Molecular Regulation of the Endothelin-1 Gene by Hypoxia HYPOXIA-INDUCIBLE p300/CBP*

Endothelin-1 (ET-1) is a peptide hormone with potent vasoconstrictor properties which is synthesized and secreted predominantly by vascular endothelial cells. Its production is regulated by numerous stimuli including ischemia and hypoxia, and the enhanced levels that oc-cur during myocardial ischemia may contribute to the progression of heart failure. We reported previously a preliminary characterization of a hypoxia-inducible fac-tor-1 (HIF-1) binding site in the human ET-1 promoter which contributed to the activation of ET-1 expression in endothelial cells. We report here that the HIF-1 binding site alone is not sufficient for the response to hypoxia but requires an additional 50 base pairs of flanking sequence that includes binding sites for the factors activator protein-1 (AP-1), GATA-2, and CAAT-binding factor (NF-1). Mutation of any one of these sites or the HIF-1 site eliminated induction by hypoxia. Mutations of the AP-1 and GATA-2 sites, but not the HIF-1 site, were complemented by overexpressing AP-1, GATA-2, HIF-1 a , or the activator protein p300/CBP, restoring the response to hypoxia. Binding studies in vitro confirmed physical associations among GATA-2, AP-1, and HIF-1 factors. Overexpression or depletion of p300/CBP

family of structurally related peptide hormones and is the most potent endogenously produced vasoconstrictor known (for review, see Ref. 1). The enhanced secretion of ET-1 during myocardial ischemia has been linked with enhanced contractility in the failing heart (2) as well as with the progression of heart failure (3). The combined actions of ET-1 and the endothelial cell relaxing factor nitric oxide may be important in regulating vascular tone and blood pressure (1). In addition to its vasomodulating activity, ET-1 has been shown to modulate multiple cell functions including proliferation, proto-oncogene and protein kinase activity (1), induction of inotropy and hypertrophy in cardiac muscle, and activation of cardiac specific genes (4,5). At least two cell surface receptors for the endothelins are present on multiple cell types, and most of the actions of the peptides are probably relayed through these receptors (for review, see Ref. 6).
Hypoxia is one of the most potent inducers of ET-1 gene expression in endothelial cells and may be the primary cause of the increased production of ET-1 during myocardial ischemia (24 -27). Hypoxia induces the synthesis and secretion of ET-1 in isolated endothelial cells by a mechanism that is antagonized by nitric oxide and carbon monoxide and mimicked by transition metals (15)(16)(17). The ET-1 promoter contains an inverted hypoxia-inducible factor-1 (HIF-1) binding site at position Ϫ118 base pairs upstream of the transcription start site which binds the factors HIF-1␣ and ARNT (HIF-1␤) and is essential for the promoter response to hypoxia (20,28). Here we report that this response also requires three adjacent transcription factor binding sites to form a functional hypoxiaresponsive complex. The response is endothelial cell-specific and is modulated but not necessarily dependent on interactions with the activator protein p300/CBP.

MATERIALS AND METHODS
Cell Culture-Human umbilical venous endothelial cells (HUVECs) were prepared from the umbilical cords of multiple donors as described previously (28). Briefly, umbilical cord veins were rinsed twice with phosphate-buffered saline and filled with 0.1% collagenase in phosphate-buffered saline containing Ca 2ϩ and Mg 2ϩ . After incubation at 37°C for 15 min, gentle flushing of the vein released a suspension of isolated cells. These were seeded onto gelatin-coated culture dishes, and the adherent ECs were cultured in M199 medium supplemented with 100 units/ml penicillin, 100 g/ml streptomycin, L-glutamine, and 10% human serum. In some experiments HUVECs were purchased from Clonetics (San Diego, CA) and cultured as described previously (19). HeLa, HepG2, and human embryonic kidney (HEK)-293 cells were also cultured as described previously (29).
Hypoxia-Our conditions for hypoxic incubations have been described in detail previously (19,28,30). Briefly, cells were incubated in a temperature-and humidity-controlled environmental chamber in an atmosphere containing 0.5% O 2 , 5% CO 2 , balance N 2 . Oxygen tension inside the chamber was monitored continuously with an oxygen-sensitive electrode (Kathaerobe Controls, Philadelphia). All cell manipulations including media changes and harvesting took place under hypoxia to avoid transitory reoxygenation.
Plasmid Insertions-A 70-base pair fragment from the enolase-1 promoter containing the enolase-1 HRE (33) was inserted into the BamHI site of pET-1Luc-WT (both Ϫ176 and -669 promoters) by blunt end ligation. Spacer DNA sequences containing 50 base pairs were inserted in between GATA-2 and HIF-1, and HIF-1 and AP-1, respectively, by polymerase chain reaction-based insertional mutagenesis using the same methods described above for site directed mutagenesis with the foreign sequence overhanging 5Ј-and 3Ј-primers of 36 and 30 base pairs each complementary to the respective site for insertion. The sequence of the inserted DNA was 5Ј-ATGCTAGGCGTCATGAGTAC-GAGGTCGGAGCTACGTACTGCCGTTGTACG-3Ј. All new constructs were confirmed by sequencing.
Northern Blot Analyses-RNA transcript levels were measured by Northern blots as described previously (19,29). Briefly, total RNA was isolated by solubilizing cells on the plate in 4 M guanidinium thiocyanate (0.25 ml/10 6 cells) and pelleting through cesium chloride. Agarose gels, blotting, and hybridizations were all as described previously (19). Complementary DNA probes including human ET-1 (purchased from ATCC) and ␤-actin (29) were labeled by random priming (Prime-It kit, Stratagene) to 10 8 cpm/g DNA. RNA bands on the autoradiographs were quantitated using a UMAX Powerlook II scanner, Power Macintosh 8500/150, and Adobe software. RNA loading on gels was monitored by ethidium staining and by probing with ␤-actin as the control.
Transfections-HUVEC, HeLa, HepG2, and HEK-293 cells were transfected by the calcium phosphate method as described previously (19). HUVECs were transfected at 80% confluence; other cell lines were transfected at 30 -40% confluence; cells were exposed to 0.5 ml of calcium phosphate precipitate containing 8 -10 g of plasmid DNA including an internal control (Renilla luciferase from Promega Biotechnology) for 8 -12 h. Transfected cultures were exposed to hypoxia after a further 24 -48 h. Expression of luciferase activity, normalized to the internal control, was determined as described previously (19).
In Pull-down Assays-Double-stranded probes biotinylated at the 5Јends were from Life Technologies, Inc. These included the wild type ET-1 HRE described above and individual mutations in the wild type sequence including GATA-2(M): TTAT to CCGA; HIF-1(M): CGT to TAC; and AP-1(M): GTGA to ACAC. Probes were also synthesized containing double mutations of AP-1 and GATA-2, and triple mutation of all three sites. The probes were purified by polyacrylamide gel electrophoresis, and equal amounts of complementary strands were annealed. Dynabeads M-280 Streptavidin (Dynal, Inc., Lake Success, NY) were prepared by washing three times in phosphate-buffered saline (pH 7.4) containing 0.1% bovine serum albumin and two times with Tris-EDTA containing 1 M NaCl. Between each wash, beads were pulled down with a Dynal magnetic particle concentrator. Double-stranded, biotinylated oligonucleotides were added to the washed beads, and the mix was rotated for 20 -30 min at 21°C. Equal cpm of proteins translated in vitro were made to 1 ϫ with binding buffer and mixed with ϳ100 g of Dynabeads containing 10 pmol of the individual oligonucleotide probe in a final volume of 250 l. The mixture was rotated at room temperature for 20 min, and proteins bound to the beads were separated from unbound proteins by successive washes, three times with 0.5 ϫ binding buffer and once with 1 ϫ binding buffer. Higher stringency wash included two washes with 2 ϫ binding buffer. Beads and bound proteins were pulled down with a magnetic concentrator, suspended in 1 ϫ sample buffer, boiled for 5 min, and resolved on SDS-polyacrylamide gels (10%) as described for immunoprecipitation reactions.
The same procedures were used to pull down p300 from nuclear extracts. In this case the protein eluted from the beads was electroblotted onto nitrocellulose membranes (Bio-Rad). Blots were stained with Ponceau Red to monitor the transfer of proteins. Membranes were blocked for 1 h at room temperature with 5% nonfat milk in TBS (25 mM Tris, 137 mM NaCl, 2.7 mM KCl) containing 0.05% Tween 20 and incubated with anti-p300 antibody for 2-4 h in the same buffer. After washing, the blots were incubated for 1 h with a 1:7,500 dilution of horseradish peroxidase-conjugated anti-rabbit IgG or horseradish peroxidase-conjugated donkey anti-goat IgG and visualized using enhanced chemiluminescence (Pierce). Fig. 1A shows the sequence of the HRE in the ET-1 promoter between base pairs Ϫ91 and Ϫ142 upstream of the transcription start site, including the positions of transcription factor binding sites. The base changes shown were introduced individually and in pairs into pET1 Ϫ176 -Luc as described under "Materials and Methods." The expression of the wild type pro-moter and each mutation in endothelial cells cultured under aerobic or hypoxic conditions is shown in Fig. 1A. The AP-1 and GATA-2 mutations reduced basal expression to Ͻ 20% of the wild type. In agreement with previous reports expression of the wild type promoter was induced by 2.3-fold (Ϯ 0.3, n ϭ 12; p Ͻ 0.001) under hypoxia (28). This response was eliminated in all constructs containing site mutations. Therefore activation of the ET-1 promoter by hypoxia in ECs requires intact GATA-2, HIF-1, AP-1, and NF-1 binding sites.

Functional Characterization of the ET-1 Promoter HRE-
It was reported previously that the decreased basal expression of AP-1 and GATA-2 site mutations in the ET-1 promoter were complemented by overexpressing the corresponding factor (23). To test for complementation of the response to hypoxia, wild type and mutated promoter constructs were cotransfected with HIF-1␣, c-Jun, or GATA-2 expression vectors. These results are shown in Fig. 1B. Expression of the wild type promoter was augmented by c-Jun (1.3-fold) and GATA-2 (2.3-fold) cotransfections but not by HIF-1␣, (all p Ͻ 0.02; n ϭ 4). The fold activation of the wild type promoter by hypoxia was not significantly effected in either case. Expression of the HIF-1 site mutation was not significantly affected by cotransfection of c-Jun, GATA-2, or HIF-1␣, and there was no hypoxia-mediated induction of this construct under any condition. The basal expression of the AP-1 and GATA-2 site mutations was fully complemented by either c-Jun or GATA-2 cotransfection, in agreement with a previous report (23), and the response to hypoxia was fully restored to both mutations by cotransfection of c-Jun, GATA-2, or HIF-1␣. Cotransfection of cDNA encoding Sp1 did not restore the hypoxia response of these mutations (data not shown).
These results demonstrate that the HIF-1␣ binding site is essential but not sufficient for activation of the ET-1 promoter by hypoxia. Homologous complementation of the AP-1 and GATA-2 site mutations and cross-complementation by the heterologous factors suggest that each of these factors can be recruited to the HIF-1 complex without directly binding DNA. Complementation of either AP-1 or GATA-2 site mutations by HIF-1␣ further suggests that these sites may modulate the DNA binding affinity of HIF-1␣ (see "Discussion").
Gel Electrophoretic Mobility Shift Assays-Gel mobility shift assays were carried out to determine whether hypoxia changed the binding activities of the HIF-1 site flanking proteins. As shown in Fig. 2, shifts corresponding to the binding of AP-1, GATA-2, and HIF-1 complexes were observed, confirming previous reports from this and other laboratories (23,28). The arrows in the upper panels indicate the positions of specific AP-1-, GATA-2-, and HIF-1-shifted bands, respectively. Using individual site sequences or the whole ET-1 HRE as the probe, hypoxia did not effect the apparent abundance or binding activity of either AP-1 or GATA-2. This is in contrast to some other cell types where AP-1 binding is strongly activated by hypoxia (36). In the top right panel, extracts from normoxic or hypoxic EC or HeLa cells were reacted with the ET-1 HIF-1 probe (first through fourth and seventh and eighth lanes) or with a probe containing the erythropoietin (Epo)-HIF-1␣ consensus (fifth, sixth, and ninth lanes). The ET-1-specific HIF-1 Constructs were transfected into HUVECs as described under "Materials and Methods," and cultures were exposed to normoxia or to hypoxia for 24 h. Luciferase expression is normalized to a control plasmid, and activities are expressed relative to the aerobic wild type. B, the wild type Ϫ176 ET-1 promoter or individual mutations (4 g each) were cotransfected into HUVECs with AP-1, GATA-2, or HIF-1␣ expression plasmids (2 g each) as indicated. Controls were cotransfected with 2 g of empty vector. Results are from at least four separate experiments in duplicate. Light bars, aerobic; dark bars, hypoxic. probe generated markedly weaker interactions than the corresponding Epo-HIF-1 probe (compare second, fourth, and sixth lanes), suggesting that the ET-1 site is a low affinity HIF-1 binding site.
Binding studies with a probe containing the complete ET-1 HRE-region (Ϫ91 to Ϫ141) are shown in the lower panels. Shifts representing AP-1 and HIF-1 were confirmed by supershifts with specific antibodies (left panel, fourth and sixth lanes). A weak supershift was observed with an antibody directed against p300/CBP (fifth lane), confirming the presence of p300 in these complexes. The third arrow on the left indicates the probable position of the GATA-2 band shift; this band was always weak, and the anti-GATA-2 antibody did not generate a supershift (not shown). Mutation of the GATA-2 site in the ET-1-HRE probe (lower right panel) did not affect the binding of AP-1 or HIF-1 factors. Therefore, a functionally disruptive mutation (that prevented the promoter activation by hypoxia) did not disrupt HIF-1 binding in vitro. Interestingly, the supershift caused by the anti-c-Jun antibody also eliminated the HIF-1-specific band (both bottom panels). These results show that the ET-1 HRE binds the HIF-1 complex weakly, the binding of AP-1 GATA-2 is not affected by hypoxia, and there is no evidence for cooperative binding in vitro.
Detection of Protein-Protein Interactions: Pull-down Assays-A biotinylated DNA pull-down assay was used to analyze protein-protein interactions in the ET-1 HRE complex. This technique was used previously by Ebert and Bunn to demonstrate cooperativity in the transcriptional assembly of HIF-1, adjacent transcription factors, and p300/CBP in the regulation of the LDH-A and Epo genes (37). In our studies, biotinylated ET-1 HRE oligonucleotides with single, double, or triple mutations were used to pull down proteins translated in vitro as described under "Materials and Methods." Fig. 3A shows the results obtained using a biotinylated ET-1 HRE with all transcription factor binding sites intact or with all mutated. The small double arrows in Fig. 3A indicate positions of HIF-1␣ (upper) and HIF-1␤ (lower) products; in subsequent assays both HIF-1 products were included in the reactions, but only the FIG. 2. Electrophoretic gel mobility shift analyses of protein binding to the ET-1 HRE. Top panels, labeled oligonucleotide probes containing the sequences corresponding to AP-1, GATA-2, and HIF-1, wild type or mutant from the ET-1 HRE were mixed with nuclear extracts from HUVECs cultured under normoxia or for 24 h under hypoxia as described under "Materials and Methods." In the right panel, fifth, sixth, and ninth lanes, the probe was the HIF-1 binding sequence from the Epo HRE; HeLa cell nuclear extracts were used in the third through sixth and ninth lanes. Where indicated, competitor (unlabeled) oligonucleotide was added to the binding reaction at 200-fold excess over the labeled probe. Arrows indicate the specific band corresponding to the indicated transcription factor. Bottom panels, labeled probe was the complete ET-1 HRE, base pairs Ϫ91 to Ϫ142 (wild type) or with the GATA-2 site mutated on the right. Where indicated, specific antibodies (4 -8 l) were added to the binding reaction. Arrows on the left of the panels indicate factor-specific shifts; arrows on the right indicate the positions of supershifts.
HIF-1␣ product was labeled with [ 35 S]methionine during translation. Interactions with the wild type probe were observed when the factors were added individually or in combination. As expected, no interactions were observed using the triple mutated oligonucleotide (Fig. 3A, fifth lane).
Individual (single) mutations of the AP-1 or GATA-2 sites in the ET-1 probe did not prevent the binding of c-Jun, GATA-2, or HIF-1 proteins (Fig. 3B). This seemingly anomalous result can be explained by cross-interactions between c-Jun and GATA-2 proteins. Previous work has established that these proteins interact in vitro, and there are high endogenous levels of both factors in rabbit reticulocyte lysates (23 and data not shown). Therefore, single mutations of the AP-1 or GATA-2 binding sites did not prevent the pull down because protein bound to the remaining intact site is sufficient to pull down both proteins. In support of this, double mutation of AP-1 and GATA-2 sites dramatically reduced the binding of both factors (Fig. 3C).
The pull down of GATA-2 and c-Jun by the double mutation probe was enhanced when HIF-1 (␣ and ␤) was included in the binding reaction (Fig. 3D), indicating that more GATA-2 and c-Jun complexed in the presence of HIF-1. It was possible to detect this interaction because the reticulocyte lysate, present in all binding reactions, does not contain detectable endogenous HIF-1-␣ (data not shown). In further support of this binding activity, when the HIF-1 site was mutated, HIF-1 binding was reduced dramatically (Fig. 3E) but was recovered when additional GATA-2 or c-Jun was added. The latter effect was sometimes masked by the high background level of HIF-1 binding even to the HIF-1-mutated probe. This background can also be attributed to cross-interactions with AP-1, GATA-2, or p300 from the lysate (see below). Importantly neither HIF-1 nor any of the other factors bound to the triple mutation probe, indicating that all interactions observed were dependent on specific binding sites. These results provide strong support for physical interactions (direct or indirect) among c-Jun, GATA-2, and HIF-1 bound to the ET-1 HRE. We were unable to demonstrate similar interactions in coimmunoprecipitation experiments using antibodies against c-Jun, GATA-2, or HIF-1 (data not shown). This suggests that the interactions require a DNA template and implicates physical associations with other factors such as p300 which may mediate complex formation and cross-interactions of the factors (see below).
Function of the ET-1 HRE Complex: Role of p300 -The studies described in Figs. 1-3 indicate that AP-1, GATA-2, HIF-1 (and NF-1) proteins interact directly or indirectly when bound to DNA to produce a functional response to hypoxia. The activator/adaptor protein p300/CBP has been show to interact with AP-1, GATA factors, and HIF-1␣ (37)(38)(39)(40)(41). Therefore one function of AP-1 and GATA-2 here may be to facilitate the recruitment of p300 to the ET-1 HRE. To test for this constructs containing the wild type ET-1 HRE or different mutations were cotransfected into endothelial cells with a p300 expression vector (Fig. 4A). Cotransfection of p300 augmented the basal and induced expression of the wild type ET-1 promoter but did not change the fold induction by hypoxia (2.01-fold with p300 compared with 2.43-fold without p300; not significant, n ϭ 5). Therefore, p300 availability may limit both basal and activated expression of the ET-1 promoter in endothelial cells. Cotransfection of p300 also augmented the expression of the HIF-1 site mutation but did not support induction by hypoxia; in fact, expression of the HIF-1 site mutation was significantly less under hypoxia under these conditions (p Ͻ 0.05). p300 overexpression augmented the basal expression of AP-1 and GATA-2 site mutations by 2-3-fold, and remarkably, the hypoxia response was fully restored to both promoters (p Ͻ 0.01; n ϭ 4 for both mutations). Therefore, the hypoxic induction of AP-1 and GATA-2 site mutations can be complemented by AP-1, GATA-2, HIF-1␣, or p300.
The restoration of function by p300 suggests that at least one function of AP-1 and GATA-2 is the recruitment of p300 to the ET-1 HRE complex. This being the case, depletion of p300 should reduce the response as it does with the LDH-A promoter and Epo 3Ј-enhancer sites (37). To test for this, p300 availability was reduced by cotransfecting the p300 binding site of adenovirus E1A (Fig. 4B). Expression of the wild type ET-1 promoter was quenched in the presence of the E1A plasmid, but again the fold induction by hypoxia did not change (2.3-fold without E1A; 2.2-fold with E1A; not significant, n ϭ 6). Cotransfection of E1A also quenched the amplified expression caused by p300 overexpression, but yet again did not change the fold induction. In contrast to this, E1A cotransfection reduced the hypoxic fold induction but not the basal expression of the Epo-HRE in HeLa cells (Fig. 4B). We also measured the effects of E1A expression on the endogenous ET-1 transcript. As shown in Fig. 4C, ET-1 transcripts were reduced by Ͼ90% in ECs infected with adenovirus, but the message was still induced by hypoxia. Both the basal expression and hypoxic induction of LDH were lost, and there was no change of ribosomal 28 S RNA. Therefore p300 appears to modulate the basal expression of the ET-1 gene promoter but not the fold induction by hypoxia. These results suggest that p300 may modify the level of expression of the ET-1 HRE possibly through interactions with AP-1 and GATA-2, but the hypoxia response is independent of this regulation.
To determine whether hypoxia activation correlated with p300 binding, biotinylated probes were used to pull down p300 from normoxic and hypoxic nuclear extracts. As shown in Fig.  5, p300 bound equally to the wild type probe and to all probes with single site mutations. There was no difference in p300 binding between normoxic and hypoxic nuclear extracts, and this pattern did not change with higher stringency washes (not shown). p300 binding to the AP-1/GATA-2 double mutation probe was reduced dramatically, but again there was no apparent difference between normoxia and hypoxia, suggesting that the ET-1 HIF-1 site is only a weak binding site for p300 compared with AP-1 or GATA-2. There was no detectable p300 bound to the triple mutation probe as expected, and binding to the phosphoglycerate kinase-HRE probe was highly hypoxiadependent, confirming that p300 binds to HIF-1 with this probe.
The Flanking Sites Are Not Strictly Position-dependent-Taken together, these results suggest that the ET-1 HRE flanking sites may stabilize HIF-1␣ binding to DNA and the interaction with p300. Because the GATA-2, HIF-1, and AP-1 sites are closely aligned in the ET-1 promoter, we sought to determine whether their functions were position-dependent. 50 base pair spacers were inserted between GATA-2 and HIF-1, and HIF-1 and AP-1 sites respectively, and the function of the HRE was determined. As shown in Fig. 6, insertion of the spacers reduced basal expression of the promoters but did not affect the fold induction by hypoxia.
Induction of Et-1 Expression by Hypoxia Is Endothelial Cellspecific-We reported previously that the induction of the endogenous ET-1 transcript by hypoxia was confined to endothelial cells with no apparent activation in HeLa cells or cardiac myocytes (19). To see if this also applied to the regulation of the promoter, the wild type ET-1 construct was analyzed in HeLa, HepG2, and HEK-293 cells. As shown in Fig. 7A, none of these cell lines supported hypoxic induction of this promoter (p Ͻ 0.001). Insertion of the ␤-enolase HRE, a non-tissue-selective HRE, into the ET-1 promoter eliminated the tissue selectivity, indicating that the selectivity was the property of the ET-1 HRE rather than other promoter elements (Fig. 7B). None of the recognized ET-1 HRE-binding factors is strictly EC-specific, although GATA-2 has been shown to contribute to ECspecific gene expression (42,43). Therefore, we analyzed the effect of cotransfecting other factors into HEK-293 cells. As HUVECs were transfected with the wild type ET-1 Ϫ176-Luc construct or individual mutations in the HRE as described in Fig. 1. Panel A, the different constructs were cotransfected with a vector expressing p300, and expression of reporter was measured after normoxic or hypoxic culture as indicated. Panel B, the wild type ET-1-Luc plasmid was cotransfected into HUVECs with a plasmid expressing the adenovirus E1A protein with or without p300. In this panel, right side, a construct containing four copies of the Epo HRE was cotransfected with or without pE1A into HeLa cells. In panel C, confluent HUVECs were infected with adenovirus YH47928 (5 Pfu/cell) expressing E1A (32) for 24 h and were exposed to hypoxia for an additional 24 h or remained aerobic. RNA was extracted and analyzed by Northern blots as described previously using ET-1 or LDH-M cDNA probes (19,48). shown in Fig. 7C, the hypoxia response was fully reconstituted by cotransfecting GATA-2 and HIF-1␣ but not by p300 or the other plasmid combinations. This supports the essential role of GATA-2 (and AP-1) factors in creating an active HRE complex and accounts at least in part for the apparent tissue specificity of this response.

DISCUSSION
In this report we have shown that individual mutations of the GATA-2, HIF-1, AP-1, or NF-1 sites in the ET-1 proximal promoter eliminated activation of the promoter by hypoxia. The HIF-1 site mutation eliminated the hypoxia response under all conditions whereas AP-1 and GATA-2 site mutations were fully complemented by overexpressing AP-1, GATA-2, HIF-1, or p300/CBP factors. These results implicate functionally important protein-protein interactions between these factors that do not necessarily require the direct DNA binding of all factors. In addition, this is the first evidence of roles for GATA factors or NF-1 in the transcriptional activation of a promoter by hypoxia. Although we have not addressed protein binding to the NF-1 site in this report, mutation of the site eliminated hypoxiainduction and it seems possible that the ubiquitous NF-1 binding proteins function in a manner similar to AP-1 and GATA-2 proteins in the HRE complex.
Pull-down binding assays of in vitro translated proteins demonstrated that c-Jun and GATA-2 proteins associated with the ET-1 HRE probe even when their individual sites were mutated, supporting the presence of strong protein-protein interactions between these factors and the HRE complex (22,23). The pull-down of these factors was dramatically decreased when the probe contained a double AP-1/GATA-2 mutation, confirming the requirement for at least one DNA binding site. The enhanced pull down of AP-1 and GATA-2 by the double mutation probe in the presence of added HIF-1 factors indicates that these factors can be pulled down by interacting directly with HIF-1 or other factors associated with DNAbound HIF-1. These results confirm the physical as well as functional relationships among HIF-1, AP-1, and GATA-2 factors. It is noteworthy that we did not observe strong proteinprotein interactions in DNA-free coimmunoprecipitation as-says of these factors under conditions where the expected interactions between c-Fos and c-Jun were seen (data not shown). This result supports the requirement of DNA binding for assembly of the complex and subsequent cross-interaction among individual factors. FIG. 6. Position dependence of the ET-1, AP-1, and GATA-2 flanking sequences. 50 base pairs of random DNA sequence were inserted in between AP-1 and HIF-1, and HIF-1 and GATA-2 sites as indicated. The expression and regulation of the promoters after transfection into HUVECs were as described in Fig. 1.   FIG. 7. Tissue-specific regulation of the ET-1 HRE. In panel A the wild type ET-1 Ϫ176 promoter was transfected into HeLa, HepG2, or HEK-293 cells as indicated, and expression was measured after exposure to normoxic or hypoxic conditions. The control plasmid, pGL3 (Promega), containing three copies of the Epo HRE was induced 8.7-fold (Ϯ 0.5, n ϭ 6) in HeLa cells. Panel B, expression of the ET-1 promoter containing one or two copies of an enolase-1 HRE transfected into HeLa cells. Panel C, effects of cotransfecting AP-1, GATA-2, HIF-1␣, and p300 vectors with the wild type ET-1 promoter in HEK-293 cells.
Hypoxic activation was restored to ET-1 promoters with GATA-2 or AP-1 mutations by overexpressing the homologous factors or by overexpressing either HIF-1␣ or p300. Complementation of AP-1 and GATA-2 site mutations with p300 demonstrates the central role of p300 in organizing the cooperative binding of these transcription factors to the ET-1 HRE site. This confirms previous work that demonstrated a similar role for p300/CBP in the cooperative binding of HIF-1 flanking site proteins to the LDH-A promoter and erythropoietin 3Ј-enhancer (37). In these latter studies mutation of the flanking cyclic AMP response element site in the LDH-A HRE reduced p300 recruitment and prevented activation by hypoxia. Complementation of the ET-1 HRE flanking site mutations by HIF-1␣ overexpression in our studies suggests that these mutations also reduce the affinity of HIF-1 for DNA binding, an effect that may be independent of p300. These studies support dual roles for the ET-1 HIF-1 flanking sites, modulating the affinity of HIF-1 binding and promoting recruitment of p300.
The ET-1 HRE differs markedly from the LDH-A and Epo HREs in having two and perhaps three strong p300 binding sites in addition to the HIF-1 complex. Whereas p300 binding to the LDH-A and Epo HRE sites is strictly dependent on hypoxia ( Ref 37; also see Fig. 4C), this is not the case for the ET-1 HRE site. p300 bound strongly to the wild type ET-1 HRE and to all single site mutations but only weakly to the double AP-1/GATA-2 mutation, and there was no evidence for increased p300 binding from hypoxic nuclear extracts. Importantly, this indicates that p300 binds principally to the ET-1 AP-1 and GATA-2 sites in a constitutive manner. This is in contrast to the LDH-A and Epo HRE sites and may account for important differences in the functions of these HREs. In particular, the absence of hypoxia-regulated p300 binding to the ET-1 probes is consistent with a similar absence of p300 influence on the fold activation of the ET-1 promoter by hypoxia (Fig. 4, A and B). This again is in contrast to the LDH-A promoters and Epo 3Ј-enhancer where fold activation correlated quantitatively with p300 availability and binding (37,41). Therefore, p300 binding controls the fold activation of the LDH-A and Epo genes by hypoxia but not the ET-1 promoter. Our data are consistent with a model whereby the HIF-1 flanking sites of the ET-1 HRE determine the affinity of both HIF-1 and p300 binding. It is also possible that the wild type ET-1 HRE has a high affinity p300 binding site that is saturated even under conditions of p300 depletion; in this case it would be predicted that this high affinity site is lost when any of the flanking sites is mutated. We have no direct evidence for the latter possibility.
The modulation of ET-1 HRE function by the flanking sites also appears to dictate EC selectivity of the response. In contrast to other HRE-dependent promoters, the wild type ET-1 promoter was not responsive to hypoxia in HeLa, HepG2, or HEK-293 cells (19). The response could be reconstituted by overexpressing GATA-2 and HIF-1␣. This result underscores the importance of these factors in regulating the response of the ET-1 HRE to hypoxia and supports a mechanism whereby flanking site factors modulate of HIF-1 binding affinity through cell-specific protein-protein interactions.
Previous studies have implicated essential contributions of AP-1 to the hypoxia-mediated activation of tyrosine hydroxylase, heme oxygenase, and possibly vascular epithelial growth factor promoters (36, 44 -47). It seems possible that AP-1 may fulfill similar roles in mediating HIF-1 and p300 binding activities also in these promoters. In each case the AP-1 site is situated further from the HIF-1 binding site than is the case with the ET-1 promoter; however, our results suggest that immediate proximity to the HIF-1 site may not be a critical parameter for the modulating roles of these factors. The ET-1 promoter may be unique in requiring not only AP-1 but also GATA-2 and NF-1 factors for a functional HRE. One important property dictated by this multiple site regulation, which conveys a degree of tissue selectivity, may be to ensure that this potent vasoconstrictor is not activated adventitiously by hypoxia in tissues other than the endothelium.