Free Fatty Acids Induce JNK-dependent Hepatocyte Lipoapoptosis*

Elevated serum free fatty acids (FFAs) and hepatocyte lipoapoptosis are features of non-alcoholic fatty liver disease. However, the mechanism by which FFAs mediate lipoapoptosis is unclear. Because JNK activation is pivotal in both the metabolic syndrome accompanying non-alcoholic fatty liver disease and cellular apoptosis, we examined the role of JNK activation in FFA-induced lipoapoptosis. Multiple hepatocyte cell lines and primary mouse hepatocytes were treated in culture with monounsaturated fatty acids and saturated fatty acids. Despite equal cellular steatosis, apoptosis and JNK activation were greater during exposure to saturated versus monounsaturated FFAs. Inhibition of JNK, pharmacologically as well as genetically, reduced saturated FFA-mediated hepatocyte lipoapoptosis. Cell death was caspase-dependent and associated with mitochondrial membrane depolarization and cytochrome c release indicating activation of the mitochondrial pathway of apoptosis. JNK-dependent lipoapoptosis was associated with activation of Bax, a known mediator of mitochondrial dysfunction. As JNK can activate Bim, a BH3 domain-only protein capable of binding to and activating Bax, its role in lipoapoptosis was also examined. Small interfering RNA-targeted knock-down of Bim attenuated both Bax activation and cell death. Collectively the data indicate that saturated FFAs induce JNK-dependent hepatocyte lipoapoptosis by activating the proapoptotic Bcl-2 proteins Bim and Bax, which trigger the mitochondrial apoptotic pathway.

Non-alcoholic fatty liver disease (NAFLD) 2 is associated with obesity and affects a third of the population of the United States. Because a subset of patients with NAFLD progresses to non-alcoholic steatohepatitis (NASH) and end stage liver disease (1), NAFLD has emerged as a substantial public health concern. This syndrome is associated with hepatocyte steatosis and elevated serum free fatty acids (2). However, the cellular mechanisms linking elevated serum free fatty acids and hepatocyte steatosis to tissue injury remain obscure. Liver cell apoptosis is a prominent feature of NASH and correlates with disease severity (3). The toxicity of lipids, or lipotoxicity, and specifically lipid-induced apoptosis, or lipoapoptosis, is a potential mechanism relating apoptosis to NASH. In support of this concept, FFAs have been reported to cause cellular steatosis and enhance expression of the apoptosis effectors tumor necrosis factor-␣ and Fas (2,4). However, direct evidence for hepatocyte lipoapoptosis by individual FFAs is lacking as is the potential mechanism by which they may engage the cellular apoptotic machinery.
c-Jun NH 2 -terminal kinase (JNK) is activated by various stress signals and interestingly also in obesity (5,6). The JNK pathway contributes to stress-induced apoptosis in several cell types, including pancreatic ␤-cells, neurons, and hepatocytes (7)(8)(9). Interactions between FFAs and JNK signaling may, therefore, be a potential link between obesity and lipoapoptosis. JNK has three isoforms of which only two are expressed in the liver, JNK1 and JNK2 (10). Sustained JNK2 and JNK3 activation are cytotoxic in neurons (11), whereas JNK1 potentiates hepatocyte apoptosis by the toxic bile acid deoxycholate (9). Thus, JNK isoform cytotoxicity appears to be cell type-and stimulus-specific. JNK may induce cell injury through both transcriptional and non-transcriptional mechanisms. Growth factor withdrawal-, UV light-, and oxidative stress-associated JNK-mediated apoptosis requires activation of the transcription factor c-Jun/AP-1 (12)(13)(14)(15). In contrast, JNK may also induce apoptosis in a transcription-independent process by activating proapoptotic members of the Bcl-2 family, including Bim and Bax, or inactivating Bcl-2 and Bcl-xL, antiapoptotic members of this family (16 -20). Activation of the proapoptotic proteins or inhibition of their antiapoptotic counterparts triggers mitochondrial dysfunction and cell death.
The dual objectives of this study were to determine whether individual free fatty acids induce hepatocyte apoptosis and whether the observed lipoapoptosis is JNK-dependent. We induced steatosis in cultured hepatocytes and demonstrated apoptotic cell death with free fatty acid treatment. Sustained JNK activation was a prominent feature of this lipoapoptosis, and inhibition of JNK abrogated hepatocyte lipotoxicity.
Fatty Acid Treatment-Oleic acid (catalog number O 1008), palmitoleic acid (catalog number P 9417), palmitic acid (catalog number P 0500), and stearic acid (catalog number S 4751) were all obtained from Sigma. Palmitic acid and stearic acid were dissolved in isopropanol at a concentration of 20 or 40 mM. Oleic acid and palmitoleic acid, commer-cially available in liquid form, were diluted in isopropanol to obtain 20 and 40 mM stock solutions. The concentration of the vehicle, isopropanol, was 1% in final incubations. Dulbecco's modified Eagle's medium containing 1% bovine serum albumin was used in all experiments. The concentration of fatty acids used ranged from 50 to 400 M.
Fat Quantitation by Nile Red-Free fatty acid-treated cells were fixed with 3.7% paraformaldehyde for 15 min at room temperature. Intracellular neutral lipid was stained with Nile Red (0.2 mg/ml) for 5 min at room temperature (24). Images were acquired by confocal microscopy with an inverted Zeiss laser scanning confocal microscope (Zeiss LSM 510, Carl Zeiss Inc., Thornwood, NJ). Cellular fluorescence was quantitated using Zeiss KS400 image analysis software (Carl Zeiss, Inc., Oberkochen, Germany). Nile Red staining, reflecting lipid content, was expressed as an increase in total cellular fluorescence intensity per cell (pixels above threshold ϫ fluorescence intensity).
Quantitation of Apoptosis-Nuclear staining with 4Ј,6-diamidino-2phenylindole (DAPI) and fluorescence microscopy were used to quantitate apoptotic cells. 400 random cells were counted in each condition. Cells with the characteristic nuclear changes of chromatin condensation and nuclear fragmentation were considered apoptotic.
Flow Cytometry-Cells were harvested by trypsinization (0.25% trypsin-EDTA, 10 min, 37°C) and fixed by resuspension in ice-cold 70% ethanol. After washing once with phosphate-buffered saline (PBS), low molecular mass DNA fragments were extracted by rinsing fixed cells in extraction buffer (0.2 M Na 2 HPO 4 , 0.1 M citric acid, pH 7.8). Cellular DNA was stained with propidium iodide (20 g/ml) in the presence of DNase-free RNase A (200 g/ml) (25). Flow cytometric data were acquired using a FACSCalibur flow cytometer and analyzed using CellQuest software (BD Biosciences Immunocytometry Systems).
Quantitation of Mitochondrial Membrane Potential (⌬)-Cells were cultured on 35-mm glass-bottomed dishes (MatTek Corp., Ashland, MA). ⌬ was measured using a fluorescence unquenching assay based on the concept of resonance energy transfer (26). The normal negative mitochondrial membrane potential results in selective uptake of the cationic MitoFluor green, which covalently binds mitochondrial proteins. On loss of mitochondrial membrane potential, MitoFluor green is still retained within mitochondria; tetramethylrhodamine methyl ester (TMRM), also a fluorescent cation, loads into mitochondria based on ⌬ but rapidly diffuses out of mitochondria upon mitochondrial depolarization. MitoFluor green fluorescence emission at fluorescein wavelengths is effectively quenched by TMRM via resonance energy transfer in fully polarized mitochondria. However, as ⌬ is lost, TMRM diffuses out of mitochondria resulting in enhanced MitoFluor green fluorescence. Cells were treated with free fatty acid for 24 h prior to loading with 200 nM MitoFluor green for 20 min at 37°C in medium. Then the cells were loaded with 1 M TMRM for 20 min at 37°C. After initial fluorescence measurement (Y), the nonfluorescent uncoupler 1799 (50 nM) was added. This uncoupler results in mitochondrial depolarization and loss of TMRM allowing determination of maximal Mit-oFluor green fluorescence (X) from depolarized mitochondria. Change in fluorescence (X Ϫ Y) was calculated for control (X C Ϫ Y C ) and treated (X T Ϫ Y T ) cells and expressed as a percent change of control (X T Ϫ Y T /X C Ϫ Y C ϫ 100). By this assay percent decrease in florescence intensity correlates with loss of ⌬. Fluorescence was recorded using an inverted fluorescence microscope, and images were collated using a cooled, charge-coupled device camera (Photometrics, Tucson, AZ) and digitized by MetaFluor software (Universal Imaging, Westchester, PA).
Cell Fractions-Whole cell, cytosolic, and mitochondrial fractions were prepared from HepG2 cells ϳ70% confluent prior to treatment with free fatty acids. To prepare whole cell lysates, cells were placed on ice, the medium was aspirated, and the cells were rinsed once with ice-cold PBS. The cells were then scraped in 1 ml of lysis buffer containing 50 mM Tris (pH 7.4), 150 mM NaCl, 1 mM EDTA, 6 mM deoxycholic acid, 1% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 1 mM Na 3 VO 4 , 1 mM NaF, and protease inhibitor mixture (Complete protease inhibitor mixture; Roche Diagnostics), transferred to microcentrifuge tubes, and incubated for 30 min at 4°C. Whole cell lysates were centrifuged at 13,000 ϫ g for 15 min at 4°C to remove insoluble material. For subcellular fractions, cultured cells were collected by scraping in homogenization buffer (70 mM sucrose, 220 mM mannitol, 3 mM EDTA, 5 mM MOPS, pH 7.4) and transferred to an ice-cold Dounce homogenizer. After 100 up and down passes in the homogenizer (on ice), the crude homogenate was centrifuged at 600 ϫ g for 10 min at 4°C to remove debris. The postnuclear supernatant was centrifuged at 7000 ϫ g for 10 min at 4°C to pellet mitochondria from cytosol. The supernatant from this step is the cytosolic fraction, and the pellet is the mitochondrial fraction. The mitochondrial pellet was rinsed in wash buffer (5 mM MOPS, 100 mM KCl, pH 7.4), collected by centrifugation, and resuspended in lysis buffer (30 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, 1% Triton X-100, 1% Nonidet P-40, and a Complete protease inhibitor tablet (Roche Diagnostics). After incubating for 30 min at 4°C with gentle shaking, the mitochondrial lysate was centrifuged at 13,000 ϫ g for 10 min at 4°C to remove insoluble protein. The protein content of samples was estimated by the Bradford assay (Sigma). The mitochondrial, cytosolic, and whole cell lysates were subjected to immunoblot analysis as described below.
Immunoblot Analysis-Protein was electrophoretically resolved by SDS-PAGE and immobilized on polyvinylidene difluoride membrane. 5% nonfat dairy milk in Tris-buffered saline (20 mM Tris, 150 mM NaCl, pH 7.4) with 0.1% Tween 20 was used to block nonspecific binding sites. were used to detect antigen-antibody complexes. Immune complexes were visualized using a chemiluminescent substrate (ECL, Amersham Biosciences) and Kodak X-Omat film (Eastman Kodak Co.). Immunoreactive areas were quantitated by densitometry using an imaging den-sitometer (Model GS-700, Bio-Rad) and the Molecular Analyst software program (Bio-Rad) to calculate the ratio of Bim-EL to ␥-tubulin.
JNK and Phospho-JNK Immunoreactivity by ELISA-JNK and phospho-JNK immunoreactivity was measured using a commercially available Fast Activated Cell-based ELISA (FACE; Active Motif, Carlsbad, CA). Briefly cells were plated in 96-well plates. Cells were fixed with 4% formaldehyde followed by quenching of endogenous peroxidase with 1% H 2 O 2 and 0.1% sodium azide in PBS containing 0.1% Triton X-100. Each well was incubated with antibody specific to total JNK or phospho-JNK. Horseradish peroxidase-conjugated secondary antibody was subsequently added and developed with the commercial reagent. Absorption was measured at 450 nm using a spectrophotometer. Cell number was quantitated by crystal violet staining. JNK and phospho-JNK immunoreactivity was normalized to cell number.
Bim Gene Silencing by siRNA-RNA interference was used to silence Bim gene expression in Huh7 and HepG2 cells. A 21-nucleotide doublestranded siRNA, 5Ј-AAT TAC CAA GCA GCC GAA GAC-3Ј, targeting human Bim was designed using proprietary software and synthesized using a Silencer siRNA construction kit (Ambion Inc., Austin, TX). Transient gene silencing was attained by transfection of siRNA into cells using siPORT lipid transfection reagent according to the manufacturer's instructions. Scrambled siRNA was used as a control. Gene silencing was verified by detecting protein with immunoblot analysis after transient transfection of Huh7 cells with siRNA. Briefly cells grown in 6-well dishes were transiently transfected with 35 nM siRNA using 4 l/ml siPORT lipid (Ambion Inc.) in a total transfection volume of 0.5 ml of Opti-MEM (Invitrogen). After incubation at 37°C in 5% CO 2 for 4 h, 1 ml of normal growth medium was added. Transfected cells were then analyzed for apoptosis, for Bax activation by immunofluorescence, or by immunoblot as described elsewhere under "Experimental Procedures." Real Time Polymerase Chain Reaction-Total RNA was extracted from cultured cells using TRIzol reagent (Invitrogen). Maloney leukemia virus reverse transcriptase (Invitrogen) and random primers (Invitrogen) were used to reverse transcribe RNA into cDNA. The cDNA template was quantified using real time PCR (LightCycler, Roche Applied Science) using SYBR green (Molecular Probes) as fluorophore. The PCR primers for Bim were forward 5Ј-AGATCCCCGCTTTTCATCTT-3Ј and reverse 5Ј-AGGACTTGGGGTTTGTGTTG-3Ј; the primers for human cytochrome P450 (CYP) 2E1 were forward 5Ј-CTACAAGGC-GGTGAAGGAAG-3Ј and reverse 5Ј-GGGAGTGCTGAGTAGGT-GGA-3Ј; the primers for murine CYP2E1 were forward 5Ј-TTCAT-CAACCTCGTCCCTTC-3Ј and reverse 5Ј-AGGCCTTCTCCAA-CACACAC-3Ј. Commercially available 18 S rRNA primers (Ambion Inc.) were used as internal control. The expected PCR product was electrophoresed in a 1% agarose gel and eluted into Tris⅐HCl using a DNA elution kit (gel extraction kit, Qiagen, Valencia, CA). The concentration of DNA in the extracted PCR product was measured (copies/l) spectrophotometrically at 260 nm. A standard curve was generated using the extracted PCR product. The inverse linear relationship between copy and cycle numbers was then determined. Each resulting standard curve was then used to calculate the copy number per microliter in each experimental sample. The target mRNA level was expressed relative to the 18 S rRNA level for each sample. The PCR conditions and primers were optimized to produce a single PCR product of the correct base pair size.
Statistical Analysis-All data are expressed as the mean Ϯ S.E. unless otherwise indicated. Differences between groups were compared using Student's t test.

RESULTS
FFA Treatment Induces Cellular Steatosis-We first confirmed our cell culture model of cellular steatosis by incubating HepG2 cells with various FFAs in medium containing 1% bovine serum albumin. Specifically HepG2 cells were incubated with oleic acid (C18:1) and palmitoleic acid (C16:1), both monounsaturated fatty acids, and stearic acid (C18:0) and palmitic acid (C16:0), both saturated fatty acids, for 18 h. Intracellular lipid vacuoles visible under phase-contrast microscopy were confirmed by Nile Red staining (Fig. 1A). To confirm equal steatotic effects of saturated and unsaturated fatty acids, cellular steatosis, expressed as a percentage of total cellular area, was quantified (Fig. 1B). Percent steatosis was 9.5 Ϯ 2.6% in oleic acid-treated cells, 14.7 Ϯ 1.9% in palmitic acid-treated cells, 10.1 Ϯ 0.5% in palmitoleic acid-treated cells, 12.4 Ϯ 1.8% in stearic acid-treated cells, and 3.7 Ϯ 0.4% in untreated controls. Thus, all the individual FFAs produced similar cellular steatosis. CYP2E1 expression is associated with human NASH (27). CYP2E1 expression in palmitic acid-treated Huh7 cells was modestly increased (ϳ2-fold) (data not shown). In contrast, free fatty acid treatment did not result in enhanced CYP2E1 expression in primary murine hepatocytes or HepG2 cells (data not shown). Thus, our reductionistic model recapitulates cellular steatosis, a cardinal feature of human NASH, and other features of human NASH albeit in a cellspecific manner.
Biochemical confirmation of apoptosis was obtained by demonstrating DNA cleavage and caspase activation, both biochemical hallmarks of apoptosis (28). In HepG2 cells, fluorescence-activated cell sorter analysis demonstrated 15.6 Ϯ 1.1% death in control cells; this was increased to 19.8 Ϯ 1.3% ( p ϭ 0.04) with oleic acid treatment and 58 Ϯ 2.7% ( p Ͻ 0.001) with palmitic acid treatment for 24 h. Fatty acid treat-ment also led to robust caspase 3/7 activation (Fig. 2E). Consistent with the morphologic assessment of apoptosis, caspase 3/7 activity was greater in palmitic acid-(7.1-fold) and stearic acid (6.5-fold)-treated HepG2 cells as compared with oleic acid-treated cells (2.6-fold). The pancaspase inhibitor Z-VAD-fmk attenuated cell death by all FFAs examined in a concentration-dependent manner (Fig. 2F). The structurally dissimilar pancaspase inhibitor IDN-6556 also reduced cell death by 50% in HepG2 cells incubated with palmitic acid or stearic acid ( p Ͻ 0.01, data not shown). Taken together, these data demonstrate that FFAs induce caspase-dependent apoptosis.
FFA-mediated Apoptosis Occurs via the Mitochondrial Pathway of Apoptosis-Mitochondrial dysfunction with loss of ⌬ and release of cytochrome c into the cytosol are prominent features of the mitochondrial pathway of apoptosis. Indeed ⌬ was 26 Ϯ 14% of the initial value following treatment with palmitic acid and 25 Ϯ 11% of the initial value following treatment with stearic acid for 18 h. Upon treatment with fatty acids, cytochrome c was also released from mitochondria into the cytosol as demonstrated by a change from punctate to diffuse cytoplasmic staining in 19% of oleic acid-treated HepG2 cells, 98% of palmitic acid-treated cells, and 96% of stearic acid-treated cells (Fig. 3, A and B). Release of cytochrome c into the cytosol by palmitic acid and stearic acid was confirmed by immunoblot analysis of cytosolic fractions (Fig. 3C). These data suggest that saturated FFAs mediate apoptosis by engaging the mitochondrial pathway of apoptosis.
Mitochondrial dysfunction in hepatocytes may be initiated via either death receptor (extrinsic) or intracellular stress (intrinsic) pathways. Death receptor-mediated mitochondrial dysfunction occurs by caspase 8 cleavage of the BH3-only Bcl-2 family protein Bid to generate an active smaller carboxyl-terminal fragment referred to as tBid (29). However, FFA treatment did not result in Bid cleavage, whereas Bid cleavage to tBid was readily identified in TRAIL plus actinomycin D-treated HepG2 cells (Fig. 4A, lane 10). Selective pharmacologic inhibition of caspase 8 with IETD-fmk at 10 and 30 M also did not diminish palmitic acid or stearic acid toxicity (data not shown). Lysosomal permeabilization has also been reported to trigger mitochondrial dysfunction in hepatocytes. Cathepsin B, a lysosomal cysteine protease, is a key mediator of this pathway (21). However, oleic acid, palmitic acid, and stearic acid treatment induced 31.9 Ϯ 2.3, 67 Ϯ 9.4, and 60.3 Ϯ 12.1% apoptosis in cathepsin BϪ/Ϫ hepatocytes (Fig. 4B), respectively, similar to that observed in wild type hepatocytes (28.6 Ϯ 2.9, 64.9 Ϯ 6.2, and 54.1 Ϯ 2.6% apoptosis with oleic acid, palmitic acid, and stearic acid treatment, respectively ( p ϭ not significant)). Thus, free fatty acid-induced apoptosis is not associated with Bid cleavage, and it is not dependent on cathepsin B. These data suggest that FFA engagement of the mitochondrial pathway of apoptosis excludes death receptor-and lysosome-initiated events.
FFA Treatment Activates c-Jun NH 2 -terminal Kinase-Because JNK activation, a key feature of the metabolic syndrome, is associated with NAFLD and has been implicated in apoptosis, we determined whether JNK activation is responsible for the lipoapoptosis observed in FFAtreated murine hepatocytes and HepG2 cells (Fig. 5, A and B). In primary murine hepatocytes stearic acid treatment led to a significant increase in JNK activity, whereas oleic acid treatment did not significantly increase phospho-JNK levels. Total JNK levels remained unchanged during FFA treatment (Fig. 5A). Stearic acid treatment led to a significant increase in phospho-JNK levels and activity by both immunoblot analysis and ELISA in HepG2 cells. A robust increase in phospho-JNK was observed by immunoblot analysis as early as 3 h following treatment with stearic acid (Fig. 5B). This increase in JNK activity was sustained and persisted at 24 h. Quantitatively as assessed by an ELISA assay, phospho-JNK was increased 2.4-, 2.5-, and 6.3-fold (compared with untreated cells) at 3, 6, and 24 h, respectively, following treatment with stearic acid (Fig. 5C). Oleic acid treatment, on the other hand, led to only modest 1-, 1.6-, and 1.8-fold increases in phospho-JNK at 3, 6 and 24 h, respectively. In primary mouse hepatocytes phospho-JNK was increased 2.3-, 2.4-, and 2.7-fold (compared with untreated cells) at 3, 6, and 24 h, respectively, following treatment with stearic acid (Fig. 5C). Oleic acid treatment, on the other hand, led only to a modest increase in phospho-JNK activity of 1.7-, 1.7-, and 1.6-fold at 3, 6, and 24 h, respectively. p44/42 mitogen-activated protein kinase, another member of the mitogen-activated protein kinase family, was phosphorylated following stearic acid treatment (Fig. 5B). In contrast, p38 MAPK was not phos-  phorylated following exposure to stearic acid. Thus, both JNK and p44/42 MAPK were strongly activated by proapoptotic free fatty acids.
To define whether JNK or p44/42 MAPK contributes to lipoapoptosis in our model, inhibitor studies were performed. A cell-permeable, small peptide JNK inhibitor reduced oleic acid toxicity by 56% ( p Ͻ 0.05), palmitic acid toxicity by 70% ( p ϭ 0.001), and stearic acid toxicity by 45% ( p Ͻ 0.05) (Fig. 6A). To further confirm this observation, SP600125, a structurally dissimilar JNK inhibitor, was also used (30). This compound also reduced oleic acid toxicity by 45% ( p Ͻ 0.005) and palmitic acid and stearic acid toxicity by 70% ( p Ͻ 0.05) (Fig. 6A). To determine the role of the two JNK isoforms expressed by hepatocytes in our model of lipoapoptosis, hepatocytes were isolated from mice deficient in JNK2 (JNK2Ϫ/Ϫ) and JNK1 (JNK1Ϫ/Ϫ). Stearic acid-and oleic acid-induced cell death was reduced in JNK2Ϫ/Ϫ mice ( p Ͻ 0.05) (Fig.  6B). In contrast, hepatocytes from JNK1Ϫ/Ϫ mice were not resistant to oleic acid or stearic acid cytotoxicity (Fig. 6C). Finally the role of p44/42 MAPK in this apoptotic pathway was assessed using the inhibitors PD98059 and U0126. At concentrations that block p44/42 MAPK activity, neither inhibitor reduced palmitic acid-or stearic acid-mediated lipoapoptosis (data not shown). Collectively these observations suggest that FFA-mediated lipoapoptosis is, in part, JNK-dependent, and at least in murine hepatocytes, the JNK2 isoform mediates this cytotoxicity.
JNK Engages the Mitochondrial Pathway of Apoptosis by Activating Bax-Bax, a proapoptotic BH3 family protein, can be activated by JNK (31). Upon activation, Bax undergoes a conformational change exposing an amino-terminal epitope, which is specifically recognized by the 6A7 monoclonal antibody (32). Indeed FFA treatment led to Bax activation ( Fig. 7A) as demonstrated by immunofluorescence microscopy for active Bax. Pretreatment of HepG2 cells with JNK peptide inhibitor and SP600125 prevented stearic acid-and palmitic acid-induced Bax activation (Fig. 7B). Bax activation led to mitochondrial permeabilization and release of cytochrome c. Indeed cytochrome c release was decreased 61% by SP600125 and 73% by the JNK peptide inhibitor (Fig. 7C) in stearic acid-treated cells, confirming inhibition of the mitochondrial pathway of apoptosis. These data place Bax activation and mitochondrial permeabilization downstream of JNK activation in stearic acidinduced lipoapoptosis.
Multidomain proteins of the Bcl-2 family, such as Bax, are known to be activated by BH3-only Bcl-2 family proteins. Of the eight known BH3-only proteins only Bim and Bid bind and activate Bax (33). As Bid activation was not observed (Fig. 4A), we focused on Bim in FFA-mediated lipoapoptosis. We first examined Bim expression in free fatty acidtreated HepG2 cells. Of three known isoforms, Bim-EL, Bim-L, and Bim-S, only the first two were detected consistently in HepG2 cells. Consistent with its toxicity profile, stearic acid led to an early and sustained increase in cellular Bim levels (Fig. 8A). Quantitative densitometry showed that stearic acid treatment led to a 2 Ϯ 0.3-fold increase in cellular Bim-EL levels at 6 h, whereas oleic acid treatment led to a 1.2 Ϯ 0.2-fold increase (Fig. 8B). The increase in Bim protein was also associated with an increase in Bim mRNA. Indeed as assessed by real time PCR, stearic acid treatment led to a 2-fold increase in Bim mRNA expression in HepG2 cells and a 3-fold increase in Huh7 cells; in contrast an increase was not observed in oleic acid-treated cells (data not shown). The observed increase in Bim is not a result of cell death as it FIGURE 5. Oleic acid and stearic acid treatment leads to JNK activation. HepG2 cells and primary mouse hepatocytes were treated with 200 M oleic acid and stearic acid for the indicated times. Whole cell lysates were analyzed for protein expression using antibodies specific to total and active (phospho) members of the mitogen-activated protein kinase family members. ␥-Tubulin was used as a control for protein loading for A and B. A, primary mouse hepatocytes treated with stearic acid show a robust activation of JNK. An increase in phospho-JNK levels was not observed by immunoblot analysis in oleic acid-treated cells (although by quantitative ELISA there was a slight (Ͻ2-fold) increase). B, stearic acid treatment led to marked JNK activation in HepG2 cells, and oleic acid-induced JNK activation was not as marked. There was no associated c-Jun and p38 mitogenactivated protein kinase activation. p44/42 was activated but to a lesser extent. C, primary mouse hepatocytes and HepG2 cells were treated with 200 M oleic acid and stearic acid for the indicated times. ELISA for total and phospho-JNK was performed using the Fast Activated Cell-based JNK ELISA (FACE) kit. -Fold change in phospho-JNK activity relative to total-JNK activity is shown. In both cell types, stearic acid led to an early and sustained increase in phospho-JNK activity. OA, oleic acid; SA, stearic acid; ERK, extracellular signal-regulated kinase. occurred within 6 h of treatment, well before detectable apoptosis. Thus, stearic acid-induced JNK-dependent lipoapoptosis is associated with Bim induction. Whether Bim induction is an epiphenomenon of stearic acid treatment or plays a role in stearic acid-induced lipoapoptosis was further dissected by siRNA-mediated Bim gene silencing in Huh7 and HepG2 cells. A reduction in Bim expression was cytoprotective; indeed stearic acid-induced lipoapoptosis was reduced by 55% (Fig.  8C). Although oleic acid toxicity was minimal per se, it was slightly reduced by decreasing Bim expression using siRNA (Fig. 8C). Furthermore Bax activation, the effector of mitochondrial permeabilization in this model, was Bim-dependent (Fig. 8D). In cells with targeted Bim knock-down, stearic acid-induced Bax activation was reduced by 83% ( p Ͻ 0.001, Fig. 8E). Collectively these data suggest that JNK mediates lipoapoptosis via Bim-dependent Bax activation.

DISCUSSION
The principal findings of this study relate to the mechanism of FFAinduced hepatocyte lipoapoptosis. The data demonstrate that (i) saturated FFAs are more cytotoxic than monounsaturated FFAs, (ii) FFA cytotoxicity occurs via caspase-dependent apoptosis, and (iii) JNK triggers the mitochondrial pathway of apoptosis by Bim-dependent Bax activation. These data provide mechanistic insight into the cellular mechanism of hepatocyte lipoapoptosis, a model relevant to human NASH.
In the current study, FFAs directly induced apoptosis in hepatocytes. Saturated FFAs were substantially more toxic than monounsaturated FFAs despite causing a similar magnitude of cellular steatosis. Our data are consistent with observations in disparate cell types, such as pancreatic ␤ cells, myocardium, skeletal muscle, neurons, and endothelial cells (34,35), where saturated FFAs, palmitic acid and stearic acid, also exhibit greater cytotoxicity than monounsaturated FFAs, oleic acid and palmitoleic acid. Our observations are in contrast to alcohol-mediated hepatotoxicity where feeding animals unsaturated FFAs potentiates hepatotoxicity (36). The mechanisms of alcohol-induced hepatotoxicity may well be quite different from those of NAFLD despite the fact that both are characterized by steatosis. Our data, indicating that specific FFAs have distinct inherent toxic potential, suggest that characterizing the FFA profile in serum and/or liver tissue may predict which individuals with NAFLD will develop more severe liver disease. Formation of reactive intermediates (reactive oxygen species), alternative lipid metabolic pathways, such as ceramide synthesis, modulation of death receptor expression, and direct activation of cellular proapoptotic machinery are putative mechanisms by which lipoapoptosis is thought to occur (2-4, 37, 38). However, none of these observations and concepts sufficiently explain how FFAs engage the core cell death machinery. The current data address this question. The results demonstrated that FFA-mediated lipoapoptosis occurs by a caspase-dependent mechanism by activating the mitochondrial pathway of apoptosis. Indeed caspase 3/7 activity was readily identified, and FFA lipoapoptosis was attenuated by caspase inhibition. The caspase activation coincided with mitochondrial dysfunction and was not blocked by inhibiting the death receptor pathway with a selective caspase 8 inhibitor or by disrupting the lysosomal pathway by genetic deletion of cathepsin B. Collectively these data indicate that cytotoxic FFAs directly engage the mitochondrial cell death pathway.
FFAs are known to affect several cellular signaling pathways, including the MAPK cascades (39,40). In the hepatocyte cell lines studies, stearic acid activated both p44/42 and JNK MAPK. Consistent with the proapoptotic properties of JNK and the prosurvival effects of p44/42 (15), only inhibition of JNK abrogated FFA-induced lipoapoptosis. These data link the known contribution of JNK to obesity-associated insulin resistance with the lipoapoptosis observed in this syndrome. In this regard JNK activation by FFAs may not only contribute to impaired insulin signaling but also contribute to cellular injury. Therefore, JNK would appear to play a pivotal role in obesity-related liver injury. Cellular responses to JNK activation are also governed by isoform-specific actions. In neurons, JNK3 mediates apoptosis (41) and cell differentiation, depending on context (42). Similarly in non-neuronal cells, JNK1 and JNK2 have proapoptotic functions or prosurvival functions in a cell-, stimulus-, and context-dependent manner (5,9,43,44). In experimental dietary obesity, JNK1 promotes weight gain and insulin resistance (5) and also promotes bile acid-mediated hepatocyte apoptosis (9). Our data suggest a predominant role for the JNK2 isoform in FFAmediated lipoapoptosis as JNK2Ϫ/Ϫ hepatocytes, but not JNK1Ϫ/Ϫ hepatocytes, were resistant to cell death. Thus, even in hepatocytes, cytotoxicity by JNK isoforms is stimulus-specific.
Mitochondrial release of cytochrome c is regulated by the proapoptotic multidomain members of the Bcl-2 family, Bax and Bak. Bax is necessary for mitochondrially mediated apoptosis depending on cellular context (45,46). For example, in hepatocytes Bax mediates ethanoland bile salt-induced mitochondrial permeabilization and apoptosis (47,48). Consistent with the crucial role of Bax in hepatocyte apoptosis, stearic acid treatment led to Bax activation and subsequent cytochrome c release. Furthermore pharmacologic inhibition of JNK prevented both Bax activation and downstream events leading to amelioration of stearic acid-induced lipoapoptosis. BH3 domain proteins are the sensors of cellular stress and are essential for engaging the mitochondrial pathway of apoptosis. To date, only Bim and Bid have been shown to directly activate Bax (33). As Bid was not activated in our model, we focused on Bim. FFA-induced JNK activation was linked mechanistically to the mitochondrial pathway of apoptosis by Bim induction. Indeed we observed an early and sustained increase in cellular Bim protein levels following treatment of HepG2 cells with stearic acid and amelioration of FIGURE 8. Stearic acid toxicity is Bim-dependent. A, HepG2 cells were treated with 200 M oleic acid and stearic acid for the indicated time. Whole cell lysates were analyzed for Bim protein expression. ␥-Tubulin was used as a control for protein loading. Stearic acid treatment led to an early and sustained increase in Bim levels. B, quantitative densitometry was performed on four independent Bim immunoblots from HepG2 cells treated with oleic acid (200 M, 6 h) and stearic acid (200 M, 6 h). Bim-EL expression was normalized to ␥-tubulin expression. Data are expressed as -fold change compared with controls. A significant increase in Bim-EL was observed in stearic acid-treated cells. C, Bim protein expression was silenced in Huh7 cells using siRNA. A Western blot for Bim-EL expression is shown in the inset with ␥-tubulin as control for protein loading. Stearic acid (SA)-induced lipoapoptosis was reduced Ͼ50% by Bim gene silencing in Huh7 cells.
Oleic acid (OA) toxicity was minimal and slightly reduced in siBim-treated cells. D, activated Bax was detected by immunofluorescence microscopy in HepG2 (shown here) and Huh7 cells (not shown). Following targeted knock-down of Bim expression, stearic acid-induced Bax activation was abrogated in both cell lines. E, stearic acid-induced Bax activation in Bim-silenced cells was quantitated by manual counting of four or more random confocal photomicrographs. Bax activation in stearic acid-treated cells was Bim-dependent. apoptosis in Bim-silenced cells. These observations are consistent with data from other models where JNK-dependent Bim induction as well as phosphorylation with downstream activation of Bax-mediated apoptosis has been reported (8,19,20,49). Thus, the current data support a model where JNK induces hepatocyte lipoapoptosis via a Bim-mediated, Bax-dependent mitochondrial pathway of cell death.
Hepatocyte lipoapoptosis is a significant public health issue given the current obesity epidemic and the relationship between apoptosis and liver disease in this syndrome (3). Our cellular model is relevant to this syndrome as toxic FFAs are thought to contribute to hepatocyte lipoapoptosis in man. Our data indicate that FFAs cause sustained JNK activation leading to engagement of the core mitochondrial proapoptotic machinery with Bim-mediated Bax activation. These observations identify JNK as a potential therapeutic target to ameliorate FFA cytotoxicity, especially the use of JNK inhibitors (50).