Functional Characterization of Evolutionarily Conserved DNA Regions in Forkhead Box F1 Gene Locus*

The Forkhead Box f1 (Foxf1) transcription factor (previously known as HFH-8 or Freac-1) is expressed in the septum transversum and splanchnic (visceral) mesoderm and is required for proper development of gut-derived organs. Sequence comparisons of mouse and human Foxf1 genes have revealed highly conserved DNA sequences located within the -5.3-kb Foxf1 promoter region and the 400-nucleotide regulatory element located 1 kb 3′ to the Foxf1 gene (3′RE). To examine their transcriptional activity during mouse embryonic development, we generated transgenic mice in which the expression of the β-galactosidase transgene was controlled by the -2.7-kb Foxf1 promoter region, the -5.3-kb Foxf1 promoter region, or the -5.3-kb Foxf1 promoter region fused to the 3′RE. The -5.3-kb Foxf1 promoter sequences induced appropriate transgene expression in the midgut and developing intestine, whereas the -2.7-kb Foxf1 promoter region was transcriptionally inactive. Addition of 3′RE to the -5.3-kb Foxf1 promoter restored proper transgene expression in the foregut, liver, and lung mesenchyme and prevented ectopic transgene expression in the developing nervous system. Cotransfection studies demonstrated that FoxA2 protein bound to the 3′RE region (+4506/+4529 bp) and was sufficient to inhibit expression of the -5.3-kb Foxf1 promoter. Furthermore, C/EBPβ and HNF-6 proteins bound to the 3′RE region (+4647/+4694 bp) and provided synergistic transcriptional activation of the -5.3-kb Foxf1 promoter in cotransfection assays. These studies demonstrated that the conserved Foxf1 3′RE region is essential for proper tissue-specific regulation of the Foxf1 promoter region during mouse embryogenesis.

Embryonic development of gut-derived organs is regulated by distinct signaling pathways, which stimulate expression of mesenchymal or endoderm-specific transcription factors. These in turn bind cooperatively to distinct promoter regions and activate target gene expression (1)(2)(3). Dynamic changes in gene expression during embryonic development are critical to mediate proper organ morphogenesis, which involves extensive cellular proliferation, migration, and establishment of appropriate positioning of epithelial cells with developing mesenchyme. Understanding the regulation of tissue-specific transcription factors will provide insight regarding organ morphogenesis during embryonic development.
Genetic studies demonstrated that Foxf1Ϫ/Ϫ embryos die by 8.5 dpc due to defects in extraembryonic mesoderm development (20,24). Foxf1 haploinsufficiency is associated with a variety of developmental abnormalities in the lung, gallbladder, esophagus, and trachea, suggesting that Foxf1 regulates mesenchymal-epithelial signaling during organ morphogenesis (21, 24 -26). Approximately 55% of the Foxf1ϩ/Ϫ newborn mice, which have only 20% of the pulmonary wild type levels of Foxf1 mRNA (low Foxf1ϩ/Ϫ), exhibited defects in the development of peripheral lung saccules and microvasculature, severe fusions of the right lung lobes, and pulmonary hemorrhage causing perinatal lethality (24 -26). These lung defects were associated with diminished expression of Flk-1, Pecam-1, Bmp4, Notch-2, and lung Kruppel-like factor, as well as delayed expression of Fgf10, implicating the Foxf1 gene in regulation of branching lung morphogenesis and vasculogenesis (24,25). Most interestingly, 40% of the newborn Foxf1ϩ/Ϫ mice displayed compensatory increases in pulmonary levels of Foxf1 mRNA (high Foxf1ϩ/Ϫ mice). High Foxf1ϩ/Ϫ mice survived and exhibited normal development of the peripheral lung microvasculature (24) but died from pulmonary hemorrhage in response to lung injury, suggesting that Foxf1 is essential for lung repair (27). Furthermore, Foxf1 is expressed in hepatic stellate cells, and haploinsufficiency of the Foxf1 gene caused defective stellate cell activation during liver regeneration (23).
Previous studies revealed that the Foxf1 promoter is highly GC-rich and is regulated by DNA methylation (28). Cotransfection assays demonstrated that the Ϫ1-kb Foxf1 upstream sequences are sufficient to drive the luciferase reporter construct in distinct mouse cell lines (28). Despite growing interest in the Foxf1 gene as a very important transcriptional regulator in developing mesoderm, the in vivo regulation of the Foxf1 promoter remains to be characterized. To investigate Foxf1 regulatory regions, we have generated transgenic mice that use ␤-galactosidase (␤-gal) reporter transgene driven by either the Ϫ2.7-kb Foxf1 promoter, the Ϫ5.3-kb Foxf1 promoter, or the Ϫ5.3-kb Foxf1 region fused to 3Ј Foxf1 conserved regulatory element (3ЈRE), which is located approximately 1 kb 3Ј to the mouse Foxf1 gene. Although the Ϫ2.7-kb Foxf1 region was insufficient to drive ␤-gal expression in vivo, the Ϫ5.3-kb Foxf1 promoter induced transgenic ␤-gal expression in the mesenchyme of midgut and developing intestine. Addition of the 3ЈRE to the Ϫ5.3-kb Foxf1 promoter restored proper ␤-gal transgene expression in the foregut mesoderm and in the mesenchyme of liver and lung, and inhibited ectopic expression of the Ϫ5.3-kb Foxf1 transgene in a developing brain and spinal cord. Our results suggest that the Foxf1 3ЈRE is essential for mesenchyme-specific Foxf1 promoter activity in developing embryos.

MATERIALS AND METHODS
Generation of Foxf1 Transgenic Mice-The Ϫ2.7-kb Foxf1 or Ϫ5.3-kb Foxf1 upstream promoter sequences were fused in-frame with a nuclear localizing ␤-galactosidase (␤-gal) gene and then cloned into an XbaI site of the pBS plasmid (Stratagene). The 3Ј Foxf1 regulatory element (ϩ4400 to ϩ4800 bp) was generated by PCR amplification from mouse genomic DNA and cloned into an XhoI site 5Ј to the Ϫ5.3-kb Foxf1-␤gal construct. The Ϫ2.7-kb Foxf1-␤-gal, Ϫ5.3-kb Foxf1-␤-gal, and Ϫ5.3-kb Foxf1 ϩ 3ЈRE-␤-gal DNA constructs were injected into pronuclei of FVB/N mouse eggs. The fertilized mouse eggs were transferred to surrogate mothers by the University of Illinois at Chicago Transgenic Mouse Facility. Transgenic mice were identified by PCR analysis of the mouse tail DNA using primers specific to the ␤-gal gene as described previously (24). Four distinct Ϫ5.3-kb Foxf1-␤-gal mouse lines, three Ϫ2.7-kb Foxf1-␤-gal mouse lines, and two Ϫ5.3-kb Foxf1 ϩ 3ЈRE-␤-gal mouse lines were established and analyzed for ␤-gal transgene activity.
We described previously the generation of heterozygous Foxf1 knock-in mice (Foxf1ϩ/Ϫ), in which the Foxf1 winged helix DNA binding domain was replaced by an in-frame insertion of a nuclear-localizing ␤-gal gene. These Foxf1ϩ/Ϫ mice were bred for six generations into the Black Swiss mouse genetic background (24). Expression of the ␤-gal gene in Foxf1ϩ/Ϫ mice was under the control of genomic Foxf1 regulatory sequences, thus allowing the use of ␤-gal enzyme staining for visualizing Foxf1-expressing cells. Foxf1ϩ/Ϫ or Foxf1-␤-gal transgenic male mice were mated with wild type female mice to generate mouse embryos at various days of gestation. Embryonic tail DNA was used for genotyping by PCR analysis as described previously (24).
In separate experiments, these LUC reporter plasmids were cotransfected with 250 ng of CMV expression vectors containing the following transcription factors: FoxA2, HNF-6, C/EBP␤, Nkx2.5, GATA-4, or a combination of them. We also included 30 ng of CMV-Renilla luciferase reporter plasmid as an internal control to normalize transfection efficiency. Thirty six hours post-transfection, cells were prepared for dual luciferase assays (Promega). Luciferase activity was normalized to a CMV-empty vector and calculated as a fold induction compared with pGL3Basic-LUC activity. Experiments were performed at least three times in triplicate, and the mean Ϯ S.D. was determined.
Infection of Hepa1-6 Cells with Recombinant Adenoviruses and Chromatin Immunoprecipitation (ChIP) Assays-Hepa1-6 cells (1 ϫ 10 7 cells per 150-mm dish), which do not express endogenous FoxA2, were infected at a multiplicity of infection of 100 infectious units/cell with adenovirus containing FoxA2 expression vector (AdFoxA2) as described (31). AdLacZ was used as a control. Infected Hepa1-6 cells were cross-linked in situ by the addition of 37% formaldehyde and were used to prepare protein extract as described (32). The resulting extract was subsequently sonicated and used for the immunoprecipitation (IP) with FoxA2 rabbit antiserum as described previously (33). Immunoprecipitation with P-selectin rabbit antiserum (BD Biosciences) was used as a control. Cross-links were reversed on all samples by the addition of TE buffer containing 10 g of RNase A (15 min at 25°C). Proteinase K (10 g) was then added, and samples were digested for 16 h at 65°C. DNA was extracted from the digested samples using PCR purification columns according to the manufacturer's instructions (Qiagen). We then used these ChIP DNA samples for PCR by using primers specific to Foxf1 3ЈRE region ϩ4504/ϩ4680 bp (sense 5Ј-acctaggaaaacaaacataataag and antisense 5Ј-actagtttattgataattttgatg).
Statistical Analysis-Student's t test was used to determine statistical significance. p values Ͻ 0.05 were considered significant. Values for all measurements were expressed as the mean Ϯ S.D.

Homology Searches between Mouse and Human Foxf1 Genes Identify Highly Conserved Regions Containing Potential Transcription Factor
Binding Sites-In order to identify conserved DNA regions outside of the Foxf1 exons, which are the most likely locations of regulatory sequences (34), we performed homology searches between the 11-kb Foxf1 human and the mouse gene (Ϫ5.5 to ϩ5.5 kb). This analysis identified two regions of homology within the Ϫ5-kb Foxf1 promoter region (I and II; Fig. 1, A and B) and a homologous 241-nucleotide region (III; ϩ4511 to ϩ4752 bp) located approximately 1-kb 3Ј to the mouse Foxf1 gene (3ЈRE; Fig. 1, A and B). We used the TFSearch, TRANSFAC, and MacVector Subsequence transcription factor binding site search programs to identify potential binding sites in the Foxf1 conserved regions. The binding sites that are conserved in both the mouse and human Foxf1 regulatory sequences have been listed in Fig. 1B. The homologous Foxf1 sequences contain potential binding sites for the following families of transcription factors: basic leucine zipper (bZIP) CCAAT/enhancer binding protein ␤ (C/EBP␤), winged helix/Forkhead Box A (FoxA), zinc finger GATA, homeodomain Nkx2.5, and cut-homeodomain HNF-6 transcription factors (Fig. 1B).
The Ϫ5.3-kb Foxf1 Promoter Drives Embryonic ␤-Gal Transgene Expression in Mesenchyme of Midgut and Developing Intestine-To determine the role of conserved Foxf1 DNA regions in vivo, we generated transgenic mice in which the ␤-gal reporter transgene was expressed under the control of either Ϫ2.7-kb Foxf1 promoter (contains region I), the Ϫ5.3-kb Foxf1 promoter (contains regions I and II), or the Ϫ5.3-kb Foxf1 promoter fused to the 3ЈRE (Ϫ5.3-kb Foxf1 ϩ 3ЈRE; contains regions I, II, and III; Fig. 2A). Three distinct transgenic Ϫ2.7-kb Foxf1 mouse lines, four Ϫ5.3-kb Foxf1 lines, and two Ϫ5.3-kb Foxf1 ϩ 3ЈRE mouse lines were established and analyzed for ␤-gal transgene expression. Similar results were obtained among the lines in each group, suggesting that the observed ␤-gal expression patterns were not because of the precise sites of transgene integration. The Ϫ2.7-kb Foxf1, Ϫ5.3-kb Foxf1, or Ϫ5.3-kb Foxf1 ϩ 3ЈRE transgenic embryos were dissected at different stages of mouse development, stained for ␤-gal enzyme activity, and then compared with ␤-gal staining of the Foxf1ϩ/Ϫ embryos, which contain a ␤-gal reporter gene knocked into endogenous Foxf1 gene locus (24). The Ϫ2.7-kb Foxf1 promoter region was insufficient to drive ␤-gal transgene in vivo, as demonstrated by the absence of ␤-gal staining in Ϫ2.7-kb Foxf1 9.5-18.5 dpc embryos (data not shown). Both Ϫ5.3-kb Foxf1 and Ϫ5.3-kb Foxf1 ϩ 3ЈRE embryos displayed transgenic ␤-gal staining in the mesoderm of midgut (Fig. 2, F, G, and I) and mesenchymal cells of developing 12.5 dpc intestine (Fig. 2, K and L), which was similar to the results from the Foxf1ϩ/Ϫ mouse embryos (Fig. 2, E, H, and J). Transgenic ␤-gal expression was also detected in muscle layers and lamina propria of the adult mouse intestine in both Ϫ5.3-kb Foxf1 and Ϫ5.3-kb Foxf1 ϩ 3ЈRE transgenic mice (Fig. 2, M--O), suggesting that the Ϫ5.3-kb Foxf1 promoter contains regulatory sequences that are sufficient for proper ␤-gal transgene expression in mesenchyme-derived intestinal tissues.
Foxf1 3ЈRE Is Essential for the Foxf1 Promoter to Drive ␤-Gal Transgene Expression in the Foregut Mesoderm-Although ␤-gal staining was not detected in septum transversum mesoderm of either Ϫ5.3-kb Foxf1 or Ϫ5.3-kb Foxf1 ϩ 3ЈRE 9.5 dpc embryos (Fig. 2, C and D), only the FIGURE 1. Potential transcription factor-binding sites within sequences conserved between human and mouse Foxf1 genes. A, mouse Foxf1 gene locus. Vista plot shows conserved DNA regulatory regions (I, II, and III) and Foxf1 exons. B, sequence homology and conserved binding sites within Ϫ3510/Ϫ3415 bp of mouse Foxf1 promoter (region II) and mouse 3ЈRE of the Foxf1 gene (region III). Sequence identity is indicated by the asterisk. Potential transcription factor DNA binding sites are underlined and determined by MacVector Subsequence searches, TFSearch: www.cbrc.jp/research/db/TFSEARCH.html, and TRANSFAC (String-based search query): www.cbil.upenn.edu/tess/. The binding site is listed only when the recognition sequence is found in both the human and mouse Foxf1 sequences. The classification of these transcription factor binding sites is as follows: zinc finger GATA, bZIP C/EBP␤, homeodomain Nkx 2.5, winged helix FoxA, and cut-homeodomain HNF-6.
The Foxf1 3ЈRE Is Essential for ␤-Gal Transgene Expression in Mesenchymal Cells of the Embryonic Liver and Lung-We have previously used Foxf1ϩ/Ϫ mice to demonstrate that ␤-gal is abundantly expressed in the mesenchyme of the embryonic lung (24) (Fig. 3A). Consistent with the absence of ␤-gal staining in the foregut of Ϫ5.3-kb Foxf1 9.5 dpc embryos (Fig. 2C), transgenic ␤-gal expression was not observed in the Ϫ5.3-kb Foxf1 12.5 dpc lungs (Fig. 3B). In contrast, Ϫ5.3-kb Foxf1 ϩ 3ЈRE 12.5 dpc lungs displayed mesenchymal ␤-gal staining, which was increased in peribronchial regions (Fig. 3C). However, these ␤-gal levels were significantly diminished when compared with age-matched lungs from Foxf1ϩ/Ϫ embryos containing a knock in of the ␤-galactosidase gene (Fig. 3A). Transgenic ␤-gal expression was also observed in smooth muscle cells surrounding pulmonary airways and trachea of 15.5 dpc lungs from Ϫ5.3-kb Foxf1 ϩ 3ЈRE transgenic embryos ( Fig. 3L and data not shown).
We demonstrated previously that Foxf1 is expressed in the mesenchyme of the gallbladder and embryonic and adult liver (21,23) (Fig.  3D). Although ␤-gal staining was not observed in Ϫ5.3-kb Foxf1 transgenic livers at all developmental time points (Fig. 3E and data not shown), we detected abundant hepatic ␤-gal expression in the mesen-  chyme of developing gallbladder and liver of Ϫ5.3-kb Foxf1 ϩ 3ЈRE transgenic embryos (Fig. 3, F and I). We used 15.5 dpc liver sections of the Ϫ5.3-kb Foxf1 ϩ 3ЈRE transgenic embryos to show that ␤-gal staining is colocalized with desmin protein, a known marker for hepatic stellate cells (Fig. 3, G and H). Transgenic ␤-gal expression was also observed in the endothelial cells of the Foxf1ϩ/Ϫ and Ϫ5.3-kb Foxf1 ϩ 3ЈRE 15.5 dpc livers as demonstrated by colocalization of ␤-gal staining with immunostaining for the endothelial specific marker isolectin B4 (Fig. 3, J and K, and data not shown). Most interestingly, ␤-gal staining was extinguished in the adult liver and lung of Ϫ5.3-kb Foxf1 ϩ 3ЈRE transgenic mice (data not shown), suggesting that the 3ЈRE Foxf1 region is essential for transient ␤-gal transgene expression in mesenchymal cells of the liver and lung during embryogenesis.
The Foxf1 3ЈRE Is Essential for Inhibiting Ectopic ␤-Gal Transgene Activity in Developing Brain and Spinal Cord-Strong ectopic ␤-gal expression was observed in the developing brain and spinal cord of the Ϫ5.3-kb Foxf1 9.5 dpc embryos (Fig. 2, C and F, and data not shown). In the Ϫ5.3-kb Foxf1 transgenic 12.5 dpc embryos, ectopic ␤-gal expression continued to be detected in diencephalon, midbrain, and tegmentum of pons (basal plate), as well as in the ventral part of the epindymal layer of the neurotube (Fig. 4, B and E). In contrast, the Ϫ5.3-kb Foxf1 ϩ 3ЈRE transgenic embryos did not exhibit ectopic expression of the ␤-gal transgene in the developing brain (Fig. 4C). Furthermore, ␤-gal staining was significantly diminished in the epindymal layer of the neurotube in Ϫ5.3-kb Foxf1 ϩ 3ЈRE transgenic embryos (Fig. 4F), suggesting that the 3ЈRE functions as a repressor of the Foxf1 promoter in the neuroectoderm of a developing brain and spinal cord. Altogether, our transgenic results suggest that the Foxf1 3ЈRE enhances the Ϫ5.3-kb Foxf1 promoter in foregut mesoderm and mesenchymal cells of the liver and lung, whereas the Foxf1 3ЈRE represses the Ϫ5.3-kb Foxf1 promoter in the neuroectoderm of a developing brain and spinal cord.
The Foxf1 3ЈRE Region Represses the Activity of Ϫ5.3-kb Foxf1 Promoter in Cotransfection Experiments-To determine the role of Foxf1 3ЈRE in the regulation of Foxf1 promoter, we generated LUC reporter constructs driven by either the Ϫ5.3-kb Foxf1 promoter alone or the Ϫ5.3-kb Foxf1 promoter fused to 3ЈRE (Ϫ5.3-kb Foxf1 ϩ 3ЈRE). Because the Ϫ2.7-kb Foxf1 region was insufficient to drive ␤-gal transgene in vivo, we did not study this promoter in our further experiments. Cotransfection studies demonstrated that the Ϫ5.3-kb Foxf1 promoter induced LUC activity in both epithelial (HepG2) and mesenchymal (U2OS and HMEC) cell lines (Fig. 5A), a result consistent with previous Foxf1 promoter studies (28). Addition of the 3ЈRE to the Ϫ5.3-kb Foxf1 promoter construct caused a significant decrease in Foxf1 promoterdriven LUC activity in all three transfected cell lines (Fig. 5A), suggesting that the 3ЈRE functions as a repressor when linked to the Ϫ5.3-kb Foxf1 promoter region.
To investigate the effect of 3ЈRE on the activity of a heterologous minimal promoter, we generated LUC reporter constructs, which contained the 3ЈRE fused with either TATA box sequences (21,23,30) or the minimal hepatocyte-specific Ϫ202-bp transthyretin (TTR) promoter region (1, 2, 11-13) driving expression of the LUC reporter gene. Transient transfection experiments demonstrated that the 3ЈRE does not influence expression of the TATA-LUC basal promoter construct in either human osteosarcoma U2OS or human hepatoma HepG2 cells (Fig. 5B). Consistent with the hepatocyte-specific activity of the TTR promoter region, the 3ЈRE sequence induced TTR promoter activity only in HepG2 cells and not in U2OS cells (Fig. 5B). These cotransfection experiments demonstrated that the 3ЈRE functions as an enhancer in HepG2 cells when linked to the Ϫ202-bp TTR promoter region.
FoxA2 Transcription Factor Binds the 3ЈRE Region and Represses the Activity of Ϫ5.3-kb Foxf1 Promoter-Our transgenic studies demonstrated that 3ЈRE functions as a repressor to Ϫ5.3-kb Foxf1 promoter in the developing midbrain and spinal cord, both of which exhibit strong expression for FoxA proteins (35)(36)(37). Because a conservative FoxAbinding site is present in the 3ЈRE region (Fig. 1B), we wanted to determine whether FoxA proteins are able to mediate the repression function of the 3ЈRE. We generated the LUC reporter construct driven by the Ϫ5.3-kb Foxf1 promoter fused to 3ЈRE ⌬ ϩ4641/ϩ4752-bp deletion mutant, which contains only the FoxA-binding site. Cotransfection studies demonstrated that either the 3ЈRE region or the 3ЈRE ⌬ ϩ4641/ ϩ4752-bp deletion mutant diminished the activity of the Ϫ5.3-kb Foxf1 promoter (Fig. 5C), suggesting that ϩ4511/ϩ4640-bp nucleotide region of the 3ЈRE was sufficient to repress the Ϫ5.3-kb Foxf1 promoter activity.
To determine the role of the FoxA2 protein in 3ЈRE function, we performed cotransfection experiments using the Ϫ5.3-kb Foxf1 or Ϫ5.3-kb Foxf1 ϩ 3ЈRE luciferase constructs and CMV-FoxA2 expression vector. Although FoxA2 transfection did not influence the Ϫ5.3-kb Foxf1 promoter activity compared with transfection with the empty CMV plasmid, transfected FoxA2 protein significantly diminished LUC activity driven by the Ϫ5.3-kb Foxf1 ϩ 3ЈRE promoter (Fig. 5D). These results suggest that FoxA2 protein represses the Foxf1 promoter through the 3ЈRE conserved region.
To determine whether FoxA proteins directly bind the 3ЈRE region, we synthesized a double-stranded oligonucleotide corresponding to the mouse 3ЈRE region ϩ4506/ϩ4529 bp, which contains a potential FoxAbinding site. EMSAs were performed with this oligonucleotide and nuclear protein extract from CMV-FoxA1-or CMV-FoxA2-transfected U2OS cells. Both FoxA1 and FoxA2 proteins formed specific DNA-protein complexes, which were inhibited by the addition of cold competitor oligonucleotide interfering with formation of the protein-DNA complexes (Fig. 5E). In contrast, the formation of FoxA-DNA complexes was not inhibited by a nonspecific oligonucleotide containing the Sp1-binding site. These results show that FoxA1 and FoxA2 proteins directly bind to the ϩ4506/ϩ4529-bp 3ЈRE region. Furthermore, we used ChIP assays to determine whether FoxA2 protein binds to the 3ЈRE Foxf1 region in the context of endogenous mouse DNA. These ChIP assays demonstrated that adenovirally expressed FoxA2 protein specifically binds to chromatin-associated Foxf1 3ЈRE region in the context of endogenous DNA (Fig. 5F).
C/EBP␤ Protein Synergizes with HNF-6 to Activate the Ϫ5.3-kb Foxf1 ϩ 3ЈRE Promoter-Because our transgenic studies demonstrated that Foxf1 3ЈRE stimulates the Foxf1 promoter in the mesenchyme of the developing lung and liver (Fig. 3), we wanted to determine whether the Ϫ5.3-kb Foxf1 promoter can be activated through the ϩ4641/ ϩ4752-bp 3ЈRE region, which contains several overlapping binding sites for known transcriptional activators (Fig. 1B). We performed cotransfection experiments in mesenchymal U2OS cells using Ϫ5.3-kb Foxf1 or Ϫ5.3-kb Foxf1 ϩ 3ЈRE LUC constructs and CMV expression vectors containing HNF-6, C/EBP␤, Nkx2.5, and GATA-4, or combinations of these proteins. Although Foxf1 3ЈRE maintained its inhibitory activity for the Foxf1 promoter after transfection with one of the CMV expression vectors (Fig. 6A), combinations of GATA-4 protein with C/EBP␤ or HNF-6 as well as C/EBP␤ protein with Nkx2.5 completely blocked the 3ЈRE repressor activity (Fig. 6A). In contrast, cotransfection of HNF-6 and C/EBP␤ proteins caused a synergistic increase in LUC activity of the Ϫ5.3-kb Foxf1 ϩ 3ЈRE reporter compared with either HNF-6 or C/EBP␤ alone (Fig. 6A), suggesting that FIGURE 5. FoxA2 protein binds to the 3RE region and retention of FoxA2-binding sequences is sufficient to repress the ؊5. 3-kb Foxf1 promoter. A, the Foxf1 3ЈRE inhibits the Ϫ5.3-kb Foxf1 promoter activity in cotransfection assays. We transiently transfected U2OS, HepG2, or HMEC cells with the Ϫ5.3-kb Foxf1 or Ϫ5.3-kb Foxf1 ϩ 3ЈRE LUC reporter plasmids. Cells were harvested at 36 h after transfection and processed for dual luciferase assays to determine luciferase activity. Transfections were performed twice in triplicate and used to calculate the fold transcriptional induction relative to pGL3 basic luciferase plasmid (Ϯ S.D.). B, the Foxf1 3ЈRE enhances TTR promoter activity in HepG2 cells. Transient transfection experiments were performed in HepG2 and U2OS cells using one of the following luciferase reporter constructs: TATA-LUC, 3ЈRE-TATA-LUC, TTR-LUC, and 3ЈRE-TTR-LUC. C, the ϩ4511/ϩ4641-bp nucleotide region of the 3ЈRE was sufficient to repress the Ϫ5.3-kb Foxf1 promoter activity. We transiently transfected U2OS cells with LUC reporter plasmid driven by the Ϫ5.3-kb Foxf1 promoter alone, Ϫ5.3-kb Foxf1 ϩ 3ЈRE, or Ϫ5.3-kb Foxf1 ϩ 3ЈRE ⌬ 4641-4752-bp deletion mutant. Transfections were performed in triplicate and used to calculate the fold transcriptional induction relative to pGL3 basic luciferase plasmid (Ϯ S.D.). A p value Յ 0.05 is shown with asterisk. D, FoxA2 protein represses the activity of Ϫ5.3-kb Foxf1 promoter through 3ЈRE. U2OS cells were transiently transfected with Ϫ5.3-kb Foxf1 or Ϫ5.3-kb Foxf1 ϩ 3ЈRE luciferase reporter plasmids as well as CMV-empty and CMV-FoxA2. The double asterisk indicates statistical significance in FoxA2 transfected cells compared with CMV-empty plasmid. E, EMSAs show that FoxA1 and FoxA2 transcription factors bind to their potential binding sites in the Foxf1 3ЈRE. Nuclear protein extract was prepared from untransfected (UN) U2OS cells or cells transfected with CMV-FoxA1 or CMV-FoxA2. EMSA was performed with the ϩ4506/ϩ4529-bp oligonucleotide containing a potential FoxA-binding site. Specificity of these protein-DNA complexes was demonstrated by the ability of the cold competitor DNA (C, 500-fold molar excess) and the inability of the nonspecific Sp1 oligonucleotide to interfere with formation of protein-DNA complexes (shown with arrows). F, ChIP assays show that FoxA2 protein bind to the 3ЈRE Foxf1 region in context of endogenous DNA. Cross-linked chromatin was prepared from Hepa1-6 cells at 24 h after infection with AdFoxA2 or AdLacZ. The cross-linked and sonicated chromatin was then immunoprecipitated (IP) with antibodies specific to FoxA2 or P-selectin (control). Immunoprecipitated genomic DNA was analyzed for the amount of mouse 3ЈRE DNA using PCR analysis with primers specific to the mouse Foxf1 3ЈRE region (ϩ4504 to ϩ4680 bp).
HNF-6 and C/EBP␤ synergize to convert 3ЈRE from transcriptional repressor into activator. Consistent with this hypothesis, both HNF-6 and C/EBP␤ displayed specific bindings to their potential binding sites in the 3ЈRE region as demonstrated by electrophoretic mobility shift assays with the ϩ4647/ϩ4694-bp oligonucleotide (Fig. 6B). Furthermore, transcriptional synergy between C/EBP␤ and HNF-6 proteins was significantly reduced after deletion of the ϩ4641/ϩ4752-bp 3ЈRE region, which contains HNF-6 and C/EBP␤-binding sites (Fig. 6C). These results suggest that the Foxf1 3ЈRE can function as an enhancer to the Ϫ5.3-kb Foxf1 promoter in the presence of both HNF-6 and C/EBP␤ transcription factors. Most interestingly, immunostaining with antibodies specific to HNF-6 or C/EBP␤ proteins demonstrated that these transcriptional regulators are coexpressed in the mesenchyme of trachea and proximal airways of 15.5 and 18.5 dpc mouse embryos (Fig.  6, D-F, and data not shown). These embryonic tissues displayed abun-dant ␤-gal staining in Foxf1ϩ/Ϫ and Ϫ5.3-kb Foxf1 ϩ 3ЈRE transgenic embryos but not in Ϫ5.3-kb Foxf1 embryos ( Fig. 6G and Fig. 3L, and data not shown). These results suggest that HNF-6 and C/EBP␤ proteins stimulate the ␤-gal transgene reporter through the 3ЈRE Foxf1 region in vivo.

DISCUSSION
Mouse genetic studies have demonstrated that the Foxf1 protein (previously known as HFH-8 or Freac-1) is an important developmental regulator of vasculogenesis and mesenchymal-epithelial cell signaling during mouse embryonic development. Foxf1Ϫ/Ϫ embryos die by 8 dpc due to extra-embryonic mesoderm defects (20,24). Foxf1 haploinsufficiency is associated with severe defects in development of the lung, esophagus, trachea, and gallbladder (24 -26). To gain insight into in vivo regulation of the Foxf1 gene, we performed a homology search between FIGURE 6. C/EBP␤ protein synergizes with HNF-6 to activate the ؊5. 3-kb Foxf1 ؉ 3RE promoter. A, cotransfection experiments show transcriptional synergy between C/EBP␤ and HNF-6. U2OS cells were transiently transfected with Ϫ5.3-kb Foxf1 or Ϫ5.3-kb Foxf1 ϩ 3ЈRE luciferase reporter plasmids as well as CMV-HNF-6, CMV-C/EBP␤, CMV-Nkx2.5, or CMV-GATA-4 expression vectors alone or in combination. Cells were harvested at 36 h after transfection and processed for dual luciferase assays to determine luciferase activity. Transcriptional induction was normalized to CMV-empty vector and expressed as a fold increase relative to pGL3 Basic luciferase plasmid (Ϯ S.D.). B, EMSAs show that HNF-6 and C/EBP␤ bind to their potential binding sites in the Foxf1 3ЈRE. Nuclear protein extract was prepared from untransfected (UN) U2OS cells or cells transfected with the CMV-C/EBP␤ or CMV-HNF-6. EMSA was performed with the ϩ4647/ϩ4694-bp oligonucleotide containing HNF-6, C/EBP␤, Nkx2.5, and GATA-4-binding sites. Specificity of these protein-DNA complexes was demonstrated by the ability of the cold competitor DNA (C, 500-fold molar excess) and the inability of the nonspecific Sp1 oligonucleotide to interfere with formation of protein-DNA complexes (shown with arrows). C, deletion of ϩ4641/ϩ4752-bp 3ЈRE region diminishes C/EBP␤/HNF-6 transcriptional synergy. U2OS cells were transiently transfected with LUC reporters driven by the Ϫ5.3-kb Foxf1 promoter alone, Ϫ5.3-kb Foxf1 ϩ 3ЈRE, or Ϫ5.3-kb Foxf1 ϩ 3ЈRE ⌬ 4641-4752-bp deletion mutant, and with CMV-HNF-6 and/or CMV-C/EBP␤ expression vectors. Transcriptional synergy was normalized to CMV-empty vector and expressed as a fold increase relative to pGL3 basic luciferase plasmid (Ϯ S.D.). The statistical significant differences (p Յ 0.05) are shown with an asterisk. D--G, HNF-6 and C/EBP␤ proteins are coexpressed in the mesenchyme of developing trachea. Wild type 18.5 dpc trachea (D--F) was dissected, paraffin-embedded, sectioned, and then stained with either isotype-control IgG, anti-HNF-6 or anti-C/EBP␤ antibody as described under "Materials and Methods." ␤-Gal is expressed in trachea of Foxf1ϩ/Ϫ 18.5 dpc embryo (G). Abbreviations used are as follows: NE, control without nuclear extract; Tr, trachea. Magnification: D--G, ϫ200. mouse and human Foxf1 DNA sequences. This analysis enabled us to identify two regions of homology within the Ϫ5.3-kb Foxf1 promoter region and a homologous 241-nucleotide region located approximately 1-kb 3Ј to the mouse Foxf1 gene (3ЈRE). To determine the role of conserved Foxf1 DNA regions in vivo, we generated transgenic mice in which the ␤-gal reporter transgene was expressed under the control of either the Ϫ2.7-kb Foxf1 promoter, the Ϫ5.3-kb Foxf1 promoter, or the Ϫ5.3-kb Foxf1 promoter fused to the 3ЈRE (Ϫ5.3-kb Foxf1 ϩ 3ЈRE). In this study, we compared ␤-gal staining in these new transgenic mouse lines with ␤-gal expression in Foxf1ϩ/Ϫ mice, in which the winged helix DNA binding domain was replaced by an in-frame insertion of a nuclear localizing ␤-gal gene to provide the true Foxf1 expression pattern from endogenous Foxf1 gene locus (24). The use of this ␤-gal gene knocked into the Foxf1 locus allowed us to determine that Foxf1 is expressed in the splanchnic mesoderm of embryonic gut and septum transversum mesenchyme (21). In this study we demonstrated that the Ϫ2.7-kb Foxf1 promoter was insufficient to drive ␤-gal transgene expression in mouse embryos. In contrast, the Ϫ2-kb Foxf1 promoter exhibited correct temporal and spatial expression of a GFP reporter in transgenic Xenopus tadpoles (38), illustrating evolutional complexity in the regulation of the Foxf1 gene among different species. Although the Ϫ5.3-kb Foxf1 promoter alone was sufficient to drive ␤-gal transgene expression in the mesenchyme of the midgut in transgenic mouse embryos, addition of the 3ЈRE was required for the Foxf1 promoter to elicit proper ␤-gal expression in the foregut mesoderm and mesenchymal cells of the developing liver and lung, suggesting that the Foxf1 3ЈRE functions as an enhancer to the Ϫ5.3-kb Foxf1 promoter in these mesenchymal tissues.
The Foxf1 3ЈRE sequence contains high affinity binding sites for transcription factors FoxA, HNF-6, C/EBP␤, Nkx2.5, and GATA. Cotransfection studies demonstrated that Foxf1 3ЈRE exhibited repressor properties when cells were untransfected or transfected with the FoxA2 protein. EMSA and ChIP assays demonstrated that the FoxA2 protein binds to the ϩ4506/ϩ4529-bp 3ЈRE region, and retention of these FoxA2-binding sequences was sufficient to repress the Ϫ5.3-kb Foxf1 promoter. In contrast, the 3ЈRE sequence functioned as an enhancer when C/EBP␤ expression vector was combined with HNF-6. Furthermore, deletion of ϩ4641/ϩ4752-bp 3ЈRE region, which includes C/EBP␤ and HNF-6 DNA-binding sites, abolished transcriptional synergy between C/EBP␤ and HNF-6 proteins in cotransfection experiments. This is the first demonstration that C/EBP␤ and HNF-6 proteins can synergize to activate mesenchyme-specific genes. Although we did not observe transcriptional synergy between C/EBP␤ and GATA-4, other investigators have shown that C/EBP␤ can directly interact with GATA-4 and GATA-1 proteins to provide transcriptional synergy for activation of the steroidogenic acute regulatory protein promoter and eosinophil granule major basic protein promoter, respectively (39,40). These results suggest that the Foxf1 3ЈRE plays an important role in the regulation of the Foxf1 promoter and that, depending on the combination of transcription factors, the 3ЈRE functions to inhibit or activate expression of the Ϫ5.3-kb Foxf1 promoter region.
Although the mechanism underlying a tissue-specific regulation of the Foxf1 3ЈRE in transgenic mice is not completely understood, several members of the Fox, zinc finger GATA, homeodomain Nkx, and bZIP C/EBP families of transcriptional regulators are expressed in foregut mesoderm as well as the mesenchyme of developing liver and lung (19,(41)(42)(43)(44). We demonstrated that both HNF-6 and C/EBP␤ proteins are coexpressed in the mesenchyme of trachea and proximal airways, and this was associated with abundant ␤-gal staining in Foxf1ϩ/Ϫ and Ϫ5.3-kb Foxf1 ϩ 3ЈRE transgenic embryos but not in Ϫ5.3-kb Foxf1 embryos. These results suggest that HNF-6 and C/EBP␤ proteins may enhance the Foxf1 promoter activity through 3ЈRE Foxf1 region.
Foxf1 is expressed in the mesenchyme of the gallbladder (21) as well as in the hepatic stellate and endothelial cells of the embryonic and adult Foxf1ϩ/Ϫ liver (see Ref. 23 and this study). Identical ␤-gal staining was observed in the liver of Ϫ5.3-kb Foxf1 ϩ 3ЈRE embryos but not in the Ϫ5.3-kb Foxf1 transgenic livers, suggesting that the Foxf1 3ЈRE region functions as an enhancer in hepatic mesenchymal cells. The Foxf1 3ЈRE may be regulated by synergistic transcriptional interactions involving C/EBP␤ transcription factor, a known coactivator of the mouse ␣1 collagen promoter in hepatic stellate cells (45). Because the ␤-gal transgene was not expressed in adult Ϫ5.3-kb Foxf1 ϩ 3ЈRE liver, the 3ЈRE may enhance Foxf1 transcription during development or maturation of the hepatic mesenchymal cells. Alternatively, the ␤-gal transgene can be inactivated by DNA methylation through potential methylation sites in the Ϫ5.3-kb Foxf1 promoter region (28). However, in our hands, in vivo inhibition of histone deacetylases by trichostatin A treatment (46) did not reactivate ␤-gal transgene expression (data not shown), suggesting that DNA methylation does not play a role in ␤-gal transgene silencing. Most interestingly, the ␤-gal transgene was not expressed in septum transversum mesoderm in both Ϫ5.3-kb Foxf1 and Ϫ5.3-kb Foxf1 ϩ 3ЈRE transgenic embryos. These results indicate that the Ϫ5.3-kb Foxf1 promoter and 3ЈRE region are still missing important regulatory sequences required for proper transgene expression in the septum transversum.
Strong ectopic ␤-gal staining was observed in the developing brain and spinal cord of Ϫ5.3-kb Foxf1 transgenic embryos. We demonstrated that the addition of 3ЈRE to Ϫ5.3-kb Foxf1 promoter caused inhibition of this ectopic ␤-gal staining, a result consistent with the ability of the Foxf1 3ЈRE to repress the Foxf1 promoter in cotransfection studies. Most interestingly, the FoxA2 transcription factor is expressed in the developing midbrain and spinal cord (35)(36)(37), and FoxA2 expression partially overlaps with ␤-gal transgene expression in the Ϫ5.3-kb Foxf1 embryos. Because FoxA2 protein directly binds to the 3ЈRE region and retention of the FoxA2 binding site is sufficient to repress the Ϫ5.3-kb Foxf1 promoter activity, the FoxA2 protein may be involved in repression of Foxf1 promoter activity in the developing brain and spinal cord.
In summary, Foxf1 3ЈRE is required for the Ϫ5.3-kb Foxf1 promoter to drive ␤-gal transgene expression in foregut mesoderm and mesenchymal cells of the developing liver and lung. The Foxf1 3ЈRE prevents ectopic ␤-gal transgene expression in developing brain and spinal cord. Taken together, our data suggest that the Foxf1 3ЈRE is involved in tissue-specific regulation of the Foxf1 promoter during mouse embryonic development.