Overexpression of Manganese Superoxide Dismutase Prevents Alcohol-induced Liver Injury in the Rat*

Mitochondria are thought to play a major role in he-patic oxidative stress associated with alcohol-induced liver injury. Thus, the hypothesis that delivery of the mitochondrial isoform of superoxide dismutase (Mn-SOD) via recombinant adenovirus would reduce alco-hol-induced liver injury was tested. Rats were given recombinant adenovirus containing Mn-SOD (Ad.SOD2) or (cid:1) -galactosidase (Ad.lacZ) and then fed alcohol enter-ally for 4 weeks. Mn-SOD expression and activity of Ad.SOD2 in liver mitochondria of infected animals was increased nearly 3-fold compared with Ad.lacZ-infected controls. Mitochondrial glutathione levels in Ad.lacZ-infected animals were decreased after 4 weeks of chronic ethanol, as expected, but were unchanged in Ad.SOD2-infected animals. Alanine aminotransferase was elevated significantly by ethanol, an effect that was prevented by Ad.SOD2. Moreover, pathology ( e.g. the sum of steatosis, inflammation, and necrosis) was elevated dramatically by ethanol in Ad.lacZ-treated rats. isolated by differential centrifugation as described above and added to a solution containing 50 m M K 2 HPO 4 , 0.1 m M Na 2 EDTA, 0.5 mg/ml cytochrome c , and 165 m M xanthine, and superoxide was generated by the addition of 0.004 units of xanthine oxidase. The reaction proceeded at room temperature for 10 min. The absorption of cytochrome c was measured at 550 nm, and SOD activity was calculated based on the millimolar extinction coefficient of 18.5. Western Blotting Mitochondria were isolated by differential centrifugation as described above. Mitochondrial protein (50 (cid:4) g) was sus- pended in Laemmli buffer, heated at 95 ° C for 5 min, and resolved by electrophoresis using 12% SDS-polyacrylamide gel. Samples were transferred to nitrocellulose and blotted with sheep anti-human Mn-SOD antibody (Oxis, Portland, OR) followed by horseradish peroxidase- conjugated anti-sheep IgG secondary antibody. (cid:3) -galactosidase was immunoblotted using a mouse anti- (cid:3) -galactosidase antibody (Chemi- con, Temecula, CA) followed by horseradish peroxidase-conjugated an-ti-mouse secondary antibody. Protein was visualized by autoradiogra- phy using ECL Western Detection Reagent (Amersham Pharmacia Biotech UK, Little Chalfont, UK). Catalase Activity Catalase activity in liver homogenate as described by Aebi (34) with some modifications. Briefly, homo- genate (10 (cid:4) g) was adjusted to a final volume of 50 (cid:4) l with phosphate buffer. The reaction was initiated by adding 3.0 ml of 12.5 m M H 2 O 2 in phosphate buffer, and the change in absorbance at 240 nm was measured at 25 ° C for 1 min. Based on a millimolar extinction coefficient for H 2 O 2 of 34.9, catalase activity was defined as (cid:4) mol of H

Alcoholic liver disease results from dose-and time-dependent exposure to alcohol (1), but precise mechanisms of pathology are still largely unknown. Endotoxin and Kupffer cells have been implicated in the mechanism of early alcoholinduced liver injury using the enteral feeding model of Tsukamoto-French (2). For example, endotoxin derived from the gut activates Kupffer cells in the liver (3). In support of this idea, gut sterilization with nonabsorbable antibiotics or inactivation of Kupffer cells by gadolinium chloride (GdCl 3 ) prevents alcohol-induced liver injury in this model (4,5). Furthermore, Kupffer cells, which release effectors and cytokines, are a major source of TNF␣ 1 in the liver (6). Indeed, TNF␣ messenger RNA in liver increased after 4 weeks of treatment with ethanol (7). Moreover, early alcohol-induced liver injury was attenuated by anti-TNF␣ antibodies and largely prevented in TNF receptor 1 knockout mice (8,9). Thus, it is clear that TNF␣ plays a critical role in early alcohol-induced liver injury.
Reactive oxygen species generated during chronic alcohol exposure may also be a major factor in liver damage (10,11). ␣-Hydroxyethyl free radicals were increased as a result of alcohol treatment (12) and were diminished by destruction of Kupffer cells with GdCl 3 (13). Furthermore, both production of TNF␣ and ␣-hydroxyethyl free radicals were decreased in the livers of NADPH oxidase knockout mice, effects that correlated with a reduction in pathology (14). Moreover, delivery of cytosolic superoxide dismutase (Cu/Zn-SOD) by adenovirus reduced early alcohol-induced liver injury as well as blunted NFB activation and TNF␣ production (15). However, it is not clear whether oxidants act as direct toxicants to hepatocytes or as signals to produce TNF␣ or other cytokines by Kupffer cells.
Under normal conditions, the mitochondrial isoform of superoxide dismutase balances the production of excess superoxide from electron transport. Hydrogen peroxide, a product of the superoxide dismutase reaction, is further reduced to water by glutathione (GSH) peroxidase using mitochondrial GSH (16). Mitochondrial glutathione is diminished in the livers of animals exposed to chronic ethanol, supporting the hypothesis that mitochondrial oxidant production plays a role in early alcohol-induced liver injury (17,18). Moreover, treatment with the glutathione precursor OTC (L-2-oxothiazolidine-4-carboxylic acid) blunted alcohol-induced liver injury in the enteral model (19). Thus, it is hypothesized that mitochondria may be a critical source of oxidants because of chronic ethanol consumption.
The role of Mn-SOD and mitochondrial oxidative stress in alcohol-induced liver injury is controversial. A recent report showed that homozygous mutations in the SOD2 gene that may lead to an increase in mitochondrial localization of SOD is a risk for severe alcoholic liver disease in humans (20). However, other reports have demonstrated decreases in Mn-SOD expression and activity in liver because of ethanol in rats (21)(22)(23). Thus, the hypothesis that overexpression of mitochondrial Mn-SOD would prevent alcohol-induced liver injury was tested here by delivering human mitochondrial Mn-SOD via recombinant adenovirus (Ad.SOD2). Indeed, overexpression of Mn-SOD reduces liver injury induced by alcohol in the enteral feeding model, suggesting that mitochondrial redox state is important in ethanol-induced liver injury.

Animals and Diets-Male
Wistar rats weighing 280 -310 g were housed in an Association for the Accreditation and Assessment for Laboratory Animal Care-approved facility on a 12-hour light/dark cycle under institutional guidelines for the humane treatment of laboratory animals. Intragastric cannulas were inserted as described by Tsukamoto et al. (2). Cannulas were tunneled subcutaneously to the dorsal aspect of the neck and connected to infusion pumps by means of a spring-tether device and swivel, allowing rats to move freely in metabolic cages. Rats received a liquid diet described by Thompson and Reitz (24), which is composed of corn oil as fat (34% of total calories), protein (23%), and carbohydrate (43%), plus minerals and vitamins, and supplemented with lipotropes as described by Morimoto et al. (25). For the ethanol diet, maltose-dextrin was replaced isocalorically with ethanol. In each group, rats received either high-fat control diet or ethanolcontaining diet by continuous infusion through an intragastric tube. The daily amount of ethanol given was gradually increased to 12.0 g/kg/day in the first week. Values were then increased progressively up to 14.5 g/kg/day. Adenoviral Synthesis and Preparation-Recombinant adenoviral vectors containing the transgene for either ␤-galactosidase (Ad.lacZ) or Mn-SOD (Ad.SOD2) were prepared as described elsewhere (26,27). Briefly, the plasmid shuttle vector pAd5-CMV-lacZ was constructed using standard cloning protocols. The adenoviral shuttle plasmids were transfected into the permissive HEK 293 host cell line to generate recombinant Ad.lacZ adenovirus. The Ad.SOD2 viral seed stock was a kind gift from Dr. John Engelhardt (University of Iowa). Virus isolates were plaque-purified and propagated in HEK 293 cells, isolated, concentrated, and titered by plaque assay to stock titers of greater than 1ϫ10 11 plaque-forming units. Rats were divided randomly into two groups and injected with Ad.lacZ or Ad.SOD2 at a concentration of 1ϫ10 9 plaque-forming units in 1 ml of lactated Ringer's solution via the penile vein 3 days before liquid diet feeding was initiated.
Urine Collection and Ethanol Assay-Ethanol concentrations in urine were measured daily. Rats were housed in metabolic cages, and urine was collected over 24 h in bottles with mineral oil to prevent evaporation. Samples were stored at Ϫ20°C for later analysis of ethanol. Ethanol concentration was determined by measuring absorbance at SCHEME 1. Working hypothesis. (a) Kupffer cells are activated by gut-derived endotoxin and release inflammatory cytokines such as TNF␣. TNF␣ causes an increase in neutrophil infiltration in liver most likely via up-regulation of important adhesion molecules such as intercellular adhesion molecule-1 on endothelial cells. TNF␣ also stimulates mitochondrial oxidant production in hepatocytes through an increase in respiration as well as an inhibition of electron transport. It is proposed that mitochondrial oxidant production is a critical point leading to injury and cell death. (b) Overexpression of SOD in mitochondria prevents oxidative stress and protects hepatocytes from ethanol-induced injury. 366 nm resulting from the reduction of NAD ϩ to NADH by alcohol dehydrogenase (28). Clinical Chemistry-Blood was collected via the abdominal aorta at sacrifice. Serum was stored at Ϫ20°C in microtubes until alanine aminotransferase (ALT) was assayed using standard enzymatic procedures (28). For measurement of plasma endotoxin, blood was taken from the portal vein and the abdominal aorta in pyrogen-free heparinized syringes during laparotomy after 4 weeks of treatment with ethanol. Blood kept in pyrogen-free glass tubes was centrifuged at 1,200 rpm for 10 min, and plasma was stored at Ϫ80°C also in pyrogen-free glass tubes until measurement of endotoxin with a Limulus amebocyte lysate test kit (Kinetic QCL, BioWhittaker, Walkersville, MD). The levels of total glutathione (GSH ϩ GSSG) and oxidized glutathione (GSSG) in whole liver and in mitochondria were measured as described elsewhere (29).
Measurement of ␣-Hydroxyethyl Free Radical Adducts in Bile-Rats were anesthetized with pentobarbital (75 mg/kg) and a 24-gauge catheter was inserted into the left femoral vein for injection of the spin trapping agent ␣-(4-pyridyl-1-oxide)-N-tert-butylnitrone (1 g/kg; Sigma). A laparotomy was performed, PE-20 tubing was inserted into the common bile duct, and the spin trap dissolved in normal saline was injected through the intravenous catheter. Bile was collected for 3 h in an Eppendorf tube containing 35 l of 5 mM deferoxamine mesylate (Sigma) to prevent ex vivo radical formation. During this procedure, 1 ml of normal saline was injected through the catheter every 30 min to compensate for loss of fluids. Samples were frozen immediately on dry ice and stored at Ϫ80°C until ESR analysis.
The ESR spectra of radical adducts was detected using a Varian E-109 spectrometer in an E-238 TM110 microwave cavity with instrument setting of 20 milliwatt microwave power, 0.5 G modulation amplitude, 0.66 s of conversion time, 0.33 time constant, 80 G scan width, and 9.785 gigahertz microwave frequency. Spectra were recorded on an IBM-compatible computer interfaced to the spectrometer, and hyperfine coupling constants were determined with a spectral simulation program (30).
Pathologic Evaluation-Autopsies were performed after 4 weeks of treatment with ethanol. Livers were fixed in formalin, embedded in paraffin, and stained with hematoxylin and eosin for the assessment of pathology. Liver pathology was scored as described by Nanji et al. (31) as follows: steatosis (the percentage of hepatocytes containing fat), Ͻ25%, 1ϩ; Ͻ50%, 2ϩ; Ͻ75%, 3ϩ; Ͼ75%, 4ϩ; inflammation and necrosis, 1 focus per low-power field, 1ϩ; 2 or more foci, 2ϩ. Pathology was scored in a blinded manner by one of authors.
Apoptosis and Infiltrating Neutrophils-Apoptosis was expressed as % of apoptotic hepatocytes showing condensation of chromatin and nuclear fragmentation/2,000 hepatocytes in high power fields (ϫ400) selected randomly (32). Caspase-3 activity was also evaluated using the chromogenic substrate DEVD-pNA (BioSource, CA) as recommended by the manufacturer. The degree of infiltrating neutrophils was expressed per 100 hepatocytes, because fat accumulation causes ballooning of hepatocytes, making sinusoidal spaces narrow, which affects the numbers of hepatocytes and sinusoidal space in each field. Values were determined by counting polymorphonuclear cells in 5 high power fields (ϫ400)/slide followed by counting the number of hepatocytes in each field. Mean values were used for statistical analysis.
Isolation of Mitochondria-Whole liver tissue was homogenized in 10 ml of buffer (40 mM Tris, 140 mM NaCl, and a protease inhibitor mixture including aprotinin, leupeptin, phenylmethylsulfonyl fluoride, and dithiothreitol). Supernatant was collected after centrifugation at 900 ϫ g for 7 min and was centrifuged at 17,000 ϫ g for 10 min to pellet mitochondria. The mitochondrial pellet was washed and resuspended in buffer, and protein concentration was measured using the Bio-Rad Protein Assay (Bio-Rad).
Hepatocyte Isolation, Infection, and Culture-Rat hepatocytes from normal, untreated rats were isolated by collagenase perfusion and Percoll centrifugation previously as described (11) and cultured at 1 ϫ 10 6 cells/well in RPMI 1640 medium containing 10% heat-inactivated fetal bovine serum and antibiotics. After 4 h of culture, hepatocytes were infected with either Ad.lacZ (100 viral infectious units/cell) or Ad.SOD2 and were cultured for 18 h to allow transgene expression to occur. Cells were then incubated in ethanol (2-200 mM) or saline for 12 h.
Measurement of SOD Activity-SOD activity in freshly isolated mitochondria was measured by the reduction of ferricytochrome c with modification of a method described elsewhere (33). Mitochondria were Ad.lacZ and Ad.SOD2. Blood samples were collected from the abdominal aorta at sacrifice after a 4-week feeding of high-fat control diet (Con) or diet containing ethanol (EtOH) as described under "Materials and Methods." Data are presented as means Ϯ S.E. (n ϭ 6). #, p Ͻ 0.05 as compared with rats fed the high-fat control diet; *, p Ͻ 0.05 as compared with Ad.lacZ-infected rats given ethanol, using ANOVA and Student-Newman-Keuls post-hoc test.
isolated by differential centrifugation as described above and added to a solution containing 50 mM K 2 HPO 4 , 0.1 mM Na 2 EDTA, 0.5 mg/ml cytochrome c, and 165 mM xanthine, and superoxide was generated by the addition of 0.004 units of xanthine oxidase. The reaction proceeded at room temperature for 10 min. The absorption of cytochrome c was measured at 550 nm, and SOD activity was calculated based on the millimolar extinction coefficient of 18.5.
Western Blotting-Mitochondria were isolated by differential centrifugation as described above. Mitochondrial protein (50 g) was suspended in Laemmli buffer, heated at 95°C for 5 min, and resolved by electrophoresis using 12% SDS-polyacrylamide gel. Samples were transferred to nitrocellulose and blotted with sheep anti-human Mn-SOD antibody (Oxis, Portland, OR) followed by horseradish peroxidaseconjugated anti-sheep IgG secondary antibody. ␤-galactosidase was immunoblotted using a mouse anti-␤-galactosidase antibody (Chemicon, Temecula, CA) followed by horseradish peroxidase-conjugated anti-mouse secondary antibody. Protein was visualized by autoradiography using ECL Western Detection Reagent (Amersham Pharmacia Biotech UK, Little Chalfont, UK).
Catalase Activity-Catalase activity was measured in liver homogenate as described by Aebi (34) with some modifications. Briefly, homogenate (10 g) was adjusted to a final volume of 50 l with phosphate buffer. The reaction was initiated by adding 3.0 ml of 12.5 mM H 2 O 2 in phosphate buffer, and the change in absorbance at 240 nm was measured at 25°C for 1 min. Based on a millimolar extinction coefficient for H 2 O 2 of 34.9, catalase activity was defined as mol of H 2 O 2 consumed/ min/mg of protein.
Glutathione Peroxidase Activity-Glutathione peroxidase activity was determined by adding 10 l of liver homogenate to 850 l of buffer containing 0.1 mM NaPO 4 , 4 mM reduced glutathione, 0.1 mM NADPH, and 2 units of glutathione reductase (35). After 30 s, 10 l of 1.2 mM t-butyl-hydroperoxide was added to initiate the reaction. The rate of decrease in absorption of NADPH at 340 nm, was measured and the amount of NADPH consumed was calculated using a millimolar extinction coefficient of 6.22.
RNase Protection Assay for TNF␣-Total RNA was isolated from whole liver using RNA STAT 60 (Tel-Test, Friendswood, TX). TNF␣ and the housekeeping gene GAPDH were detected by RNase protection assay using a mouse cytokine RNA probe template set (rCK-1, PharMingen, San Diego, CA). Riboprobes were synthesized in the presence of [ 32 P]UTP to yield labeled antisense RNA probes, and RNase protection assays were performed on 20 g of RNA using a RiboQuant TM multiprobe RNase Protection Assay Kit (PharMingen). Protected samples were separated on 5% acrylamide-bisacrylamide (19:1) urea gels, dried, and exposed to X-ray film.
Statistics-Data are presented as means Ϯ S.E. Results were compared using analysis of variance (ANOVA) followed by Student-Neuman-Keuls post-hoc test as appropriate. For comparison of pathology scores, the Mann-Whitney rank sum test was used. A p value of Ͻ 0.05 was selected prior to the study as the level of significance.

RESULTS
Body and Liver Weight-A tendency for weight loss during the 1-week recovery time after surgery was observed in all groups (data was not shown). Body weight increased steadily after feeding diets for 4 weeks, and there were no significant differences among the groups at sacrifice (Table I). However, liver to body weight ratios were increased 50 -60% by ethanol in both Ad.lacZ-and Ad.SOD2-infected animals as expected (Table I).
Urine Alcohol Concentration-In animals infected with Ad.lacZ and Ad.SOD2, alcohol levels fluctuated in a cyclic pattern with a periodicity of 5 to 6 days even though ethanol was infused continuously as reported previously (2) The cyclical pattern was described recently to be the result of fluctuations in body temperature and thyroid hormone levels which alter ethanol metabolism (36). There were no significant differences in mean urine alcohol concentrations between Ad.lacZ-and Ad.SOD2infected rats (Table I).
Western Blotting and Activity of SOD in Mitochondria-To test the hypothesis that adenoviral gene delivery of Mn-SOD resulted in localized expression, animals were infected with adenovirus (1 ϫ 10 9 plaque-forming units) containing the transgenes for human mitochondrial Mn-SOD (Ad.SOD2) or the bacterial reporter gene ␤-galactosidase (Ad.lacZ) as control. Livers were harvested at the end of ethanol treatment, subcellular compartments (i.e. mitochondria and cytosol) were isolated by differential centrifugation, and transgene expression was determined by Western analysis using antibodies against human Mn-SOD or ␤-galactosidase (Fig. 1A). ␤-Galactosidase was detected in the cytosol of Ad.lacZ-infected animals as expected. Moreover, recombinant Mn-SOD was primarily expressed in the mitochondrial compartment of the liver and was nearly 4-fold higher than the expression of endogenous Mn-SOD. A small amount of Mn-SOD was detected in the cytosol, but this is most likely because of either mitochondrial destruction during the isolation procedure or unprocessed protein that had not been transported into the mitochondria. Enzymatic activity was also evaluated to ensure that recombinant Mn-SOD had functional activity. The activity of SOD in whole liver extracts of Ad.SOD2-infected rats was increased about 2-fold compared with Ad.lacZ-infected rats (Fig. 1B).
Serum ALT and Histology-Serum ALT levels in rats fed high-fat control diet were around 20 units/liter after 4 weeks (Fig. 2). Ethanol treatment increased values in Ad.lacZ-transfected rats about 5-fold but caused only a 2-fold increase in Ad.SOD2-infected animals.
Figs. 3 and 4 show representative photomicrographs and pathology scores of livers after 4 weeks of high-fat control diet with or without ethanol. There were no pathological changes in Ad.lacZ-or Ad.SOD2-infected rats fed high-fat control diet except for very mild steatosis (Fig. 3A). In contrast, severe steatosis, inflammation, and necrosis were observed in Ad.lacZinfected rats after ethanol treatment, yielding a total pathology score of 7.8 Ϯ 0.2 (Fig. 4). High magnification photomicrographs of Ad.lacZ-infected animals fed ethanol are also shown to illustrate focal infiltrate consisting largely of neutrophils and lymphocytes, necrosis, and apoptosis (Fig. 3B). In Ad-.SOD2-infected animals, steatosis, inflammation, and necrosis were blunted by nearly 50%.
Glutathione, Glutathione Peroxidase, and Catalase Levels-Total glutathione levels of whole liver and mitochondria in Ad.lacZ-transfected rats were decreased significantly by ethanol treatment (Table II), confirming earlier work done with virus-free animals (17,18). In contrast, glutathione levels in Ad-SOD2-infected rats were maintained near control levels. The ratio of reduced glutathione to oxidized glutathione (GSH/ GSSG) was also decreased by ethanol in Ad.lacZ-infected animals but not in Ad.SOD2-infected rats (Table II).
Because ethanol inhibits GSH transport into mitochondria (37), one possible mechanism for protection against mitochondrial GSH depletion is that overexpression of Mn-SOD inhibits the effect of ethanol on GSH transport. To test this hypothesis, hepatocytes from untreated rats were isolated and infected with either Ad.lacZ (100 infectious units/cell) or Ad.SOD2 as described under "Materials and Methods." Cells were then incubated in the presence of ethanol (2-200 mM) or saline. After 12 h, cells were harvested, and mitochondrial glutathione levels were measured. Ethanol (200 mM) caused 78 Ϯ 12% depletion of mitochondrial GSH in Ad.lacZ-infected animals compared with hepatocytes incubated with vehicle alone (*, p Ͻ 0.05, four individual experiments). These findings are consistent with previously published work in which a high concentration of ethanol (Ͻ100 mM) caused a similar loss of mitochondrial GSH (38,39). In Ad.SOD2-infected rats, mitochondrial GSH levels were depleted to a similar extent by ethanol.
Glutathione peroxidase and catalase activity was measured in whole liver extracts from Ad.lacZ-and Ad.SOD2-infected rats fed either high-fat control diet or diet containing ethanol for 4 weeks (Table II). Glutathione peroxidase activity was similar in all treatment groups. In contrast, catalase activity was increased significantly by ethanol; however, because of its high activity, this change probably has no physiological significance. Interestingly, overexpression of mitochondrial Mn-SOD had no effect on the increase in catalase due to ethanol. Importantly, these findings are consistent with other reports (41).
Radical Adducts in Bile-Free radical adducts were detected by electron spin resonance spectroscopy in the bile of rats (Figs. 5 and 6). Ethanol treatment in Ad.lacZ-infected animals for 4 weeks resulted in a nearly 8-fold increase in radical adduct intensity. This increase was blunted by about 65% in animals infected with Ad.SOD2. Computer simulation identified the ␣-hydroxyethyl radical (␣ N ϭ 15.69 G and ␣ ␤ H ϭ 2.72 G) similar to the adduct described previously (12). Using 13 C-labeled ethanol, it was determined that nearly 60% of the total adduct was hydroxyethyl. However, lipid-derived radical adducts could not be distinguished from ␣-hydroxyethyl radical adducts under these conditions. mRNA level of TNF␣-Because TNF␣ is critically involved in early alcohol-induced liver injury (9) and it stimulates mitochondrial oxidative stress (42)(43)(44), TNF␣ and interleukin-1 mRNA levels in animals treated with ethanol were evaluated by RNase protection assay (Fig. 7). TNF␣ mRNA was increased about 3-fold in Ad.lacZ-infected rats after 4 weeks of ethanol diet, compared with control animals that received high-fat control diet. Treatment with Ad.SOD2 did not significantly alter ethanol-induced increases in TNF␣ mRNA levels.
Neutrophil Infiltration-Ethanol causes a significant increase of inflammatory cell influx in liver. Indeed, it was demonstrated above that ethanol caused a significant increase in inflammation, which was blunted by ϳ50% in Ad.SOD-infected animals. However, neutrophil infiltration was increased more than 2-fold by ethanol and was not affected by overexpression  6. Effect of Ad.SOD2 on ethanol-induced ␣-hydroxyethyl radical adduct formation. After 4 weeks of feeding high-fat control diet (Con) or ethanol-containing diet (EtOH) in rats infected with Ad-.lacZ and Ad.SOD2, free radical adducts were measured as described in Fig. 5. The amplitude of the first peak on the ESR radical signal in each spectrum was used for comparison of intensity. Data are presented as means Ϯ S.E. #, p Ͻ 0.05 compared with rats fed high-fat control diet; *, p Ͻ 0.05 as compared with Ad-lacZ-infected rats given ethanol, using ANOVA and Student-Newman-Keuls post-hoc test.

TABLE II
Effect of Ad.SOD2 and ethanol on glutathione in whole liver and mitochondria Glutathione, glutathione peroxidase, and catalase activities were assayed using fresh liver tissue after continuous feeding of high-fat control diet with or without ethanol for 4 weeks as described under "Materials and Methods." Data are presented as means Ϯ S.E. of Mn-SOD (Fig. 8A). To verify this finding, myeloperoxidase activity, an enzyme expressed predominantly in neutrophils, was determined (Fig. 8B). Indeed, ethanol caused a significant and similar increase in myeloperoxidase activity in both Ad.lacZ and Ad.SOD2-infected animals, confirming the histological data.
Ethanol-induced Apoptosis-Apoptosis was identified from condensed and fragmented nuclei in hematoxylin and eosinstained sections. Apoptosis was increased about 4-fold after 4 weeks of ethanol exposure in Ad.lacZ-infected liver compared with tissue from high-fat control animals (Fig. 9A). Delivery of Ad.SOD2 blunted ethanol-induced apoptosis by more than 60%. In addition, caspase-3 activity was measured in liver extracts from Ad.lacZ-and Ad.SOD2-infected animals after 3 weeks of control or ethanol-containing diet (Fig. 9B). Caspase activity determined by cleavage of the chromogenic substrate DEVD-pNA was increased nearly 50% because of ethanol in Ad.lacZ-treated animals; however, this effect was largely blunted in Ad.SOD2-infected animals, suggesting that Mn-SOD overexpression indeed protects against ethanol-induced apoptosis in hepatocytes.

DISCUSSION
Delivery of Mitochondrial SOD to the Liver-In normal liver, mitochondria efficiently reduce oxidants under normal conditions predominantly via antioxidant mechanisms including glutathione and superoxide dismutase (45). Recently, it was shown that deletion of mitochondrial SOD by about 50% results in a functional decline of oxidative phosphorylation, an increase in oxidative stress, an increase in rates of apoptosis in an age-dependent manner (46), and depletion of mitochondrial GSH (47), suggesting that Mn-SOD is important for balance of mitochondrial redox state. Moreover, these data suggest that minor changes in Mn-SOD may have a significant impact on the antioxidant status of mitochondria and support the hypothesis that overexpression of Mn-SOD may be protective against mitochondrial oxidative stress. On the other hand, a recent report showed that homozygous mutations in the SOD2 gene that may lead to an increase in mitochondrial localization of SOD are a risk for severe alcoholic liver disease in humans (20), which is in contrast to the hypothesis proposed here. Whether or not these mutations are causal is very difficult to conclude because little biochemical evidence has been presented. However, these correlation studies indeed introduce important questions about the relationship of Mn-SOD levels to early ethanol-induced liver disease. Despite a number of other studies (17,48,49), the roles of Mn-SOD and mitochondrial oxidative stress in alcoholic liver disease are still unclear. Here, delivery of adenovirus containing the transgene for human Mn-SOD resulted in a significant increase in SOD almost exclusively in the mitochondrial compartment of liver (Fig. 1A). Importantly, overexpression of Mn-SOD by about 3-fold (Fig.  1B) almost completely blocked ethanol-induced liver injury, suggesting that mitochondrial oxidative stress be involved in the mechanism of injury.
Mitochondrial Oxidative Stress Plays a Role in Alcohol-induced Liver Injury-Ethanol is known to deplete mitochondrial FIG. 7. Effect of Ad.SOD2 and ethanol on expression of TNF␣ in liver. Total mRNA was prepared from livers of Ad.lacZ-and Ad.SOD2-infected animals 4 weeks after feeding high-fat control (C) or ethanol-containing diet (EtOH). TNF␣ mRNA was measured by RNase protection assay as described under "Materials and Methods." mRNA levels of TNF␣ and interleukin-1 (IL-1␣) are expressed relative to the housekeeping genes glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and L32. glutathione levels (38,17,18,50) consistent with the hypothesis that ethanol causes oxidative stress in the mitochondria. Ethanol inhibits the mitochondrial glutathione transporter, which shuttles GSH into the mitochondria from the cytosol where it is generated (50,51). There are a number of possible mechanisms by which ethanol generates oxidants in mitochondria. It has been shown that ethanol-induced TNF␣ increases mitochondrial oxidant production by inhibiting complex II in the electron transport chain leading to the reduction of oxygen (42,43,52). Moreover, ethanol, through activation of Kupffer cells, increases respiration via prostaglandin E 2 , which may exacerbate oxidant generation in mitochondria (53,54). Indeed, ethanol feeding decreased mitochondrial glutathione in this study (Table II). Depletion of mitochondrial GSH in vivo could occur by elimination of mitochondrial oxidants through glutathione peroxidase at the expense of GSH or by inhibition of glutathione transport from the cytosol. When mitochondrial superoxide dismutase (Mn-SOD) was elevated nearly 3-fold by gene delivery using adenovirus, chronic ethanol did not reduce glutathione levels (Table II), suggesting that overexpression of Mn-SOD prevents oxidant production in vivo. It was also demonstrated here that overexpression of Mn-SOD in isolated hepatocytes did not influence depletion of mitochondrial GSH (Table II). Why Mn-SOD protected against the depletion of GSH in vivo but not in isolated hepatocytes is not completely understood; however, this apparent paradox is most likely because of the differences in ethanol concentrations between the in vivo and in vitro experiment and/or factors that promote mitochondrial oxidant production by increasing respiration or uncoupling electron transport (e.g. TNF␣ and prostaglandin E 2 ). Importantly, these data may also suggest that the defect in GSH transport during chronic ethanol exposure is a consequence of mitochondrial oxidative stress. Mitochondrial oxidative stress would lead to both GSH depletion and impaired transport and then further GSH depletion and oxidative stress. Thus, SOD2 would prevent mitochondrial GSH depletion in vivo by preventing mitochondrial oxidative stress under these conditions.
Direct evidence that Mn-SOD blunted oxidative stress comes from the fact that the increase in ESR-detectable free radical adducts because of ethanol were also blunted significantly by overexpression of Mn-SOD (Figs. 5 and 6), which also implicates mitochondria as a critical source of oxidants in early ethanol-induced liver injury. It is known that oxidant generation from mitochondria is an important factor in triggering apoptosis (55). Indeed, overexpression of Mn-SOD blunted both radical adduct formation as well as the increase in apoptosis because of ethanol (Fig. 9). These data strongly support the hypothesis that mitochondrial redox status is an important factor in the pathogenesis of early alcoholic liver disease.
An interesting point is that Mn-SOD converts superoxide to H 2 O 2 , which is eliminated primarily by GSH and catalase. Because mitochondrial GSH is not depleted in Ad.SOD2-infected animals after ethanol exposure, it is possible that catalase is important in reducing H 2 O 2 levels under these conditions. Cederbaum and colleagues (56) recently showed that overexpression of catalase in either mitochondria or cytosol was equally protective against oxidants generated in mitochondria by rotenone or antimycin A (56). These data suggest that H 2 O 2 readily diffuses into the cytosol, most likely down a concentration gradient to catalase, which is extremely abundant and highly active in hepatocytes. Moreover, others have shown (40, 41), and it is also demonstrated here, that catalase activity is increased because of chronic ethanol (Table II), supporting the conclusion that catalase levels are sufficient to metabolize H 2 O 2 in liver.
Mitochondrial Oxidative Stress Occurs Subsequent to TNF␣ Production-The role of TNF␣ in early alcohol-induced liver injury is well established (57). Recently, it was shown that mice deficient in TNF␣ receptors were resistant to ethanol-induced liver injury (9). Because Kupffer cells are the major source of TNF␣ in liver, it is hypothesized that Kupffer cells are activated by gut-derived endotoxin to produce TNF␣ as well as other inflammatory cytokines and mediators such as prostag- FIG. 9. Effect of Ad.SOD2 on apoptosis. After 4 weeks of feeding high-fat control diet (Con) or ethanol-containing diet (EtOH) in rats infected with Ad.lacZ and Ad.SOD2, apoptosis was evaluated as described under "Materials and Methods." 2000 hepatocytes were counted, and the percentage of hepatocytes exhibiting apoptosis (i.e. nuclear fragmentation and condensation) was determined. B, whole liver extracts were evaluated for caspase-3 activity as described under "Materials and Methods." Data are expressed as units/mg protein and presented as means Ϯ S.E. #, p Ͻ 0.05 compared with rats fed high-fat control diet; *, p Ͻ 0.05 compared with Ad.lacZ-infected rats given ethanol using ANOVA and Student-Newman-Keuls post-hoc test.
landins. TNF␣ induces the expression of adhesion molecules, e.g. ICAM-1 (intercellular adhesion molecule-1), which is involved in neutrophil adhesion in the liver (58). In this study, ethanol increased mRNA levels of TNF␣ in both Ad.lacZ-and Ad.SOD2-infected animals to the same extent (Fig. 7). This most likely explains the same degree of neutrophil infiltration observed in livers from these animals (Fig. 8) and is consistent with the hypothesis that Mn-SOD acts subsequent to Kupffer cell production of TNF␣ and recruitment of neutrophils.
As mentioned above, TNF␣ is also important in the production of radicals in hepatocyte mitochondria (42,43,59). Because TNF␣ levels and neutrophil infiltration are similar in Ad.lacZ-and Ad.SOD2-infected animals, but ethanol-induced free radical adduct formation and depletion of glutathione is minimized by Ad.SOD2 in vivo, it is hypothesized that mitochondrial oxidative stress occurs as a result of TNF␣ signaling. Moreover, it is hypothesized that these downstream events of mitochondrial oxidative stress are critical to alcohol-induced parenchymal cell death. This hypothesis could explain why apoptosis induced by ethanol as a result of TNF␣ was blunted in Ad.SOD2-infected animals, because mitochondria are involved in the mechanism of this process (55). The injury from ethanol, however, is due largely to necrosis. This is reflected in pathology in which most of the damage in the liver is primarily necrotic cell death with apoptosis contributing only slightly (ϳ 1% of the cells). Interestingly, Mn-SOD overexpression blunted both mechanisms of cell death, suggesting indeed that mitochondrial oxidative stress is critically involved.
In conclusion, mitochondria have a pivotal role in the mechanism in alcohol-induced liver injury. Hepatocyte mitochondria most likely generate oxidants in response to Kupffer cell TNF␣ and prostaglandin production, leading to cell death. Overexpression of Mn-SOD by gene delivery is effective against hepatocyte mitochondrial oxidative stress, and maintenance of mitochondrial GSH and is protective against alcohol-induced liver injury. Thus, preventing the accumulation of oxidants in mitochondria may be an important strategy in reducing ethanol-induced liver injury.