CCAAT/Enhancer-binding Proteins and Interact with the Silencer Element in the Promoter of Glutathione S-Transferase P Gene during Hepatocarcinogenesis

We have previously identified a silencer in the glutathione S-transferase P (GST-P) gene which is strongly and specifically expressed during chemical hepatocarcinogenesis. At least three trans-acting factors bind to multiple cis-elements in the silencer. One of them, Silencer Factor-B (SF-B), is identical with CCAAT/enhancer-binding protein β (C/EBPβ) and binds to GST-P Silencer 1 (GPS1). Many C/EBPβ binding sites are recognized by each of the C/EBP isoforms. Western blot analyses of C/EBP isoforms during chemical hepatocarcinogenesis revealed a decrease of C/EBPα expression. However, there was no change in C/EBPβ level. In the nuclear extracts from normal liver, C/EBPα was the dominant form that bound to GPS1, whereas both C/EBPα and C/EBPβ bound to GPS1 in the nuclear extracts from carcinogenic liver. Furthermore, transfection assays showed that C/EBPα not only repressed the GST-P promoter activity but also attenuated the transcriptional stimulation by C/EBPβ. These observations strongly suggest that the ratio of C/EBPα to C/EBPβ is one of the important factors for the GST-P silencer activity, and the decrease of this ratio during hepatocarcinogenesis reduces the silencer activity and, consequently, increases the GST-P expression.

Therefore, induction of GST-P gene during hepatocarcinogenesis is a rat-specific event (1).
The rat GST-P gene consists of two enhancers and a silencer (2). One of the enhancers, termed GST-P enhancer I, is a strong positive regulatory element and has two phorbol 12-O-tetradecanoylphorbol 13-acetate responsive element (TRE)-like sequences. The GST-P gene expression is mediated mainly by this GST-P enhancer I (3,4). The GST-P promoter has also TRE and GC box and shows high activities in several cell lines (2,3,5). Nevertheless, mRNA of the GST-P gene, as well as the protein, is undetectable in normal liver (6,7). These observations prompted us to investigate the silencer that had been identified just upstream of the TRE and the GC box of the promoter (2,8).
C/EBP␣ can trans-activate several genes encoding adiposespecific proteins, including a fatty acid-binding protein variously termed 422 or aP2, stearoyl-CoA desaturase, and the insulin-responsive glucose transporter (22,23). C/EBP␤ and C/EBP␦ promote the transcription of the acute phase proteins such as ␣ 1 -acid glycoprotein, serum amyloid A, and kininogen (24 -26). The basic region and the leucine zipper region (bZIP) of C/EBP isoforms show high similarity, and some cis-elements are recognized by each of the C/EBP isoforms (11,14,15,17). Therefore, the expression profile of C/EBP isoforms is important to understanding the gene expression which is controlled by them. For example, the accumulation of C/EBP␤ and C/EBP␦ reached a maximal level during the first 2 days of adipocyte differentiation and declined sharply before the onset of C/EBP␣ accumulation. The expression profiles of the adipocyte-specific mRNA are similar in pattern (14).
In this paper, the pattern of expression of C/EBP isoforms and the profile of GPS1 binding protein during chemical hepatocarcinogenesis were examined by Western blot analysis and gel mobility shift assay. We also describe functional analyses which demonstrate that C/EBP␣ not only represses the GST-P promoter but also attenuates the transcriptional stimulation by C/EBP␤ (LAP).

MATERIALS AND METHODS
Animals and Treatments-Carcinogenic experiments were according to the Solt-Farber protocol (27). Experiments were initiated by injecting 200 mg of diethylnitrosamine per kg into 6-week-old Wistar male rats. After feeding a basal diet for 2 weeks, the diet was changed to basal diet containing 0.02% 2-acetylaminofluorene. Partial hepatectomy was performed at the beginning of the 3rd week, and the rats were sacrificed at the 7th and 8th weeks after diethylnitrosamine treatment. Control rats were injected with 0.9% NaCl solution and fed a basal diet.
Preparation of Cytosol Fraction and Nuclear Extracts from Rat Liver and HepG2 Cells and Western Blot Analysis-Rat livers were excised, and 1 g was homogenized in 2 ml of cold 0.25 M Tris-HCl (pH 7.5) and centrifuged at 10,000 ϫ g for 10 min at 4°C. The supernatants were used as the cytosol fractions. Nuclear extracts from rat liver were prepared according to the method of Lichtsteiner et al. (28). Finally, the precipitates from the nuclear extracts were dialyzed against 25 mM Hepes (pH 7.6), 0.1 mM EDTA, 40 mM KCl, 10% glycerol, and 1 mM dithiothreitol. HepG2 nuclear extracts were prepared by the following procedure. Cells were washed with three times with phosphate-buffered saline, harvested, and centrifuged. The cell pellet (ϭ 1 volume) was suspended in 8 volumes of lysis buffer containing 20 mM Hepes (pH 7.9), 1 mM EDTA, 0.5 mM spermidine, 1 mM dithiothreitol, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, 1 g/ml pepstatin A, and 0.3 g/ml antipain, lysed by the addition of 3/100 volume of Nonidet P-40, and incubated for 5 min on ice. The lysate was centrifuged for 10 min at 12,000 rpm. The precipitates were suspended in 1 volume of 2 M KCl, incubated for 30 min on ice, and centrifuged for 30 min at 98,000 rpm. The supernatant was applied to Sephadex G-50 spun column equilibrated lysis buffer plus 0.05 M KCl, centrifuged, and collected. Ten micrograms of protein and 0.2 g of HepG2 nuclear extracts, as determined by the protein assay dye reagent kit (Bio-Rad), were electrophoresed by SDS-polyacrylamide gel electrophoresis containing 12% acrylamide. Proteins were electrophoretically transferred to nitrocellulose filter (Schleicher & Schuell). After transfer, GST-P and C/EBP␣ were visualized by immunoblot analysis using either GST-P antibody obtained commercially (Biotrin, Ireland) or rabbit polyclonal antibody against C/EBP␣, alkaline phosphatase-goat anti-rabbit IgG (Zymed Laboratories), and 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium phosphatase substrate system (Kirkegaard & Perry Laboratories). C/EBP␤ was also detected by immunoblot analysis using C/EBP␤ antibody (Santa Cruz Biotechnology, Inc.) and the ECL system (Amersham).
Gel Mobility Shift Assay-Nuclear extracts (10 g/reaction) were mixed with 6.25 l of 20 mM Tris-HCl (pH 7.5), 10% glycerol, 2 mM dithiothreitol, 20 mM EDTA, 2 g of poly(dI-dC), and radiolabeled GPS1 (10,000 cpm/0.2 ng). GST-P Silencer 4 (GPS4), which is a binding site for Silencer Factor A (SF-A), was used as a nonspecific competitor (9). Oligonucleotide sequences of GPS1 and GPS4 were as follows: The binding reaction was continued at 4°C for 1 h. When antibodies were used, the nuclear extracts were incubated with the antibody and other reagents at the same time. Each reaction mixture was loaded on a 6% nondenaturing polyacrylamide gel, electrophoresed at 150 V for 2 h, fixed with 10% methanol and 10% acetic acid, and autoradiographed. Radioactivity was measured using a Bioimage Analyzer BAS2000 (Fujifilm, Tokyo, Japan), and the shift ratio was calculated as counts of the DNA-protein complex divided by total counts.
Plasmid Constructions-For construction of Ϫ395GST-luciferase, the fragment from Ϫ395 to ϩ59 in the promoter and silencer region of the GST-P gene was inserted into the HindIII site in the luciferase vector, PGV-B (Toyo Ink Mfg. Co., Ltd.), according to standard protocol (29). Ϫ395(⌬GPS1)GST-luciferase generated by polymerase chain reaction was checked by sequencing by the dideoxy method (2). Deletion of nucleotides between Ϫ395 and Ϫ92 from Ϫ395GST-luciferase resulted in Ϫ91GST-luciferase.
To make 4xGPS1-91GST-luciferase, the synthesized GPS1 described above was multimerized to a 4-mer and cloned into the XbaI site of pUC18. The blunt fragment containing 4xGPS1 after SmaI and HincII digestion was ligated at the SmaI site of Ϫ91GST-luciferase.
Cell Culture and DNA Transfection-HepG2 cells, human hepatoma cell line, were cultured in minimal essential medium supplemented with 10% fetal bovine serum and passaged by trypsinization at confluence. Cells were transfected by the calcium phosphate co-precipitation technique described by Graham and van der Eb (30). For cotransfection experiments, 2 g of the luciferase reporter gene was combined with different amounts of C/EBP isoform expression plasmid. MSV-C/EBP␣, C/EBP␣ expression vector, was kindly provided by Dr. S. L. McKnight, and CMV-LAP, C/EBP␤-36 (LAP) expression vector, by Dr. U. Schibler. In each case, pBluescript (Stratagene) plasmid was used to bring the final amount of plasmid to 6 g. Luciferase activities were measured by PicaGene (Toyo Ink Mfg. Co., Ltd). All the transfection experiments were performed at least three times by using two or three different preparations of DNA, and the regulation ratio was derived from the mean values of the results. In some experiments, the transfection efficiency was checked by cotransfection with pRSVGAL, a eukaryotic expression vector which contained the Escherichia coli-␤-galactosidase (lacZ) structural gene controlled by Rous sarcoma virus long terminal repeat. ␤-Galactosidase activity was assayed as described (31). It was confirmed that the variation of transfection efficiency was less than 20%.

RESULTS
The Expression Profiles of C/EBP Isoforms during Hepatocarcinogenesis-SF-B (C/EBP␤), a member of the C/EBP family that binds specifically to GPS1, has been cloned from a normal rat liver gt11 cDNA library by a Southwestern method (9). It is known that C/EBP␤ binding sites are recognized by other members of the C/EBP family and the pattern of expression of C/EBP isoforms changes during cell differentiation (14). To examine the expression of C/EBP isoforms during hepatocarcinogenesis, carcinogenic rats were prepared according to the Solt-Farber procedure (Ref. 27 and Fig. 1). At the end of 7 or 8 weeks, rats were killed, the livers which had a large number of foci and nodules were excised, and then the cytosol fractions and nuclear extracts were prepared. As a control, rats were injected with saline and were fed basal diet. We first checked the reproducibility of the carcinogenic experiments. Western blotting analysis of the cytosol fractions with GST-P antibody indicated that GST-P protein was induced at the 7th and 8th weeks after diethylnitrosamine was injected, but no GST-P protein was detected in the control rats as described previously (Ref. 32 and Fig. 2A).
Then we performed Western blot analysis by using the nuclear extracts. For determination of expression profiles of C/EBP isoforms, antibodies specific to each of the C/EBP isoforms were used (Fig. 2, B and C). It is reported that C/EBP␣ and C/EBP␤ share the common feature that, besides the full- To examine the expression of C/EBP isoforms during hepatocarcinogenesis, carcinogenic rats were prepared according to the Solt-Farber procedure (27). Cytosol fractions and nuclear extracts of liver were prepared from the control or carcinogenic rats at the 7th and 8th weeks after NaCl or diethylnitrosamine treatment. BD, basal diet; PH, partial hepatectomy; S, time at which rats were sacrificed. length product, N-terminally truncated polypeptides are translated from the same mRNA by use of internal, in-frame AUG codons (33)(34)(35). The rat C/EBP␣ gene encodes two proteins, C/EBP␣-42 and C/EBP␣-30, whose molecular masses are 42 kDa and 30 kDa, respectively. Three kinds of C/EBP␤ are translated from a single C/EBP␤ mRNA, and these are liverenriched transcriptional-activator protein LAP FL (C/EBP␤-39), LAP (C/EBP␤-36), and liver-enriched transcriptional-inhibitory protein LIP (C/EBP␤-20).
C/EBP␣-42 was highly expressed in the control rat liver and drastically decreased during hepatocarcinogenesis (Fig. 2B). On the other hand, the expression of C/EBP␣-30, which was lower than that of C/EBP␣-42 in the normal liver, was slightly reduced in the carcinogenic liver. C/EBP␤-36 was detected in the normal rat liver, and C/EBP␤-39 and C/EBP␤-20 were undetectable. During hepatocarcinogenesis, quantities of C/EBP␤-36 proteins showed no change and C/EBP␤-39 and C/EBP␤-20 were not detected. Thus, C/EBP␤-36 is the main form of the C/EBP␤s. The C/EBP␤-36:C/EBP␤-20 ratio is estimated to be about 15 in adult normal rat liver (35). In our experiments, if C/EBP␤-20 exists in one-fifteenth the amount of C/EBP␤-36, C/EBP␤-20 may not be detectable. Nuclear extracts of HepG2 cells transfected with CMV-LAP (C/EBP␤-36) and CMV-LIP (C/EBP␤-20) were also prepared, analyzed by SDS-polyacrylamide gel electrophoresis, and immunoblotted with a C/EBP␤ peptide antibody. This antibody reacted with C/EBP␤-36 and C/EBP␤-20. One 45-kDa immunoreactive band of rat liver nuclear extracts and two bands of HepG2 nuclear extracts are presumed to be cross-reacting materials. Immunoreactive bands of HepG2 nuclear extracts may be derived from human C/EBP␤ (NF-IL6). No C/EBP␦ was detected in the normal and carcinogenic liver nuclear extracts (data not shown).
Identification of C/EBP␣ as the Major GPS1 Binding Protein in Normal Rat Liver-Members of the C/EBP family can form homodimers and heterodimers that exhibit similar binding specificities (11,14,15,17). To analyze trans-acting factors binding to GPS1 of the GST-P silencer, we prepared nuclear extracts from normal rat liver for gel mobility shift assay (Fig. 3). When the normal liver nuclear extracts were incubated with the labeled GPS1, two different mobility shift complexes (complex I and complex II) were detected. It is likely that complex II consists of more than two bands. Competition analyses using the specific competitor, GPS1, and the nonspecific competitor, GPS4, indicate that these two complexes contain binding factors specific to GPS1 (Fig. 3, lanes 2-5).
To analyze further which members of the C/EBP protein family were involved in these complexes, antibodies against each of these proteins were used in gel mobility shift analysis. When extracts were incubated with antibody specific to C/EBP␣, both complex I and complex II almost disappeared. C/EBP␤ antibody also decreased the activity of the complex I and complex II. Preimmune serum and anti-C/EBP␦ antiserum did not affect the mobility of the complexes. Owing to the high sequence conservation within the dimerization region of the different C/EBP family members, the several proteins encoded by c/ebp␣ and c/ebp␤ genes could form many homodimers and heterodimers. Although we could not clarify which isoform of C/EBP family composed two complexes, we revealed that these complexes contain both C/EBP␣ and C/EBP␤.
Profile of GPS1 Binding Proteins during Hepatocarcinogen-  3. GPS1 formed complexes with C/EBP␣ and C/EBP␤ in the normal rat liver nuclear extracts. Gel shift analyses of normal rat liver nuclear extracts were performed with GPS1 as a probe and with or without antibodies specific for three isoforms of C/EBP. GPS1 or GPS4 was used as a competitor. Lane 1, probe only; lane 2, control (without competitor); lanes 3 and 4, specific competitor GPS1 was used at 50-and 250-fold molar excess, respectively; lane 5, nonspecific competitor GPS4 was used at 250-fold molar excess; lanes 6 -9, nuclear extracts plus preimmune serum, anti-C/EBP␣, anti-C/EBP␤, and anti-C/EBP␦, respectively; lanes 10 -13, probe plus preimmune serum, anti-C/EBP␣, anti-C/EBP␤, and anti-C/EBP␦, without nuclear extracts, respectively. The relative shift ratios (control ϭ 100) of complex I and complex II were shown at the top of the panel.
FIG. 4. GPS1 formed complexes with C/EBP␣ and C/EBP␤ in the liver nuclear extracts from carcinogenic rat. Labeled GPS1 was incubated with nuclear extracts obtained at the 7th (lanes 2-6) and 8th (lanes 7-11) weeks after diethylnitrosamine treatment and with antibodies specific to C/EBP isoforms. Lane 1, probe only; lanes 2 and 7, control (without antibody); lanes 3 and 8, nuclear extracts plus preimmune serum; lanes 4 and 9, nuclear extracts plus anti-C/EBP␣; lanes 5 and 10, nuclear extracts plus anti-C/EBP␤; lanes 6 and 11, nuclear extracts plus anti-C/EBP␦. The relative shift ratio (control ϭ 100) was shown at the top of the panel.
esis-To analyze the factors which bind to GPS1 in nuclear extracts prepared from the carcinogenic rat liver, we carried out a gel mobility shift assay (Fig. 4). While incubation of radiolabeled GPS1 with normal rat liver nuclear extracts generated two retarded complexes (Fig. 3), complex I completely disappeared from hepatocarcinogenic rat livers (Fig. 4. lanes 2  and 7). The isoforms of the C/EBP family that recognize GPS1 were examined by the gel mobility shift analysis using the specific antibody. Complex II derived from hepatocarcinogenic rat liver nuclear extracts with GPS1 also involved C/EBP␣ and C/EBP␤, but not C/EBP␦ (Fig. 4, lanes 4 -6 and 9 -11).
Transcriptional Regulation of C/EBP␣ and C/EBP␤-36 (LAP) on GPS1 in the GST-P Silencer-GPS1 was identified as a negative element in the GST-P silencer (8,9). However, Figs. 3 and 4 show that trans-activators, C/EBP␣ and C/EBP␤-36 (LAP), bind to GPS1. To clarify the effect of transcriptional regulation of C/EBP␣ and C/EBP␤-36 (LAP) on GPS1 in the GST-P silencer, effector plasmids were cotransfected with Ϫ395GST-luciferase (which has the GST-P silencer and promoter) into the HepG2 cells. As a control, Ϫ395(⌬GPS1)GSTluciferase lacking the GPS1 element from the GST-P silencer was utilized (Fig. 5A). To remove the nonspecific effects of the effector plasmid to unknown sites in the reporter plasmid, the regulation ratio was calculated as follows. First, the relative activity was determined by dividing the luciferase activity with the effector plasmid by that without the effector plasmid, and then the regulation ratio was calculated by dividing the relative luciferase activity in the presence of the GPS1 element by that in its absence.
Surprisingly, C/EBP␣ repressed the transcription through GPS1 in the GST-P silencer. The repressing activity depended on the input amount of the C/EBP␣ expression vector. On the contrary, C/EBP␤-36 (LAP) acted as a transcriptional activator on GPS1 (Fig. 5A). We investigated further whether C/EBP␣ can repress the transcription activated by C/EBP␤-36 (LAP). The reporter plasmids were cotransfected with both C/EBP␣ and C/EBP␤-36 (LAP) expression vectors into the HepG2 cells. C/EBP␣ repressed not only the basal level transcription, but also C/EBP␤-36 (LAP)-activated transcription through GPS1 in the GST-P silencer (Fig. 5A).
Effect of trans-Activation of C/EBP␣ and C/EBP␤-36 (LAP) on Multimerized GPS1-C/EBP␣ generally acts as a transcriptional activator. But C/EBP␣ repressed the transcription through GPS1 in the GST-P silencer. To make sure that C/EBP␣ activates the transcription through its binding site, C/EBP␣ and C/EBP␤-36 (LAP) expression vectors were cotransfected with the reporter plasmids containing the multimerized GPS1 with GST-P promoter (Fig. 5B). Both C/EBP␣ and C/EBP␤-36 (LAP) activated transcription on multimerized GPS1. Thus, C/EBP␣ repressed the transcription through GPS1 in the GST-P silencer, but acted as a transcriptional activator on multimerized GPS1. DISCUSSION Gene expression in the liver appears to be controlled by combinatorial mechanisms derived from the liver-enriched transcription factors, including C/EBP␣ and C/EBP␤. The GST-P gene is strongly and specifically expressed during hepatocarcinogenesis, but neither mRNA of the GST-P gene nor its protein is detectable in the normal liver (6,7). We have previously cloned SF-B (C/EBP␤), which binds to GPS1, one of the multiple negative elements of the GST-P silencer (8). Members of the C/EBP family can form homodimers and heterodimers that exhibit similar binding specificities (11,14,15,17). Expression of the C/EBP proteins is important for cell differentiation and growth (14). To elucidate the function of GPS1, we analyzed the expression profiles of the C/EBP proteins during hepatocarcinogenesis and the transcriptional regulation of GPS1 by members of the C/EBP family.
To analyze trans-acting factors binding to GPS1, which is found within the GST-P silencer, we prepared nuclear extracts from normal and carcinogenic rat livers for Western blot and gel mobility shift assays. The expression of the C/EBP␣-42 was drastically reduced and C/EBP␣-30 slightly decreased during hepatocarcinogenesis. C/EBP␤-36 (LAP) was detected in the normal and carcinogenic rat liver, and C/EBP␤-39 (LAP FL ) and C/EBP␤-20 (LIP) were undetectable. Together with the results of gel shift assays, it is suggested that C/EBP␣-42 and C/EBP␣-30 mainly occupy the GPS1 site in the normal rat liver and act as negative regulators. The binding activity of C/EBP␣ to GPS1 reduced significantly in the liver during hepatocarcinogenesis, while the activity of C/EBP␤ did not change. Although little is known about the repression activity of C/EBP␣, repressor and attenuator domains have been identified within the trans-activation domains of C/EBP␣ (36,37). Furthermore, in the normal liver, C/EBP␣ occupies the acute phase-responsive element in the ␣ 1 -acid glycoprotein gene which is only expressed at a very low level (24).
We performed the transfection analysis in HepG2, a human hepatoma cell line. Expression level of C/EBP␤ is higher than that of C/EBP␣ in HepG2 cells, which was generally used to observe the effect of the transcriptional regulation of C/EBP␣ and C/EBP␤ on its binding site (35,38). The observed levels of repression and activation are weak, since endogenous factors respond to GPS1. GPS1 in the GST-P silencer showed repression activity with C/EBP␣ and activation activity with C/EBP␤-36 (LAP). These data suggest that GPS1 acts as a silencer or an enhancer depending on its binding factors. Furthermore, C/EBP␣ attenuated the transcriptional activation by C/EBP␤-36 (LAP). Therefore, C/EBP␣ acts as a negative regulator rather than a transcriptional activator on GSP1 in the GST-P silencer, although we do not know the repression mechanisms of C/EBP␣. These observations strongly suggest that the ratio of C/EBP␣ to C/EBP␤ is an important factor for silencer activity in the GST-P gene promoter, and the decrease of this ratio during hepatocarcinogenesis reduces the silencer activity and, consequently, increases the GST-P expression.
Overexpression of C/EBP␣ and C/EBP␤-36 (LAP) in HepG2 cells increased the reporter activity driven by GPS1 multimers. C/EBP␤-36 (LAP) mediates the transcriptional activation of the GPS1 located both in the silencer and multimerized GPS1. However, the function of C/EBP␣ is dependent on the factors which bind to cis-element around GPS1, since overexpression of C/EBP␣ decreased reporter activity driven by GPS1 in the GST-P silencer and mediated the transcriptional activation of the multimerized GPS1. It is important that C/EBP family members interact, directly or indirectly, with other trans-act-ing factors. For example, human C/EBP␤ (NF-IL6) and glucocorticoid receptor (GR) synergistically activate transcription of the rat ␣ 1 -acid glycoprotein gene via direct protein-protein interaction which is probably generated from the bZIP domain of C/EBP␤ (NF-IL6) and near the zinc finger region of GR (39). Analyses of the gene expression of human serum amyloid A, interleukin 6, and interleukin 8 have demonstrated that C/EBP␤ (NF-IL6) and NF-B transcriptionally synergize with each other (25,40). LeClair et al. (41) reported that the NF-B p50 subunit physically interacts with the bZIP domain of C/EBP␤ (NF-IL6). The bZIP domain, which is highly conserved among C/EBP family members, is important for the proteinprotein interaction.
In addition to SF-B binding to the GPS1 site in the GST-P silencer, SF-A and SF-C also bind to the other cis-elements in this region and function as a negative regulator. SF-A binds to GPS4 with a strong affinity and several other cis-elements in the silencer region with a weak affinity. SF-C binds to the GPS2 site, which partially overlaps GPS1 and GPS3, binding sites of SF-B and SF-A, respectively (8). We previously reported that SF-A, SF-B, and SF-C all might be important for maximum silencer activity (8). Characterization of the protein-protein interactions, including these factors, is required for further elucidation of the mechanism of repression in normal liver and activation during hepatocarcinogenesis.