Involvement of NADPH oxidase isoforms and Src family kinases in CD95-dependent hepatocyte apoptosis.

CD95 ligand (CD95L) triggers a rapid formation of reactive oxygen species (ROS) as an upstream event of CD95 activation and apoptosis induction in rat hepatocytes. This ROS response was sensitive to inhibition by diphenyleneiodonium, apocynin, and neopterin, suggestive of an involvement of NADPH oxidases. In line with this, hepatocytes expressed mRNAs not only of the phagocyte gp91phox (Nox 2), but also of the homologs Nox 1 and 4 and Duox 1 and 2, as well as the regulatory subunit p47phox. gp91phox (Nox 2) and p47phox were also identified at the protein level in rat hepatocytes. CD95L induced within 1 min ceramide formation and serine phosphorylation of p47phox, which was sensitive to inhibitors of sphingomyelinase and protein kinase Czeta (PKCzeta). These inhibitors and p47phox protein knockdown inhibited the early CD95L-induced ROS response, suggesting that ceramide and PKCzeta are upstream events of the CD95L-induced Nox/Duox activation. CD95L also induced rapid activation of the Src family kinase Yes, being followed by activation of c-Src, Fyn, and c-Jun-N-terminal kinases (JNK). Only Yes and JNK activation were sensitive to N-acetylcysteine, inhibitors of NADPH oxidase, PKCzeta, or sphingomyelinase, indicating that the CD95L-induced ROS response is upstream of Yes and JNK but not of Fyn and c-Src activation. Activated Yes rapidly associated with the epidermal growth factor receptor (EGFR), which became phosphorylated at Tyr845 and Tyr1173 but not at Tyr1045. Activated EGFR then triggered an AG1478-sensitive CD95-tyrosine phosphorylation, which was a signal for membrane targeting of the EGFR/CD95 complex, subsequent recruitment of Fas-associated death domain and caspase 8, and apoptosis induction. All of these events were significantly blunted by inhibitors of sphingomyelinase, PKCzeta, NADPH oxidases, Yes, or EGFR-tyrosine kinase activity and after protein knockdown of either p47phox, Yes, or EGFR. The data suggest that CD95L-induced apoptosis involves a sphingomyelinase- and PKCzeta-dependent activation of NADPH oxidase isoforms, which is required for Yes/EGFR/CD95 interactions as upstream events of CD95 activation.

CD95 1 (Apo-1/Fas) belongs to the death receptor family and plays an important role in apoptosis induction in many cell types. In liver, CD95 can be activated after ligation with its natural ligand (CD95L); however, CD95 can also be activated in a ligand-independent way, for example by hydrophobic bile acids or hyperosmotic cell shrinkage (1,2). Liganddependent and -independent CD95 activation in hepatocytes is a complex process, which finally results in the formation of the death-inducing signaling complex (DISC) and activation of the initiator caspase 8, which then triggers a variety of complexly regulated downstream events, which may finally result in apoptotic cell death. In hepatocytes, proapoptotic stimuli such as CD95L, hydrophobic bile acids, or hyperosmotic cell shrinkage induce a rapid oxidative stress response, which triggers ligand-independent activation of the epidermal growth factor receptor (EGFR) and subsequent c-Jun-Nterminal kinase (JNK)-dependent association of the EGFR with CD95 within 30 min (1,2). Following EGFR-catalyzed tyrosine phosphorylation of the CD95, the EGFR⅐CD95 protein complex is targeted in a microtubule-dependent way to the plasma membrane, where formation of the DISC occurs (3). The mechanisms underlying CD95L-triggered EGFR activation are not fully understood; however, EGFR activation in response to other proapoptotic stimuli involves an antioxidant-sensitive activation of Yes as upstream event (4,5). Yes, like Fyn and c-Src, are members of the Src kinase family, which are ubiquitously expressed (for a review, see Refs. 6 and 7). In rat hepatocytes, c-Src participates in integrin-dependent osmosignaling (8), whereas Yes is important for apoptosis induction through the CD95 system by CD95 ligand-independent stimuli (4,5). In nonhepatic cell types, hyperosmotic Fyn activation results in an increased phosphorylation of caveolin (9) and cortactin (10). The mechanisms of how CD95L and other proapoptotic stimuli trigger the oxidative stress response in hepatocytes remained unclear. Like tumor necrosis factor-␣, proapoptotic bile acids were suggested to induce oxygen radical formation by mitochondria (11)(12)(13)(14); however, this may represent a downstream conse-* This study was supported by the Deutsche Forschungsgemeinschaft through Sonderforschungsbereich 575 "Experimentelle Hepatologie" (Dü sseldorf). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence should be addressed: Universitä tsklinikum Dü sseldorf, Klinik fü r Gastroenterologie, Hepatologie und Infektiologie, Moorenstrasse 5, D-40225 Dü sseldorf. Tel.: 49-2118117569; Fax: 49-2118118838; E-mail: haeussin@uni-duesseldorf.de. quence but not the cause of CD95 activation. In phagocytes, reactive oxygen species are generated by an NADPH oxidase complex with its catalytic moiety gp91 phox , which is activated by assembly with regulatory proteins such as p47 phox , p67 phox , and Rac (15)(16)(17)(18). Recently, membrane oxidases similar to the phagocytic NADPH oxidase complex were found also in nonphagocytic cell types. These homologs of gp91 phox are called Nox and Duox, with the classical phagocytic gp91 phox being termed Nox 2. The isoform Nox 3 was found in fetal kidney, Nox 1 is predominantly expressed in the colon, and Nox 5 is predominantly expressed in spleen and fetal tissues (19,20). Duox 1 and 2 reflect high molecular mass gp91 phox homologs with an N-terminal peroxidase domain in addition to the C-terminal NADPH oxidase activity. Duox 1 was found primarily in the thyroid gland, whereas Duox 2 is also found in the intestine and the colon. Nox and Duox proteins were suggested to participate in signal transduction (20). Although it is dispensible for NADPH oxidase activity under cell-free conditions, p47 phox is critical for normal NADPH oxidase (gp91 phox , Nox 2) function, because p47 phox acts as an adapter protein, which facilitates stimulus-induced binding of p67 phox to the enzyme complex. Patients lacking p47 phox suffer from chronic granulomatous disease (21). In neutrophils, p47 phox is activated in response to inflammatory stimuli by multiple phosphorylations, which are thought to trigger intramolecular rearrangements that expose Src homology 3 domains for binding to proline-rich regions of other NADPH oxidase components in order to form the active enzyme complex (22,23). The present study was undertaken in order to determine the mechanisms of CD95Linduced transactivation of the EGFR and to investigate a potential contribution of NADPH oxidase isoforms for the rapid oxidative stress signal in response to proapoptotic stimuli. The data suggest that CD95L induces oxidative stress by activation of NADPH oxidase isoforms in a sphingomyelinase-and PKC-dependent way. CD95L-induced activation of NADPH oxidases leads to Yes activation, which in turn triggers ligand-independent EGFR activation as a crucial step in CD95-mediated cell death.
Cell Preparation and Culture-Hepatocytes, Kupffer cells, and hepatic stellate cells were isolated from livers of male Wistar rats, fed ad libitum with a standard diet, by a collagenase perfusion technique as described previously (1,2,24). Aliquots of 1.5 ϫ 10 6 hepatocytes were plated on collagen-coated 6-well culture plates (Falcon, Heidelberg, Germany) and cultured as published recently (1, 2) for 24 h, unless indicated otherwise, before the experiments were started. The viability of the hepatocytes was more than 95% as assessed by trypan blue exclusion. Immunocytochemical staining of the 24-h hepatocyte culture dishes with ED-2 or glial fibrillary acid protein as specific markers of Kupffer or hepatic stellate cells, respectively, revealed that the contamination of the hepatocyte culture with these nonparenchymal cells was less than 1%.
Mouse macrophage RAW 264.7 cells were cultured in Dulbecco's modified Eagle's medium at 37°C in 5% CO 2 , supplemented with 10% fetal calf serum, and 1% gentamycin. Rat peritoneal macrophages were prepared as recently published (25). Human monocytes were obtained from peripheral blood samples using BD Vacutainer CPT Cell Preparation Tubes (BD Biosciences) according to the manufacturer's recommendations.
Plasmid Construction-The nucleotide sequence encoding CD95 was amplified by reverse transcription (RT)-PCR using HepG2 mRNA and ligated into the pTOPO-TA vector (Invitrogen). After the introduction of restriction sites and removal of the stop codon by PCR, the fragment was inserted into pEYFP-N1 (Clontech) to create a fusion protein with YFP linked to the C terminus. The EGFR-CFP-construct was obtained by replacing GFP of F7 erb B1-EGFP (kindly provided by Prof. Arndt-Jovin, Ph.D. (27)) with CFP of pECFP-N1 (Clontech). All constructs were confirmed by sequencing (MWG Biotech, Ebersberg, Germany).
Immunoprecipitation-Hepatocytes were cultured on collagencoated culture plates (diameter, 10 cm; Falcon) at a density of 8 ϫ 10 6 cells/plate. They were harvested in lysis buffer as recently published (1,2). Equal protein amounts (200 g) of each sample were incubated for 2 h at 4°C with a polyclonal rabbit anti-CD95, rabbit anti-p47 phox , rabbit anti-Yes, or rabbit anti-Fyn antibody (dilution 1:100; Santa Cruz Biotechnology) in order to immunoprecipitate CD95, p47 phox , Yes, or Fyn. Then 10 l of protein A-agarose and 10 l of protein G-agarose (Santa Cruz Biotechnology) were added and incubated at 4°C overnight. Immunoprecipitates were washed 3 times as published recently (1,2) and then transferred to Western blot analysis as described above. Activation of p47 phox by serine phosphorylation was detected using an anti-phosphoserine antibody (28). The anti-phospho-Src family-Tyr 418 antibody was used in order to detect activating phosphorylation of Yes or Fyn in the respective immunoprecipitates (29), whereas Yes or Fyn/ EGFR association was detected using an anti-EGFR-antibody in the latter samples. EGFR, FADD, and caspase 8 association or tyrosine phosphorylation of the immunoprecipitated CD95 samples were detected by Western blot analysis using the respective antibodies (anti-EGFR, -FADD, -caspase 8, and -phosphotyrosine).
Subcellular Fractionation-Hepatocytes were cultured on collagencoated culture plates (diameter, 10 cm; Falcon) at a density of 8 ϫ 10 6 cells/plate. Cells were lysed in a buffer containing 10 mmol/liter Tris, 30 mmol/liter mannitol, and 10 mmol/liter CaCl 2 (pH 7,5). After centrifugation of the samples (5 min, 1200 ϫ g), the supernatants were subjected to ultracentrifugation (35 min, 40,000 g) in order to separate the plasma membrane fraction (pellet) from the cytosolic compartment including intracellular endomembranes (supernatant). This supernatant was then again subjected to ultracentrifugation (2 h, 100,000 ϫ g) in order to separate the intracellular endomembranes (pellet) from the cytosolic compartment (supernatant). The latter fractions then underwent Western blotting as described above for p47 phox , gp91 phox , GAPDH, and annexin II. GAPDH and annexin II served as markers for the cytosolic and the plasma membrane fraction, respectively. GAPDH was not detectable in the plasma membrane fraction, and annexin II was not detectable in the cytosolic fraction, indicating a high efficacy of separation.
CD95 and PKC Membrane Translocation-For determination of membrane surface trafficking of CD95 in primary rat hepatocytes, cells were cultured for 24 h on collagen-coated glass coverslips (diameter, 30 mm) in 6-well culture plates (Falcon). Permeabilized and nonpermeabilized cells were stained as published recently (1, 2), using a polyclonal rabbit anti-CD95 antibody (dilution 1:500 in PBS) and a secondary anti-rabbit Cy3-conjugated antibody. Cells were visualized using an Axioskop (Zeiss, Oberkochen, Germany), and pictures were taken with a 3CCD camera (Intas, Göttingen, Germany). Receptor membrane translocation was defined as the appearance of fluorescent spotting on the surface of the nonpermeabilized cells compared with the nonpermeabilized control cells (1,2). For each condition, a blinded observer scored at least 100 cells/independent experiment from at least three different cell preparations for CD95 membrane translocation.
CD95 translocation was also studied by detecting total CD95 amount in cytosolic and membrane fractions using Western blot analysis. For this, cells were plated on collagen-coated culture plates (ဧ 10 cm) at a density of 8 ϫ 10 6 cells/plate and were lysed as centrifuged (35 min, 40,000 ϫ g) in order to separate cytosolic from membrane fractions as recently published (1,2). Total CD95 amount in the latter fractions was then detected by Western blotting as described above.
In order to detect PKC membrane translocation, cells were cultured for 24 h on collagen-coated glass coverslips (diameter, 30 mm) in 6-well culture plates. After treatment for 1 min with either control medium or CD95L (100 ng/ml), cells were fixed for 3 min using methanol (Ϫ20°C) and then permeabilized using Triton X-100 (0.1% (v/v) in PBS, 10 min, room temperature). Cells were washed briefly with PBS (4°C) and then exposed to a rabbit anti-PKC antibody (1 h, 4°C, 1:200 in PBS), washed off, and then stained with an anti-rabbit Cy3-conjugated antibody (1 h, 4°C, 1:500 in PBS). Coverslips were mounted with diazabicyclo[2.2.2]octane 0.1% in glycine/PBS (9:1), and cells were visualized using a Leica TCS-NT confocal laser-scanning system with an argon-krypton laser on a Leica DM IRB inverted microscope (Bensheim, Germany).
Fluorescence Resonance Energy Transfer (FRET) Experiments-FRET was used in order to visualize CD95L-induced interactions between EGFR and CD95. For this purpose, Huh7 hepatoma cells were cotransfected with CD95-YFP and EGFR-CFP, as described in detail recently (3). Confocal pictures were taken using the LSM-510-META (Zeiss). All YFP/CFP cotransfections were detected using the META-scan, avoiding bleed-through of CFP in the YFP channel. CFP was excited with 405 nm, and YFP was excited with 514 nm (30). FRET efficiency was determined using LSM-Image-Examiner-3.1 software (Zeiss), and FRET pictures were normalized for the FRET efficiencies in the respective setting as indicated by the accompanying scale (FRET efficiency is given from blue/0 to red/255).
Detection of Reactive Oxygen Species (ROS)-Hepatocytes were seeded on collagen-coated 6-well culture plates (Falcon) and cultured 24 h for inhibitor experiments or 4 days for protein knockdown experiments, respectively. Cells were incubated with PBS containing 5 mol/ liter of CM-H 2 DCFDA for 30 min at 37°C and 5% CO 2 . The oxidation of the nonfluorescent 2Ј,7Ј-dichlorodihydrofluorescein diacetate, also known as dichlorofluorescin diacetate, to the highly fluorescent 2Ј,7Јdichlorofluorescein (DCF) is commonly used to detect the overall generation of reactive oxygen intermediates (31). CM-H 2 DCFDA is a chloromethyl derivative of H 2 DCFDA that exhibits a high retention in living cells. CM-H 2 DCFDA passively diffuses into cells, where its acetate groups are cleaved by intracellular esterases, and its thiol-reactive chloromethyl group reacts with intracellular glutathione and other thiols. Subsequent oxidation yields a fluorescent adduct that is trapped inside the cell, thus facilitating long term studies (32).
In order to detect ROS formation at the single cell level, measurements were performed with hepatocytes cultured on glass coverslips (ဧ 30 mm) using an inverted fluorescence microscope (Axiovert; Zeiss). For fluorescence recording, the coverslips were mounted with PBS at 37°C, equilibrated with room atmosphere. Cells were excited at 488 nm at a rate of 2 Hz by a monochromator, and emission was measured at 515-565 nm using a CCD camera as provided by the QuantiCell 2000calcium imaging setup (VisiTech, Sunderland, UK). Emission intensity of unstimulated cells was set to 1, and increased emission intensity after the addition of CD95 ligand is expressed as relative increase compared with unstimulated emission intensity.
In order to detect ROS in a hepatocyte population, after a CM-H 2 DCFDA loading period, cells were supplemented again with culture medium. Cells were then exposed to CD95L for the indicated time period. Then cells were washed briefly using ice-cold PBS, and cells were lysed in 0.1% Triton X-100 (v/v) dissolved in aqua bidest. Lysates were centrifuged immediately (10,000 ϫ g, 4°C, 1 min), and fluorescence of the supernatant was measured at 515-565 nm using a luminescence spectrometer LS-5B (PerkinElmer Life Sciences) at 488-nm excitation wavelength.
Since the NADPH oxidase inhibitor apocynin itself was recently reported to induce ROS in vascular fibroblasts (33), the effect of apocynin on ROS generation in hepatocytes was measured. At the dose used in this study (300 mol/liter), apocynin by itself increased 2Ј,7Ј-dichlorodihydrofluorescein diacetate fluorescence negligibly from 1.000 to 1.015 Ϯ 0.002 in hepatocytes (n ϭ 10).
Lipid Extraction and High Performance Thin Layer Chromatography-Cells were harvested at the indicated time points by scraping the cells off of the plate on ice. Pellets were washed and sonicated. Quantification of lipids was done using 500 g of protein for Folch extraction (34). Analysis of ceramides included a mild alkaline hydrolysis. The lower phase of Folch extraction was evaporated under nitrogen. The lipids were dissolved in chloroform/methanol (2:1, v/v).
Samples and standards were separated on silica gel high performance thin layer chromatography plates (20 ϫ 10 cm; Merck 60F 254s) prewashed for 60 min in 2-propanol and dried 30 min at 120°C. Samples and standards were applied to the TLC plates using a CAMAG Linomat IV (CAMAG, Berlin, Germany). For determination of ceramides, samples were separated using an automated multiple development procedure on an automated multiple development 2 device (CA-MAG). This procedure consisted of seven repeated developments of the chromatogram using a stepwise elution gradient with methanol, dichloromethane, and n-hexane (methanol/dichloromethane/n-hexane: 100/ 0/0; 10/90/0; 9/91/0; 8/92/0; 3/97/0; 2/98/0; 0/0/100) (35) on a CAMAG automated multiple development 2 device as described earlier (36). Visualization of separated bands was done by postchromatic derivatization after dipping in a manganese chloride solution according to Grether-Beck et al. (36) in an automated dipping device (CAMAG, Berlin, Germany). After heating the plate for 10 min at 120°C in a temperature-controlled oven the plate was dried and scanned using a CAMAG TLC Scanner II and CATS software. Quantification was done by absorption at 550 nm with a plot of peak area versus weight spotted for a series of standards using a second order polynomal calibration with four standard mixes in the range 50 -1000 ng.
Detection of Apoptosis-TUNEL of fluorescein isothiocyanate-conjugated deoxyuridine triphosphate was performed as described recently (1, 2). The number of apoptotic cells was determined by counting the percentage of fluorescein-positive cells. At least 100 cells from three different cell preparations were counted for each condition. Cells were visualized on an Axioskop (Zeiss, Oberkochen, Germany).
Statistics-Results from at least three independent experiments are expressed as means Ϯ S.E. n refers to the number of independent experiments. Results were analyzed using Student's t test; p Ͻ 0.05 was considered statistically significant.

Expression of NADPH Oxidase Isoforms in Rat Hepatocytes
and Nonparenchymal Liver Cells-The mRNA expression of NADPH oxidase isoforms was studied by real time PCR in rat hepatocytes, Kupffer cells, and quiescent and activated hepatic stellate cells (HSC). As shown in Fig. 1A, hepatocytes expressed mRNAs for Nox 1 and 4, Duox 1 and 2, and the classical NADPH oxidase isoform Nox 2 (gp91 phox ). Nox 3 mRNA was not detectable in rat hepatocytes. Nox and Duox mRNA expression in rat hepatocytes did not differ from that found in rat liver macrophages (Kupffer cells) and activated HSC, whereas quiescent HSC showed a minor signal for Duox 1 and no detectable mRNA signal for Nox 3 and 4. However, the primer pairs used to detect Nox 3 mRNA yielded a positive signal in fetal rat kidney extracts (Fig. 1A) (i.e. a tissue known to express Nox 3) (19,20). Rat hepatocytes, Kupffer cells, and HSC expressed the mRNA and protein for the regulatory subunit p47 phox , as shown by RT-PCR and Western blot, respectively (Fig. 1, A and  B). In contrast, human hepatoma cell line 7 (Huh7) cells did not express Nox 2 mRNA, whereas RT-PCR revealed Nox 1, 3, and 4 as well as Duox 1 and 2 and p47 phox mRNA expression (data not shown). As further shown in Fig. 1C, Nox 2 (gp91 phox ) protein was not only found in rat Kupffer cells but also in rat hepatocytes and hepatic stellate cells. RAW mouse and rat peritoneal macrophages served as a positive control. Rat and murine Nox 2 (gp91 phox ) protein is known to represent a deglycosylated protein with a molecular mass of approximately 55 kDa (37), whereas the human protein is highly glycosylated with a molecular mass of approximately 80 -120 kDa (37). Such differences in electrophoretic mobility between murine and human Nox 2 (gp91 phox ) were also found (Fig. 1C). The presence of p47 phox and Nox 2 in 24-h cultured rat hepatocytes was also confirmed immunocytochemically (data not shown). In line with the immunocytochemical findings, subfractionation studies on rat hepatocytes revealed an almost exclusive p47 phox expression in the cytosol (Fig. 1D). Only a small amount of gp91 phox was detected in the membrane fraction, with the majority of the gp91 phox protein being cytosolic (Fig. 1D). These findings differ from those reported for macrophages, which exhibit a predominant location of the NADPH oxidase gp91 phox subunit at the plasma membrane, whereas the p47 phox subunit is primarily cytosolic (38). However, in hepatocytes, cytosolic gp91 phox was most likely contained in very light membranes, because centrifugation of the cytosolic preparation at 100,000 ϫ g for 2 h resulted in the sedimentation of immunoreactive gp91 phox , whereas the cytosolic marker enzyme GAPDH remained soluble (Fig. 1D). This finding indicates that hepatocytic gp91 phox may be contained in membrane vesicles inside the cytosol, as one might expect for a transmembrane integral protein (15)(16)(17)(18).
Activation of NADPH Oxidase Isoforms by CD95L-p47 phox is known as an adapter molecule, which is essential for activation of Nox 2 NADPH oxidase complex (39). p47 phox activation was shown to involve serine phosphorylation, which is also required for activation of the NADPH oxidase complex in macrophages and neutrophils (22,40). It is not yet established whether p47 phox is also required for the in vivo activation of other Nox/Duox isoenzymes. As shown in Fig. 2A, CD95L induced a rapid Ser phosphorylation of p47 phox in rat hepatocytes, which persisted for more than 60 min.
In order to address a potential role of p47 phox and Nox/Duox enzymes for the generation of oxidative stress in response to CD95L, single cell fluorescence measurements were performed with DCFDA-loaded hepatocytes. DCFDA is frequently used as a probe for measurement of overall oxidative stress (31). Exposure of 24-h cultured rat hepatocytes to CD95L (100 ng/ml) induced within 1 min an oxidative stress response, as detected by DCFDA fluorescence (Fig. 2B, Table II). This CD95L-induced oxidative stress response was not only sensitive to Nacetylcysteine (1) but was also largely abolished by apocynin (Fig. 2B), diphenyleneiodonium, and neopterin (Table II) (i.e. known inhibitors of NADPH oxidase). These findings suggest a role of NADPH oxidase isoforms in triggering the CD95Linduced oxidative stress response. However, one has to keep in mind that these inhibitors may have unspecific side effects.
Therefore, the potential role of Nox/Duox enzymes in generating the rapid CD95L-induced oxidative stress response was further studied in 4-day cultured rat hepatocytes following transfection with nonsense or p47 phox antisense oligonucleotides. 4 days after transfection with the antisense but not the nonsense oligonucleotides, a substantial down-regulation of p47 phox protein was achieved (Fig. 3A). As shown in Fig. 3B, p47 phox protein knockdown, but not transfection with the nonsense oligonucleotides resulted in a strong inhibition of the CD95L-induced increase of DCFDA fluorescence. These data suggest that Nox 2 and/or other p47 phox -dependent NADPH oxidase isoforms are involved in the rapid oxidative stress response, which is triggered by CD95L.
Mechanism of CD95L-induced Activation of NADPH Oxidase Isoforms-Inhibitor studies were performed in order to gain insight into the upstream signaling events leading to CD95Linduced p47 phox phosphorylation and ROS formation. As shown in Fig. 4A, CD95L-induced p47 phox -Ser phosphorylation was sensitive to AY9944 and desipramine (i.e. inhibitors of sphingomyelinase) (41,42), suggesting that sphingomyelinase activation is upstream of CD95L-induced p47 phox phosphorylation. As shown by immunocytochemistry, acidic sphingomyelinase was not only found inside the hepatocytes in a presumably lysosomal compartment but also at the plasma membrane (Fig.  4B). In line with a CD95L-induced activation of sphingomyelinase, CD95L produced within 30 s a significant increase in , and hepatic stellate cells (HSC) were isolated as described under "Experimental Procedures." Hepatocytes were cultured for 1 or 4 days (PC d1/d4), Kupffer cells were cultured for 2 days (KC), and hepatic stellate cells were kept for 1 or 14 days in culture (HSC d1/d14), respectively. Thereafter, cells were harvested, and mRNA expression of Nox 1-4, Duox 1 and 2, and the NADPH oxidase regulatory subunit p47 phox was detected by RT-PCR (A). mRNA isolated from fetal rat kidney served as a positive control for Nox 3 (19,20). ␤-Actin mRNA expression served as loading control (n ϭ 3). All PCR products were confirmed by sequencing after isolation from agarose gels. B, the NADPH oxidase regulatory subunit p47 phox was detected by Western blotting in rat hepatocytes. GAPDH expression served as a loading control (n ϭ 3). C, gp91 phox (i.e. Nox 2) was also detected in rat hepatocytes, Kupffer cells, and hepatic stellate cells. Rat peritoneal macrophages and the mouse RAW macrophages served as positive controls. In addition, human hepatoma cell line 7 (Huh7) and hepatoma cell line G2 (HepG2) were detected for gp91 phox protein expression. Here, human blood monocytes served as a positive control. Whereas murine cells revealed a band at ϳ55 kDa, which probably reflects the unglycosylated gp91 phox , human monocytes showed a smear at ϳ85-120 kDa reflecting the gylcosylated protein, as reported previously (37). In contrast to rat hepatocytes, human hepatoma cells (i.e. Huh7 and HepG2) showed no gp91 phox expression. D, plasma membrane and cytosolic fractions of PC were obtained as described under "Experimental Procedures" (i.e. by centrifugation at 40,000 ϫ g for 35 min) and detected for gp91 phox (Nox 2) and the regulatory subunit p47 phox . gp91 phox (ϳ55 kDa) revealed a predominant expression in the supernatant (cytosol) and only small amounts of gp91 phox in the pellet (i.e. plasma membrane fraction). Higher molecular mass forms of gp91 phox were not detectable. After ultracentrifugation (100,000 ϫ g for 2 h) of this supernatant, gp91 phox was found in the pellet, whereas p47 phox and GAPDH were still found in the supernatant. These findings suggest that gp91 phox but not p47 phox is probably membrane-associated and contained in intracellular membrane vesicles. GAPDH and annexin II served as markers for the cytosolic and the plasma membrane fractions, respectively. GAPDH was not detectable in the membrane fraction, and annexin II was not detectable in the cytosolic fraction, indicating a high efficacy of separation. ceramide levels, which lasted for more than 30 min (Fig. 4C). The CD95L (100 ng/ml)-induced increase in ceramide levels within 1 min was 1.87 Ϯ 0.06-fold (n ϭ 5) compared with unstimulated control, which was arbitrarily set to 1. When AY9944 (5 mol/liter) or desipramine (5 mol/liter) was preincubated for 30 min prior to CD95L (100 ng/ml) addition for 1 min, this increase of ceramide levels was largely abolished (Fig.  4D). These findings suggest the involvement of sphingomyelinase-derived ceramide in triggering CD95L-induced p47 phox serine phosphorylation. In line with this, the addition of C 6 -and C 16 -ceramide but not of the inactive C 6 -dihydroceramide did induce p47 phox Ser phosphorylation (Fig. 4A).
Because several protein kinases, among them some PKC isoforms were reported to induce serine phosphorylation of p47 phox (40), the role of various PKC isoforms in triggering CD95L-induced p47 phox serine phosphorylation was investi-gated. Rat hepatocytes are known to express the PKC isoforms ␣, ␤ II , ␦, ⑀, and (43,44). Conventional, Ca 2ϩ -dependent PKC isoforms ␣, ␤, and ␥ and the novel Ca 2ϩ -independent isoforms ␦ and ⑀ are sensitive to inhibition by Gö6850, whereas Gö6976 and Ro-32-0432 exhibit more specificity on the classical PKC isoforms (45,46). None of these inhibitors significantly affects the PKC isoform (45,46). Neither Gö6850, Gö6976, nor Ro-32-0432, which were recently shown to inhibit phorbol 12myristate 13-acetate-induced effects on bile secretion in rat liver (47), prevented CD95L-induced Ser phosphorylation of p47 phox (Fig. 4A). However, specific inhibition of PKC by a synthetic PKC-pseudosubstrate abolished p47 phox phosphorylation in response to CD95L. Also chelerythrine, an unspecific PKC inhibitor with some effect on PKC, blunted CD95L-induced p47 phox phosphorylation. On the other hand, specific inhibitors of the PKC isoforms ⑀, , and were ineffective (Fig.  4A). The inhibitory PKC substrate also blocked p47 phox Ser phosphorylation in response to C 6 -ceramide, suggesting that PKC is involved in the signaling of ceramide toward p47 phox (Fig. 4A) (48). In line with a CD95L-induced activation of FIG. 2. CD95 ligand-induced p47 phox -Ser phosphorylation and generation of ROS in rat hepatocytes. A, hepatocytes were cultured for 24 h and then exposed to CD95L (100 ng/ml) for the time periods indicated. p47 phox was immunoprecipitated as described under "Experimental Procedures" and detected for serine phosphorylation by Western blotting and densitometric analysis. Total p47 phox served as a loading control. CD95L induced within 1 min significant p47 phox -serine phosphorylation (p Ͻ 0.05; *, n ϭ 6) suggestive of p47 phox activation. B, hepatocytes were cultured for 24 h and loaded with 5 mol/liter CM-H 2 DCFDA as described under "Experimental Procedures." When indicated, apocynin (300 mol/liter) was preincubated for 30 min prior to the CD95L (100 ng/ml) addition. Representative single cell fluorescence recordings are shown (n ϭ 10 recordings for each condition from three different cell preparations). CD95L induces a rapid ROS response, which is inhibited by apocynin, suggestive of an involvement of NADPH oxidase.

TABLE II
CD95L-induced ROS generation is mediated by sphingomyelinase and PKCz Hepatocytes were cultured for 24 h and then loaded with 5 mol/liter CM-H 2 DCFDA. Cells were then exposed to CD95L (100 ng/ml), C 6 -(10 mol/liter), dihydro-C 6 -(10 mol/liter), or C 16 -ceramide (10 mol/liter) for 1 min. When indicated, AY9944 (5 mol/liter), desipramine (5 mol/ liter), PKC inhibitor (100 mol/liter), chelerythrine (20 mol/liter), diphenyleneiodonium (10 mol/liter), apocynin (300 mol/liter), or neopterin (100 mol/liter) was preincubated for 30 min. ROS generation was measured using a luminescence spectrometer as described under "Experimental Procedures." ROS generation under untreated control conditions was arbitrarily set to 1. In another set of experiments, hepatocytes were cultured for 96 h under control conditions or treated with either nonsense or p47 phox -antisense oligonucleotides. Data are given as means Ϯ S.E. (n ϭ 3 different preparations). CD95L and C 6 -and C 16 -ceramide induced within 1 min a significant generation of ROS, whereas C 6 -dihydroceramide was ineffective. CD95L-induced ROS generation was significantly reduced by inhibition of sphingomyelinase (AY9944 or desipramine), PKC (PKC inhibitor or chelerythrine), or NADPH oxidase (diphenyleneiodonium, apocynin, and neopterin). Also, C 6 -ceramide-induced ROS generation was sensitive to PKC or NADPH oxidase inhibition. In 96-h cultured hepatocytes, p47 phox protein knockdown largely abolished CD95L-induced ROS generation. 24 PKC, this PKC isoform was translocated to the plasma membrane of hepatocytes in response to CD95L (Fig. 4E). As shown in Fig. 5 and Table II, both, inhibition of sphingomyelinase by AY9944 or desipramine and inhibition of PKC by chelerythrine or the specific PKC substrate also strongly inhibited the rapid ROS response, which is normally induced by CD95L within 1 min. PKC inhibition also inhibited ROS formation in response to C 6 -ceramide (Table II). These findings suggest that activation of sphingomyelinase and subsequent ceramide-induced PKC activation are upstream events in the CD95L-induced phosphorylation of p47 phox and activation of NADPH oxidase isoforms in rat hepatocytes.
Nox-dependent Oxidative Stress and Activation of Src Family Kinases in Response to CD95L-Next, the role of CD95L-induced NADPH oxidase activation for activation of the CD95 system in response to CD95L was addressed. Recent studies on CD95 activation by hyperosmolarity (4) or hydrophobic bile acids (5) demonstrated a rapid NAC-sensitive activation of the Src family kinase Yes as an upstream event leading to CD95 activation. As shown in Fig. 6, also the addition of CD95L to 1-day cultured hepatocytes resulted within 1 min in an activation of the Src kinase family member Yes, followed by c-Src and Fyn activation, as detected by their phosphorylation at Tyr 418 (Fig. 6, A and B). No significant change in the phosphorylation of c-Src-Tyr 529 or Lck-Tyr 505 (not shown) was found. As expected from the known inhibitor profiles, which are characteristic for the different Src kinase family members (49 -51) (see also Refs. 4 and 5), CD95L-induced Yes activation was sensitive to SU6656, but not to PP-2, whereas Fyn and Src activation was sensitive to both SU6656 and PP-2. CD95L-induced Yes and the upstream p47 phox activation were already observed at CD95L concentrations of 5 ng/ml (Fig. 6C). The similarity of dose dependences (Fig. 6C) suggests that CD95L-induced activation of NADPH oxidases may be upstream of Yes activation. NAC inhibited CD95L-induced Yes but not Fyn and c-Src activation (Fig. 6, A and B). This NAC sensitivity and the finding that Yes is also rapidly activated by externally added H 2 O 2 (Fig. 6C) again suggest that a CD95L-induced oxidative stress response may underlie CD95L-induced Yes activation. Therefore, studies on the role of CD95L-induced Nox/Duox activation for CD95L-induced Yes activation were performed. As shown in Fig. 6, D and E, AY9944, desipramine, and the inhibitory PKC substrate (i.e. inhibitors of CD95L-induced p47 phox serine phosphorylation and of the early CD95L-induced ROS response) (Fig. 4A, Table II) blunted CD95L-induced Yes activation. Inhibition of CD95L-induced Yes activation was also observed after p47 phox protein knockdown (see below) ( Fig. 10). Further, as shown in Fig. 6F, Yes activation in response to CD95L was blunted in the presence of DPI, apocynin, and neopterin. These findings suggest that the early CD95L-induced ROS response occurs via NADPH oxidase activation and represents an upstream signal for CD95L-induced Yes activation.
Nox-dependent Oxidative Stress and JNK Activation in Response to CD95L-As shown recently (1), CD95L triggers within 5-10 min an activation of JNK. CD95L-induced JNK activation was insensitive to SU6656 (Fig. 7), PP-2, or herbimycin A (not shown), indicating that Yes, Fyn, or c-Src are not involved as upstream events. However, CD95L-induced JNK activation was largely inhibited in the presence of NAC, DPI, apocynin, and neopterin ( Fig. 7) or after p47 phox knockdown (see below) ( Fig. 10). Likewise, inhibition of sphingomyelinase by AY9944, desipramine or of PKC by its inhibitory substrate or chelerythrine blunted CD95L-induced JNK activation (Fig.  7). These data suggest that not only CD95L-induced Yes activation but also JNK activation involves the action of reactive oxygen species generated by NADPH oxidase isoforms. The data further suggest that Yes is not upstream of CD95L-induced JNK activation, because the latter was not abolished after Yes inhibition by SU6656.
Yes Triggers EGFR Activation in Response to CD95L-Previous studies on ligand-independent activation of the CD95 system in rat hepatocytes by hydrophobic bile acids or hyperosmolarity revealed a Yes-dependent activation of the EGFR as another upstream event leading to CD95 activation (4,5). Therefore, experiments were performed in order to study Yes/ EGFR interactions in response to CD95L. As shown in coimmunoprecipitation studies, CD95L induced within 1 min an association between Yes and the EGFR (Fig. 8A), which gradually disappeared thereafter. No association between EGFR and Fyn or c-Src was found (data not shown). CD95L-induced Yes/EGFR association was inhibited by SU6656, indicating the requirement of active Yes for this association (Fig. 6F). In line with this, also NAC, DPI, apocynin, and neopterin, which inhibit CD95L-induced Yes activation, blunted Yes/EGFR asso-FIG. 3. p47 phox protein knockdown inhibits CD95 ligand-induced generation of ROS in rat hepatocytes. Hepatocytes were cultured for up to 4 days in order to induce p47 phox protein knockdown using antisense oligonucleotides as described under "Experimental Procedures." Transfection with nonsense oligonucleotides served as control. A, p47 phox protein expression; p47 phox and CD95 protein expression were detected by Western blot. A 4-day culture period with the antisense oligonucleotide resulted in a marked down-regulation of p47 phox protein expression (n ϭ 7). CD95 expression served as control. B, CD95L-induced ROS generation. Cells were loaded with 5 mol/liter CM-H 2 DCFDA as described under "Experimental Procedures" and exposed to CD95L (100 ng/ml). Representative single cell fluorescence recordings are shown (n ϭ 10 recordings for each condition from three different cell preparations). CD95L induces a rapid ROS response, which is inhibited by p47 phox protein knockdown. ciation (Fig. 6F). This was also seen after inhibition of sphingomyelinase or PKC (Fig. 6, D and E). CD95L-induced Yes/ EGFR association was unaffected by JNK inhibition, PP-2, herbimycin, and inhibition of EGFR-tyrosine kinase activity by AG1478 (Fig. 6F).
In line with previous data (1), CD95L induced a rapid activation of the EGFR. Here, CD95L increased EGFR phosphorylation at Tyr 845 and Tyr 1173 (i.e. known Src-kinase (52) and autophosphorylation sites (53)), respectively, but not at Tyr 1045 (Fig. 8B), which is a known Cbl docking site required for EGFR internalization (54). CD95L-induced EGFR phosphorylation at Tyr 1173 but not at Tyr 845 was sensitive to inhibition by AG1478, whereas SU6656 inhibited EGFR phosphorylation at both sites (Fig. 8B). These findings suggest that Yes activates the EGFR through phosphorylation at Tyr 845 and subsequent autophosphorylation at Tyr 1173 .
Overall, CD95L-induced EGFR-tyrosine phosphorylation was sensitive to SU6656, NAC, DPI, apocynin, and neopterin but insensitive to PP-2, herbimycin, JNK inhibitor, or AG1478 (Fig. 6F) and was also inhibited by AY9944, desipramine, and the PKC substrate (Fig. 6, D and E). This inhibitor profile and the finding that Yes, but not Fyn or c-Src, rapidly associates with the EGFR in response to CD95L suggest that CD95Linduced EGFR activation is mediated by Yes, as it was shown recently for EGFR activation in response to hyperosmolarity or proapoptotic bile acids (4,5). EGFR/CD95 Interactions in Response to CD95L-CD95L, hyperosmolarity, and proapoptotic bile salt were shown to induce an EGFR-dependent activation of the CD95 system through EGFR-catalyzed CD95-tyrosine phosphorylation, which was identified as a crucial step for CD95 trafficking to the membrane, formation of the death-inducing signaling complex (DISC), and apoptosis induction (1, 2). These proapoptotic stimuli trigger within 30 min a JNK-dependent intracellular FIG. 4. CD95 ligand-induced p47 phox -serine phosphorylation is mediated by sphingomyelinase-mediated ceramide generation and subsequent PKC activation. A, inhibitor sensitivity of CD95L-and ceramide-induced p47 phox -Ser phosphorylation; 24-h cultured hepatocytes were exposed to CD95L (100 ng/ml) or C 6 -(10 mol/liter), dihydro-C 6 -(10 mol/liter), or C 16 -ceramide (10 mol/liter) for 1 min. When indicated, AY9944 (5 mol/liter), desipramine (5 mol/liter), PKC inhibitor (100 mol/liter), or chelerythrine (20 mol/liter) were preincubated for 30 min. In another set of experiments, chelerythrine (20 mol/liter), Gö6850 (10 mol/liter), Ro-32-0432 (500 nmol/liter), Gö6976 (5 mol/liter), and PKC⑀, PKC-, PKC-, and PKC-inhibitory pseudosubstrates (100 mol/l, each) were preincubated for 30 min before CD95L (100 ng/ml) addition. p47 phox was immunoprecipitated as described under "Experimental Procedures" and detected for serine phosphorylation by Western blotting (n ϭ 4). Total p47 phox served as a loading control. CD95L and C 6 -and C 16 -ceramide induce rapid p47 phox -serine phosphorylation, which was sensitive to inhibitors of sphingomyelinase and PKC, whereas C 6 -dihydroceramide was ineffective to induce p47 phox -serine phosphorylation. B, expression of acidic sphingomyelinase in rat hepatocytes. Acidic sphingomyelinase is immunocytochemically found in hepatocytes inside the cell and at the cell membrane. Acidic sphingomyelinase expression was also detected by Western blotting in hepatocytes (PC), Kupffer cells (KC), and hepatic stellate cells (HSC). GAPDH served as loading control (n ϭ 3). C, CD95 ligand-induced ceramide generation. Hepatocytes were cultured for 24 h and then exposed to CD95L (100 ng/ml). Ceramide was measured as described under "Experimental Procedures." CD95L significantly increases ceramide levels in 24-h cultured hepatocytes (p Ͻ 0.05; #, n ϭ 5). D, inhibitor sensitivity of CD95L-induced ceramide generation. When indicated, AY9944 (5 mol/liter) or desipramine (5 mol/liter) was preincubated for 30 min before CD95L (100 ng/ml) or control medium, respectively, was added for another 1 min. Inhibition of sphingomyelinase by AY9944 or desipramine significantly inhibited the CD95L-induced increase in ceramide level (p Ͻ 0.05; *, n ϭ 3). E, CD95 ligand-induced PKC translocation to the plasma membrane. 24-h cultured rat hepatocytes were exposed to either control medium or CD95L (100 ng/ml) for 1 min. Then cells were stained for PKC as described under "Experimental Procedures." Whereas under control conditions PKC mainly shows an intracellular immunostaining, CD95L induced an enrichment of PKC immunoreactivity at the plasma membrane suggestive of a CD95L-induced PKC membrane translocation (n ϭ 3). association of the EGFR and CD95, which is followed by an EGFR-catalyzed CD95-tyrosine phosphorylation (1,2). Thereafter, translocation of the EGFR⅐CD95 protein complex to the plasma membrane and DISC formation occur and were assessed 2 h after the addition of the proapoptotic stimulus (1-3). As shown in Fig. 8A, about 30 min after CD95L addition, the Yes⅐EGFR complex starts to disappear, and EGFR increasingly associates with the CD95. No coimmunoprecipitation of Yes and CD95 was detectable, indicating that EGFR dissociates from Yes prior to its association with CD95. In line with previous data (1, 2), EGFR/CD95 association was JNK inhibitorsensitive and was followed by an EGFR-catalyzed CD95-tyrosine phosphorylation, which was sensitive to inhibition by AG1478 and genistein (Fig. 6F). This CD95-tyrosine phosphorylation was recently shown to be essential for CD95 translocation to the plasma membrane and recruitment of FADD and caspase 8 (i.e. formation of the DISC) (1-3). All of these events were strongly blunted by SU6656 but not by PP-2 (Fig. 6F). Also, inhibitors of NADPH oxidases (i.e. DPI, apocynin, and neopterin) (Fig. 6F) and inhibition of sphingomyelinase or PKC as upstream events in the CD95L-induced ROS formation blunted these responses (Fig. 6D). These inhibitors also strongly blunted the previously described (1) CD95 translocation to the plasma membrane in response to CD95L (Table III). A similar inhibitor profile of CD95L-induced membrane translocation as observed in rat hepatocytes was also found in Huh7 cells that were transfected with CD95-YFP (Table III). These data strongly support the view that CD95L-induced activation of NADPH oxidase isoenzymes is an important upstream event in the activation of the CD95 system including DISC formation.
In order to demonstrate the requirement of sphingomyelinase, PKC, and NADPH oxidases for the CD95L-induced EGFR/CD95 interactions in the living cell, FRET experiments were performed in Huh7 cells (26). For this purpose, Huh7 cells, which exhibit almost no endogenous CD95 expression (55), were cotransfected with EGFR-YFP and CD95-CFP, as described recently (3). In line with previous data (3) and as shown in Fig. 9, CD95L induced within 30 min a cytosolic FRET signal (Fig. 9A), suggestive of intracellular EGFR/CD95 association. After 2 h, the EGFR⅐CD95 complex was targeted to the plasma membrane (Fig. 9B), as evidenced by a strong FRET signal in the plasma membrane. Prevention of CD95Linduced NADPH oxidase activation by apocynin, sphingomyelinase inhibition (AY9944), or PKC inhibition by its inhibitory substrate prevented both cytosolic EGFR/CD95 association and targeting of the protein complex to the plasma membrane (Fig. 9, A and B, and Table IV).
Role of NADPH Oxidases, Yes, and EGFR in CD95L-induced Hepatocyte Apoptosis-In order to substantiate the roles of NADPH oxidases, Yes, and the EGFR in CD95L-induced hepatocyte apoptosis, knockdown experiments after transfection with respective nonsense and antisense oligonucleotides were performed. As shown in Figs. 3A and 10, after 4 days of culture, a significant down-regulation of p47 phox , Yes, and EGFR was obtained, when rat hepatocytes were transfected with corresponding antisense oligonucleotides. As shown in Fig. 10, p47 phox knockdown strongly inhibited CD95L-induced Yes, JNK, and EGFR activation as well as EGFR/CD95 association, CD95-tyrosine phosphorylation, CD95 translocation to the plasma membrane, and DISC formation. Similar results were obtained after Yes knockdown except for a preserved JNK activation and EGFR/CD95 association in response to CD95L (Fig. 10). This is an expected finding, because Yes knockdown does not affect the CD95L-induced oxidative stress response and, accordingly, JNK activation. The latter was shown to be required for EGFR/CD95 association, which occurs irrespective of the EGFR activation status (1,2). Also, EGFR knockdown resulted in an inhibition of CD95L-induced CD95-tyrosine phosphorylation, CD95 translocation to the plasma membrane, and DISC formation, whereas CD95L-induced activation of Yes and JNK were preserved (Fig.  10). Also, when the TUNEL assay was used as apoptotic readout, knockdown of either p47 phox , Yes, or EGFR significantly decreased CD95L-induced hepatocyte apoptosis by about 50%. Also, DPI, apocynin, and neopterin blunted hepatocyte apoptosis in response to CD95L, as did AY9944, desipramine, or the PKC inhibitor (Table III). These findings underline the important role of an NADPH oxidase-derived ROS signal in CD95L-induced hepatocyte apoptosis.

CD95L-induced Early Oxidative Stress
Response-In rat hepatocytes, proapoptotic stimuli, such as CD95L (1), hydrophobic bile acids (2), or hyperosmotic cell shrinkage (1) trigger a rapid oxidative stress response, as assessed by an increase in 2Ј,7Ј-dichlorodihydrofluorescein diacetate fluorescence. Inhibition of this oxidative stress response by antioxidants was accompanied by an inhibition of hepatocyte apoptosis (1, 2, 5), suggesting its requirement for CD95-dependent apoptosis induction. Mitochondria have repeatedly been shown to be a source of oxidative stress in response to proapoptotic stimuli (56,57), including ligands of the tumor necrosis factor receptor family (11,58). However, as shown in the present study, the almost immediate CD95L-induced ROS response apparently FIG. 5. CD95 ligand-induced ROS generation requires sphingomyelinase and PKC. Hepatocytes were cultured for 24 h and loaded with 5 mol/liter CM-H 2 DCFDA as described under "Experimental Procedures." When indicated, AY9944 (5 mol/liter) or PKC inhibitor (100 mol/liter) was preincubated for 30 min prior to the CD95L (100 ng/ml) addition. Representative single cell fluorescence recordings are shown (n ϭ 10 recordings for each condition from three different cell preparations). CD95L induces a rapid ROS response, which is inhibited by AY9944 and the PKC inhibitor, suggestive of an involvement of a sphingomyelinase and PKC in CD95L-induced ROS generation.
FIG. 6. CD95 ligand-induced activation of Src family kinases and activation of the CD95 system in rat hepatocytes. Hepatocytes were cultured for 24 h and then exposed to CD95L (100 ng/ml) for the given time periods. When indicated, AY9944 (5 mol/liter), desipramine (5 mol/liter), PKC inhibitor (100 mol/liter) or chelerythrine (20 mol/liter), diphenyleneiodonium (10 mol/liter), apocynin (300 mol/liter), D-(ϩ)-neopterin (100 mol/liter), NAC (30 mmol/liter), SU6656 (10 mol/liter), PP-2 (10 mol/liter), herbimycin A (1 mol/liter), L-JNKI1 (5 mol/liter), genistein (100 mol/liter), AG1478 (5 mol/liter) were preincubated for 30 min. The Src family kinases Yes and Fyn were immunoprecipitated (IP) as described under "Experimental Procedures," and their activating phosphorylation at position Tyr 418 was detected by Western blotting. c-Src phosphorylation at position Tyr 418 was detected by Western blotting using phosphospecific antibodies. A and B, CD95L induced within 1 min activation of Yes, which was sensitive to SU6656 and NAC but not to PP-2. Fyn and Src phosphorylation was sensitive to SU6656 and PP-2, whereas NAC was ineffective (n ϭ 3; A). Densitometric analysis of Yes, Fyn, and Src phosphorylation at position Tyr 418 showed a significant inhibition of CD95L-induced Yes activation by NAC (*, p Ͻ 0.05) but not of CD95L-induced Fyn and Src activation, respectively (p Ͼ 0,05; B). C, concentration dependence of CD95L-or H 2 O 2 -induced Yes-Tyr 418 phosphorylation as assessed 1 min after CD95L or H 2 O 2 addition, respectively (n ϭ 3). CD95L-induced Yes phosphorylation exhibited a similar concentration dependence as observed for CD95L-induced p47 phoxserine phosphorylation (n ϭ 3). D-F, Yes, EGFR, and CD95 were immunoprecipitated as described under "Experimental Procedures" and analyzed by Western blotting. Activating Yes-Tyr 418 phosphorylation, YES/EGFR association, and EGFR-tyrosine phosphorylation (EGFR-Tyr-P) were detected 1 min after CD95L addition; EGFR/CD95 association and CD95-tyrosine phosphorylation (CD95-Tyr-P) were detected after 60 min of CD95L exposure; and caspase 8/CD95 and FADD/CD95 association were detected 3 h after CD95L addition. Total Yes, EGFR, and CD95 served as respective loading controls. These time points were chosen based on time course studies in previous work on CD95 activation by hyperosmolarity (1, 4) or hydrophobic bile salts (2, 5). D, CD95L induced within 1 min a Yes activation and Yes/EGFR association followed by a Yes-mediated involves the action of NADPH oxidase isoforms, whose presence in rat hepatocytes could be shown at the level of mRNA expression and for Nox 2 and the regulatory subunit p47 phox also at the protein level. Interestingly, Nox-2 and p47 phox protein was located in 24-h cultured rat hepatocytes mainly in the cytosol. This distribution is different from that described in macrophages, which exhibit a predominant plasma membrane localization of gp91 phox (Nox 2) (38).
CD95L-induced ROS formation via Nox/Duox enzymes was evidenced by a rapid Ser phosphorylation of the regulatory subunit p47 phox in response to CD95 ligand and an inhibition of the CD95L-induced ROS response by inhibitors of NADPH oxidases and after knockdown of p47 phox . DPI sensitivity of CD95 (Fas) ligation-induced ROS formation was also described in Jurkat cells (59). The present findings do not rule out a contribution of mitochondria to CD95L-induced ROS generation; however, our data suggest that the initial trigger for CD95L-induced ROS formation are NADPH oxidases, whereas mitochondrial ROS generation may be a downstream consequence of subsequent CD95 activation. If this view is correct, mitochondria may be seen as an amplifier of the initial ROS response triggered by NADPH oxidase activation. Which Nox/Duox isoform(s) is responsible for the CD95L-induced ROS response in hepatocytes is not settled. However, an unequivocal requirement of p47 phox for Nox 2 (gp91 phox ) activation was demonstrated (16), whereas it is unclear whether other isoforms are also regulated in vivo by p47 phox . Thus, Nox 2 is a likely candidate for the CD95L-induced ROS response in hepatocytes; however, a contribution of other Nox/Duox isoforms is not ruled out.
In contrast to 24-h cultured rat hepatocytes, Nox 2 protein and mRNA were not expressed by the human hepatoma cell line Huh7, whereas mRNAs for p47 phox as well as Nox 1, 3, and 4 and Duox 1 and 2 were detectable. As in rat hepatocytes, also in CD95-YFP-transfected Huh7 cells, apocynin, the PKC pseudosubstrate, and AY9944 inhibited CD95L-induced CD95 translocation to the plasma membrane and apoptosis induction. These findings suggest that in Huh7 cells NADPH oxidase isoforms distinct from Nox 2 may be involved in the induction of apoptosis by CD95L.
The inhibitory action of apocynin on CD95L-induced ROS formation in rat hepatocytes deserves some comment. Previous studies have shown that inhibition of NADPH oxidase by apocynin is not mediated by apocynin itself but by a compound derived from apocynin in a peroxidase-dependent way (60). Due to this mechanism, apocynin was shown to inhibit NADPH oxidase in phagocytic cells, whereas in some (peroxidase-deficient) nonphagocytic cells, such as fibroblasts, apocynin stimulates rather than inhibits ROS formation (33). However, the inhibitory effect of apocynin was restored following treatment with H 2 O 2 and horseradish peroxidase (33). Hepatocytes are known to contain high activities of peroxidases (61,62) and to exhibit endogenous H 2 O 2 formation (61,63), which are prerequisites for the inhibitory action of apocynin on NADPH oxidases. This may explain why in rat hepatocytes after a 30-min preincubation with apocynin, CD95L-induced ROS formation via NADPH oxidase is inhibited.
Mechanism of CD95L-induced NADPH Oxidase Activation-Sphingomyelinase, ceramide, and PKC were identified as up- EGFR-tyrosine phosphorylation, which was sensitive to inhibition of sphingomyelinase and PKC. Also, EGFR/CD95 association and CD95tyrosine phosphorylation as well as DISC formation were largely prevented by inhibitors of sphingomyelinase and PKC (n ϭ 3). E, densitometric analysis of CD95L-induced Yes-Tyr 418 phosphorylation, Yes/EGFR association, and subsequent EGFR-tyrosine phosphorylation revealed a significant effect of inhibition of sphingomyelinases (AY9944 and desipramine) and PKC (PKC inhibitor and chelerythrine) (*, p Ͻ 0.05). F, also inhibitors of NADPH oxidase (DPI, apocynin, and neopterin) and Yes (SU6656) as well as NAC blunted CD95L-induced Yes activation, Yes/EGFR association, and EGFR-tyrosine phosphorylation (n ϭ 3). All maneuvers that prevented Yes activation also prevented EGFR-tyrosine kinasemediated CD95-tyrosine phosphorylation and subsequent recruitment of FADD and caspase 8 to CD95 (DISC formation) (n ϭ 3). stream events of CD95L-induced NADPH oxidase activation. In line with this, both CD95L-induced p47 phox phosphorylation and the ROS response were sensitive to inhibitors of sphingomyelinase or PKC. CD95L rapidly increased the formation of ceramide in a sphingomyelinase inhibitor-sensitive way. Although one has to keep in mind that some of these inhibitors may exert unspecific side effects, the findings indicate that sphingomyelinase-mediated ceramide formation may trigger PKC activation as an upstream event in CD95L-induced NADPH oxidase activation. It is interesting to note that p47 phox was recently shown to be a direct phosphorylation target of PKC (48), and this PKC isoform participates in neutrophil respiratory burst (39,48). Multiple roles of PKC, a member of the atypical PKC group, in cell signaling have been described (for a review, see Ref. 64), and evidence has been given that ceramide can bind to and activate PKC (for a review, see Ref. 65).
A role of ceramide formation in apoptosis has repeatedly been discussed (for a review, see Ref. 57). Evidence has been presented in leukemia cell lines and lymphocytes for an involvement of acidic sphingomyelinase in the apoptotic signaling through CD95 (66) and for a requirement of ceramide-mediated clustering for CD95-DISC formation (67). Further, ceramide was reported to overcome the apoptosis resistance of acidic sphingomyelinasedeficient mouse hepatocytes and lymphocytes (68,69) and was suggested in various cell types to mediate at least in part its apoptotic effect through activation of the JNK pathway (65, 70 -72). A role of ceramide in activating JNK in response to CD95L is suggested by the finding that CD95L-induced JNK activation was sensitive to sphingomyelinase inhibition by AY9944 and desipramine. However, CD95L-induced JNK activation was also sensitive to inhibitors of PKC and NADPH oxidases and NAC but was insensitive to Yes inhibition by SU6656. These data suggest that CD95L-induced early ROS formation may trigger JNK and Yes activation in parallel. The mechanisms underlying early CD95L-induced sphingomyelinase activation remain speculative, as are the sphingomyelinase isoenzymes, which are involved in this response. Potential activation mechanisms have been reviewed recently (65,73).
FIG. 8. CD95L-induced EGFR/Yes association is followed by EGFR-tyrosine phosphorylation and formation of the EGFR⅐CD95 complex. Hepatocytes were cultured for 24 h and then exposed to CD95L (100 ng/ml) for the time periods indicated. When indicated, SU6656 (10 mol/liter) or AG1478 (5 mol/liter) was preincubated for 30 min in order to inhibit Yes- (4,5) or EGFR-tyrosine kinase activity (1, 2), respectively. A, Yes and CD95 were immunoprecipitated (IP) as described under "Experimental Procedures." Yes immunoprecipitates were detected for CD95 and EGFR association, whereas immunoprecipitated CD95 samples were detected for Yes or EGFR association by means of Western blot (WB). Total CD95 and total Yes served as respective loading controls. CD95L-induced Yes/EGFR association precedes EGFR/CD95 association. No CD95/Yes association was detectable (n ϭ 3). B, EGFR-tyrosine phosphorylation at positions Tyr 845 , Tyr 1045 , and Tyr 1173 was detected by Western blot using phosphospecific antibodies. Total EGFR served as a loading control. CD95L (100 ng/ml) induced within 1 min a SU6656-sensitive EGFR phosphorylation at position Tyr 845 , a known target for Src family kinases (52), which was followed by EGFR autophosphorylation at position Tyr 1173 , as indicated by its AG1478 sensitivity (53). This may suggest that Yes-meditated EGFR-Tyr 845 phosphorylation leads to an activation of EGFR-tyrosine kinase activity. CD95L had little or no effect on EGFRtyrosine phosphorylation at position Tyr 1045 (n ϭ 3).
idative stress as an upstream event. An antioxidant-sensitive Yes activation was recently also found in response to other proapoptotic stimuli such as hyperosmotic hepatocyte shrinkage (4) or hydrophobic bile acids (5). In line with these latter studies, also CD95L-induced CD95 activation and DISC formation were dependent upon this Yes signal. Yes knockdown inhibited CD95L-induced CD95 activation, DISC formation, and apoptosis induction. In addition to Yes, CD95L also induced after some delay an activating phosphorylation of Fyn and c-Src, which, however, was not antioxidant-sensitive, indicating that the ROS signal is required for Yes but not for Fyn and c-Src activation. Mechanisms and significance of CD95Linduced Fyn and c-Src activation are a matter of speculation; however, in contrast to Yes inhibition by SU4466, Fyn and c-Src inhibition by PP-2 had no effect on CD95L-induced CD95 activation and apoptosis induction. Thus, Fyn and c-Src may not play a major role in apoptosis induction. In line with this, hydrophobic bile acids, such as taurolithocholate-3-sulfate, activated Yes and the CD95 system and induced apoptosis but had no effect on Fyn and c-Src phosphorylation (5).
The mechanisms underlying Yes activation in response to CD95L-induced oxidative stress or externally added H 2 O 2 are and protein knockdown of p47 phox , Yes, or EGFR Top, hepatocytes were cultured for 24 h and then exposed to CD95L (100 ng/ml) for 3 h in order to detect CD95 translocation to the plasma membrane by immunostaining or for 12 h in order to determine hepatocyte apoptosis using the TUNEL assay. When indicated, AY9944 (5 mol/liter), desipramine (5 mol/liter), PKC inhibitor (100 mol/liter), chelerythrine (20 mol/liter), NAC (30 mmol/liter), diphenyleneiodonium (10 mol/liter), apocynin (300 mol/liter), D-(ϩ)-neopterin (100 mol/liter), SU6656 (10 mol/liter), or PP-2 (10 mol/liter) was added 30 min prior to CD95L addition. In line with previous data (1), CD95L induced CD95 translocation to the plasma membrane, which was sensitive to inhibition of sphingomyelinase, PKC, NADPH oxidase, and SU6656 but not to PP-2. A similar inhibitor profile was found for CD95L-induced hepatocyte apoptosis. Data are given as means Ϯ S.E. (n ϭ 3 different preparations). Middle, cultured Huh7 cells were transfected with CD95-YFP as described under "Experimental Procedures" and then exposed to CD95L (100 ng/ml) for 3 h in order to detect CD95-YFP translocation to the plasma membrane or for 12 h in order to determine Huh7 cell apoptosis using the TUNEL assay. When indicated, AY9944 (5 mol/liter), PKC inhibitor (100 mol/liter), or apocynin (300 mol/liter) was added 30 min prior to CD95L addition. In line with previous data (3), CD95L induced CD95 translocation to the plasma membrane, which was sensitive to inhibitors of sphingomyelinase, PKC, and NADPH oxidase. A similar inhibitor profile was found for CD95L-induced Huh7 cell apoptosis. Data are given as means Ϯ S.E. (n ϭ 3 different transfections). Bottom, in another set of experiments, hepatocytes were cultured for 96 h with either nonsense oligonucleotides or p47 phox -, Yes-, or EGFR-antisense nucleotides as described under ЈExperimental ProceduresЈ and stained immunocytochemically for CD95 membrane translocation 3 h after CD95L (100 ng/ml) addition. The percentage of apoptotic cells was detected using the TUNEL assay after 12 h of exposure to CD95L (100 ng/ml). In nonsense oligonucleotide-treated hepatocytes, CD95L induced CD95 membrane translocation and apoptosis compared to untreated control, which was inhibited after p47 phox , Yes, and EGFR knockdown. Data are given as means Ϯ S.E. (n ϭ 3 different preparations).
unknown; however, oxidative stress was shown to inhibit protein phosphatases (74,75), which in turn could trigger inhibition of Yes dephosphorylation. In line with such a mechanism, inhibition of protein phosphatases by vanadate was shown to activate Yes in rat hepatocytes (5).

TABLE IV
Inhibition of CD95 ligand-induced EGFR-CFP/CD95-YFP association and subsequent membrane translocation of the CD95⅐EGFR protein complex by inhibition of sphingomyelinase, PKC, or NADPH oxidase in Huh7 hepatoma cells Huh7 hepatoma cells were cotransfected with EGFR-CFP and CD95-YFP and then exposed to CD95L (100 ng/ml) for 0 min (i.e. immediately before CD95L addition), 30 min, or 120 min, respectively. When indicated, AY9944 (5 mol/liter), PKC inhibitor (100 mol/liter), or apocynin (300 mol/liter) was added 30 min prior to CD95L addition. About 49 Ϯ 5% (n ϭ 3) of the cells expressed the transfected EGFR-CFP and CD95-YFP constructs. FRET pictures were taken as described under "Experimental Procedures" and the legend to Fig. 7. The percentage of double-transfected cells without FRET signal and FRET signal in the cellular interior or at the plasma membrane, respectively, is given. At least 100 cells per condition from three independent transfections were counted. Data are given as means Ϯ S.E. CD95L induced within 30 min an intracellular EGFR-CFP/CD95-YFP association as indicated by the FRET signal, which was followed by an enrichment of the EGFR-CFP⅐CD95-YFP protein complex in the plasma membrane within 2 h. Inhibition of sphingomyelinase, PKC, or NADPH oxidase by AY9944, PKC inhibitor, or apocynin, respectively, significantly blunted both intracellular EGFR-CFP/CD95-YFP association and subsequent membrane translocation. but not phosphorylation at Tyr 845 . These findings suggest that Yes activates the EGFR at Tyr 845 , which is followed by activating autophosphorylation at Tyr 1173 . No EGFR-Tyr 1045 phosphorylation was observed in response to CD95L. This phosphorylation site is activated in response to epidermal growth factor and offers a docking site for Cbl, which is required for EGFR internalization (54). The failure of CD95L to trigger EGFR-Tyr 1045 phosphorylation may in part explain why EGFR is not internalized but is targeted to the plasma membrane. Activated EGFR then associates with the CD95 inside the cytosol, as shown recently by fluorescence resonance energy transfer experiments (3). Yes apparently dissociates off the EGFR prior to association with the CD95, because no coimmunoprecipitation of Yes and CD95 was found. As shown previously (1, 2) and in the present study, EGFR triggers CD95-tyrosine phospho-rylation, which is one prerequisite for a microtubule-dependent translocation of the protein complex to the plasma membrane within 2 h, where DISC formation occurs (3,68,76). The important role of EGFR-induced CD95-tyrosine phosphorylation has been demonstrated in the past by inhibitor studies (1,2) and in studies on mutated CD95 (3). Huh7 cells transfected with receptors with Tyr/Phe exchanges in positions 232 and 291 (i.e. in the death domain of the CD95) are resistant toward CD95L-induced apoptosis (3). As shown in the present study, also knockdown of Yes or the EGFR strongly blunts apoptosis induction by CD95L. Also, prevention of CD95L-induced NADPH oxidase activation and subsequent ROS-dependent Yes activation, by inhibition of PKC, of sphingomyelinase by AY9944 or of NADPH oxidase isoforms by apocynin strongly blunted CD95L-induced EGFR activation, EGFR/CD95 association, CD95-tyrosine phosphorylation, membrane translocation, and apoptosis induction. These findings indicate that CD95L-signaling through NADPH oxidases is a crucial step in apoptosis induction. Fig. 11 summarizes our current view on CD95L-induced CD95 activation and apoptosis induction in rat hepatocytes.
Acknowledgments-Excellent technical assistance by Daniela Brammertz and Elisabeth Winands is gratefully acknowledged.
FIG. 10. Inhibition of CD95L-induced CD95 activation after p47 phox , Yes, or EGFR protein knockdown. p47 phox , Yes, and EGFR protein were knocked down as described under "Experimental Procedures" by use of antisense oligonucleotides. Nonsense oligonucleotides served as controls. A substantial down-regulation of these respective proteins was achieved after 4 days of hepatocyte culture (n ϭ 7). Cells were then exposed to CD95L (100 ng/ml). p47 phox expression as well as activating JNK-1 phosphorylation were detected by Western blotting (n ϭ 4). Yes and EGFR were immunoprecipitated as described under "Experimental Procedures." The immunoprecipitated Yes and EGFR samples were then detected for activating phosphorylation by Src family-Tyr 418 (Yes-Y 418 -P) or tyrosine phosphorylation (EGFR-Tyr-P) (n ϭ 4). GAPDH, total Yes, total EGFR, and total JNK-1 served as loading controls. Representative Western blots of four independent experiments are shown along with the respective loading controls. p47 phox and Yes knockdown largely prevent CD95L-induced Yes and EGFR phosphorylation (stimulation for 1 min), whereas JNK activation (stimulation for 30 min) in response to CD95L was affected by p47 phox knockdown only. These parameters were not affected after EGFR knockdown. CD95 was immunoprecipitated as described under "Experimental Procedures." CD95 samples were then detected for EGFR association, for tyrosine phosphorylation (CD95-Tyr-P), and for FADD and caspase 8 (Casp 8) association by Western blotting (n ϭ 4). Total CD95 served as loading control. The total CD95 amount was also detected in membrane and cytosolic fractions obtained by ultracentrifugation as recently described (1, 2) (n ϭ 4). p47 phox knockdown blunted CD95L-induced EGFR/CD95 association (stimulation for 60 min) in line with the reported JNK dependence of this process (1, 5) (compare Fig. 6C). All maneuvers that prevented activating EGFR-tyrosine phosphorylation (i.e. p47 phox , Yes, and EGFR knockdown) also prevented CD95-tyrosine phosphorylation (detected after 60 min of CD95L exposure), DISC formation, or CD95 enrichment in the membrane fraction (all detected after 3 h of CD95L stimulation).
FIG. 11. CD95 ligand-induced activation of the CD95 system in cultured rat hepatocytes. Fig. 11 summarizes our current view with respect to CD95 ligand-induced activation of the CD95 system. CD95L addition rapidly activates sphingomyelinase and ceramide formation, which in turn activates PKC. The latter leads to an activating serine phosphorylation of the NADPH oxidase regulatory subunit p47 phox , which is followed by generation of ROS. ROS formation triggers an activation of the Src family kinase Yes, which then associates with and activates the EGFR. EGFR then dissociates from Yes and associates with CD95 in a JNK-dependent way, which leads to EGFR-tyrosine kinase-mediated CD95-tyrosine phosphorylation, which is a prerequisite for microtubule-dependent translocation of CD95 to the plasma membrane and subsequent FADD and caspase 8 recruitment (i.e. DISC formation) (3).