The nuclear matrix protein CDP represses hepatic transcription of the human cholesterol-7alpha hydroxylase gene.

To date, the molecular mechanisms that govern hepatic-specific transcription of the human cholesterol 7alpha-hydroxylase (CYP7A1) gene are poorly understood. We recently reported that the region extending from -1888 to +46, which includes the promoter, is not capable of conferring expression to human CYP7A1 promoter lacZ transgenes in the livers of mice, but that expression is observed with transgenes containing the entire structural gene. To locate liver-specific elements in other segments of the human gene, DNase I hypersensitivity studies were performed with transcriptionally active, liver-derived HepG2 cells and with transcriptionally inactive HeLa cells. Three DNase I hypersensitivity sites were detected within the first intron of the human CYP7A1 gene, but only in HepG2 cells. Transient transfection experiments with HepG2 cells revealed a transcriptional repressor within intron 1. Five binding sites for the CAAT displacement protein (CDP) were detected within intron 1. Since CDP is a nuclear matrix protein, two methods were employed to localize nuclear matrix attachment sites within intron 1 of the human CYP7A1 gene. A matrix attachment site was found throughout the entirety of intron 1. Gel retardation experiments and cell transfection studies provided evidence for the repression mechanism. Repression is achieved by displacement by CDP of two hepatic activators, namely HNF-1alpha and C/EBPalpha, that bind to three different sites within intron 1. Additionally, CDP represses transactivation mediated by these two activators.

The rate-limiting step in the major pathway leading to bile acid biosynthesis from cholesterol in humans is catalyzed by cholesterol 7␣-hydroxylase (CYP7A1). 1 This enzyme is encoded by a single-copy gene that is expressed only in the liver, thus providing a good model for studying liver-specific gene regulation. Transcription of the human and rodent CYP7A1 gene is modulated by various hormones (for reviews, see Refs. 1 and 2), by cholesterol and its derivatives, as well as by bile acids. Additionally, in rats and rabbits, expression of this gene follows a circadian rhythm (3,4) and the rat gene displays a complex pattern of regulation during development (5). The molecular mechanisms involved in these regulatory processes remain obscure. Recently, two members of the orphan receptor family, namely the oxysterol receptor LXR and the farnesyl receptor FXR, have been implicated in dietary regulation of the rat CYP7A1 gene. Cholesterol derivatives such as oxysterols are ligands for LXR, which binds to the proximal promoter region of the rat gene, thus stimulating transcription (6 -8). Similarly, bile acids are the ligands for FXR, and binding of FXR to the rat promoter represses transcription by an as yet, undetermined mechanism. It has not been established whether similar mechanisms operate in the regulation of the human gene, although the DNA sequence of the LXR binding site is not conserved between rodents and humans. Indeed, in humans, the LXR site is replaced by a binding site for the liver-enriched transcription factor HNF-1 (9).
Our goal is to identify the liver-specific regulatory elements required for the in vivo expression of the human CYP7A1 gene. Earlier work identified functional binding sites for several liver-enriched transcription factors in the segment from Ϫ764 to ϩ1 of the human CYP7A1 gene (the promoter region). Nevertheless, the segment from Ϫ1888 to ϩ46 was not sufficient to confer hepatic expression upon a lacZ reporter gene in transgenic mice, suggesting that additional key hepatic regulatory elements may reside upstream of position Ϫ1888, within introns, or 3Ј of the structural gene.
To identify additional hepatic-specific regulatory elements, DNase I hypersensitivity (DH) studies were carried out. They revealed three DH sites within intron 1 of this gene. Transient transfection studies with HepG2 cells demonstrated that intron 1 has a repressor activity and that the main protein binding to this region is the CAAT displacement protein, CDP (10). Because CDP has been shown to be an integral component of the nuclear matrix, we examined the possibility that intron 1 may harbor nuclear matrix attachment sites (MARs). Our experiments demonstrated that intron 1 contains a MAR. In addition to the chromatin organizational function of the nuclear matrix, MARs have been shown to play important roles in the regulation of transcription. For example, transcription of the variable region of the rearranged immunoglobulin gene requires an enhancer localized in the first intron of the gene that is flanked by MARs (11). Experiments with transgenic mice demonstrated that activation of the gene promoter requires the MARs that flank the intronic enhancer (12). It was proposed that the MARs collaborate with the enhancer to generate an extended domain of "open" chromatin, accessible to regulatory proteins. Furthermore, nuclear matrix fractions from mammalian cell lines contain DNA binding sites related or identical to those for several transcription factors such as SP1, ATF, C/EBP, AP1, Oct-1, and others (13). On the other hand, an intronic nuclear matrix protein has been shown to repress transcription of the HIV-1 LTR (14). Additionally, the MAR-binding protein SATB1 (special AT-rich sequence-binding protein) has been shown to participate in negative regulation of the murine mammary tumor virus LTR in lymphoid tissues (15).
In this report, it is shown that repression of the human CYP7A1 gene is mediated by the MAR-bound repressor CDP and involves displacement of two hepatic transcriptional activators, HNF-1␣ and C/EBP␣, from their binding sites within intron 1 of the CYP7A1 gene.

EXPERIMENTAL PROCEDURES
Plasmid Constructs-Plasmids Ϫ898CAT and Ϫ764CAT have been described previously (16). To make plasmid p7␣I1-764CAT, the 1541-bp full-length human CYP7A1 intron 1 along with 23 bp of exon 1 and 71 bp of exon 2 was amplified by PCR using primers IntI5 (5Ј-GCC TGC AGA CTA GTC TTA TTC TTG GAA TTA GA GAA G) and IntI3 (5Ј-GCC TGC AGA CTA GTA GAG GTG GTT CAC CCG TTT GC). The resulting 1585-bp PCR product was digested with SpeI and ligated into SpeI-digested p-764CAT (16) to create p7␣I1-764CAT. Alternatively, to make p7␣-764-I1CAT, the SpeI ends were filled using Klenow DNA polymerase and dNTPs and intron 1 was subsequently ligated into a pOCATlink vector that had been digested with HinCII and phosphatased. The final step for the construction of p7␣-764-I1CAT involved ligation of the BamHI fragment containing the CYP7A1 promoter into the BamHI site of pOCATlink situated upstream of intron 1. The orientations of the promoter fragment and of I1 were determined by DNA sequencing. 5Ј p7␣I1-764CAT, containing 563 bp from the 5Ј half of human CYP7A1 intron 1 plus 23 bp of exon 1, was derived from p7␣I1-764CAT by digestion with ClaI, and the resulting 591-bp fragment was gel isolated and ligated into ClaI-digested p-764CAT. Orientation was determined by DNA sequencing. 3Ј p7␣I1-764CAT contains 928 bp from the 3Ј half of CYP7A1 intron 1 and 71 bp of exon 2. It was constructed by digestion of p7␣I1-764CAT with ClaI followed by religation. Intron 1 was sequenced by the dideoxy termination method of Sanger et al. (17). The CYP7A1 HCDP TATA CAT and 2C3 TATA CAT constructs as well as the apoB HCDP TATA CAT and 2C3 TATA CAT constructs were prepared by blunt-end ligation of the kinased, doublestranded oligonucleotides followed by ligation into a blunt-ended, phosphatased TATA CAT plasmid. Integrity of these constructs was confirmed by DNA sequencing.
Cell Cultures, DNase I Hypersensitivity, and Transfections-HepG2 cells were cultured in monolayers as described previously (16). DNase I hypersensitivity studies and transient transfection assays with HepG2 cells were described before (16). A control plasmid (RSV-␤gal) to monitor for differences in transfection efficiencies between plasmids, was always included at 5 g/transfection, and all experiments were performed three times in duplicate.
Preparation of Nuclear Matrices-Nuclear matrices were prepared according to the procedure of Cockerill and Garrard (18). In brief, nuclei (ϳ1 mg of nucleic acids/ml) in RSB (10 mM NaCl, 3 mM MgCl 2 , 10 mM Tris-HCl (pH 7.4)), 0.25 M sucrose, and 1 mM CaCl 2 , were digested with 100 g of DNase I/ml (Worthington) for 1-2 h at 23°C. After centrifugation at 750 ϫ g for 10 min at 4°C, the pellets were resuspended in RSB, 0.2 M sucrose, and an equal volume of a cold solution containing 4 M NaCl, 20 mM EDTA, and 20 mM Tris-HCl (pH 7.4) was added. After 10 min at 4°C, followed by centrifugation at 1500 ϫ g for 15 min, the pellets were extracted twice by suspension in a cold solution of 2 M NaCl, 10 mM EDTA, 10 mM Tris-HCl (pH 7.4), 0.5 mM phenylmethylsulfonyl fluoride, and 0.25 mg of BSA/ml and then centrifuged for 15 min at 4500 ϫ g at 4°C. The nuclear matrices were then washed with RSB, 0.25 M sucrose, and 0.25 mg of BSA/ml at 4°C and resuspended in the same solution. Although fresh preparations were used in most experiments, occasionally matrices were stored for up to 3 months at Ϫ20°C in the presence of 50% glycerol.
DNA Binding to Nuclear Matrices-DNA binding assays were performed by the protocol of Cockerill and Gerrard (18). In brief, 60 g of nuclear matrix proteins in 10 l were added to 90 l of an assay solution prepared to yield a final concentration of 50 mM NaCl, 10 mM Tris-HCl (pH 7.8), 2 mM EDTA, 0.25 M sucrose, 0.25 mg of BSA/ml, 20 ng of 32 P-end-labeled DNA fragments/ml, and sheared, denatured Escherichia coli DNA (50 -1000 g/ml) as unlabeled competitor. Under our experimental conditions, the ratio of labeled DNA fragments to matrices was such that 20 -50% of the input DNA bound specifically to matrix proteins. Incubation was for 1-2 h on a shaker at room temperature (23°C), followed by the addition of 500 l of assay buffer without DNA and centrifugation at 10,000 ϫ g for 1 min at 4°C to recover the matrices. The pellets were washed in 1 ml of assay buffer without carrier DNA, followed by solubilization of matrix-bound DNA in 0.5% SDS, overnight treatment with 0.4 mg of proteinase K/ml, phenol extraction, and ethanol precipitation after addition of 10 g of unlabeled carrier DNA. The purified matrix-bound DNA fragments were resolved on 1% agarose gels (40 mM Tris acetate and 2 mM EDTA (pH 7.8)), and the dried gels were autoradiographed.
Isolation and Digestion of Nuclear Halos-Nuclear halos were prepared as described by Mirkovitch et al. (19). Briefly, 10 A 260 units of nuclei in 10 l of isolation buffer (3.75 mM Tris-HCl (pH 7.4), 0.05 mM spermine, 0.125 mM spermidine, 1% (v/v) thiodiglycol, 20 mM KCl), with 0.1% digitonin, were heated at 37°C for 20 min to stabilize the nuclei. Seven milliliters of low salt extraction buffer (5 mM Hepes (pH 7.4), 0.25 mM spermidine, 2 mM EDTA, 2 mM KCl, 0.1% digitonin, 25 mM 3,5diiodosalicylic acid, lithium salt (Sigma) was slowly added at room temperature. After 5 min at room temperature, histone-depleted nuclei were recovered by centrifugation at 2400 ϫ g for 20 min at room temperature. The pellet was then washed four times with 8 ml of digestion buffer (20 mM Tris-HCl (pH 7.9), 0.05 mM spermine, 0.125 mM spermidine, 20 mM KCl, 70 mM NaCl, 10 mM MgCl 2 , 0.1% digitonin, 100 kallikrein inactivator units of Trasylol/ml, 0.1 mM phenylmethylsulfonyl fluoride). Appropriate restriction enzymes were then added at 1000 units/ml, and digestion proceeded for 3 h at 37°C in a shaking water bath. Solubilized DNA (S) was separated from nuclear scaffold-associated DNA (P) by centrifugation at 2400 ϫ g for 10 min at 4°C. DNA from S and P fractions was recovered after digestion with proteinase K in the presence of 1% SDS, phenol extraction, and ethanol precipitation. Naked DNA purified from intact nuclei was digested similarly (and was labeled T, for total DNA). In each experiment, the amount of DNA recovered in S and P was determined from the A 260 of each fraction. Twenty micrograms of each sample was electrophoresed on agarose gels and then transferred to nitrocellulose paper (Schleicher & Schuell BA85) by the method of Southern. The filters were subjected to UV cross-linking and prehybridized for 2-4 h at 42°C in 5ϫ SSPE (0.75 M The sizes were determined by comparison to those of radiolabeled DNA size markers run in parallel with the sample (data not shown). At the bottom, a restriction map of the area is shown, with the promoter, exons 1 and 2, and intron 1 emphasized, and the DH sites depicted by triangles above the map. The location of the probe used is shown below the map.
Gel Retardation Assays-COS cells were transfected (as described above) with 10 g of expression plasmids for either CDP (pMT2.CDP), C/EBP-␣ (pMSV.C/EBP␣), or HNF-1␣ (pGEM.HNF-1␣), and cellular lysates enriched for these proteins were prepared as described previously (20). Binding sites for transcription factors CDP, HNF-1␣, and C/EBP␣ were identified within CYP7A1 intron 1, using the online TRANSFAC 3.5 data base sequence search and analysis service (21). Single-stranded oligonucleotides were annealed, purified, and end-labeled according to procedures previously described (22). The larger oligonucleotide probes used in the C/EBP␣/CDP displacement studies were generated using PCR with either CDP site 2 in combination with the antisense I1-C/EBP␣ oligonucleotide (probe 2C) or with CDP site 3 antisense oligonucleotide (probe 2C3) and I1FCAT plasmid DNA as template. The probe for C3 was similarly made using PCR but with the I1-C/EBP␣ oligonucleotide and the antisense oligonucleotide corresponding to site 3-CDP. Antibodies specific for CDP, HNF-1␣, and C/EBP␣ were purchased from Santa Cruz Biotechnologies, Inc. Binding reactions and electrophoretic analyses of DNA-protein complexes were performed as described elsewhere (22). Typically 1 ng of labeled doublestranded oligonucleotide probe was incubated with extracts (5-10 g of protein) for 20 min at 25°C. Competition assays were performed utilizing 200 ng of unlabeled double-stranded oligonucleotides. For supershift reactions, the binding reaction was preincubated at 25°C for 15 min prior to the addition of 5-10 g of antibody, and antibody binding reactions were incubated an additional 15 min at 25°C. DNA-protein and antibody supershifted complexes were resolved on 5% nondenaturing 0.5ϫ TBE gels and were visualized by autoradiography of dried gels.
Oligonucleotides-The oligonucleotides used in this study are listed below.

Intron 1 Contains Liver-specific DNase I-hypersensitive
Sites-To localize hepatic regulatory elements of the human CYP7A1 gene in segments outside of the 5Ј proximal promoter region, DH studies were performed. DH sites reflect an open chromatin structure which could facilitate binding of transcriptional activators or repressors. To this end, nuclei from transcriptionally active HepG2 cells and from transcriptionally inactive HeLa cells were incubated with DNase I as indicated in the top panel of Fig. 1, followed by digestion with ApaI and PstI to generate a 4578-bp fragment from the region of interest. In HepG2 cells (Fig. 1, top left), the original fragment was pro-

FIG. 2. A transcriptional repressor activity in intron 1 of the human
CYP7A1 gene. Panel A shows a map of intron 1 (I1) and of its two components used below, 5ЈI1 and 3ЈI1. The left portion of panel B shows the segments of intron 1 that were inserted upstream of the CYP7A1 promoter or the apoB promoter CAT constructs and used in the transfection experiments with HepG2 cells. In the last construct, intron 1 was placed between exon 1 and the CAT gene, to mimic its location in the natural gene. The mean values of CAT activities (representing the CAT activity of the test construct divided by the ␤-galactosidase activity of the control plasmid in each case), expressed as relative CAT activity, are shown on the right, Ϯ standard deviations. These transfections were performed three times (in duplicate). An Intronic MAR-bound Repressor in the Human CYP7A1 Gene 26652 gressively digested by DNase I; concomitant with this process, four DH sites emerged and were named DH1, -2, -3, and -4. DH1 maps in the promoter region and has been previously reported by this laboratory (16). The three new DH sites (DH2, -3, and -4) all mapped within intron 1, at the positions indicated at the bottom of Fig. 1. DH2 and DH4 were very strong, whereas DH3 was weaker. These DH sites were absent in nuclei from HeLa cells (Fig. 1, top right), indicating that it is the chromatin structure of this region in HepG2 cells (but not in HeLa cells) that makes these DNA sites open and available for interaction with nuclear proteins.
A Transcriptional Repressor Activity Is Associated with Intron 1 of the Human CYP7A1 Gene-Next we asked whether intron 1 may harbor a regulatory activity. Accordingly, a 1585-bp fragment containing intron 1 was inserted in both orientations upstream of a CYP7A1 promoter CAT gene con-struct (Ϫ764CAT) and tested for its ability to modulate the activity of the promoter in transient transfection assays with HepG2 cells (Fig. 2B). For simplicity, the CAT activity of the promoter alone was set at 1.0, and the data are presented as the mean value of 6 transfections Ϯ standard deviation. In its forward orientation, intron 1 (I1F) decreased transcriptional activity of the CYP7A1 promoter almost 10-fold, whereas in the reverse orientation (I1R), the reduction was less pronounced. Intron 1 was also cloned upstream of another liver-specific promoter, that from the human apolipoprotein B gene (apoB). Again, intron 1 reduced transcriptional activity of the apoB promoter in both orientations by about 5-fold. These data establish the presence of a transcriptional repressor within the first intron of the human CYP7A1 gene. In an attempt to further localize the repressor, we took advantage of the knowledge of the locations of DH sites 2, 3, and 4 and divided intron FIG. 5. Binding of CYP7A1 DNA fragments to nuclear matrix preparations from HepG2 cells. Panels A and B show representative autoradiograms from DNA/matrix binding experiments performed by the method of Cockerill and Garrard (18). Above the autoradiograms, we show the amounts (in g/ml) of E. coli DNA used as competitor and the restriction fragments used in these binding assays. The sizes of the DNA probes assayed for MAR binding are indicated at either side of the gels. Panel C illustrates a restriction map of the area; exons 1, 2, and 3 are shown as boxes; introns 1 and 2 are also shown, with the MAR-binding fragments shown above the map, and with the 32 P-labeled fragments tested for binding shown below the map.

FIG. 6. Localization of an in vivo
MAR in the human CYP7A1 gene. Panels A, B, and C show autoradiograms from Southern blots generated using the in vivo MAR or halo method (19) to localize MARs. Restriction enzymes used are indicated above the gels, as well as the sample source (supernatant (S), pellet (P), or total HepG2 DNA (T)). The sizes in bp of the restriction fragments studied are indicated at either side of the autoradiograms. Panel D is a restriction map of the area under study; exons 1, 2, and 3 are shown as boxes; introns 1 and 2 are also shown. Below the map we show the probes used in the Southern blots depicted above in panels A, B, and C. Fragments with MAR binding activity are displayed above the map. 1 into two segments; one containing its 5Ј end (5ЈI) and including DH2 and another 3Ј segment (3ЈI) containing DH sites 3 and 4. Fragments 5ЈI and 3ЈI ( Fig. 2A) were inserted upstream of the human CYP7A1 and apoB promoters and transfected into HepG2 cells. The results in Fig. 2B show that 5ЈI and the 3ЈI segments repressed both the CYP7A1 and the apoB promoter by about 3-5-fold. The 5ЈI1R construct repressed both promoters by about 3-fold. Finally, the 3ЈI1F segment also repressed both promoters between 3-and 5-fold. These data indicate that functional negative regulatory elements are present in each of the two segments of intron 1.
To further validate our results, we made a construct in which intron 1 was cloned in between exon 1 and the reporter gene, to mimic its natural location in the human CYP7A1 gene. In this construct, intron 1 repressed the activity of the reporter gene by 10-fold, just as it did when placed upstream of the promoter. From our data in Fig. 2, we conclude that the repressor effect of intron 1 is independent of position, orientation, and promoter context, reminiscent of a silencer.
The CCAAT Displacement Protein CDP Binds to Several Sites in Intron 1 of the Human CYP7A1 Gene-In an attempt to identify the repressor protein(s) involved in the activity of intron 1, we searched a transcription factor binding site data base for matches to sequences within intron 1 (21). We noticed several matches to the recognition sequence for the CCAAT displacement protein CDP. CDP belongs to a subgroup of a homeodomain family of DNA binding factors, the cut family, because its homeodomain resembles the cut homeodomain first identified in Drosophila (23). CDP has been implicated in several systems as a transcriptional repressor of developmentally regulated genes (10,24,25). For these reasons, the five sites representing the best matches to the CDP binding site and that were situated in the vicinity of the DNase I-hypersensitive sites (shown in Fig. 1A) were studied further. Oligonucleotides encompassing these sites were synthesized and tested for binding to CDP from a COS cell extract enriched in CDP protein in gel retardation assays. First, the ability of these five potential CDP oligomers to compete for binding of CDP to the high affinity CDP binding site from the gp91 phox gene promoter was evaluated (25). The gp91 phox probe formed the expected retarded complexes with CDP (Fig. 3B, lane 1); specificity of binding was evidenced by competition with a 100-fold excess of non-radioactive gp91 phox oligomer (lane 2). A CDP antibody supershifted the retarded complexes (lane 8), providing unequivocal evidence that the retarded complexes observed with gp91 phox were due to binding to this probe by CDP. As observed in lanes 3-7, sites 1, 2, 3, 4, and 5 from the human CYP7A1 gene competed well for binding of CDP to the consensus gp91 phox probe, providing the first evidence that these four sites indeed represent CDP-binding sites. This preliminary observation was confirmed directly using each one of the oligomers for sites 1-5 as probes in gel shifts with the CDPenriched extract (Fig. 4). In Fig. 4A, binding of CDP to sites 1 (lane 1) and 2 (lane 5) is demonstrated. Specificity of binding is shown in lanes 2 and 6, respectively. The gp91 phox oligomer competes well for binding of CDP to site 1 (lane 3) and site 2 (lane 7). Finally, a CDP antibody supershifts CDP complexes representing binding to sites 1 and 2 (lanes 4 and 8). A similar gel is shown in Fig. 4B, demonstrating binding of CDP to site 3 (lanes 1-4), site 4 (lanes 5-8), and site 5 (lanes 9 -12) of intron 1 of the CYP7A1 gene. The combined data in Fig. 4 demonstrate that sites 1-5 within intron 1 of the human CYP7A1 gene are bound by CDP.
Intron 1 of the Human CYP7A1 Gene Is Attached to the Nuclear Matrix-Because CDP has been reported to be a nuclear matrix protein, we asked whether intron 1 harbored nu-clear MARs. Two methods were employed to answer this question. In the first method (18), nuclear matrix proteins are first isolated and challenged with 32 P-labeled DNA segments in the presence of increasing amounts of E. coli DNA as a nonspecific competitor. After centrifugation to remove the unbound radioactive DNA, the labeled DNA is resolved on agarose gels. Fig.  5 illustrates our data with several matrix "bound" and "unbound" DNA segments of interest. The left portion of panel A shows that a 600-bp NdeI-StuI fragment (see map in lower panel C), contained entirely within intron 1, does bind strongly to matrix proteins, even in the presence of a large excess of E. coli DNA (300 g). Similarly, the 441-bp StuI-XbaI fragment contained entirely within intron 1 is also bound by nuclear matrix proteins (Fig. 5A, lanes 5-7), indicating that the MAR region probably extends 3Ј of the StuI site toward the XbaI site (see map in panel C). Finally, a 209-bp NotI-NdeI fragment from the 5Ј-most segment of intron 1 that does not contain exon 1 sequences, also binds to the nuclear matrix (Fig. 5A, lanes  8 -10). The combined data in Fig. 5A indicate that the MARs may extend over the entire length of intron 1.
Two segments of the CYP7A1 gene that were not bound to matrix proteins are shown in Fig. 5B. The first is a 306-bp segment located 3Ј of the 5Ј AluI repeat (lanes 1-4) where it is clear that binding was totally absent in the presence of 250 g of E. coli DNA competitor (lane 2). Additionally, a 359-bp fragment from the promoter region that is sufficient for maximal transcriptional activity in transient transfection assays with HepG2 cells (26), was not bound by matrix proteins (lanes 5-8) and therefore is not a part of the MAR.

An Intronic MAR-bound Repressor in the Human CYP7A1 Gene
The second method used to identify MARs involved gentle washing of HepG2 cell nuclei with lithium diodosalycylate, a mild detergent that removes histones and loosely bound nonhistone proteins and leaves the tighter associations between DNA and matrix proteins intact (19). The resulting structures, called "halos," are washed gently and extensively to remove the detergent and then digested with one or more restriction enzymes flanking the DNA segment of interest. After centrifugation, DNA sequences bound to the pellet (P) represent MARs, while those released by restriction enzymes into the supernatant (S) do not.
Using this method, we first examined the distal EcoRI-PstI segments that flank the AluI repeat and may represent potential MAR-containing regions (Fig. 6A), by analogy to previous observations with the human apoB gene, where an AluI region in the 5Ј end of this gene harbors a MAR (27). In two separate halo preparations shown, this region was recovered in the supernatant fraction, indicating that it does not contain a MAR.
Next, we examined a 2076-bp HindIII-XbaI fragment (Fig.  6B) and then a 1300-bp HindIII-StuI fragment (Fig. 6C). Two separate experiments in panel B clearly show the presence of the HindIII-XbaI DNA in the pellet, indicating it contains a MAR. Similarly, the 1300-bp HindIII-StuI fragment was also attached to the nuclear matrix (panel C). Our results, summarized in panel D, demonstrate the presence of an "in vivo" MAR within a 1300-bp fragment that maps almost entirely within intron 1 and that there is no apparent MAR at the 5Ј end of the human CYP7A1 gene. These data are in good agreement with the "in vitro" MAR data of Fig. 5.
The sequence of intron 1 is depicted in Fig. 7 (28). Within it, there are several A/T-rich sequence motifs that may participate in the interaction with the nuclear matrix, shown as white boxes in Fig. 7. For example, there are five copies of the se-quence ATATTT, which corresponds to the "invariable core" of the topoisomerase II recognition sequence (29) and has also been found in MARs from other genes (30). Additionally, a second sequence often found in MARs, AATATTTT (18), is also present three times in intron 1 (twice with only one base pair change) and a third sequence, AATAAATAAA (with one base pair mismatch), seen in the human apoB (and in other genes) MARs (27,30), appears twice within CYP7A1 intron 1. Some of the CDP binding sites (gray boxes) coincide with the A/T-rich motifs.
Binding of Liver-specific Activators to Intron 1-As its name suggests, the CAAT displacement protein CDP repressor effect involves displacement of transcriptional activators whose binding sites overlap with that of CDP (10). To gain insight into the molecular mechanisms involved in the repression of the CYP7A1 gene by CDP, we focused on CDP sites 1, 2, 3, and 4. This was based on the observation that CDP site 1 overlaps almost completely with a binding site for the liver-enriched transcription factor HNF-1␣ (Fig. 8B), while CDP sites 2 and 3 flank the binding site of another activator, C/EBP␣ (Fig. 8B).
A second C/EBP␣ site (C/EBP␣-2) also overlaps completely with CDP site 4 (Fig. 8C). First, we studied binding of the hepatic activators HNF-1␣ and C/EBP␣ to their three putative sites within intron 1. To this end, oligonucleotides corresponding to the HNF-1␣ and C/EBP␣ sites were synthesized and tested for their ability to bind HNF-1␣ and C/EBP␣, respectively. As a source of proteins, we used COS cell extracts prepared from cells transfected with HNF-1␣ and C/EBP␣ expression vectors, respectively. Fig. 9A shows specific binding of HNF-1␣ to its site within intron 1 (lanes 5-7). HNF-1␣ formed a similar retarded complex with the HNF-1/CDP probe as that formed with a probe representing the strong HNF-1 binding site from the ␤-fibrinogen promoter used as a control in lanes 1-3. Moreover, the   FIG. 8. Overlap between CDP binding sites 1, 2, 3, and 4  CYP7A1 oligomer competed for binding of HNF-1␣ to the high affinity consensus probe (lane 3) and the consensus oligomer also competed for binding of HNF-1 to the CYP7A1 sequence (lane 7). The HNF-1␣ complexes are supershifted by an HNF-1␣ antibody (lanes 4 and 8), thus conclusively demonstrating binding of HNF-1␣ to its site within intron 1. Fig. 9B illustrates binding by C/EBP␣ to its site in between CDP sites 2 and 3 (C/EBP␣-1) (Fig. 8B). The first four lanes show a control experiment where a high affinity C/EBP consensus oligomer is bound by C/EBP present in the COS cell extract (lane 1). Both the C/EBP␣-1 oligomer (lane 3) and the C/EBP consensus oligomer (lane 2) compete effectively for binding. Similarly, the C/EBP␣-1 oligomer forms retarded complexes with C/EBP␣ (lane 5), which are competed for by itself (lane 6) and by the consensus C/EBP oligomer (lane 7). The C/EBP␣ complexes were supershifted by C/EBP␣ antibodies (lanes 4 and 8), demonstrating binding of C/EBP␣ to the C/EBP␣-1 site. Binding to site C/EBP␣-2 is depicted in Fig. 9C. Again, lanes 1-4 show binding of the consensus C/EBP probe and lanes 5-8 provide evidence that site C/EBP␣-2 can indeed be bound by the hepatic activator C/EBP␣. Two major complexes containing C/EBP␣ were observed with both C/EBP probes as well as with the consensus probe. All of these were supershifted with C/EBP␣-specific antibodies. One of these complexes corresponds to C/EBP␣ homodimers, and the other reflects binding of heterodimers between C/EBP␣ and another member of the C/EBP family, present in the COS cell extract.
Displacement by CDP Leads to Repression of Transactivation-Having demonstrated binding by the liver-enriched activators to intron 1, we designed experiments to ascertain whether CDP was capable of displacing these activators from their corresponding binding sites. First, we examined the HNF-1␣/CDP (site 1) region. The oligonucleotide probe representing this region (seen in Fig. 9A, lanes 5-8), was incubated with a COS-HNF-1␣ extract (Fig. 10, lanes 6 -9) or a COS-CDP extract (lane 10). In the absence of COS-CDP extract as a competitor (lane 6), HNF-1␣, as expected, bound to the HNF-1/CDP oligomer. This binding was greatly diminished when a 0.5ϫ dilution of a COS-CDP extract relative to that of the HNF-1␣ extract, was added as a competitor (lane 7), and totally eliminated when equal amounts of protein were added from both the COS-HNF-1␣ and COS-CDP extracts (lane 8). Addition of 2ϫ CDP extract increased CDP binding even further (lane 9). This experiment clearly shows that CDP can displace HNF-1␣ from its site within intron 1. As a control, a similar experiment was performed using the HNF-1 consensus oligonucleotide as a probe (left portion of panel A). As clearly shown on lane 5 of Fig.  10, this probe does not bind to CDP and cannot be displaced by increasing amounts of CDP (lanes 2-4), but binds strongly to HNF-1␣ (lane 1). These data demonstrate that CDP can displace HNF-1␣ from its binding site in the CYP7A1 gene. Furthermore, this displacement is specific for the HNF-1␣ site within intron-1 because CDP failed to displace HNF-1␣ from binding to the HNF-1 consensus oligonucleotide.
To ascertain whether displacement of HNF-1␣ by CDP altered the function of this site within intron 1, we made a reporter construct in which the HNF-1/CDP site was inserted upstream of the minimal promoter from the human apoB gene (TATA CAT). This promoter segment is ideally suited to test for activator function because, by itself, it has a very low transcriptional activity and it does not contain any binding sites for hepatic activators. Therefore, any activation observed with this promoter is due to the upstream sequence inserted. Earlier, it was shown that the apoB promoter was repressed in a similar manner as was the CYP7A1 promoter by intron 1 (Fig. 2). Four copies of the HNF-1/CDP site were present in the TATA CAT construct we tested (Fig. 10B). Co-transfection of the TATA CAT plasmid with 5 g of HNF-1␣ expression vector did not affect the activity of the promoter (Fig. 10B, top). In contrast, the HNF-1/CDP site stimulated the activity of the TATA CAT construct by about 2-fold, when co-transfected with 5 g of the HNF-1␣ expression vector (Fig. 10B, bottom). Co-transfections with increasing amounts of the CDP expression plasmid caused a reduction in the transcriptional activity of HNF1/CDP TATA CAT construct (Fig. 10C), thus demonstrating that CDP represses activation by HNF-1␣ of the HNF-1/CDP site within intron 1.
To verify that the activation by HNF-1 and subsequent repression by CDP was also observed with the human CYP7A1 gene, a construct was made in which the HNF-1/CDP site was cloned upstream of the segment from Ϫ34 to ϩ46 (TATA CAT) of this gene. HNF-1␣ stimulated the activity of this construct by about 2-fold. Furthermore, CDP strongly repressed the transactivation by HNF-1 (Fig. 10D).
Next, we asked whether binding of CDP to sites 2 and 3 of CYP7A1 intron 1 could displace C/EBP␣ from its binding site (shown in Fig. 8), which is located in between the two CDP sites. To answer this question, three oligonucleotides were synthesized. The first contained CDP site 2 and the C/EBP site (2/C/EBP␣-1); the second, C/EBP␣ and CDP site 3 (C/EBP␣-1/ 3); and the third, all three sites (Fig. 11A). Oligomer 2/C/ EBP␣-1 was bound by C/EBP␣ (lane 1). Increasing levels of CDP extract progressively competed away the C/EBP␣/DNA interaction (lanes 2-4). Binding of C/EBP␣ to a C/EBP␣-1/site 3 probe was weak (lane 5); it was abolished with addition of 0.5ϫ CDP extract as a competitor (lane 6). Additional amounts of CDP extract further increased CDP binding (lanes 7 and 8). Finally, C/EBP␣ bound weakly to the triple-site probe (lane 9) and was readily displaced by increasing quantities of CDP extract (lanes 10 -12). Lane 13 shows the migration of the complexes formed by CDP binding to one or both of these sites, respectively. These data demonstrate that CDP displaces C/EBP␣ from this site within intron 1 of the CYP7A1 gene.
To test for the functional consequences of CDP displacement of C/EBP␣ binding to site C/EBP␣-1 within intron 1 of the human CYP7A1 gene, one copy of the CDP/C/EBP␣-1/CDP (2C3) site, was incorporated upstream of apoB TATA CAT (Fig.  11B). Co-transfection of this construct and the TATA CAT construct into HepG2 cells with 5 g of the C/EBP␣ expression vector stimulated transcription of the 2C3 construct by 22-fold, suggesting that this C/EBP␣ binding site within intron 1 represents an important activation site in the CYP7A1 gene. Activity of the TATA CAT construct was not affected by cotransfection with C/EBP␣. Addition of increasing amounts of a CDP expression vector to the transfection of the 2C3 construct revealed a sequential and progressive reduction of CAT activity (Fig. 11C), thus revealing that displacement by CDP of C/EBP␣ bound to 2C3 leads to repression. The transcriptional activation by C/EBP binding to 2C3 and subsequent repression by CDP was also observed when using the minimal promoter from the CYP7A1 gene (Fig. 11D).
Last, we tested the effect of CDP upon C/EBP␣ binding to its C/EBP␣-2 site, that totally overlaps with CDP site 4 (for illustration, see Fig. 8C). As a control, we examined binding of C/EBP␣ to its consensus oligonucleotide, both in the absence Therefore, we concluded from the data in Fig. 12 that CDP can also displace the activator C/EBP␣ bound to site C/EBP␣-2/CDP within intron 1. By analogy to the data in Figs. 10 and 11, we predict that displacement at site C/EBP␣-2/CDP will also lead to transcriptional repression. Thus, we have presented three examples of the mechanism by which CDP represses hepatic transcription of the CYP7A1 gene by targeting activators that bind to functional sites within intron 1.

DISCUSSION
To date, our understanding of the key elements conferring liver-specific expression to the CYP7A1 gene in humans and other mammals is incomplete. One difficulty has been the lack of suitable animal models in which to carry out in vivo studies. Transgenic mice have proven invaluable in the study of tissue specific regulatory regions of many human genes. However, they may not be a suitable model system to study regulation of the CYP7A1 gene. Earlier, we prepared several lines of mice carrying human CYP7A1 transgenes (31). It was concluded that the 5Ј region of the human gene (Ϫ1888 to ϩ46) was insufficient to drive hepatic expression of a reporter lacZ gene. Human transgenes containing the entire human CYP7A1 gene were expressed in the liver, but at levels considerably lower than those of the endogenous mouse CYP7A1 gene. These low expression levels could be attributable, at least in part, to a repressive effect exerted by the large bile salt pool in these mice (5 times larger per gram of liver than the human bile acid pool).
Thus, unable to rely on transgenic mice for our studies aimed at identifying key liver-specific elements of this gene, and knowing that the 5Ј segment including the promoter was not sufficient for liver expression, we searched for additional regulatory elements in other portions of the gene. DH studies revealed three DH sites within intron 1 of the human gene. Because DH sites reflect putative binding sites for regulatory proteins, we evaluated a potential role for intron 1 in the regulation of this gene by means of transfection studies with HepG2 cells. Intron 1 repressed the activity of the human CYP7A1 promoter as well as that of a second liver-specific gene, namely the apolipoprotein B promoter. Computer analysis of the DNA sequence of intron 1 revealed several putative binding sites for the transcriptional repressor CDP. CDP was shown to bind to five sites within intron 1. Because CDP is an integral component of the nuclear matrix, the association of intron of the CYP7A1 gene with the nuclear matrix was examined by two different methods. These studies revealed that intron 1 is anchored to the nuclear matrix in HepG2 cells throughout its entirety and therefore contains MARs.
MARs are DNA segments that are defined and identified by their ability to bind proteins in histone-depleted nuclei, generally termed nuclear matrices or scaffolds (18,19). MARs typically are A/T-rich elements harboring consensus cleavage sites for topoisomerase II; in general, their DNA sequences are not highly homologous. Within the CYP7A1 intronic MAR, several short A/T-rich sequence elements resembling the topoisomerase II recognition sequences were found (Fig. 7). MARs are found within genes (11,14,18,32,33) and in intergenic regions (19,27,30,34).
Recent evidence suggests that MARs may play a direct role in transcriptional regulation. Examples include the intronic MAR of the immunoglobulin gene, which is adjacent to a tissue-specific enhancer and is required for proper regulation of the gene during development (35), and the human apolipoprotein B gene MARs that are required for high level expression of human apoB transgenes in the livers of mice (36 -38).
The correlation between the MAR and the repressor activity in HepG2 cells is reminiscent of findings with the human apolipoprotein B gene, where in HepG2 cells, a negative regulatory region is flanked by the two MARs (5Ј distal and 5Ј proximal) in the segment from Ϫ5262 to Ϫ1801 of the human apoB gene (27,39). This same segment of the apoB gene that has a weak repressor activity in cultured hepatoma cells, functions as a positive regulatory region in transgenic animal experiments (36,38). Similar results have been reported in the T cell receptor B locus (40). A second example of colocalization between a MAR and a transcriptional repressor, this time within the HIV-1 LTR region, is provided by a nuclear matrix protein that binds a specific segment of the negative regulatory element of the HIV-1 LTR, and in doing so, inhibits binding of an important activator, NF-B, to an adjacent site (14).
Recent studies with the rat CYP7A1 gene have established a complex pattern of regulation during rat liver development (5). CYP7A1 mRNA is undetectable in 18-day-old fetal livers; it increases at birth, then decreases at day 4, and then exhibits its maximum level at day 22. By day 28, the adult levels of CYP7A1 are observed. The relatively low levels of CYP7A1 mRNA commonly observed in HepG2 cells may reflect the fact that this hepatoma cell line, like most cancer cells in culture, may exhibit a developmentally early or undifferentiated pattern of expression and that the repressor described here may play a role in maintaining relatively low levels of expression in these cells. In support of this hypothesis is the observation that CDP is abundant in undifferentiated cells and is down-regulated in differentiated epithelial cells (41). For example, CDP represses transcription of the human papilloma virus type 16 in undifferentiated epithelial cells and in HeLa cells, which are rich in CDP (42).
The discovery that CDP binds to multiple sites within intron 1 (Figs. 7 and 8) as well as the displacement of HNF-1␣ and C/EBP␣ by CDP (Figs. 11 and 12) provide a plausible mechanism for the observed repression. Early in liver development, An Intronic MAR-bound Repressor in the Human CYP7A1 Gene the availability of CDP for its sites within intron 1 would be high, and, thus, binding of the liver-specific activators would be blocked. Later in development, when CYP7A1 is needed, binding by the activators would occur.
Transcriptional repression during development of the sperm histone H2B through a similar displacement mechanism involving CDP has been established (10). Furthermore, CDP has been shown to repress transcription of the immunoglobulin heavy chain gene, by displacing the activator protein Bright, from its binding to an adjacent site within intron 1 of this gene (43). Additionally, displacement by CDP of the activator CP1 of the myelomonocytic specific gp91 phox gene promoter leads to transcriptional repression (24,25).
Two possible mechanisms may help explain how the displacement of hepatic activators by CDP may be modulated during liver development. The first mechanism invokes modulation in the level of histone acetylation within intron 1. It has been reported recently that transcriptional repression of the cystic fibrosis transmembrane conductance regulation gene, mediated by CDP, is associated with histone deacetylation. Furthermore, immunocomplexes of CDP possess an associated histone deacetylase activity (44). Additionally, chromatin deacetylation blocks binding of transcriptional activators (45).
The second mechanism predicts that interaction between SATB1 and CDP will influence repression by intron 1. SATB1 is a cell-specific MAR DNA-binding protein, predominantly expressed in thymocytes. SATB1 contains an atypical homeodomain and two cut-like repeats, as well as a MAR-binding domain. A direct interaction between CDP and SATB1 has been demonstrated (46). It has been postulated that this interaction between SATB1 and CDP may sequester CDP and prevent its binding to DNA, thus allowing binding of the appropriate transcriptional activators (47). This mechanism or a combination of the mechanisms discussed above may prevail in the case of the CYP7A1 intron 1.