Contrasting Effect of Estrogen on VEGF Induction under Different Oxygen Status and Its Role in Murine ROP

Abstract

purpose. It has been reported that 17β-estradiol (E2) may enhance the proliferation of bovine retinal vascular endothelial cells (BRECs) by increasing the expression of VEGFR-2 and VEGF. The hypothesis in the current study was that estrogen may contribute to fetal vascular development and the cessation of exposure to estrogen of premature infants on birth may have an inhibitory effect on retinopathy of prematurity (ROP). Because ROP is thought to develop under relative hypoxia after exposure to high-dose oxygen, this study was conducted to investigate how estrogen modulates hypoxia-induced VEGF in BRECs and mouse ROP.

methods. Gene expression of VEGF and hypoxia-inducible factor (HIF)-1α were studied in BRECs, with or without E2, under normoxia and hypoxia (1% O₂). A binding assay was performed to determine whether estrogen interferes with HIF-1–mediated induction of VEGF. In a mouse ROP model, effects of E2 were evaluated by avascular area, subsequent extraretinal neovascularization, and retinal expression of the VEGF gene, by administering E2 during hyperoxia (75% O₂) and/or after exposure to room air.

results. Hypoxia-induced VEGF mRNA in BRECs was reduced dose dependently by 1 to 100 nM E2. E2 reduced hypoxia-induced binding of HIF-1 to the VEGF promoter site and reduced the HIF-1α mRNA level. In mouse ROP, injection of E2 during hyperoxia increased retinal VEGF mRNA and reduced the retinal avascular area at the end of hyperoxia. E2 treatment during the normoxia that followed reduced VEGF mRNA and extraretinal neovascularization. Treatment with E2 throughout both periods significantly improved retinopathy.

conclusions. Estrogen may function as a significant modulator of the level of VEGF mRNA under different oxygen conditions and could serve as a prophylactic agent for ROP.

Recent studies have revealed that the ovarian sex steroid hormone, 17β-estradiol (E2), not only targets gonadal organs but also vascular cells, lymphocytes, and neural cells and exerts different actions under various physiological and pathologic conditions. ^{1 2 3 4} In the vascular system, for instance, it is now known that E2 causes vasodilatation and increases blood flow, thereby providing protection against cardiovascular diseases, such as hypertension and atherosclerosis. ^{5 6 7} Basal release of endothelium-derived nitric oxide, which dilates vessels and protects against thrombosis by inhibiting platelet aggregation, is modulated by the functional estrogen receptors on the vascular endothelium in both sexes. ⁶ Another important feature of E2 is that it induces angiogenesis, one mechanism of which is the promotion of VEGF expression. ^{8 9 10}  

Excessive neovascularization is one of the most catastrophic ocular pathologic events in many blinding diseases, such as diabetic retinopathy, retinopathy of prematurity (ROP), and neovascular glaucoma. In these diseases, hypoxia is the underlying common condition, and the induction by hypoxia of a potent angiogenic factor such as VEGF is thought to trigger the neovascular proliferation. ^{11 12} Under hypoxic conditions, hypoxia-inducible factor (HIF)-1 plays a global role in the regulation of expression of hypoxic genes including VEGF, and temporal and spatial correlations between the expression of HIF-1 and VEGF have also been reported in an animal model of retinal neovascularization. ¹³ It is therefore of importance to determine how the HIF-VEGF system is modulated by various factors in ocular neovascular diseases. One clue for identifying such mediators may be found in ROP. During development of the fetus, retinal angiogenesis and vascular construction are well organized, whereas in premature newborns, normal vascular extension is disrupted, and angiogenesis can result in disorganized retinal neovascular proliferation. Because (1) estrogen promotes retinal endothelial cell proliferation, possibly by enhancing the expression of VEGF and its receptor ¹⁰ ; (2) serum estrogen levels, which stay high during the fetal period, are drastically reduced in premature infants; and (3) interruption of vessel growth occurs in premature infants and subsequently leads to proliferative neovascularization after cessation of the rich oxygen supply, we investigated the possibility that estrogen may in some way modulate the HIF-VEGF system. In this study, we used retinal microvascular endothelial cell cultures to study the effect of estrogen on the induction of HIF-1 and VEGF under hypoxic conditions. We also studied the effects of estrogen on retinal vascular development and regression in a murine ROP model. 

Methods

Cell Culture

Bovine retinal endothelial cells (BRECs) were cultured as previously described by us. ¹⁰ In brief, the cells were isolated from calf eyes and were grown on fibronectin-coated dishes (Iwaki Glass, Tokyo, Japan) containing phenol red–free DMEM (Sigma, St. Louis, MO) with 5.5 mM glucose, charcoal- and dextran-treated 10% plasma-derived horse serum, 50 mg/L heparin, and 50 U/L endothelial cell growth factor (Roche Molecular Biochemicals, Indianapolis, IN). The cells were cultured in 5% CO₂ at 37°C, and cells were plated at a density of 2 × 10⁴ cells/cm² and passaged when confluent (approximately 1 × 10⁵ cells/cm²). The medium was again changed every 3 days, and cells from passages 3 to 10 were used in the experiments. Cells were then incubated for 24 hours with phenol red–free DMEM containing charcoal-filtered 10% calf serum (Gibco BRL, Grand Island, NY). The medium was changed to one of the experimental conditioned media the following day. E2 (1–1000 nM) or 17α-estradiol (10 nM; Sigma) was added to the medium simultaneously with the introduction of hypoxia (1% O₂) in the cells, and in some experiments a 10-fold excess of tamoxifen (Sigma), an estrogen receptor antagonist, was added to the medium 3 hours before the stimulation with E2. For each of these experiments, the cells were incubated in the experimental medium at 37°C. 

Northern Blot Analysis

Total cellular RNA was prepared from the cultured cells at 9 or 24 hours after induction of hypoxia. The media were decanted, and the cells were lysed directly in the culture dishes with 600 μL guanidinium thiocyanate. RNA was extracted by adding 240 μL chloroform and shaking for 10 seconds, then cooling at 4°C for 5 minutes. The suspension was centrifuged at 15,000 rpm for 15 minutes at 4°C and the aqueous phase transferred to a new tube. The RNA was precipitated by adding 600 μL isopropanol and incubated on ice for 15 minutes at 4°C. The RNA pellets were washed once with 75% ethanol, dried, resuspended in 20 μL diethyl pyrocarbonate (DEPC)–treated water, and incubated for 10 minutes at 60°C. The RNA’s purity was determined by the ratio of optical density (OD) measured at 260 and 280 nm (OD₂₆₀/OD₂₈₀), and RNA quantity was estimated at OD₂₆₀. Northern blot analysis was performed on 15 μg total RNA after 1% agarose-2 M formaldehyde gel electrophoresis and subsequent capillary transfer to nylon membranes (Biodyne; Pall BioSupport, East Hills, NY) and ultraviolet cross-linking (Funa-UV-Linker, model FS-1500; Funakoshi, Tokyo, Japan). Radioactive probes were generated using labeling kits (Megaprime; Amersham, Buckinghamshire, UK) and ³²P-dCTP. Blots were prehybridized, hybridized, and washed four times in 0.5× SSC and 5% SDS at 65°C for 1 hour in a rotating hybridization oven (Taitec, Koshigaya, Japan). All signals were analyzed with a densitometer (model BAS-2000 II; FujiFilm, Tokyo, Japan). Human cDNA of VEGF was the generous gift of Lloyd P. Aiello and George L. King (Joslin Diabetes Center, Boston, MA). Bovine HIF-1α cDNA probes were made by PCR amplification using the primers 5′-TGCTTGGTGCTGATTTGTGA-3′ and 5′-GACTCTGGATTTGGTTCTAA-3′, as described by Hara et al. ¹⁴ and then by subcloning into a vector (pBluescript II; Stratagene, La Jolla, CA). The samples from each lane were normalized with 36B4 expression. ^{15 16}  

Binding Assay

Nuclear extracts were prepared from BRECs, as described by Schreiber et al. ¹⁷ An electrophoretic mobility shift assay (EMSA) was performed by incubating 6 μg nuclear extract with 10⁴ cpm (∼0.2 ng) [α-³²P] dATP-labeled, double-stranded oligonucleotide probe (Amersham) in an incubation buffer (10 mM Tris-HCl [pH 7.5], 50 mM KCl, 5 mM MgCl₂, 1 mM dithiothreitol [DTT], 1 mM EDTA, 12.5% glycerol, and 0.1% Triton X-100), as described by Semenza and Wang. ^{18 19} The oligonucleotide sequence of HIF-1 was 5′-TGCATACGTGGGCTCCAACAG-3′, as described by Forsythe et al. ²⁰ Competition experiments were performed with 50 ng of unlabeled double-stranded oligonucleotide. The mixture was loaded onto a 5% nondenaturing polyacrylamide gel, and electrophoresis was performed in 1× ΤΑΕ (40 mM Tris-acetate and 1 mM EDTA [pH 8.0]) at 4°C. The gel was then dried, and all signals were visualized with a densitometer. 

Mouse Model of ROP

C57BL/6J mice were exposed to 75% O₂ from postnatal day (P)7 to P12, along with nursing mothers. ²¹ At P12, the mice were returned to room air. The mice received daily subcutaneous injections of either 17β-estradiol (E2, 1 μg/mouse) in vehicle (dissolved in ethanol and diluted in 0.1 mL/mouse of PBS) or vehicle alone and then were killed on P12 or P17. The mice were divided into six groups according to the treatment between P7 and P17 as shown in Figure 1 : room air with vehicle injection (control, group 1), room air with E2 injection (group 2), hyperoxia with vehicle injection (group 3), hyperoxia with E2 administered for 5 days, from P12 to P16 (group 4), hyperoxia with E2 for 10 days, from P7 to P16 (group 5) and hyperoxia with E2 for 5 days, from P7 to P11 (group 6). All experiments were approved by the institutional Committee for Animal Use in Research and Education and conformed to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Fluorescein-Dextran Perfusion of the Retinal Blood Vessels and Quantification of the Avascular Area

To study the retinal vascular pattern, retinal flatmounts were obtained by using a perfusion of high-molecular-weight dextran (2,000,000) conjugated with fluorescein (Sigma), as described previously. ^{21 22} Briefly, the mice were anesthetized and a median sternotomy was subsequently performed. The left ventricle was identified, and 1 mL of a 50-mg/mL solution of high-molecular-weight fluorescein-conjugated dextran was injected with a 27-gauge needle on a 1-mL tuberculin syringe. The eyes were enucleated and placed in 4% paraformaldehyde in PBS. The lenses were removed, and the peripheral retinas were cut to allow for flatmounting with glycerol-gelatin. The flatmounted retinas were then examined with a fluorescence microscope, and the images were captured in digital format. The central capillary-free area was quantified from the digital images in masked fashion by using NIH Image (provided in the public domain by the National Institutes of Health, Bethesda, MD, and available at http://rsb.info.nih.gov/nih-image/). 

Analysis of Neovascularization

More than 50 serial 6-μm paraffin-embedded axial sections were obtained, starting at the optic nerve head, as described by Aiello et al. ²³ Staining with periodic acid-Schiff reagent and hematoxylin was followed by evaluation of 10 intact sections of equal length, each 30 μm apart, over a span of 300 μm. All retinal vascular cell nuclei anterior to the internal limiting membrane were counted in each section by means of a fully masked protocol. The mean count of all 10 sections yielded the average number of neovascular nuclei per 6-μm section per eye. 

Quantification of Gene Expression of VEGF in the Retina

Gene expression of VEGF in the retina was studied by semiquantitative PCR. The eyes were enucleated on P12 and P17 and cut in half along the limbus, and the iris-ciliary body and retina were separately collected with fine forceps. RNA was extracted from the pooled retinas of both eyes of each animal, according to the acid guanidinium thiocyanate-phenol-chloroform extraction method. ²⁴ The extracted RNA was quantified, and 5 μg of each sample was used to make cDNA with a synthesis kit (First-Strand DNA Synthesis; Amersham Pharmacia Biotech, Buckinghamshire, UK). RT-PCR quantification was performed in a linear range, as described by us. ^{25 26} The PCR conditions consisted of denaturation at 95°C for 30 seconds, annealing at 55°C for 30 seconds, and extension at 72°C for 60 seconds. The reaction was initiated by adding two units of Taq DNA polymerase (Perkin-Elmer Cetus, Norwalk, CT), followed by 35 cycles for VEGF and 30 for β-actin. The primers used in this experiment were GTAACGATGAAGCCCTGGAGTG (sense strand) and TGAGAGGTCTGGTTCCCGAAAC (antisense strand) for VEGF ²⁷ and TGTGATGGTGGGAATGGGTCAG (sense strand) and TTTGATGTCACGCACGATTTCC (antisense strand) for β-actin. Nucleotide sequencing and restriction analysis confirmed that the PCR products were derived from the target cDNA sequences. Band intensities were quantified with a fluorescence imager (FluorImager SI; Molecular Dynamics, Sunnyvale, CA) and its associated software. The cDNA concentration was first normalized by PCR with the primers for β-actin, and the relative expression of VEGF was then determined by comparing it with the expression level in mice without oxygen treatment and with vehicle injection. 

Statistical Analysis

All determinations were performed in triplicate. We used either the Mann-Whitney test to compare each group with an indicated group or the Kruskal-Wallis test. Results are expressed as the mean ± SD. P < 0.05 was considered statistically significant. 

Results

Effect of E2 on VEGF Gene Expression

The effect of exogenous treatment with 10 nM E2 on VEGF gene expression in BRECs is shown in Figure 2A . Under hypoxia, VEGF mRNA increased to 2.1 ± 0.5 times that of the control after 9 hours and to 2.5 ± 0.2 times control after 24 hours, whereas the corresponding increases after treatment with 10 nM E2 were significantly less: 1.1 ± 0.1 times and 1.6 ± 0.2 times, respectively (P < 0.05). The inhibitory effect of E2 on the expression of the VEGF gene was dose dependent between 1 and 100 nM, but 1000 nM E2 did not have a significant effect (Fig. 2B) . Ten nM E2 increased VEGF mRNA levels by 2.3 ±0.4 times at 24 hours in normoxia (Fig. 2C) . An overdose of tamoxifen (100 nM) reversed the effect of E2 (10 nM) from 1.9 ± 0.2 times to 2.6 ± 0.5 times (P < 0.05), whereas 17α-estradiol did not affect the hypoxia-induced expression of VEGF mRNA (Fig. 3) . 

Analysis of HIF-1 DNA-Binding Activity

HIF-1 DNA-binding activity was assessed by using a 21-bp oligonucleotide probe containing the HIF-1 binding site from the VEGF enhancer, as described by Forsythe. ²⁰ Nuclear extracts prepared from BRECs exposed to 1% O₂ for 9 or 24 hours were incubated with this probe. HIF-1 DNA-binding activity was detected after 9 and 24 hours of hypoxia (Fig. 4) . E2 markedly reduced binding to the mRNA of VEGF by 74% at 24 hours (P < 0.05). The addition of unlabeled oligonucleotide masked the binding activity, indicating that the reaction is specific to the HIF-1 binding site for VEGF mRNA. 

Effect of E2 on Expression of the HIF-1α Gene

The effect of E2 on expression of the HIF-1α gene is shown in Figure 5 . HIF-1α mRNA expression increased to 2.2 ± 0.1 times that of the control at 9 hours and 2.5 ± 1.4 times at 24 hours after induction of hypoxia, whereas the corresponding levels after treatment with 10 nM E2 showed reduction to 1.4 ± 0.4 times (P = 0.145) and 0.5 ± 0.3 times, respectively (P < 0.05, Fig. 5A ). The expression of HIF-1α mRNA was reduced in a dose-dependent manner between 1 and 100 nM, with the maximum reduction observed at a concentration of 100 nM E2 (0.2 ± 0.05 times, P < 0.05; Fig. 5B ), whereas E2 did not affect the expression of HIF-1α under normoxia (Fig. 5C) . 

Hyperoxia-Induced Vaso-obliteration

To assess the influence of estrogen on the regression of vessels, we compared the capillary-free area as a percentage of the total retinal surface in fluorescein-dextran perfused retina of mice (Fig. 6) at the end of hyperoxia (P12) and on P17. No capillary-free areas were observed in vehicle- and E2-treated animals in room air (Fig. 1 , groups 1 and 2). On P12, retinal flatmounts of oxygen-exposed mice treated with E2 showed significant reduction in the size of the central capillary-free area, compared with those of oxygen-exposed mice with vehicle injection (n = 8, P < 0.0001). On P17, treatment with E2 during P7 to P16 reduced the capillary-free area in oxygen-exposed mice. No difference was observed in the capillary-free area between E2-treated mice during normoxia (P12–P16) and mice with vehicle injection (n = 8). The percentage of the capillary-free area in E2-treated mice during hyperoxia (P7–P11; Fig. 6I ) was higher on P17 than that of mice with vehicle injection, but this percentage was the same as that on P12 (Fig. 6J ; n = 8). 

Neovascularization

Neovascularization was assessed on P17 by counting extraretinal vascular nuclei on serial sections (Fig. 7E) . No vascular cell nuclei were observed anterior to the internal limiting membrane in vehicle- and E2-treated animals in room air. Mice treated with E2 between P12 and P16 and between P7 and P16 both showed less extensive extraretinal neovascularization than did oxygen-exposed mice (n = 8, P < 0.01). Treatment with E2 during hyperoxia (between P7 and P11) did not reduce extraretinal neovascularization (n = 8). 

VEGF mRNA Expression in Retinal Extracts in a Murine Model

The level of mRNA expression of VEGF was evaluated in the retina by means of semiquantitative PCR. PCR products were obtained by PCR-specific primers for VEGF and β-actin cDNA in the retina. The levels of VEGF mRNA expression are shown as a ratio in relation to the level in the mice with no oxygen treatment but with vehicle injection (Fig. 8) . On P12, VEGF gene expression was reduced by 40% in oxygen-exposed mice, while treatment with E2 increased VEGF gene expression back to the control level (1.0-fold, n = 5, P < 0.05). On P17, VEGF gene expression was 1.8 times that of the control in oxygen-exposed mice and was reduced in oxygen-exposed mice receiving E2 injection from P12 to P16 (0.9 times) and P7 to P16 (0.9 times, n = 5, P < 0.05). Treatment with E2 from P7 to P11 did not reduce the expression of VEGF mRNA significantly compared with that in oxygen-exposed mice (n = 5). 

Discussion

In some general ischemic diseases such as myocardial infarction and stroke, a hypoxic condition created by the severe reduction of blood flow is followed by increased expressions of HIF, VEGF, and VEGFR. ^{28 29} In these ischemic diseases the HIF-VEGF system is thought to contribute to protective mechanisms through reconstruction of the vasculature by such means by developing collateral vessels. ^{29 30} In ocular diseases such as diabetic retinopathy and arterial and vein occlusions, however, hypoxic conditions after vascular occlusion often lead to pathologic angiogenesis, which in some cases causes severe visual loss. The HIF-VEGF system, which usually functions as a rescue mechanism, often fails to do so in the eye. ROP seems to offer another good example of this phenomenon. In premature infants, retinal vessel growth slows down or even stops, and the peripheral retina remains unperfused, which subsequently causes retinal proliferative neovascularization. One definite difference between a fetus and a premature infant is the serum level of estrogen. The serum E2 level in the fetus should be in the 10-ng/mL range ³¹ and should decrease to a minimal level after the maternal supply is interrupted. We have reported that E2 enhances the expression of both VEGF and VEGFR-2 in BRECs and E2 also promotes proliferation of these cells. ¹⁰ This observation led us to think that estrogen may contribute to the development of vessels in the fetus and that a sudden decrease in E2 levels in premature infants may have an adverse effect on the vasculature. The literature shows no consensus on the effect of E2 on VEGF expression: E2 reportedly increases the expression of VEGF in retinal endothelial cells, ¹⁰ peritoneal fluid macrophages from patients with endometriosis, ⁸ and MCF-7 cells, ⁹ but reduces the VEGF expression in HEC1A endometrial cancer cells. ³² The time course of VEGF induction by E2 also varies from one cell type to another. In rats, uterine expression of VEGF mRNA was induced 6 hours after treatment with E2, ³³ whereas in retinal endothelial cells, VEGF induction peaked 24 hours after treatment. These findings suggest that the effect of E2 on the induction of VEGF mRNA may take different routes, depending on the cell types and the situation. Another important feature of estrogen is that it can act as an antioxidant. It is reported that some antioxidants interfere with the induction of HIF-1–mediated VEGF expression in glioma cells during hypoxia. ³⁴ All these findings led us to assume that E2 could act as one of the modulators of the HIF-VEGF system, thus possibly affecting some ocular neovascularizations. 

In the present study, we first investigated the effect of E2 on hypoxia-induced VEGF gene expression. This effect was apparent from as early as 9 hours after hypoxia treatment. As previously reported by us, the increase in VEGF gene expression peaked 24 hours after E2 treatment under normoxic conditions, which indicated that the induction of the VEGF gene may not be a direct result of E2 treatment. In our study the inhibitory effect of E2 on VEGF induction during hypoxia was observed from the early stage of hypoxia, which suggests that E2 is likely to interfere with VEGF induction from the beginning. Furthermore, the effect of E2 on the induction of VEGF was dose-dependent up to 100 nM, which again suggests direct action of E2 on the process of VEGF induction by hypoxia. The opposite effect of tamoxifen on E2 treatment suggests that the inhibitory effect of E2 on VEGF induction may be receptor mediated rather than antioxidant mediated, because 17α-estradiol, which exerts an antioxidant effect, ³⁵ did not influence VEGF expression. The presence of a classic estrogen receptor (estrogen receptor α) in these culture cells was demonstrated in our previous study. At 1000 nM E2, however, the inhibitory effect was less pronounced than at 100 nM. A possible explanation for this is that E2 retains some indirect effect that increases VEGF messages, as was observed primarily under normoxic conditions in our previous study, so that the inhibitory effect of E2 could be balanced by such an opposite residual effect. 

Two major mechanisms by which hypoxia increases VEGF is the induction of VEGF mRNA through HIF-1 ^{20 36 37} and the increased stability of VEGF mRNA. ^{38 39} HIF-1 is a heterodimeric nuclear transcription factor composed of HIF-1α and -1β subunits. Expression of the HIF-1α subunit and the level of HIF-1 DNA-binding activity increase exponentially in response to a decrease in oxygen in hypoxic cells. This transcriptional activation by HIF-1 has been implicated in VEGF, erythropoietin, ^{18 40} lactate dehydrogenase, ⁴¹ phosphoglycerate kinase 1, ⁴² glucose transporter 1, ⁴³ and type II nitric oxide synthase. ⁴⁴ Our EMSA demonstrated that E2 reduced the increase in binding activity of HIF-1 to the VEGF promoter site caused by hypoxia, and Northern blot analysis showed that E2 also reduced the expression of HIF-1α in a dose-dependent manner. These observations suggest that E2 has a definite effect on HIF-mediated VEGF gene transcription. There are many reports about the effect of E2 on VEGF, whereas the effect of E2 on HIF-1 has not been reported so far. Although estrogen induces angiogenesis by enhancing the expression of VEGFR-2 as well as VEGF, our study implies that estrogen also acts inversely to reduce VEGF expression under hypoxic conditions. Although in our experiments the function of estrogen as an antioxidant was not apparent, it is reported that antioxidants interfere with the induction of HIF-1–mediated VEGF expression in glioma cells during hypoxia, ³⁴ so that the mechanism of this discrepant effect of estrogen must be further clarified. 

The possibility that estrogen may have opposite effects on VEGF expression depending on oxygen conditions offers a new insight into how estrogen may modulate angiogenesis, both physiologically and pathologically. It has been suggested that this classic steroid may help develop the vascular system and yet may also prevent pathologic neovascularization under hypoxic conditions. To test this hypothesis, we further investigated the effect of estrogen in an animal model of ROP. 

There are two important elements when we consider the pathogenesis of ROP: the growth of the avascular area and the subsequent neovascular reaction. In our study, treatment with E2 during hyperoxia (P7–P11) increased VEGF mRNA expression back to the control level (Fig. 8A) , which is compatible with the in vitro effect of E2 under normoxia (Fig. 2C) . This restored VEGF expression also seemed to help the retinal vascularization, significantly reducing the avascular area by P12 (Fig. 6I) . In contrast, on P17 after the period of room air, a relatively hypoxic period after the hyperoxia, VEGF mRNA increased in mice with oxygen and vehicle treatment, but treatment with E2 during the relative hypoxic period significantly reduced the VEGF gene’s induction in the oxygen-treated mice (Fig. 8B) . This inhibition of VEGF upregulation by E2 under hypoxia is again compatible with our in vitro data (Fig. 2) . More significantly, this decrease in VEGF level seemed to be well reflected in the extraretinal neovascular nuclei counts in retinal sections (Fig. 7E) . These suggest that estrogen may help retinal vessel growth under hyperoxia by restoring VEGF expression when hyperoxia reduces the level of VEGF genes and inhibits vessel extension, and that in relative hypoxia, estrogen may, in contrast, inhibit hypoxia-induced upregulation of the VEGF gene and may reduce extraretinal neovascularization. Avascular area and neovascularization are often interpreted as the cause and the result, but our results in mice that received E2 from P7 to P11 seem to contradict this general concept. In this group, the avascular area was significantly reduced on P12, but the number of neovascular nuclei on P17 was in a range similar to that in mice treated with vehicle (Fig. 7E) . In this group of mice, there was still 20% to 30% of avascular area remained even though reduced, which could be enough to trigger extraretinal neovascularization. Because the expression level of the VEGF gene on P17 was not statistically different from that of vehicle-treated group, the moderately elevated level of VEGF may still have contributed to enhancement of the extraretinal neovascular reaction. In other words, how much avascular area is enough to cause extraretinal neovascularization and how much excess of VEGF may enhance this angiogenic reaction appears to be a matter of threshold. These thresholds may also vary between individuals, and we think that this rather unexpected result in this group is due to the possibility that they were at some specific point in these thresholds. These led us to further postulate that under these different oxygen conditions, changes in the level of VEGF induction may critically affect extraretinal angiogenesis and therefore the role of E2 may become important as a modulator of the level of VEGF induction. Recovery of avascular area between P12 and P17 cannot be explained in a simple manner by the VEGF expression level either. In mice with E2 treatment from P7 to P11 extraretinal neovascularization was enhanced, but the avascular area was not reduced, whereas P12 to P16 mice had a good reduction in avascular area, even though VEGF expression was low (Fig 8B) . These indicate that during the relative hypoxic period in our model, the level of VEGF gene expression seems to be more closely related to the extraretinal neovascular reaction than to retinal vessel extension. It may suggest that retinal extension and/or maturation may require factors other than VEGF. At least from our present data, it is likely that to obtain the best effect of E2 on ROP, E2 seems to be required throughout the entire hyperoxic and relative hypoxic periods. 

An increase in the survival rate of infants with extremely low birth weight accounts for the recent significant increase in the number of new cases of ROP. Cryotherapy or laser photocoagulation is mainly used to treat severe types of ROP, but these treatments remain insufficient for some of these extreme cases. Results for several therapeutic agents, such as vitamin E, ⁴⁵ for premature infants, and dexamethasone ⁴⁶ and indomethacin, ⁴⁷ used in a murine model, have been reported. Considering that estrogen is essential to the fetus and its possible important role as an angiogenesis modulator, E2 could be used as a prophylactic agent for ROP. 

 Supported by Grants-in Aid for Scientific Research B2-11557128 and C2-11671735 from the Ministry of Education, Science and Culture of the Japanese Government.

 Submitted for publication August 7, 2001; revised January 24, 2002; accepted February 7, 2002.

 Commercial relationships policy: N.

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Schema of E2 treatment in a mouse ROP model. The mice were divided into six groups, according to the treatment between P7 and P17. Group 1: room air with vehicle injection (control); group 2: room air with E2 injection; group 3: hyperoxia with vehicle injection; group 4: hyperoxia with E2 administered for 5 days (P12–P16); group 5: hyperoxia with E2 for 10 days (P7–P16); group 6: hyperoxia with E2 for 5 days (P7–P11). s.c.i., subcutaneous injection.

Figure 2.

 

(A) Time course of the effect of E2 on expression of VEGF mRNA in BRECs. Total RNA was isolated at the indicated time points after E2 (10 nM) had been added to the cells simultaneously with the introduction of hypoxia. *Statistical significance at P < 0.05 between conditions with and without estrogen at each time point (Mann-Whitney). (B) Dose–response effect of E2 treatment (1–1000 nM) on expression of VEGF mRNA in BRECs. BRECs were treated with the indicated concentrations of E2 for 24 hours and harvested for the preparation of total RNA. *Statistical significance at P < 0.05 compared with the group with hypoxia with no E2 treatment (Mann-Whitney). (C) The effect of 10 nM E2 under normoxia at 24 hours. Experiments were performed in triplicate. Representative blots from three independent experiments are shown (top). Results were quantified by densitometric analysis of the autoradiograms derived from the top panels after normalization with the 36B4 control cDNA signals. Data are presented as ratios of experimental groups to the control and are expressed as the mean ± SD (bottom).

Figure 3.

 

Effects of tamoxifen and 17α-estradiol on VEGF mRNA under hypoxic conditions. Subconfluent BRECs were cultured with 17α-estradiol (10 nM), E2 (10 nM), or E2 (10 nM) combined with pretreated tamoxifen (Tx, 100 nM) for 24 hours. *Statistically significant (Kruskal-Wallis).

Figure 4.

 

EMSA of BREC nuclear extracts to study the binding activity of HIF-1 to VEGF promoter sequences. An oligonucleotide of the VEGF promoter site (5′-TGCATACGTGGGCTCCAACAG-3′) was ³²P labeled and incubated with 5-μg aliquots of nuclear extracts from hypoxic BRECs. For competition assays, 50 ng of an unlabeled oligonucleotide was included in the binding reaction mixture as indicated. Results are presented as percentages of the maximum binding activity at 24 hours after hypoxia with the normoxia level as a background. Data are expressed as the mean ± SD (bottom). *Statistically significant compared with the groups with hypoxia treatment only at each time point (Mann-Whitney).

Figure 5.

 

(A) Time course of the effect of E2 treatment on expression of HIF-1α mRNA in BRECs. Total RNA was isolated at the indicated time points, after E2 (10 nM) had been added to the cells simultaneously with the induction of hypoxia. *Statistically significant difference between conditions with and without estrogen at each time point (Mann-Whitney). (B) Dose-response effect of E2 treatment (1–1000 nM) on expression of HIF-1α mRNA in BRECs. BRECs were treated with the indicated concentrations of E2 for 24 hours and harvested for the preparation of total RNA. *Statistically significant at P < 0.05 compared with the group with hypoxia with no E2 treatment (Mann-Whitney). (C) The effect of 10 nM E2 under normoxia at 24 hours. Experiments were performed in triplicate, and the representative blots of three independent experiments are shown at the top of each panel. Results were quantified by densitometric analysis of the autoradiograms derived from the top panel after normalization with the 36B4 control cDNA signals. Data are presented as ratios of the experimental groups to the control and are expressed as the mean ± SD (bottom).

Figure 6.

 

Retinal wholemounts from normoxic (A, B) and oxygen-exposed (C–H) mice, and quantification of capillary-free area. Retinas on P12 from mice exposed to normoxia and treated with (A) vehicle and (B) E2 and from mice exposed to hyperoxia and treated with (C) vehicle and (D) E2 from P7 to P11. Retinas on P17 from mice treated with (E) vehicle injection, (F) E2 from P12 to P16, (G) E2 from P7 to P16, and (H) E2 from P7 to P11. (I, J) Capillary-free area as a percentage of total retinal surface in the FITC-dextran–perfused retina of oxygen-exposed mice (n = 8) on P12 (I) and P17 (J). *Statistically significant by (I) Mann-Whitney and (J) Kruskal-Wallis.

Figure 7.

 

Paraffin-embedded, periodic acid-Schiff reagent–, and hematoxylin-stained 6-μm serial sections prepared on P17 in a mouse model of ischemia-induced retinopathy: (A) vehicle injection, (B) E2 treatment from P12 to P16, (C) E2 treatment from P7 to P16, (D) E2 treatment from P7 to 11. Arrowheads: neovascular tufts. Bar, 20 μm. (E) Extraretinal neovascularization quantified by means of endothelial cell nucleus counts on serial sections (n = 8). *Statistically significant (Kruskal-Wallis).

Figure 8.

 

Levels of VEGF mRNA expression in the retina in mice exposed to hyperoxia on P12 (A) and P17 (B). A total of 1 μg of E2 was administered for the indicated durations. Expression was evaluated by semiquantitative PCR. Total RNA was isolated from the retina, reverse transcribed into cDNA, and normalized relative to β-actin expression. Data are presented in comparison with the expression level in mice in normoxia treated with vehicle injection and are expressed as the mean ± SD (n = 5). *Statistically significant difference between conditions with and without estrogen (A; Mann-Whitney) and among the four groups in hyperoxia (B; Kruskal-Wallis).