Peroxisome-Generated Hydrogen Peroxide as Important Mediator of Lipotoxicity in Insulin-Producing Cells

Insulin-producing RINm5F cells and INS-1E cells (provided by C. Wollheim) were cultured as described previously (30,31).

RINm5F cell clones that overexpressed different antioxidative enzymes were generated as described previously (32,33). Cellular expression of antioxidative enzymes was analyzed by Western blot or catalase enzyme activity measurement (33).

For the quantification of the catalase activity, the method originally described by Claiborne (34) was used. In brief, cells were homogenized in PBS (pH 7.4) through sonication on ice for 1 min in 15-s bursts at 90 watts. Catalase activity was measured by ultraviolet spectroscopy, monitoring the decomposition of H₂O₂ at 240 nm. Catalase activity was calculated by the following equation:

NEFAs (Sigma, St. Louis, MO) were dissolved in 90% ethanol heated to 60°C and used at different concentrations in RPMI 1640 (PAN, Aidenbach, Germany) with 1% FCS and a final BSA (MP Biomedicals, Eschwege, Germany): NEFA ratio of 2%: 1 mmol/l, according to Cnop et al. (13). All control wells received the same amount of solvent and BSA. This procedure did not cause a significant decrease in viability in the absence of added fatty acids.

Pancreatic islets were isolated from 250–300 g adult male Lewis rats by collagenase digestion, separated by Ficoll gradient, and handpicked under a stereo microscope. Isolated islets were cultured on extracellular matrix (ECM)-coated plates (35 mm) (Novamed, Jerusalem, Israel, the ECM being derived from bovine corneal endothelial cells) in RPMI-1640 medium containing 5 mmol/l glucose, 10% FCS, penicillin, and streptomycin at 37°C in a humidified atmosphere of 5% CO₂ according to Cornu et al. (35). The islets were cultured for 7–10 days on the ECM plates to adhere and spread before they were infected with HyPer-Peroxi lentivirus or treated with palmitic acid.

RINm5F mock-transfected and cytoprotective enzyme overexpressing RINm5F insulin-producing cells were seeded at 25,000 cells/well in 100-μl culture medium in 96-well plates (for propidium iodide [PI] staining, black 96-well plates) and allowed to attach for 24 h before they were incubated at 37°C with NEFAs for 24 h. Cell viability was then determined by either microplate-based MTT assay (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide) (36) or PI staining (Sigma, St. Louis, MO). PI is membrane impermeant and usually excluded from viable cells (37). To measure the rate of cell death after NEFA treatment, the PI stock solution (12.5 μg/ml) was diluted in the incubation media to a final concentration of 0.625 μg/ml. After 15 min incubation in the dark, the plates were analyzed at 520/620 nm excitation/emission using the fluorescence reader Victor² 1420 Multilabel Counter (Perkin Elmer, Wiesbaden, Germany).

To detect overall oxidative stress, 25,000 cells were seeded in 96-well black plates and cultured for 24 h. The cells were then preincubated with 10 μmol/l DCF-DA (2, 7-dichlorofluorescein diacetate, Sigma, St. Louis, MO) for 30 min at 37°C. Thereafter, medium containing dichlorofluorescein diacetate (DCF-DA) was replaced with fresh medium with or without palmitic acid. After a 24-h incubation, the plates were analyzed at 480/520 nm excitation/emission using the fluorescence reader Victor² 1420 Multilabel Counter (Perkin Elmer, Wiesbaden, Germany) (38). Data are expressed as percentages relative to untreated cells.

For in vivo analyses of mitochondrial H₂O₂ generation, the cDNA of the H₂O₂-sensitive fluorescent HyPer protein (39) was subcloned from the pHyPer-dMito (Evrogen, Moscow, Russia) into the lentiviral transfer plasmid pLenti6/V5-MCS (Invitrogen, Karlsruhe, Germany). The cDNA was excised from the pHyPer-dMito plasmid using the NheI and NotI restriction sites, blunted, and ligated into the EcoRV site of the pLenti6/V5-MCS plasmid. To construct the expression vector for the HyPer-Peroxi protein, the peroxisome target signal (PTS1) (40,41) was joined to the 3′-end of the HyPer cDNA by PCR using composite primer (HyPer-PTS1-XbaI-fw (5′-TATCTAGACGCCACCATGGAGATGGCAA-3′) and HyPer-PTS1-Bsp119I-rv (5′-GCTTCGAATTACAGCTTGGAAACCGCCTGTTTTAAAAC-3′)) and the pHyper-dCyto plasmid as template. Then, the HyPer-Peroxi cDNA was subcloned into the XbaI/Bsp119I site of the pLenti6/V5-MCS plasmid.

To express the HyPer-Mito and HyPer-Peroxi proteins, lentivirus was prepared as previously described (42): 500,000 293FT cells were transfected with the packaging plasmid pPAX2 (37.5 μg), the envelope plasmid pcDNA-MDG (7.5 μg), and the transfer plasmids pLenti6/V5-MCS-HyPer-Mito or pLenti6/V5-MCS-HyPer-Peroxi (25 μg) by calcium phosphate precipitation. The virus particles were harvested from the culture medium 48 h later and purified by ultracentrifugation (2 h at 70,000 g). The virus titers (3–5 × 10⁷ infectious particles) were quantified by Taqman qPCR assay as described elsewhere (43).

The RINm5F-control, RINm5F-Cat, RINm5F-MitoCat (33), INS-1E, and primary rat islet cells were infected with HyPer-Mito or HyPer-Peroxi lentivirus at a MOI of 10. The tissue culture cells were selected for HyPer expression using blasticidin (1 μmol/l).

RINm5F cells or primary rat islet cells that overexpressed HyPer-Mito or HyPer-Peroxi were seeded onto black 24-well glass-bottom plates (Greiner, Frickenhausen, Germany). Cells were cultured for 24 h and afterward exposed to palmitic acid for another 24 h. Live cell imaging was performed using a CFP-YFP dual filter (excitation, 427 nm and 504 nm; emission, 520 nm) with a cell^R/Olympus IX 81 inverted microscope system and CellR software (Olympus, Hamburg, Germany) for imaging and analysis.

To determine H₂O₂ production, changes in the fluorescence ratios of RINm5F, RINm5F-Cat, RINm5F-MitoCat, and INS-1E cells that overexpressed HyPer-Mito or HyPer-Peroxi were quantified spectrofluorometrically. Those cells were seeded at 25,000 cells per well onto black 96-well plates and cultured for 24 h. The fluorescence ratio was measured immediately before and after 24 h treatment with palmitic acid.

For immunocytochemical staining of primary rat β-cells, HepG2 cells, RINm5F cells, and RINm5F cells that overexpressed catalase, 100,000 cells were seeded overnight on collagen-coated glass slides and subsequently fixed with 4% paraformaldehyde. After washing, the cells were permeabilized and blocked with PBS containing 0.2% Triton X-100 and 1% BSA. The slides were incubated with primary antibodies (for detailed information see supplementary Table S1 in the online appendix at http://diabetes.diabetesjournals.org/cgi/content/full/db09-1401/DC1) diluted in PBS containing 0.1% Triton X-100 and 0.1% BSA at room temperature for 60 min. Then, the cells were washed with PBS and incubated with secondary antibodies (for detailed information see supplementary Table S1) for 60 min. For nuclear counterstaining, 300 nmol/l DAPI was applied for 5 min at room temperature. Finally, the cells were washed and mounted with Mowiol/DABCO (Sigma, St. Louis, MO) antiphotobleaching mounting media. Stained cells were examined on an Olympus IX81 inverted microscope (Olympus, Hamburg, Germany) and microscopic images were postprocessed using AutoDeblur and AutoVisualize (Autoquant Imaging, NY).

Data are expressed as means ± SEM. Statistical analyses were performed using ANOVA plus the Dunnett test for multiple comparisons, unless stated otherwise. EC₅₀ values were calculated from nonlinear regression analyses using least square algorithms of the Prism analysis program (Graphpad, San Diego, CA).

Saturated NEFAs were toxic to RINm5F insulin-producing cells. In both the MTT cell viability and the propidium iodide (PI) cytotoxicity assays, toxicity was dependent on NEFA chain length (Fig. 1). The longest chain length fatty acid analyzed, stearic acid (C18:0), was the most toxic, with a half-maximal effective concentration (EC₅₀) below 100 μmol/l. The EC₅₀ for palmitic acid (C16:0), the physiologically most abundant and important saturated NEFA, was around 100 μmol/l after a 24-h incubation. Toxicity decreased with decreasing chain length. Myristic acid (C14:0) was highly toxic (EC₅₀ ≅ 200 μmol/l), whereas tridecanoic acid (C13:0) was significantly less toxic, with an EC₅₀ well above 500 μmol/l. With further chain length shortening (C12:0 to C10:0), toxicity was negligible with EC₅₀ values exceeding 1,000 μmol/l. For even shorter chain saturated fatty acids (C8:0 to C4:0), EC₅₀ values could not be calculated because of marginal toxicity (data not shown). Toxicity was also observed in a fluorescence-activated cell sorter (FACS)-based caspase-3 assay (30% caspase-3 positive cells at 100 μmol/l palmitic acid as compared with 6% caspase-3 positive cells under control conditions), indicating an apoptotic mode of fatty acid-induced cell death.

Palmitic acid-induced cell death was also observed in isolated primary rat islet cells in the propidium iodide cytotoxicity assay. After a 24-h incubation with 500 μmol/l palmitic acid, 33 ± 4% of rat islet cells stained positive for propidium iodide, whereas in untreated cells the cell death rate was low (6 ± 1%).

Although primary rat islet cells as well as untransfected RINm5F insulin-producing cells showed virtually no immunostaining for catalase (data not shown), RINm5F insulin-producing cells transfected to overexpress catalase in the cytosol (Cat) showed a homogenous distribution of catalase in the cytoplasm and a dot-like pattern of catalase colocalization with peroxisomes (Fig. 2,A); on the other hand, catalase did not colocalize with mitochondria (Fig. 2,B). Catalase was not detectable in the cytoplasm or in the peroxisomes of insulin-producing cells that overexpressed catalase in the mitochondria (Mito-Cat; Fig. 2,C); in these cells, catalase colocalized specifically with the mitochondria (Fig. 2 D).

The localization of catalase in RINm5F-Cat cells was additionally analyzed in three dimensional (3D) images using Imaris software (Bitplane, Zurich, Switzerland). The 3D models were generated from 20 z-layers of the immunocytochemical staining of catalase and the peroxisomes (supplementary Fig. S1) or the mitochondria (supplementary Fig. S2). In these images, catalase showed a clear colocalization with the peroxisomes, whereas no significant colocalization was detectable with the mitochondria.

The results of these immunostainings were also reflected by the catalase enzyme activities of the examined cells. Primary rat islet cells and untransfected RINm5F cells showed a comparable catalase activity of 20 ± 5 units/mg and 13 ± 1 unit/mg, respectively. By contrast in RINm5F-catalase (353 ± 7 units/mg) and RINm5F-mito-catalase (367 ± 43 units/mg) cells the enzyme activity was ∼17-fold higher.

RINm5F insulin-producing cells that overexpressed different antioxidative enzymes were used as a tool to analyze the role of ROS in palmitic acid toxicity. Although overexpression of the H₂O₂-inactivating enzyme catalase in peroxisomes and in the cytosol (Cat) protected against palmitic acid (100 μmol/l) toxicity, catalase overexpression in the mitochondria (Mito-Cat) did not provide protection (Fig. 3).

Neither overexpression of the cytosolic superoxide radical detoxifying isoenzyme CuZnSOD (copper zinc superoxide dismutase) nor overexpression of the mitochondrial isoenzyme MnSOD (manganese superoxide dismutase) provided significant protection against palmitic acid toxicity.

ROS generation in RINm5F insulin-producing control cells as measured by the DCF fluorescence method was increased by about 40% in response to exposure to 100 μmol/l palmitic acid (Fig. 4). Overexpression of catalase in the peroxisomes and the cytosol, but not in the mitochondria, significantly reduced palmitic acid-induced (100 μmol/l) ROS generation.

Peroxisomes were clearly detectable by immunofluorescence as small green spots in primary rat β-cells, RINm5F insulin-producing cells, and HepG2 cells (supplementary Fig. S3). β-Cells isolated from rat pancreatic islets were identified by immunofluorescent staining of insulin; RINm5F cells were also positive for insulin. To quantify the density of peroxisomes in insulin-producing and hepatoma cells, the peroxisomes were counted in β-cells, RINm5F cells, and HepG2 cells after immunfluorescence staining of the peroxisomes by PMP-70 in 5–8 images (supplementary Fig. S4). The number of peroxisomes per cell was determined by the image analysis software Imaris (Bitplane, Zurich, Switzerland). The results clearly show that there were no significant differences in peroxisome density in hepatoma and in insulin-producing RINm5F cells as well as in primary rat β-cells.

To prove the intracellular localization of the HyPer protein targeted to the peroxisomes in RINm5F cells, fixed cells were stained for the peroxisomal membrane protein 70 (PMP-70). The fluorescence (504/520 nm) of the HyPer protein was clearly detectable as green spots after fixation with paraformaldehyde. Staining of the peroxisomes (PMP-70) showed a distinct colocalization with the HyPer-Peroxi protein (supplementary Fig. S5 A–C).

In order to obtain more information about the nature of the reactive oxygen species and the intracellular site of ROS production after palmitic acid exposure, the HyPer protein, which is a specific sensor for H₂O₂, was expressed in either the peroxisomes (RINm5F-HyPer-Peroxi) or the mitochondria (RINm5F-HyPer-Mito) of RINm5F insulin-producing cells. The HyPer protein allows ratiometric measurement of H₂O₂ by fluorescence microscopy at two different wavelengths. In the false-colored overlay images, green staining corresponds to low, yellow to medium, and red to high H₂O₂ concentrations. After 24 h of exposure to palmitic acid, RINm5F-HyPer-Peroxi cells showed a strong shift toward red fluorescence (Fig. 5,C and D). The quantification revealed a significant 240% increase of H₂O₂ production after a 24-h palmitic acid treatment (Fig. 5,E). In contrast, the RINm5F-HyPer-Mito cells showed only a slight nonsignificant increase in red fluorescence, indicating a lower H₂O₂ concentration in the mitochondria relative to the peroxisomes (Fig. 5 A, B, and E). RINm5F-HyPer-Peroxi and RINm5F-HyPer-Mito cells incubated without palmitic acid under control conditions for 24 h showed no changes in fluorescence signals (data not shown).

The key finding of H₂O₂ generation in the peroxisomes of RINm5F cells after exposure to palmitic acid was confirmed in primary rat islet cells. Single rat islet cells were lentivirally transduced with Hyper-Peroxi cDNA and incubated with palmitic acid for 24 h. In the palmitic acid treated islet cells a clear shift toward red fluorescence was detectable, which corresponds to a higher H₂O₂ concentration in comparison with untreated cells (Fig. 6). The fluorescence shift of single cells was quantified with the CellR software (Olympus, Hamburg, Germany). Palmitic acid treatment induced a significant 125% increase of H₂O₂ production in comparison with the control condition (Fig. 6 E).

Differences in peroxisomal and mitochondrial H₂O₂ concentration after 24-h exposure to palmitic acid were quantified spectrofluorometrically. Treatment with 100 μmol/l palmitic acid induced a significantly higher rate of H₂O₂ production in both the peroxisomes and the mitochondria of control RINm5F cells than did treatment with 50 μmol/l palmitic acid (Fig. 7,A and B). At 100 μmol/l palmitic acid, there was a tendency for a greater increase of the peroxisomal as compared with the mitochondrial H₂O₂ production in RINm5F cells (Fig. 7,A and B). In INS-1E cells, the H₂O₂ production was even significantly higher in the peroxisomes than in the mitochondria after palmitic acid treatment (Fig. 8).

Additional information about the intracellular site of H₂O₂ production after palmitic acid exposure was obtained by overexpressing the H₂O₂ detoxifying enzyme catalase either in the peroxisomes and the cytosol or in the mitochondria of RINm5F-HyPer-Peroxi and RINm5F-HyPer-Mito cells, respectively. Catalase expression in the peroxisomes and cytosol of RINm5F-HyPer-Peroxi cells reduced H₂O₂ production significantly in the presence of 50 and 100 μmol/l palmitic acid as compared with control cells, whereas overexpression of catalase in the mitochondria did not significantly reduce H₂O₂ production in the peroxisomes (Fig. 7 A).

Overexpression of catalase in the peroxisomes and the cytosol significantly reduced the mitochondrial H₂O₂ concentration in response to incubation with 100 μmol/l (and to a lesser extent with 50 μmol/l) palmitic acid as compared with control cells. Overexpression of catalase in the mitochondria did not significantly affect mitochondrial production of H₂O₂ even though there was a tendency toward a reduction at 100 μmol/l palmitic acid (Fig. 7 B).

Overall, the data showed that overexpression of catalase in peroxisomes and cytosol reduced H₂O₂ generation in the peroxisomes and, to a lesser extent, in the mitochondria. In contrast, overexpression of catalase in the mitochondria had no significant effect on H₂O₂ production in either organelle (compare Fig. 7 A and B).

The present data clearly show that NEFA toxicity increases with increasing chain length. NEFAs are oxidized not only in the mitochondria, but also in the peroxisomes. Fatty acid β-oxidation is a property of peroxisomes in most, if not all, organisms (27). Moreover, in yeast and plants, peroxisomes are the sole sites of fatty acid β-oxidation (44). One product of peroxisomal β-oxidation is H₂O₂, whereas mitochondrial β-oxidation generates reducing equivalents (27,28). Mitochondrial β-oxidation is tightly coupled to the respiratory chain and oxidative phosphorylation. It provides acetyl-CoA for further oxidation in the TCA cycle and ensures the production of ATP, which is not an apparent function of peroxisomal β-oxidation. In peroxisomal β-oxidation, the initial step generates H₂O₂ and energy is lost as heat. Peroxisomal β-oxidation results in chain shortening of long- and very long-chain fatty acids, which are poor substrates for mitochondrial β-oxidation. The shortened fatty acids are subsequently transported to the mitochondria in a carnitine-dependent manner for further degradation (27,28).

The observation of the chain length-dependent increase in the toxicity of saturated fatty acids to insulin-producing cells prompted us to study the role of oxygen free radicals generated by peroxisomal β-oxidation as mediators of lipotoxicity. In the first step of peroxisomal β-oxidation, FAD-containing acyl-CoA oxidase introduces a double bond at the β-position of the fatty-acyl-CoA ester and the hydrogen atoms are transferred to molecular oxygen to yield H₂O₂ (28). Remarkably, in rat liver, for example, about 20% of total oxygen consumption is accounted for peroxisomal oxidase activity (45). For the detoxification of H₂O₂, the oxidoreductase catalase, which has a high turnover rate, is expressed in the peroxisomes of most tissues (28), but not in those of pancreatic β-cells (29,30). This low catalase enzyme activity could also be found in RINm5F cells, which makes these cells well suited as model cells for H₂O₂-mediated lipotoxicity.

Experiments with insulin-producing cells overexpressing antioxidative enzymes showed that only cytosolic catalase provided protection against palmitic acid-induced toxicity; mitochondrial catalase was not protective. The superoxide radical detoxifying isoenzymes MnSOD and CuZnSOD were also not protective, indicating that the formation of superoxide radicals does not play as crucial a role in lipotoxicity as does the formation of H₂O₂. Indeed, determination of ROS production by DCF fluorescence measurements provided support for this hypothesis by showing a reduction in ROS generation in insulin-producing cells that overexpressed catalase in the peroxisomes and the cytosol, but not in the mitochondria. Using the H₂O₂-sensitive HyPer protein as a novel, specific method for detecting H₂O₂, we clearly identified H₂O₂ as the main reactive oxygen species formed during palmitic acid treatment. To determine the subcellular site of H₂O₂ formation, the HyPer protein was fused to a peroxisome- or a mitochondrion-targeting sequence to allow organelle-specific expression.

These experiments show that peroxisomes were a major site of H₂O₂ formation in insulin-producing cells, whereas the mitochondria were a minor site. Primary rat islet cells are well equipped with peroxisomes, and also showed an increased peroxisomal H₂O₂ formation in response to palmitic acid, as demonstrated in this study.

An additional argument for this hypothesis is the fact that overexpression of catalase in the peroxisomes and the cytosol (RINm5F-Cat) significantly reduced H₂O₂ production not only in the peroxisomes, but also in the mitochondria, though catalase was not overexpressed in the latter organelle. H₂O₂ as a membrane-permeable ROS (23,24) can diffuse from its site of generation in the peroxisome into the mitochondria where it is detected through the HyPer-Mito protein. Thus, the source of the elevated H₂O₂ concentration detected in the mitochondria is at least in part H₂O₂-generated in the peroxisomes.

However, in contrast to the peroxisomes in other cell types, insulin-producing cells do not appear to express catalase mRNA or protein (29,30). This lack of catalase expression leaves insulin-producing cells badly protected against potentially hazardous effects of H₂O₂ generated through peroxisomal β-oxidation. Mitochondrial β-oxidation may not be able to cope with the elevated levels of NEFAs that are associated with obesity and type 2 diabetes, resulting in a higher proportion of fatty acids being metabolized through peroxisomal β-oxidation, leading to increased H₂O₂ formation. The membrane-permeable H₂O₂ could leave the peroxisomes and harm insulin biosynthesis and secretion (11,46,47), as has been shown for insulin-secreting cells after exposure to high NEFA concentrations for extended periods (15). The results of the present study indicate that ROS generated in the peroxisomes are the major cause of lipotoxicity-mediated β-cell dysfunction. This does not detract from the fact that mitochondrial ROS formation may contribute to this phenomenon, in particular with respect to direct negative effects on mitochondrial function.

NADPH oxidase activation could also be a source of long-chain fatty acid-mediated superoxide radical generation (22) that has been postulated to cause lipotoxicity in insulin-producing cells (15,48). Palmitic acid has been shown to activate this plasma membrane enzyme in aortic smooth muscle and endothelial cells in a protein kinase C-dependent manner, leading to superoxide radical formation (49). In insulin-producing cells, palmitic acid can induce superoxide radical formation along with increased expression of the NADPH oxidase p47^phox subunit (15,48,50). However, data from the present study indicated that cytosolic CuZnSOD did not protect insulin-producing cells against palmitic acid-induced toxicity, suggesting that an increased rate of superoxide radical formation per se is insufficient to explain NEFA-induced lipotoxicity.

Thus, we hypothesize that NEFA-induced lipotoxicity is mediated by H₂O₂ generated during peroxisomal β-oxidation of palmitic acid as the physiologically most abundant long-chain saturated fatty acid. This provides an interesting new concept for lipid-induced glucose intolerance in obesity and diabetic hyperglycemia as a result of β-cell dysfunction in patients with type 2 diabetes. The lack of catalase expression in pancreatic β-cell peroxisomes (29,30) explains the exceptional susceptibility of pancreatic β-cells to lipotoxicity (14,15,23).

This work was supported by the European Union (Integrated Project EuroDia LSHM-CT-2006-518153 in the Framework Program 6 [FP6] of the European Community).

No potential conflicts of interest relevant to this article were reported.

M.E. and W.G. wrote the manuscript and researched data. S.L. wrote the manuscript.

The authors are grateful to Maren Böger for her skillful technical assistance.