Estrogen Regulation of the Apolipoprotein AI Gene Promoter through Transcription Cofactor Sharing*

Estrogen replacement therapy increases plasma concentrations of high density lipoprotein and its major protein constituent, apolipoprotein AI (apoAI). Studies with animal model systems, however, suggest opposite effects. In HepG2 cells stably expressing estrogen receptor α (ERα), 17β-estradiol (E2) potently inhibited apoAI mRNA steady state levels. ApoAI promoter deletion mapping experiments indicated that ERα plus E2 inhibited apoAI activity through the liver-specific enhancer. Although the ERα DNA binding domain was essential but not sufficient for apoAI enhancer inhibition, ERα binding to the apoAI enhancer could not be detected by electrophoretic mobility shift assays. Western blotting and cotransfection assays showed that ERα plus E2 did not influence the abundance or the activity of the hepatocyte-enriched factors HNF-3β and HNF-4, two transcription factors essential for apoAI enhancer function. Expression of the ERα coactivator RIP140 dramatically repressed apoAI enhancer function in cotransfection experiments, suggesting that RIP140 may also function as a coactivator on the apoAI enhancer. Moreover, estrogen regulation of apoAI enhancer activity was dependent upon the balance between ERα and RIP140 levels. At low ratios of RIP140 to ERα, E2 repressed apoAI enhancer activity, whereas at high ratios this repression was reversed. Regulation of the apoAI gene by estrogen may thus vary in direction and magnitude depending not only on the presence of ERα and E2 but also upon the intracellular balance of ERα and coactivators utilized by ERα and the apoAI enhancer.

Apolipoprotein AI (apoAI) 1 is the major protein constituent of plasma high density lipoproteins (HDLs), a class of lipoproteins thought to play a major role in protection against atherosclerosis (reviewed in Ref. 1). Because plasma HDL levels are correlated with plasma apoAI and liver apoAI mRNA levels (2), it is thought that factors affecting apoAI gene expression play an important role in atherosclerosis susceptibility. Although a number of pharmacological, dietary, and physiological factors affect apoAI and HDL plasma levels (3)(4)(5)(6)(7)(8), the underlying molecular mechanisms remain obscure. For example, numer-ous observational studies and a recent randomized trial have shown that estrogen replacement therapy increases apoAI and HDL plasma levels (9 -11). However, the underlying molecular mechanism(s) remains controversial. Both increased production of apoAI (12,13) and reduced HDL catabolism (14) have been suggested as potential mechanisms.
Work with animals has further complicated the issue. In cynomolgus monkeys, ethinyl estradiol or conjugated equine estrogen markedly reduces apoAI and HDL plasma levels (15,16). Similarly, ovariectomy increases hepatic apoAI mRNA levels in rats, further supporting the concept that estrogen may repress apoAI gene expression (17). Moreover, the ethinyl estradiol-induced increases in apoAI mRNA levels in these animals appears to occur via indirect dietary effects due to hormone treatment (17). Further, estrogen-induced increases in apoAI transcription rates in rats are dependent on the strains used (6).
Liver-specific expression of the apoAI gene is conferred by a powerful hepatocyte-specific enhancer located in the nucleotide region Ϫ220 to Ϫ110 upstream of the apoAI transcriptional start site (18). The activity of the enhancer depends on synergistic interactions between transcription factors bound to three distinct sites: A (Ϫ214 to Ϫ192), B (Ϫ169 to Ϫ146), and C (Ϫ134 to Ϫ119) within the enhancer (18,19). Sites A and C bind various members of the nuclear receptor superfamily including the hepatocyte nuclear factor 4, HNF-4 (20 -22), retinoid X receptor ␣ (23), and apolipoprotein AI regulatory protein-1 (24,25). Site B binds members of the hepatocyte nuclear factor 3 family, HNF-3 (19,26). Synergy between these factors during enhancer activation appears to involve interactions with uncharacterized transcription auxiliary factors (18,26). Recent evidence suggests that one or more of these factors are regulated by estrogen in heterologous nonhepatic cells (26).
The actions of estrogen are mediated primarily by the estrogen receptors (ERs) ␣ (27) and ␤ (28,29); however, only ER␣ is expressed in the liver (28,29). Estrogen signal transduction involves high affinity binding to intracellular ERs, ligand-induced conformational changes of ERs leading to the recruitment of transcriptional auxiliary factors, binding of ERs to estrogen response elements (EREs) in gene promoters, and regulation of transcriptional activity in conjunction with other transcription factors bound to their cognate sites in the promoter. Recent efforts to characterize ER transcription auxiliary factors have led to the identification of a growing number of coactivators such as Trip/Sug1 (30,31), ERAP140 and ERAP160 (32), RIP140 (33), TIF1 (34), and SRC-1 (35,36). Although all of these proteins bind ER␣ in a ligand-dependent fashion, the mechanisms by which they modulate ER signaling and "cross-talk" with other signal transduction pathways is not understood. Recent findings indicate that the related coactivators p300 (37) and CBP (38) are also involved in ER␣ function and serve as a signal integrator for several hormone-dependent and hormone-independent signal transduction pathways (Refs. 36,39,and 40 and reviewed in Ref. 41). Additional pathways of liganded ER action involve recruitment of ER␣ to gene promoters lacking EREs via protein-protein interactions with promoter-bound transcription factors (42)(43)(44)(45) and activation of the mitogen-activated protein kinase pathway (46,47).
This report shows that ER␣ and 17␤-estradiol repress apoAI promoter activity in human hepatoma HepG2 cells. This effect appears to be due to ER␣ partitioning of coactivators required for apoAI enhancer function in liver cells. The data suggest that RIP140 may play an important role in apoAI enhancer function and ER␣-mediated repression. We propose that estrogen effects on apoAI gene expression vary in direction and magnitude depending upon the balance of coactivators shared by ER␣ and the apoAI enhancer.  (18) into TK.LUC. The A.LUC, B.LUC, and ERE.LUC constructs were described previously (19,48).

Plasmid
The pMT2-ER␣ construct was created by transferring an EcoRI fragment containing the coding region of ER␣ from the HEO plasmid (49) into the pMT2 expression vector. The ER␣ mutant expression vectors AF1-DBD-X, X-DBD-AF2, and X-DBD-X were generated in pcDNA3 (Invitrogen) as described previously (50). The ER␣ X-DBD-X construct was created by introducing three nucleotide substitutions within the ER␣ DNA binding domain (DBD) (51), and the pcDNA3-ER␣ expression vector was described previously (48).
Stable Cell Line Creation-HepG2 cells stably expressing ER␣ (Hep89) were created by transfecting the pcDNA3-ER␣ expression vector into HepG2 cells by electroporation using the BTX Electro Cell Manipulator 600 according to the manufacturer's recommended settings. Stably expressing cells were selected by resistance to G418 (400 g/ml). Distinct, well isolated colonies were picked using Bellco cloning cylinders (6 ϫ 8 mm) and assessed for the presence of ER␣.
Cell Transfections-Plasmid DNAs were purified on Qiagen columns and transfected into HepG2 cells by the calcium phosphate coprecipitation method as described previously (26). The cells were seeded in deficient growth media (phenol red-free Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with heat-inactivated 10% fetal bovine serum, 1% Glutamax, 1% minimum essential medium nonessential amino acids, 100 units/ml penicillin and 100 g/ml streptomycin) at 2.5 ϫ 10 5 cells/well in a 12-well dish (Falcon) before transfection. Different amounts of the expression vectors pMT2-ER␣, pMT2-HNF-4 (52), pCMV.HNF-3␤ (53), pEFRIP (33), or RSV-CBP-HA (38) were cotransfected as indicated. Luciferase and ␤-galactosidase activity was determined as described previously (26). The data shown represent the mean Ϯ S.E. from at least three independent experiments, each in duplicate. Statistical analysis of the data was carried out using the Dunnett's method (54) to compare treated versus control samples.
Northern Analysis-HepG2 and Hep89 cells were seeded in deficient growth media and treated over 72 h in the presence or absence of 1 M E2. Total RNA was isolated (Biotecx Labs), subjected to electrophoresis, and hybridized with 32 P-labeled apoAI PstI cDNA fragment (55) or 32 P-labeled human glyceraldehyde-3-phosphate dehydrogenase cDNA (Stratagene). The relative intensities of the hybridized signals were quantitated by phosphoimaging (Molecular Dynamics).
Electrophoretic Mobility Shift Assays-Protein-DNA complexes were analyzed by incubation of bacterially expressed HNF-4, HepG2 nuclear extracts (26), or baculovirus-expressed human ER␣ (Panvera) with 32 P-labeled DNA probes corresponding to either the 110-base pair apoAI enhancer or the vitellogenin ERE followed by electrophoresis in low ionic strength polyacrylamide gels as described previously (26).
Western Blot Experiments-Nuclear extracts were prepared as described previously (26). Protein concentrations were determined by the BCA method. Proteins were transferred from a 4, 20% SDS-polyacrylamide gel to nitrocellulose and blotted using affinity-purified human monoclonal ER antibody (Stress Gen), rabbit anti-HNF-3␤ antibody (gift of R. Costa), or rabbit anti-HNF4 serum (from F. Sladek) as the primary antibodies followed by peroxidase-conjugated goat anti-rabbit IgG antibody (Zymed Laboratories Inc.). Detection was performed using the Enhanced Chemiluminescence Western blotting Detection System (Amersham Pharmacia Biotech).

Estrogen Represses apoAI mRNA Levels in HepG2
Cells Stably Expressing ER␣-Human hepatoma HepG2 cells retain many liver-specific functions; however, they no longer express ER␣. Therefore, HepG2 cells stably expressing ER␣ were created to monitor the regulation of the apoAI gene by estrogen. The resulting cell line, Hep89, expressed ER␣ by Western blot (Fig. 1A). Further, the activity of a synthetic vitellogenin estrogen response element luciferase reporter (ERE.LUC) was stimulated 25-fold by 100 nM 17␤-estradiol (E2) in Hep89 cells, whereas no E2-dependent promoter activity was observed in the parental HepG2 cells (Fig. 1B). Northern blotting demonstrated a 2-3-fold decrease in apoAI mRNA steady state levels in Hep89 cells after a 72-h treatment with 1 M E2, whereas apoAI mRNA levels remained unchanged in HepG2 cells ( Fig.  1, C and D). Therefore, apoAI mRNA steady state levels are regulated by E2 in a receptor-dependent manner.
Estrogen Represses apoAI Promoter Activity in HepG2 Cells-The ER␣ effects on apoAI mRNA levels could be due to changes in mRNA stability or apoAI transcription. To determine the potential of ER␣ to regulate apoAI gene transcription, a luciferase reporter under the control of both 5Ј-and 3Јflanking regulatory sequences of the human apoAI gene (reporter-2500 AI.LUC.CIII/AIV, Ref. 21) was cotransfected with an ER␣ expression vector into HepG2 cells. After transfection, the cells were treated with estrogen agonists and antagonists, and the reporter activity in cell extracts was determined. Doseresponse experiments with E2 resulted in a 75% maximal repression of apoAI promoter activity with an EC 50 value of approximately 12 nM ( Fig. 2A). The repression was both ligandand receptor-dependent, since it did not occur with either ER␣ or E2 alone (Fig. 2B). The estrogen receptor antagonist ICI 182,780 (1 M) also repressed promoter activity but to a lesser extent (30% repression). A 10-fold molar excess of ICI (1 M) over E2 (100 nM) effectively competed the E2-mediated repression to the level seen with ICI alone. ERE.LUC reporter activity was also regulated by ER␣ in a ligand-dependent fashion, except that in contrast to the apoAI promoter, ICI acted as a pure antagonist for ERE activation. The mechanism of ER␣ repression on the apoAI promoter may be distinct from those involved in ER␣ activation of an ERE.
Mapping of the apoAI Promoter Estrogen Response Element-A collection of apoAI promoter reporters (18) was used to delineate elements involved in ER␣-induced repression of the apoAI promoter. Deletion of the entire 3Ј-flanking region of the apoAI gene (reporter Ϫ2500 AI.LUC) or deletion of both the 3Ј region and approximately 2.25 kilobases of the 5Ј region (reporter Ϫ256AI.LUC) did not affect ER␣ and E2-induced repression (Fig. 3A). A reporter containing only the apoAI hepatocyte-specific enhancer driving the expression of the apoAI basal promoter (reporter Ϫ220/Ϫ110AI.LUC) was also repressed approximately 60% by ER␣ plus E2. The activity of the apoAI enhancer in a heterologous reporter containing the thymidine kinase promoter was also repressed by ER␣ plus E2, whereas the thymidine kinase basal promoter reporter activity remained unaffected (Fig. 3B), demonstrating that ER␣ regulation occurs directly on the apoAI enhancer.
ApoAI enhancer activity in hepatocytes depends on synergistic interactions between transcription factors bound to three distinct sites designated A (Ϫ214 to Ϫ192), B (Ϫ169 to Ϫ146), and C (Ϫ134 to Ϫ119) (18). To determine the contribution of each of these sites to the ER␣ and E2-induced repression, nucleotide mutations (denoted by X, Fig. 3B) were introduced into sites A, B, or C, rendering them incapable of binding their cognate transcription factors (18). Although the basal activity of these reporters was reduced compared with the intact enhancer, the range of repression was 48 -69% in response to ER␣ and E2. These results suggest that repression occurs at a level distinct from inhibition of individual transcription factors bound to the enhancer.
The ER␣ DNA Binding Domain and Transcription Activa-tion Functions 1 or 2 Are Necessary for apoAI Enhancer Repression-To further probe the mechanism by which ER␣ and E2 repressed the apoAI enhancer, vectors expressing ER␣ with a deletion in the N-terminal transcription activation function (AF1) or point mutations in the C-terminal transcription activation function (AF2) or in the DBD were cotransfected with either the Ϫ220/Ϫ110 ABC.LUC or ERE.LUC reporter into HepG2 cells (Fig. 4). Deletion of AF1 (mutant X-DBD-AF2) had no effect on apoAI enhancer repression but inhibited ERE activation by 80%. In contrast, inactivation of AF2 (mutant AF1-DBD-X) diminished both apoAI repression and ERE activation. Inactivation of both AF1 and AF2 completely abolished both apoAI repression and ERE activation. Finally, point mutations within the DNA binding domain that converts its binding selectivity from an ERE to a glucocorticoid response element (mutant AF1-X-AF2) eliminated repression of the apoAI enhancer. As shown previously (51), this ER␣ DNA binding domain mutant activated a reporter driven by a glucocorticoid FIG. 1. 17␤-estradiol represses apoAI mRNA levels in Hep89 cells. A, Western blot experiments were performed with nuclear extracts prepared from HepG2 and Hep89 cells and assayed for ER␣ protein levels. B, Hep89 and HepG2 cells were cotransfected with 0.5 g of ERE.LUC and 0.5 g of pRSV␤-gal reporters. After transfection, the cells were refed deficient growth medium in the absence (Ϫ) or presence (ϩ) of 100 nM E2 for 24 h. The luciferase values are reported as relative luciferase activities after correction for the corresponding ␤-galactosidase activities. C, Hep89 or HepG2 cells were treated with 1 M E2, and total RNA was isolated at the indicated time points (C, untreated control). The levels of human apoAI and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA were determined by Northern blot as described under "Materials and Methods." D, the hybridized signals from the above Northern blots were quantitated by phosphoimaging analysis and expressed as apoAI/glyceraldehyde-3-phosphate dehydrogenase ratios, with the control sample being designated 100. response element (data not shown).
These data suggest that the DNA binding domain of ER␣ cooperates with either AF-1 or AF-2 to repress the activity of the apoAI enhancer. Moreover, the observation that the X-DBD-AF2 mutant is as efficacious as wild type ER␣ in apoAI repression but not ERE reporter activation suggests that different mechanisms underlie ER␣ activation of an ERE reporter and repression of the apoAI reporter.
ER␣ Does Not Bind to the apoAI Enhancer-The requirement of an intact DNA binding domain for ER␣ repression of the apoAI enhancer suggested that ER␣ could bind directly to the enhancer and interfere with the synergistic interactions between bound transcription factors. We therefore determined if partially purified ER␣ binds to the Ϫ220/Ϫ110 apoAI enhancer fragment by electrophoretic mobility shift assays. Un-der these conditions, ER␣ did not bind to the apoAI enhancer but bound efficiently to the vitellogenin ERE as expected (Fig.  5, A and B). Control experiments showed that HNF-4 bound strongly to the apoAI enhancer as expected (20,25) (Fig. 5A) and weakly to the vitellogenin ERE (Fig. 5B). Since HNF-4 does not activate the ERE.LUC reporter in HepG2 cell transfections (data not shown), this binding interaction is unlikely to occur in the intact cell. The presence of additional cellular proteins did not overcome the inability of ER␣ to bind to the apoAI enhancer, as demonstrated in experiments comparing extracts from HepG2 cells and Hep89 cells (Fig. 5C) or when HepG2 extracts were supplemented with ER␣ (Fig. 5D). In both instances, no ER␣ binding to the enhancer could be detected, although binding to the ERE was observed. Thus even under conditions where nonspecific binding of HNF-4 to a noncognate sequence could be detected, no direct binding of ER␣ to the apoAI enhancer was observed.
ER␣ and E2 Do Not Influence HNF-4 or HNF-3␤ Abundance or Function-HNF-4, which binds to sites A and C in the apoAI enhancer (20,22,25), and HNF-3␤, which binds to site B (19), are two fundamental transcription factors involved in apoAI enhancer function. Ligand-activated ER␣ could repress apoAI enhancer activity by reducing the abundance or by inhibiting the activation properties of endogenous HNF-4, HNF-3␤, or both (see for example Refs. 42 and 56). To examine the first possibility, the effects of E2 on the relative abundance of HNF-4 and HNF-3␤ in Hep89 cells was determined by Western blotting analysis. As shown in Fig. 6, A and B, neither HNF-4 nor HNF-3␤ protein levels were altered after treatment with E2.
To determine whether ER␣ affects transcription factor activity, HepG2 cells were cotransfected with a reporter gene driven by the apoAI basal promoter and site A (reporter A.LUC) along with HNF-4 and ER␣ expression vectors. Cotransfection of the HNF-4 expression plasmid increased A.LUC activity 4-fold. This activation was not influenced by the presence of ER␣ and E2 (Fig. 6C). Similarly, a reporter driven by the apoAI basal promoter and site B (reporter B.LUC) was stimulated 8-fold by cotransfection of an HNF-3␤ expression vector, and ER␣ and E2 did not affect this activity (Fig. 6D). Thus, the repressive effects of ER␣ is not due to reduction in abundance or activity of HNF-4 or HNF-3␤. These findings raise the possibility that ER␣ may affect a non-DNA-bound factor required for optimal apoAI enhancer activity.
RIP140 Is a Coactivator for the apoAI Enhancer-Although optimal transcriptional activity of the apoAI enhancer depends on both DNA-bound transcription factors and non-DNA-bound coactivators (18,26), there is no information regarding the identity of these coactivators. We tested two well characterized ER␣ coactivators, CBP (38) and RIP140 (33) for their possible involvement in apoAI enhancer activity. Specifically, the apoAI enhancer reporter (Ϫ220/Ϫ110 ABC.LUC) was cotransfected with increasing amounts of either RIP140 or CBP expression vectors. CBP did not alter apoAI enhancer activity. In contrast, RIP140 repressed activity in a dose-dependent fashion (Fig. 7), reminiscent of its squelching affects on ER␣ activation of ERE reporters (33). RIP140 did not affect the function of an SV40 enhancer driven reporter (data not shown), demonstrating that the RIP140 effect was specific to the apoAI enhancer.
The Role of Coactivator Partitioning in ER␣-mediated Repression of the apoAI Enhancer-Transcriptional interference among nuclear receptors is due, at least in part, to partitioning of limited amounts of shared transcriptional coactivators (36,39). It is therefore possible that ER␣ and the apoAI enhancer share common cofactors and that partitioning of these cofactors to ER␣ results in repression of the enhancer. To test this possibility, HepG2 cells were cotransfected with the Ϫ220/ Ϫ110 ABC.LUC reporter, a constant amount of ER␣ expression vector, and increasing amounts of RIP140 expression vector. In the absence of E2, increasing amounts of RIP140 expression vector repressed apoAI enhancer activity as observed above (Fig. 8A). However, in the presence of E2, low amounts of RIP140 expression vector (50 ng) reversed ER␣ and E2-induced repression from 70% to approximately 35%. In fact, the reduced enhancer activity obtained at high RIP140 levels was actually stimulated 3-fold by ER␣ in an E2-dependent fashion (Fig. 8B). RIP140 does not appear to repress ER␣ production as determined by transient transfections in which different promoter/ enhancers (i.e. CMV enhancer or adenovirus major late promoter) used to drive expression of ER␣ gave similar results (data not shown). Together, these data indicate that ER␣ can affect apoAI gene expression via transcription coactivators and that RIP140 or an endogenous RIP140-like protein may be involved in apoAI enhancer regulation by ER␣ and E2. FIG. 3. ApoAI enhancer repression by ER␣ and 17␤-estradiol. A, a series of apoAI promoter deletion reporter plasmids were cotransfected with the pMT2-ER␣ expression vector into HepG2 cells. The structure of the constructs is indicated. Bent lines were used to indicate plasmid integrity. After transfection, the cells were cultured in the absence (open bars) or presence of 100 nM E2 (filled bars) for 24 h. Luciferase activities were determined as in Fig. 2B. B, a series of apoAI enhancer reporter plasmids were analyzed as above. A, B, and C denote transcription factor binding sites in the enhancer, with X indicating nucleotide substitutions that disrupt nuclear factor binding to these sites. The luciferase values are reported as relative luciferase activities with the activity of the reporter Ϫ41.LUC defined as 1. TK, thymidine kinase.

DISCUSSION
The apoAI liver-specific enhancer plays a central role in integration of diverse physiological and environmental signals affecting apoAI gene expression. Our working hypothesis is that signal-induced transient multiprotein complexes containing both DNA binding factors and factors not directly bound to DNA assemble onto the apoAI enhancer and regulate one or more steps in transcription initiation.
In this study, estrogenic signals influenced apoAI expression at the mRNA level by repressing apoAI gene transcription. More specifically, ER␣ and E2 potently repressed apoAI liverspecific enhancer activity in a ligand-and receptor-dependent manner. Although gel shift experiments did not provide evidence for ER␣ binding to the enhancer, the DBD was essential for apoAI repression. In addition, the ER␣ transcription activation functions AF1 and AF2, although not sufficient by them-  4 and 5, 0.02 and 0.1 g, respectively) were mixed with oligonucleotide probes containing the apoAI enhancer (A) or the vitellogenin ERE (B) and analyzed in electrophoretic mobility shift assays. Lane 1 contained the probe alone (Ϫ), and the reactions containing ER␣ also contained 100 nM E2. C, HepG2 and Hep89 extracts were mixed with either the apoAI enhancer or ERE probe as indicated. A monoclonal antibody (Ab) specific for ER␣ was included as indicated. D, electrophoretic mobility shift assay experiments using apoAI enhancer and ERE probes with HepG2 extracts supplemented with partially purified ER␣ and incubated in the absence (Ϫ) or presence (ϩ) of ER␣ antibody. selves, were individually essential for repression when associated with the DBD. Experiments with ER␣ mutants showed that the combination of the DBD and AF2 domain is as effective in enhancer repression as the wild type ER␣, whereas the combination of the DBD and AF1 domain is only 50% as effective as full-length ER␣. All ER␣ mutants tested were capable of ligand binding (51,57,58), and apoAI enhancer repression by them was strictly ligand-dependent. Thus, it appears that a ligand-induced change in the receptor cooperates with the DBD and either AF1 or AF2 to impart an apoAI enhancer repressing activity.
How does ER␣ and E2 repress apoAI enhancer function? Previous studies showed that maximal activity of the apoAI enhancer depends on synergistic interactions between transcription factors bound to enhancer sites A, B, and C (18). Because ER␣ does not bind to the enhancer under our gelshifting conditions, repression mechanisms involving transcription interference by quenching (59) seem unlikely. ER␣ may repress the apoAI enhancer by inhibiting the activity of transcription factors required for enhancer function (see for example Refs. 42 and 56). For example, we have recently observed that adenovirus E1A inhibits apoAI enhancer activity by selective inactivation of HNF-3, the factor that binds to site B. 2 To test this possibility, two liver-specific transcription factors, HNF-4 and HNF-3␤, involved in apoAI enhancer function were assayed. The results showed that neither HNF-3␤ nor HNF-4 activities or protein levels were influenced by ER␣ plus E2, suggesting that ER␣ represses the apoAI enhancer at some other level. Although the possibility that ER␣ activates a repressing transcription factor that binds to the enhancer and inhibits its function cannot be unequivocally excluded, the observation that ER␣ could mediate apoAI enhancer repression independent of individual cis-elements mutations within the enhancer suggests that repression occurs at a level secondary to DNA binding.
An alternative explanation for the ER␣-and E2-induced repression of the apoAI enhancer is that a coactivator common to the enhancer and ER␣ is partitioned to ER␣, leading to transcriptional interference with the enhancer. HNF-4, which binds to site A, is a member of the nuclear receptor superfamily and could share coactivators with ER␣ similar to other nuclear 2 E. Kilbourne and S. K. Karathanasis, unpublished data. receptors (reviewed in Ref. 41). For example, partitioning of p300/CBP to ER␣ was recently shown to be the major mechanism of transcriptional interference between ER␣ and the progesterone receptor (39) or the transcription factor AP-1 (36). The involvement of coactivators in nuclear receptor function and their importance in apoAI enhancer function prompted us to test the possibility that CBP or RIP140 is involved in apoAI enhancer function. The results showed that expression of CBP does not affect apoAI enhancer activity. However, similar experiments with RIP140 showed that this cofactor repressed apoAI enhancer activity, reminiscent of the RIP140 squelching effects on ER␣ activation (33). In the presence of E2, low RIP140 expression vector levels reversed the ER␣-mediated repression of apoAI enhancer activity. Therefore, RIP140-like factors may be one class of non-DNA binding coactivators in-volved in apoAI enhancer function that are partitioned to ER␣.
The observation that two different activation functions in ER␣ (i.e. AF1 and AF2) can in conjunction with the DBD repress apoAI enhancer activity independently of each other raises the possibility that ER␣ partitions an additional cofactor(s) involved in apoAI enhancer function. Consistent with this, RIP140 interaction with ER␣ requires a functional AF2 domain (60) that is inactive in the ER␣ mutant AF1-DBD-X that represses enhancer activity. Therefore a model whereby, in addition to RIP140, ER␣ shares other cofactors with the apoAI enhancer and RIP140 binding to ER␣ alters its conformation, so that some, but not all, of these factors are released and used by the enhancer is possible.
How could the disparities regarding ER␣ regulation of apoAI and HDL plasma levels between different cell and animal systems (see the Introduction) be reconciled? It is clear from our previous work that the multiprotein complexes assembled onto the apoAI enhancer are transient, and their protein composition is influenced by the prevailing developmental, physiological, and environmental factors affecting apoAI gene expression. For example, although apoAI enhancer activity does not depend on retinoids, prior repression of enhancer activity by the nuclear receptor apolipoprotein AI regulatory protein-1 converts the enhancer into a retinoid-responsive element (61). A similar phenomenon may be occurring with E2 in which the state of the apoAI enhancer due to intracellular coactivator levels determines the mode of ER␣ regulation. When cofactors shared by the apoAI enhancer and ER␣ are present in limiting amounts compared with ER␣, their partitioning by ER␣ will result in enhancer inhibition. This latter case appears to be operating in the Hep89 cell system. In contrast, when these cofactors are in excess compared with ER␣ and apoAI enhancer activity is partially repressed due to endogenous squelching, estrogen-activated ER␣ will alleviate repression by partitioning cofactors in excess and result in enhancer stimulation. Consistent with this, the low level enhancer activity obtained at high RIP140 expression vector levels was stimulated 3-fold by ER␣ in a ligand-dependent fashion. In addition, differences in ER␣ affinity for these coactivators, induced by different ligands, may explain the opposite effects of different estogen agonists and antagonists on apoAI plasma levels (62,63). Therefore, we propose that the effects of estrogen on apoAI gene transcription will depend upon the balance between cofactors shared by the apoAI enhancer and ER␣ and may explain the disparities observed for apoAI gene regulation by E2 obtained using different cell and animal systems.