Hypoxic Regulation of Angiopoietin-2 Expression in Endothelial Cells*

Exposure of endothelial cells to hypoxia-induced angiopoietin-2 (Ang2) expression. The increase in Ang2 mRNA levels occurred by transcriptional regulation and by post-transcriptional increase in mRNA stability. Induction of Ang2 mRNA resulted in an increase of intracellular and secreted Ang2 protein levels. Since the transcriptional regulation of several genes involved in angiogenesis during hypoxia is mediated by hypoxia-inducible factor-1 (HIF-1), it was conceivable that Ang2 expression might be regulated by the same oxygen-dependent mechanism. However, our data showed that pharmacological HIF inducers, CoCl2 and DFO, did not affect Ang2 expression. Moreover, HIF-1-deficient hepatoma cell (Hepa1 c4) and its wild-type counterpart (Hepa1 c1c4) up-regulates Ang2 during hypoxia. These results indicated that hypoxia-driven Ang2 expression may be independent of the HIF pathway. Using neutralizing VEGF antibody or pharmacological inhibitors of VEGF receptors, we showed that hypoxia-induced VEGF participates but could not account completely for Ang2 expression during hypoxia. In addition, hypoxia elicited an increase of cyclooxygenase-2 (COX-2) expression and a parallel increase in prostanglandin E2 (PGE2) and prostacyclin (PGI2) production. COX-2 inhibitors decreased the hypoxic induction of Ang2 and the hypoxic induction of PGE2 and PGI2 in a dose-dependent manner. Similarly, COX-2 but not COX-1 antisense treatment decreased hypoxic induction of Ang2 expression, and this effect was reversed by exogenous PGE2. Finally, exogenous PGE2 and PGI2 were able to stimulate Ang2 under normoxic conditions. These findings suggest that COX-2-dependent prostanoids may play an important role in the regulation of hypoxia-induced Ang2 expression.

A prolonged reduction in normal tissue oxygen tension triggers a cascade of compensatory responses such as angiogenesis, which results in improved tissue oxygenation. This process is regulated by multiple molecular effectors; vascular endothelial growth factor (VEGF) 1 and its cognate receptors being critical activators of angiogenesis, mediating endothelial cell proliferation, migration, and tube formation (1). The regulation and biological consequences of VEGF expression during hypoxia have been extensively studied (1,2). However, VEGF production and signaling represents only one component of the complex angiogenic response to hypoxia and several studies have shown that VEGF must work in conjunction with other angiogenic factors like the angiopoietin family to produce a stable and functional microvasculature (3)(4)(5)(6). Of the four currently known angiopoietins (Ang1-4), the best characterized are Ang1 and Ang2 (7,8). Ang1 is an agonistic ligand that induces activation of Tie2, an endothelium-specific tyrosine kinase receptor, whereas Ang2 binds to this receptor but has a contextdependent effect on its activation (agonist or antagonist under different conditions) (9 -14). Ang1 appears to be widely expressed in adult tissues and Tie2 seems to be constitutively phosphorylated in quiescent endothelium (12,15). Ang2, on the other hand, is primarily expressed in the endothelium of adult tissue undergoing vascular remodeling and in highly vascularized tumors (12,16,17). The complex role of Ang2 in vascular remodeling was initially suggested by the observations that VEGF and Ang2 showed coincidental expression patterns during angiogenic sprouting and that induced Ang2 expression was also observed in the absence of VEGF during vessel regression (16 -19). Recent reports indicate that transgenic cardiac overexpression of both VEGF and Ang2 leads to a synergistic induction of angiogenesis (6). In addition, using the papillary membrane in vivo model, Lobov et al. (4) demonstrated that Ang2 could stimulate angiogenesis or capillary regression depending on the presence of VEGF.
Like VEGF, Ang2 is also hypoxia-inducible. Its expression was found to be increased in cultured endothelial cells exposed to hypoxia (20 -22). Moreover, we and others (21,(23)(24)(25) have shown that in vivo, ischemia and hypoxia trigger overexpression of this angiogenic regulator in several tissues. These observations suggest that Ang2 may be a significant mediator of hypoxia-induced angiogenesis but the mechanisms regulating its expression are not known.
Some of the cellular responses to hypoxia involve the transcriptional activation of a variety of oxygen responsive genes through the action of the transcription factor hypoxia-inducible factor (HIF-1). For instance, HIF-1 target genes involved in angiogenesis and vasomotor control include VEGF and its receptor Flt-1, nitric-oxide synthase-2, endothelin-1, heme oxygenase, adrenomedullin, and plasminogen activator inhitor-1 (26 -34). However, it is not known whether hypoxia induced expression of Ang2 is regulated by HIF-1.
On the other hand, increasing evidence strongly supports the role of cyclooxygenase-2 (COX-2) and its metabolic product pros-taglandin E2 (PGE 2 ) as regulators of angiogenesis (35)(36)(37). For instance, an in vitro model has been used to demonstrate that overexpression of COX-2 in colon cancer cells leads to increased production of angiogenic factors (e.i.VEGF, bFGF, PDGF), which, in turn, induce angiogenesis (38). Furthermore, inhibition of COX-2 by nonsteroidal anti-inflammatory drugs suppresses tumor growth in animal models, at least in part, by inhibiting angiogenesis (39,40). In addition, it was reported that COX-2 expression is induced by hypoxia in cultured endothelial cells and this induction was not mediated by HIF-1 (41,42).
In the present study, we investigated some mechanistic aspects of hypoxia-driven Ang2 expression. We explored the role of HIF-1 as a transcriptional activator of Ang2. Moreover, the potential role of COX-2 as a regulator of angiogenesis prompted us to study its contribution to hypoxia-driven induction of Ang2 in endothelial cells.
Rofecoxib was prepared from commercial Vioxx® tablets by organic solvent extraction and recrystallization. The purity of this compound was verified by mass spectroscopy.
Endothelial Cell Culture and Hypoxia Treatment-Human umbilical vein endothelial cells (HUVEC) were obtained from Clonetics and cultured in endothelial basal medium (EBM) containing 2% fetal bovine serum and growth supplements as formulated by Clonetics (hFGF, VEGF, IGF-1, hEGF, hydrocortisone, ascorbic acid, GA-1000 and heparin). Only passages 3-5 were used and experiments were performed on cells approaching confluence. bEnd3 cells were purchased from ATCC and cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum. Exposure to normoxia was performed under standard culture conditions in a humidified incubator maintained at 5%CO 2 /95% room air at 37°C. For hypoxic exposure, cells were placed in a controlled environment chamber (Forma Scientific) flushed with 1% O 2 /5% CO 2 /94% N 2 . Before normoxic or hypoxic exposure, cells were washed with phosphate-buffered saline, and fresh medium was added. In the case of HUVECs, fresh medium contained only 2% fetal bovine serum, hydrocortisone, ascorbic acid, GA-1000, and heparin. Preliminary immunoblot experiments demonstrated that basal normoxic expression of either Ang2 or COX-2 in HUVEC and bEnd3 cells did not change at 3, 6, 12, or 24 h after new medium was added (data not shown). For simplicity, when the experiments involved a time course of hypoxic exposure, the normoxic control corresponded to cells that remained with new medium for 24 h in normoxic conditions. PGE 2 , Iloprost, PGF2␣, and U46619, COX inhibitors (indomethacin, NS-396, PTPBS, SC-560, rofecoxib), VEGF receptor inhibitors (SU1498 or VEGFR tyrosine kinase inhibitor), neutralizing antibody directed against VEGF or vehicle were added just prior to hypoxic or normoxic exposure.
Analysis of Cell Viability-Release of lactate hydrogenase (LDH) into the culture medium was measured using the in vitro toxicology assay kit (Sigma) according to the manufacturer's protocol. Levels of histoneassociated DNA fragments were assayed using the cell death enzymelinked immunoabsorbent assay (Roche Applied Science).
Nuclear Run-on Transcription Analysis-Nuclei isolation from HUVECs and run-on transcription analysis were carried out as described by Greenberg et al. (44) with few modifications. Briefly, nuclei suspension was mixed with equal volume of 2ϫ reaction medium (10 mM Tris-HCl, pH 8.0, 5 mM MgCl 2 , 0.3 M KCl, ATP, GTP, CTP (1 mM each one), 5 mM dithiothreitol, 120 units of RNase inhibitor, 100 Ci of [␣-32 P]UTP) and incubated for 30 min at 30°C. The reaction was terminated with the addition of DNase I (140 units) for 5 min at 30°C. Subsequently, 10 l of proteinase K (20 mg/ml) and 100 l of SDS/Tris buffer (5%SDS, 0.5M Tris-Cl, 0.12 M EDTA) were added and the mixture incubated for 30 min at 42°C.
Radioactively labeled RNAs were extracted with RNAgents total RNA isolation system (Promega Corp) and hybridized to cDNAs (5 g) immobilized on nitrocellulose membrane. Ang2 and ␤-actin cDNA slot blotting to nitrocellulose was performed with the Bio-Dot SF Microfiltration Apparatus (Bio-Rad) following the manufacturer's protocol. Hybridization was performed with 1-3 ϫ 10 6 cpm of nascent RNA and Quickhyb solution (Stratagene) following the manufacturer's instructions. This experiment was repeated three separate times and quantification was performed using a PhosphorImager system.
Measurement of RNA Stability-Cells were grown under normoxia or FIG. 1. Increase of Ang2 mRNA levels by hypoxia. A, HUVEC cells were exposed to normoxia (C) or 1% O 2 for 3-24 h and total RNA isolated for Ang2 Northern blot. B, RT-PCR analysis of Ang2 and Ang2 443 isoforms in HUVECs exposed to normoxia (C) or 1% O 2 (3-24 h). C, bEnd3 cells were exposed to normoxia (C) or 1% O 2 for 3-24 h and total RNA isolated for Ang2 Northern blot. ␤-actin mRNA and 18 S rRNA signals confirm equal amount of total RNA loading. hypoxia for 24 h. Thereafter, actinomycin D (ActD, 5 g/ml) was added and incubation continued under the same conditions. Cells were harvested for RNA isolation at 0, 1, 2, 4, 6, 8, 12 h after the addition of ActD. Total RNA was prepared using RNAqueous (Ambion) according to the manufacturer's protocol. Ang2 and 18 S rRNA were detected by Northern blot analysis as described above. Quantification was performed by using a PhosphorImager system. The half-life of Ang2 mRNA was calculated by using the log-linear plot of percentage of remaining Ang2 mRNA levels versus time. This experiment was repeated three separate times.
Western Blot Analysis-Cells were lysed with ice-cold lysis radioimmune precipitation assay buffer (1ϫ PBS, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS) supplemented with protease inhibitors (1 g/ml leupeptin, 0.5 g/ml aprotinin, 1.5 g/ml pepstatin, 0.1 g/ml phenylmethylsulfonyl fluoride). Lysates were centrifuged at 10,000 ϫ g for 10 min at 4°C, and supernatants were collected. Protein concentrations were determined by Bradford protein assay with bovine serum albumin as standard (Bio-Rad). Samples (30 g of protein) were electrophoresed on 7.5% or 12% SDS-PAGE under reducing conditions and transferred to nitrocellulose membranes (Bio-Rad) by standard procedures. Membranes were blocked with 10% nonfat dry milk and were incubated with either anti Ang2 (R&D Systems), VEGF (R&D Systems), HIF-1␣ (Novus Biologicals), COX-1(Santa Cruz Biotechnology), or COX-2 (Santa Cruz Biotechnology) antibodies. After washing, membranes were incubated with the appropriate horseradish peroxidaseconjugated secondary antibodies. Antigen-antibody complexes were visualized by enhanced chemiluminescence detection (ECL, Amersham). Ang2 and VEGF were identified as single bands of about 68 kDa and 23kDa, respectively, which are consistent with published sizes. HIF-1␣ migrated with a characteristic diffuse pattern (ϳ120 kDa). Recombinant human Ang2 (R&D systems) was used as positive control for Ang 2 immunoblot. Crude nuclear extracts of Hep3B cells (ATTC, CRL-1830) exposed to 1 or 21% oxygen were used as positive controls (30 g of protein) in HIF-1␣ Western blot analysis. Lysates from macrophages treated with LPS/IFNg (Transduction Laboratories) were used for positive control for COX.
ELISA and Prostanoids Measurement-The secreted Ang2 and VEGF protein levels were measured in cell culture supernatants by sandwich ELISA using the reagents and the protocol supplied with the Human VEGF and Ang2 DuoSet ELISA kit (R&D Systems). PGE 2 , PGF 2␣ , 6-keto-PGF1␣, and TXB 2 (the stable hydrolysis products of PGI 2 and TXA 2 , respectively) were determined in the cultured cell supernatants using commercial immunoassay kits (R&D systems). Prostanoids results were expressed as either pg/ml of media or pg/mg of protein.
Protein concentration in the cell culture dish was measured after washing with phosphate-buffered saline and extracting with 0.62 M NaOH.
Antisense Oligonucleotide Treatment-The phosphorothioate antisense oligonucleotides for COX-1 and COX-2 used here were identical to the ones previously used by Tsujii et al. with HUVECs (38). These sequences were as follows: COX-1 antisense: 5Ј-AGAACCGGAGCAAGA and COX-2 antisense: 5Ј-GGAAACATCGACAGT-3Ј. Corresponding FIG. 2. Hypoxic regulation of Ang2 transcriptional rate and mRNA stability. A, for nuclear run-on analysis, HUVECs were exposed for 24 h to normoxia or hypoxia (1% O 2 ), nuclei isolated and labeled nuclear RNA samples hybridized to immobilized cDNAs of Ang2 and ␤-actin. A representative blot from three independent experiments is shown. B, effect of hypoxia on Ang2 mRNA stability. HUVECs were exposed to normoxia or 1% O 2 for 24 h and then treated with actinomycin D (ActD, 5 g/ml) for the indicated times. Ang2 mRNA and 18 S rRNA levels were evaluated by Northern blot and quantified using a PhosphorImager system. Data points in the log-linear plot indicate the percentage of Ang2 signal remaining after ActD addition.

FIG. 3. Hypoxia increases Ang2 protein expression and secretion.
A, HUVEC and bEnd3 cells were exposed to normoxia (C, 24 h) or hypoxia (1% O 2 , 3-24 h), and total cell lysates prepared for Western blot analysis of Ang2. Recombinant human Ang2 was used as positive control (ϩ). B, secreted Ang2 and VEGF were measured in conditioned media of HUVECs maintained at normoxia (C, 24 h) or hypoxia (3-24 h). Data are expressed as mean Ϯ S.D. from three different experiments. * or **: p Յ 0.05 compared with normoxic control for Ang2 (*) and VEGF (**). sense oligonucleotides were used as controls. HUVECs were transfected with either sense or antisense oligonucleotide (0.5 M) with Lipo-fectAMINE transfection reagent for 4 h according to the manufacturer's protocol (Invitrogen). Twelve hours after transfection, treated cells were exposed to hypoxia or normoxia for 24 h, and then cells were lysed for immunoblot analysis.
COX Enzyme Analysis-COX activity was analyzed in HUVECs cell lysates using the COX activity assay kit from Cayman Chemical following the manufacturer instructions. This colorimetric assay measures the peroxidase activity of the enzyme using N,N,NЈ,NЈ-tetramethyl-p-phenylenediamine (TMPD) as the reducing co-substrate. COX activity was expressed as the rate of oxidation of TMPD in nmol/min/mg protein. This kit includes purified ovine COX-1 as positive control and the inhibitors SC-560 (COX-1 inhibitor) and DuP-697 (COX-2 inhibitor). Purified COX-1 activity was inhibited by SC-560 but not by DuP-697.
Statistical Analysis-All results are shown as means Ϯ S.D., and p Ͻ 0.05 was selected as the statistically significant value. Statistical analysis was carried out using SPSS 11.5 software. Analysis of variance with Tukey correction was employed for the statistical comparisons among groups. Student's t test was employed to assess differences between the two groups.

RESULTS
Hypoxia Increases Ang2 mRNA in Endothelial Cells-Subconfluent HUVEC cultures were exposed to hypoxia (1%O 2 ) for different time periods and Ang2 mRNA analyzed by Northern blot. Our results showed basal low Ang2 mRNA levels in normoxic cells that increased with hypoxia (Fig. 1A). This upregulation was evident at 6 h and continued to increase at 24 h. We also performed RT-PCR analysis in HUVEC to assess the differential response of Ang2 splicing isoforms to hypoxia. Although three human Ang2 mRNA isoforms have been described (45,46), the primers used in this study only recognized the Ang2 and Ang2 443 isoforms (also known as Ang2A and Ang2C, respectively). RT-PCR analysis showed higher basal levels of Ang2 than Ang2 443 at normoxic conditions. Exposure to hypoxia resulted in an increase of both Ang2 and Ang2 443 mRNA, which was observed at 3-24 h of hypoxic exposure (Fig.  1B). The identities of these RT-PCR products were confirmed by sequence analysis. bEnd3 cells, a mouse brain microvascular endothelial cell line, also displayed time-dependent induction of Ang2 mRNA level in response to hypoxia (Fig. 1C). Only one mouse Ang2 isoform has been reported so far.
Analysis of cell viability of HUVEC and bEnd3 cells exposed to 1% O 2 for 24 h was assessed by measuring levels of cytoplasmic histone-associated DNA fragments and extracellular lactate dehydrogenase activity. In agreement with previous observations (47,48), both assays indicated no significant cell death during 24 h of hypoxia compared with normoxic controls (data not shown).
Effect of Hypoxia on Ang2 mRNA Transcription and Stability-Induction of Ang2 mRNA expression during hypoxia could result from an increased gene transcription, increased mRNA stability, or both. To investigate the effect of hypoxia on Ang2 transcription rate, we performed nuclear run-on experiments using HUVECs exposed to normoxia or hypoxia for 24 h. As shown in Fig. 2A, normoxic cells presented basal Ang2 transcriptional activity. Hypoxia increased the rate of transcription of Ang2 more than 3-fold (3.19 Ϯ 0.4, n ϭ 3; p Ͻ 0.05). Transcription of the ␤-actin gene served as an internal control.
To test the possibility that there was also a component of post-transcriptional stabilization of Ang2 mRNA in response to hypoxia, HUVECs were maintained at normoxia or hypoxia for 24 h. Then, transcription was blocked with actinomycin D (5 g/ml), and incubation was continued under the same conditions (normoxia or hypoxia). Cells were harvested at different time points after the addition of actinomycin and the time course for the decay of Ang2 mRNA and 18S rRNA were measured by Northern blot. The stability of Ang2 mRNA was increased by hypoxia, its half-life increased from 2.7 Ϯ 0.4 h under normoxic conditions to 4.9 Ϯ 1.0 h under hypoxia (n ϭ 3; p Ͻ 0.05) (Fig. 2B). These results imply that hypoxia-mediated up-regulation of Ang2 mRNA is due to increases in the rate of transcription of Ang2 gene and in the stability of its mRNA.
Hypoxia Stimulates the Production and Secretion of Ang2 Protein-We next assessed whether induction of Ang2 mRNA levels by hypoxia resulted in a parallel increase of Ang2 protein levels. Huvec and bEnd3 cells were exposed to normoxia or hypoxia for up to 24 h and cell lysates prepared for analysis of Ang2 production by Western blot. In addition, conditioned medium from HUVECs was assayed for secreted Ang2 by ELISA.
Western blot analysis revealed undetectable Ang2 protein levels in HUVECs at normoxic conditions, whereas hypoxia induced Ang2 levels with detectable increase as early as 3 h and further stimulation at 24 h (Fig. 3A). In bEnd3 cells, basal Ang2 protein levels were found in normoxic cells, which increased with 6 to 24 h of hypoxic exposure (Fig. 3A). Similarly, hypoxia induced a progressive increase of secreted Ang2 levels in HUVECs with detectable increase as early as 6 h and further stimulation at 24 h (ϳ43-fold above control levels) (Fig. 3B). The time course of this stimulation was similar to the one for secreted VEGF (Fig. 3B).
HIF-1 Role in Ang2 Induction by Hypoxia-To investigate the transcriptional mechanism underlying Ang2 expression, we examined the role of HIF-1, a transcription factor responsible for hypoxic activation of several genes (49). HIF-1 is composed of an oxygen-regulated HIF-1␣ subunit and a constitutively expressed HIF-1␤ subunit.
We first examined the expression of HIF-1␣ in HUVEC and bEnd3 cells exposed to hypoxia (1% O 2 , 24 h) or to pharmacological HIF-1 inducers, cobalt chloride (CoCl 2 , 100 M, 24 h) and desferrioxamine (DFO, 100 M, 24 h). As expected, both cell types exhibited notable HIF-1␣ protein accumulation during hypoxic exposure, CoCl 2 or DFO treatment (Fig. 4A). Moreover, hypoxia, CoCl 2 , and DFO increased known HIF-1 target genes, Glut-1 and VEGF mRNA levels, in HUVECs (Fig. 4B). However, hypoxia but not CoCl 2 or DFO caused an increase in Ang2 mRNA and protein levels in HUVECs or bEnd3 cells (Fig.  4, B and C). Accordingly, ELISA analysis showed that hypoxia induced an increase of secreted Ang2 and VEGF while CoCl 2 and DFO induced only VEGF but not Ang2 (Fig. 4D). In order to eliminate the possibility that 24 h of CoCl 2 or DFO treatment was not sufficient time for Ang2 mRNA up-regulation, we prolonged these treatments for 48 h. Even under this condition, we were unable to observe any Ang2 up-regulation (data not shown). To explore further the role of HIF-1 in the regulation of Ang2 expression, we analyzed Ang2 mRNA levels in a mouse hepatoma cell line deficient in HIF-1␤, Hepa1 c4. These cells are derived from the parental Hepa1 c1c7 cell line and lack HIF-1 DNA binding activity (50). As shown in Fig. 5, hypoxic exposure induced Ang2 mRNA levels in wild-type Hepa1 c1c7, its mutant derivative Hepa1 c4 and VT2. The latter is a c4 cell line transfected with a plasmid expressing HIF-1␤. In contrast, Glut-1 and Epo mRNA levels, known HIF-1 targets, were in-creased by hypoxia in Hepa1 c1c7 and VT2 but not in Hepa1 c4 (Fig. 5). Taken together, these results do not support a role for HIF-1 in hypoxic expression of Ang2.
VEGF and Hypoxia Induction of Ang2 Expression-The capacity of VEGF to induce Ang2 expression in some endothelial cells raises the possibility that hypoxia-induced VEGF may be responsible for the induction of Ang2 expression during hypoxia (20,22). To assess this possibility, we first tested the capacity of recombinant human VEGF 165 (rhVEGF 165 ) to induce Ang2 protein expression in HUVECs under normoxic conditions. At 100 ng/ml, rhVEGF 165 induced a transient increase of Ang2 protein levels, which was evident at 3 h but decreased drastically at 24 h (Fig. 6A). When cells were exposed to various concentrations of VEGF for 6 h, a dose-dependent increase of FIG. 6. VEGF and hypoxia-induction of Ang2. A, representative immunoblots showing time-and concentration-dependent effect of VEGF 165 on Ang2 protein levels. HUVECs were treated with vehicle (c) or recombinant human VEGF 165 under normoxic conditions. B, effect of VEGF-neutralizing antibody (Ab-VEGF) on hypoxia-stimulated Ang2 protein levels. HUVECs were exposed to normoxia or 1% O 2 together with Ab-VEGF (0.5-10 g/ml). Inhibition of VEGF receptors partially affected hypoxia-induced Ang2 protein levels. HUVECs were exposed for 24 h to normoxia or 1% O 2 in the presence of increasing concentrations of VEGF-tyrosine kinase inhibitor (VEGF-TKinh, Flk-1 and Flt-1 inhibitor) (C) or SU-1498 (Flk-1 selective inhibitor) (D).
Ang2 protein was observed (Fig. 6A). Next, we tested whether treatment of HUVECs with a VEGF-neutralizing antibody or inhibitors of VEGF receptors affected hypoxic-mediated Ang2 expression. As shown in Fig. 3B, exposure of HUVECs to hypoxia for 24 h stimulated secretion of VEGF. Treatment with VEGF-neutralizing antibody during hypoxic exposure only decreased hypoxia-induced Ang2 protein levels partially, even though concentrations of up to 10 g/ml were tested. Similarly, SU1498 (Flk-1 inhibitor, 1-100 M) and CB676475 (Flt-1/Flk-1inhibitor, 1-100 M) partially prevented hypoxia-stimulated Ang2 expression (Fig. 6, C and D). Similar results were obtained with bEnd3 cells (data not shown). Taken together, these data suggested that hypoxia-induced VEGF production, and the subsequent VEGF receptors activation partially contribute, but could not account completely for the induction of Ang2 expression during hypoxia.
Role of COX-2 on Ang2 Expression-It has been reported that hypoxia increases COX-2 expression in HUVECs (41,42) and induces PGE 2 production in different endothelial cell types (51,52). In addition, PGE 2 was found to be able to increase VEGF expression in different cell types (53,54). As such, we sought to investigate whether activation of COX-2 is involved in the induction of Ang2 expression during hypoxia.
We first examined the effect of hypoxia (1% O 2 ) on COX isoenzymes expression, COX activity and prostanoids production on HUVECs. In agreement with previous reports (41,42), hypoxic exposure only affected mRNA and protein levels of COX-2 but not COX-1 (Fig. 7A). COX-2 protein levels increased by 2.7-fold at 6 h and reached a 4.9-fold increase at 24 h of hypoxia (Fig. 7A).
Measurement of COX enzymatic activity in cell lysates revealed a basal activity under normoxic conditions that increased by 192% after hypoxic exposure. Treatment of normoxic cells with SC-560 (preferential COX-1 inhibitor) drastically reduced the constitutive COX activity (92% reduction) whereas treatment with COX-2 preferential inhibitors, NS-398 (100 M) and PTPBS (10 M), did not have any effect. In contrast, hypoxic-induced COX activity was modestly decreased by SC-560 (22% reduction) while NS-398 (25-100 M) or PTPBS (1-10 M) drastically reduced it. In fact, hypoxic COX activity reached normoxic levels with 50 M NS-398 or 5 M PTPBS (Fig. 7B). These results suggested that COX activity in normoxic cells is driven mainly by COX-1 while, in agreement with increased expression, COX-2 seems to be the major contributor to the increased hypoxic COX activity.
The hypoxic up-regulation of COX activity and COX-2 expression was associated with an increase in the production of certain prostanoids. The amount of PGE 2 and 6-keto-PGF1␣ released in the culture medium during hypoxia rose 9.8-fold and 6.3-fold, respectively, relative to normoxic controls (Fig.  7C). Levels of secreted PGF2␣ and thromboxane B 2 (TXB 2 ) did not changed with hypoxia. 6-keto-PGF1␣ and TXB 2 are the stable forms of prostacyclin (PGI 2 ) and TXA 2 , respectively; and are produced by non-enzymatic hydration. PGD 2 was not analyzed in this study.
Next, we tested the effect of inhibition of COX-2 activity on Ang2 expression during hypoxia. HUVECs were exposed to normoxia or hypoxia for 24 h and co-treated with a non-selective COX inhibitor (indomethacin) or COX-2 preferential inhibitors (NS-398, PTPBS, and rofecoxib). Cell lysates were prepared for Ang2 immunoblot analysis and conditioned medium were assayed for PGE 2 and PGI 2 production.
As depicted in Fig. 8, all these inhibitors resulted in a dosedependent reduction of hypoxic Ang2 up-regulation. Furthermore, this effect on Ang2 expression closely correlated with the ability of these inhibitors to decrease hypoxic PGE 2 and PGI 2 FIG. 7. Effect of hypoxia on COX-2 expression and prostanoid production. A, Northern (upper panel) and Western blot (lower panel) for COX-2 and COX-1 were performed in HUVECs exposed to either normoxia (C) or 1% O 2 (3-24 h). Lysates of macrophages treated with LPS/IFNg were used as positive controls for immunoblots (ϩ). B, COX enzymatic activity was measured as described under "Experimental Procedures" in HUVECs exposed to normoxia or hypoxia (24 h) in the presence of vehicle (no inhib), COX-1 inhibitor (SC-560), or COX-2 inhibitors (NS-398 and PTPBS). Data represent the mean Ϯ S.D. (n ϭ 3). *, p Ͻ 0.05 versus untreated normoxia. Compared with untreated hypoxia, COX activity of all experimental groups were significantly lower (p Ͻ 0.05). C, levels of secreted prostanoids were measured in the culture medium of HUVECs exposed to normoxia or hypoxia (24 h). The amounts of TXA 2 and PGI 2 were measured as TXB 2 and 6-keto-PGF1␣, respectively. Data are expressed as mean Ϯ S.D. from three independent experiments. *, p Յ 0.05 compared with normoxia. production at the same doses (Fig. 9). Indomethacin at a concentration of 0.5 and 5 M reduced hypoxic Ang2 expression by 45 and 80%, respectively; while these concentrations reduced hypoxic PGE 2 by 52 and 95% (Figs. 8A and 9). Rofecoxib at a concentration of 2.5 M caused a 69% reduction of hypoxic Ang2 and a 72% decrease of hypoxic PGE 2 ; while 10 M of this drug caused an 81% reduction of hypoxic Ang2 and an 87% decrease of hypoxic PGE 2 (Figs. 8D and 9). Similar results were obtained with NS-398 and PTPBS, although higher concentrations were required for the former. NS-398 at a concentration of 25 M and 100 M reduced hypoxic Ang2 by 62 and 86%, respectively (PGE 2 was reduced by 63 and 90%, respectively) (Figs. 8B and 9). NS-398 and PTPBS also decreased hypoxia-induced Ang2 levels in bEnd3 cells while they did not affect basal normoxic Ang2 levels (data not shown). No significant cell death was seen using these concentrations of COX-2 inhibitors compared with control (data not shown).
Because these COX-2 inhibitors, at the concentrations used in this study, may have other molecular targets that are not necessarily related to their COX-2 inhibitory effect, we also decreased COX-2 expression by antisense oligonucleotides. As shown in Fig. 10A treatment of HUVECs with COX-1 or COX-2 antisense oligonucleotides under normoxic or hypoxic conditions specifically suppressed COX-1 or COX-2 protein expression, respectively. In addition, antisense inhibition of COX-2 decreased the amount of secreted PGE 2 and PGI 2 during hypoxia to normoxic levels (from 4197 to 383 pg/mg of protein for PGE 2 and from 1409 to 232 pg/mg of protein for PGI 2 ) but it did not affect normoxic levels of these prostanoids.
In agreement with our results with COX-2 pharmacological inhibitors, antisense inhibition of COX-2 but not of COX-1 expression resulted in a considerable decrease of the hypoxic Ang2 up-regulation (Fig. 10A). Both COX-2 and COX-1 sense oligonucleotides did not affect the expression of any COX isoforms, PGE 2 production or Ang2. Addition of exogenous PGE 2 (100 nM) was able to reverse the inhibitory effect of COX-2 antisense on hypoxic Ang2 protein levels and did not potentiate the inductive effect of hypoxia (Fig. 10B). Collectively, these results implicate a COX-2-dependent pathway in the regulation of Ang2 during hypoxia.
To explore further the potential capacity of prostanoids on regulating Ang2 expression, we tested whether exogenous prostanoids were capable of inducing Ang2 under normoxic conditions. We found that PGE 2 (10 nM to 2 M, 24 h) increased Ang2 protein levels in a dose-dependent manner (Fig. 11A). On the other hand, higher concentrations (Ͼ2 M) of Iloprost (a stable form of PGI 2 ) were needed to induce Ang2 (Fig. 11A). PGF2␣ and U46619 (TXA 2 receptor agonist), at 1-5 M, did not affect Ang2 expression (Fig. 11A).
Considering our results implicating VEGF in the induction of Ang2 expression and previous reports showing that COX-2 activity and PGE 2 regulates VEGF expression in some cell types (38,54,55), it is possible that the effect of PGE 2 on Ang2 induction is secondary to its effect on VEGF. However, the induction of Ang2 by PGE 2 in HUVECs was only marginally affected by inhibitors of the VEGF receptors (Fig. 11B). This result suggests that PGE 2 is capable of inducing Ang2 expression independently of VEGF signaling.

DISCUSSION
Angiogenesis is a critical physiological response to low oxygen availability. The mechanisms involved in the vascular response to hypoxia are likely to be quite complex, involving the activation of an array of genes necessary for the cascade of events that culminates in the production of mature new vessels.
The present study confirms the ability of endothelial cells to increase Ang2 expression upon hypoxic exposure. In HUVECs, which express at least two alternative splicing isoforms, both Ang2 and Ang2 443 mRNA were induced. Moreover, we showed that the increase in the steady-state of Ang2 mRNA occurred by transcriptional regulation and by post-transcriptional increase in mRNA stability.
Because hypoxia induces several HIF-1 target genes involved in angiogenesis, it was conceivable that Ang2 expression might be regulated by the same oxygen-dependent mechanism. However, our data showed that both CoCl 2 and DFO did not affect Ang2 expression. These chemical compounds mimic hypoxia by causing the stabilization of HIF-1␣ (56,57). Thus, as shown in Fig. 4B, only known HIF-target genes such as Glut-1 and VEGF are likely to be induced by both compounds. In addition, considering that tumor cells, besides endothelial cells, FIG. 9. Effect of COX-2 inhibitors on hypoxia-induced PGE 2 and PGI 2 production. HUVECs were exposed for 24 h to either normoxia (C) or 1% O 2 (H) in the absence or presence of indomethacin (non-selective COX inhibitor), NS-398, PTPBS, or Rofecoxib at the indicated concentrations. Levels of secreted PGE 2 (A) and PGI 2 (B) were measured in the culture medium. Data are expressed as mean Ϯ S.D. from three independent experiments. *, p Յ 0.05 compared with untreated hypoxia.
FIG. 10. Reduction of endogenous COX-2 using antisense oligonucleotide decreased hypoxic induction of Ang2. A, HUVECs were transfected with sense (s) or antisense (as) oligonucleotides of COX-1 or COX-2 and then exposed to normoxia (C) or hypoxia (H, 1% O 2 , 24 h). Total cell lysates were prepared for COX-1, COX-2, and Ang2 Western blots. B, HUVECs previously transfected with COX-2 sense (s) or antisense (as) oligonucleotides were exposed to normoxia (C) or hypoxia (H) for 24 h in the absence or presence of exogenous PGE 2 (100 nM). ␤-actin confirms equal amount of protein loading. also express Ang2, we used the HIF-1␤-deficient Hepa1 c4 cell to study HIF-1 role on Ang2 expression. We found that Hepa1 c4, like its wild-type counterpart, up-regulates Ang2 upon hypoxia treatment. Taken together our results indicate that hypoxia-driven Ang2 expression is independent of the HIF pathway. Therefore, hypoxia must recruit other regulatory transcription factors in the hypoxic endothelium that may regulate Ang2 transcription. HIF-1-independent hypoxia-inducible pathways have previously been reported for other genes, including COX-2, IAP-2, IL-6, and 5-aminolevulinate synthase (41,42,48,58,59).
It is reasonable to expect that hypoxia might stimulate Ang2 production through both direct and indirect signaling pathways. Besides hypoxia, some growth factors and cytokines such as VEGF, bFGF, TNF-␣, angiotensin II, and thrombin induce Ang2 expression in cultured endothelial cells (20,22,60,61,62). In this report, we showed that hypoxia stimulates production of both VEGF and Ang2 in HUVECs. However, our results using either neutralizing VEGF antibody or pharmacological inhibitors of VEGF receptors suggested that VEGF signaling participates in but could not account completely for the increased Ang2 expression during hypoxia. We have not identified conclusively which of the VEGF receptors, Flt-1 or Flk-1, is involved in regulating Ang2 expression because blocking antibodies specific for each receptor are not commercially available. Nevertheless, because SU-1498 (Flk-1 inhibitor) had a similar effect as the VEGF tyrosine kinase inhibitor (Flt-1/Flk-1 inhibitor), Flk-1 receptor appears to be the most likely candidate. It is important to mention that our results are not in agreement with a recent report showing that hypoxic stimulation of Ang2 in bovine retinal endothelial cells is independent of VEGF (22). The discrepancy between our study and this prior work could reflect intrinsic differences between endothelial cells from distinct vascular beds.
Based on growing evidence implicating COX-2 and its derived prostanoids in angiogenesis (37), we explored the possibility that this enzyme may be an important mediator of hypoxic induction of Ang2. Our findings indicate that COX-2 preferential inhibitors (NS-398, PTPBS, rofecoxib) decreased the expression of Ang-2 during hypoxia at doses that effectively inhibited the hypoxic induced PGE 2 /PGI 2 . In our study, the doses of NS-398 required to significantly decrease hypoxic Ang2 expression are higher than the ones reported for inhibition of PGE 2 by Caco-2 cells engineered to overexpress COX-2 (38). However, these doses are identical to those reported for the inhibition of COX-2-dependent HUVEC tube formation, spreading and migration (36,39). Because it is conceivable that the effect of rofecoxib, NS-398, or PTPBS, in the concentration range that we used, might be related to their effect on COX-2 independent pathways, we also tested the effect of decreasing COX-2 expression on Ang2. Reduction of COX-2 expression by antisense oligonucleotide significantly reduced the hypoxic upregulation of Ang2 strongly indicating that indeed COX-2 and derived prostanoids could be important in regulating Ang2 gene expression in endothelial cells. In addition, we observed an increase of PGE 2 and PGI 2 during hypoxia that can be reversed to normoxic levels by COX-2 inhibitors and COX-2 antisense. Notably, of several prostanglandins tested, only PGE 2 and PGI 2 were able to up-regulate Ang2 protein levels under normoxic conditions. Further studies will be required to define which COX-2 metabolite participates in the regulation of Ang2 in the hypoxic endothelium and the signaling pathways by which PGE 2 /PGI 2 regulates Ang2 expression. Nevertheless, since decreasing COX-2 activity did not abolish completely hypoxic-induced Ang2 expression, it is likely that additional COX-2 independent pathways are operative in regulating endothelium Ang2 expression.
Although endothelial cells are considered to be an important source of Ang2 during embryonic and postnatal neovascularization, other cellular types like perivascular cells also express Ang2 which may be relevant for specific tissue circumstances (10,12). It was previously reported that hypoxia modestly up-regulated Ang2 mRNA in cultured aortic smooth muscle cells but not in cultured retinal pericytes (20,63). Of note, we have found an up-regulation of Ang2 protein levels in smooth muscle cells derived from umbilical artery, which was partially decreased by NS-398 and PTPBS. 2 It remains to clearly establish if COX-2 metabolic products participate in the regulation of hypoxic Ang2 in smooth muscle cells.
In conclusion, the present investigation demonstrated a complex signaling pathway regulating Ang2 production in HUVEC and bEnd3 cells. Hypoxia not only affects Ang2 transcriptional rate and mRNA stability, but also secondarily modulates Ang2 expression through its effect on VEGF and COX-2 activity. It is intriguing to know whether these autocrine/paracrine regulatory pathways may operate in vivo, not only in hypoxic tissue, but also in neoplastic conditions.
Although we determined that HIF-1 was not involved in the hypoxic transcriptional regulation of Ang2, the molecular structure of the hypoxia-responsive element in the Ang2 promoter, the identity of the transcription factor that drive the hypoxic-expression of this gene and the role of prostanoids in the regulation of these transcription factors remain to be unveiled. tion) for the generous gift of mouse hepatoma Hepa-1c1c7 and its variants.