Involvement of COX-2/PGE2 signalling in hypoxia-induced angiogenic response in endothelial cells

Abstract To evaluate the impact of hypoxia on the angiogenic capability of endothelial cells (ECs), and further investigate whether the cyclooxygenase-2 (COX-2)/prostaglandin E2 (PGE2) signalling is involved in the angiogenic response of ECs to hypoxia. We explored the impact of various periods (1, 3, 6, 12, 24 hrs) of hypoxia (2% O2) on human umbilical vein endothelial cells (HUVECs) in vitro. We observed cell viability, migration, tube formation, analysed COX-2, vascular endothelial growth factor (VEGF), AQP1 mRNA transcription, protein expression and measured PGE2, VEGF protein concentration in cell supernatants. Then we treated HUVECs with COX-2 selective inhibitor NS398, EP1/2 combined antagonist AH6809 and exogenous PGE2 to investigate the role of COX-2/PGE2 signalling in the angiogenic response of ECs to hypoxia. The results demonstrated that short-term hypoxic treatment enhanced HUVECs proliferation, migration, tube formation, significantly up-regulated COX-2, VEGF, AQP1 mRNA level, protein expression and promoted PGE2, VEGF release. The pharmacological inhibition study revealed that exposure of HUVEC to NS398 and AH6809 under hypoxia impaired the biological responses of ECs to hypoxia. Exogenous PGE2 augments the effects of hypoxia on HUVECs, and partially reversed the inhibitory effects of NS398 on HUVECs proliferation and angiogenic capability. Short-term hypoxic treatment enhanced angiogenic capability of ECs, and COX-2/PGE2 signalling may play a critical role in the biological response of ECs to hypoxia.


Introduction
Angiogenesis, formation of new capillaries which enables delivery of oxygen and nutrients, is essential for tumour growth and metastasis [1], ischaemic disorder recovering [2] and periodontal tissue remodelling [3]. It is a highly organized process and involves a complex sequence of steps that include endothelial cells (ECs) proliferation, maturation and assembly [4]. As the main cell type involved in angiogenesis, ECs migrate from capillaries toward the pathological zone, undergoing various stress conditions including hypoxia and acidic pH [5]. Moreover, the endothelium plays a predominant role in modulating many aspects of vascular homeostasis. Dysfunction of ECs structure and function may contribute to the overall endothelial dysfunction in a range of clinic settings, including ischaemic diseases [6]. Understanding specific mechanisms underlying angiogenesis from ECs point of view offers the best approach to develop therapies for cancer, ischaemic diseases and modulate the periodontal remodelling process, where neovascularization is either impaired or activated [2].
Hypoxia is a characteristic feature of many physiological or pathological processes in vivo: driving angiogenesis in tumours [7], taking part in ischaemic diseases [8] and participating in bony or soft tissue injury [9]. It has been shown to be an important regulator of blood vessel tone, vessel structure and a potent stimulus of angiogenesis [10]. Hypoxic environment primarily affect the fundamental characteristics of ECs, which # These authors contributed equally to this paper. Tel.: ϩ86-28-85501442 Fax: ϩ86-28-85582167 E-mail: zhzhao@scu.edu.cn situates at the interface between blood and tissue and the first line to sense hypoxia [11]. Vascular ECs cope with hypoxia to maintain vascular homeostasis by expressing a number of genes, which are mediated by a variety of oxygen-sensitive signalling cascades [12]. Interestingly, the presence of hypoxia is always associated with and accompanied by inflammation [13]. For instance, the poorly vascularized regions of tumours with low oxygen tension is closely linked with inflammation in the form of tumour-associated macrophages accumulation [14]; injury in acute myocardial infarction is characterized by inflammatory event of neutrophils infiltration after a period of ischaemia [15]; the inflammatory mediators interleukin-1␤ and ␤-glucuronidase have been shown to present in the gingival crevicular fluid of children undergoing orthodontic treatment [16]. Reciprocity of inflammation, oxidative stress and neovascularization is emerging as an important mechanism underlying numerous processes from tissue healing and remodelling to cancer progression [5]. However, the mechanism of hypoxia-driven angiogenesis and the role of inflammatory mediators in the biological response of ECs to hypoxia remains unclear.
Previous study [17] revealed that short-term hypoxia can directly activate ECs towards a pro-inflammatory phenotype through the synthesis of lipid mediators. ECs produce several PGs, such as PGE2, PGI2 and thromboxane A2 in response to various stimuli [18,19] and these prostanoids, particularly PGE2, have been implicated in angiogenesis [20]. These pro-inflammatory eicosanoids directly stimulate the synthesis of angiogenic factors, promote vascular sprouting, migration and tube formation, and also enhance endothelial survival [20,21]. Specifically, it has been reported that PGE2 stimulates vascular endothelial growth factor (VEGF) expression in osteoblasts [22], fibroblasts [23], as well as in rat ECs [24] and that COX-2-dependent VEGF induction enhances angiogenesis in vivo [25].
Cyclooxygenase (COX) is a rate-limiting enzyme in the prostaglandin (PG) biosynthetic pathway. Cyclooxygenase-1 is found constitutively expressed in a wide range of tissues, while COX-2 is an inducible enzyme that produces prostaglandins during inflammatory and tumorigenic settings [26]. Furthermore, it is one of the genes induced by hypoxia and acts as a mediator of both inflammation and angiogenesis. EC-derived COX-2 is suggested important in angiogenesis [27]. Although there have been many reports regarding the mechanism of angiogenesis, few studies have explored the direct effect of COX-2/PGE2 signalling on angiogenesis.
Vascular endothelial growth factor (VEGF) is a major regulator of endothelial proliferation and migration [28], and is the most potent inducer of angiogenesis and capillary permeability [29]. It has been reported the key role of oxygen tension in regulating the expression of a variety of genes, and VEGF mRNA is induced by exposure to low oxygen tension under various pathophysiological circumstances [30]. Moreover, the hypoxiainduced COX-2 activation may augment PGE2 release, resulting in either an autocrine or paracrine action that enhances expression of VEGF through the generation of hypoxia inducible factor (HIF)-1␣ [31,32].
The aquaporins (AQP) are a family of proteins which mediate water resorption in mammalian renal tubules and several other tissues [33]. Aquaporin-1 is one family member of AQP with a molecular weight of 28 kDa. Previous studies validated that AQP1 is responsible for the high vascular permeability and interstitial fluid pressure. AQP1 is reported to be required for hypoxiainducible angiogenesis in human retinal vascular ECs [34]. However, the role of COX-2/PGE2 signalling in the VEGF and AQP1 expression of ECs are rarely reported. Therefore, the mechanisms involved in the impact of hypoxia on VEGF and AQP1 expression of ECs remains to be further elucidated from the perspective of inflammation.
In order to better understand how ECs adapt to hypoxic environment, we explored the impact of hypoxia on the human umbilical vein endothelial cells (HUVECs) by observing cell viability, migration, tube formation, analysing COX-2, VEGF and AQP1 expression and measuring PGE2, VEGF concentration in supernatants of HUVECs with different periods of hypoxic treatment. Then to further investigate into the potential role of COX-2/PGE2 signalling in the biological response of ECs to hypoxia, we used COX-2 selective inhibitor NS398, E-prostanoid receptor 1/2 (EP1/2) combined antagonist AH6809 and exogenous PGE2 to treat HUVEC under hypoxia. The findings of present study may elucidate the relationship between inflammation and angiogenesis in human ECs response to hypoxia, and provide theoretical basis for further understanding of the mechanisms involved in cancer, ischaemic disorder and periodontal tissue remodelling.
Cell characterization was evaluated by their morphology on reaching confluence. Besides, immunostaining was performed in order to detect factor VIII-related antigen and CD31 in HUVECs raised to confluence. Briefly, HUVECs were fixed in 4% (w/v) paraformaldehyde for 20-30 min., washed with PBS three times and air-dried at 4ЊC. Then treated with 3% hydrogen peroxide for 15 min. at room temperature to block intrinsic peroxidase and washed with PBS three times. Fixed cells were incubated for 1 hr at 37ЊC with primary antibody anti-factor VIII-related antigen or anti-CD31 (mouse anti-human; Abcam, Cambridge, UK). Subsequently, cells were washed in PBS and incubated with biotin-conjugated rabbit anti-mouse IgG (Santa Cruz, CA, USA) for 1 hr at 37ЊC. Finally, they were incubated with horseradish peroxidase labelled streptavidin for 20 min. at 37ЊC. At least five independent experiments were performed and cultures were examined with a microscope (Nikon, Japan).

Hypoxic treatment of HUVECs
At 80% confluence, HUVECs were placed in 6 well plates at 1 ϫ 10 4 /cm 2 and incubated with 5% CO2, 100% humidity at 37ЊC for two days. The cells were assigned to two groups as hypoxic group (2% O2) and normoxic control group (20% O2), while the former divided into five subgroups (1, 3, 6, 12, 24 hrs) according to different periods of hypoxic exposure. Then, the cells were subjected to hypoxia maintained using a three gas modular hypoxic incubator with CO2/O2 monitoring and CO2/N2 gas sources (Binder, Camarillo, CA, USA). Figure 1 is schematical diagram of the device. Culture medium was pre-equilibrated overnight prior to cell exposure. Cell culture plates were placed in the incubator and saturated with a gas mixture containing 2% oxygen, 5% CO2 and 93% nitrogen for the generation of hypoxia at 37ЊC for defined time periods (1, 3, 6, 12, 24 hrs), each with three replicates and the experiments were done for at least three times. In some experiments, the combined EP1/2 antagonist AH6809 (Sigma-Aldrich), or exogenous PGE2 (Sigma-Aldrich) and/or NS398 (Sigma-Aldrich), a selective inhibitor of COX-2, were also added into the HUVECs medium 4 hrs before hypoxic treatment.

Transwell migration assay
Human umbilical vein endothelial cell motility was measured using 24multiwell insert system (Millipore, CA, USA) with 8 m pore size polycarbonate filter insert that divides the chamber into upper and lower portions. The filters were coated with human recombinant fibronectin. Briefly, HUVECs that were previously starved overnight in 5% foetal calf aerum (FCS; Sigma-Aldrich), were trypsinized and resuspended in 0.1% FCS medium at a density of 4 ϫ 10 5 cells/ml. Two hundred and fifty microlitres of this suspension were then added to the upper chamber of the insert while 750 l of 0.1% FCS was added to the lower chamber. After 3 hrs of adhesion, cells were allowed to migrate across the 8 m pore size Fig. 1 Schematic diagram of the Binder, a three gas modular hypoxic incubator used to simulate hypoxic conditions in vivo. Ambient oxygen concentrations of 2% were maintained using this device with CO2/O2 monitoring and CO2/N2 gas sources. A pure N2 entrance is set for the nitrogen replacement of air in incubator to control the oxygen concentration and achieve the purpose of hypoxia.
polycarbonate membrane under normoxia or hypoxia for 1, 3, 6, 12 and 24 hrs. Cells on the top of the filter were removed by gentle swabbing and the remaining cells on the bottom side of the filter were stained using 0.25% cresyl violet, and counted under a light microscope with an eyepiece grid to visualize set fields. At least four fields in duplicate wells were counted for each condition.

Tube formation assay
Matrigel (BD Biosciences, New Bedford, MA, USA) was added to wells of a cold 96-well plate (45 l/well), and then incubated at 37ЊC for 1 hr to allow gelling. Human umbilical vein endothelial cells were removed from confluent cultures by treatment with trypsin and 0.05% EDTA. The cells were washed in serum-containing medium and then resuspended to 10 5 cells/ml. Into each culture well was added 100 l cell suspension. Cell culture was carried out at 37ЊC for 48 hrs in a humidified 5% CO2 atmosphere for complete penetration of cells into the Matrigel. Then after 12 and 24 hrs of normoxic or hypoxic treatment, cells were observed directly and scored on a scale of 0-5 for tube formation using dark field illumination on a light microscope. Based on quality and number of the tubes, they were assigned numeric values: 0, no real tubes; 1, some poorly formed tubes; 2, some formed tubes; 3, network of tubes both formed and poorly formed; 4, network of formed tubes; and 5, network of well formed tubes. Score results from four random fields in duplicate wells were averaged.

Pharmacological COX-2 inhibition
This pharmacological inhibition study was employed to further investigate the biological response of HUVECs to hypoxia. NS398 was added into the HUVECs medium 4 hrs before hypoxic treatment. Minimum effective concentration (MEC) of inhibitor necessary for significant inhibition of COX-2 protein expression was determined by a concentration gradient test, in which NS398 was used at final concentrations of 0, 5, 10 and 20 M, respectively.

RNA isolation and real-time quantitative PCR
After hypoxic treatment, cells were washed twice with PBS. The total RNA was extracted using TRIzol regent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's protocol. Total RNA was quantified, in a spectrophotometer, at an absorbance  Table 1. The PCR program was initiated by 10 sec. at 95ЊC for pre-degeneration before 40 thermal cycles, each of 5 s at 95ЊC (degeneration), 31 s at 60ЊC (annealing) and 30 s at 72ЊC (elongation). The starting copy numbers of unknown samples were calculated by the 7300 System SDS Software from the standard curve. The housekeeping gene, glyceraldehyde 3-phosphate dehydrogenase (GAPDH), was concurrently amplified in each sample as control and was used for normalization. The cDNA of control HUVECs untreated of hypoxia normalized to the level of GAPDH mRNA have been ascribed a fold induction of 1. Melting curves for each PCR reaction were generated to ensure the purity of the amplification product. Data analysis was performed using the 2 Ϫ⌬⌬Ct method described previously, where GAPDH was used as the reference gene [36].

Western blotting
To obtain whole-cell extracts, cells that were treated with or without hypoxia were washed twice with ice-cold PBS and then lysed and sonicated in a lysis buffer (Keygen total protein extraction kit; Keygen Biotech., Nanjing, China). The cytosolic fraction was collected as the supernatant after centrifugation at 14,000 ϫ g at 4ЊC for 15 min. and assayed it quantitatively with the BCA method. After boiling for 5 min., 20-25 ml of the lysate (50 mg of protein) was applied to SDS-12% PAGE at 80 V for 30 min. and 120 V for 1 hr. The proteins in the gel were then transferred to a PVDF membrane (Millipore). After blocking, the membranes were probed with 1:1000 dilutions of the anti-AQP1 and 1:100 dilutions of the anti-VEGF (Abcam, Hong Kong), followed by the addition of horseradish peroxidase (HRP)-conjugated secondary antibody (diluted 1:5000) at 37ЊC for 1 hr. Immunoreactive proteins were visualized using a chemiluminescence kit (Immobilon Western Chemiluminescent HRP Substrate; Millipore). Band intensities were determined using the ChemiDoc XRS Gel documentation system and Quantity One software (Bio-Rad, CA, USA).

Enzyme-linked immunosorbent assay assays
The supernatant media of HUVECs, stored in aliquots at Ϫ70ЊC, were thawed on ice before the measurement of VEGF and PGE2 using specific

Statistical analysis
All experiments were performed at a minimum of three times. Measurements are expressed as mean Ϯ S.D. Statistical comparisons were made using factorial analysis of variance (ANOVA), for comparing treatments from controls. In MTT, cell migration and tube formation assay, Student's t-test was used to compare between the two groups at same time point. A value of P Ͻ 0.05 was statistically considered significant.

Characterization of HUVECs
The cells exhibited typical cobblestone morphology on reaching confluence ( Fig. 2A), and immunocytochemical staining showed that they expressed factor VIII-related antigen and CD31 ( Fig. 2B and C), which initially suggested that the cells are ECs.

Hypoxia enhances HUVECs proliferation, migration and tube formation
To ensure that hypoxia used in this study (2% oxygen, 5% CO2 and 93% nitrogen in a modular hypoxic incubator) induces typical cell responses to hypoxic stress, we initially verified HUVECs proliferation and viability following hypoxic exposure by MTT assay. As shown in Figure 3A, an up-regulation of HUVECs viability occurs after temporary exposure to hypoxia (1, 3 and 6 hrs), the cells viability was increased gradually with increasing time of hypoxia. At 3 and 6 hrs, it was significantly augmented by hypoxia compared with normoxia control (P Ͻ 0.05) and reached maximum at 6 hrs. With increasing time of hypoxia (12 and 24 hrs), it was decreased, less than those of the normoxia control group (P Ͻ 0.05). Then, we assessed in vitro migration of HUVECs using fibronectin-coated 8 m filters under either normoxic or hypoxic conditions for 1, 3, 6, 12 and 24 hrs. We found that hypoxia induced migration of HUVECs in a time-dependent manner. As compared with the migration value under normoxia of 12 Ϯ 0.8, 63 Ϯ 4.9, 96 Ϯ 7.3, 147 Ϯ 9.4 and 264 Ϯ 17.7 cells/field at 1, 3, 6, 12 and 24 hrs, respectively, and the amount of cell migration induced by various periods of hypoxia were 67 Ϯ 4.2, 118 Ϯ 8.3, 173 Ϯ 12.6, 295 Ϯ 19.5, 347 Ϯ 21.6, which were significantly higher than that of normoxia at each time point (P Ͻ 0.05) (Fig. 3B).
Finally, to study the effect of hypoxia on capacity of HUVECs to form functional capillaries in vitro, wells containing Matrigel were used in a tubeforming assay. Human umbilical vein endothelial cells were observed after 12, 24 hrs of normoxic or hypoxic treatment and analysed for the quality and number of the tubes formed in the gel. Overall, hypoxia led to a significantly higher level of tube formation than that of normoxia (P Ͻ 0.05) (Fig. 3C).

Hypoxia increases angiogenesis-related gene transcription and protein expression after temporary exposure
To find out the angiogenic capability in ECs under hypoxia, we proceeded to demonstrate the VEGF, AQP1 mRNA synthesis, protein expression and VEGF release of HUVECs after hypoxic treatment. As shown in Figure 4A, the VEGF mRNA was significantly increased by hypoxia at all time points (P Ͻ 0.05). During cultivation, it transiently increased about 2.3 folds, then peaked at 3 hrs and decreased after 6 hrs of the culture, still higher than those of  (B) HUVECs were assayed for migration under normoxia or hypoxia for indicated periods of time. The cells migrate across the membrane were counted under a light microscope with an eyepiece grid to visualize set fields. At least four fields in duplicate wells were counted for each group to generate the bar chart. (C) The tube formation of HUVECs after 12 and 24 hrs of normoxia or hypoxia was viewed by phase-contrast microscopy at ϫ100 magnification, tube formation of each group was scored on a scale of 0-5 based on quality and number of the tubes, score results from four random fields in duplicate wells were averaged to generate the bar chart. *P Ͻ 0.05 versus normoxia group. the control group. Vascular endothelial growth factor protein expression increased significantly at 1 hr after hypoxic exposure and reached summit at 3 hrs (P Ͻ 0.05), at 6 hrs, it decended and there was no significant difference between 6 hrs group and control (P Ͼ 0.05), then declined to the level lower than control at 12 and 24 hrs (P Ͻ 0.05) (Fig. 4B). The accumulation of VEGF protein in ECs supernatants treated by hypoxia was also measured. Data showed that VEGF secretion was significantly increased by hypoxia at all time points. (P Ͻ 0.05), and reached the summit at 6 hrs, with the concentration of 255.29 pg/10 5 cells/ml (Fig. 4C).
Aquaporin-1 mRNA level was transiently increased about 6.4 folds and peaked at 3 hrs by 10.26 folds. After 6 hrs, it decreased and remained higher than the control (P Ͻ 0.05) (Fig. 4D). AQP1 protein expression was significantly up-regulated by hypoxia in short periods (1, 3, 6 hrs, P Ͻ 0.05), and declined to the level lower than the control at 12 and 24 hrs, while the differences were not significant (P Ͻ 0.05) (Fig. 4E).

Hypoxia up-regulates COX-2 mRNA level, protein expression and promotes PGE 2 release in short periods
In order to better understand the effect of hypoxia on the function of HUVECs, we carried out real-time quantitative PCR and Western blot experiments to measure COX-2 mRNA level and protein expression of HUVECs with different periods of hypoxic exposure. The results showed that COX-2 mRNA was at very low levels in unstimulated cells, early minutely induced by 1.35 folds at 1 hr after hypoxic treatment, and reached top at 3 hrs by 2.67 folds, then afterwards declined to level below 2 folds and remained higher than the untreated control group (P Ͻ 0.05) (Fig. 5A). Similarly, COX-2 protein expression also increased significantly at 1 hr and maximized by 1.34 folds at 3 hrs, thereafter declined to levels significantly lower than control (P Ͻ 0.05) (Fig. 5B). Having proved that hypoxia regulates COX-2 at transcription and expression level, we further verified whether it induces COX-2 expression by the production of prostaglandin, determined as PGE2 accumulation in the supernatants of hypoxia treated ECs by ELISA assay. The results indicated that hypoxia stimulated a significant short-term increase in PGE2 release that peaked at 3 hrs, with a concentration of 137.83 Ϯ 11.66 pg/10 5 cells/ml. Then the amount of PGE2 in supernatants went down, possibly following the pattern of COX-2 expression. Late PGE2 release was decreased, making the quantity of it accumulated in the culture medium significantly lower than the control (P Ͻ 0.05), which was consistent with the progression pattern of COX-2 protein expression (Fig. 5C).

Concentration-dependent effect of NS398 on COX-2 expression
The data show that NS398 regulated COX-2 expression in a concentration-dependent manner (Fig. 6). When NS398 was used at 5 M, COX-2 protein expression was decreased to level about that of control. And when NS398 was used at concentrations of 10 and 20 M, COX-2 protein expression was decreased to 24.1% and 21.5% of control group. This indicated that 10 M of NS398 was more effective to inhibit COX-2 activity than 5 M, but had similar effects to 20 M under current experimental conditions. To avoid inflicting excessive injury to the cells and unwanted effects, we selected 10 M as final concentration in the following pharmacological inhibition study.

Selectively inhibition of COX-2 activity by NS398 impaires the biological response of HUVECs to hypoxia
In order to verify whether the endogenous COX-2/PGE2 signalling contributes to the biological response of ECs to hypoxia, we pretreated cell cultures with 10 M of COX-2-selective inhibitor, NS398 for 4 hrs prior to indicate periods of hypoxic treatment. Fig. 6 Effect of the selective COX-2 inhibitor NS398 on COX-2 activity by concentration gradient test. HUVECs were exposed to hypoxia and supplemented with NS398 at indicated concentrations of 0, 5, 10 and 20 M. Cells without hypoxia or NS398 treatment served as control. The COX-2 activity was determined by Western blot analysis. *P Ͻ 0.05, **P Ͻ 0.01 versus control.
First, MTT assay was employed to examine whether NS398 has any effects on HUVECs proliferation. We found that NS398 had an inhibitory effect on HUVECs proliferation under normoxic culture conditions, and the increased proliferation of HUVECs by temporary hypoxia exposure was completely blocked by NS398 (P Ͻ 0.05) (Fig. 7A). Moreover, the difference between the cell viability of NS398 ϩ hypoxia-treated group and the hypoxia-treated group was small especially at earlier hypoxic time points of 1 hr (P Ͼ 0.05), with prolongation of hypoxic time (3, 6, 12 and 24 hrs), the cell viability of NS398-treated groups significantly decreased compared to the untreated groups.
To find out whether NS398 attenuated prostanoid production induced by hypoxia, supernatants were collected and PGE2 concentrations were measured. Results showed that PGE2 synthesis and release were strongly inhibited by NS398 in hypoxia-treated cultures (P Ͻ 0.05) (Fig. 7B). NS398 at the concentration of 10 M completely abolished the inductive effect of PGE2 production by hypoxia, suggesting that the enhancing effect of hypoxia on PGE2 release may be COX-2-dependent.
In order to study the effect of COX-2/PGE2 signalling on angiogenic capability of HUVECs, we examined the effects of NS398 on HUVECs VEGF mRNA transcription, secretion and AQP1 mRNA level. The findings demonstrated that VEGF mRNA transcription was obviously blocked by NS398 (P Ͻ 0.05), and its transcription levels gradually rose to a summit (6 hrs) before declination (P Ͻ 0.05) (Fig. 7C), and VEGF release was also significantly inhibited (P Ͻ 0.05), indicating that NS398 abrogated the inductive effect of hypoxia on VEGF synthesis and secretion (Fig. 7D). Moreover, NS398 significantly down-regulated AQP1 mRNA transcription under normoxic or hypoxic condition (P Ͻ 0.05) (Fig. 7E).

Combined blockade of EP1/2 by AH6809 attenuates the biological response of HUVECs to hypoxia
Then, we proceeded to evaluate the effects of AH6809 (combined EP1/2 antogonist) on HUVECs VEGF mRNA transcription, secretion and AQP1 mRNA level to confirm whether PGE2 receptor EP1/2 is involved in pro-angiogenic reaction of ECs to hypoxia. The data demonstrated that VEGF mRNA transcription and secretion were obviously blocked by AH6809 (P Ͻ 0.05), but the inhibitory effect was much smaller than that of NS398 (P Ͻ 0.05) and hypoxia stimuli could still promote VEGF mRNA transcription and secretion (Fig. 7C and D). In addition, AH6809 significantly decreased AQP1 mRNA transcription to a value which was still higher than that of NS398-treated group under normoxia or hypoxia (P Ͻ 0.05) (Fig. 7E).

Exogenous PGE 2 enhances and partially reversed the inhibitory effects of NS398 on COX-2 expression, cell viability and VEGF, AQP1 mRNA transcription of HUVECs
From the above mentioned results, we could find that the inhibitory effects of COX-2 inhibitor NS398 on VEGF, AQP1 transcription and VEGF release were stronger than that of combined EP1/2 antagonist AH6809. To better understand the effect of PGE2 as a COX-2 downstream effector, we examined the combinational treatment of NS398 (10 M) and exogenous PGE2 (10 M) on COX-2 expression, comparing with that of NS398 or exogenous PGE2 treatment alone. Specifically, HUVECs were treated with 10 M of NS398 and/or 10 M of exogenous PGE2 under normoxia for 4 hrs. Results demonstrated that 10 M of NS398 significantly blocked COX-2 expression, similar to the situation observed in hypoxic condition, while exogenous PGE2 prominently augments the protein expression of COX-2. Interestingly, when treated together with NS398, exogenous PGE2 partially reversed the inhibitory effects of NS398 on COX-2 expression of HUVECs (Fig. 8).
For cell viability, the MTT value of HUVECs under hypoxia was significantly higher than that under normoxia. NS398 exerted a significantly inhibitory effect on HUVECs proliferation both under hypoxia and normoxia, while exogenous PGE2 stimulated them, and combinational treatment of NS398 and PGE2 resulted in a value between the NS398 and PGE2, with hypoxic group significantly higher than the normoxic group (P Ͻ 0.05) (Fig. 9A). Similarly, for VEGF and AQP1 mRNA transcription, NS398 alone also significantly blocked while exogenous PGE2 significantly augmented them, and exogenous PGE2 (10 M) partially reversed the inhibitory effect of NS398 ( Fig. 9B and C).

Discussion
The endothelium plays a predominant role in modulating many aspects of vascular homeostasis. When facing a stimulus like hypoxia, this barrier is able to orchestrate a protective cellular response, including angiogenesis, cell proliferation and apoptosis [37,38]. In the present study, we explored how hypoxia exerts its angiogenic activities on ECs by measuring HUVECs viability, migration, tube formation and angiogenesis related genes/protein (VEGF, AQP1) expression. Vascular endothelial growth factor is a potent angiogenic stimulator with the ability to promote ECs growth [39], enhance ECs survival [40] and induce vasodilatation [41]. Aquaporin-1 is abundantly presented in endothelia of nonfenestrated capillaries [42] and maybe responsible for the vascular hyper-permeability and increases in the hydrostatic interstitial pressure of tissue [43]. Study showed that high degree of AQP1 expression was preferentially associated with enhanced angiogenesis and ECs migration [44,45]. It has been reported that up-regulation of AQP1 by hypoxia may worked as an O2 transporter in ECs and accelerate intracellular hypoxia directly and induce pro-angiogenic molecules in response to hypoxia, thereby leading to hypoxia inducible angiogenesis [46]. In the present  study, we demonstrated specifically that short-term hypoxic treatment: (i) enhanced HUVECs proliferation, motility, tube formation capability; (ii) increased VEGF mRNA level, protein expression and promoted VEGF secretion; (iii) augmented AQP1 mRNA transcription and protein expression. The findings have suggested that temporary hypoxia may promote ECs proliferation and migration, facilitate the angiogenic capability of ECs, and high expression of AQP1 may be associated with the pro-inflammatory phenotype. The adaptive responses of ECs to hypoxia could increase regional blood vessel number, collateral circulation formation and vascular permeability, thus ameliorate local hypoxic microenvironment in ischaemic and remodelling tissue.
Understanding of the pathways involved in ECs hypoxic response and identifying strategies to activate or inhibit this process in tissue remodelling promotion, ischaemic diseases and cancer therapy would have important clinical implications. The important master regulator of angiogenesis under hypoxic conditions, hypoxia-inducible factor-1 (HIF-1), is a dimeric transcription factor composed of HIF-1␣ and HIF-1␤ [47]. It promotes the synthesis of protein that increases the cellular supply with O2 and initiates the defence against hypoxia at different levels [48]. Various traditional studies have focused their efforts on unravelling the importance of HIF-1␣ pathway in tumour progression and angiogenesis, as well as in the adaptive response of ECs to low oxygen availability [49]. Recent data [50,51] also suggest that ECs are highly specialized in acquiring a pro-inflammatory phenotype when encountering stressors. They cope with hypoxia by expressing a number of inflammation-related genes, which are mediated by a variety of signalling cascades [52,53]. However, the reports from the perspective of inflammatory mediators on the relationship between angiogenesis and hypoxia are rare.
Numerous molecules have been implicated in ECs proliferation and angiogenesis. COX-2/PGE2 signalling may be a key regulator of angiogenesis and mediate angiogenesis through multiple mechanisms, including an increase in VEGF transcription and expression [54,55]. Previous investigations validated that amplification of the COX-2/PGE2 pathway in hypoxic colorectal cancer cells might have important implications for stimulating angiogenesis [56]. There is now ample evidence that COX-2 is involved in controlling cell growth and implicated as a key regulator of angiogenesis. The hypoxia-induced COX-2 activation may augment PGE2 release, resulting in either an autocrine or paracrine action that enhances expression of VEGF [57,58]. Stimulation of VEGF production by PGE2 has also been demonstrated in previous studies in various cellular systems [59][60][61]. Our data documented that short-term hypoxia significantly up-regulated COX-2 mRNA transcription, protein expression and promoted PGE2 release (the production of secreted PGE2 is an appropriate measure of COX activity). Accompanied with COX-2 activation and PGE2 secretion, the VEGF and AQP1 expression also increased. Hence, based on the data herein, we demonstrated a potential parallel in the activation way of COX-2 and VEGF to hypoxic stress, and elucidated a positive correlation between COX-2 and VEGF expressions in angiogenesis. The present data also indicated that hypoxia signalling may transfer to COX-2/PGE2 and VEGF signalling, playing cooperative role in the angiogenic action of ECs. Furthermore, we could deduce a possible parallel relationship between inflammation and angiogenesis under hypoxia from the results above, and they might have a positive interaction with each other when faced with hypoxia.
To further study the function of COX-2/PGE2 in the biological response of ECs to hypoxia, we used NS398 (a selective COX-2 inhibitor to inhibit the activity of COX-2) and AH6809 (a combined antogonist of EP1/2) to pre-treat HUVECs prior to indicated periods of hypoxic exposure. Overexpression of COX-2 has been described among various human malignancies [62][63][64] and blood vessels in pathological conditions [65,66], suggesting a critical role of COX-2 in tumorigenesis and inflammation. By contrast, selective pharmacological inhibition or genetic depletion of COX-2, which abolishes the eicosanoid synthesis in stimulated ECs, may abrogate an important homeostatic response to hypoxic stress, leading to inhibition of angiogenesis and induction of apoptosis [67]. Previous study has demonstrated that inhibition of COX-2 activity and thus resulting in significantly down regulation of VEGF expression and angiogenesis inhibition in vivo [68]. The present study demonstrated that exposure of HUVECs to NS398: (i) impaired proliferative response to hypoxia; (ii) strongly inhibited PGE2 release; (iii) attenuated VEGF mRNA and AQP1 mRNA transcription, and diminished VEGF secretion. Furthermore, AH6809 modestly abrogated the expression of hypoxia-induced angiogenesis related factors (VEGF and AQP1). Comparing to NS398, the blockade of the PGE2 receptors EP1/2 with a selective antagonist AH6809 received a much smaller effect on angiogenic capability of HUVECs under hypoxia. These results indicate that COX-2 induction ought to be a critical determinant of the angiogenic response of ECs to hypoxia and hypoxia probably exerts its effects in a COX-2-dependent manner. Moreover, the function of EP1/2 is not as crucial as that of COX-2 in the angiogenic capability of HUVECs, but a study has reported that the EP1/2 receptor contributed to growth factor-mediated tubular formation in angiogenesis [69].
To better understand the role of PGE2 in angiogenic response of ECs to hypoxia, we compared NS398, exogenous PGE2 and the combinational treatment of NS398 (10 M) and exogenous PGE2 (10 M) on cell viability, COX-2 expression and VEGF, AQP1 mRNA level. The results indicated that PGE2 augmented cell viability and angiogenic capability of HUVECs, and it could partially reversed the inhibitory effect of NS398. From the findings that inhibition of COX-2 activity resulted in attenuated PGE2 secretion, and exogenous PGE2 enhanced COX-2 expression, we described a positive feedback loop of COX-2-PGE2-COX-2. This was consistent with the over-production of PGE2 observed in COX-1 and COX-2 knockout cell lines, and also supported by previous studies using human synovial fibroblasts, prostatic carcinoma cell line (PC-3) and mouse lung fibroblast cells [70][71][72]. We speculated that PGE2 may lead to increased cell viability and angiogenic capability of HUVECs through its own action and/or the signal amplification of COX-2.
To sum up, the aforementioned results indicated that COX-2 products produced endogenously may serve to facilitate hypoxia mediated cell proliferation and angiogenic capability, and the involvement of endogenously produced prostanoids in promoting ECs proliferation and angiogenic capability occurs dependently of changes in COX-2 activity, and exogenous PGE2 augments the angiogenic effect of hypoxia on HUVECs. Therefore, the hypoxiainduced functional responses and expression of VEGF and AQP1 in ECs is dependent of PGE2 mediated by COX-2 activation. In conclusion, (i) hypoxia could significantly enhance the angiogenic capability of HUVECs; (ii) COX-2 may be a potential therapeutic target for ischaemic disorder and cancer management, and also the critical factor in tissue remodelling as the COX-2/PGE2 pathway influences most, if not all, of the hallmarks of these events; (iii) inflammation and angiogenesis maybe two respects of one response of HUVECs to hypoxic exposure, and they might have a positive interaction with each other.