A novel promoter for vascular endothelial growth factor receptor (flt-1) that confers endothelial-specific gene expression.

The human transmembrane fms-like receptor tyrosine kinase Flt-1 is one of the receptors for vascular endothelial growth factor, a growth factor which induces endothelial proliferation and vascular permeability. Flt-1 is expressed specifically in endothelium and is likely to play a role in tumor angiogenesis and embryonic vascularization. To elucidate the molecular basis for the endothelial specific expression of Flt-1, the promoter region has been isolated and functionally characterized. The promoter region contains a TATA box, a GC-rich region, and putative transcription factor binding elements such as cAMP response element binding protein/activating transcription factor (CREB/ATF) and ets. Adenovirus-mediated transient expression of the flt-1 promoter/luciferase fusion gene in endothelial cells and other cell types demonstrated that a 1-kilobase fragment of the 5'-flanking region of flt-1 is involved in the endothelial-specific expression. A CREB/ATF element was found to be essential for basal transcription of the flt-1 expression. In addition, we also showed that the first intron negatively regulates flt-1 promoter activity. The flt-1 promoter will be useful in functional studies on the regulation of endothelial-specific gene expression and also as a tool in targeting the expression of exogenously introduced genes to the endothelium.

The human transmembrane fms-like receptor tyrosine kinase Flt-1 is one of the receptors for vascular endothelial growth factor, a growth factor which induces endothelial proliferation and vascular permeability. Flt-1 is expressed specifically in endothelium and is likely to play a role in tumor angiogenesis and embryonic vascularization. To elucidate the molecular basis for the endothelial specific expression of Flt-1, the promoter region has been isolated and functionally characterized. The promoter region contains a TATA box, a GC-rich region, and putative transcription factor binding elements such as cAMP response element binding protein/ activating transcription factor (CREB/ATF) and ets. Adenovirus-mediated transient expression of the flt-1 promoter/luciferase fusion gene in endothelial cells and other cell types demonstrated that a 1-kilobase fragment of the 5-flanking region of flt-1 is involved in the endothelial-specific expression. A CREB/ATF element was found to be essential for basal transcription of the flt-1 expression. In addition, we also showed that the first intron negatively regulates flt-1 promoter activity. The flt-1 promoter will be useful in functional studies on the regulation of endothelial-specific gene expression and also as a tool in targeting the expression of exogenously introduced genes to the endothelium.
Vascular endothelial growth factor is not only a specific mitogen for vascular endothelial cells but also a potent mediator of vascular permeability (1,2). We as well as other groups have shown that Flt-1 1 (fms-like tyrosine kinase) and Flk-1 (fetal liver kinase-1; mouse homologue of kinase insert domaincontaining receptor (KDR)) are receptors for vascular endothelial growth factor (3)(4)(5)(6). These receptors and Flt-4 (7,8) are members of a family of tyrosine kinases, which is characterized by proteins containing seven immunoglobulin-like domains, a single transmembrane region, and a kinase insert sequence.
We have recently shown that flt-1 is expressed specifically in the endothelium in adult mouse tissues by in situ hybridization (9). We have also shown that flt-1 is expressed in the endothelium during neovascularization of healing skin wounds and during early vascular development in mouse embryos. Therefore, expression of flt-1 is highly restricted to vascular endothelial cells. However, little is known about the molecular regulation of endothelial-specific gene expression as yet.
As a first step to address this issue, we identified the promoter region of flt-1 and characterized this promoter in a transient expression assay. Here we show in a series of transfection assays that a 1-kb DNA fragment of a 5Ј-flanking sequence of flt-1 demonstrates functional activity in vascular endothelial cells but not in epithelial cells, vascular smooth muscle cells, or fibroblasts. The study also demonstrates that the flt-1 CREB/ATF element is essential for basal transcription and the first intron of flt-1 contains negative regulatory elements.

EXPERIMENTAL PROCEDURES
Materials-The recombinant adenovirus Adex1CAlacZ, the adenovirus cosmid vector pAdex1W and EcoT22I-digested adenoviral DNAterminal protein complex were obtained from Dr. Izumi Saito at the Institute of Medical Science, the University of Tokyo.
Cell Cultures-Bovine adrenal endothelial cells (BAEC) were obtained from Dr. Richard Weiner at the University of California, San Francisco, and maintained in Dulbecco's modified essential medium supplemented with 1 mg/ml glucose, 1 ng/ml basic fibroblast growth factor, and 10% fetal bovine serum (FBS). Human umbilical vein endothelial cells (HUVEC), human aortic endothelial cells, human pulmonary arterial endothelial cells, human aortic smooth muscle cells, and human mammary epithelial cells were obtained from Clonetics and maintained according to the manufacturer's recommendation. NIH-3T3 cells and human foreskin fibroblasts (HFF) were maintained in Dulbecco's modified essential medium supplemented with 10% FBS. NCI-H292 human pulmonary mucoepidermoid carcinoma cells were maintained in RPMI 1640 supplemented with 10% FBS. Rat aortic smooth muscle cells (Sprague-Dawley rats) were isolated from explants as described previously (31) and maintained in Dulbecco's modified essential medium supplemented with 10% FBS.
Cloning of the 5Ј-Flanking Region of the Human flt-1 Gene-A human placenta genomic library in EMBL-3 phage (Clontech) was screened with a 600-bp EcoRI/AccI fragment from the 5Ј-end of the flt-1 cDNA. After three rounds of screening, 13 positive clones were isolated. Two sets of overlapping synthetic oligonucleotides, 5Ј-GGACACTC-CTCTCGGCTCCTCCCCGGCAGCGGCGGCGGCTCGG-3Ј (oligo-E) and 5Ј-CGCTGGCCGCTGCACCCGAGCCCCGGAGCCCGCTCCG-AGCCGCCGC-3Ј (oligo-F), corresponding to the 5Ј-end of the flt-1 cDNA between positions ϩ3 and ϩ79 (designated as probe A) and 5Ј-

G G T C T T T G C C T G A A A T G G T G A G T A A G G A A A G C G -
AAAGGCTGAGCATAACT-3Ј (oligo-J) and 5Ј-CAGAATTGTTTGC-CATTTCTTCCACAGGCAGATTTAGTTATGCTCAGCCT-3Ј (oligo-K) corresponding to the sequence of the flt-1 cDNA between positions ϩ427 and ϩ502 (designated as probe B) were annealed, followed by filling-in with Klenow fragment in the presence of [␣-32 P]dCTP. Four of the 13 clones hybridized with probe A, but not with probe B. In contrast, the other clones hybridized with probe B, but not with probe A. Three different clones which hybridized with probe A were selected for restriction endonuclease and Southern blot analyses. The 3-kb EcoRI/XhoI fragments from all three clones and a 7-kb EcoRI fragment from clone 5-11 were subcloned into Bluescript KSϩ (Stratagene) to generate pBKS3.0 and pBKS7.0. These plasmids were used for further restriction enzyme mapping, nucleotide sequencing analysis, subcloning, and expression studies as described below.
Construction of Luciferase-fusion Plasmids Containing a 5Ј-Flanking Region, Exon 1, and a Hybrid Intron-The plasmid p(Ϫ2.5k/ϩ550)spluc which contains a hybrid intron composed of the 5Ј portion of the first intron of flt-1 and the 3Ј portion of mouse immunoglobulin heavy chain gene (10) was constructed by cloning annealed complementary oligonucleotides corresponding to a mouse immunoglobulin heavy chain variable region (5Ј-TCGAGGCTTGAGGTCTGGACATATACATGGGTGA-CAATGACATCCACTTTGCCTTTCTCTCCACAGGTGTCCACTCCC-AGGTCCAACTGCAG-3Ј and 5Ј-CTGCAGTTGGACCTGGGAGTGGA-CACCTGTGGAGAGAAAGGCAAAGTGGATGTCATTGTCACCCAT-GTATATGTCCAGACCTCAAGCC-3Ј) into the XhoI and SrfI sites of p(Ϫ2.5k/ϩ550)-luc. The resulting plasmid p(Ϫ2.5k/ϩ550)sp-luc was treated with XhoI, Klenow fragment, and NcoI and ligated with a 4-kb NcoI/EcoRV fragment from pBKS7.0 which contains the 5Ј portion of the first intron of flt-1.
Transient Transfection Experiments and Enzyme Assays-For transfection analyses, plasmids were purified with Wizard Megaprep (Promega) followed by cesium chloride gradient ultracentrifugation. BAEC and NIH-3T3 cells were seeded onto 6-well plates or, for HFF, onto 100-mm dishes at a density adjusted so that they reached 40 -60% confluence prior to transfection. For BAEC and NIH-3T3 cells, 5 g of DNA of test plasmid, 5 g of pSV-CAT (pCAT-promoter, Promega), and 10 l (5 l for NIH-3T3 cells) of Lipofectin (Life Technologies, Inc.) were incubated in 0.1 ml of OptiMEM (Life Technologies, Inc.) for 20 min. Similarly, for HFF, 25 g of plasmid construct DNA and 5 g of DNA of pSV-CAT, and 50 l of Lipofectin were incubated in 0.4 ml of OptiMEM. The resulting transfection mixture was added to the medium and incubated for 6 -10 h at 37°C. Then, the medium was replaced by complete medium for an additional 3 days. For the stimulation experiments, 10 M forskolin and 0.5 mM 3-isobutyryl-1-methylxanthine (IBMX) or 0.5 mM dibutyryl-cAMP and 0.5 mM IBMX were added directly to media 16 -17 h prior to harvest. Cells were washed with ice-cold phosphatebuffered saline twice, and lysed with 80 l (6-well plate) or 400 l (10-cm dish) of 100 mM potassium phosphate (pH 7.8), 0.5% Triton X-100. After removing the insoluble cell debris by centrifugation, each cell lysate was used to measure luciferase and CAT activities. The luciferase activity was measured with a Monolight 2010 luminometer in the presence of 1 mM dithiothreitol using Luciferase Assay Reagent (Promega). CAT activity was determined by the phase-extraction procedure using [ 3 H]chloramphenicol (DuPont NEN) and xylenes after endogenous deacetylating activity was destroyed by heating the lysates for 10 min at 65°C (11). The efficiency of transfections was normalized with activities of CAT assay.
Construction of a Recombinant Adenovirus Containing the flt-1 Promoter-Luciferase Fusion Gene and Enzyme Assays-The plasmid p(Ϫ748/ϩ284)-luc was digested with SalI, treated with T4 DNA polymerase, and digested with BamHI to generate a 3.7-kb fragment. The fragment was cloned into the SwaI site of pAdex1W adenovirus cosmid vector to generate pAdexFLTP-luc. pAdexFLTP-luc and the EcoT22Idigested adenoviral DNA-terminal protein complex were co-transfected into 293 cells to prepare a replication-negative recombinant adenovirus AdexFLTP-luc (32). A recombinant adenovirus Adex1CAlacZ which contains a CAG promoter (modified chicken ␤-actin promoter with CMV-IE enhancer) (33) and a ␤-galactosidase gene and AdexFLTP-luc were used to coinfect with various cell lines. Briefly, AdexFLTP-luc (1.1 ϫ 10 5 plaque-forming units/ml) and Adex1CAlacZ (8.4 ϫ 10 4 plaqueforming units/ml) were incubated with cells in 0.5 ml of Dulbecco's modified essential medium supplemented with 10% FBS in 24-well plates. Luciferase activity was measured as described above and ␤-galactosidase activity was measured with chlorophenol red ␤-D-galactopyranoside (Boehringer Mannheim Biochemica). The efficiencies of transfections were normalized with activities of ␤-galactosidase assays.
Primer Extension and S1 Mapping-Primer extension analysis was carried out according to described methods (12). Briefly, oligo-F was end-labeled with T4 polynucleotide kinase. Approximately 5 ng of labeled primer was hybridized to 50 g of total RNA from HUVEC, human lung tissue (Clontech), and yeast tRNA in hybridization buffer (80% formamide, 40 mM PIPES (pH 6.4), 400 mM NaCl, 1 mM EDTA) at 30°C overnight. The extension reaction was carried out with 50 units of avian myeloblastosis virus reverse transcriptase (Promega) in 50 mM Tris-HCl (pH 7.6), 60 mM KCl, 10 mM MgCl 2 , 1 mM dNTPs, 1 mM dithiothreitol, 1 units/l of RNase Block (Stratagene), 50 g/ml actinomycin D for 2 h at 37°C. The extended products were analyzed on denaturating 6% gel polyacrylamide gels. Sequence reactions on flt-1 with the same primer were run in parallel for accurate determination of the extension termination site. S1 mapping analysis was carried out as described (13). Briefly, end labeled oligo-F was hybridized with pBKS3.0 and incubated with 4 units of Klenow fragment in the presence of 4 mM dNTPs for 30 min at 37°C. After heat inactivation, the extended product was digested with SmaI, separated on an alkaline agarose gel, and purified by phenol extraction and ethanol precipitation. The probe (5 ϫ 10 4 Cerenkov counts) was then hybridized to 50 g of total RNA from HUVEC, human lung tissue (Clontech), or yeast tRNA in hybridization buffer at 30°C overnight. The reaction mixture was digested with 300 units of S1 nuclease in 280 mM NaCl, 50 mM sodium acetate (pH 4.5), 4.5 mM ZnSO 4 for 60 min at 30°C. The protected products were analyzed on denaturating 6% polyacrylamide gels.

Restriction Map and Exon-Intron Organization of 5Ј-Specific
Human flt-1 Genomic Clones-Genomic clones from a human placental genomic library were obtained by using the human flt-1 cDNA 5Ј-end 600-bp DNA fragment (described under "Experimental Procedures"). Three overlapping but not identical genomic clones were selected for further analysis based on the result of Southern analyses using the human flt-1 cDNA 5Ј-end oligo DNA probe. The restriction maps of these clones were determined by the partial restriction method (Fig. 1). The 3-kb EcoRI/XhoI fragments from all three clones were subcloned into pBluescript-KS(ϩ). Detailed restriction maps and partial sequences showed that these 3-kb fragments were identical. Although the restriction pattern of the 5Ј region of clone 4-18 was different from that of 5-21A and 5-11, the reason for this diversity remains unclear.
Sequence Analysis of the Promoter Region of flt-1-The nu-cleotide sequence of a 1.8-kb BstXI/XhoI fragment from clone 4-18 ( Fig. 1) was determined by the Sanger method. This fragment contains exon 1, a 5Ј portion of intron 1, and the 5Јflanking region of flt-1 containing putative transcription factor binding sites such as a TATA box, a CREB/ATF element, and an ets binding site (Fig. 2). The first intron contains a putative transcription arrest site as discussed below. Transcription Initiation Site-To identify the transcription initiation site of flt-1, primer-extension analysis was performed with total RNA from HUVEC and human lung tissue (Fig. 3). The transcription initiation site was mapped to an adenosine residue 25-bp downstream from the TATA box. This result was confirmed by S1 mapping analysis (data not shown).
The flt-1 Promoter Activity Is Endothelial Cell Specific-To determine the sequences essential for efficient transcription of the flt-1 promoter, a DNA segment extending from ϩ284 bp to Ϫ2.5 kb was fused to a luciferase gene in the pMC-luc vector. The construct designated as p(Ϫ2.5k/ϩ284)-luc contains 2.5 kb of the promoter region, 230 bp of exon 1, and 54 bp of the 5Ј-end of the first intron. This construct was used to generate a series of 5Ј-end deletions (Figs. 4 and 6A). The resultant constructs are referred to as p(X/Y)-luc. For each, X and Y represent the 5Ј-and 3Ј-end positions in nucleotides. Each construct was transfected into BAEC.
Deletion mutant p(Ϫ748/ϩ248)-luc showed the highest activity. Transfection of the series of constructs deleted from Ϫ2.5 kb to ϩ151 suggested the presence of at least two regions, 2500 to Ϫ1195 and Ϫ356 to Ϫ333, containing negative regulatory sequences, and two regions, Ϫ748 to Ϫ583 and Ϫ239 to Ϫ75, containing positively regulatory sequences. Deletion to ϩ151 decreased luciferase activity to the level of the promoterless plasmid pMC-luc.
To determine if the 2.5-kb promoter region used in these experiments was sufficient to confer cell-type specificity, the deletion constructs were also transfected into NIH 3T3 cells and HFF. Relative luciferase activities in these cells were much weaker than those in BAEC (Fig. 4).
To clarify endothelial specificity of the flt-1 promoter further, we employed various human primary cells. However, because we could not efficiently transfect primary cells by conventional methods, we introduced the construct using a replication-deficient recombinant adenovirus. The recombinant adenovirus AdexFLTP-luc carrying the flt-1 promoter (Ϫ748/ϩ284)-luciferase fusion gene was used to infect various human primary cells and established cell lines. Following infection, relative luciferase activities seen in human primary endothelial cells such as aortic endothelial cells, pulmonary arterial endothelial cells, umbilical vein endothelial cells, and bovine adrenal endothelial cells were much higher than activities seen in human primary mammary epithelial cells, human primary aortic smooth muscle cells, NCI-H292 cells, human foreskin fibroblasts, rat primary aortic smooth muscle cells, and NIH-3T3 cells (Fig. 5). These results showed that the flt-1 promoter region between positions Ϫ748 and ϩ284 conferred endothelial-specific gene expression.
The CREB/ATF Element Is Essential for flt-1 Promoter Activity-In some genes, a CREB/ATF element has been shown to be involved in not only transcriptional activation by a cAMPdependent protein kinase A but also basal transcription. To characterize the CREB/ATF element of the flt-1 promoter, we constructed a deletion mutant in the CREB/ATF element. Deletion of 4 internal bases in the CREB/ATF element of the flt-1 promoter (ACGT out of TGACGTCA) diminished relative luciferase activity in BAEC by 85% (Fig. 6B). However, we also failed to detect any stimulation of luciferase activity in response to forskolin/IBMX and dibutyryl-cAMP/IBMX in BAEC transfected with either p(Ϫ962/ϩ284)-luc or p⌬CRE(Ϫ962/ ϩ284)-luc (data not shown). Therefore, the CREB/ATF element of the flt-1 promoter is important for basal transcription of flt-1, but may not be important in the transcriptional activation in FIG. 1. Structure and restriction map of the flt-1 genomic  clones containing a 5-flanking region, exon 1, and a 5  The transcription initiation site identified by primer extension and S1 mapping is indicated by an asterisk and designated as ϩ1 (Fig. 3). The consensus sequence of a TATA box, the putative binding sites for CREB/ATF, and ets, and the putative transcription arrest site are underlined. A unique separated palindromic sequence is boxed. The nucleotide sequence of the synthetic oligonucleotide oligo-F used for primer extension is also underlined. The 5Ј-end of intron 1 is indicated by an arrow.
response to cAMP elevation.
The First Intron of flt-1 Negatively Regulated Transcription-Transfection of the p(Ϫ2.5k/ϩ550)-luc construct, which contains 220 bp of the first intron of flt-1 (containing the 5Ј splice site but not the 3Ј splice site), resulted in no luciferase activity (Fig. 7). This may be due to the production of an undesirable protein instead of luciferase since the first intron contains an ATG at ϩ286 which is upstream of the initiation codon of the luciferase gene and may not be spliced out because of the lack of a 3Ј splice site. When a 3Ј splice site from a mouse immunoglobulin gene (10) was introduced downstream of the first intron to generate a hybrid intron (p(Ϫ2.5k/550)sp-luc), luciferase activity was partially restored. The idea to make this construct was based on results obtained with a hybrid intron consisting of a 5Ј splice site from the first exon of the adenovirus tripartite leader and a 3Ј splice site from a mouse immunoglobulin gene on pMT2 expression vector (14). These studies showed that the hybrid intron was completely spliced out when eukaryotic initiation factor 2 was expressed on pMT2 vector. Thus, it appears likely that the decrease of luciferase activity seen in p(Ϫ2.5k/550)sp-luc (Fig. 7) does not result from a deficiency in splicing. Therefore, we conclude that the first intron of flt-1 negatively regulated the transcription. DISCUSSION In this report, we identified the flt-1 promoter region and showed that a 1-kb fragment of the 5Ј-flanking sequence of the flt-1 gene demonstrated functional activity in vascular endothelial cells but limited activity in epithelial cells, vascular smooth muscle cells, and fibroblasts. Deletion studies of the flt-1 sequence indicated that the regions from Ϫ2500 to Ϫ1195 and Ϫ356 to Ϫ333 contained negative regulatory sequences and regions from Ϫ748 to Ϫ583 and Ϫ239 to Ϫ75 contained positive regulatory sequences.
We determined that the transcription initiation site is located 25 bp downstream of a TATA box by primer extension and S1 mapping. The regions surrounding the transcription initiation site is similar to other initiator sequences. These results indicated that the initiation site of flt-1 is a typical one.
The 5Ј-flanking sequence of flt-1 exhibits common features of a promoter because it contains a TATA box, a GC-rich region, and potential binding sites for transcription factors. Internal deletion studies showed that the consensus CREB/ATF element (TGACGTCA) located at Ϫ74 was essential for basal transcription of flt-1. However, many attempts have failed to transactivate the flt-1 promoter in response to cAMP elevation by treatments with forskolin/IBMX and dibutyryl-cAMP/ IBMX, suggesting that CREB is not involved in regulation of the flt-1 expression. As many transcription factors are known to bind the consensus CREB/ATF element, other members of the basic region-leucine zipper protein family may bind and activate transcription of flt-1 (15). The CREB/ATF-like element of the TGF-␤2 promoter has been shown to be essential for basal level promoter activity, but does not confer responsiveness from either protein kinase A or C (16).
The promoters of genes whose expression is limited to endothelial cells and certain other cells, such as endothelin-1 (17), endothelial-leukocyte adhesion molecule 1 (18), P-selectin (19), vascular cell adhesion molecule 1 (20 -22), and thrombomodu-lin (23) contain ets-binding site(s) (24). Similarly, flt-1 expression is limited to endothelial cells and its promoter contains at least four putative ets-binding sites at Ϫ36, Ϫ49, Ϫ125, and Ϫ141. Ets-1 expression is observed in endothelial cells during the early stages of blood vessel formation (25) and tumor angiogenesis (26). flt-1 is expressed in endothelial cells from the early stages of mouse embryo development (9) and is up-regulated in tumor endothelial cells (27). These results suggest that Ets-1 plays a role in regulation of the flt-1 expression during embryonic vascularization and tumor angiogenesis.
We have also shown that the first intron of flt-1 negatively regulated gene expression. There are several mechanisms by which introns have been shown to regulate gene expression: 1) transcriptional attenuation by a silencer, 2) formation of double-stranded RNA by antisense transcripts (28,29), and 3) transcription arrest (30). In the case of flt-1, the first intron contains a sequence which is very similar to the transcription arrest site in the first intron of the adenosine deaminase gene (Fig. 8). Thus, we predict that the negative regulation conferred by intron 1 may be due to transcriptional arrest.
In conclusion, we showed in a series of transfection assays that a 1-kb DNA fragment of the 5Ј-flanking sequence of flt-1 has functional activity in vascular endothelial cells but limited activity in epithelial cells, vascular smooth muscle cells, and fibroblasts. We also showed that the flt-1 CREB/ATF element was essential for basal transcription and the first intron of flt-1 negatively regulated gene expression. Asterisks indicate identical bases. The core sequence of the transcription arrest site is boxed. The underlined bases have been shown to be important for full transcription arrest activity by point mutational analysis (30).