Identification of hypoxia-response element in the human endothelial nitric-oxide synthase gene promoter.

The human endothelial nitric-oxide synthase gene (heNOS) is constitutively expressed in endothelial cells, and its expression is induced under hypoxia. The goal of this study was to search for regulatory elements of the endothelial nitric-oxide synthase (eNOS) gene responsive to hypoxia. Levels of eNOS mRNA, measured by real time reverse transcriptase-PCR analysis, were increased, and heNOS promoter activity was enhanced by hypoxia as compared with normoxia control experiments. Promoter truncation followed by footprint analysis allowed the mapping and identification of the hypoxia-responsive elements at position -5375 to -5366, closely related to hypoxia-inducible factor (HIF)-responsive element (HRE). To test whether known HIF-1 and HIF-2 are involved in hypoxia-induced heNOS promoter activation, HMEC-1 and HUVEC were transiently transfected with HIF-1alpha and HIF-1beta or HIF-2alpha and HIF-1beta expression vectors. Exogenous HIF-2 markedly increased luciferase reporter activity driven by the heNOS promoter in its native location. The induction of luciferase was conserved with the antisense construct and was increased in cotransfection experiments when this fragment was cloned 5' to the proximal 785-bp fragment of the eNOS promoter. Deletion analysis and site-directed mutagenesis demonstrated that the two contiguous HIF consensus binding sites spanning bp -5375 to -5366 relative to the transcription start site were both functional for heNOS promoter activity induction by hypoxia and by HIF-2 overexpression. In conclusion, we demonstrate that heNOS is a hypoxia-inducible gene, whose transcription is stimulated through HIF-2 interaction with two contiguous HRE sites located at -5375 to -5366 of the heNOS promoter.

diffuses to vascular smooth muscle cells and stimulates guanosine 3Ј,5Ј-cyclic monophosphate formation, which leads to vasodilation. The enzymatic production of NO in endothelial cells relies on the constitutively expressed eNOS whose activity is regulated by two major pathways, calcium/calmodulin binding or phosphorylation by AKT (4,5). In addition, different physiological and pathological conditions can modulate the level of eNOS expression in endothelial cells (6) by acting at both the transcriptional (7,8) and post-transcriptional levels (9 -11).
In response to hypoxia, systemic arteries vasodilate, allowing more blood to be delivered to peripheral tissues. This vasodilation occurs within seconds after hypoxia and is maintained for hours (12). The molecular basis of hypoxic vasodilation is not fully understood. Possible mechanisms include direct relaxation of vascular smooth muscle cells induced by changes in pH and ion channel conductance or a decrease in ATP levels (13). Endothelial cells also reduce the amount of released vasoconstrictors such as endothelin or thromboxane and release increased amounts of vasodilators such as adenosine, prostacycline, endothelial hyperpolarizing factor, and NO (14,15). Various effects of hypoxia on eNOS expression and NO synthesis in endothelial cells have been described. In some studies the activity (16) and expression (17)(18)(19)(20)(21) of eNOS have been shown to be increased by hypoxia, whereas in others it was found to be either decreased (22,23) or not modified (24). The discrepancies between these studies could be explained first by differences in the duration of exposure to hypoxia, and second by the differences in endothelial cell types used. The decrease observed in eNOS mRNA levels during prolonged hypoxia could be due, at least in part, to mRNA destabilization by Rho kinase because it is inhibited by a Rho kinase inhibitor (25).
No known hypoxia-sensitive element has been reported previously in the eNOS promoter (1600 bp). In two previous studies, a hypoxia-induced up-regulation of the eNOS gene expression in endothelial cells (19,20) was reported but without demonstration of an involvement of an HIF-1 related pathway. In the first study, the 5Ј-region flanking the eNOS gene (1600bp) used in transfection experiments for demonstrating hypoxia response does not contain any HIF-1 consensus sequence (19). The second study showed that the enhanced eNOS expression observed is because of a redox-sensitive AP1-mediated transcription (20).
In the present study, we searched for other elements of the eNOS promoter responsible for transcriptional induction under hypoxia. We identified a hypoxia-response element at position Ϫ5375 to Ϫ5366-bp of the heNOS promoter, and we show that transcription factor HIF-2 preferentially binds this response element under hypoxia and activates eNOS transcription.
Cell Culture-HMEC-1 cells 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) (26). Human umbilical vein endothelial cells (HUVEC) were isolated as described by Jaffe et al. (27), and cells were used at passage 2 and 3. Both cell types were cultured in MCDB-131 medium supplemented with 10% 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 with 10% heat-inactivated fetal calf serum, 2 mM L-glutamine, 100 units/ml penicillin, 100 g/ml streptomycin. Cells were maintained at 37°C and 5% CO 2 in an humidified incubator.
Hypoxia Treatment-Cells were incubated in a temperature and humidity-controlled environmental chamber IG750 (Jouan) in an atmosphere containing 2% O 2 , 5% CO 2 , 94% N 2 . Hypoxia was also mimicked by treatment with 130 M of iron chelator DFO.
Extraction of RNA and cDNA Synthesis-For preparation of RNA, total RNA was extracted from cells using Trizol® reagent (Invitrogen) according to the manufacturer's instructions. After isolation the RNA was used for cDNA synthesis using 0.5 g of total RNA from each sample, pd(N) 6 random hexamer (Amersham Biosciences) and dNTPs (1 mM), 200 units of Moloney murine leukemia virus-reverse transcriptase, and 20 units of RNase inhibitor. Reverse transcription was carried out for 50 min at 37°C and stopped by heating to 65°C for 5 min.
Real Time PCR Analysis-The levels of eNOS transcript were determined by real time reverse transcriptase (RT)-PCR using SYBR Green chemistry (28). To normalize for input load of cDNA between samples, Ribosomal Protein L32 (RPL32) was used as an endogenous standard. Specific primers were designed with Primer3 (www-genome.wi.mit. edu/cgi-bin/primer/primer3_www.cgi): eNOS1674S, 5Ј-TGTATGGATG-AGTATGACGTG-3Ј; eNOS1834R, 5Ј-TGTTCCGGCCGAGGG-3Ј; RPL-32forward, 5Ј-CCCAAGATCGTCAAAAAGA-3Ј; and RPL32 reverse, 5Ј-TCAATGCCTCTGGGTTT-3Ј. cDNA corresponding to 50 ng of RNA was added to the SYBR-Green JumpStart Taq Ready Mix (0.375 M of each specific primer, 0.2 l of internal reference dye, and 10 l of SYBR-Green JumpStart Taq Ready Mix) in a total volume of 20 l. PCRs were carried out in a real time PCR cycler (Mx4000, Stratagene). The thermal profile was 95°C for 2 min, 40 cycles of 95°C for 30 s, and 60°C for 1 min. At the end of each phase at 60°C, fluorescence was measured and used for quantitative purposes. eNOS mRNA expression data were normalized to RPL32 content, and relative transcript level between two samples (1 and 2) was calculated using the formula: 2 (Ctenos1-CtRPL1)Ϫ(Ctenos2-CtRPL2) . Independent experiments were performed with different HUVEC preparations derived from different subjects, and then for each preparation two reverse transcriptions were performed, and two PCRs with each reverse transcription were done.
DNA Constructs and Site-directed Mutagenesis-A 6047-bp fragment of the heNOS promoter (numbered according to GenBank TM accession number AC092466 sequence) and deletion mutants were cloned upstream from the luciferase gene (29) and used in transient transfection experiments. As a control reporter plasmid, the oligonucleotide HRE of the erythropoietin gene 5Ј-GGAGCTTGCCCTACGTGCTGTCT-CAG-3Ј (30) was subcloned 5Ј to the luciferase reporter gene driven by the SV40 promoter (vector pGL3-SV40, Promega), creating a construct containing four putative hypoxia-response element (HRE) motifs (HRE-pGL3-SV40).
Other constructs were generated by sub-cloning the corresponding Pfu polymerase amplification product 5Ј to the 1704-or 785-bp promoter fragments of the heNOS gene. The fragment Ϫ5569 to Ϫ5005 was subcloned immediately 5Ј to the 1704-bp promoter and 5Ј to the 785-bp promoter. Reporter plasmids with a 565-bp fragment (Ϫ5569 to Ϫ5005) divided up into four overlapping parts (part A, Ϫ5569 to Ϫ5395; part B, Ϫ5434 to Ϫ5305; part C, Ϫ5338 to Ϫ5153; and part D, Ϫ5203 to Ϫ5005) were also generated.
The putative eNOS HRE 41-bp sequence was inserted upstream the SV40 promoter driving the luciferase reporter gene (heHREpGL3SV40).
Mutations were introduced into the pGL3-6047 plasmid, using an oligonucleotide-directed mutagenesis system (QuickChange Site-directed Mutagenesis kit, Stratagene) according to the manufacturer's instructions. Constructs and mutants were sequenced using the DY-Enamic TM ET Terminator cycle sequencing kit (Amersham Biosciences) and an Applied Biosystems 377 DNA sequencer.
Transient Transfection Assays-Cell transfections of plasmid DNA were carried out using polyethyleneimine suspension in a commercially available solution (EXGEN 500). Briefly, HUVECs were plated on a 6-well plate at 2.5 ϫ 10 5 cells/well, incubated for 24 h, and then incubated for 4 h by adding to 3 ml of fresh medium, 100 l of transfection mix containing 0.37 pmol of the relevant reporter gene vector, 0.37 pmol of pRenilla luciferase gene vector (Promega) as transfection standardizing control, and 20 l of EXGEN 500 in 100 mM NaCl. HMEC-1 and HeLa cells were seeded on 2-cm 2 multidish plates at 5 ϫ 10 4 cells/well, incubated 24 h, and then incubated for 4 h by adding to 1 ml of medium 100 l of transfection mix containing 0.13 pmol of the relevant reporter gene vector, 0.13 pmol of pRenilla luciferase gene vector (Promega), and polyethyleneimine suspension (4 l for 2 g of total DNA). Coexpression experiments were performed using expression vectors containing cDNA for various transcription factors, pcDNA3.1-HIF-2␣ and pcDNA3.1-HIF-1␤ or pcDNA3.1-HIF-1␣ and pcDNA3.1-HIF-1␤. Similar experimental procedures were used, except that relevant expression vectors were included, in polyethyleneimine suspension at a molar ratio reporter vector/expression vector equal to 10:3. Luciferase activity was measured by luminometry using the dual luciferase revelation system (Promega) as described by the manufacturer. Luciferase signals were normalized for transfection efficiency to the signals obtained with the Renilla luciferase reporter driven by the SV40 promoter.
Preparation of Nuclear Protein Extracts-HMEC-1 or HUVEC were grown to subconfluence and then treated or not with 130 M DFO for 2 h before cell lysis. For hypoxic cell extracts, cell lysis was performed with the cells still under hypoxia to avoid reoxygenation effects. Nuclear extracts were prepared as described previously by Khachigian et al. (31).
DNase I Footprinting Analysis-The Ϫ5569 to Ϫ5305 heNOS promoter fragment was obtained by PCR. For radioactive labeling, one primer (1.5 pmol) was end-labeled with [␥-32 P]ATP by T 4 DNA polynucleotide kinase (Invitrogen) before a 25-cycle PCR amplification using 10 pmol of the second primer. DNase I footprinting analyses were performed by incubating 5000 cpm of probe with 40 g of nuclear extracts for 30 min on ice 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, 10% glycerol before submission to increasing amounts of DNase I for 2 min. The reactions were performed in the presence of nuclear extracts of cells treated for 2 h by 130 M DFO compared with control cells. Reactions were stopped by addition of stop buffer (200 g of proteinase K, 50 mM EDTA, 100 g/ml tRNA, 1% SDS) and incubated for 1 h at 50°C. Samples were phenol-extracted, ethanol-precipitated, and recovered in loading buffer (98% formamide, 10 mM EDTA, 0.1% bromphenol blue, 0.1% xylene cyanol), denatured, and applied, together with a G ϩ A Maxam-Gilbert sequence of the probe, on a 6% acrylamide, 8 M urea sequencing gel.
Competition and supershift experiments were performed by preincubating nuclear extract with relevant competitor oligonucleotides or purified antibodies prior to addition of labeled probes.
Computer and Statistical Analysis-eNOS promoter regions were analyzed for hypoxia-binding sites using TRANSFAC data base (transfac.gbf.de/TRANSFAC/) and MatInspector 2.0 software (32). All transfection data are expressed as mean Ϯ S.E. All transfection data were analyzed by the analysis of variance test.

Induction of heNOS mRNA Expression in Endothelial Cells
Treated by DFO-The response to DFO was studied for the endogenous heNOS gene by real time semi-quantitative RT-PCR analyses using RPL32 as a reference mRNA. Iron chela- Reporter constructs were transfected into HUVEC as described under "Experimental Procedures," and cultures were exposed to normoxia or hypoxia (2% O 2 ) for 24 h. Luciferase activity was measured using the dual-luciferase reporter assay system. The fold induction is defined as the ratio between hypoxia-induced firefly luciferase activity and basal firefly luciferase activity (both normalized with the internal control Renilla luciferase activity). All data represent mean Ϯ S.E. of four experiments performed in duplicate. *, p Ͻ 0.05 relative to pGL3-5523.
FIG. 3. DNase I footprint analysis using antisense strand of the region involved in the hypoxia response. A DNA fragment spanning 265 bp was PCR-amplified with a labeled 3Ј primer and submitted to increasing amounts of DNase I endonuclease in the absence of nuclear extracts (lanes [3][4][5] or in the presence of HMEC-1 normoxic nuclear extracts (lanes 6 -8) or DFO-treated nuclear extracts (lanes 9 -11). A ϩ G sequence marker (lane 1) was obtained by Maxam-Gilbert sequencing of the end-labeled fragment. Results with HUVEC nuclear extracts are shown in lanes 12 and 13 (normoxic nuclear extracts) and lanes 14 and 15 (DFO-treated nuclear extracts). The protected region is underlined, and the corresponding sequence is shown. Asterisk indicates one DNase I-hypersensitive site. tion by DFO is used to mimic hypoxia (33) and induces expression of target genes in particular through the HIF-1/HRE mechanism (34,35). Compared with non-treated cells, the he-NOS/RPL32 ratio was significantly increased in HUVEC exposed to DFO for 2 h (1.9-fold Ϯ 0.2, n ϭ 3) and decreased after 8 h (Fig. 1A). Hypoxia time course experiments on HMEC-1 cells revealed similar results (Fig. 1B), showing a rapid increase of eNOS expression after DFO exposure.
Functional Regulation of the heNOS Promoter by Hypoxia-To test whether this response was due to a transcriptional effect, three truncated promoter regions located upstream of the heNOS gene were analyzed for their ability to respond to hypoxia (2% O 2 for 24 h) in fusion constructs with the luciferase gene. Under hypoxia conditions, an induction was observed in primary cultured endothelial cells (HUVEC) with the 6047-bp heNOS construct (2.3-fold Ϯ 0.25, n ϭ 4) and with the 5523-bp construct (2.3-fold Ϯ 0.18, n ϭ 4), but this response was abolished with the 5039-bp fragment, indicating that a hypoxiaresponse element (called heHRE for hypoxia-response element of the heNOS gene) is localized between bp Ϫ5523 and Ϫ5039 of the heNOS promoter (Fig. 2). Similar results were obtained with non-endothelial cells (HeLa cells) (data not shown).
In Vitro Footprint Analysis of the Hypoxia-responsive Region of the heNOS Promoter-DNase I footprint experiments were performed to localize sites of hypoxia response in the region of the heNOS distal promoter identified as functional. Computerassisted analysis of the region allowed us to select a region containing the core sequence of a consensus HIF-1-response element (HRE) 5Ј-RCGTG-3Ј (36) located at Ϫ5371. An endlabeled 265-bp PCR product from Ϫ5569 to Ϫ5305-bp of the heNOS distal promoter was used as a probe in footprint experiments performed with nuclear extracts from DFO-treated or  3 and 4). The same DNA-protein specific complexes were observed with mutants E1 (lanes 6 and 7) and E2 (lanes 12 and 13) but not with the mutants E4 (lanes 9 and 10) or E3 (lanes 15 and 16).
non-treated endothelial cells (HMEC-1 or HUVEC). As shown in Fig. 3, nuclear proteins protected the antisense probe in a large region in DFO-treated cells. This region was not protected in extracts from normoxic cells. The protected element spans 27 bp from Ϫ5382 to Ϫ5356 and is limited in its 3Ј-part by a DNase I-hypersensitive site. The same protection was observed using the sense probe (data not shown). The protection observed is centered on two potential HRE sites (site 1, 5Ј-CGTG-3Ј located at Ϫ5375, and site 2, 5Ј-TACGTG-3Јlocated at Ϫ5371 ; Fig. 3).
Binding of Hypoxia-inducible Proteins to the eNOS Promoter Region (Ϫ5382/Ϫ5356)-To identify proteins that bind to the heHRE region, we performed gel-shift experiments. A 41-bp oligonucleotide (eHRE) was designed (Fig. 4A) to be centered on the protected region of the heNOS promoter identified by footprint analysis and contains the two potential overlapping HREs (Ϫ5375 to Ϫ5366). The double-stranded oligonucleotide eHRE was 32 P-labeled and incubated with 5-g aliquots of nuclear extract from DFO-treated or non-treated HMEC-1 cells.
EMSA showed that a specific DNA-protein complex, absent with control extracts from normoxic cells, is formed with nuclear extracts from DFO-treated HMEC-1 (Fig. 4B), because it is displaced by an excess of cold probe. Competition assays with a 10-and 100-fold molar excess of the unlabeled eHRE wildtype (lanes 3 and 4) or mutated probes were performed (Fig.  4B). Mutating only one of the two potential HREs as in mu-tants E1 (lanes 5 and 6) and E2 (lanes 7 and 8) maintained competition with the probe because of the presence of a DNAprotein interaction involving the other site. However, mutating both sites as in mutants E3 (lanes 9 and 10) and E4 (lanes 11 and 12) abolished the competition with the wild-type probe. Therefore, each of these sites is able to form specific DNAprotein complexes. The first two nucleotides (TA) of the 3Ј HRE site located at Ϫ5371 are essential for the binding as shown by the E4 mutant (lanes 11 and 12), but the two nucleotides before the 5Ј-CGTG-3Ј of the Ϫ5375 HRE site are not essential for the binding as shown by the E5 mutant (lanes 13 and 14) which competed with the probe. As deduced from these experiments, the core sequences of the two sites are CGTG and TACGTG. Other mutations within the 41-bp region but not altering the two core sites resulted in a competition with the wild-type probe and confirmed the essential role of the two core sequences (data not shown).
To identify proteins involved in the formation of these complexes, we performed supershift assays. These experiments showed that antiserum raised against HIF-2␣ (Fig. 4B, lane 3), HIF-1␣ (lane 5), or HIF-1␤ (lane 6) but not by the preimmune serum (lane 4) decreased the formation of the hypoxia-inducible complex and provoked the appearance of a supershifted complex. These results indicate that both HIF-1 and HIF-2 are present in these hypoxia-induced complexes.
To further confirm these results, we performed EMSA experiments with in vitro translated proteins HIF-2␣, HIF-1␣, and HIF-1␤ and wild-type or mutant eHRE probes. DNA-proteinspecific complexes with wild-type eHRE probe were observed in the presence of in vitro translated proteins HIF-2␣ and HIF-1␤ or HIF-1␣ and HIF-1␤ but not with HIF-1␣ and HIF-2␣ alone (Fig. 4D). The series of mutant probes used for EMSA experiments showed that only the mutants (E3 and E4) that altered both HRE-like response elements abolished the binding. Therefore, both HIF-1 and HIF-2 complexes are able to bind (on their own) to the heHRE sequence in vitro.
Effect of Coexpression of HIF-1 and HIF-2 on the Promoter Activity of the heNOS Gene-To investigate whether the increase in the heNOS promoter activity depends on HIF-1 or HIF-2, we performed cotransfection experiments in different cell types. The reporter plasmid containing 6047 bp of the heNOS promoter upstream to the firefly luciferase coding sequence was cotransfected into HMEC-1, HUVEC, and HeLa cells in the absence or the presence of expression vectors encoding HIF-1␣ and ARNT (i.e. HIF-1) or HIF-2␣ and ARNT (i.e. HIF-2) (Fig. 5B). The activation control reporter plasmid HREpGL3-SV40, containing the four putative HRE motifs from the 3Ј-flanking sequence of the human erythropoietin gene (30), was also tested in cotransfection experiments (Fig. 5A).
Characterization of the HIF-2-response Element from the he-

NOS Promoter by Cotransfection Experiments-
The identification of sequences required for HIF-2-induced activation of the heNOS promoter was done in HMEC-1 and HUVEC by cotransfection experiments, as described above, with different reporter vectors corresponding to deletion mutants of the he-NOS promoter obtained by exonuclease III digestion of the vector containing 6047 bp (29). As shown in Fig. 6, cotransfections of HMEC-1 and HUVEC performed with pGL3-6047 and HIF-2 expression vector showed, respectively, a 2.6 (Ϯ0.5, n ϭ 6) and 2.5 (Ϯ0.4, n ϭ 4)-fold increase of the promoter activity as compared with the transfection with the empty pcDNA3.1 vector. Progressive deletions from Ϫ6047 to Ϫ5879 and Ϫ5523 did not show significant reduction of transcriptional induction. Deletion from Ϫ5523 to Ϫ5039 significantly reduced the transcription levels, confirming the presence of a HIF-2-HRE located between positions Ϫ5523 and Ϫ5039 of the heNOS promoter. Further successive deletions allowed induction of transcription levels to increase slightly in HMEC-1 cells and HUVEC.
Mutational Analysis of the heNOS Gene Promoter Activity-HMEC-1 and HUVEC were cotransfected with the 6047-bp fragment of heNOS promoter and with HIF-2 or HIF-1 expression plasmids. Functional analysis of the two adjacent HREbinding sites of the heNOS promoter was performed with mutants E1pGL3-6047, E2pGL3-6047, and E3pGL3-6047 (Fig.  7A). Wild-type or mutated heNOS promoter regions were tested for their ability to respond to HIF-2 or HIF-1 expression vectors in fusion constructs with the luciferase gene (Fig. 7A). No induction was observed with HIF-1 expression vectors. The 2.6-fold (Ϯ0.2, n ϭ 4) induction observed with HIF-2 expression vector was diminished to a 2.4 induction when the mutation involved only one of the two potential HREs (E1pGL3-6047 or E2pGL3-6047), but the induction was abolished when the two potential HRE sites were mutated (E3pGL3-6047) in HUVEC (Fig. 7B). These findings demonstrate the requirement of both intact heHRE in position Ϫ5375 to Ϫ5366 for obtaining full HIF-2-mediated transactivation of the eNOS promoter.
A previous study (20) described that two AP1-binding sites (Ϫ1530 and Ϫ661) were involved in redox-sensitive enhanced eNOS expression. We confirmed, by using HRE and AP1 site mutants, that both AP1 and eHRE sites are at play for hypoxiamediated induction of the Ϫ6047 heNOS promoter (data not shown), and we tested whether the HIFs-mediated transactivation is modified in the presence of AP1 mutants (Mu-tAP1Ϫ1530, MutAP1Ϫ1530/Ϫ661, and M4pGL3Ϫ6047, corresponding to site-directed mutants at both the heHREs and/or the AP1-binding sites introduced in the context of the pGL3Ϫ6047 construct). The HIF-2-mediated transactivation of the Ϫ6047 heNOS promoter was slightly decreased with constructs including mutated AP1 sites (Fig. 7B). In contrast, it was unaltered in HMEC-1 cells (Fig. 7C).
The 41-bp heHRE (Ϫ5390 to Ϫ5350 bp) was then displaced upstream from the 785-bp heNOS promoter and tested in cotransfection experiments with HIF-2 and HIF-1 expression vectors (Fig. 7D). Cotransfection of HIF-2 expression vectors showed an 11.8 (Ϯ0.4, n ϭ 7) increase of the luciferase activity (compared with the transfection with the empty pcDNA3.1 vector), whereas no induction was observed with HIF-1 expression vectors (Fig. 7D). A weaker induction (7.7 Ϯ 0.9, n ϭ 6) was maintained when the 41-bp heHRE (Ϫ5390 to Ϫ5350 bp) was placed in the antisense orientation. Separately mutating each of the eHREs positioned upstream from the 785-bp heNOS promoter did not abrogate enhancer activity, whereas mutation of both sites abolished enhancer activity in cotransfection experiments (Fig. 7D).
Functional Analysis of the heHRE in the Context of a Heterologous Promoter-The 41-bp heHRE was cloned upstream the SV40 promoter driving the luciferase reporter gene and transfected into HeLa cells to test the hypoxia response in the context of a heterologous promoter (Fig. 8A). After 24 h of hypoxia treatment (2% O 2 ), the hypoxia-induced transcriptional activation reached 3.5 (Ϯ0.3, n ϭ 6) for the heHRE pGL3-SV40 construct compared with 8.2 (Ϯ0.7, n ϭ 6) with the control vector containing four HREs of the erythropoietin gene (HRE pGL3SV40). These results show that the heHRE is a functional HRE acting autonomously as a hypoxia-sensitive enhancer.
The same constructs were used in cotransfection experiments with either HIF-2 or HIF-1. When placed upstream the SV40 promoter driving the luciferase reporter gene, the he-HRE induced transcriptional activation after cotransfection with either HIF-2 (3.3 Ϯ 0.3, n ϭ 4) or HIF-1 (2.2 Ϯ 0.35, n ϭ 4) (Fig. 8B). These values were increased when two heHREs were inserted upstream of the SV40 promoter (2heHREpGL3SV40). Coexpression of HIF-2 or, although to a lesser extent, HIF-1, activated transcription of the heHRE reporter plasmid. So, in the context of a heterologous promoter, the heHRE is able to be activated by both HIF-1 and HIF-2.

DISCUSSION
Our results confirm the occurrence of an early up-regulation of the eNOS gene mRNA under hypoxia mimicked in vitro by DFO in both primary cultured and immortalized endothelial cells. This initial up-regulation lasts a few hours and is followed by a decrease in eNOS mRNA level. The range of induction observed for heNOS expression under hypoxia is similar to other endothelial genes such as the VEGF receptor Flt-1 (37) or endothelin (38).
By using a long promoter fragment (Ϫ6047 bp upstream from the ATG translation initiation codon) of the heNOS, we confirmed a promoter-driven increased transcription under hypoxia that used shorter promoter fragments, which was not identified in previous studies (19,20). The elements that drive up-regulation of heNOS gene transcription were mapped between nucleotides Ϫ5523 and Ϫ5039 by progressive deletion of the promoter. DNase footprinting experiments showed that DFO-treated nuclear extracts of endothelial cells protected the probe in a large region (27 bp from Ϫ5356 and Ϫ5382 bp) centered by a potential HRE (TACGTG). EMSA experiments, site-directed mutagenesis, and sequence analysis showed the presence of two contiguous hypoxia-response elements with closely related sequences and arranged in tandem (Ϫ5375 to HIF-1␣ and HIF-2␣ are two highly related basic helix-loop-Helix/Per-Arnt-Sim homologous transcription factors that undergo increased function at low oxygen levels by ␣ subunit stabilization (39 -41). Transcriptionally active heterodimers composed of HIF-1␣ and HIF-1␤ (also named ARNT), or HIF-2␣ and HIF-1␤, play a crucial role in cellular adaptation to hypoxia by inducing the transcription of several genes (42)(43)(44) after binding to HREs.
EMSA experiments and transfection experiments with hypoxia or induced expression of HIF-2 demonstrated the independent protein-DNA interaction on the two contiguous HRE (Ϫ5375 to Ϫ5366). The core sequence of the two sites (CGTG and TACGTG) fits the consensus sequence of the HIF-1-binding site 5Ј-RCGTG-3Ј (45). Our results show that the initial T within the sequence, TACGTG, is required to obtain a response to hypoxia and HIF-2 as shown previously (46) in other genes. Hypoxia induces a number of genes whose promoters or enhancer regions contain more than one HIF-1-binding site like EPO, VEGF, glycolytic enzymes, and plasminogen activator inhibitor-1 (36,47,48). They contain two functionally essential HIF-1 sites arranged as direct or inverted repeats. This second HIF-1-binding site (HBS) is named the HIF-1 ancillary sequence (HAS) (46). This HAS sequence realizes an imperfect inverted repeat with HBS with a spacing of 8 or 9 nucleotides between HBS and HAS motifs, crucial for its activity. In our case, as in the enolase 1 gene, the two adjacent HBS-like sites are in the same orientation and without spacing (36). A search for an imperfect inverted repeat in close proximity within the heNOS promoter sequence was unsuccessful. For several hypoxia-responsive genes, in contrast to the heNOS where the two HRE sites can mediate independently the transcriptional response, the presence of a simple HBS is not sufficient to mediate transcriptional response to hypoxia (36,46).
To evaluate the role of specific hypoxia-response transcription factors as assessed by their overexpression, HIF-1␣, HIF-2␣, and HIF-1␤ expression vectors were transfected in different cell types (endothelial and non-endothelial cells) (Fig. 5). The two different dimers, HIF-1 or HIF-2, were not equally effective on the 6047-bp heNOS promoter construct, as HIF-2 expression vectors are able to induce the eNOS promoter at higher levels than HIF-1, in both endothelial cells and non-endothelial cells. The preferential transcriptional induction of the eHRE by HIF-2 is much more pronounced when the HRE is in its native location than when inserted into a shorter promoter fragment or in a heterologous promoter. In the last two cases, the observed induction by HIF-1 could result from the higher levels of promoter activation in these conditions, allowing its detection in the assay. The unusual structure of the HRE site as compared with other HIF-1-responsive promoters could contribute to the preferential HIF-2 transactivation, although both HIF-1 and HIF-2 bind to the eHRE in EMSA with DFO-treated en- Results are mean Ϯ S.E. of six experiments. *, p Ͻ 0.05 relative to the pGL3-SV40 construct. B, the control reporter HREpGL3SV40 or the one or two copies heHRE 41-bp sequence upstream of the SV40 promoter driving the luciferase reporter gene (heHREpGL3-SV40 or 2 heHREpGL3-SV40) were cotransfected into HMEC-1 cells with HIF-1 or HIF-2 expression vectors. Firefly luciferase activity is normalized to the pRenilla luciferase activity. The fold induction reported is luciferase activity observed in cotransfection with HIF-1 or HIF-2 expression vectors relative to the empty expression vector. Results are mean Ϯ S.E. of six experiments. *, p Ͻ 0.05 relative to the pGL3-SV40 construct. dothelial cells nuclear extracts or when synthesized in vitro. The specific physiological consequences attributable to HIF-2 transactivation are not known and would require disruption of HIF-1 or HIF-2 signaling pathways to identify their specific targets. Although HIF-2 was initially considered as an endothelial specific protein (41), it has been detected in several cell lines and normal or tumoral tissues (49,50). Higher levels of induced HIF-2␣ as compared with HIF-1␣ under hypoxia were described in human lung microvascular endothelial cells (51) suggesting a greater involvement of HIF-2 in hypoxia response of endothelial cells. However, this observation could not explain the HIF-2 specificity in the context of the heNOS promoter because we show that HIF-1 is also able to bind the heHRE according to our EMSA experiments performed with nuclear extracts from hypoxia-treated cells. HIF-2 can bind the same site as HIF-1 and can transactivate known HIF-1 target genes (30), like erythropoietin (41), but can also specifically stimulate expression of genes such as TIE-2 (41) or FLK-1 (52) through binding to DNA-response elements distinct from HRE. Our results suggest another mechanism where HIF-2 seems able to preferentially induce eNOS expression through its binding to HRE elements, as observed for the HRE present in the VEGF promoter that is also preferentially induced by HIF-2 (49).
A 2-fold decrease of the transcription level in HMEC-1 and at a weaker level in HUVEC is observed when the Ϫ5039-bp promoter is cotransfected with HIF-2 ( Fig. 6), suggesting the presence of a repressor element which is removed with proximal promoter deletions. This hypothesis is supported by the increase of the HIF-2-mediated transactivation when the eHRE is displaced upstream by a shorter eNOS promoter fragment pGL3-785 (Fig. 7D). We raise the hypothesis that HIF-2 induces the expression of transcription factors able to inhibit eNOS promoter activity through binding to a response element located between Ϫ5523 and Ϫ3555. One candidate is the transcription inhibitor DEC1 that binds the E box (53,54). Indeed inspection of the DNA sequence of the eNOS promoter showed E boxes present at this location of the eNOS promoter. This hypothesis remains to be tested.
In the case of the VEGF (55), endothelin-1 (38), and tyrosine hydroxylase (56) promoters, a cooperation between AP1 and HIF-1 was observed, as attested by the functional interdependence of the transcription factors. Because we confirmed that the Ϫ1530 and Ϫ661 AP1 sites of the eNOS promoter participate in the hypoxia response of the heNOS promoter, we searched for a cooperation between the two sites. We studied the functional transcriptional effect of the overexpressed HIF-2 in the context of the large 6047-bp eNOS promoter wild type and different mutants on eHREand AP1-response elements. A slight decrease in the transcriptional induction was observed in HUVEC after overexpression of HIF-2 with the wild-type promoter construct as compared with the AP1 mutant. In contrast a similar transcriptional activation was obtained with mutant AP1 sites in HMEC-1 (Fig. 7C). The absence of cooperation could be either due to the absence of specific factors for this interaction in HMEC-1 or to the putative binding of the HIF-1/2 coactivator p300 with the large T antigen of SV40 expressed in HMEC-1, which could alter the protein interaction between AP1 and HIF-1/2 (57).
Our results shed light on the molecular mechanisms underlying the transcriptional up-regulation of eNOS under hypoxia, which implicate more specifically HIF-2. These results further suggest the conclusion that there is a particular set of genes whose transcriptions are induced specifically by HIF-2, and it would be of importance to assess the physiological role of HIF-2-modulated transcription.