Basal reactive oxygen species determine the susceptibility to apoptosis in cirrhotic hepatocytes

Free Radical Biology & Medicine 41 (2006) 1645–1654www.elsevier.com/locate/freeradbiomed

Original Contribution

Basal reactive oxygen species determine the susceptibility toapoptosis in cirrhotic hepatocytes

Jay Raval a, Suzanne Lyman a, Takashi Nitta b, Dagmara Mohuczy b, John J. Lemasters c,Jae-Sung Kim b, Kevin E. Behrns b,⁎

a Department of Surgery, University of North Carolina, Chapel Hill, NC 27599, USAb Department of Surgery, Division of General Surgery, University of Florida, P.O. Box 100286, 1600 SW Archer Road, Gainesville, FL 32610, USA

c Department of Pharmaceutical Sciences and Biochemistry & Molecular Biology, Medical University of South Carolina, Charleston, SC 29425, USA

Received 27 August 2005; revised 10 July 2006; accepted 24 July 2006Available online 9 September 2006

Abstract

Hepatocytes from cirrhotic murine livers exhibit increased basal ROS activity and resistance to TGFβ-induced apoptosis, yet when ROS levelsare decreased by antioxidant pretreatment, these cells recover susceptibility to apoptotic stimuli. To further study these redox events, hepatocytesfrom cirrhotic murine livers were pretreated with various antioxidants prior to TGFβ treatment and the ROS activity, apoptotic response, andmitochondrial ROS generation were assessed. In addition, normal hepatocytes were treated with low-dose H2O2 and ROS and apoptotic responsesdetermined. Treatment of cirrhotic hepatocytes with various antioxidants decreased basal ROS and rendered them susceptible to apoptosis.Examination of normal hepatocytes by confocal microscopy demonstrated colocalization of ROS activity and respiring mitochondria. Basalassessment of cirrhotic hepatocytes showed nonfocal ROS activity that was abolished by antioxidants. After pretreatment with an adenovirusexpressing MnSOD, basal cirrhotic hepatocyte ROS were decreased and TGFβ-induced colocalization of ROS and mitochondrial respiration waspresent. Treatment of normal hepatocytes with H2O2 resulted in a sustained increase in ROS and resistance to TGFβ apoptosis that was reversedwhen these cells were pretreated with an antioxidant. In conclusion, cirrhotic hepatocytes have a nonfocal distribution of ROS. However, normaland cirrhotic hepatocytes exhibit mitochondrial localization of ROS that is necessary for apoptosis.© 2006 Elsevier Inc. All rights reserved.

Keywords: Reactive oxygen species (ROS); Hepatocytes; Apoptosis; Transforming growth factor beta (TGFβ); Mitochondria

Introduction

Transforming growth factor beta (TGFβ) induces apoptosisin normal murine hepatocytes through an apoptotic pathwaythat requires reactive oxygen species (ROS) generation, themitochondrial permeability transition (MPT) with cytochrome c

Abbreviations: AdCat, adenovirus expressing catalase; AdLuc, adenovirusexpressing luciferase; AdMnSOD, adenovirus expressing MnSOD; DMNQ,2,3-dimethoxy-1,4-naphthoquinone; H2-DCFDA, 2′,7′-dichlorofluorescein dia-cetate; DPPD, N,N-diphenyl-1,4-phenylenediamine; MTR, MitoTracker Red;MPT, mitochondrial permeability transition; NAC, N-acetylcysteine; PI,propidium iodide; ROS, reactive oxygen species; TGFβ, transforming growthfactor beta.⁎ Corresponding author. Fax: +1 352 338 9810.E-mail address: [email protected] (K.E. Behrns).

0891-5849/$ - see front matter © 2006 Elsevier Inc. All rights reserved.doi:10.1016/j.freeradbiomed.2006.07.023

release, and caspase activation [1]. The increase in ROS afterTGFβ is an early event that occurs within 90 min, lastsapproximately 3 h, and precedes the MPT and caspaseactivation [1,2]. Furthermore, inhibition of a ROS burstabolishes the apoptotic response and related intracellular events[1–3]. The source and mechanism of TGFβ-induced ROS havebeen attributed to the mitochondria, microsomes, and mem-brane-associated NADPH oxidase-like systems, yet the eva-nescent nature of ROS has made definitive source identificationdifficult [4,5]. In addition, TGFβ-induced down-regulation ofthe antioxidant, glutathione, further complicates the balance ofROS production versus scavenger activity [4,6]. Therefore,although ROS play an integral role in hepatocyte deathfollowing TGFβ administration, the necessity of ROS and theintracellular mechanisms through which ROS-mediated eventsoccur remain unclear.

mailto:[email protected]

http://dx.doi.org/10.1016/j.freeradbiomed.2006.07.023

1646 J. Raval et al. / Free Radical Biology & Medicine 41 (2006) 1645–1654

Despite the requirement of ROS generation for TGFβ-induced hepatocyte apoptosis in normal cells, increasedintracellular ROS in chronic inflammatory states do notinevitably induce parenchymal cell death, and, in fact, mayallow an adaptive state that protects against cell death [1].Previous work demonstrated that in a carbon tetrachloride(CCl4)-induced murine model of liver cirrhosis, hepatocytesisolated from this chronically inflamed liver have a greater than1.5-fold increase in ROS under basal conditions, fail to generatea ROS burst in response to TGFβ, resist apoptosis, yet uponpretreatment with the antioxidant, trolox, recovered respon-siveness to TGFβ-induced programmed cell death [1]. Theassociation between increased cellular ROS and resistance tocell death has been noted not only in chronic inflammatoryconditions, but also in neoplastic cells, and changes in ROS maybe associated with a malignant phenotype [7,8]. The source ofROS generation in chronic inflammation and neoplasia isunknown in these disease states in which chronic hypoxia mayinstigate free radical generation. Moreover, initiation of a singleoxygen-derived free radical pathway within a given cellularlocale can propagate rapidly and exponentially to multipleintertwined oxidant-generating pathways within various cellularcompartments, thereby rendering identification of the primaryROS-generating pathway difficult.

The cellular ROS state represents the balance of free radicalproduction and maintenance versus antioxidant scavengingactivity, and, therefore, the cellular expression of antioxidantenzymes such as the catalase, superoxide dismutases (SOD) 1and 2, and the glutathione peroxidase systems should beexamined in chronic inflammation and neoplasia. Previousstudies have documented decreased antioxidant gene expressionboth in noninflammatory and in inflammatory and neoplasticconditions [4]. Furthermore, other studies have suggested thatantioxidants such as SOD2 (MnSOD) may act as tumorsuppressors by controlling the cellular ROS state [8]. Becausethe expression of antioxidant enzymes at the time of ROSgeneration is often unknown, it is difficult to discern ifdecreased antioxidant expression is the cause of or resultsfrom increased ROS [4]. Furthermore, the variability inquantifying ROS at a given time and the relative specificity ofvarious exogenous antioxidants for oxidative pathways leadsone to question the importance of decreased antioxidants as aprimary cause for the increased ROS in disease state likeinflammation and neoplasia.

Because our previous data [1] suggested that an increase inROS resulted in TGFβ-induced apoptosis in normal hepato-cytes but prevented apoptosis in hepatocytes from a cirrhoticliver, we sought to examine these hepatocytes from normal andcirrhotic livers to further elucidate the importance of ROS inhepatocyte responsiveness to proapoptotic stimuli such asTGFβ. We found that in normal hepatocytes, TGFβ-inducedROS were initiated in the mitochondria and that inhibition ofROS precluded apoptosis. In addition, under basal conditions,cirrhotic hepatocytes demonstrated a diffuse increase in ROSwhich was abolished with trolox and in response to TGFβ theseantioxidant treated cells demonstrated an acute increase inmitochondrial derived ROS. Finally, normal hepatocytes treated

with low-dose H2O2 developed a sustained increase in ROSwhich inhibited TGFβ-induced apoptosis in these convertedhepatocytes.

Materials and methods

Materials

Adult, 8-week-old, male BALB/c mice were obtained fromHarlan Laboratories (Indianapolis, IN). Deferoxamine, 2,3-dimethoxy-1,4-naphthoquinone (DMNQ), glutathione, N,N-diphenyl-1,4-phenylenediamine (DPPD), N-acetylcysteine(NAC), and trolox were acquired from Sigma Chemical (St.Louis, MO). Catalase adenovirus (AdCat) was a kind giftfrom the University of Iowa Vector Core. SOD2 (AdMnSOD)and luciferase (AdLuc) adenoviruses were obtained from theUNC Chapel Hill Vector Core. Anti-caspase 3 rabbitpolyclonal antibody was purchased from Cell Signaling Tech-nology (Beverly, MA). 2′,7′-Dichlorofluorescein diacetate(H2-DCFDA) and MitoTracker Red (MTR) were obtainedfrom Molecular Probes (Eugene, OR).

Hepatocyte isolation and culture

Eight-week-old, male BALB/c mice weighing 20–25 g wereinjected in the peritoneum twice weekly with 2 ml/kg of 50%carbon tetrachloride (CCl4; cirrhotic mice) in sterile mineral oilor an equal volume of mineral oil alone for a total of 8 weeks[1]. All hepatocyte isolations were performed following a 7-dayrecovery. Hepatocytes were isolated through an abdominalincision that allowed cannulation of the inferior vena cava,clamping of the suprahepatic inferior vena cava, and transectionof the portal vein. The liver was perfused in retrograde fashionthrough the hepatic veins. Initially, the liver was perfused with asolution containing 25 mM Hepes, 115 mM NaCl, 5 mM KCl,1 mM KH2PO4, and 0.5 mM EGTA at pH 7.4 for a total volumeof 50–100 ml. The perfusate was changed to a solution withoutEGTA but containing 1 mM CaCl2 and 0.4 mg/ml type Icollagenase (0.6 mg/ml for cirrhotic livers, WorthingtonBiochemical, Lakewood, NJ) at pH 7.4 for a total volume of100–150 ml. The liver was excised, combed manually todisperse the cells, and subjected to differential centrifugation.Cell viability was determined using Trypan blue exclusion witha viability of >90% accepted for experiments. Hepatocyteswere plated in Waymouth’s medium supplemented with 10%fetal calf serum, 5 μg/ml insulin, and 100 nmol/L dexameth-asone. After 2 h, the medium was changed to hormonallydefined medium (HDM) containing insulin (5 μg/ml, SigmaChemical), transferrin (5 μg/ml), selenium (30 nM), and freefatty acid (1.52 μM palmitic acid, palmitoleic acid, stearic acid,oleic acid, linoleic acid, and linolenic acid; Sigma Chemical).

Adenovirus purification and infection

The replication-deficient adenoviruses expressing the lucif-erase (AdLuc), as a control, catalase (AdCat), or manganesesuperoxide dismutase (AdMnSOD) were prepared and stored as

1647J. Raval et al. / Free Radical Biology & Medicine 41 (2006) 1645–1654

described previously [1]. Twenty-four hours prior to treatmentcontrol and cirrhotic hepatocytes were infected at a multiplicityof infection of 100. Expression of the transgene was confirmedby immunoblot or luciferase assay and transfection efficiencywas routinely greater than 80%.

Morphologic assessment of apoptosis

Propidium iodide (PI) staining and fluorescence microscopywere used for morphologic assessment of apoptosis [1].Hepatocytes were fixed with methanol-acetic acid (3:1) for10 min at 4°C, washed twice, and stained with PI (0.33 mg/ml),and visualized under green excitation light using a IX-Olympusmicroscope (Olympus, Tokyo, Japan). The number of con-densed nuclei in five high-powered fields (X400) wasdetermined as a percentage of the total number of nuclei.

ROS determination

Cells were assayed in triplicate at a density of 1×105 perwell in a 12-well plate and after removal of media 1 ml H2-DCFDA solution (10 μM in DMSO) was added and incubatedat 37°C for 20 min. H2O2-treated cells served as positivecontrols. Fluorescence was determined in a Fluostar Spectro-fluorometer (BMG Labtech, Durham, NC) and read withwavelengths of excitation of 488 nm and emission of 525 nm,respectively. Cell lysates were the harvested for determinationof protein concentration using the Bradford assay.

Immunoblot analysis

To obtain whole cell extracts, cells were rinsed twice in PBSand lysed in buffer containing 0.05 M Tris, pH 7.3, 0.15 MNaCl, 1% NP-40, 0.5% deoxycholate, and the protease inhibitorcocktail (Sigma) for 10 min at 4°C. Samples were centrifuged at14,000 rpm to remove debris and the protein concentration wasdetermined by Bradford assay. Following SDS-polyacrylamidegel electrophoresis, samples were transferred to PVDFmembranes and blocked in 5% nonfat milk in TBS-T. After a1-h incubation with primary antibodies at concentrationsrecommended by the manufacturers, blots were washed for15 min in TBS-T, incubated with HRP-conjugated secondaryantibody (1:1000) for 30 min, washed for 15 min in TBS-T,visualized by enhanced chemiluminescence (Amersham Bios-ciences, Piscataway, NJ), and exposed to a Biomax-MS film(Perkin-Elmer, Boston, MA).

Caspase activity

Caspase-3 activity was assessed by a colorimetric enzymeassay (BD Biosciences, San Jose, CA) [1]. Each assay wasperformed in triplicate with 2×106 hepatocytes. Cell lysatesfrom treated or untreated hepatocytes were incubated with50 μl of 2X reaction buffer/DTT and 5 μl of 1 mM caspase-3 substrate (DEVD-pNA) was added. Following incubationat 37°C for 1 h absorbance was determined at 405 nm in aspectrophotometer.

Confocal microscopy

Localization of TGFβ-induced ROS production to mito-chondria was investigated in hepatocytes dual-labeled withH2-DCFDA (green fluorescence) to detect ROS and Mito-Tracker Red (MTR; Molecular Probes, Eugene, OR) to detectmitochondria [1]. MTR is electrophoretically taken up by themitochondria. After uptake, MTR becomes covalently boundto sulfhydryl groups of mitochondrial proteins and remains inthe mitochondria even if the mitochondrial depolarize [9,10].Colocalization of these signals yields a yellow signal thatindicates mitochondria actively producing ROS. Hepatocytes,3×105, were plated overnight in a MatTek culture dish(MatTek, Ashland, MA) and treated or untreated for 20 min at37°C and H2-DCFDA and MTR were added to achieve finalconcentrations of 2 μM and 500 nM, respectively. After a1-h incubation, the medium was aspirated, hepatocytes weresubjected to two wash steps with PBS, and confocal imagingwas performed. Confocal images were obtained with a ZeissLSM 510 laser scanning confocal microscope (Thornwood,NY). Detector gain was equal in both experimental and controlgroups. Cells were randomly selected and analyzed forcolocalization of H2-DCF fluorescence and mitochondrialstaining with MTR. The images of green H2-DCF and redMTR were superimposed and distinct regions of yellowprovided direct evidence of mitochondria derived ROS. Thedetector gains used on the confocal microscope for both the dyeswere selected so as to not obscure visualization of subcellularfeatures of the cell while still allowing discrimination of signalsproportional to activity. These gain levels were not alteredduring the course of the experiments. However, by keeping gainsconstant, MTR intensity levels varied slightly between theexperiments. Images were background-corrected before analysisand average fluorescence was calculated using Adobe Photo-shop (San Jose, CA).

H2O2-treated hepatocytes

Normal hepatocytes were treated with 10 μM H2O2 for 10–20 min and ROS activity was assessed with H2-DCF for 72 h.ROS activity in these hepatocytes was increased approximatelytwo-fold and maintained this level of activity for 72 h. ThisROS activity is similar to that seen in cirrhotic hepatocytes atbaseline. These cells were then subjected to TGFβ treatmentand apoptosis was determined at 48 h as described previously.

Determination of protein tyrosine phosphatase activity

Protein tyrosine phosphatase activity was spectrometricallydetermined using p-nitrophenyl phosphate [11].

Data analysis

All experiments were performed in at least triplicate. Dataare reported as mean plus or minus the standard deviation.Statistical analysis was performed using ANOVA and Dunnett’st test and a P value of 0.05 was considered significant.

Fig. 1. (A) Cirrhotic hepatocytes were pretreated with exogenous antioxidantsincluding deferoxamine (DEF; 1.5 mM), glutathione (GLUT; 5 mM), DPPD(5 μM), or N-acetylcysteine (NAC; 2.5 mM) for 1 h prior to administration ofTGFβ (5 ng/ml) and ROS activity was measured 90 min after TGFβadministration. The ROS activity, expressed as fold difference, returned tobaseline at time zero (data not shown), but increased in response to TGFβtreatment for all antioxidants tested. These findings suggest involvement ofmultiple ROS-generating pathways in increased cirrhotic hepatocyte basal ROSactivity. (B) The percentage of condensed nuclei, indicative of morphologicapoptosis, was determined after cirrhotic hepatocytes were pretreated withexogenous antioxidants (agents and dose identical to Fig. 1A) for 1 h prior totreatment with or without TGFβ (5 ng/ml) and apoptosis was determined at 48 h.Pretreatment with the antioxidants alone did not induced apoptosis, butantioxidant treatment followed by TGFβ administration increased markedlycirrhotic hepatocyte apoptosis for each antioxidant tested.


Results

Exogenous antioxidants and cirrhotic hepatocyte apoptosis

Previous work from our laboratory demonstrated thatcirrhotic hepatocyte resistance to TGFβ-induced apoptosiswas mediated through ROS and that pretreatment with troloxdecreased ROS and rendered these hepatocytes susceptible toTGFβ-induced apoptosis [1]. To determine if antioxidantsselective for specific ROS-generating pathways would provideinformation about the source of ROS production in cirrhotichepatocytes, these cells were treated with 1.5 mMdeferoxamine,5 mM glutathione, 5 μM DPPD, or 2.5 mM NAC for 1 h. Eachtreatment group exhibited a significant decrease that returnedROS to control levels in response to the various antioxidants(data not shown). Furthermore, following antioxidant pretreat-ment each group demonstrated a significant increase in ROSgeneration after TGFβ administration (Fig. 1A). This ROS spike90min after TGFβ administration corresponds to generation of aROS burst that occurs in normal hepatocytes [1]. Importantly,the ROS burst following TGFβ administration in cirrhotichepatocytes pretreated with the various antioxidants wasassociated with a significant increase in morphologic apoptosisat 48 h (Fig. 1B). These data suggest that cirrhotic hepatocytesrespond to several exogenous antioxidants with decreasedROS and that a reduction in baseline ROS in cirrhotic hepa-tocytes permits a TGFβ-induced ROS burst with subsequentapoptosis. However, exogenous administration of multipleantioxidants does not specify the ROS-initiating pathway incirrhotic hepatocytes.

Adenovirus expression of antioxidants and cirrhotic hepatocyteapoptosis

To determine if adenoviruses expressing antioxidantenzymes had an effect on cirrhotic hepatocytes similar toexogenously applied antioxidants, adenoviruses expressingluciferase as a control (AdLuc), catalase (AdCat), or MnSOD(AdMnSOD) were administered. The adenoviruses expressingMnSOD and catalase both decreased ROS in cirrhotichepatocytes (Fig. 2A), and these cells were capable of attaininga ROS spike 90 min following TGFβ treatment (Fig. 2B). Theadenovirus expressing MnSOD was particularly effective indecreasing ROS and permitting a robust ROS response toTGFβ. In addition, assessment of apoptosis (Fig. 2C) demons-trated that cirrhotic hepatocytes transduced with either theMnSOD or the catalase adenoviruses underwent significantapoptosis 48 h following treatment with TGFβ. These findingsindicate that cirrhotic hepatocytes infected with antioxidantenzymes behave similarly to cirrhotic hepatocytes treated withexogenous antioxidants; however, adenoviral expression ofMnSOD appeared particularly effective in reducing ROS andpermitting a TGFβ-induced ROS burst.

To confirm that caspase-mediated apoptosis was the mode ofcell death in cirrhotic hepatocytes, cell lysates from control andadenoviral infected cells were subjected to immunoblots foractivated caspase-3. The primary antibody has affinity for the

inactive, precleaved, 35-kDa caspase-3 zymogen and thecleaved, activated, 17-kDa fragment. In Fig. 3A, cirrhotichepatocytes transduced with the luciferase alone (AdLuc; lane1) or luciferase followed by TGFβ (lane 2) did not exhibit thecleaved caspase-3 product. Likewise, infection with AdMnSODalone (lane 3) failed to result in caspase-3 cleavage. However,when cells infected with AdMnSOD for 24 h were subsequentlytreated with TGFβ, caspase-3 cleavage occurred at 48 hfollowing TGFβ treatment (lane 4). Similar findings wereevident for cirrhotic hepatocytes treated with AdCat (Fig. 3B).These immunoblot findings were substantiated by a caspaseassay that showed only cirrhotic hepatocytes transduced withantioxidant enzymes and subsequently treated with TGFβ for48 h exhibited caspase-3 activation (Fig. 3C). Additionally,cirrhotic hepatocytes transduced with AdMnSOD and AdCat for24 h and pretreated for 1 h with zVAD, a pan-caspase inhibitor,failed to undergo apoptosis in response to TGFβ, furtherconfirming an apoptotic mode of cell death (Fig. 3D).Collectively, these data suggest that TGFβ-mediated cell deathin cirrhotic hepatocytes pretreated with adenoviruses expressingantioxidants is caspase dependent.

Fig. 2. (A) Cirrhotic hepatocytes were transfected with adenoviruses (MOI100) expressing luciferase (AdLuc; control), catalase (AdCat), or MnSOD(AdMnSOD) for 24 h and ROS activity was fluorometrically determined usingDCF. Some hepatocytes were untreated with adenovirus (control). Transfectionwith AdCat and AdMnSOD decreased significantly ROS formation (P<0.05 vscontrol). (B) Cirrhotic hepatocytes were transfected with adenoviruses (MOI100) expressing luciferase (AdLuc), catalase (AdCat), or MnSOD (AdMnSOD)24 h prior to TGFβ (5 ng/ml) treatment and incubated with DCF. Generation ofROS, expressed as fold difference, was fluorometrically determined at 90 minfollowing TGFβ administration. The adenoviruses expressing the antioxidants,catalase and MnSOD, permitted a ROS burst at 90 min, indicating ROSresponsiveness to TGFβ administration in cirrhotic hepatocytes. (C) Thepercentage of condensed nuclei, indicative of morphologic apoptosis, wasdetermined at 48 h in cirrhotic hepatocytes in response to treatment with orwithout TGFβ (5 ng/ml) after pretreatment for 24 h with adenoviruses (100MOI) expressing luciferase (AdLuc), catalase (AdCat), or MnSOD(AdMnSOD). Transfection of cirrhotic hepatocytes with adenoviruses expres-sing the antioxidants, catalase and MnSOD, followed by treatment with TGFβincreased significantly the percentage of apoptotic hepatocytes compared tocontrol (P<0.05 vs control).


Colocalization of TGFβ-induced ROS and mitochondrialfunction

Because AdMnSOD administration suggested a prominentrole for mitochondria in ROS production, confocal microscopywas used to investigate TGFβ-induced ROS production inmitochondria. In these experiments, green fluorescent H2-DCF(2 μM) was used to identify ROS and red fluorescent MTR(500 nM) was used to label mitochondria. The ROS responsewas monitored by confocal microscopy in both normal andcirrhotic hepatocytes that were untreated or pretreated withvarious antioxidants prior to administration of TGFβ. Asignificant increase in baseline ROS activity was evident inuntreated cirrhotic hepatocytes compared to normal hepatocytes(Fig. 4A).

Similar to our previous study [1], normal hepatocytes treatedwith TGFβ underwent a ROS burst at 90 min (Fig. 4B);however, when these normal hepatocytes were pretreated withAdMnSOD and then administered TGFβ, a ROS burst did notoccur. In normal hepatocytes, ROS levels did not changesignificantly in response to antioxidant treatment alone. Similarresults were noted when AdCat and trolox were used asantioxidants.

In cirrhotic hepatocytes, AdMnSOD infection decreasedsignificantly the basal ROS activity (Fig. 4C). Cirrhotichepatocytes treated with TGFβ alone did not undergo a ROSspike at 90 min, but cirrhotic cells pretreated with AdMnSODunderwent a ROS burst similar to that of untreated control cells(Fig. 4D).

These data confirm that a reduction in ROS levels in cirrhotichepatocytes sensitizes these previously resistant cells to TGFβ-induced mitochondrial ROS generation and subsequent apo-ptosis. Conversely, treatment of normal hepatocytes withvarious antioxidants renders these formerly responsive he-patocytes resistant to TGFβ-induced mitochondrial ROSgeneration.

Role of antiapoptotic proteins and protein tyrosinephosphatase in cirrhotic livers and hepatocytes

To test if resistance to apoptosis in cirrhotic hepatocytescould result from modified apoptotic machinery, lysates fromnormal and cirrhotic livers were immunoblotted for antiapopto-tic proteins, including Bcl-xL and MCL-1 [12]. Levels of bothBcl-xL and MCL-1 from cirrhotic livers were similar to thosefrom normal livers (Fig. 5A), suggesting that upregulation ofantiapoptotic proteins is not the mechanism underlyingresistance to apoptosis in cirrhotic livers.

ROS can modify proteins, lipids, and nucleic acids, leadingto tissue injury. Since protein tyrosine phosphatase is a keysignaling molecule associated with TGFβ transduction and isalso known to be sensitive to oxidative stress [13,14], weexamined the possibility that enhanced basal ROS in cirrhotichepatocytes could perturb the activity of protein tyrosinephosphatase. Measurement of total tyrosine phosphataseactivity demonstrates that there was no significant differencein tyrosine phosphatase activity between normal and cirrhotic


hepatocytes (Fig. 5B). Moreover, treatment with TGFβ for 48 hdid not change the activity of tyrosine phosphatase in bothgroups (Fig. 5B). These findings suggest that protein tyrosinephosphatases are not significantly involved in cirrhotichepatocyte resistance to apoptosis, and the mechanisms otherthan altered tyrosine phosphatase may contribute to resistanceto apoptosis.

H2O2-treated normal hepatocytes mimic cirrhotic hepatocyteROS activity and apoptotic response

Our previous work [1] and the findings in this study suggeststrongly that basal ROS activity in cirrhotic hepatocytesmediates responsiveness to TGFβ-induced apoptosis. Todetermine if normal hepatocytes, which are responsive toTGFβ-induced apoptosis, could be rendered resistant toapoptosis by increased basal ROS activity, these cells wereexposed to H2O2 and ROS activity monitored over 72 h.Treatment with 10 μM H2O2 produced sustained ROS activitycomparable to cirrhotic hepatocyte basal ROS activity (Fig. 6A).Additionally, hepatocyte viability was 97% over 72 h, indicatingthat this low concentration of H2O2 is not cytotoxic (data notshown). When these H2O2-converted hepatocytes were subse-quently treated with TGFβ, a ROS burst was not apparent (datanot shown). Furthermore, these H2O2-converted hepatocyteswere resistant to TGFβ-induced apoptosis (Fig. 6B). Troloxpretreatment of normal hepatocytes resulted in the expectedinhibitory response to TGFβ-induced apoptosis compared tonormal hepatocytes not treated with antioxidant. However,trolox pretreatment of H2O2-converted hepatocytes resulted in adecrease in ROS activity and return of the TGFβ-inducedapoptotic response. To further evaluate the role of enhancedROS in cirrhotic hepatocytes, normal hepatocytes were treatedwith 30 μM 2,3-dimethoxy-1,4-naphthoquinone (DMNQ), an

Fig. 3. Cleavage of caspase-3 by immunoblot was assessed at 48 h to confirmapoptotic cell death in cirrhotic hepatocytes transfected with adenoviruses (100MOI) expressing luciferase (AdLuc), catalase (AdCat), or MnSOD(AdMnSOD) for 24 h prior to treatment with TGFβ (5 ng/ml). (A) Cirrhotichepatocytes were transfected with AdLuc alone (lane 1), AdLuc followed byTGFβ (lane 2), AdMnSOD alone (lane 3), or AdMnSOD followed by TGFβ(lane 4). As expected, only infection with AdMnSOD followed by TGFβtreatment resulted in the cleaved product indicating caspase-3 activity. (B)Cirrhotic hepatocytes were transfected with AdLuc alone (lane 1), AdLucfollowed by TGFβ (lane 2), AdCat alone (lane 3), or AdCat followed by TGFβ(lane 4). The cleaved caspase-3 fragment was present only in AdCat-infectedhepatocytes treated with TGFβ. (C) Caspase-3 activity, expressed as folddifference, was measured at 48 h to confirm apoptotic cell death in cirrhotichepatocytes transfected with adenviruses (100 MOI) expressing luciferase,catalase, or MnSOD for 24 h prior to treatment with or without TGFβ (5 ng/ml).Infection with adenoviruses alone did not alter caspase-3 activity whereastransduction of cirrhotic hepatocytes with antioxidant expressing adenoviruses(AdCat and AdMnSOD) followed by TGFβ increased significantly caspase-3activity. (D) The percentage of condensed nuclei was determined 48 h followingTGFβ treatment in cirrhotic hepatocytes infected with adenoviruses (MOI 100)expressing luciferase (AdLuc), catalase (AdCat), or MnSOD (AdMnSOD) for24 h prior to TGFβ treatment (5 ng/ml). These hepatocytes were also pretreatedwith the pan-caspase inhibitor, zVAD (5 μM), for 1 h prior to treatment with orwithout TGFβ. Pretreatment with zVAD inhibited apoptosis in cirrhotichepatocytes infected with AdCat and AdMnSOD confirming a caspase-mediated form of cell death.

intracellular redox-cycling agent [15], for 20 min. After washingonce, hepatocytes were further incubated in HDM (withoutDMNQ) for 24 h in the presence of TGFβ. Development ofapoptosis was assessed by chromatin condensation and nuclearfragmentation in PI-stained nuclei, as described under Materialsand methods (Fig. 7). DMNQ at this concentration was notcytotoxic (data not shown). In normal hepatocytes, TGFβtreatment substantially increased apoptosis (Fig. 7A). Incontrast, DMNQ-converted hepatocytes were resistant toTGFβ-mediated apoptosis (Fig. 7B). Taken together, these

Fig. 4. Normal and cirrhotic hepatocytes were loaded with the fluorophores H2-DCFDA (2 μM; green) to assess ROS formation and MitoTracker Red (MTR500 nM; red) to localize respiring mitochondria. Confocal microscopy wasperformed at 90 min following TGFβ (5 ng/ml) administration. (A) Normal(left) and cirrhotic (right) hepatocytes were imaged without treatment. Note alow mitochondrial ROS formation in normal hepatocytes while diffused ROSformation in cirrhotic hepatocytes. (B) Normal hepatocytes were treated withTGFβ for 90 min. Some hepatocytes were transfected with AdMnSOD for 24 hprior to TGFβ treatment. Confocal imaging revealed a ROS burst in themitochondria (yellow fluorescence, left panel), which was suppressed byAdMnSOD transfection (middle and right panel). (C) Confocal imaging ofcirrhotic hepatocytes. TGFβ alone did not caused a mitochondrial ROS burst(left panel). Although AdMnSOD alone (middle panel) decreased ROSgeneration, subsequent TGFβ administration induced a mitochondrial ROSburst (yellow, right panel). (D) Quantification of yellow fluorescence, anindication of mitochondrial generation of ROS for control and cirrhotichepatocytes treated with TGFβ, AdMnSOD, or TGFβ and AdMnSOD.

Fig. 5. (A) Whole tissue lysates were prepared from normal and cirrhotic liversand levels of Bcl-xL and MCL-1 were determined by Western blot analysis.Actin levels were also immunoblotted to ensure an equal protein loading. (B)Hepatocytes isolated from normal and cirrhotic livers were incubated in HDM inthe presence or absence of 5 ng/ml TGFβ, as described under Materials andmethods. After 48 h, cell lysates were prepared and total protein tyrosinephosphatase activity was determined spectrometrically.


findings confirm the importance of the basal ROS activity inpreventing TGFβ-induced hepatocyte apoptosis.

Discussion

ROS serve as intermediaries in cellular signaling pathwaysand recent studies have demonstrated the importance ofgenerating these molecules in apoptotic death pathways[1,4,16]. Our previous work examining cirrhotic hepatocytessuggests that chronic elevation of basal ROS activity inhibitsmitochondrial pathway-dependent hepatocyte apoptosis in-duced by TGFβ, TNFα, and UV, and the findings in thisstudy document the crucial role for ROS in TGFβ-inducedhepatocyte apoptosis. The aim of this study was to investigatefurther the necessity of ROS in hepatocyte apoptosis, andspecifically to examine the importance of mitochondrial-generated ROS activity. The major findings of this studycorroborate our previous work and extend our findings tosuggest that the basal increase in cirrhotic hepatocyte ROSlikely is not limited to a single ROS-generating pathway.Furthermore, we found that the mitochondria are a primarysource of TGFβ-induced ROS activity in normal hepatocytesand in cirrhotic hepatocytes that have been pretreated withantioxidants. Finally, normal hepatocytes exposed to H2O2 andDMNQ can be “converted” to an oxidative state that confers

Fig. 6. (A) ROS fluorescent units (FU) were determined over time in normal,cirrhotic, and normal-H2O2-converted hepatocytes. Normal hepatocytes (opencircles) have a low basal level of ROS activity whereas cirrhotic hepatocytes(black diamonds) have a high steady-state level of ROS activity. Normalhepatocytes exposed to 10 μM H2O2 for 10 min (black triangles) failed tomaintain a sustained ROS response. However, normal hepatocytes exposed to10 μM H2O2 for 20 min (black squares) demonstrate sustained increased ROSactivity similar to basal cirrhotic hepatocytes over the study period. Thesehepatocytes had 97% viability (data not shown). (B) The percentage ofcondensed nuclei was determined in normal and H2O2-converted hepatocytes at48 h after pretreatment (or not) with trolox (2 μM) followed by treatment with orwithout TGFβ (5 ng/ml). Treatment with H2O2 alone did not increase apoptosis,and, similar to cirrhotic hepatocytes, H2O2-converted cells were resistant toTGFβ-induced apoptosis at 48 h. However, H2O2-converted hepatocytes thatwere pretreated with trolox prior to TGFβ exposure underwent apoptosis similarto normal hepatocyte controls and antioxidant-treated cirrhotic hepatocytes.

Fig. 7. Hepatocytes isolated from normal livers were incubated with 5 ng/mlTGFβ for 24 h. Some hepatocytes were treated with 30 μM 2,3-dimethoxy-1,4-naphthoquinone (DMNQ) for 20 min prior to TGFβ administration. Afterwashing once, hepatocytes were further incubated with TGFβ. Apoptosis wasevaluated by chromatin condensation and nuclear fragmentation of PI-stainednuclei (arrows), as described under Materials and methods. TGFβ aloneinduced a substantial apoptosis (A), which was reversed by a brief treatmentwith DMNQ (B).


resistance to TGFβ-induced apoptosis. Cumulatively, thesefindings suggest that ROS play an integral role in hepatocyteresponsiveness to apoptotic stimuli.

The cirrhotic liver represents the end-stage morphologicresult from chronic inflammatory injury related to viralhepatitis, alcohol ingestion, and other metabolic causes.Chronic CCl4-induced liver injury in the mouse producesbridging fibrosis with up-regulation of fibrotic stimuli such asTGFβ [17]. Despite increased hepatic levels of TGFβ and otherproapoptotic cytokines, the cirrhotic liver does not have achronically increased rate of apoptosis [1], which suggests thatthe hepatocyte exposed to chronic inflammation has adapted anantiapoptotic mechanism. The antiapoptotic phenotype hasbeen noted in several forms of liver injury including alcohol-induced liver injury which is associated with increased ROS[18]. In that study, hepatocellular apoptosis was decreased witha resultant increase in dysplastic hepatocytes and thesephenotypic changes were likely related to p53 expression.Furthermore, increased ROS activity has been linked tomalignant transformation [19] which may be related to cytokine

receptor profiles and resistance to apoptosis [20] or to decreasedexpression of antioxidants [8,21]. In addition to decreasingresponsiveness to apoptosis, increased ROS may result innegative regulatory changes in the cell cycle and alter thebalance between proliferation and apoptosis [22]. Collectively,these findings suggest that chronically increased ROS maymediate changes in susceptibility to apoptosis and as demon-strated by our findings and those of Herrera et al. [4,23,24]multiple ROS-generating pathways and downstream cellularsignaling pathways may be involved.

Because the mitochondria are a primary source of ROS, we[1] and others [25] have focused on the importance ofmitochondrial ROS generation in chronic hepatic inflamma-tory states. The current study demonstrates that TGFβ-inducedROS activity in normal hepatocytes predominantly arises fromthe mitochondria as evidenced by colocalization of H2-DCFDA fluorescence activity and respiring mitochondria onconfocal microscopic imaging. Moreover, in cirrhotic hepato-cytes that are pretreated with an antioxidant and subsequentlytreated with TGFβ, significant ROS activity again arises in themitochondria, suggesting that the mitochondria are animportant source of TGFβ-induced ROS. These findings alsosuggest that even though cirrhotic hepatocytes have elevatedbasal ROS activity, the mitochondria in these cells are


functionally intact, respond appropriately after treatment withan antioxidant, and do not have an irreversible injury related toCCl4 administration. The mechanisms that may influencemitochondrial ROS production have been examined closelyand may be related to the Bcl family of proteins [26–28].However, our results suggest that up-regulation of antiapopto-tic proteins, Bcl-xL and MCL-1, is not the mechanismunderlying resistance to apoptosis in cirrhotic hepatocytessince levels of both proteins in cirrhotic livers were similar tothose in normal livers. In receptor-dependent apoptosismitochondrial ROS production may be related to a Bid-mediated induction of mitochondrial ROS [27]. Alternatively,Bax, another proapoptotic Bcl family member, may associatewith the mitochondrial voltage-dependent anion channel andmodulate cytochrome c release and apoptotic cell death in theacute phase [28]. In the setting of chronic inflammation,mitochondrial function and Bcl family function have been lessstudied, but in nonalcoholic steatohepatitis (NASH) changes inthe P450 system may up-regulate ROS production [25]. Thesestudies suggest that several pathways may be involved inmitochondrial ROS generation, and that further investigation isneeded to determine how these pathways may mediate ROSproduction in chronic inflammatory states.

Our experiments in which normal murine hepatocytes wereexposed to H2O2 and exhibited sustained ROS activity andantioxidant-reversible resistance to apoptosis suggest that theredox state of hepatocytes mediates the TGFβ-apoptoticresponse. The importance of redox state is further supportedby our findings that a brief treatment of DMNQ, a redox-cyclingagent [15], to normal hepatocytes reversed the sensitivity toTGFβ-induced apoptosis. Similar findings were noted byTejima et al. [29] who used H2O2 as a preconditioning moleculeand demonstrated that the hepatotoxicity could be reduced bylow-dose H2O2. Likewise, hepatocytes treated with nontoxicdoses of menadione, a superoxide generator, resisted oxidant-induced cell death through an ERK-dependent pathway whilethe JNK pathway was proapoptotic [30]. Interestingly, theproapoptotic function of the JNK pathway appeared to bemitochondrial-independent. Other mechanisms that may renderhepatocytes resistant to H2O2-mediated cell death involveintramitochondrial changes in caspase processing [31]. In thisstudy, cells treated with H2O2 demonstrated a slight decrease inthe inner mitochondrial membrane potential but withoutcytochrome c release. In addition, within the mitochondria,procaspase-9 underwent autocleavage, but overexpression ofthe antiapoptotic protein, Bcl-2, caused accumulation ofcaspase-9 and prevented oxidant-induced cell death. Theintramitochondrial processing of procaspase-9 was associatedwith disulfide-bonded dimers of caspase-9 that appeared toprevent mitochondrial release. These studies suggest thatexposure to nontoxic doses of oxidants can prevent cell deathby up-regulation of antiapoptotic cellular mechanisms.

In conclusion, these experiments demonstrate that the ROSstate of hepatocytes mediates responsiveness to TGFβ-inducedapoptosis. In normal hepatocytes, which have low basal ROSlevels, a mitochondrial-derived ROS burst is required forTGFβ-induced apoptosis whereas in cirrhotic hepatocytes,

which have an increased basal level of ROS, pretreatment withan antioxidant permits a TGFβ-induced mitochondrial ROSresponse with subsequent apoptosis. These studies furthersuggest that the mitochondria in cirrhotic hepatocytes are notirreversibly damaged and that adequate energy stores arepresent to permit apoptosis. The mechanisms that regulate theseROS-mediated responses are unknown and warrant furtherinvestigation.

References

[1] Black, D.; Bird, M. A.; Samson, C. M.; Lyman, S.; Lange, P. A.; Schrum,L. W.; Qian, T.; Lemasters, J. J.; Brenner, D. A.; Rippe, R. A.; Behrns,K. E. Primary cirrhotic hepatocytes resist TGFbeta-induced apoptosisthrough a ROS-dependent mechanism. J. Hepatol. 40:942–951; 2004.

[2] Sanchez, A.; Alvarez, A. M.; Benito, M.; Fabregat, I. Apoptosis inducedby transforming growth factor-beta in fetal hepatocyte primary cultures:involvement of reactive oxygen intermediates. J. Biol. Chem. 271:7416–7422; 1996.

[3] Herrera, B.; Alvarez, A. M.; Sanchez, A.; Fernandez, M.; Roncero, C.;Benito, M.; Fabregat, I. Reactive oxygen species (ROS) mediates themitochondrial-dependent apoptosis induced by transforming growth factor(beta) in fetal hepatocytes. FASEB J. 15:741–751; 2001.

[4] Herrera, B.; Murillo, M. M.; Alvarez-Barrientos, A.; Beltran, J.;Fernandez, M.; Fabregat, I. Source of early reactive oxygen species inthe apoptosis induced by transforming growth factor-beta in fetal rathepatocytes. Free Radic. Biol. Med. 36:16–26; 2004.

[5] Albright, C. D.; Salganik, R. I.; Craciunescu, C. N.;Mar,M.H.; Zeisel, S. H.Mitochondrial and microsomal derived reactive oxygen species mediateapoptosis induced by transforming growth factor-beta1 in immortalized rathepatocytes. J. Cell Biochem. 89:254–261; 2003.

[6] Franklin, C. C.; Rosenfeld-Franklin, M. E.; White, C.; Kavanagh, T. J.;Fausto, N. TGFbeta1-induced suppression of glutathione antioxidantdefenses in hepatocytes: caspase-dependent post-translational and caspase-independent transcriptional regulatory mechanisms. FASEB J. 17:1535–1537; 2003.

[7] Cullen, J. J.; Hinkhouse, M. M.; Grady, M.; Gaut, A. W.; Liu, J.; Zhang,Y. P.; Weydert, C. J.; Domann, F. E.; Oberley, L. W. Dicumarol inhibitionof NADPH:quinone oxidoreductase induces growth inhibition ofpancreatic cancer via a superoxide-mediated mechanism. Cancer Res. 63:5513–5520; 2003.

[8] Weydert, C.; Roling, B.; Liu, J.; Hinkhouse, M. M.; Ritchie, J. M.;Oberley, L. W.; Cullen, J. J. Suppression of the malignant phenotype inhuman pancreatic cancer cells by the overexpression of manganesesuperoxide dismutase. Mol. Cancer Ther. 2:361–369; 2003.

[9] Buckman, J. F.; Hernandez, H.; Kress, G. J.; Votyakova, T. V.; Pal, S.;Reynolds, I. J. MitoTracker labeling in primary neuronal and astrocyticcultures: influence of mitochondrial membrane potential and oxidants.J. Neurosci. Methods 104:165–176; 2001.

[10] Elmore, S. P.; Nishimura, Y.; Qian, T.; Herman, B.; Lemasters, J. J.Discrimination of depolarized from polarized mitochondria by confocalfluorescence resonance energy transfer. Arch. Biochem. Biophys.422:145–152; 2004.

[11] Montalibet, J.; Skorey, K. I.; Kennedy, B. P. Protein tyrosine phosphatase:enzymatic assays. Methods 35:2–8; 2005.

[12] Michels, J.; Johnson, P. W.; Packham, G. Mcl-1. Int. J. Biochem. Cell Biol.37:267–271; 2005.

[13] Meng, T. C.; Fukada, T.; Tonks, N. K. Reversible oxidation and inactivationof protein tyrosine phosphatases in vivo. Mol. Cell 9:387–399; 2002.

[14] Greenwel, P.; Hu, W.; Kohanski, R. A.; Ramirez, F. Tyrosine dephos-phorylation of nuclear proteins mimics transforming growth factor beta 1stimulation of alpha 2(I) collagen gene expression. Mol. Cell. Biol.15:6813–6819; 1995.

[15] Dypbukt, J. M.; Ankarcrona, M.; Burkitt, M.; Sjoholm, A.; Strom, K.;Orrenius, S.; Nicotera, P. Different prooxidant levels stimulate growth,trigger apoptosis, or produce necrosis of insulin-secreting RINm5F cells.


The role of intracellular polyamines. J. Biol. Chem. 269:30553–30560;1994.

[16] Ribeiro, A.; Bronk, S. F.; Roberts, P. J.; Urrutia, R.; Gores, G. J. Thetransforming growth factor beta(1)-inducible transcription factor TIEG1,mediates apoptosis through oxidative stress. Hepatology 30:1490–1497;1999.

[17] Tsukada, S.;Westwick, J. K.; Ikejima, K.; Sato, N.; Rippe, R. A. SMADandp38 MAPK signaling pathways independently regulate alpha1(I) collagengene expression in unstimulated and transforming growth factor-beta-stimulated hepatic stellate cells. J. Biol. Chem. 280:10055–10064; 2005.

[18] Pani, G.; Fusco, S.; Colavitti, R.; Borrello, S.; Maggiano, N.; Cravero,A. A.; Farre, S. M.; Galeotti, T.; Koch, O. R. Abrogation of hepatocyteapoptosis and early appearance of liver dysplasia in ethanol-fed p53-deficient mice. Biochem. Biophys. Res. Commun. 325:97–100; 2004.

[19] O'Byrne, K. J.; Dalgleish, A. G. Chronic immune activation andinflammation as the cause of malignancy. Br. J. Cancer 85:473–483; 2001.

[20] Yoon-, D. Y.; Cho, Y. S.; Park, J. W.; Kim, S. H.; Kim, J. W. Up-regulationof reactive oxygen species (ROS) and resistance to Fas-mediated apoptosisin the C33A cervical cancer cell line transfected with IL-18 receptor. Clin.Chem. Lab. Med. 42:499–506; 2004.

[21] Liu, J.; Hinkhouse, M. M.; Sun, W.; Weydert, C. J.; Ritchie, J. M.; Oberley,L. W.; Cullen, J. J. Redox regulation of pancreatic cancer cell growth: roleof glutathione peroxidase in the suppression of the malignant phenotype.Hum. Gene Ther. 15:239–250; 2004.

[22] Horimoto, M.; Fulop, P.; Derdak, Z.; Wands, J. R.; Baffy, G. Uncouplingprotein-2 deficiency promotes oxidant stress and delays liver regenerationin mice. Hepatology 39:386–392; 2004.

[23] Herrera, B.; Fernandez, M.; Alvarez, A. M.; Roncero, C.; Benito, M.; Gil,J.; Fabregat, I. Activation of caspases occurs downstream from radicaloxygen species production, Bcl-xL down-regulation, and early cytochrome

C release in apoptosis induced by transforming growth factor beta in ratfetal hepatocytes. Hepatology 34:548–556; 2001.

[24] Herrera, B.; Fernandez, M.; Roncero, C.; Ventura, J. J.; Porras, A.;Valladares, A.; Benito, M.; Fabregat, I. Activation of p38MAPK by TGF-beta in fetal rat hepatocytes requires radical oxygen production, but isdispensable for cell death. FEBS Lett. 499:225–229; 2001.

[25] Fromenty, B.; Robin, M. A.; Igoudjil, A.; Mansouri, A.; Pessayre, D. Theins and outs of mitochondrial dysfunction in NASH. Diabetes Metab.30:121–138; 2004.

[26] Kroemer, G.; Reed, J. C. Mitochondrial control of cell death. Nat. Med.6:513–519; 2000.

[27] Ding, W. X.; Ni, H. M.; DiFrancesca, D.; Stolz, D. B.; Yin, X. M. Bid-dependent generation of oxygen radicals promotes death receptoractivation-induced apoptosis in murine hepatocytes. Hepatology 40:403–413; 2004.

[28] Adachi, M.; Higuchi, H.; Miura, S.; Azuma, T.; Inokuchi, S.; Saito, H.;Kato, S.; Ishii, H. Bax interacts with the voltage-dependent anion channeland mediates ethanol-induced apoptosis in rat hepatocytes. Am. J. Physiol.Gastrointest. Liver Physiol. 287:G695–G705; 2004.

[29] Tejima, K.; Arai, M.; Ikeda, H.; Tomiya, T.; Yanase, M.; Inoue, Y.;Nagashima, K.; Nishikawa, T.; Watanabe, N.; Omata, M.; Fujiwara, K.Ischemic preconditioning protects hepatocytes via reactive oxygen speciesderived from Kupffer cells in rats. Gastroenterology 127:1488–1496;2004.

[30] Czaja, M. J.; Liu, H.; Wang, Y. Oxidant-induced hepatocyte injury frommenadione is regulated by ERK and AP-1 signaling. Hepatology 37:1405–1413; 2003.

[31] Katoh, I.; Tomimori, Y.; Ikawa, Y.; Kurata, S. Dimerization and processingof procaspase-9 by redox stress in mitochondria. J. Biol. Chem. 279:15515–15523; 2004.

Documents

Basal reactive oxygen species determine the susceptibility to apoptosis in cirrhotic hepatocytes