Introduction

Hepatocyte nuclear factor 4 (HNF4), a member of the nuclear receptor superfamily, is essential for metabolic processes, as revealed by analyses of Hnf4 gene disruptions in the mouse (see below). Originally purified from rat liver nuclear extracts (Sladek et al, 1990), HNF4 is also present in the intestine, pancreas, kidney and stomach, as well as in the visceral endoderm of the yolk sac during mouse embryogenesis (Duncan et al, 1994; Taraviras et al, 1994; Nakhei et al, 1998). Absence of HNF4 is embryonic lethal (Chen et al, 1994; Duncan et al, 1997; Parviz et al, 2003) and disruption of the gene, specifically in adult liver, leads to death, due to severe dyslipidemia, high-serum bile acid levels and ureagenesis defects (Hayhurst et al, 2001; Inoue et al, 2002). In addition, HNF4 is involved in glucose homeostasis, especially in neoglucogenesis (Rhee et al, 2003), and disruption of the gene in pancreatic -cells leads to impaired secretagogue-stimulated insulin secretion (Gupta et al, 2005), as reported for Hnf4-associated non-insulin-dependent diabetes mellitus MODY1 in humans (Yamagata et al, 1996). Extra-pancreatic abnormalities have been reported in MODY1 patients showing low plasmatic triglyceride, apoAII and apoCIII levels, most probably due to hepatocyte defects (for reviews, see Ryffel, 2001; Sladek and Seidel, 2001). Thus, HNF4 is necessary for metabolic functions in liver and for proper insulin secretion in the pancreas. The central role of HNF4 is also highlighted by the huge number of putative HNF4 target genes, as reported in analysis combining chromatin immunoprecipitation (ChIP) from hepatocytes and pancreatic islets with promoter microarrays (Odom et al, 2004).

HNF4 can be divided into six domains (A–F), the A/B and F domains being the most variable among nuclear receptors. The Hnf4 gene gives rise to several isoforms differing within these domains, via internal splicing and transcription from two alternative promoters. HNF41 and HNF47 are the prototypes of the isoforms derived from the P1 and P2 promoters, respectively, and differ only by their first exons (Nakhei et al, 1998). Just like many other nuclear receptors (Warnmark et al, 2003), the HNF41 protein contains two activating functions (AF): the conserved AF-2 motif in the C-terminal ligand-binding domain and an AF-1 motif contained at the N-terminus, encoded by the 1-specific first exon (Hadzopoulou-Cladaras et al, 1997; Sladek et al, 1999). These motifs are essential for HNF4 transactivation capacity in vitro and act via recruitment of cofactors (e.g., Wang et al, 1998; Sladek et al, 1999; Yoon et al, 2001). In contrast with HNF41, no AF-1 activity was found for the HNF47 A/B domain using one-hybrid assays. Hence, HNF41 and HNF7 display distinct capacities to interact with cofactors (Torres-Padilla et al, 2002; Eeckhoute et al, 2003). In addition, the two promoters show different tissue-specific activities: P1-derived HNF41 is almost exclusive in the liver, whereas P2-derived HNF47 may be the predominant form in the pancreas (Nakhei et al, 1998; Boj et al, 2001; Eeckhoute et al, 2003).

Since HNF41 and HNF47 differ only by the presence or not of a functional AF-1 motif, these isoforms should present both redundancy and specificity, depending on the cell type and the target gene. The structure of the Hnf4 gene permitted us to design the first direct test in vivo of HNF4 AF-1 function, making use of a naturally occurring AF-1-deficient variant rather than targeted deletion, as reported for retinoid X receptor (RXR) (Mascrez et al, 2001). We created two 'knock-in' mouse lines expressing only HNF41 or HNF47 (and their splice-derived variants) under control of both promoters, replacing by homologous recombination either the HNF41 first exon-coding sequence by that specific for HNF47 or the reciprocal. The resulting mice present hepatic and possible pancreatic defects, revealing the physiological role of the HNF41 AF-1 motif. Furthermore, analysis of the gene expression pattern in livers of these mice has permitted identification of target genes whose expression is AF-1 dependent.

Results and discussion

HNF41 and HNF47 transcript accumulation in HNF4-expressing tissues

To investigate the roles of P1 and P2 promoter-derived isoforms in the mouse (Figure 1A and B), it was essential to know which isoform is present in each HNF4-expressing tissue. By replacing the 1 isoform by 7 in a tissue expressing predominantly or exclusively 1 in the wild-type (WT) mouse, we expected to diminish those functions for which the 1-specific AF-1 motif is required. In reciprocal fashion, the expression of target genes in tissues expressing mainly HNF47 could be affected by ectopic addition of the AF-1 motif.

The liver contains only traces of HNF47 transcripts and kidney expresses exclusively HNF41, whereas both isoforms are expressed in the intestine (Figure 1C; Nakhei et al, 1998; Briancon et al, 2004). While information in the literature concerning HNF4 isoforms in the pancreas and islets is contradictory (Boj et al, 2001; Eeckhoute et al, 2003), we found essentially HNF47 transcripts, while those for HNF41 were hardly detectable in either whole mouse pancreas or isolated islets (Figure 1C). Similarly, only HNF47 was detected for the stomach, contradicting earlier results (Nakhei et al, 1998).

Hnf41/7 reciprocal replacement

The most direct approach to determine the roles of the 1 and 7 isoforms would be to generate isoform-specific knockouts. However, since P1 and P2 promoter usage is tissue specific, in order to maintain HNF4 levels constant, we expressed either the 1 or the 7 isoform from both promoters. Targeting constructs contained the coding sequence (CDS) of the alternative first exon flanked by promoting/5'untranslated regions (5'UTR) and intronic sequences from the endogenous locus (Figure 2A and B). In one line, the exon 1D CDS was deleted and replaced by that of exon 1A, and the reciprocal for the other targeting event. This minimal intervention strategy permits conservation of the ratio among the internal splicing isoforms (see below). Recombinant ES cells were obtained (Supplementary Figure S1) and injected into blastocysts to obtain founder chimeras. Mice that are homozygous for exon 1D and exon 1A replacements are referred to as '1-only' and '7-only' mice, respectively.

It is worth noting that insertion of the floxed phosphoribosyltransferase gene (neo) in the HNF47-specific first intron (Figure 2A) was not lethal in the homozygous mice, contrary to the same insertion downstream of the HNF41-specific first exon (Figure 2B): this suggests that the knockout of P2-derived transcripts would be viable (see Results section in Supplementary Figure S1). Neo cassettes were deleted by mating with pgk-cre mice.

The 1-only and 7-only mouse livers express the expected isoform under control of both P1 and P2 promoters

The 1-only and 7-only mice are viable, fertile and do not present any obvious phenotype. We investigated whether the profile of the HNF4 isoforms was as expected in the isoform replacement mice, focusing on the liver. In the adult 7-only mouse liver, HNF47 is strongly expressed from the P1 promoter and HNF41 transcripts are not detectable (Figure 2C). No HNF47 transcripts could be detected in the 1-only mouse liver. We also verified that the ratio between the C-terminal splicing-derived isoforms, 2 and 8, compared to 1 and 7, was maintained in the mutant mice (Figure 2D). This confirms that the exon replacement does not affect downstream splicing events.

In Western blots (Figure 2E), HNF4 proteins from liver extracts of the 7-only mice migrate faster than those in WT and 1-only samples, consistent with the expected smaller protein. In addition, total amounts of HNF4 proteins appear grossly equivalent among the different genotypes. On 7-only and WT liver sections, using adequate antibodies, the isoform replacement was validated at the protein level, HNF4 being restricted to hepatocytes and excluded from bile duct cells (Figure 2F).

The 7-only mice present a nonlethal dyslipidemia

To reveal potential subtle defects in the mutant mice, a panel of electrolytes, metabolites, nutrients, enzymes and hormones was investigated (Figure 3A). As the adult-liver specific Hnf4 disruption causes ureagenesis and lipid and bile acid metabolism defects (Hayhurst et al, 2001; Inoue et al, 2002), we anticipated that the 7-only mice would present an attenuated form of these metabolic dysfunctions. Indeed, serum cholesterol and triglyceride contents were lower in these mice compared to WT (41% and 53% decrease, respectively), whereas FFA levels were slightly diminished (by 24%) and ketone body contents were increased (by 82%). The cholesterol content in all lipoprotein fractions was diminished in the 7-only mice, as shown on fast-performance liquid chromatography (FPLC) lipoprotein profiles (Figure 3B). These results reveal that the HNF4 AF-1 domain is implicated in lipid metabolism, since its absence in 7-only mice leads to dyslipidemia. However, blood bile acid and urea levels were not altered.

In addition to lipid metabolism defects, serum bilirubin levels were higher in 7-only females. This could reflect a hepatic defect, and is consistent with low levels of constitutive androstane receptor (CAR), a main regulator of bilirubin clearance (see below and Huang et al, 2003). Iron amounts were reduced in the 7-only mouse serum and this might be due to intestinal absorption and/or hepatic defects (excluding a possible alteration of hepatic transferrin expression; Figure 5D). Other serological parameters associated with hepatic functions (albumin, ALAT and coagulation tests) were unchanged.

Since kidney expresses only HNF41, 7-only mice could present renal defects, but serum creatinin and urea levels were not affected. However, since HNF4 is expressed in renal proximal tubules and is absent from glomeruli (Chabardes-Garonne et al, 2003), renal re-absorption defects might occur. No glucosuria was detected (n=8) but potassium concentrations were reduced by 16% in 7-only females, indicating possible gender-dependent defects in renal re-absorption. Alternatively, this hypokaliemia may be due to intestinal absorption defects.

Lipid accumulation in the 7-only mouse liver

Focusing on dyslipidemia, we investigated whether 7-only mice accumulated lipids in liver as in the liver-specific Hnf4 null mice. Indeed, the 7-only mice present slight hepatic steatosis. Small lipid droplets, homogeneously distributed within the parenchyma, were revealed by Oil Red O staining of liver sections from 5-month-old 7-only mice (Figure 3C). Interestingly, this steatosis was very subtle in 9-week-old mice (not shown), indicating a chronic amplification of metabolic defects with age. To exacerbate phenotypic differences, animals were fasted for 24 h to stimulate release of fatty acids from adipose tissue and their accumulation/catabolism in hepatocytes. As expected, fat droplets, much larger than in controls, were revealed in 7-only livers.

Expression of some key genes of lipid metabolism/transport is affected in 7-only mouse liver

To understand the basis for the dyslipidemia presented by the 7-only mice, we investigated expression of a panel of hepatic genes, mostly known to be transactivated by HNF4 and/or for which expression was altered in the liver-specific HNF4 KO (Hayhurst et al, 2001). Strikingly, very few genes involved in lipid transport/metabolism were strongly deregulated in the 7-only livers (Figure 4A–F). 1-only mice were checked for some of these deregulated genes and presented normal transcript levels (Supplementary Figure S2).

Hepato-specific downregulation of apoAIV and apoCIII expression. In the 7-only livers, transcripts of the HDL component apoAIV were nearly undetectable, and components apoCII and apoCIII of VLDL were reduced by 64 and 19%, respectively (Figure 4A). This suggests that the 1-specific AF-1 domain is required for full expression of these putative HNF4 direct targets (Hayhurst et al, 2001; Odom et al, 2004 and references therein).

Disruption of the apoAIV gene in the mouse triggered a decrease in VLDL/HDL-cholesterol levels and also in triglyceride levels due to altered expression of the apoCIII gene within the same gene cluster (Maeda et al, 1994; Weinstock et al, 1997). Hence, in the 7-only mice, combined diminished hepatic expression of the apoAIV and apoCIII genes could contribute to low serum cholesterol and triglyceride levels. However, since apoAIV and apoCIII are expressed both in the liver and the intestine, with apoAIV being mainly expressed in the intestine (Elshourbagy et al, 1985), we investigated expression of both genes in this tissue. Their downregulation in the 7-only mice is hepato-specific: no decline in transcript levels was revealed in the 7-only mouse intestine (Figure 4G). Thus, expression of the apoAIV gene in liver and intestine is regulated by HNF4, but only in the former is AF-1 necessary (Ktistaki et al, 1994; Sauvaget et al, 2002).

VLDL secretion may be altered. The two genes essential for VLDL secretion MTP and apoB (Figure 4A and E) were both downregulated by 25% in 7-only liver compared to WT. In the endoplasmic reticulum, the abetalipoproteinemia-associated gene product MTP is absolutely required for lipid assembly with apoB and, thus, for VLDL secretion (Leung et al, 2000). In addition, mice heterozygous for a disrupted apoB allele presented reduced cholesterol levels in all lipoprotein fractions (Farese et al, 1995; Huang et al, 1995), and in another study (Leung et al, 2000) reduced triglyceride levels and higher hepatic triglyceride accumulation. Thus, the combined small decreases in apoB and MTP transcript levels in the 7-only mouse livers may contribute to low-serum triglyceride and cholesterol levels and to steatosis.

Lipid uptake from blood may be increased. While the rate-limiting enzyme for hydrolysis of VLDL and chylomicron triglyceride, LPL, is barely detectable in WT liver, its expression was weakly induced in 7-only liver (Figure 4B). Hepatic LPL gene expression has been reported to be induced by cytokines (Merkel et al, 1998 and references therein), suggesting that inflammatory signals could be present in 7-only liver. Transgenic mice overexpressing LPL specifically in the liver on a LPL null background presented hepatic steatosis and increased circulating ketone body levels (Merkel et al, 1998), as is the case for the 7-only mice. Moreover, the major HDL receptor gene SR-B1 seemed induced in the 7-only livers and may participate in steatosis and in low HDL-cholesterol (Figure 4A).

Fatty acid/cholesterol synthesis and catabolism are not likely to trigger 7-only dyslipidemia. We queried whether lipid synthesis and catabolism were disturbed in 7-only mouse liver. Transcript levels of two enzymes involved in the rate-limiting steps of mitochondrial and peroxisomal -oxidation pathways (MCAD and AOX, respectively) were increased or unchanged compared to WT, both in fed and fasted mice, and those of CPTII, allowing the incorporation of long-chain fatty acids into mitochondria for -oxidation, were increased (Figure 4C). In addition, transcripts of HMG-coA-synthase, the rate-limiting enzyme in ketogenesis, were increased (Figure 4B), consistent with higher serum levels of ketone bodies. These genes are targets for a key player in lipid metabolism, PPAR (Gulick et al, 1994; Rodriguez et al, 1994; Hashimoto et al, 1999; Barrero et al, 2003), whose variation in expression levels was not correlated with the induction of some of its target genes (Figure 4B and C). In contrast, transcript level of the key enzyme for fatty acid synthesis (FAS) is decreased in the 7-only liver, and the difference between the genotypes is accentuated with aging (Figure 4C). This indicates an inverse correlation between lipid and FAS transcript levels and suggests a negative feedback loop. These observations, although fragmentary, suggest that increased FAS or defects in -oxidation/ketogenesis are not involved in the observed hepatic steatosis.

Transcript levels of HMG-coA-reductase, the enzyme responsible for the committed step in cholesterol synthesis, were not significantly affected in the 7-only livers (Figure 4B) as for those of a few genes involved in bile acid synthesis (cyp7A1, but also cyp8B1 and cyp7B1; not shown) and transport (Figure 4D; except for NTCP, which was mildly elevated). Thus, these preliminary observations did not reveal defects in cholesterol synthesis and transformation (via bile acid synthesis) to account for dyslipidemia. Further investigations are required to decipher the molecular aspects of the 7-only mouse phenotype.

Expression of some key genes of glucose and amino-acid metabolism is not affected

HNF4 is also a well-known regulator of carbohydrate and amino-acid metabolism. Transcripts of enzymes involved in these metabolic pathways (Figure 5A–C) were not altered in the 7-only livers, including ornithine transcarbamylase (OTC), a known target gene of HNF4 whose deregulation is responsible for the ureagenesis defects reported in the Hnf4 null livers (Inoue et al, 2002). Since, in the absence of HNF4, the induction of neoglucogenic genes by a fast was impaired (Rhee et al, 2003), we investigated whether PEPCK induction was altered in the fasted 7-only mouse livers. This induction was normal (Figure 5C), as expected since neoglucogenic genes are reported to be induced through coactivation with PGC1, which interacts with the AF-2 domain common to both HNF4 isoforms (Yoon et al, 2001).

CAR expression is drastically reduced in the 7-only mouse livers

Expression of cytochrome P450 genes, central in xenobiotic detoxification, is induced by several nuclear receptors, including the CAR and the pregnane-X receptor (PXR). In the absence of the HNF41 AF-1 motif, the expression profile of both factors (Figure 6A) was reminiscent of that of the liver-specific HNF4 KO (Hayhurst et al, 2001; Tirona et al, 2003). Whereas PXR was not affected, CAR transcripts in the 7-only mice were drastically reduced compared to WT (Figures 6B and 7B). This was associated with a lack of induction by the CAR-specific ligand TCPOBOP of the CAR inducible gene, cyp2b10, in the 7-only mice (Figure 6C).

Since no HNF4-binding sites have been described in the mouse CAR promoter, we scanned for sites with the MatInspector program. In the 10 kb region upstream of the ATG, eight sites were identified, three of which strongly bound HNF4 in electrophoretic mobility shift assay (EMSA) (oligos -1341, -3624 and -7598; Figure 6D). However, these sites seemed to bind 1 and 7 homodimers at equivalent levels, as expected from previous in vitro work performed with HNF41 and a deletion construct lacking the A/B domain (Sladek et al, 1999). In vivo, only the most proximal site (-1341) bound HNF4, and both 1 and 7 isoforms were found at this site, as shown in ChIP assays using 7-only versus WT and 1-only livers (Figure 7). Thus, the drastic reduction of CAR expression in the 7-only livers is likely mediated by differences other than differences in the affinity of the HNF4 isoforms for this site. We propose that the diminished capacity of HNF47 to regulate CAR gene expression is caused by differences in cofactor recruitment capacity of the 7 isoform (Torres-Padilla et al, 2002).

From a physiological perspective, loss of CAR in the mouse increases sensitivity to zoxazolamine-induced paralysis, while decreasing sensitivity to acetaminophen or cocaine-induced acute hepatic response (Wei et al, 2000; Zhang et al, 2002). Thus, it can be predicted that the 7-only mice will show interesting behavior towards some pharmaceuticals.

The 1-only mice present a slight glucose intolerance

In humans, mutations affecting HNF4 activity are known to be associated with MODY1 (Yamagata et al, 1996). In addition, mice deleted for HNF4 in -cells display a weak hyperinsulinemia and paradoxically an impaired response to a glucose tolerance test associated with an insulin secretion defect that was correlated with low transcript levels of the ATP-dependent potassium channel subunit Kir6.2 (Gupta et al, 2005). Since normal pancreatic cells express only HNF47 (Figure 1), we investigated whether the 1-only mice could also present insulin secretion defects.

While the 1-only mice did not display abnormal insulin and/or glucose levels (Figure 3A), they showed a weak but significant glucose intolerance to glucose injection (Supplementary Figure S3A). In order to distinguish between an insulin secretion defect and insulin resistance from peripheral tissues, we performed insulin sensitivity tests. The 1-only mice did not show significant variations in glycemia following insulin injection (Supplementary Figure S3B), as expected for an islet defect in the absence of peripheral resistance.

We evaluated Kir6.2 transcripts in pancreas and isolated islets of 1-only mice, as well as other genes susceptible to account for the phenotype. However, preliminary results (Supplementary Figure S3C–D) did not permit to highlight genes whose expression was clearly disturbed in 1-only islets, even if Kir6.2 transcript levels tended to be lower.

Concluding remarks

In this study, we have reported HNF4 knockin mice expressing only one type of HNF4 isoform (HNF41 or HNF47 and their splice-derived variants) under control of the two promoters, P1 and P2. The 1-only and 7-only mice do not present any evident phenotype, contrasting with the embryonic lethality due to visceral endoderm defects of the constitutive HNF4 deletion (Chen et al, 1994; Duncan et al, 1997): this indicates functional redundancy of the isoforms in the yolk sac. In addition, contrasting with the severe architecture defects reported in the fetal-liver-specific HNF4 KO (Parviz et al, 2003), 7-only fetal livers appear indistinguishable from WT (see epithelial markers in Supplementary Figure S4). In the adult, the 7-only mice exhibit a nonlethal dyslipidemia associated with slight hepatic steatosis, reminiscent, in more subtle form, of the phenotype of the liver-specific Hnf4 null mice. This indicates not only that HNF47 is sufficiently redundant functionally with HNF41—the main isoform in the WT liver—to enable long-term survival of the 7-only mice, but also that there are specificities inherent to each isoform due to the presence or absence of the AF-1 domain. Although the molecular mechanisms supporting the 7-only mouse phenotype were not fully elucidated in this study, alterations of the expression of apoAIV, apoCIII, apoB, MTP and LPL genes are likely to be involved.

To our mind, the most important finding of this study is the ability to discriminate between genes whose expression is strictly or mainly dependent on the presence of a functional AF-1 domain in the HNF4 protein (CAR, apoAIV, apoCII), and genes whose expression is independent of this motif, although dependent upon HNF4 (i.e., OTC and apoAII).

The cases of differential regulation by HNF4 isoforms imply that different cofactors are recruited in vivo on some promoters by HNF41 versus HNF47. Indeed, the AF-1 domain has been shown to interact in vitro with general factors associated with the basal RNA polymerase II machinery, elements of the mediator complex TRAP220/DRIP205 and TRAP170, and coactivators such as SRC-1, GRIP1 and CBP/p300 (Green et al, 1998; Sladek et al, 1999; Malik et al, 2002; Torres-Padilla et al, 2002). The affinity and coactivation efficiency of CBP/p300 and GRIP1 for HNF41 has been shown to be higher than for HNF47 since these cofactors interact with both AF-1 and AF-2 motifs (Eeckhoute et al, 2003). It will be a challenge to define which interactions are disturbed in hepatocytes in the absence of the AF-1 domain on the regulatory regions of the apoAIV or CAR genes, whose expression levels are most affected in 7-only livers. In addition, for apoAIV expression, we showed that the presence of a functional AF-1 domain is required in liver but apparently not in intestine, suggesting that recruitment of different cofactors by HNF4 may occur in these tissues.

It is perhaps surprising to observe enhanced expression of some HNF4 target genes in the 'loss of function' 7-only mice. This could be due to indirect physiological effects. However, it could also reflect the differential capacities of HNF4 isoforms to act as competitors for more robust activators at regulatory sites or to repress expression of target genes via recruitment of corepressors. Indeed, the gene for HMG-coA synthase was repressed by HNF41 in transfection assays, HNF41 competing with PPAR for a binding site in the promoter (Rodriguez et al, 1998). Thus, HNF47 may not compete efficiently with PPAR, which may be a more potent activator for the regulation of the HMG-coA synthase gene, leading to gene de-repression. In addition, histone deacetylases (HDAC) are constitutively associated with nuclear receptors and inhibition of HDAC activity in transfection assays for HNF4 targets has been shown to result in enhanced target gene expression (Ruse et al, 2002; Torres-Padilla et al, 2002). HNF47 activity was shown to be less potently repressed in vitro than HNF41 by the corepressor SMRT, and the 7 isoform may act as a less potent repressor than 1 on some target genes (see Briancon et al, 2004).

The mice described here will be useful to define the ensemble of the HNF4 target genes which require AF-1 for expression in the 'loss of function' tissues, as well as those that are misregulated by a 'gain of function' like the expression of HNF41 in the pancreas. Array analyses of HNF4-expressing tissues from mutant and WT mice should permit identification of the full spectrum of AF-1-sensitive target genes. This knowledge will permit 'clustering' of the targets and identification of common regulatory elements, paving the way for identification of cofactors mediating AF-1 dependency.

Materials and methods

Plasmid constructs for homologous recombination at the Hnf4 locus

To replace the coding sequence of the 7-specific exon 1D by that of the 1-specific exon 1A, a plasmid construct was prepared carrying exon 1A CDS cloned 3' of promoter sequences/5'UTR of exon 1D, and 5' of G418-resistance cassette, intron 1D sequences and DT-A cassette. The reciprocal construct was created for exon 1A replacement. Details are given in Supplementary Methods.

ES cell screening and mouse breeding

CK35 ES cells were a gift of C Kress. Transfectants were selected in 300 g/ml G418 (Calbiochem-Novabiochem Corp.). Two ES clones of each construct were microinjected into C57Bl/6 blastocysts and founder male chimeras obtained. The floxed neo cassette was deleted following mating with pgk-Cre mice (Lallemand et al, 1998), and a subsequent cross with C57Bl/6 mice eliminated the Cre transgene. Since no differences in phenotype were observed between Cre⁺ and Cre^- or mice from different ES clones, we used mice of all genotypes (not shown). PCR genotyping of mice is detailed in Supplementary Methods. Homozygous 1-only and 7-only mice were born in mendelian proportions (Supplementary Methods). WT littermates served as controls.

Northern blots and RT–PCR/-quantitative real-time PCR

See Supplementary Methods for RNA extraction and Northern blot. Conditions for RT–PCR and sequences of the primers specific for the 1 and 7 exon 1 CDS have been described (Briancon et al, 2004). Quantitative real-time PCR assays were performed with SYBR Green Master Mix (Applied Biosystems) and analyzed following the standard curve method to take into account the amplification efficiency of primers (for HNF41/7, see details in Briancon et al, 2004). All real-time PCR results are normalized to -actin (s.d.). Primer sequences can be obtained upon request.

Western blot analysis and HNF4 immunodetection on histological sections

Nuclear protein extracts were prepared and Western blots performed as described in Supplementary Methods. Membrane was probed with a HNF4 C-terminal peptide antibody (sc-6556; Santa Cruz Biotechnology), and reprobed with a TFIIB antibody (sc-225). Bound antibody was revealed by peroxidase-conjugated secondary antibody (Caltag, DakoCytomation) detected with the ECL Plus reagent (Amersham Biosciences). For immunohistochemistry, the HNF41 (N1–14) and 7-specific primary antisera were provided by Sladek et al (1999) (and unpublished data). Immune complexes were detected with 3,3'-diaminobenzidine (DakoCytomation) and sections counterstained with Mayer's hematoxylin.

Serum and bile chemistry

Blood collection and assays were performed on overnight fasted cohorts of male and female adult mice, mainly at the Mouse Clinical Institute, Illkirch, France. Details are given in Supplementary Methods.

Ligand-induced activity of CAR

5-Month-old mice were injected intraperitoneally with the CAR agonist 1,4-bis[2-(3,5-dichloropyridyloxy)]benzene (TCPOBOP) at a dose of 3 mg/kg body weight, or vehicle (5% DMSO in sunflower oil), and killed 6 h later (Wei et al, 2000).

Electrophoretic mobility shift assays

Details and oligonucleotide sequences are given in Supplementary Methods/Tables.

Chromatin immunoprecipitation

Nuclei were prepared from livers and chromatin crosslinking, sonication and immunoprecipitation were performed as described in Gresh et al (2004), except that nuclei from different animals were not pooled. After sonication, the soluble chromatin was precleared and subjected to immunoprecipitation using -HNF4 (sc-6556) or -IL-1ra (sc-8482) antibody in the presence of 1 g/ml salmon sperm DNA and 1 mg/ml BSA. Immune complexes were collected by adsorption to protein G-Sepharose (Sigma). Relative enrichments at HNF4-binding sites were determined by real-time PCR using the standard curve method and were normalized to CAR gene 3'UTR, which is devoid of HNF4-binding sites. The same enrichments were obtained relative to a nonrelevant sequence within the -actin promoter (not shown). Primer sequences are given in Supplementary Table.

Supplementary data

Supplementary data are available at The EMBO Journal Online.

Acknowledgements

We are grateful to C Kress, M-A Nicola, D Rocancourt and D Vallois for tutoring in mouse engineering and in islet isolation. We thank A Bailly for sharing unpublished data and primers for ChIP assays, and G Hayhurst, GP Navarro, H Strick-Marchand, S Tajbakhsh and FM Sladek for advice and for sharing materials, AM Catherin, C Mulet and P Keating for technical help and B Laine and M Pontoglio for helpful discussions. NB was supported by a fellowship from the Ministère de l'Enseignement Supérieur, de la Recherche et de la Technologie. The Association pour la Recherche contre le Cancer (France) provided a fellowship (NB) and funding.

References

Barrero MJ , Camarero N , Marrero PF , Haro D (2003) Control of human carnitine palmitoyltransferase II gene transcription by peroxisome proliferator-activated receptor through a partially conserved peroxisome proliferator-responsive element. Biochem J 369: 721–729  | Article | PubMed | ISI | ChemPort |

Boj SF , Parrizas M , Maestro MA , Ferrer J (2001) A transcription factor regulatory circuit in differentiated pancreatic cells. Proc Natl Acad Sci USA 98: 14481–14486  | Article | PubMed | ChemPort |

Briancon N , Bailly A , Clotman F , Jacquemin P , Lemaigre FP , Weiss MC (2004) Expression of the 7 isoform of hepatocyte nuclear factor (HNF) 4 is activated by HNF6/OC-2 and HNF1 and repressed by HNF41 in the liver. J Biol Chem 279: 33398–33408  | Article | PubMed | ISI | ChemPort |

Chabardes-Garonne D , Mejean A , Aude JC , Cheval L , Di Stefano A , Gaillard MC , Imbert-Teboul M , Wittner M , Balian C , Anthouard V , Robert C , Segurens B , Wincker P , Weissenbach J , Doucet A , Elalouf JM (2003) A panoramic view of gene expression in the human kidney. Proc Natl Acad Sci USA 100: 13710–13715  | Article | PubMed | ChemPort |

Chen WS , Manova K , Weinstein DC , Duncan SA , Plump AS , Prezioso VR , Bachvarova RF , Darnell Jr JE (1994) Disruption of the HNF-4 gene, expressed in visceral endoderm, leads to cell death in embryonic ectoderm and impaired gastrulation of mouse embryos. Genes Dev 8: 2466–2477  | PubMed | ISI | ChemPort |

Duncan SA , Manova K , Chen WS , Hoodless P , Weinstein DC , Bachvarova RF , Darnell Jr JE (1994) Expression of transcription factor HNF-4 in the extraembryonic endoderm, gut, and nephrogenic tissue of the developing mouse embryo: HNF-4 is a marker for primary endoderm in the implanting blastocyst. Proc Natl Acad Sci USA 91: 7598–7602  | PubMed | ChemPort |

Duncan SA , Nagy A , Chan W (1997) Murine gastrulation requires HNF-4 regulated gene expression in the visceral endoderm: tetraploid rescue of Hnf-4(-/-) embryos. Development 124: 279–287  | PubMed | ISI | ChemPort |

Eeckhoute J , Moerman E , Bouckenooghe T , Lukoviak B , Pattou F , Formstecher P , Kerr-Conte J , Vandewalle B , Laine B (2003) Hepatocyte nuclear factor 4 alpha isoforms originated from the P1 promoter are expressed in human pancreatic beta-cells and exhibit stronger transcriptional potentials than P2 promoter-driven isoforms. Endocrinology 144: 1686–1694  | Article | PubMed | ISI | ChemPort |

Elshourbagy NA , Boguski MS , Liao WS , Jefferson LS , Gordon JI , Taylor JM (1985) Expression of rat apolipoprotein A-IV and A-I genes: mRNA induction during development and in response to glucocorticoids and insulin. Proc Natl Acad Sci USA 82: 8242–8246  | PubMed | ChemPort |

Farese Jr RV , Ruland SL , Flynn LM , Stokowski RP , Young SG (1995) Knockout of the mouse apolipoprotein B gene results in embryonic lethality in homozygotes and protection against diet-induced hypercholesterolemia in heterozygotes. Proc Natl Acad Sci USA 92: 1774–1778  | PubMed | ChemPort |

Green VJ , Kokkotou E , Ladias JA (1998) Critical structural elements and multitarget protein interactions of the transcriptional activator AF-1 of hepatocyte nuclear factor 4. J Biol Chem 273: 29950–29957  | Article | PubMed | ISI | ChemPort |

Gresh L , Fischer E , Reimann A , Tanguy M , Garbay S , Shao X , Hiesberger T , Fiette L , Igarashi P , Yaniv M , Pontoglio M (2004) A transcriptional network in polycystic kidney disease. EMBO J 23: 1657–1668  | Article | PubMed | ISI | ChemPort |

Gulick T , Cresci S , Caira T , Moore DD , Kelly DP (1994) The peroxisome proliferator-activated receptor regulates mitochondrial fatty acid oxidative enzyme gene expression. Proc Natl Acad Sci USA 91: 11012–11016  | PubMed | ChemPort |

Gupta RK , Vatamaniuk MZ , Lee CS , Flaschen RC , Fulmer JT , Matschinsky FM , Duncan SA , Kaestner KH (2005) The MODY1 gene HNF-4alpha regulates selected genes involved in insulin secretion. J Clin Invest 115: 1006–1015  | Article | PubMed | ISI | ChemPort |

Hadzopoulou-Cladaras M , Kistanova E , Evagelopoulou C , Zeng S , Cladaras C , Ladias JA (1997) Functional domains of the nuclear receptor hepatocyte nuclear factor 4. J Biol Chem 272: 539–550  |