Characterization of an Upstream Enhancer Region in the Promoter of the Human Endothelial Nitric-oxide Synthase Gene*

The endothelial nitric-oxide synthase gene is constitutively expressed in endothelial cells. Several transcriptionally active regulatory elements have been identified in the proximal promoter, including a GATA-2 and an Sp-1 binding site. Because they cannot account for the constitutive expression of endothelial nitric-oxide synthase gene in a restricted number of cells, we have searched for other cell-specific regulatory elements. By DNase I hypersensitivity mapping and deletion studies we have identified a 269-base pair activator element located 4.9 kilobases upstream from the transcription start site that acts as an enhancer. DNase I footprinting and linker-scanning experiments showed that several regions within the 269-base pair enhancer are important for transcription factor binding and for full enhancer activity. The endothelial specificity of this activation seems partly due to interaction between this enhancer in its native configuration and the promoter in endothelial cells. EMSA experiments suggested the implication of MZF-like, AP-2, Sp-1-related, and Ets-related factors. Among Ets factors, Erg was the only one able to bind to cognate sites in the enhancer, as found by EMSA and supershift experiments, and to activate the transcriptional activity of the enhancer in cotransfection experiments. Therefore, multiple protein complexes involving Erg, other Ets-related factors, AP-2, Sp-1-related factor, and MZF-like factors are important for the function of this enhancer in endothelial cells.

Several studies were aimed at characterizing the 5Ј-upstream sequences that drive transcriptional activity of the promoter up to 3500 bp. A major transcriptional effect was identified for Sp-1/Sp-3 and GATA-2 transcription factors, for which binding sites are located respectively at Ϫ103 and Ϫ230 bp upstream from the major transcription start site (9 -11). A second positive regulatory domain was detected between Ϫ140/ Ϫ120 bp in the proximal promoter (12). It was also shown that a 1600-bp human eNOS (heNOS) promoter fragment allows the endothelial expression of a reporter gene in transgenic mice but is not sufficient to observe a full expression of the gene in all endothelial territories and to reproduce the endogenous pattern of expression (13).
In this study, we characterized cis-acting sequences, located 4.9 kb upstream from the major transcription start site, that increase constitutive expression of the heNOS gene in endothelial cells. Analysis of the enhancer sequence by DNase I footprinting and linker-scanning mutants led us to identify five major binding sites, in particular for Erg and other Ets family member proteins, for Sp-1-related factors, and for MZF-like transcription factors.
Cell Culture-HMEC-1 are human dermal microvascular endothelial cells immortalized by transfection with a pBR322-derived plasmid containing the coding region for the simian virus 40 A gene product, large T antigen, and were a gift from Thomas J. Lawley (Emory University, School of Medicine, Atlanta, GA). Human umbilical vein endothelial cells were isolated as described by Jaffe et al. (14). Both cell types were cultured in MCDB-131 medium supplemented with 20% heat-inactivated fetal calf serum, 2 mM L-glutamine, 100 units/ml penicillin, 100 g/ml streptomycin, 10 ng/ml human recombinant epidermal growth factor, and 1 g/ml hydrocortisone. HeLa cells were cultured in RPMI medium supplemented by 10% heat-inactivated fetal calf serum, 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. Cells were maintained at 37°C and 5% CO 2 in an humidified incubator.
DNase I Hypersensitivity Assay-In situ DNase I digestion method was performed as described previously by Stewart et al. (15), using 0.2% and 0.05%, respectively, of Nonidet P-40 for permeabilization of HMEC-1 or HeLa cells. After phenol extraction, DNA samples (30 g) were subsequently submitted to Southern blot analysis. DNase I hypersensitivity assay was also performed on isolated nuclei with standard procedures (16).
DNA Subcloning and Deletions Mutants-A PCR product extending from 5Ј-upstream from the DNase I hypersensitive site HS1 to downstream from the initiator methionine codon, was obtained by PCR using a commercially available long PCR kit (Taqϩ Precision PCR system, Stratagene La Jolla, CA) and a human eNOS cosmid clone as template (17). The PCR product was subcloned between the NheI and HindIII * This work was supported by INSERM, by the Fondation de France, and by an unrestricted 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.
sites of the firefly luciferase reporter gene vector pGL3-basic (Promega, WI). For generation of deletion mutants, the construct was partially digested by exonuclease III/S1 nuclease using a commercial kit (Erasea-base system, Promega). All constructs were sequenced using Thermosequenase dye terminator cycle sequencing (Amersham Pharmacia Biotech) and an Applied Biosystems 373 DNA sequencer. Sense (pGL3-p2) and antisense (pGL3-p42) constructs containing the 269 bp activator element (Ϫ4907/Ϫ4638) in both orientations were generated by subcloning a 269-bp Pfu polymerase amplification product 5Ј to the 1703-bp promoter fragment of the heNOS gene. For point mutation generation, a commercial kit (Quick Change mutagenesis kit, Stratagene) was used with pGL3-p2 as template. Constructs and mutants were checked by sequencing.
Preparation of Protein Extracts-HMEC-1 nuclear extracts were prepared as described previously (21). For whole cell extract preparation, HeLa cells were seeded on a 2-cm 2 multidish plate and transfected with 2 g of relevant expression vectors or empty control vectors. After 40 h, the cells were harvested in 40 mM Tris-HCl, pH 7.4, 1 mM EDTA, 150 mM NaCl, collected by low speed centrifugation (800 ϫ g, 4°C), and then resuspended in 40 l of buffer containing 20 mM Tris-HCl, pH 7.4, 0.4 M KCl, 2 mM dithiothreitol, 10% glycerol. Cells were broken by freezing and thawing (three times), cell debris were removed by centrifugation (800 ϫ g, 10 min., 4°C), and the supernatant (whole cell extract) was aliquoted and stored at Ϫ70°C.
DNase I Footprinting Analysis-A fragment of the heNOS gene promoter region (Ϫ4907 to Ϫ4638) was obtained by PCR amplification using primers flanking the enhancer. For radioactive labeling, one primer (1.5 pmol) was end-labeled with [␥-32 P]ATP by T 4 DNA polynucleotide kinase (Life Technologies, Inc.) before a 20-cycle PCR amplification using 10 pmol of the second primer. DNase I footprinting analyses were performed by incubating 5,000 cpm of probe with 20 -30 g of nuclear extracts for 15 min, at room temperature, in 25 mM Hepes, 50 mM KCl, 0.1 mM EDTA, 5 mM MgCl 2 , 5 mM CaCl 2 , 0.5 mM phenylmethylsulfonyl fluoride, 1 mM dithiothreitol, and 10% glycerol before submission to increasing amounts of DNase I for 2 min. Control reactions were performed in the absence of proteins. Reactions were stopped by addition of stop buffer (200 g of proteinase K, 50 mM EDTA, 100 g/ml tRNA, and 1% SDS) and incubated 1 h at 50°C. Samples were phenol extracted, ethanol precipitated, and recovered in (98% formamide, 10 mM EDTA, 0.1% bromphenol blue, 0.1% xylene cyanol), denatured, and loaded, together with a GϩA Maxam-Gilbert sequence of the probe, on a 6% acrylamide, 8 M urea sequencing gel.
Linker-scanning Mutants Generation-Seven linker-scanning mutations were introduced in the pGL3-p2 construct. For each mutant, two amplicons, the 5Ј and the 3Ј, were synthesized using a PCR-based method with Pfu polymerase. Primers used to create the linker-scanning mutants are listed in Table I. The 5Ј amplicon was generated using two primers designated MluB10 and Brx. MluB10 is a gene-specific, sense primer that corresponds to the Ϫ4907/Ϫ4883 region of the heNOS promoter and has a Mlu I restriction site at its 5Ј end. Brx primers, where x corresponds to the location of the specific bases to be mutated, are 24 -29 base antisense primers containing a sequence identical to the promoter sequence linked to a mutated region at the 5Ј-end, heterologous to the heNOS sequence and containing a restriction site. The 3Ј amplicon was generated similarly using a common antisense primer HSFP3Ј (positions Ϫ4684/Ϫ4638) and different Scx primers containing a sequence identical to heNOS promoter and a 5Ј-end heterologous to the heNOS sequence and containing a restriction site identical to the corresponding Brx oligonucleotide. After restriction digestion and purification, each 5Ј amplicon was ligated to its corresponding 3Ј amplicon, and the ligation product was submitted to a further Pfu amplification reaction. The 269-bp mutated product was then subcloned in front of the 1703-bp promoter in pGL3-1703 construct. Clones obtained were checked by sequencing.
Electrophoretic Mobility Shift Assay-Oligonucleotides corresponding to DNase I protected regions were designed according to the sequence established previously (see Table II). Crude nuclear extracts (5-7 g) or whole cell extracts of cells transfected with expression vector (4 g) were incubated with approximately 0.15 pmol of [␥-32 P]ATP end-labeled double-stranded oligonucleotides in 20 mM Hepes, pH 7.9, 50 mM KCl, 3 mM MgCl 2 , 10% glycerol, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 1 g of poly(dI⅐dC), and 10 g of bovine serum albumin. Complexes were resolved by electrophoresis on prerun acrylamide:bisacrylamide (29:1) native gels. Supershift experiments were performed by preincubating nuclear or cell extracts with relevant polyclonal antiserum or purified polyclonal antibody prior to addition of labeled probes. Antibodies against MZF-1/1b or MZF-1b from J. Morris (Indiana University Medical Center, Indianapolis, IN), anti-NERF 1/1b/2 and anti-Erg 1/2 antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA.).
Computer Analysis and Statistical Analysis-Enhancer sequences were checked for transcription binding sites using Transfac data base and MatInspector 2.0 software (22). All transfection data are expressed as the means Ϯ S.E., quantitative results were analyzed by the Student's t test. Analysis of cotransfections experiments in HMEC-1 and HeLa cells were performed using a two-way analysis of variance, with a test of interaction.

Identification of Endothelial Specific DNase I-hypersensitive
Sites in the Promoter Region of the heNOS Gene-DNase I hypersensitivity experiments were performed on permeabilized cells as well as on isolated nuclei. Results obtained with permeabilized cells are shown, but similar results were obtained with the two methods. Using a probe located at the 3Ј edge of a HMEC-1 were permeabilized using 0.2% Nonidet P-40 and submitted in situ to increasing quantities of DNase I (0 -400 IU). After DNA purification and digestion by HindIII, Southern blot analysis using B13 probe shows three major DNase I digestion products respectively 9, 6, and 4 kb long, indicating the presence in endothelial cells of three hypersensitive sites located approximately 5, 1.9, and 0.9 kb from the major transcription start site. C, a similar analysis performed using HeLa cells permeabilized using 0.05% Nonidet P-40, shows no DNase I-hypersensitive sites in heNOS promoter.
large HindIII restriction fragment located in the 5Ј region of the heNOS gene (Fig. 1A), three hypersensitive sites were detected by Southern blot experiments in the heNOS gene in HMEC-1 (Fig. 1B). From the size of these fragments, it can be deduced that these sites are located approximately at 5 kb (HS1), 1.9 kb (HS2), and 0.9 kb (HS3) upstream from the major transcription start site (Fig. 1A). The distal hypersensitive site (HS1) does not correspond to any previously identified regulatory regions of the heNOS promoter. Using HeLa cells that do not express eNOS, as observed by reverse transcription-PCR (data not shown), none of these hypersensitive sites could be detected in similar experiments (Fig. 1C).
Functional Analysis of the HS1-hypersensitive Site-A 6-kb promoter fragment was obtained, using a PCR-based method, and cloned 5Ј to a luciferase reporter gene (pGL3-basic). After exonuclease III digestion of this plasmid (pGL3-6090), different deletion mutant constructs of the heNOS promoter were obtained and transfected into HMEC-1. As shown in Fig. 2A, transfections of HMEC-1 performed with pGL3-6090 construct showed a 5-fold increase of the promoter activity, compared with that obtained with the 1703-bp promoter fragment (pGL3-1703). Progressive deletions from Ϫ6090 to Ϫ4907 did not show significant reduction of transcriptional activity. Deletions from Ϫ4907 to Ϫ4638 returned transcription to the level obtained with pGL3-1703, indicating the existence of an activator element located between positions Ϫ4907 and Ϫ4638 of the promoter. Deletion of the promoter from Ϫ4638 to Ϫ1703 did not significantly influence the promoter activity (data not shown). Deletions from Ϫ1703 to Ϫ782 showed a slight decrease in the promoter activity (30%; data not shown). To investigate the enhancer characteristics of the distal activator element (Ϫ4907/Ϫ4638), it was cloned in both orientations immediately 5Ј to the 1703-bp promoter (pGL3-p2 and pGL3-p42) or 5Ј to an heterologous SV40 promoter (pGL3-SV40-Enh). Transfection experiments performed in HMEC-1 with pGL3-p2 (sense construct) and pGL3-p42 (antisense construct) showed a transcriptional activity similar to pGL3-4907 (Fig. 2B). Similarly, transfection experiments using pGL3-SV40-Enh showed an enhanced transcriptional activity as compared with the SV40 promoter (Fig. 2B). A similar pattern of transcriptional activation was observed with the different constructs in human umbilical vein endothelial cells (Fig. 2C), demonstrating simi-

FIG. 2. Functional analysis of the human eNOS promoter.
A, large scale analysis of the heNOS promoter using deletion mutants obtained by exonuclease III digestion of a vector containing 6090 bp upstream from the firefly luciferase coding sequence. HMEC-1 were transfected with the different deletion constructs and pRenilla plasmid as an internal control for transcription efficiency. After 24 -36 h, cells were harvested and firefly luciferase and renilla luciferase activities were measured using Dual-luciferase reporter assay system (Promega). B, transcriptional activities in HMEC-1 of the different plasmids containing the activator element cloned in front of the 1703-bp heNOS promoter or the heterologous SV-40 promoter (pGL3-SV40). Activities of the sense (pGL3-p2) and the antisense (pGL3-p42) constructs were compared with the activity of the native 4907-bp promoter. The 269-bp element is represented by a dark gray arrow. Similarly, transcriptional activity of the construct pGL3-SV-40-Enh was compared with promoter vector pGL3-SV-40. C, comparative activity of heNOS promoter in two endothelial cell types. HeNOS promoter deletion mutants and pGL3-p2 were transfected either in HMEC-1 or human umbilical vein endothelial cells. D, transcriptional activities in HeLa cells of different promoter fragments or pGL3-p2 compared with the 1703-bp promoter. Similar analysis comparing pGL3-SV40-Enh, containing the enhancer, to the native SV40 promoter. All data represent transcriptional activities relative to pGL3-1703 (mean Ϯ S.E., n ϭ 4, duplicates determination). *, p Ͻ 0.05, relative to pGL3-1703. lar mechanisms of activation in the two endothelial cell models. In contrast, transfection experiments using HeLa cells showed no transcriptional activation of the native promoter by the heNOS distal enhancer, because no difference in transcriptional activity was observed between pGL3-4638 and pGL3-4907 (Fig. 2D). However, the pGL3-p2 construct showed a 3-fold increase in transcriptional activity in HeLa cells compared with the 1703-bp promoter (Fig. 2D). A large increase of transcriptional activity was also observed when using the pGL3-SV40-Enh construct compared with pGL3-SV40 in HeLa (Fig. 2D).
In Vitro Analysis of the Enhancer Region of heNOS Promoter-DNase I footprint experiments were performed to identify regulatory elements involved in the heNOS distal enhancer activity. An end-labeled PCR-product corresponding to the 269-bp activator element of the promoter was used as a probe in footprint experiments with HMEC-1 nuclear extracts. As shown in Fig. 3, nuclear proteins protected the sense probe in four regions, named A, B, C, and D (Fig. 3A). The protected element A spans 8 bp and is limited in its 3Ј part by a DNase I-hypersensitive site. Element B spans approximately 15 bp, and element C spans 30 bp. Element D spans 6 bp and exhibits an hypersensitive site immediately downstream. Using an antisense probe, we observed protection over elements A, B, and C, but element D was not protected (Fig. 3A). To understand the role of each DNase I protected element, a systematic mutational analysis was conducted by mutating 8 -10 bp of each element (Table I). These linker-scanning studies allowed functional binding sites to be identified among the four DNase I protected elements. Five linker-scanning mutants (pGL3-p2-LSA, pGL3-p2-LSB2, pGL3-p2-LSC2, pGL3-p2-LSC3, and pGL3-p2-LSD) showed a clear, significant decrease in transcriptional activity compared with the pGL3-p2 wild type construct in HMEC-1 (Fig. 4). The mutants pGL3-p2-LSA, pGL3-p2-LSC2, and pGL3-p2-LSD conserved 30 -35% of the enhancer activity, whereas pGL3-p2-LSB2 and pGL3-p2-LSC3 no longer exhibited an enhancer effect. In contrast, the linkerscanning mutant pGL3-p2-LSB1 kept 60% of the enhancer activity, and the linker-scanning mutant pGL3-p2-LSC1 had no effect on enhancer activity. These data demonstrate the functional role of five regions of the enhancer in the regulation of basal heNOS expression.
Nucleoprotein Complexes Formed by Element A-A 34-bp oligonucleotide (FPA-WT) was designed to be centered on footprint element A, overlapping 15 bp of footprint B (Fig. 5A). Several DNA-protein complexes were observed when using a labeled FPA-WT probe in EMSA experiments (Fig. 5C, lanes 2 and 12). Addition of 10 -100-fold molar excess of cold competitor FPA-WT oligonucleotide, suppressed all the complexes (lanes 3, 4, 13, and 14). Computer analysis of the FPA-WT sequence revealed putative binding sites for several transcription factors located in element A, in particular for MZF-1 zinc finger transcription factors and for Krü ppel-like transcription factors family members (Fig. 5A). FPA-derived oligonucleotides were designed ( Fig. 5B and Table II). Oligonucleotides mutated for the entire A site (FPA-M) or for the MZF-1 consensus sequence 5Ј-TCCCCA-3Ј (FPA-m2) failed to compete for complexes formation (Fig. 5C, lanes 5, 6, 9, and 10, respectively), suggesting a major effect of this core element in the protein binding. An FPA-m1 oligonucleotide, mutated in the EKLF consensus sequence 5Ј-CACCC-3Ј, competed formation of all complexes (Fig. 5C, lanes 7 and 8). An oligonucleotide, bearing a consensus MZF-1 binding site (MZF consensus), with unrelated flanking sequences, prevented the formation of complex I (Fig. 5C, lanes 15 and 16). The same oligonucleotide with a 3-bp mutation in the MZF binding site (MZF consensus mutant) failed to compete (lanes 17 and 18). These results suggest that MZF-like factors bind element A, whereas EKLF factors do not.
Despite the fact that HMEC-1 cells seem to express MZF-1 (as seen with reverse transcription-PCR and Western blot analysis; data not shown) and MZF-1b (Western blot analysis; data not shown), we were unable to observe any supershift using either the MZF-1/1b antibody or an antibody specifically targeted against the N-terminal domain of MZF-1b (data not shown). Similarly, coexpression of MZF-1 expression vector with pGL3-p2 plasmid had no effect on the enhancer transcriptional activation (data not shown).
Nucleoprotein Complexes Formed by Element C-Footprint element C was functionally restricted to base position Ϫ4755/ Ϫ4735 by linker-scanning analysis (linker-scanning mutants LSC2 and LSC3). A labeled FPC 1-3 -WT probe, containing the total footprint element C sequence (Ϫ4765/Ϫ4735), formed multiple complexes in the presence of crude nuclear extracts (Fig. 6B, lanes 2 and 8), among which several were specifically competed by 100-fold molar excess of cold FPC 1-3 -WT oligonucleotide (Fig. 6B, lanes 3 and 9). Computer analysis of Ϫ4755/ Ϫ4738 region revealed putative binding sites for AP-2, Sp-1, MZF-1, and Ets. Competition experiments performed using an AP-2 consensus oligonucleotide showed competition for complex IV formation (Fig. 6B, lane 6). In addition an Sp-1 consensus oligonucleotide competed for complexes IV-V formation (lane 5). Finer analysis of the sequences of AP-2 and Sp-1 consensus oligonucleotides revealed putative Sp-1 binding sites within the AP-2 oligonucleotide. Inversely the Sp-1 oligonucleotide had no binding site for AP-2. This suggests that the competition of AP-2 oligonucleotide for complex IV formation may be in fact Sp-1dependent. Therefore, the relative contribution of AP-2 and Sp-1 for complex IV formation should be further investigated.
MZF-1 consensus oligonucleotide exhibited a clear competition for complex I (lane 4), suggesting the binding of an MZFlike protein. Lastly, an FPC 1-3 -WT oligonucleotide bearing a mutation in the Ets-binding site still competed away all complexes with efficiency equivalent to that of the wild type FPC 1-3-WT oligonucleotide (lane 10). Altogether, these results suggest an implication of Sp-1-related and MZF-like transcription factors in the formation of DNA-protein complexes in element C but not of Ets family members.
Nucleoprotein Complexes Formed by Element B and D-Incubation of the labeled FPB2-WT probe centered on the region B2 (Fig. 7iA) with HMEC-1 nuclear extracts resulted in a DNA-protein complex formation (Fig. 7B, lane 2), which was completely prevented with cold FPB2-WT competitor (Fig. 7B,  lane 3). Computer analysis of FPB2-WT revealed a putative Ets-binding site possibly involved in the complex formation. Indeed, we observed a clear competition using a 100-fold molar excess of a consensus Ets-1 oligonucleotide (Fig. 7B, lane 5), whereas an FPB2 oligonucleotide, mutated for the Ets-binding site (FPB2m) failed to prevent the complex formation (Fig. 7B,  lane 4). Using antibodies directed against Erg 1/2 transcription factors or NERF 1/1b/2 factors, we did not observe any supershift on DNA-protein complexes formed by HMEC-1 nuclear extracts and FPB2-WT probe (data not shown), suggesting that neither Erg nor NERF transcription factors were present in DNA-protein complexes.
The same study performed with an FPD-WT probe centered on the functional element D resulted in the formation of one major specific complex and several minor complexes (Fig. 8B,  lanes 3 and 5). Computer analysis of FPD-WT revealed also an Ets-binding site. Consensus Ets-1 oligonucleotide showed a clear competition for complex formation (Fig. 8B, lane 7), whereas FPD oligonucleotide, mutated for the Ets-binding site (FPDm), failed to compete (lane 6). Therefore, these results suggest the binding of Ets-related proteins to the Ets-binding sites located in positions Ϫ4775 (region B2) and Ϫ4688 (region D).
Supershift experiments performed with FPD-WT probe us-  (Table I). Boxes represent DNase I protected elements A-D. Black crosses represent linker-scanning mutations. Sequence of mutations are described in Table I. Transcriptional activities of enhancer construct pGL3-p2 and of the different linker-scanning mutants were compared with that of pGL3-1703 promoter. Data are expressed as the means Ϯ S.E. of five independent experiments (duplicate determination). *, p Ͻ 0.05, relative to pGL3-p2.

5Ј-CTAGGCTAGCGCGGGGAAAGGGAAGGGAG-3Ј WT 5Ј-GCTCCCCACCCGTCT-3Ј
Sc ing an antibody specifically directed against NERF 1/1b/2 factors did not allow supershifted bands to be observed (data not shown). In contrast, an antibody specifically targeted against Erg 1/2 transcription factors abolished formation of a minor complex formed with FPD-WT probe and nuclear extract (Fig.  8B, lane 4).
Complementary data were obtained with HeLa cells transfected with several expression vectors encoding different Etsrelated transcription factors. Only whole cell extracts of HeLa cells transfected with pSG5-p55-Erg and pCMV5-Elk-1 expression vectors led to specific DNA-protein complexes in EMSA experiments using FPB2-WT (Fig. 7B, lanes 2 and 3) or FPD-WT probes (Fig. 8C, lanes 3 and 4). Cotransfection experiments in HMEC-1 showed a slight potentializing effect of Erg transcription factor on transcriptional activity of the enhancer compared with an empty vector (Fig. 9A). Elk-1 overexpression led to a pronounced inhibition of transcriptional activity (Fig.  9C) when coexpressed at a 1/1 ratio with pGL3-p2 but had no effect when used at a lower ratio (1/50) (data not shown). Similar experiments using Ergexpression vector, and pGL3-p2 in HeLa cells showed a significant effect of Erg on unidentified Ets-binding sites located in the first 1703 bp of the promoter sequence (Fig. 9B, lanes 1 and 2). It is therefore difficult to assess the effect of p55Erg on the enhancer (Fig. 9B). However, the effect of Erg, as tested by a test of interaction, was significant both in HMEC-1 and HeLa cells (Fig. 9, A and B, 5. Identification of multiple DNA-protein complexes in footprint element A. A, localization of footprint element and functional region A as determined by footprint (black) and linker-scanning analysis (gray). The sequence protected in DNase I footprint analysis is boxed, and the region mutated by linker-scanning is underlined. B, three oligonucleotides derived from FPA wild type oligonucleotide (FPA-WT) were designed. They contain a complete element A mutation (FPA-M), a mutation abolishing the Krü ppel-like binding site core element 5Ј-CACCC-3Ј (FPA-m1), or a mutation abolishing MZF-like binding site core element 5Ј-TCCCCA-3Ј (FPA-m2). C, labeled FPA-WT oligonucleotide was incubated with 5 g of crude nuclear extracts from HMEC-1 for 15 min at 4°C. Complexes were resolved by electrophoresis in 8% acrylamide, 0.5ϫ TBE native gel,. Competition assays were performed to identify specific nucleo-protein complexes, using different oligonucleotides, from a 10-fold (lanes 3, 5, 7, 9, 13, 15, and 17) to a 100-fold (lanes 4, 6,8,10,14,16, and 18) molar excess. Oligonucleotides used are defined in Table II. Specific DNA-protein complexes (I-IV) are indicated by arrows.

FIG. 6. Identification of complex DNA-protein interactions in
Footprint C element. A, localization of the different linker-scanning regions C1, C2, and C3 (gray underlining) along footprint element C (black box). B, the FPC 1-3 WT probe (Ϫ4765/Ϫ4731) was incubated with 5 g of HMEC-1 nuclear extracts for 30 min at room temperature in the presence of 0.01% Nonidet P-40 and then submitted to electrophoresis in 7% acrylamide, 0.5ϫ TBE native gel. Different competition studies (lanes 3-6, 9, and 10) were performed using a 100-fold molar excess of indicated cold oligonucleotides (oligonucleotides sequences are in Table  II). Arrows indicate specific nucleo-protein complexes that are named I-V. ATTCGATCGGGGCGGGGCGAG tion factors could participate in the regulation of the heNOS gene expression by binding to the two Ets-binding sites identified in B2 and D. To validate this hypothesis, mutations of Ets-binding site located in position Ϫ4775 (B2) or Ϫ4688 (D) were introduced by a PCR-based mutagenesis, and enhancer transcriptional activity was measured by transfection. Fig. 10 shows that independent disruptions of Ets site in position Ϫ4775 (pGL3-p2-Ets 4775 mut) or Ets site located in position Ϫ4688 (pGL3-p2-Ets 4688 mut) led to a 50% decrease of activity compared with the wild type enhancer. The effects of these point mutations were significantly weaker than the corresponding linker-scanning mutations (pGL3-p2-LSB2 for pGL3-p2-Ets 4775 mut and pGL3-p2-LSD for pGL3-p2-Ets 4688 mut, respectively), which totally abolished enhancer activity. These data suggest that Ets-related transcription factors and possibly Erg participate in the activity of the distal enhancer of heNOS. Endothelial Specificity of the Transcription Factor Interactions-To test the hypothesis of an endothelial specificity of transcription factors implicated in the enhancer function, we performed EMSA experiments using HeLa nuclear extracts and the different EMSA probes, and we compared the migration patterns with those obtained using HMEC-1 nuclear extracts. No difference could be observed using the two nuclear extracts and the FPA-WT, FPB2-WT, and FPC 1-3 -WT probes (data not shown). However, the Erg-related complex observed using FPD-WT probe was not present when using HeLa cell nuclear extracts, even at longer exposure (Fig. 11).

DISCUSSION
To identify new cis-regulatory elements of the heNOS gene, we surveyed 16.5 kb of the 5Ј promoter region for the presence of DNase I-hypersensitive sites that are frequently associated with open transcriptionally active chromatin (23). Three hypersensitive sites (HS1, HS2, and HS3) were observed specifically in HMEC-1 cells and were absent in HeLa cells, which do not express heNOS, suggesting that they may reflect endothelial specific regulation of heNOS gene expression. HS1 is located 4.5-5 kb upstream from the major transcription start site, and deletion mutants supported the hypothesis that this hypersensitive site corresponds to a major transcriptional regulatory element. Indeed, the results obtained with the deletion mu- tants show that a transcription activator is localized between positions Ϫ4907 and Ϫ4638. The HS2 site, located at position Ϫ1.9 kb, does not correspond to a regulatory region identified so far. Furthermore, we did not observe any modification of the transcriptional activity of the promoter when this region was present or absent. The HS3 site, located at Ϫ900 bp, may correspond to a previously identified functional region of the promoter, located between Ϫ1033 and Ϫ779 bp (9 -11) and also detected using our deletion mutants between positions Ϫ1703 and Ϫ782 (data not shown). Surprisingly the main positive regulatory domain of the basal promoter, located Ϫ104/Ϫ95 bp and involving an Sp-1 binding site (9,12) was not detected by DNase I hypersensitivity experiments.
Deletion studies allowed us to identify a 269-bp activator region spanning from nucleotides Ϫ4638 to Ϫ4907 with respect to the transcription start site. This activator behaves as a classical enhancer acting independently of its position and orientation and also functioning when placed upstream from an heterologous SV-40 promoter. Interestingly, the 269-bp fragment showed a lower enhancer activity in HMEC-1 when transposed close to the SV40 promoter compared with its enhancer activity when placed 5Ј to the 1703-bp heNOS promoter, suggesting specific interactions between cis-acting sequences located in the enhancer and the regulatory sequences contained by the 1703-bp promoter. Transfection data obtained in HeLa cells showed an endothelial specificity of interactions between the enhancer and the promoter region because no difference in transcriptional activity was detected when the enhancer was normally positioned between positions Ϫ4907 and Ϫ4638. However, transcriptional activity of the enhancer, when moved directly upstream from the 1703-bp heNOS promoter (pGL3-p2 construct) or when placed in front of the heterologous SV-40 promoter, is detectable in HeLa cells. This suggests that interactions may exist between transcription factors present in HeLa cells and the enhancer in its non-native configuration. It also suggests that the endothelial specificity of the enhancer in its native position may be due to the existence, between positions Ϫ1703 and Ϫ4638, of inhibitory sequences active in nonendothelial cells or of a structurally important region, opening the chromatin specifically in endothelial cells.
We identified four sites (A, B, C, and D) of DNA-protein interactions by DNase I footprinting experiments within the 269-bp region, with A and D sites showing the most pronounced protection. Among the seven linker-scanning mutants created to analyze these sites, four decreased the enhancer activity from 25 to 60% (B1, A, C2, and D), and two others (B2 and C3) nearly abolished the transcriptional activation. Only the mutant C1 did not modify the transcriptional activity of the 269-bp region, demonstrating the absence of regulatory domain in this sequence. These results indicate that the 269-bp region contains multiple transcriptionally active elements probably interacting in a complex fashion.
Our data and computer analysis suggest that two half-binding sites for MZF transcription factors could be present in the 269-bp activator element, in agreement with previous studies showing that MZF transcription factors transactivate promoters through two half-sites (24,25). Indeed, we identified a MZF-1 consensus binding sequence in footprint A and in region C2 (footprint C), whose binding specificity were confirmed by competition in EMSA experiments using an oligonucleotide containing an unrelated MZF-1 binding site (26). It is also striking that the two linker-scanning mutants disrupting each MZF half-site in the 269-bp enhancer (mutants LSA and LSC2) show a similar decrease in transcriptional activity, suggesting FIG. 9. Cotransfection experiments using expression vectors in endothelial or nonendothelial cells. A, endothelial cells were cotransfected either with the heNOS promoter construct with or without the enhancer (pGL3-1703 or pGL3-p2) and the expression vector alone (pSG5) or containing the Erg cDNA (pSG5-p55-Erg). Transcriptional activities were measured. Data represent the means Ϯ S.E. of four independent experiments. *, p Ͻ 0.05. ns, not statistically significant. B, Elk-1 inhibits the enhancer activity in HMEC-1. pCMV5-Elk-1-FLAG or pCMV5-FLAG were cotransfected with pGL3-1703 or pGL3-p2, at 1/1 ratio (n ϭ 2). *, p Ͻ 0.05. C, HeLa cells were cotransfected either with the heNOS promoter construct with or without the enhancer (pGL3-1703 or pGL3-p2) and the expression vector alone (pSG5) or containing the Erg cDNA (pSG5-p55-Erg). Transcriptional activities were measured. Data represent the means Ϯ S.E. of three independent experiments. *, p Ͻ 0.05. Two-way analysis of variance was performed. activity of the Ets core binding site mutations (pGL3-p2-Ets4775 mut, pGL3-p2-Ets4688 mut) and of the corresponding linker-scanning mutants (respectively pGL3-p2-LSB2 and pGL3-p2-LSD) were measured and compared with the 1703 bp promoter activity. Full enhancer activity was also measured (pGL3-p2). Data represent the means Ϯ S.E. of four experiments. *, p Ͻ 0.05, relative to pGL3-p2. that the full activation requires an intact two-half-site structure. The long distance between these half-sites (50 bp) is consistent with a recent model for another two zinc finger domains transcription factor whose binding sites are separated by 44 bp (27) and also with the identification, in the Factor XIII promoter, of two active MZF-1 half sites almost 140 bp apart (25). However, using antibodies directed against these MZF isoforms, we did not observe any supershift or modification of the complexes formed with probe A. In addition, cotransfections of an MZF-1 expression vector did not show a transactivating effect of MZF-1 on the enhancer activity (data not shown). Therefore, although our data support the implication of an MZF-like factor, based on consensus sequence homology and competition experiments, they do not favor a direct role for MZF-1 or MZF-2 in the enhancer activity. However, we can hypothesize that an MZF-related factor is at play, as an example, MZF-3, an MZF-related transcription factor, which was recently described in the mouse (GenBank TM accession number AF082568). Its potential implication remains to be evaluated when more information is available.
Region C2 is likely to bind Sp-1-related factor, because complexes are competed away with an Sp-1 consensus oligonucleotide. Sp-1 binding site is also a major regulatory element of the proximal heNOS promoter for constitutive (9 -12) and inducible expression, such as with lysophosphatidylcholine (28).
Along the enhancer region, two functional Ets-binding sites, located at positions Ϫ4775 and Ϫ4685, were identified by linker-scanning analysis and site-directed mutagenesis of the core consensus sequence. Endothelial cells express several members of the Ets family of transcription factors under basal or activated conditions (20, 29 -31). Among all the different Ets factors that we have tested, Erg is the only Ets factor for which we found evidence of involvement. First, EMSA experiments using extracts from transfected HeLa cells show that only Erg and Elk-1, but not Ets-1, Ets-2, or Elf-1, are able to bind B2 and D element through their binding site. In addition, an anti-Erg antibody inhibits the binding of Erg to element D, whereas we were not able to detect any supershift or inhibition in experiments using a probe corresponding to element B2. Data obtained by cotransfections in HMEC-1 with various Ets expression vectors showed that only Erg could slightly transactivate the enhancer activity. This activation was also observed in HeLa cells, but in this case it was very difficult to estimate the quantitative effect of Erg on the enhancer because of its effect on the promoter. Using similar cotransfection experiments, Elk-1 strongly inhibits the enhancer activity when large amounts of vector were cotransfected, whereas Ets-1, Ets-2, or Elf-1 did not show any effect at any cotransfection ratio. The Elk-1-induced inhibition seems to be due to the expression of high amounts of Elk-1 (as observed in EMSA experiments), which suggests that it is the consequence of a strong competition of Elk-1 for native HMEC-1 Ets-related factors that bind B2 or D sites. Among the other Ets-related factors tested, Ets-1 and Ets-2 transactivated only the proximal promoter of heNOS (data not shown), as was shown in previous studies (12).
Taken together, these results support the hypothesis that Erg-like factors could participate in the heNOS enhancer activity, although other still unidentified Ets factors are likely to be also involved. Supershift experiments using element D probe showed that a single faint complex was displaced by an anti-Erg antibody, whereas several other complexes were not affected. These complexes likely involve Ets-like factors because they are competed away by an oligonucleotide bearing the Ets consensus sequence. However, none of these could be detected in the complexes by using an antibody against several Ets factors. In contrast, an Ets factor does not seem to be implicated in region C3, although it contains an Ets core consensus sequence.
When results obtained using constructs where mutations are restricted to the core Ets site are compared with results obtained using corresponding linker-scanning mutants that involve flanking nucleotides (pGL3-p2-LSB2 and pGL3-p2-LSD), the transcriptional effect of the latter appears more pronounced. It is therefore likely that linker-scanning mutations affect the binding ability of proteins adjacent to the core Ets site.
Interestingly, using a 5.2-kb fragment of the mouse eNOS promoter driving the ␤-galactosidase expression, Teichert et al. (32) obtained transgenic mice lines in which the expression pattern was similar to that of the endogenous gene. If regulatory regions of the eNOS gene are conserved between mice and humans, this could imply that the distal enhancer we have identified participates in vivo in the control of both the quantitative level and the restricted pattern of expression. The absence of this enhancer in the shorter human promoter construct (1600-bp 5Ј-flanking sequence) used by Guillot et al. (13) could conversely explain the incomplete pattern of expression observed in this study.
In conclusion, an enhancer was identified at distance from the transcription start site. Complex interactions between several elements located in the activator region can be inferred from our results. Among all transcription factors tested, Erg was the single transcription factor able to bind the enhancer and transactivate the promoter. Other important transcription factors involved in the transactivating activity and probably belonging to the MZF and Ets family remain to be identified.