Hepatocyte nuclear factor-4 is responsible for the liver-specific expression of the gene coding for hepatocyte growth factor-like protein.

In an attempt to understand the molecular mechanism regulating the expression of the gene coding for human hepatocyte growth factor-like protein/macrophage stimulating protein (HGFL), our laboratory has isolated and characterized approximately 4200 bp of the 5′-flanking region of the HGFL gene. To determine the location of sites which may be critical for the function of the HGFL gene promoter, we constructed a series of hybrid genes containing serial deletions of this region attached to the coding sequences for chloramphenicol acetyltransferase. Expression of these chimeric plasmids was examined by transient transfection of HepG2 and 293 cells. Our results suggest that the transcriptional activity of the HGFL promoter is modulated in HepG2 cells by one positive element at position −135 to −105 (−135/-105). In contrast, only background levels of chloramphenicol acetyltransferase expression have been detected in 293 cells. The −135/-105 region appears to bind a liver-specific transcription factor essential for expression of this gene. Gel mobility shift experiments with antibodies against hepatocyte nuclear factor-4 (HNF-4) and transactivation of the HGFL promoter by a HNF-4 cDNA expression vector suggest that HNF-4 binds to the −135/-105 region and is responsible for the liver-specific expression of HGFL.

In an attempt to understand the molecular mechanism regulating the expression of the gene coding for human hepatocyte growth factor-like protein/macrophage stimulating protein (HGFL), our laboratory has isolated and characterized approximately 4200 bp of the 5-flanking region of the HGFL gene. To determine the location of sites which may be critical for the function of the HGFL gene promoter, we constructed a series of hybrid genes containing serial deletions of this region attached to the coding sequences for chloramphenicol acetyltransferase. Expression of these chimeric plasmids was examined by transient transfection of HepG2 and 293 cells. Our results suggest that the transcriptional activity of the HGFL promoter is modulated in HepG2 cells by one positive element at position ؊135 to ؊105 (؊135/؊105). In contrast, only background levels of chloramphenicol acetyltransferase expression have been detected in 293 cells. The ؊135/؊105 region appears to bind a liver-specific transcription factor essential for expression of this gene. Gel mobility shift experiments with antibodies against hepatocyte nuclear factor-4 (HNF-4) and transactivation of the HGFL promoter by a HNF-4 cDNA expression vector suggest that HNF-4 binds to the ؊135/؊105 region and is responsible for the liver-specific expression of HGFL.
We previously isolated a human gene which is located at the D3F15S2 locus on human chromosome 3 (3p21), a region believed to code for one or more tumor suppressor genes because this area is deleted in DNA from small cell lung carcinomas, renal carcinomas, and other forms of cancer (1,2). The corresponding cDNA for this gene codes for a protein with a domain structure similar to that of a known growth factor, hepatocyte growth factor (HGF). 1 Both proteins contain four kringle do-mains followed by a serine protease-like domain and are approximately 50% identical. Based on these similarities, the protein encoded by this cDNA was designated as "hepatocyte growth factor-like protein" (HGFL).
HGF is a multifunctional protein that elicits different biological responses in a cell-type and tissue-specific manner (3). HGF can function as either a growth factor or tumor suppressor for a broad spectrum of tissues and cell types including various epithelial cells. HGF has been shown to be identical to scatter factor (SF), a mesenchymal cell-derived cytokine that dissociates cohesive colonies of epithelial cells into individual units (4). The cellular responses of HGF/SF are elicited by c-met, a tyrosine kinase receptor involved in signal transduction (5). Recently, HGF/SF was shown to be required in mice for liver and placental development, since mice lacking this gene die in utero with defects primarily affecting these organs (6,7).
The human HGFL gene is 4960 base pairs (bp) in length (from the codon for the putative initiator methionine to the polyadenylation site), containing 18 exons and 17 intervening sequences (2). The expression pattern of HGFL has been shown by in situ hybridization analysis of mouse tissues to be restricted to hepatocytes in the liver (8). The translated amino acid sequence of the gene and cDNA coding for human HGFL predict a protein of 80,325 dalton molecular mass containing 711 amino acids (2). Western analysis using polyclonal antibodies to HGFL has shown that this protein is secreted and is present in human, mouse, and rat plasma with a molecular mass of approximately 90,000 daltons (9).
Although the precise biological function(s) of HGFL remains to be elucidated, two functions for HGFL have been reported. The HGFL protein has been shown to function as an inflammatory mediator based on the identity of its translated amino acid sequence with the amino acid sequence of macrophagestimulating protein (MSP) (9,10). This function is further enforced by the observation that conditioned medium from COS-7 cells expressing human HGFL mRNA specifically contained a factor which activates macrophages (10). Based on the structural homology to a known growth factor (HGF), HGFL may also exert it effects as a regulator of cell growth. Gaudino et al. (11) have shown that recombinant HGFL induces phosphorylation of RON, a tyrosine kinase receptor homologue to the HGF receptor gene (c-met), in the epithelial cell line T47D. This phosphorylation is followed by a stimulation in DNA synthesis.
Apart from the progress on identifying a putative function for the HGFL protein, no information regarding the regulation of the expression of the HGFL gene is yet available. A wide range of liver-specific gene products have been found to be controlled at the level of transcription (12). This control may be governed by the interaction of cis-acting DNA sequences in the 5Ј-flanking region of many liver-specific genes and their cognate trans-acting factors. In this report, we have used this strategy to investigate the regulation of the expression of the HGFL gene. We have cloned and characterized the 5Ј-flanking region of the HGFL gene. A positive regulatory element has been identified using transient transfection analyses into various cell lines with HGFL promoter-CAT chimeric plasmids. Sequence and transfection analyses suggested that HNF-4 may play a critical role in HGFL regulation. HNF-4 is a liverenriched transcription factor in the steroid/thyroid/retinoic acid superfamily which has been shown to regulate liver-specific expression of a variety of genes. Antibody reactivity and competition gel mobility shift assays demonstrate that HNF-4 binds to the HGFL promoter and transactivation experiments show that HNF-4 is sufficient for HGFL reporter gene expression.

EXPERIMENTAL PROCEDURES
Cloning and Sequencing of the Promoter Region of the Human HGFL Gene-The plasmid pL5Bam6 was previously isolated as described by Han et al. (2). The DNA sequence of the 5Ј-flanking region of the human gene coding for HGFL was determined by both the chemical modification procedure (13) and the quasi-end-labeling modification of the dideoxy chain termination method (14). Sequence was determined on both strands for 79% of the sequence and 97% was determined two times or more for nucleotides Ϫ4154 to Ϫ274 in Fig. 1 (GenBank TM accession no. U37055). The remainder of the sequence (nucleotides Ϫ273 to 146) has previously been determined (2). It was determined that nucleotide 168 in Han et al. (2) was a sequencing error and is not present in the actual sequence (this has been corrected in accession no. M74179 in GenBank TM ); this sequencing error was between nucleotides Ϫ105 and Ϫ106 in Fig. 1. The sequence was analyzed by both the Microgenie (15) and the MacVector (IBI) DNA sequencing programs.
Construction of Plasmids-Chimeric human HGFL promoter-CAT plasmids were generated by cloning PCR amplification products containing various lengths of the 5Ј-flanking sequence from the HGFL gene, into the multiple cloning site of the promoterless CAT vector-pBLCAT6 (16). pL5Bam6 was used as a template for all PCR amplification reactions. The oligonucleotides listed in Table I were used to create the chimeric constructs. Restriction endonuclease extensions which were added on to the primers for cloning purposes are shown in boldface type. The region identical to the HGFL promoter is indicated.
The pL5CAT(Ϫ3049/ϩ1) plasmid which contains nucleotides from Ϫ3049 to ϩ1 of the HGFL promoter, was constructed by PCR amplification using primers a and t (Table I), followed by digestion with BamHI and XhoI and ligation into the corresponding sites of pBLCAT6. Plasmids pL5CAT(Ϫ1554/ϩ1) and pL5CAT(Ϫ1348/ϩ1) were cloned into pBLCAT6 after PCR amplification using primers b and t followed by restriction endonuclease digestion with XbaI and XhoI or BglII and XhoI, respectively. Clones pL5CAT(Ϫ1233/ϩ1), (Ϫ1075/ϩ1), (Ϫ913/ϩ1), (Ϫ793/ϩ1) and (Ϫ554/ϩ1) were all PCR-amplified with the reverse primer, t, and primers c, d, e, f, and g respectively, followed by digestion with XbaI and XhoI and ligation into pBLCAT6. pL5CAT(Ϫ272/ϩ1) was created by digestion of the Ϫ793/ϩ1 PCR product with PstI and XhoI and contains nucleotides Ϫ272 to ϩ1 of the HGFL promoter. Constructs pL5CAT(Ϫ135/ϩ1), (Ϫ104/ϩ1), (Ϫ69/ϩ1) and (Ϫ56/ϩ1) were made by PCR amplification using the primers h, i, j, and k, respectively with primer t and digestion with XbaI and XhoI. For pL5CAT(Ϫ25/ϩ1), primers l and m were annealed and digested with XbaI and XhoI followed by insertion into pBLCAT6. All of the above mentioned constructs have the same end terminus at ϩ1 in the HGFL promoter (see Fig. 2). Plasmids containing PCR products were confirmed by restriction enzyme and DNA sequence analyses.
The HGFL promoter regions for plasmids pL5CAT(Ϫ1554/Ϫ43) and (Ϫ43/Ϫ1554) were amplified using the n and o primers followed by digestion with SstI and XbaI and ligation into pBluescript. The insert was excised by digesting with HindIII and cloned into pBLCAT6 in both orientations. This same fragment, containing nucleotides Ϫ1554 to Ϫ43 was cloned into pBluescript downstream of a BamHI/XbaI 2.5-kb fragment from pL5Bam6 containing nucleotides Ϫ4154 to Ϫ1555. The entire 4.1-kb insert was excised by digestion with HindIII and cloned into pBLCAT6 to create pL5CAT(Ϫ4154/Ϫ43). pL5CAT5(Ϫ135/Ϫ105) was created by annealing oligonucleotides p and q, followed by digestion with HindIII and XbaI and ligation into pBLCAT5 (16). pBLCAT5 contains the herpes simplex virus tk promoter upstream of the CAT reporter gene. pL5CAT5(Ϫ135/Ϫ105mut) was synthesized in an analogous manner with primers r and s. The plasmid pMT2.HNF4 (17) was described elsewhere and was a generous gift from Dr. Francis M. Sladek (University of California, Riverside).
Cell Culture, DNA Transfections, and CAT Assays-HepG2 (human hepatocellular carcinoma) cells and 293 (transformed human primary embryonal kidney) cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 2 mM L-glutamine, and 50 g/ml gentamicin. For transient transfection analysis, the cells were seeded at approximately 40% confluency 24 h prior to transfection. The cells were transfected with the various human HGFL promoter-CAT chimeric plasmids using the DNA calcium phosphate method (18). In each experiment, the cells were co-transfected with equal molar amounts (4.75 ϫ 10 Ϫ12 M) of each HGFL-CAT promoter construct and 5 g of the ␤-galactosidase reference plasmid pRSV␤gal as an internal standard for transfection efficiency. The cells were incubated in the presence of the DNA-calcium phosphate co-precipitate for 24 h. After removal of the precipitate, the cells were incubated for an additional 40 -48 h. HepG2 and 293 cells were harvested and resuspended in 0.25 M Tris⅐HCl. Whole cell extracts were prepared by disrupting the cells with three freeze-thaw cycles. The supernatant was assayed for the amount of CAT protein produced using CAT enzyme-linked immunosorbent assay kits (5 Prime 3 3 Prime, Boulder, CO) according to the manufacturer's instructions. All experiments were repeated at least four times (in duplicate), and CAT protein production was normalized for ␤-galactosidase activity (19).
RNA Preparation and Primer Extension Analysis-Total RNA from cells transiently transfected with HGFL promoter-CAT constructs was isolated using Trizol reagent (Life Technologies, Inc.). A 30-base oligo-  (16) was 5Ј-end-labeled with T4 polynucleotide kinase and [␥-32 P]ATP. The end-labeled primer (500,000 cpm) was hybridized in 2 ϫ hybridization salts (20 mM Tris⅐HCl pH 7.5, 0.6 M NaCl, 4 mM EDTA) to 20 g of total HepG2 RNA transfected with either pL5CAT(Ϫ1348/ ϩ1) or pL5CAT(Ϫ1554/ϩ1) at 44°C overnight after denaturation for 5 min at 90°C. Nucleic acids were purified and redissolved in 20 l of reverse transcriptase mix (20). Complementary DNA was synthesized using avian myeloblastosis virus reverse transcriptase (Life Science Inc., St. Petersburg, FL). The cDNA products were separated on a 6% denaturing polyacrylamide gel along with a sequencing ladder for accurate size determination. The results were visualized by autoradiography after a 16-h exposure at Ϫ80°C with intensifying screens.
Preparation of Nuclear Extracts-For nuclear extract preparation, tissue culture cells were washed twice with ice-cold phosphate buffered saline. The cells were collected and resuspended in 5 volumes of a buffer containing 10 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM dithiothreitol, and 25 g/ml phenylmethylsulfonyl fluoride. Dithiothreitol and phenylmethylsulfonyl fluoride were added just prior to use. The cells were incubated on ice for 10 min, pelleted, and resuspended in 3 volumes of the same buffer. Nonidet P-40 was added to 0.05%, and the cells were homogenized by 10 strokes with a tight fitting type B Dounce homogenizer. The suspension was centrifuged for 10 min at 4°C at 2,000 rpm, and the supernatant was discarded. The crude nuclear pellet was resuspended in 1 volume of a buffer containing 5 mM HEPES, pH 7.9, 26% glycerol, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.5 mM dithiothreitol, and 25 g/ml phenylmethylsulfonyl fluoride. NaCl was added to a final concentration of 300 mM. This mixture was incubated on ice for 45 min, and the nuclear extract was cleared by centrifugation at 15,000 rpm for 20 min. The supernatant was quick frozen in aliquots and stored at Ϫ80°C. The protein concentration (as determined by Bradford assays) ranged from 2 to 6 mg/ml.
Gel Retardation Assay-In order to obtain the Ϫ135/Ϫ105 fragment, complementary oligonucleotides encompassing this region of the HGFL promoter were synthesized, annealed, and end-labeled with T4 polynucleotide kinase and [␥-32 P]ATP. Nuclear extracts (2 to 5 l) were incubated in a binding buffer containing 15% glycerol, 20 mM HEPES pH 7.9, 5 mM MgCl 2 , 50 mM KCl, 0.1 mM EDTA, and poly(dI⅐dC) (0.1 mg/ml final concentration) for 10 min. Subsequently, 2000 -5000 cpm of probe were added (together with the unlabeled competitors in the case of competition experiments), and the binding reaction was continued for another 30 min. For supershift experiments, antiserum was added 10 min after the addition of probe, and the reactions were continued from this point. The reactions were loaded on precooled 0.25 ϫ TBE acrylamide gels containing 0.5 ϫ TBE as running buffer. The gels were run in the cold at 150 V, dried, and exposed to film. For competition experiments, the sequence of the HNF-4 oligonucleotide is 5Ј-GGCAAGGT-TCATATTTGTGTAG-3Ј (21), and for the ␣ 1 -antitrypsin gene the sequence is 5Ј-CCCAGCCAGTGGACTTAGCCCCTGTTTGCT-3Ј (22). The 28-bp HNF-1 primer had the following sequence: 5Ј-CAAACTGT-CAAATATTAACTAAAGGGAG-3Ј and is from the human fibrinogen ␤-chain gene (23). Mutant oligonucleotides encompassing the Ϫ135/ Ϫ105 region are shown in Table II.

Comparison of the 5Ј-Flanking Regions of the Mouse and
Human HGFL Genes-Sequence for 4154 bp of 5Ј-flanking sequence, immediately upstream from the initiator methionine, has been determined for the human gene coding for HGFL protein ( Fig. 1; including 272 bp previously reported by Han et al. (2)). Three regions of repetitive DNA were identified by searching GenBank TM ; including one Alu repetitive sequence ( Fig. 1). Comparison of the 5Ј-flanking region of the human gene coding for HGFL protein with 1285 bp of 5Јflanking sequence of the mouse gene coding for the same protein (1) identified five regions of homology ( Fig. 1). Three of these regions (nucleotides Ϫ1018 to Ϫ836, Ϫ560 to Ϫ342, and Ϫ164 to Ϫ1) are in the same orientation between the two genes, while nucleotides Ϫ1860 to Ϫ1734 and Ϫ1702 to Ϫ1631 are found on the complementary strand in the mouse gene (nucleotides 757-827 and 847-972 in Degen et al. (1)). These five regions are between 67.3 and 78.3% identical between the human and mouse genes.
Analysis of the Human HGFL 5Ј-Flanking Region-Examination of the 5Ј-flanking region of the HGFL gene reveals a number of potential regulatory elements (those found within the first 1 kb of the iMet are shown in Fig. 1). Most prominent, within approximately the first 1 kb upstream of the HGFL gene, are six potential liver factor A1 (LF-A1/HNF-4) nucleotide binding sites. LF-A1/HNF-4 has been shown to be important for the expression of a number of liver specific promoters including the ␣ 1 -antitrypsin (␣ 1 -AT) gene (22). Furthermore, no apparent TATA box was found in this region. The precise role of the multiple potential regulatory sites in the flanking DNA of the HGFL gene remains to be determined. However, the numerous potential regulatory elements in the promoter region of the human HGFL gene suggests that the expression of this gene may be regulated by a variety of cytokines and steroid hormones, analogous to HGF (24).
Deletion Analysis of the HGFL Promoter Region-To elucidate the sequences responsible for the liver-specific transcription of the human HGFL gene, a series of HGFL promoter-CAT plasmids were constructed. Fig. 2A depicts the different regions of the HGFL promoter cloned into the promoterless CAT vector pBLCAT6 (16). Equimolar amounts of each chimeric clone were examined by transient transfection into HepG2 cells (which express HGFL) (8) and 293 cells. No hybridization to HGFL is observed following Northern analyses of RNA isolated from 293 cells. 2 Depicted in Fig. 2B are the average transfection data for each of the HGFL promoter-CAT constructs. In each cell line, the parent plasmid used for chimeric construction, pBLCAT6, gave only minimal activity. As a positive control, the plasmid pD5CAT which contains the SV40 promoter/ enhancer region upstream of the CAT gene was used (25). The most active HGFL promoter-CAT construct had approximately 12% of the amount of CAT protein produced compared to the positive control in HepG2 cells. None of the HGFL promoter-CAT plasmids was active after transfection into 293 cells.
With the exception of pL5CAT(Ϫ4154/Ϫ43) and pL5CAT(Ϫ1554/Ϫ43) which have their 3Ј end point at Ϫ43, all of the other HGFL promoter-CAT plasmids have the same end point at nucleotide ϩ1 (Fig. 1). pL5CAT(Ϫ43/Ϫ1554) has the same HGFL promoter sequence as pL5CAT(Ϫ1554/Ϫ43) but cloned in the reverse orientation. The HGFL promoter region of pL5CAT(Ϫ43/Ϫ1554) was inactive in either cell type. pL5CAT(Ϫ1554/Ϫ43) gave results similar to pL5CAT(Ϫ1554/ 2 R. Muraoka and S. J. F. Degen, unpublished data.  1 The wild type Ϫ135/Ϫ105 HGFL promotor sequence is shown in boxes on the first line. Homologous sequences are shown by ---and variations are indicated. The oligonucleotides were designed based on the sequence reported in Han et al. (2). The G residue reported at Ϫ106 was found to be deleted in this study (Refer to Experimental Procedures for more detail). Oligonucleotides with or without this residue gave similar binding and competition results (data not shown). 2 For the competition assays, ϩϩϩ indicates strong competition, ϩϩ intermediate competition, ϩ slight competition, and Ϫ indicates a lack of competition. Refer to text for more details. ϩ1) which contains an additional 43 bp suggesting that these extra bases were not important for CAT protein production from these constructs.
As shown in Fig. 2B, the region containing nucleotides from Ϫ135 to ϩ1 of the HGFL promoter appeared to contain the minimal amount of flanking DNA required to give strong liverspecific promoter activity. When the sequence between Ϫ135/ Ϫ105 was deleted to create pL5CAT(Ϫ104/ϩ1), CAT protein production decreased to approximately background levels. Another less prominent positive regulatory element may also be present in the region between nucleotides Ϫ1075 and Ϫ914. CAT protein production was reduced by approximately 25% when this region is deleted (Fig. 2B). Further upstream, the nucleotides between Ϫ4154 to Ϫ3050 appear to contain a negative regulatory element since removal of this region relieves transcriptional suppression.
Primer Extension Analysis-In order to determine the transcription initiation site of transfected HGFL-CAT plasmids, primer extension analysis was performed. These chimeric constructs were utilized since the transcriptional start point of the endogenous locus could not precisely be mapped. Fig. 3 shows the primer extension results obtained from HepG2 cells transfected with pL5CAT(Ϫ1348/ϩ1) using an oligonucleotide primer complementary to part of the coding region of the CAT gene. Two bands of approximately equal intensity were obtained of 117 and 118 bp in length (lane 2). These results place the transcription start site at Ϫ75 and Ϫ76 bp upstream of the iMet (Fig. 1). Identical results were obtained with RNA prepared from HepG2 cells transfected with other HGFL-CAT chimeric plasmids. No primer extension products were ob-  (2) with the first nucleotide of this sequence represented by an arrow. Asterisks represent putative transcription start sites. The first exon is identified with the translated amino acid sequence above the DNA sequence. The splice site for the first intervening sequence is between nucleotides 52 and 53 and interrupts the codon between the first and second base. Underlined sequences represent repetitive DNA; nucleotides Ϫ3613 to Ϫ3321 code for an Alu repetitive element. Sequence represented by boldface type is similar to regions in the 5Ј-flanking region of the mouse gene coding for HGFL. Nucleotides Ϫ1860 to Ϫ1734 and Ϫ1702 to Ϫ1631 are both 70.8% identical to nucleotides 847-972 and 757-827, respectively, on the complementary strand in the mouse gene (1). Nucleotides Ϫ1018 to Ϫ836, Ϫ560 to Ϫ342 and Ϫ164 to Ϫ1 are 78.3, 67.3, and 77.2% (respectively) to nucleotides 12-189, 465-683, and 1118 -1277 in the mouse gene, respectively. Putative transcription factor binding sites are boxed and represented within the first kilobase of sequence upstream of iMet: ERE-half, estrogen response element half-site (identical to type 2 nuclear receptor half-sites); IRE, ␥-interferon response element; RXR, 9-cis-retinoic acid response element; T3, thyroid hormone response element; GMCSF, granulocyte/macrophage colony-stimulating factor; LF-A1, HNF-4/LF-A1 binding site; C/EBP, CCAAT-enhancer binding protein; CRE, cAMP response element; alpha INF, ␣-interferon binding site; NF-Il 6 , a response element for a nuclear factor that stimulates interleukin-6; and GRE, glucocorticoid receptor element.
tained from RNA isolated from untransfected HepG2 cells (data not shown) or when tRNA was used as a negative control with the same primer (lane 1).
Localization of Nuclear Factor Binding Sites by Gel Retardation Analysis-To investigate the possible interaction of trans-acting factors with the Ϫ135/Ϫ105 region of the HGFL gene, oligonucleotides were synthesized spanning both strands of the Ϫ135/Ϫ105 region and gel electrophoretic mobility shift assays were performed with crude nuclear extracts from HepG2 and 293 cells (Fig. 4). A specific protein/DNA complex is formed with this region in HepG2 cells extracts (lanes 2 and 8) but is not present in 293 cells (lane 9). This complex can be specifically competed by a nonradiolabeled self-competitor (lanes 3-5) and not with a nonrelated fragment homologous to an HNF-1 binding site (lanes 6 and 7). These results suggest that the region encompassing nucleotides Ϫ135/Ϫ105 is involved in the liver-specific expression of HGFL.
Since the region between nucleotides Ϫ135/Ϫ105 appeared to be crucial for the expression of the HGFL promoter-CAT constructs in HepG2 cells, the nucleotide sequence in this region was examined for known protein binding motifs. The most striking putative transcription factor binding site in this area was for LF-A1 (Fig. 1) (22). LF-A1 is thought to be identical to HNF-4 based on DNA binding data and on antiserum reactivity (26).
To test indirectly if the protein binding to the Ϫ135/Ϫ105 was HNF-4, oligonucleotides were synthesized complementary to a reported HNF-4 consensus sequence (21) and to a region of the ␣ 1 -AT promoter containing nucleotides Ϫ130 to Ϫ101  Fig. 1. B, the pBLCAT6 constructs containing variable amounts of the 5Јflanking region of the human HGFL gene (as shown in A) were transfected into HepG2 (striped bars) and 293 (solid bars) cells. After transfection, the amount of CAT protein produced was quantified by an ELISA and normalized for ␤-galactosidase activity. In each assay, the relative activity of pL5CAT(Ϫ1554/Ϫ43) was set at 100.

FIG. 3. Primer extension analysis.
Primer extension was performed using RNA isolated from HepG2 cells transfected with pL5CAT(Ϫ1348/ϩ1) and a radiolabeled oligonucleotide complementary to nucleotides 59 -88 of the CAT gene of pBLCAT6 (lane 2). Lane 1 is a primer extension reaction with the same primer as in lane 2 with tRNA as a template. The cDNA products were resolved on a 6% denaturing polyacrylamide gel along with a sequencing ladder (GATC) for exact size determination of resulting cDNA products. The sizes of the primer extension products are indicated in base pairs. which has been shown to bind HNF-4 (22). These oligonucleotides were used in competition gel mobility shift experiments shown in Fig. 5A. The ␣ 1 -AT competitor almost completely competed for the wild type Ϫ135/Ϫ105 region at 50-and 100fold molar excess (lanes 4 and 5). The HNF-4 consensus competitor competed to a lesser extent (lanes 6 -8).
To examine the DNA sequence requirements for protein binding to the Ϫ135/Ϫ105 region, mutant oligonucleotides were synthesized and tested for their ability to compete for trans-acting factor binding in gel mobility shift assays. Table II shows the sequence of the mutant oligonucleotides used in the assay, and Fig. 5B shows the results of the competition experiments. At 100-fold molar excess, mutants 4, 5, and 6 (lanes 6 -8) were unable to compete for protein binding to the Ϫ135/ Ϫ105 region of the HGFL promoter. Mutant 2 almost completely competed for specific protein binding (lane 4) and mutants 1, 3, 7, 8, 9, and 10 competed to a lesser extent (lanes 3, 5, 9, 10, 13, and 14, respectively). Therefore it appears that the most important sequence involved in protein binding to the Ϫ135/Ϫ105 region is TCAGGTCAG at nucleotides Ϫ126 to Ϫ118 (Table II; Fig. 1). However, the remaining nucleotides in this region also appear to be involved in protein binding but to a lesser extent.
HNF-4 Binds to the Ϫ135/Ϫ105 Region of the HGFL Promoter-To directly demonstrate that HNF-4 binds to the Ϫ135/ Ϫ105 region of the HGFL promoter, gel mobility shift experiments were performed with the Ϫ135/Ϫ105 region and antiserum directed against a synthetic peptide derived from the carboxyl terminus of the HNF-4 protein (26). The antiserum retarded the mobility of the protein-DNA complex formed with the Ϫ135/Ϫ105 probe and HepG2 cell extracts (Fig. 6A, lane 4). Antiserum against HGFL was used as a negative control and did not react with the protein-DNA complex (lane 3). Preimmune serum also had no effect (data not shown).
As further evidence for HNF-4 binding to the Ϫ135/Ϫ105 region of the HGFL promoter, 293 cells were transfected with an expression vector for HNF-4, pMT2.HNF-4 (17). Nuclear extracts were prepared from the transfected 293 cells and tested for binding to the Ϫ135/Ϫ105 region (Fig. 6B). Extracts prepared from untransfected 293 cells were unable to retard the mobility of the Ϫ135/Ϫ105 region (lane 2), whereas extracts from 293 cells transfected with pMT2.HNF-4 (lane 3) produced a mobility shift identical to that of extracts from HepG2 cells (lane 1).
Transactivation by HNF-4 -In order to test the Ϫ135/Ϫ105 region for heterologous promoter activity and transactivation by HNF-4, oligonucleotides spanning this region were synthesized and cloned into the multiple cloning site of pBLCAT5 (16) creating pL5CAT5(Ϫ135/Ϫ105). pBLCAT5 contains the herpes simplex virus tk promoter 5Ј to the CAT gene. As depicted in Fig. 7, this region was able to stimulate transcription from the heterologous promoter approximately 16-fold over the parental plasmid in HepG2 cells. Additionally, a heterologous promoter construct, pL5CAT5(Ϫ135/Ϫ105mut), containing three base pair changes compared to the wild type Ϫ135/Ϫ105 HGFL sequence (GGT to TTG at nucleotides Ϫ123 to Ϫ121; mutant 5 in Table II)  mutant Ϫ135/Ϫ105 sequence showed that protein-DNA complexes were formed from both HepG2 and 293 cell extracts (data not shown), suggesting that the mutant sequence may bind another protein(s).
An expression vector for HNF-4, pMT2.HNF-4 (17), was also cotransfected with either the wild type or mutant heterologous promoter constructs (Fig. 7). In HepG2 cells, increased HNF-4 expression was only moderately able to increase transcription from the wild type Ϫ135/Ϫ105 promoter. In 293 cells, the expression of HNF-4 was able to stimulate transcription from pL5CAT5(Ϫ135/Ϫ105) over 10-fold compared to pL5CAT(Ϫ135/Ϫ105) alone, whereas the mutant was only slightly stimulated. The mutant Ϫ135/Ϫ105 sequence in either cell type was not stimulated to an appreciable extent with the HNF-4 expression vector compared to the wild type sequence. DISCUSSION The HGFL gene was identified based on its similarity to a probe coding for the kringle domains of bovine prothrombin (2) and named for its structural similarity to HGF. Based on the translated cDNA sequence for this gene, it has become apparent that HGFL protein is identical to MSP (9,27). HGFL/MSP has been shown to be a chemoattractant for mouse resident peritoneal macrophages (28), to induce morphological changes in macrophages (10), and to inhibit induction of nitric oxide production by lipopolysaccharide-or cytokine-induced macrophages (29). Furthermore, the expression of HGFL has been shown to be up-regulated during liver regeneration and inflammation (30). This investigation has focused on understanding the factors involved in the regulation of the HGFL gene.
Transient transfection analyses with sequential deletions of the HGFL flanking DNA fused 5Ј to the CAT gene were performed to identify potential regulatory sequences governing the expression of this gene. Since the endogenous transcriptional start point (tsp) of this gene could not be precisely determined due to the duplicated copies of this locus 3 (32), the 3Ј end point of the majority of deletion constructs was set at the A nucleotide of the potential iMet (ϩ1 in Fig. 1) with a few exceptions. The plasmids pL5CAT(Ϫ4154/Ϫ43) and pL5CAT(Ϫ1554/Ϫ43) have their 3Ј end point at the A nucleotide (nucleotide Ϫ43 in Fig. 1) of an inframe ATG codon upstream of the ATG at ϩ1. An upstream ATG codon is also present in the mouse; however, this codon is not in frame. Transient transfection analyses of pL5CAT(Ϫ1554/ϩ1) and pL5CAT(Ϫ1554/Ϫ43) gave indistinguishable results (Fig. 2B), suggesting that these first 43 bp were not necessary for the transcription of CAT. From these results, it may be inferred that the tsp is not within the first 43 bp, however, an alternative start site can not be excluded.
In order to determine the tsp of the HGFL gene, primer extension analyses were performed using RNA isolated from HepG2 cells transfected with the chimeric HGFL promoter-CAT construct pL5CAT(Ϫ1348/ϩ1). This allowed us to assay for RNA initiated specifically from the HGFL promoter and not from RNA initiated from highly homologous regions of the amplified chromosome 1 loci (32). The sequence of one copy of a homologous gene has been determined 3 and was found to be 97% identical to the gene coding for HGFL, including 2200 bp of the 5Ј-flanking sequence. All but the 3Ј end of the HGFL gene is homologous to sequences on chromosome 1 (2). Primer extension experiments performed on total RNA from HepG2 cells using endogenous HGFL sequence primers resulted in a multitude of putative start sites 3 most likely originating from the combination of HGFL and related chromosome 1 sequence start points. The two bands obtained in Fig. 3 correspond to RNA transcripts potentially initiated from Ϫ75 and Ϫ76 bp upstream of the iMet of the HGFL promoter (Fig. 1). This same combination of bands was observed using several HGFL promoter-CAT constructs. The tsp identified in the human HGFL flanking DNA is within 20 bp of the unique tsp identified in the (nonamplified) mouse gene (1) when the two sequences are aligned at the iMet residue.
Transient transfection analyses implicated several regions which may be critical for expression of HGFL. One region, between nucleotides Ϫ135 to ϩ1, contained the minimal sequence of the human HGFL flanking DNA that could drive cell type-specific expression (Fig. 2B). Given the results of the primer extension experiments, this suggests that the minimal promoter required may only span 60 bp from nucleotides Ϫ135 to Ϫ75. Furthermore, there is no apparent TATA box or Sp1 site near the tsp (Fig. 1). TATA-less promoters are found in many housekeeping genes (e.g. genes encoding the enzymes of intermediary metabolism). Genes with TATA boxes in the promoter region generally initiate transcription at well defined sites, whereas transcription of TATA-less promoters generally occurs over an extended region. The primer extension results presented in Fig. 3 suggest a well defined tsp contrary to the ambiguous tsp of many promoters devoid of TATA boxes.
When comparing the mouse and human HGFL sequences, several areas of homology were observed; one of which overlaps with the Ϫ135 to ϩ1 region. Nucleotides Ϫ164 to ϩ1 have a high degree of similarity, approximately 76%, between the two species. Furthermore, the Ϫ135/Ϫ105 region is 71% identical between the mouse and human. Similarities in DNA sequence among different species are thought to represent important regulatory regions.
The Ϫ135/Ϫ105 region appeared to be the most critical region implicated in our transient transfection studies since it is both required for promoter activity and is able to confer this activity in a cell type-specific manner. Among the multiple putative binding motifs in the Ϫ135/Ϫ105 region, sequence analyses identified a putative recognition site for LF-A1/HNF-4 in the Ϫ135/Ϫ105 region. HNF-4 is a potent transcriptional activator that controls the expression of a variety of genes including the liver-specific expression of the transthyretin (33), ␣ 1 -AT (22), and apolipoprotein CIII (26) genes. Due to the presence of this binding sequence, oligonucleotides were made complementary to a reported HNF-4 consensus sequence (21) and to the 5Ј region of the ␣ 1 -AT promoter where HNF-4 has been reported to bind (22). Competition experiments were performed with both of these regions against the Ϫ135/105 promoter region of HGFL. The results depicted in Fig. 5A show that both competitors were capable of binding the same protein as the Ϫ135/Ϫ105 region indicating that HNF-4 was binding the HGFL promoter region. Recently, however, a more refined consensus sequence for HNF-4 binding sequence has been reported by Jiang et al. (34). This sequence is a direct repeat of AGGTCA separated by one nucleotide, referred to as a DRϩ1 element. Close inspection of the Ϫ135/Ϫ105 region shows that a very near match to this consensus sequence is found at nucleotides Ϫ131 to Ϫ121 containing the sequence 5Ј-AG-GTCTCAGGTCA-3Ј. This may explain why the consensus sequence (5Ј-GGCAAGGTTCATATTTGTGTAG-3Ј) chosen in our competition experiments for HNF-4 may not compete to a larger extent.
In an attempt to define the nucleotides involved in the binding of protein to the Ϫ135/Ϫ105 region, several mutant oligonucleotides were obtained and used in the competition gel mobility shift experiments shown in Fig. 5B. Although most the nucleotides appeared to be important for binding, the sequence encompassing nucleotides Ϫ126 to Ϫ118 containing the sequence TCAGGTCAG was found to be most crucial. The most important nucleotide sequence involved in potential HNF-4 binding contains one copy of the AGGTCA motif.
In order to confirm that HNF-4 was responsible for binding and conferring liver-specific expression to the HGFL promoter, a HNF-4 cDNA expression vector was transfected into 293 cells. Extracts from these cells were now capable of binding to the Ϫ135/Ϫ105 region as shown in Fig. 6B. Furthermore, antibodies directed specifically against HNF-4 were able to completely supershift the protein-DNA complex formed in endogenous HepG2 cell extracts (Fig. 6A), suggesting that all the protein complexes contain at least one HNF-4 monomer. These results indicate that HNF-4 does in fact bind to this region. HNF-4 thus far has been reported to bind exclusively as a homodimer (26,34). The results of our supershift experiments, in which all of the protein-DNA complex is recognized, are consistent with this observation, although heterodimer formation with another protein(s) cannot be excluded. Interestingly, HNF-4 is not exclusively found in the liver, but its expression has also been reported in intestine and kidney tissue (21,35). Preliminary northern analyses of HepG2 and 293 RNA indicate that HNF-4 is not present in the kidney derived 293 cells used in our studies. 2 Further experiments were undertaken to demonstrate that not only could HNF-4 bind to the Ϫ135/Ϫ105 region, but that binding of this protein resulted in transcriptional activation. Cotransfection experiments with the vector pBLCAT5 and pMT2.HNF-4 gave results identical to those of pBLCAT5 alone in both cell types. Transactivation experiments shown in Fig. 7 demonstrated that the Ϫ135/Ϫ105 region was sufficient to promote activation of a heterologous promoter construct. More importantly, HNF-4 expression was solely able to activate transcription of CAT through the Ϫ135/Ϫ105 promoter region in 293 cells. These results demonstrate that HNF-4 is necessary and sufficient to both bind to (Fig. 6) and activate transcription from the Ϫ135/Ϫ105 region of the HGFL promoter.
A mutant Ϫ135/Ϫ105 sequence which changes the important HNF-4 AGGTCA motif to ATTGCA, decreased transcription of CAT in HepG2 cells compared to the wild type sequence. This mutant sequence was shown not to compete, even at 100-fold molar excess, with the wild type sequence in gel mobility shift assays (mutant 7, Table II and Fig. 5B). In contrast to the wild type Ϫ135/Ϫ105 heterologous promoter construct which did not stimulate transcription in 293 cells, the mutated sequence was able to confer some transcriptional activity in 293 cells. The activity in both cell types, however, appeared not to be stimulated to a significant extent by HNF-4. Gel mobility shift experiments performed with this mutant sequence showed that protein was capable of binding to the mutated sequence in both cell extracts (data not shown) and may account for the activation seen. The protein-DNA complex formed with the wild type FIG. 7. Heterologous promoter activity of the ؊135/؊105 region. The Ϫ135/Ϫ105 region of the HGFL promoter was inserted upstream of the herpes simplex virus tk promoter in pBLCAT5 to create pL5CAT5(Ϫ135/Ϫ105). A mutant Ϫ135/Ϫ105 region which changes 3 bp (GGT to TTG at nucleotides Ϫ123 to Ϫ121; Table II, mutant 7) was also cloned 5Ј to the tk promoter in pBLCAT5 to create pL5CAT5(Ϫ135/Ϫ105mut). The amount of CAT protein produced was determined after transfection into HepG2 (stripped bars) and 293 (solid bars) cells and normalized for ␤-galactosidase activity. In each cell type, the activity of pL5CAT5 was set at 1. The fold stimulation of CAT protein production of the Ϫ135/Ϫ105 and Ϫ135/Ϫ105mut clones over pBLCAT5 after transfection into is shown. The experiments were performed with and without co-transfections of an HNF-4 cDNA expression vector, pMT2.HNF-4. sequence had a different electrophoretic mobility (larger) then the mutant complex. Furthermore, the activity was approximately equivalent (5.7-versus 4.2-fold) in both HepG2 and 293 cells. This was not entirely unexpected since weak binding to direct repeats (DRϩ1 elements) or near matches to these repeats (as is apparent in this case) has been reported for family members of the type 2 nuclear receptor family, specifically 9-cis-retinoic acid response element homo-and heterodimers (31,34,36). Initial attempts investigating the protein/DNA complexes formed on the Ϫ135/ϩ1 promoter region using DNa-seI footprinting analyses with extracts from 293 cells with and without overexpression of HNF-4 have been unsuccessful thus far. However, it appears as though the majority of the Ϫ135/ϩ1 region of the promoter may be occupied in various cell types. Efforts are underway to purify HNF-4 and to more precisely map the important contact residues for HNF-4 binding.
After examination of the DNA sequence of the 5Ј-flanking region of the human HGFL gene, various putative regulatory elements found in inducible genes were identified (Fig. 1). Of these elements, there are several potential liver-specific C/EBP transcription factors and multiple potential HNF-4 binding sites. Both of these proteins have been found to regulate a number of liver-specific promoters. There are a number of potential regulators in the Ϫ135 to Ϫ105 region. There are various hormone responsive elements including: a putative estrogen response element half-site, a site for regulation by retinoic acid and/or derivatives of retinoic acid through the 9-cis-retinoic acid response element, a site for the thyroid hormone binding and a ␥-interferon response element. Since all of these putative binding sites occur within or overlap the Ϫ135/ Ϫ105 region, it is tempting to speculate that the levels of HGFL may be regulated by the competition or availability of these factors with their recognition sequences.
Based on our results, it can be concluded that HNF-4 is necessary and sufficient for the liver-specific expression of HGFL in HepG2 cells. Antibody reactivity and transactivation experiments conclude that HNF-4 binds to the Ϫ135/Ϫ105 region and is the sole factor required for stimulating transcription from the wild type Ϫ135/Ϫ105 region in 293 cells. Furthermore, mutations of this sequence result in the lose of HNF-4 binding and tissue specificity. These studies represent an initial effort to unravel the mechanisms governing expression of HGFL and provides a basis for further study of transcriptional regulation of this gene. Current studies are under way to determine the inducibility of this gene, to more precisely map protein/DNA contact sites and to further map upstream regulators.