Expression of gastrointestinal glutathione peroxidase is inversely correlated to the presence of hepatitis C virus subgenomic RNA in human liver cells.

There is great medical need to develop novel therapies for treatment of human hepatitis C virus (HCV). By gene expression analysis of three HCV-subgenomic RNA replicon cell lines, we identified cellular proteins whose expression is affected by the presence of HCV and therefore may serve as drug targets. Data from cDNA array filter hybridization, as well as from Northern and Western blotting, revealed that the gastrointestinal-glutathione peroxidase (GI-GPx) was drastically down-regulated (up to 20-fold) in all replicon cell lines tested. Concomitantly, total cellular glutathione peroxidase activity was drastically reduced, which rendered these human liver cells more susceptible toward oxidative stress. Interferon alpha caused down-regulation of the HCV-replicon followed by recovery of GI-GPx expression to nearly normal levels. Furthermore, expression of GI-GPx in replicon cells by gene transduction caused down-regulation of HCV RNA in a dose-dependent manner. Moreover, activating the endogenous gene coding for GI-GPx by all-trans-retinoic acid (RA) was sufficient to cause down-regulation of the HCV replicon. A small interfering RNA duplex abrogated GI-GPx up-regulation by RA and concomitantly suppression of HCV. The RA effect was dependent on the presence of sodium selenite, was reversible, and was independent of RNA-activated protein kinase. Taken together, these results show that HCV inhibits the expression of GI-GPx in replicon cells to promote its intracellular propagation. Modulation of GI-GPx activity may open new avenues of treatment for HCV patients.

The hepatitis C virus (HCV) 1 infects ϳ2% of the population worldwide (1), and more than 80% of these patients become chronically infected causing recurring hepatitis, liver cirrhosis, and cancer. HCV-related end-stage liver disease is now the leading reason for liver transplantation. Interferon (IFN) monotherapy is effective in only 10% of HCV patients (2)(3)(4), and the combination of pegylated interferon and ribavirin can eradicate the virus in about 50% of patients (reviewed in Ref. 5). The mechanism of action of ribavirin has not been fully determined, but incorporation of ribavirin triphosphate into the HCV genome seems to induce error-prone replication to a nontolerable mutation rate (6). Binding of IFN␣/␤ leads to the dimerization of the respective receptors and activation of the Jak/STAT pathway, which turns on cellular genes known to be involved in antiviral responses (7). Among these are the Mx proteins, major histocompatibility antigens, 2Ј,5Ј-oligoadenylate synthase, and the double-stranded RNA-specific protein kinase (PKR). The precise mechanisms by which HCV can resist the IFN response is not fully understood, but viral proteins E2 and NS5a seem to play a role in evasion of the cellular defense system by blocking PKR (reviewed in Ref. 8).
Molecular studies on HCV-host cell interaction have been hampered by the lack of small animal model systems (reviewed in Ref. 9). To overcome these limitations, a human hepatoma cell line (HuH7) was transfected with subgenomic HCV RNAs, called replicons ( Fig. 1A) (10). Stable cell lines containing high levels of replicon RNA and viral proteins proved useful to study HCV replication as well as translation and processing of HCV proteins present in the system (10 -12).
To examine the influence of the HCV replicon on the cellular transcription, we utilized microarray technology. Differential hybridization of more than 400 cDNAs with probes derived from RNA of mock-and replicon-transfected HuH7 cell lines resulted in the identification of several de-regulated genes. In the present study we focus on the analysis of the gastrointestinal-glutathione peroxidase (GI-GPx) gene, whose mRNA was down-regulated in the replicon cell lines up to 20-fold compared with control cell lines.
GI-GPx is one of four selenium-containing glutathione peroxidases, and its expression in rodents appears to be restricted to the epithelium of the gastrointestinal tract (13), whereas in humans it is also found in liver (14). Glutathione peroxidases are functioning primarily to counteract oxidative attack but also play roles in signaling (reviewed in Refs. 15 and 16). Reactive oxygen species (ROS), such as superoxide, hydrogen peroxide, and hydroxyl radicals, are produced during aerobic metabolism. Glutathione peroxidases, along with superoxide dismutases and catalases, are considered the main antioxidant enzymes to encounter oxidative stress, which can cause severe cell damage (17). Selenoproteins like GPx incorporate selenocysteine (Sec) co-translationally into the polypeptide chain by a complex mechanism (reviewed in Refs. 18 and 19). Sec has its own code word, UGA, that serves a dual function of dictating Sec or the cessation of protein synthesis. The feature that distinguishes UGA as a Sec codon is the presence of a stem-loop structure called a Sec insertion sequence (SECIS) element within the 3Ј-UTR (18 -21). About 25 selenoproteins with mostly unknown functions have been identified so far in humans and mice (22).
We report that, compared with control HuH7 cells, HCVreplicon cells expressed less GI-GPx and were more susceptible toward ROS-producing agents. Ectopic expression of GI-GPx in replicon cells resulted in a decrease of the HCV RNA and protein NS5a. We show that up-regulation of GI-GPx by retinoic acid and selenite caused drastic down-regulation of HCV in the replicon system. Therefore, retinoids, which exert their action by influencing gene networks, may be useful for treating HCV patients.
Screening of Arrays-Atlas TM (Clontech) "apoptosis" and "stress" cDNA expression arrays include 205 and 234 human cDNAs, respectively, spotted in duplicate dots (10 ng of cDNA per dot) on positively charged nylon membranes. Probes were generated in a volume of 15 l by reverse-transcribing 5 g of total RNA in the presence of 50 Ci of [␣-33 P]dATP, dNTPs, Moloney murine leukemia virus reverse transcriptase, and a cDNA synthesis primer mix, which is a combination of primers specific for each type of Atlas TM array. After removal of the non-incorporated nucleotides on a Sepharose spin column, efficiency of labeling was controlled by scintillation counting. Filters were prehybridized for 1 h and then hybridized with the radioactive probe (2 ϫ 10 6 cpm/ml) for 15 h at 68°C. After washing two times with 2ϫ SSC, 1% SDS and four times with 0.1ϫ SSC, 0.5% SDS at 68°C, membranes were exposed to x-ray films or to a PhosphorImaging screen (0.5-2 days). Data were utilized when the nine housekeeping genes of the filters generated similar signals for all four HuH7 cell lines. Only those signals, which were at least three times above background of filters, were considered as specific.
Northern Blot Analysis-Cultures were washed twice with ice-cold phosphate-buffered saline (PBS), and cellular RNA was extracted using the RNeasy Total RNA kit (Qiagen, Germany). 5 g of total RNA per lane was separated in 1.2% agarose, 2.2 M formaldehyde gels and transferred onto Hybond TM -N nylon membrane (Amersham Biosciences) by capillary blot with 10ϫ SSC. Pre-hybridization was performed in 6ϫ SSC containing 5ϫ Denhardt's solution, 0.5% SDS, and 250 g/ml denatured salmon sperm DNA at 65°C for 4 h. For detection of mRNAs, specific antisense oligonucleotides (10 ng) were radioactively labeled with terminal transferase and 50 Ci of [␣-32 P]dATP according to the manufacturer's protocol (Roche Applied Science).
The 32 P-labeled probes were added to freshly prepared pre-hybridization solution at 10 6 cpm/ml and incubated with the membranes overnight at 65°C. Filters were washed three times with 2ϫ SSC, 0.2% SDS at 65°C for 15 min each, followed by three washes with 0.2ϫ SSC, 0.2% SDS at 60°C for 15 min each. The blots were exposed overnight at Ϫ80°C to Kodak X-Omat AR films with intensifier screens.
Glutathione Peroxidase Assay-For measuring cellular glutathione peroxidase activity, we followed the manufacturer's instructions (Calbiochem). Briefly, cells were washed with ice-cold PBS, harvested with a rubber policeman in 5 mM EDTA, 1 mM dithiothreitol, and 50 mM Tris-HCl, pH 7.5, and lysed by three cycles of freezing and thawing. After spinning for 15 min at 10,000 ϫ g (4°C), protein concentration of the supernatant was determined with the BCA reagents. 180 g of protein were applied per assay.
tert-Butyl hydroperoxide was used as substrate, and GPx activity was estimated indirectly. Oxidized glutathione, produced upon reduction of the peroxide by GPx, is recycled to its reduced state by glutathione reductase by oxidation of NADPH ϩ H ϩ to NADP ϩ . The oxidation of NADPH ϩ is accompanied by a decrease in absorbance at 340 nm, which provides a spectrophotometric means of monitoring GPx activity. The rate of decrease in the A 340 is directly proportional to the GPx activity in the sample. This assay does not discriminate between the activities of individual isozymes of the glutathione peroxidase gene family.
Cell Viability Assay-For quantification of the degree of cell death in cell culture, we employed the viability assay based on the reduction of tetrazolium salt to formazan by mitochondrial dehydrogenase activity. The assay was performed in 96-well microtiter plates (Greiner, Frickenhausen, Germany) as described previously (24), but Alamar Blue (Roche Applied Science) was used instead of 3-(4,5-dimethylthiazol-2yl)-2,5-diphenyltetrazolium bromide. Eight wells per sample point were analyzed, and each experiment was repeated independently at least three times.
Transfection of HuH7 Cells-A 2.0-kb promoter fragment of the gpx2 gene was amplified by PCR utilizing genomic DNA isolated from 9-13 replicon cells and HEK 293 cells. The upstream primer (5Ј-GGG GAT TAA TTC CTA CAA ACA ATT CTC ATT TGC-3Ј) contained an AsnI restriction site, and the downstream primer (5Ј-GGG GCT CGA GGC CTG AGC CCA CTT ATG-3Ј) contained an XhoI restriction site. This fragment was fused with the luciferase gene, which was cloned into the pcDNA3 vector (Invitrogen) via the restriction sites PstI and XbaI, resulting in the pGPx-Luc vector. Sequence analysis revealed that the gpx2 promoter amplified from 9-13 replicon and HE293 cells were identical with published sequences (25). The pGPx-Luc plasmid was transfected into HuH7 cells and co-transfected with the vector pBabe⍀6 (26) in 9-13 replicon cells, and stable transfectants were selected for 3 weeks with 1 mg/ml geneticin (G-418, Calbiochem) and 10 g/ml puromycin (Sigma). Pools of drug-resistant cells were used for experiments.
siRNA Preparation and Transfection-We tested siRNAs targeting two positions of the human GPx2 mRNA (GenBank TM accession number X68314). The sense sequence of duplex 1 siRNA was r(CCC UCU GGU UGG UGA UUC A)dTdT and of duplex 2 was r(GGA UGA UGG CAC CUU CCU A)dTdT. The control, nonsilencing siRNA was r(UUC UCC GAA CGU GUC ACG U)dTdT. The GPx2 and control siRNAs were designed by Qiagen-Xeragon (Germantown, MD). One control siRNA, r(GAC ACU GAG ACA CCA AUU GAC)dTdT targeting the NS5b coding sequence, was shown to suppress HCV effectively (27), although the 5Ј-UTR siRNA r(UUC UCC GAA CGU GUC ACG U) dTdT had little effect on HCV (28). All siRNAs were synthesized by Qiagen-Xeragon.
For annealing, a 20 M siRNA solution was prepared, heated up to 90°C for 1 min, and incubated at 37°C for 60 min. The day before transfections, 5 ϫ 10 5 HuH7 5-15 cells were plated per well of a 6-well plate. Transfections were carried out with Lipofectamine 2000 Reagent (Invitrogen) in serum-free medium. The final siRNA concentration was 50 nM. After 3 days, transfections were repeated, and cells were incubated for another 3 days in the presence or absence of 1 M alltrans-retinoic acid and 50 nM sodium selenite. Cultures were rinsed, harvested, and lysed, and 20 g of protein was analyzed by Western blotting.
Adenovirus Construction, Purification, and Infection-The adenovirus used here were all E1-and E3-defective derivatives of adenovirus type 5 (reviewed in Ref. 29). The coding region for GI-GPx (0.7 kb) was amplified by PCR by using an upstream primer containing a HindIII recognition site (5Ј-GCG CAA GCT TAT GGC TTT CAT TGC CAA GTC CTT C-3Ј, start codon underlined) and a downstream primer containing an XbaI site (5Ј-GTT CAT CTA GAT ATG GCA ACT TTA AGG AGG CGC TTG-3Ј). The 3Ј-UTR (0.3 kb) of the GI-GPx mRNA, containing a SECIS element, was amplified using the upstream primer 5Ј-GCC CTC GAG ATG TGA ACT GCT CAA CAC ACA G-3Ј with an XhoI recognition site and the downstream primer 5Ј-CCA CGC GGC CGC TTT ATT GGT CTC TTC TAG CAG AGT GGC-3Ј covering the polyadenylation site (AAUAAA) and containing a NotI recognition site for cloning. RNA isolated from HuH7 cells was reverse-transcribed by random priming and used as template for PCR. The cDNA coding for human GI-GPx was cloned with and without 3Ј-UTR into the transfer plasmid (pPM7) between the cytomegalovirus immediate early promoter/enhancer and the rabbit ␤-globin intron/polyadenylation signal. This expression cassette was inserted into a bacterial plasmid-borne adenovirus genome using recombination in bacteria (30,31). A cloned version of the novel genome was identified by DNA analysis, and the viral genome was released from the plasmid by restriction enzyme digestion, and virus replication was initiated by transfecting the genome into HEK 293 cells. Virus was amplified in modified HEK 293 cells and purified from cell lysates using CsCl density gradient centrifugation as described (31). Virus was quantified by protein content using the conversion factor 1 mg/ml pure virion protein ϭ 3.4 ϫ 10 12 viral particles/ml (31). The control viruses AdJ5 and AdLuc were described previously (32).

GI-GPx Is Down-regulated in Replicon
Cells-Interference of HCV with cellular signaling events can be reflected in altered gene expression. Possible influences and dependences of HCV RNA replication and HCV proteins on host cell transcription are accessible to analysis with cDNA arrays. Three replicon cell lines (HuH 9-13, HuH 5-15, and HuH 11-7) (10, 11) were analyzed for alterations in cellular RNA expression patterns. HuH7-pcDNA3 cells are HuH7 cells resistant to G-418 by integration of a neo gene-carrying plasmid (pcDNA3, Invitrogen) and served as control. Radioactively labeled complex cDNA probes were derived from the RNA of these four cell lines and were hybridized with the cDNA array membranes (stress and apoptosis arrays, Atlas TM , Clontech). Results from this novel signal tranduction microarray analysis revealed significant down-regulation (10 -20-fold) of GI-GPx message in the three replicon cell lines (data not shown).
To verify the results obtained by cDNA array screening, we performed Northern blot analyses utilizing 10 g of total RNA of each cell line and two different radioactively labeled DNA oligonucleotides as probes. The sequences were deduced from the coding region (GI-GPx; probe 1) and from the 3Ј-UTR (GI-GPx; probe 2) (Fig. 1B). Densitometric scanning of autoradiograms of three independent experiments established that the mRNA coding for GI-GPx is about 10-(HuH 5-15) and 20-fold (HuH 11-7 and HuH 9-13) down-regulated in replicon cell lines compared with control cell line HuH7 pcDNA3 (Fig. 1B). We performed Northern blot analyses with RNA isolated from four naive HuH7 cell lines with various numbers of passages (passage 15, 41, 80, and 136) to rule out that due to clonal selection the HuH7 pcDNA3 reference cell line overexpressed the GI-GPx mRNA. These cell cultures expressed similar levels of GI-GPx mRNA as the HuH7 pcDNA3 reference cell line (data not shown). Thus, the low amount of GI-GPx mRNA in the replicon cell lines was linked to the presence of HCV-RNA and/or HCV-proteins.
Furthermore, we wondered whether the cytoplasmic member of the glutathione peroxidase family expressed in hepatocytes, the cellular glutathione peroxidase (cGPx), compensates for the deficiency of the GI-GPx message. Stripping the membranes (Fig. 1B) and hybridizing them with radioactively labeled DNA oligonucleotides specific for human cGPx revealed similar cGPx mRNA levels in control cells and replicon cells (Fig. 1B, lower panels). These data demonstrate that the GI-GPx gene is specifically affected by the HCV replicon.
The level of GI-GPx mRNA (Fig. 1B) was also reflected on the protein level. GI-GPx protein was readily detectable in pcDNA3 cells, was hardly detectable in 5-15, and undetectable in 9-13 and 11-7 replicons (Fig. 1C, upper panel). Furthermore, we studied the expression of MnSOD, another enzyme involved in detoxification of cellular radicals. Western blotting demonstrated that levels of MnSOD protein were similar in the three replicon cell lines and in pcDNA3 control cells (Fig. 1C, middle panel).
Replicon Cells Are Susceptible to Oxidative Stress-We addressed physiological consequences of the transcriptional down-regulation of the GI-GPx gene. Therefore, cellular extracts of pcDNA3 control cells and the three replicon cell lines (HuH 5-15, 11-7, and 9-13) were prepared to estimate activity of the cellular glutathione peroxidase. The data show that indeed the three replicon cell lines exhibited only about half of the GPx activity of the pcDNA3 control cells (Fig. 2A). These findings prompted us to investigate whether the replicon cells were more susceptible toward treatment with radical and oxidative stress-producing compounds. Replicon cell lines and control cells (HuH7 pcDNA3) were incubated for 24 h with various concentrations of paraquat (methyl viologen), and the viability of the cultures was measured using the Alamar Blue assay. Paraquat, a bipyridol herbicide that uses molecular oxygen to produce superoxide anion radical and subsequently hydrogen peroxide (33), impaired viability of replicon cells more severely than of HuH7 pcDNA3 control cells (Fig. 2B). The estimated LD 50 values for paraquat calculated from three independent experiments were 260 Ϯ 50 M for HuH 9-13, 270 Ϯ 75 M for HuH 5-15, 310 Ϯ 65 M for HuH 11-7, and 3 mMϮ120 M for HuH pcDNA3. Thus, the HCV replicon cells were about 10-fold more susceptible toward this radical-producing compound. Similar results were obtained with the ROS-generating agent 2,3-dimethoxy-1,4-naphthoquinone (data not shown).
Effect of Interferon on HCV and GI-GPx Expression-To explore whether the presence of the HCV replicon was responsible for the down-regulation of the GI-GPx mRNA in HuH7 cells, we treated replicon cells with interferon ␣ (1000 units/ml) to decrease HCV RNA and protein. It is known that incubation of replicon cells with IFN␣ causes rapid depletion of the viral HCV RNA and consequently of HCV proteins (11). For HCV RNA detection, we hybridized the blots with a radioactively labeled oligonucleotide whose sequence was derived from the neomycin phosphotransferase (neo) gene. This gene is located on both the HCV replicon RNA and the pcDNA3 vector used for establishing the control cell line HuH pcDNA3. Accordingly, bands of 9 and 1 kb were detected in replicon cells and in HuH pcDNA3 cells, respectively (Fig. 3). IFN␣ treatment of 5-15, 11-7, and 9-13 cultures for 2 and 4 days caused eradication of the HCV RNA (Fig. 3) and NS5a protein (data not shown) to undetectable levels but had no effect on expression of the neo gene in HuH pcDNA3 control cells (Fig. 3). The IFN␣-induced down-regulation of HCV followed an increasing up-regulation of GI-GPx mRNA in HuH 5-15, 11-7, and 9-13 replicon cells. After 4 days of IFN␣ treatment, GI-GPx mRNA levels of 9-13 cells almost met the levels of control cells (HuH7 pcDNA3) (Fig.   3). IFN␣ is known to cause an up-regulation of the IFN-response gene PKR (double-stranded RNA-specific protein kinase). Indeed, transcription of the two PKR-specific mRNAs was increased in all HuH7 cell lines after addition of IFN␣ (Fig.   FIG. 1. Replicon cells express less GI-GPx mRNA and protein than control HuH7 cells. A, HCV is a single-stranded RNA virus of about 9.6 kb. The 5Ј-and 3Ј-UTRs flank a single open reading frame encoding at least 10 structural and nonstructural proteins (56). Viral proteins are generated as a polyprotein of about 3000 amino acids that is translated via the IRES located in the 5Ј-UTR (57). In the subgenomic replicon, the region that encodes the core proteins was replaced by the neomycin phosphotransferase gene (neo) and the IRES of the encephalomyocarditis virus. This IRES drives the translation of the polypeptide for the nonstructural proteins, whereas the selectable marker neo is expressed under the control of the original HCV IRES of the 5Ј-untranslated region (11,12). B, HuH7 control cells (pcDNA3) and the HuH replicon cell lines 5-15, 11-7, and 9-13 were plated in 10-cm culture dishes (5 ϫ 10 5 cells) and harvested after 3 days when cells were actively progressing through the cell cycle. Total RNA was isolated and 10 g separated in a 1.2% agarose gel and used for Northern blot analysis. Blots were hybridized with radioactively labeled oligonucleotides (Probe 1 and Probe 2, upper panels) complementary to coding region and 3Ј-UTR, respectively, of the human GI-GPx mRNA. Blots were stripped and re-hybridized with two oligonucleotides (Probe 1 and Probe 2, lower panels) recognizing the cellular glutathione peroxidase (cGPx) mRNA. Membranes were exposed to Kodak x-ray films for 1 day (GI-GPx) and 2 days (cGPx), respectively, at Ϫ80°C with intensifier screens. The blots shown are representatives of three independent experiments. C, replicon cells were grown for 3 days, when extracts were prepared and 20 g of protein was separated by 12.5% PAGE for Western blotting using a GI-GPx-specific rabbit antiserum (upper panel) and an MnSOD antibody (middle panel). For loading control, membranes were re-probed with an ␣-tubulin-specific antibody (lower panel). Bound antibody was detected by enhanced chemiluminescence (Amersham Biosciences). Positions of protein size marker are indicated on the left.
3). Furthermore, the HuH7 pcDNA3 data show that GI-GPx was not a direct IFN␣-responsive gene.
IFN␣ treatment activated the innate immune response leading to the complete removal of the HCV replicon within 2 days. The up-regulation of GI-GPx in replicon cells by IFN␣ after 4 days seemed to be a consequence of the depletion of HCV rather than a direct effect of IFN␣ signaling (Fig. 3). Taken together, these data revealed an inverse correlation of the HCV replicon and GI-GPx.
Effect of Overexpression of GI-GPx on HCV Replicon-To gain more direct proof that GI-GPx prevents the replication of genomic HCV RNA, we overexpressed the human GI-GPx cDNA in replicon cells. Therefore, the GI-GPx coding region was cloned into expression vector pPM7. To allow efficient readthrough of the stop codon coding for selenocysteine, we fused the 3Ј-UTR of the GI-GPx mRNA containing the SECIS element to the GI-GPx coding region. To control the effectiveness of these constructs, HEK 293 cells were transiently transfected for 1 day with plasmids containing the GI-GPx coding region with and without the 3Ј-UTR of GI-GPx. Only the construct with 3Ј-UTR of GI-GPx produced evident amounts of GI-GPx protein (Fig. 4A). Measuring activities of the transfected cultures disclosed a 50% increase in glutathione peroxidase activity when the GI-GPx ϩ3Ј-UTR construct was used proving the functioning of the constructs (Fig. 4A).
The two pPM7 GI-GPx constructs (ϩ3Ј-UTR and Ϫ3Ј-UTR) were recombined into the adenovirus genome, and virus particles were produced. For overexpression of GI-GPx mRNA in all three replicon cell lines, 1000 virus particles per cell were used for transduction (Fig. 4B). As control, adenovirus with the gene coding for the green fluorescence protein was used. After 4 days, cultures were harvested, and expression of the HCV protein NS5a and of GI-GPx was analyzed by Western blotting. Only the construct containing the SECIS element (ϩ3Ј-UTR) allowed expression at detectable levels. The expression of the GI-GPx gene resulted in a dramatic reduction of the NS5a signal in the cell lines 11-7 and 5-15. The effect was weaker in 9-13 cells probably due to lower GI-GPx levels (Fig. 4B).
Seven days after transduction, expression of GI-GPx was still high and NS5a remained down-regulated in these cells (data not shown). These data demonstrate again the inverse correlation between presence of GI-GPx and HCV replicon.
Effect of Retinoic Acid on Replicon-In order to utilize a mechanism to activate the endogenous gpx2 gene, coding for GI-GPx, we examined the promoter of the human gene (Gen-Bank TM accession number AF199441). We identified several putative retinoic acid-response elements in the promoter and in the first intron. Indeed, treatment of the breast tumor cell line MCF-7 with all-trans-retinoic acid (RA), a classical morphogen that can induce differentiation and apoptosis (24,53), increased both GI-GPx mRNA level and enzymatic activity (34). The induction by RA was specific for the expression of the gpx2 gene, but not of the gpx1 gene coding for the cellular cGPx (34).
In order to assess potent activators of the gpx2 promoter, we stably transfected naive HuH7 and 9-13 replicon cells with a construct consisting of the human gpx2 promoter (2 kb that all-trans-RA was the most potent inductor of the gpx2 promoter (2.5-fold), followed by 9-cis-RA, 13-cis-RA, and 4-hydroxyphenylretinamide (Fig. 5A). The pattern was similar in HuH7 and 9-13 replicon cells (data not shown) and implies that RA receptors, but not retinoid X receptors, were involved in gpx2 promoter regulation.
We then incubated the replicon cell line 9-13 for 3 days with the same retinoids (10 M each in the presence of 50 nM sodium selenite). Only those compounds efficiently activating the gpx2 promoter (Fig. 5A) were able to induce significant down-regulation of NS5a. All-trans-RA, 13-cis-R, and 4-hydroxyphenylretinamide caused reduction of NS5a by about 80%, whereas methoprene acid and AM-480 had hardly any effect on NS5a expression (Fig. 5B).
In order to elucidate the effect of retinoids on HCV in more detail, HuH7 replicon cells were incubated with increasing concentrations (from 10 Ϫ4 to 10 M) of all-trans-retinoic acid in the presence of 50 nM sodium selenite for 1 week. The GI-GPx mRNA (data not shown) and protein levels increased in a dose-dependent manner (Fig. 5C, lanes 4 -9). The half-maximal induction was at 1 M RA, which is in good agreement with effects mediated by retinoic acid receptors (24). In parallel, we studied the effect of activating GI-GPx expression via RA treatment on the HCV replicon by monitoring NS5a levels. In fact, 1 and 10 M RA (Fig. 5C, lanes 8 and 9) caused significant down-regulation of the NS5a protein. At this concentration we did not observe any cytotoxic effects of RA either on naive HuH7 cells or on replicon cells. Furthermore, growth curves of all cell lines over 1 week in the presence or absence of 1 M RA (Ϯ50 nM sodium selenite) were indistinguishable from control cultures (data not shown).
Notably, RA (10 M) without sodium selenite hardly induced GI-GPx protein expression (Fig. 5C, lane 3). This finding implies that selenite has to be present in the culture medium for efficient translation of selenocysteine proteins as already described before for hepatocytes (35). Fig. 5C also shows that selenite itself caused a slight increase of GI-GPx protein levels which, however, was not sufficient to affect NS5a levels (lane 2). Even cultivating the replicon cell lines for several passages (3 weeks) in the presence of 50 nM sodium selenite did not cause down-regulation of HCV (data not shown). The finding that RA caused NS5a down-regulation only in the presence of selenite argues that the effect was indeed mediated by the selenocysteine protein GI-GPx.
To strengthen this theory siRNA experiments were performed. Therefore, we ordered two sets of Gpx2 siRNA (Qiagen) and found that only duplex 1 was efficient in down-regulating the GI-GPx expression (see Fig. 5D, middle panel, lane 7). Replicon 5-15 cultures were transfected with GI-GPx siRNAs and again after 3 days. After incubation for an additional 3 days in the presence or absence of all-trans-retinoic acid (1 M) and sodium selenite (50 nM), expression of NS5a, GI-GPx, and PKR was analyzed. Although cultures transfected with a nonsilencing control siRNA responded to the RA/sodium selenite treatment by down-regulation of NS5a (Fig. 5D, lanes 3 and 4) as the control cells (transfection procedure only) did (lanes 1 and 2), this down-regulation was severely reduced by Gpx2 siRNA (compare lane 8 with lanes 2 and 4). This inhibitory effect was not complete, which may be because not all cells took up the siRNA in sufficient amounts or because minor effects of RA on HCV were independent of GI-GPx. The siRNAs targeting the 5Ј-UTR and the NS5b coding sequence of HCV served as controls. Although the 5Ј-UTR siRNA had minor consequences on NS5a levels (Fig. 5D, lane 5) as described before (27), the NS5b siRNA effectively suppressed NS5a expression (lane 6), as shown previously (28). The nonsilencing control siRNA and the 5Ј-UTR siRNA demonstrate that transfection of the siRNAs shown here did not trigger the IFN response (36). Furthermore, neither RA nor transfection of siRNAs caused induction of PKR expression (data not shown).
The RA Effect on the Replicon Is Reversible-Because it is known that RA can have significant effects on cell cycle and differentiation, we investigated whether the replicon cell lines experienced permanent alterations upon RA treatment. The   FIG. 4. A, transient expression of GI-GPx requires 3Ј-UTR. The pPM7 plasmid without additional insert (mock) or with the human hemagglutinin-tagged GI-GPx coding region with 3Ј-UTR (ϩ3UTR) of the GI-GPx gene including the SECIS or without 3Ј-UTR (Ϫ3UTR) was transfected into HEK 293 cells, which were seeded in a 6-well plate (4 ϫ 10 5 cells per well) the day before transfection. After 1 day, cultures were harvested, and 20 g of cellular protein was analyzed by Western blotting utilizing a monoclonal hemagglutinin antibody. Pronounced expression of GI-GPx (arrow) was only observed when the SECIS element of the 3Ј-UTR was present. In parallel, glutathione peroxidase activity assays were performed by using 50 g of cellular protein of the transfected cells. The values for the relative activities are presented (mock control ϭ 100%). The HEK 293 culture transfected with the GI-GPx ϩ3Ј-UTR construct showed significant higher levels (1.5-fold) of glutathione peroxidase activity than the mock-transfected culture or cultures transfected without the SECIS element. B, expression of GI-GPx causes down-regulation of replicon. The replicon cell lines 5-15, 11-7, and 9-13 were plated at a density of 10 5 cells per well of a 6-well plate and infected with 10 3 adenoviral particles/cell, containing either the green fluorescent protein (GFP) as negative control, the GI-GPx cDNA without the 3Ј-UTR (Ϫ 3Ј-UTR), and the GI-GPx cDNA with the 3Ј-UTR containing the SECIS (ϩ3Ј-UTR), as indicated. After 4 days post-infection cells were harvested, lysed, and 10 g of protein separated on a 12.5% polyacrylamide gel for Western blot analysis. Monitoring expression of the transduced GI-GPx cDNA without 3Ј-UTR (Ϫ3Ј-UTR) revealed no (cell line 5-15 and 9-13) or only weak expression (cell line 11-7). However, drastic GI-GPx expression was achieved with the GI-GPx ϩ3Ј-UTR construct. Western blot analyses show a strong downregulation of the HCV protein NS5a in all replicon cell lines infected with the GI-GPxϩ3Ј-UTR construct. Thus, expression of NS5a was inversely correlated to levels of GI-GPx. Loading efficacy and integrity of proteins were controlled by re-blotting with an ␣-tubulin antibody. The figures depicted exemplify results from three independent experiments. 5-15 cells were cultivated for four passages (each 4 days) in the presence (ϩ) and absence of all-trans-RA (Ϫ) (Fig. 6). RA caused up-regulation of GI-GPx expression and concomitantly down-regulation of NS5a after 3 days (passage 1) (Fig. 6A, lane   2). The decline of NS5a expression to barely detectable levels was even more pronounced during the next three passages (Fig.  6A, lanes 4, 7, and 10). Removal of RA after one passage (Fig.  6A, ϩ/Ϫ) caused reduced expression of GI-GPx and recovery of NS5a levels (Fig. 6A, compare lanes 4 and 5). The GI-GPx expression was completely lost after one more passage in the absence of RA (Fig. 6A, lane 8). Densitometric scanning of the Western blot analyses demonstrated first that the RA effect on up-regulation of GI-GPx and down-regulation of NS5a was reversible and second that the expression of NS5a and GI-GPx was strictly and inversely correlated (Fig. 6, B and C).
The RA Effect on the Replicon Is PKR-independent-Furthermore, to rule out that the RA effect on the replicon was caused by known replicon antagonists like interferon-dependent antiviral host response, we investigated the effect of RA on the double-stranded RNA-activated protein kinase PKR (Fig. 7). PKR is a ubiquitously expressed protein kinase that is induced by interferon, growth factors, and different stress signals and constitutes the major IFN-regulated antiviral pathway in mammals. Thus, PKR is a key player in the cellular response to stress (37,38). RA in combination with selenite (RA/Se) caused drastic down-regulation of NS5a by 80% (Fig. 7, upper panel) but did not induce activation of the PKR gene as did interferon (Fig. 7). On the other hand, IFN caused complete down-regulation of NS5a in the presence of PKR. The up-regulation of GI-GPx after IFN treatment was a consequence of the removal of the replicon from the cells as shown before (see Fig. 3).
As additional control to exclude an IFN response by RA, we investigated the effect of RA on the IFN-responsive genes infar-1 and infar-2, coding for the type I IFN receptor (38,55). Neither the INFAR-1 nor the -2 levels were increased by RA treatment (data not shown).
These data are in favor of a novel PKR-independent pathway, by which RA leads to HCV down-regulation. Using subtherapeutic concentrations of IFN␣ alone and in combination with all-trans-RA substantiated this hypothesis. Interferon ␣ caused a dose-dependent down-regulation of NS5a, which was further enhanced by addition of all-trans-RA (data not shown). The finding that RA acts independently of interferon implies the possibility to cure patients who do not respond to interferon standard therapy with RA. These three cultures (ϪRA, ϩRA, and ϩ/ϪRA) were passaged two more times every 4 days (P3 and P4). Western blotting revealed that long term treatment with RA drastically reduced NS5a levels (lanes 7 and 10). Furthermore, removal of RA led to a reduction of GI-GPx to undetectable levels and a concomitant recovery of NS5a levels after two passages (lanes 8 and 11). B, the levels of NS5a were quantified by densitometric scanning of the autoradiograms (A). Whereas NS5a levels remained quite constant of control cultures (ϪRA, circles), the NS5a levels in RA-treated cultures decreased down to 5% of control (ϩRA, diamonds). Most interestingly, NS5a levels of cultures treated with RA recovered within two passages, when RA was left out (ϩ/ϪRA; squares). C, the levels of GI-GPx protein were quantified as described for NS5a (B). Removal of RA (ϩ/ϪRA; square) caused rapid decline of GI-GPx levels within two passages. One typical example out of three is depicted.

DISCUSSION
The present study demonstrates that human liver cells harboring a subgenomic HCV replicon RNA express drastically reduced levels of gastrointestinal glutathione peroxidase. The inverse correlation between HCV and GI-GPx was revealed by several lines of evidence (summarized in Fig. 8). 1) Cells housing the HCV replicon express 10 -20-fold less GI-GPx than cells without replicon. 2) Treating replicon cells with interferon ␣ caused disappearance of HCV RNA and HCV protein NS5a, as described previously (11). HCV down-regulation was followed by a recovery of GI-GPx mRNA levels (this study). 3) Forced expression of GI-GPx in replicon cells by adenoviral gene transfer induced down-regulation of HCV RNA and NS5a. 4) Activating the promoter of the cellular gpx2 gene (coding for GI-GPx) by nuclear receptor ligand retinoic acid in a time-and dose-dependent manner caused down-regulation of HCV. Wash out experiments demonstrated that this effect of RA on GI-GPx expression and HCV replicon was reversible. The retinoic acid effect depends on selenium and is abrogated by Gpx2 siRNA.
The transcriptional down-regulation of GI-GPx in replicon cells was reflected on protein and enzymatic levels and seemed to be responsible for a 10-fold higher sensitivity than control cells to radical-producing agents like paraquat. It was already shown in knock-out mice that glutathione peroxidase protects against the lethal oxidative stress caused by high levels of paraquat (39). Thus, the GI-GPx down-regulation permits selective killing of HCV replicon cells by paraquat and may open new routes for treatment of HCV-infected patients.
Selenocysteine proteins do not only function to protect against oxidative stress but seem to have other crucial roles in maintaining a healthy physiology of the liver (40,41). Selenium was proposed to have anticarcinogenic functions, and recently it was shown that in progressed stages of colorectal cancer expression of selenoprotein GI-GPx was decreased (42,58). Selenium has also been implicated in enhancing immune functions and thereby slowing the progression of AIDS in human immunodeficiency virus-positive patients (reviewed in Ref. 40). Our data now point to a more direct antiviral function of the selenoprotein GI-GPx in HCV infection, and it remains to be seen if the same applies to HIV. Because deficiency in dietary selenium results in decreased levels of selenoproteins, thus compromising biological processes that are maintained by these proteins, it will be interesting to investigate whether selenium lack supports HCV spreading in patients.
Here we show for the first time that the selenocysteine protein GI-GPx plays a crucial function in limiting the HCV replicon in human liver cells, and one might speculate that GI-GPx carries antiviral properties. Our novel findings support the hypothesis that GI-GPx, apart from being a barrier against hydroperoxide absorption, may also be involved in cell growth, differentiation, and ultimately in regulating signal transduction (reviewed in Refs. 16 and 58). In a recent study, it was shown that endogenously produced ROS did not lead to NF-B activation but instead lowered the magnitude of its activation (43). Therefore, it may be an advantage for HCV to abolish GI-GPx expression in order to prevent an NF-B-driven anti- viral host response. Although it is not known how the HCV replicon system interferes with the cellular signal transduction machinery to induce down-regulation of the gpx2 gene, it is conceivable that NS5a is involved. Transfection of NS5a itself seems to be sufficient to trigger the elevation of ROS leading to the activation of several cellular transcription factors like NF-B and STAT-3 (44). Several studies demonstrated that HCV-infected liver cells in patients (45) and in vitro (46,47) are characterized by a high level of oxidative stress.
A major contribution to nucleic acid damage is caused by oxidative stress due to H 2 O 2 and other endogenous reactive oxygen species (48). Therefore, it is conceivable that the downregulation of GI-GPx expression is accountable for, first, the high rate of mutations in HCV and development of quasispecies and, second, the development of liver fibrosis and cancer (49). Over a long term period, an inappropriate composition of the cellular antioxidant defense system might facilitate DNA damage and contribute to the increased cancer risk in HCV patients. Not surprisingly, we identified in our array analysis several genes, which are involved in DNA repair, up-regulated in replicon cells. Among these were the growth arrest and DNA damage-inducible protein GADD153, the DNA repair protein RAD54, the DNA excision repair protein ERCC1, and the damage-specific DNA-binding protein p48 subunit. 2 Another consequence of the down-regulation of GI-GPx may be the promotion of synthesis of HCV proteins. Under conditions of cellular stress, the rate of translation for most cellular mRNAs is reduced. However, mRNAs containing an internal ribosomal entry site (IRES) are selectively translated under these conditions. In particular, translation mediated by the viral IRES is enhanced following prolonged exposure to anoxia (50). Consistent with this model is the observation that antioxidants block replication of HCV in the subgenomic replicon system (46).
As stated above, it is conceivable to cure HCV patients by treatment with ROS-producing agents in order to selectively kill the highly susceptible HCV-positive hepatocytes. Furthermore, our data obtained with the adenoviral gene transduction system or by activation of the endogenous promoter of the cellular GI-GPx gene by retinoic acid demonstrate that enhanced expression of GI-GPx supports the host cell in limiting HCV replication. As shown here in human replicon cells, RA can activate the putative retinoic acid-response element(s) of the gpx2 gene. Increasing the GI-GPx activity to reduce cellular radical levels may form the basis for valuable treatments of HCV patients. The differential regulation and the tissue-specific expression of GI-GPx may allow finding means to up-regulate the GI-GPx protein in HCV-infected liver cells.
It is worth mentioning that more than additive effects in inhibiting the replicon were obtained when RA was applied in the presence of ribavirin or subtherapeutic concentrations of interferon (data not shown), both belonging to the standard regimen for treatment of HCV patients and shown to be effective in the replicon system (11,51). Because the inhibitory effect of RA on the replicon acts independently of an interferon response (Fig. 7), RA may be useful for treating patients who do not react on interferon treatment, the so-called "nonresponders." About 50% of chronic HCV patients are nonresponders facing cirrhosis, cancer, and, finally, liver transplant. Furthermore, the inhibitory effect of RA on HCV infection in vivo may be even stronger than in the subgenomic replicon system. In recent studies it was determined that the HCV core protein, which is absent in replicon cells (Fig. 1A), can increase expression of retinoic acid-response elements containing genes and RA-mediated apoptosis (52)(53)(54).
The effects of RA on replicon cells described here were recently confirmed by a study in humans. A patent application (59) describes examples for successful application of retinoic acid to HCV-positive patients. Within 3 weeks of treatment with all-trans-RA, the viral load dropped to undetectable levels. It is likely that the underlying mechanism is based on the up-regulation of GI-GPx as described here.
Our results present new insight into the pathogenesis of hepatitis C and provide an experimental rationale for investigating retinoic acid as a stand-alone therapy or in combination with interferon. Therefore, it will be crucial to elucidate and understand in more detail the virus/host cell interaction and the controlling elements regulating expression of GI-GPx.