Two mutations in the promoter region of the human protein C gene both cause type I protein C deficiency by disruption of two HNF-3 binding sites.

Protein C is a vitamin K-dependent zymogen of a serine protease that inhibits blood coagulation by the proteolytic inactivation of factors Va and VIIIa. Individuals affected with protein C deficiency are at risk for thrombosis. Genetic analyses of affected individuals, to determine the cause of the protein C deficiency, revealed a large variety of mutations in the protein C gene, including several in the promoter region of this gene. Comparison of the region around two of these mutations, A G and T A, with transcription factor consensus sequences suggested the presence of two overlapping and inversely oriented HNF-3 binding sites. Direct evidence for the presence of the two HNF-3 binding sites in the protein C promoter was obtained using electrophoretic mobility shift assays and UV cross-linking experiments. These experiments revealed that HNF-3 can bind specifically to both putative HNF-3 sites in the wild-type protein C promoter. Due to the T A mutation, one binding site is completely lost, while the other site still binds HNF-3, but with strongly reduced affinity. As a consequence of the A G mutation, the protein C promoter loses all its HNF-3 binding capacity. Transient transfection experiments demonstrated that the binding of HNF-3 to the protein C promoter is of physiological significance. This followed from experiments in which the introduction of the A G or T A mutation resulted in a 4-5-fold reduced promoter activity in HepG2 cells. Furthermore, transactivation of the wild-type protein C promoter construct with HNF-3 showed a 4-5-fold increased promoter activity in HepG2 cells. In HeLa cells, significant wild-type promoter activity was only observed after transactivation with HNF-3. When a promoter construct containing the T A mutation at position −27 was used, the transactivation potential of HNF-3 was 2-fold reduced in HepG2 cells, whereas in HeLa cells no transactivation was observed. With the promoter construct containing the A G mutation, no transactivation by HNF-3 was found either in HepG2 or in HeLa cells.

Protein C, which is synthesized in the liver as a vitamin K-dependent zymogen of a serine protease, plays an important role in the regulation of the hemostatic system. After activation by the thrombin-thrombomodulin complex, activated protein C inhibits blood coagulation in the presence of protein S (1), phospholipids, and calcium ions through the proteolytic inactivation of factors Va and VIIIa (2,3). Furthermore, activated protein C stimulates fibrinolysis through the neutralization of plasminogen activator inhibitor-1 (4).
The physiological significance of the protein C anticoagulant activity is clearly shown in individuals homozygous or compound heterozygous for protein C deficiency. These individuals suffer from massive disseminated intravascular coagulation or neonatal purpura fulminans (5)(6)(7). Individuals affected by heterozygous protein C deficiency, although more mildly affected, are at risk of thrombophlebitis, deep vein thrombosis, or pulmonary embolism (8,9).
The human protein C gene, located on chromosome 2q13-q14 (10), contains nine exons spanning 11 kilobase pairs of genomic DNA (11,12). Of these nine exons the first and part of the second exon (21 bp) 1 consist of non protein coding sequences. Genetic analyses of individuals with hereditary protein C deficiency revealed a large variety of mutations in the protein C gene, including several in the promoter region of the gene (13).
Transcription of eukaryotic genes by RNA polymerase II involves DNA elements located within the promoter region and transcription factors that associate with these DNA elements (14). Certain transcription factors, such as the TATA box-binding protein (TBP), specify the transcription initiation site (15), whereas others regulate the efficiency of transcription initiation. This latter group of transcription factors comprises both ubiquitous and tissue-specific factors. Because tissue-specific regulation in the liver has been studied extensively, a significant number of liver-specific or liver-enriched transcription factors have been cloned and characterized, including HNF-1␣ (16) and -1␤ (17); HNF-3␣, -3␤, and -3␥ (18); HNF-4 (19); C/EBP␣ (20) and -␤ (21); DBP (22); and LAP (23).
Analyses of promoter mutations causing inherited disease is very useful in identifying transcription factors involved in the transcriptional control of a gene of interest (24 -29). In this study we examined the promoter region of the human protein C gene around two mutations, A Ϫ32 3 G (13, 30) and T Ϫ27 3 A (31), which are known to cause reduced protein C antigen and activity levels in the blood (type I protein C deficiency). We show that the wild-type promoter contains two binding sites for HNF-3 (␣, ␤, or ␥) and that HNF-3 (␣, ␤, or ␥) transactivates the wild-type protein C promoter in both cells of hepatic (HepG2) and epithelial (HeLa) origin in transient transfection assays. Furthermore, we show that both mutations influence the binding of HNF-3 to the protein C promoter and significantly reduce (T Ϫ27 3 A) or completely abolish (A Ϫ32 3 G) transactivation of this promoter by HNF-3.

MATERIALS AND METHODS
Amplification of Genomic DNA-Fragments of the human protein C promoter region, containing 386 bp of 5Ј-flanking sequence, the complete first nontranslated exon (53 bp) and 54 bp of the first intron, were amplified from genomic DNA of different individuals. DNA from an individual not deficient in protein C was used to obtain the wild-type PCR fragment. DNA from the proband of the PC-43-010 pedigree (31) was used to obtain a PCR fragment with the T Ϫ27 3 A mutation, while a patient from the PC-1-La Jolla IV pedigree (13,30) was used to obtain a PCR fragment with the A Ϫ32 3 G mutation. To perform the amplifications we used the following oligonucleotides: 5Ј-CAGCGTC-CCCGGGCTTGTATGGTGGCACATAAATACATGT-3Ј (Ϫ396 to Ϫ357; all nucleotide numbering is relative to the transcription start site (12)) and 5Ј-CTCTTCTCTTCTCCCGGGGGCAGCCCTCCCTCCACACCCC-TCATA-3Ј (ϩ122 to ϩ78). The underlined nucleotides in the oligonucleotide sequence represent modified nucleotides that are not present in the protein C promoter region. These modified nucleotides introduce a SmaI site in the PCR fragment at position Ϫ386. Amplifications were performed in a 50-l reaction mixture containing 10 mM Tris-HCl, pH 8.0, 1 mM MgCl 2 , 50 mM KCl, 350 ng of primers, 100 ng of genomic DNA, 250 M dNTPs, 60 g/ml bovine serum albumin, and 0.3 unit of Taq polymerase. After an initial incubation at 91°C for 4 min, 32 cycles were carried out at 91°C for 1 min, 56°C for 1 min, and 72°C for 1 min.
Plasmid Constructions-The pCAT00 vector (25) was used to make reporter constructs containing wild-type and mutated protein C promoter sequences driving chloramphenicol acetyltransferase (CAT) gene expression. PCR-amplified DNA fragments of 518 bp, spanning nucleotides Ϫ396 to ϩ122 of the human protein C gene, from different individuals were digested with SmaI. The resulting 493-bp fragments, spanning nucleotides Ϫ386 to ϩ107, were cloned into the SmaI site of pCAT00. The reporter construct, containing the wild-type protein C sequence, was named pCwtCAT. The reporter constructs, containing the protein C sequence with the T Ϫ27 3 A or A Ϫ32 3 G mutation, were named pC-27CAT and pC-32CAT, respectively. The integrity of all constructs was verified by sequencing.
Transient Transfection-The differentiated human hepatoma cell line HepG2 (ATCC HB8065) and human epithelial HeLa cells (ATCC CCL2) were cultured in minimal essential medium containing Earle's salts and nonessential amino acids supplemented with 15% heat-inactivated fetal calf serum. Cells were seeded at a density of approximately 1 ϫ 10 5 cells/60-mm tissue culture dish. After 24 h a DNA mixture, containing 6 g of protein C CAT reporter construct, 2 g of ␤-galactosidase expression vector (pCH110; Ref. 32), and 1.5 g of nonspecific plasmid pUC13, was transfected into the cells by the calcium phosphate coprecipitation method (33). For cotransfection experiments, 0.5 g of pUC13 was replaced by 0.5 g of HNF-3 expression vector. Forty-eight hours after transfection, cells were harvested and ␤-galactosidase activity was measured (34). The CAT activity of each construct was determined essentially as described by Seed and Sheen (35) and normalized to ␤-galactosidase activity. All transfections were repeated two to four times in duplicate, with at least two different plasmid preparations, and data from representative experiments are shown.
In Vitro Transcription/Translation-HNF-3 was in vitro transcribed and translated using the TnT-coupled reticulocyte lysate system (Promega) according to the manufacturer's protocol, in a final volume of 50 l. The crude reticulocyte lysate containing translated proteins was used directly in electrophoretic mobility shift assays (EMSAs).
Oligonucleotides Used in EMSAs-The following nucleotides for human protein C gene were used. The following nucleotides for human transthyretin gene were used.
Nucleotide changes compared to the protein C and transthyretin wild-type sequence, respectively, are indicated in boldface type. In the TTR-2x oligonucleotide, we changed seven nucleotides compared to the TTR-1x oligonucleotide to introduce a second HNF-3 consensus sequence (36).
EMSA-EMSAs were performed with 4 l of in vitro transcribed/ translated HNF-3 in a 10-l reaction mixture containing 10 mM HEPES (pH 7.9), 100 mM KCl, 1 mM dithiothreitol, 2.5 mM MgCl 2 , 0.05 mM EDTA, 0.1% (v/v) Nonidet P-40, 10% (v/v) glycerol, 0.03 mg/ml bovine serum allbumin, and 1 ng of 32 P-end-labeled oligonucleotide. In competition experiments different amounts of unlabeled oligonucleotides were included. After an incubation of 25 min at room temperature, free DNA and DNA-protein complexes were separated by electrophoresis on an 8% polyacrylamide gel with 0.33 ϫ TBE buffer at 4°C. Subsequently, the gel was subjected to autoradiography at Ϫ80°C for 16 h.
Antisera produced in rabbits against HNF-3␣ and -3␤ (a gift from R.H. Costa) were used for "supershift" assays, while antiserum against HNF-3␥ (a gift from E. Lai) was used for competition assays. As a negative control we used antiserum against NF-1 (a gift from P. J. Rosenfeld). In each case, 1 l of 5-fold diluted antiserum was added to the reaction mixture prior to the 25-min incubation at room temperature.
Oligonucleotides Used in UV Cross-linking Experiments-The following nucleotides for human protein C gene were used. pCBrU2x: TCCCGGTTCGTTTATAAACACCAATACCT Indicated in boldface type are 5Ј-bromouracil nucleotides, which are incorporated instead of thymidine nucleotides present in the wild-type protein C promoter region at position Ϫ27 (sense strand) or position Ϫ32 (antisense strand).
UV Cross-linking-A fusion protein between glutathione S-transferase (GST) and the HNF-3␤ DNA binding domain (a generous gift from D. G. Overdier) was affinity purified using RediPack GST purification modules (Pharmacia Biotech Inc.) according to the manufacturer's protocol. UV cross-linking experiments were performed with 1 g of the fusion protein (GST-HNF-3␤) in a final volume of 10 l. The reaction mixture contained 10 mM HEPES (pH 7.9), 100 mM KCl, 1 mM dithiothreitol, 2.5 mM MgCl 2 , 0.05 mM EDTA, 0.1% (v/v) Nonidet P-40, 10% (v/v) glycerol, 1 g of dI-dC, and 1 ng of 32 P-end-labeled oligonucleotides. After an incubation of 25 min at room temperature, the reaction mixture was irradiated for 5 min with UV (255 nm) to irreversibly cross-link DNA and protein. Subsequently, the molecular mass of the cross-linked complexes was determined by SDS-PAGE on a 10% polyacrylamide gel and autoradiography at Ϫ80°C for 16 h. As molecular weight markers, the prestained markers of Sigma were used (M r 6,500 -205,000).

RESULTS
To investigate whether the T Ϫ27 3 A and A Ϫ32 3 G mutations influenced the protein C promoter activity, we analyzed the transcriptional activity of the wild-type and mutated promoters in transient transfection assays. Therefore we made constructs containing 493-bp protein C promoter fragments cloned 5Ј to the CAT reporter gene. This protein C fragment contains three polymorphic sites, which are associated with variation in plasma protein C levels (37). Sequencing of both the pC-27CAT and pC-32CAT constructs revealed that the pC-27CAT construct contains the haplotype associated with low protein C plasma levels (CGT), whereas the pC-32CAT construct contains the haplotype associated with high protein C plasma levels (TAA). However, the association is not the consequence of differences in promoter activity. This followed from three independent transient transfection experiments with two different DNA isolations in which no difference in promoter activity was observed between the two haplotypes in the wild-type context (data not shown). Therefore we used the promoter constructs irrespective of the haplotype. After introduction into HepG2 cells, the pC-27CAT construct had an approximately 4-fold reduced promoter activity compared to the wild-type promoter ( Table I). The use of pC-32CAT reduced the promoter activity to about 20% of wild-type activity. When we repeated the transfection experiments with HeLa cells, no CAT activity could be detected with any of the three constructs.
To identify the transcription factor(s) binding to the promoter sequence around the Ϫ32/Ϫ27 region, we compared this sequence to a number of transcription factor consensus sequences. This suggested that the human protein C promoter region contains two putative HNF-3 binding sites, which are overlapping and in reverse orientation (Fig. 1). The first binding site (bs1), located from nucleotide Ϫ26 to Ϫ37, is 100% identical to the HNF-3 consensus sequence (36), while the second binding site (bs2), located from nucleotide Ϫ33 to Ϫ24, showed an identity of 83% with the consensus sequence. The mutations at Ϫ32 or Ϫ27 each occur at nucleotide positions that are invariant among numerous HNF-3 binding sites. Furthermore, they decrease the identity of the promoter sequence with the HNF-3 consensus sequence to 92% and 67% for bs1 and bs2, respectively.
To obtain evidence for the existence of the two putative HNF-3 binding sites, EMSAs were performed. As shown in Fig.  2, HNF-3␣, -3␤, and -3␥ each formed two distinctive complexes with the pCwt oligonucleotide. In the case of HNF-3␣ and -3␤, the majority of HNF-3 binding was localized in a low mobility complex. For HNF-3␥, however, the formation of a high mobility complex seemed slightly favored over the formation of the low mobility complex. The binding activities were all specific since both complexes could be effectively competed only by the corresponding unlabeled oligonucleotides (shown in Fig. 3 for HNF-3␤).
In order to ascertain whether the complex formation is due to HNF-3 binding, we tested specific antisera against HNF-3␣, -3␤, and -3␥. As a consequence of the addition of anti-HNF-3␣ and anti-HNF-3␤ to the reaction mixture, both the low and high mobility complexes were "supershifted" (Fig. 4). Anti-HNF-3␥ mainly competed for HNF-3 binding, while some fraction of the complexes was also "supershifted." The same experiments with NF-1 antiserum showed no "supershifted" complex or competition.
Next, we determined the influence of the mutations on the binding affinity of HNF-3 to the promoter. This showed that as a consequence of the introduction of the Ϫ27 mutation, only the high mobility complex could be formed by HNF-3␣, -3␤, and -3␥ (Fig. 5A). In addition, for all HNF-3 isoforms the amount of high mobility complex formed did not increase as a consequence of the loss of formation of the low mobility complex, indicating a significantly reduced binding affinity. Neither HNF-3␣, -3␤, nor -3␥ formed a complex with the oligonucleotide containing the Ϫ32 mutation (Fig. 5B).
To investigate whether the two complexes formed by HNF-3 with the pCwt oligonucleotide are the consequence of the binding stoichiometries of one and two HNF-3 protein monomers, respectively, in the high and low mobility bands, we performed EMSAs with HNF-3 and the TTR-1x oligonucleotide. The TTR-1x oligonucleotide contains the HNF-3 consensus sequence from the transthyretin promoter (38) and binds one HNF-3 monomer (39). As shown in Fig. 6, only the high mobility complex was formed, indicating that the high mobility complex is formed due to binding of one HNF-3 protein. We next modified the TTR-1x oligonucleotide to create an oligonucleotide containing two HNF-3 consensus sequences in the same context as in the protein C promoter (TTR-2x). EMSAs with   this TTR-2x oligonucleotide showed the appearance of high and low mobility complexes, just as observed with the pCwt oligonucleotide (Fig. 6).
To explore further the nature of the low and high mobility complexes, we performed UV cross-linking experiments. In these experiments we used a GST-HNF-3␤ fusion protein and oligonucleotides containing 5Ј-bromouracil nucleotides in both strands (pCBrU2x), in one strand only (pCBrU1x-27 or pCBrU1x-32), or in neither of the two strands (pCBrU0x). The 5Ј-bromouracil was incorporated in place of the thymidines at the locations of the two mutations associated with protein C deficiency, i.e. at position Ϫ27 in the sense strand and/or position Ϫ32 in the antisense strand, on the assumption that these thymidines play important roles in HNF-3 binding. After UV irradiation the cross-linked complexes were separated on a SDS-PAGE gel. As shown in Fig. 7 (lane 1), the oligonucleotide containing two 5Ј-bromouracil nucleotides was cross-linked into complexes of approximately 55 and 110 kDa. When only a single modified nucleotide was present in either strand, only a 55-kDa complex was formed (lanes 2-4), whereas no complex was formed when the oligonucleotide did not contain 5Ј-bromouracil nucleotides. Boiling of the samples prior to SDS-PAGE resulted in diminishment of the 110-kDa complex, whereas the intensity of the 55-kDa band was not affected (lanes 5-8).
Finally, we investigated whether HNF-3 was able to transactivate the protein C promoter. HepG2 and HeLa cells were cotransfected with the protein C reporter constructs pCwtCAT, pC-27CAT, or pC-32CAT and expression vectors for HNF-3␣, -3␤, or -3␥. As shown in Fig. 8A, cotransfection of HepG2 cells with HNF-3␣ and -3␥ increased transcription from the wildtype protein C promoter about 4-fold, whereas HNF-3␤ transactivated the promoter activity about 5-fold. In contrast when the pC-27CAT construct was used, the transactivation potential of HNF-3␣, -3␤, and -3␥ was reduced to about a factor of 1.5, 3, and 1.3, respectively (Fig. 8A). Essentially no transactivation with HNF-3␣ and -3␥ was observed using the pC-32CAT construct, while HNF-3␤ transactivated this construct approximately 2-fold. In Fig. 8B it can be seen that HNF-3␣, -3␤, and -3␥ all transactivated the wild-type protein C promoter in HeLa cells. As shown in HepG2 cells, HNF-3␤ seems to be the most potent transactivater. Finally, in HeLa cells, none of the HNF-3 family members could transactivate the promoter constructs containing one of the mutations (Fig. 8B). DISCUSSION The T Ϫ27 3 A and the A Ϫ32 3 G mutations in the protein C promoter region are both associated with type I protein C deficiency (13,30,31). The aim of this study was to determine the cause of this association. We show that both mutations significantly reduce protein C promoter activity in hepatic HepG2 cells, whereas in non-hepatic HeLa cells no constitutive    1 and 5), pCBrU1x-32 (lanes 2  and 6), pCBrU1x-27 (lanes 3 and 7), and pCBrU0x (lanes 4 and 8) were incubated with affinity-purified GST-HNF-3␤ fusion protein. The complexes formed after UV cross-linking for 5 min were either directly loaded (lanes 1-4) or boiled for 5 min and then loaded (lanes 5-8) onto a 10% reducing SDS-PAGE gel. The complexes of interest are indicated by arrows.
promoter activity is observed. This tissue-specific promoter activity indicates the necessity of liver-enriched (or liver-specific) transcription factors in the expression of the protein C gene.
In the protein C promoter region, two putative HNF-3 binding sites were identified by comparing the sequence of the promoter region around the Ϫ27 and Ϫ32 mutations with transcription factor consensus sequences. Both binding sites were largely overlapping and reversely orientated. The presence of the T Ϫ27 3 A mutation, which decreases the identity of the promoter sequence with both HNF-3 consensus sequences, abolishes one binding site, whereas the other binding site is still capable of binding HNF-3, although with clearly reduced affinity. Due to the presence of the A Ϫ32 3 G mutation, which also decreases the identity between promoter region and HNF-3 consensus sequence, no HNF-3 binding was observed. Therefore, we conclude that the adenosine at position Ϫ32 in the protein C promoter is essential for HNF-3 binding. In vitro, with relative large amounts of HNF-3 present the thymidine at position Ϫ27 seems very important but not essential for HNF-3 binding.
Evidence for the simultaneous binding of two HNF-3 monomers to the wild-type protein C promoter was obtained from UV cross-linking experiments. In these experiments the naturally occurring nucleotides at position Ϫ32 and Ϫ27 were replaced by 5Ј-bromouracil nucleotides, which dramatically enhance the irreversible cross-linking of proteins to DNA after UV irradiation (40). After separation on a SDS-PAGE gel, we observed complexes of approximately 110 kDa, corresponding to the oligonucleotide with two HNF-3 monomers and complexes of 55 kDa, corresponding to the oligonucleotide with one HNF-3 monomer. Boiling of the samples prior to loading resulted in the disappearance of the 110-kDa complex, whereas the 55-kDa complex was not affected. We interpret the loss of the 110-kDa complex as due to denaturing of the doublestranded oligonucleotides with as a consequence the formation of two complexes of approximately 55 kDa, consisting of one HNF-3 monomer linked to a single-stranded oligonucleotide.
With the use of x-ray crystallography, HNF-3 has been found to interact with the transthyretin promoter as a monomer over a linear distance of about 40 Å along the axis of the double helix (39). As stated above, the wild-type protein C promoter binds two HNF-3 monomers at largely overlapping binding sites. Since these binding sites are separated 5 bp (or about half a turn of the double helix), we hypothesize that the HNF-3 monomers bind at opposite faces of the DNA helix (Fig. 9).
The binding of two or more HNF-3 monomers close to each other has been reported previously (41)(42)(43). Furthermore, it has been shown that HNF-3 transcription factor binding sites often overlap with other transcription factor binding sites (41,(43)(44)(45). However, this report is novel in demonstrating the binding of two HNF-3 monomers to two largely overlapping HNF-3 binding sites. The functional role of this motif of overlapping binding sites remains unknown.
Both the T Ϫ27 3 A and A Ϫ32 3 G mutations influence the binding of HNF-3 to the protein C promoter, and the transactivation capacity of the promoter is also influenced. As shown by transfection assays in HepG2 cells, the Ϫ27 mutated promoter is transactivated to a lesser extent by HNF-3 than the wild-type promoter. This indicates that the residual binding of HNF-3 to the Ϫ27 mutated promoter is still sufficient to transactivate the protein C promoter. Transfection experiments in HeLa cells show that only the wild-type promoter is transactivated by HNF-3. The observed stimulation of protein C transcription by exogenous HNF-3 in HepG2 cells, despite the presence of HNF-3 proteins in these cells, could be attributed to the limiting amounts of HNF-3 in these cultured cells (46,47).
Another liver-enriched transcription factor known to be involved in protein C gene expression is HNF-1 (48). This transcription factor is, according to Berg et al., necessary but not sufficient for protein C gene expression. However, our observations in HeLa cells, which do not express endogenous HNF-1, argue against the necessity of HNF-1. Importantly, HNF-1 is expressed in HepG2 cells as opposed to HeLa cells. This differ- ence in HNF-1 expression might explain the different response of the protein C promoter containing the Ϫ27 mutation to HNF-3 in HepG2 and HeLa cells. In HepG2 cells complex formation between HNF-1 and HNF-3 can occur with transactivation of the promoter as a consequence. In HeLa cells such complex formation between HNF-1 and HNF-3 cannot occur, and consequently no transactivation takes place. Therefore, we hypothesize that in vivo a functional interplay occurs between HNF-1 and HNF-3 in the transcriptional regulation of the protein C promoter. This hypothesis is supported by the fact that the HNF-1 binding site in the protein C promoter is localized directly downstream of the HNF-3 binding sites. Cooperation between HNF-3 and other factors bound to closely situated DNA binding sites have been reported previously (43,49,50).
HNF-3 consists of a family of liver-enriched transcription factors, called HNF-3/forkhead. During liver differentiation the members of this family show a different expression pattern. In fetal cells HNF-3␣ and HNF-3␥ are more abundant than HNF-3␤, whereas in adult liver cells HNF-3␤ and HNF-3␥ are most abundant (47). In vitro HNF-3␣, -3␤, and -3␥ seem to bind with a similar affinity to the protein C promoter. This excludes the different expression pattern of the HNF-3 proteins as a possible regulatory mechanism for protein C gene expression. More likely, the fact that all HNF-3 proteins bind with the same affinity to the protein C promoter ensures the presence of protein C independent of the age of the organism.
Recently, Qian et al. (45) showed that acute-phase livers exhibit a dramatic reduction of HNF-3␣ (95% decrease) and HNF-3␤ (20% decrease) expression. Furthermore, they showed that this reduced HNF-3␣ and -3␤ expression coincided with reduced expression of one of their target genes, the transthyretin gene. In this report we showed that the protein C gene is also a target gene of HNF-3␣ and -3␤. This indicates that the observed reduction of protein C activity during septic shock (51)(52)(53)(54), severe infection (53), and after major surgery (55) might well be caused by a decrease in protein C transcription.
Sequence comparison of the protein C promoter region around position Ϫ27/Ϫ32 with transcription factor consensus sequences revealed more than just the presence of putative HNF-3 sites. The sequence also contains a core sequence that resembles the TBP consensus sequence (TATA(A/T)A(A/T)) (56). This transcription factor, which is essential for basal transcription (57), might therefore bind to the core sequence. However, due to the Ϫ27 mutation the identity with the TBP consensus is increased (from 71 to 86%), which argues against the hypothesis that this mutation interferes with TBP binding and thereby would be responsible for the protein C deficiency in the patients.
In conclusion, we have identified the liver-enriched transcription factor HNF-3 as a potential physiologic regulator of protein C gene expression. Furthermore, we propose a functional interplay between HNF-3 and HNF-1 to drive protein C gene expression.