Cloning and functional analysis of the promoter for KDR/flk-1, a receptor for vascular endothelial growth factor.

KDR/flk-1 is one of two receptors for vascular endothelial growth factor, a potent angiogenic peptide. KDR/flk-1 is an early marker for endothelial cell progenitors, and its expression is restricted to endothelial cells in vivo. To investigate the molecular mechanisms regulating expression of KDR/flk-1, we cloned and characterized the promoter of the human KDR/flk-1 gene. The transcription start site was localized by primer extension and ribonuclease protection to a nucleotide 303 base pairs (bp) 5' of the initiation methionine codon. The 5'-flanking sequence is rich in G and C residues and contains five Sp1 elements but no TATA consensus sequence. By reporter gene transfection experiments, we found that approximately 4 kilobases of KDR/flk-1 5'-flanking sequence directed high level luciferase activity in bovine aortic endothelial cells; further deletion analysis revealed positive regulatory elements between bp -225 to -164, -95 to -77, -77 to -60, and +105 to +127. Mutation of an atypical GATA sequence between bp +105 and +127 did not affect promoter activity, suggesting that GATA elements are not essential for the high level promoter activity of this gene. Consistent with endothelial cell-restricted expression of KDR/flk-1 mRNA, we found that the 4-kilobase flanking sequence directed high level promoter activity in endothelial cells but not in other cell types. To our knowledge this is the first report characterizing the KDR/flk-1 promoter. Understanding the KDR/flk-1 promoter will allow us to investigate endothelial cell-specific gene regulation and to uncover methods for targeting gene delivery specifically to endothelial cells.

KDR/flk-1 is one of two receptors for vascular endothelial growth factor, a potent angiogenic peptide. KDR/ flk-1 is an early marker for endothelial cell progenitors, and its expression is restricted to endothelial cells in vivo. To investigate the molecular mechanisms regulating expression of KDR/flk-1, we cloned and characterized the promoter of the human KDR/flk-1 gene. The transcription start site was localized by primer extension and ribonuclease protection to a nucleotide 303 base pairs (bp) 5 of the initiation methionine codon. The 5-flanking sequence is rich in G and C residues and contains five Sp1 elements but no TATA consensus sequence. By reporter gene transfection experiments, we found that ϳ4 kilobases of KDR/flk-1 5-flanking sequence directed high level luciferase activity in bovine aortic endothelial cells; further deletion analysis revealed positive regulatory elements between bp ؊225 to ؊164, ؊95 to ؊77, ؊77 to ؊60, and ؉105 to ؉127. Mutation of an atypical GATA sequence between bp ؉105 and ؉127 did not affect promoter activity, suggesting that GATA elements are not essential for the high level promoter activity of this gene. Consistent with endothelial cell-restricted expression of KDR/flk-1 mRNA, we found that the 4-kilobase flanking sequence directed high level promoter activity in endothelial cells but not in other cell types. To our knowledge this is the first report characterizing the KDR/flk-1 promoter. Understanding the KDR/flk-1 promoter will allow us to investigate endothelial cell-specific gene regulation and to uncover methods for targeting gene delivery specifically to endothelial cells.
Vascular endothelial growth factor (VEGF) 1 is a potent and specific endothelial cell mitogen (1,2). Through interactions with its receptors KDR/flk-1 and flt1, VEGF has critical roles in the growth and maintenance of vascular endothelial cells and in the development of new blood vessels in physiologic and pathologic states (3)(4)(5). The patterns of embryonic expression of VEGF suggest that it is crucial for differentiation of endothelial cells from hemangioblasts and for development of blood vessels at all stages of growth (6,7). Among many potentially angiogenic factors, VEGF is the only one whose pattern of expression, secretion, and activity suggests a specific angiogenic function in normal development (8).
High affinity receptors for VEGF are found only on endothelial cells, and VEGF binding has been demonstrated on macroand microvascular endothelial cells and in quiescent and proliferating endothelial cells, suggesting that these receptors are important for both growth and maintenance of all endothelial cells (6,9). The tyrosine kinases KDR/flk-1 and flt1 have been identified as candidate VEGF receptors by affinity cross-linking and competition binding assays (10 -12). These two receptor tyrosine kinases contain seven similar extracellular immunoglobulin domains and a conserved intracellular tyrosine kinase domain interrupted by a kinase insert (10,13,14); they are expressed specifically by endothelial cells in vivo (11,(15)(16)(17). In situ hybridization in the developing mouse has demonstrated that KDR/flk-1 is expressed in endothelial cells at all stages of development, as well as in the blood islands in which endothelial cell precursors first appear (11), and that KDR/ flk-1 specifies endothelial cell precursors at their earliest stages of development (17).
The vascular endothelium is critical for physiologic responses including thrombosis and thrombolysis, lymphocyte and macrophage homing, modulation of the immune response, and regulation of vascular tone. The endothelium is also intimately involved in the pathogenesis of vascular diseases such as atherosclerosis (18). Although a number of genes expressed in the endothelium have been characterized (19 -22), expression of these genes is either not limited to vascular endothelium (e.g. the genes encoding von Willebrand factor, endothelin-1, vascular cell adhesion molecule-1, platelet/endothelial cell adhesion molecule-1) or is restricted to specific subpopulations of endothelial cells (e.g. the gene for endothelial-leukocyte adhesion molecule-1). (A fragment of the promoter for Tek/ Tie2, another developmentally regulated endothelial cell receptor tyrosine kinase, has recently been shown to direct transgene expression in subpopulations of endothelial cells during mouse embryonic development but not in endothelial cells of adult mice (23). This suggests that the Tek/Tie2 promoter fragment used in this study is sufficient to direct gene expression to subpopulations of endothelial cells during specific periods of development, although functional elements within this promoter have not yet been identified.) In contrast with cells derived from the skeletal muscle and hematopoietic lineages, little is known about the mechanisms of specification and differentiation of endothelial cells. To understand the molecular mechanisms regulating cell type specificity and activation of * This work was supported in part by a grant from Bristol-Myers Squibb. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EMBL Data Bank with accession number(s) X89776 and X89777.
‡ Recipient of a National Research Service Award from the National Institutes of Health.
As a first step we cloned and characterized the promoter for the human KDR/flk-1 gene. We report here the sequence of the 5Ј-flanking region of the gene and identify a single transcription start site located 303 bp 5Ј of the initiation methionine codon. Four kilobases of KDR/flk-1 5Ј-flanking sequence were found to have promoter activity similar to that of the potent SV40 promoter/enhancer in reporter gene transfection experiments in endothelial cells. Deletion analysis in endothelial cells showed the presence of positive regulatory elements in regions from bp Ϫ225 to Ϫ164, Ϫ95 to Ϫ77, Ϫ77 to Ϫ60, and ϩ105 to ϩ127. We found that KDR/flk-1 mRNA was expressed specifically in endothelial cells in culture and that 4 kb of the KDR/ flk-1 5Ј-flanking sequence had cell type-specific promoter activity in transient transfection assays.

EXPERIMENTAL PROCEDURES
Screening of Human and Mouse Genomic Libraries-A 567-bp human KDR/flk-1 cDNA fragment was generated from human umbilical vein endothelial cell (HUVEC) total RNA by the reverse transcriptase polymerase chain reaction (PCR) (24). This fragment was radiolabeled with [␣-32 P]dCTP and used to screen a phage library of human placenta genomic DNA in the vector FixII (Stratagene, La Jolla, CA) as described (24). Likewise, a 451-bp mouse KDR/flk-1 cDNA was generated by reverse transcriptase PCR from mouse lung total RNA and used to screen a phage library of mouse placenta genomic DNA in the vector DashII (Stratagene). Hybridizing clones were isolated and purified from each library, and phage DNA was prepared according to standard procedures (24).
Cell Culture and mRNA Isolation-Bovine aortic endothelial cells (BAEC) were isolated and cultured in Dulbecco's modified Eagle's medium (JRH Biosciences, Lenexa, KS) supplemented with 10% fetal calf serum (HyClone Laboratories, Logan, UT), 600 g of glutamine/ml, 100 units of penicillin/ml, and 100 g of streptomycin/ml as described (22,25). Cells were passaged every 3-5 days, and cells from passages 4 -8 were used for transfection experiments. Saos-2 human osteosarcoma cells, HeLa human epidermoid carcinoma cells, HepG2 human hepatoma cells, human fibroblasts, U937 human histiocytic lymphoma cells, RD human embryonal rhabdomyosarcoma cells, MCF7 human breast adenocarcinoma cells, JEG-3 human choriocarcinoma cells, A7r5 fetal rat aortic smooth muscle cells, and NIH 3T3 mouse fibroblasts were obtained from the American Type Culture Collection. Primary culture HUVEC were obtained from Clonetics Corp. (San Diego) and were grown in EGM medium containing 2% fetal calf serum (Clonetics Corp.). Primary culture human aortic and intestinal smooth muscle cells were also obtained from Clonetics Corp. All cells were cultured in conditions identical to those for BAEC, with the exception that medium used for smooth muscle cells was supplemented with 25 mM HEPES (Sigma) and that HUVEC were cultured as above. Primary culture cells were passaged every 4 -6 days, and cells from passages 3-5 were used in the experiments described here. Total RNA from cells in culture was prepared by guanidinium isothiocyanate extraction and centrifugation through cesium chloride (24).
DNA Sequencing-Restriction fragments derived from the human and mouse KDR/flk-1 genomic phage clones were subcloned by standard techniques into pSP72 (Promega, Madison, WI) or pBluescript II SK (Stratagene) and sequenced from alkaline-denatured double-stranded plasmid templates by the dideoxy chain termination method (26) with Sequenase 2.0 DNA polymerase (U. S. Biochemical Corp.). DNA was sequenced from both directions at least twice, and both dGTP and dITP sequencing protocols were used to resolve compression artifacts in the highly GC-rich 5Ј-flanking region of the human and mouse KDR/flk-1 genes. Sequence analysis was performed with the GCG software package (Genetics Computer Group, Madison, WI).
Primer Extension Analysis-Primer extension analysis was performed as described (25). A synthetic oligonucleotide primer (5Ј-CT-GTCTAGAGAAGGAGGCGCGGAGGTGGAACT-3Ј) complementary to the 5Ј end of the human KDR/flk-1 cDNA (Fig. 1A) was end labeled with [␥-32 P]ATP and hybridized to 20 g of each RNA sample, which was then subjected to reverse transcription. Extension products were analyzed by electrophoresis on an 8% denaturing polyacrylamide gel.
Ribonuclease Protection Assay-We used a 559-bp PstI-PstI fragment of the human KDR/flk-1 gene (Fig. 2B) cloned in pSP72 as the template Potential cis-acting elements are underlined and discussed under "Results." PstI sites used to generate the riboprobe are double underlined, and the oligonucleotide used for primer extension is underlined with an arrow. Panel B, restriction map and nucleotide sequence of the mouse promoter. Nucleotide sequences and potential cisacting elements are indicated as above. An asterisk marks the end of the cDNA.
for in vitro transcription of an ␣-32 P-labeled antisense RNA with T7 RNA polymerase (Boehringer Mannheim). Gel-purified riboprobe (5 ϫ 10 5 cpm) was hybridized with 20 g of total RNA or 3 g of poly(A) RNA plus 17 g of yeast tRNA at 55°C for 16 h in an annealing buffer containing 20 mM Tris-HCl, pH 7.40, 400 mM NaCl, 1 mM EDTA, and 0.1% sodium dodecyl sulfate in 75% formamide. After the RNA had been annealed the unhybridized RNA was digested for 45 min at room temperature with 200 units of RNase T1 (Boehringer Mannheim) and 0.3 unit of RNase A (Boehringer Mannheim) in a buffer containing 10 mM Tris-HCl, pH 7.50, 300 mM NaCl, 5 mM EDTA. The digestion products were then treated with proteinase K, extracted with phenol: chloroform, and analyzed by electrophoresis on a 4% denaturing polyacrylamide gel.
Northern Analysis-RNA blots were hybridized as described (27). Total RNA (10 g) from cells in culture was fractionated on a 1.3% formaldehyde-agarose gel and transferred to a nitrocellulose filter. The human KDR/flk-1 cDNA probe was labeled with 32 P by random priming and used to hybridize the filter. The filter was then autoradiographed for 16 h on Kodak XAR film at Ϫ80°C.
Plasmids-Plasmids pGL2 Basic and pGL2 Control contained the firefly luciferase gene (Promega). pGL2 Basic had no promoter, whereas pGL2 Control was driven by the SV40 promoter and enhancer. The plasmid pSV␤gal (Promega) contained the ␤-galactosidase gene driven by the SV40 promoter and enhancer.
Reporter constructs containing fragments of the human KDR/flk-1 5Ј-flanking region were inserted into pGL2 Basic and named according to the length of the fragment (from the transcription start site) in the 5Ј and 3Ј directions. (For example, plasmid pGL2Ϫ4kbϩ296 contained a human KDR/flk-1 promoter fragment extending from approximately Ϫ4 kb 5Ј of the transcription start site to position ϩ296 inserted into pGL2 Basic.) Plasmids pGL2Ϫ4kbϩ296 and pGL2Ϫ900ϩ296 were created by restriction digestion of purified phage DNA by using 5Ј BamHI and PvuII sites, respectively, and the 3Ј XhoI site at position ϩ296. Plasmids pGL2Ϫ716ϩ268, pGL2Ϫ570ϩ268, pGL2Ϫ323ϩ268, pGL2Ϫ225ϩ268, pGL2Ϫ164ϩ268, pGL2Ϫ37ϩ268, pGL2Ϫ225ϩ127, pGL2Ϫ225ϩ105, pGL2Ϫ225ϩ56, and pGL2Ϫ225ϩ5 were created from promoter fragments generated by PCR of human KDR/flk-1 phage DNA. Plasmids pGL2Ϫ116ϩ268, pGL2Ϫ95ϩ268, pGL2Ϫ77ϩ268, pGL2Ϫ60ϩ268, and pGL2Ϫ12ϩ268 were created by digesting the promoter fragment contained in plasmid pGL2Ϫ164ϩ268 from the 5Ј end with exonuclease III (Pharmacia Biotech Inc.). Plasmid pGL2 GATA-MUT was identical to pGL2Ϫ225ϩ268 except that bp ϩ108 to ϩ110 were mutated in the former (see below). All constructs were sequenced from the 5Ј and 3Ј ends to confirm orientation and sequence.
Mutagenesis-Site-directed mutagenesis of the atypical GATA sequence located in the first exon of the human KDR/flk-1 5Ј-flanking region was performed by PCR according to the method of Higushi et al. (28). A DNA fragment containing human KDR/flk-1 bp Ϫ225 to ϩ268 was used as a template. The sequence TGGATATC was mutated to TGGTCGTC by using one set of mismatched primers (5Ј-TCTGGCAGC-CTGGTCGTCCTCTCCTA-3Ј and 5Ј-TAGGAGAGGACGACCAGGCT-GCCAGA-3Ј) and one set of primers flanking both ends of the template (5Ј-TGCCTCGAGTTGTTGCTCTGGGATGTT-3Ј and 5Ј-TGTAAGCTT-GGGAGCCGGTTCTTTCTC-3Ј). The sequence of the mutated PCR fragment was confirmed by the dideoxy chain termination method.
Transfections-All cell types were transfected by the calcium phosphate method as described (22) with the exception of A7r5 cells, which were transfected with DOTAP (Boehringer Mannheim) as instructed by the manufacturer. In all cases, 20 g of the appropriate reporter construct was transfected along with 2.5 g of pSV␤gal to correct for variability in transfection efficiency. Cell extracts were prepared 48 h after transfection by a detergent lysis method (Promega). Luciferase activity was measured in duplicate for all samples with an EG&G Autolumat 953 luminometer (Gaithersburg, MD) and the Promega luciferase assay system. ␤-Galactosidase activity was assayed as described (22).
The ratio of luciferase activity to ␤-galactosidase activity in each sample served as a measure of the normalized luciferase activity. The normalized luciferase activity was divided by that of pGL2 Control and expressed as relative luciferase activity. Each construct was transfected at least six times, and data for each construct are presented as the mean Ϯ S.E. Relative luciferase activity among constructs was compared by a factorial analysis of variance followed by Fisher's least significant difference test. Statistical significance was accepted at p Ͻ 0.05.  Fig. 1A was hybridized to 20 g of total RNA from HUVEC and HeLa cells or 3 g of poly(A) HUVEC RNA and yeast tRNA. Extension products were analyzed on an 8% polyacrylamide gel. A Sanger sequencing reaction primed on a plasmid DNA template (with the same primer) was run next to it. Panel B, strategy for mapping the transcription start site of the KDR/flk-1 gene by ribonuclease protection. The predicted length of the protected fragment based on the results of primer extension is 145 bp. Panel C, ribonuclease protection analysis of the KDR/flk-1 transcription start site. Total RNA from HUVEC and HeLa cells or poly(A) HUVEC RNA and yeast tRNA was incubated with a 559-bp 32 P-labeled riboprobe spanning the immediate 5Ј region of the human KDR/flk-1 gene, and the annealing products were digested with RNase. The size marker (bp) was prepared by radiolabeling ⌽X174 replicative form DNA digested with HaeIII. Protected fragments were analyzed on a 4% polyacrylamide gel. The arrow denotes a single protected fragment of approximately 145 bp which is observed only in endothelial cells, confirming the results of primer extension. zyme DNA mapping, subcloning, and sequencing. The 780-bp sequence of the promoter and first exon is shown in Fig. 1A. Likewise, a murine KDR/flk-1 cDNA probe was used to screen a murine placental phage library, and one clone was similarly identified and characterized. The sequence of the mouse KDR/ flk-1 promoter is shown in Fig. 1B.

Isolation and Characterization of Human and
Identification of the Transcription Start Site of Human KDR/flk-1-To identify the transcription start site of the human KDR/flk-1 gene, we performed primer extension with a complementary oligonucleotide probe corresponding to bp ϩ212 to ϩ243 (underlined with arrow, Fig. 1A). Primer extension was performed on total RNA from HUVEC and HeLa cells and on poly(A) RNA from HUVEC. Gene transcription was initiated only in endothelial cells (Fig. 2A). A single transcription start site, corresponding to an A nucleotide located 303 bp 5Ј of the site of translation initiation, was identified. We designated this nucleotide as ϩ1 for our remaining experiments involving the human KDR/flk-1 gene. The 5Ј-CA-3Ј nucleotide pair at this position is the most common site for transcription initiation (29).
To confirm the results of the primer extension studies, we performed ribonuclease protection analysis with an antisense riboprobe generated from a 559-bp genomic PstI-PstI fragment extending 5Ј from position ϩ145 (Fig. 2B; the PstI sites are double underlined in Fig. 1A). Incubation of this probe with HUVEC poly(A) RNA and HUVEC total RNA, but not with total RNA from HeLa cells, resulted in protection of a single fragment corresponding in length to the distance between the 3Ј PstI site and the transcription start site identified by primer extension (Fig. 2C). Despite the absence of a TATA consensus sequence, transcription of the human KDR/flk-1 gene appears to begin from a single site located 303 bp 5Ј of the translation initiation codon (Fig. 1A, curved arrow).
Identification of Potential cis-Acting Sequences-The 5Јflanking sequence of the human KDR/flk-1 gene contains islands rich in G and C residues and lacks TATA and CCAAT boxes near the transcription start site (Fig. 1A). Comparison of this 5Ј-flanking sequence with sequences in the Transcription Factors Data Base revealed a series of five Sp1 sites (30) located between human KDR/flk-1 nucleotides Ϫ124 and Ϫ39. There are two AP-2 consensus sites (31,32) at positions Ϫ95 and Ϫ68 and two inverted NFB-binding elements (33,34) at Ϫ130 and Ϫ83 interspersed among the Sp1 sites. Two atypical GATA consensus sequences (both GGATAT) are present in the KDR/flk-1 promoter, one at position Ϫ759 and the other at position ϩ107 within the untranslated portion of the first exon. In addition, multiple CANNTG elements are found in the promoter at positions Ϫ591, Ϫ175, ϩ71, and ϩ184; CANNTG elements can be bound by E-box-binding proteins (35)(36)(37). The sequence AAACCAAA, which is conserved among genes expressed preferentially in keratinocytes (38), is present at human KDR/flk-1 position Ϫ508.
We also compared the human and mouse KDR/flk-1 promoters to identify conserved consensus sequences for nuclear proteins (Fig. 1B). Elements conserved between the two species include two Sp1 sites located at positions Ϫ244 and Ϫ124 relative to the 5Ј end of the reported mouse cDNA sequence (13), two AP-2 sites at positions Ϫ168 and Ϫ148, a noninverted NFB site at position Ϫ153, and the keratinocyte element AAACCAAA at position Ϫ195. An atypical GATA element (GGATAA) is found in the untranslated portion of the first exon of the mouse promoter at position ϩ18; an atypical GATA element (GGATAT) is located similarly in the human promoter. Also, a CANNTG sequence is present 12 bp 5Ј of the G-and C-rich sequences of the promoter at mouse KDR/flk-1 position Ϫ257, a location analogous to that of the CANNTG element at position Ϫ175 of the human promoter. Conservation of these elements across species suggests that some may have functional significance.
Deletion Analysis of the Human KDR/flk-1 Promoter-To identify DNA elements important for basal expression of KDR/ flk-1 in endothelial cells, we constructed a series of luciferase reporter plasmids containing serial 5Ј deletions through the promoter region (Fig. 3). These plasmid constructs in pGL2 Basic were cotransfected into BAEC with pSV␤gal (to correct for differences in transfection efficiency), and the luciferase activity was normalized to that of the pGL2 Control vector driven by the SV40 promoter/enhancer. The activity of the longest human KDR/flk-1 genomic fragment, spanning bp Ϫ4 kb to ϩ296, was similar to that of the powerful SV40 promoter/ enhancer and consistent with the high level of KDR/flk-1 mRNA expression in endothelial cells. Similar levels of activity were produced in constructs containing as much as 15.5 kb of 5Ј-flanking sequence (data not shown). Serial 5Ј deletions from bp Ϫ4 kb to Ϫ225 caused no significant change in promoter activity, implying that elements in this region are not important for basal activity of the KDR/flk-1 promoter. Deletion of sequences between bp Ϫ225 and Ϫ164 significantly reduced KDR/flk-1 promoter activity to 63% that of the full promoter fragment (p Ͻ 0.05), suggesting the presence of positive regulatory elements in this region. Deletion of bp from Ϫ95 to Ϫ77, a sequence that contains one AP-2 site and one NFB site, resulted in a further significant decrease in activity to 20% that of pGL2Ϫ4kbϩ296 (p Ͻ 0.05). Further deletion of bp from Ϫ77 to Ϫ60, an area containing an overlapping AP-2/Sp1 site, significantly reduced KDR/flk-1 promoter activity to less than 5% that of pGL2Ϫ4kbϩ296 (p Ͻ 0.05). Thus, 5Ј deletion analysis revealed that many positive regulatory elements in the KDR/ flk-1 promoter are necessary for high level expression of the gene.
To determine whether sequences in the first exon of human KDR/flk-1 are important for basal expression, we created a series of 3Ј deletion constructs from the vector pGL2Ϫ225ϩ268, which is the smallest construct that possessed full promoter activity (Fig. 4). A fragment was identified between bp ϩ105 and ϩ127 which, when deleted, caused a 5-fold reduction in promoter activity (p Ͻ 0.05), indicating the presence of a positive regulatory element in this region.
Because GATA-2 is a key regulatory factor in endothelial cell-specific gene expression (21,39), we examined the functional importance of the atypical GATA site located between bp ϩ105 and ϩ127 of human KDR/flk-1. Three bp of the GATA motif in the fragment Ϫ225 to ϩ268 were mutated to GTCG by PCR (28) to create pGL2 GATA-MUT. Mutation of these bp in the GATA motif has been observed to eliminate GATA-2 binding activity in the endothelin-1 gene promoter. 2 In comparison with the native pGL2Ϫ225ϩ268 promoter construct, the pGL2 GATA-MUT construct containing the mutated atypical GATA sequence did not have significantly decreased promoter activity in BAEC (p Ͼ 0.05; Fig. 5). (11,16), it does not necessarily follow that its expression would be limited to endothelial cells in culture. To determine whether a tissue culture system is suitable for studying cell type-specific regulation of the KDR/flk-1 gene, we performed Northern analysis of RNA extracted from various cells in culture. KDR/flk-1 message was detected in HUVEC but not in primary culture cells (human aortic and intestinal smooth muscle cells and fibroblasts) or human cell lines (RD, HeLa, HepG2, MCF7, and U937; Fig. 6). Similarly, we have not detected the presence of a KDR/flk-1 message by reverse transcriptase PCR in HeLa, A7r5, or 3T3 cells (data not shown). Thus, expression of KDR/flk-1 message in tissue culture appears to be restricted to endothelial cells, as it is in vivo.

High Level Expression Induced by the KDR/flk-1 Promoter Is Specific to Endothelial Cells-Although KDR/flk-1 expression is restricted to endothelial cells in vivo
To determine whether 5Ј-flanking sequences of the KDR/ flk-1 gene confer endothelial cell-specific expression in cultured cells, we transfected pGL2Ϫ4kbϩ296, which contains more than 4 kb of the human KDR/flk-1 5Ј-flanking sequence and includes most of the untranslated portion of the first exon, into a variety of cell types in culture (Fig. 7). In accord with our previous experiments in BAEC, reporter gene expression driven by the pGL2Ϫ4kbϩ296 promoter fragment was similar When transfected into BAEC, the plasmid pGL2Ϫ225ϩ268 directed luciferase expression comparable to that directed by pGL2 Control, which contains the SV40 promoter and enhancer. When 3 bp of the GATA motif at ϩ107 were mutated to create pGL2 GATA-MUT there was no significant difference in promoter activity.
to that driven by the potent SV40 promoter/enhancer. In JEG-3, Saos-2, A7r5, 3T3, and HeLa cells, however, expression driven by the pGL2Ϫ4kbϩ296 promoter was markedly lower, demonstrating that induction of high level expression by this promoter is specific to endothelial cells. We observed a similar expression pattern with a reporter plasmid containing 15.5 kb of KDR/flk-1 5Ј-flanking sequence (data not shown).
Finally, in an attempt to establish the function of regulatory elements within the KDR/flk-1 5Ј-flanking sequence in other cell types, we transfected into JEG-3 and Saos-2 cells the promoter constructs that defined positive regulatory elements in endothelial cells. In JEG-3 cells, promoter activity was reduced significantly (p Ͻ 0.05) when elements from bp Ϫ77 to Ϫ60 and ϩ127 to ϩ105 were removed (Fig. 8). Because similar reductions were obtained in endothelial cells, these two positive regulatory elements do not appear to be endothelial cell specific. In contrast, no significant changes were noted after deletion of the elements from Ϫ225 to Ϫ164 and Ϫ95 to Ϫ77, suggesting that these fragments may define endothelial cellspecific regulatory elements. (Deletion of the region from Ϫ164 to Ϫ95 resulted in a reduction in promoter activity in JEG-3 cells but not BAEC, which may reflect differential usage of core promoter elements in nonendothelial cells.) Identical studies were done in Saos-2 cells, and the results were similar (data not shown). Because promoter activity in nonendothelial cells is so low, we are reluctant to overinterpret the cell type speci-ficity of regulatory elements in the KDR/flk-1 promoter. However, these results exclude the possibility that the cell type specificity of this promoter is due to the presence of silencer elements in the 5Ј-flanking region of this gene. DISCUSSION As a receptor for VEGF, KDR/flk-1 plays an essential role in angiogenesis and endothelial cell growth, and it is among the earliest markers of endothelial cell differentiation during development. Moreover, in situ analysis and immunocytochemistry have shown that KDR/flk-1 expression is restricted to endothelial cells in vivo; presumably this restricted pattern of expression determines the pattern of VEGF activity. Despite the importance of the KDR/flk-1 gene in endothelial cell growth, the mechanisms that regulate and restrict its expression are not known. We report for the first time the cloning and characterization of the human and mouse KDR/flk-1 promoters, and we identify regions containing positive regulatory elements within the 5Ј-flanking region of the human gene.
Analysis of the human KDR/flk-1 5Ј-flanking region reveals that the transcription start site is located 303 bp 5Ј of the methionine initiation codon. Like constitutive endothelial nitric oxide synthase (40), another gene expressed in endothelial cells, KDR/flk-1 lacks a TATA box, is rich in G and C residues, and has numerous putative binding sites for Sp1, a ubiquitous nuclear protein that can initiate transcription of TATA-less genes (41). We identified by deletion analysis three sequences within the 5Ј-flanking region of the KDR/flk-1 gene which appear to contain elements important for its expression in endothelial cells. Deletion of sequences between bp Ϫ225 and Ϫ164 reduced activity to 63% that of the full-length promoter, deletion between Ϫ95 and Ϫ77 further reduced promoter activity to 20%, and deletion from Ϫ77 to Ϫ60 reduced promoter activity to less than 5%. Because potential binding sites for Sp1, AP-2, NFB, and E-box proteins located within these three positive regulatory elements in the human KDR/flk-1 gene are also present in the mouse 5Ј-flanking sequence, they may represent functional binding domains. AP-2 is a developmentally regulated trans-acting factor (42) without a demonstrated role in endothelial cell gene regulation. NFB, however, trans-activates the inducible expression of vascular cell adhesion molecule-1 and tissue factor in endothelial cells (20,43) and is known to be a mediator of tissue-specific gene regulation (33). Nuclear proteins that bind the E-box motif include the basic helix-loop-helix family of trans-acting factors. E-box-binding proteins have not been clearly associated with endothelial cell gene expression, although members of this family are critical for proper maturation of many cell types, including skeletal muscle and B lymphocytes (36,44). Further experiments will be necessary to determine if these or other unidentified nuclear proteins specifically trans-activate the KDR/flk-1 gene.
Four zinc finger-containing transcription factors in the GATA protein family bind to the consensus sequence (A/ T)GATA(A/G) and regulate cell type-specific gene expression in many cell lineages (45); among these GATA-2 has been most closely linked to endothelial cell gene expression. GATA-2 functions as an enhancer of endothelin-1 gene expression (39) and acts to restrict expression of von Willebrand factor to endothelial cells (21). We observed that the human KDR/flk-1 5Ј-flanking region has two potential GATA-binding sequences, at positions Ϫ759 and ϩ107, and that loss of the element located at Ϫ759 had no effect on expression of KDR/flk-1 in endothelial cells. The potential GATA element at position ϩ107 is in a region of the first exon which we have identified as a powerful positive regulatory element. Although this GATA sequence (GGATAT) differs from the GATA-binding sequences of endothelin-1 and von Willebrand factor and from the consensus GATA sequence (A/T)GATA(A/G), we speculated that it might be the functional motif in the region between ϩ105 and ϩ127 because the functional GATA site in the von Willebrand factor gene is located similarly in the first exon and because a similar GATA element is found in the first exon of the mouse KDR/flk-1 gene. To our surprise, mutation of 3 bp in this element (GATA to GTCG), which had been observed to prevent trans-activation of the GATA cis-acting element in the endothelin-1 promoter, 2 had no significant effect on KDR/flk-1 promoter activity (Fig.  5). Thus, our deletion analysis and mutagenesis studies do not support a functional role for the two GATA sequences in the human promoter in its high level activity in endothelial cells. These observations are consistent with the finding that early stages of endothelial cell development are normal in mice deficient in GATA-2 (46) and suggest that other transcription factors are necessary for expression of the human KDR/flk-1 gene.
We demonstrate in this study that expression of KDR/flk-1 is restricted to endothelial cells in culture, as it is in vivo. Moreover, we show that the activity of the KDR/flk-1 promoter in endothelial cells is similar to that of the potent SV40 promoter/ enhancer and that this high level activity is specific to endothelial cells: activity in other cell types is markedly diminished. We do not yet understand why we observed low but detectable promoter activity in transient transfection assays of cell types that do not express the KDR/flk-1 gene in vivo; however, this situation is not unique among cell type-specific genes (47). It is possible that other silencer elements outside of the 15.5-kb 5Ј-flanking region are necessary to block promoter activity completely in nonendothelial cells. Alternatively, the context of the promoter in relation to normal chromatin structure may be essential for precise regulation of the gene. An example of this type of regulation can be found in the control of MyoD expression. MyoD, like KDR/flk-1, is developmentally regulated, and it marks skeletal muscle precursors at an early stage (48). The MyoD 5Ј-flanking region contains an enhancer element that increases MyoD expression in many cell types in culture, even though MyoD expression is specific to skeletal muscle in vivo (47). In contrast, transgenic constructs containing the MyoD enhancer are skeletal muscle-specific, implying that chromatin structure modifies the activity of this enhancer and regulates cell type specificity. Our results suggest that tissue-specific regulation of KDR/flk-1 involves a complex interaction between known, widely distributed nuclear factors and other, unknown elements. Therefore a complete explanation of the mechanisms of endothelial cell-specific expression of KDR/flk-1 may require integration of in vivo and in vitro observations. Identification of the regulatory mechanisms responsible for KDR/flk-1 expression is likely to provide important information about the specification and differentiation of endothelial cells early in embryogenesis. Moreover, knowledge about DNA elements that restrict gene expression to endothelial cells may be useful for deciphering the function of proteins in this cell type and, potentially, for directing or preventing expression of genes specifically in endothelial cells.