A potent enhancer made of clustered liver-specific elements in the transcription control sequences of human alpha 1-microglobulin/bikunin gene.

alpha 1-Microglobulin (A1M) and bikunin are plasma proteins which are present both as free molecules and as complexes with either IgA heavy chains for A1M or the H1, H2, and H3 heavy chains of the inter-alpha-inhibitor family for bikunin. Mature A1M and bikunin originate from the cleavage of an A1M/bikunin precursor (ABP) synthesized from a single gene with liver-specific expression. Five kilobases of the 5'-flanking region of the human ABP gene were sequenced. Deletion mutants of this region subcloned upstream of a CAT reporter gene were transfected into HepG2 hepatoma cells. A segment covering the -2.7- to -2.8-kb area is required for full activity of the ABP gene. This segment contains a cluster of six elements (boxes 1-6, 5' to 3') which are potential binding sites for the liver-enriched trans-acting factors HNF-1, HNF-4, HNF-3, HNF-1, HNF-3, and HNF-4, respectively. This cluster enhances the activity of heterologous minimal promoters in a position- and distance-independent fashion in HepG2 cells. This enhancer activity is restricted to liver cells as the cluster is unable to activate promoters in Chinese hamster ovary (CHO) or HeLa cells. By band-shift experiments we have shown that the liver-enriched transcription factors HNF-1, or HNF-3, do bind to boxes 1 and 4, or 3, respectively. The combination of a weak promoter and a strong distant and liver-specific enhancer distinguishes the ABP gene from most other plasma protein genes expressed in hepatocytes.

a1-Microglobulin (AlM) and bikunin are plasma proteins which are present both as free molecules and as complexes with either IgA heavy chains for A I M or the H1, HZ, and H3 heavy chains of the inter-a-inhibitor family for bikunin. Mature A1M and bikunin originate from the cleavage of an AlMbikunin precursor (ABP) synthesized from a single gene with liver-specific expression. Five kilobases of the 6"flanking region of the human ABP gene were sequenced. Deletion mutants of this region subcloned upstream of a CAT reporter gene were transfected into HepGZ hepatoma cells. A segment covering the -2.7-to -2.8-kb area is required for full activity of the ABP gene. This segment contains a cluster of six elements (boxes 1-6, 5' to 3') which are potential binding sites for the liverenriched trans-acting factors HNF-1, HNF-4, HNF-3, HNF-1, HNF-3, and HNF-4, respectively. This cluster enhances the activity of heterologous minimal promoters in a position-and distance-independent fashion in HepGZ cells. This enhancer activity is restricted to liver cells as the cluster is unable to activate promoters in Chinese hamster ovary (CHO) or HeLa cells. By band-shift experiments we have shown that the liverenriched transcription factors HNF-1, or HNF-3, do bind to boxes 1 and 4, or 3, respectively. The combination of a weak promoter and a strong distant and liver-specific enhancer distinguishes the ABP gene from most other plasma protein genes expressed in hepatocytes.
al-Microglobulin (AlM)' is a plasma glycoprotein found in * This work was supported in part by the University of Rouen and the European Economic Community Bridge program. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisernent" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted X67082.
to the GenBankTM/EMBL Data Bank with accession number(s) Recipient of a doctoral fellowship from Ministbre de la Recherche e t de la Technologie. a free state (Mr 31,000) as well as covalently bound to some immunoglobulin A heavy chains (reviewed in Ref. 1). Although a precise function has not yet been ascribed to AlM, this protein is a member of the lipocalin superfamily which includes carrier proteins with an affinity for small hydrophobic ligands such as retinol, cholesterol, steroids, or odorant molecules (1,2). As such, A1M has been proposed as a carrier for porphyrin (3) or retinol (4). Bikunin is another plasma glycoprotein found in a free state (Mr 45,000) as well as covalently bound to the heavy chains of inter-a-inhibitor and pre-a-inhibitor. Inter-a-inhibitor and pre-a-inhibitor are two large molecules with serine protease inhibitor activity which originates from two functional tandemly arranged Kunitztype inhibitory domains in the bikunin chain (reviewed in Refs. 5 and 6). Although A1M and bikunin have no apparent structural or functional similarity, both molecules originate from a shared precursor polypeptide and are released by cleavage of a short connecting peptide within the precursor (7). In mammals a single AlM/bikunin precursor (ABP) mRNA (1.25 kb) (7) is transcribed by a single copy gene (8,9) with hepatocyte-restricted expression (10). The human ABP gene is comprised of 10 exons and spans about 20 kb (8). The largest intron in the gene (7 kb) separates exon 6 coding for the C-terminal end of A1M from exon 7 coding for the connecting peptide and the N-terminal end of bikunin (8). This feature has been assumed to reflect an assembly of two quite distinct ancestral genes coding for a lipocalin and a Kunitz-type protease inhibitor, respectively (8).
The primary function of A1M or bikunin is suggested by their lipocalin-or Kunitz-type-specific amino acid sequence, respectively. Furthermore, they both may be involved in a number of regulatory pathways, including immunoregulation by A1M (1) or stimulation of endothelial cell growth by bikunin (11). The ABP gene is therefore a source for two plasma proteins which could each be important for several biological events. Pathologic changes in the level of either A1M or bikunin could shed light upon their normal role(s), and therefore we have been interested in studying the overall regulation of expression of the corresponding gene under abnormal conditions.' We further aim to elucidate the major mediators involved in the physiological and pathological expressions of this gene. As a first step toward this goal, we now present a structural and functional study of the 5'flanking region of the gene. We have observed that the expression of the ABP gene is controlled by a distal cluster of liver-specific elements which strongly enhances transcription i n a tissue-specific manner.
PCR-PCR was performed with the Taq polymerase and GeneAmp 10 X PCR buffer from Perkin-Elmer Cetus in a Cetus thermocycler 480 (30 cycles: 90 s at 94 "C, 90 s at 55 "C, 120 s at 72 "c).
Plasmids and Constructs-pUC18, -19, and -BM21 were from Boehringer Mannheim. pCHllO (Pharmacia) contains the 8-galactosidase gene under the control of the SV40 early promoter. pSVZCAT, pRSVCAT, and pTKCAT have the chloramphenicol acetyltransferase (CAT) gene under the control of SV40 early region promoter or Rous sarcoma virus 3' long terminal repeat or herpes simplex virus (HSV) thymidine kinase (TK) promoter (-109 to +51), respectively, and were used as positive controls with a high (pSVZCAT, pRSVCAT) or medium (pTKCAT) promoter strength. pTK50 is a CAT plasmid with a minimal 50-bp-long HSV TK promoter including an SP-1 box, a TATA box, and a SalI/HindIII/ XbaI multiple cloning site at the 5' end of the promoter sequence (16). pFIXCAT is a CAT plasmid with a minimal promoter (416 bp) of the human coagulation Factor IX gene (17), and it was used as a control with a weak promoter strength. pUMSVOCAT is a low C. N'Guyen, unpublished data. background promoterless CAT expression vector with a unique SmaI cloning site at the 5' end of CAT gene (18), and it was used as a negative control. The plasmids pTK50, pFIXCAT, and pUMSVOCAT were also used for further constructs as described below. The latter were all controlled by sequencing the 5' and 3' ends of inserted DNAs.
An EcoRI-BamHI segment covering the region -692/+335 (sequence numbering according to Fig. 1) of the human ABP gene was purified from the XCh4A-11 clone (8) and subcloned into pUC18. Various constructs with nested deletions at insert 3' end (i.e. upstream or downstream to ABP gene transcription start site) were prepared by cleavage at the BamHI site followed by digestion with Bal-31 exonuclease. The variably deleted inserts were released by EcoRI digestion, filled-in with Klenow polymerase, added with SmaI linkers, digested with SmaI, and purified by agarose gel electrophoresis prior to ligation at the SmaI site of pUMSVOCAT and screening by clone sequencing. Among the plasmids thus generated, the p-692/ 57CAT clone (ABP promoter from -692 to +57) was further used to prepare a set of nested deletions at insert 5' end, as follows. The insert was cut at its unique HindIII site (-595), digested with Bal-31, filled in with Klenow polymerase, added with SmaI linkers, pUMSVOCAT at the SmaI site, generating p-594/57CAT, p-433/ digested with SmaI, purified by electrophoresis, and subcloned into 57CAT, p-348/57CAT, p-246/57CAT, p-l63/57CAT, p-83/57CAT, and p-36/57CAT, as well as similar control clones with the ABP promoter in an antisense orientation.
A SalIIBamHI fragment of the AlM/bikunin gene covering the 5'flanking sequence from -4964 to +335 was subcloned into pUC19. A partial digest of this construct with HindIII provided a linearized plasmid cut at the HindIII site in the insert (-595) but not at the insert 5' end in the plasmid polylinker; a further digest at the SmaI site in the polylinker released a HindIIIISmaI segment covering the 3' end of the gene promoter. The remaining linear plasmid was then ligated with a HindIII/SmaI 3' segment of the promoter isolated from p-692/57CAT, thus generating a pUC18 construct containing the ABP gene from -4964 (SalI) to +57 (SmaI). This ABP gene segment was cut with SmaI and SalI, filled-in at the SalI site, and subcloned in a sense orientation at the SmaI site of pUMSVOCAT, generating p-4964/57CAT. The pUC18 construct was also digested with EcoRI in the polylinker 3' to the ABP insert and with PstI at a PstI site (-3327) in the ABP gene as well as in the polylinker 5' to the insert. The largest insert segment (-3327/+57, PstIIEcoRI) was subcloned in pUCBM21. This construct was linearized by EcoRV in the polylinker 5' to the insert and nested deletions at the insert 5' end were then prepared by an exonuclease III/Sl nuclease technique (19), followed by end filling-in, SmaI linker addition, SmaI digest, insert purification by electrophoresis, and subcloning at the SmaI site of pUMSVOCAT, generating p-3299/57CAT, p-2929/57CAT, p-2422/ 57CAT, p-l939/57CAT, p-l305/57CAT, and p-859/57CAT as well as similar clones with the ABP promoter in an antisense orientation.
Short segments of the ABP promoter were prepared and subcloned in several plasmids as follows. A SmaI (-2929) to NcoI (-2549) 381bp enhancer (see "Results") was purified from p-2929/57CAT, filledin with Klenow polymerase, and ligated at a filled-in HindIII site upstream to CAT cDNA in the pTK50 plasmid, generating ~3 8 1 1 TK50H (enhancer in a sense orientation relative to CAT gene) and pA381/TK50H (enhancer in an antisense orientation). Likewise, constructs with the 381-bp enhancer ligated 3' to CAT gene at the unique BarnHI filled-in site in pTK50 were designated p381/TK50B and pA381/TK50B. The same 381-bp enhancer in sense or antisense orientation was ligated 3' to CAT gene at the unique BamHI filledin site in pFIXCAT, generating p381/FIXB and pA381/FIXB, respectively. Finally, a short segment of the 381-bp enhancer covering only 107-bp (-2806 to -2700) was amplified by PCR with oligonucleotides 1 and 10 (see above) and p-2929/57CAT as a template and Hind111 site of pTK50, generating p107/TK50H to p107x3/TK50H. ligated as a mono-, di-, or trimer in either orientation at the unique Mammalian Cell Cultures and Transfections-Established cell lines were cultured as monolayers in 75-cm2 flasks at 37 'C in a 5% CO, atmosphere and fed with 20 ml of Minimal essential medium supplemented with 25 mM NaHC03, 10% heat-inactivated fetal calf serum, L-glutamine, 2 mM, penicillin, 50 units/ml, and streptomycin, 50 rg/ ml. Fresh culture medium was added every 2nd day until cell subconfluency was reached. The cells were then trypsinized with a 0.5 mg/ml trypsin (GIBCO-BRL) solution in Na,EDTA, 0.2 mg/ml and transferred into 6-cm dishes at an initial concentration of lo6 cells/ dish. After 24 h the cells were at 50% confluency and were used for transfection experiments. These cells were transfected by a Capo4 Human a1 -Microglobulin/Bikunin Gene Enhancer 20767 precipitation procedure modified from Ref. 20 with a mixture of plasmids (banded twice on CsCl gradient) made of CAT plasmid (4 pg) + pCHllO plasmid (1 pg). The cells were exposed for 4 h to the CaPO,/DNA precipitate and then shocked with 15% glycerol in culture medium for 3 min and rinsed with culture medium. Fourtyeight hours later the (usually confluent) cells were washed with an ion-free Dulbecco's phosphate-buffered saline (Sigma) and harvested by gentle scraping with a rubber policeman. The cell extracts were immediately processed for @-galactosidase and CAT assays. Each set of CAT constructs to be compared was studied in at least two independent transfection experiments. Within an experiment each construct was studied in triplicate.
@-Galactosidase and CATAssays-Our miniaturized fast procedure for simultaneous measurements of P-galactosidase and CAT activities is reported elsewhere (21). For CAT assays, the two-phase partition of [3H]acetyl-CoA and 3H-acetylated chloramphenicol followed by scintillation counting was carried out essentially as in Ref. 22. All values of CAT activity in cell extracts were normalized to the 0galactosidase activity in the same extracts and finally expressed as: counts/min [3H]acetylchloramphenicol/~-galactosidase unit.
R N A Isolation and Primer Extension Annlysis-RNA from cultured hepatoma cells were obtained by in situ lysis according to Ref. 23 followed by centrifugation onto a CsCl cushion (24). Poly(A)+ RNAs were isolated onto oligo(dT) columns (Pharmacia kit). RNA integrity was monitored by visual inspection of the 28 and 18 S ribosomal bands in an agarose gel electrophoresis. A primer extension reaction was carried out onto poly(A)' RNAs (10 pg) with reverse transcriptase and a cDNA synthesis kit (Amersham) as recommended by the manufacturer, in the presence of 1 pl of [w3*P]dCTP. The size of the extended product was determined onto a sequencing gel using a known dideoxy sequence as a size ladder.

Preparation of Nuclear Extracts and Gel Mobility Shift Assay-
Nuclei from different cell lines were prepared as described previously (25) with solutions containing 0.5 mM phenylmethylsulfonyl fluoride, 2 mM benzamidine, and 5 pg/ml aprotinin, pepstatin, and leupeptin. Binding reactions for band shift assays were performed in 16 pl of a reaction mixture containing 10 mM Hepes buffer, pH 7.9,30 mM KC1, 9 mM MgC12, 9 mM spermidine, 0.5 mM dithiothreitol, 10% glycerol, 5 pg/ml of the protease inhibitors mentioned above, 1.5 pg of poly(d1-dC), and 1 pg of sonicated salmon sperm DNA. Four pg of nuclear protein extract were preincubated in this mixture for 5 min at 4 "C. Then 1 ng of kinase-labeled double-stranded oligonucleotide as a probe and competitor oligonucleotides, if any, were added and incubated for 15 min on ice. The DNA-protein complexes were loaded onto a low ionic strength (0.25 X TBE (89 mM Tris, 89 mM borate, 2.5 mM EDTA)) 5% acrylamide (acrylamide/bis: 29/1) gel and electrophoresed at 12 V/cm.

RESULTS
Nucleotide Sequence of the 5"Flanking Region in the Human ABP Gene-A XEMBL-3 clone containing part of the human ABP gene was isolated by screening a genomic library with a human 179-bp cDNA probe corresponding to ABP gene exon I. By comparing a restriction map of this clone to the published map of the entire gene (8), we isolated a SalI/ BamHI 5.3-kb segment. Its 3' end sequence (not detailed) corresponded to a known ABP gene sequence (8) and indicated that this SalIIBamHI segment contained the 5'-flanking sequence of the gene. The SalIIPstI segment (-4964 to -3328) on the 5' side of this flanking sequence was sequenced after subcloning various restriction fragments into a pUC plasmid (strategy not detailed).
The PstI/BamHI 3' side (-3327 to +335) of the flanking sequence was subcloned into pUCBM21 and sequenced by a nested deletion strategy with exonuclease III/S1 nuclease. Some of the resulting clones were subsequently used for constructing deletion mutants in CAT plasmids.
The resulting sequence is presented in Fig. 1. It covers 4964 bp upstream of the published transcription start site numbered (+1) (8). Two shorter sequences spanning the region from -693 to exon I (8) or from -1583 (our numbering) to exon I (9) are included in the present sequence. Compared with (8) our sequence displays an extra G at -66 and a GC inversion at -25. The sequence of the entire 5"flanking region was scanned on both strands for the presence of potential protein-binding motifs. This search was based upon published libraries of such motifs (26,27) and a periodically up-dated list of consensus in our laboratory. No obvious TATA box at the usual -201-25 position was found. Two potential CCAAT boxes are located at -162 and -316. The other, most prominent findings are indicated in Fig. 1 (see  legend). Among them, an extensive series of potential IL-6aresponsive elements (28) were scattered through the entire sequence (Fig. 1). An IL-6P-responsive element, which is a weak affinity binding site for the HNF-1 transcription factor (29), was also found. Notably, numerous potential liver-specific sequences are present. They are detailed in Fig. 1 and Table I. A tight cluster of six such elements is located between -2802 and -2659 (boxed in Fig. 1). These six elements were designated boxes 1-6 (5' to 3') (see Fig. 1 and Table I). They contain potential binding sites for the hepatocyte-enriched nuclear factors HNF-1, HNF-4, HNF-3, HNF-1, HNF-3, and HNF-4, respectively ( Fig. 1 and Table I).
Effect of Progressive 5' Deletions on Transcriptional Activity of the ABP 5"Flanking Region-We wished first to localize potential transcriptional control sequences in the 5"flanking region of the ABP gene. To this end, various CAT constructs were made, all containing an identical ABP promoter 3' end (+57) but deleted to a variable extent at their 5' end (-2929 to -36). The promoter was positioned in a sense or antisense orientation relative to the CAT gene. These constructions were transfected into HepG2 cells. The relative CAT activities of the constructs with the promoter in a sense orientation compared with the control plasmids pFIXCAT or pTKCAT (see "Experimental Procedures") are shown in Fig. 2. The longest three constructs, i.e. p-4964/57CAT to p-29291 57CAT, exhibited the same strong transcriptional activity, which was 50-fold higher than the Factor IX promoter (pFIXCAT) and slightly higher than the HSV TK promoter (pTKCAT). Deleting 2.5 kb at the 5' end in the 5 kb of ABP promoter, i.e. generating p-2422/57CAT, resulted in a dramatic (50-fold) decrease in CAT activity. A further 1.7-kb deletion in ABP gene sequence (i.e. p-692/57CAT) resulted in a further 10-fold drop (activity = 0.1 relative to pFIXCAT) in CAT activity. This activity was partially recovered (4-fold above pFIXCAT activity) in p-348/57CAT; this construct corresponds to the "minimal" ABP promoter, since all constructs deleted further (p-246/57CAT to p-36157CAT) lacked any significant CAT activity. Several control constructs made with the variably deleted ABP promoter in a reverse orientation relative to CAT gene did not exhibit significant CAT activities (data not shown). Overall, our results suggested the presence of a strong positive regulatory region located between -2929 and -2423 and a negative regulatory region between -692 and -349.
To verify the transcription start site of the ABP gene from an ABPICAT construct, poly(A)+ RNAs isolated from HepG2 cells transfected with p-2929/57CAT were hybridized with a CAT gene-specific primer (Fig. 3, left panel) and treated with reverse transcriptase. The extended product was 136 bp in length (Fig. 3, rightpanel), a value which is in close agreement with the published transcription start site (+1) of the ABP gene (8).
A Strong Enhancer Containing Several Liver-specific Elements Is Located in a Distal 5"Flanking Region of the ABP Gene-We further analyzed the activity of the positive regulatory element described above. Various CAT constructs were prepared from a 381-bp segment (-2929 to -2549) of the ABP gene ligated in sense or antisense orientation upstream Human 01 -Microglobulin/Bikunin Gene Enhancer -4964 GTCGACGGAT CTTGGCTCAC TGCAACCTCC GCCTCCTGGG TTGAAATGAT TCTTCTGCCT """"""" """"""__"__""""""""""""""""  When several slightly different consensus are available, the one that best fits the ABP gene sequence is considered. Consensus ambiguities Complementary strand similar to consensus. HNF-1 is also designated LF-B1 or HNF-la (31). of a minimal promoter or downstream to CAT gene in the plasmids pTK50 or pFIXCAT which both contain a weak minimal promoter. As shown in Fig. 4 A , all such constructs transfected in HepG2 cells exhibited a strong increase in CAT activity compared with the controls, independent of the orientation or location of the 381-bp segment relative to the heterologous promoter. Specifically, the pTK50 plasmid activity was enhanced 5-fold or more whatever the orientation/ position of the 381-bp sequence; likewise the pFIXCAT plasmid activity was enhanced up to 30-fold. Therefore, this 381bp sequence meets all the standard criteria for an enhancer. Furthermore, this sequence encompasses the set of six clustered binding sites for liver-specific factors mentioned above ( Fig. 1 and Table I). T o investigate whether these boxes could account for the enhancer activity, additional constructs were made with a PCR-generated 107-bp segment covering boxes 1-5 (box 6 is apparently nonfunctional as judged from band shift experiments, see below). Up to three copies of this 107bp segment were ligated upstream of the minimal T K promoter in pTK50. As seen in Fig. 4B, this 107-bp segment increased TK promoter activity about 16-fold and therefore fully retained the enhancer activity first seen with the 381-bp (-2929 to 2549) segment. Additivity in enhancement was observed when two copies of this 107-bp segment were ligated to pTK50. Finally, no further significant increase was observed with three copies of the enhancer (construct p107x3/ TK50H).
Liver-restricted Activity of the ABP Enhancer-To test whether the enhancer activity detected in HepG2 cells is cellspecific, the enhancer/ABP promoter-containing construct p-2929157CAT was transfected into two nonhepatic cell lines, namely immortalized cervical carcinoma cells of human origin (HeLa) and an established Chinese hamster ovary (CHO) cell line. T o allow for comparisons between lines, two internal standards were used for normalization, namely pTKCAT, a plasmid with the ubiquitous HSV T K promoter, as well as pTK50, a construct which contains a minimal T K promoter (proximal SP-1 box and TATA motif). As seen in Fig. 5 , p-2929/57CAT activity was 2.5-fold higher than pTKCAT activity in HepG2 cells, whereas it was 2or 20-fold weaker than pTKCAT activity in CHO or HeLa cells, respectively. Therefore p-2929/57CAT activity was about 6-or 50-fold lower in CHO or HeLa cells, respectively, as compared with its activity in HepG2 cells. Likewise, the 6-fold enhancement of transcription provided in the construct p381/TK5OH containing the 381-bp enhancer in pTK50 was observed in HepG2 but not in CHO or HeLa cells (Fig. 5). In contrast, the enhancerless construct with the minimal ABP promoter, i.e. p-348/57CAT, used as a negative control, exhibited a weak and similar activity (below pTK50 activity) in the three cell lines.
Characterization of the DNA Binding Proteins Recognizing the ABP Enhancer-To verify that the putative binding sites for liver-specific factors which we localized by homology search within the ABP enhancer ( Fig. 1 and Table I) are indeed recognized in oitro, we performed hand shift experiments. Nuclear extracts from the differentiated hepatoma HepG2 or H4II cell lines (human or rat origin, respectively). from the dedifferentiated rat hepatoma H5 cell line, or from the human cervical carcinoma HeLa cell line were used. The probes were double-stranded labeled oligonucleotides spanning (i) one of the ARP boxes 1-6 or (ii) estahlished targets for the HNF-1, HNF-3, or HNF-4 transcription factors. 'The results are presented in Fig. 6.
The potential HNF-1 boxes 1 (oligonucleotides ARP 1/2) and 4 (oligonucleotides ARP 7/8) were both hound hv the HNF-1 transcript,ion factor (Fig. 6 A ) . First, thev hoth inhihited the binding of HNF-1 present in HepC2 nuclear extract to its target present in the rat albumin promoter (PE56 probe) (25) as shown in Fig. 6A, middle panels. Furthermore. they formed a complex corresponding to the electrophoretic migration of HNF-1 (Fig. 6A, right panel.?) which was slower than the complex formed with v-HNF-1 present in the dedifferentiated H5 hepatoma cell line (25) (Fig. 6.4. far left panel). Finally, boxes 1 and 4 were not recognized by any protein present in HeLa extracts (data not shown).
From ( i ) the competition of PE56, ARP 1/2, or ARP 7/X with the complex formed by HNF-1 and PES6 and (ii) the competition of PE56 with complexes formed hy HNF-1 and A R P 1/2 or ARP 7/X, it is clear that HNF-1 displays a similar affinity for PES6 and ABP 7/8 (hox 4) but a weaker affinity for ARP 1/2 (hox 1).
douhlet. Finally, with both HepG2 and HeLa extracts, weak nonspecific complexes were observed with both ABP 5/6 and 9/10 probes, since the shifted bands remained unchanged in the presence of homologous competing oligonucleotides ABP 5/6 and ABP 9/10, respectively ( Figure 6B, middle and right panels). We conclude that only ABP box 3 (oligonucleotides ABP 5/6) does bind HNF-3. Finally, we failed to confirm the tentative assignment of boxes 2 and 6 as HNF-4 binding sites. The corresponding oligonucleotides ABP 3/4 (box 2) and ABP 11/12 (box 6) were both unable to displace the binding of an HepG2 nuclear protein to a functional PKHNF4 probe (15) (Fig. 6C, left  panel). Furthermore, when used as radioactive probes they failed to form a complex of the HNF-4 type. The complex they formed with a nuclear protein present in HepG2, H5, and HeLa cell lines (Fig. 6C, middle and right panels) was nonspecific as it was not displaced by any of the competitors tested, including the homologous oligonucleotides.

DISCUSSION
Although several reports have dealt with the sequence and organization of the human ABP gene and its 5"flanking region (8,9), no functional analysis of the ABP gene promoter has yet been reported. Given the strong variations in bikunin mRNA in pathological states, and the potential influence of the bikunin level on inter-a-inhibitor and pre-a-inhibitor levels in plasma, we have been interested in clarifying which cis-acting elements (CAE), trans-acting factors, and mediators drive ABP gene expression. A computer-aided search for the presence of CAEs in the 5"flanking sequence of the ABP gene provided us with a large series of putative CAEs (Fig. 1). Investigating each potential CAE was beyond the task of this work and hence a number of 5' deletion mutants subcloned in a CAT expression vector were used to localize the major CAEs. A basal weak promoter activity required one or more elements located between -348 and -247, possibly including a putative CCAAT box at -316 and/or a putative HNF-4 box at -306 (detailed in Table I). The activity of the minimal ABP promoter is not tissue-specific as it reached a similar level in various cell lines when compared with pTKCAT activity (Fig. 5). A proximal negative regulatory element located in the -692 to -349 area abolishes this basal ABP promoter activity. As several constructs (p-433/57CAT to p-692/57CAT in Fig. 2) were required to progressively abolish this activity, this negative element likely covers a rather broad area, approximately from -600 to -349. Linker scanning mutations could eventually provide the location(s) of the negative element(s), but they are beyond the scope of the large construct (p-2929/57CAT) containing the 5"flanking present study. This proximal negative element is prohably sequence up to -2929 was tested. This suggests that. among not a silencer, since the corresponding DNA segment sub-the potential CAEs found in the area extending from -69.1 to cloned, in either orientation, into the pTK50 plasmid did not -2422 ( Fig. 1 and Table I), none is functionally important, at significantly decrease the activity of a minimal T K promoter least under the conditions used in this studv (heterologous (results not shown). Finally, the basal ABP promoter activity reporter gene, hepatoma cell, no hormone/c.ytokine inducwas not significantly exceeded or even fully restored until a tion). However, given the down-regulation of the ARP gene during the acute inflammation response? further experiments will be necessary to clarify which of the numerous potential IL-6a-responsive sites (IL-6 DNA binding protein site) found in the ABP 5"flanking sequence are indeed functional under IL-6 induction. The strong CAT activity observed with p-2929/57CAT could be explained by a distal grouping of liver-specific elements clustered within an enhancer. This was demonstrated with various CAT constructs in which this ABP enhancer drove a heterologous TK or Factor IX promoter in a orientation-and location-independent fashion. This strong CAT activity was accounted for by an increased CAT gene transcription as judged from a Northern blot experiment (not shown). When the enhancer activity was studied in the context of its own ABP gene promoter, the location of the transcription start site in the ABP/CAT construct was in agreement with the location previously found (+1) by a primer extension experiment made with human liver RNAs (8). Our results did not confirm another proposal for a transcription start site at +42 (9). Finally, the ABP enhancer is tissuespecific, since among the cell lines tested, it was active only in an hepatoma cell line. By a computer-aided search, we have identified in the liver-specific enhancer six potential binding sites for hepatocyte nuclear factors. Specifically, two HNF-1-, two HNF-3-, and two HNF-4-binding boxes were present in the order HNF-1, -4, -3, -1, -3, -4 (boxes 1-6, 5' to

3').
Band shift experiments clearly demonstrated that boxes 1 (HNF-I), 3 (HNF-3), and 4 (HNF-1) do indeed bind their cognate nuclear factor, whereas we failed to show a specific binding for boxes 2,5, and 6. It is worth noting that for each box this binding capacity, or lack thereof, did not obviously correlate with the extent of mismatch(es) between the ABP gene sequence and the reference consensus (Table I). It is likely that box 6 is indeed nonfunctional, since the enhancer capacity of the entire cluster (boxes 1-6) did not significantly differ from the enhancer activity of a narrowed cluster made of boxes 1-5 only, as judged from the CAT activities of constructs containing either a 348 bpor 107-bp-long ABP segment, respectively (Fig. 5). In contrast, the lack of binding capacity of boxes 2 and 5 for HNF factors in band shift experiments should not be regarded as conclusive evidence that they are nonfunctional, for two reasons. First, a weak affinity of a given box for its cognate protein may have not been detected under our experimental conditions, but it may be of functional significance in vivo. Second, our band shift experiments were performed with probes and competitors which contained only a single ABP box. Thus, we did not allow for a possible cooperativity in binding of several transcription factors, a situation which may significantly determine the overall affinity of each factor for its DNA target (15,32). Further experiments aimed at analyzing such a possible interplay of nuclear factors are currently in progress.
Some features of the transcription control sequences in the ABP gene are similar to those found in many other liverspecific genes (30), including the presence of multiple CAEs which contribute a high level of expression in hepatocytes compared with other cells and the involvement of trans-acting factors that are expressed with a broader tissue distribution than the target gene (33). Recently, however, the liver-restricted transcription of a number of mammalian genes, particularly those coding for plasma proteins, have been studied. These genes are driven by the interplay of CAEs in their 5'-flanking region and usually are under the control of a strong promoter containing tissue-specific elements driven by HNF factors (e.g. Refs. 15,28,34,and 35). The ABP gene is distinct from such genes, as its overall activity results from the rather unusual combination of a weak promoter of potential ubiquitous expression, enhanced by a remote cluster of liver-specific elements for HNF factors, an arrangement recently described also for the human transferrin gene (36).