Transcription factor HNF-6/OC-1 inhibits the stimulation of the HNF-3α/Foxa1 gene by TGF-β in mouse liver


  • Nicolas Plumb-Rudewiez,

    1. Hormone and Metabolic Research Unit, Institute of Cellular Pathology and Université Catholique de Louvain, Brussels, Belgium
    Search for more papers by this author
    • N.P.-R. and F.C. contributed equally to this work.

  • Frédéric Clotman,

    1. Hormone and Metabolic Research Unit, Institute of Cellular Pathology and Université Catholique de Louvain, Brussels, Belgium
    Search for more papers by this author
    • N.P.-R. and F.C. contributed equally to this work.

  • Hélène Strick-Marchand,

    1. Unité de Génétique de la Différenciation, Institut Pasteur, Paris, France
    Search for more papers by this author
  • Christophe E. Pierreux,

    1. Hormone and Metabolic Research Unit, Institute of Cellular Pathology and Université Catholique de Louvain, Brussels, Belgium
    Search for more papers by this author
    • C.E.P. was senior research assistant of the FNRS (Belgium)

  • Mary C. Weiss,

    1. Unité de Génétique de la Différenciation, Institut Pasteur, Paris, France
    Search for more papers by this author
  • Guy G. Rousseau,

    1. Hormone and Metabolic Research Unit, Institute of Cellular Pathology and Université Catholique de Louvain, Brussels, Belgium
    Search for more papers by this author
  • Frédéric P. Lemaigre

    Corresponding author
    1. Hormone and Metabolic Research Unit, Institute of Cellular Pathology and Université Catholique de Louvain, Brussels, Belgium
    • Hormone and Metabolic Research Unit, Université catholique de Louvain and Institute of Cellular Pathology, Avenue Hippocrate 75, Box 7529, 1200 Brussels, Belgium
    Search for more papers by this author
    • fax: (32) 2-764-7507


A network of liver-enriched transcription factors controls differentiation and morphogenesis of the liver. These factors interact via direct, feedback, and autoregulatory loops. Previous work has suggested that hepatocyte nuclear factor (HNF)-6/OC-1 and HNF-3α/FoxA1 participate coordinately in this hepatic network. We investigated how HNF-6 controls the expression of Foxa1. We observed that Foxa1 expression was upregulated in the liver of Hnf6−/− mouse embryos and in bipotential mouse embryonic liver (BMEL) cell lines derived from embryonic Hnf6−/− liver, suggesting that HNF-6 inhibits the expression of Foxa1. Because no evidence for a direct repression of Foxa1 by HNF-6 was found, we postulated the existence of an indirect mechanism. We found that the expression of a mediator and targets of the transforming growth factor beta (TGF-β) signaling was increased both in Hnf6−/− liver and in Hnf6−/− BMEL cell lines. Using these cell lines, we demonstrated that TGF-β signaling was increased in the absence of HNF-6, and that this resulted from upregulation of TGF-β receptor II expression. We also found that TGF-β can stimulate the expression of Foxa1 in Hnf6+/+ cells and that inhibition of TGF-β signaling in Hnf6−/− cells down-regulates the expression of Foxa1. In conclusion, we propose that Foxa1 upregulation in the absence of HNF-6 results from increased TGF-β signaling via increased expression of the TGF-β receptor II. We further conclude that HNF-6 inhibits Foxa1 by inhibiting the activity of the TGF-β signaling pathway. This identifies a new mechanism of interaction between liver-enriched transcription factors whereby one factor indirectly controls another by modulating the activity of a signaling pathway. (HEPATOLOGY 2004;40:1266–1274.)

A network of liver-enriched transcription factors (LETFs) controls differentiation and morphogenesis of the liver.1–3 It comprises factors from several families that have been identified through biochemical and genetic studies. These include the CCAAT/enhancer binding proteins, the proline acid–rich factors, the homeodomain proteins of the hepatocyte nuclear factor (HNF)-1 family, winged helix proteins (HNF-3/FoxA), nuclear receptors (HNF-4, LXR, FXR and FTF/hB1F/CPF families), and the Onecut (OC) factors (HNF-6/OC-1, OC-2 and OC-3). These LETFs regulate the expression of genes whose products control the development of the liver and its physiological functions. Evidence for a network of LETFs stemmed from studies based on transfection of hepatoma lines, which showed that HNF-1α and HNF-4 can control each other.4–7 Other groups extended this notion by showing that a combination of LETFs acting in synergy can directly regulate the expression of a transcription factor gene.8, 9 Later, transgenic mice were generated in which genes for LETFs were knocked out or in which LETFs were overexpressed. Recently, chromatin immunoprecipitation studies have shown how LETFs can directly regulate the expression of another LETF via an ordered binding to its promoter10 or cell-specific chromatin-modifying activities.11 These experiments define in vivo hierarchical relationships between the factors that participate in this hepatic network.

HNF-6, a LETF of the OC family,12, 13 is essential for proper development of the biliary tract14 as well as the expression of hepatic genes.15–17 It participates in the network of LETFs.18 Indeed, HNF-6 controls directly the expression of Hnf1β,14 cooperates with HNF-1α to regulate the expression of Hnf4α,9 and interacts with FoxA2 without binding to DNA to stimulate the Foxa2 gene.19

Transcriptional regulation of LETFs is also controlled by extracellular signals, including extracellular matrix components,20, 21 fibroblast growth factors,22 bone morphogenic proteins,23 retinoic acid,24, 25 cytokines,26 growth hormone,18 and transforming growth factor beta (TGF-β).27 However, how these signaling pathways are integrated in the LETF network is poorly understood. In the present work, we show how HNF-6 indirectly controls the expression of Foxa1 via modulation of TGF-β signaling. This identifies a new mode of action of HNF-6 in the LETF regulatory network.


HNF, hepatocyte nuclear factor; BMEL, bipotential mouse embryonic liver; TGF-β, transforming growth factor beta; LETF, liver-enriched transcription factor; OC, Onecut; RT-PCR, reverse-transcriptase polymerase chain reaction; TβR, TGF-β receptor; mRNA, messenger RNA; Erk, extracellular signal-regulated kinase.

Materials and Methods

Reverse-Transcriptase Polymerase Chain Reaction (RT-PCR) and Real-Time Quantitative RT-PCR.

Total RNA was extracted from mouse embryonic liver or from bipotential mouse embryonic liver (BMEL) cells with TriPure reagent (Roche, Basel, Switzerland). Total RNA (500 ng for real-time PCR) was incubated for 10 minutes at room temperature and for 2 hours at 37°C in a volume of 25 μL containing 50 mmol/L Tris-Cl (pH 8.3), 75 mmol/L KCl, 3 mmol/L MgCl2, 10 mmol/L dithiothreitol, 200 units of Moloney murine leukemia virus reverse transcriptase (Life Technologies, Merelbeke, Belgium), 40 units of RNAsin (Promega, Leiden, The Netherlands), 3 μg of random hexamers (Life Technologies), and 4 mmol/L dNTPs (Amersham Pharmacia Biotech, Brecht-St. Lennarts, Belgium). For PCR amplification, 3 μL of a 3-fold dilution of the reverse transcription mix were incubated in a volume of 50 μL containing 10 mmol/L Tris-Cl (pH 9.0), 50 mmol/L KCl, 0.1% (vol/vol) Triton X-100, 1.5 mmol/L MgCl2, 40 μmol/L dNTP, 10 pmol of each primer, and 2.5 units of Taq polymerase (Promega). Unlabeled dCTP concentration was 20 μmol/L when the reaction was supplemented with 2.5 μCi of 3000 Ci/mmol [32P]dCTP (Amersham Pharmacia Biotech). The number of cycles (94°C, 30 seconds; 55°C, 30 seconds; and 72°C, 30 seconds) was determined for each amplicon to reach a midlogarithmic phase of amplification. Real-time PCR was performed with SYBR Green PCR Core Reagents (Perkin Elmer, Lennik, Belgium) on a MyiQ cycler (Bio-Rad, Eke, Belgium). FoxA1, TGF-β receptor (TβR)-II, PAI-1, and P15 messenger RNA (mRNA) were quantitated in at least three samples and normalized to β-actin mRNA.

Primer Sequences.


Knockout Mice.

In Hnf6−/− mice,28 the first exon of the Hnf6 gene, which contains essential transcriptional regulatory regions and the cut DNA-binding domain, was replaced with a neomycin resistance cassette to generate a null allele. All animals received humane care according to the criteria outlined in the Guide for the Care and Use of Laboratory Animals prepared by the National Academy of Sciences.

BMEL Cell Lines.

BMEL cell lines were obtained as described previously.29 Briefly, embryonic day 14.5 fetuses from Hnf6+/− mice breedings (129SVJ × Swiss background) were harvested, livers were dissociated in hepatocyte attachment medium (Invitrogen-Life Technologies, Merelbeke, Belgium), and liver cells were plated onto 10-cm plates coated with collagen I (Sigma, Bornem, Belgium). The cells were cultured in a humidified atmosphere with 5% CO2 at 37°C. The next day, and weekly thereafter, the medium was replaced and cultured under basal conditions—that is, with RPMI 1640 (Invitrogen) supplemented with 10% fetal calf serum, 50 ng/mL epidermal growth factor (PeproTech, London, England), 30 ng/mL insulin-like growth factor II (PeproTech), 10 mg/mL insulin (Roche, Vilvoorde, Belgium), Fungizone (Invitrogen), and antibiotics. After 13 to 19 weeks, islands of cells displaying epithelial morphology were isolated and further cultured. Seven wild-type and 12 Hnf6−/− colonies were expanded, from which one Hnf6+/+ line and one Hnf6−/− line were selected for this study. Aggregate cultures or cultures on Matrigel were as described.29

Transient Transfections and Luciferase Reporter Gene Assays.

BMEL cells (50,000 per well on 24-well plates) were transfected in RPMI 1640 medium supplemented with 10% fetal calf serum using lipofectamine 2000 (Life Technologies), 400 ng pFoxa1(−479/+66)-luc or 100 ng pCAGA12-MLP-luc (a gift from C. Hill, Imperial Cancer Research Fund, London, United Kingdom) reporter plasmid, 25 ng or 100 ng of pGK-TβRII-expression vector (a gift from C. Mummery, Netherlands Institute for Developmental Biology, Utrecht, the Netherlands),30 and 15 ng of pRL-SVK3 internal control, as indicated in the figures. After 3 hours of transfection, the cells were washed and further incubated in serum-free medium in the absence or presence of TGF-β1 R&D Systems, Abingdon, UK). Luciferase activities were measured after 48 hours via dual luciferase assay in a TD-20/20 luminometer (Promega). Luciferase activities were expressed as the ratio of reporter activity (firefly luciferase) to internal control activity (Renilla luciferase). The mouse Foxa1 gene promoter (from −479 to +66) was isolated via PCR with oligonucleotides containing PstI and BglII restriction sites and was cloned upstream of the luciferase coding sequence in pGL3-basic (Promega). To evaluate the effect of HNF-6 on TβRII expression, BMEL cells (100,000 per well on 12-well plates) were transfected with 3 μg of pCMV-HNF-6 expression vector.

Cell Extracts and Western Blotting.

BMEL cells (300,000 per 6-cm dish) were cultured for 12 hours in serum-free RPMI 1640 medium and washed twice with phosphate-buffered saline before preparation of cytoplasmic and nuclear extracts as described previously.31 Cytoplasmic extracts (20 μg protein) or nuclear extracts (30 μg protein) were loaded on sodium dodecyl sulfate–polyacrylamide gels for Western blotting with anti-FoxA1 (a kind gift from W. Schmid, German Cancer Research Center, Heidelberg, Germany), anti–β-actin (Sigma, #A-5441), anti–phospho-Smad2 (Upstate, Cell Signaling #07-392, Leusden, The Netherlands), anti-Smad2/3 (BD Biosciences, #610842, Erembodegem, Belgium), anti-activated mitogen-activated protein kinase (Sigma, #M8159), anti-Erk1 (Santa Cruz Biotechnology Cat:sc-93, Tebu-Bio, Boechout, Belgium), anti-Erk2 (Santa Cruz Biotechnology Cat:sc-154), or anti-proliferating cell nuclear antigen (Biognost, #MAB424, Heule, Belgium) antibodies. Secondary antibodies were HRP-Conjugated Goat Anti–Rabbit Ig–Specific (BD Biosciences, #554021) and HRP-Conjugated Goat Anti–Mouse Ig–Specific (BD Biosciences, #554002). Secondary antibodies were detected using electrochemiluminescence (Amersham Pharmacia Biotech).


Foxa1 Gene Expression is Upregulated in the Livers of Hnf6−/− Mouse Embryos.

Our previous work suggested that the expression of Foxa1 is upregulated in the livers of Hnf6−/− mouse embryos.14 We confirmed these observations using quantitative RT-PCR. These experiments, performed at embryonic day 12.5, showed that inactivation of the Hnf6 gene leads to a 2-fold upregulation of Foxa1 (Fig. 1A), suggesting that HNF-6 inhibits Foxa1 expression during liver development. To investigate whether this inhibition is direct or indirect, we searched for putative HNF-6 binding sites in the sequence of the mouse Foxa1 gene promoter32–34 but found none. This excluded that HNF-6 represses the Foxa1 gene through direct binding to its promoter. In this gene, we found HNF-6 binding sites in a 1-kb region located from −5250 to −6298 and conserved in the human gene. However, HNF-6 did not inhibit Foxa1 gene expression via this region in transient transfection experiments (data not shown). This suggested that the negative regulation of Foxa1 expression by HNF-6 could result from an indirect effect.

Figure 1.

Inactivation of the Hnf6 gene increases expression of Foxa1 and of components of the TGF-β signaling pathway in developing liver. The expression of the indicated gene products was measured in livers from Hnf6+/+ and Hnf6−/− mouse embryos at embryonic day 12.5. (A) Quantitative RT-PCR of Foxa1 mRNA. Data are the mean ± SEM (n = 5). *P < .02. (B) RT-PCR of mRNA from genes involved in TGF-β signaling as revealed by ethidium bromide-staining of agarose gels. The TATA box-binding protein mRNA (Tbp) served as control. H2O refers to RT-PCR without input RNA. Numbers to the right of the gel refer to the expression level in Hnf6−/− relative to wild-type liver, as determined in the microarray (MA) analysis. nd, not determined.

The Expression of Components of the TGF-β Signaling Cascade is Increased in Hnf6−/− Embryonic Liver.

To search for a mechanism through which HNF-6 could indirectly inhibit the Foxa1 gene, we analyzed DNA microarray data obtained on embryonic day 12.5 Hnf6+/+ and Hnf6−/− livers.35 This analysis revealed that the mRNA levels of TβRII and of P15, a gene known to be stimulated by TGF-β,36 were approximately twofold higher in Hnf6−/− livers than in Hnf6+/+ livers. These data, which were confirmed via RT-PCR (Fig. 1B), suggest that TGF-β signaling is increased in Hnf6−/− livers. This is supported by the observation that the expression level of PAI-1, another target of the TGF-β cascade37 not analyzed in the microarray experiment, was upregulated in Hnf6−/− livers (Fig. 1B). Taken together, these data suggest that HNF-6 represses TGF-β signaling in developing liver.

Generation of BMEL Cell Lines.

To confirm that TGF-β signaling is increased in the absence of HNF-6 and to determine whether this relates to increased Foxa1 gene expression, we generated cell lines from Hnf6+/+ and Hnf6−/− embryos according to the procedure of Strick-Marchand and Weiss.29 As expected, Hnf6−/− and Hnf6+/+ BMEL cells expressed LETFs typical of hepatic cells (Fig. 2A). Consistent with the expected bipotential character of BMEL cells,29Hnf6+/+ and Hnf6−/− cells could be induced to express hepatic or biliary differentiation markers. When cultured as aggregates, the cells were induced to express the hepatocyte marker alcohol dehydrogenase, and when grown on Matrigel the cells expressed the biliary marker γ-glutamyl transpeptidase-4 (γ-Gt4) (Fig. 2A). The induction of γ-glutamyl transpeptidase-4 was stronger in the Hnf6−/− cells than in the wild-type cells, which is in line with the known biliary phenotype of Hnf6−/− mice.14

Figure 2.

The increases in gene expression resulting from inactivation of the Hnf6 gene are cell-autonomous, and the gene expression profile of Hnf6−/− BMEL cell lines mimics that of Hnf6−/− embryonic liver. The expression of the indicated gene products was measured in Hnf6+/+ and Hnf6−/− BMEL cell lines. (A) RT-PCR showing that the two cell lines express LETFs, as expected for hepatic cells (left panel). The induction of the hepatocyte marker alcohol dehydrogenase in aggregate cultures and of γ-glutamyl transpeptidase-4 in the cells grown on Matrigel is also shown, demonstrating that the cell lines are bipotential (right panel). The two markers were not detected under basal conditions. (B) Quantitative RT-PCR showing upregulation of Foxa1 mRNA in Hnf6−/− BMEL cells compared with Hnf6+/+ BMEL cells after 12 hours of culture in RPMI 1640 without serum and growth factors (left panel). The right panel shows a Western blotting experiment indicating that the FoxA1 protein level is increased in nuclear extracts of Hnf6−/− BMEL cells cultured for 12 hours under the same conditions. The pairs of lanes correspond to separate experiments. (C) Quantitative RT-PCR showing that the increase in TβRII, Pai1, and P15 gene expression detected following inactivation of the Hnf6 gene persists in BMEL cells cultured for 12 hours in RPMI 1640 without serum and growth factors. (D) Quantitative RT-PCR showing that transfection of HNF6 in Hnf6−/− BMEL cells results in a decrease of TβRII mRNA 24 hours after transfection. Data are the mean ± SEM for at least 4 experiments. *P < .02; **P < .001.

To determine if BMEL cells are a valid model for investigating the mechanism of Foxa1 regulation by HNF-6, we first compared Foxa1 gene expression in Hnf6−/− and Hnf6+/+ cells. As shown in Fig 2B, the concentration of Foxa1 mRNA, measured with quantitative RT-PCR, was higher in the absence of HNF-6, as seen in embryonic liver (Fig. 1A). Western blot analysis indicated that FoxA1 protein levels were also increased in Hnf6−/− BMEL cells (Fig. 2B). We compared the expression of genes coding for factors involved in the TGF-β pathway, namely TβRII, Pai1, and P15. Fig. 2C shows that, as in Hnf6−/− embryonic liver (Fig. 1B), expression of these genes was upregulated in Hnf6−/− cells. In addition, when Hnf6−/− cells were transfected with a HNF-6 expression vector, TβRII mRNA levels decreased, confirming that upregulation of TβRII mRNA in untransfected Hnf6−/− cells is HNF-6–dependent (Fig. 2D). Taken together, our data indicate that Hnf6−/− cells mimic perturbations in gene expression observed in vivo. Therefore, they are a good model for studying the upregulation of Foxa1 expression and the potential increase in TGF-β signaling that result from inactivation of the Hnf6 gene.

The TGF-β Pathway is Stimulated in Hnf6−/− BMEL Cells.

TGF-β signaling involves the ligand-induced formation of a heteromeric TβRI/TβRII receptor complex that phosphorylates the proteins Smad2 and Smad3. Receptor-activated Smad2 and Smad3 bind to Smad4, and these complexes translocate to the nucleus, where they stimulate transcription of genes via binding to CAGA-containing cis-acting sequences.38–40 TGF-β signaling also leads to extracellular signal-regulated protein kinase (Erk) activation.41, 42

To test whether TGF-β signaling is increased in Hnf6−/− BMEL cells, we performed Western blotting experiments (Fig. 3A). Total and phosphorylated Smad2 or Erk proteins were visualized in cytoplasmic and nuclear extracts from Hnf6+/+ and Hnf6−/− cells. No exogenous TGF-β was added to the culture medium. Total amounts of Smad2 and Erk, as quantitated with ImageJ software (National Institutes of Health, Bethesda, MD), were similar in Hnf6+/+ and Hnf6−/− cells. Phosphorylated Smad2, found only in the nuclear fraction as expected, was two times more abundant in the Hnf6−/− nuclei. Similarly, higher amounts of phosphorylated Erk protein were present in the Hnf6−/− cells compared with Hnf6+/+ cells. These observations indicate that the activation of TGF-β signaling mediators is increased in the absence of HNF-6.

Figure 3.

The TGF-β signaling pathway is stimulated in Hnf6−/− BMEL cells. (A) Western blots showing that Smad2 and Erk1 phosphorylation is increased in Hnf6−/− BMEL cells. Nuclei of Hnf6−/− cells contains 2 times more phosphorylated Smad2 than nuclei of Hnf6+/+ BMEL cells (lines 1 and 2). The cytoplasm of Hnf6−/− cells contains 2 times more phosphorylated Erk1 than the cytoplasm of Hnf6+/+ cells (lines 3 and 4). Total quantities of Smad2 and Erk (as measured with ImageJ software) are similar in Hnf6+/+ cells and in Hnf6−/− cells. The pairs of lanes correspond to separate experiments. (B) The activity of a TGF-β–luciferase reporter construct (CAGA-luc) is higher in transiently transfected Hnf6−/− BMEL cells than in Hnf6+/+ cells (n = 7). Data are the mean ± SEM. *P < .001. PCNA, proliferating cell nuclear antigen.

To confirm that the activation of TGF-β signaling mediators observed in Hnf6−/− BMEL cells results in increased expression of target genes, we compared the activity of a TGF-β reporter construct containing the luciferase coding sequences downstream of a promoter containing 12 Smad-binding CAGA sites (pCAGA12-MLP-luc) in Hnf6+/+ and in Hnf6−/− BMEL cells. In the absence of exogenous TGF-β, the relative activity of the TGF-β reporter construct was 2-fold higher in the Hnf6−/− cells than in Hnf6+/+ cells (Fig. 3B), demonstrating that the TGF-β signaling pathway is more active in the absence of HNF-6.

Upregulation of Foxa1 Gene Expression in the Absence of HNF-6 Results From Increased TGF-β Signaling.

Because both the expression of the Foxa1 gene and the activity of the TGF-β pathway are upregulated in the absence of HNF-6, we hypothesized that upregulation of the Foxa1 gene results from increased TGF-β signaling. As a first step to test this hypothesis, we verified if the Foxa1 gene could be stimulated by TGF-β in BMEL cells. Incubation of Hnf6+/+ BMEL cells with TGF-β1 induced a 3.5-fold increase in Foxa1 mRNA levels as measured with quantitative RT-PCR (Fig. 4A). To determine if the TGF-β pathway targets the Foxa1 promoter itself, we tested whether a Foxa1 promoter/luciferase construct transfected in Hnf6+/+ BMEL cells could be stimulated by treating the cells with TGF-β. Indeed, TGF-β1 induced a 2.7-fold increase in relative luciferase expression (Fig. 4B). Because we showed that the TGF-β pathway is stimulated in Hnf6−/− BMEL cells, we also expected to find a higher activity of a transfected Foxa1 promoter in these cells compared with Hnf6+/+ BMEL cells. Fig. 4C shows that the relative activity of the Foxa1 promoter was 2.2-fold higher in the Hnf6−/− BMEL cells. From this set of experiments, we concluded that the Foxa1 gene is stimulated by TGF-β in liver cells and that this is—at least in part—promoter dependent.

Figure 4.

Foxa1 gene transcription is stimulated by TGF-β. This can be mimicked by increasing TβRII levels in Hnf6+/+ BMEL cells and depends on both the Smad and Erk cascades. (A) Incubation of Hnf6+/+ BMEL cells for 10 hours with TGF-β1 in the absence of serum increases Foxa1 mRNA, as measured with quantitative RT-PCR (n = 5). (B) Treatment of Hnf6+/+ BMEL cells with TGF-β1 for 24 hours stimulates the activity of a transiently transfected Foxa1 promoter (−479/+66)-luciferase construct (Foxa-Luc) (n = 5). (C) The activity of the Foxa1 promoter, measured in transient transfections, is higher in Hnf6−/− than in Hnf6+/+ BMEL cells (n = 7). Transient transfection of Hnf6+/+ cells with a TβRII expression vector stimulates Foxa1 promoter activity (n = 3). (D) Transient transfection of Hnf6+/+ BMEL cells with increasing amounts of a TβRII expression vector stimulates the activity of the TGF-β reporter construct (n = 3). (E) Incubation of Hnf6−/− BMEL cells for 48 hours with an inhibitor of Smad phosphorylation (SB431542) or with the MEK inhibitor PD98059 reduces Foxa1 mRNA levels, as measured with quantitative RT-PCR and normalized for cells incubated with vehicle only (n = 3). Data are the mean ± SEM. *P < .02; **P < .001. TGF-β1, transforming growth factor beta 1; Co, control.

In Hnf6−/− liver or Hnf6−/− BMEL cells, the expression of TβRII is increased as compared with Hnf6+/+ liver or Hnf6+/+ cells (Figs. 1, 2). Because Hnf6+/+ and Hnf6−/− BMEL cells produce similar amounts of TGF-β1, TGF-β2, and TGF-β3 (data not shown), we asked whether increasing TβRII concentration in liver cells was sufficient to increase the TGF-β response in the absence of exogenously added TGF-β. Hnf6+/+ cells were cotransfected with the TGF-β reporter construct pCAGA12-MLP-luc and increasing amounts of a TβRII expression vector. Figure 4D shows that transfection of TβRII induced a dose-dependent increase in reporter activity, demonstrating that an increase in TβRII in liver cells is sufficient to increase TGF-β signaling. The same experiment was performed to test whether upregulation of TβRII expression would be sufficient to stimulate Foxa1 expression. Indeed, cotransfection of a TβRII expression vector with the Foxa1 promoter-luciferase construct in Hnf6+/+ BMEL cells resulted in increased Foxa1 promoter activity (Fig. 4C). We conclude that, in liver cells, an increase in TβRII levels can stimulate the TGF-β pathway and Foxa1 expression.

To demonstrate that the increase of Foxa1 expression in Hnf6−/− BMEL cells results from increased TGF-β signaling, we investigated whether incubating Hnf6−/− BMEL cells with inhibitors of TGF-β signaling reduces Foxa1 expression. Hnf6−/− BMEL cells were incubated with SB431542, a compound known to inhibit phosphorylation of Smads by TβRI,43, 44 or with PD98059, an inhibitor of the Erk pathway. Both compounds reduced Foxa1 mRNA levels (Fig. 4E). We concluded that increased Foxa1 expression in Hnf6−/− BMEL cells results from increased TGF-β signaling, and that Foxa1 lies downstream of both the Smad and the Erk components of TGF-β signaling.


From the analysis of Hnf6−/− embryonic liver and cell lines presented here, we conclude that HNF-6 indirectly inhibits Foxa1 expression. We cannot exclude that this repression is exerted in part through direct binding of HNF-6 to the Foxa1 gene. However, this is unlikely given that neither the Foxa1 promoter region, which is known to confer in vivo gene activity,33, 34 nor an upstream conserved region identified in the present work, were directly responsive to HNF-6 in transient transfection assays. In addition, re-expression of HNF-6 in Hnf6−/− BMEL cells for 24 hours in a transient transfection experiment was not sufficient to restore normal Foxa1 mRNA levels (data not shown). Therefore, our data point to an indirect regulation of Foxa1 expression by HNF-6 mediated by the TGF-β pathway.

To investigate the mechanism of this regulation, we resorted to BMEL cells.29 Both Hnf6+/+ and Hnf6−/− cells behaved as hepatoblasts because they expressed the expected set of LETFs and they could be induced to express hepatocytic or biliary differentiation markers. We therefore confirm that such cell lines can be established from various mouse strains, including knockout mice. Not only did we obtain BMEL cells from wild-type and Hnf6−/− embryos with a similar efficiency, we also noted that the gene expression profile of the Hnf6−/− line mimicked that of the Hnf6−/− embryonic liver. This demonstrates the usefulness of establishing BMEL cell lines to study mechanisms regulating liver development. It should be noted that the down-regulation of Hnf1β expression that was previously detected in Hnf6−/− liver was not observed in Hnf6−/− BMEL cells, suggesting that BMEL cells do not fully recapitulate the Hnf6−/− hepatic phenotype. The lack of Hnf1β down-regulation in Hnf6−/− BMEL cells may result from the fact that the down-regulation of HNF-1β in Hnf6−/− liver is not fully penetrant and is transient.14

We have shown here that TGF-β can stimulate Foxa1 expression and that this action is promoter dependent. The minimal promoter of the Foxa1 gene has been characterized,32–34 but no TGF-β response elements have been described. Our analysis of this promoter region revealed a CG-rich sequence with many potential Sp1 and Smad-binding sites. The sequence has similarity with the P15 promoter, which is known to be stimulated by both Smad and Erk signal transducers36, 45 and in which Sp1 binding is critical for its response to Smad proteins.46 It is known that Sp1 and other transcription factors can physically interact with Smad proteins and are required for the expression of TGF-β target genes47; whether this is the case for Foxa1 remains to be determined.

Another conclusion of the present work is that HNF-6 represses Foxa1 expression by inhibiting TGF-β signaling. The mechanism through which HNF-6 inhibits TGF-β signaling depends, at least in part, on down-regulation of TβRII expression. Indeed, in the absence of HNF-6, BMEL cells and embryonic liver show increased TβRII expression, and overexpression of TβRII in Hnf6+/+ cells induces a stimulation of TGF-β–responsive genes. Moreover, blocking TGF-β signaling at the receptor level in Hnf6−/− cells reduces the stimulation of the TGF-β pathway. Further work is needed to investigate how HNF-6 controls TβRII expression. A potential HNF-6 binding site has been found in the sequence of the human TβRII gene,48 but this region is not conserved in the mouse TβRII gene, and we found no candidate HNF-6 binding sites in the region 3 kb upstream of the initiator ATG. As HNF-6 inhibits TGF-β signaling in the liver, some effects of HNF-6 in this organ could result from decreased expression of TGF-β target genes other than Foxa1. We show here that the expression of the P15 gene is indeed controlled by HNF-6. An HNF-6–dependent inhibition of the P15 gene, which codes for the cyclin-dependent kinase inhibitor P15(INK4B), is expected to promote hepatocyte proliferation.46 Consistent with this, we found that the liver of Hnf6−/− mice is hypoplastic (F.C., G.G.R., and F.P.L., unpublished data, December 2002).

Many genes are directly regulated by both HNF-6 and FoxA1. We speculate that the HNF-6 → Foxa1 cascade constitutes a failsafe mechanism through which an increase in FoxA1 protein may at least in part compensate for reduced HNF-6 activity. This might for instance explain why the genes coding for alpha fetoprotein and transthyretin, which are regulated by both HNF-6 and FoxA1, are only marginally affected in Hnf6−/− embryonic liver (14% and 16% loss of expression, respectively; data not shown). When two factors of the same family that share DNA binding and transcriptional properties are coexpressed, one member can compensate for the absence of the other,49 a process called paralogous compensation. The present work points to another mechanism, namely heterologous compensation, in which the loss of one transcription factor (HNF-6) would be compensated for by the activity of a factor belonging to a different family (FoxA1).

Previous work on HNF-6 has shown that it controls the transcription of its target genes in four different ways (Fig. 5). The first is direct regulation (Fig. 5A), whereby HNF-6 binds to and stimulates the promoter of a LETF gene, as has been illustrated for HNF-1β.14 A second way (Fig. 5B) is transcriptional synergy, whereby HNF-6 cooperates with another LETF to stimulate the target gene promoter, as is the case for the stimulation of the Cyp2c12 gene by HNF-6 and FoxA2.50 These first two mechanisms may be combined (Fig. 5D) as when HNF-6 and HNF-1α stimulate transcription of the HNF-4α gene synergistically.9 On the other hand, Rausa et al.19 have shown that HNF-6 may not bind DNA directly but that it interacts with FoxA2 bound to the Foxa2 gene promoter to recruit CBP/p300 and therefore acts as a transcriptional coactivator (Fig. 5C). The present work identifies a fifth mechanism, in which HNF-6 indirectly controls expression of a LETF by modulating the response of the LETF gene to an extracellular signal (Fig. 5E). To our knowledge, this is the first report of two factors of the LETF network that are indirectly connected via modulation of a signaling pathway.

Figure 5.

Five mechanisms through which HNF-6 participates in the network of LETFs. HNF-6, hepatocyte nuclear factor 6; LETF, liver-enriched transcription factor; TF, transcription factor.


The authors thank C. Hill, C. Mummery, and W. Schmid for reagents; A. Zwijsen and members of the HORM unit for discussions; and M.-A. Gueuning for technical assistance.