 25] and wild-type 270 ± 39 pmol/l [n = 20]; P < 0.001), with no difference in blood glucose between the groups (transgenic 7.6 ± 0.3 mmol/l and wild-type 7.7 ± 0.3 mmol/l). There were no significant differences in insulin or glucose for female mice at any of the time points.

For high-fat-fed male mice, there were no differences in body weight between transgenic and wild-type mice, while for female mice the increase in body weight was greater for transgenic than for wild-type mice (Table 1). In transgenic male mice, basal insulin levels were lower than in wild-type mice, while no differences in glucose levels were observed (Table 1). In female transgenic mice, insulin levels were not different from high-fat-fed control mice. As for the low-fat-fed animals, there were no differences between the transgenic and wild-type high-fat-fed mice with regard to glucagon, leptin, and FFA (Table 1 and data not shown). Thus, male transgenic mice in both diet groups appeared to exhibit increased insulin sensitivity. This may reflect a possible feedback between β-cells and insulin sensitivity, which deserves further studies.

Intravenous glucose tolerance tests (IVGTTs) performed on mice fed a low- or high-fat diet showed a significant reduction in acute insulin release in both male and female transgenic mice (Fig. 2A–D). In female transgenic mice at both feeding conditions and in high-fat-fed transgenic males, this was accompanied by impaired glucose elimination, while in low-fat-fed transgenic males, glucose elimination was normal (Fig. 2E–F). Both male and female transgenic mice were found to respond adequately to the arginine challenge (Fig. 2G–J), indicating that the exocytotic machinery is not affected in the transgenic mice (Fig. 2K–L). The insulin response to arginine was increased in high-fat-fed transgenic mice due to the elevated blood glucose.

Insulin secretion from isolated islets was analyzed after stimulation with glucose at different concentrations. Islets from transgenic female mice exhibited a significantly lower insulin response to high glucose concentrations compared with islets from wild-type mice at both feeding conditions (Fig. 3A). In islets from male mice, the blunted response to glucose was less pronounced during low-fat feeding (Fig. 3B) and reached statistical significance only in the high-fat-fed group.

There was no significant difference in triglyceride content between transgenic and wild-type islets from mice fed the low-fat diet (Fig. 3C). High-fat feeding induced a more than threefold increase in islet triglyceride content in the wild-type group; this increase was significantly lower in the transgenic group.

There were no differences in glucose oxidation among the four experimental groups at 2.8 mmol/l glucose (Fig. 3D). At 16.7 mmol/l glucose, glucose oxidation was slightly reduced in islets from transgenic animals of the low-fat-fed group compared with islets of wild-type mice of the same group. This effect was much more pronounced and statistically significant in islets from mice fed the high-fat diet. With regard to palmitate oxidation, this was slightly increased at low glucose in high-fat-fed wild-type islets and in transgenic islets of both feeding groups (Fig. 3E). There was no difference in palmitate oxidation at high glucose between transgenic and wild-type mice fed either low- or high-fat diet.

The expression levels of proteins of particular relevance were investigated by Western blot analyses. In view of the decreased glucose oxidation observed in islets of transgenic animals, we measured the levels of GLUT2 and glucokinase. Both of these were found to be upregulated in islets of transgenic mice compared with wild-type islets and accompanied by an increased expression of pancreatic/duodenal homeobox-1 (PDX-1), a known transactivator of these genes (25,26) (Fig. 4A).

The expression levels of carnitine palmitoyltransferase I (CPT I), which transports long- and medium-chain acyl-CoA into the mitochondria, was found to be upregulated in wild-type mice fed the high-fat diet compared with the low-fat diet, while high-fat-fed transgenic mice had a similar expression of CPT I in islets as low-fat-fed wild-type and transgenic mice (Fig. 4A).

It was recently demonstrated that UCP2 negatively regulates insulin secretion (13,27). Upregulation of UCP2 by fatty acids was furthermore suggested to play an important role in obesity-induced β-cell dysfunction (13). We thus examined UCP2 protein levels in the islets of HSL transgenic mice and found these to be higher than in islets from wild-type mice (Fig. 4A). Recent studies point to the involvement of PPARs in the upregulation of UCP2 by fatty acids in β-cells (28–30), and acyl CoA oxidase, a known target for PPARα regulation (31), was shown to be upregulated in islets from transgenic mice of both feeding groups compared with the respective wild-type controls. Upregulation of PPARα-responsive genes in transgenic islets was further confirmed at the mRNA level using quantitative real-time PCR. Acyl-CoA oxidase, MCAD, and UCP2 were all found to be upregulated in transgenic islets from both diet groups compared with wild-type islets (Fig. 4B). The increased UCP2 mRNA levels in transgenic islets were further verified in Northern blot experiments (Fig. 4C). To confirm that the upregulation of UCP2 has a functional relevance, we measured the levels of cellular ATP and found these to be reduced in transgenic islets compared with wild-type islets under both feeding conditions (Fig. 4D).

No gross abnormalities in general islet architecture were observed in HSL transgenic mice. As expected, HSL immunofluorescence was markedly increased in islets of transgenic mice, with a staining pattern compatible with overexpression in β-cells (Fig. 5B and D). There was no staining of exocrine pancreas. Staining for insulin (Fig. 5E–H) and glucagon (Fig. 5I–L) revealed no differences between transgenic and wild-type islets. Moreover, insulin content was similar in freshly isolated islets from both low-fat-fed (wild-type 104 ± 17 ng/islet and transgenic 115 ± 11 ng/islet) and high-fat-fed (wild-type 90 ± 10 ng/islet and transgenic 93 ± 7 ng/islet) mice. Immunofluorescence for GLUT2 was similar in wild-type and transgenic islets from low-fat-fed mice, and the pattern of staining was indistinguishable in transgenic and wild-type islets, displaying the typical plasma membrane localization of the transporter (Fig. 5M–P). In islets from high-fat-fed mice there was a disrupted, sometimes cytoplasmic, distribution of GLUT2, leaving only few islet cells with a normal plasma membrane localization, which is an early sign of cellular dysfunction following triglyceride accumulation described in islets of ZDF rats (32). This redistribution of GLUT2 was observed in both transgenic and wild-type islets.

To rule out the possibility that the phenotype described above is due to the disruption of an endogenous gene at the site of insertion or transgenic expression at some other site than β-cells, critical aspects of the phenotype were confirmed in an additional transgenic line with the same degree of overexpression (i.e., 50-fold). Also, mice of this second line developed severely blunted GSIS, when challenged with a high-fat diet, as evident from both IVGTT (Fig. 6A) and batch incubations of isolated islets (Fig. 6B). Furthermore, UCP2 was found to be upregulated in transgenic islets compared with wild-type islets (Fig. 6C). Thus, these studies confirm that the observed phenotype is the result of overexpression of HSL specifically in β-cells.

The present study demonstrates that increased expression of HSL in β-cells results in impaired GSIS. Transgenic mice overexpressing HSL specifically in β-cells exhibit glucose intolerance and abrogated GSIS. The phenotype was mild in mice fed a low-fat diet but further accentuated in mice fed a high-fat diet, despite the fact that overexpression of HSL partly prevented the accumulation of islet triglycerides that occurs upon fat feeding. Thus, the rate of influx of fatty acids into the islet triglyceride pool and the capacity to mobilize this pool are both important determinants of islet lipotoxicity. In fact, the recently demonstrated inverse correlation between cytotoxicity and triglyceride accumulation in β-cells led the authors to propose that the ability of normal β-cells to form and accumulate cytoplasmic triglycerides might serve a cytoprotective mechanism against FFA-induced islet dysfunction and apoptosis (33). The phenotype of our mice strongly supports this notion. Further support comes from a separate study (34) in which we show that the expression of islet HSL is downregulated during prolonged high-fat feeding of C57BL/6J mice. Thus, downregulation of HSL and possibly other, as yet unidentified, lipases in β-cells might be part of a defense mechanism against development of lipid-induced islet dysfunction.

HSL is the only known mammalian lipase for which hormonal regulation, mediated by reversible protein phosphorylation, has been firmly established (11). Thus, elevations in cAMP levels, resulting from glucose metabolism and binding of GLP-1 and other cAMP-elevating incretins to their respective receptors, may activate HSL via phosphorylation by protein kinase A, thus resulting in increased hydrolysis of stored acylglycerols. In the proposed model (Fig. 7), the generation of fatty acids and other lipolytic products are under normal conditions tightly controlled and hence dynamic and may play a role in providing the lipid signal that appears to be necessary for GSIS (35). Support for a role of HSL in GSIS has been provided by studies of islets of HSL knockout mice, which displayed increased basal insulin secretion and blunted GSIS (36). When the size of the triglyceride pool increases substantially, as during high-fat feeding, the supply of lipolytic products as a result of basal, i.e., cAMP-independent, lipolysis becomes significant. Even further increases in the mobilization of fatty acids may be caused by increased expression of HSL, as is the case in the HSL transgenic mice, and/or by dysregulation of the enzyme. At a certain point, the chronic increase in fatty acid flux, caused by triglyceride overstorage and/or increased HSL action, exceeds a threshold level, a point at which the β-cell enters the lipotoxic pathway (Fig. 7).

A possible mechanism behind lipotoxicity is changes in the expression of critical genes, induced by fatty acids, acyl CoA, or other fatty acid metabolites (3,37). In a hypothesis-driven attempt to identify such changes, the expression of UCP2 was found to be increased in transgenic compared with wild-type islets under both feeding conditions. Recent data have clearly established UCP2 as a negative regulator of insulin secretion. Impaired GSIS has been demonstrated in rat islets overexpressing UCP2 (27), and in UCP2 null mice, GSIS was found to be increased (13). According to the proposed mechanism, UCP2 activation decreases mitochondrial membrane potential, resulting in reduced ATP levels and increased activity of ATP-sensitive K⁺ channels, which in turn impairs GSIS (27). It has been proposed that UCP2-dependent mechanisms play an important role in obesity-induced β-cell dysfunction, and there is much to suggest that fatty acids may be responsible for the upregulation of UCP2 observed in this disease (13). Upregulation of UCP2 by fatty acids has been demonstrated in cultured β-cells (38–40) as well as in isolated rat islets (29). Moreover, UCP2 has been shown to be upregulated in ob/ob mice as well as in rodents fed a high-fat diet (13,27). PPARα (41,42) as well as PPARγ (29) are expressed in pancreatic islets, and both have been implicated in mediating the fatty acid-induced upregulation of UCP2 (28,30). Sterol regulatory element-binding protein is another transcription factor that has been implicated in the regulation of UCP2 by fatty acids (40,43), presumably operating downstream of PPAR (40). In islets of transgenic HSL mice, decreased ATP levels accompanied the increased UCP2 levels. The UCP2 protein levels were increased by the high-fat diet in the HSL transgenic mice. Several PPARα-responsive genes, as well as PPARα itself, were upregulated in islets from HSL transgenic mice compared with the respective wild-type controls. Thus, we hypothesize that a PPARα-mediated upregulation of UCP2 at least partly explains the impaired GSIS observed in mice with β-cell-specific overexpression of HSL. Furthermore, we propose that fatty acids, mobilized from the triglyceride stores through the action of HSL, directly or after derivatization to some other lipid component, are responsible for the upregulation of UCP2 (Fig. 7). It is possible that additional alterations in gene expression, induced by the increased flux of ligands to lipid-activated transcription factors or by alternative routes for fatty acid-induced changes in gene expression (37), also contribute to the phenotype.

The expression of both GLUT2 and glucokinase were found to be upregulated in islets of transgenic mice compared with wild-type islets. The reason for this, as well as the mechanism(s) involved, is not known. It seems unlikely that fatty acids are responsible for the upregulation because exposure of islets to palmitate has been shown to decrease the expression of both GLUT2 and glucokinase via downregulation of the transcription factor PDX-1 (26), a known transactivator of both GLUT2 (25) and glucokinase (44). Since we found PDX-1 levels to be elevated in transgenic islets, it is possible that the effects on GLUT2 and glucokinase observed in the HSL transgenic mice are mediated via PDX-1. Despite the increase in GLUT2 and glucokinase expression, we found glucose oxidation to be significantly decreased in islets of high-fat-fed transgenic mice. There was an increase in palmitate oxidation in high-fat-fed wild-type islets, which was less pronounced in transgenic islets. With regard to transportation of fatty acids into the mitochondria, CPT I was found to be upregulated in islets from high-fat-fed wild-type mice, while CPT I expression was not upregulated in HSL transgenic islets. It is thus possible that the mechanism behind the observed reduction in GSIS in HSL transgenic mice is due to reduced ATP levels, caused not only by increased expression of UCP2 but also by reduced glucose oxidation. Increased generation of intracellular long-chain acyl-CoA, due to increased activity of HSL, may result in allosteric activation or inactivation of metabolic enzymes leading to deteriorated β-cell function.

This work was supported by the Swedish Research Council (project nos. 112 84 [to C.H.], 121 86 [to S.E.], 4499 [to F.S.], 1034 [to P.A.], and 6834 [to B.A.]); the Center of Excellence Grant of the Juvenile Diabetes Foundation, U.S.; the Knut and Alice Wallenberg Foundation, Sweden; the Swedish Diabetes Association; and the following foundations: Novo Nordisk, A. Påhlsson, Arne and IngaBritt Lundberg, Crafoord, Thuring, Swedish Nutrition Foundation, Socialstyrelsen, and T. Zoega.

We thank Sara Larsson, Ann-Helen Thorén, Lena Kvist, and Lillian Bengtsson for excellent technical assistance and Dr. Giovanni Patane at the Department of Internal Medicine, University of Catania, Ospedale Garibaldi, Italy, for helpful advice on how to prepare islets for Western blot analysis for UCP2.