Mitochondrial Reactive Oxygen Species Signal Hepatocyte Steatosis by Regulating the Phosphatidylinositol 3-Kinase Cell Survival Pathway*

Abnormal dietary intake of macronutrients is implicated in the development of obesity and fatty liver disease. Steatosis develops in cultured hepatocytes exposed to medium containing either a high concentration of long chain free fatty acids (HFFA) or medium deficient in methionine and choline (MCD). This study examined the mitochondrial reactive oxygen species (ROS)-dependent regulation of the phosphoinositol (PI) 3-kinase pathway in steatosis induced by exposure of AML-12 mouse hepatocytes to MCD or HFFA medium. Exposure to either MCD or HFFA medium resulted in increased production of superoxide anions and H2O2, transduction of the PI 3-kinase pathway and steatosis. Inhibition of PI 3-kinase with LY294002 prevented steatosis. Pharmacologically inhibiting electron transport chain complex III production of ROS prevented activation of PI 3-kinase during macronutrient perturbation, whereas pharmacologically promoting electron transport chain complex III ROS production activated PI 3-kinase independent of nutrient input. The data suggest that H2O2 is the ROS species involved in signal transduction; promoting the rapid conversion of superoxide to H2O2 does not inhibit PI 3-kinase pathway activation during nutrient perturbation, and exogenous H2O2 activates it independent of nutrient input. In addition to transducing PI 3-kinase, the ROS-dependent signal cascade amplifies the PI 3-kinase signal by maintaining phosphatase and tensin homolog in its inactive phosphorylated state. Knockdown of phosphatase and tensin homolog by small interfering RNA independently activated the PI 3-kinase pathway. Our findings suggest a common path for response to altered nutrition involving mitochondrial ROS-dependent PI 3-kinase pathway regulation, leading to steatosis.

Innumerable epidemiologic studies demonstrate that inappropriate intake of macronutrients and a sedentary lifestyle conspire to produce rampant obesity in the United States. Hepatic steatosis (the accumulation of excess neutral fat within hepatocytes) in obese persons is thought to result from metabolic disturbances attendant to obesity. An estimated 30 million adults and 1.6 million children in the United States have nonalcoholic fatty liver disease (NAFLD), 2 of which an estimated 30% have nonalcoholic steatosis (NASH), which can lead to chronic debilitating liver injury, including cirrhosis (1). Hepatic steatosis is the defining feature of NAFLD (2). It is commonly held that steatosis is the precursor of events that lead to the transition of NAFLD to NASH, including the induction of oxidative stress and activation of inflammatory pathways (3,4). The development of hepatic steatosis is therefore considered to be the necessary first effect of disturbed nutrition that induces the cascade of events leading to NASH. How disturbed nutrition causes hepatic steatosis is not well understood.
Experimental NASH can be produced in laboratory rodents by dietary manipulation, including both restricted intake (i.e. methionine-choline deficient (MCD) diet) (5) and excess intake (i.e. various high fat diets) (6,7). How nutritional excess and deprivation can both culminate in experimental NASH is not clear (8).
The MCD diet-fed mouse is a well accepted and reliable model of experimental NASH (9,10). Our findings using this model have suggested that this is not so much a model of toxicity induced by deficiency of methyl donor molecules, as is commonly thought, but a model of dietary imbalance that incites an orchestrated response leading to experimental NASH. In a detailed time course study of female A/J mice fed the MCD diet, we found that these mice developed significant steatosis and massive overexpression of the Th1 cytokine osteopontin (OPN) within 1 week of starting the diet (11). The hepatic expression of OPN protein was 16-fold increased after 1 day of diet treatment and 40-fold increased after 2 days, whereas mRNA expression was not increased until 2 weeks, suggesting translational regulation of early OPN protein expression. These events occurred at a time when the diet is unlikely to have produced toxic deficiency. Furthermore, no oxidative injury developed until 4 -8 weeks of dietary depriva-tion. Therefore, it seems reasonable to conclude that the dietary change directly influenced the induction of the metabolic response represented by OPN overexpression.
To further explore the possibility that the MCD diet represented a signal leading to steatosis and other expressions of experimental NASH, including the overexpression of OPN, we examined the effects of substituting medium deficient in methionine and choline (MCD medium) on AML-12 mouse hepatocytes in vitro (12). Substituting MCD medium for control medium produced significant hepatocyte steatosis by 18 h of exposure. Exposure to MCD medium induced significant increases in OPN protein expression by 6 h of treatment. Thus, the in vitro situation mimicked the early events in the in vivo MCD diet model. We further showed that exposure to the MCD medium induced a significant increase in the PI 3-kinase p85 regulatory subunit expression beginning at 1 h of incubation and a sustained increase in pAkt expression during 1-18 h of incubation. Inhibiting PI 3-kinase with LY294002 blocked the MCD medium-induced increases of OPN expression and prevented cellular steatosis from developing. Taken together, these findings suggested that there is an orchestrated cellular response to the stress induced by exposure to MCD medium in which OPN expression and steatosis are dependent upon the activation of the PI 3-kinase pathway.
In the present study, we sought to determine the mechanisms by which nutrient perturbation produces steatosis in hepatocytes in culture. Feldstein et al. (13) demonstrated that human hepatocytes (HepG2 cells) in culture treated with medium containing excess long-chain fatty acid (HFFA) medium develop steatosis. We hypothesized that nutrient perturbation in the form of macronutrient deprivation (MCD medium) or excess (HFFA medium) activates and modulates the PI 3-kinase cell signaling pathway to produce steatosis. We proposed that its activation results from increased production of mitochondrial reactive oxygen species (ROS) induced by the change in macronutrient input. Mitochondrial dysfunction and injury are thought to be important in the pathogenesis of NAFLD (14 -18). Mitochondria are a major source of ROS in cells (19). ROS are produced by mitochondria during normal oxidative metabolism, and we hypothesized that a change in macronutrient input could induce a metabolic stress that would increase the rate of ROS production by mitochondria. ROS species, in particular H 2 O 2 , can act as signals by increasing tyrosine phosphorylation of receptor tyrosine kinases even in the absence of growth factor binding, thus independently inducing the activation of downstream signaling cascades like that of PI 3-kinase (20). Showing that perturbation of macronutrient input to hepatocytes, either deprivation or excess input, produces steatosis by mechanisms that are dependent upon mitochondrial ROS regulating the PI 3-kinase pathway would provide evidence for a novel link between abnormal dietary intake and NASH.
We further proposed that phosphatase and tensin homolog (PTEN) is mechanistically involved in the positive modulation of the PI 3-kinase signaling pathway during nutrient perturbation. PTEN is the natural inhibitor of the PI 3-kinase pathway, acting to degrade PIP 3 produced by PI 3-kinase before it can activate Akt through phosphorylation. PTEN can be inacti-vated by exposure to ROS. Specifically, exposure of purified PTEN or cells to H 2 O 2 can result in time and concentration-dependent oxidation of PTEN and reduction of phosphatase activity (21). Further, PTEN function is regulated by its membrane interaction, which is determined in large part by the degree of phosphorylation of the C-terminal tail (22). The constitutively phosphorylated protein (pPTEN) must be dephosphorylated before it can bind to membrane PDZ-containing proteins and assume full functionality as a lipid phosphatase (23,24). However, the dephosphorylated protein is unstable and rapidly degraded (23). Thus, the phosphatase activity of PTEN is entirely dependent upon the rate of dephosphorylation of pPTEN, providing tight regulation of this important suppressor of the PI 3-kinase/Akt pathway. PTEN activity is modulated in this way during ischemia/perfusion stress (25), which suggests that ROS signaling may be involved. Finally, the importance of PTEN in hepatic steatosis was demonstrated by showing that liver-specific PTEN knockouts develop steatosis and NASH-like disease (26). If nutrient perturbation activates PI 3-kinase and inhibits PTEN function, it would provide evidence of a novel signal amplification mechanism with potential importance in the evolution of fatty liver disease.
The overall aim of the present study was to determine the signaling mechanisms involved in the development of hepatocyte steatosis resulting from altered macronutrient input, specifically the role of mitochondrial ROS in the downstream activation and modulation of the PI 3-kinase pathway.

EXPERIMENTAL PROCEDURES
Hepatocyte Cell Culture-AML-12 cells, a mouse hepatocyte cell line that has been stably transfected with human transforming growth factor-␣ (American Type Culture Collection, Manassas, VA), were treated as previously described (12). In brief, these cells are grown to 70 -80% confluence in Dulbecco's modified Eagle's medium/Ham's nutrient mixture F-12 containing 10% fetal bovine serum (for growth factors) and then made quiescent by incubating for 24 h in standard medium without fetal bovine serum. Quiescent cells were then exposed either to control growth medium (MCϩ; methionine (0.116 mM) and choline (0.064 mM)) or identical medium that was manufactured to be completely deficient in methionine and choline (MCD medium) purchased from Invitrogen. In separate experiments, cells were treated with control medium and HFFA medium (addition of 1.0 mmol/liter fatty acid mixture of 2:1 oleate/palmitate) with 1% bovine serum albumin supplemented with identical glucose and serum concentrations. This HFFA construct was used because it has been successfully used in vitro to produce steatosis and experimental NASH in cultured primary mouse hepatocytes and HepG2 cells (13). Cells were then maintained at 37°C for various times before harvesting for analysis. Inclusion of pharmacologic agents with control or test media was used as described under "Results." Steatosis Analysis-Cellular triglyceride content was determined after 18 h of treatment as described (12). Cells were scraped from culture plates that had been rinsed with cold PBS and were subsequently pelleted by centrifugation. The cell pellet was suspended in 20 mM Tris buffer, pH 7.5, and homogenized. Cell triglyceride content in the whole homogenates was measured using a spectrophotometric kit procedure (Thermo DMA, Arlington, TX) and expressed as g of triglyceride/mg of protein. Hepatocyte steatosis was also determined in situ using standard Oil-Red O dye staining.
Confocal Time Lapse Microscopy for Analysis of ROS-AML-12 cells grown on glass bottom Petri dishes (Matek, Ashland, MA) and made quiescent were exposed to control medium containing 4Ј,6-diamidino-2-phenylindole (DAPI) dye (Sigma) for 20 min, rinsed with sterile PBS once, and then incubated with control medium containing the ROS-sensitive dyes dihydroethidium (DHE) (Invitrogen) or 2Ј,7Ј-dichlorofluorescin diacetate (DCF) (Sigma) in PBS for 30 min at 37°C. Dishes were once again rinsed with PBS and subsequently exposed to control and test media containing 10% serum and DHE. Some experiments were done in the presence of the superoxide dismutase Mn(III) tetrakis (4-benzoic acid) porphyrin (MnTBAP) (A. G. Scientific, San Diego, CA) or electron transport chain (ETC) complex III inhibitors stigmatellin, myxothiazol, or antimycin A (Sigma). In some experiments, AML-12 cells were exposed to exogenous H 2 O 2 (25 M or 1.0 mM) by its addition to the medium. For experiments involving oxidation of PTEN, we chose 1 mM H 2 O 2 as the lowest concentration resulting in nearly maximal oxidation of PTEN when applied to HeLa and NIH 3T3 cells (21). Preliminary studies showed this concentration to not be lethal to AML-12 cells. If experiments were extended beyond 15 min, H 2 O 2 was replenished by injection every 15 min to maintain the concentration Ն1 mM. Real time fluorescent images were obtained using a Zeiss 510 META confocal laser-scanning microscope while maintaining dishes at 37°C.
pPTEN Immunofluorescence-AML-12 cells were grown and treated as described above for confocal microscopy with the exception that DAPI and DHE were not included. Cells were treated for 30 min with control and test media and then fixed in situ using 4% paraformaldehyde and washed with 0.2% Triton X-100 followed by PBS. This was followed by incubation at room temperature for 1 h with 3% bovine serum albumin and overnight incubation with the primary pPTEN (Ser 380 /Thr 382 / Thr 383 ) antibody (Cell Signaling, Danvers, MA) at 4°C, followed by incubation with an anti-rabbit IgG fluorescein isothiocyanate antibody and DAPI. Fluorescence was recorded with green (fluorescein isothiocyanate) and blue (DAPI) filters using an upright LEICA DMR-HC microscope.
Western Blot Analysis-Immunoreactive levels of various proteins and phosphoproteins were assessed by Western blot analysis. Cell homogenates containing 10 -25 g of protein were resolved by 7.5-10% SDS-polyacrylamide gel electrophoresis and transferred overnight onto nitrocellulose membranes by electrophoresis. Specific primary antibodies to PI 3-kinase p85 subunit, pAkt1 (Ser 473 ), and Akt2 (Upstate Biotechnology, Lake Placid, NY); Akt, pAkt, PTEN, and pPTEN (Cell Signaling, Danvers, MA); and Akt1 (Chemicon, Temecula, CA) were used to assess their expression. ␤-Actin antibodies were used to confirm equal protein loading among samples; these results always showed equal loading and are not shown. The bound primary antibodies were detected with a respective horseradish peroxidase-conjugated secondary antibody (1:1000 dilution; Amersham Biosciences) and visualized with an enhanced chemiluminescence method. Protein expression levels were determined by densitometric analysis.
PTEN siRNA-AML-12 cells at 60 -75% confluence were transfected with a range of concentrations of small interfering RNA (siRNA) targeted against PTEN (target sequence ATC GAT AGC ATT TGC AGT ATA) using HiPerFect transfection reagent and protocol (Qiagen, Cambridge, MA). After 18 h, cells were harvested, and total cellular RNA was purified. The RNA was then reverse transcribed, and the resultant cDNA was used for quantitative real time RT-PCR for PTEN (primers available upon request). In separate experiments, total cellular protein was purified and used to load Western blots as described above.
Statistical Analysis-Comparison between groups was performed using Student's t test for unpaired samples or by analysis of variance. A p value of Ͻ0.05 was considered statistically significant.

Treating with MCD or HFFA Medium Produces Hepatocyte
Steatosis-Treating cultured AML-12 hepatocytes with MCD or HFFA medium for 18 h resulted in steatosis, as demonstrated by Oil-Red O staining (Fig. 1A). Average cell triglyceride content was significantly increased from control by treatment for 18 h with MCD (p Ͻ 0.05 versus control) and HFFA media (p Ͻ 0.01 versus control) (Fig. 1B). Cell viability as determined by trypan blue exclusion was not reduced by incubating with either MCD or HFFA medium for 18 h (data not shown).
PI 3-Kinase/Akt Pathway Transduction Is Necessary for Development of Hepatocyte Steatosis during Treatment with MCD or HFFA Medium-We previously demonstrated the activation of the PI 3-kinase pathway in AML-12 cells exposed to MCD medium with maximal activation at 1 h of exposure and that steatosis was prevented by treating cells exposed to MCD medium with LY294002 (12). In the present study, Western blot analysis of cell homogenates with specific antibodies against the p85 regulatory subunit and pAkt demonstrated transduction of the PI 3-kinase pathway to a similar degree after 1 h of exposure to either MCD or HFFA medium ( Fig. 2A). The isoforms of Akt, Akt1 and Akt2, have distinct roles in regard to growth and glucose homeostasis (27). Their relative importance in liver metabolism is not well known. To determine which Akt isoform is involved in this signaling cascade, we used specific Akt1 and Akt2 antibodies to demonstrate that Akt1 is the predominant isoform in AML-12 cells, and thus is the isoform predominantly involved (Fig. 2B). Moreover, expression of pAkt1 increased with exposure to MCD or HFFA medium in parallel with the expression of pAkt (Fig. 2, A and B). Finally, the increase in triglyceride content seen with either treatment at 18 h was reversed by treating cultures with the pharmacologic inhibitor of PI 3-kinase LY294002 (Fig. 2C). Thus, either treatment resulted in activation of the PI 3-kinase/Akt pathway, cellular steatosis, and reversal of this physiological effect when the PI 3-kinase pathway was inhibited. These data suggest that these conditions of macronutrient deprivation and excess produce steatosis by a mechanism that is dependent upon the PI 3-kinase/Akt pathway and predominantly involves the Akt1 isoform.
Treating with MCD or HFFA Medium Increases Hepatocyte Superoxide Production-We hypothesized that altered macronutrient input could produce metabolic stress that would stimulate hepatocyte mitochondria to increase ROS production. The ROS-sensitive vital dye DHE was used to detect the production of superoxide in real time using live cell confocal time lapse fluorescence microscopy. Incubating with either MCD or HFFA medium produced marked increases in red DHE fluorescence within 30 min (Fig. 3, A- FIGURE 2. Transduction of the PI 3-kinase/Akt pathway during exposure to MCD and HFFA medium and dependence of steatosis upon PI 3-kinase. A, Western blots for p85 and pAkt in AML-12 cells exposed to either HFFA or MCD medium for 1 h show increased p85 expression and increased phosphorylation of Akt relative to control cells. These blots show that either macronutrient perturbation resulted in activation of the PI 3-kinase cell survival pathway. B, Western blots for Akt1 and Akt2 in AML-12 cells exposed to control, HFFA, or MCD medium for 1 h show Akt1 to be the predominant Akt isoform in AML-12 cells and show that its expression does not change with exposure to nutrient stress. Increased expression of pAkt1 after exposure to macronutrient perturbation parallels the increased expression of pAkt. C, preventing PI 3-kinase/Akt signal transduction with the specific PI 3-kinase inhibitor LY294002 (LY) abrogates the steatosis produced by macronutrient perturbation. Treatment with 10 M LY (LY [10]) completely eliminates the increase in cellular triglyceride content produced by MCD treatment, whereas LY294002 at 10 -50 M concentration markedly reduces the increase produced by HFFA. These experiments provide confirmatory data on the involvement of the PI 3-kinase cell survival pathway in the development of steatosis resulting from macronutrient perturbation. *, p Ͻ 0.05. FIGURE 3. Production of superoxide anions by AML-12 hepatocytes exposed in culture to either MCD or HFFA medium. Time lapse images were captured by confocal fluorescence microscopy staining for nuclei (blue, DAPI stain) and superoxide (red, DHE stain). A, the superoxide content of cells exposed to control medium does not change over 30 min of incubation. B, the superoxide content of liver cells increases significantly when exposed to MCD medium for 30 min. C, the superoxide content of liver cells increases significantly when exposed to HFFA medium for 30 min. The effect seems to be somewhat greater than that seen with MCD medium, although direct quantitative comparisons are difficult. D, the increase in superoxide content seen after exposure to HFFA medium is decreased significantly when similar cells are co-treated with the exogenous superoxide dismutase MnTBAP (50 M). Similar findings in cells exposed to MCD medium and MnTBAP are not shown. 3D), indicating the rapid elimination of superoxide in the formation of H 2 O 2 . However, transduction of the PI 3-kinase pathway was not altered in cells exposed to either MCD or HFFA medium along with the permeable superoxide dismutase mimetic MnTBAP as evidenced by no reduction in expression of p85 and pAkt (Fig. 4, A and B). Thus, signal transduction with either nutrient perturbation was unaffected by the rapid conversion of superoxide to H 2 O 2 . The ROS-sensitive vital dye DCF was used to detect the production of H 2 O 2 in real time using live cell confocal time lapse fluorescence microscopy. Incubating with either MCD or HFFA medium produced marked increases in green DCF fluorescence within 10 min of exposure (Fig. 4C) relative to control medium and similar to cells exposed to exogenous H 2 O 2 , indicating that either change in nutritional milieu prompted an increase in H 2 O 2 production. The H 2 O 2 appeared to be distributed throughout the cytoplasm of the cells, as evidenced by the distribution of fluorescence (Fig. 4C), and persisted for at least 30 min after exposure, when experiments were terminated (data not shown). Cells grown in control medium with inclusion of H 2 O 2 demonstrated an increase in p85 and pAkt expression to a similar degree as with nutrient perturbation (Fig. 4D). These data show that accelerating the conversion of superoxide to H 2 O 2 does not reduce the effect of nutrient change on PI 3-kinase activation, H 2 O 2 is rapidly generated and distributed to the cytoplasmic compartment with either nutrient change, and H 2 O 2 can initiate transduction of the PI 3-kinase pathway independent of nutrient perturbation.
Mitochondria Are the Source of Increased Superoxide Production Seen with MCD or HFFA Medium-Complexes I and III of the mitochondrial ETC leak superoxide anions during normal respiration. Superoxide is specifically produced in large part by complex III of the mitochondrial ETC. To confirm complex III as the source of superoxide production during acute nutrient perturbation, cells exposed to either MDC or HFFA medium were treated with specific ETC complex III inhibitors stigmatellin and myxothiazol, both of which decrease mitochondrial superoxide production. As expected, the increase in DHE fluorescence with HFFA exposure was abrogated by treatment with either compound (Fig. 5, A-C). Similarly, treatment with either compound eliminated the increase in DHE fluorescence with exposure to MCD medium (data not shown). Moreover, both ETC complex III inhibitors eliminated the increased expressions of p85 and pAkt that were produced by either nutrient perturbation alone (Fig. 5, D-F).
Antimycin A is a complex III inhibitor that paradoxically increases superoxide production (28). Treating AML-12 cells in control medium with antimycin A for 30 min increased DHE fluorescence (Fig. 6A) and produced a dose-dependent increase in p85 and pAkt expression to a similar degree as nutrient perturbation (Fig. 6C). To determine whether the increased ROS production induced by treatment with antimycin A could produce steatosis in the absence of nutrient perturbation, AML-12 cells grown in control medium were exposed to very low dose (10 -20 nM) antimycin A (necessary because of its toxicity at higher dose) for 5 days, which resulted in steatosis (Fig. 6B). Taken together, these data indicate the source of increased superoxide production resulting from changes in nutrient input to be largely complex III of the ETC.

PTEN Is Inactivated during Nutrient Perturbation, Leading to Positive Modulation of the PI 3-Kinase/Akt Pathway-PTEN
is a lipid phosphatase naturally involved in inhibiting the PI 3-kinase pathway by degrading PIP 3 . PTEN is constitutively phosphorylated, and failure to dephosphorylate it leaves it inactive and unable to effectively degrade PIP 3 (24). Exposure of PTEN to H 2 O 2 can oxidize it, which also reduces its ability to degrade PIP 3 (21). If exposure to MCD or HFFA resulted in either of these changes in PTEN, it would result in loss of inhibition of the PI 3-kinase/Akt pathway by preventing PIP 3 degradation. AML-12 cells exposed to MCD or HFFA medium expressed substantial increases in cellular pPTEN, as demonstrated by immunofluorescence microscopy (Fig. 7A) and Western blot (Fig. 7B). We were able to demonstrate oxidation of PTEN when cells were exposed to 1 mM H 2 O 2 , but not with either nutrient treatment (Fig. 7C). These findings suggest that in this setting, phosphorylation is the main mechanism of

. Evidence for a role of H 2 O 2 in signaling PI 3-kinase transduction
in AML-12 hepatocytes exposed in culture to nutritional stress. A, Western blot shows increased p85 and pAkt expression in cells exposed to MCD medium for 1 h relative to control. Treatment with the superoxide dismutase MnTBAP (25 M) did not reduce p85 and pAkt expression resulting from exposure to MCD medium. B, treatment with HFFA medium for 1 h results in increased p85 and pAkt expression in cells relative to control. Treatment with MnTBAP (25 M) did not reduce p85 and pAkt expression resulting from exposure to HFFA medium. C, images were captured by confocal fluorescence microscopy staining for nuclei (blue, DAPI stain) and H 2 O 2 (green, DCF stain). Cells exposed to either MCD or HFFA medium for 10 min show H 2 O 2 generation far in excess of cells treated with control medium and DCF staining equivalent to that of cells treated with exogenous 25 M H 2 O 2 (data not shown). D, Western blots for p85 and pAkt in cells exposed to exogenous H 2 O 2 (25 M) for 1 h show increased p85 expression (ϳ1.5 times control) and massively increased pAkt expression relative to untreated cells.
PTEN inactivation. This could be the result of increased active phosphorylation or failure to dephosphorylate. They also suggest that the functional intracellular concentration of H 2 O 2 produced by nutrient treatment is less than that resulting from exposure to 1 mM exogenous H 2 O 2 . To demonstrate that PTEN functions to inhibit the PI 3-kinase pathway specifically in hepatocytes, AML-12 cells cultured in control medium were transfected with siRNA against PTEN. This treatment resulted in dose-dependent reduction in PTEN mRNA expression and activation of the PI 3-kinase pathway, as evidenced by increased cellular concentration of pAkt (Fig. 8, A-C). Taken together, these data indicate that PTEN functions in hepatocytes to inhibit the PI 3-kinase pathway and that during nutrient perturbation, the inhibition of PTEN function through maintaining a relatively phosphorylated state potentially provides amplification of PI 3-kinase signaling.
A Critical Threshold Concentration of Free Fatty Acid Is Needed to Produce PI 3-Kinase Signal Transduction-In the foregoing experiments, HFFA was defined as a 1.0 mmol/liter fatty acid mixture of 2:1 oleate/palmitate with 1% bovine serum albumin supplemented with identical glucose and serum concentrations as taken from Feldstein's work (13). We sought to determine exactly what concentration of fatty acid is needed to exert the effect of exposure to HFFA medium and to determine if the effect of increasing FFA concentrations is graded versus a critical threshold concentration to produce the effect. The data FIGURE 5. Effect of specific pharmacologic mitochondrial ETC complex III inhibitors on AML-12 hepatocytes exposed in culture to nutritional stress. Myxothiazol and stigmatellin are compounds well known to decrease superoxide production from complex III of the ETC. A, the superoxide content of liver cells as evidenced by red DHE fluorescence increases significantly when exposed to HFFA medium for 30 min. B, co-treatment of AML-12 cells with HFFA medium and the ETC inhibitor myxothiazol (5 M) substantially reduces DHE fluorescence compared with cells treated with HFFA alone. C, co-treatment of AML-12 cells with HFFA medium and the ETC inhibitor stigmatellin (5 M) substantially reduces DHE fluorescence compared with cells treated with HFFA alone. D, co-treating cells exposed to MCD medium for 1 h with stigmatellin (dose effect 1, 2, and 5 M) reduces the increased p85 expression and eliminates the increased pAkt expression seen in cells treated with MCD medium alone. E, a similar effect is seen when co-treating cells exposed to MCD medium for 1 h with myxothiazol (5 M). F, similarly, co-treating cells exposed to HFFA medium for 1 h with the ETC inhibitors stigmatellin (5 M) or myxothiazol (5 M) eliminates the increased p85 and pAkt expressions observed with HFFA medium alone. shown in Fig. 9 demonstrates that a critical threshold concentration ϳ0.25 mmol/liter fatty acid mixture of 2:1 oleate/palmitate produces the PI 3-kinase transduction ascribed to HFFA medium.

DISCUSSION
The present study examined the signaling mechanisms involved in the development of hepatocyte steatosis resulting from altered macronutrient input to hepatocytes in culture and resulted in several important findings. 1) Both nutrient deprivation in the form of MCD medium and nutrient excess in the form of HFFA medium resulted in a rapid increase in superoxide production. 2) Either nutrient perturbation resulted in transduction of the PI 3-kinase signaling pathway and steatosis that can be prevented by inhibiting PI 3-kinase. 3) We show through a variety of manipulations that the mitochondrial ETC III is the source of ROS and that H 2 O 2 is the probable ROS signaling the transduction of the PI 3-kinase pathway during macronutrient perturbation. 4) Increasing superoxide production during normal cellular respiration by means other than macronutrient perturbation has the identical effects of transducing the PI 3-kinase pathway and producing steatosis. 5) Inhibition of PTEN through phosphorylation appears to be involved in positive modulation of the PI 3-kinase signal during macronutrient perturbation. These findings link abnormal macronutrient input to a signaled response, leading to hepatocyte steatosis.
We construe the abrupt increase in ROS production during nutrient perturbation to be the result of an acute metabolic stress imposed by the change in macronutrient input. This find- FIGURE 7. PTEN inactivation in AML-12 hepatocytes exposed in culture to MCD or HFFA medium. PTEN can be inactivated by either maintaining phosphorylation or by oxidation. A, immunofluorescence staining with a specific antibody to pPTEN (inactive phosphorylated form) at 30 min of exposure to either MCD or HFFA medium demonstrates significant increases in pPTEN (green fluorescence) relative to control medium-treated cells (blue, DAPIstained nuclei). B, cells exposed to MCD or HFFA medium for 30 min show a marked increase in pPTEN expression by Western blot relative to control, whereas there is no increase in total PTEN (note that HFFA Western blot was performed separately from the MCD blot). These data indicate that PTEN is inactivated predominantly by phosphorylation during short term macronutrient perturbation. C, Western blot run with a nonreducing gel shows greater mobility of the oxidized species. Treating AML-12 cells with 1 mM H 2 O 2 resulted in significantly increased expression of oxidized PTEN, whereas no oxidized PTEN was detected after treating with MCD or HFFA for 30 min. These data suggest that the level of oxidative stress and/or H 2 O 2 levels achieved during macronutrient perturbation is less that produced by treating with 1 mM H 2 O 2 . FIGURE 8. Effect of siRNA against PTEN on AML-12 hepatocytes maintained in control culture medium. A, relative quantitative mRNA expression by real time PCR for PTEN with untransfected control arbitrarily being normalized to 1 shows significant knockdown of PTEN mRNA in a dose-dependent manner by siRNA against PTEN. No significant difference is observed when cells are transfected by nonsilencing "scrambled" control siRNA. B, cells exposed to siRNA against PTEN show increased pAkt expression in a dose-dependent manner, whereas cells treated with untransfected control medium remain unaffected. Densitometry analysis (n ϭ 3) for each condition is expressed in arbitrary units (A.U.) in a graph. C, cells exposed to siRNA against PTEN show significantly decreased PTEN protein expression in a dose-dependent manner. ***, p Ͻ 0.001; **, p Ͻ 0.01; *, p Ͻ 0.05. ing is important because it suggests that nutrient deprivation and excess have identical effects in producing ROS and thus initiating a signal cascade leading to steatosis. A growing literature involving ROS signaling of various cellular events is confused by the failure to discriminate between injury induced by oxidative stress (a toxic effect) and the signals resulting from it that serve to protect the cell (ROS-stimulated response). These two effects may be one and the same although differing in the degree of ROS involved (29,30). Recent work has shown that mitochondrial ROS are required for physiologic hypothalamic sensing of glucose (mediating brain-dependent insulin release) and fat (mediating postprandial satiety) (31,32). In cardiovascular biology, it has been shown that the effect of hyperglycemia on inhibiting endothelial nitric-oxide synthase is fully mediated by mitochondrial ROS signaling through the PI 3-kinase/Akt pathway (33) and that leptin induces endothelial production of monocyte chemoattractant protein-1 by ROS-dependent activation of other kinase pathways (34), providing important insight into the vascular complications of type-2 diabetes. In monocytes in culture, serum withdrawal induces immediate ROS production and ROS-dependent activation of the PI 3-kinase/Akt signal pathway (35). Our work suggests that a change in macronutrient input to hepatocytes produces increased mitochondrial ROS production, which in turn acts as a signal to regulate the PI 3-kinase/Akt pathway cell survival pathway. In this way, mitochondria may act as a nutrient sensor for the hepatocyte that is independent of insulin signaling. We must ask how nutrient deprivation and excess act to produce the same ROS signal that results in steatosis. One hypothesis might be that the increase in superoxide generation with either perturbation involves increased fatty acid oxidation. Indeed, a recent study showed that feeding mice the MCD diet increases hepatic fatty acid oxidation possibly via a decrease in stearoyl-coenzyme A desaturase-1 expression (10). This mechanism seems unlikely in the setting of the present study, since increased superoxide production was observed to occur within 30 min of exposure to the MCD diet. A pronounced increase in fatty acid oxidation in the HFFA construct seems more likely, since it would represent the oxidation of the fuel made available in the medium. This would mirror the observation that mice fed conjugated linoleic acid for 12 weeks showed an increase in both fatty acid oxidation and steatosis (36). However, finding that there was no increase in the magnitude of signal transduction above an FFA concentration of ϳ0.25 mmol/liter speaks against this available fuel theory. Clearly, there is need for more work to determine the mechanism by which macronutrient perturbation excites the production of mitochondrial ROS.
The accumulation of superoxide and the potential for membrane damage resulting from it are limited by its rapid conversion to H 2 O 2 by the action of mitochondrial superoxide dismutase. In contrast to superoxide, H 2 O 2 is readily permeable across the mitochondrial membrane into the cytoplasm (37). Our data suggest that the signal emanating from the mitochondria involved in signaling the PI 3-kinase/Akt pathway is H 2 O 2 . Other data support the idea that ROS-mediated signaling is involved in hepatocyte function. Hepatocyte resistance to oxidative injury is mediated by ROS-dependent activation of protein kinase C, which inhibits proapoptotic signaling (38). More-over, constitutive overexpression of CYP2E1, which is expected to cause increased ROS production, causes activation of extracellular signal-regulated kinase 1/2 with resultant protection from oxidative injury (39). Furthermore, H 2 O 2 treatment of primary cultures of rat hepatocytes activates extracellular signal-regulated kinase in a dose-dependent manner and provides protection against ROS-mediated injury/cell death (40). We show that exogenous H 2 O 2 can increase p85 expression, inhibit PTEN, and increase pAkT expression. Together, these data suggest that ROS signaling may be intimately involved in signal transduction, leading to steatosis in experimental NASH. The mechanism of transduction of PI 3-kinase remains to be established, although protein kinase C and extracellular signal-regulated kinase seem likely candidates as intermediates in the process given the aforementioned studies.
Intracellular H 2 O 2 is potentially toxic and is rapidly degraded by catalase and other peroxidases to prevent intoxication. The balance between its production and degradation provides an opportunity for H 2 O 2 to act as a signal molecule. Epidermal growth factor signal transduction in cancer cells involves generation of H 2 O 2 , which is involved in inhibition of proteintyrosine phosphatase activity (41). Incorporation of excess catalase into cancer cells by electroporation completely abolishes epidermal growth factor signal transduction, showing the critical importance of the balance between production and degradation in H 2 O 2 -mediated signaling (41). Of interest, ROS exposure down-regulates catalase expression in renal mesangial cells by a mechanism that is dependent upon PI 3-kinase/Akt signaling through phosphorylation of the FoxO1 transcription factor (42). Further work will be needed to show if this mechanism actually accentuates H 2 O 2 signaling in experimental steatosis. It is unlikely to have had an effect in these short term experiments. More likely is that the abrupt and large excess of H 2 O 2 production overwhelmed degradation, which led to increased signal.
A novel finding in the present study is that both nutrient deprivation and excess function to produce hepatocyte steatosis by the same mechanism involving ROS-dependent transduction and modulation of the PI 3-kinase pathway. Most of the effects of activating PI 3-kinase result from PIP 3 -dependent activation of Akt. PTEN is the natural inhibitor of Akt activation, acting by degrading PIP 3 . We found evidence that PTEN is functionally inactivated through phosphorylation during nutrient perturbation, which would provide amplification of the PI 3-kinase signal transduction. We also found that inhibiting the expression of PTEN in hepatocytes resulted in increased pAkt expression, providing evidence that PTEN provides a brake for PI 3-kinase signaling during normal cellular respiration. This is supported by in vivo data showing that liver-specific PTEN knock-out mice develop steatosis and liver cancer (26). These mice also have increased hepatic H 2 O 2 content, which suggests that interacting with PTEN provides a natural detoxification pathway for ROS. H 2 O 2 can inactivate PTEN through its reversible oxidation (21), which would consume H 2 O 2 in the process. These findings together suggest that PTEN oxidation may help to regulate cellular H 2 O 2 during normal respiration. Our data suggest that acute regulation of PTEN function during nutrient perturbation occurs mainly by maintaining phospho-rylation. As with the pathway for PI 3-kinase transduction, the mechanism by which PTEN is not effectively dephosphorylated in this model is unclear. p85 itself has been shown to regulate PTEN. Liver-specific p85 knock-out mice show increased hepatic PTEN expression and activity (43). Further in vitro work will be required to determine the mechanisms involved in maintaining PTEN in a relatively phosphorylated state during nutrient perturbation, and in vivo work will be required to determine if longer term dietary perturbation results in PTEN oxidation as well.
In summary, these findings support a novel pathway involving cellular metabolic stress and a patterned response leading to steatosis in this model of experimental NASH. In our conceptual model of how an abnormal diet stimulates the development of hepatic steatosis, abnormal dietary macronutrient (fuel) input or abrupt change in input to the liver results in increased generation of mitochondrial ROS, which are normally produced during aerobic energy metabolism. There is plentiful evidence that mitochondria can act as sensory organelles and send signals to the cell to adjust metabolism. In the case of a change in macronutrient input to the liver, the mitochondria "sense" the change by increasing ROS production. Our data show that an abrupt change in the ambient macronutrient composition of medium to which hepatocytes in culture are exposed results in a prompt increase in mitochondrial ROS production. This results in a signal being sent to adjust cellular metabolism to accommodate the stress. This signal is H 2 O 2 produced by the action of superoxide dismutase on the burst of ROS just produced.
The PI 3-kinase-dependent cell signaling pathway has been called a "survival" pathway, because its activation promotes cell growth, activates metabolic pathways involved in cytoprotection, activates the cell cycle and promotes proliferation, and inhibits apoptosis (44,45). H 2 O 2 acts to activate and modulate the PI 3-kinase/Akt cell survival pathway, which is responsible for the many adjustments in cell metabolism needed to accommodate the stress incurred by the change in macronutrient input. The acquisition of fat (steatosis) is but one of the adjustments made. This cell survival pathway is employed as a thrift mechanism to save up fuel for a time when it is less available. Hibernating animals activate it when gorging before the winter fast (46). We believe that it is important in malfunction when people eat too much and fail to fast or otherwise use the fuel stored. The crux of this work shows the potential central role of the mitochondrial ROS/PI 3-kinase/Akt nutrient-stimulated signaling pathway in the development of steatosis. These studies provide insight into the pathogenesis of human NAFLD/ NASH resulting from inappropriate nutrient intake.