The Mouse Fetoprotein Transcription Factor (FTF) Gene Promoter Is Regulated by Three GATA Elements with Tandem E Box and Nkx Motifs, and FTF in Turn Activates the Hnf3 b , Hnf4 a , and Hnf1 a Gene Promoters*

Fetoprotein transcription factor (FTF) is an orphan nuclear receptor that activates the a 1 -fetoprotein gene during early liver developmental growth. Here we sought to define better the position of FTF in transcriptional cascades leading to hepatic differentiation. The mouse FTF gene was isolated and assigned to chromosome 1 band E4 (one mFTF pseudogene was also found). Exon/intron mapping shows an mFTF gene structure similar to that of its close homologue SF1, with two more N-terminal exons in the mFTF gene; exon mapping also delimits several FTF mRNA 5 * - and 3 * -splice variants. The mFTF transcription initiation site was located in adult liver at 238 nucleotides from the first translation initiator codon, with six canonical GATA, E box, and Nkx motifs clustered between 2 50/ 2 140 base pairs (bp) from the cap site; DNA/protein binding assays also pinpointed an HNF4-binding element at 1 36 bp and an FTF-binding element at 2 257 bp. Transfection assays and point mutations showed that the mFTF promoter is activated by GATA, HNF4 a , FTF, Nkx, and basic helix-loop-helix factors, with marked cooperativity between

At 8 -8.5 days of mouse embryogenesis, endodermal cells of the ventral foregut interact with the cardiac mesoderm and become committed to the hepatic differentiation program; these newly specified cells then migrate and proliferate in the mesenchyme of the septum transversum where liver morphogenesis becomes apparent at ϷE10.5 (1,2). Initial induction of hepatic functions is driven in part by growth factors of the FGF 1 family, secreted by cardiac mesodermal cells and acting via transmembrane receptor kinases at the endodermal cell surface (3). The process also involves potentiating transcription factors among which GATA factors appear essential to endodermal determination across vertebrates as well as invertebrates (Ref. 4 and references therein). Following liver specification, transcriptional activation cascades further develop among early hepatic transcription factors creating interactive regulatory networks that amplify the induction signals and imprint the liver phenotype; the HNF4␣ gene product, for example, activates the Hnf1␣ gene whose product further enhances the Hnf4␣ gene promoter, and both HNF1␣ and HNF4␣ activate a broad spectrum of liver functions (5,6).
One of the earliest events marking endodermal specification to liver function is the activation of the ␣ 1 -fetoprotein (AFP) locus, one of four albumin-related genes tandemly organized in the genome but differentially expressed during development (7). The AFP gene is the first to be activated in the foregut endoderm (8), and therefore transcription factors that transduce early cell specification signals to unfold AFP chromatin are likely to exert high ranking liver differentiation functions. One prime candidate in that regard is an orphan nuclear receptor originally pinpointed as a potent, highly specific and mandatory activator of the proximal AFP gene promoter (9 -11). The fetoprotein transcription factor (FTF) (NR5A2) (FTF designation as per Genome Data Base Nomenclature Committee (accession number 9837397); NR5A2 as per Nuclear Receptor Nomenclature Committee (12)) belongs to a subgroup of nuclear receptors related to the archetypal Drosophila segmentation gene Ftz-F1 (13) and that bind as monomers to the DNA motif PyCAAGGPyCPu (where Py is C or T and Pu is A or G) (11). FTF is widespread and tightly conserved among vertebrates (14 -17); its closest mammalian relative is SF1 (18) that is mainly expressed in steroidogenic cell lineages, whereas FTF is mainly expressed in gut derivatives.
Early developmental patterns of FTF expression in several species (14,17,19) and its activation of the AFP locus in hepatocyte progenitors suggested an important role for FTF in endodermal differentiation pathways. To gain better insight into FTF functions and its position in liver induction pathways, we isolated and functionally characterized the mouse FTF gene promoter in search for upstream FTF gene regulators, and we also searched for new FTF downstream gene targets. The combined results make a compelling case for FTF as a key intermediary between initial signals of liver specification and cascade activations of other transcription factors that enhance hepatic differentiation. cis-activating elements pin-pointed in the mFTF promoter may lead to new important effectors of endodermal differentiation pathways.

EXPERIMENTAL PROCEDURES
Isolation and Mapping of the Mouse FTF Gene-A mouse 129 SV genomic DNA library in bacteriophage DASH II (Stratagene) was screened by plaque hybridization using rat FTF cDNA (11) as a probe. The PCR method was used to screen a mouse 129 SV P1 bacteriophage library (Genome Systems) with the following primers from mouse FTF gene introns: 5Ј-AGTTGAATCTCTGCTGCCCGTGTCC-3Ј (intron 2) and 5Ј-TAGCCCGAGAGTGTAAAACCAGGAA-3Ј (intron 3). DASH II DNA from the FTF-positive clones was prepared by the plate lysate method, and the P1 clone DNA was prepared by the alkaline lysis procedure. DNA was digested with various restriction enzymes, and DNA fragments were ligated into pBluescript SK ϩ (Stratagene) and used to transform Escherichia coli DH5␣; plasmid DNA was isolated, and positive subclones were identified by Southern analysis using mouse FTF/LRH-1 (11) cDNA fragments. DNA was sequenced on both strands by the dideoxynucleotide method, and exon/intron boundaries were mapped against the mFTF/LRH1 cDNA sequence (11); mFTF genomic sequences were deposited in GenBank TM under accession number AF239709 (BankIt 322861).
The mFTF gene transcription start site was mapped by primer extension of mouse liver RNA using a 25-mer oligonucleotide (FTF-PE1) spanning mouse FTF gene sequence ϩ37 to ϩ61 (underlined in Fig. 1B). FTF-PE1 was radiolabeled (10 6 cpm/ng) with [␥-32 P]ATP and T4 polynucleotide kinase (Life Technologies, Inc.) and purified by gel electrophoresis and electroelution; 5 ϫ 10 5 cpm of the FTF-PE1 primer was annealed to 200 g of total RNA from adult mouse liver overnight at 50°C in 32 l of Superscript II reverse transcriptase buffer (Life Technologies, Inc.); the reaction was cooled to 42°C and pursued for 2 h in the presence of 10 mM dithiothreitol, 1 mM dNTPs, and 100 units of Superscript II reverse transcriptase. Reverse transcription products were subjected to RNase A digestion, phenol/chloroform extraction, and ethanol/sodium acetate precipitation and analyzed by electrophoresis on denaturing 6% polyacrylamide gels; reference DNA sequences used the FTF-PE1 primer with DNA from FTF gene plasmid 4F-CAT (below).
Chromosomal localization of the mouse FTF gene was carried out by fluorescence in situ hybridization (Genome Systems). Mouse FTF genomic DNA (P1 clone) was labeled with digoxigenin dUTP by nick translation, mixed with sheared mouse DNA, and hybridized to metaphase chromosomes from normal mouse embryo fibroblasts in 50% formamide, 10% dextran sulfate, and 2ϫ SSC. Hybridization signals were revealed with fluoresceinated anti-digoxigenin antibodies followed by counterstaining with 4,6-diamidino-2-phenylindole. Of 80 metaphases analyzed, 72 exhibited specific labeling; FTF gene assignment was further confirmed by cohybridization with a telomeric probe spe-cific to chromosome 1.
FTF Gene Constructs-FTF gene promoter activity was analyzed with FTF/CAT reporter constructs spanning 4 kb or 280 bp of contiguous DNA 5Ј-adjacent to the mouse FTF transcription start site. A HindIII-BamHI fragment from pSV0-CAT (9) was cloned into HindIII-BamHI-digested pBluescript SK ϩ to generate vector SK-CAT; the EcoRI FTF gene fragment Ϫ3.9 kb to ϩ79 bp (leftmost fragment in Fig.  1A) was isolated, blunted, and inserted in HindIII-blunted SK-CAT to generate construct 4F-CAT. To obtain pF-CAT, the 4F-CAT vector was digested with PstI to leave only FTF gene segment Ϫ184/ϩ79 upstream from CAT, and a PstI Ϫ280/Ϫ185 FTF segment was inserted at Ϫ184 in the correct orientation. Mutations in the FTF gene promoter were derived by PCRs using plasmid 4F-CAT and oligonucleotides listed in Table I; two complementary oligonucleotides (125 ng of each) overlapping the targeted region were mixed with 50 ng of 4F-CAT DNA and amplified with Pfu DNA polymerase (Stratagene) for 30 s at 95°C, 1 min at 55°C, and 2 min/kb of DNA template at 68°C (18 cycles), followed by 1 h at 37°C with 10 units of DpnI (Stratagene). The PCR products were transformed into E. coli DH5␣, and mutations were confirmed by sequencing. Mutant 4FmG123 was derived from 4FmG3.
Electrophoretic Mobility Shift Assays and Transient Cell Transfections-Bandshift assays were conducted as described before (10,11) with total nuclear proteins from adult rat liver and oligonucleotides listed in Table II, using 0.2 ng of 32 P-labeled probe, 3 g of nuclear proteins, and 5-500-fold molar excess of cold competitor oligonucleotides. For FTF binding assays, probe and competitors were coincubated with the nuclear extract for 30 min at 4°C; for HNF4␣ supershift assays, 1 l of anti-HNF4␣ serum (Santa Cruz Biotechnology) was preincubated with the nuclear extract 1 h at room temperature, and then the probe and competitors were added for an additional 30 min.

RESULTS
Genetic Mapping of the Mouse FTF Locus-We retrieved three independent clones (ZF2, ZF6, and ZF25) from the mouse genomic library screened with a rat FTF cDNA probe; clones ZF2 and ZF25 gave stronger signals on Southern blots and were further characterized. Clone ZF2 contained a contiguous sequence closely matching the mouse FTF/LRH-1 cDNA sequence (11) but with many base changes, deletions, insertions, and no introns, all indicative of a retro-pseudogene. Clone ZF25 contained FTF coding and interspersed DNA sequences corresponding to the exon 3-6 domain in Fig. 1A. Intronic oligonucleotides were used for further screening of a mouse P1 library, which yielded an insert extending 35 kb upstream from ZF25. The P1 insert revealed additional exons and a cluster of transcription factor recognition sites 5Ј-flanking exon 1 and suggestive of a promoter domain (Fig. 1B). Intron/exon boundaries mapped from or P1 clones all conformed to the GT/AG splicing rule. Mouse FTF exons 3-6 correspond to SF1 exons 1-4 (18), and two additional 5Ј FTF exons encode the longer FTF N-terminal (A/B) domain (11); the conserved exon/intron junctions of SF1 and FTF predict that the FTF 3Ј-gene domain contains three more exons (7-9 in the mouse) encoding C terminus amino acids 390 -560 (11). mFTF exons 1 and 4 -6 are similar in size to zebrafish ff1 exons 1 and 3-5 (17).
The mFTF gene transcription initiation site was identified by primer extension of adult mouse liver RNA using antisense primer ϩ61/ϩ37 (underlined in Fig. 1B); a predominant transcription start site was mapped (Fig. 1B) at a G residue 79 nt upstream from the previously reported 5Ј-end of mouse FTF cDNA and 238 nt upstream from the first translation initiation codon (11) (other extension primers downstream of ϩ79 bp yielded smears of reverse transcription products terminating around ϩ79; some mRNA structure perhaps blocked polymerase extension but alternative start sites around ϩ79 cannot be excluded).
Chromosomal assignment of the mFTF gene was carried out by fluorescence in situ hybridization with a P1-FTF genomic probe minimizing cross-hybridization with the mouse FTF pseudogene. Out of 80 metaphases analyzed, 72 showed single specific labeling of the middle portion of chromosome 1 (identified by 4,6-diamidino-2-phenylindole staining and cohybridization with a specific telomeric probe); measurements of chromosomal distances on 10 fluorescence in situ hybridization metaphases located the mouse FTF gene at chromosome 1 band E4 (Fig. 1C).
Functional Analysis of the mFTF Gene Promoter-Functional analyses were initiated with mFTF gene construct 4F-CAT carrying FTF gene segment Ϫ4 kb/ϩ79 bp. Transfection assays showed strong reporter gene activity in HepG2 and Hep3B hepatoma cells (which express endogenous FTF) (21) and no detectable activity in (FTF-negative) HeLa or F9 cells (inset, Fig. 2B). Reporter CAT activity was also higher in Hep3B than HepG2 cells, correlating with higher expression of endogenous FTF in Hep3B cells (21). These results thus indicated that mFTF gene segment Ϫ4 kb/ϩ79 bp correctly reproduced hepatic specificity and relative activities of endogenous FTF gene promoter functions. Electromobility shift assays were then conducted with presumptive HNF4 and FTF recognition sequences located at ϩ36 and Ϫ257 (Fig. 1B), slightly divergent from consensus binding sites. Radiolabeled DNA probe ϩ36/ϩ50 (oligonucleotide FH4, Table II) yielded a single retarded liver complex that was competed by cold oligo FH4 and more efficiently by an optimized HNF4-binding site (oligo DR1) ( Fig. 2A, lanes 1-7) but not at all by 100-fold excess of a mutated FH4 sequence (oligo mFH4 in Table II) (Fig. 2A, lane  10). Furthermore, the retarded complex was completely supershifted by anti-HNF4␣ antibodies ( Fig. 2A, lane 11). These results showed that FTF promoter segment ϩ36/ϩ50 is a highly specific HNF4␣ recognition site, of lower affinity than a canonical HNF4 DR1 element. Similar assays using as a probe the strong FTF-binding element of the AFP gene promoter (10, 11) (oligonucleotide FTF␣ in Table II) indicated that mFTF gene segment Ϫ254/Ϫ268 (oligo FF, Table II) can also efficiently bind FTF, with lower affinity than the FTF␣ site ( Fig.  2A, lanes 12-17).
mFTF gene activation by HNF4␣ or FTF was tested by transient transfection in Hep3B and HepG2 cells. Cotransfection of construct 4F-CAT with an HNF4␣ expression vector resulted in 5-6-fold enhancement of reporter gene activity in both cell lines (Fig. 2B, lane 3); cotransfection with an FTF expression vector also resulted in significant increase of CAT activity in Hep3B cells (Fig. 2B, lane 4) but not in HepG2 cells (perhaps because HepG2 cells express abundant SF1, which binds and activates the same DNA motif as FTF) (11,21). Three consensus GATA-binding sites (WGATAR) were conspicuous in the proximal mFTF gene domain Ϫ80/Ϫ138 (Fig. 1B), and we tested activatory effects of GATA expression vectors including the GATA6 and GATA4 variants expressed in the foregut endoderm and liver primordium (22,23). Cotransfection of GATA6 or GATA4 (and GATA1, GATA2, or GATA5 as well) raised 4F-CAT activity 2-3-fold in HepG2 or Hep3B cells (Fig. 2B, lane 2); coexpression of GATA6 or GATA4 with HNF4␣ showed additive or cooperative effects enhancing FTF promoter activity as much as 12-fold in Hep3B cells (Fig. 2B, lane 6). We also found significant activation of the 4F-CAT construct with the Nkx2.5 expression vector (Fig. 2B, lane 5), which is consistent with the recognition of a cognate high affinity (24) Nkx2.5-binding element at Ϫ99/Ϫ105 bp (Fig. 1B); coexpression of GATA4 and Nkx2.5 enhanced 4F-CAT activity only slightly more than each factor alone (not shown). No stimulation of mFTF promoter activity was detected with expression vectors encoding SP1, HNF1␣, HNF3␣, HNF3␤, HNF6, or C/EBP␣ (Fig. 2B, lanes 7-12) (no binding sequences for these factors were apparent in the mFTF promoter sequence), nor with c-Myc, USF1, or USF2 expression vectors (not shown). Generally similar results were obtained in cotransfections using the shorter FTF gene construct pF-CAT carrying only 280 bp of 5Ј-flanking DNA (Fig. 2B, lanes 13-17); this suggested a direct action of the activating factors on their cognate DNA motifs clustered around the cap site. Basal activity of pF-CAT was similar to that of 4F-CAT and provided no indication for regulatory components operating between Ϫ280 bp and Ϫ4 kb.
To prove their role in mFTF promoter activity, point mutations were introduced in the HNF4, GATA, and Nkx elements GGGTCAAAGGTCAAT in the natural 4-kb context of construct 4F-CAT; mutations were also introduced in tandem E box motifs at Ϫ66 and Ϫ58 bp. Mutation of the HNF4 sequence (FmH4 in Table I) reduced basal mFTF promoter activity in HepG2 or Hep3B cells (Fig.  3B, lane 2), and in cotransfection assays it abolished mFTF promoter activation by HNF4␣ (Fig. 3C, lanes 1 versus 4).
Mutation of the Nkx element also eliminated promoter activation by Nkx2.5 (Fig. 3C, lanes 3 versus 11). These combined data showed that mFTF promoter activation in cotransfection assays using HNF4␣, GATA, and Nkx2.5 vectors resulted from a direct action of these factors on their cognate DNA-binding elements surrounding the FTF cap site. Notably, in contrast to the proximal GATA and HNF4 mutants (mG3, mH4), the distal GATA and Nkx mutants (mG12, mNk) showed significant increase of basal promoter activity in both HepG2 and Hep3B cells (Fig. 3B, lanes 2 and 3 versus 4  and 6). Also, when the two E box motifs were mutated (mE12), strong reduction of basal FTF promoter activity was observed (Fig.  3B, lane 9), whereas a single mutation of either the distal or the Left, the mFTF mRNA start site was located by primer extension of mouse liver RNA with FTF antisense primer ϩ37/ϩ61 (sequencing ladders used the same primer with 4F-CAT plasmid DNA); the major reverse transcription product is marked by the arrow. Right, mFTF promoter sequence compared with the corresponding zebrafish FTF gene sequence (zff1) (17); dots indicate conserved nucleotides; transcription factor recognition elements are marked in bold; arrows ϩ1, mRNA initiation sites; underlined sequence: oligonucleotide used for primer extension. C, chromosomal mapping of the mFTF gene by fluorescence in situ hybridization using a mouse P1 FTF probe (middle panel, cohybridization with a chromosome 1 telomeric probe); arrows point specific signals that assign the mFTF gene to chromosome 1 band E4.
proximal E box had opposite effects on basal promoter activity (mE1 versus mE2 in Fig. 3B, lanes 7 and 8). Double mutant mE12 maintained similar relative responses to exogenous GATA4 or HNF4␣ (Fig. 3C, lanes 1 versus 9 and 2 versus 10), like GATA mutants to HNF4␣ and vice versa (Fig. 3C, lanes 1 versus 6 and 2 versus 5, and data not shown), suggesting that each factor can activate the mFTF gene promoter independently.

FTF Gene Structure and Splice
Products-This is the first molecular cloning of a mammalian FTF genomic locus. The partial mFTF gene structure established here indicates an intron/exon organization similar to that of the Sf1 gene, which was duplicated from the same ancestral gene (14). Like in man, the mouse Sf1 and FTF genes reside on different chromosomes (16,29,30). The mouse FTF gene structure is largely conserved in zebrafish (17) but with an apparent evolutionary remodeling in the exon 2-3 domain, as discussed below.
The mouse exon mapping indicates 5Ј-exon heterogeneity between FTF cDNAs cloned from different species. Clones of human FTF cDNAs uniformly lack a sequence corresponding to  Table  II) (mFH4, mutated FH4 sequence; DR1, optimized HNF4-binding site), and the binding of FTF to promoter sequence Ϫ254/Ϫ268 (oligo FF, Table  II) (FTF␣ and mFTF␣ are the wild-type and mutated FTF-binding sequence from the AFP promoter). Reactions used 3 g of rat liver nuclear proteins (arrows point the four rat FTF isoforms) (10, 11) and 5-100-fold excess (numbers above lanes) of cold competitors (C, no competitor; lane 11 shows a supershift reaction by preincubation of the extract with anti-HNF4␣ antiserum; all bandshift is removed indicating that the FH4 element is highly restricted to HNF4␣). B, transient transfection assays conducted in HepG2 and Hep3B hepatoma cells with mFTF gene reporter constructs 4F-CAT and pF-CAT cotransfected with transcription factor expression vectors (C, control cotransfections with void vector pCI). Results are averages Ϯ 1 S.D. from three sets of duplicate or triplicate transfections using two plasmid preparations and referred to pCI controls given an independent value of 1 for HepG2 or Hep3B transfections. Inset, autoradiograms are CAT assays using FTF gene construct 4F-CAT transiently transfected in the four cell lines indicated (arrows, chloramphenicol-acetylated products; Ͻ0.05%, undetectable activity). mouse exon 2 (the hFTF cDNA clone we initially reported in Ref. 16 corresponds to human FTF cDNA variant hB1F2 in Ref. 29 and CPF variant 1 in Ref. 31; we reviewed our original sequence data and found one compressed C missing at position 344 in Ref. 16, disrupting the hFTF N-terminal reading frame).
FTF also exists as C-terminal splice variants; human and Xenopus "short" variants (hFTFs and xFF1rAs) diverge to noncoding sequences at the 3Ј-splice junction of exon 6 (16,33). Other FTF cloning products include human FTF clone CPF-2 (31) truncated in exon 6 to restart with exon 7, and zebrafish ff1 clones "B" which also interrupt a C-terminal exon (17). Such transcripts devoid of the AF2 activatory domain could in principle exert dominant negative functions (11,21,33); however, no significant amounts of such products have been detected in vivo (17,33). The FTF exons 2-3 domain is further considered below.
Regulation of the mFTF Gene Promoter-This report is also the first functional characterization of an FTF gene promoter. Many cis-acting elements pinpointed here are stringently conserved in zebrafish (17); such evolutionary conservation suggests that these elements fulfill important regulatory functions. Given the broad phylogenetic role of GATA factors in endodermal specification (4), a triple GATA-binding domain also suggests tight connections of the GATA-FTF pathway with endodermal differentiation, which conceivably could extend to invertebrate FTF/Ftz-F1 homologues (to the best of our knowledge, FTF is the first liver-enriched transcription factor gene directly activated by GATA).
In cotransfection assays, GATA and HNF4␣ were highly efficient partners in coactivating the mFTF promoter, consistent with abundant coexpression of GATA4, HNF4␣, and FTF in mammalian liver. However, FTF is also abundant in other tissues such as the adult exocrine pancreas in which HNF4␣ is not significantly expressed. Furthermore, during development FTF gene expression precedes that of HNF4␣ (FTF can be unambiguously detected by in situ hybridization at E7.0 in the foregut endoderm, 2 which is consistent with AFP gene transcription beginning at least from E8.0 (8)), and HNF4␣ knockout experiments also indicate that the FTF (and AFP) genes are efficiently expressed without HNF4␣ at early stages of liver differentiation (25). Therefore the FTF promoter segment analyzed here would bypass HNF4␣ at prehepatic endodermal stages.
Point mutations selecting the downstream GATA site or its adjacent E box element resulted in lower levels of basal FTF promoter activity, indicating that in hepatoma cells the proximal FTF regulatory domain Ϫ85/Ϫ61 is subject to up-regulation by GATA Table I) introduced in the GATA (G1, G2, and G3), E box (E1 and E2), Nkx, and HNF4 promoter sites (Fig. 1B). B, mutation effects on basal mFTF promoter activity in transient transfection assays; results are averages Ϯ 1 S.D. from three sets of transfections with three plasmid preparations, referred to wild-type 4F-CAT activity given an independent value of 1 for HepG2 or Hep3B cells. C, mutation effects on the 4F-CAT response to cotransfected expression vectors; results are averages Ϯ 1 S.D. from three transfections with two preparations of plasmids, referred to void vector pCi transfected in parallel and given an independent value of 1 for HepG2 or Hep3B cells. also intriguingly similar to a composite element of the hematopoietic lineage, with the same orientation and spacing, and coactivated by a protein complex made of GATA1, bHLH factor Tal1, and several bridging molecules that include the LIM domain protein Lmo2 (34). GATA1, Tal1, and Lmo2 are each required for proper development of the erythropoietic lineage, and genes activated by the composite GATA/E box motif are predicted to play critical differentiation functions in the blood cell lineage. FTF could thus plausibly qualify as a first target gene for such a GATA/E box multiprotein activatory complex in the endodermal lineage. Furthermore, a second FTF promoter E box element is present two nucleotides downstream from the first one and appears to down-regulate FTF promoter activity (Fig. 3B, lane 8); this strikingly reproduces a variant hematopoietic GATA/E box motif in which the second E box serves to attenuate the adjacent GATA1-Tal1 multiprotein activatory complex (35). Differential mFTF promoter activity between tissues or developmental stages might then finely respond to changing combinations of activatory or repressor components forming the GATA/E boxes complex. We note here that the second E box is not conserved in the zebrafish promoter (see Fig. 1B), the usage of which is much less restricted among adult tissues (abundant FTF transcripts ff1-I are detected in ovary, testis, and muscle) (17).
The more distal GATA/Nkx mFTF promoter domain between Ϫ99 and Ϫ138 bp conveys clear inductive responses to exogenous GATA and Nkx2.5 in cotransfection assays, but mutations in these elements do not reduce but rather enhance basal FTF promoter activity in hepatoma cells. This suggests that the distal GATA/Nkx segment may be less efficiently used in hepatocytes than in other cell types or liver differentiation states. Like bHLH factors, Nk2-related homeoproteins such as Nkx2.5 broadly serve mammalian cells to execute differentiation programs; Nkx factors are also represented in hepatic cells (36). Notably, the combination of GATA4/Nkx2.5 factors that enhanced the mFTF promoter are well known to induce pleiotropic differentiation program in the cardiogenic lineage, and they cooperatively activate heart-specific gene promoters via tan-dem motifs similar to that in the distal FTF promoter (37). Interestingly, cardiogenic pathways use GATA4/Nkx2.5 as terminal transducers of differentiation cascades triggered by cellsurface TGF-␤-related receptors (38); similar TGF-␤ signals operate in endodermal specification (39), and TGF-␤ or FGF relays (3) could plausibly reach the distal FTF promoter domain and activate its preferential use in specific cells and times.
An Alternative mFTF Promoter?-FTF translational analyses in vitro and in vivo (11) have clearly established that rodent FTF exon 1 is transcribed in prenatal liver, and therefore that the promoter characterized here is used at fetal stages of development (hFTF exon1 transcripts are also present in fetal human liver) (29). But data in lower vertebrates point to a second FTF promoter downstream from exon 1 and used preferentially at early stages of development. FTF cloning from embryonic cDNA libraries has yielded transcripts with a translation start site corresponding either to the Met 62 codon of mouse exon 3 (amphibian xFF1rA and rrFTZ-F1␣) (14,40) (zebrafish ff1 variant II (17) also initiates translation at Met 62 ) or to the Met 94 codon of mouse exon 4 (frog rrFTZ-F1␤, chicken OR2.0) (15,40). Furthermore, the N-terminal exon of xFF1rA, ff1-II, and rrFTZ-F1␣ contains a 3Ј-end sequence matching mouse exon 3 segment Met 62 -Val 89 (see Fig. 1A) (17,40), but its 5Ј-side sequence becomes unrelated to mouse exon 3 (17,33,40), and it spans Ն500 nucleotides in xFF1rAs (14,33). This readily suggests that FTF gene transcription starts in an intronic domain in which the exon 2-3 structure has been evolutionarily rearranged. In addition, fish/amphibian transcripts containing exon 1 are undetectable at embryonic stages, whereas transcripts excluding exon 1 increase abundantly in early embryos and then decline (14,17). Thus, all indicate that at least in lower vertebrates FTF gene transcription proceeds from both the upstream promoter characterized here and from an internal "early" promoter yet to be mapped and independently regulated. The early promoter also seems broadly activated in fish, amphibian, or avian embryonic cells (especially brain and steroidogenic tissues), whereas the upstream pro-  Fig. 2) with rat liver nuclear extract and oligonucleotides shown in Table II (H1FTF-1 and H1FTF-2 correspond to FTF sites 1 and 2 in the Hnf1␣ promoter). Right, diagram of FTF site mutations introduced in Hnf1␣ promoter construct pH1-CAT (mutated sequences H1mF1 and/or H1mF2 are shown in Table I). B, transfection assays in Hep3B or HepG2 cells showing the response of pH1-CAT or its FTF site mutants to coexpressed transcription factors; results are averages Ϯ 1 S.D. from three transfections using two plasmid preparations, referred to parallel transfection with void vector pCI (column C) given an independent value of 1 for HepG2 or Hep3B cells (basal activities of pH1-CAT, H1mF1, H1mF2, or H1mF12 did not vary significantly). moter is less promiscuous and more active in gut derivatives (14,15,17,40). We note that no SF1 transcripts have been found in fish (41); an alternate or ancestral FTF gene promoter possibly evolved in early vertebrates to better fulfill both FTF and SF1 functions in time and space. Whether a second mFTF gene promoter may actively operate between exons 1 and 3 remains conjectural. But FTF isoforms referred to translational variants, and fluctuating between adult liver and fetal liver or embryonic stem cells (11) might also indicate modulations in both promoter and splicing products; for instance, the four rat FTF isoforms (10) (see Fig. 2A) interpreted as translational variants Met 1 , Met 62 , Met 83 , and Met 87 (11) could also be accounted for by exon variants 1:2, 1:3, 1:4, and 3:4. An alternative mFTF promoter might take charge of FTF transcription detected as early as the 2-cell stage embryo and in all cells of the E3.5 blastocyst (our unpublished results with a lacZ-marked mFTF genomic locus). 2 Novel FTF-activated Gene Targets-The Hnf3␤, Hnf4␣, and Hnf1␣ genes encode transcription factors critically involved in early liver differentiation, and the promoters of all three genes were found here to be activated by FTF. Combined with temporal patterns of expression of these four factors (19,25,42,43) and the fact that the FTF promoter, but not the Hnf3␤, Hnf4␣, and Hnf1␣ promoters, contain bona fide GATA-inducible sites, these results clearly suggest that FTF lies between GATA and HNF3␤, HNF4␣, and HNF1␣ in transcriptional cascades of liver cell differentiation. Thus, as FTF initiates hepatic functions at the AFP promoter, it would also enhance other transcription factors engaged into liver-induction networks. We recall here that AFP gene activity is restricted to early proliferative stages of liver development and that the FTF regulation domain in the AFP promoter is deactivated by hormonal signals that interrupt liver cell growth (9,11,44); this implies that the FTF function may be dependent upon or enhanced by cell growth stimuli. Although other potential FTF liver gene targets have recently been proposed (21,29,31,45,46), it is uncertain whether all FTF effects can actually proceed in nongrowing hepatocytes since most results were obtained by cotransfection in growing cultured cells. For one, the hepatitis B viral core promoter, strongly activated by three FTF-binding elements (21,29), could have evolved FTF activation sites to adapt sustained hepatitis B virus replication with proliferative stages of hepatitis. FTF action on the Hnf3␤, Hnf4␣, and Hnf1␣ promoters may then be exerted only transiently at early devel-  Table I) introduced in reporter construct pH3-CAT. Autoradiogram, FTF binding to Hnf3␤ promoter sites H3FTF-1 and H3FTF-2, shown by bandshift reactions using a radiolabeled AFP/FTF probe (FTF␣) incubated with 3 g of rat liver nuclear proteins and 20-fold molar excess of unlabeled oligonucleotides (sequences are given in Table II) (C, no competitor). Histograms, transfection assays in HepG2 cells showing the effect of FTF site mutations on basal pH3-CAT activity (lanes C) and on HNF3␤ promoter inducibility by cotransfected FTF and/or HNF3␤; results are averages Ϯ 1 S.D. from four sets of transfections using two plasmid preparations referred to pH3-CAT/pCI cotransfection (lane 1) given a value of 1. B, autoradiogram, bandshift reactions showing FTF binding to Hnf4␣ promoter element FTF-2 (oligo H4FTF-2 in Table II) (results were similar with oligo H4FTF-1); reactions were conducted as in A with 20 -500-fold molar excess of cold competitors. Histogram, transfection assays in HepG2 cells showing enhancement of pH4-luc activity by coexpressed FTF, HNF1␣, or GATA6 (no additive effects were found between these three factors); results are averages Ϯ 1 S.D. from three transfections using two plasmid preparations, referred to void vector pCI (C) given a value of 1. opmental growth stages and be substituted by alternative factors in mature nondividing hepatocytes; the Hnf3␤, Hnf4␣, and Hnf1␣ promoters all contain substitute proximal elements expected to operate efficiently under adult stage conditions. But FTF clearly plays a role in hepatocyte bile acid synthesis pathways; FTF activates the rate-limiting Cyp7A gene promoter in adult hepatocytes (47), and FTF is a direct target for CYP7A feedback repression by FTF heterodimeric partner SHP induced by bile acid receptor FXR (48 -50). FTF could thus use alternative regulatory means to modulate differentially early developmental and late metabolic liver program. It is interesting that the Hnf1␣, Hnf3␤, Hnf4␣, hepatitis B virus, and AFP promoters, but not Cyp7A, all contain a close doublet of FTFbinding sites ( Fig. 4 and 5, Ref. 21). 2 This may point to novel regulatory means for DNA-bound FTF monomers operating during cell growth; transient FTF-FTF interactions via a cell cycle-dependent FTF modification, ligand, or coregulator is an intriguing possibility.
The proximal FTF site in the HNF1␣ promoter is particularly interesting with regard to developmental growth signaling. That site is stringently conserved in the Xenopus Hnf1␣ promoter (51), which undergoes developmental enhancement (52) closely parallel to that of the Xenopus FTF gene (xFF1rA) (14). Furthermore, the FTF site in the xHnf1␣ promoter lies in a 30-bp domain essential to promoter activity (deletant 238/207 in Ref. 51), and it is contained in a short 69-bp promoter segment inducible by activin A (53), also a cell surface receptor ligand of the TGF-␤ superfamily. The 69-bp domain carries the conserved HNF4-binding site (5,51) which, however, is dispensable for activation of the 69-bp domain in endodermal cells (54). It is thus very tempting to infer that the endodermal activin A response element in the xHnf1␣ promoter is indeed FTF (xFF1rA). Given the fast kinetics of xHnf1␣ promoter induction by activin A (Ͻ6 h) (53), this may then imply that TGF-␤ and/or FGF signaling could reach both the FTF gene promoter (as discussed previously) and the FTF protein itself.
In conclusion, this combined analysis of FTF promoter regulatory elements and gene targets suggests that FTF is an early intermediary in the endodermal chain of events leading to hepatic differentiation. Our results are consistent with a simplified scheme (Fig. 6) where GATA cooperates with bHLH or homeodomain factors to activate the FTF gene in hepatocyte precursors, and the FTF gene product in turn activates early liver functions (AFP) and other transcription factors enhancing cross-regulatory networks to set the hepatic phenotype. FTF gene disruption in the mouse is embryonic lethal at ϷE7.5, 2 and it will be important to see if FTF-mutant ES cells contribute to the hepatocyte lineage of chimeric mice.