Identification and characterization of two enhancers of the human albumin gene.

A 12.5-kilobase pair (kb) segment upstream of the human albumin gene was analyzed for transcription enhancing activity using transient transfection analysis, gel mobility shift assays, DNase I footprinting, and site-specific mutagenesis. Two enhancer regions were identified, one 1.7 kb upstream of the transcription initiation site (E1.7) and the other 6 kb upstream (E6). In E1.7, a nuclear protein from HuH-7 hepatoma cells binds to an AT-rich sequence, GTTACTAATTGAC. Competition gel mobility shift assays suggested that this protein is HNF-1, which regulates the promoter of the albumin gene and several other liver-specific genes. A 60-base pair E1.7 fragment carrying the AT-rich sequence stimulates a heterologous (alpha-fetoprotein) promoter in a dose-dependent manner. In E6, a HuH-7 nuclear protein binds to a GT-rich sequence, TGTTTGGC.A 27-base pair E6 fragment carrying this sequence is able to stimulate the SV40 promoter in an orientation-independent manner. An alteration of this sequence by site-specific mutagenesis resulted in the loss of transcriptional activity as well as binding to the HuH-7 nuclear protein. Competition gel mobility shift assays showed that homologous elements exist in the albumin promoter. These results show that the promoter and enhancer of the human albumin gene are regulated by two common transcription factors through two shared cis-acting elements, one AT-rich and the other GT-rich.

Studies of transcription of the mouse and rat albumin genes in uivo and in vitro have shown that the 170-bp' region immediately upstream of the transcription initiation site is sufficient for tissue-specific expression of the albumin gene (Ott et al., 1984;Gorski et al., 1986;Cereghini et al., 1987;Heard et al., 1987;Babiss et al., 1987;Izban and Papaconstan-* This work was supported by the National Cancer Institute of Canada, the Medical Research Council of Canada, and Japan Immunoresearch Laboratories Co., Ltd. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. The nucleotide sequence(s) reported in thispaper has been submitted M92816 and M93105. The abbreviations used are: bp, base pair(s); kb, kilobase pair(s); CAT, chloramphenicol acetyltransferase; AFP, a-fetoprotein; HEPES, N-2-hydroxyethylpiperazine-N'2-ethanesulfonic acid. tinou, 1989;Maire et al., 1989). At least six cis-acting elements have been identified in this region that interact with various transcription factors including HNF-1 (LF-B1, PAF, AFPl), C/EBP, DBP, CTF/NFl, and NFY (Lichtsteiner et al., 1987;Johnson, 1990;Tronche et al., 1990). The corresponding region of the human albumin gene shows a 90% sequence identify and therefore is likely to be regulated by the same DNA-binding proteins.
Enhancers of the rodent albumin genes have been shown to exist far upstream (-10.5 to -8.5 kb) of the transcription initiation site (Pinkert et aL, 1987;Herbst et al., 1989). In contrast, the human albumin enhancer identified so far has been mapped adjacent to the promoter (-486 to -221 bp) (Frain et al., 1990). This difference in the position of the enhancers between the rodent and human albumin genes is at odds with the striking similarity of their promoters. In this study, we examined whether additional enhancers exist further upstream of the human albumin gene. We report here the localization of two enhancers in upstream regions (-1.7 and -6 kb) and the identification of a nucleotide element responsible for the enhancer activity in each region. These elements are also found in the albumin promoter, indicating that two common factors participate in the regulation of the enhancer and promoter of the human albumin gene.

EXPERIMENTAL PROCEDURES
Cell Cultures-The human hepatoma cell line HUH-7 was maintained in a chemically defined medium, IS-RPMI (Nakabayashi et al., 1984). HeLa cells were grown in IS-RPMI containing 5% fetal calf serum.
Five CAT plasmids carrying progressively shorter albumin 5'-pAL1.9-CAT, pAL1.8-CAT, pALl.G-CAT, and pAL1.5-CAT) were flanking sequences from the HindIII site at -3 kb (pAL2.6-CAT, prepared as follows. pAL3.0-CAT was linearized at the ClaI site, and the ends were filled in by the Klenow fragment of DNA polymerase I and then treated with exonuclease 111. Aliquots of the reaction mixture were removed at several time points, treated with mungbean nuclease to form blunt ends, and digested with SstI. The smaller fragments released were ligated to pALO.3-CAT prepared from pAL3.0-CAT by digestion with ClaI and SstI to remove the -3 to -1.1-kb albumin 5"flanking DNA. CAT plasmids containing the 169-bp AFP promoter and one or three copies of (enhancer at -1.7 kb; see below) were constructed as follows. pAFO.l7(Bg)-CAT (Nakabayashi et al., 1989) was cleaved by EglII and blunt-ended by the Klenow fragment of DNA polymerase I. One or three copies of the 60-bp fragment (-1796 to -1737 bp) of E1.7 were ligated to form pAF0.17[E,.T],-CAT or pAF0.17[E1.7]3-CAT, respectively. pSV[Es]N-CAT and pSV[E6]R-CAT, in which E6 (enhancer at -6 kb; see below) is linked to the SV40 promoter in normal and reverse orientations, respectively, were constructed as follows. E6 was treated with the Klenow fragment of DNA polymerase I to form blunt ends, and EglII linkers were attached and inserted into the EglII site of pSV1'-CAT in normal and reverse orientations.
Cell Transfection and CAT Assays-HUH-7 cells were transfected with 20 pg of plasmid DNA/75-cm2 flask using the calcium phosphate precipitation method (Graham and van der Eb, 1973). Two days later, cells were harvested and lysed by five cycles of freezing and thawing. The lysate was heated at 60 "C for 10 min and centrifuged at 15,000 rpm for 5 min, and the supernatant was removed to measure CAT activity as described previously (Watanabe et al., 1987).
Gel Mobility Shift Assays-Partial purification of nuclear extracts and gel mobility shift assays were conducted as described previously (Sawadaishi et al., 1988). The nuclear extract was preincubated with 5 pg of poly(d1-dC).poly(dI-dC) for 10 min on ice and then incubated with 5 fmol of DNA end-labeled with 32P for 20 min on ice in 10 mM Tris-HC1 (pH 7.5), 45 mM KCl, 2 mM MgC12, 1 mM EDTA, 1 mM dithiothreitol, and 8% glycerol. The reaction mixture was electrophoresed on a 6% polyacrylamide gel containing 6.7 mM Tris-HC1 (pH 7.5), 3.3 mM sodium acetate, and 1 mM EDTA. The gel was dried and autoradiographed at -70 "C.
For competition experiments, a 100-or 200-fold molar excess of competitor DNA was preincubated with nuclear extracts for 5 min before the addition of end-labeled DNA.
This mixture was incubated on ice for 30 min and then at 25 "C for 1 min with 0.5 pg/ml DNase I. The reaction was stopped by the addition of EDTA (final concentration, 15 mM). DNA was extracted with phenol/chloroform/isoamyl alcohol (25:24:1) and electrophoresed on an 8% polyacrylamide sequencing gel.

Localization of Two Enhancer
Regions between -10.8 and -1.1 kb of Human Albumin Gene- Frain et al. (1990) have reported the presence of an enhancer at -486 to -221 bp of the human albumin gene that stimulates the albumin promoter -5-fold in Hep3B cells. In our assays using HUH-7 cells, however, this region exhibited little enhancer activity, possibly reflecting the difference in cell lines used. T o examine whether additional enhancer activities are associated with further upstream regions, transfection assays using the CAT reporter gene were conducted in HUH-7 cells. Initial analysis showed that there was no enhancer activity in the region between -1.1 and 0.5 kb. To quickly detect enhancer activity that might be present upstream of -1.1 kb, we linked the CAT gene to 12.5 kb of the 5"flanking sequence with or without internal deletions between -10.8 and -1.1 kb (Fig.   1A). Analysis of CAT expression in HUH-7 cells showed that the intact 12.5-kb fragment supported 10-20-fold higher CAT expression than the 1.1-kb fragment (Fig. lB, cf. lanes 1 and 5). The deletion from -10.8 to -5 kb (pALlP[Al]-CAT) resulted in an -50% decrease in transcription stimulatory activity (Fig lB, cf. lanes 1 and 2 ) . The deletion from -5.1 to -1.1 kb (pAL12 [A2]-CAT) also resulted in an -50% decrease in CAT activity (Fig. lB, lane 3). The deletion from -10.8 to -1.1 kb (pAL12[A3]-CAT) resulted in the complete loss of transcription stimulatory activity (Fig. lB, lane 4 ) . These results suggest that at least two enhancer regions exist between -10.8 and -1.1 kb, one between -10.8 and -5.1 kb and the other between -5.1 and -1.1 kb.
Delimitation of Enhancer Activity Present between -5.1 and -1.1 kb-To determine whether the 5'-half of this region contains the enhancer activity, we compared CAT activities supported by 3 and 5 kb of the albumin 5'-flanking DNA. No significant differences were observed, indicating that the enhancer is not present between -5 and -3 kb (Fig. 2). To localize the enhancer activity between -3 and -1.1 kb, we constructed CAT plasmids containing progressively shorter fragments from -3 kb and analyzed for their ability to support CAT expression. No significant changes in CAT activity were observed by reducing the size from 3 to 1.8 kb (Fig. 2). However, further deletion of -200 bp from -1.8 kb (-1867 to -1647 bp) (pAL1.6-CAT) resulted in a 60% decrease in CAT activity. Additional reduction in size to 1.1 kb had little effect on CAT activity. These results show that an enhancer is present between -1867 and -1647 bp. This enhancer region is referred to as E1.7. The nucleotide sequence of is shown in Fig. 3A.
Identification of Enhancer Element in E1.7-TO determine a regulatory element(s) responsible for the enhancer activity of E1.7, we conducted DNase I footprint analysis. The results showed that HUH-7 nuclear proteins bind to a 27-bp region from -1796 to -1770 bp in E1.7 (Fig. 4). The sequence 5'-TTGTTACTAATTGACAA-3' contained in the protection region is similar to the HNF-1-binding site that is present in the albumin promoter (Table I) (Courtois et al., 1987(Courtois et al., , 1988Hardon et al., 1988;Sawadaishi et al., 1988). To test whether HNF-1 in fact binds to the E1.7 element, we conducted competition gel mobility shift assays using an albumin promoter fragment (-90 to +17 bp) carrying the HNF-1 site as a competitor. The binding of protein to E1.7 was effectively prevented by this competitor (Fig. 5). A distal albumin promoter fragment (-275 to -91 bp) that lacks the HNF-1binding site had no effect.
To test whether the HNF-1 binding sequence in E1.7 has transcription stimulatory activity, we inserted one or three copies of the 60-bp fragment carrying this element (-1796 to -1737 bp) 5' to the 169-bp AFP promoter that is fused to the CAT gene (Fig. 6A). In HUH-7 cells, one and three copies of the enhancer fragment led to 2-and 5-fold increases in CAT activity, respectively (Fig. 6B), whereas no significant CAT expression was observed in HeLa cells (data not shown).
These results show that the HNF-1 binding sequence is responsible for the enhancer activity of E1.7.
Delimitation of Enhancer Activity Present between -10.8 and -5.1 kb-To localize the enhancer activity associated with the 5.7-kb region from -10.8 to -5.1 kb, we digested this DNA with Ssp1 to yield five fragments of 1700, 2100, 520, 757, and 640 bp (Fig. 7). In transfection analysis, only the 757-bp fragment was found to be active in stimulating the albumin promoter. The 757-bp fragment was then divided into two fragments (497 and 260 bp) by HpaI digestion, and each fragment was tested for enhancer activity. The 497-bp fragment supported CAT expression to the same level as the 757-bp fragment (Fig. 8A, lune 5) delimit the enhancer activity, the 497-bp fragment was digested with Hind111 and HincII to yield three fragments of 174, 242, and 81 bp (Fig. 7). The enhancer activity was found to be associated with the 81-bp fragment (Fig. 8B). The 81bp fragment was also able to stimulate the SV40 early promoter in either normal or reverse orientation in HUH-7 cells (Fig. 9, A and B ) . In HeLa cells, on the other hand, no enhancer activity was observed (Fig. 9C), indicating that the element contained in the 81-bp fragment is cell-specific. This enhancer region is termed Es, and its nucleotide sequence is shown in Fig. 3B.
Identification of Enhancer Element in Es-Inspection of the nucleotide sequence of E6 revealed the presence of a sequence (GCCAAACA) that is the reverse complement of a proteinbinding site (5"TGTTTGGC-3') in the albumin promoter (Cereghini et aL, 1987;Godbout et al., 1988), hepatitis B virus enhancer (Shaul and Ben-Levy, 1987), and domain A of the human AFP enhancer (-3.8 kb) (Table I).* Competition gel mobility shift assays showed that the binding of HUH-7 nuclear proteins to E6 was effectively prevented by fragments of the albumin promoter, hepatitis B virus enhancer, and AFP enhancer domain A carrying the GT-rich element (Fig. 10). This suggests that the same factor regulates all these regulatory regions.
T o confirm that the GT-rich element is responsible for the enhancer activity of Efi, we changed four nucleotides in the GT-rich region and analyzed for the ability to support CAT expression and to bind to HUH-7 nuclear proteins. The results show that the normal 27-bp E6 fragment was active in supporting CAT expression, whereas the same fragment carrying M. Motomura, H. Nakabayashi, and T. Tamaoki, unpublished data.  . E,., I II

FIG. 5. Competition between Ej.7 and human albumin promoter for HUH-7 nuclear protein binding. Gel mohility shift
assays were conducted using the El.: DNA end-labeled with '"P and HUH-7 nuclear extracts in the presence or absence of a 100-fold molar excess of unlaheled El., DNA, promoter fragment I (without the HNF-1-hinding site), or fragment I1 (with the HNF-1-binding site) as a competitor. A, the alhumin 5"flanking sequence showing El.; and promoter fragments I and 11 (heau-y lines) used in competition mohility shift assays. Numbers indicate the nucleotide positions relative to the transcription initiation site. R, inhibition of binding of HUH-7 nuclear proteins to El.? by E, DNA and alhumin promoter lragment 11, hut not by fragment I. the mutated GT-rich sequence was not (Fig. 1lA). Similarly, the mutant sequence failed to compete with the wild-type sequence for binding to HUH-7 nuclear proteins (Fig. 11R). These results show that the GT-rich sequence is the enhancer element in E6.

DISCUSSION
In this study, we have defined two enhancers that are present 1.7 (E1.7) and (Efi) 6 kb upstream of the human albumin gene. Together with the enhancer proximal to the promoter (-486 to -221 bp) (Frain et al., 1990), the human albumin gene contains at least three enhancers in the 5'flanking region. We have identified an AT-rich sequence (HNF-1-binding site) as the element responsible for the enhancer activity of El.7. Similar elements have been shown to exist in the proximal enhancer (-358 to -342 bp) (Frain et al., 1990) as well as in the promoter (-65 to -49 bp) of the human albumin gene (Table I). These results indicate that HNF-1 functions as a common factor regulating the promoter and two enhancers (proximal and El.,) of the human albumin gene. The AT motif is also shared by the enhancer and promoter of the human AFP gene (Sawadaishi et al., 1988). In the case of rodent albumin and AFP genes, the AT motif is present in the promoters (Table I), but not in the enhancers (Godbout et al., 1988al., , Herbst et al., 1989. This difference between the species is puzzling in view of the fact that the rat and human HNF-1 molecules are highly similar in structure (94% amino acid sequence homology) and alike in function (Courtois et al., 1987(Courtois et al., , 1988Hardon et al., 1988;Lichtsteiner and Schibler, 1989;Bach et al., 1990). Thus, although HNF-1 has been shown to regulate the promoters of a number of other liver-specific genes (Mendel and Crabtree, 1991), the regulation of the enhancer as well as the promoter by HNF-1 appears to be a characteristic feature of the human albumin and AFP genes. We have identified a GT-rich sequence as the functional element in Es. This sequence is also found in the human albumin promoter (-128 to -144 bp) (Table I), forming a second element shared by the enhancer and promoter. In the case of the mouse albumin gene, a sequence closely related to  5 and 6 ) , and AFP enhancer domain A (78 bp) (lanes 7 and 8). A 15-bp synthetic oligonucleotide corresponding to the AFPl (HNF-1)-binding site in human AFP enhancer domain B (Sawadaishi et al., 1988) was used as a competitor without the GT element (lanes 9 and IO). 100 and 200 indicate 100-and 200fold molar excesses of competitor DNA, respectively. + andindicate the presence and absence of HUH-7 nuclear proteins or competitors, respectively. the Es element (TGTTTGC) is present in the enhancer at -10 kb and binds to a mouse hepatocyte nuclear protein, eH-TF (Zaret et al., 1990). Two homologous sequences are also present in the mouse albumin promoter (Table I). Thus, as far as GT-rich elements are concerned, the human and mouse albumin genes are similar in that these sequences are shared by the enhancers and promoters.
In the case of the AFP gene, there is a difference between humans and mice with respect to the distribution of GT-rich elements. In mice, GT-rich sequences are present in the promoter and two of the three enhancers (Table I) (Godbout  et al., 1988), whereas in humans, it is present in the enhancer? but not in the promoter.
What is the significance of the presence of AT-and GTrich elements in the enhancer and promoter? We have shown that multimerization of the enhancer AT-rich element results in an increase in transcription stimulatory activity (Fig. 6). Similarly, Lichtsteiner and Schibler (1989) have reported that a synthetic promoter containing two B elements (albumin promoter HNF-1-binding site) is strongly activated by purified HNF-1 in uitro. It has been shown that a protein bound hancer, would allow a wide range of expression of the albumin and AFP genes in developing liver.
At present, little is known about the protein(s) that regulates GT-rich elements. Some of the GT-rich elements so far identified exhibit a dyad symmetry, suggesting that they are also regulated by dimerization of factors. The exact mechanism of regulation by GT-rich elements and the significance of their presence in both the promoter and enhancer must await the isolation and characterization of the binding protein(s). to a cis-acting element often facilitates the binding of a second protein to another site and that factors bound far apart can also interact with each other, enhancing transcriptional activity (Ptashne, 1988). It is therefore possible thet factors bound at the AT-and GT-rich elements in the promoter and enhancer exhibit homologous and/or heterologous protein-protein interactions leading to transcriptional enhancement. It should be noted that the AT-or GT-rich sequences in the promoter and enhancer are not identical. Thus, although the same factors bind to the promoter and enhancer elements, their affinity and probably specificity for these sites may differ. In addition HNF-1 has been shown to recognize DNA elements as a dimer (Courtois et al., 1990;De Simone et al., 1991;Rey-Campos et al., 1991;Mendel et al., 1991); and consequently, it can exhibit different binding or activation potential depending on the partner protein that it dimerizes (Mendel and Crabtree, 1991;Mendel et al., 1991). It is possible that HNF-1 and the partner factors change in the relative amount during liver development. This, combined with the presence of AT-rich elements in both the promoter and en-