Report

Modulation of Hepatic Gene Expression by Hepatocyte Nuclear Factor 1

See allHide authors and affiliations

Science  04 Jul 1997:
Vol. 277, Issue 5322, pp. 109-112
DOI: 10.1126/science.277.5322.109

Abstract

Hepatocyte nuclear factors 1 and 4 (HNF-1 and HNF-4) are liver-enriched transcription factors that function in the regulation of several liver-specific genes. HNF-1 activates genes containing promoters with HNF-1 binding sites. However, this factor negatively regulates its own expression and that of other HNF-4–dependent genes that lack HNF-1 binding sites in their promoter region. This repression is exerted by a direct interaction of HNF-1 with AF2, the main activation domain of HNF-4. The dual functions of gene activation and repression suggest that HNF-1 is a global regulator of the transcriptional network involved in the maintenance of hepatocyte-specific phenotype.

Liver-specific gene expression is governed by the combinatorial action of a small set of liver-enriched transcription factors, including HNF-1, C/EBP, HNF-3, and HNF-4 (1). The expression patterns of HNF-1 and HNF-4 closely correlate with the differentiation state of hepatic cells. HNF-4 is an activator of the HNF-1 gene, defining a transcriptional hierarchy involved in both the determination and maintenance of hepatic phenotype (2). In transient transfection experiments, HNF-1 negatively regulates its own and other HNF-4–dependent promoters that are not directly recognized by HNF-1 (3). These findings suggested the functioning of an indirect negative autoregulatory mechanism that is triggered by increased intracellular concentrations of HNF-1. HNF-1 did not affect several other promoters, and fusion proteins containing different NH2- and COOH-terminal parts of the HNF-1 molecule failed to inhibit HNF-4–mediated transcription (3,4). These findings argue against a squelching effect.

To investigate the potential role of HNF-1 on the transcription of its own gene in the in vivo chromosomal context, we generated stable HepG2 cell lines (H1A and H1B) expressing different amounts of HNF-1 protein. The expression of endogenous HNF-1 and various target genes was analyzed by Northern (RNA) blot analysis. A glycerol aldehyde phosphate dehydrogenase (GAPDH) probe that produced constant amounts of mRNA was used as a control. Hybridization with a probe encompassing the coding region of the rat HNF-1 cDNA (rHNF-1CR) produced signals of 3.6 and 3.0 kb that corresponded to endogenous and transgene-derived HNF-1, respectively (Fig. 1). The amount of endogenous HNF-1 transcript was reduced in both H1A and H1B cell lines. This decrease was also observed by hybridization with the use of a 3′ untranslated (3′ UTR) fragment of the human HNF-1 as a probe that detected only endogenous HNF-1 mRNA (Fig. 1). Expression of apolipoprotein C-III (apoC-III) is dependent on HNF-4 (5). The amount of apoC-III mRNA was decreased in the HNF-1–overproducing cell lines. However, the hybridization signal for mRNA transcribed from the HNF-1–dependent albumin gene (6) was increased. The amount of HNF-4 transcript remained constant, which implied a lack of positive reciprocal activation of HNF-1 and HNF-4 in the chromosomal context. Consistent with this notion is the observation that some HNF-4 promoter constructs that contain the putative HNF-1 binding site do not drive liver-specific expression in transgenic mice (7). Moreover, wild-type and null mutant mice that are devoid of HNF-1 express HNF-4 in similar amounts (8).

Figure 1

Ectopic expression of HNF-1 represses endogenous HNF-1 and apoC-III transcription. HepG2 cells were transfected with pCB-HNF-1 expression vector (3), and stably expressing cell lines were selected and expanded in a medium containing G418 (150 μg/ml; Geneticin, BRL). Polyadenylate RNAs from cell lines expressing different amounts of HNF-1 (H1A and H1B) were prepared and compared with wild-type HepG2 mRNA by Northern blot hybridization using the following probes: rHNF-1CR (containing the entire coding region of rat HNF-1 cDNA), hHNF-1 3′ UTR (containing nucleotides 2305 to 2783 of the 3′ untranslated region of human HNF-1 cDNA), and cDNAs coding for human apoC-III, mouse albumin, human HNF-4 (hHNF-4), and GAPDH as control. Hybridization and washing conditions were as described in (13). Positions of radioactive signals are shown at the right. With HNF-4 a second signal above the 4.5-kb band can also be seen, which corresponds to cross-hybridization with contaminating 28S ribosomal RNA.

In the HNF-1–overproducing cell lines, the activities of the albumin, apoC-III, and HNF-1 promoters were affected in the same way as the amount of their steady-state mRNA, indicating that the observed changes were the result of altered transcription rates (Fig.2). In addition, the activity of the chimeric promoter construct 4×A TK-CAT, which contains four copies of the HNF-4 binding site of the HNF-1 promoter, was also reduced. On the other hand, the activity of the control promoter (RSV-CAT) was not changed (Fig. 2). This suggests that HNF-1 exerts its negative effect by counteracting HNF-4 activation on the corresponding regulatory regions.

Figure 2

Negative regulation of HNF-4–dependent promoters by HNF-1. Wild-type (WT) and HNF-1–overexpressing HepG2 cell lines (B and A) were transfected by the calcium phosphate precipitation method (13) with 2 μg of the indicated reporters containing the mouse albumin (Alb-CAT) (3), human apolipoprotein C-III (ApoC-III-CAT) (5, 13), rat HNF-1 (HNF-1-CAT) (3), or Rous sarcoma virus (RSV-CAT) (3) promoters, or a chimeric reporter construct containing four copies of the HNF-4 binding site of the HNF-1 promoter fused to the minimal promoter region (nucleotides −85 to +51) of the herpes simplex virus thymidine kinase gene (4×A TK-CAT) (3). The bars represent means ± SE of normalized CAT (chloramphenicol acetyltransferase) activities from at least four independent experiments, and these values are expressed as relative activation (Alb-CAT) or as a percentage of the activity measured in wild-type cells.

To understand the molecular mechanism responsible for the above observations, we performed electrophoretic mobility shift assays to compare the amounts of active DNA binding protein in the stable cell lines. As a control, Sp1 binding activity was monitored and found to be similar in all extracts (Fig. 3A). Alb-PE (3, 6) and site A (3) were used as probes for HNF-1 and HNF-4, respectively. In the H1A and H1B cell lines, DNA binding to the Alb-PE probe was 11 and 4 times that of the wild type, respectively (Fig. 3A). This is much lower than the observed increase in total amounts of HNF-1 mRNA. The difference might result from additional translational control mechanisms or limiting intracellular concentrations of DCoH (dimerization cofactor for HNF-1), which is required for HNF-1 dimerization and stability (9). No difference in DNA binding activity on the HNF-4 probe was observed with the different HepG2 cell lines (Fig. 3B). An antibody raised against HNF-4 almost quantitatively supershifted the DNA-protein complex formed on the site A probe, whereas an HNF-1 antibody failed to supershift the complex (Fig. 3B). Moreover, no difference in HNF-4 binding affinity to site A was detected with the use of extracts from the HepG2, H1A, and H1B cell lines or extracts from HNF-4– and HNF-1–transfected COS-1 cells (10). Thus, neither HNF-1 nor another factor that may have been induced by HNF-1 interacts directly with site A, and HNF-1 does not affect the DNA binding activity of HNF-4.

Figure 3

Site A of the HNF-1 promoter binds HNF-4 but not HNF-1. Nuclear extracts from wild-type HepG2 (H) and the HNF-1–overexpressing (H1A and H1B) cell lines were prepared and analyzed in electrophoretic mobility shift assays using (A) Alb-PE and Sp1 or (B) site A oligonucleotide probes, as described (13). In some assays, 1 μl of antibody to HNF-4 (4) or HNF-1 (1) at 1:6 dilution were also included. The identity of HNF-1 that bound to the Alb-PE probe was verified by supershifts with HNF-1 antibody (10).

Although mobility shift experiments did not reveal interactions between HNF-1 and HNF-4, weak protein-protein interactions may exist that are unable to survive electrophoretic conditions but could explain the down-regulation of HNF-4–dependent genes by HNF-1. To test this idea, we mapped the HNF-4 protein domains necessary for HNF-1–mediated down-regulation. Fusion proteins containing the Gal4 DBD (the DNA binding domain of yeast Gal4 protein) and parts of HNF-4 bind to the Gal4 response element as dimers through the Gal4 DBD (11), and their expression is not affected by HNF-1 (10). The Gal4 HNF-4(E) construct that contains the complete ligand binding–dimerization domain (12) of HNF-4 was a potent activator of the Gal4-responsive reporter in both HepG2 and COS-1 cells (Fig. 4). This activation was strongly inhibited by increased intracellular amounts of HNF-1 derived from its ectopic expression in the stably transfected HepG2 cell lines (H1B and H1A) or from cotransfected expression vector (COS-1 cells). Similar results were obtained with cotransfected HNF-1(440), which lacks the COOH-terminal activation domains, whereas HNF-1(280), which contains the dimerization and DNA binding domains, failed to exhibit repressor activity (Fig. 4). Partial deletion of the main activation domain of HNF-4 [Gal4 HNF-4(E354)], which is located between amino acids 337 and 368, resulted in loss of activity. No significant change was observed in experiments with Gal4 VP16, which was used as an unrelated control (Fig. 4).

Figure 4

HNF-1 represses HNF-4 activity through interaction with the HNF- 4 E domain in vivo. COS-1 and HepG2 lines (WT, H1B, H1A) were transfected with 2 μg of G4-CAT reporter containing four copies of the 17-nucleotide oligomer Gal4 binding site, together with 0.5 μg of the indicated Gal4 expression plasmids. In COS-1 cells, 0.5 μg of pCB-HNF-1 (FL), pCB-HNF-1(280) (M280), or pCB-HNF-1(440) (M440) was also included where indicated. The numbers represent mean values of β-galactosidase–normalized CAT activities from at least six independent experiments with SEs of <8% and are expressed as activation relative to the activity obtained with the Gal4 DBD. Maximal activity in both HepG2 and COS-1 cells was obtained with a fusion construct containing the entire E domain [Gal4 HNF-4(E)]. No activity was observed when other combinations of HNF-4 domains (such as amino acids 337 to 368, 337 to 455, 368 to 455, and 227 to 455) were tested, suggesting that the HNF-4 AF2 domain is active only in the context of an intact E domain (14).

These results indicated that HNF-1 may repress gene expression through physical interaction with HNF-4. In vitro evidence for such protein-protein interaction was provided by pull-down assays with glutathione-S-transferase (GST)–HNF-4 fusion proteins and in vitro synthesized 35S-labeled HNF-1. HNF-1 associated with TFIIB (Fig. 5A); this interaction may be important for HNF-1–facilitated formation of preinitiation complexes. Comparable amounts of bound HNF-1 protein were recovered by GST-HNF-4(130–368), containing the entire E domain, and by GST-HNF-4(AF2), containing the main activation region (amino acids 337 to 368) of HNF-4. In contrast, no interaction was observed with an HNF-4 derivative lacking the AF2 domain [GST-HNF-4(ΔAF2)] or with the GST-Gal4 fusion protein that was used as an unrelated control (Fig.5A).

Figure 5

HNF-1 interacts with the AF2 domain of HNF-4 in vitro and in vivo. (A) In vitro synthesized [35S]methionine-labeled HNF-1 was incubated with the indicated GST fusion proteins, and the bound proteins were analyzed by 10% SDS–polyacrylamide gel electrophoresis (PAGE). Growth and expression of GST fusion proteins in Escherichia colistrain JM109 were performed as described (15).35S-labeled full-length recombinant HNF-1 was synthesized in vitro from the corresponding constructs in Bluescript KS (Stratagene) using the TNT coupled reticulocyte lysate system (Promega). Glutathione-Sepharose beads containing 2 μg of each fusion protein were incubated with 35S-labeled proteins in interaction buffer [100 mM KCl, 20 mM Hepes (pH 7.9), 0.1% NP-40, 5 mM MgCl2, 0.2% bovine serum albumin (BSA), 10% glycerol, 0.1 M phenylmethylsulfonyl fluoride, and aprotinin (10 μg/ml)] for 1.5 hours at 4°C with constant agitation. After extensive washing with the same buffer minus BSA and glycerol, the beads were resuspended in 20 μl of SDS-loading buffer and analyzed by SDS-PAGE; 8% of the input 35S-labeled HNF-1 is shown in the first lane. (B) COS-1 cells were transfected with 500 ng of nuclear localization–deficient mutant pMT-HNF-4(227–455) alone (pCB6 vector) or with 1 μg of pCB-HNF-1 FL, pCB-HNF-1(440), pCB-HNF-1(280), or pBx-Gal4 expression vectors, transferred to cover slips, and stained with polyclonal peptide antibody raised against the COOH-terminal 11–amino acid epitope of HNF-4, as described (14). The number of cells examined showing [nuclear]:[nuclear plus cytoplasmic]:[cytoplasmic] staining was 0:0:54 in HNF-4(227–455)–transfected cells. In cotransfected cells this ratio was 56:2:1 (HNF-1 FL), 51:1:2 [HNF-1(440)], 0:0:42 [HNF-1(280)], and 0:0:51 (Gal4). Typical examples of the immunofluorescent images are shown (magnifications, ×284).

Interaction between HNF-4 and HNF-1 in intact cells was determined by nuclear cotranslocation assays with the use of a mutant form of HNF-4 [HNF-4(227–455)], which lacks specific nuclear localization signals but contains the domain required for in vitro interaction with HNF-1. This mutant was detected exclusively in the cytoplasm of transfected COS-1 cells (Fig. 5B). Coexpression of HNF-4(227–455) with either full-length HNF-1 (FL) or HNF-1(440), but not with HNF-1(280) or Gal4 protein, resulted in its translocation to the nucleus (Fig. 5B). Thus, HNF-1–HNF-4 interaction required the HNF-1 domain located between amino acids 280 and 440, but not the COOH-terminal activation domains of HNF-1.

Taken together, our results indicate that the AF-2 domain of HNF-4 is sufficient and necessary for physical interaction with HNF-1 and for repression. This association may block the HNF-4 activation domain in a way that prevents either its interaction with coactivators that transduce AF2 activity to the transcription machinery or its direct interaction with general transcription factors. As a consequence, increased amounts of HNF-1 induce a regulatory mechanism that leads to the general down-regulation of HNF-4–dependent liver-specific genes, including the HNF-1 gene itself. This promoter-dependent dual function of HNF-1 suggests its central role in the coordination of the regulatory network that defines the hepatic phenotype.

REFERENCES AND NOTES

View Abstract

Stay Connected to Science

Navigate This Article