Potential Role of the Angiopoietin/Tie2 System in Ischemia-Induced Retinal Neovascularization

Abstract

purpose. Ischemia-induced neovascularization can cause catastrophic loss of vision in retinal disorders such as diabetic retinopathy. Recent studies have shown that the angiopoietin-Tie2 system is a major regulator of vascular integrity and is involved in pathologic angiogenesis. In the study described herein, the role of these molecules in ischemic retinal disorders was investigated.

methods. Human epiretinal membranes were examined by immunohistochemistry, In situ hybridization, and reverse transcription-polymerase chain reaction (RT-PCR) analysis. Effects of angiopoietins on tube formation were studied in vitro in bovine retinal capillary endothelial cells (BRECs) and in a murine model of ischemia-induced retinal neovascularization.

results. In human epiretinal membranes surgically obtained from eyes with ischemic retinal disorders, substantial upregulation of angiopoietin 2 (Ang2) and the receptor Tie2 was recorded than in those from eyes with nonischemic diseases, whereas expression of Ang1 was constant in all membranes. Both Ang1 and Ang2 promoted tube-forming activity and enhanced the effects of vascular endothelial growth factor (VEGF) in cultured BRECs. Soluble Tie2 fusion protein (sTie2-Fc), which precluded modulation of VEGF-dependent tube formation by the angiopoietins, suppressed both VEGF and hypoxia-conditioned, medium-induced tube-forming activity in BRECs. Intravitreal injection of sTie2-Fc, soluble Flt-1 fusion protein (sFlt-1-Fc), and both chimeric proteins suppressed retinal angiogenesis in a murine model of retinal ischemia in the order of sTie2-Fc < sFlt-1-Fc < sTie2-Fc+sFlt-1-Fc.

conclusions. These results reinforce the substantial role of the angiopoietins/Tie2 system in ischemia-induced angiogenesis as well as the VEGF system and suggest that combined inhibition of Tie2 and VEGF signaling may be more effective in halting or preventing pathologic angiogenesis in ischemic retinal disorders.

Pathologic growth of new blood vessels is the common final pathway in diabetic retinopathy, retinopathy of prematurity, and age-related macular degeneration, and often leads to catastrophic loss of vision. The primary stimulus in these disease states is hypoxia. Vascular endothelial growth factor (VEGF) mediates such ischemia-induced ocular neovascularization, ^{1 2 3 4} as well as tumor angiogenesis ⁵ and formation of collateral vessels in cardiovascular diseases. ⁶ Previous studies in which inhibition of either VEGF or its receptor resulted in suppression of pathologic angiogenesis ^{7 8} validate the hypothesis that the VEGF signal transduction system is a viable target for antiangiogenic therapeutic intervention. 

Angiopoietins and the Tie2 receptor constitute another recently identified endothelial cell-specific, ligand-receptor system that is crucial not only in vascular development but also in postnatal angiogenesis. Mice without angiopoietin 1 (Ang1) or the Tie2 receptor die later than do those without VEGF or VEGF receptors, indicating that this family exerts its effect in the later stages of formation of embryonic blood vessels. ^{9 10 11} The phenotype of Ang1- and Tie2-knockout mice is distinct from that of mice without VEGF receptors. Endothelial cells are detected in normal numbers and tube formation occurs, but the distinction between large and small vessels is obscure and encapsulation by periendothelial cells is absent. ¹⁰ These findings suggest that the Ang1-Tie2 system plays a role in endothelial-stromal cell communication and regulates the maturation and stability of vessel structures. The four known angiopoietins all bind to Tie-2. ¹² The affinities of Ang1 and Ang2 are similar, ^{11 13 14} but Ang2 competitively inhibits Ang1-induced autophosphorylation and chemotactic effects in endothelial cells. ^{13 14 15} Moreover, transgenic mice that overexpress Ang2 mimic the phenotype of Ang1- and Tie2-knockout mice, suggesting that Ang2 is a natural antagonist for Tie2. ¹³ Widespread expression of Ang1 and Tie2 and phosphorylation of Tie2 in the quiescent vasculature of adult tissues have been reported, ¹⁶ suggesting their role in postnatal angiogenesis as well as in prenatal vascular development. In contrast to Ang1, Ang2 is highly expressed only at sites of vascular remodeling in the adult, most notably in the female reproductive tract. ¹³ Ang2 is also upregulated by hypoxia and angiogenic cytokines, including VEGF, ^{17 18} and in pathologic angiogenesis associated with tumors ^{19 20} and ischemia in the retina in an animal model. ¹⁸ A study of a model of corneal angiogenesis revealed that Ang1 and -2 facilitate VEGF-induced neovascularization. ²¹ These data support the notion that angiopoietins and Tie2 may contribute substantially to the pathologic angiogenesis observed in ocular neovascular disorders, in which angiogenic stimuli such as hypoxia or VEGF are abundant. Involvement of this system, however, has not been investigated in detail. 

In the study described herein, we investigated the role of the angiopoietin-Tie2 system in ischemia-induced retinal neovascularization in ischemic ocular disorders such as diabetic retinopathy. We demonstrated a more marked presence of Ang2 and Tie2 in surgically excised epiretinal membranes (ERMs) from eyes with ischemic retinal disorders compared with that in membranes associated with nonischemic retinal diseases, whereas we found that expression of Ang1 was similar in both types of membranes. Furthermore, we demonstrated that recombinant soluble Tie2 fusion protein (sTie2-Fc), which inhibited angiopoietin modulation of VEGF-dependent tube formation in cultured retinal endothelial cells, suppressed retinal angiogenesis both in vitro and in a murine model of retinal ischemia. With the combination of soluble Flt-1-fusion protein (sFlt-1-Fc), it suppressed retinal neovascularization more efficiently. These results not only reinforce that the angiopoietin-Tie2 system and the VEGF system play a substantial role in ischemia-induced angiogenesis but also suggest that the combined inhibitory influence of both VEGF and Tie2 signaling may be effective in halting or preventing pathologic angiogenesis in ischemic retinal disorders. 

Materials and Methods

Reagents

Human Ang1*, Ang2, sTie1-Fc, and sTie2-Fc were kindly supplied by Regeneron Pharmaceuticals, Inc. (Tarrytown, NY). The recombinant protein was formulated in buffer solution consisting of 0.05 M Tris-HCl (pH 7.5), 150 mM NaCl, and 0.05% 3-([3-cholamidopropyl]dimethylammonio-2-hydroxy-1-propanesulfonate (CHAPS). Ang1* and -2 recombinant proteins were produced with baculovirus vectors and purified with protein A-Sepharose, as described previously. ¹⁴ Ang1* is a genetically engineered variant of naturally occurring Ang1 that maintains similar properties in all assays. ¹³ The purity of proteins was greater than 95%, as judged by reducing and nonreducing silver-stained SDS-polyacrylamide gel electrophoresis. sTie2-Fc or sTie1-Fc are recombinant fusion proteins consisting of the ectodomain of Tie2 or Tie1 receptors fused to the Fc portion of human IgG₁. The fusion proteins were produced according to standard protocols in Sf-21AE cells infected with baculovirus vectors bearing the respective fusion constructs. Recombinant fusion proteins were then purified by protein A-Sepharose chromatography. ¹⁴ Previous studies have shown that sTie2-Fc binds to Ang1, Ang2, and mouse angiopoietins. ¹³ VEGF was obtained from Genzyme (Cambridge, MA). 

Human Ocular Tissue

Human tissues were handled according to the tenets of the Declaration of Helsinki. All specimens were pathologic samples, and consent to the study of these surgical samples was obtained from every patient. Specimens were obtained from 31 eyes of 31 patients. Specimens from 22 eyes were used for immunohistochemistry, and specimens from 9 eyes were processed for RNA extraction for reverse transcription-polymerase chain reaction (RT-PCR) analysis. Because of the small size of the ERMs, the first series of samples was used for immunohistochemistry, and the second series of samples was used thereafter for RT-PCR, to determine the expression of angiopoietin genes. The patients ranged in age from 34 to 74 years (mean ± SD, 59.5 ± 4.1) at the time of vitrectomy. Among the 22 membranes used for immunohistochemistry, 13 epiretinal membranes were obtained from patients with ischemic retinal disorders (12 with diabetic retinopathy and 1 with retinal vein occlusion [RVO]), and nine epiretinal membranes were obtained from patients with nonischemic retinal disorders (idiopathic macular pucker [MP]). The membranes were removed with intraocular forceps during vitrectomy, fixed in 3.7% formalin with phosphate-buffered saline (PBS; pH 7.4) for 1 hour at 4°C, dehydrated with a graded alcohol series, and embedded in paraffin. The paraffin-embedded specimens were serially sectioned at 5-μm thickness and placed on aminopropyltriethoxysilane-coated glass slides (Dako, Glostrup, Denmark) for immunohistochemical staining. Sections were rehydrated with a graded series of alcohol and rinsed with PBS. Due to the small size of the sample, some ERMs were immediately placed in optimal cutting temperature compound (Tissue Tek; Miles, Elkhart, IN) and frozen on dry ice. Frozen sections 4 to 6 μm thick were fixed in 100% acetone for 30 seconds, dried, and briefly hydrated in PBS. Hydrogen peroxide-methanol (0.3%) was applied to each paraffin-embedded specimen and OCT-embedded specimen for 10 minutes to block endogenous peroxide activity. 

Immunohistochemistry

After incubation with blocking serum for 20 minutes, the specimens were incubated overnight at 4°C with one of the primary antibodies: goat polyclonal anti-Ang1, 1:400 dilution; goat polyclonal anti-Ang2, 1:400 dilution; rabbit polyclonal anti-Tie2, 1:100 dilution; rabbit polyclonal anti-VEGF, 1:200 dilution (Santa Cruz Biotechnology, Santa Cruz, CA); mouse monoclonal anti-CD-68, 1:50 dilution (Elm, Rome, Italy); mouse monoclonal anti-glial fibrillary acidic protein (GFAP), 1:50 dilution; or rabbit monoclonal anti-CD34, 1:40 dilution (Dako). Specimens were then washed for 10 minutes with PBS. A standard indirect immunoperoxidase procedure (Vectastain Elite ABC kit; Vector Laboratories, Burlingame, CA) was performed with 3-amino-9-ethylcarbazole (AEC; Dako) as the substrate. All incubation steps were performed in a moist chamber. Finally, the slides were washed with PBS for 30 minutes, and coverslipped with antifade medium (Vectashield; Vector Laboratories) for viewing. To confirm antibody specificity, the primary antibody preincubated with the immunizing peptide for Ang1 or -2 (100 ng/mL; Santa Cruz Biotechnology) was used. Other staining procedures were the same as described previously. Immunohistochemical staining of each ERM with one of cytokine antibodies was graded as: ±, if few cells stained; +, if only occasional cells stained or if cell clusters of cells, but weakly; and ++, if most cells stained or if clusters of cells stained intensely, as previously reported. ²²  

Double immunofluorescence staining was performed by overnight incubation with the primary antibodies followed by a second incubation for 30 minutes with the corresponding fluorescent dye-conjugated IgG, that is, donkey anti-goat IgG (H+L) (Alexa Fluor 546), rabbit anti-mouse IgG (H+L) (Alexa Fluor 488), goat anti-rabbit IgG (H+L) (Alexa Fluor 546), and goat anti-mouse IgG (H+L) (Alexa Fluor 488; all from Molecular Probes, Inc., Eugene, OR). Slides were washed with PBS for 30 minutes and mounted with antifade medium (Vector Laboratories). Slides were examined and photographed in a laser scanning microscope (LSM 10 BioMedical; Carl Zeiss, Oberkochen, Germany). 

RT-PCR and Sequencing Analysis

For RT-PCR analysis, five membranes were obtained from eyes with ischemic retinal disorder (proliferative diabetic retinopathy), and four membranes were obtained from eyes with nonischemic retinal disorder (idiopathic MP). The membranes were put directly into an RNA extraction solution (Isogen; Nippon Gene, Toyama, Japan) immediately after removal during surgery. Total cellular RNA was prepared according to the manufacturer’s protocol. In brief, surgical materials were homogenized in 1 mL of the solution, and 200 μL chloroform was added. After centrifugation at 4°C, the aqueous phase was collected, and total RNA was precipitated with an equal volume of isopropanol. RNA was then dissolved in 10 μL water treated with diethyl pyrocarbonate. 

We calculated the relative amount of RNA in each case by quantifying the amplified β-actin cDNA fragment, because the total amount of RNA extracted in each case was below the limit of the ordinary measurement with an ultraviolet photometer because of the minute size of the tissue. Two microliters of the solution containing total RNA was reverse transcribed with a cDNA synthesis kit (First-Strand; Pharmacia Biotech, Uppsala, Sweden) at 37°C for 1 hour in a 15-μL reaction volume containing a random hexadeoxynucleotide primer and Moloney murine leukemia virus reverse transcriptase. A 2-μL aliquot of the reaction product was subjected to 35 cycles of PCR for amplification of β-actin cDNA. The density of the band of amplified β-actin cDNA was measured in each case, and the relative amount of total RNA extracted from each tissue was calculated. Based on these results, we adjusted the starting amount of RNA for further RT-PCR analysis on the expression of Ang1, Ang2, and β-actin. 

RNA was reverse transcribed as described previously, and PCR was performed at 35 cycles in a 50-μL reaction volume containing 800 nM of each primer, 100 μM dNTP, and 5 U Taq DNA polymerase (Toyobo, Tokyo, Japan) in a thermal cycler (Mini Cycler; MJ Research, Watertown, MA). The thermal cycle was 1 minute at 94°C; 2 minutes at 64°C (Ang1), 64°C (Ang2), or 67°C (β-actin); and 3 minutes at 72°C followed by 3 minutes at 72°C. The nucleotide sequences of the PCR primers were 5′-AGAACCACACGGCTACCATGCT-3′ (Ang1 sense primer corresponding to nucleotides +671 to +692), 5′-TGTGTCCATCAGCTCCAGTTGC-3′ (Ang1 antisense primer corresponding to nucleotides +1059 to +1080), 5′-AGCTGTGATCTTGTCTTGGC-3′ (Ang2 sense primer corresponding to nucleotides +377 to +396), 5′-GTT CAAGTCTCGTGGTCTGA-3′ (Ang2 antisense primer corresponding to nucleotides +802 to +821), 5′-TGA CGG GGT CAC CCA CAC TGT GCC CAT CTA-3′ (β-actin sense primer corresponding to nucleotides +541 to +570), and 5′-CTA GAA GCA TTT GCG GTG GAC GAT GGA GGG-3′ (β-actin antisense primer corresponding to nucleotides +1171 to +1201). An aliquot of the PCR product was electrophoresed in a 1.5% agarose gel and stained with ethidium bromide. These cDNAs were cloned with the RT-PCR method recommended by the manufacturer. The PCR products were then subcloned into a vector (pCRII; Invitrogen, San Diego, CA) and sequenced in their entirety. Comparison with the published human sequences revealed complete sequence identity. For the positive control of PCR, total RNA was harvested from cultured human umbilical vein endothelial cells (HUVECs), and RT-PCR was performed in the same manner. For the negative control, PCR was performed with same amounts of RNA samples without RT. 

In Situ Hybridization

Slides of paraffin-embedded specimens were treated with 0.2 M HCl for 20 minutes, followed by washing in PBS containing 0.01% diethyl pyrocarbonate, digestion with 20 μg/mL proteinase K at 37°C for 10 minutes, and fixation in 4% paraformaldehyde for 5 minutes. Blocking was performed in PBS containing 50% formamide and 2× SSC at room temperature for 1 hour. Sense and antisense Ang2 cRNA probes were generated from the plasmid made by RT-PCR in our previous study ¹⁸ and labeled with digoxigenin-dUTP (DIG RNA labeling kit, Roche Molecular Biochemicals, Mannheim, Germany), as recommended by the manufacturer. The efficiency of labeling was confirmed by agarose gel electrophoresis. The probe was used at a concentration of 50 ng/section. Hybridization was performed at 45°C for 16 hours. After extensive sequential washings in 2×, 1×, and 0.5× SSC, the unhybridized probe was digested with RNase (Promega, Madison, WI) in 0.5× SSC. The hybridization product was detected after incubation with an alkaline phosphatase-conjugated anti-digoxigenin antibody (1:500 dilution; Roche Molecular Biochemicals) overnight at 4°C, followed by development in 4-tetrazolium chloride (1:50 dilution; Roche Molecular Biochemicals) overnight at room temperature. 

Tube-Formation Assay in Retinal Vascular Endothelial Cells

Primary cultures of bovine retinal endothelial cells (BRECs) were isolated by homogenization and a series of filtration steps, as previously described. ²³ Bovine eyes were purchased from an abattoir. Cells were grown on fibronectin (Sigma, St. Louis, MO)-coated dishes (Iwaki Glass, Tokyo, Japan) containing Dulbecco’s modified Eagle’s medium (DMEM) with 5.5 mM glucose, 10% platelet-derived horse serum (PDHS; Wheaton, Pipersville, PA), 50 mg/mL heparin, and 50 U/mL endothelial cell growth factor (Roche Molecular Biochemicals). Cells were characterized for their endothelial homogeneity by immunoreactivity for factor VIII antigen and remained morphologically unchanged under these conditions, as confirmed by light microscopy. The tube formation assay was performed as previously reported. ^{23 24} An 8:1:1 (400 μL) mixture of Vitrogen 100 (Celtrix, Palo Alto, CA), 0.2 N NaOH and 200 mM HEPES in 10× RPMI medium (Gibco BRL, Gaithersburg, MD), containing 5 μg/mL fibronectin and 5 μg/mL laminin, was made and added to 24-well plates. After polymerization of the gels, 1.0 × 10⁵ BRECs were seeded and incubated for 24 hours at 37°C with DMEM containing 20% PDHS. The cell number was chosen to optimize the shape and tube length, based on the results from previous studies. ^{23 24} The medium was removed, and additional collagen gel was introduced onto the cell layer. Growth factors (VEGF, 50 ng/mL; Ang1*, 200 ng/mL; Ang2, 200 ng/mL; or combinations) and hypoxia-normoxia-conditioned media, with or without Tie2-Fc (2 μg/mL), were then added to the cultures in the amounts indicated for each experiment. Before making the collagen gel, six points were randomly marked in the center area of the bottom of each well. Five days later, the density per surface area of the tubelike structures was determined in each of six fields randomly preselected by computer (Winroof; Mitani Corp., Osaka, Japan). All groups were studied in quadruplicate in three independent experiments. 

Hypoxia-Conditioned Medium

Confluent cell monolayers of BRECs in DMEM with 5.5 mM glucose and 1% PDHS were exposed to 1% ± 0.5% oxygen in a water-jacketed mini-CO₂/multigas incubator with reduced oxygen control (model BL-40M; Jujikagaku, Tokyo, Japan). All cells were maintained at 37°C in a constant 5% CO₂ atmosphere with oxygen deficiency induced by replacement with nitrogen. Hypoxia-conditioned medium was collected from confluent cultures after 24 hours and was filtered before use to remove any cellular components. Medium from dishes cultured under normoxic conditions (95% air, 5% CO₂) served as the control. 

Tie2 and VEGF Receptor Inhibition in a Mouse Model of Ischemia-Induced Retinal Neovascularization

The well-established mouse model of ischemia-induced retinal neovascularization was created as previously described. ^{2 4 25} All animals were handled according to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Briefly, litters of 7-day-old (postnatal day [P]7) C57BL/6J mice were exposed to 75% ± 2% oxygen for 5 days and then returned to room air at P12 to produce retinal neovascularization. Mice of the same age, maintained in room air, served as the control. Maximal retinal neovascularization was observed at P17, 5 days after return to room air (data not shown), in good agreement with previous reports. ²⁵ A solution (0.5 μL) containing 0.67 μg sTie2-Fc, 0.25 μg sFlt-1-Fc, or both fusion proteins was injected into the vitreous of one eye with a 32-gauge needle (Hamilton, Reno, NV) on P12 and P14, as previously described. ² To attempt to block VEGF and angiopoietins, which may induce angiogenesis as a result of retinal hypoxia, days 12 and 14 were selected for the injections. As a control, an equivalent amount of human IgG was injected into the contralateral eye. At P17, the mice were killed by cardiac perfusion of 1 mL 4% paraformaldehyde in PBS, and the eyes were enucleated and fixed in 4% paraformaldehyde overnight at 4°C before paraffin embedding. Serial 6-μm paraffin-embedded axial sections were obtained from the optic nerve and stained with hematoxylin and periodic acid-Schiff, according to a standardized protocol. All retinal vascular nuclei anterior to the internal limiting membrane were counted in each section by a fully masked protocol. For each eye, 10 intact sections of equal length, each 30 μm apart, were evaluated. The mean number of neovascular nuclei per section per eye was then determined. No retinal detachment or other damage related to the needle puncture was observed. 

Statistical Analysis

All determinations were performed in triplicate, and experiments were repeated three times, unless otherwise indicated. Results are expressed as the mean ± SD. One-way ANOVA followed by the Fisher t-test was used to evaluate significant differences, and P < 0.05 was selected as the statistically significant value. For evaluation of in vivo retinal angiogenesis, the χ² test for categorical data and the paired Student’s t-test or the Mann-Whitney rank sum test for quantitative data with unequal variance are used. 

Results

Immunostaining of Angiopoietins and Tie2 in Human Ocular Specimens

Immunohistochemistry was performed to identify the expression of angiopoietins and Tie2 in epiretinal membranes obtained from eyes with ischemic retinal disorders and to compare their expression with that in eyes with nonischemic retinal disorders. Immunostaining by the anti-Ang1 antibody was almost abolished by the Ang1 immunizing peptide but not by the Ang2 immunizing peptide (Figs. 1a 1b 1c) . Regarding Ang2 staining, the staining disappeared during preincubation with Ang2 immunizing peptides but not with Ang1 immunizing peptides (Figs. 1e 1f 1g) . These data suggest that these antibodies specifically stain Ang1 and Ang2. 

In ischemic retinal disorders, immunostaining of Ang1 was observed in most ERMs from eyes with proliferative diabetic retinopathy (PDR, 8/12) and retinal vein occlusion (RVO, 1/1). Prominent staining was observed in only 1 of the 13 ERMs (Table 1) . In nonischemic retinal disorder (idiopathic MP), Ang1 immunostaining was observed in more than half of all ERMs (5/9) and prominent staining was not observed (0/9; Table 1 ). Patchy Ang1 immunostaining was observed that did not correlate with vascular lumens, and no remarkable differences were observed between staining of both types of retinal disorders (Figs. 1a 1d) . 

In contrast, Ang2 immunostaining was more prominent in ERMs from eyes with ischemic retinal disorders (Figs. 1e 1h) . The staining was well localized to the vascular lumen and was most marked in highly vascularized regions. Ang2 immunostaining was prominent in 10 of the 13 membranes from eyes affected by ischemic retinal disorders (9/12 in PDR, 1/1 in RVO; Table 1 ), whereas prominent Ang2 immunostaining was observed in only one of nine membranes from eyes affected by nonischemic retinal disorders (Table 1) . Similar to Ang2, Tie2 immunostaining was more prominent in membranes from eyes with ischemic retinal disorders than in those with nonischemic membranes (Fig. 2) . Tie2 immunostaining was prominent in 12 of the 13 ischemic membranes, whereas 2 of 9 nonischemic membranes were prominently stained (Table 1) . 

To investigate further the correlation of the presence of Ang and Tie2 with that of VEGF, we stained the same specimens with anti-VEGF antibody. VEGF is a primary mediator of ischemic retinal neovascularization. As expected, immunostaining of VEGF was more prominent in eyes with ischemic retinal disorders (9/12 in PDR, 1/1 in RVO) than in those with nonischemic retinal disorders (3/9; Table 1 ). The localization of VEGF was very similar to Ang2, with staining observed in the highly vascularized region (Fig. 2) . As shown in Table 1 , staining of Ang2 had intensity similar to VEGF staining in most (8/13) of the fibrovascular membranes from eyes with ischemic retinal diseases (eyes 12, 13, 15, 18–22). In contrast to Ang1, the immunoreactivity of Ang2, VEGF, and Tie2 tended to be higher in the highly vascularized regions in ERMs from eyes with ischemic retinal disorders (Table 1 , Figs. 1 2 ). 

To determine the specific cell type that expresses Ang2 and Tie2, a double-immunofluorescence study was performed using monoclonal antibodies against cellular markers: CD34 for vascular endothelial cells, CD68 for macrophages, and GFAP for glial cells. Most vascular endothelial cells that were immunopositive for CD34 were stained with Ang2, particularly in highly vascularized membranes (Fig. 3a) . Immunofluorescence of Ang2 was not detected in the CD68-positive cells or in the GFAP-positive cells (data not shown). Tie2 was originally identified as an endothelium-specific receptor tyrosine kinase. As expected, all vascular endothelial cells were positive for Tie2 (Fig. 3b) . In addition, an overlap between Tie2-positive cells and GFAP- and CD68-positive cells was not observed (data not shown). 

RT-PCR and In Situ Hybridization Analysis of Angiopoietin Genes

RT-PCR was performed to determine the expression of angiopoietin genes in surgically excised specimens. The Ang2 gene was expressed in most of the ERMs from eyes with ischemic retinal diseases (5/6). In contrast, no PCR product was observed in the membranes of eyes with nonischemic retinal diseases (0/5; Fig. 4 ). Negative control experiments without RT did not show any PCR products. Regarding Ang1, the same PCR cycle that revealed the Ang2 gene did not produce a PCR product of Ang1 in either ischemic or nonischemic membranes, although it produced a PCR product of Ang1 from cDNA of cultured HUVECs (data not shown). Because of the low level of Ang1 mRNA expression, we focused on expression of Ang2 in the in situ hybridization study. As shown in Figure 2 , Ang2 mRNA was localized mainly to the cells lining the vascular lumen, suggesting the predominance of the gene in vascular endothelial cells. 

Effects of Angiopoietins on In Vitro Retinal Angiogenesis Induced by VEGF and Hypoxia

To investigate the effects of angiopoietins on in vitro retinal angiogenesis, tube-formation assays were performed using cultured BRECs. Both Ang1 and -2 individually induced significant tube-forming activity, which mimicked sprouting of endothelial cells (Fig. 5) . VEGF induced a 15.4-fold increase in tube-forming activity, which was enhanced by Ang1* and Ang2 by 69% (P < 0.001) and 15% (P < 0.001), respectively. Excess sTie2-Fc precluded the modulation by Ang1* and Ang2. The angiogenesis induced by VEGF alone was also suppressed by sTie2-Fc (16%; P < 0.01; Fig. 5 ). To determine the effect of sTie2-Fc on hypoxia-induced retinal angiogenesis in vitro, tube formation was induced by conditioned medium from BRECs exposed to hypoxia. Hypoxia-conditioned medium stimulated tube-forming activity threefold. The addition of sTie2-Fc suppressed tube-forming activity by 16% (P < 0.01), in contrast to sTie1-Fc, which did not significantly suppress tube formation (Fig. 6) . 

Effects of sTie2-Fc and sFlt-1-Fc on Ischemia-Induced Retinal Angiogenesis in a Murine Model

To investigate the effects of Tie2 inhibition on ischemic retinal angiogenesis in vivo, a highly reproducible murine model of retinal ischemia was used. The mice exposed to 75% oxygen from P7 to P12 exhibited extensive retinal capillary obliteration (data not shown), which correlated well with observations in previous reports. ²⁵ When the mice were returned to room air, the inner retina became hypoxic, expression of Ang2 was upregulated, ¹⁸ and retinal neovascularization occurred in 100% of the animals by P17. Intraocular injections of sTie2-Fc at a dose of 0.67 μg/eye at P12 and P14 reduced histologically evident retinal neovascularization at P17 in 12 (92%) of 13 eyes (P < 0.0001), compared with an equivalent injection of control protein in the contralateral eye (Figs. 7A 7B) . Histology of the mouse retina did not demonstrate any remarkable deterioration of vessel integrity in the retina, as has been reported in the knockout mouse. ¹¹ Because VEGF has been shown to be a predominant regulator of ischemia-induced retinal neovascularization, the effect of Tie2 inhibition was compared with that of VEGF inhibition. The mean magnitude of inhibition compared with the contralateral eye injected with IgG was 23.0%, 37.0%, and 50.0% by injection of sTie2-Fc, sFlt-1-Fc, and both, respectively. Suppression of the neovascular response was evident in histologic examination of paraffin-embedded ocular cross sections (Fig. 7C) . No retinal detachment or other damage related to the needle puncture was observed. 

Discussion

Recent studies have reported a major role of the angiopoietin-Tie2 system in postnatal pathologic angiogenesis and in vascular development in embryogenesis through the regulation of vascular integrity. Ischemia-induced retinal neovascularization often results in catastrophic loss of vision in the final stage of various ocular diseases, including diabetic retinopathy, retinal vein occlusion, and retinopathy of prematurity. In the present study, we showed the potential role of the angiopoietin-Tie2 system in ischemic retinal neovascularization. We found that Ang2 and Tie2 are more markedly expressed in retinal vascular proliferative membranes in human ischemic retinal disorders, such as PDR, than in nonischemic retinal diseases. We also showed that inhibition of Tie2 by sTie2-Fc suppresses in vitro retinal angiogenesis induced by hypoxia-conditioned medium and the retinal angiogenesis in a murine model of retinal ischemia. The inhibition of Tie2, when combined with inhibition of VEGF, more efficiently suppressed retinal angiogenesis than did inhibition of VEGF alone, suggesting that signaling of both Tie2 and VEGF plays a potential role in ischemia-induced retinal angiogenesis. 

Of the two ligands for Tie2, Ang2 was recently reported to play a critical role in pathologic angiogenesis. In contrast to widespread expression of Ang1 and Tie2, ¹⁶ Ang2 is highly expressed only at sites of vascular remodeling in the adult, notably in the female reproductive tract. ¹³ Expression of Ang2 is also selectively enhanced in pathologic angiogenesis associated with tumors, ^{19 20} retinal ischemia in an animal model, ¹⁸ and choroidal neovascularization associated with age-related macular degeneration. ²⁶ In the present study in which we used human retinal specimens, Ang2 staining was more marked in vascular proliferation in eyes with ischemic retinal disorders than that in those without ischemic retinal diseases, whereas Ang1 staining did not differ greatly in the two groups. Ang2 was more prominently stained in highly vascularized regions, as was the receptor Tie2. We further investigated whether the preferential expression of Ang2 in ischemic retinal neovascular membranes results from the abundant mRNA in the cells anchoring the membranes and whether the factor is locally produced. RT-PCR analysis clearly demonstrated that Ang2 mRNA is abundant in ERMs in ischemic retinal diseases, whereas no substantial expression of the Ang2 gene was observed in nonischemic retinal diseases. These data agree with previous reports and reinforce the role of Ang2 in pathologic angiogenesis in ischemic retinal disorders. 

A variety of cell types, including glial cells and macrophages as well as vascular cells, ^{27 28} may be involved in the development of vascular proliferative membranes in ischemic retinal disorders. We performed double immunofluorescent staining to determine the cell types that express both Ang2 and Tie2. The results showed that that both Ang2 and Tie2 were colocalized in most vascular endothelial cells, particularly in those in highly vascularized regions. In situ hybridization demonstrated gene expression of Ang2 in vascular endothelial cells (Fig. 3) . Stratmann et al. ²⁹ demonstrated expression of Ang2 mRNA in angiogenic vascular endothelial cells in glioblastomas by in situ hybridization, and expression of Ang2 mRNA was identified in several types of vascular endothelial cells in vitro. ^{15 17 18} The present data further support production of Ang2 by vascular endothelial cells and local autocrine action of the protein. 

Recent in vitro findings regarding the bioactivity of Ang1 showed that this ligand can induce potent chemotaxis, weak but positive mitogenesis, ¹⁵ capillary sprouts, ³⁰ and an antiapoptotic effect on endothelial cells, ^{20 31} confirming its critical role in angiogenesis. In tube-formation assays, we found that not only Ang1 but also Ang2 enhances tube-forming activity in retinal microvascular endothelial cells. Because Ang2 inhibits Ang1-induced Tie2 signaling in vascular endothelial cells, ^{13 15} the observed response may be paradoxical. A recent report by Teichert-Kuliszewska et al. ³² demonstrated Ang2-dependent tube formation and Tie2 autophosphorylation in HUVECs cultured in three-dimensional fibrin matrices. Another report also showed that Ang2 at a high concentration elicits Tie2-dependent intracellular signaling linked to endothelial cell survival. ³³ In addition, the report of a corneal pocket assay showed that Ang1 and -2 facilitate neovascularization when coadministrated with VEGF. ²¹ Although the molecular mechanism was not investigated in detail, these data suggest that in active angiogenesis, particularly in microvascular endothelial cells, Ang2 can probably induce at least some level of Tie2 signaling, which contributes to endothelial angiogenic functions. 

To investigate the role of the angiopoietin-Tie2 system further, we determined the effect of Tie2 inhibition, by using recombinant soluble sTie2-Fc. The ability of this protein to inhibit Tie2 signaling was confirmed by a tube-formation assay, which showed that sTie2-Fc precluded the modulation by Ang1 and -2 of VEGF-induced tube-forming activity. The application of sTie2-Fc in the same doses suppressed hypoxia-conditioned, medium-induced tube formation. Although Ang1 more effectively enhanced VEGF-dependent tube formation than Ang2, hypoxia enhanced secretion of Ang2, but did not affect expression of Ang1 in the same cell type. ¹⁸ The observed inhibitory response may result from suppression of increased Ang2 binding to Tie2 receptor, as well as stable binding of Ang1 to Tie2 on BRECs. In vivo experiments, intravitreous injection of the protein reduced formation of retinal neovascularization. A recent report using intramuscular infection of the adenovirus-carrying extracellular domain of the Tie2 receptor also demonstrated a suppression of retinal neovascularization by systemically applied Tie2 inhibitor. ³⁴ In addition, we have demonstrated that Ang2 mRNA is upregulated in the inner retinal layer and in neovascular cells in an in vivo model. ¹⁸ These observations indicate a substantial contribution of Ang2-Tie2 signaling in ischemia-induced retinal neovascularization. Stratmann et al. ²⁹ demonstrated that Ang2 is expressed in a subset of angiogenic vasculature with few periendothelial cells in glioblastomas. This suggests that Ang2 probably lessens the interaction between endothelial cells and the periendothelial component as an antagonist of Ang1 in pathologic angiogenesis, as observed in the vascular development of Ang2 transgenic mice. ¹³ It may lead to a favorable environment for endothelial cells to diminish antiangiogenic regulation by pericytes ³⁵ or to make contact with additional angiogenic cytokines. sTie2-Fc probably suppresses such effects of Ang2. Furthermore, the findings from the present in vitro studies and previous reports ^{21 32 33} suggest the existence of Tie2 signaling by Ang2 in retinal microvascular endothelial cells. Because sTie2-Fc entraps Ang2 as well as Ang1, it may suppress not only Ang1- but also Ang2-dependent Tie2 signaling linked to angiogenic activity in endothelial cells. 

Because VEGF is a dominant mediator of ischemic retinal neovascularization, ^{1 2 3 4} we investigated a correlation between expression of VEGF and that of Ang2 and Tie2. By immunohistochemistry, staining for both Ang2 and Tie2 was well colocalized to sites of VEGF staining in vascular proliferation in ischemic retinal disorders. Both Ang1 and -2 enhanced VEGF-induced neovascularization in an in vivo model ²¹ and in the tube-formation assay in the present study. Colocalization of Ang2 and Tie2 with VEGF further supports the notion that Ang2 and VEGF cooperatively contribute to ischemic retinal neovascularization. Conversely, this colocalization may indicate induction of Ang2 by VEGF, because VEGF induces Ang2. ^{17 18} To investigate further the cooperative role of Ang2 and Tie2 with VEGF in retinal angiogenesis, sFlt-1-Fc, an inhibitor of VEGF, was used to inhibit VEGF signaling and was compared with inhibition of Tie2 in an in vivo model of retinal ischemia. Inhibition of Tie2 was less effective than inhibition of VEGF; however, it additively suppressed retinal angiogenesis when combined with inhibition of VEGF. These data suggest that both Tie2 and VEGF signaling play a major role in retinal angiogenesis and that VEGF signaling is predominant. 

Although its exact role remains an enigma, not only VEGF-VEGFR but also the Ang2-Tie2 interaction probably plays a key role in ischemia-induced retinal neovascularization, because Ang2 was upregulated not only in an experimental model but also in human ischemic retinal disorders, Ang2 induced angiogenic activity as did Ang1, and there was an enhanced VEGF effect in retinal microvascular endothelial cells, and because Tie2 inhibition suppressed ischemia-induced retinal neovascularization. Because this system has a substantial role in ischemia-induced neovascularization, these findings may strongly indicate that this system along with the VEGF system can be targeted to treat and/or prevent various ischemic retinal disorders including diabetic retinopathy. Coinhibition of the VEGF and Tie2 systems may be a promising strategy to cure these disorders. 

 Supported by a Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology and the Ministry of Health, Labor and Welfare of the Japanese government.

 Submitted for publication March 19, 2002; revised June 27, 2002; accepted July 12, 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.

 

Immunohistochemistry for Ang1 and -2 in surgically excised ERMs. (a–d) Immunostaining for Ang1; (e–h) immunostaining for Ang2. (a–c, e–g) ERM from an eye with the ischemic retinal disorder PDR; (d, h) ERM from an eye with the nonischemic retinal disorder, idiopathic MP. (a) Immunostaining by the anti-Ang1 antibody was patchy and not well localized to the vascular lumen in the PDR-induced ERM. (b) Staining by the anti-Ang1 antibody was almost abolished by Ang1 immunizing peptide in the same ERM. (c) Staining by the anti-Ang1 antibody in the same ERM was not abolished by Ang2 immunizing peptide. (d) Immunostaining of Ang1 in an MP-induced ERM. Patchy staining was observed, and there was no marked difference between the PDR-induced (a) and the MP-induced (d) ERMs. (e) Immunostaining by the anti-Ang2 antibody in a PDR-induced ERM. Prominent Ang2 staining was observed. Ang2 staining was well localized to the vascular lumen and was most marked in highly vascularized regions. (f) Prominent Ang2 staining was not altered by preincubation of the primary antibody with Ang1 immunizing peptide in the same ERM. (g) The staining virtually disappeared with preincubation with Ang2 immunizing peptides in the same ERM. (h) Immunostaining by the anti-Ang2 antibody was patchy in an MP-induced ERM. Immunostaining of Ang2 was more prominent in a PDR-induced (e) than that in an MP-induced (h) ERM. (a–h) Magnification, ×400; bar (a): 10 μm.

Table 1.

 

Immunohistochemical Staining for Angiopoietins and the Tie2 system in ERMs Secondary to Ischemic or Nonischemic Retinal Disorders

Eye	Diagnosis	Ang1	Ang2	Tie2	VEGF
1	MP	−	+	−	+
2	MP	−	−	−	−
3	MP	+	++	+	+
4	MP	+	+	++	+
5	MP	−	−	−	−
6	MP	+	+	+	++
7	MP	+	+	++	+
8	MP	+	+	+	+
9	MP	−	−	−	−
10	PDR	−	+	+	++
11	PDR	−	+	++	−
12	PDR	+	++	++	++
13	PDR	+	++	++	++
14	PDR	+	++	++	+
15	PDR	++	++	++	++
16	PDR	−	++	++	+
17	PDR	−	+	++	+
18	PDR	+	++	++	++
19	PDR	+	++	++	++
20	PDR	+	++	++	++
21	PDR	+	++	++	++
22	RVO	+	++	++	++

Figure 2.

 

Immunohistochemistry for Tie2 and VEGF and in situ hybridization for Ang2 in surgically excised ERMs. (a, b) Immunostaining for Tie2; (c, d) immunostaining for VEGF. (a, c) ERM from a PDR-affected eye. (b, d) ERM from an eye with idiopathic MP. (e, f) In situ hybridization of Ang2 mRNA in an ERM from an eye with PDR. Immunohistochemistry for Tie2 and VEGF showed more prominent staining in ERM from (a, c) a PDR- than in that from an (b, d) MP-affected eye. Tie2 staining was well localized in the vascular lumen and most marked in highly vascularized regions (a). VEGF staining was very similar to Ang2 and Tie2 staining and was prominent in highly vascularized regions (c). In situ hybridization demonstrated that Ang2 mRNA was localized mainly to the cells lining the vascular lumen, suggesting the predominance of the gene in vascular endothelium (e). The sense probe showed no staining (f). Magnification, ×400; bar (a): 10 μm.

Figure 3.

 

Double immunofluorescence immunohistochemistry of an ERM from a PDR-affected eye. (a) Double labeling revealed an overlap between cells positive for Ang2 (red) and those positive for CD34 (green). Labeling for Ang2 (red) overlapped extensively the labeling for CD34, a marker for vascular endothelial cells. (b) Double labeling showed overlap between cells positive for Tie2 (red) and those positive for CD34 (green). Magnification, ×180.

Figure 4.

 

RT-PCR analysis and in situ hybridization of the angiopoietin gene in retinal proliferative tissue. RT-PCR was performed with total RNA harvested from surgically excised ERMs. Expression of Ang2 was detected in five of six eyes with ischemic retinal diseases (lanes 1–6: PDR) but in none of five eyes with nonischemic retinal disorders (lanes 7–11: idiopathic MP). Expression of the β-actin control was constant in all membranes. Lane 12: negative control of PCR with no reverse transcriptase.

Figure 5.

 

Effects of Ang1 and -2 on tube formation in BRECs. BRECs were seeded in three-dimensional collagen gels and incubated with medium containing vehicle, Ang1, Ang2, VEGF, or sTie2-Fc. Five days later, the density per surface area of the tube-like structures was determined in each of six randomly preselected fields. All groups were studied in quadruplicate in three independent experiments. (A) Representative phase-contrast micrograph of tube formation: (a) unstimulated control or stimulated with (b) Ang1 (200 ng/mL), (c) Ang2 (200 ng/mL), (d) VEGF (50 ng/mL), (e) Ang1 (200 ng/mL) and VEGF (50 ng/mL), (f) Ang2 (200 ng/mL) and VEGF (50 ng/mL), (g) VEGF (50 ng/mL) and sTie2-Fc (2 μg/mL), (h) VEGF and Ang1 with sTie2-Fc (2 μg/mL), or (i) VEGF and Ang2 with sTie2-Fc. (B) Summarized results. Error bars: SD. *P < 0.01, **P < 0.001.

Figure 6.

 

Effects of sTie1-Fc or sTie2-Fc on hypoxia-conditioned, medium-stimulated tube formation in BRECs. BRECs were seeded in three-dimensional collagen gels and incubated with medium conditioned by bovine retinal pericytes exposed to normoxia or hypoxia with vehicle, sTie1-Fc, or sTie2-Fc. Five days later, 10 different fields (10× objective) were chosen in each well, and the density per surface area of the tube-like structures was determined. All groups were studied in quadruplicate in three independent experiments; summarized results are shown. *P < 0.01, **P < 0.001

Figure 7.

 

Inhibitory effects of Tie2 and VEGF signaling with intravitreous injection of sTie2-Fc and sFlt-1-Fc in retinal neovascularization of ischemic murine retina. (A) Mice were injected with (a) human recombinant sTie2-Fc (0.67 μg/eye), (b) sFlt-1-Fc (0.25 μg/eye), or (c) both sTie2-Fc and sFlt-1-Fc in one eye. Equivalent doses of control human IgG were injected into the contralateral eye. The average number of neovascular cell nuclei per 6-μm histologic section per eye was then determined. Solid lines connect eyes from the same animal. Arrows: mean of each group. (B) Retinal neovascularization determined from neovascular nuclei counts of chimeric protein-injected eyes is expressed as a percentage of the contralateral control eye in each treatment group. Error bars indicate SE in all animals in each group. *P < 0.05; **P < 0.01; ***P < 0.001. (C) Typical histologic findings in the corresponding retinal locations in eyes injected with (a) control human IgG, (b) sTie2-Fc, (c) sFlt-1-Fc, or (d) both sTie2-Fc and sFlt-1-Fc. Arrows: an area of retinal neovascularization with vascular cells internal to the inner limiting membrane. G, ganglion cell layer; I, inner nuclear layer; O, outer nuclear layer; and R, retinal pigment epithelium.

The authors thank George D. Yancopoulos, Regeneron Pharmaceuticals, Inc. (Tarrytown, NY), for the generous gifts of human Ang1*, Ang2, sTie1-Fc, and sTie2-Fc and for helpful discussions and Kazuo Nishimura for technical assistance and helpful discussions.