Activating Transcription Factor 2 (ATF2) Down-regulates Hepatitis B Virus X Promoter Activity by the Competition for the Activating Protein 1 Binding Site and the Formation of the ATF2-Jun Heterodimer*

The hepatitis B viral X promoter is known to be positively autoregulated by its own HBx protein, which also interacts with many cellular regulatory proteins. We investigated the effect of activating transcription factor 2 (ATF2) on the activity of the X promoter. Cotransfection of the ATF2 expression vector with a X promoter-chloramphenicol acetyltransferase plasmid repressed the X promoter activity in HepG2 cells. HBx activated activating protein 1 (AP-1)-mediated transcription through the hepatitis B virus E element by 35-fold, while its activation activity was inhibited in the presence of ATF2, suggesting that ATF2 inhibited the autoactivation of X promoter by HBx and basal transcription mediated by AP-1. Since the binding sites of AP-1 and ATF2 in the hepatitis B virus E element overlap, the repression of X promoter activity by ATF2 is exerted by the competition for the AP-1 binding site and the formation of the ATF2-Jun heterodimer as in the case of the consensus AP-1 element. However, the small X promoter had a ATF2 binding site and was activated by ATF2. These results suggest that the syntheses of X proteins are differentially regulated by ATF2.

The hepatitis B viral X promoter is known to be positively autoregulated by its own HBx protein, which also interacts with many cellular regulatory proteins. We investigated the effect of activating transcription factor 2 (ATF2) on the activity of the X promoter. Cotransfection of the ATF2 expression vector with a X promoterchloramphenicol acetyltransferase plasmid repressed the X promoter activity in HepG2 cells. HBx activated activating protein 1 (AP-1)-mediated transcription through the hepatitis B virus E element by 35-fold, while its activation activity was inhibited in the presence of ATF2, suggesting that ATF2 inhibited the autoactivation of X promoter by HBx and basal transcription mediated by AP-1. Since the binding sites of AP-1 and ATF2 in the hepatitis B virus E element overlap, the repression of X promoter activity by ATF2 is exerted by the competition for the AP-1 binding site and the formation of the ATF2-Jun heterodimer as in the case of the consensus AP-1 element. However, the small X promoter had a ATF2 binding site and was activated by ATF2. These results suggest that the syntheses of X proteins are differentially regulated by ATF2.
Hepatitis B virus (HBV) 1 is closely related not only to acute and chronic hepatitis, but also to the development of hepatocellular carcinoma (reviewed in Refs. 1 and 2). HBV, unlike the other DNA viruses, encodes reverse transcriptase and replicates through reverse transcription of pregenomic RNA (3,4). The production of pregenomic RNA is regulated by a combined interaction of HBV enhancers (5,6). One of the HBV genes, X, encodes a basic protein of 154 amino acids. It has been implicated in the carcinogenicity of this virus and is regarded as a major causative factor due to its ability to induce hepatocellular carcinoma in transgenic mice (7). The X protein exhibits a transcriptional activation function for many viral and cellular genes without binding to DNA (8) (reviewed in Ref. 9). In addition, a 3Ј-truncated X gene-cell fusion product from integrated HBV DNA in chronic hepatitis tissue (10) and a 3Јelongated X product of 193 amino acid (21 kDa) from mutant HBV DNA (11,12) were shown to exert a transactivation function. Smaller forms of the X protein, initiated at the second and the third ATG codons of the X ORF, are also important in the transacting function. All three forms of the X protein can individually transactivate a class III promoter, which transcribes the tRNA and 5 S rRNA by RNA polymerase III. In contrast, transactivation of several different class II promoters, which transcribe the mRNAs by RNA polymerase II, displays various requirements for the different X proteins (13). The small X promoter expresses the smaller X gene transcripts that arise within the X ORF and whose heterogeneous 5Ј ends straddle the second ATG codon. A positive regulatory element is also known to be necessary for the efficient transcription from the small X promoter (14).
One of the HBV enhancers, the EnI enhancer, partially overlaps with the X gene promoter and is located between the S gene and the X gene coding sequences (15,16) (Fig. 1). This enhancer is composed of at least five different factor binding sites named 2C, GB, EP, E, and NF1 sites (17)(18)(19)(20)(21). A liverenriched transcription factor, HNF3, binds to the 2C site, and HNF4, RXR␣ and COUP-TF bind to the GB element in vitro (20). Nuclear factor EF-C and NF1 bind to the inverted repeat of the EP element and NF1 site, respectively (22). The E element is bound by ATF2/CREB and AP-1, and also mediates the transactivation of X promoter activity by X protein. Deletion, linker scanning, and point mutation analyses of the E element revealed that mutation of the E element reduced X promoter activity by 20 -50% (23,24).
We investigated the functional interaction between binding proteins on the E element and the functional implication of X protein in the regulation of the X promoter. Cotransfection of an ATF2 expression vector with a X promoter-chloramphenicol acetyltransferase (CAT) construct repressed the X promoter activity. Etkcat, a heterologous plasmid containing the E element in front of the tk promoter, was used as a reporter for combinational cotransfection of ATF2, c-Jun, c-Fos, and HBx expression vectors, which revealed that ATF2, by interfering with AP-1, inhibits basal transcription of the X promoter and the transactivation activity of HBx through AP-1. In addition, we also tested whether the small X promoter activity is influenced by ATF2. Although the X promoter was repressed by ATF2, the small X promoter had a ATF2 binding site and was stimulated by ATF2. We conclude that syntheses of X proteins are differentially regulated by ATF2. * This work was supported in part by grants from the Korea Ministry of Education and from the Korea Science and Engineering Foundation (KOSEF) through the Research Center for Cell Differentiation at Seoul National University. 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.

EXPERIMENTAL PROCEDURES
Plasmid Construction-The X promoter-CAT reporter plasmid (Xpcat) was constructed by insertion of the AccI (nt 1069)-BamHI (nt 1397) fragment of HBV-adr-k (25) in front of the CAT gene. The mutant plasmid (XMpcat) has mutations in the ATF2 binding site of E element which were generated by oligonucleotide-directed in vitro mutagenesis (EM, 5Ј-CTGCCAAGTATTTGCTAATTCAACCCCCA-3Ј). The mutated sequences are underlined. The Etkcat was constructed by inserting pentamerized double stranded E (5Ј-CTGCCAAGTATTTGCTGACG-CAACCCCCA-3Ј) oligonucleotide into the SmaI site of pBLcat2. APtkcat and CREtkcat were constructed by inserting trimerized double stranded consensus AP-1 (5Ј-CGCTTGATGAGTCAGCCGGAA-3Ј) and CRE (5Ј-AGAGATTGCCTGACGTCAGAGAGCTAG-3Ј) oligonucleotides, purchased from Promega (Madison, WI), into the SmaI site of pBLcat2, respectively. Eukaryotic expression plasmids of HBx, the small X promoter-CAT construct and serial deletion constructs were described previously (14). The ATF2 eukaryotic expression vector was constructed by subcloning the full-length ATF2 cDNA into the pECE expression vector (26). The c-Jun and c-Fos expression vectors were kindly provided by K. Kim (Seoul National University, Korea). Bacterial expression vectors for ATF2 and c-Fos were constructed by inserting their cDNAs in frame into pGEX-3X. MBP-fused X, middle X, and small X expression vectors were obtained by inserting full-length and deleted fragments of X genes into the pMAL-c2 plasmid, resulting in proteins which have amino acid residues 1-154, 70 -154, and 105-154 of X protein, respectively. For the construction of expression plasmid of truncated c-Jun (pETc-jun⌬), the c-jun gene was subcloned into the vector pET3b (Stratagene) to generate the plasmid pETc-jun and subsequently deleted the DNA fragment between the two AvaI sites (nucleotides 220 and 679 of the c-jun gene), which resulted in the deletion of amino acids between 74 and 232 of c-Jun protein.
Immunoblot Assay-Proteins separated by SDS-polyacrylamide gel electrophoresis were electrophoretically transferred to nitrocellulose filters for 1 h at 20°C using 5 V/cm 2 . Blocking was carried out in phosphate-buffered saline (PBS) containing 5% defatted dry milk and 0.1% Tween 20, followed by washing and incubation of the membranes with antibodies in the same buffer without defatted dry milk. Polyclonal X antibodies were diluted 1:2,000 and used as primary antibodies. After incubation for 1 h at 25°C, the blot was further incubated for 1 h at 25°C with a horseradish peroxidase-linked anti-rabbit mouse antibody. Protein/antibody complexes were visualized by the enhanced chemiluminescence Western blotting detection system (Amersham Corp.) according to the manufacturer's instructions.
Transient Transfection and CAT Assay-Transient transfection and CAT assay were performed as described previously (14). Approximately 12 h prior to transfection HepG2 cells were plated at a density of 1 ϫ 10 6 /60-mm diameter plate. Transfection of plasmid DNA into HepG2 cells was carried out by the calcium phosphate coprecipitation method as described previously (27). Typically, 2 g each of reporter and effector plasmids were used. The total amount of transfected DNA was always adjusted to 12 g with pUC19. After 42 h, cells were harvested, and CAT assays were performed. Cell extracts were normalized to the total amount of protein, as determined by the Bradford assay (Bio-Rad). CAT enzymatic activity was quantitated by measuring conversion of chloramphenicol to its acetylated forms using the BAS radio analytic imaging system according to manufacturer's instructions. All experiments were repeated at least three times.
Mobility Shift Assay-Escherichia coli cells were grown in LB medium containing 50 g/ml ampicillin. Subsequently the cells were induced under vigorous shaking conditions with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside for 3 h at 30°C. GST fusion proteins were purified according to the protocol from Pharmacia Biotech Inc., and MBP-fused X proteins were purified according to the Short Protocols in Molecular Biology (28). DNA binding reactions were carried out in a 15-l volume, which typically contains 10 mM Hepes (pH 7.9), 60 mM KCl, 5 mM MgCl 2 , 1 mM dithiothreitol, 1 mM EDTA, 10% glycerol, 500 ng of poly(dI-dC) (Sigma), and 2 g of bovine serum albumin. Prior to the reaction with DNA, 100 ng each of c-Jun, GST-Fos, GST-ATF2, and MBP-HBx proteins were mixed and incubated for 20 min. Antibodies against ATF2 and c-Jun were from Santa Cruz Biotech. Inc. (Santa Cruz, CA). DNA binding was started by adding 10,000 cpm of probe to the preincubated reaction mixture with incubation for 15 min at room temperature. Samples were loaded on a 4% polyacrylamide gel (acrylamide:bisacrylamide ϭ 60:1) in 0.5 ϫ TBE (44 mM Tris, 44 mM boric acid and 1 mM EDTA). After electrophoresis, gels were dried and exposed to x-ray film.
In Vitro Transcription and Translation-The pETc-jun⌬ plasmid was linearized and used for the coupled in vitro transcription and translation reaction. Truncated c-Jun protein was synthesized in rabbit reticulocyte lysates programmed with the in vitro transcribed mRNA in the presence of unlabeled methionine according to the protocol recommended by the manufacturer (Promega). DNase I Footprinting Analysis-DNase I footprinting analysis was performed as described previously (14) with some modifications as follow. Briefly, the SacII site (nt 1445) was cut and then radiolabeled by incubation with [␥-32 P]ATP (specific activity, 7000 Ci/mmol) after dephosphorylation with calf intestinal phosphatase. The radiolabeled DNA fragment was cut with DraI (nt 1719), fractionated on an 6% polyacrylamide gel and isolated by electroelution for DNase I footprinting.

ATF2 Represses the Hepatitis B Viral X Promoter through the E Element of Enhancer I-
To study the effect of ATF2 on the activity of the HBV X promoter, an expression plasmid encoding ATF2 cDNA was cotransfected into HepG2 cells together with a X promoter-CAT construct (Xpcat), or a construct having mutations in the ATF2 binding site of the E element (XMpcat). As shown in Fig. 1A, ATF2 repressed CAT expression from the X promoter two fold in HepG2 cells, but not with a construct having a mutation in the ATF2 binding site. These data suggest that ATF2 repressed X promoter activity through the ATF2 binding site. Since ATF2 is known to be an activator, we investigated the effect of ATF2 on the consensus CRE ciselement in a heterologous promoter system. The oligonucleo-

FIG. 1. The effects of ATF2 on the hepatitis B viral X promoter.
A, schematic diagram of EnI and X promoter is shown at the top, in which representative cis-elements of 2C, GR, EP, E, and NF1 are shown. The histogram represents the relative levels of CAT activity in extracts of HepG2 cells cotransfected with 2 g each of reporter plasmids and 5 g of ATF2 expression plasmid as indicated. B, CAT assay using the heterologous reporter plasmids. The ATF2 expression plasmid was cotransfected with heterologous reporter plasmids, Etkcat and CREtkcat, which have five copies of E oligonucleotide and three copies of consensus CRE oligonucleotide, respectively, in front of tk promoter. Two micrograms and one microgram of Etkcat and CREtkcat, respectively, were cotransfected into HepG2 cells with 5 g of ATF2 expression vector as indicated.
tides of E and consensus CRE were multimerized and inserted in front of the tk promoter. The resulting plasmids were cotransfected with the ATF2 expression vector. ATF2 activated CRE-mediated expression from the tk promoter 8-fold, whereas it repressed E element-mediated expression from the tk promoter (Fig. 1B). It is well known that ATF2 activates CREmediated transcription (29). These results imply that other transcription factors may be the target for repression of the X promoter by ATF2.
ATF2 Represses AP-1 and HBx-mediated Transcriptional Activation of the HBV X Promoter-The sequence of the E element (5Ј-TGACGCAA-3Ј) is almost identical to consensus CRE and AP-1 binding sites. Also, the bindings of AP-1 and ATF2 on this element were reported previously (30,31). To see the repression mechanism of ATF2, we cotransfected c-Jun, c-Fos, ATF2, and HBx expression vectors in various combinations into HepG2 cells with Etkcat as a reporter plasmid, and asked whether ATF2 can inhibit AP-1 activity and HBx-mediated transcriptional activation. The Jun homodimer and Jun/Fos heterodimer activated the reporter plasmid about 9-and 7-fold, respectively ( Fig. 2A, lanes 3 and 5, respectively). The activation property was increased with coexpression of HBx 25-and 35-fold, respectively ( Fig. 2A, lanes 11 and 13, respectively). But the expression of Fos alone exhibited no significant activation with or without the expression of HBx protein ( Fig. 2A,  lanes 12 and 4). Cotransfection of ATF2 expression vector repressed the transactivation by Jun and Jun/Fos ( Fig. 2A, compare lanes 3 and 5 with lanes 7 and 9, respectively). This repression of Jun and Jun/Fos by ATF2 was also shown in the additional activation by HBx (Fig. 2A, compare lanes 11 and 13  with lanes 14 and 16, respectively). It is of interest that ATF2 repressed the activity of the Jun/Fos heterodimer ( Fig. 2A,  lanes 9 and 16) more than that of the Jun homodimer ( Fig. 2A,  lanes 7 and 14) especially with additional activation by HBx.
To assess this observation, we explored the physical interaction among binding proteins on the E element with a mobility shift assay. The same amounts of GST-Fos, GST-ATF2, c-Jun, and MBP-HBx were used for the assays. Jun or Jun/Fos by themselves bound to the E element, although binding efficiency was 5-to 7-fold lower (Fig. 2B, lanes 1 and 2) than that of the consensus AP-1 element (Fig. 3A, lanes 1 and 2). But the addition of ATF2 protein resulted in the diminished binding of Jun or Jun/Fos protein (Fig. 2B, lanes 4 and 5). In contrast, Jun alone decreased the binding of ATF2 on the E element (Fig. 2B,  lane 4), whereas Jun/Fos did not (Fig. 2B, lane 5). Addition of anti-Jun antibody to the mixtures of ATF2 and Jun, or ATF2 and Jun/Fos resulted in the unchanged binding of ATF2 and no supershifted complex (Fig. 2B, lanes 6 and 8). But the DNAprotein complexes were supershifted by anti-ATF2 antibody (Fig. 2B, lanes 7 and 9). These data demonstrate that the binding of ATF2 to the E element hindered the binding of AP-1 (Jun/Fos) probably due to the overlapping binding sites in the E element. For the demonstration of the ATF2's stronger affinity for the E element than that of AP-1, we examined the effect of increasing amounts of ATF2 on the binding of AP-1. When the amount of ATF2 increase from 20 ng to 100 ng to the fixed 100 ng of AP-1 protein, less than 50 ng of ATF2 could completely abolish the binding of AP-1 to the E element (Fig. 2C). These results clearly showed that the ATF2 has a stronger affinity for the E element than that of AP-1. MBP-HBx augmented the binding of ATF2 on the E element, but could not induce the binding of Jun or Jun/Fos in the presence of ATF2 (Fig. 2B, lanes 10 and 11).
ATF2 Represses AP-1 and HBx-mediated Transactivation by the Protein-Protein Interaction and the Competition for the Overlapping Sites-To examine the change of AP-1-mediated transcription by ATF2, which does not bind well to the consensus AP-1 site, transient transfection and mobility shift assays were performed (Fig. 3). Jun and Jun/Fos bound to the AP-1 site efficiently (Fig. 3A, lanes 1 and 2), which was also confirmed by a supershift by anti-Jun antibody (Fig. 3A, lane 5), but incubation of ATF2 with Jun prior to the addition of the probe did not alter the mobility of the complex (Fig. 3A, lanes 3  and 4) and no super-shifted band was detected with anti-ATF2 antibody (Fig. 3A, lane 6). This result suggests that ATF2/Jun heterodimer as well as ATF2 homodimer did not bind well to the consensus AP-1 site. However, the addition of increasing amounts of ATF2 decreased the binding of Jun homodimer on the AP-1 site (Fig. 3A, lanes 7-10). These results suggest that the formation of the ATF2/Jun heterodimer decreased the formation of the Jun homodimer and consequently diminished the binding of Jun homodimer on the AP-1 site. We next examined the effect of ATF2 on the transcription mediated by AP-1 binding sites (Fig. 3B). The decreased expression from the tk promoter was observed by cotransfection of the ATF2 expression

FIG. 2. ATF2 represses the activation activity of AP-1 and HBxmediated activation.
A, transcriptional activities of Etkcat were measured in HepG2 cells. Two micrograms of Etkcat reporter plasmid were cotransfected into HepG2 cells with 2 g each of plasmids expressing c-Jun, c-Fos, ATF2, and HBx proteins as indicated. Cell extracts from transfected cells were normalized to the total amount of protein as determined by Bradford assay (Bio-Rad). CAT assays were repeated four times and a representative CAT assay is shown. B, binding assay with c-Jun, c-Fos, and ATF2 on the HBV E element. Bacterially expressed and purified c-Jun, GST-c-Fos, GST-ATF2, and MBP-HBx were incubated in the presence of 2 g of bovine serum albumin and 500 ng of poly(dI-dC) for 15 min at room temperature as indicated, followed by adding a probe. The same amount (100 ng) of each protein was used in the assay. For antibody supershift analysis, 1 l of each antibody was added and further incubated for 15 min at room temperature prior to electrophoresis. C, competition assay between ATF2 and AP-1 for the binding on HBV E element. Mobility shift assay was accomplished for the E element as a probe with increasing amounts of ATF2 (0, 20, 50, and 100 ng) to the fixed 100 ng of AP-1 (50 ng each of c-Jun and GST-Fos).
vector with the APtkcat reporter, even though ATF2 alone could not bind to the AP-1 site. Coexpression of Jun/Fos augmented the AP-1-mediated transcription from tk promoter four-fold (Fig. 3B, lane 3). The activation by Jun/Fos was further enhanced by cotransfection with the HBx expression vector (Fig. 3B, lane 4). However, coexpression of ATF2 diminished the activation by Jun/Fos and the additional activation by HBx (Fig. 3B, lanes 5 and 6). In summary, ATF2 could also repress AP-1 activity in a manner consistent with a proteinprotein interaction.
It was also tested if the binding activity of ATF2 and c-Jun on the HBV E element is affected by protein-protein interaction. To discriminate the DNA-protein complexes of c-Jun, ATF2, and ATF2/c-Jun heterodimer, the truncated c-Jun between amino acids 74 and 232 was obtained by in vitro translation and used for the analysis. The truncated c-Jun (c-Jun⌬) contains the intact DNA-binding domain and leucine zipper domain which is responsible for dimerization with other proteins. The ATF2/c-Jun⌬ heterodimer, which migrates between c-Jun⌬ and ATF2, bound well on the CRE element (Fig. 3C,  lanes 3 and 4) but did not bind efficiently on the HBV E element (Fig. 3C, lanes 8 and 9). Therefore, ATF2 could repress transcription mediated by cis-elements such as consensus AP-1 site and HBV E element on which ATF2/c-Jun heterodimer do not bind efficiently.
Domain Analysis of the HBx Protein Responsible for Interaction with ATF2-We investigated whether HBx affects the binding of Jun, Jun/Fos and ATF2 on the E element (Fig. 2B,  lanes 10 and 11). HBx enhanced the binding affinity of ATF2 to the E element 5-10-fold (Fig. 4A, lanes 4 and 5). Smaller forms of X protein, which have the C-terminal domain in common, were reported to activate class II and III promoters (13,32).
Smaller forms of X protein were also examined for interaction with ATF2 (Fig. 4). Increasing amounts of full-length X (Fig.  4A, lanes 4 and 5) and middle X proteins (Fig. 4A, lanes 6 and  7) increased the binding of ATF2 on the E element. However, small X protein (Fig. 4A, lanes 8 and 9) increased the ATF2binding only slightly. To see whether ATF2 and X proteins can interact with each other in the absence of a cis-element, X proteins were incubated with glutathione-agarose beads containing immobilized GST-ATF2, and then washed extensively with PBS three times. The interacted proteins were eluted with PBS containing 0.1% triton X-100, and electrophoresed in a 8% SDS-polyacrylamide gel. Bands corresponding to the fulllength X and middle X proteins were detected with anti-HBx antibody (Fig. 4B, lanes 2 and 3). However, only a faint band of small X protein was detected by the Western analysis (Fig. 4B,  lane 4). These observations agree with that of the mobility shift assay (Fig. 4A). Middle X and small X proteins have amino acid residues 70 -105 and 105-154 of full length X protein, respectively. Therefore, amino acid residues 70 -105 of X protein participates in the interaction with ATF2. This domain of X protein corresponds to one of three domains that are known to interact with cellular proteins (33). These data also indicate that X protein can interact with ATF2 in the absence of a proper cis-element.
ATF2 Transactivates the Small X Promoter-We next investigated the effect of ATF2 on the activity of the small X promoter. A series of small X promoter deletion constructs, described previously (14), was cotransfected into HepG2 cells with the ATF2 expression vector. As shown in Fig. 5A, expression from pHH115, which has only a minimal promoter region, was still activated by ATF2 (Fig. 5A, lane 10), suggesting that FIG. 3. The effects of ATF2 on the consensus AP-1 binding site. A, binding of Jun/Fos on the consensus AP-1 binding site in the presence of ATF2. Bacterially expressed GST-ATF2, c-Jun, and GST-Fos were incubated with 32 P-labeled AP-1 oligonucleotide as a probe. The amounts of each protein used were 100 ng for lanes 1-6. In lanes 7-10, the amounts of ATF2 used were 0, 50, 100, and 150 ng, respectively, to the fixed 50 ng of c-Jun. The detailed procedure of mobility shift assay and antibody supershift assay were the same as described in the legend of Fig.  2. B, the effect of ATF2 on the AP-1-mediated tk promoter activity. One microgram of APtkcat reporter plasmid was cotransfected into HepG2 cells with 2 g of each plasmid expressing c-Jun, c-Fos, ATF2, and HBx as indicated. C, mobility shift assay using ATF2/c-Jun⌬ heterodimer with CRE and HBV E element as probes. Five microliter of in vitro translated c-Jun⌬, 100 ng of ATF2 protein (lanes 2, 3, 7, and 8) and 200 ng of ATF2 (lanes 4, 5, 9, and 10) were used. The mixture of ATF2 and c-Jun⌬ was incubated for 20 min at 30°C for the efficient formation of heterodimer, followed by adding a probe. For supershift analysis, anti-ATF2 antibodies were added after addition of the probe (lanes 5 and 10).
the region between nt 1567 and 1679 is responsible for the activation by ATF2. This region corresponds to the minimal small X promoter and was already known to have several transcription factor binding sites, including HNF4 and Sp1 (34,35). To test whether this region has the binding site for ATF2, a 115 base pair HpaII-HincII DNA fragment was radiolabeled and used as a probe for a mobility shift assay (Fig. 5B). Incubation of purified GST-ATF2 with DNA probe resulted in a specific DNA-ATF2 complex which competed with a 50-fold molar excess of E element, but not with the same amount of mutant oligonucleotide, EM, indicating that ATF2 binds specifically to the minimal small X promoter. To know the exact binding site of ATF2 on the small X promoter, DNase I footprinting analysis was performed with SacII-DraI DNA fragment (nt 1145-1719) as a probe. The protected region by ATF2 spanned from nt 1638 to 1656, and have a sequence (TTA-CATAA) similar to the ATF2 binding site of E-selectin promoter (TGACATCA) (36). These results suggest that this protected region of the small X promoter is responsible for the activation by ATF2.

DISCUSSION
Two enhancers, EnI and EnII, have been identified in the HBV genome (15,16,35,37,38). EnI is located in the region downstream of the HBsAg ORF and upstream of the X ORF. Therefore, the transcription from the X promoter was largely dependent on the EnI activity. EnI is composed of multiple functional elements that act synergistically to stimulate transcription. Two regions of E and EF-C in particular appear to be important for EnI to exhibit a transacting function (17,22,39). In this report, we analyzed the transcriptional regulation of X promoter mediated by the E element of EnI. ATF2, one of the binding proteins of the E element, repressed AP-1 and HBxmediated transcriptional activation of the X promoter. It was especially of interest that EnI activity was negatively regulated by the transcription factor ATF2, which is known to be a general activator. Since AP-1 and ATF2 could bind to the same site of the E element independently and were known to interact with each other, it is likely that the down-regulation of the X promoter by ATF2 occurs via the direct competition for factor binding sites and the formation of the ATF2/c-Jun heterodimer B, each of MBP-HBx, MBP-Mx, MBP-Sx, and MBP proteins was incubated with glutathione-agarose beads containing immobilized GST-ATF2, and then washed extensively with PBS three times. The interacted proteins were eluted with PBS containing 0.1% Triton X-100, and electrophoresed in a 8% SDS-polyacrylamide gel. After electrotransfer of proteins to a nitrocellulose membrane, the proteins were detected by anti-X antibody as described under "Experimental Procedures." FIG. 5. The effect of ATF2 on the small X promoter. A, a series of deletion constructs containing the regulatory region of small X promoter was cotransfected into HepG2 cells with the ATF2 expression vector as indicated. For the relatively low promoter activity, 2 g each of pBH284 and pXH245 plasmids, and 5 g each of pXH215, pXH172, and pHH115 reporter plasmids were cotransfected into HepG2 cells with 10 g of ATF2 expression vector. CAT assay was performed as described in Experimental Procedures. B, binding of ATF2 on the HpaII-HincII (nt 1567-1679) DNA fragment. The HpaII-HincII DNA fragment, which corresponds to the promoter region of pHH115 reporter plasmid, was excised, eluted and radiolabeled with [␥-32 P]ATP after dephosphorylation with calf intestinal alkaline phosphatase. The GST-ATF2 was incubated with the radiolabeled HpaII-HincII DNA fragment as a probe. For competition analysis, 50 ng each of E (lane 2) and EM (lane 3) oligonucleotides was added in the reaction mixture. C, DNase I footprinting analysis of small X promoter region with ATF2. The SacII site (nt 1445) was cut and then radiolabeled by incubation with [␥-32 P]ATP after dephosphorylation with calf intestinal phosphatase. The radiolabeled DNA was excised with DraI (nt 1719), eluted and used as a probe. Serial increasing amounts of ATF2 (0.5 and 60 g) was used for analysis as indicated. Protected region by ATF2 is denoted as a solid bar.
which can not bind efficiently to the E element. Though some genes have an autonomous silencer element (40), others may be repressed by the displacement of transactivator proteins from the promoter element. In the case of the sea urchin histone H2B-1 gene, a displacement protein appears to sterically prevent interaction of the CAAT-binding protein (41). Direct protein-protein interaction between members of different transcription factor families may be a general mechanism by which a limited number of transcription factors can specifically regulate a large number of genes. Gene repression by proteinprotein interaction was also observed in the HBV genes. Addition of Jun, but not of Jun/Fos, in the incubation mixture reduced the binding of ATF2 on the E element, indicating that the reduction of ATF2 binding was caused not by the competition for binding sites but by protein-protein interaction. A variant of the AP-1 binding site and of CRE within the EnI of HBV could bind ATF2 and AP-1, respectively, but the Jun/ ATF2 heterodimer did not bind well (Fig. 3C, lanes 8 and 9). The AP-1 binding sites within the c-Jun (TTACCTCA) and ELAM promoters (TGACATCA) appear to specifically bind a Jun/ATF2 heterodimer, whereas only AP-1 binds on the consensus AP-1 binding site (TGAGTCA). Therefore, it appears that variants of both the CRE and AP-1 consensus binding sites determine the binding specificity of these sites for members of both the ATF/CREB and Jun/Fos families.
The X gene product may stimulate its own synthesis through the E element (42). If so, this would obviously require a mechanism to down-regulate X expression after a desired level of X protein has been reached. In agreement with such an autoregulatory scheme, a reduction of HBV gene expression was observed if a high level of X gene product was supplied in trans (43). A broad spectrum of X gene activity on transcription suggests the possibility that its expression during viral replication may have an influence on cellular genes that are involved in the regulation of HBV transcription. We postulate that ATF2 is one of the best candidates for the regulatory factor involved in the auto-regulation of HBx production. Overexpression and activation of ATF2 resulted in the enhanced binding of ATF2 and consequent repression of the X promoter. ATF2 can be activated both by regulatory proteins and through phosphorylation. Protein activators of ATF2 include adenovirus E1a, HTLV-I Tax, and HMGI(Y) (44 -47). The second ATF2 activation mechanism involves phosphorylation (48,49). Phosphorylation of ATF2 would lead to a conformational change in ATF2, resulting in an increase of its DNA binding affinity. This possibility is consistent with the observation that ATF2 binding to DNA can be correlated with its phosphorylation state (50). In this context, it could be postulated that HBx protein activates ATF2 both through direct interaction with ATF2 and the phosphorylation of ATF2 by activating some kinases. These observations suggest that activation of ATF2 by any pathway results in the repression of X promoter, and the production of X protein can be regulated automatically by ATF2.
Kwee et al. (13) found that smaller forms of X protein, initiated at the second and the third ATG codons of the X ORF, also have the transactivating function. Small X promoter and its positive regulatory element was reported to be located in the 5Ј-distal half of the X ORF (14). ATF2 activated the small X promoter through the region responsible for the minimal small X promoter activity, while repressing the X promoter through the E element of EnI, suggesting that differential regulation of X proteins was exerted by ATF2. In our previous report (14), we suggested that the X promoter may be regulated by downstream DNA sequences corresponding to the small X promoter.
Although all three forms of the X protein can individually transactivate the class III promoter, transactivation of several different class II promoters displays various requirements for the different X proteins. Therefore, it is conceivable that the X protein and smaller forms of X protein may have different regulatory activities. In summary, differential regulation of X proteins by ATF2 would result in the expression of different forms of X protein and a different set of class II promoter would be modulated by X proteins.