Identification of a novel cis-acting element participating in maximal induction of the human low density lipoprotein receptor gene transcription in response to low cellular cholesterol levels.

In this paper, we present both in vivo and in vitro evidence for the presence of a novel cis-acting regulatory element that is required for maximal induction of the human low density lipoprotein (LDL) receptor gene following depletion of cellular sterols in HepG2 cells. First, in vivo dimethyl sulfate footprinting of the human LDL receptor promoter before and after transcriptional induction in HepG2 cells revealed protection from −145 to −126, 5′-GAGCTTCACGGGTTAAAAAG-3′ (referred to as FP1 site). Second, transient transfections of HepG2 cells with promoter luciferase reporter constructs containing the FP1 site resulted in significant enhancement (approximately 375%) of reporter gene expression in response to low levels of sterols compared with parallel plasmid without the FP1 site. In addition, this response was markedly attenuated on nucleotide substitutions within the FP1 site. Third, by electrophoretic mobility shift assays, the FP1 sequence was found to bind protein(s) from HepG2 nuclear extracts in a sequence-specific manner. In vitro binding of the FP1 mutants paralleled the results obtained for their in vivo transcription. On the basis of competition profiles, the FP1-binding factor is different from the known transcription factors binding to the AT-rich CArG and GArC motifs. Furthermore, the FP1-binding protein is not specific to HepG2 cells because nuclear factor(s) with the same specificity was observed in nuclear extracts of non-hepatic HeLa cells. We conclude that transcriptional induction of the LDL receptor gene in response to sterol depletion is mediated, in part, by an highly conserved novel cis-acting element through the binding of specific nuclear protein(s).

In this paper, we present both in vivo and in vitro evidence for the presence of a novel cis-acting regulatory element that is required for maximal induction of the human low density lipoprotein (LDL) receptor gene following depletion of cellular sterols in HepG2 cells. First, in vivo dimethyl sulfate footprinting of the human LDL receptor promoter before and after transcriptional induction in HepG2 cells revealed protection from ؊145 to ؊126, 5-GAGCTTCACGGGTTAAAAAG-3 (referred to as FP1 site). Second, transient transfections of HepG2 cells with promoter luciferase reporter constructs containing the FP1 site resulted in significant enhancement (approximately 375%) of reporter gene expression in response to low levels of sterols compared with parallel plasmid without the FP1 site. In addition, this response was markedly attenuated on nucleotide substitutions within the FP1 site. Third, by electrophoretic mobility shift assays, the FP1 sequence was found to bind protein(s) from HepG2 nuclear extracts in a sequence-specific manner. In vitro binding of the FP1 mutants paralleled the results obtained for their in vivo transcription. On the basis of competition profiles, the FP1-binding factor is different from the known transcription factors binding to the AT-rich CArG and GArC motifs. Furthermore, the FP1-binding protein is not specific to HepG2 cells because nuclear factor(s) with the same specificity was observed in nuclear extracts of non-hepatic HeLa cells. We conclude that transcriptional induction of the LDL receptor gene in response to sterol depletion is mediated, in part, by an highly conserved novel cisacting element through the binding of specific nuclear protein(s).
Low density lipoprotein (LDL) 1 receptor is a key component of the mechanism by which animal cells maintain balanced cholesterol homeostasis. The transcription of the LDL receptor gene is maintained under tight feedback regulation by cellular levels of sterols (1,2). In cells with excess sterols, transcription is repressed; in contrast, transcription is accelerated in cells requiring cholesterol. In vivo this feedback regulation is most important in the liver. Ingestion of cholesterol in the diet decreases hepatic levels of the LDL receptor mRNA, and consequent decline in the hepatic LDL receptor causes LDL to accumulate in the circulation (3)(4)(5). The target for sterol regulation lies within a stretch of 10 base pairs that has been designated sterol regulatory element-1 (SRE-1) (6,7). In addition to SRE-1, there are two Sp1-like sequences in the human LDL receptor gene promoter which bind to purified Sp1 (4,6,8). SRE-1 bind to SRE-1 binding proteins (SREBPs), which undergo proteolytic cleavage at the C-terminal membrane-associated domain (125 kDa) and are converted to functionally active nuclear forms (68 kDa) (9 -11). The nuclear form of SREBP is transcriptionally active because it contains an acidic transcriptional activation domain and a basic helix-loop-helix leucine zipper region that mediates protein dimerization and DNA binding. All essential nucleotides of SRE-1 and Sp1 sites are conserved in evolution (4).
In addition to regulation of the LDL receptor gene by cellular cholesterol levels, transcription is modulated by a variety of mitogenic and nonmitogenic signals in multiple cell types. Insulin and platelet-derived growth factor can stimulate LDL receptor gene transcription in quiescent mesenchymal cells (12)(13)(14). Serum factors stimulate LDL receptor gene transcription in HepG2 cells, and mitogenic stimulation increases LDL receptor gene transcription in lymphocytes (15,16). cAMP, protein kinase C agonists, calcium ionophores, and arachidonic acid metabolites have been shown to affect LDL receptor expression in HepG2 cells (17)(18)(19). Cytokines have been shown to modulate the LDL receptor pathway activity in endothelial cells, arterial smooth muscle cells, and HepG2 cells (20 -22). Many of these stimuli increase the LDL receptor transcript, and some have been shown to enhance LDL receptor gene transcription. Induction of LDL receptor gene transcription by platelet-derived growth factor and insulin have been ascribed to the participation of Sp1 and SRE-1 sequences, respectively (23,24). The mechanisms by which LDL receptor gene transcription respond to a variety of other humoral signals are not clearly understood.
In the studies presented here, we have identified a novel regulatory element (FP1; Fig. 1) in the promoter of the gene for the human LDL receptor gene that is required for maximal induction of the receptor gene following depletion of sterols using both in vivo and in vitro approaches. Maximal induction of the LDL receptor gene may result from synergistic interactions of the nuclear factor(s) binding to this regulatory element with the other known transcription factors Sp1 and SREBP involved in controlling LDL receptor expression in the animal cell.

MATERIALS AND METHODS
Enzymes and Biochemicals-Standard molecular biology techniques were used (25). Vent DNA polymerase and restriction enzymes were purchased from New England Biolabs. 25-Hydroxycholesterol and cholesterol were purchased from Sigma and Steraloids, Inc., respectively. Dimethyl sulfate, piperidine, and hydrazine were obtained from Aldrich Chemical Co. [␥-32 P]ATP was purchased from ICN. Antibody to serum response factor (SRF) was purchased from Santa Cruz. DNA ligase, TRIzol, Superscript II reverse transcriptase, and all tissue culture supplies were purchased from Life Technologies. Fetal bovine lipoprotein-deficient serum (LPDS) was obtained from PerImmune, Inc. Zeta probe blotting membrane was purchased from Bio-Rad. pGL2-Basic and pSV-␤-galactosidase (pSV-␤-Gal) vectors were purchased from Promega. The pGL2-Basic vactor lacks eukaryotic promoter and enhancer sequences and contain the luciferase gene coding firefly luciferase which allows sensitive and rapid quantitation of reporter activity. pSV-␤-Gal vector was used as a positive control for monitoring transfection efficiencies of HepG2 cells. In this plasmid, SV40 early promoter and enhancer drive transcription of the lacZ gene which encodes the ␤-galactosidase enzyme. Dual-Light chemiluminescent reporter gene assay system for the combined detection of luciferase and ␤-galactosidase was purchased from TROPIX, Inc. Lovastatin was a gift from Merck, Sharp and Dohme.
Cell Culture-HepG2 (gift from Dr. Pitor Zimniak) cell line was routinely grown in Dulbecco's modified Eagle's medium with 10% fetal calf serum (Life Technologies). HeLa cell line (gift from Dr. Sam Goldstein) was grown under identical conditions in 8% newborn calf serum. Both the cell lines were grown at 37°C in 5% CO 2 , 95% air atmosphere. For suppressed and induced conditions, the cells were switched to medium containing 10% LPDS in the presence of either sterols (5 g/ml 25-hydroxycholesterol plus 10 g/ml cholesterol; suppressed) or 30 M lovastatin (induced). After incubation for 16 -20 h, the cells were either treated with 0.1% dimethyl sulfate for isolation of in vivo partially methylated genomic DNA (26,27) or used for RNA isolation (5,28,29).
Ligation-mediated Polymerase Chain Reaction-4 g of partially in vivo methylated genomic DNA from induced and suppressed HepG2 cells were subjected to the ligation-mediated polymerase chain reaction (LMPCR) method of in vivo footprinting essentially as described by Wold and colleagues (30,31). The specific primers used to screen the human LDL receptor gene promoter are described below. LMPCRs were performed with slight modifications as follows: (i) first round denaturation was for 5 min at 95°C followed by extension for 30 min and 10 min at 60 and 76°C, respectively; (ii) amplification was carried out using 95°C for denaturation, 65°C for annealing, and 76°C for extension; and (iii) labeling reaction was done with primer 3 at 95°C (3.5 min), 69°C (2 min), and 76°C (10 min) for 5 cycles; for primer 6, which was used to label the top strand, 95, 60, and 65°C steps were repeated for 4 cycles. Following amplification and ethanol precipitation, DNA fragments were resuspended in 10 l of formamide dye and 1-2 l was applied onto a 8% sequencing gel. Gels were dried and exposed to Kodak X-AR film with an intensifying screen at Ϫ70°C for 15-36 h. Each of the experiments described were performed at least three times from independently methylated DNA samples, with consistent results.
Reporter Gene Constructs-HindIII-linked oligonucleotides were used to amplify regions of interests in the human LDL receptor gene 5Ј-flanking region, and the amplified fragments were subcloned in the sense orientation into the HindIII site of the pGL2-Basic vector. The sequence of the entire insert was verified by double-strand sequencing. All luciferase constructs have a common 3Ј-end located at ϩ35 with respect to the start site of transcription and contain the TATA sequences and transcription start sites of the human LDL receptor gene. Plasmid DNA was isolated using Qiagen columns for transfection experiments (Qiagen, Inc.).
Cell Culture and Transient Transfection of HepG2 Cells-For transfection experiments, HepG2 cells were seeded at 1 ϫ 10 6 /60-mm dish 1 day in advance. Transfections were performed in triplicate with 6.0 g of DNA for each construct and 2.0 g of pSV-␤-Gal vector by the calcium phosphate method (25). After 16 h, the cells were washed with phosphate-buffered saline and refed with Dulbecco's modified Eagle's medium containing 10% fetal calf serum. Approximately 1 day later, transfected cells were switched to media supplemented with either 10% LPDS (Ϯlovastatin) or with 10% LPDS plus 25-hydroxycholesterol/ cholesterol, and the cells were incubated for an additional 16 -20 h. Finally, dishes were washed with phosphate-buffered saline and lysed with 150 l of luciferase lysis buffer (100 mM potassium phosphate, pH 7.8, 0.2% Triton X-100, and 0.5 mM dithiothreitol). Luciferase-generated light signals were detected with an automated luminometer (Model 2010, Analytical Luminescence Laboratory) using a 2-s delay and 5-s measurement, and ␤-galactosidase initiated light signals were measured similarly after the addition of accelerator-II. The luciferase activity was normalized to ␤-galactosidase activity for each plasmid. Data are expressed as the means Ϯ S.D.
Preparation of Nuclear Extracts and Gel-mobility Shift Assays-HepG2 nuclear extracts were prepared from 20 confluent 150-mm dishes, and HeLa cells were grown in suspension until they reached a concentration of 5 ϫ 10 5 /ml. Nuclei were prepared from both cell lines by the method described by Dignam et al. (32) except that buffer A was supplemented with 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 1 g each of pepstatin A and leupeptin per ml (Sigma), and the nuclear extracts were prepared from nuclei as described by Gorski et al. (33). Nuclear extracts were quick frozen and stored in liquid nitrogen in aliquots. Protein concentrations were determined using a modified Bradford assay with bovine ␥-globulin as standard (Bio-Rad) and were typically 3-7 mg/ml. Oligonucleotides were annealed and gel-purified prior to end labeling with T4 polynucleotide kinase in the presence of [␥-32 P]ATP. Nuclear extract (5 g) was incubated in a final volume of 15 l of 10 mM Tris, pH 7.5, 50 mM NaCl, 5% glycerol, 1 mM dithiothreitol, 1 mM EDTA, 1 mM spermidine, 1 g of poly(dI-dC) with each probe (20,000 cpm; 0.5 to 1 ng) for 20 min at room temperature. In competition analysis, extracts were incubated with the indicated molar excess of unlabeled oligonucleotides for 5 min prior to the addition of the labeled oligonucleotides. The DNA-protein complexes were resolved on 4.5% nondenaturing polyacrylamide gels (29:1 cross-link) in 0.25 ϫ TBE (1 ϫ TBE is 89 mM Tris, pH 8, 89 mM boric acid, and 2 mM EDTA). The gels were dried and analyzed by autoradiography using Kodak X-AR film.

RESULTS
In Vivo Footprinting of the Human LDL Receptor Gene Promoter-The optimal conditions for the suppressed and induced states of LDL receptor gene expression in HepG2 cells were determined by growing them in LPDS media containing different concentrations of sterols and lovastatin, and then by quantitating the LDL receptor transcript by using the reverse transcription-polymerase chain reaction method (5). We found that maximal suppression of the LDL receptor gene was obtained in LPDS media with 25-hydroxycholesterol concentration above 1 g/ml, and maximal induction was obtained with 30 M lovastatin (Fig. 2). To determine the occupancy status of different genetic elements in vivo, primers (primers 1-3) for footprint analysis were first designed to target nucleotides Ϫ1 to Ϫ117 of the human LDL receptor promoter. G ladders from the suppressed and induced HepG2 are shown in Fig. 3A and a comparison of reactivities of G residues revealed Gs which are protected, enhanced, or unaffected with these treatments. SRE-1 region showed an hypersensitive G at Ϫ59, an observation consistent with earlier in vivo footprinting studies (34,35). However, G residues of the proximal Sp1 site (Ϫ40 to Ϫ49) and the distal Sp1 site (Ϫ92 to Ϫ100) exhibited slightly more protection from dimethyl sulfate attack in the induced than suppressed HepG2 cells as compared to previous reports (34,35). In addition, weak protection was observed at nucleotide Ϫ106.
To rule out the possibility that this selective pattern is not specific for this particular combination of oligonucleotides, a different oligonucleotide (primer 4) closer to SRE-1 was used for labeling the PCR products from an independent experiment. As expected, an identical pattern of footprinting over SRE-1 and the adjacent Sp1 site was observed (data not shown).
Investigation of the protein-DNA interactions in the region further upstream of these repeats using primers 5-7 had revealed significant protections of nucleotides from Ϫ126 to Ϫ145 (designated Footprint 1 (FP1)), and a slight protection of nucleotides from Ϫ175 to Ϫ187 (designated Footprint 2 (FP2)) ( Fig. 3B). The protected Gs within the FP1 site included gua-  9). Total cellular RNA was subjected to RT-PCR analysis using human LDL receptor and ␤-actin specific primers. Primer LDLR 1, 5Ј-GGCTGGGTGATGTTGTGGAA-3Ј (ϩ2083 to ϩ2102) and primer LDLR 2, 5Ј-GGCCGCCTCTACTGGGTTGA-3Ј (ϩ1715 to ϩ1734) were used for amplification, and primer LDLR 3, 5Ј-GAAGCCATTTTCAGT-GCCAA-3Ј was used for probing human LDL receptor-specific amplified product. To verify that equivalent quantities of cDNA from both conditions were used in the PCR, amplification of ␤-actin cDNA was used as a control. Actin 1, 5Ј-TACAATGAGCTGCGTGT-3Ј (ϩ312 to ϩ329) and actin 2, 5Ј-AAGGCCAACCGCGAGAAGAT-3Ј (ϩ423 to ϩ406) were used for amplification, and actin 3, 5Ј-AAGGCCAACCGCGAGAAGAT-3Ј (ϩ377 to ϩ395) was used for probing the amplified product. to Ϫ117 in suppressed (S) and induced (I) HepG2 cells as determined by using primers 1 to 3. On the right-hand side is a second gel loaded with 25% of the samples shown on the left and is exposed for equal time period. This gel is shown to indicate hypersensitivity at Ϫ59 within SRE-1 (indicated by an arrow) and slightly higher protection of both the Sp1 sites in induced than suppressed HepG2 cells. B, in vivo protein-DNA interactions in the region further upstream of distal Sp1 site (Ϫ120 to Ϫ213) of human LDL receptor promoter under both sterols conditions. C, summary of protected Gs (indicated by open circles above nucleotide sequence) within FPs 1 and 2.
nines at Ϫ126, Ϫ134, Ϫ135, Ϫ136, Ϫ143, and Ϫ145, whereas Gs at nucleotides Ϫ175, Ϫ177, Ϫ178, Ϫ183, and Ϫ187 were protected in the FP2 (Fig. 3C). Interestingly, the in vivo protected FP1 site (Ϫ126 to Ϫ145) coincided with the in vitro DNase I protected region (Ϫ127 to Ϫ159) observed earlier for the human LDL receptor gene promoter (36). The observed in vivo footprints were very reproducible, even when slightly different modifications or cleavage conditions were used.
Footprint I Is Required for Maximal Induction of the LDL Receptor Gene in Response to Low Levels of Sterols-As a first step in determining the significance of the in vivo protected sites, varying lengths of the human LDL receptor gene promoter were ligated to the 5Ј end of the promoterless luciferase reporter gene in plasmid pGL2-Basic. These constructs were transiently transfected into cultured HepG2 cells, the transfected cells were grown under induced and suppressed conditions for LDL receptor expression, and luciferase activity was determined as described under "Materials and Methods." The luciferase activity is expressed relative to the activity of the construct labeled A (Fig. 4A). The negative control, promoterless plasmid pGL2-Basic, consistently exhibited negligible levels of luciferase expression (Fig. 4B). The co-transfection of pSV-␤-Gal control vector allowed correction of luciferase values for transfection efficiency. Transfection results demonstrated that presence of Ϫ273 to Ϫ127 region, that includes both the FP1 and FP2 sites in reporter construct containing the Ϫ126 to ϩ35 sequence of human LDL receptor promoter, resulted in at least 375% increase in luciferase expression in the absence of sterols compared to a parallel plasmid without this additional region; levels of both constructs were almost suppressed to a similar extent by sterols (compare plasmid A versus B, Fig. 4). During the course of this work, Streicher et al. (24) have recently reported that constructs containing a fragment that spans the Ϫ222 to ϩ88 region of the human LDL receptor gene

FIG. 4. Expression of human LDL receptor promoter-luciferase fusion genes in HepG2 cells. Panel A, at the
top is an extended map of the human LDL receptor promoter with the locations of FPs 1 and 2, the SRE-1 and Sp1 sites, and the TATA and transcription start sites. Several plasmid constructs with the luciferase gene under the control of DNA fragments spanning different portions of the Ϫ273 to ϩ35 region of the human LDL receptor promoter were tested for luciferase activity by transient transfections into HepG2 cells in the absence of sterols (see "Materials and Methods"). Each construct (6 g) was co-transfected with plasmid pSV-␤-Gal (2 g) to correct for variations in DNA uptake by the cells. Relative luciferase activity values represent luciferase/␤-galactosidase enzymatic activity ratios relative to that of construct A, and are averages of at least five to seven independent experiments. In the case of plasmids H and I, luciferase activity shown in brackets are relative to the parallel plasmid F; letter M in the box represents mutated site and specific nucleotide substitutions are indicated in lower case letters below these constructs. Panel B, schematic representation of the sterolmediated regulation of LDL receptor promoter-luciferase constructs in transfected HepG2 cells. Each bar represents mean of three to five independent transfection experiments performed in duplicate; thin lines indicate S.D. HepG2 cells were transiently transfected with the indicated plasmid together with pSV-␤-Gal. After incubation for 20 h in the absence of sterols (Ϯlovastatin) or presence of 10 g/ml cholesterol plus 2 g/ml 25-hydroxycholesterol, the cells were harvested for duplicate measurements of luciferase and ␤-galactosidase activities. Corrected luciferase activities were calculated as described in Panel A. Fold regulation is the ratio of corrected luciferase activity in the absence of sterols (plus lovastatin) divided by the corrected luciferase activity in the presence of sterols and are indicated on the top of bars for each constructs. 5Ј-flanking region showed 2-3-fold higher reporter gene level expression compared to plasmids containing either Ϫ105 to ϩ88 (both Sp1 sites and SRE-1) or Ϫ69 to ϩ88 (proximal Sp1 site and SRE-1) region. These results are in agreement with our results presented here. To identify the cis-element in more detail, a series of 5Ј deletion constructs without the FP2 site were prepared and transfected into HepG2 cells (Fig. 4A). As shown in Fig. 3A, deletion of sequence between Ϫ273 to Ϫ173, which includes the FP2 site, consistently resulted in an approximately 2-fold elevation in the expression of the reporter gene in the absence of sterols relative to plasmid A (compare plasmid A versus C, Panel A). Further 5Ј deletions between nucleotides Ϫ173 and Ϫ139 (plasmids D, E, F, and G) did not affect the enhancement observed for plasmid C in response to depletion of sterols. The reason for this elevation has not been investigated, and the attention was focused in the present study on understanding the role of the FP1 site in LDL receptor gene expression. To further delineate the regulatory element present within Ϫ145 to Ϫ126 of promoter region, reporter plasmids were constructed with the FP1 sequence specifically changed at nucleotides shown in Fig. 4 (plasmids H and I), and tested in the transient transfection assays. Introduction of transversion mutations of all the nucleotides between Ϫ134 and Ϫ146 (plasmid H) abolished the FP1-dependent enhancement in the reporter gene level. Interestingly, specific changes of four nucleotides (guanines at Ϫ135/Ϫ136 and cytosines at Ϫ145/Ϫ146 to thymines and adenines, respectively; plasmid I) also drastically reduced the ability of the FP1 region to enhance transcription. The observed reporter levels are not statistically different from the construct without the FP1 site (compare plasmid I versus B, Fig. 4A). Fold regulation by sterols of different constructs were also tested in the HepG2 cell line and the results are summarized in Fig. 4B. Mostly, constructs showing higher induced levels showed a slight and insignificant elevation in the basal levels (suppressed) of the reporter gene in the presence of sterols (Fig. 4B). The above study shows that the FP1 site in conjunction with the SRE-1 and both Sp1 sites showed maximal regulation by sterols in HepG2 cells. Finally, the presence of both the FP1 and FP2 sites alone is not sufficient to elicit transcription of the reporter gene (plasmid J, Fig. 4A) because no transcription was observed with a promoter fragment containing these sequences without the other known regulatory elements of the human LDL receptor gene. In short, the above results indicate that other sequences within the Ϫ273 to Ϫ126 region contribute very little and that nearly all of the 3-4-fold elevation in reporter gene expression is mediated by sequences within the Ϫ139 to Ϫ126 region of the human LDL receptor promoter in HepG2 cells.
In Vitro Binding of Nuclear Factor(s) to the FP1 Sequence-To begin to understand the nature of the protein interactions that occur over the FP1 site, EMSA were performed with nuclear extracts from induced HepG2 cells. Oligonucleotide pair A (nucleotides Ϫ120 to Ϫ146) that includes all nucleotides within the in vivo dimethyl sulfate protected region were used in the EMSA (see Fig. 5 for sequences of oligonucleotide pairs used). As seen in Fig. 6A, nuclear extracts from HepG2 cells produced three slowly migrating shift bands, labeled I, II,  -10) with radiolabeled wild-type oligonucleotide pair A encoding the FP1 site in the presence of poly(dI-dC) as nonspecific competitor. To test for the sequence specificity of the protein-DNA complex, competition assays were performed. Binding of nuclear extract to the 32 P-labeled oligonucleotide pair were assayed in the presence of indicated fold molar amounts of unlabeled competitors. Complex I is the specific protein-DNA complex formed with oligonucleotide pair A, and complexes II and III represent nonspecific binding of nuclear proteins. The position of free probe is indicated (FP). Sequences of the competing oligonucleotides are shown in Fig. 5. Lane 1, labeled oligonucleotide pair A without nuclear extract; lane 2, labeled oligonucleotide pair A with nuclear extract; lanes 3-6 competition analysis with indicated fold molar excess of unlabeled oligonucleotide pair A; lanes 7 and 8, competition with unlabeled oligonucleotide pair B carrying the same mutation present in plasmid I (Fig. 4A); lanes 9 and 10, competition with unlabeled oligonucleotide pair C containing nucleotide substitutions shown for plasmid H (Fig. 4A). Panel B, competition assays using oligonucleotide pair A, C, D, SeRE, or GArC at the indicated molar excess. Lane 1, labeled oligonucleotide pair A without nuclear extract; lane 2, labeled oligonucleotide pair A with nuclear extract; lane 3, excess unlabeled oligonucleotide pair A; lanes 4 and 5, excess unlabeled oligonucleotide pair D; lanes 6 and 7, excess unlabeled oligonucleotide pair E; lane 8, excess unlabeled oligonucleotide pair CArG; lane 9, excess unlabeled oligonucleotide pair GArC. and III. Addition of increasing amounts of unlabeled oligonucleotide pair A competed for binding and strongly inhibited the formation of band I. In contrast, bands II and III are not competed and thus appear to represent nonspecific binding of proteins in nuclear extracts (Fig. 6A, lanes 3-6). The mutant oligonucleotide pairs B and C containing mutations of plasmids I and H, respectively (lanes 7-10), did not compete with oligonucleotide pair A, in agreement with the transfection data shown in Fig. 4. The lack of competition of oligonucleotide pair D or E for binding to the FP1 site suggests that sequences flanking the AT-rich core are required for the binding to the FP1 sequence (Fig. 6B). In reciprocal experiments in which these sequences were tested as labeled probes, complex I did not form with the mutated or deleted oligonucleotide pairs (data not shown). Furthermore, as shown in Fig. 6B, oligonucleotide pairs corresponding to the consensus sequences of two AT-rich motifs, the CArG (c-fos; Ref. 37) and GArC (␣-myosin; Ref. 38), did not compete for complex I even with a 100-fold molar excess. Thus, the above results indicate that HepG2 cells express a nuclear factor that specifically binds to the FP1 sequence and the FP1 binding activity is distinct from the earlier reported nuclear factors binding to the CArG and GArC motifs.
To compare the relative amounts of the FP1-binding factor in nuclear extracts of induced and suppressed HepG2 cells, EMSA were performed. To optimize comparisons between bands, different concentrations of nuclear extracts prepared from HepG2 cells were used at the same time in the binding reactions. No significant differences in the formation of the specific complex I were observed (data not shown).
To test whether the FP1-binding factor is specific to the hepatic cell line, HepG2, EMSA were also performed with non-hepatic HeLa nuclear extracts. When amounts of the HeLa nuclear extract similar to those used for the HepG2 cells were tested, a specific retarded band of similar electrophoretic mobility as complex I was observed (Fig. 7). Furthermore, binding specificity of HeLa nuclear protein was similar to that of HepG2 cells because unlabeled oligonucleotide pair A effectively competed in the binding assays while mutant oligonucleotide pair B or C showed no significant competitions. Similar results were also obtained for other mutant oligonucleotide pairs tested with HepG2 nuclear extract (data not shown).

DISCUSSION
This paper describes the identification of a novel regulatory element (FP1) in the human LDL receptor promoter that is necessary for maximal enhancement of the receptor gene following the depletion of sterols. The functional significance of this regulatory element is supported by the following observations: (i) sequences within the FP1 site showed specific protection from dimethyl sulfate attack in induced HepG2 cells, as detected by the LMPCR in vivo footprinting; (ii) the presence of the FP1 site resulted in an approximately 375% increase in reporter gene expression in HepG2 cells in response to sterol starvation compared to a construct without this sequence; (iii) mutagenesis of the FP1 site showed that specific nucleotide substitutions within this region abolished the enhanced expression of the reporter gene after sterol depletion without affecting the basal levels in presence of sterols; (iv) consistent with the in vivo results, the FP1 sequence showed specific binding to nuclear protein(s) from HepG2 cell extracts; (v) in vitro binding of oligonucleotides with specific nucleotide substitutions paralleled completely the results obtained for their in vivo transcription; (vi) alignment and comparison of the LDL receptor gene 5Ј-flanking sequences in different species (human, rat, and hamster) showed remarkable conservation of the position and sequence of the FP1 site (Fig. 8) (39 -41). Taken together, the above experiments define a 20-base pair FP1 site as a functionally relevant enhancer sequence that mediates its effect through a nuclear protein in response to the low cellular levels of sterols.
A search of the NIH transcription factor data bank with the FP1 site showed similarities between the AT-rich region within the FP1 sequence and the AT-rich regulatory elements called the CArG box (sequence motif CC(A/T-rich) 6 GG) and the GArC box (preference sequence G(A/T-rich) 6 -7 CTC) (37,38). The CArG box motif forms the core of the serum response element (SeRE), a DNA element that is required for the transient transcriptional response of many nonmuscle-specific immediateearly response genes, such as c-fos and ␤-actin, upon serum or growth factor stimulation. These sequences have been shown to bind SRF (serum response factor), a ubiquitous protein that activates the transient transcriptional responses of a number of genes (42). On the other hand, the exact role of the GArC motif is not very clear but has been found to be important in regulating the ␣-myosin heavy chain gene expression (38). Despite the fact that the AT-rich core is present in the FP1 site, it is not likely that either the CArG or GArC-binding factor play any role in LDL receptor expression for the following reasons. First, the FP1 sequence exhibits an obvious difference from the CArG motif (SeRE) such as the inversion of the 2 C and 2 G residues surrounding the AT-rich core. These residues in the SeRE are crucial for the binding of SRF. This difference in sequence together with the lack of competition of the c-fos SeRE in the EMSA argues against the FP1-binding protein being SRF. Furthermore, lack of inhibition of FP1 binding (Fig.  6B) and no supershifting of the specific retarded complex I in EMSA using anti-SRF argues against this possibility (data not shown). Second, the lack of competition for the FP1 sequence binding in EMSA using the GArC oligonucleotide pairs as competitor demonstrates that the FP1 sequence bind to a factor different from the earlier reported GArC factor. In addition, GArC oligonucleotide pair resulted in the formation of two specific retarded shift bands as opposed to one specific complex observed with the FP1 sequence in our study (38). Third, correlation between the in vitro binding and the in vivo transfection results of mutant constructs are consistent with the argument that a nuclear factor other than the CArG and GArCbinding factors is responsible for the formation of the retarded complex I. The role of sequences within the FP1 site including the AT-rich region is not clear yet. Further mutational analysis of the AT-rich core and surrounding sequences will determine which residues within the FP1 site are required for the binding and whether the AT-rich core actually interacts with the transacting factor(s) or simply provides an environment which facilitates the binding through other contacts in the adjacent region.
A similar search of the NIH transcription factor data bank for the FP2 site revealed homology of part of the FP2 sequence (nucleotides Ϫ179 to Ϫ174) to the binding site, TGGCGA of a known transcription factor F-ACT-1 (43,44). In fact, there is an additional sequence, AGGCGA, in which there is an A instead of T, present in the other strand of the FP2 site. F-ACT-1 is also known to bind to the SeRE (CArG motif) and binding has been shown to be mutually exclusive with that of SRF. F-ACT-1 appears to be a negative regulator of ␣-actin gene expression (44). In cardiac muscle cells, SRF and F-ACT-1 are the transacting factors involved in the basal transcription of the skeleton ␣-actin gene and for its induction by transforming growth factor ␤-1 and other trophic signals (45). The role of the FP2 site has not been the center of investigation in the present study, however, deletions including the FP2 site and upstream region were found to slightly increase reporter gene expression, suggesting that the FP2 site either alone or in conjunction with other regulatory elements play a role in LDL receptor expression.
In summary, the studies presented here demonstrate that transcriptional control of the LDL receptor is mediated, in part, by an highly conserved novel cis-acting element present in the 5Ј-flanking region of human LDL receptor gene promoter. The close proximity of the FP1 site to the downstream Sp1/SRE-1 sites suggests that transcriptional modulation of the LDL receptor gene may result from cooperative interaction of FP1binding protein with other transcription factors involved in the regulation of LDL receptor gene expression. In fact, the presence of an additional regulatory element may help explain transcriptional induction of the human LDL receptor gene promoter by a variety of agents. An assessment of the potential interactions between the FP1/Sp1/SRE-1 elements will ultimately require the purification or molecular cloning of the trans-acting factor(s) that interacts with the FP1 sequence. Modification of LDL receptor transcription by pharmacologic intervention, with the intent of managing hypercholesterol-emia, will require definition of each of the factors critical to this process.