Characterization of hepatic-specific regulatory elements in the promoter region of the human cholesterol 7alpha-hydroxylase gene.

Cholesterol 7α-hydroxylase is the rate-limiting enzyme in the degradation of cholesterol to bile salts and plays a central role in regulating cholesterol homeostasis. The mechanisms involved in the transcriptional control of the human gene are largely unknown. HepG2 cells represent an appropriate model system for the study of the regulation of the gene. To identify liver-specific DNA sequences in the promoter of the human CYP7 gene, we first examined the DNase I hypersensitivity in the 5′-region of the gene. An area of hypersensitivity was observed in the region from −50 to −200 of the human gene in nuclei from transcriptionally active HepG2 cells, but was absent in transcriptionally inactive HeLa cell nuclei or in free DNA. Various 5′-promoter deletion constructs were made and transfected into HepG2 cells. About 300 base pairs of upstream sequence are required for high level promoter activity of the human CYP7 gene in HepG2 cells. DNase I footprinting of the hypersensitive region revealed nine protected sequences. Gel retardation experiments demonstrated binding of HNF-3 to the segment from −80 to −70 and of hepatocyte nuclear factor HNF-4 (and ARP-1) to the segment from −148 to −127 of the human CYP7 promoter. Deletion of either of these sites depressed promoter activity in HepG2 cells. A third region from −313 to −285 is bound by members of the HNF-3 family and acts as an enhancer. Additionally, the segment from −197 to −173 binds a negative regulatory protein that is present in Chinese hamster ovary cell extracts and in HepG2 cell extracts. These experiments define the key control elements responsible for basal transcription of the human CYP7 gene in HepG2 cells.

zyme in the degradation of cholesterol to bile salts and plays a central role in regulating cholesterol homeostasis. The mechanisms involved in the transcriptional control of the human gene are largely unknown. HepG2 cells represent an appropriate model system for the study of the regulation of the gene. To identify liverspecific DNA sequences in the promoter of the human CYP7 gene, we first examined the DNase I hypersensitivity in the 5-region of the gene. An area of hypersensitivity was observed in the region from ؊50 to ؊200 of the human gene in nuclei from transcriptionally active HepG2 cells, but was absent in transcriptionally inactive HeLa cell nuclei or in free DNA. Various 5-promoter deletion constructs were made and transfected into HepG2 cells. About 300 base pairs of upstream sequence are required for high level promoter activity of the human CYP7 gene in HepG2 cells. DNase I footprinting of the hypersensitive region revealed nine protected sequences. Gel retardation experiments demonstrated binding of HNF-3 to the segment from ؊80 to ؊70 and of hepatocyte nuclear factor HNF-4 (and ARP-1) to the segment from ؊148 to ؊127 of the human CYP7 promoter. Deletion of either of these sites depressed promoter activity in HepG2 cells. A third region from ؊313 to ؊285 is bound by members of the HNF-3 family and acts as an enhancer. Additionally, the segment from ؊197 to ؊173 binds a negative regulatory protein that is present in Chinese hamster ovary cell extracts and in HepG2 cell extracts. These experiments define the key control elements responsible for basal transcription of the human CYP7 gene in HepG2 cells.
Cholesterol 7␣-hydroxylase catalyzes the rate-limiting step in the pathway that leads to the catabolism of cholesterol to bile acids (for review, see Ref. 1). Cholesterol 7␣-hydroxylase is a microsomal enzyme member of the cytochrome P-450 family. In human and rat, the major products of this metabolic pathway are cholic acid and chenodeoxycholic acid. Bile acids have an important role in cholesterol homeostasis; their synthesis and excretion cause a decrease in hepatic cholesterol levels, while their presence in the intestine facilitates the solubilization of dietary fats and is required for the absorption of cholesterol and fat-soluble vitamins. Because of the importance of these functions, bile acid synthesis in the liver is carefully regulated to maintain cholesterol homeostasis (1). To date, little is known about the molecular mechanisms that control cholesterol catabolism and bile acid synthesis.
The cDNAs and genes for cholesterol 7␣-hydroxylase have been isolated from rat (2)(3)(4), human (5,6), hamster (7), and mouse (8). CYP7 mRNA is found exclusively in the liver (9), making this gene a target for the study of the molecular mechanisms implicated in hepatic-specific gene expression. Work by several groups has demonstrated that CYP7 mRNA levels are modulated in cultured cells by a number of effectors. For example, in cultured rat hepatocytes (10) and human hepatoma (HepG2) cells, the addition of bile acids to the culture media suppresses CYP7 mRNA levels (11), and dietary cholesterol and dexamethasone increase CYP7 mRNA levels (12). In vivo, however, cholesterol feeding increases but dexamethasone reduces CYP7 mRNA levels in rats (12,13). The elegant work by Lavery and Schibler (14) demonstrated that in rats, CYP7 gene expression follows a strict diurnal rhythm, with mRNA levels peaking in the evening; this phenomenon is controlled at the level of transcription by a specific transcription factor, DBP.
Our long-term goal is to elucidate the molecular mechanisms that operate to regulate transcription of the human CYP7 gene in the liver as well as those that promote modulation by diet and hormones. As a first step toward our goal, we have pursued the identification of liver-specific elements that regulate basal transcription of the human CYP7 gene in hepatic cells. HepG2 cells have been used successfully as a model system to study CYP7 gene expression in a number of laboratories (11,15,16). Up-regulation of CYP7 mRNA by cholesterol as well as downregulation by bile acids have been demonstrated (11).
Our rationale has been to use DNase I hypersensitivity as a tool to map liver-specific elements that are relevant in vivo, followed by a thorough analysis of the underlying hepatic control elements in HepG2 cells prior to testing them in in vivo animal models. Using this approach, we established that liverspecific promoter elements of the CYP7 gene lie between Ϫ213 and ϩ1. Within this region are functional binding sites for the liver-enriched transcription factors HNF-3 (Ϫ80 to Ϫ70), HNF-4, and ARP-1 (Ϫ144 to Ϫ127). Furthermore, a ubiquitous transcription factor that binds to the region from Ϫ197 to Ϫ173 reduces promoter activity of the human CYP7 gene, and an enhancer resides within an HNF-3-like binding site at Ϫ300 to Ϫ293.

EXPERIMENTAL PROCEDURES
Plasmid Construction-All constructs were derived from plasmid D230 20.5-1 (kindly provided by the late Dr. Mike Komaromy). This plasmid contains the segment from Ϫ780 to ϩ133 of the human cho-lesterol 7␣-hydroxylase gene and was constructed by PCR 1 using oligonucleotide primers derived from the GeneBank TM data base sequence. The genomic sequence is flanked by BamHI sites, allowing it to be excised by digestion with BamHI. Construct Ϫ764CAT was made by insertion of the Ϫ764/ϩ46 BamHI fragment into a pOCAT vector (17) that had been digested with BamHI and treated with calf intestinal phosphatase. Plasmid Ϫ764CAT was used as a template for the remaining 5Ј-promoter deletion plasmids. The 5Ј-primers used for amplification were as follows: Ϫ341 to Ϫ317 for the Ϫ341CAT construct, Ϫ313 to Ϫ289 for the Ϫ313CAT construct, Ϫ285 to Ϫ261 for the Ϫ285CAT construct, Ϫ268 to Ϫ244 for the Ϫ268CAT construct, Ϫ227 to Ϫ203 for the Ϫ227CAT construct, Ϫ213 to Ϫ189 for the Ϫ213CAT construct, Ϫ91 to Ϫ67 for the Ϫ91CAT construct, and Ϫ65 to Ϫ41 for the Ϫ65CAT construct. In every case, the sequence of the primer started with CGCG-GATTC (the BamHI site). The 3Ј-primer for all these constructs extended from ϩ46 to ϩ22 and also contained a BamHI site at its 5Ј-end. In each case, the amplified product was digested with BamHI, followed by ligation to the pOCAT vector that had been digested with BamHI. To prepare plasmid Ϫ313CAT⌬(Ϫ80/Ϫ70), the following PCR primers were used: primer 1, from Ϫ313 to Ϫ292 (with the BamHI site at the 5Ј-end); primer 2, from Ϫ55 to Ϫ95, but without the sequence from Ϫ70 to Ϫ80; primer 3, from Ϫ95 to Ϫ55 without Ϫ80 to Ϫ70; and primer 4, from ϩ46 to ϩ22 with the BamHI site at the 5Ј-end. The first PCR utilized primers 3 and 4 and generated an intermediate product spanning from Ϫ95 to ϩ46 without Ϫ80 to Ϫ70. The two intermediate PCR products were mixed, denatured, reannealed, and used as primers on a third PCR to generate a product spanning from Ϫ313 to ϩ46 without Ϫ80 to Ϫ70. This BamHI fragment was then cloned into pOCAT. The orientation of the insert was determined by PCR. Constructs Ϫ227CAT⌬-(Ϫ197/Ϫ173), Ϫ213CAT⌬(Ϫ144/Ϫ127), and Ϫ313CAT⌬(Ϫ300/Ϫ293) were made in a similar manner as the construct described above. Construct Ϫ313mut-1 was made by introducing a primer in which five point mutations (CGTAC instead of AAACA at Ϫ298 to Ϫ294) were made to disrupt HNF-3 binding. The DNA sequence of every promoter deletion construct was verified by sequencing.
Tissue Culture and Transient Transfection Assays-Human hepatoma cells (HepG2), Chinese hamster ovary (CHO) cells, and HeLa cells were grown as described previously (18). Transient transfections with the various plasmid constructs were performed by the calcium phosphate coprecipitation method as described previously (19) with 7 g of the CAT gene expression plasmid and 6 g of an internal reference plasmid (pRSV␤-gal). In some experiments, 6 g of an expression plasmid for HNF-3␣ (20) or control vector was included. The CAT assays were performed according to the protocol of Gorman et al. (21), and CAT activities were normalized according to the results of the ␤-galactosidase activity in order to correct for differences in transfection efficiency.
Gel Mobility Shift Assays and Preparation of Nuclear Extracts-Nuclear extracts from HepG2 and CHO cells were prepared by the method of Dignam et al. (22). Extracts from mouse liver nuclei were made by the procedure of Gorski et al. (23). Whole cell extracts from COS cells were prepared as described previously (20). In some cases, cells were transfected with 20 g of an expression plasmid for HNF-3 (24) using the calcium phosphate coprecipitation technique as described above. For gel mobility shift assays, 1-4 g of nuclear extract was incubated for 30 min at room temperature with 0.5-1 ng of 32 P-labeled double-stranded oligonucleotide and 4 g of poly(dI-dC) in buffer containing 15 mM Hepes (pH 7.9), 15% glycerol, 0.6 mM EDTA, 60 mM KCl, 5 mM MgCl 2 , and 0.6 mM dithiothreitol. Incubation mixtures were then fractionated on 5% native polyacrylamide gels in 0.5 ϫ Tris borate/ EDTA. Gels were soaked in 10% glycerol for 10 min, dried, and exposed to x-ray films.
DNase I Hypersensitivity-DNase I hypersensitivity studies were performed as described by Levy-Wilson et al. (18).
DNase I Footprinting-DNase I footprinting was performed as described before (19).
Oligonucleotides-Oligonucleotides were purchased from the Stanford University Digestive Disease Center. Complementary sets of single-stranded oligonucleotides were annealed to form double-stranded oligonucleotides and then purified on nondenaturing 15% acrylamide gels.

DNase I Hypersensitivity in the 5Ј-End of the Human CYP7
Gene-To determine whether there are any DNase I-hypersensitive sites in the 5Ј-end of the human CYP7 gene, several sets of experiments were performed. A representative example is shown in Fig. 1. Nuclei from HepG2 cells or HeLa cells or free DNA was treated with increasing amounts of DNase I, followed by digestion of the purified DNA with EcoRV. The Southern blots were hybridized with the probe indicated at the bottom of Fig. 1. The original 2917-bp EcoRV fragment was progressively digested by incubation of HepG2 nuclei with increasing amounts of DNase I. Concomitantly, a broad band appeared and was designated DH1 (DNase I-hypersensitive). From the mobility of the DH1 band relative to known restriction fragment markers run in parallel with the samples, DH1 was localized to between Ϫ50 and Ϫ200 of the CYP7 gene, as shown above the map in the bottom part of Fig. 1. When nuclei from HeLa cells (in which the CYP7 gene is transcriptionally inactive) were treated in an identical manner with DNase I, the EcoRV fragment was resistant to degradation, and no DNase I-hypersensitive sites were detected (data not shown). The hypersensitive region was not observed when free DNA from HepG2 cells was subjected to the same procedure ( Fig. 1, right panel), suggesting that chromatin structural features in the vicinity of the transcriptional start site of the human CYP7 gene in HepG2 nuclei are selectively open and available for DNase I cutting.
DNase I Footprinting of the 5Ј-End of the Human CYP7 Gene-The DNase I hypersensitivity studies suggest that hepatic-specific transcription factors bind to the region from Ϫ50 to Ϫ200, thus causing its hypersensitivity. Binding of hepaticspecific nuclear proteins to the 5Ј-proximal region of the human CYP7 gene was examined by DNase I footprinting. Nine protected regions were observed in the DNA segment from Ϫ341 to ϩ46 using nuclear extracts from HepG2 cells (Fig. 2, A and B). Footprint 1 extends from Ϫ35 to Ϫ48, footprint 2 from Ϫ54 to Ϫ62, footprint 3 from Ϫ67 to Ϫ81, footprint 4 from Ϫ91 to Ϫ104, footprint 5 from Ϫ129 to Ϫ144, footprint 6 from Ϫ174 to Ϫ191, footprint 7 from Ϫ213 to Ϫ227, footprint 8 from Ϫ268 to Ϫ285, and footprint 9 from Ϫ313 to Ϫ341. The protected regions have been evolutionarily conserved, as evidenced by a correspondence among the footprints in human, mouse, and rat liver cells as shown in Fig. 2 (C and D), suggesting that the nuclear proteins that bind to these sequences have an important functional role in CYP7 gene transcription.
Functional Assays to Test the Role of the Footprints in Promoter Activity-The combined results from the hypersensitivity studies and DNase I footprinting strongly suggest that key liver-specific promoter elements reside in the 300 bp immediately upstream of the transcriptional start site of the CYP7 gene. This hypothesis was tested as follows. Several constructs were made in which 5Ј-segments of the human CYP7 gene of varying lengths were cloned upstream of the reporter CAT gene, and their promoter activity was tested in transient transfection assays in HepG2 cells. A schematic illustration of the constructs and their promoter activities are shown in Fig. 3.
The construct with the largest 5Ј-extension was Ϫ764CAT, followed by Ϫ341CAT. To elucidate the potential functional role of the footprinted regions, subsequent constructs were designed to exclude either one footprint at a time or the region between two footprints. Transient transfections into HepG2 cells revealed a small but reproducible reduction of CAT activity (ϳ20%) upon deletion of sequences between Ϫ764 and Ϫ341, suggesting that weak positive control elements reside in this region. On the other hand, a weak negative element may reside between Ϫ341 and Ϫ313 (footprint 9), as judged by the 30% increase in CAT activity of the Ϫ313CAT construct as compared with the Ϫ341CAT construct. Deletion of the segment from Ϫ313 to Ϫ285 caused a 40% decrease in CAT activity, suggesting that this segment may contain binding sites for a positive regulatory element. Further deletions from Ϫ285 to Ϫ227 and from Ϫ227 to Ϫ213 increased CAT activity by 30% each, respectively, suggesting that footprints 7 and 8 may harbor binding sites for negative regulatory elements. The Ϫ213CAT construct exhibited the highest promoter activity, suggesting that most of the important elements are localized within 213 bp upstream of the start site, in agreement with the DNase I data of Fig. 1. Deletion of the segment from Ϫ213 to Ϫ91 that includes footprints 4 -6 caused a 40% reduction in promoter activity, suggesting the removal of a positive element. The activity of the Ϫ65CAT deletion was 2-fold lower than that of the Ϫ91CAT construct, suggesting that a positive element resides between Ϫ91 and Ϫ65. Therefore, from the data in Fig.  3, we conclude that the segment from Ϫ213 to ϩ1 contains the minimal regulatory elements required for transcription from Gel Retardation Experiments Identify the Nuclear Factors Involved in CYP7 Promoter Activity-The decline in transcriptional activity observed when sequences between Ϫ213 and Ϫ91 were deleted (Fig. 3) suggests that DNA sequences in this region play a key role in basal promoter activity. Three nuclear protein-binding sites, namely footprints 4 -6, are located in this region.
To identify the protein factors responsible for footprints 4 -6, gel retardation experiments were conducted. A doublestranded oligonucleotide corresponding to the segment from Ϫ197 to Ϫ173 (representing footprint 6) was incubated with nuclear extracts from HepG2 and CHO cells (Fig. 4). Two specific complexes were formed with HepG2 extracts (lane 1) that were competed for by an excess of unlabeled oligonucleotide (lane 2). The same specific complexes were seen when using CHO extracts (lanes 3 and 4), suggesting that the transcription factor binding to this region may be ubiquitous. The specific complex of lower molecular weight may represent a proteolytic fragment of the major binding protein since the intensity of this band varied from extract to extract. However, we cannot rule out the possibility that two different proteins bind to this sequence. An excess of unlabeled oligonucleotides representing in each case the binding site for a known transcription factor such as HNF-1, C/EBP, HNF-3, and HNF-4 failed to abolish complex formation (data not shown). Comparison of the DNA sequence from Ϫ197 to Ϫ173 to a data base of transcription factor-binding sites revealed only some similarity to the binding site for the GATA factor that plays a role in erythroid expressed genes (25,26).
The segment from Ϫ148 to Ϫ127 (representing footprint 5) exhibited sequence similarity to the binding sites for the liverenriched transcription factors HNF-4 (27) and ARP-1 (28). These two proteins bind to the same DNA sequence with different affinities. When the labeled Ϫ148/Ϫ127 oligonucleotide was incubated with an extract from COS cells that had been transfected with an HNF-4 expression vector (Fig. 5, lane 2), we observed a complex that was not detected with control extract. This complex was competed for by an excess of unlabeled oligonucleotide (lane 3), confirming its identity as the HNF-4 complex. A weaker but specific HNF-4⅐DNA complex was also detected with HepG2 nuclear extracts (lanes 6 and 7). When the Ϫ148/Ϫ127 probe was incubated with an extract from COS cells that had been transfected with an ARP-1 expression plasmid, we observed formation of a retarded complex (lane 4), not evident with control extract, that was also specific (lane 5). However, the affinity of ARP-1 for the footprint 5 sequence appears to be severalfold lower than the affinity of HNF-4 for that same sequence.
Incubation of an oligonucleotide representing footprint 4 with nuclear extracts yielded one weak specific complex of high molecular weight (data not shown). Computer analysis of footprint 4 revealed no similarities to the binding sites of known transcription factors.
Analysis of the Function of the Footprinted Sequences-To ascertain the functional importance of the proteins binding to footprints 5 and 6, the effect of deleting these regions upon promoter activity was determined. The Ϫ144/Ϫ127 deletion (footprint 5) was made in the context of the Ϫ213CAT construct, and the Ϫ197/Ϫ173 deletion (footprint 6) was made in the context of the Ϫ227CAT construct (Fig. 6). Deletion of the sequence from Ϫ197 to Ϫ173 (footprint 6) increased promoter activity by 2-fold, suggesting that it binds a negative regulatory element. On the other hand, deletion of the Ϫ144/Ϫ127 segment (footprint 5) decreased CAT activity by ϳ2.5-fold, indicating that the protein(s) binding to footprint 5 also play a role in the transcriptional activation of the CYP7 gene. Having demonstrated that both HNF-4 and ARP-1 can bind to footprint 5 (Fig. 5), we examined the effect of an excess of each of these transcription factors upon the CAT activities of the wild-type and deletion constructs. To this end, the reporter constructs were cotransfected with expression plasmids for either HNF-4 or ARP-1. Cotransfection of the wild-type Ϫ213CAT construct with HNF-4 failed to alter promoter activity, suggesting that HNF-4 may not be limiting in our HepG2 cells. As expected, the additional HNF-4 protein did not affect the CAT activity of the Ϫ213CAT construct in which the HNF-4 site had been deleted (Fig. 6). An excess of ARP-1 protein reduced the transcriptional activity of the wild-type Ϫ213CAT construct by 2-fold, a result consistent with the fact that ARP-1, when binding to its cognate sequence, usually exerts a repressor effect (28). As expected, an excess of ARP-1 protein had no effect in the absence of the ARP-1-binding site.
These data suggest that both HNF-4 and ARP-1 can bind to the Ϫ144/Ϫ127 region. When HNF-4 binds, it promotes activation of transcription; when ARP-1 binds, transcription is decreased. This is analogous to the situation in the proximal promoter of the human apoB gene, where an HNF-4 (ARP-1) site resides (20).
Role of HNF-3 in CYP7 Promoter Activity-Computer analysis of the sequence of the proximal promoter region of the human CYP7 gene revealed a perfect match to the HNF-3binding site (TGTTTGCT) (20) in the segment from Ϫ80 to Ϫ70, a sequence that is 100% conserved among the human, rat, and mouse genes. Binding of nuclear proteins from HepG2 cells and mouse liver to an oligonucleotide encompassing this sequence (Ϫ88 to Ϫ65) was examined. In Fig. 7, we observe that the labeled Ϫ88/Ϫ65 oligonucleotide forms two major retarded complexes with HepG2 extracts, designated as A and B (lane 3). The mobility of these complexes is identical to that of complexes A and B formed between a labeled oligonucleotide representing a consensus HNF-3-binding site and HepG2 nuclear proteins (lane 1). Specificity of binding is established in lane 2. Furthermore, the Ϫ88/Ϫ65 probe also formed two specific complexes with mouse liver nuclear extracts (lane 4) that were competed for by the homologous oligonucleotide (lane 5) as well as by the HNF-3 consensus oligonucleotide (lane 6). The HNF-3 consensus oligonucleotide formed retarded complexes with mouse liver proteins that were identical in mobility to those formed by the Ϫ88/Ϫ65 oligonucleotide (lane 7) and were specific (lane 8) and also competed for by the Ϫ88/Ϫ65 oligonucleotide (lane 9). These observations confirm that HNF-3 does bind to the Ϫ88/Ϫ65 segment of the CYP7 promoter.
The functional significance of this binding was studied by making a promoter construct in which the HNF-3 site at Ϫ80 to Ϫ70 was deleted in the context of the Ϫ313CAT construct (Fig.  8). The Ϫ313CAT construct has an activity of 1.10 as compared with the Ϫ764CAT construct. Deletion of the Ϫ80/Ϫ70 segment reduced promoter activity by 3-fold, demonstrating that binding of HNF-3 to this sequence is important for promoter activity in HepG2 cells. When an HNF-3␣ expression vector was cotransfected with our constructs, the activity of the wild-type Ϫ313CAT construct was 5-fold higher, implying that HNF-3␣ is limiting in HepG2 cells. When the region from Ϫ80 to Ϫ70 was deleted from the construct, transcriptional activity in the presence of an excess of HNF-3␣ was again reduced, thus underscoring the importance of this region in hepatic transcription of the CYP7 promoter.
Based on computer analysis, there are three other potential HNF-3-binding sites within the region from Ϫ320 to Ϫ240 of the human CYP7 promoter, namely Ϫ316 to Ϫ306, Ϫ288 to Ϫ278, and Ϫ235 to Ϫ243 (29). The footprinting data of Fig. 2 revealed a protected region from Ϫ285 to Ϫ268. However, deletion of this segment did not affect promoter activity (data not shown), suggesting that the putative HNF-3 site at Ϫ285 to Ϫ268 is not functional in our assays. Similarly, the segment from Ϫ255 to Ϫ245 was not protected by nuclear proteins from liver or HepG2 cells (Fig. 2), suggesting that HNF-3 may not bind to that sequence, and deletion of the region from Ϫ268 to Ϫ227 did not significantly affect the activity of the human CYP7 promoter. To evaluate the possible role of HNF-3 in these regions, we cotransfected our promoter deletion constructs together with an HNF-3␣ expression vector (Fig. 9). The transcriptional activity of the Ϫ764CAT, Ϫ341CAT, and Ϫ313CAT constructs was 5-7-fold higher in the presence of excess HNF-3␣. On the other hand, the activity of the Ϫ285CAT construct dropped 2-fold as compared with that of the Ϫ313CAT construct in the presence of cotransfected HNF-3␣, suggesting that a functional HNF-3-binding site may reside in the segment from Ϫ313 to Ϫ285 and that the putative site at Ϫ243 to Ϫ235 may not be functional. Examination of the DNA sequence from Ϫ313 to Ϫ285 revealed two (5 out of 8 bp) matches to the HNF-3 recognition sequence (Ϫ297 to Ϫ293 and Ϫ290 to Ϫ286). The sequence of this segment of the human CYP7 gene is shown in Fig. 10A, with the two 5-bp matches to the HNF-3 recognition sequences indicated by brackets.
Binding of the Ϫ313/Ϫ285 sequence to nuclear proteins from HepG2 cells was examined. As shown in Fig. 11A, the Ϫ313/ Ϫ285 oligonucleotide probe formed five specific retarded complexes with HepG2 nuclear proteins, namely complexes A, B, C, and E and a lower molecular weight complex (lanes 1 and 2). The HNF-3 consensus oligonucleotide from the transthyretin promoter competes for formation of complexes A, B, and C, but not complex E or the smaller complex (lane 3). On the other hand, the HNF-3 consensus probe forms two major specific retarded complexes with HepG2 proteins, namely complexes C and D, and a lower molecular weight complex migrating between the nonspecific complex and complex E (lanes 4 and 5) that are partially competed for by the Ϫ313/Ϫ285 oligonucleotide (lane 6).
These data suggest that the Ϫ313/Ϫ285 segment does bind members of the HNF-3 family. The question arises as to whether these DNA/protein interactions are functionally significant. To elucidate the functional role of these sequences, we first mutagenized the putative HNF-3-binding site centered at Ϫ297 to Ϫ293 by altering those 5 bp as shown for the Ϫ313mut-1 construct in Fig. 10A. The CAT activity of the Ϫ313mut-1 construct was 3-fold lower than that of the wildtype construct, thus validating the importance of those 5 bp in the transcriptional activity of the human CYP7 promoter (Fig.  10B). The involvement of HNF-3␣ in binding to this segment was demonstrated by cotransfection of the wild-type and mutant Ϫ313CAT constructs with an HNF-3␣ expression plasmid. The transcriptional activity of the Ϫ313mut-1 construct in the presence of an excess of HNF-3␣ was reduced by Ͼ2-fold as compared with the wild-type construct (Fig. 10B). In Fig. 11B, the DNA binding properties of the Ϫ313mut-1 and wild-type sequences are compared. As shown above in Fig. 11A, the wild-type oligonucleotide forms four major specific retarded complexes with HepG2 nuclear proteins, namely complexes A, B, C, and E, with complex B being the strongest (lanes 1 and 2). The mutant oligonucleotide as well as the HNF-3 consensus oligonucleotide compete mainly for the formation of complex C

FIG. 7. An HNF-3-binding site is present between ؊88 and ؊65.
The layout is similar to that described in the legend to Fig. 4. Retarded complexes are designated A and B. 3 and 4). The Ϫ313mut-1 oligonucleotide forms two major specific complexes, complex C and a lower molecular weight complex (lanes 5 and 6). Interestingly, both the HNF-3 oligonucleotide and the wild-type Ϫ313/Ϫ285 sequence compete for formation of complex C (lanes 7 and 8). The HNF-3 oligonucleotide forms complexes C and D (lanes 9 and 10). Thus, a 5-bp mutation in the segment from Ϫ297 to Ϫ293 disrupts binding of the proteins responsible for the formation of complexes A and B, but not complex C, suggesting that the second HNF-3 site at Ϫ290 to Ϫ286 may be responsible for complex C formation and for the residual CAT activity of the Ϫ313mut-1 construct (Fig. 10). In parallel experiments, we deleted the segment from Ϫ300 to Ϫ293 in the context of the Ϫ313CAT construct. Gel retardation experiments with the Ϫ313CAT⌬(Ϫ300/Ϫ293) oligonucleotide revealed no retarded complexes (data not shown), suggesting that the 8-bp deletion can disrupt binding of nuclear proteins to the adjacent Ϫ290/ Ϫ286 sequence. Transfections of the wild-type and mutant Ϫ313CAT⌬(Ϫ300/Ϫ293) constructs into HepG2 cells revealed a 5.5-fold decrease in CAT activity of the deletion construct as compared with the wild-type construct, thus confirming that the HNF-3 site at Ϫ300 to Ϫ293 is important for CYP7 promoter activity (Fig. 10). Cotransfection of the HNF-3␣ expression plasmid with the wild-type and Ϫ313CAT deletion constructs yielded similar results, i.e. a 4-fold reduction in promoter activity of the deletion construct.  10. Deletion or mutagenesis of the sequence from ؊302 to ؊293 severely reduces transcriptional activity. Shown in A are the DNA sequences of the segment from Ϫ313 to Ϫ285 of the human CYP7 gene as well as the sequences of the Ϫ313mut-1 oligonucleotide and the Ϫ313CAT⌬(Ϫ300/Ϫ293) oligonucleotide. Shown in B are the CAT activities of the wild-type and mutant constructs either in the presence or absence of cotransfected HNF-3␣, with the constructs shown on the left and the CAT activities on the right. In summary, mutagenesis or deletion of the segment from Ϫ300 to Ϫ293 in the context of a Ϫ313CAT promoter construct severely impairs promoter activity, demonstrating that protein binding to this segment is functionally significant. Results of the gel retardation experiments show that members of the HNF-3 family of transcription factors are involved in binding to the Ϫ313/Ϫ285 region and thus play an important functional role in CYP7 gene regulation in HepG2 cells. Our overall data can be reconciled by postulating that, in transient assays, ϳ213 bp of upstream sequence are required for basal promoter activity of the human CYP7 gene. However, in the context of the entire gene, the sequence from Ϫ313 to Ϫ285 binds HNF-3 and related proteins, and it functions as a transcriptional enhancer because deletion or mutagenesis of the HNF-3 sequence at Ϫ300 to Ϫ293 severely reduces transcription. DISCUSSION The mechanisms by which expression of the CYP7 gene is regulated are worthy of study because of the important role that cholesterol 7␣-hydroxylase plays in regulating overall cholesterol homeostasis. However, information available to date is fragmentary. Using transgenic mice, Ramirez et al. (30), using large constructs encompassing Ϫ1633 bp of the rat CYP7 gene 5Ј-upstream region, ligated to a mouse albumin enhancer and linked to the reporter LacZ gene, demonstrated that the regulation of the reporter gene by bile salts that is observed in vivo can be reproduced. However, in the absence of the mouse albumin enhancer, expression of the reporter gene in the liver was not detectable. Using a line of transformed cultured mouse hepatocytes and stable transfections, a liver-specific enhancer was localized some 7 kilobase pairs upstream of the transcriptional start site of the rat CYP7 gene and was required for high level transcription from the CYP7 promoter. But, this enhancer may not function in vivo in transgenic mice. Using HepG2 cells and transient transfection assays, Molowa et al. (31) implicated the transcription factor HNF-3 as playing a role in transcription of the human CYP7 gene by binding to a site located between Ϫ432 and Ϫ220. Chiang and Stroup (15), studying the proximal promoter region of the rat CYP7 gene, described two regions protected from DNase I digestion that they designated footprints A and B. Rat footprint A (Ϫ81 to Ϫ35) corresponds to human footprints 1-3 of Fig. 2, and footprint B is equivalent to our footprint 5. These investigators have suggested (based on computer analysis of the DNA sequence) that HNF-3, C/EBP, HNF-4, retinoic acid receptor, COUP, and glucocorticoid receptor may bind to the rat promoter region and influence transcription. They have also proposed that recognition sites for binding proteins that respond to corticosteroids as well as insulin and bile acids may also reside in this region. Lavery and Schibler (14) have shown that the liver-enriched basic leucine zipper protein DBP binds to an element centered at Ϫ225 of the rat CYP7 gene and plays an important role in the circadian transcription of the gene. A similar sequence is present in footprint 7 (Ϫ227 to Ϫ213) of the human gene (Fig. 2). Thus, the apparent complexity and multiplicity of interactions among various transcription factors that may regulate this gene warrant a comprehensive and systematic approach.

(lanes
Our long-term goal with the human CYP7 gene is to identify all of the control elements that regulate this gene and that modulate its in vivo expression. Encouraged by our earlier results with the apoB gene (30) and by the suitability of HepG2 cells as a model system for hepatic-specific expression of the CYP7 gene (11,32,33), we began our studies by identifying a key DNase I-hypersensitive region in the segment from Ϫ50 to Ϫ200 of the human CYP7 gene. This DNase I-hypersensitive region is absent in cells in which the CYP7 gene is transcriptionally inactive, such as HeLa cells, and in free DNA, suggesting that it is the chromatin structural features of the CYP7 gene that make this region open and available for interaction with transcription factors.
Transfections of 5Ј-CAT deletion constructs illustrated in Fig. 3 yielded results that were consistent with those from the DNase I hypersensitivity studies. Together, both approaches FIG. 11. Gel retardation studies show that the segment from ؊313 to ؊285 binds to members of the HNF-3 family. The labeled DNA probe used in each experiment as well as the source of nuclear extract and competitor DNA are indicated above the lanes. The retarded complexes are designated by letters. NS, nonspecific complex. point to the location of key hepatic-specific elements between Ϫ213 and ϩ1. Within this region, we found functional binding sites for HNF-3 (Ϫ80 to Ϫ70), HNF-4 (and/or ARP-1) (Ϫ144 to Ϫ127), and a ubiquitous transcription factor (Ϫ197 to Ϫ173) present also at high concentrations in CHO cell extracts. Deletion of either the HNF-3 or HNF-4 sites reduced promoter activity, while deletion of the Ϫ197/Ϫ173 sequence increased promoter activity. The HNF-3 and HNF-4 sites are also present in the rat gene, but their functional role has not been determined. HNF-4 and HNF-3 may act synergistically, as has been shown to occur in the promoter of another liver-specific gene, the apoA-I gene (34). The negative influence of the sequence from Ϫ197 to Ϫ173 upon transcription is of interest. It is not uncommon to find that a ubiquitous factor, working in conjunction with other cell-specific factors, plays a key role in transcriptional activation or repression. For example, in the human apoC-III gene, an interaction between SP1, bound to distal regulatory sites, and HNF-4, bound to a more proximal promoter site, is required for transcriptional activation (35).
Another interesting observation relates to the role of the sequence from Ϫ300 to Ϫ285 in CYP7 promoter activity. This sequence binds one or more proteins related to the HNF-3 family of transcription factors. Deletion of the 8-bp sequence (Ϫ300 to Ϫ293) in the context of the Ϫ313CAT construct decreased promoter activity by 70%. Mutagenesis of the HNF-3binding site within this sequence also decreased promoter activity, suggesting that the segment from Ϫ313 to Ϫ285 functions as a hepatic-specific enhancer for the CYP7 promoter. This observation is in good agreement with the hypothesis formulated by Molowa et al. (31), suggesting that a liver-specific enhancer harboring HNF-3-binding sites is present in the segment from Ϫ432 to Ϫ220 of the human CYP7 gene.
It has recently been proposed (29) that a number of regulatory elements responsive to many effectors such as phorbol ester, insulin, glucocorticoids, and thyroid hormone may reside within the segment of the human CYP7 promoter from Ϫ764 to ϩ46 studied in this work. However, the precise location of these control sequences has not yet been determined. Our studies provide detailed information regarding the hepatic-specific elements required for basal transcription of the human gene in HepG2 cells. It would be interesting to elucidate whether the DNA sequence elements responsive to thyroid hormone, for example, interact with one or more of the hepatic elements described here or whether different mechanisms operate to modulate the expression of this gene.
Because our ultimate goal is to identify all of the regulatory elements of the human CYP7 gene that are functional in vivo, we want to ascertain the role that the hepatic-specific sequences characterized in HepG2 cells play in liver expression of these constructs in transgenic mice. Future experiments will address these questions.