Identification and Characterization of Transcriptional Regulatory Regions Associated with Expression of the Human Apolipoprotein E Gene*

Multiple cis-acting regulatory elements have been mapped within a 1-kilobase fragment spanning nucleotides -651 through +356 of the human apolipoprotein E gene using a transient expression system based on the chloramphenicol acetyltransferase gene as well as DNase I footprinting techniques. A 651-base pair 5’-flanking region of the human apolipoprotein E gene was capable of directing chloramphenicol acetyltransferase gene expression over a 48-fold range among the various cultured cell lines tested. Deletion analysis of this 651-base pair upstream region linked to either the chloramphenicol acetyltransferase gene or the intact apolipoprotein E structural sequences revealed at least three regulatory domains within the proximal 383 nucleotides. One of these domains contained a GC box transcriptional control element. Further analysis demonstrated that the other two domains contained enhan-cer-like activity. These enhancer-like elements were located within the nucleotides spanning -366 to -246 and -193 to -124. A third enhancer element was identified in the first intron, within nucleotides +44 to +262. Changing the distance of the three enhancer elements from the transcription start site and revers-ing their orientation did not significantly alter their effect on transcription rates. However, enhancer activity was mM CaC12. The amount of DNase I was adjusted empirically to obtain even digestion patterns, and the digestion was terminated by adding 100 pl of "stop solution" (100 mM Tris, pH 8.0, 100 mM NaCl, 1% sodium dodecyl sulfate, 10 mM EDTA, 100 pg of proteinase K/ml, and 50 pg of tRNA/ ml). The reactions were incubated at 37 "C for 15 min, extracted with an equal volume of phenokchloroform (l:l), and precipitated with 3 volumes of ethanol. The fragments were analyzed on 8% polyacryl-amide, 8 M urea sequencing gels (36). Materiak-[14C]Chloramphenicol (50-57 mCi/mmol) was purchased from Du Pont-New England Nuclear. All restriction endonu- cleases, Ba131 nuclease, T4 DNA ligase, linkers, and Klenow fragments of DNA polymerase I were purchased from either New England Biolabs or Boehringer Mannheim. Acetyl-coenzyme A (lithium salt) and o-nitrophenyl-#?-D-galactopyranoside were purchased from Phar- macia LKB Biotechnology Inc. and Sigma, respectively. Cell culture media and nutrients were obtained from GIBCO. Unless stated oth- erwise, procedures involving recombinant DNA, enzymes, and re-agents were used under the conditions recommended by the suppliers.

Multiple cis-acting regulatory elements have been mapped within a 1-kilobase fragment spanning nucleotides -651 through +356 of the human apolipoprotein E gene using a transient expression system based on the chloramphenicol acetyltransferase gene as well as DNase I footprinting techniques. A 651-base pair 5'flanking region of the human apolipoprotein E gene was capable of directing chloramphenicol acetyltransferase gene expression over a 48-fold range among the various cultured cell lines tested. Deletion analysis of this 651-base pair upstream region linked to either the chloramphenicol acetyltransferase gene or the intact apolipoprotein E structural sequences revealed at least three regulatory domains within the proximal 383 nucleotides. One of these domains contained a GC box transcriptional control element. Further analysis demonstrated that the other two domains contained enhancer-like activity. These enhancer-like elements were located within the nucleotides spanning -366 to -246 and -193 to -124. A third enhancer element was identified in the first intron, within nucleotides +44 to +262. Changing the distance of the three enhancer elements from the transcription start site and reversing their orientation did not significantly alter their effect on transcription rates. However, enhancer activity was influenced by the promoter and cell line that were used. DNase 1 footprinting assays showed that specific sequences within two of these elements (-193 to -124 and +44 to +262) bind proteins in nuclear extracts from HepG2 and Chinese hamster ovary cells. A protein footprint also was identified for a GC box element at nucleotides -59 to -45. Thus, control of apolipoprotein E gene expression is the result of a complex interaction of several different regulatory elements. *This work was supported in part by Grant HL37063 from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Human apolipoprotein (apo)' E is a major component of various lipoprotein classes in mammals (1-3). It is a singlechain polypeptide of 299 amino acids (M, = 34,000) (4), and it is synthesized mainly in the liver (5). Studies of its tissue distribution have shown that various extrahepatic tissues, including brain, adrenal, spleen, ovary, testis, and kidney, contain abundant levels of apoE mRNA (6-8). Mouse peritoneal macrophages (9) and human monocyte-derived macrophages (10) also have been shown to contain apoE mRNA. Related studies have demonstrated that transfection of the intact human apoE gene into various cultured mammalian cells of both hepatic and non-hepatic origin leads to production of a biologically active apoE protein (11).
Apolipoprotein E mediates the cellular uptake of cholesterol, and it serves as a ligand for the apoB,E (LDL) receptor (3,12,13). There are three major isoforms of apoE (E2, E3, and E4), which represent different alleles at a single genetic locus (14,15). These isoforms are the result of specific cysteine-arginine interchanges in the primary structure (1, 3).
The nucleotide sequences of the mRNAs for human (16,17), rat (18), and mouse apoE (19) are known. The human apoE gene has been mapped to chromosome 19 (20), and the nucleotide sequences of the gene for human apoE (20,21) and rat apoE (22) have been determined. Each sequence has four exons and three introns, and the human gene is 3597 nucleotides in length. Recently, it was found that the apoE gene is linked closely to the apoC-I gene (23).
The 5'-flanking region of the human apoE gene has several striking structural features (21); however, the function or specific roles of these sequence elements in apoE gene expression have not been determined. To understand the potential role of these elements in the transcriptional regulation of the apoE gene, we have examined the 5'-flanking region as well as the first intron of the apoE gene using transient expression systems involving chimeric apoE gene/chloramphenicol acetyltransferase (CAT) gene recombinants and DNase I footprinting assays. The identification and partial characterization of specific sequences that modulate apoE gene expression are reported here.  (-651 to +73) of the human apoE gene sequence was excised by BglII-Sac1 digestion of the genomic clone pUCE4 (21) and ligated into plasmid pLSl to produce pHAE-CAT1. This plasmid (pLS1) was derived from pTKl (same as pTEl in Ref. 24) by removal of the TK promoter (24) by BglII-Sac1 digestion. Panel B, a 1001-bp fragment of the human apoE gene was excised by BglII and ligated into the BglII site of the vector pTK1, which is located 617 bp upstream of the TK promoter, producing two plasmids, pHAEN-10 (correct orientation) and pHAEN-11 (reverse the plasmid vector pLSl was used. The pLSl vector was derived from pTEl (24) by removal of the thymidine kinase (TK) promoter. These vectors contain the bacterial (CAT) gene coding sequence, simian virus 40 (SV40) splice sites, polyadenylation signals, the ampicillin resistance gene, and the origin of replication from pBR322 (24). Details of the construction of the pTEl vector are described in Ref. 24. A BglII-Sac1 fragment from a cloned human apoE gene containing 651 base pairs (bp) of the 5'-flanking DNA, the 44-bp first exon, and 29 bp of the first intron was inserted into pLSl immediately upstream of the CAT gene to produce pHAE-CAT1 (Fig. L4). Nucleotide positions of apoE gene fragments refer to the previously determined sequence of the apoE gene (21).

EXPERIMENTAL PROCEDURES
Plasmids containing progressively shorter sequences at the 5' end were generated by Bal31 exonuclease treatments. Briefly, pHAE-CAT1 was digested with BglII and treated with BaL31 for different lengths of time. The Bal31 reactions were stopped by adding 20 mM EGTA, heating the solution to 65 "C for 15 min, and then precipitating with ethanol. The resulting fragments were ligated with BglII linkers (8-mer, New England Biolabs) under conditions recommended by the supplier. The ligation products, which had various lengths of the 5"flanking sequence, were excised by BglII-Sac1 digestion and purified on a 5% acrylamide gel and then were ligated into the pLSl vector. The precise end points of these deletion mutants were determined by DNA sequencing using the method of Maxam and Gilbert (25) or dideoxynucleotide chain termination (26).
For the analysis of the enhancer activity of regions of the human apoE gene, three CAT vectors were used pTKl (Fig. lB), pTKlO (same as pTKl but lacking the 600-bp pBR322 sequence between polylinker sites and the TK promoter), and pA,,CAT, (see Fig. 4). In both the pTKl and pTKlO plasmids, the expression of the CAT gene is under the control of the TK promoter of the herpes simplex virus. In pTKl (same as pTE1 in Ref. 24), fragments of the apoE gene were inserted into the BglII site located 600 bp upstream of the TK promoter, whereas in pTKlO the inserts were placed immediately 5' to the TK promoter (BglII site). A starting plasmid, pHAEN-10, was constructed by insertion of a 1-kilobase (kb) fragment of the apoE gene (-651 to +356) into pTKl in the normal transcriptional orientation relative to the TK promoter. Plasmid pHAEN-11 was constructed similarly, but the same fragment was inserted in the reverse orientation. Subfragments within the 1-kb insert of pHAEN-10 were generated by restriction endonuclease digestion and Bal31 treatment, followed by addition of the BglII linkers. These subfragments were th-n inserted into pTKl and pTKlO vectors in both orientations.
To compare the activity of the enhancer elements when positioned 5' or 3' of the CAT gene, the plasmid pA,,CAT, was used. The CAT gene in this plasmid was under the control of the enhancerless SV40 promoter (27). The pA,,CAT, vector was digested with BglII for the 5' end-positioned inserts and with BamHI for the 3' end-positioned inserts, respectively, and then was ligated to apoE gene fragments that contained BglII linkers at each end (see Fig. 4). The control plasmids, pSVTK-CAT and pSVE-CAT, were constructed by ligation of the SV40 enhancer segment (a 400-bp insert isolated from pSV2-CAT (28) by HindIII-PuuII digestion) into the polylinker region (HindIII-BglII sites) of pTKl and pHAE-CAT1, respectively. The SV40 enhancer segment in pSVTK-CAT and pSVE-CAT is situated in the antisense orientation against the CAT coding sequence.
Gene Transfer of Recombinant Plasmids-Twenty-four h before orientation). Panel C, a 5.2-kb apoE gene-containing fragment was excised by BamHI-Hind111 digestion of a 10.5-kb cloned fragment (11) and inserted into BamHI-Hind111 sites of pAT153 to produce pHEG-10. This plasmid was used for construction of the 5' deletions of the apoE gene as described under "Experimental Procedures." The resulting recombinants are shown in Fig. 2B (30), cells were subjected to a 15% glycerol shock for 2 min (31) and were incubated for another 20-24 h before being harvested. Maintenance of Tissue Culture-All cells except Chinese hamster ovary (CHO) cells were grown in Dulbecco's modified Eagle's medium. All media were supplemented with 10% fetal bovine serum, 100 units of penicillin/ml, and 100 pg of streptomycin/ml. Chinese hamster ovary cells were grown in F-12 and Dulbecco's modified Eagle's medium; Hela, L, BaGL, U-937, and 5774.2 cells were grown in Dulbecco's modified Eagle's medium with 10% fetal bovine serum; HepG2 and Fu5AH cells were grown in modified Eagle's medium with 10% fetal bovine serum; and CaCo-2 cells were grown in modified Eagle's medium with 15% fetal bovine serum.
Chloramphenicol Acetyltransferase and #?-Galactosidase Assays-Transfected cells (at about 80% confluency) were harvested approximately 36-40 h after transfection by scraping, washed twice with phosphate-buffered saline, and resuspended in 200 p1 of CAT assay buffer (0.25 M Tris, HCl, pH 7.8, 10% glycerol, and 5 mM dithiothreitol). Cell lysates were prepared by 3 cycles of freezing (-70 'C) and thawing (65 "C), incubation at 65 "C for 10 min, and centrifugation at 12,000 rpm for 15 min. Supernatant fluids were collected and tested for protein content, CAT activity, and #?-galactosidase activity. Protein was quantified by the method of Bradford (32) using bovine serum albumin as a standard. The CAT assay was carried out as described previously (28). Our standard assay contained 50-100 pg of protein, 0.2 pCi of [*4C]chloramphenicol (specific activity, 50-57 mCi/ mmol), and 5.0 mM acetyl-coenzyme A (lithium salt) in a final reaction volume of 200 pl. Reactions were allowed to proceed for 15 min to 2 h at 37 "C. Chloramphenicol acetylation was linear for up to 2 h under these conditions (more 5.0 mM acetyl-coenzyme A was added at 60 min). The CAT reaction products were extracted with 1 ml of ethyl acetate and chromatographed on silica gel thin-layer chromatography plates using ch1oroform:methanol (95:5, v/v) and then quantified by liquid scintillation spectrometry. The #?-galactosidase assay was performed using 50 pg of protein, 0.5 mM o-nitrophenyl-(3-D-galactopyranoside in 600 pl of total reaction volume as described (24).
Preparation and Analysis of RNA-Total cellular RNA was isolated and RNA dot blot analysis was performed (6). Aliquots of total RNA from cultured cells according to the procedure of Chirgwin et al. (33), (1.0, 2.0, and 3.0 pg) were adjusted to 3 pg of RNA/sample with yeast RNA, denatured in 1.0 M formaldehyde, 0.9 M NaCl, and 0.09 M sodium citrate at 60 "C for 15 min, and then loaded onto nitrocellulose filters under a mild vacuum as described (6). Primer Extension Analysis-Primer extension analysis was performed essentially as described (24,34). Radioactively labeled primer was prepared by a polynucleotide kinase reaction using [y3'P]ATP and a 20-mer single-stranded oligonucleotide primer corresponding to nucleotides 15-34 of the coding sequence of the CAT gene. Twenty ng of the 32P-labeled primer was hybridized to 20 pg of total cellular RNA and extended as described (24,34). The extension products were analyzed on 6% polyacrylamide, 8 M urea gels (25). The relative amounts of the primary transcripts on autoradiograms were measured by densitometric scanning.
Isolation and Preparation of Nuclear Extracts-Nuclear extracts from the human hepatoma cultured cell line HepG2 and from CHO cells were prepared essentially as described by Dignam et al. (35). Nuclei were suspended at a concentration of 2 X 10s/ml in a buffer consisting of 20 mM Hepes, pH 7.9, 25% glycerol, 0.42 M NaC1, 1.5 mM MgCl,, 0.2 mM EDTA, 0.5 mM dithiothreitol, and 0.5 mM phenylmethylsulfonyl fluoride. The mixture was stirred gently for 30 min at 0 "C and then centrifuged at 12,500 rpm for 30 min at 2 "C. The supernatant fluid was dialyzed against 250 volumes of dialysis buffer (20 mM Hepes, pH 7.9, 20% glycerol, 0.1 M KCl, 0.2 mM EDTA, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride) for 2 h with one change of buffer. The dialysates then were centrifuged at 12,000 rpm for 10 min at 4 "C, and the supernatant fluid was stored in aliquots at -80 "C. The protein concentration of the extract was determined by the method of Bradford (32) using bovine serum albumin as a standard.
DNase Z Footprinting-The footprinting reactions were performed in a 100-p1 volume in a buffer containing 10 mM Hepes, pH 7.9, 50 mM KCl, 0.1 mM EDTA, 0.25 mM dithiothreitol, 0.25 mM phenylmethylsulfonyl fluoride, and 10% glycerol (35,36). Various amounts of the nuclear extracts were preincubated with 1 pg of poly(d1-dC) in the reaction mixture for 20 min at room temperature. Then 5 ng of either 5' end-labeled (by T4 polynucleotide kinase) or 3' end-labeled (by Klenow DNA polymerase catalyzed fill-in reaction) DNA fragment was added, and the mixture was incubated for 40 min on ice.
DNAse I digestion was initiated by the addition of a MgCl,-CaCl, buffer to a concentration of 5 mM MgCl, and 2.5 mM CaC12. The amount of DNase I was adjusted empirically to obtain even digestion patterns, and the digestion was terminated by adding 100 pl of "stop solution" (100 mM Tris, pH 8.0, 100 mM NaCl, 1% sodium dodecyl sulfate, 10 mM EDTA, 100 pg of proteinase K/ml, and 50 pg of tRNA/ ml). The reactions were incubated at 37 "C for 15 min, extracted with an equal volume of phenokchloroform (l:l), and precipitated with 3 volumes of ethanol. The fragments were analyzed on 8% polyacrylamide, 8 M urea sequencing gels (36).
Materiak-[14C]Chloramphenicol (50-57 mCi/mmol) was purchased from Du Pont-New England Nuclear. All restriction endonucleases, Ba131 nuclease, T4 DNA ligase, linkers, and Klenow fragments of DNA polymerase I were purchased from either New England Biolabs or Boehringer Mannheim. Acetyl-coenzyme A (lithium salt) and o-nitrophenyl-#?-D-galactopyranoside were purchased from Pharmacia LKB Biotechnology Inc. and Sigma, respectively. Cell culture media and nutrients were obtained from GIBCO. Unless stated otherwise, procedures involving recombinant DNA, enzymes, and reagents were used under the conditions recommended by the suppliers.

RESULTS
Expression of Apolipoprotein EIChloramphenicol Acetyltransferase Hybrid Gene in Cultured Cells-It had been shown previously that the whole human apoE gene associated with 5.0 kb of the 5"flanking sequence and 1.6 kb of the 3"flanking sequence was sufficient for expression in cultured cell lines (11). To identify boundaries of the promoter region of the apoE gene, a fragment of the gene that contained proximal 5'-flanking sequences was inserted into a test vector at the 5' end of the bacterial CAT gene, which was employed as a general marker for promoter activity (Fig. lA). After the plasmid pHAE-CAT1 (which contained 651 bp of the 5'flanking region, the 44-bp first exon, and 29 bp of the first intron) was constructed and transfected into 10 different cultured mammalian cell lines, the level of CAT activity in cell extracts was measured. To correct for differences in DNA H u m a n Apolipoprotein E Gene Regulation 13343 A. The values in the columns on the right represent the means of results that were obtained from two separate experiments. The relative CAT activities of the different deletion constructs are expressed as a percentage of that of pHAE-CAT1 (-651 to +73). To obtain relative CAT activities, the specific activity of each construct was normalized to the @-galactosidase activity before comparison (i.e. normalized CAT activity = CAT specific activity/@-galactosidase activity). Panel B, deletion of portions of the 5"flanking region of the intact apoE gene. The gene is not drawn to scale. Chinese hamster ovary cells were transfected with 15 pg of each plasmid and processed for preparation of RNA and dot blot analysis as described under "Experimental Procedures." The RNA blot was hybridized with a :'*P-labeled full length human apoE cDNA probe (HinfI fragment, 1.1 kb, Ref. 17), and the blots were quantitated by densitometric scanning of the autoradiograms. uptake within a cell line, the cultured cells were cotransfected with a vector containing the RSV-directed &galactosidase gene (pRSV-&galactosidase) (24). The P-galactosidase activity in each cell extract was measured, and CAT activity (determined by apoE gene promoter effectiveness) was normalized to this value. To compensate for differences in transfection efficiency among cell lines, each cell line was transfected separately with another control vector containing the RSV promoter ligated to the CAT gene and cotransfected with pRSV-@-galactosidase. These genes were assayed as described above. The promoter activity of the apoE gene fragment then was determined relative to RSV promoter activity. The RSV promoter was chosen for both control vectors because it displays little cell specificity (28).
The results of this experiment are summarized in Table I, which shows the relative apoE promoter-driven CAT activities of the different cell lines. Overall, CAT expression over a 48-fold range was observed, with promoter activity being greatest in CHO cells. These results indicate that a CAT gene recombinant containing 651 bp of the 5"flanking sequences of the human apoE gene allows relatively efficient expression of the gene in CHO, HepG2, L, and CaCo-2 cells. In a separate control experiment, the apoE gene fragment inserted in the antisense orientation in a second test vector, pHAE-CAT10, showed no expression in any of the cell lines examined (data not shown). The CHO and HepG2 cells were used subsequently for more detailed analyses of the regulatory elements within the 5"flanking region of the human apoE gene. The CHO cells were employed because of their high apoE promoter activity, and the hepatoma-derived HepGP cells were used because of the normally high expression of the apoE gene in the liver.
Identification of the Sequences That Are Required for Gene Expression-To identify specific sequences in the 5"flanking region that are significant for human apoE gene expression, portions of the 5' ends of this region were progressively deleted by Bal31 nuclease treatment. The structures resulting from the deletions are outlined in Fig. 2 A , as are their activities. In both CHO and HepG2 cells, deletion of the 268 nucleotides between -651 and -383, which contains part of an AluI sequence, had little effect on apoE promoter-directed CAT activity. Deletion of the region between nucleotides -383 and -212 resulted in almost a %fold reduction in CAT activity. This region contains two types of directly repeated elements: 5'-TCCAGAT-3' (-355 to -349, -335 to -329, and -268 to -262) and 5'-CAGGAAAGGA-3' (-312 to -303 and -296 to -287). In addition, this region contains the hexanucleotide core sequences of an Spl protein binding "GC box" (-279 to -274) (36), but the identities of the neighboring nucleotides indicate that it may be a low affinity binding site (37). Deletion of the sequence between nucleotides -212 and -81 resulted in about a 4-fold reduction in CAT activity in both cell types. The removal of the region between nucleotides -81 and -39 resulted in another 2.5-4-fold reduction in promoter activity. Since this region contains two GC boxes, these results suggest that one or both GC boxes are functional components of the apoE promoter complex. Deletion of the region from nucleotides -39 to -14, which contains a consensus TATA box element, results in an additional 6-to 10-fold decrease in activity. In summary, the results from these deletion studies with the chimeric CAT gene suggest that, in addition to the TATA box in the proximal 383 nucleotides of the 5'-flanking region of human apoE gene, at least three different domains, possibly containing several elements, are involved in the regulation of its transcription.

w .
Deletion of the nucleotides between -3000 and -383 resulted in a slight increase in the expression of the apoE gene, which suggested that an upstream negative transcription element might have been removed. However, when fragments from this region were tested subsequently with a test vector that was employed for enhancer assays (described below), a negative effect on gene expression could not be demonstrated (data not shown).

Identification of Apolipoprotein E Gene Enhancer Elements-To determine whether regulatory elements in the 5'-
flanking region of the apoE gene, as well as sequences in the intragenic regions of the apoE gene, can function as transcription enhancers, a series of recombinants in the vectors pTK1 and pTKlO were constructed (see "Experimental Procedures"). In these vectors, the expression of the CAT gene is under the control of the TK promoter of the herpes simplex virus, which is active in most cultured cells (24). The use of these two vector systems enabled us to determine the effects of putative enhancer sequences on the TK promoter located either adjacent to (pTK10) or 600 bp distant from (pTK1) the putative enhancer. In addition, several fragments were inserted in either orientation relative to the direction of transcription to examine the effect of orientation on the potential enhancer fragments.
First, we examined the CAT activity of plasmids pHAEN-10 and pHAEN-11, which have a I-kb insert spanning nucleotides -651 and +356 (including the first exon and 312 nucleotides of the first intron) ligated into vector pTK1. In both CHO and HepG2 cells this fragment stimulated the activity of the TK promoter by about 5-fold in both orientations relative to the control pTKl vector (Fig. 3). Then, to delineate enhancer sequences in this fragment, a series of smaller fragments from this region was tested. The results shown in Fig. 3 suggest that there are three elements with enhancer-like properties located within this 1-kb fragment. One element, termed upstream regulatory element 2 (URE2), is located between residues -366 and -246. Another element, termed upstream regulatory element 1 (UREl), was found in a fragment located between residues -246 and -81; UREl is located within a 69-bp segment between residues -193 and -124 of this fragment. The third element, termed intron regulatory element 1 (IREl), was found within the first intron and is located in a 219-bp segment between residues +44 and +262. These three elements had promoter-enhancing activity in both orientations. Other 5"flanking and intragenic regions, including distal upstream sequences (-3000 to -651), portions of the first and second introns (+561 to +1439), and the third intron (+1852 to +2907), were tested, but they failed to show any enhancer activity in these assays (data not shown).
Although enhancers can often act over very large distances, there are many examples of enhancers that exhibit substantially reduced activity when separated from the test promoter (24,(38)(39)(40)(41)(42). To examine the effect of distance on the activities of the apoE enhancer elements, the appropriate fragments were ligated into plasmid pTKlO in a location immediately proximal to the TK promoter (ie. there was no 600-nucleotide spacer). The observed enhancement was essentially identical to that observed when the spacer was present (data not shown), suggesting that the action of these enhancer-like elements is relatively independent of distance under these conditions.
To determine whether these enhancer-like elements can act when positioned at the 3' end of the CAT gene and with a different promoter, another test vector (pAIOCAT2), which contained an enhancerless SV40 promoter (27), was employed. The three putative enhancer elements, URE1, URE2, and IRE1, were inserted either immediately 5' to the enhancerless SV40 promoter or at the 3' end of the CAT gene and transfected into CHO cells. The results shown in Fig. 4 indicate that UREl was able to enhance SV40 promoter activity from both the 5' and the 3' positions. IRE1 also exhibited enhancer activity for the SV40 promoter, although the effect from the 3' position (about a 2-fold enhancement) was about half that from the 5' position. URE2 possessed little activity at either position, suggesting that it may be influenced by the associated promoter, perhaps through specific interactions of trans-acting factors. Thus, it was concluded that URE2 has conditional enhancer-like activity.
Because the cloning of many of these inserts into the CAT vectors involved adding BglII linkers on either end, we examined whether the synthetic linkers themselves had an effect on CAT gene activity. Fragments from the second intron of the apoE gene as well as 300-and 600-bp fragments from bacteriophage DNA (ie. 4X 174 RF DNA) were prepared and ligated with the BglII linkers and then inserted into the same CAT gene vectors. None of these control fragments exhibited enhancer activity or any other effect on the transcription of the CAT gene (data not shown).

Enhancer activity of apolipoprotein E gene fragments in different cultured cell lines
Results represent the mean f S.D. from four independent experiments that were corrected individually for B-ealactosidase activitv. The auoE gene fraements were inserted in uTK1. Primer Extension Analysis of Chloramphenicol Acetyltransferase Gene Recombinants-To verify that the products of the CAT gene reflected correctly initiated mRNA transcripts and that CAT activity measurements in this system do indeed reflect events at the RNA level, total cellular RNA from cells that were transfected with the CAT gene constructs were examined by primer extension analysis. Total RNA from CHO cells transfected with the promoter test vectors pHAE-CAT1 (containing 651 nucleotides of proximal 5'-flanking sequence of the apoE gene) or pSVE-CAT (SV40 enhancer in pHAE-CAT1 located 600 bp upstream from the promoter) produced a prominent band 150 nucleotides in length (79 nucleotides of the apoE gene sequence and 71 nucleotides of the CAT gene sequence) (lanes 1 and 2 in Fig. 5), corresponding to the length predicted for the correct RNA transcription initiation site. The same results were obtained when a shorter promoter fragment, containing 81 nucleotides of the apoE

5' Flanking Region
gene 5-flanking sequence, was examined (data not shown). The SV40 enhancer stimulated the RNA transcription signal by greater than nine times (compare lanes 1 and 2 in Fig. 5), facilitating detection without influencing the initiation site. The enhancer test vector, pHAEN-2, which contains UREl 600 bp upstream of the TK promoter, generated a 134-bp transcript (71 nucleotides of the CAT gene sequence and 63 nucleotides of the TK gene sequence) that also indicated correct transcription initiation (lane 6 in Fig. 5) (24). Furthermore, the relative intensity of the transcript band produced by DNA-containing UREl is about 6-fold stronger than that produced by the TK promoter itself (compare lanes 5 and 6 in Fig. 5). This finding is consistent with the enhancer test results obtained from the CAT enzyme activity assay (Fig. 3).
Cell-type Specificity of the Apolipoprotein E Enhancers-An important property of some viral and cellular enhancers is that they often exhibit host, or cell-type, specificity (24,35, 39-42). To test whether the apoE enhancer sequences also exhibit cell-type specificity, the TK promoter/CAT constructs containing regulatory elements URE1, URE2, and IRE1, as well as the DNA segment spanning -81 to -15 as a negative control, were transfected individually into four different cultured cell lines (Table 11). A control plasmid, pSVTK-CAT, which contained the relatively strong SV40 enhancer located 600 bp upstream from the TK promoter, also was tested. The UREl and URE2 segments were active in all cell lines tested. The IRE1 element showed equivalent activity in CHO and HepG2 cells, no activity in HeLa cells, and little activity in L cells. Thus, the three apoE gene enhancer-like elements showed variations in their activity among different cultured cells. However, the biological roles of these elements remain to be determined. It is noteworthy that the ratio between the magnitude of stimulation by the apoE UREl domain enhancer and the SV40 enhancer in each cell line remained relatively constant among all four cell lines (3-4-fold). In this regard, the activities of the other apoE enhancers showed much greater variation in the different cell lines. In addition, the UREl-containing fragment was able to stimulate TK promoter activity more than 4-fold overall in several other cell lines tested (data not shown).

DNase I Footprinting of the Apolipoprotein E Enhancer
Regions-To test the possibility that nuclear protein factors may interact with the apoE gene regulatory elements that were identified by the above studies, nuclear extracts prepared from HepG2 cells were assayed for the presence of sequencespecific binding activities by using the DNase I footprinting

TABLE I11 Sequence homology of portions of the protein binding regions in UREl and IRE1 to SRE-42 (repeat 3) and the GC box consensus sequence
' , indicates nucleotides that are identical in the corresponding positions.

' -A G G A G C G G G G G T -3 '
* * * * * * technique (43). A DNA fragment spanning UREl (nucleotides -365 to -15) was labeled at position -15 on the 5' strand and incubated with two different concentrations of nuclear extract. It was then subjected to partial cleavage with pancreatic DNase I. A sequence-specific protected region was observed between nucleotides -165 and -144 (Fig. 6A), which lies within the UREl enhancer-containing fragment (-193 to -124). When the corresponding 3' strand was end-labeled and tested for DNase I protein footprinting, a similar protected sequence (-160 to -144) was observed (Fig. 6A). Also protected from DNase I digestion on the 3' strand were nucleotides -59 to -45, where the proximal GC box element is located (Fig. 6A). Tjian and co-workers (37,44,45) have demonstrated that a GC box element is recognized specifically by the transcription factor Spl. Nucleotides around the distal GC box (core hexanucleotide at -74 to -69) were not protected from DNase I digestion.
DNase I footprints were also observed when the same DNA fragment (-365 to -15) was incubated with CHO nuclear extracts. Within URE1, nucleotides -161 to -141 on the 5' strand and nucleotides -160 to -144 on the 3' strand were found to be protected from DNase I digestion (Fig. 6B). Thus, nuclear proteins from HepG2 and CHO cells appear to bind to the same sequence within URE1. Also, binding of Spl (or an Spl-like protein) to the proximal GC box (nucleotides -59 to -45) was observed on both strands (Fig. 6B).
DNA fragments spanning IREl and UREZ also were examined for DNaseI-resistant protein footprints.
A weakly protected region spanning nucleotides +166 to +195 in the IREl domain was detected using both HepG2 and CHO nuclear extracts (data not shown). However, titration experiments with additional nuclear extract did not improve the clarity of the footprint. Thus, the protected region in IREl may be bound by proteins with low affinities and/or relatively low concentrations in the nuclear extracts, or they may require associated factors dependent upon metabolic conditions to mediate binding. The sequences within UREZ did not reveal any protein binding regions with the crude HepG2 and CHO nuclear extracts under the conditions employed.

DISCUSSION
The location of cis-acting regulatory regions within the human apoE gene has been investigated. The results suggest that the proximal 383 nucleotides of the 5"flanking sequence permit apoE gene transcription in several cell lines. It was surprising to find that CHO cells were the most active among those tested for the activity of the 651-nucleotide promoter fragment since CHO cells do not normally express the endogenous apoE gene. However, the promoter fragment may have been active because of the elimination of either upstream or downstream negative control elements. In this regard, examination of various 5'-and 3"proximal and distal apoE gene fragments with the pTKl vector did not demonstrate negative elements in this test system. Further studies with alternative test vectors may be required to determine additional control elements.
In promoter deletion studies, there is a good correlation between the results obtained from the CAT enzyme assay using the expression of the chimeric apoE promoter-CAT gene and the measurement of cellular RNA levels obtained from transfection of the intact apoE genes containing various lengths of the 5"flanking region. These results (Fig. 2, A and B) suggest that the region between -383 and -15 contains at least three domains, possibly with multiple elements, that affect apoE gene expression. An additional regulatory domain was found subsequently in the first intron. From the analysis of various fragments in these regions, a 166-bp fragment between -246 and -81 was found to have relatively strong enhancer activity. The level of basal expression of control vectors was enhanced by this DNA sequence up to nearly 10fold in HepG2 and CHO cells. Much of this enhancer activity was found within a 69-bp fragment URE1). Furthermore, this upstream regulatory element (URE1) acted in a distance-, orientation-, and position-independent manner, characteristics seen in many viral enhancers (24, 46-50). The UREl-containing DNA sequence was active in all cells tested. This lack of cell-type specificity of UREl may correlate with the wide tissue distribution of apoE mRNA in various tissues of mammals (6). In contrast to URE1, URE2 and IREl exhibited partial cell-type specificity (IRE1) and position or promoter dependence (UREZ).
Although the general mechanism of enhancer action is not clear, it appears to be coupled with the binding of specific protein factors in the nuclei of cells in which they are active (24,39,51,52). The results presented here show that nuclear proteins bind specifically to a region within URE1. The sequence of the UREl protein binding region (Fig. 7) contains inverted repeated sequences (-164 to -159, -152 to -147, ~'-L$CCTCTATGCCCCACCTC~TI'C-~'). It is not clear whether this feature plays a role in these seuuences being recognized by the proiein-factor(s).
The protein binding region within UREl appears to be associated directly with the enhancer activity of this domain. A synthetic oligonucleotide of 30 nucleotides spanning region -169 to -140 and containing the UREl footprint sequence acted as an enhancer when examined with the pTKl test vector (data not shown). However, the 30-mer was only 60% as active as the UREl-containing fragment (-246 to -81). This finding suggests that the enhancer activity of this domain may be the result of complex interactions between adjacent sequences, not just a simple protein-DNA binding interaction.
Both HepG2 and CHO cells showed a prominent protein footprint in the same portion of URE1. Thus, this region of Regulation the apoE promoter may contribute significantly to its activity in these two diverse cell types. However, the lack of expression by the endogenous apoE gene in CHO cells suggests that additional protein factors and sequences are involved in gene expression.
A prominent protein footprint (-59 to -45) was found at the proximal GC box, indicating that the GC box is likely to be active in stimulating apoE gene transcription. In contrast, no protein binding was observed for the two distal GC boxes (-77 to -68 and -280 to -271), suggesting that they may not play a role in apoE expression. This probability correlates with the observation that the nucleotides surrounding the hexamer GC core would influence these GC boxes to be weak binding sites for Spl (53).
There is a striking homology between parts of the protein binding domains of UREl and IREl and the Spl binding sequence (Table 111). Nine nucleotides in 12-nucleotide sequences from UREl (-157 to -146) and IREl (+169 to +180) are identical, and 8 of 10 nucleotide positions in UREl and 7 of 10 positions in IREl are identical to the corresponding positions of the Spl-binding GC box (10 nucleotides) consensus sequence (37). Nevertheless, it is unlikely that Spl binding by itself is responsible for the enhancer activities of UREl and IRE1, for two reasons. First, there is no evidence that GC box sequences alone have enhancer activity in any viral or cellular genes. In the case of the apoE gene, the fragment containing the proximal GC box (-59 to -45), whose footprint as seen in Fig. 6 suggests a high affinity for Spl binding, did not stimulate TK promoter-directed CAT activity when placed in enhancer test vectors (Fig. 3). Second, the protein binding regions of UREl and IREl are much longer than observed for the proximal GC box, indicating that other proteins in addition to Spl-like proteins may be bound to these two regions. However, there may be an interaction between Spl, which might be bound to these GC box-like sequences, and adjacent UREl-specific and IREl-specific binding proteins that are directly responsible for enhancer activity.
The identity between 9 of 12 nucleotide positions in portions of UREl and IREl and repeat 3 of the sterol regulatory region (SRE-42) of the LDL receptor gene (53, 54) is also interesting. The SRE-42 element in the 5'-flanking region of the LDL receptor gene is the cholesterol-responsive sequence associated with the down regulation of the LDL receptor gene. It is surprising to find a homologous component of part of the SRE-42 sequence in the apoE gene, since increased intracellular cholesterol levels have been associated with increased levels of apoE mRNA (55, 56). Perhaps the activities of these sequences are modulated by closely related proteins that bring about diverse effects through interactions with additional unrelated proteins and by different affinities of DNA binding. It is also possible that the repeat 3 element in the SRE-42 domain is not involved directly in the cholesterol responsiveness of the LDL receptor gene. It may be that repeat 3 of SRE-42 and sequences of UREl and IREl bind a protein in common that modulates the overall activity of these regulatory domains. In this regard, the Spl binding sequence (Table  111) suggests that an Spl-like protein might be involved in the activities of these regulatory domains.