Transcription Factor AP-2 Functions as a Repressor That Contributes to the Liver-specific Expression of Serum Amyloid A1 Gene*

We previously identified transcription factor AP-2 as the nuclear factor that interacts with the tissue-specific repressor element in the rat serum amyloid A1 (SAA1) promoter. In this report, we provide evidence for a second AP-2-binding site and show that both AP-2 sites participate in mediating the transcription repression of SAA1 promoter. This proximal AP-2 site overlaps with the NFκB-binding site known to be essential for SAA1 promoter activity. Protein binding competition experiments demonstrated that AP-2 and NFκB binding to these overlapping sites were mutually exclusive. Furthermore, the addition of AP-2 easily displaced prebound NFκB, whereas NFκB could not displace AP-2. These results thus suggest that one mechanism by which AP-2 negatively regulates SAA1 promoter activity may be by antagonizing the function of NFκB. Consistent with a repression function, transient expression of AP-2 in HepG2 cells inhibited conditioned medium-induced SAA1 promoter activation. This inhibition was dependent on functional AP-2-binding sites, since mutation of AP-2-binding sites abolished inhibitory effects of AP-2 in HepG2 cells as well as resulted in derepression of the SAA1promoter in HeLa cells. In addition to SAA1, we found that several other liver gene promoters also contain putative AP-2-binding sites. Some of these sequences could specifically inhibit AP-2·DNA complex formation, and for the human complement C3 promoter, overexpression of AP-2 also could repress its cytokine-mediated activation. Finally, stable expression of AP-2 in hepatoma cells significantly reduced the expression of endogenous SAA, albumin, and α-fetoprotein genes. Taken together, our results suggest that AP-2 may function as a transcription repressor to inhibit the expression of not only SAA1 gene but also other liver genes in nonhepatic cells.

We previously identified transcription factor AP-2 as the nuclear factor that interacts with the tissue-specific repressor element in the rat serum amyloid A1 (SAA1) promoter. In this report, we provide evidence for a second AP-2-binding site and show that both AP-2 sites participate in mediating the transcription repression of SAA1 promoter. This proximal AP-2 site overlaps with the NFB-binding site known to be essential for SAA1 promoter activity. Protein binding competition experiments demonstrated that AP-2 and NFB binding to these overlapping sites were mutually exclusive. Furthermore, the addition of AP-2 easily displaced prebound NFB, whereas NFB could not displace AP-2. These results thus suggest that one mechanism by which AP-2 negatively regulates SAA1 promoter activity may be by antagonizing the function of NFB. Consistent with a repression function, transient expression of AP-2 in HepG2 cells inhibited conditioned medium-induced SAA1 promoter activation. This inhibition was dependent on functional AP-2-binding sites, since mutation of AP-2-binding sites abolished inhibitory effects of AP-2 in HepG2 cells as well as resulted in derepression of the SAA1 promoter in HeLa cells. In addition to SAA1, we found that several other liver gene promoters also contain putative AP-2-binding sites. Some of these sequences could specifically inhibit AP-2⅐DNA complex formation, and for the human complement C3 promoter, overexpression of AP-2 also could repress its cytokinemediated activation. Finally, stable expression of AP-2 in hepatoma cells significantly reduced the expression of endogenous SAA, albumin, and ␣-fetoprotein genes. Taken together, our results suggest that AP-2 may function as a transcription repressor to inhibit the expression of not only SAA1 gene but also other liver genes in nonhepatic cells.
Expression of cell type-specific genes is tightly controlled by the combined actions of positive and negative transcription regulators that interact with specific DNA sequences (1)(2)(3). In addition, the tissue-specific expression of transcription factors themselves can contribute significantly to their functional role in regulating cell-type determination or differentiation (4,5). To achieve accurate tissue-specific transcription, a complex array of signals must be integrated through the actions of transcription activators and repressors, culminating at the gene promoter to regulate initiation of RNA synthesis (6 -8).
For instance, expression of MyoD is restricted to the skeletal muscle cells and contributes to myogenesis by interacting with other positive transcription factors to activate the transcription of muscle-specific genes (9). In contrast, neuron-restrictive silencer factor is expressed in most nonneuronal tissues and in undifferentiated neuronal progenitors (10,11). Because of its tissue distribution and its repressive effects on neuronal-specific genes, this silencer factor may function as a master negative regulator for neurogenesis (10,12).
A growing list of transcription factors has been shown to function as either transcription activators or transcription repressors, depending on the promoter and cellular context of their target genes (13,14). Examples of this group of transcription factors include the Drosophila transcription factor Kruppel, which converts from an activator to a repressor in a concentration-dependent manner (15,16); human thyroid hormone receptor ␤, which converts from a repressor to an activator upon ligand binding (17,18); and YY1, which exerts positive or negative effects on transcription depending on the presence or absence of the E1A protein (19,20). Transcription factor AP-2 is yet another transcription regulator with dual functions. First identified in HeLa cell nuclear extracts (21,22), AP-2 was named for its transcription activation function. It acts as an activator in regulating many genes, including those involved in the morphogenesis of peripheral nervous system, face, limbs, skin, and nephric tissues (23)(24)(25). Recently, AP-2 has been shown to negatively regulate the transcription of stellate cell Type I collagen, K3 keratin, acetylcholinesterase, prothymosin, ornithine decarboxylase, retina fatty acid-binding protein, and CCAAT/enhancer binding protein ␣ (C/EBP␣) 1 (26 -31). Exactly when a particular transcription factor with dual functions acts as an activator or as a repressor depends not only on its intrinsic features and concentration and on the presence of other transcription factors but perhaps also on the structure, relative position, and orientation of the promoter sequence itself (13,14).
Serum amyloid A (SAA), one of the major acute-phase pro-teins (32,33), represents an excellent model system to study cytokine-induced regulation and liver-specific expression. Its serum concentration increases 1,000-fold after acute inflammation, being regulated primarily by the 200 -300-fold increase in SAA gene transcription (32,34,35). In mice, the SAA gene family consists of four genes, SAA1, SAA2, SAA3, and SAA5, and a pseudogene (36,37). Whereas expression of SAA1, SAA2, and SAA3 is dramatically induced after inflammation and each contributes equally to the increased SAA mRNA levels in the liver, SAA5 expression is induced to a much lower level and with different induction kinetics (38,39). In addition to being highly regulated by the inflammatory cytokines, expression of SAA1 and SAA2 are also highly cell type-specific, restricted to the liver hepatocytes (32,34,35). Studies of the rat SAA1 promoter demonstrate that 304 bp of its 5Ј-flanking sequence are sufficient not only for its liver cell-specific expression but also for its cytokine-induced expression in response to inflammatory mediators (40). Further deletion analyses identified a 66-bp DNA fragment spanning bp Ϫ138 to Ϫ73 that could confer its cytokine responsiveness. Within this cytokine response unit (CRU) reside binding sites for transcription factors NFB, C/EBP, and YY-1 (41). Whereas NFB and C/EBP function cooperatively to induce SAA1 promoter activity in response to cytokine stimulation, YY-1 functions as a repressor in opposing NFB-mediated transcription activation and contributes to the low basal expression of SAA1 and perhaps the transience of its expression after inflammation (41). Subsequent transient transfection analyses revealed a tissue-specific repressor element distal to the CRU that conferred repression on the SAA1 promoter in HeLa cells but had no such inhibitory activity in liver cells (42). This repressor element formed an intense DNA-protein complex with nuclear extracts prepared from HeLa cells but not from liver cells.
We have recently purified this DNA-binding protein from HeLa nuclear extracts and shown by protein sequencing and biochemical and immunological analyses that it is identical to the transcription factor AP-2 (43). Here, we report on the repressive effects of AP-2 on the expression of transfected and endogenous liver genes and suggest that it may function as a negative regulator to repress the expression of some liver genes in nonhepatic cells.

EXPERIMENTAL PROCEDURES
Cell Culture and Conditioned Medium-HeLa, HepG2, and Hep3B cells were grown in basal medium that contained minimum essential medium (Life Technologies, Inc.) supplemented with 10% fetal calf serum, 0.1 mM nonessential amino acids, 1 mM pyruvate, 100 units/ml penicillin, and 100 g/ml streptomycin. HepG2 and Hep3B cells were used for our transient transfection and stable cell line studies because they have been demonstrated to confer proper regulation of many endogenous and transfected genes, including SAA (44 -47). Conditioned medium (CM) known to contain a mixture of inflammatory mediators capable of inducing acute-phase gene expression was prepared from activated mixed lymphocyte cultures as described (48) and used as a mixture with an equal volume of basal medium.
Transient Transfection Assay-HepG2 cells (1.5-2 ϫ 10 5 ) were seeded into 60-mm culture dishes 16 h before they were transfected with plasmid DNAs using the FuGENE TM 6 transfection kit (Roche Molecular Biochemicals). Cells were incubated for 16 -20 h before being stimulated with basal medium or with 50% CM to induce the reporter gene activities. Sixteen to 20 h after treatment, cells were harvested for cell extract preparation and CAT or luciferase assays.
CAT and Luciferase Assays-CAT assays were performed essentially as described (48). The conversion of acetylated forms of chloramphenicol was calculated by measuring the radioactivity with a PhosphorImager (Molecular Dynamics). Luciferase activity was measured with luciferase assay kit (Tropix, Inc.) using a Lumat LB 9507 luminometer (EG & G Berthold). The averages and S.E. from 3 or 4 independent experiments were calculated relative to the activities for the control samples, to which a value of 1.0 was assigned.
EMSA-For EMSA, 2 ϫ 10 4 cpm (ϳ1 ng) of 32 P-labeled DNA fragments were mixed with indicated amounts of protein samples in reaction buffer (48). After incubation for 30 min at room temperature, the samples were loaded on a 5% polyacrylamide gel and subjected to electrophoresis at 200 V for 100 min. The gel was then dried and autoradiographed. The intensities of shifted protein-DNA complexes were quantified using a PhosphorImager (Molecular Dynamics). In oligonucleotide competition experiments, the protein was incubated simultaneously with the labeled probe and an excess amount of unlabeled competitor oligonucleotides. Recombinant AP-2 and the p50 subunit of NFB were purchased from Promega.
Establishment of AP-2-expressing Stable Hep3B Cell Lines-Hep3B cells were plated at 70 -80% confluence in 100-mm culture dishes 16 -24 h before transfection with pIND/AP-2 or empty vector (3 g/ dish). Cells were trypsinized 48 h after transfection, washed with basal medium, and split into four 100-mm culture dishes in selection medium containing 400 g/ml G418. Cells were fed with fresh selection medium every 3-4 days. After culturing for 2-3 weeks, individual resistant clones were picked and transferred to 24-well plates. Clones that express AP-2 were screened by EMSA using the AP-2d sequence as a probe and further verified by RT-PCR. Three AP-2-expressing clones, designated H3B.A1, H3B.A2, and H3B.A6, were selected for further analysis. For comparison, three control clones that contain the empty pIND vector were selected similarly and were designated H3B.V1, H3B.V2, and H3B.V3.
RT-PCR-Total cellular RNAs were isolated from the cloned Hep3B cell lines using the High Pure RNA Isolation kit (Roche Molecular Biochemicals). Titan TM One Tube RT-PCR system (Roche Molecular Biochemicals) was used to optimize conditions for PCR (55°C for 30 min, 94°C for 30 s, 64°C for 1 min, and 68°C for 1 min) and to quantify mRNA levels. The amplification reactions contained 1 Ci of [␣-32 P]dCTP for the detection of reaction products and were carried out to ensure that they are within the linear range of the amplification. The reaction products were resolved on 5% native polyacrylamide gels and quantified using a PhosphorImager (Molecular Dynamics).

RESULTS
The Proximal Region of SAA1 Promoter Contains Two AP-2binding Sites-We had previously identified a repressor element in the rat SAA1 promoter that could inhibit its promoter activity in HeLa cells (42). Furthermore, we showed this repressor element contained a high affinity binding site for AP-2 (43). In this study, we further investigated the mechanisms for the repressive effects of AP-2 by searching for additional AP-2-binding sites in the SAA1 promoter. In addition to the AP-2-binding site (bp Ϫ272 to Ϫ265) identified initially, one additional sequence (5Ј-CCCCAGGG-3Ј) located at bp Ϫ85 to Ϫ78 was found to be a potential AP-2-binding site (Fig. 1A). These two sites were designated, relative to their positions in the SAA1 promoter, the distal AP-2-binding site (AP-2d) and the proximal AP-2-binding site (AP-2p) (Fig. 1A). It is interesting to note that the sequences at the AP-2p site are highly conserved between rat, human, and mouse SAA1 genes (40,52). To determine whether the AP-2p site could indeed interact with AP-2, oligonucleotide competition experiments were carried out using radiolabeled AP-2d sequence and purified AP-2 in EMSA. In the absence of any competitors, a strong AP-2⅐DNA complex was detected (Fig. 1B). As expected, both the consensus AP-2 binding sequence from the metallothionein IIA promoter and the AP-2d site very effectively competed for AP-2⅐DNA complex formation, whereas the mutated AP-2d-site oligonucleotides could not compete (Fig. 1B). When the wild-type and mutated AP-2p oligonucleotides were used as competitors, the wild-type AP-2p oligonucleotides showed dose-dependent competition of the complex formation, whereas the mutant AP-2p site did not. This result indicates that the AP-2p site functioned as a binding site for AP-2, although it had lower binding affinity than the AP-2d site.
Interestingly, the position of the AP-2p site overlaps that of NFB, which is known to be essential for SAA1 promoter function (40,50). The overlapping nature of these two transcription factor-binding sites raises an intriguing question, that is, whether binding of one factor to the DNA would affect the binding of the other factor. To address this question, gel shift assays were performed using the AP-2p probe, which contains both AP-2-and NFB-binding sites, and recombinant AP-2 and NFB. The reactions were carried out under probelimiting conditions to allow AP-2 and NFB to compete for limited amounts of target DNA. When added individually to the reactions, both AP-2 and NFB formed specific DNA-protein complexes (Fig. 2). However, when AP-2 and NFB were added together, no DNA-protein complexes with slower mobilities were observed, arguing against simultaneous binding of these two transcription factors to the same DNA molecule. To examine their relative binding affinities, radiolabeled AP-2p probe was first incubated with a constant amount of AP-2 or NFB. After a brief incubation, increasing amounts of NFB or AP-2 were added to the reaction mixture. As shown in Fig. 2, the addition of AP-2 readily displaced the prebound NFB from this overlapping region (lanes 8 -12). At a 2.3-fold molar excess of AP-2, no NFB remained bound to the probe. In sharp contrast, NFB had little effect on the prebound AP-2, even at a 28-fold molar excess of NFB (lanes 2-6). Taken together, our results indicate that AP-2 and NFB bind to this overlapping region in a mutually exclusive manner. Furthermore, AP-2 has much higher DNA binding affinity than NFB. It is therefore tempting to speculate that one mechanism by which AP-2 may inhibit cytokine-induced SAA1 promoter function is by antagonizing the ability of activated NFB to bind to its cognate binding site.
Binding of AP-2 to the SAA1 Promoter Is Required for Its Inhibitory Effects-Site-specific mutation of NFB-or C/EBPbinding sites in the CRU completely abolished SAA1 promoter cytokine responsiveness (40,50), indicating the importance of these two transcription factors in SAA1 gene transcription. Since we showed that the NFB-binding site overlaps with that of AP-2 and the presence of AP-2 effectively prevents NFB from binding to its cognate site, we sought to determine whether AP-2 binding to the AP-2p site is required for its repressive effects on the SAA1 promoter. Two reporter constructs, pSAA1/CAT(Ϫ120) (wild-type AP-2p site) and pSAA1/ CAT(Ϫ120) mAP-2p (mutant AP-2p site), were generated. As determined by EMSA, the 2-bp mutation at the AP-2p site completely abolished AP-2 binding but still retained full NFB binding activities (data not shown). These two reporter constructs were cotransfected into HepG2 cells with the AP-2expression vector, pSAP2, to assess the effects of AP-2p site mutation on the ability of AP-2 to repress the reporter gene expression. As shown in Fig. 3, the CAT activities for both the wild-type and the mutant constructs could be induced 5-6-fold by CM. However, cotransfection with AP-2 expression vector completely inhibited CM-induced expression of the wild-type pSAA1/CAT(Ϫ120) reporter. In contrast, expression of AP-2 had no significant effects on the cytokine responsiveness of the mutant pSAA1/CAT(Ϫ120) mAP-2p reporter. As controls, cotransfection with the frameshift AP-2 mutant construct had no inhibitory effects on either reporter construct. These results indicate that AP-2-mediated repression on a pSAA1/ CAT(Ϫ120) reporter requires a functional AP-2p site.
Mutation of AP-2-binding Sites Results in Derepression of SAA1 Promoter in HeLa Cells-Because liver and liver-derived cell lines such as HepG2 do not normally express AP-2, a study of the effects of AP-2 in these cells requires cotransfection with AP-2 expression vectors. In contrast, HeLa cells express abundant amounts of AP-2, thus making cotransfection unnecessary as well as providing an alternative approach to examine the specificity of AP-2 repression effects. To investigate whether the lack of SAA1 promoter activities in HeLa cells is, at least in part, due to repression by the endogenous AP-2 through its binding at AP-2d and AP-2p sites, we generated a double AP-2-site mutant construct, pTK/SAA1(Ϫ285/Ϫ70) mAP-2d/p , in which both the distal and the proximal AP-2 sites were mutated. As expected, the construct containing the wild-type AP-2 sites, pTK/SAA1(Ϫ285/Ϫ70), was nonresponsive to CM stimulation when transfected into HeLa cells, presumably due to inhibition by the endogenous AP-2. Mutation of AP-2-binding sites, however, resulted in derepression of SAA1 promoter activity (Fig. 4). This result indicates that endogenous AP-2 in HeLa cells can indeed inhibit the cytokine-induced SAA1 activation and that binding of AP-2 is required for its repressive effects on the SAA1 promoter.
Other Liver Gene Promoters Contain Potential AP-2-binding Sequences-Two AP-2-binding sites were identified in the SAA1 promoter region, a high affinity AP-2d site and a weaker AP-2p site. To determine whether AP-2 may also regulate the transcription of other liver genes, we searched for potential AP-2-binding sites using the AP-2d sequence 5Ј-ATACCTCAG-GCAGC-3Ј and the AP-2 consensus sequence 5Ј-GGCN 3 GCC-3Ј (25,53). As shown in Table I, at least 12 genes that are expressed mainly in the liver were found to contain putative AP-2-binding sites. To determine whether these sequences in the liver gene promoters could interact with AP-2 and therefore compete for its binding, oligonucleotides corresponding to six of the genes, chosen for their acute-phase response or liver-enriched expression, were synthesized and used as competitors in EMSA. HeLa cell nuclear extracts were used as the source of AP-2 binding activity, and radiolabeled AP-2d sequence was used as a probe. As shown in Fig. 5, all six oligonucleotides showed dose-dependent competition for AP-2 binding. Although they contained highly homologous core AP-2 binding sequences, they nevertheless exhibited considerable differences in their abilities to compete for AP-2 binding. For example, although the sequence from the human complement C3 promoter was 20 times more effective in competing for AP-2 binding than the SAA1 AP-2d site, sequences from the ␣-fibrinogen and apolipoprotein E promoters were 4 and 20 times, respectively, less effective. The fact that these oligonucleotides could inhibit AP-2⅐DNA complex formation suggests that AP-2 may bind to these sequences and perhaps regulate their expression in vivo.
AP-2 Inhibits Cytokine-induced Human Complement C3 Promoter Activity in HepG2 Cells-The fact that human complement C3 promoter contains such a high affinity AP-2-binding site suggests AP-2 may play a functional role in C3 gene regulation as well. To determine whether AP-2 can similarly repress C3 promoter activity, we transfected HepG2 cells with the cytokine-inducible pC3/Luc(Ϫ199) reporter with or without the AP-2-expression vector. When cotransfected with the empty vector, the luciferase activity was dramatically induced (25-fold) in response to CM stimulation. However, cotransfection with AP-2 expression vector pSAP2 resulted in more than 90% inhibition of the pC3/Luc(Ϫ199) reporter gene activity (Fig. 6A), whereas cotransfection with the frameshift AP-2 mutant pSAP2/FS21 had no inhibitory effects. To further demonstrate that the inhibitory effects were specific for AP-2, HepG2 cells were also transfected with the dominant-negative AP-2 mutant pSAP⌬166 -277 (53). As shown in Fig. 6B, coexpression of this mutant effectively neutralized the inhibitory effects of AP-2 on human C3 promoter. SAA1 or C3 promoter. Together with the oligonucleotide competition results that suggest many other liver gene promoters contain AP-2-binding sites, we sought to examine whether AP-2 may have similar effects on the expression of endogenous liver genes in their normal chromosomal context. Stable Hep3B cell lines that constitutively express moderate levels of AP-2 were generated by transfecting AP-2 expression vector, pIND/ AP-2, into cells followed by selection with G418. Among the 11 stable cell lines isolated, three clones, H3B.A1, H3B.A2, and H3B.A6, were chosen for further studies. As controls, three Hep3B cell lines, H3B.V1, H3B.V2, and H3B.V3, that had the vector DNA stably integrated were selected in a similar manner. The AP-2 mRNA levels in these cell lines were determined by RT-PCR. As with the parental Hep3B cells, no detectable AP-2 expression was observed in H3B.V1, H3B.V2, and H3B.V3 cells (Fig. 7A). In contrast, AP-2 mRNA was easily detected in H3B.A1, H3B.A2, and H3B.A6 cells. Likewise, AP-2 protein can be detected by Western blot in stably transfected AP-2-expressing cells but not in the cells transfected with vector only (data not shown). To determine AP-2 DNA binding activities in these stable cells, EMSAs were performed using radiolabeled AP-2d probe. Significant AP-2-binding activities were detected in H3B.A1, H3B.A2, and H3B.A6 cells, whereas no detectable AP-2⅐DNA complexes were observed in H3B.V1, H3B.V2, and H3B.V3 cells (Fig. 7B). For comparison, HeLa cell extracts were also incubated with the AP-2d probe. A strong AP-2⅐DNA complex with the same electrophoretic mobility as those with the stable cell lines was detected. This protein-DNA  5. Competition of various oligonucleotides from liver gene promoters for AP-2 binding. The distal AP-2-binding sequence from the SAA1 promoter was radiolabeled and used as probe in EMSA. HeLa cell nuclear extracts were incubated with labeled probe in the presence of increasing amounts of double-stranded oligonucleotide competitors. The competing oligonucleotides were from the promoters of rat SAA1 (AP-2d site), human C3, human apolipoprotein B (ApoB), rat phosphoenolpyruvate carboxykinase (PEPCK), rat ␣ 1 -acid glycoprotein (␣ 1 -AGP), ␣-fibrinogen (␣-FG), and apolipoprotein E (ApoE). The efficiency of competition was calculated and presented as the amount of oligonucleotides required for 50% inhibition of AP-2⅐DNA complex formation.
complex could be specifically competed by wild-type but not mutant AP-2 oligonucleotides. When compared with HeLa cell extracts, the levels of AP-2 in the stable cell lines were 2-3-fold lower (Fig. 7B).
To determine whether AP-2 expressed in Hep3B cells could inhibit the expression of endogenous liver genes, we first examined its effects on the expression of endogenous SAA genes. To induce the endogenous SAA genes, control (H3B.V1, V2, and V3) and AP-2-expressing (H3B.A1, A2, and A6) cells were treated with basal medium or with 50% CM for 12 h. Total RNA was isolated, and the levels of SAA mRNA were determined by RT-PCR and quantified. In nonstimulated cells, SAA mRNA was undetectable in both control and AP-2-expressing cells (Fig. 7C). Upon cytokine stimulation, the expression of endogenous SAA mRNA in control cells was dramatically induced. In contrast, in AP-2-expressing cell lines, an increase in SAA mRNA was also detected after stimulation, albeit the magnitude of induction was much lower than that in control cells (Fig. 7, C and D).
To determine whether AP-2 could also exert inhibitory effects on other liver genes, we chose to examine the expression of albumin and AFP because they are two prototypic liver genes that are constitutively expressed in Hep3B cells and because putative AP-2-binding sites were also found in their promoters. When compared with control cells, albumin and AFP expression levels were consistently 2-fold lower in the AP-2-expressing cells (Fig. 7D). This reduced albumin and AFP expression in H3B.A1, A2, and A6 cells was not due to nonspecific inhibition since the expression of control genes, GAPDH, and ␤-actin were not altered. Taken together, our results indicate that AP-2 when expressed in a liver-cell background is still capable of conferring repression on inducible-and constitutively expressed liver-specific genes but has no effect on the expression of genes that are also expressed in nonliver cells.

DISCUSSION
Transcription factor AP-2 joins a growing list of transcription regulators that are able to function as either transcription activators or transcription repressors depending on the cellular and promoter context of their target genes (13,14,54). Initially characterized as an activating protein in SV40 gene transcription (21,22), AP-2 has since been shown to participate in the transcription activation of many other genes, including those related to neuronal and epidermal development (23-25, 55, 56). More recently, AP-2 was also shown to function as a transcription repressor in repressing the expression of cell type-specific genes (29,31).
Expression of SAA1 is highly inducible by the proinflammatory cytokines and is highly tissue-specific, being expressed primarily in the liver. Our earlier studies identified a 66-bp CRU in the SAA1 promoter that is responsible for its cytokineinduced regulation (41,50). Within the CRU, two cis-regulatory elements, the NFB-and C/EBP-binding sites, were essential for its promoter responsiveness to cytokine stimulation (40,50). In addition to the CRU, we identified two DNA regulatory elements that have the characteristics of transcription silencers and can serve as binding sites for the transcription factor AP-2. Interestingly, AP-2 is expressed abundantly in HeLa and many other nonliver tissues but is absent in liver or liverderived hepatoma cells (57,58). This inverse correlation between AP-2 and SAA1 expression is consistent with the repressive function of AP-2 on SAA1 gene transcription and may contribute to liver-specific expression pattern of SAA1 by diminishing or abrogating its expression in nonliver cells.
This study together with our earlier studies leads us to formulate a model for the regulatory mechanisms that control the inflammation-induced and tissue-specific expression characteristics of SAA1 (Fig. 8). In nonstimulated hepatic cells such as HepG2, YY1 functions as a constitutively expressed transcription repressor that occupies its site in the CRU and accounts for the low basal expression of SAA1. In response to stimulation, inflammatory mediators dramatically increase the activities of the positive transcription factors NFB and C/EBP in the nucleus. In the absence of AP-2, as in liver cells, these two potent transactivators could displace YY1 and bind to their respective binding sites in the SAA1 promoter and function cooperatively to activate SAA1 gene transcription (Fig. 8). In HeLa cells, however, in addition to YY1, AP-2 is also constitutively expressed and could bind to its binding site in the CRU. Because of its higher binding affinity, AP-2 prevents activated NFB from binding to the overlapping binding site. Consequently, without NFB binding, the SAA1 promoter cannot respond to cytokine induction despite the presence of activated NFB and C/EBP in the nucleus. Thus, in this case, the tissuespecific expression of AP-2 itself plays an important role in determining the expression pattern of SAA1. It is noteworthy that removal of the repressive effects of AP-2 in HeLa cells by mutating the two AP-2 sites did not lead to full activation of the SAA1 promoter. This may reflect the nature of the cellular milieu in HeLa cells, which is not optimum for SAA1 gene transcription. For example, compared with HepG2 cells, HeLa has relatively low levels of C/EBP, a critical transcription factor for SAA1 gene activation. In addition, results from AP-2expressing cell lines suggest that AP-2 is likely not the only repressor that inhibits SAA1 promoter activity in nonhepatic cells.
Five lines of evidence support the notion that AP-2 plays a central role in repressing SAA1 expression. First, two high affinity AP-2-binding sites were localized to within 300 bp of the SAA1 gene transcription initiation site. Moreover, protein binding studies indicated that the proximal AP-2 binding sequence that overlaps with that of NFB binds to AP-2 with significantly higher affinity than NFB at these overlapping sites. The fact that AP-2 binds with higher affinity and the constitutive presence of AP-2 in the nucleus plus the normally inactive state of NFB in nonstimulated cells gives AP-2 a distinct advantage over NFB in competing for the overlapping binding sites. Consequently, in cells that express AP-2, i.e. many nonliver cells, this overlapping site would normally be occupied by AP-2 and prevent subsequent NFB binding. Thus, AP-2 may mediate repression on the SAA1 promoter by antagonizing the function of NFB. Second, a 20-bp region spanning the NFB site to the AP-2p site is 100% conserved between rat and human SAA1 genes and 80% with that of mouse SAA1 gene (40,52). Such high sequence conservation suggests functional importance of this region. Third, forced expression of AP-2 in HepG2 cells efficiently inhibited CM-mediated induction of the SAA1 promoter. This inhibition was highly specific because mutation of the AP-2-binding sites or expression of a dominant-negative mutant of AP-2 relieved AP-2-mediated inhibitory effects. Fourth, whereas the wild-type SAA1 promoter is nonresponsive to cytokine stimulation when transfected into HeLa cells, mutation of AP-2-binding sites led to derepression of the promoter and responsiveness to cytokine induction. Finally, stable expression of AP-2 in Hep3B cells significantly reduced endogenous SAA gene expression in response to inflammatory cytokines.
Transcription repression may be achieved by various mechanisms. In principle, negative regulators could inhibit transcription by interfering with any of the several steps in the transcription initiation pathway (60,61). A repressor might interfere with the activity of a DNA-bound activator by quenching or masking its activation potential (62). Alternatively, it may block nuclear localization of an activator (63). In addition, negative regulators might interfere with the function of general transcription machinery (6,64,65). Finally, repressors might block transcription activation by binding to DNA at a region that overlaps that of transcription activators, thus excluding the binding of activators (26,41). In the rat SAA1 promoter, one possible mechanism by which AP-2 represses the SAA1 promoter in HeLa cells is by blocking the activity of the transactivator NFB. In this regard, it is also interesting to note that although the 3Ј region of NFB-binding site overlaps with the AP-2p site, its 5Ј region overlaps with the binding site of another transcription repressor YY1 (41). However, unlike AP-2, YY1 is expressed in all tissues examined, including the liver, and modulates the cytokine responsiveness of the SAA1 promoter rather than controls its cell-specific expression. Another   FIG. 8. A model for inflammationinduced and tissue-specific regulation of SAA1 gene expression. A, schematic diagram of four known transcription factors binding at the proximal region of the SAA1 promoter with NFBbinding site flanked by two repressors, YY1 and AP-2. B, in HepG2 cells, YY1 normally binds and represses SAA1 gene transcription, accounting for its low basal expression. Upon stimulation, activated NFB and C/EBP displace YY1 from the promoter and confer synergistic transcription activation. In HeLa cells, however, the presence of AP-2 binding to its high affinity site prevents activated NFB from binding to the overlapping site. Consequently, SAA1 gene transcription remains in the repressed state after induction despite the presence of NFB and C/EBP in the nucleus. distinguishing feature between these two repressors is that whereas NFB is unable to displace prebound AP-2 at the overlapping region, it can readily displace prebound YY1 (41). Because the binding between AP-2 and NFB and between YY1 and NFB at their respective overlapping sites is mutually exclusive, the function of NFB is therefore opposed by two repressors, with relative binding affinities at these overlapping sites of AP-2 Ͼ NFB Ͼ YY1.
Transcription repressors that can function as general negative regulators and contribute to cell differentiation have been reported (10, 12, 66 -68). One such negative regulator is the neuron-restrictive silencer factor, which could potentially repress a large number of neuron-specific genes in nonneuronal cells (10,12). In this regard, the function of AP-2 in repressing liver gene expression in nonliver cells may resemble that of neuron-restrictive silencer factor. In addition to the SAA1 and C3 promoters, we identified potential AP-2-binding sites in at least 10 other liver gene promoters. Several of these sequences could effectively compete for AP-2 binding in EMSA, implying AP-2 may also bind to these promoters and perhaps regulate their expression. Thus, AP-2 may, by restricting its expression in AP-2-expressing nonliver cells, help regulate the expression of many liver genes in a cell-specific manner.
Our results indicate that AP-2, when expressed in a liver cell background, is capable of conferring repression not only on cytokine-inducible genes but also on constitutively expressed liver-specific genes. The reduced expression of SAA, albumin, and AFP was not due to nonspecific effects caused by the manipulation of cells during G418 selection because (a) the comparisons were made with control Hep3B cell lines that were transfected with an empty vector and were selected in a similar manner, (b) consistent repression was observed in three independent AP-2-expressing Hep3B cell lines, and (c) the expression of two housekeeping genes, GAPDH and ␤-actin, in these cell lines was not altered. Therefore, together with our transient transfection studies, these results further support the notion that AP-2 may have broad functional role as a general negative regulator in repressing the expression of some liver genes in nonhepatic cells. Although AP-2 may possess the ability to repress the expression of certain liver genes, it is unlikely to be the only regulatory factor responsible for the tissue-specific repression of these liver genes. An indication is that although the levels of SAA, albumin, and AFP were reduced in AP-2-expressing Hep3B cells, they were nevertheless significantly higher than that in HeLa or other nonliver cells. Therefore, repression of liver genes in nonliver cells is likely to be determined by multiple factors, including other repressors and chromatin structures. Together with AP-2, these determinants contribute to the silencing of SAA1 and other liver genes in nonliver cells. It should be noted that although overexpression of AP-2 in hepatic cells could repress the expression of a number of liver genes, our results could not conclusively prove the converse would be true. That is, removal or reduction of AP-2 in nonhepatic cells will lead to enhanced expression of SAA or other liver genes.
In summary, we report that transient expression of AP-2 in HepG2 cells effectively inhibited cytokine-induced expression of SAA1 and human C3 promoter activities. This inhibition was dependent on functional AP-2-binding sites and could be neutralized by a dominant-negative mutant of AP-2. In addition to SAA1 and C3, 10 other liver gene promoters were found to contain potential AP-2-binding sites, some of which are very effective competitors for AP-2 binding. Finally, stable expression of AP-2 in a liver cell background significantly reduced the expression of several endogenous liver genes. Although it is tempting to speculate that the observed inhibition in human SAA1 expression in AP-2-expressing Hep3B cells was due to the repressive effects of the conserved AP-2p site, additional studies are required to specifically address this issue. Taken together, our studies suggest that AP-2 may function as a general transcription repressor that not only inhibits SAA1 gene transcription in nonhepatic cells but also as a contributor to the cell-specific expression of other liver genes.