Vitamin D-binding Protein Gene Transcription Is Regulated by the Relative Abundance of Hepatocyte Nuclear Factors 1α and 1β*

Vitamin D-binding protein (DBP)/Gc-globulin, the major carrier of vitamin D and its metabolites in blood, is synthesized predominantly in the liver in a developmentally regulated fashion. By transient transfection analysis, we identified three regions in the 5′-flanking region of the rat DBP gene, segments F-2, B, and A, that contain tissue-specific transcriptional determinants. Gel mobility shift and DNase I footprinting analyses showed that all three regions contained binding sites for the hepatocyte nuclear factor 1 (HNF1), a transcriptional regulator composed of HNF1α and HNF1β hetero- and homodimers. The activity of the most proximal segment A (coordinates −141 to −43) was DBP promoter-specific, position-dependent, and positively controlled by HNF1α. In contrast, the two more distal determinants (segments F-2 and B; coordinates −1844 to −1621 and −254 to −140, respectively) acted as classical enhancers in transfected hepatocyte-derived HepG2 cells; their activities were promoter- and orientation-independent, and disruption of their respective HNF1-binding sites resulted in marked loss of DBP gene expression. Remarkably, the activities of these two distal elements depended upon the relative levels of HNF1α and HNF1β; HNF1α had a major stimulatory effect, whereas HNF1β acted as a trans-dominant inhibitor of HNF1α-mediated enhancer activity. These results suggested that the net expression of the DBP gene reflected a balance between the two major HNF1 species; the relative abundance of HNF1α and HNF1β proteins in a cell may thus play a critical role in determining the pattern of gene expression.

Vitamin D-binding protein (DBP)/Gc-globulin, the major carrier of vitamin D and its metabolites in blood, is synthesized predominantly in the liver in a developmentally regulated fashion. By transient transfection analysis, we identified three regions in the 5-flanking region of the rat DBP gene, segments F-2, B, and A, that contain tissue-specific transcriptional determinants. Gel mobility shift and DNase I footprinting analyses showed that all three regions contained binding sites for the hepatocyte nuclear factor 1 (HNF1), a transcriptional regulator composed of HNF1␣ and HNF1␤ hetero-and homodimers. The activity of the most proximal segment A (coordinates ؊141 to ؊43) was DBP promoter-specific, position-dependent, and positively controlled by HNF1␣. In contrast, the two more distal determinants (segments F-2 and B; coordinates ؊1844 to ؊1621 and ؊254 to ؊140, respectively) acted as classical enhancers in transfected hepatocyte-derived HepG2 cells; their activities were promoter-and orientation-independent, and disruption of their respective HNF1-binding sites resulted in marked loss of DBP gene expression. Remarkably, the activities of these two distal elements depended upon the relative levels of HNF1␣ and HNF1␤; HNF1␣ had a major stimulatory effect, whereas HNF1␤ acted as a trans-dominant inhibitor of HNF1␣-mediated enhancer activity. These results suggested that the net expression of the DBP gene reflected a balance between the two major HNF1 species; the relative abundance of HNF1␣ and HNF1␤ proteins in a cell may thus play a critical role in determining the pattern of gene expression.
Vitamin D-binding protein (DBP) 1 is a monomeric, multifunctional glycoprotein first identified as the group-specific component of serum or Gc-globulin (1). It is essential to the transport of vitamin D sterols in the blood and to the removal of plasma actin monomers released to the blood subsequent to tissue damage. DBP also contributes to complement C5a-mediated chemotaxis, macrophage activation, and fatty acid transport (reviewed in Refs. 2 and 3). The DBP gene is a member of the multigene family that includes albumin, ␣-fetoprotein, and ␣-albumin (4,5). The members of this gene family are tightly linked on chromosome 4 in human (6) and chromosome 14 in rat (r) (7) and encode proteins with conservation of both primary and secondary structures. All four genes in this family are predominantly expressed in the liver. During embryonic development, expression of the rat albumin, ␣-fetoprotein, and DBP genes is induced in the yolk sac and maintained in the fetal liver (8), whereas ␣-albumin expression begins in the liver during the subsequent perinatal period. ␣-Fetoprotein is selectively silenced at the end of the fetal period, whereas ␣-albumin, albumin, and DBP expression remains high in the liver throughout adult life (9). 2 Although mechanisms underlying the transcriptional control of the albumin and ␣-fetoprotein genes have been studied by a number of laboratories, the basis for the tissue-specific and developmentally regulated expression of the DBP gene remains unexplored.
Tissue-specific gene expression is predominantly regulated at the level of transcription initiation (reviewed in Ref. 10). Four families of transcription factors have been identified that are involved in liver-specific gene expression: hepatocyte nuclear factor 1 (HNF1), CCAAT/enhancer-binding protein (C/ EBP), HNF3, and HNF4 (reviewed in Refs. 11 and 12). Expression of these factors is not fully restricted to the liver, and the specificity and levels of gene expression in adult hepatocytes appear to be mediated by specific combinations of trans-activators. Whereas the HNF3 proteins bind DNA as a monomer, members of each of the remaining hepatocyte-enriched transcription factor families bind DNA as homo-and heterodimers. The composition of these dimers within a family and their potential competition for shared DNA-binding sites may impose additional complexity on hepatocyte-gene regulation. In addition to the transcription-enhancing activity usually attributed to these hepatocyte-enriched factors, the presence of repressor-like molecules has been suggested by somatic cell hybrid studies demonstrating extinction of liver-specific gene expression when differentiated hepatoma cells are fused to dedifferentiated hepatoma cells or fibroblasts (13). Therefore, the fidelity and level of liver-specific expression is ensured by both positive and negative gene regulation.
The hepatocyte-enriched transcription factor HNF1 has been studied in detail. HNF1 was initially identified as a nuclear protein that binds to an element required for liver-specific transcription of the ␤-fibrinogen gene (14). It has been subsequently found to interact with sites in promoters as well as enhancer elements in a number of hepatocyte-restricted genes including albumin and ␣-fetoprotein (15)(16)(17)(18)(19). The HNF1 proteins have an amino-terminal dimerization domain, POU-like homeodomain with DNA binding activity, and carboxyl-termi-nal transcriptional activation domain. Two members of the HNF1 transcription factor family, HNF1␣ (also known as HNF1, LF-B1, HP-1, APF, and A box factor) and HNF1␤ (also known as vHNF1, vAPF, and LFB3) (20 -22), have been characterized in detail, and they bind DNA as hetero-or homodimers. In addition to hepatic expression, HNF1␣ and HNF1␤ are also found in nonhepatic tissues including kidney, intestine, and stomach (23,24); the relative concentrations of these two proteins differ markedly from tissue to tissue and are affected by developmental stimuli (21). Whether the relative levels of these two proteins impact on expression of specific genes has not been defined.
The rDBP gene including 2,196 base pairs (bp) of the 5Јflanking region has been previously isolated and structurally characterized (25). Here three segments involved in rDBP expression within this 5Ј-flanking region have been identified by transient transfection studies. Gel mobility shift assays and DNase I footprinting analyses demonstrated that their functional activities can be attributed to the presence of three corresponding HNF1-binding sites. The functional importance of these HNF1 sites was tested by mutational analyses, and opposing effects of HNF1␣ and HNF1␤ on rDBP gene expression were demonstrated.  (26) and manufacturer's specifications were used for the isolation and manipulation of DNA. Expression vectors for mouse HNF1␣ (27) and HNF1␤ (21) were gifts from Dr. G. R. Crabtree (Stanford University School of Medicine); the carboxylterminal transcriptional activation domain of HNF1␤ expression vector was sequenced and confirmed to be identical to the published sequence (21) and homologous to human vHNF1-B (28). The rat pAFP-CAT vector (15) was a gift from Dr. J.-L. Danan (Center National de La Recherche Scientifique, France). The pSVAO(X)-CAT vector was a gift from Dr. T. R. Kadesch (University of Pennsylvania). pSVAO(X)-CAT is a derivative of pSVAO-CAT (29) where SV40 sequences between PstI and BamHI from pSVAO-CAT were removed, and the polylinker (PstI, SalI, XbaI, and BamHI) was added. Oligonucleotides used for gel mobility shift assays and PCR were synthesized by Life Technologies, Inc. PCR was performed with Taq DNA polymerase and 10ϫ PCR buffer containing 15 mM MgCl 2 (Boehringer Mannheim), using a Perkin-Elmer thermal cycler. Autoradiographic signals were quantitated by PhosphorImager (Molecular Dynamics).

Materials-Restriction
Plasmid Constructions-The 5Ј-flanking region of the rDBP gene (2.7-kilobase pair EcoRI/HindIII fragment of rDBP genomic clone C1) (25) was digested with BstNI (coordinate ϩ53) and the following secondary restriction enzymes: SstI (Ϫ1926), NsiI (Ϫ1465), BamHI (Ϫ1077), NheI (Ϫ670), RsaI (Ϫ253), BstNI (Ϫ141), and RmaI (Ϫ41). Coordinates used in this study are the same as published (25). Double digestions generated fragments with identical 3Ј termini and a series of nested 5Ј termini. Each fragment was blunt-ended with mung bean exonuclease and T4 DNA polymerase and then ligated to HindIII linkers (New England Biolabs). After inactivation of ligase, the DNA fragments were digested with HindIII and ligated to the pSVAO-CAT(X) vector previously linearized with HindIII. Eight different CAT constructs containing 5Ј-deletions of the rDBP promoter region with same 3Ј-end at ϩ53 were generated (Fig. 1A). Seven subsegments of the 5Ј-flanking region (A, Ϫ141 to Ϫ43; B, Ϫ254 to Ϫ140; C, Ϫ667 to Ϫ149; D, Ϫ667 to Ϫ259; E, Ϫ1077 to Ϫ664; F, Ϫ1899 to Ϫ1490; G, Ϫ2196 to Ϫ1927, Fig. 1A) were amplified by PCR using the appropriate specific primers each containing HindIII recognition sequences (5Ј-CCCAAGCTT-3Ј) at their 5Ј-ends. These fragments were cloned into the HindIII site downstream of the SV40 promoter and CAT gene in plasmid pCAT promoter (Promega Corp.) in order to study their activities with a heterologous promoter. pCAT promoter was also digested with BglII upstream of the CAT gene, blunt-ended with Klenow, and ligated to each of the blunt-ended PCR fragments. To generate a second series of CAT constructs in which the seven segments of the rDBP 5Ј-flanking region were linked to the rDBP minimal promoter driving CAT, the pDBP/CAT vector was constructed by cloning the rDBP minimal promoter (Ϫ39 to ϩ53) into the XbaI site of pCAT Basic (Promega Corp.).
Each of the seven amplified fragments was then similarly cloned into either HindIII (upstream of CAT gene) or BamHI (downstream of CAT gene) sites of the pDBP/CAT vector. The pDBP/CAT constructs, containing the putative cis-elements (fragments A-G and F-1 to F-4; Fig.  3A) cloned upstream of the CAT gene, were sequenced by the DNA and Protein Core, University of Pennsylvania Veterinary School. Only pDBP/CAT containing fragment E differed from the published sequence; Ϫ856 CGTCTT Ϫ860 was CTTGCGCTT. Likewise, four overlapping subsegments of F (F-1, Ϫ1899 to Ϫ1757; F-2, Ϫ1844 to Ϫ1621; F-3, Ϫ1758 to Ϫ1564; and F-4, Ϫ1628 to Ϫ1490) were cloned into the pDBP/CAT vector at the HindIII site upstream of the CAT gene (Fig.  3A) and confirmed by sequencing.
HNF1-binding site mutations in region B (fragment mutB and plasmid Ϫ2196DBPmB/CAT) or F-2 (fragment mutF-2 and plasmid Ϫ2196DBPmF-2/CAT) (see Fig. 7B) were generated using the splicingby-overlap extension method (30). To generate mutB, 5Ј (2025 bp) and 3Ј (238 bp) fragments for the overlap splicing were amplified using Pfu DNA polymerase. Each fragment was amplified with oligos Ϫ2196/ 5ЈmB and 3ЈmB/ϩ53 as primer pairs using the Ϫ2196DBP/CAT construct as the template (Table I and Fig. 1A). The conditions were as follows: denaturation at 94°C for 3 min, followed by 25 cycles at 94°C for 1 min, 50°C for 1 min, and 72°C for 2 min with a final 5-min extension at 72°C. The two amplified fragments were gel-purified, mixed, denatured at 94°C for 3 min, annealed at 40°C for 1 min, and extended at 72°C for 2 min. The full-length mutB fragment was then generated by amplification between outside primers, oligos Ϫ2196 and ϩ53 (Table I), with annealing at 60°C. MutF-2 was similarly generated using oligos 5ЈmF-2 and 3ЈmF-2 (Table I). These two mutant fragments were digested with HindIII and ligated into the HindIII site of pSVAO-CAT(X). The plasmid Ϫ2196DBPmF-2&B/CAT (Fig. 7B) containing HNF1 site mutations at both regions F-2 and B was generated by releasing the 1562-bp fragment containing mutation in region F-2 from Ϫ2196DBPmF-2/CAT by SacI and BstXI digestion and ligating it to the 5.4-kilobase pair SacI/BstXI doubly digested fragment of Ϫ2196DB-PmB/CAT containing the mutated region B and the vector sequences.
Cell Culture and Transfections-HepG2 and NIH3T3 cells were obtained from American Type Culture Collection and cultured to 60% confluency in Eagle's minimal media supplemented with 2 mM L-glutamine, 100 IU/ml penicillin/streptomycin (Mediatech), and 10% fetal bovine serum (Life Technologies, Inc.). Five to 20 g of CAT construct and 4 g of pCH110 containing the ␤-galactosidase coding sequence under control of the SV40 early promoter were transfected into cells by the calcium phosphate precipitation method (31,32). Cells were harvested 48 -72 h after the transfection and were lysed with 0.9 ml of reporter lysis buffer (Promega Corp.), and lysates were used for enzyme assays. The amount of cell lysate used in CAT assays was adjusted based on the transfection efficiency determined by ␤-galactosidase activity (33). CAT activities were measured as described previously (29,34) and quantitated by PhosphorImager.
Preparation of Nuclear Extracts and Northern Analysis-Nuclear extracts were prepared as described previously (35), except that all the solutions contained a protease inhibitor mixture (Boehringer Mannheim). Protein concentrations were measured using Bio-Rad protein assay solutions following the manufacturer's instructions (Bio-Rad). A Northern blot analysis of liver RNA from wild type, HNF1␣ Ϫ/Ϫ , and HNF1␣ ϩ/Ϫ mice was generously carried out by Drs. M. Pontoglio and M. Yaniv (36), using a 32 P-labeled 875-bp EcoRI rDBP cDNA fragment (37). Equal loading of lanes was documented by ethidium bromide staining for rRNA.
Gel Mobility Shift Assays-Gel retardation assays were performed using 1-14 g of nuclear proteins. Probe fragments, ranging in size from 98 to 223 bp, were prepared as follows. The appropriate ciselements cloned at the HindIII site of the pDBP/CAT plasmid were released by HindIII digestion and labeled using Klenow enzyme, [␣-32 P]dCTP, and [␣-32 P]dATP by incubating at room temperature for 30 min. The labeled fragments were subjected to 6% polyacrylamide gel electrophoresis. DNA bands with the expected sizes were cut out of the gel and purified (26). The oligonucleotide ␤28 (Table I) was labeled using T4 polynucleotide kinase and [␥-32 P]ATP by incubating at 37°C for 30 min. Unincorporated radioactive nucleotides were removed by a G-25 spin column (Amersham Pharmacia Biotech). Nuclear extracts were incubated with 20,000 cpm of the labeled DNA probe in a 20-l reaction mixture containing 12 mM HEPES, pH 7.8, 0.1 M KCl, 1 mM EDTA, 1 mM dithiothreitol, 5% glycerol, 50 ng of bovine serum albumin (BSA; Boehringer Mannheim), and 1-2 g of poly(dI-dC) (Amersham Pharmacia Biotech). The binding mixture was incubated at room temperature for 20 min and run on a 5% nondenaturing polyacrylamide gel in 1ϫ TBE buffer at 10 V/cm. Gels were dried and exposed to an x-ray film at room temperature without an intensifying screen.
Competitors and Probes in the Gel Mobility Shift Assay-The sequences of oligos ␤28, wtF-2, wtB, and wtA containing HNF1-binding sites present in the ␤-fibrinogen gene, the rDBP gene region F-2, B, and A, respectively, are listed in Table I. mutF-2 and mutB are the corresponding oligonucleotides with mutations in their HNF1-binding sites. Sense and antisense oligonucleotides used as competitors or as a probe were synthesized and annealed in 100 mM NaCl, 50 mM Tris-HCl, pH 7.5, by heating at 90°C for 3 min and then cooling down to 25°C at 1°C per 3 min in the thermal cycler. Nonspecific competitor, "␤-lac," was a 222-bp fragment of the coding region of the ␤-lactamase gene isolated from the pCAT Basic (Promega Corp.) vector by AvaII digestion. Fragments "F-2a" and "F-2b" were PCR-amplified subregions of fragment F located between Ϫ1844 to Ϫ1757 and between Ϫ1758 to Ϫ1621 of the rDBP 5Ј-flanking region, respectively (Fig. 4B).
DNase I Footprinting-pDBP/CAT plasmids containing fragment F-2, B, or A at the HindIII site were digested with SalI (for F-2 and B) or XbaI (for A) and labeled with Klenow enzyme, [␣-32 P]dCTP, and [␣-32 P]dATP by incubating for 30 min at room temperature. Unincorporated radioactive nucleotides were removed by a G-50 spin column, and the DNAs were precipitated with ethanol and digested with BanI. The labeled fragments were separated from the contaminants by polyacrylamide gel electrophoresis, excised, and purified (26). The binding reaction consisted of the following components in a final volume of 65 l: 10 mM Tris-HCl, pH 8.0, 5 mM MgCl 2 , 1 mM CaCl 2 , 2 mM dithiothreitol, 50 g/ml BSA, 100 mM KCl, 5% glycerol, 2 g of poly(dI-dC), end-labeled DNA fragment, and 50 -100 g of HepG2 nuclear proteins. The mixture was incubated at room temperature for 20 min and treated with DNase I (Boehringer Mannheim) for 30 s at room temperature. The amount of DNase I was adjusted empirically for each extract to produce an even pattern of partial cleavage products, generally between 0.5 and 0.02 units of the DNase I. Reactions were stopped by the addition of 2 volumes of 30 mM EDTA, 375 mM NaCl, and 0.75% SDS. Excess proteins were digested by adding 2 g of proteinase K (Boehringer Mannheim) followed by an incubation for 30 min at 42°C. Nucleic acids were extracted with 1 volume of phenol:chloroform (1:1) and ethanol-precipitated. They were dissolved in 99% formamide and 10 mM EDTA with tracking dyes, heated at 90°C for 5 min, and loaded on 8% polyacrylamide, 7 M urea gels. The DNA sequence in adjacent lanes was established by chemical cleavage of phosphodiester bonds 3Ј of A and G as described previously (38).

Identification of Two
Tissue-specific Enhancer Segments 5Ј to the DBP Promoter-To identify determinants important in rDBP transcriptional control, a series of seven sequential 5Јdeletions of the previously sequenced rDBP 2196-bp 5Ј-flanking region contiguous with the promoter and 5Ј-untranslated region (25) were generated and fused to a promoterless CAT reporter gene (Fig. 1A). Plasmids containing these deletions were introduced into human hepatoma HepG2 cells or into NIH3T3 fibroblasts, and CAT activity (normalized for transfection efficiency) was determined. The full-length Ϫ2196DBP/ CAT construct was highly expressed in the HepG2 cells but inactive in NIH3T3 (see below). Deletions of sequences between Ϫ1926 and Ϫ1465 and between Ϫ253 and Ϫ141 each resulted in a marked decrease in activity in the HepG2 cells (Fig. 1B). These data suggested the presence of at least two positive regulatory elements 5Ј to the rDBP promoter.
To confirm and extend the 5Ј-deletion study, seven subsegments of the DBP 5Ј-flanking region were generated with termini roughly coincident with the end of the deletion constructs (fragments A-G; Fig. 1A). Each fragment was linked to the rDBP basal promoter (Ϫ39 and ϩ53) and tested for transcriptional activity. Consistent with the 5Ј-deletion results, fragments F (Ϫ1899 to Ϫ1490) and B (Ϫ254 to Ϫ140) increased CAT gene expression 2-6-fold in HepG2 cells ( Fig. 2A). This effect was independent of the position of the fragment. In contrast, fragment A (Ϫ141 to Ϫ43) was active only in its native position, 5Ј to the basal promoter ( Fig. 2A). Linkage of fragments F and B to a heterologous SV40 promoter (Fig. 2B) also resulted in 2-4-fold increases in CAT activity in HepG2 regardless of its position, whereas fragment A was inactive in this context. Fragment E containing half of a rodent B2 small repetitive element (25), fragment G, and fragment D all demonstrated negative activity on the rDBP promoter ( Fig. 2A). Fragment C encompassed negative element D and positive element B. The net effect of fragment C on the DBP promoter was negative. However, in the deletion series, elimination of Ϫ670 to Ϫ253, which corresponds to D, did not result in marked enhancement. This suggested that the negative activity in D might not be of major importance. All regions of the rDBP 5Ј-flanking region lacked activity in transfected NIH3T3 cells (Fig. 2B). The tissue-specific, position-and promoter-independent enhancer activity identified in fragments F and B, and a tissue-specific, position-and promoter-dependent activity identified in fragment A were further studied.
Prior to detailed characterization, the active region in the 409-bp fragment F was further sublocalized by linking each of four overlapping subfragments (F-1 through F-4) to the DBP promoter-driven CAT reporter plasmid (Fig. 3A). Analyses of activities in transfected HepG2 cells mapped the enhancer activity to the F-2 subfragment; a marginal enhancing effect by F-3 was not significant (Fig. 3B). The activity of F-2 was independent of its position relative to the promoter (data not shown). Of note, a repressor activity was found in the adjacent F-4 fragment (Fig. 3B). Activity of the F-2 subfragment was 2-fold greater than the activity of its parental F fragment ( 2A versus Fig. 3B), consistent with elimination of the negative activity in fragment F-4. The F-4 fragment lacked a negative effect when oriented 3Ј to the reporter gene (data not shown), possibly explaining the greater activity of the parental fragment F located 3Ј (5-fold increase) versus 5Ј (2-fold) to the DBP promoter ( Fig. 2A). Thus, segments with rDBP promoter transcription-enhancing activity were localized to the 223-bp F-2 fragment, the 114-bp B fragment, and the 98-bp A fragment.
Functional HNF1-binding Sites within Regions F-2, B, and A-HNF1 is one of the major transcriptional activators in hepatocytes. The HNF1 consensus binding site is a partially degenerate palindromic sequence consisting of two 7-base pair inverted repeats separated by a single variable nucleotide, 5Ј-GGTTAATnATTAAC(a/c) -3Ј (11,39). The 5Ј-half of the pal-FIG. 2. Internal fragments B and F contained tissue-specific, positionand promoter-independent enhancer elements. A, constructs containing rDBP 5Ј-flanking fragments A-G (see Fig. 1A) were cloned either 5Ј or 3Ј to the minimal rDBP promoter (Ϫ39/ϩ53) linked to the CAT reporter gene. Each construct was transfected into HepG2 cells, and mean normalized CAT activities were determined (the number of repeat experiments is noted in each case). B, the same set of seven fragments was cloned into a CAT reporter gene driven by the SV40 early promoter. Each construct was transfected into HepG2 or NIH3T3 cells. The CAT activities in A and B were expressed in arbitrary units relative to the parental construct containing minimal rDBP or SV40 promoter alone (indicated as 100).
FIG. 3. Sublocalization of fragment F enhancer activity to subfragment F-2 and identification of repressor activity in subfragment F-4. A, the top line represents the Ϫ2196 rDBP promoter and contiguous 5Ј-flanking region, each of the seven internal fragments is shown below. Fragment F is shown in expanded scale on the next line (Ϫ1899 to Ϫ1490) and below that are the four overlapping subsegments that were linked to the rDBP minimal promoter 5Ј to the CAT reporter. B, each construct was transfected into HepG2 cells, and normalized CAT activities were expressed relative to the activity of the rDBP minimal promoter (control, indicated as 100).
indrome is highly conserved, and the 3Ј-half contains substantial sequence variation. Computer-assisted analysis of the entire 2196-bp 5Ј-flanking sequence of the rDBP gene identified five putative HNF1-binding sites. These sites were located in fragments G, F-2, D, B, and A (Table II). Fragments G and D failed to show enhancing activity with either the SV40 promoter or the rDBP promoter (Fig. 2). The coincidence of transcriptional activation activity and the presence of putative HNF1-binding sites in fragments F-2, B, and A was further investigated.
Gel mobility shift assays were performed to determine whether HNF1 could bind to regions F-2, B, and A. When incubated with HepG2 nuclear extract, fragment F-2 formed two predominant retarded complexes on the native gel ( Fig. 4A; the bracketed doublet and single band). Both complexes were sensitive to self-competition by F-2 but were not competed by nonspecific competitor ␤-lac. F-2 was divided into two nonoverlapping halves, F-2a and F-2b (Fig. 4B). F-2b, which contains the putative HNF1-binding site, inhibited formation of both complexes. A 28-mer containing a known functional HNF1binding site present in the ␤-fibrinogen gene promoter (␤28) selectively competed only the more slowly migrating complex identifying this as the HNF1 complex. The identity of the faster migrating band was not determined in this study. F-2a showed partial competition for the slower migrating complex; this effect was variable and not further pursued. NIH3T3 nuclear extract did not form any specific retarded complexes with frag-ment F-2. The DNA-protein doublet formed by fragment B and HepG2 nuclear extracts were specifically inhibited by F-2 and F-2b, suggesting that the same proteins that bound to F-2 were involved in formation of the B fragment complexes. The B fragment complexes were also competed by ␤28 confirming that they involved HNF1. Binding of HNF1 to fragment A was similarly confirmed by efficient competition of the retarded band by F-2, F-2b, and ␤28. These mobility shift studies demonstrated that HNF1 bound in vitro to fragments F-2, B, and A.
To delineate precise protein-binding sites in fragments F-2, B, and A, DNase I footprinting was performed on end-labeled coding strands of each fragment (Fig. 5). HepG2 nuclear extracts protected a 16-bp region (Ϫ1645 to Ϫ1661) in fragment F-2 (left panel) corresponding to the HNF1 site. This was a weak footprint (see below). A somewhat larger 29-bp region (Ϫ186 to Ϫ158) was protected on fragment B extending 10 bp 3Ј from the HNF1 consensus site (middle panel) to include the half-palindrome of an adjacent predicted glucocorticoid response element consensus sequence (see Table II). A 23-bp region (Ϫ65 to Ϫ43) including the HNF1-binding site in fragment A was also specifically protected by HepG2 nuclear extracts (right panel). Specificity controls including NIH3T3 nuclear extracts and BSA did not result in protection (Fig. 5 and  data not shown). The relative positions of HNF1 consensus sites from the 5Ј-end of each footprint were similar in all fragments. The footprint on the F-2 fragment appeared weaker in intensity, smaller in size, and shortened at its 3Ј-end compared with that on fragments B and A.
HNF1 Bound to Fragment B with Higher Affinity Than to Fragments F-2 or A-A modified competition binding study was carried out to extend the observation that the HNF1 footprints on fragments F-2, B, and A differed in relative intensities. Double-stranded 25-mer oligonucleotides containing the HNF1-binding sequences from each fragment (wtF-2, wtB, and wtA, respectively; Table I) were synthesized and used as competitors in a gel mobility shift assay with the 32 P-labeled ␤28 HNF1 oligonucleotide as a probe. The purity and amounts of each of the annealed oligonucleotides used as cold competitors were verified by native polyacrylamide gel electrophoresis after 32 P-end labeling (data not shown). Incubation of ␤28 with HepG2 nuclear extract generated a set of retarded bands consistent with the presence of HNF1 in the extract (Fig. 6, 1st lane) (40). Assignment of the major bands was aided by analysis of complexes formed from nuclear extracts of NIH3T3 cells transfected with HNF1␣ and/or HNF1␤ expression vectors as follows: the HNF1␣ homodimers formed the two most slowly migrating bands, and HNF1␤ homodimers formed the most rapid complex, and HNF1␣␤ heterodimers were intermediate. The predominance of the HNF1␣ bands was consistent with the predominance of this HNF1 isoform in HepG2 cells. Competition for the 32 P-␤28 complexes was carried out with 1-, 4-, and 16-fold molar excess of unlabeled wtF-2, wtB, and wtA oligonucleotides, and signal intensities of the residual bands were quantified. These data revealed that an equimolar excess of wtF-2, wtB, or wtA reduced the labeled HNF1␣ homodimer complex signal by 12, 87, and 29%, respectively. This result ordered the relative affinities of each HNF1-binding site for HNF1␣ homodimer binding as follows: fragment B Ͼ Ͼ fragment A Ն fragment F-2. The lowest affinity of fragment F-2 was consistent with its weaker footprint.
Functional Importance of HNF1 Binding for rDBP Gene Expression-The above results suggested that HNF1 binds to the F-2 and B enhancer fragments as well as to the rDBP gene promoter (fragment A), although with different affinities. To determine the relative functional importance of HNF1 binding to the F-2 or B enhancer elements, each of the two respective HNF1 sites was disrupted individually or in combination, and the effect on expression was determined (Fig. 7). The HNF1binding sites were disrupted by introducing base substitutions into the more conserved 5Ј-half of the HNF1 palindromic sequence, and loss of HNF1 binding was confirmed by gel mobility shift and competition assays (Fig. 7A). The mutF-2 and mutB substitutions were then introduced at their respective positions in the Ϫ2196DBP/CAT construct, and the activities of the resultant genes were tested by transfection into HepG2 cells (Fig. 7B). Mutation of the HNF1-binding site in region F-2 or B reduced rDBP promoter activity to 23 and 7% of wild type Ϫ2196DBP/CAT expression, respectively (Fig. 7B). When both F-2 and B were mutated, the DBP promoter activity was further decreased to 1% of wild type. These data demonstrated that the HNF1-binding sites in the F-2 and B segments mediated positive enhancer function in HepG2 cells, the higher affinity B site having a greater effect on gene expression than the F-2 site. This enhancing activity was most likely mediated a The sequence of the footprinted region is shown in italics and the conserved half of the putative HNF1-binding site is indicated in bold.
b The mutated sequence in each HNF1-binding site is shown in bold. F-2, B, and A. Fragments F-2, B, and A were labeled at their 3Ј-ends and mixed with either BSA, NIH3T3, or HepG2 nuclear extracts at the protein concentrations noted above the respective lanes. The regions protected from DNase I digestion are bracketed, and the sequences within these regions are indicated. In fragment B, the sequences of the predicted glucocorticoid-responsive element (GRE) are also indicated. The consensus sequence for HNF1 is shown below each footprint site (see also by HNF1␣ homodimers because they are the predominant form of HNF1 in the HepG2 cells (see above; Fig. 6). Furthermore, the levels of transcriptional enhancement measured when these two sites were functioning in concert was greater than their additive effects, suggesting synergistic interaction.

FIG. 5. DNase I footprinting confirmed HNF1 binding to fragments
The above data demonstrated a major contribution of the HNF1 sites to transcriptional enhancement of DBP gene expression in tissue culture cells. To determine whether HNF1␣ also has an enhancing effect on DBP expression in vivo, Northern analysis for DBP mRNA levels was carried out on mRNA from the livers of mice with targeted disruption of the endogenous HNF1␣ gene (36). The level of DBP mRNA in the livers of the homozygous null mice (HNF1␣ Ϫ/Ϫ ) was reduced to 50% that of the heterozygotes (HNF1␣ ϩ/Ϫ ) and wild type littermates (data not shown). The magnitude of this decrease, similar to that previously reported for expression of the albumin and FIG. 6. HNF1 bound to fragment B with higher affinity than to fragment F-2 or A. The affinities of HNF1 for fragments F-2, B, and A were compared using an electrophoretic mobility shift-based competition assay. End-labeled probe, ␤28, was incubated with HepG2 nuclear extracts, and competition for complex formation was carried out with unlabeled oligonucleotides containing the corresponding HNF1-binding sites in fragments wtF-2, wtB, and wtA (Table I). Binding reactions included each of the three oligonucleotides at 1-, 4-, or 16-fold molar excess over ␤28 (indicated above respective lanes). Positive controls for the migration of the HNF1␣ homodimer and the HNF1␤ homodimer were generated by using nuclear extracts from NIH3T3 cells transfected with an HNF1␣ or HNF1␤ expression vector. The position of the HNF1␣/HNF1␤ heterodimer was deduced from the appearance of the intermediate band in cells expressing both HNF1␣ and HNF1␤.

FIG. 7. Mutation of the HNF1-binding sequences in fragments F-2 and/or B disrupted transcription from the rDBP promoter.
A, end-labeled fragments F-2 and B were used as a probe in an electrophoretic mobility shift assay with HepG2 nuclear extracts. In each case competition was performed with either the wild type sequence of the HNF1-binding sites in fragments F-2 and B (wtF-2 and wtB) or with HNF1-mutated sequences (mutF-2 and mutB). The sequences of the oligonucleotides are shown below the autoradiograph. The right and left palindrome halves are underlined, and the mutated sequences are shown in bold. B, the diagram shows the Ϫ2196DBP/CAT with the three mutations introduced. These constructs (5 g) were transfected into HepG2 cells. The normalized CAT activities are shown relative to Ϫ2196DBP/CAT (indicated as 100). (36), confirmed the importance of HNF1␣ to DBP gene expression in the liver. These data, along with the mutation studies, documented the functional importance of HNF1␣ in enhancement of DBP gene expression.

HNF1␣ and HNF1␤ Displayed Opposing Effects on Transcriptional Activation of the rDBP Gene-HNF1␣ and HNF1␤
share conserved amino-terminal DNA-binding domains but diverge in their carboxyl-terminal transcription activation domains (21). HNF1␣ is the predominant isoform in hepatocytes and the HepG2 cell line, and it mediates an enhancement of DBP gene transcription. HNF1␤ is the predominant isoform in kidney and several other organs (21). The functional impact of each of these two HNF1 isoforms on DBP gene expression was further explored by expressing each of these proteins individually or together in NIH3T3 cells along with the wild type Ϫ2196DBP/CAT reporter gene. NIH3T3 cells have no endogenous HNF1 activity and were unable to transcribe the Ϫ2196DBP/CAT gene (Fig. 8, lane 1). Gel mobility shift assays demonstrated that nuclear extracts prepared from NIH3T3 cells transfected with either the HNF1␣ and/or HNF1␤ expression vectors formed new DNA-protein complex(es) with the ␤28 HNF1-binding site oligomer (Fig. 6). When co-transfected with an HNF1␣ expression vector, theϪ2196DBP/CAT gene was expressed in the fibroblasts (Fig. 8, lane 2). A positive, although much less intense, response was also detected when the HNF1␤ expression vector was co-transfected (Fig. 8, lane 5). Unexpectedly, co-expression of HNF1␤ with HNF1␣ was not additive in transcriptional enhancement, but instead HNF1␤ decreased HNF1␣ enhancement in a dose-dependent manner (Fig. 8, lanes 2-4). Thus each of the two isoforms of HNF1 had a positive effect on rDBP gene expression independently but differed in the potency of transgene activation. In addition, the two HNF1 isoforms appeared to act antagonistically when co-expressed.

The Negative Effect of HNF1␤ on HNF1␣-induced DBP Gene Transcription Was Mediated by the F-2 and B-binding Sites-
The fragments of the DBP promoter and 5Ј-flanking region that mediated the positive and/or negative effects of the two HNF1 isoforms were determined by co-transfecting HNF1␣ or HNF1␤ expression vectors into HepG2 cells along with the wild type Ϫ2196DBP/CAT construct or with the constructs containing the defined HNF1-binding site mutations. The levels of HNF1␣ in HepG2 cells are lower than in adult liver, and transfection of the HNF1␣ expression vector has been demonstrated previ-ously to result in enhancement of co-transfected albumin or ␣-fetoprotein gene expression (22). With the increased levels of HNF1␣ generated by the expression vector, Ϫ2196DBP/CAT expression increased 2-fold in the HepG2 cells (Fig. 9A, panel  i). HNF1␣ also stimulated the expression of the Ϫ2196DBP/ CAT constructs containing the mutations of the HNF1-binding sequences at the F-2 site, B site, or both (Fig. 9A, panels ii, iii,  and iv). Disruption of the F-2 site (Ϫ2196DBPmF-2/CAT) decreased base-line expression to 20% of the wild type construct, and overexpression of HNF1␣ returned expression to full wild type levels (Fig. 9A, panel ii). In contrast, loss of the B site (Ϫ2196DBPmB/CAT) resulted in a more substantial loss of expression to 8% of the wild type construct and overexpression of HNF1␣ only increased levels to 25% of wild type (Fig. 9A,  panel iii). Combined loss of the A and B site (Ϫ2196DBPmF-2&B/CAT) further depressed base-line expression to 5% of wild type, and the ability of HNF1␣ overexpression to enhance expression was as limited as that seen with the single B site mutation (Fig. 9A, panel iv). These results demonstrated the following: (a) that HNF1␣ was a positive transcription factor for rDBP gene expression, (b) that HNF1-binding sites B and F could each mediate a positive response to HNF1␣, and (c) that site B was the main mediator of overall HNF1␣ enhancing activity. It was further inferred that site A could mediate a positive response by HNF1␣ based upon residual positive activity in the double mutant (Fig. 9A, panel iv), but this was not rigorously proven.
In contrast to the positive enhancing effect of HNF1␣, HNF1␤ mediated a dose-dependent trans-dominant repressor activity in transfected HepG2 cells (Fig. 9B). The level of Ϫ2196DBP/CAT gene activity was decreased by 90% at the highest level of HNF1␤ provided (Fig. 9B, panel i). This contrasted with the known positive HNF1␤ effect on the rat ␣-fetoprotein gene (Fig. 9B, panel i, inset, and see Ref. 15). Cotransfection of HNF1␤ with each of the three HNF1 mutant constructs demonstrated that the dominant-negative effect of HNF1␤ could be mediated by the F-2 and B sites. HNF1␤ markedly decreased the expression of Ϫ2196DBPmF-2/CAT at all levels of co-transfection (Fig. 9B, panel ii). In contrast, expression of the Ϫ2196DBPmB/CAT gene was slightly enhanced at low amounts of HNF1␤, whereas at higher levels HNF1␤ exerted a negative effect (Fig. 9B, panel iii). The lack of a negative effect by HNF1␤ on the expression of Ϫ2196DBPmF-2&B/CAT (Fig. 9B, panel iv) excluded the site A as a negative-effect mediator. Thus HNF1-binding sites F-2 and B mediated both positive and negative effects on rDBP gene transcription, reflecting the relative levels of the two HNF1 isoforms HNF1␣ and HNF1␤. The B site was the major effector of both the enhancer and trans-dominant repressor activity.

DISCUSSION
This study demonstrated multiple roles for HNF1 in transcription of the vitamin D-binding protein gene. Three functional HNF1-binding sites were identified within the proximal 2 kilobase pairs of the rDBP 5Ј-flanking region. These three sites differed in the details of HNF1 binding and in their impact on rDBP expression. Moreover, the sites mediated unique transcriptional responses to alterations in the balance of HNF1␣ and HNF1␤ subunits. The profiles of HNF1 binding and activity for the rDBP gene were distinct from those previously reported for the closely related albumin and ␣-fetoprotein genes. The overall impact of these three binding sites is likely to underlie the distinct tissue and developmental pattern of rDBP gene expression.
Fragment A (HNF1-binding site; Ϫ65 to Ϫ41) appeared to constitute part of the rDBP promoter. It was active only when directly linked to the DBP transcription start site and adjacent basal promoter region (first 40 bases of 5Ј-flanking region). DNA sequence alignments of the promoters of the albumin, ␣-fetoprotein, and DBP genes showed conservation of this proximal HNF1 site in all these genes and in species ranging from human to rat, mouse, and Xenopus laevis (41). In the cases of albumin and ␣-fetoprotein, the proximal HNF1 site was necessary for activity of the respective promoters (42,43), and in both cases the proximal HNF1 site mediated a positive effect on gene expression. From these previous data in closely related genes and the residual activity in response to HNF1␣ by the DBP double mutant, a positive regulatory role for the footprinted site in DBP fragment A could be inferred. These data suggested that the proximal HNF1 site is a highly conserved positive element for hepatic gene expression within the multigene family.
The two 5Ј HNF1-binding sites located in fragments F-2 and B behaved like classical enhancer elements. The F-2 and B fragments both increased the activity of homologous and heterologous promoters in a position-independent manner in transfected HepG2 cells (Figs. 1-3). Individual mutation of the HNF1-binding sites at these two positions severely decreased DBP gene expression in HepG2 cells, and the two sites also appeared to act in a synergistic fashion (Fig. 7). Individually, loss of the B site had a more profound effect on the expression of the DBP promoter in HepG2 cells than loss of the F-2 site (Fig. 7). The data were consistent with a predominant role of site B in enhancing expression; the decrease in expression in HepG2 cells due to loss of the F-2 site could be fully compensated by an increase of HNF1␣ levels acting via the intact B site, whereas loss of the B site could not be compensated in the reciprocal manner (Fig. 9A, panels ii and iii). Thus regions F-2 and B both mediated HNF1-dependent enhancement of DBP promoter activity.
The distinct effects of the two major forms of HNF1, HNF1␣, and HNF1␤ on DBP promoter action were unanticipated. Although both forms enhanced expression of Ϫ2196DBP/CAT in fibroblasts, the action of HNF1␤ was only 5% that of HNF1␣ (Fig. 8). More surprisingly, co-transfection of the minimally active HNF1␤ with the highly active HNF1␣ resulted in a dose-dependent loss of HNF1␣-mediated transcriptional en-hancement. HNF1␤ appeared to exert the same antagonistic effect on the HNF1␣ enhancement in the HepG2 cells, suppressing expression of the Ϫ2196DBP/CAT reporter by over 10-fold (Fig. 9B, panel i). The ability of HNF1␤ to enhance expression from the rDBP promoter in the NIH3T3 fibroblasts but to suppress expression in the HepG2 hepatocytes suggested that the suppressive effect was due to direct interference with preexisting HNF1␣-mediated enhancement. This dominantnegative effect of HNF1␤ on HNF1␣ enhancement was mediated through both F-2 and B sites (Fig. 9B, panels ii and iii). Thus, in the context of the rDBP promoter, the two 5Ј HNF1binding sites, F-2 and B, were each capable of mediating opposing effects of HNF1␣ and HNF1␤ on rDBP gene transcription.
Although fragments F-2, B, and A each bound HNF1, the details of these interactions were different. The footprints at the HNF1 sites were distinct as were the affinities for HNF1 as measured in vitro by the competition assay. The relative affinities of the three sites may parallel their distinct HNF1 footprint patterns as both properties probably reflect differences in the less conserved 3Ј-half of the HNF1-binding site as well as the sequences surrounding the respective sites. The B site, with a 28-bp footprint, bound HNF1 more avidly than the A or F-2 sites that had 22-or 16-bp footprints, respectively. The DNase I footprints generated on fragments F-2 and A by the HepG2 nuclear extracts included only the HNF1 consensus sequences. This does not rule out the possibility of other transcription factor(s) binding to these regions because binding conditions might not be optimal for the association of additional factors or the amount of such factors might not be sufficient in HepG2 cells for detection by DNase I footprinting analysis. For example, HepG2 cells have been shown to contain at least an order of magnitude less C/EBP than adult liver, and a C/EBP site of the hepatitis B virus enhancer is not footprinted by HepG2 nuclear extracts (44). The HNF1 footprint in fragment A had an 11/13 base identity to the ␣-fetoprotein enhancer referred to as the "AT-rich motif" (5Ј-ATTAATAAT-TACA-3Ј). A 306-kDa protein, AT motif-binding factor-1 (ATBF1) has been identified to bind to this motif more efficiently than HNF1 (45,46). By co-transfection studies ATBF1 selectively suppresses the activity of the ␣-fetoprotein enhancer and promoter regions containing functional AT-rich motifs, apparently by com- petition between ATBF1 and HNF1 for the same binding site. The relatively weak enhancing activity of region A might reflect the action of ATBF1 in this region.
The protected region in fragment B was larger than that of fragment F-2 and A, extending 10 nucleotides downstream from the HNF1 consensus site. A sequence similar to a glucocorticoid response element (GRE: 5Ј-GGTACAnnnTGT-TCT-3Ј (47)) was detected 3Ј to the HNF1-binding site of fragment B (5Ј-CTCCCAGACTGTCCT-3Ј; Table II). The extended protection toward this putative GRE and previous reports showing positive regulation of DBP by glucocorticoid (48) 3 suggested that glucocorticoid receptors might also bind to this region, and composite response elements containing GREs have been previously observed (49). The protected region of fragment B also included the sequence 5Ј-CCTCCC-3Ј which is found at the 3Ј boundary of many HNF1 sites and has been shown to be important for enhancing the activity of the human prothrombin gene (16). Either of these elements, if functional, may contribute to the higher affinity of the B fragment for HNF1, perhaps by stabilizing the complex.
In this study, HNF1␤ showed a trans-dominant negative effect on HNF1␣-stimulated activity of the DBP promoter. This was mediated by an apparent interference with the positive effects of HNF1␣ at the same sites. The human vHNF1-C isoform in which the activation domain has been interrupted by alternative splicing was previously identified as a trans-dominant repressor as well (28). In contrast, HNF1␤ has been demonstrated to have a positive effect on some binding sites such as those in albumin and ␣-fetoprotein, or more commonly no functional effect at all, such as in the ␤-fibrinogen promoter (21), the sucrase-isomaltase promoter (50), and the CYP2E1 promoter (51). The HNF1␤ expression vector used in the present study had been cloned from mouse (21), and DNA sequence analysis of this clone and comparison to published sequences confirmed that it was the murine counterpart of the full-length human vHNF1-B, clearly distinguishing it from vHNF1-C (data not shown). Previously, HNF1 has been inferred to mediate a negative feedback loop on its own expression, but this effect was indirect, via down-regulation of HNF4 (52,53). Several additional reports have documented direct trans-dominant repressor activity by HNF4 as well as C/EBP (54 -57). None of these reports appear to relate to the presently described effect of HNF1␤.
Several mechanisms might be considered to explain the observed negative regulation by HNF1␤ on DBP gene expression observed in the current report. It was unlikely that the negative effect of HNF1␤ on the DBP promoter resulted from squelching of the general transcriptional machinery or from direct transcriptional down-regulation of HNF1␣ expression because such mechanisms would be expected to affect the expression from HNF1␣-dependent ␣-fetoprotein promoter similarly, and this was not the case (Fig. 9B, panel i, inset). It is feasible that overexpression of HNF1␤ may alter the transcrip-tional activity of HNF1␣ by sequestration of cofactors such as the 11-kDa dimerization cofactor of HNF1 (DCoH) required for the positive activity of HNF1␣ homodimers via stabilization of dimer binding (58,59). Since DCoH can also bind HNF1␤, HNF1␤ might sequester DCoH and decrease the transcriptional activity of HNF1␣ homodimer by destabilizing it. This mode of inhibition of HNF1␣, however, would also be expected to affect ␣-fetoprotein expression and thus appears unlikely. Direct competition of a nonfunctional or less functional HNF1␤ homodimer or ␣␤ heterodimer with an active HNF1␣ homodimer for DNA binding at the DBP enhancer sites might be the most simple mechanism for this observed gene-specific trans-dominant repression. A differential ability of the competing dimers to assemble at the various promoters (DBP versus ␣-fetoprotein) might explain the disparate effect on these two genes.
The trans-dominant effect of HNF1␤ on the DBP promoter activity was different than its effect on the albumin as well as the ␣-fetoprotein promoter. The enhancement of the ␣-fetoprotein promoter by HNF1␤ is greater than by HNF1␣, whereas HNF1␤ and HNF1␣ have comparable enhancing effects on the albumin promoter (15). However the positive effect of HNF1␤ on the extended Ϫ2196DBP promoter is only 5% that of HNF1␣ in cotransfected NIH3T3 cells (Fig. 8), and there is no enhancement by HNF1␤ on the promoter containing the HNF1 A site alone (Fig. 9B, panel iv). Moreover, HNF1␤ abolished the HNF1␣-mediated enhancement of DBP promoter activity (Fig.  7), but transcriptional activation by HNF1␣ on the albumin promoter is unaffected by HNF1␤ (22). Finally, the same level of HNF1␤ that decreased DBP promoter activity to 10% in transfected HepG2 cells resulted in a 4-fold increase in expression from the ␣-fetoprotein promoter (Fig. 9B, panel i, inset). The basis for the distinct, gene-specific effects of HNF1␤ on these various promoters may reflect several potential mechanisms. It has been shown that the composition of core promoters can differentially regulate the transcriptional response to a given transcription factor (60,61). For example, tumor suppressor p53 protein specifically represses the activity of TATA box-dependent promoters, but it has no effect on promoters directed by a pyrimidine-rich initiator element (62). Various TATA sequences have been reported to vary in their responses to different upstream regulators (63). In this regard, it is interesting to note that the DBP promoter does not contain a classic TATA box, instead it has the sequence TGTA, whereas a classic TATA is present in the promoters of the ␣-fetoprotein and albumin genes (25). This duality of the transcriptional activity of HNF1␤ on various promoters may depend upon protein conformation that can be influenced by the nature of the DNA-binding site and/or may trigger differential interactions with distinct cofactors.
The functional importance of HNF1␣ in DBP expression was supported by the 50% decrease in expression of the endogenous DBP gene in the liver of HNF1␣ Ϫ/Ϫ mice. This decrease was remarkably similar to that observed for the albumin and ␤-fibrinogen genes. As shown by the fibroblast co-transfection ex-3 Y.-H. Song and N. E. Cooke, unpublished observations.

TABLE II
Predicted HNF-1-binding sites in the rat DBP 5Ј-flanking region The putative HNF1-binding sites in region G, F-2, D, B, and A are shown in bold. Sequences that match the HNF1 consensus are shown in capital letters, and the divergent nucleic acids are shown in lowercase letters. The region protected by HepG2 nuclear extracts in DNase I footprinting experiments are underlined. The putative glucocorticoid response element in fragments B is shown in italics. Fragment HNF-1 consensus sequence, GGTTAATNATTAAC(A/C) Ϫ76 CTTGCTTCTGTGCAGAGaTTAATAATTgAtgAATTTCTAGTT Ϫ35 periments (Fig. 8), HNF1␤ directly enhanced DBP promoter function in an environment lacking HNF1␣ although this enhancement was only 5% of the effect observed with an equivalent amount of HNF1␣. Therefore in the HNF1␣-deficient mice, the net effect of HNF1␤ on DBP expression might be positive and sustain a minor level of DBP expression. The doubling of the HNF1␤ mRNA level in the livers of the HNF1␣-deficient mice (36) might further accentuate this positive effect. The fact that the endogenous mouse DBP gene was still expressed in liver at appreciable levels may also reflect control by other, as yet unidentified, transcription factors acting on the DBP gene. The partial loss of endogenous mouse DBP gene expression in HNF1␣ Ϫ/Ϫ mice is thus consistent with the finding in cell culture that HNF1␤ homodimers can support DBP promoter activity in the absence of HNF1␣. A model to explain liver-restricted expression of the DBP gene can be proposed based upon the observation that the DBP gene is oppositely regulated by HNF1␣ and HNF1␤ and on the different tissue distributions of HNF1␣, HNF1␤, and DCoH. Gel mobility shift assays showed that HNF1␣ homodimers are more highly represented than the HNF1␣/HNF1␤ heterodimers in the adult liver, whereas the heterodimers are more abundant than both types of homodimers in the kidney where DBP is expressed at much lower levels (21). Like HNF1, the DCoH gene is also expressed in a tissue-specific way, being most abundant in liver and kidney with less detected in intestine and stomach (58). Although HNF1␣ in the kidney can be stabilized by DCoH, the trans-dominant negative activity of HNF1␤ present at high levels in that organ (21) may overcome the positive effect of low levels of HNF1␣ and suppress DBP gene expression. Therefore, high level expression of DCoH and HNF1␣ compared with HNF1␤ would result in expression in the liver, and the reversal of this balance would result in a relative repression of DBP expression in the kidney. Although this model fits the presently available data, it is no doubt overly simplistic and does not take into account the expected contributions and/or interactions with other transcription factors involved in rDBP gene expression in vivo.
The 5Ј-flanking region of the rDBP gene was studied to determine the basis for its liver-specific expression. As was the case for other members of the DBP multigene family, HNF1 was identified to be a major regulator of DBP gene expression. Importantly, the mechanism of DBP gene regulation was clearly unique; the relative abundance of HNF1␣ and HNF1␤ played a critical role in determining the pattern of DBP gene expression in a genespecific fashion. Based on these findings it is reasonable to predict that this reciprocal regulation by the two subunits of HNF1 may contribute to expression patterns of certain other genes in HNF1-containing cell types and that the mechanism(s) that mediate their opposing effects can now be investigated.