Angiostatic Effect of Penetrating Ocular Injury: Role of Pigment Epithelium-Derived Factor

Abstract

purpose. To characterize the angiostatic effect of penetrating ocular injury and to begin to explore its mechanism, with an emphasis on the role of pigment epithelium–derived factor (PEDF).

methods. Using the rat model of oxygen-induced retinopathy (OIR), single or multiple dry needle injuries were made, penetrating the globe of one eye; the opposite eye served as a control. Eyes were harvested from rats killed 1, 3, and 6 days after injury, and retinas were dissected and processed for assessment of neovascularization and microglial activation or were processed for genetic and proteomic analysis. Temporal and spatial expression patterns of PEDF were analyzed by in situ hybridization.

results. Penetrating ocular injury resulted in a 30% decrease in neovascular area in the retinas of OIR rats. At day 1 after injury, needle insertion caused a 4.1-fold increase in retinal PEDF mRNA and a 1.5-fold increase in retinal PEDF protein. Vitreous PEDF protein increased 3.4-fold in injured eyes compared with noninjured eyes. In situ hybridization showed an increase in PEDF mRNA in areas surrounding the puncture site. Concentrated vitreous protein from injured eyes caused a 60% decrease in retinal neovascularization when injected into the vitreous cavity of OIR rats. Preincubation of vitreous samples with anti-PEDF partially abolished this efficacy.

conclusions. The pattern of angiostasis resulting from penetrating ocular injury is consistent with the release of an endogenous antiangiogenic factor from the wound site. Preliminary studies show a possible role for PEDF in this effect. Further characterization of this role and the identification of other factors may lead to new therapeutic strategies for angiogenic eye conditions.

Angiogenesis is the hallmark of a number of blinding ocular conditions, including diabetic retinopathy, macular degeneration, retinopathy of prematurity, and vein occlusion retinopathy, to name a few. In angiogenesis, new blood vessels originate from existing venules by a series of events involving the disruption of the endothelial cell basement membrane (BM) and extracellular matrix (ECM) by proteolytic enzymes, ^{1 2} endothelial cell proliferation and migration, ^{3 4} and formation and closure of the nascent vascular tubes. ⁵ The inhibition of any one of these events or a combination is of potential therapeutic value for blinding conditions in which angiogenesis participates. 

Current therapies for retinal and subretinal angiogenesis include ablative photocoagulation methods. ^{6 7 8 9} These techniques are inherently destructive, and improved therapeutic strategies are desirable. Accordingly, recent clinical trials have tested the efficacy of antiangiogenic drugs in patients with neovascular or “wet” macular degeneration. The agents pegaptanib ¹⁰ (Macugen; Eyetech Pharmaceuticals, Inc., New York, NY) and ranibizumab ¹¹ (Lucentis; Genentech, San Francisco, CA) are delivered by intravitreal injection. Clearly, this is a case in which the risks of procedure-related adverse events (eg, endophthalmitis, retinal detachment, altered intraocular pressure) are balanced by the potential beneficial effects and by the lack of viable alternatives. 

Intravitreal injection is frequently used in preclinical trials to screen classes of agents or to indicate the merit of a specific therapeutic strategy. In experiments using a rat model of oxygen-induced retinopathy (OIR), we noted a consistent and significant angiostatic influence of vehicle injection, an important experimental control in these preclinical experiments. ^{12 13 14 15} We have now systematically examined the potential of dry needle intravitreal penetration to inhibit abnormal retinal and preretinal neovascularization (NV). The results of these experiments support the notion that the most effective angiostatic drug might be found within the retina’s natural complement of proteins. Identification of endogenous retinal angiostatic factors will allow for the development of novel and less invasive therapies for retinal neovascular abnormalities. 

Herein, we describe the angiostatic effect of penetrating ocular injury and report our preliminary efforts to uncover its mechanism. Our findings strongly suggest that one contributing factor is pigment epithelium–derived factor (PEDF). PEDF shares homology with the serpin superfamily, but its known bioactivities are apparently not related to serpin activity. ¹⁶ It has been found in the interphotoreceptor matrix, in numerous locations in the brain, in the vitreous and aqueous humors, in the spinal cord and pineal gland, in association with fibroblasts, in ciliary epithelium, and in cultured retinal pigment epithelium (RPE). ¹⁷ PEDF binds to heparin and other glycosaminoglycans in the ECM through lysine residues at a novel binding site for members of the serpin family. ^{18 19} PEDF is among the most potent known natural antiangiogenic factors, based on the inhibition of endothelial cell migration in vitro. It is even more active than angiostatin, thrombospondin-1, and endostatin. ²⁰ PEDF inhibits endothelial cell migration toward many angiogenic factors, including platelet-derived growth factor, vascular endothelial growth factor (VEGF), IL-8, fibroblast growth factor (FGF), and lysophosphatic acid. ²⁰ It shows endothelial cell specificity, inhibiting the migration of bovine adrenal microvascular endothelial cells and human umbilical vein endothelial cells, but it does not inhibit the migration of human foreskin or lung fibroblasts, aortic smooth muscle cells, oral keratinocytes, or neutrophils toward stimulatory cytokines. ²⁰ In rat cornea, the addition of neutralizing antibodies to PEDF, in the absence of exogenous angiogenic inducers, stimulates the invasion of new vessels. In the vitreous, the removal of PEDF also induces angiogenic activity. ²⁰  

Experiments using the mouse model of OIR ²¹ have successfully used adenoviral vectors to overexpress PEDF or direct injections of PEDF protein to inhibit retinopathy. ^{22 23 24} The significant decrease in neovascular area in these experiments demonstrates the effectiveness of this endogenous protein as an antiangiogenic therapy and its potential importance in the influence of a penetrating ocular injury, a procedure that we show increases PEDF levels. 

Methods

Oxygen Exposure and Penetrating Ocular Injury

Litters of Sprague-Dawley rats and their mothers were transferred within four hours after birth to infant incubators, where they were exposed to 50% oxygen and then 10% oxygen in alternating 24-hour periods. Control rats were raised simultaneously in room air. On postnatal day (P)14, oxygen-exposed rats were removed to room air. At this time rats were sedated with methoxyflurane vapors, eyelids were separated, and 0.5% proparacaine ophthalmic solution was applied. A 30-gauge dry needle was inserted into the mid-vitreous cavity in one eye of oxygen-exposed (n = 15) and room air-raised rats (n = 14). Insertion occurred approximately 0.5 mm posterior to the ora serrata (posterior approach) at the nasal or the temporal pole, and the needle was advanced to an area just above the posterior pole. Care was taken to avoid contact with the posterior surface of the lens during penetration. This procedure was followed by topical application of neomycin/polymyxin B/gramicidin ophthalmic drops (Alcon, Fort Worth, TX). On P17 (day 3 after exposure in room air), oxygen-exposed (n = 16) and room air-raised groups (n = 15) received the same treatment, again in one eye only. Untreated fellow eyes in all rats served as controls. To determine whether effects were specific to puncture wounds that penetrated the sclera, choroid, and retina, some oxygen-exposed rats (n = 12) were injured by a more anterior needle penetration in one eye on P14. In this case, the needle was inserted immediately anterior to the limbus and was advanced through the ciliary body and into the vitreous cavity (anterior approach). To determine the cumulative effect of multiple injuries, a small cohort of rats (n = 24) received more than one needle penetration at P14. All experiments were performed in accordance with institutional guidelines and with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Assessment of Normal and Pathologic Vessel Growth

On P20 (day 6 after exposure), rats were killed, both eyes were enucleated, and retinas were dissected and stained for adenosine diphosphatase (ADPase) activity to illustrate normal and pathologic retinal blood vessels. ²⁵ Images of retinas were then digitized and captured (Adobe PhotoShop; Adobe Systems, San Jose, CA) and stored on a computer (Macintosh G4 or G5; Apple, Cupertino, CA). Using software (Enhance 3.0; MicroFrontier, Des Moines, IA), areas of interest within the images were traced on the monitor screen with an interactive stylus (FTG Data Systems, Stanton, CA). The preretinal nature of the NV was confirmed, where required, using simultaneous viewing of the tissue under 200× magnification on a microscope (Zeiss Standard 16; Carl Zeiss, Thornwood, NY). In this way the avascular area, total retinal area, total vascular area, and total area of pathologic vessel growth were measured. Values were converted from pixels to square millimeters for all area measurements. The percentage of vascular area relative to total retinal area was calculated. All treatment groups were compared for total pathologic growth and vascularization. Geographic distribution of pathologic growth relative to puncture site was determined by dividing retinas into quadrants and summing area of preretinal NV for each of the four regions. Statistical comparisons were made between injured and noninjured eyes using analysis of variance and Dunnett’s post hoc analysis. ²⁶  

Assessment of Retinal Microglial Activation and Recruitment

Variable oxygen-exposed and room air-raised rats received a single, posterior penetrating ocular injury in the temporal quadrant on P14 according to the method described previously. Opposite eyes served as noninjured controls. Retinas were harvested at P20 from variable oxygen-exposed and room air-raised rats, dissected, whole mounted, and stained for major histocompatibility complex class 2 antigen (MHC 2) activity with mouse anti–I-A OX-6 (Serotec, Oxford, UK), and mouse anti–OX-42 (Serotec). OX-6 is directed against a monomorphic determinant of rat MHC 2 antigens, expressed by activated microglia but not resting cells. OX-42 is directed against the type 3 complement receptor (CR3) and recognizes macrophages and microglial cells. Most proangiogenic or antiangiogenic effects ascribed to microglia rely on microglial activation, a process that is characterized by the diversity in microglial activation markers and morphologic features. Thus, a complete analysis of the microglial population was facilitated by combining markers for activated (OX-6) and resting (OX-42) cells. ²⁷ After overnight fixation in periodate-lysine-paraformaldehyde (PLP), retinas were washed with phosphate-buffered saline (PBS) and soaked in 1.5 mL endogenous blocking solution (ImmunoPure Peroxidase Suppressor; Pierce, Rockford, IL) for 30 minutes and, after two 10-minute washes in PBS, were transferred to 12-well plates. Tissue was blocked for 2 days with 10% normal horse serum (NHS) in a 1% solution of triton-100 in PBS and then was incubated in primary antibody (mouse anti-rat OX-6 at 1:200 or OX-42 at 1:100) or negative control (mouse IgG) in 0.1% triton and 10% NHS in PBS for 2 days. After five washes in 0.1% triton/PBS for 30 minutes each, retinas were rinsed with PBS and incubated in secondary antibody (biotinylated horse anti-mouse IgG, rat adsorbed [Vector Laboratories, Burlingame, CA]) diluted 1:250 in 0.1% triton/NHS/PBS for 2 days at room temperature. Three 30-minute washes in 0.1% triton/PBS were followed by incubation for 90 minutes at room temperature in horseradish peroxidase (HRP)–conjugated streptavidin, diluted to 2 mg/mL with 0.1% triton/PBS (1:250). Retinas were then washed twice in 0.1% triton/PBS for 30 minutes each and in PBS alone for 30 minutes. Finally, retinas were flattened on glass slides, and one drop of substrate (AEC; BioGenex, San Ramon, CA) was placed on the tissue. The reaction was stopped with PBS rinses, and the retinas were coverslipped, photographed, and analyzed by counting stained microglia in random fields within the injured and opposite quadrants. 

Assessment of Retinal Gene Expression and Protein Level

To investigate what, if any, changes might have occurred in retinal gene expression or in protein levels after penetrating injury, other litters of rats were exposed to the 50%/10% oxygen paradigm. At the time of removal from the exposure chamber, penetrating injuries were given through the temporal pole of the left eyes at P14, as described. Rats were killed at P15, P17, or P20, and the injured and noninjured retinas from three eyes were collected, pooled, and frozen in liquid nitrogen. 

Total RNA and protein were isolated from these samples using standard methods and reagent (Trizol Reagent; Invitrogen, Carlsbad, CA). Each RNA sample was quality controlled for DNA and protein contamination and integrity. Microfluidic assay (Bioanalyzer; Agilent Technologies, Palo Alto, CA) was used to assay integrity, and spectrophotometric and fluorometric methods were used to quantify the protein and nucleic acids in the sample. By combining specific fluorescent assays with standard spectrophotometry, the amount of RNA, DNA, and protein were specifically quantified for rigorous quality control of each sample. After quality control, the RNA was prepared for microarray analysis using a standard protocol (Affymetrix protocol; Affymetrix, Santa Clara, CA). Briefly, 5 μg total RNA was reverse transcribed to double-stranded cDNA using an oligo-dT primer coupled to a T7 promoter. In vitro transcription from the double-stranded cDNA was then carried out using T7 polymerase and incorporating biotin-modified CTP and UTP ribonucleotides. Biotinylated cRNA samples (15 μg) were fragmented and hybridized (GeneChip Rat Expression Set 230; Affymetrix) that contained full coverage of the transcribed rat genome (30,200 transcripts). Hybridized cRNA was detected using streptavidin coupled to phycoerythrin and was visualized using a laser scanner. Image data were quantified to generate gene expression values and ratios of gene expression between the two hybridized samples from injured and noninjured eyes. 

Retina samples were homogenized for two-dimensional gel electrophoresis using a tissue grinder (Kontes, Vineland, NJ) in 700 μL homogenization buffer containing 320 mM sucrose, 10 mM HEPES (pH 7.5), 4 mM dithiothreitol (DTT), 0.5 mM MgSO₄, 2 mM EDTA, 2 mM EGTA, 0.1 mM sodium vanadate, 5 mM potassium fluoride, 10 mM benzamidine, 100 μg/mL phenylmethylsulfonyl fluoride (PMSF), 50 μg/mL aprotinin, 10 μg/mL leupeptin, and 10 μg/mL pepstatin A. Amounts of protein in the homogenates were assayed with the bicinchoninic acid protein assay (BCA) kit (Pierce, Rockford, IL). Varying amounts of the total proteins (100, 200, 300, and 400 μg) were mixed with isoelectric focusing (IEF) lysis buffer with final concentrations of 9.5 M urea, 5% β-mercaptoethanol, 2% Nonidet P-40, and 2% ampholyte (Bio-Lyte 310; Bio-Rad, Hercules, CA). After centrifugation of the IEF samples at 10,000g for 10 minutes, soluble fractions were collected and applied to the anode of IEF gels. IEF was carried out at 350 V for 14 hours and then at 500 V for 30 minutes. After focusing, IEF gels were applied to the second dimension on 10% slab gel sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), as previously described. ²⁸ The isoelectric point in the IEF dimension and the molecular weight in the SDS-PAGE dimension were estimated by running marker proteins. 

When expression level consistently differed by fourfold in three consecutive microarrays and was supported by a consistent difference of gene product in three consecutive two-dimensional gels, the product was targeted for additional study. Those targets that had a history of angiostatic or angiogenic capacity in some biologic system were examined with Western and Northern blot analyses. 

For Northern blot analysis, total RNA was isolated, as described, from three pooled retinas. Twenty micrograms total RNA from each sample was resolved on a 1% agarose- formaldehyde denaturing gel. The RNA was blotted to 0.2 μm neutral nylon membranes (Schleicher & Schuell, Keene, NH) and hybridized to a ³²P-labeled probe (3 × 10⁶ cpm/mL) specific to the gene of interest. The membrane was washed with 1× SSC, dried, and developed by autoradiography. ³²P-labeled probes specific for PEDF mRNA were prepared by random primer labeling (Rediprime II DNA Labeling System; Amersham, Piscataway, NJ) of PEDF cDNAs (generous gift from Jim McGinnis, Oklahoma University, Oklahoma City, OK). Each Northern blot was repeated at least three times. 

Relative basic FGF (bFGF) and VEGF mRNA expression levels in isolated retinal quadrants from oxygen-exposed rats were determined by real-time reverse transcription–polymerase reaction (RT-PCR). Eyes from oxygen-exposed rats were injured in the temporal retinal quadrant on P14. The injured temporal and the opposite nasal retinal quadrants were dissected on P15, and total RNA was isolated as described. For bFGF amplification, cDNAs were reverse transcribed using a commercial kit (High-Capacity cDNA Archive Kit; Applied Biosystems, Foster City, CA) according to the manufacturer’s protocol. For VEGF amplification, cDNAs were reverse transcribed using another product (Superscript First-Strand Synthesis System for RT-PCR; Invitrogen) according to the manufacturer’s protocol. Quantitative real-time RT-PCR was performed in duplicate by coamplification of rat bFGF and β-actin (endogenous normalization control) in separate tubes using gene-specific according to the manufacturer’s protocol (assays TaqMan Gene Expression Assays; Applied Biosystems; primer and probe sequences used in this assay are proprietary). For relative VEGF expression, VEGF and β-actin (endogenous normalization control) were coamplified in separate tubes in duplicate using the iQ SYBR Green Supermix (Bio-Rad) at a final concentration of 1×, with up and down primers at a concentration of 200 nM each in a total volume of 50 μL. The rate of accumulation of amplified DNA was monitored by continuous measurement of fluorescence (SYBR Green I; Molecular Probes, Eugene, OR). Primer sequences were: VEGF upstream, 5′-CAA TAG CTG CGC TGG TAG ACG TCC-3′; VEGF downstream, 5′-CAA TAG CTG CGC TGG TAG ACG TCC-3′; β-actin upstream, 5′-CCA GGC ATT GCT GAC AGG ATG CAG-3′; β-actin downstream, 5′-GAG GCC AGG ATA GAG CCA CCA ATC-3′. The protocol used for amplification was an initial denaturation at 94°C for 2 minutes, followed by 35 cycles of 15 seconds at 94°C, 30 seconds at 58°C, and 1 minute at 72°C. Melt-curve analysis was performed immediately after amplification by increasing the temperature in 0.5°C increments starting at 50.0°C for 80 cycles of 10 seconds each. Based on the analysis of the melting curves and agarose gel electrophoresis, a single PCR product of the correct size was observed. RT-PCR data were analyzed and expressed according to the comparative C_T method described elsewhere (User Bulletin 2; ABI Prism 7700 Sequence Detection System; Applied Biosystems). 

For Western blot analysis, either the vitreous from six eyes or from the retinas of three eyes were pooled in 300 μL cold lysis buffer (150 mM NaCl, 1.0% TritonX-100, 0.1% SDS, 50 mM Tris-HCl, 100 μg/mL PMSF, 1 mM orthovanadate, 0.3 μg/mL EDTA, 0.5% deoxycholate acid, 50 μΜ NaF, 0.5 μg/mL leupeptin, 0.7 μg/mL pepstatin A, and 1.0 mg/mL aprotinin) and homogenized by sonication at 4°C. The samples were incubated at 4°C for 30 minutes and then centrifuged at 5000 rpm for 15 minutes at 4°C. Protein concentrations of the supernatants were determined with the BCA kit (Pierce). The volume of each sample was adjusted to a protein concentration of 2.5 μg/μL with cold lysis buffer containing protease inhibitors. Twenty microliters (50 μg) was mixed with 20 μL of 2× Laemmli buffer (Sigma, St Louis, MO) and heated at 95°C for 5 minutes. The samples were resolved by SDS-PAGE and were transferred to 0.2 μm nitrocellulose membranes (Bio-Rad). Nitrocellulose membranes were blocked with TBST-1% bovine serum albumin (Sigma) and were probed with primary antibodies. Either goat anti-mouse IgG HRP (Chemicon, Temecula, Ca), goat anti-rabbit IgG-HRP (Chemicon), or donkey anti-goat IgG-HRP (Santa Cruz Biotechnology, Santa Cruz, CA) secondary antibodies were applied to the membranes and were developed with enhanced chemiluminescence (ECL; Amersham, Piscataway, NJ). The following primary antibodies were used in this study: anti-TIMP-1, -TIMP-2 and -TIMP-3 (Chemicon); anti–PAI-1 and anti-VEGF (Santa Cruz Biotechnology); anti-endostatin (R&D Systems, Minneapolis, MN); anti-PEDF (generous gift from Noel Bouck, Case Western Reserve University, Cleveland, OH). Each Western blot was repeated at least three times. 

Retinal VEGF was also measured by ELISA (VEGF Quantikine M Colormetric Sandwich ELISA; R&D Systems) at 1, 3, and 6 days after injury in injured and control eyes from oxygen-treated rats according to the manufacturer’s protocol. 

In Situ Hybridization

Mouse PEDF cDNA (1374 bp, GenBank accession number AF017055 donated by Ignacio Rodriguez) subcloned into the pBluescript SK vector (Stratagene, La Jolla, CA) was linearized with HindIII and XbaI to produce templates for in vitro transcription of antisense and sense riboprobes. This transcription was performed using a DIG RNA labeling kit (Roche Diagnostics, Indianapolis, IN) to produce a single-stranded digoxigenin-11-uridine-triphosphate-labeled probe. The probe was then hydrolyzed to 500 bp according to a limited alkaline hydrolysis procedure (www.roche-applied-science.com) and checked for size by agarose gel electrophoresis and by ethidium bromide staining. 

Eyes from oxygen-treated rats were injured by dry needle puncture in the temporal retinal quadrant on P14. The eyes were enucleated on P15 and embedded in paraffin. Seven-micrometer paraffin sections of rat eyes were placed onto silane-coated slides and deparaffinized with xylene. The slides were then rehydrated with an alcohol gradient and washed with PBS containing 0.3% Triton X-100 and PBS. Sections were permeabilized with a proteinase K treatment (25 μg/mL in TE buffer at 37°C for 30 minutes) and were fixed with 4% paraformaldehyde at 4°C for 5 minutes. After a wash with PBS, the sections were acetylated in a solution of 0.1 triethanolamine in TEA buffer containing 0.25% acetic anhydride (added just before use). They were then incubated in a 50% formamide solution of 4× SSC, and then a hybridization solution (40% formamide, 10% dextran sulfate, 4× SSC, 0.2 mg/mL BSA, 10 mM DTT, 1 mg/mL tRNA, and 1 mg/mL salmon sperm DNA) containing the riboprobe was placed on each section and left overnight at 42°C. The sections were then washed with 1× SSC and were incubated for 30 minutes at 37°C in NTE buffer containing 20 μg/mL RNase. The slides were rinsed with 0.1× SSC, and the probe was detected using the DIG Nucleic Acid Detection Kit (Roche Diagnostics). PEDF mRNA activity was observed and photographed on a light microscope (Olympus AX70; Olympus, Tokyo, Japan). 

Extraction of Retinal and Vitreous Protein, Partial Purification, Antibody Blocking, and Injection

On P15, 1 day after the removal of rats (n = 20) from the oxygen exposure chamber and after penetrating ocular injury, the vitreous and retina were removed and were separately pooled from injured and noninjured eyes. Pooled samples were placed in 400 μL PBS with 200 μg/mL PMSF, 1.4 μg/mL pepstatin A, and 0.6 μg/mL EDTA. The resultant solution was homogenized with 15 strokes using a handheld homogenizer at 4°C. Retinal protein samples were centrifuged at 30,000g at 4°C for 2 hours, and the protein concentrations of the supernatants were determined by BCA assay (Pierce). Vitreous protein samples were partially purified to remove the large fraction of the type 2 collagen that dominates vitreous protein composition but that would not contribute substantially to vitreous bioactivity. A 400-μL volume of 100 mM Tris-HCl/2 M NaCl, pH 7.4, at 4°C was added to the vitreous samples before centrifugation at 30,000g at 4°C for 2 hours. Excess salt was removed from the supernatants by repetitive cycles of dilution with PBS followed by concentration with centrifugal filter devices with a 3000 molecular weight cut-off (Microcon; Millipore, Bedford, MA). Protein concentrations of the supernatants were then determined by the BCA assay (Pierce). This process was repeated three times (total, n = 60). 

To determine the angiostatic potential of the partially purified vitreous and nonpurified retinal protein samples, the samples were injected into the left eyes of other oxygen-exposed rats (n = 8 for retina protein; n = 19 for vitreous protein). Right eyes received PBS. Injection of these rats occurred at the time of removal from oxygen (P14), and assessment of retinal angiogenesis occurred at P20. Samples were concentrated and normalized so that 5-μL injection volumes contained 0.25 μg retinal or vitreous protein. Another experiment (n = 11) involved injection of similar vitreous protein samples that were preincubated with excessive polyclonal anti-PEDF antibody (BioProducts, Middletown, MD). Antibody–antigen complexes were then precipitated and were removed before injection. 

Results

Effect of Penetrating Injury on Normal and Pathologic Vessel Growth

Retinas of oxygen-treated animals that received penetrating ocular injuries by the anterior approach at P14 were assessed at P20. No difference was seen in area of preretinal NV between these eyes and noninjured eyes (2.83 ± 0.36 vs. 2.91 ± 0.40 mm², respectively). Conversely, eyes injured using the posterior approach at P14 exhibited only 2.01 ± 0.29 mm² of retinal NV at P20 (P < 0.001). Multiple penetrating injuries resulted in a stronger antiangiogenic effect (Fig. 1) . Remarkably, rats receiving penetrating injuries in each of the four quadrants exhibited an 84% (3.1 ± 0.9 mm² for noninjured eyes versus 0.5 ± 0.3 mm² from eyes receiving 4 injuries) reduction in NV area compared with noninjured eyes (P < 0.001). 

In all experiments combined, the area of preretinal NV in eyes injured at P14 using the posterior approach was reduced by 28% (P < 0.05) compared with noninjured eyes at P20. However, there was no statistical difference in average area of retinal vascularization between injured and fellow eyes at this time (27.5 ± 2.5 mm² [76.2% of total retinal area] versus 28.1 ± 2.8 mm² [75.5% of total retinal area], respectively). Room air-raised rats exhibited no vascular abnormalities at P20 as a result of penetrating injury at P14, and injured and noninjured eyes were fully vascularized, as expected for this age. Timing of injury (P14 vs P17) did not produce a significant difference in NV area (1.93 ± 0.42 vs 2.10 ± 0.61 mm²; P > 0.1) or vascularization (28.5 ± 2.1 mm² [77.7% of total retinal area] versus 27.4 ± 2.7 mm² [74.7%]; P > 0.5). 

Nasally and temporally injured eyes were assessed for geographic distribution of retinal NV in relation to injury site. Figure 2illustrates a near-average result of a penetrating injury to the temporal pole of a left eye. Tufts of preretinal vessels are visible in the nasal, superior, and inferior quadrants. The temporal quadrant contains some abnormal vessel buds but no preretinal growth. To quantify the distribution of NV, each injected retina was schematically divided into four quadrants (superior, nasal, inferior, and temporal), and area of preretinal NV was determined for each quadrant. Figure 3shows the effect of nasal and temporal injuries on the distribution of oxygen-induced retinal NV. Eyes injured at the nasal pole showed least NV area in the injured nasal quadrant. NV area in the injured quadrant was significantly smaller than in the temporal (P < 0.025) or inferior (P < 0.05) quadrant. Similarly, temporally injured eyes had least NV in the temporal retinal quadrant, with a statistically significant difference between the temporal and inferior quadrants (P < 0.025). Moreover, each retinal quadrant of injured eyes exhibited smaller NV area than the same quadrant in contralateral, noninjured eyes, regardless of the site of penetration. In eyes injured in the nasal quadrant, reduction in NV area ranged from a value of 72.2% (P < 0.001) in the nasal quadrant to 7.1% (P > 0.1) in the opposite, temporal quadrant when compared with noninjured eyes. In eyes injured in the temporal quadrant, reduction in NV area ranged from 58.3% (P < 0.001) in the temporal quadrant to 6.1% (P > 0.1) in the opposite, nasal quadrant. Quadrants adjacent to the wound site exhibited intermediate levels of reduction in NV area. Notably, on comparison of the injured and opposite quadrants, there was an apparent asymmetry in avascular area with increased development of the intraretinal vasculature in the injured quadrant (Fig. 2) . The mean avascular area for injured temporal retinal quadrants was 1.69 ± 0.79 mm² (18.8% of the retinal quadrant) compared with 2.44 ± 0.800 mm² (27.1%) for the opposite nasal quadrants (n = 22; P < 0.03). The mean avascular area for injured nasal quadrants was 1.48 ± 0.795 mm² (16.4%) compared with 2.88 ± 1.22 mm² (32.0%) for the opposite temporal quadrants (n = 12; P < 0.025). 

Measurements of Microglial Activation and Recruitment

To determine whether the angiostatic effect of penetrating ocular injury was caused by the recruitment of activated microglia to the wound site or by the activation of resident retinal microglia near the wound site, we stained retinas using appropriate immunocytochemical markers for major histocompatability complex expression. Counts of activated (OX-6–positive) microglia showed no difference between retinal areas near sites of injury and areas from the opposite quadrant or between injured and noninjured retinas (Fig. 4) . The same was true of the microglial/macrophage complement stained by OX-42. There was, however, a significant difference between the numbers of cells exhibiting OX-6 immunoreactivity when comparing oxygen-exposed with room air-raised rats (10.8 ± 7.1 vs. 2.7 ± 5.2/mm², respectively; P < 0.05). The morphology also differed, with retinas from room air-raised rats rarely containing ramified cells, which were common in the retinas of oxygen-exposed rats. This increase in ramified microglia after oxygen exposure is under further investigation. 

Retinal Gene Expression and Protein Level

Rats were killed at 1, 3, and 6 days after oxygen and ocular injury, and retinas were harvested for RNA and protein isolation. Injured and opposite retinal quadrants from some retinas were dissected and isolated for total RNA extraction. Either total RNA or protein was analyzed by gene microarray, two-dimensional gel electrophoresis, quantitative real-time RT-PCR, and Northern and Western blotting. Genomic and proteomic surveys suggested that a number of changes in gene expression occurred as a result of penetrating ocular injury. Results from the two methods were compared, and a list of candidates was created for which mRNA or protein level was elevated or decreased. Among these candidates were at least six with a history of angiostatic potential in some biologic system. 

Expression profiles of the candidate angiostatic proteins, as measured by Western blot analysis, revealed increases of TIMP-1, -2, and -3, ²⁹ endostatin, ³⁰ plasminogen-activator inhibitor-1 (PAI-1), ³¹ and PEDF ²⁰ in retinas from injured eyes compared with control eyes (Fig. 5) . The greatest difference in retinal PEDF expression came 1 day after injury (P < 0.02). In injured retinal tissue, there was a 1.5-fold increase in the protein level and a 4.1-fold increase in mRNA level. In the vitreous, the increase in PEDF protein was 3.4-fold (Fig. 6B) . TIMP-3, endostatin, and PAI-1 also showed the greatest differences in retinal protein levels 1 day after injury (P < 0.001). Significant differences were also seen 3 days after injury for TIMP-1 (P < 0.001), TIMP-2 (P < 0.001), TIMP-3 (P < 0.001), PEDF (P < 0.02), PAI-1 (P < 0.005), and endostatin (P < 0.001). TIMP-3, PEDF, PAI-1, and endostatin demonstrated the greatest upregulation 1 day after injury, and TIMP-1 and TIMP-2 showed the greatest difference in protein levels 3 days after injury. Notably, there was an increase 1 day after injury in the proangiogenic VEGF and bFGF proteins (Figs. 7A 8) . There was a relative 1.44 increase in bFGF mRNA expression (P < 0.05) in isolated, injured, temporal retinal quadrants compared with the isolated, opposite quadrants 1 day after injury. However, there was no difference in VEGF mRNA expression between injured (temporal) and opposite quadrants 1 day after injury (Fig. 7B) . Differences were observed in the levels of several less-defined retinal messages and proteins. We are working to identify these candidates using other methods, including vitreous protein fractionation, followed by assays of fraction bioactivity and identification of individual proteins through mass spectrometry. 

In Situ Hybridization

In situ hybridization was performed on retinal tissue from the eyes of oxygen-treated rats that were injured in the nasal retinal quadrant on P14 by dry needle puncture. Eyes were harvested and processed on P15. Retinal PEDF expression showed an uneven distribution in injured eyes (Fig. 9) . The retinal quadrant containing the injury site exhibited increased PEDF message in the ganglion cell, retinal pigment epithelium, and inner nuclear layers and in the outer limiting membrane relative to that of the retinal quadrant opposite the injury. This is consistent with the findings of others ^{32 33} who have described retinal sites of PEDF expression under normal conditions and during subretinal angiogenesis. Figure 9illustrates the differential expression in injured (Figs. 9A 9D)and opposite (Figs. 9B 9E)quadrants of the same eye. Figures 9C and 9Fdisplay a negative control resulting from tissue interaction with the sense riboprobe. 

Isolation of Retinal and Vitreal Protein, Partial Purification, Antibody Blocking, and Injection

Treatment of oxygen-exposed rats with partially purified and concentrated vitreous protein from injured eyes inhibited retinal angiogenesis by 61% (P < 0.01). Preincubation of the vitreous protein samples with excess polyclonal anti-PEDF partially, but significantly (P < 0.05), abolished the angiostatic activity (Fig. 10) . Intravitreal injection of retinal protein from injured eyes inhibited retinal angiogenesis in oxygen-exposed rats by 19% (NS; not illustrated). 

Discussion

Efforts to understand and intervene in the potentially destructive and irreversible effects of retinal NV often have included intraocular injections to deliver experimental drugs. Herein we report a previously unidentified effect of intravitreal needle penetration—inhibition of retinal NV. Notably, the angiostatic effect was most prominent near the site of injury, suggesting that a diffusible factor released from the wound site was responsible. Based on this distinct pattern of angiostasis, a role for altered intraocular pressure (IOP) secondary to globe penetration appears unlikely. Intraocular injection can cause an alteration in IOP. In a 14-day-old rat, virtually any injection volume leads to transient retinal blanching that is secondary to hypertony. Breaching the globe with a 30-gauge needle, with or without injection, produces longer-lived hypotony. IOP alterations of this sort can profoundly affect retinal blood flow and can result in compensatory vascular responses. However, these changes cannot explain the present results; IOP alterations exert a pan-retinal influence, ³⁴ but the angiostatic effect observed in this study exhibited a distinctly asymmetrical pattern. Similarly, recruitment and activation of microglia appeared unaffected by the penetrating wound. Although there was a distinct difference in the number of activated retinal microglia observed in room air-raised compared with oxygen-treated rats, no difference resulted from the penetrating injury. Moreover, on balance, microglia appeared more likely to provide a proangiogenic influence. ³⁵ Finally, the angiostatic effect apparently required damage to the retina or RPE-choroid complex because our anterior approach, penetrating cornea and iris/ciliary body, did not yield similar results. These findings raise the notion that eye tissues have the capacity to manufacture angiostatic cytokines, a feature that holds tremendous therapeutic potential. 

These experimental findings bear some similarity to those of Faktorovich and colleagues ^{36 37} describing the neuroprotective effect of bFGF administration on photoreceptor cell survival in light-damaged and spontaneously degenerating rodent retinas. These studies demonstrated that dry-needle insertion into the subretinal space produced a similar effect. Moreover, greater rescue resulted from subretinal insertion than from vitreous insertion. This difference was attributed to more extensive cellular damage to RPE and neuroretina from the needle path. The pattern of protection resulting from needle insertion led the authors to conclude that a diffusible substance maintaining highest concentration at the injury site was responsible. 

Inspection of the injured and opposite retinal quadrants showed consistent asymmetry with respect to retinal avascular area, with enhanced intraretinal vascular development in the injured quadrant (Fig. 2) . Western blot analysis of protein samples from injured retinas clearly showed an upregulation of bFGF and VEGF 1 day after injury (Fig. 7) . However, bFGF mRNA was increased by approximately 40% in the injured retinal quadrant compared with the opposite noninjured quadrant, but VEGF mRNA levels were the same. Given the angiostatic effect of penetrating injury demonstrated by our experiments, it may seem paradoxical that these strong inducers of angiogenesis ³⁸ are upregulated in response to retinal injury. These observations led us to hypothesize that a local upregulation of bFGF in response to the wound might induce rapid intraretinal vascular growth, thus leading to smaller avascular retinal regions and decreasing the ischemia-induced hypoxia that drives retinal NV. VEGF may not be involved in this process because upregulation of VEGF protein appears pan-retinal. Prostaglandin E₂ (PGE₂) is abundant in retinal tissue and increases in response to injury and various pathologic conditions. ^{39 40 41} Cheng et al. ⁴² demonstrated that cultured Müller cells treated with PGE₂ respond with increased expression of bFGF and VEGF, and several studies suggest that Müller cells are an important in vivo source of retinal VEGF. ^{43 44 45 46 47} Therefore, we hypothesize that angiogenesis leading to preretinal neovascularization was briefly diverted to normal intraretinal vasculogenesis by a local bFGF upregulation. Furthermore, local bFGF upregulation and pan-retinal VEGF up-regulation, in response to penetrating ocular injury, may occur through a PGE₂-dependent mechanism. This increased local vasculogenesis might have complemented the effects of injury-induced angiostatic proteins that were longer lived. Although NV is likely driven by elevated VEGF or bFGF levels in several systems, the induction of angiogenesis compared with vasculogenesis is also likely to depend on factors such as tissue and cellular localization or isoform distribution and the presence or absence of other growth factors/cytokines. 

Our surveys pointed to a small group of factors that were clearly induced by the injury and had longstanding histories of angiostatic potential. One of these is PEDF, a factor that is produced by RPE (a site of injury in this and the Faktorovich studies ^{36 37} ) and one that, in addition to its well-known angiostatic capacity, exhibits strong neuroprotective activity. It is interesting to speculate that PEDF might have played a role in the effect reported by Faktorovich. ³⁶ However, PEDF had not been identified at the time of the Faktorovich report, ³⁶ and no probes were developed for its study. 

In our initial search for endogenous retinal antiangiogenic molecules related to penetrating injury, we focused on the contribution of PEDF because of its clear upregulation and its angiostatic potency. ²⁰ After needle penetration, retinal PEDF message and protein were increased by 4.1-fold and 1.5-fold, respectively, on P15 (Fig. 6) . Furthermore, the spatial distribution of retinal PEDF mRNA correlates with its induction in proximity to the wound site (Fig. 9) . Notably, the greatest increase in PEDF protein is seen in the vitreous, implying that PEDF is secreted from sites of retinal synthesis or is released from damaged retinal cells. That PEDF remains high in the vitreous three days after the injury argues for active secretion. Relatively free diffusion of factors through the vitreous may explain our finding of significant angiostasis in the retinal quadrants adjacent to the injury and of consistent, if not always significant, angiostasis in the opposite quadrant. The limited angiostatic effect seen in the opposite quadrant could be explained by rapid turnover of the factor(s) in the vitreous, limiting its/their activity to a region surrounding the wound site. A vitreous diffusion route may also help to explain why normal intraretinal vessel growth is not affected by penetrating injury. Perhaps once angiostatic factors are released into the vitreous their retinal levels remain relatively low, thereby limiting their bioavailability to intraretinal vascular endothelium. 

Recently, two large clinical trials used intravitreal drug administration to patients with neovascular or “wet” macular degeneration. ^{10 11} Ethical considerations limited the use of vehicle or sham injection controls. Our findings raise the possibility that a fraction of the efficacy shown by these agents actually may be attributed to the intraocular route of delivery rather than to their bioactivity. This question can only be addressed by extensive and careful dose-response experiments designed to discern the effect of therapeutic agents from that of the injection injury. Presently, retinal photocoagulation remains the most widely used method of treating subretinal and preretinal NV. The angiostatic mechanism of laser application remains to be precisely defined. Our findings raise the notion that the laser may induce an effect similar to that of penetrating injury. 

A review of current literature yielded several publications from which comparisons of OIR in vehicle-injected compared with noninjected eyes could be made. ^{12 14 15 22 48 49 50} Of these, only one report, by Mori and colleagues, ²² did not demonstrate a reduction in NV after vehicle injection. This report showed no significant difference in vehicle-injected and noninjected eyes among 12 OIR mice. The discrepancy between our present findings and those of others may be the result of the relatively small sample size used in the Mori et al. study. ²² The discrepancy cannot be attributed to species because of the six studies examined, ^{12 14 15 48 49 50} two used the same mouse model Mori et al. used. ²² Small differences in injection method might have yielded disparate effects, but this variable could not be evaluated because usually detailed descriptions of the injection method were not provided. 

Therapeutic strategies that seek to modulate the synthesis and release of PEDF or other candidate endogenous factors may provide distinct advantages over administration of exogenous agents. Therapy can be made site specific and can be less prone to the potentially toxic adverse effects of local or systemic administration of exogenous agents. We will continue to use the penetrating ocular injury model to explore the expression patterns of other known endogenous angiostatic factors, including the six identified herein, and to search for and identify novel factors that may provide future insight into the pathogenesis of retinal NV or that may harbor other therapeutic value. 

 Supported by National Eye Institute Grants EY07533 and EY08126 (JSP) and a Challenge Grant from Research to Prevent Blindness.

 Submitted for publication May 27, 2005; revised August 2, 2005; accepted November 30, 2005.

 Disclosure: J.S. Penn, None; G.W. McCollum, None; J.M. Barnett, None; X.Q. Werdich, None; K.A. Koepke, None; V.S. Rajaratnam, None

 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.

 

Dose-response curve describing the relationship between the number of ocular needle penetrations and retinal NV in eyes of oxygen-exposed rats. Injury occurred on P14, and assessment took place 6 days after oxygen exposure and after injury. Contralateral eyes served as controls.

Figure 2.

 

Representative flat-mount of an ADPase-stained retina. This retina was from an oxygen-exposed rat injured by dry-needle puncture in the temporal retinal quadrant on P14. The site of injury is indicated by the circle in the temporal quadrant. Areas of preretinal NV are indicated by the arrows. Assessment was made 6 days after oxygen exposure and after injury.

Figure 3.

 

Oxygen-exposed rats were injured with a dry needle on P14 either in the nasal or in the temporal retinal quadrant. Neovascular areas for each retinal quadrant of injured and control eyes were assessed on P20.

Figure 4.

 

Flat-mounts of rat retinas stained with mouse anti–I-A OX-6 and mouse anti–OX-42 to detect the presence of resting and activated microglial cells. Retinas shown are near the needle puncture site, approximately 0.5 mm posterior to the temporal ora serrata. (A) Oxygen-treated, non-injured. (B) Oxygen-treated retina was injured by dry-needle puncture. (C) Injured retina from a rat raised in room air. (D) Oxygen-treated, noninjured retina processed in the absence of primary antibodies. Retinas were dissected on P20.

Figure 5.

 

Postinjury expression profiles of angiostatic proteins as measured by Western blot analysis for whole retinas. Eyes from oxygen-treated rats were uninjured or were injured in the nasal retinal quadrant on P14 by dry-needle puncture. Bands of interest were quantitated by densitometric analysis. The y-axis of each graph represents the fold change in optical density. Each bar represents the mean ± SD of three independent determinations. *P < 0.02; †P < 0.005; ‡P < 0.0001.

Figure 6.

 

Retinal and vitreal PEDF levels. (A) Northern blot analysis of retinal PEDF mRNA at day 1 after injury. (B) Western blot analysis of retinal and vitreal PEDF. Oxygen-treated rats were injured by dry needle puncture in the nasal retinal quadrant on P14. Whole retina and vitreous samples were harvested for analysis on P15.

Figure 7.

 

Eyes of oxygen-treated rats received a penetrating ocular injury on P14. Retinal specimens were collected on P15. (A) Western blots comparing postinjury expression profiles of VEGF and bFGF from injured and uninjured whole retinas. (B) Relative expression of bFGF and VEGF mRNAs in injured versus opposite, uninjured retinal quadrants within injured eyes by quantitative real-time RT-PCR. Each bar represents the mean ± SEM of three independent determinations. *P < 0.05.

Figure 8.

 

Postinjury expression profile of VEGF determined by ELISA. Eyes from oxygen-treated rats were uninjured or were injured in the nasal retinal quadrant on P14 by dry-needle puncture.

Figure 9.

 

PEDF mRNA in situ hybridization of retinal tissue from an oxygen-treated rat injured in the nasal retinal quadrant on P14. The eye was enucleated and processed on P15. (A, D) Transverse sections from the injured retinal quadrant. (B, E) Transverse sections from the opposite uninjured retinal quadrant. (C, F) Transverse sections that are negative controls resulting form retinal tissue interaction with a sense riboprobe.

Figure 10.

 

Treatment of oxygen-exposed rats with partially purified and concentrated vitreous protein. Partially purified vitreous protein extracts were administered to oxygen-treated rats on P14 by intravitreal injection. Some eyes of the oxygen-treated rats were noninjected (no inj.), vehicle injected (veh.), or injected with partially purified protein extract that had been preincubated with anti-PEDF antibody. Neovascular response was assessed on P20. *P < 0.01.