Binding specificity and modulation of the ApoA-I promoter activity by homo- and heterodimers of nuclear receptors.

Three proximal regulatory elements, AIB, AIC, and AID, of the apoA-I gene are necessary and sufficient for its hepatic expression in vivo and in vitro. DNA binding and competition assays showed that elements AIB and AID contain hormone response elements composed of imperfect direct repeats that support the binding of the hepatic nuclear factor-4, other nuclear orphan receptors, and the ligand-dependent nuclear receptors retinoic X receptor (RXRalpha), RXRalpha/RARalpha, and RXRalpha/T3Rbeta. Substitution mutations on repeats 1 and 2 in the hormone response sites of elements AIB and AID, respectively, abolished the binding of all nuclear receptors and reduced promoter activity to background levels, indicating the importance of both hormone response elements for the hepatic expression of the apoA-I gene. Cotransfection experiments in HepG2 cells with normal and mutated promoter constructs and plasmids expressing nuclear hormone receptors showed that RXRalpha homodimers transactivated the wild type promoter 150% of control, in the presence of 9-cis-retinoic acid (RA), whereas RXR alpha/T3R beta heterodimers repressed transcription to 60% of control, in the presence of T3. RXR alpha/RAR alpha and hepatic nuclear factor-4 did not affect the transcription, driven by the proximal apoA-I promoter. Potassium permanganate and dimethyl sulfate interference experiments showed that RXRalpha homodimers, RXRalpha/RARalpha, and RXRalpha/T3Rbeta heterodimers participate in protein-DNA interactions with 12, 13, and 11 out of the 14 nucleotides, respectively, that span repeats 1 and 2 and the spacer region separating them on the hormone response element of element AID. The binding of RXRalpha homodimers and RXRalpha/T3Rbeta heterodimers is associated with ligand-dependent activation by 9-cis-RA or repression by T3. Upon deletion or mutation of repeat 1, homodimeric binding of RXRalpha is lost whereas heterodimeric binding is retained. This heterodimeric binding to the mutated element AID is mediated solely by interactions with repeat 2 and one adjacent nucleotide and is confined to a heptameric core recognition motif. The interactions of the RXRalpha heterodimers with repeat 2 are associated with low levels of ligand-independent transcriptional activity. The findings suggest that the specific types of homo- and heterodimers of nuclear hormone receptors occupying the hormone response elements of apoA-I and the availability of the ligand may play an important role in the transcriptional regulation of the human apoA-I gene.

Epidemiological data and transgenic animal experiments have shown that increased apoA-I and high density lipoprotein levels are protective against atherosclerosis (1,2). Thus the mechanisms which regulate the synthesis of apoA-I are important. Previous studies have shown that the regulatory elements AIB (Ϫ128/Ϫ77), AIC (Ϫ175/Ϫ148), and AID (Ϫ220/Ϫ190) are sufficient for hepatic expression of the human apoA-I gene in tissue culture (3,4) and in transgenic mice (5). It has been shown that the regulatory element AIC is recognized by heat stable factors related to CCAAT/enhancer-binding protein (C/ EBP) which act as positive regulators and by heat labile activities, one of which acts as a negative regulator (4). In addition, the regulatory element AID contains a hormone response element (HRE) 1 that is recognized by ARP-1, a transcriptional repressor of apoA-I (6), hepatic nuclear factor-4 (HNF-4) (7), RXR␣ homodimers, RXR␣/RAR␣, RXR␣/ARP-1 (8,9), and RXR␣/PPAR heterodimers (10). All these factors are members of the steroid/thyroid receptor superfamily, that includes receptors for steroids, thyroid hormone, and retinoic acids as well as orphan nuclear receptors with unidentified ligands. In the present study we have demonstrated the existence of an additional HRE on the regulatory element AIB between nucleotides Ϫ132/Ϫ119 and we have investigated the role of both HREs and of different types of hormone nuclear receptors on the transcriptional regulation of the human apoA-I gene. We demonstrate that both HREs are recognized by homodimers of HNF-4, ARP-1, EAR-2, EAR-3, and RXR␣ as well as heterodimers of RXR␣ with RAR␣ or T 3 R␤, and that both elements are essential for the hepatic expression of the apoA-I gene.
Binding of the RXR␣ homodimers on the HRE of element AID requires direct repeats 1 and 2 and leads to ligand-dependent transcriptional activation whereas binding of the RXR␣/RAR␣ and RXR␣/T 3 R␤ heterodimers on this HRE may occur on direct repeats 1 and 2 or only on repeat 2. When bound on both repeats 1 and 2, the RXR␣/T 3 R␤ heterodimers repress transcription in the presence of T 3 , whereas the RXR␣/RAR␣ heterodimers and HNF-4 do not affect the transcription. In addition, binding of the RXR␣ heterodimers to only one repeat on the HRE of element AID is associated with low levels of ligandindependent transcriptional activity. The findings demonstrate that hormone nuclear receptors can modulate the transcription of the human apoA-I gene and thus may affect plasma apoA-I and high density lipoprotein levels.
Plasmid Constructions and CAT Assays-The apoA-I promoter region was derived from an apoA-I genomic clone as described previously (11). The proximal promoter region spanning nucleotides Ϫ264/ϩ5 was subcloned into the Asp718 and SmaI restriction sites of the pUCSH-CAT plasmid to produce the CAT derivative designated [Ϫ264/ ϩ5]apoA-I-CAT (4). This reporter construct was used to create mutated apoA-I promoter constructs. The mutated Ϫ264/ϩ5 CAT constructs containing deletions or nucleotide substitutions in elements AIB and AID were generated by amplification of the parent pUCSH[Ϫ264/ϩ5] apoA-I CAT construct (4). For example, in order to generate the mutation AIDM1, the region upstream of nucleotide Ϫ186 was amplified by PCR using the PCR-264AI 5Ј primer which contains an Asp718 and EcoRV restriction sites and the mutagenic AIDM1 primer ( Table I). The region downstream of nucleotide Ϫ225 was amplified by using the AIDM1C mutagenic primer and the PCR-19AI 3Ј primer which contains an SspI restriction site. An aliquot containing 5% of the two amplified regions was used as a template for further amplification by the 5Ј and 3Ј primers PCR-264AI and PCR-19AI. The amplified DNA was digested with Asp718 and SspI and cloned into the Asp718 and SmaI sites of pUCSH-CAT (4). The remaining mutations were constructed in a similar manner using the amplification primers shown in Table I. PCR reactions were performed using the Perkin Elmer automated thermocycler according to the manufacturer's specifications. The sequence of the final constructs was determined by DNA sequencing. Oligonucleotides used were synthesized by the solid-phase phosphite triester method using an automated AB-380B oligonucleotide synthesizer.
Preparation of Nuclear Extracts and Whole Cell Extracts from Transfected COS-1 Cells-Nuclear extracts were prepared from livers of 10 rats (approximately 120 g of liver) as described (12). Extracts from COS-1 cells transfected with the pMT2 vector carrying full-length cDNAs for HNF-4, ARP-1, EAR-2, EAR-3, RXR␣, and T 3 R␤ were prepared as described (6). Similarly prepared were extracts for COS-1 cells transfected with the pSG5 expression vector carrying a flag epitope fused to the NH 2 -terminal of T 3 R␤.
Gel Electrophoretic Mobility Shift Assay-This analysis was performed using either crude hepatic nuclear extracts or COS-1 whole cell extracts as described (13). Competitors were used at 50 -100-fold excess. In supershift assays, various dilutions of polyclonal or monoclonal antibodies were added to the reaction mixture prior to the addition of the probe.
Dimethyl Sulfate and Potassium Permanganate Interference Assays-For the methylation interference assay, single stranded DNA (5 pmol) was end-labeled with T 4 polynucleotide kinase and annealed to its complementary unlabeled strand. Double stranded DNA (10 7 cpm) was treated with dimethyl sulfate for 3 min at room temperature in the presence of 2 g of salmon sperm DNA. For permanganate interference assay, single-stranded DNA (10 7 cpm) was treated with potassium permanganate (KMnO 4 ) for 10 min at room temperature in the presence of 4 g of salmon sperm DNA followed by annealing to its complementary strand (14). The treated probes were incubated with COS-1 whole cell extracts expressing the indicated nuclear receptors; the complexes were analyzed by a preparative mobility shift assay. Following electrophoresis the protein-DNA complexes and the free probe were excised from the gel, electroeluted, and treated with 1 M piperidine for 30 min at 95°C. The samples were then dried, the dry pellets were counted and dissolved in 98% formamide dye. Equal counts from all the samples were fractionated by electrophoresis and the bands were visualized by autoradiography.
Transient Transfection Experiments and CAT Assays-Monolayers of COS-1 or HepG2 cells were maintained as stocks in Dulbecco's modified Eagle's medium supplemented with either 10% fetal calf serum or charcoal stripped 5% fetal calf serum, respectively. 0.5 ϫ 10 6 HepG2 cells were plated on 30-mm dishes and the following day were transfected using the calcium-phosphate DNA co-precipitation method. A total of 6 g of plasmid DNA was used, containing the wild type or mutant apoA-I-CAT plasmids, phosphoglycerol kinase ␤-galactosidase plasmid (generous gift of Dr. F. Mavilio) as internal control, and various concentrations of pMT2 expression plasmids carrying the cDNAs of different hormone nuclear receptors. Post-transfection HepG2 cells were subjected to 30 and COS-1 cells to 60 s of glycerol shock. Cotransfection experiments with the ligand-dependent nuclear receptors RXR␣, RAR␣, and T 3 R␤ were performed either in the absence or presence of 10 Ϫ6 M 9-cis-RA (generous gift of Dr. H. Gronemyer), 10 Ϫ6 M all-trans-RA, or 10 Ϫ7 M T 3 . The cells were harvested 40 h post-transfection and lysed in 200 l of 1 ϫ reporter lysis buffer. An aliquot of the cell extracts was heated at 65°C for 10 min prior to the CAT assays. The assays were performed in a 7-ml plastic scintillation vial, in a total volume of 250 l, containing 100 mM Tris-HCl, pH 7.8, 1 mM chloramphenicol, 0.25 Ci of 3 H-labeled acetyl coenzyme A, and 20 -30 l of extracts. The reaction mixture was also supplemented with 100 M cold acetyl coenzyme A. One blank sample and one sample containing 5 milliunits of purified CAT enzyme were always included. The reaction mixture was overlaid with 4 ml of water-immiscible scintillation fluid and incubated at 42°C for a maximum of 45 min before it was counted in a liquid scintillation counter. The units of active CAT enzyme in the cell extracts was determined by comparison of the radioactivity counts obtained in the samples containing the cell extracts and the counts obtained in samples containing 5 milliunits of purified CAT enzyme. The background values of the blank sample were always subtracted from the counts/min values of the different samples. Each experiment was repeated two to three times in quadruplicate and the mean values were calculated. The ␤-galactosidase activity of the cell lysates was determined spectrophotometrically by monitoring the hydrolysis of the synthetic substrate onitrophenyl galactoside at 410 nM and was utilized to normalize for variability in the efficiency of transfection. Control samples containing 5, 7.5, and 10 milliunits of purified ␤-galactosidase allowed the conversion of the OD units of the different samples into ␤-galactosidase units.

Binding of Orphan and Ligand-dependent Nuclear Receptors to the Regulatory Elements AIB and AID of the Human ApoA-I: Effect of Promoter Mutations on Binding
The regulatory element AID (Ϫ220/Ϫ190) contains sequences that share high similarity with an AGG/TTCA motif (half-site) found in HREs on the promoter regions of a variety of genes (15)(16)(17). Examination of the HRE on element AID showed the presence of three putative direct repeats between nucleotides Ϫ190 to Ϫ210, whereas the element AIB contains two putative direct repeats between nucleotides Ϫ132 to Ϫ119 (Fig. 1A). DNA binding assays have shown that both regulatory elements AIB and AID can support the homodimeric binding of HNF-4, ARP-1, EAR-2, EAR-3, and RXR␣ as well as the heterodimeric binding of RXR␣/RAR␣ and RXR␣/T 3 R. Monomers of T 3 R␤ or homodimers of either T 3 R␤ or RAR␣ do not bind to either regulatory element (Fig. 1, B-C). The findings establish that the regulatory elements AIB and AID of apoA-I contain functional HREs which can bind a variety of orphan and ligand-dependent nuclear receptors. We have performed DNA binding supershift experiments to characterize the putative RXR␣/RAR␣ and RXR␣/T 3 R␤ heterodimers which bind to the regulatory element AID. For this analysis we have utilized a monoclonal anti-RAR␣ antibody and an anti-flag antibody against T 3 R␤ which contains an amino-terminal flag sequence encoding for Asp-Tyr-Lys-Asp-Asp-Asp-Asp-Lys. The expression vector carrying a flagged version of T 3 R␤ was transiently expressed in COS-1 cells. These COS-1 extracts, as well as extracts of COS-1 cells expressing RXR␣ or RAR␣ were utilized in the supershift assays. Fig. 1D shows that the monoclonal anti-RAR␣ antibody supershifted the RXR␣/RAR␣ heterodimers bound to the wild type oligonucleotide AID. Similarly, the anti-flag antibody supershifted the RXR␣/T 3 R␤ heterodimers, bound to oligonucleotide AID (Table I).
To explore potential differences in the binding requirements of different members of the hormone nuclear receptor family we generated a series of nucleotide substitution mutations in the elements AIB and AID, designated AIBM, AIDM, AIDM1, AIDM2, AIDM3, AIDM4, and AIDM5 (Table II). AIBM and AIDM mutations span part of both repeats of element AIB or only the second repeat of element AID, respectively. Mutation AIDM1 is localized upstream of the HRE on element AID, whereas mutation AIDM2 contains nucleotide substitutions in repeat 1. Finally, the AIDM3 and AIDM4 mutations contain substitutions in repeat 2, and the AIDM5 mutation contains substitutions in the putative repeat 3. We also generated two deletion mutations, one of repeat 1 and one of putative repeat 3, in the HRE of element AID, which were designated AI⌬REP1 and AI⌬REP3, respectively. The oligonucleotides containing these mutations, shown in Table II, were tested in DNA binding and competition experiments. These analyses showed that the AIDM1 mutation had no effect on the binding of rat hepatic nuclear extracts (compare lane 1 with lane 7 of Fig. 2A). The AIDM2 mutation increased substantially the binding of the slower migrating activity present in the rat liver nuclei (compare lane 1 with lane 9 of Fig. 2A). The AIDM3 and AIDM4 mutations did not bind substantially or compete for the binding of hepatic nuclear activities to element AID ( Fig. 2A, lanes 5, 6,  11, and 13). The DNA-protein complexes formed with AIDM4 are not competed out by the wild type AID sequences (data not shown). Thus, these complexes must originate from the binding to the mutated probe of activities unrelated to nuclear hormone nuclear receptors which do not normally bind to the wild type probe. The AIBM mutation and the AIDM mutation which altered drastically the HREs of element AIB and AID, respectively, abolished the binding of all nuclear activities to element AID (Fig. 2D, lanes 8 and 16), whereas deletion or mutation of the putative repeat 3 and deletion of repeat 1 (AI⌬REP3, AIDM5, and AI⌬REP1) did not affect qualitatively the binding of hepatic nuclear activities to this site (Fig. 2, E and F).
We have also tested the effects of the deletion and oligonucleotide substitution mutations within the two regulatory elements AID and AIB (Table II) on the binding of orphan and ligand-dependent nuclear receptors. This analysis showed that the AIDM1 mutation did not affect the binding of the homodimers of the orphan receptors HNF-4, ARP-1, EAR-2, and EAR-3 (Fig. 2B). The remainer of the mutations affected the binding of the orphan receptors differentially. Binding of HNF-4 was moderately affected by changes in repeat 1 (AIDM2 mutation) and it was greatly affected by changes in the first half of repeat 2 (AIDM3 mutation). Finally, binding was dimin- ished by changes in the second part of repeat 2 (AIDM4 mutation). Binding of ARP-1, a known negative regulator of apoA-I (6), was not greatly affected by the first three mutations (AIDM1, AIDM2, and AIDM3), whereas binding of the orphan receptors EAR-2 and EAR-3 was mainly affected by the AIDM3 mutation (Fig. 2B). Deletion of repeat 1 of element AID (AI⌬REP1) reduced the binding of HNF-4 but did not affect qualitatively the binding of the other orphan receptors ARP-1, EAR-2, and EAR-3 ( Fig. 2E), whereas deletion or mutations in the putative repeat 3 of element AID (AI⌬REP3 and AIDM5) did not affect the binding of any of the orphan receptors (Fig. 2, E and F). Finally the mutations AIBM and AIDM, which altered drastically the HREs of elements AIB and AID, respectively, abolished the binding of all the orphan nuclear receptors to this site (Fig. 2D). The findings suggest that the orphan nuclear receptors have different binding specificities for the regulatory element AID of apoA-I, although in most cases bind-ing is affected considerably by changes in repeat 2. The analysis with the ligand-dependent nuclear receptors showed that the AIDM2 and AI⌬REP1 mutations which affect the first repeat and the AIDM3 and AIDM4 mutations which affect the second repeat in the HRE of element AID abolished the binding of RXR␣ homodimers (Fig. 2, C and E). The AIDM4 mutations affected mainly the binding of RXR␣/RAR␣ heterodimers. The AIDM2, AIDM3, and AIDM4 mutations decreased the binding of RXR␣/T 3 R␤ heterodimers. Deletion of repeat 1 of element AID (AI⌬REP1) did not affect qualitatively the binding of RXR␣/RAR␣ and RXR␣/T 3 R␤ heterodimers (Fig. 2E). Deletion or mutations in the third repeat of the element AID (AIDM5 and AID⌬REP3) did not affect qualitatively the binding of RXR␣ homo-and heterodimers (Fig. 2, E and F). In addition, the AIDM2 mutation formed a DNA-protein complex of higher mobility than the complex formed with the wild type oligonucleotide or those carrying the AIDM3 and AIDM4 mutations   . Finally, the AIBM and AIDM mutations, which altered drastically the HREs of the regulatory elements AIB and AID, respectively, eliminated the binding of all combinations of ligand-dependent nuclear receptors to the mutated probe ( Fig. 2D).
In summary, the findings of Fig. 2, A-F, demonstrate that binding of RXR␣ homodimers requires intact repeats 1 and 2 in the HRE of element AID, whereas binding of RXR␣/RAR␣ and RXR␣/T 3 R␤ heterodimers can still occur when short mutations are introduced in either repeat 1 or repeat 2. Similarly, the binding of the orphan nuclear receptors is differentially affected by the various mutations. Binding of HNF-4 requires the intact second part of repeat 2, whereas binding of EAR-2 and EAR-3 requires the intact first part of repeat 2 of element AID. Additionally, binding of ARP-1 is affected only by extensive mutagenesis of repeat 2. Finally, the binding of crude rat liver nuclear extracts parallels that of HNF-4.

Contribution of the HREs to the Strength of the ApoA-I Promoter in HepG2 Cells
To assess the importance of the two HREs for the apoA-I promoter strength we introduced the mutations of Table II in the apoA-I promoter and generated mutant Ϫ264/ϩ5 A-I CAT constructs. Transient transfection assays in HepG2 cells showed that drastic mutations in the regulatory element AIB (AIBM mutation) or the regulatory element AID (AIDM mutation) or both elements (AIBDM mutation) reduced the promoter activity to 3-7% of the control (Fig. 3). This finding, combined with the DNA binding data of Fig. 2D establishes the importance of both HREs for the hepatic transcription of the apoA-I gene since promoter mutations which preclude the binding of hormone nuclear receptors to any of the two HREs essentially abolish the hepatic expression of apoA-I.
More detailed analysis of the effect of nucleotide substitution or deletion mutations in the regulatory element AID on the promoter strength showed that the AIDM2 mutation which altered nucleotides within the first repeat and the deletion (AI⌬REP1) of the first repeat of element AID reduced the promoter strength to approximately 50 and 35% of control, respectively. The AIDM3 and AIDM4 mutations which altered nucleotides within the second repeat reduced the promoter strength to 30 and 15% of control, respectively. A mutation 5Ј of repeat 1 (AIDM1) and a mutation within the putative third repeat (AIDM5) reduced the promoter strength to 75 and 90% of control, respectively, and deletion of the putative repeat 3 (AI⌬REP3) increased slightly the promoter strength to 115% of  Table II. The extracts utilized are indicated on the top of the figure and have been described in the legend to Fig. 1. Panel A, binding and competition assays of hepatic nuclear extracts to the wild type (AID) and the mutated (AIDM1, AIDM2, AIDM3, and AIDM4) probes. Panel B, binding of orphan nuclear receptors to the wild type (AID) and the mutated (AIDM1, AIDM2, AIDM3, and AIDM4) probes. Panel C, binding of ligand-dependent nuclear receptors to the wild-type (AID) and the mutated (AIDM2, AIDM3, and AIDM4) probes. Panel D, binding of rat liver nuclear extracts, orphan and ligand-dependent nuclear receptors to probes having mutations in both repeats of element AIB (AIBM) or in the second repeat of element AID (AIDM). Panel E, binding of rat liver nuclear extracts, orphan and ligand-dependent nuclear receptors to probes having deletions of either the putative repeat 3 (A1⌬REP3) or repeat 1 (A1⌬REP1) of element AID. Panel F, binding of rat liver nuclear extracts, orphan, and ligand-dependent nuclear receptors to a probe having nucleotide substitution in the putative repeat 3 (AIDM5) of element AID. control (Fig. 3). The combined data of Figs. 2 and 3 indicate that mutations in the regulatory element AID and AIB, which diminish the binding of hepatic nuclear activities and all types of nuclear receptors to the two HREs also decrease proportionally the strength of the apoA-I promoter. Mutations in repeat 2 of the HRE of element AID affected more severely the promoter strength as compared to those in repeat 1 (Fig. 3).

Mode of Binding of RXR␣ Homodimers, RXR␣/RAR␣, and RXR␣/T 3 R␤ Heterodimers on the Wild Type Regulatory Element AID as Determined by Dimethyl Sulfate and KMnO 4 Interference Assays
The observation that the RXR␣ homo-and heterodimers have different binding specificities on the regulatory element AID prompted us to determine the nucleotides which are involved in these DNA-protein interactions by permanganate (KMnO 4 ) and methylation (dimethyl sulfate) interference analysis using the wild type element AID as probe. This analysis showed that nucleotides at position 5 and 6 of repeat 1 and repeat 2, and nucleotides at position Ϫ1, located in the spacer region between repeats 1 and 2 of the coding strand, participate in strong DNA-protein interactions with RXR␣ homodimers. In addition, the nucleotide at position 6 of the putative third repeat, of the coding strand, participates in weak DNA-protein interactions with RXR␣. The KMnO 4 and dimethyl sulfate interference pattern of the noncoding strand showed that nucleotides at position 1 and 2 of repeat 1 as well as nucleotides at position 1, 2, 3, and 4 of repeat 2 participate in strong DNA-protein interactions with RXR␣, and the nucleotide at position 3 of repeat 1 participates in weak interactions with RXR␣ (Fig. 4, A-C). Overall, 12 out of the 14 nucleotides that form repeats 1 and 2 and the putative spacer region on the HRE of element AID participate in DNA-protein interactions with RXR␣ homodimers. Five of these nucleotides are on repeat 1, six on repeat 2, and one in the spacer region between repeats 1 and 2. Eleven nucleotides participate in strong and one in weak interactions. In addition, one nucleotide in the putative third repeat participates in weak DNA-protein interactions with the RXR␣ homodimers (Fig. 4C). These findings demonstrate the requirement of both repeat 1 and repeat 2 on the HRE of element AID for the binding of RXR␣ homodimers and the minimal involvement of the putative third repeat in this binding. The findings are also consistent with the DNA binding assays of homo-and heterodimers of RXR␣ to the mutated AID sequences (Fig. 2, C-F).
The KMnO 4 and dimethyl sulfate modification analysis was also utilized to determine the mode of binding of RXR␣/RAR␣ or RXR␣/T 3 R␤ heterodimers on the wild type element AID. The analysis with the RXR␣/RAR␣ heterodimers showed that all the nucleotides of repeats 1 and 2, which participate in DNAprotein interactions with the RXR␣ homodimers also participate in DNA-protein interactions with the RXR␣/RAR␣ heterodimers. In addition, nucleotide 4 of the noncoding strand of repeat 1 participates in weak DNA-protein interactions with the RXR␣/RAR␣ heterodimers (Fig. 4, D-F). Overall 13 of the 14 nucleotides that form repeats 1 and 2 and the putative spacer region between them participate in interactions with this heterodimer. Six of the nucleotides are in repeat 1, six in repeat 2, and one in the spacer region between repeats 1 and 2. Ten of the nucleotides participate in strong, and three participate in weak DNA-protein interactions. The nucleotides which participate in weak interactions are nucleotides 3 and 4 of the noncoding strand and nucleotide Ϫ1 of the coding strand. Repeat 3 is not involved in the binding of RXR␣/RAR␣ heterodimers (Fig. 4F). The analysis with the RXR␣/T 3 R␤ heterodimers showed that all but one of the nucleotides of repeats 1 and 2 which participate in DNA-protein interactions with the RXR␣ homodimers also participate in interactions with the RXR␣/T 3 R␤ heterodimers. The oligonucleotide which does not participate in DNA-protein interactions with the RXR␣/T 3 R␤ heterodimers is nucleotide 3 of the noncoding strand of repeat 1 (Fig. 4, G and H). Overall, four nucleotides of repeat 1 and seven nucleotides of repeat 2 and the putative spacer region between them participate in DNA-protein interactions with the RXR␣/T 3 R␤ heterodimers. Six of the nucleotides participate in strong and the remaining five in weak DNA-protein interactions. The nucleotides which participate in strong interactions are nucleotides 5 of repeat 1 and 5 and 6 of repeat 2 of the coding strand and nucleotides 1, 2, and 3 of repeat 2 of the noncoding strand. The remaining nucleotides as well as the nucleotide Ϫ1 in the putative spacer region between repeats 1 and 2 participate in weak DNA-protein interactions. Residues of the putative repeat 3 are not involved in the binding of the RXR/T 3 R␤ heterodimers (Fig. 4, G-I).

Mode of Binding of the RXR␣/RAR␣ and the RXR␣/T 3 R␤ Heterodimers on the Regulatory Element AID Carrying Mutation or Deletion of Repeat 1, as Determined by KMnO 4 and Dimethyl Sulfate Interference Assays
As shown in Fig. 2, C and E, mutation or deletion of the first repeat in the HRE of the regulatory element AID (AIDM2 and AID⌬REP1 mutations) prevents the binding of RXR␣ homodimers but allows the binding of RXR␣/RAR␣ and RXR␣/ T 3 R␤ heterodimers. To investigate further the mode of binding of the heterodimers to the mutated sequence we performed KMnO 4 and dimethyl sulfate interference analysis using the oligonucleotides AIDM2 and AID⌬REP1 as probes (Table II). The KMnO 4 and dimethyl sulfate modification pattern of the RXR␣/RAR␣ heterodimers with the mutated AIDM2 probe is shown in Fig. 5, A and B, respectively, and summarized in Fig.  5C. This analysis showed that the binding of the RXR␣/RAR␣ heterodimers to the mutated AIDM2 probe is confined to a heptameric core recognition motif. The seven oligonucleotides which participate in DNA-protein interactions are the six nucleotides of repeat 1 and the nucleotide Ϫ1 localized in the putative spacer region between repeats 1 and 2. Nucleotides 5 and 6 of the coding strand and nucleotide 4 of the noncoding strand participate in strong, and the remaining in weak, DNAprotein interactions with the RXR␣/RAR␣ heterodimers.
The KMnO 4 and dimethyl sulfate modification pattern of the RXR␣/T 3 R␤ heterodimers with the mutated AIDM2 probe is  Table II.

Role of the Hormone Response Elements in ApoA-I Gene Transcription
shown in Fig. 5, D and E, respectively, and summarized in Fig.  5F. This analysis showed that the binding of the RXR␣/T 3 R␤ heterodimers to the mutated AIDM2 probe is confined to a heptameric core recognition motif. The seven nucleotides of the coding and noncoding strand which participate in DNA-protein interactions with the RXR␣/T 3 R␤ heterodimers are identical to those which participate in DNA-protein interactions with the RXR␣/RAR␣ heterodimers. The exception here is that only 2  (Table II). The RXR␣ homo-and heterodimers were produced by expression of the corresponding cDNAs in COS-1 cells. The KMnO 4 and dimethyl sulfate modification pattern of RXR␣ homodimers with the coding and noncoding strand of element AID is shown in Panels A and B, respectively. Panel C is a summary of the interference pattern deduced from the findings of Panels A and B. The KMnO 4 and dimethyl sulfate modification pattern of RXR␣/RAR␣ heterodimers with the coding and noncoding strand of element AID is shown in Panels D and E, respectively. Panel F is a summary of the interference pattern deduced from the findings of Panels D and E. The KMnO 4 and dimethyl sulfate modification pattern of RXR␣/T 3 R␤ heterodimers with the coding and noncoding strand of element AID is shown in Panels G and H, respectively. Panel I is a summary of the interference pattern deduced from the findings of Panels G and H. F indicates free probe; and B indicates probe recovered from the DNA-protein complex after chemical treatment. Strong interactions are illustrated with large rectangles for the RXR␣ homodimers, ovals for the RXR␣/RAR␣ heterodimers, and diamonds for the RXR␣/T 3 R␤ heterodimers. Weak interactions are illustrated with small rectangles, ovals, or diamonds. The nucleotide sequence of the coding and noncoding strand of element AID is indicated on each side of Panels A, B, D 5. A-H, KMnO 4 and dimethyl sulfate modification pattern of the DNA-protein complexes formed with the RXR␣/RAR␣ or RXR␣/T 3 R␤ heterodimers using the mutated element AIDM2 or AI⌬REP1 as probes (Table II). The RXR␣/RAR␣ and RXR␣/T 3 R␤ heterodimers were produced by expression of the corresponding cDNAs in COS-1 cells. The KMnO 4 and dimethyl sulfate modification pattern of RXR␣/RAR␣ heterodimers with the coding and noncoding strand of element AIDM2 is shown in Panels A and B, respectively. Panel C is a summary of the interference pattern deduced from the findings of Panels A and B. The KMnO 4 and dimethyl sulfate modification pattern of RXR␣/T 3 R␤ heterodimers with the coding and noncoding strand is shown in Panels D and E, respectively. Panel F is a summary of the interference pattern deduced from the findings of Panels D and E. The KMnO 4 and dimethyl sulfate modification pattern of RXR␣/RAR␣ heterodimers with the coding and noncoding strand of the AI⌬REP1 probe is shown in Panels G and H. Panel I is a summary of the interference pattern deduced from the findings of Panels G and H. F indicates the free probe, and B indicates the probe recovered from the DNA-protein complex after chemical treatment. Strong interactions are illustrated with ovals for the RXR␣/RAR␣ heterodimers, and diamonds for the RXR␣/T 3 R␤ heterodimers. Weak interactions are illustrated with small ovals or diamonds. The nucleotide sequence of the coding and noncoding strand of the mutated element AID is indicated on each side of Panels A, B, D, E, G, and H. Nucleotides which participate in DNA-protein interactions are indicated by an asterisk (*). AID⌬REP1 probe is shown in Fig. 5, G and H, and summarized in Fig. 5I. The pattern is similar to that obtained with the AIDM2 probe which carries mutations in the first repeat. The only exception is that nucleotide 5 on the coding strand of the putative third repeat also participates in weak DNA-protein interactions.
The combined data of Figs. 4, A-I, and 5, A-I, indicate that a heptameric core recognition motif is the minimum sequence required for the binding of RXR␣/RAR␣ and RXR␣/T 3 R␤ heterodimers to the regulatory element AID. The binding of the RXR␣ homodimers requires the presence of both direct repeats of element AID. Finally, the putative repeat 3 has minimal participation in DNA-protein interactions with homo-or heterodimers of RXR␣. Fig. 6A shows a summary of all the nucleotides which participate in DNA-protein interactions with the various combinations of the ligand-dependent nuclear receptors. Fig. 6B summarizes the mode of binding of the ligand-de-pendent nuclear receptors RXR␣, RXR␣/RAR␣, and RXR␣/ T 3 R␤ on the wild type and the mutated regulatory element AID of the human apoA-I promoter. This summary is based on the DNA binding data of Figs. 1 and 2

Effect of Ligand-dependent Nuclear Receptors and HNF-4 on the ApoA-I Promoter Strength
Transactivation by RXR␣ Homodimers in the Presence of 9-cis-RA-We have performed cotransfection titration experiments in HepG2 cells with the wild type or mutated Ϫ264/ϩ5 apoA-I CAT constructs and plasmids expressing various combinations of ligand-dependent nuclear receptors in the presence or absence of their corresponding ligands. The experiments with RXR␣ were performed in the presence or absence of its ligand 10 Ϫ6 M 9-cis-RA and increasing amounts of an RXR␣ expression plasmid, ranging from 50 to 750 ng. This analysis showed that cotransfection with RXR␣ transactivated moderately (1.5-fold) the apoA-I promoter, in the presence of its ligand 9-cis-RA. Optimal ligand-dependent transactivation was observed in the range of 100 -250 ng of plasmid. In the absence of exogenously added ligand there was no significant transactivation in this range of concentrations whereas at higher concentrations there was a trend toward transcriptional repression (Fig. 7A).

Lack of Transactivation by RXR␣/RAR␣ Heterodimers-
The experiments involving RXR␣/RAR␣ heterodimers were performed with constant amounts (100 ng) of RXR␣ expression plasmid and increasing amounts of RAR␣, ranging from 50 to 500 ng. It is expected that the higher RAR␣ concentrations will favor the formation of heterodimers rather than homodimers of RXR␣. The experiments were performed in the absence of any ligand or in the presence of 10 Ϫ6 M 9-cis-RA or 10 Ϫ6 M all-trans-RA. This analysis showed that the RXR␣/RAR␣ heterodimers did not transactivate significantly the apoA-I promoter in the presence of any of the ligands. In the absence of both ligands and at higher concentrations of RAR␣, there was a trend toward transcriptional repression (Fig. 7B). The findings suggest that the RXR␣/RAR␣ heterodimers abolished the 1.5-fold transactivation achieved by RXR␣ homodimers in the presence of 9-cis-RA.

Repression of Transactivation by RXR␣/T 3 R␤ Heterodimers in the Presence of T 3 -
The experiments involving RXR␣/T 3 R␤ heterodimers were performed with constant amounts of RXR␣ (100 g) and increasing amounts of T 3 R␤, ranging from 50 to 500 ng for the reasons described above. The experiments were performed in the absence of any ligand or in the presence of either 10 Ϫ6 M 9-cis-RA or 10 Ϫ7 M T 3 . This analysis showed that the RXR␣/T 3 R␤ heterodimers repressed transcription to 60% of control in the presence of T 3 . In the presence of 9-cis-RA, cotransfection with 50 ng of T 3 R␤ and 100 ng of RXR␣ expression plasmids caused a 1.5-fold increase in transcription, similar to the increase observed with the RXR␣ homodimers (Fig.  7A). Most likely, this increase is the result of the formation of RXR␣ homodimers, promoted by the higher concentration of RXR␣ as compared to the T 3 R␤ vector, in the presence of 9-cis-RA. When the two receptor-expressing plasmids were used in equal concentration (100 ng), this increase was no longer apparent. transcription which reached 75% of the control value at 750 ng of HNF-4 expression plasmid (Fig. 7D).

Effect of Selected Mutations within the Repeats of the Regulatory Element AID on the ApoA-I Promoter Strength and Its Transactivation by HNF-4 and the Ligand-dependent Nuclear Receptors
Cotransfection experiments with HNF-4 and the AIDM1 to AIDM4 mutated promoter constructs showed that the apoA-I promoter strength remained similar in the presence and absence of HNF-4 (Fig. 8A). This indicates that diminished binding of HNF-4 to the regulatory element AID impairs the promoter strength despite the fact that other nuclear receptors may still bind to this element. Similar cotransfection experiments were performed in HepG2 cells using the AIDM2, AI⌬REP1, and AI⌬REP3 mutated promoter constructs and combinations of plasmids expressing RXR␣, RAR␣, or T 3 R␤. This analysis showed that the promoter which lacks repeat 3 (AI⌬REP3) behaves in all cases similarly to the wild type promoter (data not shown). The promoter which lacks repeat 1 (AI⌬REP1) had 30 -35% activity and was not affected by RXR␣ in the presence or absence of 9-cis-RA. Interestingly the promoter which carries a mutation in the first repeat of element AID (AIDM2) was transactivated 2-fold by RXR␣ in the presence or absence of 9-cis-RA (Fig. 8B).
Cotransfection experiments with RXR␣/RAR␣ heterodimers showed that the promoter which lacks repeat 1 (AI⌬REP1) exhibited 30 -35% activity and was not affected by RXR␣/RAR␣ heterodimers in the presence or absence of 9-cis-RA or alltrans-RA. The promoter which carries a mutation in the first repeat of element AID (AIDM2) was transactivated 2-fold by RXR␣/RAR␣ heterodimers in the presence or absence of any of the two ligands (Fig. 8C). Finally, cotransfection experiments with RXR␣/T 3 R␤ heterodimers showed that the promoter which lacks the first repeat had the same 30 -35% activity and was not affected by the RXR␣/T 3 R␤ heterodimers and their ligands. The promoter which carries a mutation in the first repeat of element AID (AIDM2) was transactivated 1.3-fold by RXR␣/T 3 R␤ heterodimers in the absence of any of the two ligands, was transactivated 2.1-fold in the presence of 9-cis-RA, and was not affected by T 3 (Fig. 8D). It should be noted that binding of the RXR␣/T 3 R␤ heterodimers to this mutated promoter sequence generates a slow migrating DNA-protein complex (Fig. 2C).
The combined data of Figs. 4, 5, and 8 indicate that binding of RXR␣/RAR␣ and RXR␣/T 3 R␤ heterodimers to repeat 2 of element AID is associated with low levels of ligand-independent transcriptional activity whereas binding of homo-or heterodimers of RXR␣ to repeats 1 and 2 can lead to ligand-dependent activation or repression of transcription.

The Two Proximal HREs of the Human ApoA-I Promoter Are
Essential for Hepatic Expression-The present study has focused on the functional significance of the regulatory elements AID and AIB of apoA-I, and the potential contribution of hormone nuclear receptors to the transcriptional activation of this promoter in hepatic cells. Sequence comparisons showed that the regulatory elements AID and AIB contain sequences with high similarity to an AGG/TTCA motif found on the promoter sites of genes responsive to members of the steroid/thyroid receptor superfamily (15, 16, 18). The HRE present on element AID is composed of three putative direct repeats with the sequence A/GGG/TTCA on the noncoding strand, whereas the HRE on element AIB is composed of two putative direct repeats with the sequence A/GGT/ATCA on the noncoding strand. In both cases there is a 1 to 2-nucleotide spacer region between the repeats. Drastic mutagenesis which altered either part of both repeats in the HRE of element AIB or repeat 2 and the adjacent spacer region in the HRE of element AID, eliminated the binding of hepatic activities present in rat liver nuclei and reduced the promoter strength to approximately 5-7% of control. These findings suggest that both HREs are essential for optimal hepatic expression of the apoA-I gene and that the factors which occupy them act synergistically to increase transcription.
The HREs on the Regulatory Elements AID and AIB of ApoA-I Are the Binding Sites of Orphan and Ligand-dependent Nuclear Receptors: Mutations in the HRE of Element AID Affect Differently the Binding of the Orphan and Ligand-dependent Nuclear Receptors-Previous studies have identified some of the factors occupying the regulatory element AID (3,4,6). In the present study we demonstrate that both regulatory elements AIB and AID of apoA-I are the binding sites of the orphan nuclear receptors HNF-4, ARP-1, EAR-2, and EAR-3, as well as of homodimers of RXR␣ and heterodimers of RXR␣ with RAR␣ or T 3 R␤. Binding of the RXR␣ heterodimers was also verified by supershift assays. We did not observe binding of RAR homodimers as suggested previously (8) as well as binding of RAR␣/T 3 R␤ heterodimers shown previously to recognize the direct repeat of myosin heavy chain (18) or monomers of homodimers of T 3 R␤ shown previously to recognize inverted or palindromic repeats (17,19).
Drastic mutations on the second repeat in the HRE of element AID which eliminated the binding of orphan and liganddependent nuclear receptors also diminished the apoA-I promoter strength and its ability to be transactivated or repressed by them. On the other hand, limited mutations in repeat 2 (AIDM3 and AIDM4) which reduced the promoter strength to 30 and 15% of control, respectively, resulted in the diminished binding of liver nuclear extracts and of HNF-4 as well as other orphan receptors and ligand-dependent nuclear receptors to the mutated sites. The findings suggest that an intact repeat 2 in the HRE of element AID is required for optimal hepatic expression of the apoA-I gene. Interestingly, certain mutations which diminished the binding of HNF-4 did not affect considerably the binding of ARP-1, and of EAR-2 and EAR-3 to the mutated probes. ARP-1, EAR-2, and EAR-3 which usually act as repressors, exhibit a wider tissue distribution than HNF-4 but are also expressed, at lower concentrations, in liver and intestinal cells as compared to HNF-4 (20). The promoter mutations which allow preferential binding of ARP-1, EAR-2, and EAR-3 but diminish binding of liver nuclear extracts and HNF-4 decreased substantially the apoA-I promoter strength, thus supporting further the role of HNF-4 as a positive regulator of the hepatic expression of the human apoA-I gene. The mutagenesis analysis also established that the binding RXR␣ homodimers have a strict requirement for intact repeats 1 and 2 in the HRE of element AID. On the other hand, binding of the RXR␣/RAR␣ heterodimers is affected mostly by alterations in repeat 2, and binding of the RXR␣/T 3 R␤ heterodimers is affected by alterations in both repeats 1 and 2 of this HRE.
Interestingly, mutations in repeat 1 (AIDM2) of element AID produced a slower migrating DNA-RXR␣/T 3 R␤ complex, compared to that formed with the wild type or other mutated AID probes. The mobility of this complex is similar to that formed with hepatic nuclear extracts using the same mutated probe. The origin of this complex remains unclear. DNA binding assays with the wild type element AID and 1:1 ratio of COS-1 extracts enriched in RXR␣ or T 3 R␤ have provided a tentative explanation for the observed higher mobility DNA-RXR␣/T 3 R␤  (Table II)  complex. We have found that supplementation of the COS-1 extracts with 9-cis-RA favors the formation of both the higher mobility RXR␣/T 3 R␤ complex and a RXR␣ homodimeric complex that displays intermediate mobility between the high and the low DNA-RXR␣/T 3 R␤ complex. In contrast, supplementation of the COS-1 extract mixed with T 3 favors the formation of the faster migrating complex (data not shown). Thus, it is possible that the fast-migrating complex may represent a DNA-RXR␣/T 3 R␤ dimer which is accessible to T 3 (21) and the slowmigrating complex may represent a higher order complex between RXR␣/T 3 R␤ and a third protein, and that the formation of this complex is enhanced by the presence of 9-cis-RA (22,23).
Nucleotides in the HRE of Element AID Which Participate in Protein-DNA Interactions with Homo-and Heterodimers of RXR␣: Spacing Requirements for Binding to AID and Ability of RXR␣/RAR␣ and RXR␣/T 3 R␤ Heterodimers to Bind to a Heptameric Core Motif-A series of KMnO 4 and dimethyl sulfate interference experiments using normal and mutated AID sequences as probes demonstrated that the RXR␣ homodimers and the RXR␣/RAR␣ and RXR␣/T 3 R␤ heterodimers share almost identical contact points with the direct repeat 1 and 2 in the HRE of element AID. These interactions involve four or five nucleotides of repeat 1, six nucleotides of repeat 2, and one nucleotide in the putative spacer region between repeat 1 and repeat 2. The putative repeat 3 is not involved in binding of the RXR␣ heterodimers and shows only a weak contact point with the RXR␣ homodimer. Overall, the RXR␣ homodimers and the RXR␣/RAR␣ heterodimers appear to bind more strongly to the intact HRE of element AID than the RXR␣/T 3 R␤. The RXR␣ homodimers have 11 strong and 2 weak DNA-protein contact points, the RXR␣/RAR␣ heterodimers have 10 strong and 3 weak DNA-protein contract points and the RXR␣/T 3 R␤ heterodimers have 6 strong and 5 weak DNA-protein contact points.
An important feature of the HREs is the number of nucleotides separating the two repeats (spacer region). It has been proposed that spacing determines the type of homo-or heterodimers of receptors that bind to an HRE (15,16,24). It has been suggested that RXR homodimers require direct repeats with a spacing of one nucleotide (DR1) for binding (25)(26)(27)(28), RXR/RAR heterodimers can bind to DR1, DR2, or DR5s (28 -31), whereas RXR/T 3 R heterodimers prefer DR4s for binding (16,32,33). Nevertheless, exceptions to this rule have been noted (15,18). Repeats 1 and 2 on element AID of apoA-I have been classified as a DR2, whereas repeats 2 and 3 can be classified as a DR1 (15). Similarly, repeats 1 and 2 of element AIB can be classified as a DR1. As shown in this study, contrary to the rule, the HREs of apoA-I are recognized both by homodimers of RXR␣ as well as heterodimers of RXR␣ with RAR␣ or T 3 R␤. Furthermore, the nucleotides which participate in the binding of these homo-and heterodimers are nearly identical. It has been shown that binding of homodimeric RXR␣ is dramatically influenced by the nature of the nucleotide preceding both AGG/TTCA motifs (34). According to the data, RXR␣ homodimers preferentially interact with direct repeats containing either an A or a G immediately upstream of the AGG/TTCA motif, whereas repeats which contain either a T or a C at the same position have greatly reduced binding (34). As shown in Fig. 1A, both repeat 1 and repeat 2 in the HRE on the noncoding strand of element AID are preceded by a G nucleotide. Similarly repeats 1 and 2 of the HRE on the noncoding strand of element AIB are preceded by G and A, respectively. In contrast, the putative repeat 3 of the HRE on the noncoding strand of element AID which does not participate appreciably in protein-DNA interactions is preceded by a C nucleotide. Interestingly, the RXR␣/RAR␣ and RXR␣/T 3 R␤ heterodimers also have the ability to interact with repeat 2 and one or two neighboring nucleotides, suggesting that their minimum requirement for binding to this HRE is a heptameric core recognition motif rather than a hexameric motif, as suggested previously (18). In this latter case, one should not exclude the possibility of extensive nonspecific but nevertheless stabilizing interactions of the heterodimers with the phosphate backbone of the mutated probe (35,36).
Modulation of ApoA-I Promoter Strength by the Combination of Homo-and Heterodimers of Orphan and Ligand-dependent Nuclear Receptors Occupying the HREs-To ascertain the effects of HNF-4 and RXR␣ homo-and heterodimers we performed cotransfection titration experiments using different combinations of nuclear receptor plasmids and apoA-I promoter CAT plasmids. The titration experiments were essential in order to establish the minimum concentration of expression vector required for optimal activation or repression of transcription. This approach enabled us to establish 9-cis-RA dependent activation of the apoA-I promoter by RXR␣ homodimers and T 3 -dependent repression by RXR␣/T 3 R␤ heterodimers. Optimal activation or repression was achieved at concentrations of expression vector in the range of 100 -250 ng. In this range of concentrations, the RXR␣/RAR␣ heterodimers and HNF-4 homodimers did not affect significantly the apoA-I promoter strength. At higher concentrations, both the HNF-4 and the RXR␣/RAR␣ heterodimers showed a trend toward transcriptional repression. The optimal apoA-I transactivation obtained, in the context of the proximal (Ϫ264/ϩ5) apoA-I promoter, by RXR␣ homodimers was 1.5-fold and the repression by RXR␣/T 3 R␤ heterodimers was 60% of control.
It has been shown previously that the HRE in element AID is a RXR response element (8,9,19). This element, when placed in front of a minimal promoter was transactivated 12-fold by RXR␣, in the presence of 9-cis-RA in CV-1 cells (19) and 8-fold by all-trans-RA in HepG2 cells (9). The minimal thymidine kinase promoter carrying the HRE of element AID was transactivated 10-fold by the RXR␣/RAR␣ heterodimers, in the presence of all-trans-RA, and 25-fold in the presence of 9-cis-RA in CV-1 cells (19). HREs from other promoters linked to the minimal thymidine kinase promoter could also confer transactivation of the heterologous promoter by RXR␣/RAR␣ heterodimers, whereas larger promoter constructs containing other HREs could be either activated or repressed by RXR␣/ RAR␣ heterodimers (19). In general, it has been proposed that the RXR/RAR complexes can function both as repressors or activators of gene expression, depending upon the nature of the direct repeats. Repression was generally observed by DR1s, whereas activation can be observed by certain DR1 as well as by DR2 and DR5 motifs (22,23,28,29). Finally the minimal thymidine kinase promoter under the control of the HRE of element AID was not transactivated by RXR/T 3 R heterodimers. These heterodimers in the presence of T 3 (37) were shown previously to bind mainly on DR4 motifs (38 -40) and to activate transcription in the presence of T 3 .
Our findings show that the extent of transactivation observed using minimal promoters carrying the HRE of element AID are much greater than the extent of transactivation observed by the RXR␣ homodimers and the RXR␣/RAR␣ heterodimers, when this HRE exists in the context of the proximal human apoA-I promoter (Ϫ264/ϩ5). Consistent with these findings, a recent study has shown that retinoids can increase 100 to 150% above control the apoA-I promoter activity and by 25 to 30% the apoA-I mRNA and protein in HepG2 cells (41). Similar to other promoter systems we suggest that in the context of the entire apoA-I promoter the nuclear hormone receptors which occupy elements AIB and AID form stereospecific DNA-protein complexes which interact directly or indirectly via the TATA box binding protein associated factors with the proteins of the basal transcription system (42,43). Other proteins bound to the proximal apoA-I promoter may help to orient properly the nuclear receptor molecules bound to the two sites and optimize their interactions as well as their interactions with the proteins of the basal transcriptional machinery. It is expected that the complex formed with the heterologous promoters which contain the HRE are different than the complexes formed with the intact apoA-I promoter and this may explain the relatively large increase in transactivation observed in the heterologous promoter constructs. We must also point out that the minimal reporter constructs containing the HREs utilized in the previous studies have very low levels of promoter activity. Thus the 8-to 15-fold transactivation achieved by the RXR␣ homodimers and RXR␣/RAR␣ heterodimers using the heterologous promoters is low compared to the activity of the intact apoA-I promoter. Previous studies have established that the proximal Ϫ264/ϩ5 apoA-I promoter region is sufficient for hepatic transcription both in vivo (5) and in vitro (3,4), therefore the transactivation levels observed in this study using the proximal apoA-I promoter by homo-and heterodimers of RXR␣ in hepatic cells, reflect more closely the physiological situation.
Alterations in the HRE of Element AID Affect the Ability of the Homo-and Heterodimers of RXR␣ to Transactivate the ApoA-I Promoter-Another strong indication that the ability of the RXR␣ homo-and heterodimers to activate or repress transcription depends on their precise protein-DNA interactions with the intact apoA-I promoter came from transactivation studies of mutated promoters. This analysis showed that the promoter which lacks repeat 3 could be transactivated or repressed by RXR␣ homo-and heterodimers in the presence of their ligand to the same extent as the wild type promoter. The findings suggest that deletion of repeat 3 did not affect the optimal DNA-protein interactions of these receptor combinations on the apoA-I promoter and their ability to be activated by their ligands. These findings are consistent with the dimethyl sulfate and KMnO 4 interference experiments which showed that the putative third repeat of the HRE of element AID participates minimally in DNA-protein interactions only with RXR␣ homodimers. On the other hand, a mutated promoter which lacks repeat 1 lost its ability to respond to the RXR␣ homo-and heterodimers in the presence or absence of their ligands. The KMnO 4 and dimethyl sulfate interference studies showed that deletion of repeat 1 resulted in association of the RXR␣ heterodimers with a motif which consists of repeat 2 and one or two neighboring nucleotides. Thus, it is possible that this generalized lack of responsiveness to the different hormonal signals may reflect altered DNA-protein and protein-protein interactions between the nuclear hormone receptors, other factors (22,23), and/or the proteins of the basal transcription system (44,45). An interesting pattern of transactivation was observed with the promoter carrying mutations in repeat 1 of element AID (AIDM2). This mutation eliminated binding of the RXR␣ homodimers but permitted binding of the RXR␣/RAR␣ and RXR␣/T 3 R␤ heterodimers as demonstrated by the DNA binding and interference assays. This promoter was transactivated 2-fold by RXR␣ and RXR␣/RAR␣ heterodimers, in a ligand-independent fashion. Since RXR␣ homodimers do not bind to this promoter, most likely the observed transactivation in both cases is mediated by RXR␣/RAR␣ heterodimers formed with the endogenous RAR␣ present in HepG2 cells or the exogenously added RAR␣, respectively. The observed ligandindependent transactivation by the heterodimers may be caused by conformational changes induced upon its binding to the heptameric motif which consists of repeat 2 and one neigh-boring nucleotide. Such conformational change may be caused by the altered DNA-protein interactions observed by in vitro KMnO 4 and dimethyl sulfate studies as well as possible altered interactions with the phosphate groups of the backbone. This may explain the fact that the promoter which lacks repeat 1 (AI⌬REP1) does not respond to heterodimers of RXR␣ in the presence or absence of their ligands. The mutated promoter in repeat 1 (AIDM2) was activated by RXR/T 3 R␤ heterodimers in the presence of 9-cis-RA but was not affected by T 3 . The lack of the T 3 -mediated repression may again reflect conformational changes of the RXR/T 3 R␤ heterodimers induced by altered DNA-protein and protein-protein interactions, which may relieve the transcriptional repression. As discussed, binding of the RXR/T 3 R␤ heterodimers to a probe carrying mutations in repeat 1 generated a slower migrating DNA-protein complex as compared to the complex formed with the wild type probe.
The present as well as previous studies demonstrate that a large number of transcription factors that belong to the steroid/ thyroid receptor superfamily bind to the HREs on the regulatory elements AIB and AID of apoA-I promoter. It is believed that the major role of RXR is to modulate a number of different hormonal signaling pathways through the formation of heterodimers. In addition, RXR has the ability to form homodimers that respond to 9-cis-RA. The dual function of RXR would always be under the tight control of intracellular signals such as the retinoids, thyroid hormone, or other unidentified ligands, that will ensure the activation of one or more specific signaling pathways. Regarding apoA-I gene regulation the present study indicates that increase in the concentration of RXR␣ and 9-cis-RA will accelerate transcription. On the other hand, at a given RXR␣ concentration, an increase in the concentration of T 3 R␤ and T 3 or an increase in ARP-1, EAR-2, and EAR-3 will repress transcription. Repression in transcription may occur either by sequestering RXR or by displacing the binding of other positive regulators (46). Finally, an increase in the concentration of RAR␣ and all-trans-RA or of HNF-4 will lead to intermediate levels of transcription. Overall, the specific types of homo-and heterodimers which can occupy the HREs of apoA-I and the availability of ligands in the cell nucleus may determine the extent of transcriptional activation of the human apoA-I promoter.