Lysophosphatidylcholine as a death effector in the lipoapoptosis of hepatocytes.

The pathogenesis of nonalcoholic steatohepatitis (NASH) is unclear, despite epidemiological data implicating FFAs. We studied the pathogenesis of NASH using lipoapoptosis models. Palmitic acid (PA) induced classical apoptosis of hepatocytes. PA-induced lipoapoptosis was inhibited by acyl-CoA synthetase inhibitor but not by ceramide synthesis inhibitors, suggesting that conversion products other than ceramide are involved. Phospholipase A2 (PLA2) inhibitors blocked PA-induced hepatocyte death, suggesting an important role for PLA2 and its product lysophosphatidylcholine (LPC). Small interfering RNA for Ca2+-independent phospholipase A2 (iPLA2) inhibited the lipoapoptosis of hepatocytes. PA increased LPC content, which was reversed by iPLA2 inhibitors. Pertussis toxin or dominant-negative Gαi mutant inhibited hepatocyte death by PA or LPC acting through G-protein-coupled receptor (GPCR)/Gαi. PA decreased cardiolipin content and induced mitochondrial potential loss and cytochrome c translocation. Oleic acid inhibited PA-induced hepatocyte death by diverting PA to triglyceride and decreasing LPC content, suggesting that FFAs lead to steatosis or lipoapoptosis according to the abundance of saturated/unsaturated FFAs. LPC administration induced hepatitis in vivo. LPC content was increased in the liver specimens from NASH patients. These results demonstrate that LPC is a death effector in the lipoapoptosis of hepatocytes and suggest potential therapeutic values of PLA2 inhibitors or GPCR/Gαi inhibitors in NASH.

from visceral fat of obese subjects are strong culprits in insulin resistance. Recent papers have suggested a potential role for FFAs in insulin deficiency as well as in insulin resistance, consistent with FFA-mediated injury of pancreatic b-cells and several other tissues (4).
FFAs may be related to NAFLD through their increased flux from visceral fat to the liver. Because of strong epidemiological and in vivo data suggesting the relationship between FFAs and NAFLD, several in vitro experiments studying hepatocyte injury by FFAs have been conducted, which suggested possible roles for the mitochondrial pathway, lysosomal pathway, endoplasmic reticulum stress, and c-Jun N-terminal kinase (JNK) activation in the lipoapoptosis of hepatocytes (5)(6)(7)(8)(9). However, the intracellular metabolites of FFAs that are directly responsible for such intracellular events are not clearly identified. We found evidence supporting the role of lysophosphatidylcholine (LPC) produced by phospholipase A 2 (PLA 2 ), which catalyzes the hydrolysis of the fatty acyl ester bond at the sn-2 position of glycerophospholipids and has been implicated in several types of cell death (10)(11)(12)(13), as an effector in the lipoapoptosis of hepatocytes (Fig. 1).

Cell death assay
Chang cells grown in DMEM-10% FBS were treated with FFAs or other reagents for 24 h unless stated otherwise. We chose a 24 h treatment protocol because our time course study revealed that cell death .50% with full-blown features of apoptosis, such as nuclear condensation/fragmentation and sub-G1 DNA peak, occurred at 24 h after treatment. Palmitic acid (PA) solution was made according to a previously published protocol with modifications (14). Briefly, PA stock solution (50 mM) was prepared by dissolving in 70% ethanol and heating at 50jC. Working PA solution was made by diluting stock solution in 2% fatty acidfree BSA-DMEM-10% FBS, because the concentration of FFAs, particularly that of PA, is quite low in 10% FBS, whereas cells were healthier with serum for prolonged experiments. 3-[4,5-Dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide (MTT) assay, Hoechst 33258/propidium iodide (PI) staining, and DNA ploidy analysis were conducted as described (15). We used the MTT assay throughout the study because the results from the MTT assay were consistent with those of other cell death assays such as Hoechst/PI staining or DNA ploidy analysis in our pilot study. In some experiments, cells were pretreated with inhibitors for 1 h before PA treatment. To assess the death of primary murine hepatocytes, a lactate dehydrogenase release assay was used (12). Percentage cell death was calculated as (experimental release 2 spontaneous release)/(total release 2 spontaneous release) 3 100.

Small interfering RNA and RT-PCR
Small interfering RNAs (siRNAs) for human Ca 21 -independent phospholipase A 2 b (iPLA 2 b) (AAT TGC GCG GAG AAC GAG GAG), iPLA 2 g (AAA ATG AAC ATT TCC GGG ACA), and an irrelevant siRNA purchased from Invitrogen were transfected using Oligofectamine (Invitrogen) according to the manufacturer's protocol. Expression of iPLA 2 b and iPLA 2 g at the RNA level was determined by RT-PCR using specific primer sets. Expression of markers for hepatocytes was also tested by RT-PCR using specific primer sets (16).

Transfection
Chang cells stably expressing a dominant-negative Ga i2 mutant (a i2 G203T) (18) or wild-type Ga i2 (University of Missouri-Rolla cDNA Resource Center) were produced by transfection with FuGENE 6 (Roche) and selection with G418 for 3 weeks.

Cell fractionation
Cells lysed in an isotonic buffer were fractionated as described (20). The heavy membrane fraction and cytosolic fraction were subjected to Western blotting using specific antibodies against cytochrome c (PharMingen) or Bid (Santa Cruz) (15). The purity of the heavy membrane and cytosolic fractions was confirmed by Western blotting using anti-Cox4 and anti-IkB antibody, respectively. Expression of phosphorylated JNK and total JNK in the whole cell lysate was studied by Western blotting using specific antibodies (Cell Signaling).

Oil Red O staining
Cells were fixed with 10% formaldehyde for 1 h. After staining with 3 mg/ml Oil Red O solution for 15 min, dye was extracted by isopropanol and A 540 was measured.

Human liver biopsy
Human liver tissue was obtained from the nontumor part of the livers from patients undergoing hepatic resection for metastatic colon cancer at Samsung Medical Center. The absence of cancer and cirrhosis was confirmed grossly and microscopically. The isolation of hepatocytes was performed using a two-step collagenase perfusion technique as described previously (21). After isolation, primary hepatocytes were cultured in a modi-fied hormonally defined medium (William's E medium supplemented with 20 mg/l epidermal growth factor, 10 mg/l insulin, 1.7 mg/l hydrocortisone, 24.97 mg/l CuSO 4 I5H 2 O, 14.38 pg/l ZnSO 4 I7H 2 O, 3 mg/l H 2 SeO 3 , 50 mg/l linoleic acid, 1.05 g/l NaHCO 3 , and 1.19 g/l HEPES) (22). Aspartate aminotransferase/ alanine aminotransferase (AST/ALT) levels were measured using the FPC 3000 Blood Chemistry Analyzer (Fuji). Ultrasonographyguided liver biopsy was conducted in patients with clinical findings compatible with NAFLD without history of alcoholism (alcohol consumption , 140 g/week) using 18-gauge Solco needles inserted through the intercostal space. All patients had serum AST/ALT levels . 60 U/l and body mass index . 25 without recent history of diet control or parenteral nutrition. All were negative for viral markers, autoantibodies related to liver diseases, clinical evidence of cirrhosis, lipodystrophy, or Wilson's disease, and were not taking drugs related to lipid metabolism. Informed consent was obtained from all participants. All human studies were approved by the Institutional Review Board of Samsung Medical Center. An experienced hepatopathologist blinded to subject details scored the liver biopsy specimens according to a published classification (2).

Terminal deoxynucleotidyl transferase mediated dUTP nick end labeling staining
Terminal deoxynucleotidyl transferase mediated dUTP nick end labeling (TUNEL) staining of mouse liver tissues was conducted as described (15). Briefly, formalin-fixed sections were deparaffinized and microwaved in 0.01 M sodium citrate buffer (pH 6.0) and 0.1% Triton X-100. They were incubated with TUNEL reagents (Roche Molecular Biochemicals) at 37jC for 60 min. After washing, they were further incubated with Convert-POD for 30 min. Diaminobenzidine was used as a color substrate.

Mouse hepatocytes
Primary murine hepatocytes were isolated from C57BL/6 mice using the retrograde two-step collagenase perfusion technique (23). LPC (kindly provided by K-S. Song, Doosan Pharmaceutical; purity . 99%) dissolved in PBS-2% BSA by sonication was injected into the tail vein of ICR mice. All animal experiments were conducted in accordance with the institutional guidelines of the Osaka University Animal Facility and the Samsung Medical Center Animal Facility.

Measurement of iPLA 2 activity
Activity of iPLA 2 was measured using a commercial kit (Cayman) with modifications (24). Briefly, cells were collected with a cell scraper, sonicated, and centrifuged at 20,000 g at 4jC for 20 min. The supernatant was removed, and the concentration of proteins was determined. iPLA 2 activity was assayed by incubating the samples with arachidonoyl thio-PC for 1 h at 25jC in a Ca 21 -free buffer (300 mM NaCl, 0.5% Triton X-100, 60% glycerol, 4 mM EGTA, 10 mM HEPES, pH 7.4, and 2 mg/ml BSA). The reaction was terminated by adding 5,5 ¶-dithio-bis-2-nitrobenzoic acid for 5 min, and A 405 was measured. The specific activity of iPLA 2 was calculated and expressed as absorbance per milligram of protein. The background activity was subtracted from all readings.

Statistical analysis
All values are expressed as means 6 SD from more than three independent experiments performed in triplicate. Two-tailed Student's t-test was used to compare values between two groups. ANOVA was used for multiple comparisons. The Scheffé test was used to compare two groups once ANOVA showed significant differences. P , 0.05 were considered to represent statistically significant differences.

Apoptosis of Chang liver cells by FFAs
We studied the effect of PA, the most abundant saturated fatty acid in vivo, on the Chang "normal" hepatocyte line. We chose Chang cells because they represent a more physiological system as nonmalignant human hepatocytes compared with hepatoma cells (25). RT-PCR analysis showed that Chang cells expressed markers for hepatocytes such as a-fetoprotein, ALT, phosphoenolpyruvate carboxykinase, albumin, tryptophan 2,3-dioxygenase, cytochrome enzymes (Cyp7A1, Cyp3A4), a-1-antitrypsin, and cytokeratin 8 ( Fig. 2A) (16). Furthermore, Chang cells cleared 78.8% of added NH 4 Cl, which was significantly higher than the clearance by control HeLa cells (8.0%) (P , 0.05) ( Fig. 2A), suggesting that Chang cells represent authentic hepatocytes. PA induced Chang cell death, as assessed by MTT assay (Fig. 2B). Hoechst/PI staining demonstrated that Chang cell death by PA was a classical apoptosis characterized by nuclear condensation/fragmentation ( Fig. 2C). Sub-G1 DNA peak was also observed by DNA ploidy analysis (Fig. 2D). A time course study showed that cell death .50% occurred at 24 h after treatment ( Fig. 2E).
Because lipid intermediates produced from PA, such as ceramide, are well-known inducers of apoptosis (26), we investigated whether agents inhibiting the conversion of PA to lipid intermediates could affect Chang cell lipoapoptosis. Triacsin C, which inhibits acyl-CoA synthetase, the first enzyme in the conversion of PA to lipid intermediates through palmityol-CoA (27) (Fig. 1), effectively blocked Chang cell death by PA, suggesting that conversion products of PA but not PA itself induce lipoapoptosis (P , 0.000005) (Fig. 2F). We then studied the effect of fumonisin B1, which blocks ceramide synthesis by inhibiting sphingosine N-acyltransferase ( Fig. 1) (28). Contrary to our expectation, fumonisin B1 did not affect Chang cell death by PA (Fig. 2G). Because this finding was inconsistent with previous reports by others using pancreatic islet cells (4), we next studied the effect of myriocin, which blocks the synthesis of ceramide/sphingosine by inhibiting serine palmitoyltransferase (29) (Fig. 1). Myriocin also did not inhibit Chang cell death by PA (Fig. 2H), suggesting that the conversion of FFAs to lipid intermediates in the sphingolipid-ceramide pathway and finally to ceramide is not involved in the lipoapoptosis of hepatocytes.
We next explored the possible involvement of protein kinase C (PKC) in Chang cell lipoapoptosis, because recent papers showed a role for PKC activation in apoptosis (30) and that PKC activators such as diacylglycerol (DAG) and LPC can be synthesized from palmitoyl-CoA (31) (Fig. 1). Both calphostin C and GF109203X, well-known PKC inhibitors, significantly attenuated Chang cell death by PA, supporting a role for PKC in Chang cell lipoapoptosis (P , 0.000005 for both) (see supplementary Fig. I).

Roles of PLA 2 and LPC in lipoapoptosis
We next studied the possible role of PLA 2 in Chang cell lipoapoptosis, because PLA 2 is involved in several cell death models (12,13) and phosphatidylcholine (PC), a conversion product of DAG, can be a substrate for PLA 2 , leading to the production of LPC, another PKC activator ( Fig. 1). Intriguingly, aristolochic acid, a general inhibitor of PLA 2 , and bromoenol lactone (BEL), a specific inhibitor of iPLA 2 (12,13), markedly inhibited Chang cell death by PA, suggesting important roles for PLA 2 and the conversion products of PA downstream of PLA 2 , rather than of DAG upstream of PLA 2 , in lipoapoptosis and PKC activation (P , 0.0001 and 0.00005, respectively) (Fig. 3A). Because these results also suggested the possible involvement of iPLA 2 among diverse PLA 2 members, we used other PLA 2 inhibitors that are specific for each type of PLA 2 . PACOCF 3 , another iPLA 2 -specific inhibitor of a different class (12,32), dramatically blocked Chang cell death by PA (P , 0.00005). MAFP, an inhibitor of cytoplasmic phospholipase A 2 (cPLA 2 ) (12), also significantly blocked Chang cell death by PA (P , 0.005). However, the protection by the maximum tolerable MAFP level was less compared with that by PACOCF 3 (P , 0.005), suggesting that iPLA 2 rather than cPLA 2 is involved in the lipoapoptosis of hepatocytes (Fig. 3B). Consistent with these pharmacological data, iPLA 2 activity was increased significantly after treatment of Chang cells with PA (P , 0.05), which was inhibited by BEL (P , 0.01) (see supplementary Fig. II).
To confirm the role of iPLA 2 , we conducted an RNA interference experiment. Transfection of siRNA for iPLA 2 b or iPLA 2 g, two major types of iPLA 2 , significantly attenuated Chang cell death by PA, substantiating the roles of iPLA 2 in lipoapoptosis (P , 0.00005 for both comparisons) (Fig. 3C). Transfection of iPLA 2 b and iPLA 2 g siRNA markedly decreased the expression of iPLA 2 b and iPLA 2 g, respectively, at the RNA level (Fig. 3D). However, the combination of both siRNAs did not induce a further decrease in lipoapoptosis (Fig. 3C).
Next, we studied whether LPC, which is liberated from PC by PLA 2 as the most abundant lysophospholipid in vivo, could indeed be produced from PA. Intracellular LPC content was increased significantly by PA (P , 0.01), and the LPC content after PA treatment was attenuated significantly by BEL (P , 0.05) (Fig. 3E). In contrast, LPC content in the culture supernatant was not increased after treatment of Chang cells with PA (see supplementary Fig. III). Treatment of Chang cells with exogenous palmitoyl-LPC (?10-50 mM) for 24 h also induced significant cell death in a dose-dependent manner (P , 0.000001) (Fig. 3F).
We next tested the effect of pertussis toxin (PTX) on Chang cell lipoapoptosis, because PTX inhibits LPC signaling through G-protein-coupled receptor (GPCR)/Ga i (33,34). PTX (?100-400 ng/ml) remarkably inhibited Chang cell death by exogenous LPC or PA (P , 0.00005 and 0.000000005, respectively) (Fig. 3F). We further studied the role of GPCR/Ga i in lipoapoptosis by transfecting a dominant-negative mutant of Ga i (a i2 G203T) (18). Stable Chang cell transfectants expressing a i2 G203T were resistant to death by PA or LPC, strongly supporting the role of LPC and GPCR/Ga i in lipoapoptosis of Chang cells (Fig. 3G). Transfection with a control wild-type Ga i2 did  not significantly affect Chang cell death by PA or LPC, probably because of endogenous expression of Ga i . We next examined the effect of other lysophospholipids that exist in vivo at much lower concentrations compared with LPC but that could be affected by PLA 2 inhibitors. Lysophosphatidylethanolamine (LPE), lysophosphatidylglycerol (LPG), lysophosphatidylinositol (LPI), or lysophosphatidylserine (LPS) up to 50 mM, which is much higher than their physiological concentrations, induced negligible cell death even at 48 h after treatment, suggesting that the effects of PLA 2 inhibitors are most likely through LPC (Fig. 3H).
We next studied intracellular events associated with lipoapoptosis. We measured the intracellular content of cardiolipin, which is crucial for cytochrome c binding to the mitochondrial inner membrane as a mitochondriaspecific phospholipid and could be affected by FFA treatment. Cardiolipin content measured by NAO fluorescence began to decrease at ?3-6 h after PA treatment and decreased further at ?12-24 h after treatment (Fig. 4A). The PA-induced decrease in cardiolipin content was attenuated significantly by iPLA 2 inhibitors such as BEL and PACOCF 3 , suggesting important roles for PLA 2 or LPC in the decrease of cardiolipin content and lipoapoptosis (P , 0.0001 and 0.0005, respectively) (Fig. 4B). We also studied whether LPC could directly induce changes in Bid apart from its receptor-mediated effect, because previous papers reported that Bid could perturb mitochondrial membrane in association with LPC (35). Bid in the mitochondrial fraction was increased by PA treatment, which was inhibited by BEL (Fig. 4C), suggesting a possible role for full-length Bid in cell death by LPC, as reported (35). Probably because of the decrease in cardiolipin content and the increase in Bid in the mitochondria, cytochrome c was translocated from mitochondria to cytoplasm and mitochondrial potential was decreased between 3 and 24 h after PA treatment (Fig. 4D). Chang cell death by PA was inhibited significantly by a pancaspase inhibitor, zVAD-fmk, suggesting that effector caspases were activated downstream of cytochrome c translocation (P , 0.000005) (Fig. 4E). We also studied whether PA activates JNK, because LPC is a well-known activator of JNK (36). Treatment of Chang cells with PA induced JNK activation, which was inhibited by PACOCF 3 , suggesting that LPC produced from PA is involved in JNK activation (Fig. 4F). Exogenous LPC also induced JNK activation in Chang cells, as expected (Fig. 4F). Stable Chang cell transfectants expressing dominant-negative a i2 G203T mutant showed attenuated JNK activation after treatment with PA or LPC, suggesting that LPC produced from PA activates JNK through GPCR/Ga i (Fig. 4G). SP600125 (?10-20 mM), a JNK inhibitor, induced a small but significant decrease in PA-induced death (?25%) (P , 0.005), suggesting that JNK activation plays a certain role in the lipoapoptosis of Chang cells (Fig. 4H).

Inhibition of PA-induced lipoapoptosis by oleic acid
Because DAG can be converted to triglyceride (TG) (Fig. 1), we examined the role of TG in lipid injury. PA did not increase intracellular TG content, estimated by Oil Red O staining followed by spectrophotometry or microscopic examination. In contrast, oleic acid (OA), the most abundant unsaturated fatty acid in vivo, markedly increased TG content. Treatment with OA in combination with PA further increased TG content (Fig. 5A, B).
Regarding cell viability, OA did not induce Chang cell death. On the contrary, OA significantly blocked Chang cell death by PA (P , 0.00005), similar to previous reports (37) (Fig. 5C). OA also abolished the increase in LPC content by PA treatment (P , 0.00005) (Fig. 5D), suggesting that increased FFAs lead to either steatosis of hepatocytes attributable to increased TG content or death of hepatocytes attributable to increased LPC content, depending on the absolute/relative abundance of saturated/unsaturated fatty acids (Fig. 1).

Effect of PA on primary hepatocytes
We next used primary hepatocytes instead of cell lines. Hoechst/PI staining showed that PA induced the death of primary human hepatocytes as well, which was a classical apoptosis characterized by nuclear condensation/fragmentation (Fig. 6A). Lipoapoptosis of primary human hepatocytes by PA was inhibited by BEL (P , 0.0005) but not by fumonisin B1 or myriocin, suggesting that the iPLA 2 /LPC pathway rather than the ceramide pathway is crucial, similar to Chang cells (Fig. 6B, C). Exogenous LPC also induced the death of primary human hepatocytes, which was inhibited by PTX (P , 0.0005) (Fig. 6D). PTX also significantly blocked the death of primary human hepatocytes by PA, suggesting that endogenous LPC induces the death of primary human hepatocytes through GPCR/Ga i (P , 0.000001) (Fig. 6D). OA attenuated the death of primary human hepatocytes by PA, supporting our hypothesis that unsaturated fatty acids divert saturated fatty acids into the TG pathway and attenuate primary human hepatocyte death by saturated fatty acids (Fig. 6E). Consistent with the death of primary human hepatocytes, PA induced the release of AST from cultured primary human hepatocytes, which was inhibited by BEL, PTX, or OA. LPC also induced AST release, which was blocked by PTX (Fig. 6F). PA induced the death of primary murine hepatocytes also, as assessed by lactate dehydrogenase release assay. Lipoapoptosis of primary murine hepatocytes by 750 mM PA (33.9 6 3.8%) was significantly inhibited by 30 mM BEL (26.4 6 3.6%), 100 mM PACOCF 3 (20.5 6 3.1%), or 300 ng/ml PTX (26.9 6 1.4%) (P , 0.05, 0.01, and 0.05, respectively).

LPC content in NAFLD liver specimens
We next studied whether LPC is able to induce liver injury by directly injecting LPC in vivo. When 60 mg/kg LPC was administered to ICR mice through tail veins, AST/ALT levels were notably increased at 24 h after injection (P , 0.005 for both comparisons) (Fig. 7A). Hematoxylin and eosin and TUNEL staining of histological sections obtained at 3 days after LPC injection showed lobular hepatitis with a histological score of 1 (focal lytic necrosis, one focus or less per 3100 field) without evidence of steatosis (38) and TUNEL 1 apoptotic cells in the Fig. 4. Intracellular events associated with the lipoapoptosis of hepatocytes. A: Chang cells were treated with PA for the indicated time periods, and 10-N-nonyl acridine orange (NAO) fluorescence was determined by flow cytometry as a measure of mitochondrial cardiolipin content. B: Chang cells were treated with PA for 24 h after pretreatment with BEL or PACOCF 3 , and NAO fluorescence was measured as in A. C: Chang cells were treated with PA with or without BEL pretreatment. After cell fractionation, the presence of Bid in the heavy membrane fraction was examined by Western blotting. D: Chang cells were treated with PA for the indicated time periods and then incubated with JC-1 for 10 min. Emission at 530 and 590 nm was measured, and absorption at 590 nm (A 590 )/A 530 was calculated as an indicator of mitochondrial potential. Treated cells were fractionated, and the presence of cytochrome c in the cytosolic fraction was examined by Western blotting (inset). E: Chang cells were treated with PA after pretreatment with zVAD-fmk, and MTT assay was done. F: Chang cells were treated with PA for 24 h with or without PACOCF 3 pretreatment, and cell lysate was subjected to Western blotting using antibodies specific for phosphorylated c-Jun N-terminal kinase (JNK) or total JNK (upper panel). Chang cells were also treated with LPC for the indicated time, and cell lysate was subjected to Western blotting (lower panel). G: Stable transfectants expressing dominant-negative Ga i2 mutant were treated with PA or LPC, and cell lysate was subjected to Western blotting as in F. H: Chang cells were treated with PA after pretreatment with SP600125, a JNK inhibitor, and MTT assay was done. Values shown are means 6 SD. ** P , 0.01, *** P , 0.001. area of hepatitis, respectively, supporting our hypothesis that LPC plays a role as a mediator of hepatocyte injury in vivo (Fig. 7B).
We finally measured LPC content in liver biopsy specimens. ANOVA revealed that LPC content in the liver specimens from patients with NAFLD was significantly higher than that in the nontumor part of the liver from patients with metastatic colon cancer (P , 0.005). LPC content in the liver specimens from patients with NASH (NAFLD class 3-4) appears to be higher compared with that from patients with NAFLD class 2 (lobular hepatitis only); however, the difference was not statistically significant by the Scheffé test (P . 0.1) (Fig. 7C).

DISCUSSION
We observed that PA induces the apoptosis of hepatocytes through conversion to LPC, which led to the activation of GPCR, mitochondrial events, and JNK (Fig. 8). Although Chang cells and primary hepatocytes from mice or human all underwent apoptosis after treatment with PA, the susceptibility to lipoapoptosis was different, which could be attributable to differences in cell types, species, or the method of cell death assay. Although the identity of Chang cells has been questioned, recent papers have demonstrated the hepatocyte nature of Chang cells based on the expression of a-fetoprotein and albumin (39,40), which was again supported by our results showing the expression of various markers for hepatocytes and the removal of ammonia by Chang cells. In contrast to our expectation, the conversion of PA to ceramide/sphingosine did not play a role in Chang cell lipoapoptosis, because both fumonisin B1 and myriocin inhibiting ceramide synthesis at two different steps did not block Chang cell lipoapoptosis. These results differ from previous findings suggesting essential roles for ceramide in the lipoapoptosis of pancreatic islet cells or hematopoietic cells (4,41). However, papers reporting the ceramide-independent lipoapoptosis of various types of cells, including hepatocytes, have also been published (5,7,42,43). These discrepancies might be attributable to differences in cell types or experimental conditions.
Instead of the ceramide pathway, conversion to DAG, PC, and then LPC appears to play an important role in Chang cell lipoapoptosis. Palmitoyl-CoA can be incorpo- rated to phosphatidic acid, which can be converted to DAG, a well-known PKC activator (30) (Fig. 1). Downstream of DAG, we noticed that DAG could be converted to PC and then to LPC, another PKC activator, by PLA 2 (44). Because PLA 2 participates in several cell death models (12,13), we investigated whether PLA 2 is involved in the lipoapoptosis of hepatocytes. Our results showing a remarkable inhibition of Chang cell lipoapoptosis by BEL or PACOCF 3 suggest the involvement of iPLA 2 and LPC rather than DAG upstream of PLA 2 . Although BEL has been reported to inhibit phosphatidic acid phosphohydrolase in addition to iPLA 2 (24), our finding that another iPLA 2 -specific inhibitor, PACOCF 3 , which is structurally unrelated to BEL (24), dramatically inhibited Chang cell death by PA suggests that iPLA 2 plays a critical role in lipoapoptosis. Moreover, LPC content was increased significantly after PA treatment of Chang cells and reverted to a normal level with BEL, suggesting an important role of LPC as a death effector in PA-induced lipoapoptosis. However, it is not clear which among several types of iPLA 2 that could be inhibited by BEL (45,46) is responsible for the lipoapoptosis of hepatocytes. Because we observed significant decreases in Chang cell lipoapoptosis with both iPLA 2 b and iPLA 2 g siRNA, a single type of iPLA 2 may not Fig. 6. Death of primary hepatocytes by PA. A: Primary human hepatocytes isolated by two-step collagenase perfusion were treated with PA for 24 h, and Hoechst/PI staining was conducted as in Fig. 2C. B: Primary human hepatocytes pretreated with BEL were incubated with PA as in Fig. 3A, and MTT assay was done. C: Primary human hepatocytes pretreated with fumonisin B1 or myriocin were incubated with PA as in Fig. 2G, H, and MTT assay was conducted. D: After pretreatment with PTX, primary human hepatocytes were treated with PA or exogenous LPC as in Fig. 3F. E: After treatment of primary human hepatocytes with various combinations of PA and OA, MTT assay was conducted as in Fig. 5C. F: After treatment of human primary hepatocytes with PA in the presence or absence of BEL, PTX, or OA, aspartate aminotransferase (AST) level in the supernatant was measured using a blood chemistry analyzer. Values shown are means 6 SD. *** P , 0.001. be fully responsible for the production of LPC from FFAs. Further work will be necessary to elucidate the roles of specific iPLA 2 types in lipoapoptosis.
The role of PLA 2 in cell death/survival has been reported in diverse models of hypoxia, reperfusion injury, and reactive oxygen species damage (12,13,47,48). Our results are similar to previous findings showing a significant role of iPLA 2 in hypoxic cell injury (12,47); however, our findings are different from those results in that LPC appears to be an effector downstream of iPLA 2 activation. Our data that iPLA 2 activity was significantly (nearly 2fold) increased by PA and that the increased activity was completely reversed by BEL are also similar to previous data that iPLA 2 activity was increased in the course of hypoxic cell death (12). The mechanism of the activation of iPLA 2 activity by PA is not clearly understood and may involve translocation to target organelles, displacement of inhibitor molecules by FFA-induced endoplasmic reticulum stress, or reactive oxygen species (12,24,49). Incomplete protection of PA-induced hepatocyte death by a combination of iPLA 2 b and iPLA 2 g siRNAs or iPLA 2 inhibitors such as BEL and PACOCF 3 suggests a potential role of iPLA 2 -independent mechanisms in the lipoapoptosis of hepatocytes. For instance, the role of fatty acid metabolites other than LPC, such as DAG and ceramide, cannot be totally eliminated, because we used mostly pharmacological inhibitors that could have overlapping substrate specificities.
Exogenous LPC has been reported to induce apoptosis (44,50) and stimulate inflammatory cells (17). However, the proapoptotic role of endogenous LPC has not been reported, to our knowledge. As iPLA 2 does not have a Fig. 8. Proposed model for NAFLD. DAG produced by the incorporation of FFA into the glycerol backbone may be converted to LPC or TG depending on the absolute/relative abundance of saturated/unsaturated fatty acids. Steatosis or hepatitis may occur when the TG pathway or the LPC pathway, respectively, prevails. LPC may induce apoptosis through the activation of G-proteincoupled receptor (GPCR/Ga i ), mitochondrial events, and JNK. Fig. 7. LPC, liver injury in vivo, and nonalcoholic fatty liver disease (NAFLD). A: LPC was administered into the tail vein of ICR mice, and AST/ALT levels were measured. B: Histological grading of the liver injury was assessed according to a published criteria (38). TUNEL staining was done on the adjacent sections to identify apoptotic hepatocytes in the area of hepatitis. H&E, hematoxylin and eosin. Magnification 3100. C: LPC content was measured in liver biopsy specimens from patients with NAFLD and in the nontumoral part of the liver from patients undergoing hepatic resection for metastatic colon cancer (Normal). The severity of NAFLD was assessed according to a published classification (2). Values shown are means 6 SD. ** P , 0.01. specificity toward PC, the roles of other lysophospholipids cannot be completely eliminated. However, in vivo concentrations of other phospholipids, such as phosphatidylethanolamine, phosphatidylglycerol, phosphatidylinositol, and phosphatidylserine, are around or less than one-tenth of that of PC (51,52). Much higher in vivo concentrations and much stronger apoptotic activity of LPC compared with LPE, LPG, LPI, or LPS essentially eliminates the possibility that other minor lysophospholipids play significant roles in lipoapoptosis. Previous papers also showed obvious apoptotic activity of LPC, whereas that of other lysophospholipids has not been clearly demonstrated (44,50). Products of PLA 2 released from PC together with LPC, such as arachidonic acid or its derivatives (10,11), have diverse effects in vivo. Whereas dipalmitoyl-phosphatidic acid is predominantly produced by PA treatment (53), a certain amount of arachidonic acid may be produced by repeated deacylation and reacylation at the sn-2 position and may contribute the phenotypes associated with FFA excess. However, a recent paper showed much weaker activation of stress molecules such as JNK by a variety of unsaturated FFAs, including arachidonic acid, compared with saturated FFAs (54), consistent with our results suggesting that saturated FFAs exert much stronger cytotoxicity compared with unsaturated FFAs.
Intracellular events associated with LPC-induced apoptosis are not clearly defined. Recent papers reported that GPCRs mediate various actions of LPC (33,34), but it remains controversial whether GPCRs are LPC receptors or mediate LPC actions indirectly. The inhibition of PA-or LPC-induced Chang cell apoptosis by PTX or a dominant-negative Ga i mutant (18) suggests that certain GPCR/Ga i is involved in apoptosis by endogenous or exogenous LPC (55). The effect of a dominant-negative Ga i mutant observed in this study is not mediated by lysophosphatidic acid (LPA), which could be produced from LPC by autotaxin (56), because Chang cell death by PA or LPC was not inhibited by Ki16425, an LPA 1 /LPA 3 receptor antagonist (56), and LPA up to 50 mM did not induce Chang cell death (M. S. Han et al., unpublished data). Furthermore, LPA is a well-known survival or growth factor and has been shown to protect cells from apoptosis (57).
Because GPCRs recognize extracellular ligands, LPC might be released to the extracellular space and bind to GPCRs (58). Instead, LPC might enter the receptor binding site in a lateral manner between transmembrane regions of the receptor without leaving the membrane (59), which is more consistent with the absence of the increase of LPC content in the culture supernatant after PA treatment. However, it is not proven whether such a model proposed for LPA is valid for LPC binding to its receptors. Downstream of receptor binding, we observed decreased cardiolipin content after PA treatment, consistent with a previous report (53). The PA-induced decrease in cardiolipin content could be attributable to the inhibition of CDP-DAG synthase, the rate-limiting step in cardiolipin biosynthesis, by LPC, as suggested previously (60).
Although a different mechanism, such as poor substrate availability of dipalmitoyl-phosphatidic acid for CDP-DAG synthase, has been proposed (53), the inhibition of lipoapoptosis and the reversal of the decrease in cardiolipin content by PLA 2 inhibitors downstream of phosphatidic acid support a role for LPC rather than dipalmitoylphosphatidic acid upstream of PLA 2 . The loss of cardiolipin, a mitochondria-specific phospholipid that maintains mitochondrial membrane stability by binding to cytochrome c (61), is probably responsible for cytochrome c translocation, the loss of mitochondrial potential, and cell death after PA treatment. Besides receptor binding, LPC might directly induce mitochondrial events by recruiting Bid, a well-known initiator of apoptosis, to the mitochondrial membrane (35). JNK activation by saturated FFAs may also contribute to the cytochrome c release and lipoapoptosis in hepatocytes, whereas the contribution of JNK activation in the lipoapoptosis of Chang cells does not appear to be dominant. Although we focused on the mitochondrial events associated with LPC accumulation in the course of hepatocyte death by PA, previous papers showed the involvement of other organelles, such as endoplasmic reticulum and lysosome, in the lipoapoptosis of hepatocytes (5)(6)(7)(8). Further work will be required to address the potential relationship between the iPLA 2 /LPC pathway and endoplasmic reticulum stress or lysosomal permeabilization (7,8,24).
The inability of OA to induce lipoapoptosis is consistent with JNK activation by saturated FFAs but not by unsaturated fatty acids (54). The increased TG content by OA but not by PA and the further increase by combined OA/ PA treatment suggest that an increased channeling of PA into TG plays a protective role against PA-induced lipoapoptosis. Although channeling into TG protects against lipotoxicity, excessive TG may impose deleterious effects on cellular function, such as impaired insulin production in pancreatic b-cells. An increased TG content in the liver leads to steatosis, which is one of the two main features of NAFLD. The preferential increase of TG content by unsaturated fatty acids might be attributable to the increased stability of lipid droplets containing a higher percentage of unsaturated acyl chains (37).
The death of hepatocytes constitutes another main feature of NAFLD, leading to inflammation, fibrosis, and ultimately cirrhosis (NASH). Cells may enter the steatosis track if the DAGYTG pathway prevails or the death track if the DAGYPCYLPC pathway predominates, which might depend on the amount and types of available FFAs or the "second hit," such as inflammatory cytokines (62,63), that may determine the conversion of steatosis to hepatitis (Fig. 8). Consistent with our in vitro models, we observed increased LPC content in patients with NAFLD depending on the severity of the disease, yet we cannot completely eliminate the possibility that the difference in the methods of tissue sampling between NAFLD tissue and control tissue might have affected the results. The development of hepatitis and increased liver enzymes after LPC administration also provides another line of in vivo evidence supporting the role of LPC in liver injury.
Our results demonstrate that FFAs lead to steatosis or lipoapoptosis according to the absolute/relative abundance of FFAs, consistent with previous epidemiological data reporting more prevalent NASH in patients with a high intake of saturated fatty acids (64). Our data suggest the potential therapeutic values of the modulation of dietary fatty acids and inhibitors of PLA 2 and GPCR/Ga i in the prevention/treatment of hepatocyte injury in NASH. However, caution is advised for PLA 2 and GPCR/Ga i inhibitors as therapeutic agents, because they could have diverse biological effects on various tissues.