INTRODUCTION

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, 10, 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, 21, 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

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. [-³²P]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.

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% CO₂, 95% air atmosphere. For suppressed and induced conditions, the cells were switched to medium containing 10% LPDS in the presence of either sterols (5 µg/ml 25-hydroxycholesterol plus 10 µg/ml cholesterol; suppressed) or 30 µM 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).

4 µg 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 µl of formamide dye and 1-2 µl was applied onto a 8% sequencing gel. Gels were dried and exposed to Kodak X-AR film with an intensifying screen at 70 °C for 15-36 h. Each of the experiments described were performed at least three times from independently methylated DNA samples, with consistent results.

Oligonucleotides were synthesized on an Applied Biosystems DNA synthesizer model 380B or obtained from commercial sources and gel purified prior to use. Primers 1-3 were utilized for scanning the region spanning the SRE-1 and both Sp1 sites (see Fig. 1 for relative positions). Primer 1 was 5-TGAGGGGGCGTCAGCTCTTCACCGGAG-3 (236 to 210; extension step). Primer 2 was 5-AAGGACTGGAGTGGGAATCAGAGCTTCA-3 (165 to 138; PCR amplification step). Primer 3 was 5-AGTGGGAATCAGAGCTTCACGGGTTA-3 (156 to 131; labeling step). Primer 4 was 5-GCCGATGTCACATCGGCCGTTC-3 (126 to 104; labeling step). Primers 5-7 were utilized to scan the top strand of the promoter region farther upstream of the distal Sp1 site. Primer 5 was 5-GCGAGGAGCAAGGCGACGGTCCAGCG-3 (+123 to +98; extension step). Primer 6 was 5-ATGCTCGCAGCCTCTGCCAGGCAGT-3 (+76 to +52; PCR amplification step). Primer 7 was TGTCTTCACCTCACTGCAAG-3 (73 to 91; labeling step). Primers 1, 2, and 4 were used to show that the footprinting patterns obtained in the induced and suppressed states are not specific to primers 1-3.

HindIII-linked oligonucleotides were used to amplify regions of interests in the human LDL receptor gene 5-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 3-end located at +35 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.).

For transfection experiments, HepG2 cells were seeded at 1 × 10⁶/60-mm dish 1 day in advance. Transfections were performed in triplicate with 6.0 µg of DNA for each construct and 2.0 µg 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 (±lovastatin) 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 µl 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 ± S.D.

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 × 10⁵/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 µg 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 [-³²P]ATP. Nuclear extract (5 µg) was incubated in a final volume of 15 µl of 10 mM Tris, pH 7.5, 50 mM NaCl, 5% glycerol, 1 mM dithiothreitol, 1 mM EDTA, 1 mM spermidine, 1 µg 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 × TBE (1 × 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

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 µg/ml, and maximal induction was obtained with 30 µM 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 1 to 117 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 59, an observation consistent with earlier in vivo footprinting studies (34, 35). However, G residues of the proximal Sp1 site (40 to 49) and the distal Sp1 site (92 to 100) 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 106. 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 126 to 145 (designated Footprint 1 (FP1)), and a slight protection of nucleotides from 175 to 187 (designated Footprint 2 (FP2)) (Fig. 3B). The protected Gs within the FP1 site included guanines at 126, 134, 135, 136, 143, and 145, whereas Gs at nucleotides 175, 177, 178, 183, and 187 were protected in the FP2 (Fig. 3C). Interestingly, the in vivo protected FP1 site (126 to 145) coincided with the in vitro DNase I protected region (127 to 159) 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.

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 5 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 273 to 127 region, that includes both the FP1 and FP2 sites in reporter construct containing the 126 to +35 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 222 to +88 region of the human LDL receptor gene 5-flanking region showed 2-3-fold higher reporter gene level expression compared to plasmids containing either 105 to +88 (both Sp1 sites and SRE-1) or 69 to +88 (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 5 deletion constructs without the FP2 site were prepared and transfected into HepG2 cells (Fig. 4A). As shown in Fig. 3A, deletion of sequence between 273 to 173, 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 5 deletions between nucleotides 173 and 139 (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 145 to 126 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 134 and 146 (plasmid H) abolished the FP1-dependent enhancement in the reporter gene level. Interestingly, specific changes of four nucleotides (guanines at 135/136 and cytosines at 145/146 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 273 to 126 region contribute very little and that nearly all of the 3-4-fold elevation in reporter gene expression is mediated by sequences within the 139 to 126 region of the human LDL receptor promoter in HepG2 cells.

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 120 to 146) 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, 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 5-flanking sequences in different species (human, rat, and hamster) showed remarkable conservation of the position and sequence of the FP1 site (Fig. 8) (39, 40, 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.

Comparison of the nucleotide sequence of the 5-flanking regions of the human, mouse, and hamster LDL receptor genes.

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)₆GG) and the GArC box (preference sequence G(A/T-rich)_6-7CTC) (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 immediate-early 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 GArC-binding 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 trans-acting 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 179 to 174) 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 trans-acting 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 5-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 FP1-binding 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.