Orphan nuclear receptor HNF-4 binds to the human coagulation factor VII promoter.

The human coagulation protease factor VII plays a pivotal role in the initiation of the coagulation cascade by both the extrinsic and the intrinsic pathway. Although the gene, encoding factor VII, is expressed predominantly in the liver, the mechanisms underlying this tissue-specific expression have not been elucidated. In this study, we have analyzed the contribution of 5 kilobases upstream of the ATG translational initiation codon upon hepatic factor VII gene transcription. Transient transfection assays of a set of nested deletions in both liver and non-liver cell lines, HepG2 and HeLa respectively, indicate that several regions are involved in liver-specific expression. A slight negative effect on factor VII promoter activity in HepG2 cells is mediated by sequences upstream of position -1212. DNase I protection experiments reveal six footprints, FPVII1 through FPVII6, within the proximal 714 base pairs but a minimal promoter of 165 base pairs containing only FPVII3-6 is sufficient to confer liver-specific expression in HepG2 cells. Interestingly, FPVII6, at position -14 to +10 on the sense strand, would indicate that an as yet unknown transcription factor covers the ATG translational initiation codon. Gel retardation experiments show that the liver-enriched transcription factor HNF-4 binds specifically to footprint FPVII4 at position -71 to -49. Furthermore, a T --> A transversion, that in the HNF-4 binding site of factor IX causes a severe bleeding disorder, was introduced into the HNF-4-binding site of factor VII and reduced promoter activity by 20-50%. Coordinate HNF-4-mediated regulation of several blood protease genes as well as genes involved in lipid metabolism might account for the positive correlation of these factors with increased risk of occlusive heart diseases.

The human coagulation protease factor VII plays a pivotal role in the initiation of the coagulation cascade by both the extrinsic and the intrinsic pathway. Although the gene, encoding factor VII, is expressed predominantly in the liver, the mechanisms underlying this tissue-specific expression have not been elucidated. In this study, we have analyzed the contribution of 5 kilobases upstream of the ATG translational initiation codon upon hepatic factor VII gene transcription. Transient transfection assays of a set of nested deletions in both liver and non-liver cell lines, HepG2 and HeLa respectively, indicate that several regions are involved in liver-specific expression. A slight negative effect on factor VII promoter activity in HepG2 cells is mediated by sequences upstream of position ؊1212. DNase I protection experiments reveal six footprints, FPVII1 through FPVII6, within the proximal 714 base pairs but a minimal promoter of 165 base pairs containing only FPVII3-6 is sufficient to confer liver-specific expression in HepG2 cells. Interestingly, FPVII6, at position ؊14 to ؉10 on the sense strand, would indicate that an as yet unknown transcription factor covers the ATG translational initiation codon. Gel retardation experiments show that the liver-enriched transcription factor HNF-4 binds specifically to footprint FPVII4 at position ؊71 to ؊49. Furthermore, a T 3 A transversion, that in the HNF-4 binding site of factor IX causes a severe bleeding disorder, was introduced into the HNF-4-binding site of factor VII and reduced promoter activity by 20 -50%. Coordinate HNF-4-mediated regulation of several blood protease genes as well as genes involved in lipid metabolism might account for the positive correlation of these factors with increased risk of occlusive heart diseases.
Factor VII is a vitamin K-dependent protease of the coagulation cascade present in plasma in trace amounts (1) which, when complexed with tissue factor and calcium ions, can cleave both factor X and factor IX initiating the extrinsic pathway and the intrinsic pathway of blood coagulation, which results finally in the formation of a fibrin clot.
The human factor VII gene is a single copy gene on chromosome 13q34, 3 kb 1 upstream of the coagulation protease factor X gene (2,3), which consists of eight exons spread over 12 kb of genomic DNA, which produces a 2.4-kb mRNA encoding a mature protein of 254 amino acids (4).
A positive correlation between high plasma factor VII antigen levels (FVIIag), procoagulant activity (FVIIc), and an increased risk of coronary heart disease has been demonstrated (5). Although both increased FVIIag and FVIIc levels are associated with other risk parameters such as higher age, use of oral contraceptives, and plasma triglyceride and cholesterol (see Ref. 6 and references therein), the picture is complicated as both FVIIag and FVIIc are strongly correlated with plasma triglyceride levels (7), while the correlation between cholesterol and FVIIag is rather weak. More detailed studies emphasized the strong correlation between triglycerides present in very low density lipoprotein and low density lipoprotein particles and FVIIag levels (8 -10). Intriguingly, males heterozygous for an Arg 353 to Gln 353 polymorphism found in about 20% of the caucasian population have FVIIag concentrations reduced by 20 -25%, which presumably reduces the risk of coronary heart disease and thrombosis by lowering the proportion of factor VII molecules being in the activated state (11,12).
From these prospective studies it is possible that a drug that interferes with factor VII expression levels could be an antithrombotic, and the first step toward such a drug would be to understand the regulatory mechanisms of factor VII expression. As only 522 nucleotides of the factor VII 5Ј-flanking region have been published (4), we have cloned and sequenced a further 4291 bp and functionally characterized this sequence in reporter gene assays. We have identified cis-acting regulatory sequences in the promoter by DNase I footprinting analysis, and we show that HNF-4 is a positive regulator of factor VII expression.

Isolation and Sequencing of a Recombinant Cosmid Clone-Sequence
from the promoter previously published extends only 522 bp upstream of the ATG (4). A 654-bp fragment was amplified from human genomic DNA (Clontech) by PCR and sequenced. This showed some differences with the published sequence which included a 10-bp insertion (CCTATATCCT) located at position Ϫ324 with respect to the ATG that has been previously described as a polymorphism by Marchetti and co-workers (13). To obtain additional 5Ј-flanking sequences, six aliquots of 1 ϫ 10 4 clones each of a human placenta genomic DNA cosmid library (Clontech) were screened by PCR using the same primers. One positive aliquot was subjected to screening by colony hybridization using the PCR fragment as probe, and one positive cosmid clone was obtained. Based on restriction digests and Southern blotting the clone was estimated to contain at least 5 kb of 5Ј-flanking sequences. A 3.5-kb EcoRI fragment and a 0.8-kb SmaI fragment overlapping each other in the already known sequence were subcloned into vector pBIIKSϩ (Stratagene). To recombine both fragments, the 3.5-kb EcoRI fragment was cloned into the EcoRI sites in the vector containing the 0.8-kb SmaI fragment (Fig. 1). Additional 5Ј-flanking sequences were subcloned as a 2.5-kb ApaI fragment that overlaps the EcoRI fragment. Enzymatic sequencing of both strands was performed by both ExoIII/S1-generated nested deletion templates (Erase a base kit, Promega) or primer walking techniques.
Construction of Plasmids-The luciferase reporter gene plasmids of the pGL2 series were obtained from Promega, and reporter plasmids were constructed in several steps. First the factor VII open reading frame was deleted by subcloning the proximal sequence as a 280-bp EcoRI/HindII fragment into EcoRI/SmaI of pBIIKSϩ. In the next step the 3.5-kb EcoRI fragment was inserted in order to obtain a long consecutive promoter fragment.
To generate the plasmids pLUC-237/-34 and pLUC-165/-34, the PstI or NsiI sites in the progenitor plasmid pLUC-474/-34 were treated with T4 DNA-polymerase and fused to the SmaI site of the vector. In pLUC-46/-34 the single NcoI site was filled up by Klenow-polymerase and fused to SmaI. Construct pLUC-2.5/ϩ40 was generated by blunt end fusion of a PCR product amplified from promoter and luciferase parts of pLUC-712/ϩ131 and replacing the promoter-luciferase EcoRI fragment in pLUC-2.5/-34. Plasmid pLUC-712/ϩ131 itself originates from inserting the originally cloned SmaI fragment into the filled-in HindIII site of pGL2-basic. All PCR fragments were sequenced to rule out amplification artifacts.
The plasmids pLUC-474/-34"Leyden" and pLUC-165/-34"Leyden" were generated by replacing the StyI-NcoI fragments containing the putative HNF-4 site by a double-stranded oligonucleotide that contained the Leyden-specific point mutation (Table I). The mutation was also introduced into plasmid pLUC-1.6/-34 by replacing the EcoRI fragment that contained the HNF-4-binding site by the corresponding fragment from pLUC-474/-34"Leyden." All plasmid derivatives containing the SV40 enhancer were generated by replacing the BamHI-HindIII vector fragment, which contains luciferase gene followed by the SV40 T antigen intron and polyadenylation signal, with the identical part from pGL2-control (Promega), that also contains the 250-bp SV40 enhancer downstream of the polyadenylation signal.
Cell Culture and Transfections-Human hepatoma cells HepG2 (14) and HeLa cells were obtained from the American Type Culture Collection. COS-7 cells and L132 cells were taken from our in-house collection. HepG2 cells were cultured in Dulbecco's modified Eagle's medium/ F12 nutrient mix 1:1 (Life Technologies Inc.) supplemented with 10% fetal calf serum (Boehringer Mannheim). HeLa cells, L132 cells, and COS-7 cells were grown in minimal essential medium with 5% fetal calf serum. Both media were supplemented with gentamicin and cells were grown in a 10% CO 2 atmosphere. COS-7 cells were transfected with Lipofectin Reagent (Life Technologies Inc.) following the manufacturer's protocol. Transient transfections into HepG2 and HeLa cells were performed by the calcium phosphate coprecipitation technique as described by Ausubel et al. (15). Plasmid DNAs were isolated with DNA purification columns purchased from Qiagen. Cultured cells, replated 24 h before transfection to about 70% confluence, were cotransfected with 0.48 pmol of luciferase plasmid (e.g. 3 g of pLUC-3.9/-34) and 1.5 g of pCMV␤ used as an internal control. The precipitate was removed 16 h later, and the cells were exposed for 3 min to 0.4 ml of 10% glycerol in medium, washed three times with phosphate-buffered saline, and grown in fresh medium for 48 h. In HeLa cell transfections the glycerol shock was omitted. Cells were harvested using 150 l of reporter lysis buffer (Promega) according to the manufacturer's recommendations.
Luciferase and ␤-Galactosidase Assays-Luciferase assays were carried out in microtiter plates using the Luciferase assay system (Promega) as described in the manufacturer's protocol. ␤-Galactosidase assays were performed in microtiter plates as follows: 100 l of prewarmed assay buffer (40 mM Tris-HCl, pH 7.5, 2 mg/ml O-nitrophenyl ␤-D-galactopyranoside, 84 mM ␤-mercaptoethanol) were added to 20 l of cell extract and incubated at 37°C. After 2-45 min the reaction was stopped by adding 75 l of 1 M sodium carbonate and the A 420 determined. Luciferase data were normalized with respect to ␤-galactosidase values in order to correct for differences in transfection efficiency.
RNA Isolation and Northern Blotting-Single step RNA isolation from cultured cells and Northern blotting were done as described in Ref. 15. Radioactive hybridization was performed with [ 32 P]dCTP-labeled DNA fragments in 5 ϫ SSPE, 10 ϫ Denhardt's, 100 g/ml singlestranded DNA, 50% formamide, and 2% SDS at 42°C. After 16 h of incubation, the filters were washed with 2 ϫ SSC, 0.5% SDS for 5 min at room temperature and 2 times 30 min at 42°C. For non-radioactive hybridization DIG-labeled RNA probes were used according to the manufacturer's recommendations (Boehringer Mannheim).
DNase I Footprint Assays-Nuclear extracts from HepG2, HeLa, L132, and COS-7 cells were prepared by the method of Dignam et al. (16) with minor modifications as described by Ausubel et al. (15).
DNase I footprints were performed in a total volume of 50 l, containing 20 mM HEPES, pH 7.5, 30 mM KCl, 4 mM MgCl 2 , 0.5 mM EDTA, 0.5 mM dithiothreitol, 4% glycerol, and 1.5 g of poly(dI-dC)⅐poly(dI-dC). Nuclear extracts containing 10 -60 g of total protein were incubated for 15 min at room temperature. One to two ng of end-labeled fragments (1-2 ϫ 10 5 counts/min) were added and incubated for another 15 min at room temperature. Limited digestion was achieved by adding 5 l of Ca 2ϩ /Mg 2ϩ solution (final concentration 1 mM MgCl 2 , 0.5 mM CaCl 2 ) and 0.33-1 units of freshly diluted DNase I for 1 min at room temperature. The reaction was stopped by 140 l of DNase I stop buffer (192 mM sodium acetate, 32 mM EDTA, 0.14% SDS, and 64 g/ml yeast RNA). The DNA was treated with phenol, ethanol precipitated, and analyzed on 6% polyacrylamide, 7 M urea sequencing gels.
Electrophoretic Mobility Shift Assay-Duplex oligonucleotides were labeled at both ends by filling in with Klenow fragment of DNA-polymerase I. Ten to 30 g of crude nuclear extracts were preincubated in presence of 2 g of poly(dI-dC)⅐poly(dI-dC) in a total volume of 25 l of DNase I footprint buffer at room temperature. After 10 min, 0.25 ng of labeled oligonucleotide (3 ϫ 10 4 counts/min) were added, and incubation was continued for an additional 10 min. Protein⅐DNA complexes were fractionated on 5% polyacrylamide gels (29:1, 0.5 ϫ Tris-borode, 5% glycerol) with 0.4 ϫ TBE running buffer at 4°C.
Antibody Supershift Assay-The electrophoretic mobility shift assays were performed with HNF-4 antiserum kindly provided by Dr. Frances Sladek (University of California, Riverside). One l of undiluted antiserum was added to 10 g of crude nuclear extract preincubated with labeled oligonucleotide.

5Ј-Flanking Sequence of the Human Factor VII Gene-Based
on the limited published sequence (4) we cloned a PCR fragment of 523 bp immediately upstream of the factor VII translational initiation codon from human genomic DNA (Clontech). This clone identified one positive clone cosfVII on screening a human placenta cosmid library. The cosmid DNA, which was shown to be colinear with genomic DNA by Southern blotting using the PCR fragment as a probe, contains at least 5 kb of new 5Ј-flanking sequence which was sequenced on both strands. Compared to the published sequence two additional nucleotides were found, an additional C at Ϫ140 and an additional G at position Ϫ460 ( Fig. 1), which were probably not detected previously due to band compressions typically found in GC-rich sequences.
A homology search of the 4813-bp 5Ј-flanking sequence against the EMBL databank identified three Alu repeats at position Ϫ4742 to Ϫ4440, Ϫ2739 to Ϫ2518, and Ϫ942 to Ϫ642. Sequence alignment of the factor VII and factor X promoters revealed a similarity of 86% for a small 37-bp element located in factor VII promoter at position Ϫ2340 to Ϫ2304 and in the factor X promoter in the same orientation at position Ϫ520 to Ϫ485. The functional relevance, if any, of this element is unclear as deletion of these sequences in the factor X or factor VII promoters did not alter reporter gene expression significantly (see Fig. 3 and Refs. 3,18).
Tissue Distribution of Factor VII Transcription-In order to determine the tissue distribution of factor VII expression, we probed a human multiple tissue Northern blot from Clontech (catalog no. 7760 -1) with a randomly labeled fragment encoding amino acids Ϫ60 to ϩ152 in factor VII. Factor VII expression is clearly restricted to liver, where a strong signal corresponding to transcripts of 2.4 kb of length was obtained while in all other tissues transcripts can be detected only at very low levels (Fig. 2). Under the low stringency washing conditions, another band of about 1.2 kb was detected in pancreas, and two additional bands of 1.3 and 1 kb were observed in skeletal muscle.
Transcriptional Control of the Factor VII Gene in HepG2 Cells-To investigate the transcriptional regulation of factor VII, a promoterless luciferase reporter gene in the vector pGL2basic was fused to a variety of promoter fragments at position Ϫ34 as a HindII/SmaI fusion which is 3Ј of the transcriptional start site (Fig. 3). The constructs were all transfected into human hepatoma HepG2 cells, known to express factor VII (22), as well as HeLa and COS-7 cells, and transient promoter activity was monitored by measuring luciferase activity in cell extracts. Differences in transfection efficiencies were corrected for by cotransfection of pCMV␤ that carries the ␤-galactosidase reporter gene under the control of the CMV promoter. As negative control the promoterless plasmid pGL2-basic was used while as a positive control plasmid the luciferase gene was transcribed from the SV40 early promoter-enhancer in pGL2control which gave rise to very high luciferase values in all three cell lines tested. In HepG2 cells, expression from the factor VII promoter in plasmid pLUC-3.9/-34 was significantly higher than the negative control pGL2-basic, but, even in hepatocytes, the factor VII promoter is remarkably weak as compared to the SV40 promoter, which gave about 50-fold higher luciferase values.
Deletion of sequences from Ϫ4813 down to position Ϫ1601 did not significantly alter luciferase expression, but a longer deletion to position Ϫ1212 doubled luciferase expression (Fig.  3). Further truncation of the promoter down to position Ϫ165 retained the high promoter activity of pLUC-981/-34, but a longer deletion to position Ϫ46 drastically reduced promoter activity down to values seen with the promoterless control vector. These results suggested that the first 165 bp upstream from the translational initiation codon are sufficient to confer expression in hepatocytes. Activity was restored to 73% on plasmid pLUC-474d3/-34 by sequences from Ϫ474 to Ϫ355, although the presence or absence of this fragment on the longer constructs pLUC-474/-34, pLUC-237/-34, and pLUC-165/-34 had no impact on expression.
Hepatocyte Specificity of Factor VII Expression-Direct comparison of expression data from HepG2 and HeLa cells was difficult because transfection efficiencies differed by a factor greater than 20, as judged from ␤-galactosidase values, but expression from the factor VII promoter was detected in both cell lines. Virtually no promoter activity was detectable in COS-7 cells as none of the factor VII promoter constructs ex-pressed luciferase better than the negative control pGL2-basic (data not shown). In HeLa cells sequential truncation of the promoter led to a steady reduction in activity down to about 50% of the original values in pLUC-165/-34 and to 10% in plasmid pLUC-46/-34 (Fig. 3) which does not identify any specific promoter in the 5Ј-flanking sequence. Constructs containing the heterologous SV40 enhancer were more informative as only the constructs containing sequences upstream of position Ϫ1212 can be activated by a factor of three to four by the SV40 enhancer, which suggests that the enhancer interacts with some promoter-like structures located upstream from position Ϫ1212 (Fig. 3).
Identification of Regulatory Sequences in the Factor VII Promoter-A homology search for putative transcription factorbinding sites, using the Findpatterns algorithm of the GCG software package on the data base Transcription Factor Sites Release 6.5 (23,24), revealed four putative Sp1 binding sites at positions Ϫ2295, Ϫ1862, Ϫ1847, and Ϫ1785 and consensus sequences for AP-1 binding at position Ϫ2488, Ϫ1006, and Ϫ750 although no inducibility of factor VII expression by phorbolesters was observed (data not shown). A putative C/EBP-␤/ NF-IL6-binding site at position Ϫ690 to Ϫ682 seems not to be FIG. 2. Tissue distribution of factor VII transcripts determined by a multiple tissue Northern blot. The multiple tissue Northern blot purchased from Clontech was hybridized with a randomly labeled factor VII cDNA fragment corresponding to codons Ϫ60 to ϩ152 (panel A) and with a ␤-actin-specific probe (panel B), respectively. Lane 1, heart; 2, brain; 3, placenta; 4, lung; 5, liver; 6, skeletal muscle; 7, kidney; 8, pancreas. A strong signal corresponding to a transcript of the correct size is visible in liver (lane 5). relevant as adding IL6, IL1-␤, and combinations of IL6 with IL1-␤ or dexamethasone did not effect transient transfection assays nor RNA levels measured by Northern blots. This is consistent with a report by Hassan et al. (25) that IL6 does not affect factor VII mRNA levels in Hep3B cells.
The findpatterns search for cis-acting regulatory sequences did not identify any further putative binding sites for liverenriched transcription factors, but when repeated with more degenerated consensus sequences a potential binding site for the liver-enriched transcription factor HNF-4 at position Ϫ67 to Ϫ56 was seen.
HNF-4 Binds to the Factor VII Promoter-The transcription factor HNF-4 (26) has been shown to be involved in liverspecific expression of coagulation factors IX (17,27) and X (3,18). So, to investigate whether the presence of the putative HNF-4-binding site is of any significance for factor VII transcriptional regulation, electrophoretic mobility shift assays (EMSA) were performed with double-stranded 32 P-labeled oligonucleotides that correspond to wt factor VII sequence from Ϫ77 to Ϫ43 and a mutated oligonucleotide carrying a T 3 A transversion (Table I). EMSA analysis with nuclear extracts from HepG2 cells revealed a concentration-dependent shift of protein⅐DNA complexes as compared to free probe (Fig. 4A,  lanes 1-4, C, and F). This shift could be specifically competed for by preincubation with unlabeled wt oligonucleotide (lanes 5 and 6), but not by the T 3 A transversion mutant oligonucleo-tide (lane 7) which when present in the factor IX promoter causes the Leyden phenotype of factor IX deficiency (27). The affinity of a DNA-binding protein to this mutated oligonucleotide was strongly reduced in our experiments (lane 9) and could be competed for much more efficiently by the wt sequence as compared to the mutated oligonucleotide itself (lanes 10 and 11) demonstrating that protein binding is highly specific and dependent on the sequence of the binding site.
To prove unambiguously that the protein binding is indeed HNF-4, we repeated the EMSA assays in the presence or absence of an antiserum raised against HNF-4 (26). The slower migrating complex obtained with HepG2 nuclear extracts (Fig.  4B, lane 2, C) was supershifted upon addition of the HNF-4 antiserum (lanes 3 and 4, S). In contrast, the slower migrating complexes obtained with crude nuclear extracts from HeLa, L132, and COS-7 cells (lanes 5, 7, and 9) that could be competed for by cold oligonucleotide (data not shown) were not recognized by the anti HNF-4 antiserum (lanes 6, 8, and 10).
To further investigate the functional role of the HNF-4 site in HepG2 cell experiments, we replaced the StyI/NcoI fragment containing the wt-binding site in plasmid pLUC-165/-34 and pLUC-474/-34 by an oligonucleotide that contains the Leydenspecific point mutation (Table I) mutation reduced the promoter activity to about 80 -50% of the particular wt plasmid in HepG2 cells (Fig. 3) which confirms the functional importance of the HNF-4-binding site for liverspecific expression.
The Factor VII Promoter Is Regulated by Additional Cisacting Regulatory Sequences-To identify further cis-acting regulatory sequences in the promoter, DNase I protection assays were performed. Besides the footprint FPVII4 originating from HNF-4 on the antisense strand at position Ϫ50 to Ϫ76, five more regions protected against DNase I digestion were detected (Fig. 5). All footprints were confirmed on the sense strand, on which HNF-4 protects position Ϫ49 to Ϫ71 (data not shown), but none of additional sites were similar to known cis-acting regulatory sequences. The strong footprint FPVII6, at position Ϫ13 to ϩ9, interestingly covers the ATG translational initiation codon, and we have tested the specificity of this footprint in EMSA experiments using two complementary oligonucleotides spanning nucleotides Ϫ20 to ϩ12 on the sense strand and nucleotides ϩ14 to Ϫ18 on the antisense strand (Table I). A protein⅐DNA complex produced on incubation with HepG2 nuclear extracts (Fig. 6, lanes 1-3, C) was shown to be specific as unlabeled double-stranded oligonucleotide competed successfully (lane 4), but the HNF-4 oligonucleotide did not (lane 5). In COS-7 cell nuclear extracts a distinct but different binding activity was present (lane 6), leading to a complex with different migration pattern.
None of the promoter-reporter fusion discussed above contained this cis-acting site. To test the functional importance of this site, we constructed the new plasmid pLUC-2.5/ϩ40, where the translational initiation codon of the luciferase gene was fused to position ϩ40 in the factor VII coding region. In transient transfection experiments, luciferase expression from construct pLUC-2.5/ϩ40 was 1.9-fold higher than from plasmid pLUC-2.5/-34 (Fig. 7). Furthermore, when the first factor VII exon and part of the first intron up to position ϩ131 were included in plasmid pLUC-712/ϩ131 reporter gene expression was about 2.7 times higher than from plasmid pLUC-712/-34. These experiments suggest that sequences downstream of the putative HNF-4-binding site and the first exon are also involved in expression of factor VII. DISCUSSION During the past few years, a number of studies on promoters of coagulation proteases like prothrombin (20,21,28), factor IX (27, 29 -39), and factor X (3, 18) have been published. The regulation of factor IX gene expression is of particular interest, since the Leyden phenotype of the severe bleeding disorder hemophilia B is caused by several point mutations within the promoter region, of which a number map in the HNF-4-binding site, reducing the affinity of HNF-4 for the promoter. Both the factor IX and the factor X promoter are positively regulated by the liver-specific transcription factor HNF-4.
In this work we show that measurable amounts of factor VII transcripts are present only in liver suggesting that regulation takes place at the transcriptional rather than at the translational level (Fig. 2), and we have defined promoter elements involved in liver-specific expression of the coagulation factor VII present in the 4813 bp of factor VII 5Ј-flanking sequence.
Factor VII Expression Is Tissue-specific-Factor VII promoter transient transfections showed cell line specificity of luciferase expression to human hepatocyte cell line HepG2 (Fig. 2). In contrast, the factor VII promoter is not active at all when compared with the promoterless vector pGL2-basic in non-human COS-7 cells. Although HeLa cells did express luciferase from the factor VII promoter, it seems likely that in HeLa cells the slight decrease of expression upon successive truncations is the effect of a rather unspecific fortuitous transcription initiation in HeLa cells, which is further substantiated by the observation that the heterologous SV40 enhancer only increases expression in HeLa cells when present with 1.6 kb of promoter whereas it increases expression in HepG2 cells with a promoter fragment of only 165 bp (Fig. 3). Thus it is unlikely that identical sequences act as promoters in HepG2 and HeLa cells.
A number of studies show that triglyceride and cholesterol levels in plasma are positively correlated with the risk for coronary heart diseases and thrombosis (5) and the plasma levels of lipoproteins and activity of vitamin K-dependent coagulation proteases are positively correlated with lipid concentrations in plasma (8 -10). The apolipoprotein genes AI, CIII, and AIV form a gene cluster, and, since coregulation has been shown, it has been hypothesized that apoCIII, AI, AII, AIV, and E have evolved from a common ancestor (40,41). Interestingly, the genes encoding factor VII and factor X are also clustered and separated by only 2823 nucleotides (3). The significant similarity of vitamin K-dependent blood coagulation proteins suggests that the genes could have also evolved from a common ancestral gene and that they might be regulated by a common mechanism. The close correlation of FVIIag level and activity of factor IX, factor X, and prothrombin with the lipoprotein metabolism tempted us to compare the factor VII promoter and promoters of apolipoprotein genes and coagulation proteases factor IX and X with respect to common regulatory mechanisms.
Factor VII Expression Is Regulated by Several Factors-Functional studies of promoter deletions in the HepG2 cell line showed that deletion of sequences from Ϫ1601 down to Ϫ1212 increases expression about 2-fold which suggests a negative element is present upstream of Ϫ1212 (Fig. 3). A 25-bp sequence element at position Ϫ1230 to Ϫ1214 that has a similarity of 76% to position Ϫ122 to Ϫ98 of the apoAIV promoter element AIVC is a good candidate for a binding site of such a negatively acting factor. The element AIVC was defined by a footprint covering position Ϫ148 to Ϫ92 in the apoAIV promoter, and proteins binding to the 5Ј-half of the motif at For EMSA assays the 5Ј-protruding ends were blunted and labeled with Klenow polymerase. A, wild-type sequence comprising the HNF-4 binding site. B, analogous to the "Leyden"-phenotype of factor IX deficiency the thymidin at position Ϫ59 was replaced by an adenosin (bold letter). C, oligonucleotide comprising the binding site for the as yet unidentified factor causing footprint FPVII6. The translational initiation codon is underlined.

GTGACGTCTCTAAAGTAGTACCAGAGGGTCCG
Factor VII Promoter position Ϫ142 to Ϫ128 have been identified but not to the 3Ј-half (41).
In our experiments the proximal 165 bp upstream from the ATG translational initiation codon were sufficient for liverspecific expression (Fig. 3), and DNase I protection assays on the proximal 714-bp promoter fragment revealed a set of six footprints of which four map in the first 145 nucleotides (Fig.  5). Removal of the protein-binding sites VII1 and 2 at a distance from 380 to 450 from the translational initiation codon had no effect on the transcriptional activity (compare pLUC-474/-34 to pLUC-237/-34). However, the presence of these bind-ing sites might account for the higher promoter activity in plasmid pLUC-474d3/-34 as compared to pLUC-46/-34 (Fig. 3).
Including sequences up to position ϩ40 in the promoterluciferase construct pLUC-2.5/ϩ40 raised expression levels to about two times the values obtained with the plasmid pLUC-2.5/-34 (Fig. 7). The effect is cell line specific as almost no expression from this plasmid is detectable in HeLa cells. As demonstrated by the strong footprint FPVII6 in DNase I protection experiments (Fig. 5) and by EMSA experiments (Fig. 6), an as yet unidentified regulatory protein binds specifically to this sequence which covers the ATG translational initiation codon.
Inclusion of the entire factor VII exon I and the 5Ј-terminal part of the first intron in plasmid pLUC-712/ϩ131 has a similar effect; plasmid pLUC-712/-34 and pLUC-712/ϩ131 differ 2.7-fold (Fig. 7). This effect is slightly more pronounced than in the plasmid pair pLUC-2.5/-34 and pLUC-2.5/ϩ40 which argues for an additional regulatory function located within exon I or the 5Ј-end of intron I. This is similar to transcriptional regulation of von Willebrand factor which has a positive regulatory region located within the first exon (42) and in apoB where HNF-1 and C/EBP bind to an enhancer in the second intron (43).
C/EBP-␣ and related factors that recognize the same binding sites as C/EBP-␤ play an important role in expression of apolipoproteins AI (44 -46), AII (47,48), B (49 -51), and coagulation factor IX (32,33,38), and we anticipated a similar function of C/EBP-␣ at position Ϫ690 to Ϫ682 at the putative C/EBP-␤/ NF-IL6-binding site in the factor VII expression. However, the DNase I protection assay performed on the factor VII promoter did not show a footprint over the putative C/EBP-binding site (see below), and in transient transfections deletion of this binding site in plasmid pLUC-474/-34 does not alter expression when compared to pLUC-712/-34 (Fig. 3) and probably does not mediate acute phase response of factor VII expression by C/EBP-␤ as examined by induction and cotransfection experiments (data not shown). One explanation for the lack of this interaction could be the very low C/EBP concentration in HepG2 cells, which also accounts for the weak factor IX expression in HepG2 cells (32).
Transcription Factor HNF-4 Binds to the Factor VII Promoter-One of the footprints which we detected corresponds to the liver-enriched transcription factor HNF-4. HNF-4 is an orphan member of the steroid hormone receptor superfamily which has been highly conserved throughout evolution, suggesting that it might respond to an as yet unidentified ligand. Initially, it was characterized as a factor that activates apolipoprotein CIII and transthyretin (TTR) gene expression (26,52) and it also positively regulates numerous genes (for a review, see Ref. 52). HNF-4 exists in two different isoforms as a result of differential splicing from which the largest is by far the most common in HepG2 cells (53). Despite the presence of binding activity specific for the HNF-4 site in all cells investigated, HNF-4 is mainly restricted to hepatocytes. With high levels in liver and moderate levels in kidney and intestine, the expression pattern of HNF-4 reflects the tissue distribution of genes regulated by HNF-4 (54). HNF-4 has been demonstrated to confer liverspecific expression on coagulation proteases factor IX (30,33,34) and factor X (3,18). A number of point mutations in the factor IX promoter between position Ϫ21 and ϩ13 cause the Leyden phenotype of the severe bleeding disorder hemophilia B (Table I; (26,31, and references therein), and the mutations at position Ϫ21 and Ϫ20 disrupt the HNF-4-binding site. As shown by our DNase I protection experiments and EMSA assays, HNF-4 specifically binds to position Ϫ71 to Ϫ49 on the factor VII promoter (Figs. 4 and 5).
In order to prove the functional importance of the HNF-4binding site for factor VII expression, we introduced the Leyden-specific Ϫ20 T to A mutation (ACTTTG 3 ACTTAG ;  Tables I and II) into the HNF-4-binding site of the factor VII promoter which lowered promoter activity in HepG2 cells by 20 -50% (Fig. 3) and reduced the binding of HNF-4 in crude FIG. 6. Electrophoretic mobility shift assay of sequences from position ؊20 to ؉12. The end-labeled double-stranded oligonucleotide comprising nucleotides Ϫ20 to ϩ12. Free oligonucleotide (lane 1, F) migrated at a slower rate after incubation with 10 and 30 g of HepG2 nuclear extract (lanes 2 and 3, C). A 400-fold molar excess of unlabeled oligonucleotide efficiently competed for binding (lane 4). In contrast, the HNF-4 oligonucleotide fails to compete for binding (lane 5). In COS-7 nuclear extracts a distinct but different binding activity is present (lane 6). extracts (Fig. 4). The only slight promoter activity reduction of about 20% seen in pLUC-1.6/-34"Leyden" and pLUC-474/-34"Leyden" as compared to the 50% reduction in pLUC-165/-34"Leyden" could be caused by the positive effect of promoter sequence Ϫ474 to Ϫ355 which also restores promoter activity in plasmid pLUC-474d3/-34 (Fig. 3). The identical point mutation introduced into the HNF-4-binding site of the factor X promoter reduced activity by 80.8%, whereas the totally random mutagenesis of ACTTTG to GACAAT reduced factor X promoter activity by 82.8% (21). The apolipoprotein genes apoAI (55), CIII (56,57), and AIV (41,58) as well as AII (56) and B (51,56) are also positively regulated by HNF-4. Mutagenesis of the HNF-4-binding site in the apoB and apoCIII promoter elements BA1 and CIIIB reduced activity of reporter gene constructs by 98.5 and 92%, and deletion of the elements A1D and AIIJ from the apoAI and apoAII promoters reduced promoter activity by 50% and 30% (56) which is in the range of the effect seen in the factor VII promoter. Since all apolipoprotein genes and the majority of coagulation protease genes are positively regulated by HNF-4 we hypothesize that the HNF-4 concentration is one possible link causing the correlation between lipid metabolism and vitamin K-dependent coagulation proteases. Although the positive effect of HNF-4 in apolipoprotein expression is dependent on additional factors as shown by the synergism between HNF-4 and C/EBP binding to overlapping sites in the apoB promoter (51) and the requirement for additional transcription factors as suggested for the apoAII and apoCIII promoter, this dependence can be overcome in the apoCIII promoter by high level expression of HNF-4 (56). One or more of the strong footprints in the factor VII promoter could also be involved in mediating HNF-4 transactivation or, in analogy to the apoB gene regulation, this function could also be fulfilled by C/EBP, or a hypothetic C/EBP-like factor, binding to the putative site at position Ϫ690 to Ϫ682. None of the additional binding sites for liver-enriched transcription factors like HNF-3 or HNF-1, which have been shown to be involved in, for example, apolipoprotein or prothrombin gene regulation, have a counterpart in the factor VII promoter. Although coordinated liver-specific gene expression certainly involves a complex regulatory network, HNF-4 may play a key role. It is known that HNF-4 is a positive regulator of the liver-specific transcription factor HNF-1, which is involved in expression of genes like prothrombin, ␣1-antitrypsin, and transthyretin (20, 21, 59 -61), and HNF-1 in turn down-regulates expression from its own promoter and other HNF-4-regulated genes like apo-CIII and apoAI (62). The genes apoB, CIII, AI, and AII can be down-regulated by some or all of the steroid receptor superfamily transcription factors ARP-1, EAR-2, and EAR3/COUP-TF which bind with different affinity to HNF-4-binding sites (51, 56 -58), but, as Mietus-Snyder and colleagues concluded, the amount of HNF-4 is significantly higher than the level of ARP-1 and EAR-3 in liver and intestine, and thus the transcriptional activity of the apoCIII gene is at least in part dependent upon the intracellular balance of HNF-4, ARP-1, and EAR-3 (57).
We demonstrate in this paper that, like factor IX and factor X, expression of factor VII is probably regulated by HNF-4. These genes are part of a complex regulatory network that lead finally to coordinated liver-specific expression or repression. In analogy to the regulatory mechanism described for the apolipoprotein genes, additional orphan steroid receptors may also be involved which we are currently investigating.