The Fetoprotein Transcription Factor ( FTF ) Gene Is Essential to Embryogenesis and Cholesterol Homeostasis and Is Regulated by a DR4 Element*

The fetoprotein transcription factor ( FTF ) gene was inactivated in the mouse, with a lacZ gene inserted in-frame into exon 4. LacZ staining of FTF (cid:1) / (cid:2) embryos shows that the mFTF gene is activated at initial stages of zygotic transcription. FTF gene activity is ubiquitous at the morula and blastocyst stages and then follows expression patterns indicative of multiple FTF functions in fetal development. FTF (cid:2) / (cid:2) embryos die at E6.5–7.5, with features typical of visceral endoderm dysfunction. Adult FTF (cid:1) / (cid:2) mice are hypocholesterolemic, and express liver FTF at about 40% of the normal level. Overexpression of liver FTF in transgenic mice indicates in vivo that FTF is an activator of CYP7A1. However, CYP7A1 expression is increased in FTF (cid:1) / (cid:2) liver. Gene expression profiles indicate that higher CYP7A1 expression is caused by attenuated liver cell stress signaling. Diet experiments support a model where FTF is quenched both by activated c-Jun, and by SHP as a stronger feedback mechanism to repress CYP7A1. quantitated in triplicate of or 20 (cid:3) g of nuclear proteins against standard parallel with (cid:3) g of liver nuclear proteins from groups. Transfection Assays— Transfection assays were conducted with FTF reporter vector 4F-CAT, which carries mFTF gene segment (cid:3) 3.9 kb to (cid:4) 79 bp cloned into pBluescript SK-CAT Point mutations were introduced in the FTF DR4-like element using PCR (11) and oligonu- cleotide primer GAAACTGGAGACATcaTTTAgAGCAttTCATACATGC-TGGAAAAAGTGC; mutations (lower case letters) were confirmed by sequencing. Transient transfections followed a calcium phosphate pro-cedure described before 11), and were conducted with HepG2 or Hep3B human hepatoma cells. Cells were cultured in DMEM containing 10% charcoal-stripped fetal bovine serum (Wisent) (DMEM/FBS). Cells 6 , DMEM/FBS) were plated in and transfected 24 h later with (cid:3) g 4F-CAT transcription factor expression (pSG5-hRXR , pCMX-hLXR (cid:1) , pCMX-T (cid:2) (pSG5 or vectors and (cid:3) g pCMV- internal cells were h and or -retinoic 22( (10 (cid:3) M or (cid:1) ,5-triiodothyronine (100 (cid:3) M NruI site of SK (cid:4) -derived vector MT5 (cid:1) /3 (cid:1) (30), placing the HA-FTF sequence under the control of the mouse metallothionein I ( MTI ) promoter. The MTI-HA-FTF recombinant vector, linearized with SalI, was microinjected into fertilized mouse eggs from C57BL/6J (cid:2) C3H F1 mice (B6C3F1), followed by transfer to pseudopregnant B6C3F1 mice. Eight transgenic lines were characterized for basal and zinc-induced HA-FTF levels in adult liver nuclei. Transgenic line TgF35 cosegregated 3 copies of the HA-FTF transgene, nearly silent in basal conditions, and was derived to homozygosity. Zinc injection in adult TgF35 mice induces dose-dependent increase of liver HA-FTF protein level, reaching a maximum within 3 h and homogeneously distributed in all hepatocytes. For the present study, we used 2-month-old male TgF35 and wild-type B6C3F1 mice; they received a single intraperito- neal injection of saline or 60 mg/kg ZnSO 4 , and liver RNA was prepared after 6 h.

Development of the mammalian embryo relies upon nutritive functions fulfilled by the visceral endoderm and then by the liver (1). A part of these functions is accomplished by nutrient carrier proteins of the albumin gene family, a multigene locus expressed by the liver and subject to precise developmental controls. One albumin-related gene, the ␣ 1 -fetoprotein (AFP) 1 gene, is acti-vated at the onset of liver differentiation and operates tightly coupled with liver growth (2,3). In 1988, our group circumscribed a proximal AFP promoter element essential to AFP gene activity in hepatocytes, and distinct from promoter components regulating the other albumin loci (4). The AFP-specific activator was then identified as orphan receptor fetoprotein transcription factor (5-7), so named for its first identified target locus (genome data base nomenclature, 2 NR5A2 in the nuclear receptor nomenclature, Ref. 9); also referred to as LRH1 or CPF). FTF belonged to a primitive class of nuclear receptors and emerged as a critical lead to connect AFP gene activation with early embryonic growth and differentiation processes.
Subsequent studies indicated that developmental FTF functions even preceded its activation of the AFP locus in hepatocytes. In situ hybridization analysis in the mouse at embryonic day 8 -9 showed abundant FTF transcripts in the foregut endoderm, before liver morphogenesis (10). Characterization of the FTF gene promoter also revealed a cluster of regulatory motifs conserved in distant species and potential targets of cell lineage specification factors (11). Among these were three proximal binding sites for GATA factors, known to be essential for visceral endoderm function (12,13). Furthermore, three HNF genes important to liver differentiation, HNF1␣, HNF4␣, and HNF3␤, were found to each contain double FTF-binding sites in their proximal promoter and to be activated by FTF in transfection assays (11). Thus, a pivotal role was suggested for FTF in a transcriptional cascade using determination factors to activate FTF in prehepatic endodermal cells, and then using FTF to drive AFP and other effectors of the hepatic program.
Besides functions in early development, FTF was expected to maintain important nutritive roles in the adult, FTF being maintained at high levels in liver, intestine, and acinar pancreas. A landmark finding was the identification of FTF as one factor that binds and regulates the CYP7A1 gene promoter, the rate-limiting enzyme in the conversion of cholesterol to bile acids (BA) (14 -17). A model emerged whereby FTF would act as a "competence" factor driving the BA pathway, subject to negative feedback regulation by interaction of FTF with orphan receptor SHP, activated by the BA receptor FXR and by FTF itself (15,16,18,19). Additional reports associated further FTF action with membrane transporters of the enterohepatic BA recycling system (20,21) and with pancreatic cholesterol esterase (22), all of which pointing to an important systemic role for FTF in adult cholesterol/BA homeostasis.
Here we document an FTF-lacZ gene knockout in the mouse, delineating FTF expression sites at all developmental stages and confirming crucial developmental and metabolic FTF functions from early embryogenesis to adulthood. We also identify a promoter DR4 regulatory element directly connecting FTF gene regulation with hormonal and metabolic homeostasis.

EXPERIMENTAL PROCEDURES
FTF Gene Inactivation-A 7.5-kb SpeI mFTF gene fragment encompassing exons 4 -6 was retrieved from a mouse 129/SV genomic library screened with rat FTF cDNA (11). The 7.5-kb fragment was used to generate a knockout vector carrying a lacZ gene inserted in-frame into mFTF exon 4. Briefly, a 4-kb lacZ fragment was derived from vector pCH110 (Amersham Biosciences) using PstI, S1 nuclease, and KpnI, and spliced into pSVK3 (Invitrogen) pretreated with EcoRI, Klenow, and KpnI. Recombinant pSVK3-lacZ was digested with SmaI/XbaI and ligated with the 5Ј-end mFTF gene fragment obtained with BglII (which cuts exon 4 at position 462 in the sequence shown in Ref. 6) and treated with Klenow. This placed lacZ (with a stop codon and polyadenylation signal at its end) in phase with the N-terminal portion of mFTF exon 4, interrupting the FTF sequence upstream from the first zinc finger of the DNA binding domain (11). An MC1-neo cassette was ligated to the 3Ј-BglII mFTF fragment and inserted downstream from lacZ in the opposite polarity, and an MC1-tk cassette was inserted at the 5Ј-SpeI end of the FTF gene fragment. The construct was then used for electroporation into ES cells. This knockout vector predicted that an FTF-lacZ allele would transcribe a 4-kb FTF-lacZ mRNA in all FTF-permissive cells (i.e. from either the mFTF promoter 5Ј-flanking exon 1, or any other putative promoter elements between exons 1 and 3) (11), and that FTF-lacZ translation products would carry 19, 40, or 101 residues of the mFTF N-terminal domain (from three in-frame initiation codons in exons 1 and 3), with no DNA binding capacity and no other apparent regulatory function (6).
For gene targeting in embryonic stem (ES) cells, mouse WW6 (129SV) ES cells (23) were grown on STO fibroblasts in high glucose Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal bovine serum and recombinant LIF (24). ES cells were electroporated with FTF-lacZ plasmid DNA linearized with NotI, and cultured on neomycin-resistant STO fibroblasts for 14 days in selection medium containing 200 g/ml G418 and 2 M ganciclovir. Resistant colonies were genotyped by Southern blot using a mFTF probe flanking the SpeI site in intron 3 and detecting a 3.8-kb StuI fragment from the wild-type FTF allele and a 7.8-kb fragment from the mutant allele.
Four ES clones carrying the lacZ-FTF allele were injected into mouse MF1 blastocysts which were then transferred to pseudopregnant C57BL/6J ϫ CBA mice. The chimeric male progeny (129SV ϫ MF1) was crossed with MF1 females to establish germ line transmission, and subsequent breeding was maintained in the 129SV ϫ MF1 genetic background. Mice were housed in filter-topped microisolator cages, under 50 -70 air changes per hour and 10:14 h dark/light cycle, and fed on Teklad standard chow TD2018. Neonatal mice were genotyped using tail DNA, by Southern blot as above or by PCR with neo gene primers 5Ј-GGGATCGGCCATTGAACAAGATGG-3Ј and 5Ј-CGTAAAGCACGA-GGAAGCGGTCAG-3Ј. Embryos from E9.5 onward were genotyped by Southern blot using yolk sac DNA. At E7.5 and E8.5, whole-embryo DNA was used with PCR primers from FTF exon 4 (5Ј-flanking the BglII lacZ insertion site) (ATGGTGAATTACTCCTATGATGAAG) and FTF exon 5 (GTAGGGACATCGTTTTCTCTGCC), generating a 1.7-kb product only from a wild-type FTF allele; the FTF primers also amplified a 204-bp segment from the FTF pseudogene (11), common to all genotypes and providing an internal PCR control. At E5.5 and E6.5, we used in situ hybridization (below) to distinguish FTFϪ/Ϫ embryos from FTFϩ/ϩ or FTFϩ/Ϫ embryos.
In Situ Hybridization-Conditions were adapted from Aubin et al. (25) and Rausa et al. (26). FTF riboprobes were generated from a 1.2-kb ClaI/HindIII mFTF cDNA fragment (nt 661-1856 in Ref. 6) cloned in pBluescript SKϩ: the sense probe was obtained with HindIII-digested DNA and T7 RNA polymerase and the antisense probe with ClaIdigested DNA and T3 RNA polymerase, with radiolabeling using [␣-33 P]UTP. HNF3␤ probes were obtained as before (10). Embryos were fixed 15-20 h in 4% paraformaldehyde/PBS, dehydrated in ethanol and toluene, and embedded in paraffin. Whole mount sections (8 m) were treated with xylene, ethanol and 4% paraformaldehyde, washed with PBS, and then treated for 7.5 min with proteinase K (20 g/ml in 50 mM Tris, pH 8.0, 5 mM EDTA), washed with PBS and treated again with 4% paraformaldehyde and PBS. This was followed by a 10-min incubation in 0.25% acetic anhydride in 0.1 M triethanolamine, pH 8.0, washes in PBS, saline, and water, and dehydration in ethanol gradients (30 -100%). Slides were then overlaid with 150 l of 1:1 mix of formamide (Fisher Super Pure) and hybridization buffer (20 mM Tris, pH 7.5, 1.2 M NaCl, 4 mM EDTA, 2ϫ Denhardt's, 1 mg/ml yeast tRNA, 200 g/ml salmon sperm DNA) and incubated for 2 h at room temperature. Hybridization was conducted with 5 ϫ 10 6 cpm of riboprobe overlaid on tissue sections in 75 l of a 2:2:1 mix of hybridization buffer:formamide: 50% dextran sulfate and overnight incubation at 65°C with a coverslip. Slides were then washed twice at 50°C with 5ϫ SSC and 10 mM dithiothreitol, once at 65°C with 2ϫ SSC, 50% formamide, and 20 mM dithiothreitol, followed by RNase A treatment (50 g/ml), washes at 65°C in 2ϫ SSC and 50% formamide then at room temperature in 2ϫ SSC and 0.1ϫ SSC, and dehydration in ethanol gradients in 0.3 M NaOAc, pH 5.5. Slides were embedded in a Kodak emulsion and left 2-4 weeks in a dark dessicator at 4°C. They were developed with the Kodak D19 reagent and stained with hematoxylin-eosin.
LacZ Activity-Expression from the FTF/lacZ transgene was monitored by ␤-galactosidase staining on whole embryos or cryostat tissue sections, essentially as described by Larochelle et al. (27). Embryos were fixed 5-30 min at 4°C with 0.2% glutaraldehyde and 4% paraformaldehyde in PBS, washed three times with PBS, and stained 15-20 h at 30°C in 4 mM K 4 Fe(CN) 6 ⅐3H 2 O, 4 mM K 3 Fe(CN) 6 , 2 mM MgCl 2 , and 1 mg/ml X-gal in PBS. Morulae and blastocysts were treated with 0.5% Pronase 10 min at 37°C before fixation. For tissue sections, organs were incubated overnight in 30% sucrose in PBS and embedded in OCT before freezing.
Microarray Analysis-Liver gene expression profiles were established in 2-4-month-old male FTFϩ/Ϫ and FTFϩ/ϩ mice (groups of 4 -6 littermates from 4 or 5 litters) fed a normal diet or a diet supplemented with cholesterol or cholic acid (below). Total liver RNA was purified using the RNeasy Mini Kit (Qiagen, Valencia, CA), quantitated by spectroscopy and pooled (equal amounts of RNA from each mouse). RNA labeling, hybridization, and image analysis were performed at the McGill University Genome Quebec Innovation Center, using the Affymetrix U74Av2 and 430A GeneChip arrays and GENECHIP data processing software (Affymetrix, Santa Clara, CA). Fold changes in mRNA levels were calculated for fluorescence intensities greater than 250.
Northern Blots-Total liver RNA was purified with guanidine thiocyanate (28), and poly(A) ϩ RNA was isolated with the Oligotex mRNA Maxi Kit (Qiagen). Poly(A) ϩ RNA (15 g) was run on a 1% agarose/10 mM guanidine thiocyanate gel, transferred onto a positively charged nylon membrane (Roche Applied Science), and hybridized with FTF, SHP, CYP7A1, and GAPDH cDNA probes PCR-labeled with digoxigenin-11-dUTP (Roche Applied Science). mRNA bands were detected with a chemiluminescence kit (Roche Applied Science) and quantitated with IMAGEQUANT (Molecular Dynamics).
Western Blots-Western blot assays were used to determine the relative amounts of FTF, HNF1␣, HNF4␣, and C/EBP␣ proteins in liver nuclei from rats fed a high-cholesterol diet or treated with thyroid hormone (T 3 ) (below). Total liver nuclear proteins (pooled from 5 rats) were resolved by SDS-PAGE and transferred onto Immobilon-P membranes (Millipore). Blotted membranes were blocked in TBS (40 mM Tris-HCl, pH 8.0, 0.2 M NaCl, 5% w/v nonfat dry milk, 0.1% Tween-20) 1 h at room temperature under agitation, and incubated 1 h with antisera against mouse FTF (1:1000 dilution in TBS), rat C/EBP␣ (1:10,000), human HNF4␣ (1:10,000) (Santa Cruz 6556x), or mouse HNF1␣ (1:1000) (BD Biosciences H69220). Membranes were washed four times in TBS and incubated for 1 h with a 1:10,000 TBS dilution of secondary antibody linked to horseradish peroxidase (Jackson Immu-noResearch Laboratories). After four more washes in TBS, reactions were developed using the Amersham Biosciences Enhanced Chemiluminescence System, and signals were quantitated using the NIH Image software. Samples from high cholesterol and T 3 groups were analyzed in triplicate sets of 10 or 20 g of nuclear proteins against standard curves run in parallel with 2.5-25 g of liver nuclear proteins from control groups.
Diets and Hormone Treatment-Serum cholesterol levels and liver gene expression profiles were analyzed in adult male FTFϩ/ϩ and FTFϩ/Ϫ mice (5 or more per group) fed for 5 days on 2% cholesterol diet Teklad TD86295 against control diet Teklad 7001, or fed for 5 days on 1% cholic acid diet Teklad TD00629 against control diet Teklad TD94045. Cholesterol was measured on a Beckman Coulter LX-20 analyzer with reagent 467825; group values were compared by Student's two-tailed unpaired t test. Changes in liver FTF levels were assessed in male Fisher rats (75-100 g, 5 per group) fed for 5 days on 2% cholesterol TD86295 diet versus control M/R7001 diet, and in male Sprague/Dawley rats (300 g, 5 per group, fed on regular Teklad chow TD97350) 5 or 24 h following intraperitoneal injection of 375 g/kg triiodothyronine (Sigma) or saline vehicle.
FTF Transgenic Mice-To assess FTF effect on CYP7A1 expression in vivo, we used a transgenic mouse line expressing FTF under the control of an inducible promoter. This system will be documented in detail elsewhere. Briefly, mFTF cDNA was ligated at its 5Ј-end with a hemagglutinin (HA) epitope sequence, and the HA-FTF cDNA was cloned in the NruI site of SK ϩ -derived vector MT5Ј/3Ј (30), placing the HA-FTF sequence under the control of the mouse metallothionein I (MTI) promoter. The MTI-HA-FTF recombinant vector, linearized with SalI, was microinjected into fertilized mouse eggs from C57BL/6J ϫ C3H F1 mice (B6C3F1), followed by transfer to pseudopregnant B6C3F1 mice. Eight transgenic lines were characterized for basal and zinc-induced HA-FTF levels in adult liver nuclei. Transgenic line TgF35 cosegregated 3 copies of the HA-FTF transgene, nearly silent in basal conditions, and was derived to homozygosity. Zinc injection in adult TgF35 mice induces dose-dependent increase of liver HA-FTF protein level, reaching a maximum within 3 h and homogeneously distributed in all hepatocytes. For the present study, we used 2-month-old male TgF35 and wild-type B6C3F1 mice; they received a single intraperitoneal injection of saline or 60 mg/kg ZnSO 4 , and liver RNA was prepared after 6 h.
FTF Expression Sites Throughout Development-Mice expressing lacZ inserted at the FTF locus allowed us to combine ␤-galactosidase staining and in situ hybridization to delineate FTF expression sites at all developmental stages. For ␤-galactosidase analysis, FTFϩ/Ϫ males were mated with FTFϩ/ϩ females. At 2.5 dpc (morula stage), FTFϩ/Ϫ embryos displayed homogeneous ␤-galactosidase staining, and at 3.5 dpc (blastocyst stage) all inner cell mass and trophectoderm cells were positive (Fig. 2, A and B) (lacZ activity was also easily detected at the 4-cell stage) (not shown). In the egg cylinder (E5.5), FTF expression segregated in the visceral endoderm (embryonic and extra-embryonic); ectodermal cells were uniformly negative (Fig. 2, G and H). At gastrulation (E6.5, E7.5), strong FTF signals persisted in the endodermal layers, and now also marked the node giving rise to the mesoderm layer and definitive endoderm (Fig. 2, K and L, and O and P). At mid-gestation (E8-E15), FTF gene activity expanded to endodermal derivatives (yolk sac, liver, pancreas, intestine), with strong transient ␤-galactosidase staining recorded in extra-endodermal structures, growing bone, testis, lung mesenchyme, sensory ganglions, posterior brain, and thalamus (Fig. 2, C and D and data not shown). At postnatal stages, FTF gene activity was detected in all hepatocytes and pancreas acinar cells (not shown), and it persisted in the intestine (prominently in the crypts of Lieberkuhn) (Fig. 2E) and in the granulosa layer of ovarian follicles (Fig. 2F).
Phenotype of FTFϪ/Ϫ Mice-Crosses among FTFϩ/Ϫ mice yielded no homozygous FTFϪ/Ϫ newborns, indicating an embryonic lethal phenotype. A total of 186 embryos were genotyped between E5.5 and E12.5 (Table I). At E5.5 and E6.5, FTFϪ/Ϫ embryos were recovered at the expected mendelian ratio of 25%, which then declined abruptly between E7.5 and E9.5, to 0% at E10.5, indicating fetal death prior to liver differentiation. At E5.5, FTFϪ/Ϫ embryos showed no histological abnormalities (Fig. 2, G and H, and I and J). At E6.5 however, they showed conspicuous anomalies, with a disorganized ectoderm and the amniotic cavity absent or malformed (Fig. 2, K  and L, and M and N). At E7.5, FTFϪ/Ϫ embryos appeared as cell clumps typical of uterine resorption; a few embryos had recognizable structures with a constricted extra-embryonic junction and no mesodermal cells outside the junction (Fig. 2, O and P, and Q and R); at E8.5, no structures were identifiable. The constriction and mesoderm clustering at E7.5 was typical of a defective node organizer or primitive line (31). E6.5 embryos were further analyzed for expression of HNF3␤, a marker of the node and visceral endoderm (32,33). FTFϪ/Ϫ embryos expressed HNF3␤ in the visceral endoderm but not in the anterior portion of the primitive line (Fig. 2, S and T, and U and V): this suggests that the node does not form in FTF knockout embryos, and consequently they cannot pass gastrulation.
Phenotype of FTFϩ/Ϫ Mice-Heterozygous FTFϩ/Ϫ mice had no apparent defect, their growth and reproduction were normal, they had no visible changes in liver histology, and their plasma levels of AFP, glucose, triglycerides, bilirubin, hepatic enzymes, and amylase were normal (data not shown). However, in each of four pilot experiments comparing adult FTFϩ/ϩ with FTFϩ/Ϫ littermates (5-8 per group), fasting level of total plasma cholesterol was consistently 14 -17% lower in the FTFϩ/Ϫ group (data not shown). Further measurements were made in larger groups of 2-month-old male littermates fed a 2% cholesterol diet or a control diet for 5 days, and confirmed a 10-12% lower serum cholesterol level in FTFϩ/Ϫ mice (Table  II). Scatchard assays on several pools of adult FTFϩ/Ϫ versus FTFϩ/ϩ livers indicated little changes in FTF dissociation constants (K d values in Fig. 3), but they suggested that FTF in the haploid group could be lower than 50% of normal (y-axis intercepts y 0 in Fig. 3) (HNF1 and HNF4 appeared unchanged in EMSAs) (Fig. 3). Further qPCR assays on three pools of livers from FTFϩ/ϩ and FTFϩ/Ϫ littermates indicated an average 94 FTF mRNA copies per cell (assuming 20 pg of RNA per cell) in FTFϩ/ϩ liver, and an average 35 copies per cell (37% of normal) in FTFϩ/Ϫ livers (data not shown). These initial observations thus indicated a role for FTF in adult cholesterol homeostasis, and suggested FTF-dependent feedback on FTF gene activity.
To gain further insight into hepatic changes incurred in the FTFϩ/Ϫ state, liver microarray analyses were performed. Gene   FIG. 1. FTF gene inactivation. A, the upper map locates the 7.5-kb mFTF gene fragment used to generate a recombination vector disrupting mFTF exon IV with lacZ-neo inserted at a BglII restriction site, and the resulting mutant FTF allele. Open arrows, direction of transcription. Intronic probe A, used in Southern blots, detects a 3.8-kb StuI restriction fragment from a wild-type FTF allele and a 7.8-kb fragment from a recombinant FTF-lacZ allele. PCR primers 1-2 and 3-4 (arrowheads) amplify a 1.7-kb fragment from the mFTF gene, a 204-bp product from the mFTF pseudogene (11), and a 700-bp neo gene fragment. B, Southern blot analysis of genomic DNA from embryonic stem cell clones obtained by transfection with the FTF targeting vector. C, Southern blot genotyping of littermates born to an FTFϩ/Ϫ mouse. D, embryonic genotyping by PCR analysis, an amplification product from the neo gene signals one or two mutant FTF alleles, and the lack of a 1.7-kb FTF gene product identifies an FTF-null embryo. profiling using the Affymetrix U74Av2 GeneChip array displayed reduced expression (4.3-fold) of FTF target gene mrp3 (20,34), but a strong increase in CYP7A1 mRNA levels (8.4fold) and a slightly increased SHP mRNA level (array 1 in Table III). Northern blot analysis confirmed the CYP7A1 and SHP results (Fig. 4). Gene profiling was repeated twice using the Affymetrix 430A GeneChip array: although not as pro-nounced, again mrp3 mRNA level was lower, and CYP7A1 mRNA higher in FTFϩ/Ϫ liver (arrays 2 and 3 in Table III). These results seemed to conflict with the postulated activating role of FTF on CYP7A1 gene activity. To clarify the point, we used transgenic mice (TgF35) carrying an FTF transgene controlled by the inducible MTI promoter. qPCR analysis of liver mRNAs was performed 6 h following zinc or saline injection to TgF35 transgenic versus B6C3F1 wild-type mice. Control transgenic mice showed some increase in total FTF mRNA level compared with wild-type controls, with slightly increased CYP7A1 mRNA, nearly 5-fold higher SHP mRNA, and no change in HNF1␣ or HNF4␣ mRNAs (Fig. 5A). Upon zinc injection, total FTF mRNA was raised 5-fold in transgenic liver relative to wild-type liver, and CYP7A1 and SHP mRNA levels were over 6-and 18-fold higher in the transgenic group (Fig.  5A). Northern blot quantitation gave comparable results, and also showed lower amounts of endogenous FTF mRNA following induction of HA-FTF transgene expression by zinc injection (Fig. 5B). These results supported the activating role of FTF on CYP7A1 gene activity and thus pointed to indirect mechanisms for higher CYP7A1 levels in FTFϩ/Ϫ mice. Gene expression profiles further revealed striking pleiotropic changes in FTFϩ/Ϫ liver, indicative of broadly attenuated cell stress signaling pathways, including reduced levels of c-Jun (arrays 1-3 in Table III). Apart from CYP7A1, mrp3, and cell signaling markers, only a few other genes showed consistent changes of expression in FTFϩ/Ϫ liver (arrays 1-3 in Table III). Liver gene expression profiles were then analyzed in FTFϩ/Ϫ mice fed on a diet containing 2% cholesterol versus a control diet (array Chol in Table III) or on a 1% cholic acid diet versus a control diet (array CA in Table III). As expected, cholesterol feeding reduced the level of mRNAs involved in cholesterol synthesis, and cholic acid increased SHP and decreased CYP7A1 and CYP8B1 mRNA levels (Table III). Notably, the hepatic content of c-Jun mRNA was markedly enhanced by either cholesterol or cholic acid feeding (Table III). CYP7A1, SHP, and c-Jun results were confirmed by qPCR analysis (not shown).
DR4 Regulatory Element in the FTF Gene Promoter-The FTF gene promoter contains a sequence closely matching a direct AGGTCA repeat with a 4-nt spacer (DR4 motif), at ϩ6 in the mFTF promoter and conserved from fish to man (Fig. 6A). This DR4-like sequence was characterized for its potential regulation by LXR-RXR and TR-RXR heterodimeric receptors. First, electromobility shift assays were conducted with an FTF DR4 probe and rat liver nuclear extracts. Two retarded complexes were specifically competed by low molar excess of cold DR4 oligos from the FTF, CYP7A1, or ABC1 gene promoters (Fig. 6B, lanes 2-4), and not by a mutated FTF DR4 oligo (lane 5). Antibodies against RXR attenuated both complexes (lanes 8 and 12), anti-LXR attenuated the slower complex (lanes 7 and  14), anti-TR attenuated the faster complex (lanes 9 and 13), and anti-FXR had no effect (lane 10). These results indicated FIG. 3. EMSA analyses on liver nuclear proteins from FTF؉/؉ and FTF؉/؊ mice. Livers were pooled from 5 adult mice in each group, and equal amounts of nuclear proteins were run in EMSA reactions with increasing amounts of FTF probe; free and bound probes were then resolved by electrophoresis and counts were processed by Scatchard plot transformation. Autoradiograms illustrate EMSA reactions using FTF, HNF4, or HNF1 probes with constant amounts of liver nuclear proteins (C, no competitor; wt and m, 50-fold molar excess of cold wild-type or mutant oligos; FTF mutant oligo was TGTTCAAt-GAaA; split FTF bands are translational isoforms; lower-case letters indicate base changes from wild-type FTF sequence).   (11), carrying 4 kb of contiguous 5Ј-flanking mFTF promoter sequence. We started with stable transfection in the well differentiated rat hepatoma 7777.6 subline (4). This stringent system, in which transfected genes are integrated in the genome, has shown much higher sensitivity to mutations affecting basal AFP gene activity, compared with transient expression systems, in which reporter genes function as episomes (4,5). The activity of 4F-CAT stably transfected in 7777.6 cells was reduced by about 75% upon mutation of the FTF DR4 motif (Fig. 6C) (Fig. 6D, lanes 3-4), and over 3-fold in HepG2 cells upon addition of 9-cis-RA and 22(R)-OH-cholesterol (lane 6).
Cotransfection of RXR␣ and TR␤ in the presence of 9-cis-RA and T 3 also enhanced 4F-CAT activity 2-2.5-fold in Hep3B or HepG2 cells (lanes 9 and 10). The DR4 mutant reporter was not enhanced in any of these conditions (lanes 13-20), indicating   (Table III)) were run in duplicates on 1% agarose/10 mM guanidine thiocyanate gels, transferred onto a nylon membrane, and hybridized with digoxigenin-labeled FTF, CYP7A1, SHP, or GAPDH cDNA probes; hybridization signals were quantitated by chemiluminescence. Quantitation of the bands is expressed as fold change in FTFϩ/Ϫ liver relative to FTFϩ/ϩ liver, after correction for GAPDH levels. FTF ratio 0.4 refers to transcripts originating from wild-type FTF alleles. that receptor and ligand effects were mediated by the DR4 element. Transfections with control vectors pSG5 or pCMX (lanes 1 and 2 and 11 and 12) were unreactive to ligands (not shown).
Finally, we tested if FTF would react in vivo to cholesterolderived LXR␣ ligands or to TR hormone ligands. We measured liver FTF levels in Fisher rats fed for 5 days on a 2% cholesterol diet, and in Sprague-Dawley rats 5 and 24 h following an intraperitoneal injection of T 3 thyroid hormone (serum T 3 levels were maximal at 5 h, but still 10-fold over control levels (2.5 nM) at 24 h (data not shown)). With the high cholesterol diet, liver FTF protein level increased about 2-fold, C/EBP␣ about 3.5-fold, HNF4␣ Ϸ1.4-fold, and HNF1␣ was diminished (Fig.  7A). Injection of T 3 increased liver FTF level about 2.3-fold after 5 h, and 3-fold after 24 h; C/EBP␣ showed some increase at 5 h (Fig. 7B). T 3 injection also enhanced the relative amount of TR-RXR complexes bound to the FTF DR4 element in bandshift assays (Fig. 7C). Northern blot measurements at 5 h showed about 2-fold increase in liver FTF mRNA in T 3 -injected animals (Fig. 7D). Liver FTF responses in vivo were thus consistent with a functional DR4 response element at the rat FTF gene locus. DISCUSSION FTF in the Embryo-The FTF/lacZ mouse shows that FTF gene activation is part of prime genomic events surrounding initiation of zygotic transcription at the 2-cell stage. FTF is also ubiquitously expressed in the early embryo, like its homolog in Xenopus (38), and FTF is essential to mouse embryogenesis, like its distant relative Ftz-F1 in Drosophila (39). In contrast, SF1, the closest FTF homolog, is specifically required for development of steroidogenic organs (40). This supports the notion of FTF being a more primitive receptor (41), which also fulfills SF1 functions in lower vertebrates like fish (42). FTFnull mouse embryos provide no signs of essential FTF functions at preimplantation stages. FTF relatives in fly or frog are maternal genes (39,43), whereas FTF is not expressed in mouse oocytes, which could suggest other more ancient functions in lower species. FTFϪ/Ϫ mouse embryos die at E6.5-7.5 with morphological features typical of aborted gastrulation and visceral endoderm dysfunction. This mimics similar phenotypes obtained by inactivating upstream activators (GATA) (12,13,44,45) or downstream targets (HNF4␣, HNF3␤) (31-33, 46, 47) of the FTF gene, suggesting the possibility of interdependent lethal phenotypes in the GATA-FTF-HNF cascade. GATA6 knockout, for instance, is also lethal at E6.5-7.5 (13), and FTF expression could be expected to be severely curtailed in GATA6Ϫ/Ϫ embryos. Reciprocally, FTFϪ/Ϫ lethality occurs somewhat earlier than with HNF3␤ or HNF4␣ knockouts (32,33,46), possibly because the FTF-null state depletes both HNF3␤ and HNF4␣. The FTF/lacZ mouse also shows expression patterns indicative of transient FTF functions in multiple fetal structures outside endodermal derivatives. High FTF gene expression in actively growing cell compartments, such as ossification zones, reinforces the suggestion that the FTF gene may generally react to cell surface growth receptor signaling (11). It also predicts extensive differential use of FTF promoter elements in a developmentally restricted manner. As a case in point, mouse Pdx1 is coexpressed with FTF in fetal pancreas and intestine, but Pdx1 binds to the FTF promoter only in pancreas, and only from E13.5 through E16.5 (48). The FTF DR4 regulatory element, responding to retinoid and thyroid ligands, also opens multiple signaling pathways through which FTF could be transiently up-regulated during fetal development. Finally, FTF physically interacts with the homeodomain protein Ftz (41), a mandatory partner of FTF ancestor Ftz-F1 (39), suggesting that homeodomain proteins may be important FTF coeffectors in developmental events.
FTF in the Adult-Converging evidence suggests that FTF basically operates the enterohepatic cholesterol/BA cycle by activating liver BA production and export via CYP7A1 and mrp3 (14 -17,34), intestinal BA absorption via ASBT (21), and BA reentry from intestinal cells into portal blood via mrp3 (20). Short term overexpression of hepatic FTF in transgenic TgF35 mice provides compelling in vivo support for the activating effect of FTF on CYP7A1 and SHP. FTF-haploinsufficient mice allowed us to investigate further the net systemic effect of reducing FTF production at both hepatic and intestinal FTF expression sites. FTFϩ/Ϫ mice express higher levels of CYP7A1 mRNA than wild-type mice. Recently, a dual mechanism was brought to light in the feedback repression of CYP7A1. One proceeds via BA activation of FXR leading to SHP induction and FTF quenching (15,16), while the other involves BA-FXR activation of an FGF-JNK-c-Jun signaling cascade (49 -52). Basal CYP7A1 activity is thus higher in FXRϪ/Ϫ (18), SHPϪ/Ϫ (49,50), or FGFR4Ϫ/Ϫ (53) mice, and BA repression of CYP7A1 is eliminated in FXRϪ/Ϫ mice (18) but not in SHPϪ/Ϫ mice (49,50). Under FTFϩ/Ϫ steady-state conditions, liver SHP expression is not diminished, but c-Jun level and cell signaling are reduced, thus enhancing CYP7A1 activity. This resembles ASBT-deficient mice, in which ileal BA FIG. 5. Quantitative analysis of liver mRNAs from TgF35 MTI-HA-FTF transgenic and wild-type B6C3F1 mice. Five adult male mice in each group received an intraperitoneal injection of saline or 60 mg/kg ZnSO 4 . After 6 h, livers from each group were pooled, and RNA was prepared. A, real-time qPCR analysis using total RNA expressed as fold change in transgenic liver relative to wild-type liver, normalized to GAPDH. Bars are 1 S.D. in triplicate assays from two preparations of RNA. B, Northern blot analysis using equal amounts of liver poly(A) ϩ RNA from zinc-treated TgF35 versus B6C3F1 mice (RNA samples processed and run in duplicates as in Fig. 4). Quantitation of mRNA bands is fold change in zinc-induced transgenic liver relative to zinc-treated control, corrected for GAPDH levels. FTF ratios 0.5 or 3.3 are values calculated for mRNA transcripts originating from the endogenous FTF gene, or from the endogenous FTF gene plus the HA-FTF transgene.
recycling is eliminated, also resulting in 3-fold increase in CYP7A1, with no change in SHP (54). Like CYP7A1, the AFP gene is a negative acute phase reactant (5), and has long been known to be repressed by c-Jun via the FTF-binding AFP promoter element (5,55,56); furthermore, c-Jun strongly interacts with FTF in GST pull-down assays (57). 3 FTF is thus a plausible feedback target for both SHP and c-Jun.
When FTFϩ/Ϫ mice were fed a cholesterol-or cholic acidenriched diet, liver c-Jun expression was increased (Table III). However, only cholic acid resulted in CYP7A1 repression, showing that c-Jun enhancement per se is not sufficient to mount a feedback response. Robust BA activation of FXR, indicated by higher SHP and lower CYP8B1 in the cholic acid group (array CA, Table III), thus appears required to activate c-Jun into an effective repressor. On the other hand, a pure FXR agonist cannot repress CYP7A1 in the absence of SHP (49,50) and therefore activation of the FXR-FGF-JNK pathway is also not sufficient to mount a significant repressive response. We conclude that a threshold of repressive c-Jun is contributed by both the absolute level and the activation state of c-Jun. Consistent with this, coexpression of c-Jun in transfection assays enhances the repressive effect of JNK on CYP7A1 (52). Cholic acid does not completely repress CYP7A1 in the absence of SHP (49,50), suggesting that SHP is a more effective silencer in using both competitive and active repression mechanisms to quench FTF activity (8).
Cell signaling and the FXR-SHP loop provide first-line feedback mechanisms to reduce BA production in normal conditions. Under cholestasis such as produced by ligation of the bile duct, a second-line hepatoprotective mechanism is activated whereby intense cytokine signaling exacerbates the JNK pathway, resulting in enhancement of hepatic FTF level, with a strong increase of liver mrp3 to expel BA from hepatocytes (34). Since FXRϪ/Ϫ mice are refractory to CYP7A1 repression by 3 (11,35) and the position of the DR4 motif conserved in human, mouse, and zebrafish FTF genes; mDR4: base changes in DR4-mutant oligo mF (panel B) and in DR4-mutant 4F-CAT (panels C and D). B, EMSA analyses using an FTF DR4 probe and total liver nuclear proteins (3 g) from adult male Sprague-Dawley rats. Lanes 1-10, pool of 5 normal livers; lanes 11-14, pool of 3 livers taken 24 h following intraperitoneal injection of 375 g/kg triiodothyronine; lanes C and T 3 , no competitor; lanes F, 7A, AB, mF, 20-fold molar excess of cold FTF, CYP7A1, and ABC1 DR4 oligos, and FTF DR4-mutant oligo. C, functional assay of FTF reporter construct 4F-CAT and its DR4 mutant, using stable transfection in 7777.6 hepatoma cells. Results are CAT activity in cells transfected with mDR4, relative to CAT activity in cells transfected with 4F-CAT and given a value of 1 (triplicate assays in two sets of transfections with pools of 200 -300 cell clones per transfection; bar is 1 S.D.). D, functional assays of 4F-CAT and mDR4 constructs, using transient transfection in HepG2 and Hep3B hepatoma cells. Reporter constructs were cotransfected with a RXR␣ expression vector, and with LXR␣ or TR␤ expression vectors, and cells were cultured without (Ϫ) or with (ϩ) ligands added to the medium (22(R)-OHC and 9-cis-RA in LXR-RXR transfections, T 3 and 9-cis-RA in TR-RXR transfections). Results are CAT activities from three or more transfections, relative to 4F-CAT activity in control transfections (lanes 1 and 2) using vectors pSG5 or pCMX and given a value of 1 (bars are 1 S.D.). excess BA feeding (18), cytokine shortcuts to JNK activation may have limited bearing on normal physiology. Nevertheless, it illustrates that FTF-regulated genes respond differently to homeostatic changes. In FTFϩ/Ϫ liver, FTF target genes mrp3 and CYP7A1 showed opposite changes in steady-state conditions (arrays 1-3, Table III), and CYP7A1, but not mrp3, was inhibited by cholic acid feeding (array CA, Table III). Likewise, in isolated intestinal cells, bile acids activate mrp3 and deactivate ASBT, via their FTF-regulatory domains (20,21). Thus, mrp3 and CYP7A1/ASBT respond to bile acids in opposite directions, as a coherent systemic way to reduce cellular toxicity. This differential regulation seems to be provided by the particular arrangement of tandem FTF activation sites in the mrp3 promoter (20), whereas CYP7A1 and mouse ASBT promoters contain a single FTF activation site.
A simplified model of FTF regulation and function is presented in Fig. 8. In essence, FTF activates CYP7A1 and BA production in hepatocytes (steps 1 and 2), and BA export from enterocytes via mrp3 (step 3). FTF activation of CYP7A1 is subject to feedback repression by SHP (step 4) and by JNKactivated c-Jun (step 5). In the FTFϩ/Ϫ mouse, BA recycling (step 3) and signaling (steps 6 and 7) is attenuated, resulting in higher CYP7A1 activity, higher cholesterol elimination, and lower plasma cholesterol levels. The model can explain why FXR is essential for BA feedback repression of CYP7A1 (in the absence of FXR, steps 7 and 8 are not operational), while the loss of SHP can be bypassed by BA via step 5, but not by an FXR-specific ligand (step 4 is not operational, and step 6 is not activated). The FTF DR4 element further integrates metabolic functions of FTF and LXR␣ (step 9) in reducing hepatic cholesterol stores and enhancing cholesterol secretion. LXR␣ can also indirectly activate genes carrying an FTF response element but devoid of a DR4 element, such as the human CYP7A1 gene. FTF levels somewhat lower than predicted in FTFϩ/Ϫ liver, and negative feedback on the endogenous FTF gene following acute induction of transgenic FTF (Fig. 5B) may be explained by reduced signaling to the DR4 element as a result of enhanced CYP7A1 activity and reduced availability of cholesterol-derived LXR ligands (step 2).
By affecting all FTF-regulated genes in a mildly disruptive way, the FTFϩ/Ϫ mouse establishes a new systemic model closely relevant to physiological conditions, and to better control of atherogenic processes. The FTF-inducible transgenic mouse (TgF35) also provides a powerful tool to identify primary FTF gene targets: this work is in progress in our laboratory. FIG. 8. Model derived from the present and other studies, integrating FTF regulation and function in cholesterol-bile acid homeostasis. FTF exerts positive control on BA synthesis and recycling, and CYP7A1 activation by FTF is subject to repression by SHP and activated c-Jun. In FTFϩ/Ϫ mice, hepatic FTF level is reduced but cell signaling is attenuated, resulting in net enhancement of CYP7A1. See "Discussion" for reference to numbers 1 to 9. L, FXR-specific ligand.
FIG. 7. Liver FTF response in vivo to cholesterol or thyroid hormone. A, Western blot analysis of liver nuclear proteins from Fisher rats fed for 5 days on a control diet (C) or on a 2% cholesterol diet. These chemiluminescence reactions illustrate results obtained with duplicate runs of 10 g total liver nuclear proteins (pools of 5 rats per group). Quantitation of protein bands is expressed as fold change of liver protein levels in rats fed the cholesterol diet, relative to rats fed the control diet, calculated as described under "Experimental Procedures"; split FTF bands are translational isoforms. B, measurement of liver nuclear proteins from Sprague-Dawley rats, 5 and 24 h following intraperitoneal injection of 375 g/kg triiodothyronine, using Western blots quantitated in reference to liver nuclear proteins from rats injected with saline (C); bars are 1 S.D. in triplicate assays from two sets of experiments. C, autoradiogram of duplicate EMSA reactions using the mFTF DR4 probe and liver nuclear proteins from Sprague-Dawley rats (pools of 5), 5 or 24 h following T 3 injection, or 5 h following saline injection (C); quantitation is the ratio of radioactivity counted in TR-RXR complexes relative to LXR-RXR complexes. D, Northern blot analysis of liver FTF mRNA from Sprague-Dawley rats 5 h following T 3 injection, quantitated in reference to liver FTF mRNA from rats injected with saline (C), and normalized to GAPDH (mRNA pooled from 5 livers per group and run in duplicate).

Sylvie Roy, Denis Allard and Luc Bélanger
Jean-François Paré, Daniel Malenfant, Chantal Courtemanche, Mariève Jacob-Wagner, and Cholesterol Homeostasis and Is Regulated by a DR4 Element ) Gene Is Essential to Embryogenesis FTF The Fetoprotein Transcription Factor (