Protein-DNA Interactions at a Drug-responsive Element of the Human Apolipoprotein A-I Gene*

Previously, we demonstrated that when two human hepatoma cell lines, Hep3B and HepG2, were exposed to gemfibrozil, a hypolipidemic drug, a 2-fold induction in apolipoprotein A-I (apoA-I) mRNA levels resulted. To determine if mRNA stabilization was responsible for the changes in apoA-I mRNA levels, the half-lives for apoA-I mRNA were measured in the presence of actinomycin D with and without gemfibrozil. These experiments revealed no differences in stability. However, nuclear run-off assays indicated that the transcription rate of the apoA-I gene was increased 2-fold in gemfibrozil-treated cells. Transient transfection experiments also indicated that the induction of apoA-I mRNA level in response to gemfibrozil is mediated at the transcriptional level. We have identified two copies of the “drug-responsive ele-ment” (DRE) in the apoA-I promoter region that may be responsible for the increase in apoA-I transcriptional activity by gemfibrozil. Using gel mobility shift assays with a synthetic DRE oligonucleotide, we have demonstrated that exposure of Hep3B and HepG2 cells to gemfibrozil resulted in strong induction of a protein-DNA complex. The formation of this complex is highly se-quence-specific as indicated by the DNA competition experiments. The drug-inducible nuclear proteins bind to the DRE of the human apoA-I gene with an apparent K d of 4.1 n M . Methylation interference experiments have localized the contact sites of nuclear factors to the DRE region. Southwestern blot analyses have identified two groups of drug-inducible nuclear proteins with molecular

ApoA-I 1 is the major constituent of plasma high density lipoprotein (HDL) and participates in cholesterol ester formation as a cofactor for lecithin-cholesterol acyltransferase (1). It has been proposed that apoA-I, together with lecithin-cholesterol acyltransferase, promotes cholesterol efflux from peripheral tissues and carries the excess cholesterol to the liver for catabolism (1)(2)(3). This process is known as "reverse cholesterol transport." Epidemiological and genetic studies have indicated that levels of plasma HDL are inversely correlated with atherosclerotic risk (4 -8). Finally, evidence that apo AI expression is a major determinant of atherosclerosis comes from the work of Rubin et al. (9), which indicated that transgenic mice expressing high amounts of human apoA-I were significantly protected from the development of fatty streak lesions after exposure to high fat diets.
There are a number of drugs widely used for reducing the progression of coronary heart diseases by altering lipoprotein metabolism. One of the most extensively studied is gemfibrozil. The Helsinki Heart Study (10) demonstrated that the elevation of HDL cholesterol resulting from use of this drug had a more protective effect in reducing atherosclerotic risk than the lowering of LDL cholesterol. Recently, we have demonstrated that exposure of two human hepatoma cell lines, HepG2 and Hep3B, to gemfibrozil resulted in a 2-fold induction of apoA-I mRNA (11). The cis-and trans-acting elements involved in apoA-I gene transcription have been extensively studied (12)(13)(14). Based on transient transfection studies, it was concluded that the DNA region located between nucleotides Ϫ256 and Ϫ41 upstream from the transcription start site (ϩ1) of the human apoA-I gene contains regulatory elements that are necessary and sufficient for maximal expression in HepG2 cells (12). This DNA region acts as a powerful liver-specific enhancer (13). The mechanism by which drugs such as gemfibrozil regulate apoA-I mRNA levels has not been determined. Evidence summarized below suggests that expression of the apoA-I gene may be influenced directly by drug-induced transcription factors.
It is well known that the cytochrome P-450 system is responsible for the metabolism of a variety of drugs (15). The induction of the CYP1A1 gene is mediated by the interaction of the aryl hydrocarbon receptor (AhR) together with its ligand to several DNA recognition sites within a dioxin-responsive enhancer upstream of the CYP1A1 gene (16 -18). While examining the structure of the human apoA-I gene promoter a sequence match with the 5Ј-flanking region of the CYP1A1 gene was detected (19). The match involves a decanucleotide motif, 5Ј-(G/C)(N(T/G)(A/G)GCTGGG-3Ј, corresponding to a putative drug response element (DRE) first identified by Fujii-Kuriyama and co-workers (19). It was suggested initially that this element may be recognized by the AhR. This has now been shown not to be the case. At present, factors that bind to the element have not been identified, and its role in mediating induction in response to inducers of the dioxin type remains unclear. Two copies of the element, spaced 10 nucleotides apart in an inverted repeat, are present in the human apoA-I promoter between nucleotides Ϫ77 and Ϫ45, relative to the transcription start site of the gene. One of the elements is a 100% match with the DRE consensus, and the other is a 90% match.
Recently, it has been demonstrated that a polymorphism of the human apoA-I promoter region at 75 bp (a guanine to adenine substitution) upstream from the transcription start site of the gene is associated with elevated levels of HDL cholesterol and apoA-I (20,21). This genotype was also greatly affected by smoking (22). Here, we describe experiments showing that gemfibrozil can induce a group of nuclear proteins in Hep3B and HepG2 cells and that these factors bind specifically and with high affinity to the DRE region of human apoA-I gene. Transient transfection assays also demonstrate that the DRE is necessary for gemfibrozil-mediated induction of human apoA-I gene expression.

EXPERIMENTAL PROCEDURES
Materials-Complementary pairs of oligonucleotides were synthesized on a Biosearch model 8600 DNA synthesizer at Queen's University, Department of Biochemistry, Core Facility for Protein/DNA Chemistry. Sequences of the synthetic oligomers are indicated in Table I.
Cell Culture and Drug Treatments-Human hepatoma cell lines, Hep3B and HepG2, were obtained from the American Type Culture Collection (Rockville, Maryland). Cells were grown in T75 flasks containing 20 ml of Eagle's minimal essential medium (MEM) supplemented with 10% fetal bovine serum (FBS) as described previously (23). Freshly confluent monolayers were washed twice with MEM and then incubated with fresh medium for 24 h in the presence of gemfibrozil, dissolved in ethanol, to give a final concentration of 40 g/ml. Control cells were incubated with an equal volume (20 l) of ethanol. In some experiments, gemfibrozil and cycloheximide were added to the cells to give final concentrations of 40 and 10 g/ml, respectively. Where noted, Hep3B or HepG2 cells were also treated with or without gemfibrozil and in the presence of actinomycin D, ranging from 0.625 to 10.0 g/ml, for various time periods as described in the figure legends. Cell viability was routinely monitored by trypan blue exclusion and lactate dehydrogenase leakage as described previously (23). In all experiments the number of dead cells never exceeded 5% of the total number of cells.
RNA Isolation and Northern Blot Analyses-Total cellular RNA was isolated by extraction with guanidine-HCl, as described (11). For Northern blotting, 20 g of total RNA were denatured by treatment with glyoxal, subjected to electrophoresis on a 1.5% agarose gel, and transferred to Zeta-probe GT membrane (Bio-Rad) as described previously (11). Blots were prehybridized and hybridized with nick-translated apoA-I and ␤-actin cDNA probes as described (18). ␤-actin was used as an internal control, since its cognate mRNA levels are not affected by gemfibrozil (11). All results were normalized using densitometric scans of Northern blots probed with radiolabeled ␤-actin to correct for loading variations.
Specific gene transcripts were quantified by hybridization to filterbound DNA. The recombinant pBS apoA-I used in this study was prepared by ligating the 0.6-kb human apoA-I insert from pkT apoA-I into the PstI site of the pBS vector. Preparation of Zeta-probe GT nylon filters was as outlined (24). Hybridization, washing, and elution of bound, labeled RNA were carried out as described previously (26). Nonspecific hybridization to each filter was determined by carrying out transcription with labeled pGEM 3Z control vector. The relative rates of transcription were calculated in terms of parts per million, after subtraction of control values. Total incorporation into RNA per assay ranged from 1 ϫ 10 7 dpm to 5 ϫ 10 7 dpm, all of which was used for each hybridization selection. Throughout the analysis, 3 H sense strand RNAs produced by transcription with Sp6 polymerase were used to normalize for differences in hybridization efficiency. The efficiency ranged from 30 to 35%.
Preparation of Nuclear Extracts-Hep3B and HepG2 cells were grown as described above, and 3-6 ϫ 10 8 cells were used to prepare nuclear extracts. Nuclear extracts were prepared by a modification of the method of Dignam et al. (27). The modification was as follows: cells were suspended in 2 volumes of buffer (10 mM HEPES (pH 8.0), 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM DTT, and 1 mM phenylmethylsulfonyl fluoride), and incubated on ice for 15 min. The swelled cells were then passed through a 23 1 ⁄2-gauge needle five times followed by a 26 1 ⁄2-gauge needle for another five strokes instead of using a Dounce homogenizer. Nuclear extracts were dialyzed against binding buffer (25 mM HEPES at pH 8.0, 12.5 mM MgCl 2 , 20% glycerol, and 50 mM KCl) and stored as aliquots at Ϫ80°C. Nuclear protein content was determined by the method of Lowry et al. (28).
Gel Mobility Shift Assays-For mobility shift assays, nuclear extracts were incubated with 100 g of poly (dI-dC) in binding buffer containing 5 mM DTT and 5 M ZnCl 2 on ice for 30 min. Then 2 fmol (10,000 cpm) 5Ј-end-labeled synthetic oligonucleotide corresponding to the drug response element or other consensus binding elements were added to the reaction mixtures and incubated on ice for another 30 min. Free DNA and protein-DNA complexes were resolved on a 6% polyacrylamide gel in a buffer containing 10 mM Tris, 5 mM sodium acetate and 2.5 mM EDTA, pH 7.2. The gel was preelectrophoresed at 80 V for 1 h. The samples were then loaded onto the gel and electrophoresed at 80 V for 4 -6 h at 4°C. Gels were vacuum-dried and exposed to Kodak x-ray film at Ϫ80°C with a DuPont Hi-plus intensifying screen. Competition assays were performed by adding the unlabeled competitor DNA 15 min prior to the addition of labeled DNA as indicated in the figure legends. The band representing the protein-DNA complex was identified by autoradiography, excised from the gel, and quantitated by Cerenkov counting.
Southwestern Blot Analyses-Equal volumes of nuclear extract and sample buffer (5% SDS, 5 mM Tris/HCl, pH 6.8, 200 mM DTT, 20% glycerol, and 0.5% pyronin Y) were mixed and subjected to electrophoresis on a 10% SDS-polyacrylamide gel at room temperature. Protein samples were electrotransferred to a Millipore Immobilon P membrane in 25 mM Tris-HCl, 192 mM glycine, 20% (v/v) methanol, and 0.002% SDS at room temperature for 4 h. Hybridization of 5Ј-end-labeled oligonucleotides to nuclear proteins immobilized on the membrane was performed according to the method of Singh and co-workers (29).
Construction of Plasmids-A 491-bp DNA fragment of the human apoA-I promoter between nucleotides Ϫ491 and ϩ1 was generated by polymerase chain reaction (PCR) amplification. Two synthetic oligonucleotide primers were made complementary to the apoA-I gene between nucleotides Ϫ491 and Ϫ460 and between nucleotides Ϫ25 and ϩ1, respectively. PCR amplification was carried out using these two primers together with HepG2 high molecular weight genomic DNA prepared as described (30,31). The PCR protocol was as follows: 30 cycles of 95°C for 45 s, 52°C for 45 s, and 72°C for 30 s. A 491-bp DNA fragment resulted, and its sequence was confirmed by DNA sequencing. The fragment was inserted into the XhoI site of the pGL2-Basic vector (Promega), upstream of the luciferase gene. This plasmid is hereafter called pGL2 (apoA-I Ϫ491)luc. Plasmid pGL2 (apoA-I Ϫ250)luc was constructed by releasing a DNA fragment (Ϫ491 to Ϫ251 of the apoA-I promoter) from pGL2 (apoA-I Ϫ491) using SmaI, and the vector was gel-purified and religated.
Plasmid pGL2 (apoA-I Ϫ250 mutant DRE)luc was prepared by a PCR-based protocol for extensive mutagenesis of a target DNA sequence as described by Morrison and Desrosiers (32). Briefly, we selected seven G and three A residues in the DRE region for mutagenesis, since these residues might participate in the interaction with nuclear proteins as determined by methylation interference. To generate this mutated DRE plasmid, two sets of primers were used. One set of primers, designated GL and Luc, hybridized to specific regions of the plasmid pGL2. The other set of primers, which contained the mutagenic DRE residues, were named primers A and B. The primers GL (5Ј-TGTATCTTATGGTACTGTAACTG-3Ј) and B (5Ј-GACCAGTGAGCAG-CAACAGGGCCGGGTATTGCTTCTCCTCCTCCACTACCAGACC-3Ј) were complementary to the noncoding strand of DNA, while primers Luc (5Ј-GGCGTCTTCCATTTTACC-3Ј) and A (5Ј-CTGGTAGTGGAG-GAGGAGAAGCAATACCCGGCCCTGTTGCTGCTCACTG-3Ј) were complementary to the coding strand of DNA. The plasmid pGL2 (apoA-I Ϫ250)luc was used as the PCR template. Amplification was carried out in an Eppendorf MicroCycler E for 30 cycles (consisting of 1 min of denaturing at 94°C, 1 min of annealing at 54°C, and 1 min of extension at 72°C) followed by a final extension for 10 min at 72°C. The first round of PCR using GL ϩ A primer pairs in one tube and B ϩ Luc primer pairs in another tube produced fragments of 241 and 153 bp, respectively. These products were purified from agarose gels using the Sephaglas Band Prep Kit (Pharmacia Biotech Inc.) combined in equimolar amounts and used as templates for the second round of PCR amplification in the presence of GL and Luc primers to generate a 342-bp product, which contained the DRE mutation. The DNA was then digested with KpnI and HindIII, and the resulting DNA fragment was purified from an agarose gel. The purified DNA fragment was cloned into the KpnI and HindIII sites of the pGL2-basic (Promega) and used to transform competent E. coli cells. DNA was prepared from individual clones by the alkaline lysis miniprep procedure, and the entire DNA insert including the area of mutagenesis was sequenced using the Sequenase version 2.0 system (U.S. Biochemical Corp.).
The vector designated pGL2 TK/luc (kindly provided by Dr. M. Petkovich, Department of Biochemistry, Queen's University) was derived from the pGL2-Basic vector in which the thymidine kinase promoter was inserted upstream of the luciferase gene. Plasmid pGL2 (apoA-I-DRE)TK/luc was constructed by cloning a synthetic oligo-DRE (between nucleotide positions Ϫ78 and Ϫ43 from the transcriptional start site of the apoA-I gene) into the SmaI site of the pGL2 TK/luc. The vector pGL2 (apoA-I-mutant DRE)TK/luc was constructed by cloning a synthetic mutated oligo-DRE 5Ј-GGCCGTTTACTTTTTACTAGCCTAAA-GATTTAGACC-3Ј into the SmaI site of pGL2 TK/luc. The inserts in all of the above plasmids have the same 5Ј 3 3Ј orientation as found in the natural human apoA-I promoter.
Transfections and Luciferase Assay-The human hepatoma cell line, Hep3B, was maintained as monolayers on 100-mm plates in MEM supplemented with 10% FBS. Transient DNA transfections were performed by the calcium phosphate precipitation procedure described by Gorman et al. (33). Ten g of total DNA was used per plate, including 4 g of reporter DNA and 6 g of carrier pGL2 Basic Vector. After glycerol shock, the cells were washed twice with phosphate-buffered saline and once with MEM and cultured in MEM ϩ 10% FBS in the absence or presence of gemfibrozil (40 g/ml) for 24 h with changes of fresh medium.
Transfected Hep3B cells were harvested by washing three times in phosphate-buffered saline and assayed for luciferase activity as described (34). In all transfections, 5 g of an internal control plasmid (pSG⌬Lac Z) containing the Escherichia coli lacZ gene under the control of the SV40 early promoter and enhancer was included in order to correct for differences in transfection and harvesting efficiency. Transfected cells were harvested as described above, and ␤-galactosidase activities in the cell lysates were determined (34). The pGL2 promoter vector (Promega), which contains an SV40 promoter, is used as a reference for both transfection and luciferase assays. Parallel transfection experiments and luciferase assays were also performed using HepG2 cells by the procedures described above. All luciferase activities are reported as mean Ϯ S.E. Significance of group differences was determined by Student's t test, using two-tailed p values.
Methylation Interference-Single-stranded synthetic oligonucleotides corresponding to the upper or lower strands of the DRE from the human apoA-I gene were radiolabeled at their 5Ј-ends by [␥-32 P] ATP and T4 polynucleotide kinase. Double-stranded oligonucleotides were formed by heating the labeled oligonucleotide with an equimolar amount of the corresponding unlabeled complementary strand at 90°C for 5 min. in 0.05 M Tris-HCl (pH 8.0), 0.01 M MgCl 2 , followed by a slow cooling to room temperature over a period of 2 h. The double-stranded oligonucleotides were partially methylated with dimethyl sulfate as described (35). The methylation experiment was carried out according to the method of Baldwin and Sharp (35). The recovered DNA sample was cleaved with 1 M piperidine and subjected to polyacrylamide gel electrophoresis on a 20% acrylamide DNA sequencing gel. An equivalent amount of radioactivity for each sample was loaded in each lane. The gel was exposed to Kodak XAR-5 film at Ϫ70°C. The labeled oligonucleotides were subjected to Maxam and Gilbert sequencing (36), and the G ϩ A reactions serve as markers.

RESULTS
Effect of Actinomycin D on the Induction of Apolipoprotein AI mRNA Levels by Gemfibrozil-Previously, we demonstrated that exposure of HepG2 and Hep3B cells to gemfibro-zil resulted in a 2-fold induction in apoA-I mRNA levels (11).
To determine whether inhibition of RNA synthesis affects the induction of apoA-I mRNA levels, Hep3B cells were treated both in the absence and presence of increasing amounts of actinomycin D for 30 min prior to the addition of gemfibrozil. Actinomycin D at concentrations greater than or equal to 0.625 g/ml were able to block the increase in apoA-I mRNA levels that was observed when the cells were treated with gemfibrozil alone. Under the same regimen of drug treatment, similar results were also obtained using HepG2 cells (data not shown).
To examine if mRNA stabilization is responsible for the observed induction of apoA-I mRNA levels by gemfibrozil, the half-lives for apoA-I mRNA were determined in the presence of actinomycin D (1 g/ml) with and without gemfibrozil. Hep3B cells were pretreated with gemfibrozil (40 g/ml) or ethanol (drug vehicle) for 24 h. Cells were then cultured for additional periods ranging from 0 to 8 h in the presence of actinomycin D. As shown in Fig. 1, the level of apoA-I mRNA was approximately 2-fold higher in gemfibrozil-treated cells compared with control cells. After the addition of actinomycin D the levels of apoA-I mRNA decreased at the same rate in the absence or presence of gemfibrozil, while levels of actin mRNA remained relatively constant (Fig. 1, top panel). The apparent half-lives obtained for apoA-I mRNAs following the addition of actinomycin D in the absence or presence of gemfibrozil were approxi- mately 5.6 Ϯ 0.5 h and 5.7 Ϯ 0.4 h, respectively. These data were analyzed by multivariate statistics and F test using the SAS (Statistical Analysis Systems) computer program. The results indicated that there was no significant difference (p Ͼ 0.05) between the two half-lives.
Analysis of the Stimulation of Transcription by Gemfibrozil Treatment-The transcription rate of the apoA-I gene was measured using isolated nuclei from Hep3B cells cultured in the absence or presence of gemfibrozil treatment. The effect of gemfibrozil on apoA-I transcription rates at various times is shown in Fig. 2. The rate of transcription of the apoA-I gene increased approximately 2-fold between 18 and 24 h after stimulation.

Interactions of Drug-inducible Nuclear Factors Isolated from Hepatoma Cells with the DREs of the Human apoA-I Gene-
The possibility that induction of apoA-I transcription by gemfibrozil involved protein-DNA interactions at the DREs located between nucleotides Ϫ77 and Ϫ45, was examined by mobility shift experiments using a 5Ј-end-labeled double-stranded oligonucleotide, oligo-DRE, corresponding to the apoA-I promoter region between nucleotides Ϫ79 and Ϫ44. Extracts prepared from control Hep3B or HepG2 cells do contain a factor that will bind to the oligonucleotide (Fig. 3). However, treatment with gemfibrozil resulted in strong induction of a retardation complex. The induction of this complex was blocked by inhibiting protein synthesis with cycloheximide (data not shown). The question arose as to whether or not the induction of this complex by gemfibrozil was due to a general increase in transcriptional activation during drug treatment. In an attempt to answer this question, we examined the levels of protein binding to oligonucleotides corresponding to Sp1 and AP1 binding sites before and after gemfibrozil treatment. No change in binding activity to either site was observed with nuclear extracts prepared from untreated or gemfibrozil-treated cells (Fig. 3).
Competition Gel Mobility Shift Studies-In order to determine the specificity of binding of the drug-inducible nuclear proteins, competition gel retardation assays were performed with several different double-stranded oligomers (Table I). In nuclear extract from gemfibrozil-treated Hep3B cells, binding to the labeled apoA-I-DRE probe was efficiently blocked by competition with a 50 -100-fold molar excess of unlabeled apoA-I-DRE (Fig. 4). The DNA fragments ((P-450(I) and P-450(II)) of the CYP1A1 gene (19), which contain a DNA sequence very similar to apoA-I-DRE, also competed efficiently with the probe. Furthermore, the AP1 and AP2 consensus binding elements (37) were able to compete efficiently with the probe for binding to the nuclear factor(s). However, no competition was observed with a 200-fold molar excess of the GC-rich, Sp1 consensus sequence (37) as well as the core recognition motif for the AhR (16,17). Mutation of 10 nucleotides within the DRE motif of the human apoA-I gene also showed no competition with the apoA-I-DRE probe (Fig. 4). Results of these competition studies are summarized in Table I.
Measurement of Relative Binding Affinity of the Drug-inducible Nuclear Proteins for Oligo-DRE-The relative DNA binding affinity of oligo-DRE was measured by saturation binding assays using gel mobility shift analysis in which constant amounts of nuclear extracts were titrated with increasing amounts of labeled oligo-DRE (Fig. 5). The data were plotted by the method of Scatchard (38). The drug-inducible nuclear proteins bound to the DRE of the human apoA-I gene with an apparent K d of 4.1 nM.
Methylation Interference Analyses-To more precisely localize the contact sites of nuclear factors to the DRE region, methylation interference studies were performed. Protein binding was impaired by premethylation of guanines in any of seven positions: Ϫ52, Ϫ60, Ϫ69, and Ϫ72 of the coding strand, and Ϫ51, Ϫ54, and Ϫ71 of the noncoding strand (Fig. 6). Experiments that detected methylation of adenines in the minor groove of the DNA revealed adenines at positions Ϫ53, Ϫ61, and Ϫ64 of the coding strand were also involved in protein binding (Fig. 6). Similar methylation interference patterns were observed with either nuclear extract isolated from untreated cells or gemfibrozil-treated cells.
Southwestern Blot Analyses-To further characterize the nu-

FIG. 2. Time course of transcription rates of apoA-I genes in Hep3B cells cultured in the absence or presence of gemfibrozil.
Transcription rates were determined by nuclear run-off assays using nuclei isolated from Hep3B cells cultured in the absence or presence of gemfibrozil (40 g/ml) as described under "Experimental Procedures." Nuclei (2-3 ϫ 10 7 ) were isolated at 0, 2, 4, 10, 18, and 24 h. Hybridization of the newly synthesized [ 32 P]RNA to plasmids containing an apoA-I insert were carried out in triplicate. Nonspecific hybridization to each filter was determined by performing transcription with labeled pGEM3Z control vector. Filters were washed extensively, and bound radioactivity was measured by liquid scintillation counting. Relative transcription rates were calculated as parts per million (ppm)/filter ϭ ((counts per min (cpm)/filter) Ϫ cpm background)/amount of [ 32 P] RNA used in hybridization (input count). These numbers were then normalized for hybridization efficiency as measured by binding of known amounts of 3 H riboprobes added to the hybridization mixture. This was then corrected for the size to the specific gene: ppm/gene ϭ ((ppm/filter)/ percentage of hybridization) ϫ (gene size/cDNA insert size). Sizes of the apoA-I gene and apoA-I cDNA fragments were 2 and 0.6 kilobases, respectively. Results are mean Ϯ S.E. expressed in percentage of the O-h time point of three independent experiments.
FIG. 3. Gel mobility shift analysis. Synthetic double-stranded oligonucleotides corresponding to the DNA sequence between nucleotides Ϫ79 and Ϫ44 upstream from the transcription start site (ϩ1) of the human apoA-I gene, designated as oligo-apoA-I-DRE, were used as probes to study protein-DNA interactions. In addition, synthetic double-stranded Sp1 and AP1 consensus sequences were also utilized as probes to determine their activities before and after gemfibrozil treatment. 32 P-Labeled oligo-apoA-I-DRE, oligo-Sp1, and oligo-AP1 (1 ϫ 10 4 cpm) were incubated with no protein (Blank) or with nuclear extracts from Hep3B cells untreated (Control) or treated with gemfibrozil (40 g/ml) as described under "Experimental Procedures." Similar results were also observed when nuclear extracts isolated from HepG2 cells were used. clear factors that bind to the oligo-DRE, Southwestern analyses were carried out. As indicated by a solid horizontal bar beside lane 2 in Fig. 7, the labeled oligo-DRE bound to a protein band of approximately 28 kDa in size. This band appeared to be inducible by gemfibrozil treatment, since no such band was detectable in the nuclear extracts of control Hep3B cells. The 32 P-labeled oligo-DRE also bound to several proteins that were present in both gemfibrozil-treated and untreated samples. Two of these proteins have molecular masses of approximately 68 and 42 kDa, indicated by an open square and a star, respectively (Fig. 7). These two protein bands gave similar signals on the x-ray autoradiograph. Therefore, they could be used as internal controls for protein loading. In addition, three other protein bands with molecular masses of 30 kDa (open circle), 15 kDa (open triangle), and 14.5 kDa (solid square) were also induced by gemfibrozil treatment.
To study the specificity of these drug-inducible protein bands toward the oligo-DRE, Southwestern blot analyses were performed using 32 P-labeled oligo-P-450(I), oligo-P-450(II), and oligo-AhR separately. It was observed that two protein bands with masses of approximately 30 and 15 kDa could bind to oligo-P-450(I) and oligo-P-450(II) (Fig. 6, lanes 3-6). Furthermore, these two bands were highly inducible upon gemfibrozil treatment. It was also noteworthy that no nuclear proteins from control and gemfibrozil-treated cells interacted with oligo-AhR (Fig. 7, lanes 7 and 8). The data demonstrated that the drug-inducible bands bound specifically to the DRE consensus sequence.
Functional Analysis of ApoA-I-DRE-To analyze the role of apoA-I-DRE in the regulation of human apoA-I gene expression in response to gemfibrozil, we carried out transient transfec-tion experiments using a series of pGL2-derived luciferase reporter plasmids. The construction of these plasmids is described in detail under "Experimental Procedures." Plasmids were transfected into Hep3B cells in the absence or presence of gemfibrozil (40 g/ml), and luciferase activity was measured. As shown in Fig. 8A, both pGL2 (apoA-I Ϫ491)luc and pGL2 (apoA-I Ϫ250)luc constructs showed a significant 2-fold induction of luciferase activity in the presence of gemfibrozil. By contrast, pGL2 (apoA-I Ϫ250 mutant DRE) in which the 10 nucleotides (Fig. 6) involved in possible contact with nuclear proteins were substituted (G 3 T, C 3 A, T 3 G, A 3 C) demonstrated no increase in luciferase activity upon gemfibrozil treatment.
In order to further demonstrate that the DRE was able to confer responsiveness to gemfibrozil, transient transfection experiments were carried out using a series of pGL2 TK/luc constructs. The control vector pGL2TK/luc demonstrated no change in luciferase activity in response to gemfibrozil treatment (Fig. 8). To determine if binding of nuclear factors to the DRE is necessary for gemfibrozil-mediated induction of apoA-I gene expression, one copy of synthetic DRE oligonucleotide was inserted upstream of the thymidine kinase promoter, and the construct was tested for inducibility by gemfibrozil. After transient transfection of these constructs into Hep3B cells, pGL2 (apoA-I DRE)TK/luc showed luciferase activity approximately 4-fold higher than that observed by the control vector pGL2 TK/luc. This suggests that the DRE can act as an enhancer that may increase expression of the reporter gene. Furthermore, the luciferase activity of pGL2 (apoA-I-DRE)TK/luc was significantly increased by an additional 2-fold in the presence of gemfibrozil. However, a plasmid containing one copy of the  Table I. Gel mobility shift assays were performed by using constant amounts of nuclear proteins (5 g) isolated from gemfibrozil-treated Hep3B cells and labeled oligo-apoA-I-DRE (25 ng) as well as various amounts of unlabeled oligonucleotides as indicated in the figure. The concentrations of the competitors are in molar excess, relative to the concentration of the labeled DNA. The bands representing specific protein-DNA complexes were identified by autoradiography, excised from the gel, and quantitated by Cerenkov counting. 32 P-Labeled oligo-apoA-I-DRE was also incubated with with nuclear extracts from control Hep3B cells. Similar results were also observed when nuclear extracts isolated from HepG2 cells were used (data not shown). Sp1 consensus binding element 5Ј-CTGCGGGGCGGGGCAGA-3Ј 0 Oligo-AhR Xenobiotic response element at which the AhR is known to bind 5Ј-AGTTGCGTGAGAAGA-3Ј 0 ** * * ** Oligo-M1 Mutated apoA-I-DRE 5Ј-GGCCGGGTATTGCTTCTCCTCCTCCACTACCAGACC 0 **** a The specificity of the drug-inducible nuclear proteins toward the labeled oligo-apoA-I-DRE was determined by comparative competition experiments using various unlabeled synthetic oligomers. The competition efficiency is defined as the amount of competitor required to reduce the protein-DNA complex formation by 50% and arbitrarily set at 100 for unlabeled apoA-I-DRE. Results are the mean Ϯ S.E. of three separate experiments. Mutated nucleotides (G3 T, C3 A, T3 G, and A3 C) in the mutated apoA-I-DRE are indicated by an asterisk (*). The consensus sequences for various elements are underlined. mutated apoA-I-DRE oligomer, pGL2 (apoA-I-mutant DRE)TK/luc, lost the enhancer ability and responsiveness to gemfibrozil treatment (Fig. 8). DISCUSSION Previously, we have demonstrated that exposure of two cultured human hepatoma cell lines, Hep3B and HepG2, to gemfibrozil resulted in a 2-fold induction in apoA-I mRNA level (11). Based on the results of the present report, the increase in apoA-I mRNA level upon gemfibrozil treatment is due primarily to increased rates of transcription of this gene. First, in actinomycin D-treated cells, the half-life of apoA-I mRNAs was not significantly affected by gemfibrozil treatment (Fig. 1). Second, both in vitro nuclear run-off transcription assays (Fig.  2) and transient transfection experiments (Fig. 8) indicated a significant 2-fold increase in apoA-I gene expression in the presence of gemfibrozil.
In the presence of actinomycin D, the half-life of apoA-I mRNA was 5.6 Ϯ 0.5 h in control Hep3B cells, in agreement with our previous studies, which measured apoA-I mRNA halflife by pulse-chase methods (25). By contrast, a recent report by Vandenbrouck and co-workers (39) has shown that the half-life of apoA-I mRNA is about 15 h in control HepG2 cells. Consequently, we also have performed kinetic studies in the presence of actinomycin D to provide a measurement of apoA-I mRNA half-life in HepG2 cells. Our results indicated that in HepG2 cells, apoA-I mRNA degraded with a half-life of 12.0 Ϯ 1.5 h. 2 Thus, the data suggest that the rate of degradation of apoA-I mRNA is at least 2-fold faster in control Hep3B cells compared with control HepG2 cells. The difference between the half-lives in these two cell lines might partially explain the 2-fold higher steady state level of apoA-I mRNA in control HepG2 cells compared with control Hep3B cells (11).
Recently, many studies have been focussed on the use of fibric acid derivatives to treat diet-resistant hyperlipidemic patients (40 -44). Fibrates are used widely to lower plasma cholesterol and triglyceride levels as a treatment for coronary heart disease. However, these compounds also cause peroxisome proliferation and, in some cases, hepatocarcinogenesis in rodents (45,46). It has been proposed that fibrates and other peroxisome proliferators activate a member of the steroid hormone receptor superfamily, the peroxisome proliferator-activated receptor (PPAR) (47)(48)(49)(50)(51). The first PPAR was isolated from a mouse liver cDNA library by Isseman and Green (47). It appears that one of the physiological roles of PPAR is to regulate fatty acid homeostasis. PPAR binds to a specific response element (PPRE) located upstream of the target genes (48,(52)(53)(54)(55), which consists of an almost perfect direct repeat of the sequence TGACCT separated by one base pair. PPAR binds to these PPREs through heterodimerization with the 9-cis-retinoic acid receptor, retinoid X receptor ␣ (56 -58). Functional PPREs have been identified in several of the genes encoding enzymes involved in the peroxisomal ␤-oxidation pathway, such as acyl-CoA oxidase and the trifunctional enzyme (53)(54)(55)59) as well as in the cytochrome P-450 4A6 gene (60,61) in the 3-hydroxy-3-methylglutaryl-CoA synthase gene (62), in the medium chain acyl-CoA dehydrogenase gene (63), in the apolipoprotein A-I gene (64), and in the AP2 gene (51). In view of these observations, the possibility exists that regulation of human apoA-I gene expression by gemfibrozil might involve the interaction of PPAR at the PPRE of the gene. However End-labeled, partially methylated synthetic oligo-DRE (Ϫ79 to Ϫ44, coding or noncoding strand) was incubated with nuclear extracts isolated from gemfibrozil-treated Hep3B cells (Gemfibrozil) or untreated cells (Control). Complexed and free fractions were separated on a nondenaturing low ionic strength mobility shift assay gel. Bound and free DNA fractions were eluted from the 5% acrylamide gel prior to piperidine cleavage at methylated sites. Products were then separated on a 20% sequencing gel. Methylated regions that interfered with protein binding are indicated, and binding is summarized at the bottom. Lane G ϩ A indicates Maxam and Gilbert G ϩ A reaction. *, guanines methylated at the N 7 position (major groove); ٙ, adenines methylated at the N 3 position (minor groove). Similar results were observed in three independent experiments.
there is evidence to suggest this may not be the case. First, the consensus sequence of PPRE is different from the DRE described in the present study. Second, the hepatic peroxisome proliferation seen in rodents receiving gemfibrozil does not occur in humans, as evaluated by light and electron microscopy of liver biopsy samples from patients on long term gemfibrozil therapy (65). Third, PPAR has a molecular mass of 52.4 kDa, while the gemfibrozil-induced proteins have sizes of approximately 30 and 15 kDa as observed by Southwestern blot analysis (Fig. 7). Finally, the in vitro translated mouse PPAR␣ synthesized from pSG4-mPPAR␣ expression vector (kindly provided by Dr. S. Green, Alderley Park, Macclesfield, United Kingdom) using the rabbit reticulocytes in vitro transcription system (Promega) shows specific DNA binding activity to a PPRE but not to the DRE of the human apoA-I promoter. 3 Studies from both humans and rodents indicate that gemfibrozil causes an increase in the levels of plasma HDL and/or apoA-I (10, 40, 66 -69). Saku et al. (66) observed a 36% increase in HDL cholesterol and a 29% rise in apoA-I levels when their patients with familial hypertriglyceridemia were treated with gemfibrozil. This was accompanied by an increase in the synthetic rate of apoA-I by 27% with no change in the fractional clearance rate. The results from our studies support, in part, Saku's findings and demonstrate that gemfibrozil may act by selectively increasing expression of the apoA-I gene in human liver. Elucidation of the mechanism by which gemfibrozil modulates apoA-I gene expression requires the identification of promoter elements and transcription factors responsible for mediating the biological response to these agents. We have identified two copies of the DRE in the apoA-I promoter that may be responsible for the increase in apoA-I transcriptional activity by gemfibrozil. These two decanucleotide elements, spaced 10 nucleotides apart in an inverted repeat, are located between nucleotides Ϫ77 and Ϫ45, relative to the transcriptional start site of the apoA-I gene. We have demonstrated that exposure of Hep3B or HepG2 cells to gemfibrozil results in a strong induction of a protein-DNA complex (Fig. 3). Similar results have been observed with the rat hepatoma cell line H4IIE (obtained from ATCC), indicating that the binding specificity of the factors is conserved across species (data not shown). We observe no difference in the binding specificities of the nucleotides that are involved in the protein-DNA interactions using either control or gemfibrozil-treated nuclear extracts as determined by methylation interference analysis. There exists at least one possible explanation for this observation. If the DRE in the apoA-I promoter is mediating the gemfibrozil action, the activity rather than the abundance or binding affinity of the factor must be regulated by gemfibrozil. Recently, Leighton and Kemper (70) have demonstrated that phenobarbital treatment increases the expression of CYP2C1 and CYP2C2 in rat liver. However, indistinguishable DNase I digestion patterns were obtained using liver nuclear extracts from either control or phenobarbital-treated rats (71). Furthermore, another study has indicated that the DNase I hypersensitivity pattern at the rat liver xenobiotic response elements do not alter when aryl hydrocarbon receptor translocation and DNA bindings are induced by 2,3,7,8-tetrachlorodibenzo-p-dioxin treatment (72).
The competition mobility shift experiments presented in Fig.  4 shows that the nuclear factor(s) that bind(s) apoA-I-DRE is highly sequence-specific. A 25-fold molar excess of a doublestranded oligonucleotide homologous to apoA-I-DRE competed strongly for binding with the 32 P-labeled apoA-I-DRE probe, while an equivalent molar excess of mutated apoA-I-DRE resulted in virtually no competition with the apoA-I-DRE probe. Furthermore, the DRE consensus sequence of CYP1A1 (P-450(I) and P-450(II)) competed efficiently for factor binding to the apoA-I-DRE. However, no competition was observed with 200-fold molar excess of the Sp1 consensus sequence or the xenobiotic consensus sequence for the AhR. It is worth noting that both the AP1 and AP2 consensus sequences compete efficiently for protein binding to the apoA-I-DRE. One possible explanation for this observation is that nuclear factor(s) that recognize(s) the apoA-I-DRE can also bind to the AP1 and AP2 consensus sequences. Alternatively, it is possible that the protein-DNA complex observed in the mobility shift assay consisted of multiple proteins of which AP1 and AP2 binding proteins are two of the potential candidates. Results from the present studies suggest that the former may be the more likely possibility. First, our methylation interference experiments indicated that the contact sites of the nuclear factors with the DRE region did not resemble those of the AP1 (73) or AP2 (74, 3 S.-P. Tam and S. Green, unpublished results.

FIG. 7. Southwestern blot analyses.
Nuclear proteins were isolated from Hep3B cells cultured in the absence (C) or presence of gemfibrozil (G), fractionated by 10% SDS-polyacrylamide gel electrophoresis, transferred onto an Immobilon P membrane, and probed with 32 P-labeled oligo-DRE (lanes 1 and 2), oligo-P-450(I) (lanes 3 and 4), oligo-P-450(II) (lanes 5 and 6), and oligo-AhR (lanes 7 and 8), respectively. Molecular weight markers are indicated by solid lines beside lane 1. Similar results were observed in three separate studies. 75) DNA recognition sites. Second, Williams and Tjian (74) have demonstrated that HepG2 cells express neither AP2 mRNA nor protein. Third, the two groups of gemfibrozil-induced nuclear proteins detected by Southwestern blot analyses have molecular masses of 15 and 30 kDa, whereas AP1 and AP2 proteins have molecular masses between 40 and 75 kDa (75). Finally, we have performed supershift experiments to determine whether c-Jun and c-Fos protein interact with the DRE binding site. For supershift assays, the nuclear extracts isolated from gemfibrozil-treated cells were preincubated with affinity-purified polyclonal antibodies raised against c-Jun and c-Fos (Oncogene Science) at 4°C for either 2 h or 18 h before the standard gel mobility shift studies. Both sets of experiments showed no effect on the binding activity of DRE oligonucleotide to the gemfibrozil-treated nuclear extract when either anti-c-Jun or anti-c-Fos antibodies were used (data not shown).
The specific protein binding to oligo-DRE was observed with nuclear extracts from gemfibrozil-treated cells but not from cytosolic extracts (data not shown). Consequently, it was not clear whether inducible binding was due to a drug-induced translocation of a preexisting factor from the cytosol to the nucleus or was attributable to its de novo synthesis. The results of inhibiting protein synthesis with cycloheximide suggest the latter is the case. At the moment, we are uncertain whether this involves synthesis of the factors themselves or of ancillary proteins or enzymes required for binding activity.
In order to establish a functional role for the DRE in apoA-I gene transcription in response to gemfibrozil treatment, we have carried out transient transfection experiments using pGL2-derived luciferase reporter plasmids. Constructs that contain nucleotides Ϫ491 to ϩ1 and Ϫ250 to ϩ1 upstream from the transcription start site (ϩ1) of the human apoA-I gene show a significant 2-fold increase in luciferase activity in the presence of gemfibrozil (Fig. 8A). The involvement of the DRE in gemfibrozil-mediated induction of apoA-I gene expression was established by using the pGL2 (apoA-I Ϫ250 mutant DRE) in which the 10 nucleotides involved in possible contact with nuclear proteins were substituted. Results from these studies indicated that this plasmid had lost the enhancing ability in response to gemfibrozil. In addition, we also determined whether or not the DRE could function as a transcriptional enhancer using a heterologous promoter. These experiments were carried out by inserting one copy of synthetic DRE oligonucleotide upstream of the thymidine kinase promoter, and the construct was tested for inducibility by gemfibrozil. The luciferase activity of pGL2 (apoA-I-DRE)TK/luc was increased significantly by 2-fold in the presence of the drug. However, a plasmid containing one copy of mutated apoA-I-DRE oligomer, pGL2 (apoA-I-mutant DRE)TK/luc, was unresponsive to gemfibrozil treatment. The data indicated that the DRE located in the apoA-I promoter region can confer drug responsiveness on a heterologous promoter.
In conclusion, we have identified a group of gemfibrozilinducible nuclear proteins that bind specifically to the DRE region of the human apoA-I gene. This protein-DNA interaction may provide some insights into a novel drug-response mechanism by which drug-inducible trans-acting nuclear factors modulate apoA-I gene expression. Further experiments will be required to further understand the relationship between the structure and the function of these induced nuclear proteins in the regulation of human apoA-I transcription.