Expression of the alpha7 isoform of hepatocyte nuclear factor (HNF) 4 is activated by HNF6/OC-2 and HNF1 and repressed by HNF4alpha1 in the liver.

The hepatocyte nuclear factor (HNF) 4alpha gene possesses two promoters, proximal P1 and distal P2, whose use results in HNF4alpha1 and HNF4alpha7 transcripts, respectively. Both isoforms are expressed in the embryonic liver, whereas HNF4alpha1 is almost exclusively in the adult liver. A 516-bp fragment, encompassing a DNase I-hypersensitive site associated with P2 activity that is still retained in adult liver, contains functional HNF1 and HNF6 binding sites and confers full promoter activity in transient transfections. We demonstrate a critical role of the Onecut factors in P2 regulation using site-directed mutagenesis and embryos doubly deficient for HNF6 and OC-2 that show reduced hepatic HNF4alpha7 transcript levels. Transient transgenesis showed that a 4-kb promoter region is sufficient to drive expression of a reporter gene in the stomach, intestine, and pancreas, but not the liver, for which additional activating sequences may be required. Quantitative PCR analysis revealed that throughout liver development HNF4alpha7 transcripts are lower than those of HNF4alpha1. HNF4alpha1 represses P2 activity in transfection assays and as deduced from an increase in P2-derived transcript levels in recombinant mice in which HNF4alpha1 has been deleted and replaced by HNF4alpha7. We conclude that although HNF6/OC-2 and perhaps HNF1 activate the P2 promoter in the embryo, increasing HNF4alpha1 expression throughout development causes a switch to essentially exclusive P1 promoter activity in the adult liver.

The hepatocyte nuclear factor (HNF) 4␣ gene possesses two promoters, proximal P1 and distal P2, whose use results in HNF4␣1 and HNF4␣7 transcripts, respectively. Both isoforms are expressed in the embryonic liver, whereas HNF4␣1 is almost exclusively in the adult liver. A 516-bp fragment, encompassing a DNase I-hypersensitive site associated with P2 activity that is still retained in adult liver, contains functional HNF1 and HNF6 binding sites and confers full promoter activity in transient transfections. We demonstrate a critical role of the Onecut factors in P2 regulation using site-directed mutagenesis and embryos doubly deficient for HNF6 and OC-2 that show reduced hepatic HNF4␣7 transcript levels. Transient transgenesis showed that a 4-kb promoter region is sufficient to drive expression of a reporter gene in the stomach, intestine, and pancreas, but not the liver, for which additional activating sequences may be required. Quantitative PCR analysis revealed that throughout liver development HNF4␣7 transcripts are lower than those of HNF4␣1. HNF4␣1 represses P2 activity in transfection assays and as deduced from an increase in P2-derived transcript levels in recombinant mice in which HNF4␣1 has been deleted and replaced by HNF4␣7. We conclude that although HNF6/OC-2 and perhaps HNF1 activate the P2 promoter in the embryo, increasing HNF4␣1 expression throughout development causes a switch to essentially exclusive P1 promoter activity in the adult liver.
Hepatocyte nuclear factor (HNF) 1 4␣, an orphan member of the steroid/thyroid receptor superfamily, is highly expressed in the adult liver (1). During mouse development, it is one of the first liver-enriched transcription factors (LETF) to be expressed, HNF4␣ transcripts being detected at E (embryonic day) 4.5 in primitive endoderm, in the visceral endoderm (E 5.5), and in the liver bud (E 8.5) (2,3). The crucial role of this factor in the embryo was demonstrated by the inactivation of the HNF4␣ gene, causing perigastrulation lethality (4) because of failure to activate visceral endoderm functions (5). When null embryos were transiently rescued by tetraploid complementation with wild-type morulae-derived visceral endoderm, HNF4␣ was dispensable for hepatic specification but not for activation of hepatic functions (6). In addition, liver-specific HNF4␣ deletion in the embryo (7) led to disorganization of liver architecture (8), and its forced expression in cultured hepatic cells resulted in the reexpression of some liver-specific genes (9) and restoration of epithelial morphology (10,11). HNF4␣ plays a pivotal role, regulating expression of genes involved in nutrient metabolism and transport (for review, see Ref. 12), and its induced disruption in the adult liver provokes lethal defects in lipid and bile acid metabolism and ureagenesis (13,14). Hence, HNF4␣ is essential for both the induction and the maintenance of hepatic functions.
The HNF4␣ nuclear receptor contains six domains (A-F) and possesses two activation functions (AF): AF1 corresponds to the 24 N-terminal amino acids (15), and AF2 is located in the E domain. Different isoforms of HNF4␣ arise by alternative splicing. The most important for the adult liver appear to be ␣1, the prototype protein (1), and ␣2, which possesses a 10-amino acid insertion in the C-terminal F domain (16; for review, see Ref. 12).
The isoforms considered in this work result from alternative promoter usage. An N-terminal variant of HNF4␣ was cloned from the mouse (17). It arises from a distal upstream promoter, P2 (18), located about 40 kb upstream from the P1 promoter from which the ␣1 and ␣2 isoforms are transcribed (3,19). The HNF4␣7 protein derived from P2 usage differs from ␣1 (and its splicing variants) only by the first exon, which in the case of ␣1 but not ␣7 contains a functional AF1. The differences in activation functions of these isoforms result from differences in * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains Figs. S1-S4 and additional experimental procedures and results.
§ Recipient of a fellowship from the Ministère de l'Enseignement Supérieur, de la Recherche et de la Technologie and from the Association pour la Recherche contre le Cancer, France.
ʈ cofactor recruitment (20). In the liver, the ␣7 isoform is expressed mainly in the embryo and is nearly extinguished in the adult. In accordance with this expression profile, the transactivation capacity of the HNF4␣7 isoform is efficient for fetal functions such as ␣-fetoprotein, whereas that of ␣1 is higher than ␣7 for most hepatocyte promoters (21). The HNF4␣7 isoform is also expressed in the adult stomach, intestine, and pancreas (17). In human, mutations in the HNF4␣ coding sequence that is common to P1 and P2-derived isoforms are associated with maturity-onset diabetes of the young (22,23). Because predominantly P2 is active in the pancreas (17,24,25), mutant HNF4␣7 is probably incriminated in the insulinsecretion defects of diabetes, whereas the associated hepatic dysfunctions (26,27) can likely be ascribed to mutant ␣1 (for review, see Ref. 12).
Because the HNF4␣7 and HNF4␣1 isoforms show qualitative differences in their AF domains, it can be speculated that the regulation of expression of each is critical for ontogenesis of HNF4␣-expressing tissues. In addition, because HNF4␣7 transcripts are present in embryonic liver but are mainly extinguished in the adult where HNF4␣1 transcripts are particularly high, it is possible that the products of one promoter influence activity of the other, providing another link in the transcription factor network regulating liver-specific gene expression. Although regulation of expression of the P1 promoter and its enhancer in liver cells has been analyzed (28,29), the factors driving expression of the P2 promoter in embryonic liver and its down-regulation in the adult remain to be determined.
Here we demonstrate the existence of a single DNase I-hypersensitive site specifically associated with P2 activity. This site is present in embryonic liver and is retained in the adult even though HNF4␣7 transcripts are barely detectable in adult liver, suggesting that chromatin at the promoter region remains in an open conformation after birth. Using a combination of embryonic liver-derived cultured cells and knock-out mice, we show that two Onecut factors, HNF6 and OC-2, are essential for HNF4␣7 expression in the developing liver. HNF1␣, and to a lesser extent HNF1␤, also activate the P2 promoter. In addition, quantitative real time PCR revealed that, from E 12.5 and throughout liver development, P2-derived transcripts are always less abundant than those from P1. Finally, using knock-in mice expressing only one of the two N-terminal HNF4␣ isoforms under control of both promoters, we describe a dramatic repressive effect of HNF4␣1 on the P2 promoter, providing a novel mechanism to explain HNF4␣7 postnatal repression in the liver.

EXPERIMENTAL PROCEDURES
Library Screening for Mouse Genomic Sequence Upstream from Exon 1D-Genomic clones were obtained by screening a mouse 129/Sv BAC library (Incyte, Inc.) for exon 1D sequences. One clone called m21 was used as a template for most experiments. This clone contained at least 12 kb upstream from the P2 promoter, the P1 promoter and its enhancer region, and exon 2.
Analysis of DNase I-hypersensitive Sites-Nuclei were prepared from freshly excised organs of adult mice and liver from E 18.5 fetuses as described previously (35) or from cultured cells scraped into phosphatebuffered saline on ice. Purified nuclei were resuspended in 15 mM Tris-Cl, pH 7.5, 15 mM NaCl, 60 mM KCl, 5 mM MgCl 2 , 2 mM EDTA, 0.15 mM spermine, 0.5 mM spermidine, treated for 10 min at 0°C with 0 -10 units of DNase I (Roche Applied Science)/g of DNA. Nuclei were lysed, and DNA was purified and cleaved with the appropriate restriction enzyme prior to Southern blot analysis (36). Probes abutting the restriction cleavage sites (a 644-bp EcoRI-BamHI genomic fragment for EcoRI or BamHI and a 566-bp AlwNI-HindIII fragment for HindIII digestions) were [␣-32 P]dCTP labeled by random priming (Amersham Biosciences kit).
RNA Isolation-Total RNA was isolated from cells or tissues either by lysis in guanidinium thiocyanate followed by centrifugation through a CsCl cushion (37) or by the method of Chomczynski and Sacchi (38) using TRIzol reagent (Invitrogen). Poly(A) ϩ RNA was selected by use of PolyATtract® poly(dT) columns according to the manufacturer's instructions (Promega).
RT-PCR and Quantitative Real-time PCR-Reverse transcription (RT) was performed using SuperScript II TM Moloney murine leukemia virus reverse transcriptase and random hexamers on 5 g of DNase I-treated RNA following the manufacturer's instructions (Invitrogen). Semiquantitative PCR (Bioline Taq) was performed with 1 Ci of [␣-32 P]dCTP and quantified when necessary with ImageQuant software. Forward primers for mouse HNF4␣7 and HNF4␣1 were 5Ј-catggtcagtgtgaacgcg-3Ј and 5Ј-catggatatggccgactaca-3Ј, respectively, and a common reverse primer was used (21). Reverse and ␣7 forward primers matched rat cDNA perfectly, whereas the ␣1 forward primer contained two mismatches with the rat sequence, leading to poor amplification of HNF4␣1 in FGC4 rat cells. The 5Ј-UTR P2 primer (5Ј-tgggtacccttggtcatgg-3Ј) was used with the same reverse primer for as many as 36 cycles because it was difficult to detect traces of ␣7 transcripts in the wild-type adult liver. Primers for 28 S rRNA (21) and HNF6 (39) have been published, and OC-2 primers are as follows: 5Ј-gccacgccgctgggcaac-3Ј and 5Ј-cagctgcccggacgtggc-3Ј. Real-time PCR was performed with SYBR Green Master Mix (Applied Biosystems) using HNF4␣1 (5Ј-gaaaatgtgcaggtgttgacca-3Ј and 5Ј-agctcgaggctccgtagtgttt-3Ј), HNF4␣7 (5Ј-ggtacccttggtcatggtcagt-3Ј and 5Ј-tggatgaattgaggttggcac-3Ј), or ␤-actin primers (5Ј-gcacagcttctttgcagctcct-3Ј and 5Ј-gtacggcctcggcaacag-3Ј), on a PerkinElmer Life Sciences ABI PRISM® 7700 Sequence Detection system. To take into account the amplification efficiency of primers, we used the absolute standard curve method. Partial mouse HNF4␣1 and HNF4␣7 cDNA were amplified by PCR, cloned in pCR®2.1-TOPO® plasmid (Invitrogen) and sequenced. These plasmids were diluted in series to serve as standard dilutions on each 96-well plate. For ␤-actin, gel-purified PCR products were used as well. Data were analyzed by using Sequence Detector Systems software (Applied Biosystems). The numbers of target molecules in cDNA samples were estimated using the linear regression curve equation (R 2 Ͼ 0.99) of the logarithm of the target molecules number plotted versus Ct values (threshold) in the corresponding standard.
Reporter Constructs-The 516-bp and 4.1-kb luc constructs were obtained by subcloning of a Ϫ516/Ϫ13 bp and a Ϫ4096/Ϫ13 bp restriction fragment (relative to the ATG) into XhoI-HindIII (blunted) and NheI-HindIII (blunted) sites upstream from the pGL2-Basic (Promega) promoterless luciferase gene. Mutations in HNF1/6 sites were performed using a QuikChange site-directed mutagenesis kit (Stratagene). All promoter constructs were verified by sequencing. The apoB-luc reporter has been described previously (40).
Transient Transfections-MMH, 3TDM-1, and NIH3T3 cells (1.7 ϫ 10 5 cells/6-cm dish) were transfected using 3.3 g of reporter construct and 1.7 g of internal control RSV-lacZ (pMJM20) by the calcium phosphate procedure as described previously (41) unless specified. For MMH cells, DNA precipitates were added in Ham's F-12 medium containing 10% fetal calf serum, epidermal growth factor, insulin-like growth factor II, and insulin. Cells were harvested about 64 h later. FGC4 cells were transfected by electroporation and ␤TC1 cells by the calcium phosphate procedure or electroporation using 10 g of reporter construct and 5 g of internal control RSV-lacZ as described previously (28). Cotransfection experiments of NIH3T3 and MMH cells were per-formed using pCB6tagHNF6␣, 3 pXJ42-OC-2 (42), RSV-vHNF1 (43), RSV-HNF1 (44), and pMT7-rHNF4␣1 (45) as indicated in the figures, appropriately completed as needed with the corresponding empty vectors. Luciferase activity was determined as described previously (46) and ␤-galactosidase activity by the standard colorimetric method. Luciferase activity was normalized to ␤-galactosidase activity for transfection efficiency and calculated relative to that of pGL2-Basic unless specified otherwise. Values are the means of three to ten independent experiments (ϮS.D.).
Electrophoretic Mobility Shift Assays (EMSA)-Nuclear extract preparations and EMSA were performed as detailed previously (28). Sequences of the ␣7H1 and ␣7H6 oligonucleotides were 5Ј-gtgactggttactgtttaacgtatc-3Ј and 5Ј-ccacccaccttgggtgattagaagaatcaataagataaccgggcg-3Ј. Bases in bold were replaced in ␣7H1mut with GT and GG (47) and T and G in ␣7H6mut oligonucleotide. For competitions, a 50ϫ molar excess of unlabeled oligonucleotides was used. The antibody raised against the HNF6 peptide GTAREPNPSVTG in rabbit (UC-6), the antibody raised against the OC-2 amino acids 36 -311 in rat (UC-1), and the HNF3␣ antibody (Santa Cruz) were used at dilutions of 1/10 and 1/3.
Transient Transgenesis-The 4.096-kb sequence upstream from the HNF4␣7 ATG was subcloned in front of an nls-lacZ cassette (pSKT-nls-lacZ plasmid, a kind gift from S. Tajbakhsh, Institut Pasteur, Paris) taking care to keep the entire ␣7 5Ј-UTR intact. The HindIII-NotI restriction fragment (ϳ7.4 kb) containing the construct was microinjected into zygote pronuclei (Centre d'Ingenierie Genetique Murine, Institut Pasteur). Embryos were harvested at E 17.5, fixed for 1 h in 4% paraformaldehyde following a sagittal section, and X-gal stained at 30°C overnight. Genotyping was performed by duplex PCR with lacZspecific primers and primers specific for an endogenous gene (a kind gift from J. Barra, Institut Pasteur). Results are representative of three independently obtained transgenic embryos, all sharing identical staining profiles.
Recombinant Mice-OC-2 knock-out mice were obtained by deleting the OC-2 first exon by homologous recombination in embryonic stem cells. Double knock-out HNF6Ϫ/Ϫ OC-2Ϫ/Ϫ E 12.5 embryos were harvested at the expected proportions by crossing double heterozygous mice. 4 HNF4␣ knock-in mice were obtained by replacement of the ␣1 first exon coding sequence (1A) by exon 1D coding sequence for "␣7only" mice and vice versa for "␣1-only mice" using homologous recombination in embryonic stem cells. In all cases, the entire 5Ј-UTR and Kozak consensus were conserved. Both types of homozygous mice are viable and fertile. 5 Terminology-Because up to nine isoforms of HNF4␣ have been described or proposed (for review, see Ref. 12), each time that HNF4␣7 is mentioned, this includes not only ␣7 but also the constitutively produced splicing-derived isoforms, ␣8 and ␣9. Likewise, HNF4␣1 is meant to encompass at least the ␣2-␣4 isoforms.

RESULTS
A Single Hypersensitive Site Is Associated with P2 Activity and Retained in the Adult Liver-We began by defining the transcription initiation site of the murine HNF4␣7 gene because the 5Ј-end of the cDNA published sequence (17) did not match either the sequence of our BAC clone or that of a mouse chromosome 2 genomic contig. To verify in addition that the first exon of the HNF4␣7 gene, exon 1D, is not preceded by an additional exon in hepatic cells, we performed both primer extension and RNase protection assays (see figures in the supplemental material; Refs. 83, 84). The major start site was situated at Ϫ103 bp relative to the ATG.
DNase I-hypersensitive site analysis was undertaken to identify genomic regions implicated in hepatic regulation of the HNF4␣7 gene in the vicinity of exon 1D. The experiments were performed with fetal liver and cultured cells expressing ␣7 (Fig.  1A) and with tissues that have been described (17,19,21) to express only traces of ␣7 (adult liver) or none at all (kidney, spleen). Using the E-B probe (Fig. 1B), which hybridized to an 8.7-kb EcoRI genomic fragment spanning exon 1D, we identi-fied at least seven sites. However, only one was present exclusively in ␣7-expressing samples, including the ␤TC1 insulinoma cell line (Fig. 1A), and absent from kidney and spleen. This site was positioned ϳ300 bp upstream from the ATG, within the promoter region. Strikingly, this site was also strongly digested in adult liver chromatin, suggesting that the reduction in HNF4␣7 transcript amounts compared with fetal liver is not the consequence of a dramatic change in chromatin at the promoter. The signal corresponding to this site was less pronounced in fetal liver than in MMH-pal cells or adult liver, probably because of dilution by hematopoietic cells that do not express the HNF4␣ gene (48).
Positions of the hypersensitive sites were verified on HindIII-digested DNA, using the probe abutting the 3Ј-HindIII site (Fig. 1B), permitting analysis of the same genomic region but from the other direction. The specific site was observed again in fetal and adult liver (data not shown). We also searched for hypersensitive sites upstream from the 5Ј-EcoRI Left, hypersensitive sites detected in all tissues and MMH-pal cells are designated with lines. Only one site (black triangle) is present in both tissues and cells that express HNF4␣7 transcripts and is absent from spleen and kidney, which do not express this isoform. Bottom, the presence or absence of HNF4␣7 transcripts in tissues/cell lines. Gray triangle, undegraded EcoRI-EcoRI band. *, nonspecific band. B, estimated positions of ubiquitous (white ovals) DNase I-hypersensitive sites and of the site associated with HNF4␣7 expression (black oval) near the transcription initiation site (arrow). The boxes below represent the positions of hybridization probes used with EcoRI or BamHI (hatched) or HindIII (white) digestions. C, a Blast alignment of the 4.1-kb region upstream from the mouse HNF4␣7 ATG (Ϫ4096 bp/ϩ721 bp) was performed against the human genome. Two elements are conserved within the corresponding region on the homologous human chromosome 20 (gray and black bars, 86 and 80% identity, respectively). B, BamHI; E, EcoRI; H, HindIII. site using BamHI and the E-B probe, but only ubiquitous sites were found (data not shown and Fig. 1B).
As a complement to this analysis, we performed a Blast search against the human genome (chromosome 20) of the sequence upstream from the murine HNF4␣7 ATG to pinpoint conserved regions that could represent regulatory elements. In the 4.1-kb upstream region, two homology stretches were indeed found, at Ϫ520/ϩ54 bp (containing exon 1D and the specific hypersensitive site shown previously) and a smaller stretch at Ϫ3523/Ϫ3388 bp relative to the ATG (Fig. 1C). The same alignment was performed against the rat genome, and the two conserved elements were found within longer homology stretches (not shown).
In all cell lines expressing the endogenous HNF4␣7 gene, activity levels of the 4.1-kb and 516-bp constructs were close to (MMH-ep) or superior to that of the ubiquitous SV40 promoter used as a control (Fig. 2B). This was in contrast to NIH3T3 fibroblasts for which reporter activity levels represented only about 5% of the SV40 promoter. This suggests that cell typespecific transcription factors are necessary for reporter activity of the HNF4␣7 constructs.
We expected to observe only weak promoter activity of the HNF4␣7 constructs in FGC4 cells because these cells mimic an adult hepatocyte phenotype, including high expression of HNF4␣1 and only very low HNF4␣7 transcript amounts. We initially anticipated that transfection tests would reproduce the extinction of HNF4␣7 expression characteristic of adult liver, but this turned out not to be the case because these constructs showed high reporter activity levels in FGC4 cells (see "Experimental Procedures").
The activity levels of both HNF4␣7 constructs were similar in all cells, suggesting that the distal homology stretch (Fig.  1C) is not implicated in HNF4␣7 expression in the transfected cells and that the 516-bp element is sufficient to drive consistent cell type-specific expression. When isolated and tested by transfection (Fig. S2), this distal homology stretch showed at best only a weak enhancer effect.
Functional HNF1 and HNF6 Sites Activate the P2 Promoter-The 516-bp fragment was analyzed for consensus binding sites. Others have localized Pdx1 and HNF1 sites implicated in HNF4␣7 expression in pancreatic cells (18,24). As shown in Fig. 3A, there is also a perfect HNF6 consensus that we found to be strictly conserved in the human sequence (GenBank accession number AF509467).
EMSA was employed to verify that the relevant transcription factors bind to the HNF1 and HNF6 sites in hepatic cells. Fig.  3B shows that both HNF1␣ and HNF1␤ bind to the HNF1 site. In adult liver, adult-type FGC4 hepatoma cells and in the differentiated embryo-derived MMH-ep cells, HNF1␣ is the major form, whereas the undifferentiated MMH-pal cells express mainly HNF1␤ (Fig. 3B). In line with these observations, HNF1␤ is the first isoform to be expressed at the onset of liver development, whereas HNF1␣ rapidly becomes predominant in hepatocytes (49,50). The specific complexes were abolished by competition with a well known HNF1-binding oligonucleotide but not a mutant oligonucleotide. Fig. 3C provides a similar analysis of the HNF6 site. MMHpal, FGC4 cells, and adult liver nuclear extracts all showed binding to the putative HNF6 site (␣7H6 oligonucleotide), and this was effectively competed with an excess of the ␣7H6 but not the mutant oligonucleotide (not shown), which does not bind any complex. It was observed consistently that the MMHpal band migrated slightly more slowly than that of FGC4 or adult liver extract. This could indicate that a different but related protein is bound to the HNF6 site in MMH-pal extracts. There are three different members of the Onecut family: HNF6, present as two splicing-derived isoforms, ␣ and ␤ (51, 52), as well as OC-2 (42) and OC-3 (53). In the mouse, the OC-2 protein is larger than HNF6␣ (39) and it is known that both OC-2 and HNF6 are expressed very early in the developing liver (51,54,55), a stage that is mimicked by MMH-pal cells. Fig. 3D shows that MMH-pal cells do not harbor detectable HNF6 transcripts, but they do express OC-2. In addition, Fig. 3E reveals that the protein bound to the ␣7H6 oligonucleotide in MMH-pal cells is displaced by anti-OC-2, whereas that in adult liver reacts mainly with anti-HNF6. In MMH-pal cells, attenuation of the complex was also observed with anti-HNF6, probably because of antibody cross-reaction. These results demonstrate that the HNF6 site is bound by HNF6 and by OC-2.
Analysis of 5Ј-deletion mutants of the 516-bp fragment transfected in hepatic MMH-pal, MMH-ep, FGC4 cells, and fibroblastic NIH3T3 cells (Figs. S3 and 4A) revealed that when the region upstream from 221 bp was deleted, all cell types showed a similar reduction in reporter activity, implying that ubiquitous factors are involved in transcriptional activity of this region. However, when a 56-bp fragment containing the HNF1 site was deleted, the majority of tissue-specific promoter activity was lost. This is in agreement with previous studies in INS-1 insulinoma cells and FTO-2B hepatoma cells (18). Indeed, the 165-bp construct retained very low reporter activity level, indicating that the HNF6 site, when separated from upstream sequences, is not sufficient to confer significant promoter activity. This suggests that Onecut factors interact with yet to be identified factors binding within these upstream sequences. No hepato-specific complex in addition to HNF1 and Onecut (see above) could be detected by EMSA using oligonucleotides spanning the Ϫ450/Ϫ122 bp region (Fig. S4).
Site-directed mutagenesis in the context of the Ϫ516 and Ϫ221 promoter fragments was used to test the functionality of the HNF1 and HNF6 sites in hepatic cells. In MMH-pal and MMH-ep cells, mutation of the HNF6 site resulted in a more dramatic loss of activity than did that of the HNF1 site, indicating that in models of embryonic hepatic cells the HNF6 site plays a predominant role (Fig. 4A). In these cells, disruption of the HNF1 site showed a very weak effect, which was surprising compared with results of 5Ј-deletions. In contrast, in the adult hepatocyte model (FGC4), mutation of the HNF1 site destroyed the majority of activity, whereas that of the HNF6 site had a smaller effect. In all hepatic lines, the simultaneous mutation of both sites compromised promoter activity more severely than single mutations. These mutations had no effect upon promoter activity in NIH3T3 cells, but they abolished transactivation by HNF1␤ and HNF6 (data not shown).
The effects of combinations of transactivators upon promoter activity in fibroblasts are depicted in Fig. 4B. It can be seen that HNF1␣ is a more potent transactivator than HNF1␤ and that HNF6 is more robust than OC-2. When used in combination, HNF1␣ with HNF6 produced an additive enhancement, whereas HNF1␤ with HNF6 gave activity similar to that of HNF6 alone. The effect of OC-2 was not additive with that of HNF1 family members, although together with HNF1␣ the effect was higher than with either alone. We conclude that both sites are indeed functional in the context of the HNF4␣7 promoter and that members of the HNF1 and HNF6 families are active, with the strongest contribution being observed for HNF6 and HNF1␣.
In Vivo Role of HNF6/OC-2 in HNF4␣7 Expression in the Developing Liver-Targeted mutation of the HNF6 gene has been described (56), and that of the OC-2 gene has been obtained recently. 6 E 12.5 livers of both types of mutants as well as of the double mutant HNF6Ϫ/Ϫ OC-2Ϫ/Ϫ (Fig. 5) were examined by quantitative real-time RT-PCR for HNF4␣ transcripts originating from the P1 or P2 promoter. In no case were the transcript levels of HNF4␣1 affected, but the HNF4␣7 transcripts were reduced by two-thirds in the double mutant, without being affected in the single mutants. This reduction in HNF4␣7 transcripts amount was also observed at E 13.5 and 15.5 (not shown). This attests that the Onecut family indeed plays a central role in P2 promoter regulation in vivo and demonstrates redundancy between HNF6 and OC-2 in activation of the HNF4␣7 gene.
Four Kilobase Pairs of the HNF4␣7 Promoter Region Is Sufficient to Drive Expression of a Reporter Gene in the Stomach, Intestine, and Pancreas, but Not the Liver, of Transgenic Embryos-To test whether the region containing the two conserved elements (Fig. 1C) is sufficient to ensure faithful expres-6 F. Clotman, manuscript in preparation. sion of HNF4␣7 in vivo, a fragment containing the 4.1-kb upstream from the ATG was fused to an nls-lacZ reporter, injected into mouse zygotes, and embryos harvested at E 17.5 were stained for ␤-galactosidase activity. Fig. 6A shows the stained viscera of a transgenic (left) and wild-type (right) embryo. No staining was obtained in tissues outside of the abdominal cavity (data not shown). Intense staining of the intestinal loops and pancreas of the transgenic embryo is clearly visible, as is the total absence of staining in the liver, whereas in the nontransgenic embryo background staining is present in the intestine. However, because the bacterial ␤-galactosidase activity is targeted to the nucleus, the transgene-directed staining has a punctate rather than homogeneous appearance, making it possible to distinguish transgenic staining from endogenous background (Fig. 6B, b). Fig. 6B shows higher magnifications of the stained tissues (above) compared with the wild-type control (below). In addition to intestine (b), the stomach (a) shows punctate staining throughout the glandular re-gion, and the pancreas (c) presents intensely stained clusters of cells that correspond to the ducts (acini were not stained on cryosections; data not shown).
In no case was staining of the liver observed, either in toto or on liver cryosections, and gallbladder, kidney, and spleen were also negative (not shown). At first sight these results imply that the 4.1-kb fragment contains the elements sufficient to direct expression to the stomach, the intestine, and the pancreatic ducts, but not to the liver. Alternatively, because quantitative analysis of HNF4␣7 expression levels in these tissues has never been reported, it could be imagined that expression levels in the liver are below the sensitivity level of the staining. To address this question, real-time RT-PCR was employed to determine transcript levels of the two HNF4␣ isoforms in embryonic and adult mice.
Expression Levels of P2 and P1 Promoter-directed Transcripts during Liver Development and in Different Tissues-To determine levels of hepatic HNF4␣1 and HNF4␣7 transcripts during liver development, quantitative real-time RT-PCR was employed. In addition, in E 17.5 embryos, transcripts corresponding to the two isoforms were also evaluated for each of the X-gal-stained tissues of Fig. 6. Fig. 7A shows the results obtained for hepatic transcripts corresponding to the two isoforms, normalized to ␤-actin. HNF4␣1 transcripts are present at E 12.5 and increase thereafter in an exponential fashion nearly until birth, reaching a maximum in adult liver at a level that is 40-fold greater than that at E 12.5. Hepatic HNF4␣7 transcripts increase with time during development (3-fold between E 12.5 and E 17.5) followed by a decline to very low values in the adult. HNF4␣7 transcript levels are always lower than those of HNF4␣1. This contradicts previous RT-PCR results, but the latter had been obtained with rat primers for which the ␣1 sequence contains mismatches with the mouse sequence leading to underestima- tion of HNF4␣1 transcripts (21). Indeed, the ratio of HNF4␣1 to HNF4␣7 transcripts is never lower than 10 and increases to about 1,000 in the adult (Fig. 7B). Fig. 7C shows a comparison of HNF4␣7 transcript levels in intestine, stomach, and pancreas, relative to the liver. Among all of the ␣7-expressing tissues, liver is the lowest, but only by a factor of 3 compared with the highest (pancreas). In addition, the ratio of ␣1 to ␣7 transcripts was calculated, revealing that liver and intestine show similar ratios corresponding to similarly fairly abundant amounts of total HNF4␣ transcripts, whereas stomach contains low transcript levels, all derived from P2. The pancreatic levels are intermediate. Among all of the HNF4␣-expressing tissues, only stomach contains exclusively ␣7 transcripts, and only kidney exclusively presents ␣1 transcripts.
The expression patterns of the reporter gene in the transgenic embryos indicate that only a small fraction of the cells in stomach, pancreas, and intestine, which show moderate to strong expression of the reporter gene, are positive for ␤-galactosidase staining. In contrast, it would be expected that in the liver the transcripts would be expressed by all hepatocytes, which occupy at this stage about 60% of the tissue (57). Taking together this anticipated widespread expression in the liver and the fact that the expression levels are always weak, it is possible that the expression level per cell would not be within the limits of detection, although we do not think that this is the case. However, taking together the promoter analysis, the results from HNF6/OC-2 double knock-out embryos, and quantitation of P2 expression levels, we favor the interpretation that the tissue-specific factors important for expression in the liver have indeed been identified and that additional elements necessary for maximal expression remain to be identified. In conclusion, with the exception of the liver, the staining of the tissues is a faithful reflection of the HNF4␣7 expression pattern in E 17.5 embryos.
HNF4␣1 Represses P2 Promoter Activity by a Mechanism That Is Not Dependent upon HNF4 Binding Sites-Given the high levels of HNF4␣1 expression in postnatal liver, we investigated whether this isoform could participate to the postnatal repression of HNF4␣7. We show here that HNF4␣1 inhibits P2 promoter activity in transfection assays. Indeed, using concentrations of HNF4␣1 expression vector demonstrated to activate the apoB reporter (Fig. 8B), we found a 2.2-3.7-fold repression of P2 promoter activity, using either the 516-or 221-bp promoter in MMH-ep cells (Fig. 8A). Under these conditions, the effect should not be a reflection of squelching (58,59). ChIP assays have shown binding of HNF4␣ protein at the P2 promoter in adult hepatocytes (60), but no consensus binding site was detected in this region (Fig. 3A). Thus, to define the site through which the repression operates, two approaches were used. First, the Ϫ516 bp/Ϫ122 bp region was scanned for HNF4␣ binding activity, using a series of 44 -47-bp-long oligonucleotides overlapping by 22-35 bp, in the presence of liver nuclear extracts with or without anti-HNF4␣ antiserum, as well as with extracts from transfected COS-7 cells expressing an HNF4␣1 expression vector. No bands were observed that could be displaced with the antibody (Fig. S4), implying that the region is devoid of sites that bind HNF4␣ in our EMSA conditions, even with weak affinity.
Second, the 516-bp promoter fragment harboring or not mutations of the HNF1 and/or HNF6 sites were transfected into MMH-ep cells in the presence or absence of HNF4␣1 expression vector, to test whether repression could act via indirect binding of HNF4␣1 at these sites, acting like a cofactor through FIG. 7. HNF4␣7 transcripts are not predominant during hepatic development. A, evolution of HNF4␣1 and ␣7 transcript amounts during liver development as determined by quantitative realtime PCR assays using the standard curve method. Data are normalized relative to ␤-actin as an internal control and are the means of at least 3 Ϯ S.D. independent experiments. Note that values on the ordinate are represented with a logarithmic scale. d.p.c., day postcoitum. B, developmental evolution of the ratio of HNF4␣1 and ␣7 transcripts in the liver. P6, postnatal day 6. C, amounts of HNF4␣7 transcripts at E 17.5 in the intestine (duodenum and upper part of the small intestine), the whole stomach, pancreas, and kidney relative to liver after normalization to ␤-actin. Kidney was used as a negative control for HNF4␣7 expression. Bottom, relative amounts of HNF4␣1/␣7 transcripts in the same embryonic tissues. The stomach does not express detectable HNF4␣1 transcripts.
FIG. 8. HNF4␣1 acts as a repressor on the HNF4␣7 promoter in transfection assays. A, MMH-ep cells were cotransfected with 1.7 g of 516-bp, 221-bp, and 165-bp luciferase constructs and pMT7-rHNF4␣1 vector completed to 300 ng with empty pMT7 vector. B, in parallel, apoB-luc reporter plasmid, which is known to be activated by HNF4␣1 (20), was cotransfected in the same conditions as in A. C, same as in A on the 516-bp construct mutated at the HNF1 and HNF6 sites. Activities are given relative to those in the absence of the HNF4␣1 expression vector. another LETF. The mutated constructs continued to be repressed by HNF4␣1, suggesting that it acts via a yet to be identified factor bound to the P2 promoter (Fig. 8C).
The HNF4␣7 P2 Promoter Is Repressed in Vivo by HNF4␣1-To determine the physiological roles of the HNF4␣ isoforms originating from different promoters, we have prepared recombinant mice that express only one isoform, but under control of the two cognate promoters (␣1-only and ␣7only mice). 5 To achieve this, it was necessary to replace the first exon coding sequence of one isoform with the first exon coding sequence of the other, without modifying the sequence transcribed from the latter promoter. To monitor transcriptional activity at the two promoters, a forward primer in each 5Ј-UTR sequence had to be used. Fig. 9A presents a scheme of these mutations.
RNA was prepared from adult liver of wild-type, homozygous ␣7-only, and homozygous ␣1-only mice. Semiquantitative radioactive RT-PCR that measures only transcription from P2 was carried out. Fig. 9B demonstrates that when HNF4␣1 is replaced by HNF4␣7, transcription from the P2 promoter is markedly higher than in livers that do express HNF4␣1 (wildtype and ␣1-only). Although there is variability in P2 transcripts from one mouse to another, there is no overlap between ␣7-only and the two other genotypes. These results strongly suggest that the P2 promoter is actively repressed by HNF4␣1 in the adult liver. DISCUSSION The use of alternative promoters is a prevalent mechanism (around 9% of genes in mice and up to 18% in humans (61)) employed for the differential regulation of genes with complex spatio-temporal expression patterns. In some cases, like that of HNF4␣, the resulting protein is modified by alternative promoter usage, producing an additional level of complexity. Moreover, HNF4␣ is not an exception among nuclear receptors because alternative promoters have been reported for the peroxisome proliferator-activated receptor ␥ (62) and the retinoic acid receptor genes (63). Furthermore, cross-and autoregulatory loops are a commonly admitted feature among transcription factors implicated in ontogenesis and maintenance of differentiated cell status and in particular in the liver via the LETF (64 -68). Here we have analyzed regulation of the distal promoter P2 of the HNF4␣ gene which gives rise to HNF4␣7 transcripts. In the pancreas, the P2 promoter appears to be regulated by the homeoproteins Pdx1 and HNF1␣ (18,24), and an important role for the HNF1 binding site within this promoter is suggested by the cosegregation of maturity-onset diabetes of the young in a Czech family with a mutation in this site (69). Here we describe that the Onecut factors HNF6/OC-2 and the HNF1 homeoproteins are involved in activation of the HNF4␣7 gene in the developing liver and that the HNF4␣1 isoform can mediate repression of P2, as depicted in the model for regulation of both the P1 and P2 promoters in Fig. 9C.
To detect regulatory cis-elements in the vicinity of exon 1D, we performed DNase I-hypersensitive site analysis. This led us to focus on the proximal promoter region because the only site found to be associated with P2 activity was revealed in this region. Ex vivo analysis of the P2 promoter required cell lines that could be used as faithful models of early liver development stages and expressing the transcription factors suspected to be important for P2 activity, which was the case for the nontransformed bipotential MMH-pal cells and their epithelial committed progeny MMH-ep. In transfection assays, a highly conserved 516-bp promoting element encompassing the hypersensitive site conferred high reporter activity. EMSA and site-directed mutagenesis analysis identified within this element two functional binding sites for LETF, HNF1 and HNF6.
We showed here that both HNF1␣ and HNF1␤ bind to the HNF1 site, yet only HNF1␣ was a strong transactivator of the P2 fragment, contrary to previous results (18). However, mutagenesis of the site only weakly affected reporter activity in MMH cells expressing HNF1␤ exclusively or low levels of HNF1␣. This argues against a critical role of the HNF1 binding site in these models. However, we do not exclude a role of HNF1 factors in vivo on P2 activity. Given that HNF1␤ in the embryo is expressed at the onset of liver ontogenesis and becomes almost exclusively restricted to biliary cells by E 14 (70,71), we suggest that if this factor played a role upon regulation of the P2 promoter, it should be mainly restricted to early stages of development. In line with a lack of such a role, HNF4␣7 expression is not affected either in HNF1␤Ϫ/Ϫ embryonic stem cell-derived embryoid bodies (72) or in HNF6Ϫ/Ϫ embryonic liver in which HNF1␤ is significantly down-regulated (70). In FIG. 9. Absence of HNF4␣1 in vivo results in an increase in HNF4␣7 transcripts in the adult liver. A, scheme of the HNF4␣ knock-in alleles. Boxes represent exons (coding sequences), and arrows mark major transcription start sites. The forward primer used in B was designed in the P2 5Ј-UTR (one-branch arrow), which is unchanged in the ␣1-only allele compared with the ␣7-only and wild-type (wt) alleles. B, semiquantitative RT-PCR assays showing an increase in P2-derived transcripts in knock-in mice expressing the ␣7 isoform under control of the P1 and P2 HNF4␣ promoters (␣7-only homozygous mice) compared with wild-type and knock-in mice expressing the ␣1 isoform under both promoters (␣1-only homozygous mice). 28 S rRNA was amplified as an internal control. Numbers below are Storm quantification analyses using ImageQuant software of P2-derived transcript amounts normalized to 28  contrast with HNF1␤, the more potent transactivator HNF1␣ is highly expressed in hepatocytes from E 14.5 onward (49), and mutation of its binding site within the promoter construct provoked a strong decrease in reporter activity in FGC4 cells, which express high levels of HNF1␣. In addition, ChIP assays have shown that in adult hepatocytes HNF1␣ is bound to promoter P2 ((60) but see also contradictory results in (24)) in accordance with our observation of retention of a prominent hypersensitive site in adult liver. Thus, we conclude that the HNF1 site may be functional in liver development via binding of HNF1␣ and probably not HNF1␤, at least after E 14.5 when HNF1␣ is highly expressed in the liver. In the adult liver, the low HNF4␣7 transcript levels could result from the stimulatory effect of HNF1␣ being strongly attenuated by the effect of repressor factors (see below).
The HNF6 consensus site in the P2 promoter can bind HNF6 and OC-2, both of which are expressed from the onset of liver ontogenesis (51,54,55). HNF6 is then retained in hepatocytes during development, although it becomes highly enriched in biliary cells (70). The HNF6 binding site is transactivated by both factors, and its mutation caused a decrease in P2 activity in hepatic cell lines expressing either OC-2 or HNF6. In addition to this ex vivo argument, the essential role of the Onecut factors was demonstrated in vivo by the use of HNF6Ϫ/Ϫ OC-2Ϫ/Ϫ double knock-out E 12.5 embryos where a drastic reduction in HNF4␣7 transcripts in the liver was observed. Furthermore, because neither of the single deficiencies showed decreased ␣7 expression, we conclude that the Onecut factors act redundantly. The residual ␣7 expression may be attributed to HNF1 factors because the HNF1␣ transcript amount is not affected in double mutant livers. 6 Although the model depicted in Fig. 9C shows that some of the same factors appear to regulate expression of P1 and P2, clear differences in promoter response have been observed. In contrast to HNF4␣7, we observed no differences in the HNF4␣1 transcript levels in any of the Onecut knock-out embryos, even though HNF6 has been shown to stimulate the P1 promoter in vitro (29,51). In addition, the only HNF4␣ transcripts affected in HNF1␤ Ϫ/Ϫ embryonic stem cell-derived embryoid bodies are from P1 and not P2 (72). A synergistic induction of the P1 promoter has been reported in transactivation assays for HNF1␣ and HNF6 (29), whereas in the case of P2 the effect is additive. Taken together, these observations imply that the P1 and the P2 promoters of the HNF4␣ gene are regulated by distinct mechanisms during liver development, either by different combinations of factors within the LETF network and/or by the same factors using different activation modes.
We then asked whether the identified elements were sufficient to reproduce in vivo the expression pattern of HNF4␣7. A 4.1-kb fragment upstream from the HNF4␣7 ATG was shown to be sufficient to drive nls-lacZ expression in E 17.5 embryos in the intestine, pancreas, and stomach, tissues that express the endogenous ␣7 transcripts at this stage. With the exception of the liver, the transgene expression profile is consistent with the RT-PCR results. Expression of the transgene in the intestine could result from stimulation of the P2 promoter by Cdx2 (73), for which putative binding sites were found in silico, and by HNF1␣, which have been described to act together in intestinespecific gene transcription (74,75). In the pancreas, the candidate regulators HNF1␤ and HNF6 are both expressed in duct cells and/or their likely precursors, HNF6 acting upstream from HNF1␤ in this tissue (51,55,56,71,76). The lack of transgene expression in the fetal liver is likely the result of a lack of cis-elements that could be essential to visualize transgene activity in this tissue, where the HNF4␣7 expression level is low. Nevertheless, the ensemble of data argues that a critical regulatory region and relevant binding sites have been identified for this tissue. Still, we cannot exclude the possibility that the ␣7 transcript levels per hepatocyte are below the detection threshold of the transgene product. The quantitative results obtained from real-time PCR showed in addition that, contrary to previous results (21), in the liver these transcripts are always less abundant than ␣1 transcripts from E 12.5 onward.
No information has been available on mechanisms involved in repression of ␣7 expression in the adult liver. Here we provide in vitro and in vivo evidence that HNF4␣1 is implicated in repression of P2 activity. We showed that overexpression of HNF4␣1 in ␣7-expressing hepatic cells leads to significant down-regulation of a reporter gene driven by the P2 promoter, at concentrations that activate a control reporter construct. This repression appears to be independent of a direct binding of HNF4␣ to DNA within the proximal promoter region because we were not able to identify a complex by EMSA, although we know from ChIP assays (60) that HNF4␣ is bound to the P2 promoter in adult hepatocytes. This suggests an indirect binding of HNF4␣1 via interaction with another factor. We tested whether an interaction between HNF4␣1 and HNF1 and/or Onecut factors could mediate the repression using P2 constructs mutated for these binding sites. Indeed, a functional interaction of HNF4␣1 with HNF1␣ (77,85) and other proteins such as ACBP, the repressor SHP, or transforming growth factor-␤ downstream effector Smad3 has been demonstrated (78 -80). However, HNF1␣ and HNF6 do not seem to be involved because the repression was not substantially relieved by mutation of their binding sites. Hence, the sites and intervening protagonists implicated in the repression mechanism remain to be identified.
The repression by HNF4␣1 was reproduced in recombinant mice engineered to express in all HNF4␣-positive tissues only one kind of isoform under control of both the P1 and P2 promoters. Thus, in mice in which ␣1 is replaced by ␣7 (so that no ␣1 is produced at all), we found a large increase of transcript levels expressed from P2 in the adult liver, strongly suggesting that ␣1 protein is also a repressor of the P2 promoter in vivo. These results led us to propose that the presence of a functional AF1 in HNF4␣1 compared with the AF1 lacking ␣7 isoform is implicated in the repressive effect. In transfection assays, HNF4␣1 is able to mediate a stronger repression than ␣7, via recruitment of SMRT, which may be caused by an increased ability of the SMRT⅐HNF4␣1 complex to recruit histone deacetylase activity (20). Taking together the results of P2 promoter activity in the ␣7-only mice and the quantitative PCR analyses of HNF4␣1 and ␣7 expression in the liver at different stages of development, the HNF4␣1-mediated repression of the P2 promoter may begin when the level of HNF4␣1 transcripts exceeds a critical threshold and is essentially total in the adult liver where the HNF4␣1 levels are very high. After birth, this repressor effect would chronically override the stimulatory effects mediated by the Onecut and HNF1 factors. This active mechanism, strictly HNF4␣1 concentration-dependent, is entirely consistent with retention of the P2 hypersensitive site in adult liver and with results of ChIP analysis implicating binding of HNF4␣1 to the P2 promoter in hepatocytes (60).