Requirement of a GT box (Sp1 site) and two Ets binding sites for vascular endothelial cadherin gene transcription.

Vascular endothelial cadherin (VE cadherin) gene encodes a Ca2+-dependent cell adhesion molecule required for the organization of interendothelial junctions. This gene is exclusively and constitutively expressed in endothelial cells. Previous data with transgenic mice revealed that the transcriptional regulatory elements present within a -2486/+24 DNA fragment of mouse VE cadherin gene mimic the tissue-specific activity of the endogenous promoter. In this study, we analyzed elements implicated in the function of the proximal regulatory region. Electrophoretic mobility shift assay identified a GT-rich sequence (positions -49/-39) interacting with factors related to the Sp1 family. Point mutations abolished the binding of nuclear proteins in vitro and drastically diminished the activity of the promoter in transient transfection assay. Supershift assays with antibodies against proteins of the Sp1 family revealed that Sp1 and Sp3 interact with this region of the VE cadherin promoter. Furthermore, two GGAA motifs, located at positions -93/-90 and -109/-106, were shown to interact with nuclear factors. Site-directed mutagenesis of these sequences demonstrated that these Ets binding sites are essential for promoter activity. In vitro binding assays in the presence of various antisera suggest that Erg is one of the proteins interacting with the -109/-106 site.

Vascular endothelial cadherin (VE cadherin, 1 CD 144) is a member of the calcium-dependent cell adhesive protein family that is involved in the formation and maintenance of cell-cell contacts (1)(2)(3). Cadherins promote a homophilic type of binding through their extracellular domains. The cadherin cytoplasmic region interacts with proteins called catenins (␣-catenin, ␤-catenin, plakoglobin, and p120), which mediate the linkage of the cadherins to the actin cytoskeleton (4).
VE cadherin is strictly specific to endothelial cells and is located at intercellular junctions (1,5). In embryonic stem cell-derived embryoid bodies, VE cadherin null mutation inhibited the assembly of vascular-like structures, thus indicating a crucial role for this protein in vascular morphogenesis (6). In mouse embryos, VE cadherin transcripts were detected at the time of endothelial cell differentiation from mesodermal derivatives (embryonic day 7.5) (5). VE cadherin expression was detected in the developing vascular tree and in all vessel types in the adult, although a down-regulation of VE cadherin transcription level has been observed in brain capillaries (5). The constitutive and exclusive expression of VE cadherin in endothelial cells is a unique feature among the proteins known to be expressed in this cell type. Despite much effort to determine the transcription factors interacting with endothelial promoters, the mechanisms that direct gene expression in the endothelium are still poorly understood. Accordingly, we decided to undertake a functional dissection of the VE cadherin promoter.
VE cadherin locus (Cdh5) has been localized in the middle of mouse chromosome 8, in a region where other cadherin loci have been mapped (7). The gene contains 12 exons spanning more than 36 kilobase pairs, and the first exon, which is entirely untranslated, starts at one transcriptional site. In another report, 2 we show that a DNA fragment containing nucleotides Ϫ2486 to ϩ24 was sufficient to confer endothelial specific expression of a reporter gene in transgenic mice, indicating that most if not all the sequences regulating VE cadherin gene expression are present in this fragment. Furthermore, in cultured cells, a 5Ј-deletion strategy, used to map the functional domains located within the Ϫ2486/ϩ24 sequence, indicated that the following three well defined regulatory regions were involved: a proximal Ϫ139/ϩ24 region, which contains a ubiquitous promoter, and two fragments, Ϫ289/Ϫ140 and Ϫ2226/Ϫ1191, silencing the transcriptional activity of the proximal promoter in non-endothelial cells. These findings suggest that the combination of these three domains may form the basis for the tissue-specific expression of VE cadherin gene.
In this report, we investigated the role of proximal cis-acting elements of VE cadherin promoter as well as the transcription factors that interact with these sequences. By means of gel shift assay and transient transfection of mutated forms of the Ϫ139/ϩ24 fragment, we determined three motifs required for full promoter activity. These elements represent the positive cis-acting partners of VE cadherin gene transcriptional mechanisms.

EXPERIMENTAL PROCEDURES
Electrophoretic Mobility Shift Assay (EMSA)-Nuclear extracts were prepared according to the method of Ohlson and Edlund (8) with the exception that cells were treated 5 min with 0.4% Nonidet P-40 instead of Dounce homogenization. The oligonucleotides were obtained from Genome Express (Grenoble, France). For EMSA, 0.5-0.9 ng of 32 Plabeled double strand oligonucleotide (2 ϫ 10 4 cpm) was mixed with 20 * 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 /EBI Data Bank with accession number(s) Y10887.
Cell Culture-Cell lines were obtained from the American Type Cul-ture Collection, except the mouse endothelioma cell line, eEND, which was obtained through the courtesy of E. F. Wagner. Bovine aortic endothelial cells (BAEC) were prepared by collagenase treatment as described previously (9) and were used from passages 3 to 12. All cells were maintained in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% (15% for BAEC) fetal calf serum (Seromed), 100 units of penicillin/ml, and 100 g of streptomycin/ml (Life Technologies, Inc.).
Generation of Mutations-The mutated Ϫ139CAT constructs were produced by site-directed mutagenesis as described by Mikaelian and Sergeant (10). Briefly, two simultaneous PCR reactions, using Ϫ139CAT 2 as matrix, were performed. The first one used two primers homologous to the pBLCAT3 vector (11) in which the Ϫ139/ϩ24 promoter fragment was inserted, 5Ј-AAGTTGGGTAACGCCAGGGT-3Ј from position 321 (11) and 5Ј-GAGCTAAGGAAGCTAA-3Ј from position 471, containing a mismatched 3Ј-end. The second one used the mutated primer described in Fig. 2 (mutation M2) or in Fig. 4C (mutations EBS2M, EBS4M, and EBS234M24), and a more external primer in the chloramphenicol acetyltransferase (CAT) gene, 5Ј-GGCATTTCAGT-CAGTTGCTC-3Ј from position 556 in pBLCAT3. Amplified fragments from each PCR reaction were purified, mixed, and subjected to a second round of PCR using the two most external primers (341 and 556). The amplified fragment was inserted into pCR2.1 vector (Invitrogen) and verified by sequencing. The Ϫ139/ϩ24 fragment was liberated by digestion with PstI and XhoI endonucleases (see Fig. 1) and inserted into the PstI and XhoI sites of pBLCAT3.
Transfection and CAT and Luciferase Assays-BAEC and NIH-3T3 cells were transfected by the calcium phosphate method (12). Each transfection was performed with 10 g of appropriate CAT construct and 5 g of luciferase reporter plasmid (pGL3-control, Promega) to correct variability in transfection efficiency. Forty-eight hours after transfection, cell extracts were prepared, and luciferase activity was determined on an aliquot with the luciferase assay system (Promega) on a luminometer (Pharmacia). The CAT assays were performed as described (13) with the equivalent of a given number of arbitrary light units of extracts as described in the figure legends. Data were expressed as the percentage of acetylated chloramphenicol relative to total chloramphenicol.

RESULTS
Features of the Ϫ139/ϩ24 VE cadherin Promoter Sequence-VE cadherin 5Ј-flanking sequence does not contain any canonical TATA, CAAT, or AP1 box usually encountered just upstream of the first exon ( Fig. 1 and Ref. 7). Nevertheless, a unique transcriptional start site was detected (7), indicating that a precise transcriptional initiation mechanism occurs. Moreover, the Ϫ139/ϩ24 fragment promoted a high level of reporter gene expression in both bovine aortic endothelial cells (BAEC) and the mouse fibroblastic cell line, NIH-3T3 (Fig. 3). 2 This activity was comparable to that of the herpes simplex thymidine kinase promoter in both cell types. To determine which elements within this fragment could support this activ- FIG. 1. Nucleotide sequence of the ؊139/؉24 region of the VE cadherin gene. All numbering is relative to the transcriptional start site (ϩ1). The sequence belonging to the first exon is in boldface. The PstI and XhoI sites delimiting the fragment are underlined, and the putative Sp1 (GT box) and Ets binding sites (GGAA or GGAT; EBS 3 is in an inverted orientation) are boxed.
ity, we searched for putative transcription factor binding sites. This analysis revealed six potential cis-acting elements as follows: a GT box between positions Ϫ49 and Ϫ39, which represents a recognition motif for Sp1 family members (20), although more classical Sp1 binding sites generally consist of GC boxes (21); and five consensus sites, GGAA or GGAT, for the Ets family of oncoproteins (22), localized at positions Ϫ72/Ϫ69, Ϫ93/Ϫ90, Ϫ102/Ϫ99, Ϫ109/Ϫ106, and Ϫ121/Ϫ118 and denoted EBS1 to EBS5, respectively.
Sp1 and Sp3 Transcription Factors Interact with the GT Box-To determine whether the GT box could bind transcription factors, we analyzed by electrophoretic mobility shift assay (EMSA) the interaction of a VE cadherin promoter oligonucleotide containing this motif, with nuclear proteins present in the murine endothelioma cell line eEND (23), which expresses high level of VE cadherin. Results showed that the DNA probe (Fig.  2B, WT) formed four complexes, designated I to IV, that were competed by excess amounts of unlabeled wild type oligonucleotide (Fig. 2, lanes 2 and 3). Complexes I-IV were competed by unlabeled oligonucleotides containing mutations outside of the GT box (Fig. 2, lanes 6 and 7); however, two oligonucleotides carrying mutations within the Sp1 site failed to prevent DNA-protein interactions (Fig. 2, lanes 4 and 5), indicating that all four complexes involved the GT box. Moreover, complexes I-IV were competed by a consensus Sp1 binding site (GC box) but not by an irrelevant oligonucleotide (Fig. 2, lanes 8 and 9), further suggesting the identity of binding factors as Sp1-like proteins. Identical patterns were observed with nuclear extracts from various cell types, including BAEC and NIH-3T3 (data not shown), suggesting that these transcription factors are ubiquitous. EMSA in the presence of monospecific antibodies to different members of the Sp1 family revealed that complex I was formed by Sp1 and complexes II and III by Sp3, as indicated by the decrease of the respective signal intensity and the apparition of supershifted complexes (Fig. 2, lanes 11 and  12). Complex IV was not affected by the presence of these antibodies or by antibodies against Sp2 (data not shown) or Sp4 (Fig. 2, lane 13). The identity of proteins present in complex IV remains to be determined.
The GT Box Is Critical for Promoter Activity-We investigated the functional importance of the Sp1/Sp3 site by mutational analysis. Mutant M2 (Fig. 2B), which could not support the formation of complexes with nuclear extracts in gel shift assay (data not shown), was introduced in the Ϫ139CAT construct containing the Ϫ139/ϩ24 VE cadherin promoter fragment driving the expression of the chloramphenicol acetyltransferase (CAT) gene. Transient transfection analysis of BAEC and NIH-3T3 cells showed that this mutation caused approximately 80% reduction in reporter activity in both cell types (Fig. 3), indicating that the GT motif is a major determinant in promoter activity.
EBS2 and EBS4 Interact with Nuclear Proteins-The potential interaction of the five consensus Ets binding sites with endothelial nuclear proteins was evaluated by EMSA. EBS1, EBS3, and EBS5 did not form any specific complex, indicating an absence of binding with nuclear proteins (data not shown), whereas DNA-protein interactions were observed with EBS2 and EBS4 (Fig. 4). With the EBS2 probe (Fig. 4A), a specific complex was formed and could be competed by unlabeled wild type EBS2 oligonucleotide (Fig. 4A, lanes 2 and 3) as well as oligonucleotides containing EBS4, EBS234 (Fig. 4A, lanes 5  and 8), and two functional Ets binding sites derived from other promoters (Fig. 4A, lanes 5 and 10). Conversely, oligonucleotides bearing a mutated EBS2, EBS1, EBS5, or an irrelevant sequence (Fig. 4A, lanes 4, 6, 7, and 11) were unable to compete with EBS2 for nuclear factor binding. Similar results were obtained with EBS4 (Fig. 4B), except that two specific complexes were observed, the slower migrating one (complex a) yielding a much stronger signal than the faster migrating band (complex b). These data establish the capacity of EBS2 and EBS4 to interact with Ets-related proteins. Cross-competition of EBS2 and EBS4 for complex formation (Fig. 4) as well as co-migration of complex a, generated by EBS4, with that obtained with EBS2 (Fig. 5), suggest that these complexes are formed by the same protein(s). EMSA, conducted with EBS2 or EBS4 and nuclear extracts from cells of endothelial (BAEC), fibroblastic (NIH-3T3), erythro-megakaryocytic (HEL), epithelial (HeLa), or hepatocytic (HepG2) origin, formed complexes co-migrating with those obtained with eEND extracts (Fig. 5). Of note, a weaker DNA binding activity was consistently observed with HeLa extracts. These data strongly suggest that the nuclear factors involved are not cell-specific.
EBS2 and EBS4 are Both Required for Optimal Function of VE Cadherin Promoter-To characterize the contribution of these elements to the promoter activity, we mutated EBS2 and EBS4 individually or in combination within the Ϫ139CAT construct. The mutations were known to disrupt any binding activity (Fig. 4B). When transfected in BAEC, mutations of EBS2 or EBS4 resulted in a 78% or an 88% decrease, respectively, in activity compared with the wild type construct; mutation of both sites reduced transcription to background level (Fig. 6). In NIH-3T3, mutations of EBS2, EBS4, or both provoked a 60 -70% decrease in promoter activity (Fig. 6). These data clearly demonstrate that both elements are important and equivalent for VE cadherin gene transcription. It is interesting to note that the loss of one site cannot be compensated by the other. Moreover, results with mutations in both sites show that transcriptional activity of the basal promoter in BAEC is completely  Fig. 2B) was introduced in the Ϫ139CAT construct. The wild type (WT) and the mutated (M2) constructs were transfected into BAEC or NIH-3T3 cells, and CAT activities were measured 48 h later. In each assay, a luciferase expression plasmid was cotransfected, and CAT assays were normalized according to the luciferase activity; 1500 arbitrary units were used for CAT assays with BAEC extracts and 3000 arbitrary units for NIH-3T3 extracts. The promoterless plasmid pBLCAT3 was used as a negative control and measured the experimental background. All data are the average of five to seven independent experiments. The standard deviations are indicated in the figure. dependent on EBS2 and EBS4, whereas these sites are less critical for gene expression in NIH-3T3. These variations in mutation susceptibility may indicate differential transcriptional mechanisms in BAEC and NIH-3T3.
Characterization of Ets Gene Expression in Endothelial Cells-To identify the Ets family members that could poten-tially regulate the VE cadherin gene through EBS2 and EBS4, we analyzed by Northern blotting the expression of various ets genes in endothelial cells (eEND and BAEC). We chose either genes that had already been shown to be expressed in human umbilical vein endothelial cells, such as ets-1, ets-2, and erg (24), or genes that were known to be relatively ubiquitously expressed such as ERM (17) and ER81 (25). The probes used in this study were derived from selected portions of the cDNAs (see "Experimental Procedures") to be specific for each gene, and we verified that the probes cross-hybridized between species (data not shown). Hybridization profiles are shown Fig. 7. All these genes were expressed in endothelial cells except ets-2, for which no signal could be detected even after a long exposure. Of note, low transcript levels of ets-2 were observed in human umbilical vein endothelial cells by Khachigian et al. (24). erg and ERM transcripts were detected at high levels, whereas ets-1 and ER81 genes were weakly expressed. These data are in agreement with those from Khachigian et al. (24) for ets-1 and erg expression in human endothelial cells; we extended this study to two other family members, ERM and ER81, that could also potentially be involved in endothelial genes expression.
Erg Binds to EBS4 in EMSA-To test whether the Ets members defined above could interact with the VE cadherin promoter, antibodies against these proteins were used in EMSA in addition to eEND nuclear extracts using either EBS2 or EBS4 probe. As shown in Fig. 8, two antibodies made against different parts of Erg blocked complex b formation with EBS4; the presence of anti-Erg antibodies did not affect the pattern observed with EBS2 probe (data not shown). These data suggest that Erg is one of the factors binding to VE cadherin promoter. None of the available antibodies against Ets-1, Ets-2, ERM, or ER81 was able to induce the formation of a supershifted complex or to disrupt DNA-protein binding in EMSA using EBS2 or EBS4 probe (data not shown), suggesting that another Etsrelated protein, possibly uncharacterized, participates in the formation of the major complex. DISCUSSION Promoting activity of the Ϫ139/ϩ24 fragment is the basis for VE cadherin gene transcription. Within this region, three boxes have been shown to be essential for promoter activity as follows: a GT box (positions Ϫ48 to Ϫ40) and two Ets binding sites FIG. 4. EMSA of eEND nuclear proteins binding to the EBS2 or EBS4 sites. A, the EBS2 probe (positions Ϫ100 to Ϫ83) (lane 1) was incubated with 20 g of eEND nuclear extracts (lanes 2-11). Competitions with the unlabeled oligonucleotides indicated at the top of A (lanes 3-11) were performed with 200-fold excess amounts. The specific complex is indicated by an arrow; a nonspecific band is also present. B, the EBS4 probe (positions Ϫ116 to Ϫ99) (lane 1) was incubated with eEND nuclear extracts (lanes 2-11). Two complexes were observed; the major band was designated complex a and the minor band was designated complex b. The competitors added are shown at the top of B (lanes  3-11). C, sequence of the double-stranded oligonucleotides used in A and B; the consensus Ets binding site (cons EBS) and the Ets binding site of the polyoma enhancer (EBS PyE) are derived from Gégonne et al. (39) and Monté et al. (17), respectively; the sequence of HNF1 is indicated in Fig. 2A.   FIG. 5. Binding properties of EBS2 and EBS4 are not cellspecific. Nuclear extracts from endothelial (eEND and BAEC), fibroblastic (3T3), erythro-megakaryocytic (HEL), epithelial (HeLa), or hepatocytic (HepG2) cells, incubated with EBS2 or EBS4 probes, displayed similar binding pattern in EMSA. Additional bands could be observed with HEL and HepG2 nuclear proteins.
(positions Ϫ93 to Ϫ90 and Ϫ109 to Ϫ106). Furthermore, our data suggest that these sites act through binding with specific nuclear proteins.
The GT box is a consensus binding site for the Sp1 family of transcription factors, which are expressed in various cell types and organs (20,26). In this paper, we show that the GT box present within VE cadherin promoter can interact with Sp1, Sp3, and possibly a third, yet unidentified, factor. In the case of TATA-less genes, Sp1 has been shown to be able to recruit TATA-binding protein and associated factors, thereby positioning the transcription initiation complex (27). The lack of TATA box upstream from the VE cadherin transcriptional start site and the presence of the GT box at proximity support the notion that Sp1 could be involved in VE cadherin transcription initiation. It is well documented that Sp1 activates transcription while Sp3 acts as a repressor in most cell types studied (28). The fact that both factors interact with VE cadherin basal promoter suggests that the GT box may be the target of a transcriptional regulation that would modulate VE cadherin expression. The relative concentration of Sp1 and Sp3 in endothelial cells may be a determining factor in the amount of VE cadherin gene expression.
Sp1 family proteins are known to be DNA bending factors, and it has been hypothesized that the induced conformational change in the DNA structure could facilitate the recruitment of distal DNA-bound transcription factors and the assembly of the transcription initiation complex via protein-protein interactions (29). The GT box, which occupies a key position in VE cadherin promoter, may play a role in the linkage of the upstream regulatory regions (Ϫ289/Ϫ140 and Ϫ2226/Ϫ1191) to the basic promoter. Sp1 has been shown to regulate the expression of other endothelial genes such as those coding for the platelet-derived growth factor B-chain (24), the platelet-derived growth factor A-chain (30), the endothelial nitric oxide synthase (31), the vascular cell adhesion molecule-1 (32), and KDR/flk-1 (33). Together with our data, this indicates that Sp1 plays an important role in the acquisition of the endothelial phenotype.
Our results provide evidence for the existence of two functional binding sites for the Ets family of transcription factors. Ets proteins are characterized by a unique DNA binding domain recognizing the core GGA(A/T) motif, although adjacent All the blots were subsequently hybridized to ␤-actin probe (see "Experimental Procedures" for description of the probes). Detection of radioactive signal was performed in a PhosphorImager, and signals shown for ets-1 and ER81 probes correspond to a 10-fold enhancement in intensity compared with those obtained with erg and ERM. No signal was observed with ets-2 probe (data not shown). The arrow shows the faint ER81 signal in eEND.
FIG. 8. Erg interacts with EBS4. eEND nuclear extracts were incubated with EBS4 alone, or in the presence of 1 l of antisera that recognize either Fli-1 and Erg or Erg, or 1 l of non-immune serum (N.I.) and were subjected to EMSA. Addition of either antisera disrupted complex b formation, strongly suggesting that Erg binds to EBS4. sequences may influence binding affinities. This extending family contains proteins involved in cell growth, differentiation, or transformation (22). Some factors are relatively ubiquitous (Ets-2, ER81, ERM, or GABP␣), and others are cellspecific such as those expressed in the hematopoietic lineage (Spi-1/PU.1, Spi-B, Elf-1, and Fli-1) or are developmentally regulated (Ets-1) (34). No endothelial specific Ets protein has been identified yet; however, expression of several ets genes has been detected by Northern blot analysis in cultured endothelial cells (ets-1, ets-2, erg, ERM, and ER81) (Ref. 24 and this paper). EMSA with anti-Erg antibodies strongly suggest that Erg, which is abundantly expressed in endothelial cells, can bind EBS4 and therefore regulate VE cadherin transcription. However, Erg does not appear to be the major Ets regulator of VE cadherin expression for the following two reasons: 1) the signal intensity of complex a is much higher than that of complex b (Erg-dependent), and 2) Erg does not form a complex with EBS2, whereas individual mutations have shown that the contribution of EBS2 and EBS4 in VE cadherin promoter activity is comparable. GABP␣ is another abundant and ubiquitous Ets factor that could potentially bind to VE cadherin promoter. However, we did not retain this possibility as the electrophoretic profiles obtained with EBS2 and EBS4 are not consistent with the typical two-band pattern due to heterodimeric interaction of GABP␣ with GABP␤ (35). Some other endothelial genes contain functional Ets binding sites in their promoter region (24, 36 -38), but attempts to identify the transacting factor(s) were not successful (24,36,37). Accordingly, characterization of the protein involved in complex a formation is underway, and its trans-acting ability on Ets-regulated endothelial genes will be evaluated.
It is well established that Ets factors bind DNA as monomer (except for GABP␣⅐GABP␤ complex) but can also associate with other unrelated transcription factors bound to their cognate motifs in the vicinity of the Ets binding site (34). In this respect, Ets-1 and Sp1 were shown to synergize for promoter activity of the long terminal repeat of human lymphotropic virus type 1 (39) and of the megakaryocytic ␣IIb integrin (40). The complete inhibition of CAT activity in transfected BAEC, observed when both Ets sites were mutated, supports the notion that a similar cooperation between the Ets-related factor and Sp1 may exist in the context of VE cadherin promoter.
In conclusion, using a mutational analysis and nuclear factor DNA binding assays, we have identified three essential cisacting elements in the proximal VE cadherin promoter. We showed that Sp1 and Sp3 interact with the GT box at positions Ϫ48/Ϫ40 and that a nuclear factor, possibly related to the Ets family, was involved in the major interaction with the two functional GGAA sites at positions Ϫ93/Ϫ90 and Ϫ109/Ϫ106.