Effects of Ischemic Preconditioning and Bevacizumab on Apoptosis and Vascular Permeability Following Retinal Ischemia–Reperfusion Injury

Abstract

Purpose.: Using transient ischemia followed by reperfusion (IR) to model ischemic retinal disease, this study compares the effects of ischemic preconditioning (IPC) and therapies targeting vascular endothelial growth factor (VEGF) and tumor necrosis factor (TNF)-α on retinal apoptosis, vascular permeability, and mRNA expression.

Methods.: Rats were subjected to 30 or 45 minutes of retinal ischemia followed by reperfusion for up to 48 hours. Neurodegeneration was quantified by caspase-3 (DEVDase) activity and by measuring nucleosomal DNA content (cell death ELISA). Vascular leakage was quantified by the Evans Blue dye method. A set of IR-responsive mRNAs was identified by whole-genome microarray and confirmed by RT-PCR analyses. VEGF protein was measured by Western blot analysis. IPC was accomplished with 10 minutes of ischemia 24 hours before IR. VEGF and TNFα signaling was inhibited by intravitreal injection of bevacizumab or etanercept, respectively.

Results.: IR caused significant retinal cell apoptosis and vascular permeability after 4 and 48 hours. Whereas IR decreased VegfA mRNA, VEGF protein was significantly increased. IPC effectively inhibited neurodegeneration, bevacizumab effectively inhibited vascular permeability, and etanercept failed to affect either outcome. IPC significantly altered the IR responses of 15 of 33 IR-responsive mRNAs, whereas bevacizumab had no significant effect on these mRNAs.

Conclusions.: IR provides an acute model of ischemic retinopathy that includes neurodegeneration and VEGF-dependent vascular permeability and is amenable to rapid drug therapy testing. The distinct effects of IPC and bevacizumab demonstrate that the apoptotic and vascular responses to IR may be separated and that therapeutics targeting each pathologic endpoint may be warranted in treating ischemic retinal diseases.

Retinal neovascularization and macular edema are associated with vision loss in several retinal diseases, including age-related macular degeneration (AMD) and ischemic retinopathies, such as retinal vascular occlusive diseases and diabetic retinopathy. Although the vascular components are most obvious, these retinopathies are also associated with various degrees of inflammation and neurodegeneration. For example, development of AMD is linked to complement activation and inflammation. ^1,2 Diabetic retinopathy is associated with microglial activation, increased expression of inflammatory cytokines, adherence of leukocytes to the retinal microvasculature (leukostasis), and apoptotic death of neurons. ^{3 –6} The functional interactions between inflammatory, neuronal, and vascular components of retinopathies are just now beginning to be explored. For example, inflammation has been hypothesized to contribute to the development of vascular abnormalities through leukostasis and subsequent microvascular occlusion, production of vasoactive compounds such as VEGF, and subsequent vascular permeability and vascular cell drop out. ^5,7 Vascular occlusion and vessel drop out can lead to focal retinal ischemia, which may be an initiating event for both neurodegeneration and neovascularization. ^8,9 Alternatively, adaptive responses to neurodegeneration may include the expression of neurotrophic factors that, in turn, cause vascular permeability and angiogenesis. ¹⁰  

TNFα and VEGF have been identified as therapeutic targets for treating inflammatory and ischemic retinal diseases. ^{11 –13} TNFα-targeted treatments, including etanercept and infliximab, have been applied with promising results in ocular inflammatory disease, AMD, and diabetic macular edema. ^{14 –16} In a rat endotoxin-induced uveitis model of inflammatory retinopathy, systemic administration of the TNFα inhibitor etanercept diminished leukostasis, endothelial and neuronal apoptosis, and vascular permeability. ¹⁷ Systemic etanercept suppressed retinal inflammatory markers (ICAM-1, eNOS, and NF-κB), while reducing leukostasis and vascular dysfunction in rats with short-term diabetes. ¹⁸ Systemic etanercept inhibited retinal apoptosis in a rat model of short-term diabetes and retinal vascular cell loss in a mouse model of longer-term diabetes. ¹⁹ Intravitreal injection of the TNFα inhibitor pegsunercept inhibited retinal microvascular cell death in rat models of both type 1 and type 2 diabetes. ²⁰  

VEGF contributes to the vascular angiogenesis and vascular permeability associated with many retinopathies (for a review, see Ref. 21). Its effect is supported by case studies and small clinical trials demonstrating that treatment with VEGF antagonists can alleviate edema and prevent neovascularization in ischemic and inflammatory retinopathies. ^22,23 However, there is concern that inhibition of VEGF function could cause neurodegeneration. ²⁴ VEGF acts as a potent neurotrophic factor and retinal neurons express VEGF receptors. ^25,26 Blocking VEGF function by repeated application of a soluble VEGF receptor protein or neutralizing VEGF antibody caused the loss of retinal ganglion cells in mice and rats. ²⁵ However, clinical trials and other animal experiments suggest that the risks of ocular anti-VEGF therapies are minimal. ^{27 –29}  

The intraocular pressure-induced ischemia–reperfusion (IR) model involves temporary ischemia followed by natural reperfusion, which causes an inflammatory and neurodegenerative response in the intact retina. This procedure models the neuronal damage observed in diseases with transient vessel occlusions. Most IR studies employ periods of ischemia lasting 45 to 120 minutes and examine retinal function and histology after 7 to 10 days of reperfusion (for a review, see Ref. 30). Electroretinogram (ERG) analysis reveals significant decreases in neuronal function at 1 week after IR, with reduced a- and b-wave amplitudes. ³¹ IR also induces the loss of retinal neurons indicated by decreased thicknesses of retinal layers, including the ganglion cell layer (GCL), inner nuclear layer (INL), and inner plexiform layer (IPL) as observed by histologic analysis. ^32,33 Terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling (TUNEL) of ischemic retinas demonstrates the presence of apoptotic neurons in all retinal layers. A recent study also suggested that the IR model replicates vascular abnormalities observed in the diabetic retinopathy. ³⁴ Neuronal apoptosis and loss of retinal ganglion cells occurs within 2 days of ischemia, whereas the loss of vascular cells occurs after 7 to 14 days. ³⁴ However, little is known regarding the impact of IR on vascular permeability. Ischemic preconditioning (IPC), in which a brief (8- to 10-minute) ischemic episode and recovery period precede the IR procedure, effectively prevents subsequent neurodegeneration. ^35,36 However, although the effect of IR on retinal neurodegeneration has been documented and the effectiveness of IPC in reducing neural cell damage has been shown, this model has seldom been used to address the relationship between neuronal cell death and vascular changes. 

In this study, we compared the effects of IPC and intravitreal injection of bevacizumab or etanercept on retinal vascular permeability, neurodegeneration, and mRNA expression in rat retina after intraocular pressure–induced IR. The results demonstrate that IPC, while effective at preventing IR-induced neurodegeneration, has little effect on vascular permeability. Bevacizumab prevented IR-induced vascular permeability, while having no effect on overall caspase activation or internucleosomal DNA cleavage. In contrast, blocking TNFα had little effect on either vascular permeability or neurodegeneration. A set of 33 IR-responsive mRNAs were used to examine the effects of the treatments on the gene expression response to IR, revealing a unique subset of genes that correlated with the prevention of neurodegeneration but not with vascular permeability. This study demonstrates that neurodegeneration and vascular dysfunction in response to retinal IR may be functionally separated and suggests that diseases that include an ischemic retinal response may require combination therapies protective of both vascular and neural function. 

Methods

Animal Model

Male Sprague-Dawley rats (Charles River Laboratories, Wilmington, MA) weighing 150 to 175 g (unless otherwise noted) were maintained in specific pathogen-free conditions, monitored by quarterly sentinel testing, and were treated in accordance with the guidelines of the Institutional Animal Care and Use Committee of the Penn State Hershey College of Medicine and in compliance with the ARVO Statement for the Use of Animals in Ophthalmic and Visual Research. 

Drugs

A pharmaceutical-grade formulation of humanized VEGF antibody, bevacizumab (25 mg/mL solution; Avastin; Genentech Inc., South San Francisco, CA), and a pharmaceutical-grade formulation of a dimeric fusion protein of the extracellular portion of the human 75-kDa (p75) TNF-α receptor (TNFR) linked to the Fc portion of human IgG1, etanercept (Enbrel, 50 mg/mL solution; Immunex Corp., Thousand Oaks, CA), were obtained from the Hershey Medical Center Pharmacy. Each drug was injected intravitreally (2 μL/eye) with a 32-gauge needle 48 hours before IR or sham treatment. Phosphate-buffered saline (PBS, 2 μL/eye) was injected in control eyes. 

Ischemia–Reperfusion

Deep anesthesia was induced in the rats with intramuscular injection of ketamine and xylazine (66.7 mg/kg and 6.7 mg/kg body weight, respectively). Ischemia was applied to the eye by increasing the intraocular pressure and thus cutting off the blood supply from the retinal artery. Increased pressure was achieved by introduction of sterile saline through a 32-gauge needle that was inserted into the anterior chamber of the eye through the cornea. The needle was attached by Tygon tubing linked to a syringe pump (Braintree Scientific, Braintree, MA), and the flow rate was set at 40 μL/min. The retina was monitored for blanching, indicating loss of blood flow. Intraocular pressures were measured with a rebound microtonometer designed for use on rodent eyes (TonoLab; Icare, Helsinki, Finland). Published comparison of rat intraocular pressure measurements with the microtonometer and direct micromanometry suggested that the measurements are approximately 5 mm Hg lower than actual pressures. ³⁷ The microtonometer showed pressures ranging from 9 ± 1 to 11 ± 1 mm Hg in sham eyes and from 90 ± 3 to 94 ± 4 mm Hg in eyes undergoing ischemia. Unless otherwise stated, the pressure was held for 45 minutes and then released, allowing the eye to reperfuse naturally for 48 hours. Sham-treated eyes were treated by briefly inserting a 32-gauge needle into the anterior chamber of the eye through the cornea. IPC was accomplished by performing the IR procedure with a 10-minute period of ischemia of and then allowing natural reperfusion for 24 hours before the 45-minute period of ischemia. 

Caspase-3 Activity

Caspase-3 activity was measured in retinal protein homogenates by using a fluorometric assay system (CaspACE; Promega, Madison, WI). Immediately after euthanatization, the retina was excised, carefully cleaned of all vitreous, and bisected, and a half portion was placed in 60 μL of cold lysis buffer (25 mM HEPES [pH 7.5], 5 mM MgCl₂, 5 mM EDTA, 5 mM DTT, 2 mM PMSF, 10 μg/mL leupeptin, and 1% NP40). The retina was gently sonicated followed by a 30-minute incubation at 4°C and a 20-minute centrifugation at 16,000g at 4°C. Protein content in cleared lysate was measured by Bradford protein assay (Bio-Rad, Hercules, CA). Caspase-3 activities in the supernatants were measured in a 96-well plate format, according to the manufacturer's protocol, using 50 μg retinal protein and a 60-minute incubation at 37°C. 

Apoptotic DNA Cleavage ELISA

Apoptotic DNA cleavage was assayed with an ELISA kit (Cell Death Detection; Roche Applied Science, Indianapolis, IN) and normalization to retinal wet weight. This assay measures cytoplasmic nucleosome-associated DNA fragments. Immediately after euthanatization, the retina was excised, carefully cleaned of all vitreous, and placed in a mass-tared microtube containing chilled lysis buffer supplied in the kit (200 μL/retina), which was then weighed to obtain the retinal mass. Retinal tissue was homogenized with a rotating pestle. The resulting homogenate was vortexed and incubated for 30 minutes at room temperature with gentle rocking. After centrifugation for 10 minutes at 10,000g and 4°C, the supernatant was collected and placed on ice. For each sample, duplicate 20-μL aliquots of retinal supernatant, along with positive control (provided) and negative control (lysis buffer), were subjected to DNA-fragmentation ELISA according to the kit manufacturer's instructions. After color reaction to detect captured DNA fragments, relative DNA fragmentation was expressed as optical density (light absorbance at 405 nm with a 490-nm reference wavelength) normalized by retinal mass in each aliquot of retinal supernatant. 

Terminal Deoxynucleotidyl Transferase dUTP Nick-End Labeling

Eyes were excised, fixed for 20 minutes in 2% paraformaldehyde, and frozen in OCT embedding compound (Tissue Tek; Sakura Finetek USA, Torrance, CA). Two 10-μm sections were obtained through the plane of the optic nerve and were analyzed for DNA strand breaks by the TUNEL technique (ApopTag Apoptosis Detection Kit; Millipore, Bellerica, MA) and using 3-amino-9-ethylcarbazole (AEC; Sigma-Aldrich, St. Louis, MO) to detect digoxigenin-labeled dNTPs incorporated in the DNA strands. Images of noncounterstained tissue were obtained with differential interference contrast (DIC) microscopy (BH-2 microscope with 20× objective; Olympus America, Inc. Lake Success, NY). 

Retinal Permeability

Blood–retinal barrier leakage in male Sprague-Dawley rats (200–250 g) was measured according to the method of Xu et al., ³⁸ with 2 hours of circulation time and blood drawn from the inferior vena cava at the end of the circulation to obtain the concentration of Evans blue dye in the plasma. Retinal leakage was calculated on the basis of the accumulation of dye in the retinal tissue and the plasma dye concentration and was expressed as microliters of plasma equivalent per gram dry retina weight per hour of circulation. 

RNA Isolation

Total RNA was isolated as described previously ³⁹ (Tri-Reagent/BCP; Molecular Research Center, Cincinnati, OH), followed by a purification column procedure (RNeasy; Qiagen, Valencia, CA), to further purify and remove contaminating organics. Quality and quantity were assessed on a bioanalyzer (RNA 6000 Nano LabChip with a 2100 Expert Bioanalyzer; Agilent, Palo Alto, CA). Only RNA samples with an RNA integrity number (RIN) greater than 8 were used for further analyses (based on the Agilent RIN software algorithm assigning a 1 to 10 integrity scale, with 10 being totally intact and 1 being totally degraded). 

Microarray Analysis

Microarray analysis was performed with microarrays (RatRef12; Illumina, San Diego, CA) in the Penn State College of Medicine Functional Genomics Core Facility, according to standard procedures. These arrays contained 22,519 probes corresponding to rat mRNA sequences in the NCBI RefSeq database (provided in the public domain by the National Center for Biotechnology Information, Bethesda MD, available at www.ncbi.nlm.nih.gov/locuslink/refseq/). Briefly, 250 ng of RNA was transcribed to cDNA according to the manufacturer's instructions (Ambion-Applied Biosystems, Inc. [ABI], Foster City, CA). After second-strand synthesis and purification, biotin-labeled cRNA was generated by in vitro transcription. cRNA from each sample was fragmented, denatured, and hybridized to microarray slides. The slides were incubated with Cy3-labeled streptavidin and washed. Microarrays were scanned (BeadStation scanner; Illumina), and images were imported into the allied software (BeadStudio; Illumina). Initial quality control for sample preparation, labeling, and hybridization was performed (GenomeStudio software; Illumina) with quality control probes built into the microarray design for positive and negative controls. The data were then exported (GeneSpring GX; Agilent Technologies) for subsequent statistical analyses. The genes were filtered by expression level, as described elsewhere, ³⁹ to eliminate genes that were below detection or were not expressed in the tissue. Differential expression was determined with a combination of P-value (P < 0.05, two-tailed t-test) and magnitude change (≥1.4-fold increase and ≤0.7-fold decrease), as published ³⁹ and in accordance with standards of data analysis. ⁴⁰  

Quantitative RT-PCR

Quantitative PCR analysis was performed according to a published method ³⁹ (7900HT Sequence Detection System, 384-well optical plates, with Assay-On-Demand gene-specific primers and probes, and SDS 2.2.2 software; all from ABI). The 2^−ΔΔCt analysis method ⁴¹ was used to quantify relative amounts of mRNAs, with β-actin as an endogenous control. β-Actin levels were determined to be unchanged in an absolute quantification experiment (data not shown). For a full listing of primer and probe sets see Table 1. 

Table 1.

 

IR-Responsive Transcripts Validated by qRT-PCR

Symbol	Name	Alias	AOD No.	RefSeq ID
Bbs2	Bardet-Biedl syndrome 2		Rn00586096_m1	NM_053618.1
C1S	Complement component 1, s subcomponent	r-gsp	Rn00594278_m1	NM_138900.1
Carhsp1	Calcium regulated heat stable protein 1	Crhsp24	Rn00596083_m1	NM_152790.2
Cntf	Ciliary neurotrophic factor		Rn00755092_m1	NM_013166.1
Cx3cl1	Chemokine (C-X3-C motif) ligand 1	Scyd1, Cx3c	Rn00593186_m1	NM_134455.1
Dcamkl1	Doublecortin-like kinase 1	DCLK1, Ania4, Cpg16	Rn00584294_m1	NM_053343.2
Ddit3	DNA-damage inducible transcript 3	Chop10, Gadd153, MGC124604	Rn00492098_g1	NM_024134.2
Edn2	Endothelin 2	ET-2	Rn00561135_m1	NM_012549.1
Elovl4	Elongation of very long chain fatty acids-like 4	FEN1/Elo2, SUR4/Elo3	Rn01403757_m1	XM_236476.4
Gbp2	Guanylate binding protein 2		Rn00592467_m1	NM_133624.1
Gfap	Glial fibrillary acidic protein		Rn00566603_m1	NM_017009.2
Hspb1	Heat shock protein 1	Hsp25, Hsp27	Rn00583001_g1	NM_031970.3
Igf2	Insulin-like growth factor 2	somatomedin A	Rn00580426_m1	NM_031511.1
Jak3	Janus kinase 3		Rn00563431_m1	NM_012855.1
Kcne2	Potassium voltage-gated channel, Isk-related family, member 2	Mirp1	Rn02094913_s1	NM_133603.2
Lgals3	Lectin, galactoside-binding, soluble, 3	gal-3	Rn00582910_m1	NM_031832.1
Lgals3bp	Lectin, galactoside-binding, soluble, 3 binding protein	Ppicap	Rn00478303_m1	NM_139096.1
Litaf	Lipopolysaccharide-induced TNF factor	EET-1, Pig7	Rn01424675_m1	NM_001105735.2
Mct1	Monocarboxylate transporter 1	Slc16a1	Rn00562332_m1	NM_012716.2
Nppa	Natriuretic peptide precursor A	ANF, ANP, Pnd, RATANF	Rn00561661_m1	NM_012612.2
Pcgf1	Polycomb group ring finger 1	Nspc1	Rn01425394_g1	NM_001007000.1
Prkcb1	protein kinase C, beta		Rn00562312_m1	NM_012713.2
Serping1 (C1INH)	Serine (or cysteine) peptidase inhibitor, clade G, member 1	C1Inh	Rn01485600_m1	NM_199093.1
Slc6a11 (Gat3)	Solute carrier family 6 (neurotransmitter transporter, GABA), member 11	Gabt4, Gat3	Rn00577664_m1	NM_024372.2
Spp1	Secreted phosphoprotein 1	Sialoprotein (osteopontin)	Rn00681031_m1	NM_012881.2
Stat1	Signal transducer and activator of transcription 1		Rn00583505_m1	NM_032612.3
Stat3	Signal transducer and activator of transcription 3		Rn00562562_m1	NM_012747.2
Syn1	Synapsin I		Rn00569468_m1	NM_019133.1
Syp	Synaptophysin	Syp1	Rn00561986_m1	NM_012664.1
Timp1	Tissue inhibitor of metallopeptidase 1	Timp	Rn00587558_m1	NM_053819.1
Tnfrsf12a	Tumor necrosis factor receptor superfamily, member 12a	Fn14, MGC72653	Rn00710373_m1	NM_181086.2
Vamp2	Vesicle-associated membrane protein 2	Synaptobrevin 2, RATVAMPB, RATVAMPIR, SYB, Syb2	Rn00360268_g1	NM_012663.2
Vegfa	Vascular endothelial growth factor A	VEGF164	Rn00582935_m1	NM_031836.2

 Gene names, aliases, and Applied Biosystems (Foster City, CA) qRT-PCR assay catalogue numbers (AOD No.).

Western Blot Analysis

The rat retinas were removed and the vitreous separated before snap freezing in liquid nitrogen. Western blot analysis for VEGF was performed as described elsewhere, ⁴² with 100 μg protein per sample separated on 4% to 12% Bis-Tris gels (NuPAGE; Invitrogen, Carlsbad, CA) and a mouse monoclonal anti-VEGF antibody (clone C-1, sc-7269, 1:500; Santa Cruz Biotechnology, Santa Cruz, CA). To control for protein loading, the membranes were probed with a mouse monoclonal anti-actin antibody (clone C4, Mab1501, 1:5000; Millipore). After incubation with anti-mouse secondary antibody linked to horseradish peroxidase (GE Health Care, Piscataway, NJ), the signal was detected by a Western blot analysis detection reagent (Amersham ECL; GE Health Care, Piscataway, NJ), read by a bioimaging system (Genegnome; Syngene, Frederick, MD), and quantified (ImageQuant 5.0 software; Molecular Dynamics, Sunnyvale, CA). 

Statistical Analysis

A mixed-effects, two-way analysis of variance (ANOVA) model that varied both treatment and ischemia status between animals was fit for the data from all experiments. The fixed effects were the treatment status (treatment versus control) of the animal and the ischemia status (IR versus sham) of the eye. Individual animal differences were modeled by using random effects, which accounted for the positive correlation between the eyes of the same animal. The ANOVA model was completely specified by including model parameters for all main effects and their interaction. Tests for differences between groups (e.g., treatment-IR versus control-IR) were based on t-statistics from the ANOVA model. For the experiments examining mRNA at 4 and 48 hours of reperfusion, a mixed-effects, one-way ANOVA model was fit separately for each time point, with ischemia status used as the only fixed effect and a random effect used for individual animal differences. Similarly, a t-statistic from the one-way ANOVA model was used to test the difference between the sham and ischemia groups. The outcomes modeled in this framework were permeability, caspase activity, DNA fragmentation, and mRNA expression. The mRNA data points were normalized to a reference group (control-sham group) and were log transformed before statistical analysis (SAS ver. 9.1 software; SAS, Cary, NC). 

Results

IR-Induced Retinal Apoptosis and Vascular Permeability

The neurodegenerative and vascular effects of IR were examined by inducing a relatively mild ischemic insult. The rat retinas were subjected to a 45-minute period of complete ischemia induced with elevated intraocular fluid pressure by pumping sterile saline into the anterior chamber of the eye followed by natural reperfusion of blood for periods of 4 or 48 hours. The effects of IR on retinal cell death were determined by measuring caspase-3/7 activity (Fig. 1A) and nucleosomal DNA fragmentation (Fig. 1B). Caspase activity was not significantly changed at 4 hours, but was significantly increased (24%) at 48 hours. Nucleosomal DNA fragmentation was significantly increased (5.8- and 4.1-fold) at 4 and 48 hours of reperfusion, respectively. Vascular leakage of plasma, as measured by Evans blue dye accumulation in the retina, was significantly increased (4.9- and 4.6-fold) at 4 and 48 hours after ischemia, respectively (Fig. 1C). Thus, biochemical measures confirmed that significant apoptosis was caused by 45 minutes of ischemia. Furthermore, IR caused a rapid and sustained increase in vascular permeability. TUNEL of retinal sections was used to identify the location of apoptotic cells in IR retinas after 48 hours of reperfusion. Whereas sham retinas exhibited virtually no TUNEL-positive cells in any layer, numerous positive cells were identified in clusters located predominantly within the outer nuclear layer (ONL) of the IR retinas (Fig. 2). Isolated TUNEL-positive cells were also observed within the INL at this time point. No TUNEL-positive cells were observed to be associated with the retinal vasculature. 

Figure 1.

 

Retinal IR caused neurodegeneration and vascular leakage. One eye of each animal was subjected to retinal ischemia for 45 minutes and reperfused for a specific period (4 or 48 hours) before being assayed for (A) caspase-3 activity, (B) DNA fragmentation, and (C) Evans blue dye leakage. The contralateral eye was subjected to needle puncture and served as the sham control. Results shown are the means and SE of means obtained from four to eight animals in each group. Comparisons between sham and IR eyes were calculated with the mixed-effects, two-way ANOVA model: **P < 0.01, and ***P < 0.001.

Figure 2.

 

Retinal IR caused DNA strand breaks in cells of the ONL of retinas 2 days after ischemia reperfusion. Representative differential interference contrast (DIC) microscopy images of TUNEL-stained sham (A) and IR (B) retinas. The IR eye was subjected to retinal ischemia for 45 minutes and reperfused for 48 hours before being sectioned and mounted for TUNEL after 48 hours of reperfusion. The contralateral eye was subjected to needle puncture and served as the sham control. Images are oriented with the inner limiting membrane toward the top. TUNEL-positive nuclei were detected in groups in the ONL (arrows) and occasionally in the INL (arrowhead) of the IR retina. Retinal layers were easily discernable and are labeled as: IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer; OS, photoreceptor outer segments.

Given that VEGF is necessary for vascular permeability in numerous settings, including retinal diseases, we sought to determine whether IR induces VEGF protein expression. Rat retinas were subjected to 45 minutes of ischemia followed by 4 or 48 hours of reperfusion, at which time whole retinal lysates were obtained and subjected to immunoblot analysis, to compare VEGF protein contents (Fig. 3). A 19- to 20-kDa form of VEGF (presumably VEGF-165) and a band with electrophoretic mobility corresponding to approximately 60 kDa (presumably a VEGF-165 dimer) were significantly increased by comparable extents in response to IR. Quantification of these two bands suggested that total VEGF-165 protein levels in retinas significantly increased after 4 (∼1.7-fold) and 48 (1.3-fold) hours of reperfusion. This result suggests that retinal permeability after IR is mediated by increased VEGF expression. 

Figure 3.

 

Retinal IR caused increased expression of VEGFA protein. One eye of each animal was subjected to retinal ischemia for 45 minutes and reperfused for a specific period (4 or 48 hours) before being assayed for VEGF protein content by Western blot analysis, normalized to the β-actin protein content. The contralateral eye was subjected to needle puncture and served as the sham control. The results are the means and SE of means obtained from eight animals in each group. Comparisons between sham and IR eyes were calculated by using the mixed-effects, two-way ANOVA model: *P < 0.05, and ***P < 0.001.

Identification and Validation of mRNAs Responsive to IR

Whole-genome microarrays were used to identify alterations in retinal mRNA levels caused by 45 minutes of ischemia followed by 48 hours of reperfusion (Fig. 4). Transcriptomic analysis was performed with RNA from six sham and six IR retinas. After the results were filtered to remove mRNAs that are not appreciably expressed in retina, mRNA corresponding to 7904 genes remained for statistical analysis. According to statistical analysis (P < 0.05, t-test) and magnitude change (≥1.4-fold increase or ≤0.7-fold decrease), 1099 probes were identified as differentially expressed in the IR group relative to the sham group (848 decreased and 251 increased). The full set of microarray gene expression data has been deposited in the NIH/NLM Gene Expression Omnibus ⁴³ (GEO accession number GSE20521; http://www.ncbi.nlm.nih.gov/projects/geo/ provided by the National Center for Biotechnology Information, National Institutes of Health, Bethesda, MD). The mRNA alterations were examined by using pathway-analysis software (Ingenuity Systems, Redwood CA) to detect relationships to existing data on disease states, cellular functions, and canonical pathways (see Supplementary Data). An overrepresentation of genes associated with endocrine system disorders (69 genes), metabolic disease (77 genes), and neurologic disease (195 genes) was observed (P < 0.001, Fisher exact test). The three most common cellular functions of the differentially expressed genes were cell death (207 genes), cell–cell signaling (74 genes), and cellular movement (106 genes). Phototransduction (14 genes) and complement pathways (9 genes) were the two most highly regulated canonical pathways in the database. 

Figure 4.

 

Retinal IR caused numerous alterations of mRNA expression levels. Retinal mRNA expression was analyzed by microarray profiling for differences between sham control and IR retinas (45-minute ischemia followed by 48-hour reperfusion). Results were filtered for genes detected as present, and differential expression was determined as statistically significant (t-test P < 0.05) with an increase of 1.4-fold or greater or a decrease to 0.7-fold or less. All the 1099 differentially expressed genes are plotted for each of the six animals per group. Each abundance value is scaled to provide a mean sham control value of unity.

To validate the microarray findings and examine the temporal course of gene expression changes, we performed qRT-PCR confirmation on retinal RNA samples from sham and IR experiments, using samples from both 4 and 48 hours of reperfusion. Thirty-three mRNAs (Table 1) were confirmed to exhibit IR-responsive expression by qRT-PCR analysis (Table 2). These 33 mRNAs included 17 mRNAs increased with IR treatment (C1S, Carhsp1, Cntf, Edn2, Gbp2, Gfap, Hspb1, Jak3, Lgals3, Lgals3bp, Litaf, Serping1 [C1INH], Spp1, Stat1, Stat3, Timp1, and Tnfrsf12a), and 16 mRNAs decreased with IR treatment (Bbs2, Cx3cl1, Dcamkl1, Ddit3, Elovl4, Igf2, Kcne2, Mct1, Nppa, Pcgf1, Prkcb1, Slac6a11 [Gat3], Syn, Syp, Vamp2, and VegfA). The confirmation by qRT-PCR was performed in the same set of samples used in the microarray analysis and in samples from additional independent IR experiments employing 4- and 48-hour reperfusion times. All 33 mRNAs examined demonstrated statistically significant changes in retinal content in at least three of four of the independent experiments with 48 hours of reperfusion (Tables 2, 3; Supplementary Data). These genes exhibited changes that were lesser in magnitude and less consistent after 4 hours of reperfusion. For example, in the experiment shown in Table 2, only five genes (Hsp1, Lgals3, Mct1, Syn, and Tnfrsf12a) were significantly altered by IR after 4 hours of reperfusion. In a second experiment, nine genes were significantly altered by IR at this time point (Supplementary Data). It should be noted that VegfA mRNA exhibited a consistent decrease in abundance in response to IR (Tables 2, 3; Supplementary Data), suggesting that the observed increase in VEGF protein content is due to posttranscriptional regulation. 

Table 2.

 

Quantitative RT-PCR Validation of IR-Responsive mRNA Expression Changes at 4 and 48 Hours of Reperfusion

Gene	4 h Reperfusion			48 h Reperfusion
	Sham 4 h (S4)	Ischemia 4 h (IR4)	IR4 vs. S4	Sham 48 h (S48)	Ischemia 48 h (IR48)	IR48 vs. S48
Bbs2	1.00 ± 0.05	0.81 ± 0.03	↔	1.00 ± 0.08	0.50 ± 0.04	↓**
C1S	1.00 ± 0.30	0.88 ± 0.14	↔	1.00 ± 0.17	1.91 ± 0.19	↑*
Carhsp1	1.00 ± 0.16	1.49 ± 0.20	↑	1.00 ± 0.09	2.94 ± 0.23	↑↑***
Cntf	1.00 ± 0.09	1.15 ± 0.05	↔	1.00 ± 0.19	2.75 ± 0.10	↑↑**
Cx3cl1	1.00 ± 0.17	0.75 ± 0.07	↔	1.00 ± 0.11	0.45 ± 0.04	↓↓**
Dcamkl1	1.00 ± 0.08	0.90 ± 0.08	↔	1.00 ± 0.16	0.58 ± 0.03	↓*
Ddit3	1.00 ± 0.27	0.52 ± 0.04	↓	1.00 ± 0.15	0.12 ± 0.02	↓↓↓↓***
Edn2	1.00 ± 0.21	1.47 ± 0.31	↑	1.00 ± 0.64	3.50 ± 0.67	↑↑*
Elovl4	1.00 ± 0.11	0.88 ± 0.09	↔	1.00 ± 0.10	0.39 ± 0.03	↓↓***
Gbp2	1.00 ± 0.56	1.50 ± 0.40	↑	1.00 ± 0.14	2.48 ± 0.16	↑↑**
Gfap	1.00 ± 0.21	0.71 ± 0.09	↔	1.00 ± 0.21	5.49 ± 0.34	↑↑↑**
Hspb1	1.00 ± 0.25	2.21 ± 0.31	↑↑*	1.00 ± 0.21	3.42 ± 0.98	↑↑*
Igf2	1.00 ± 0.07	0.89 ± 0.03	↔	1.00 ± 0.08	0.26 ± 0.06	↓↓***
Jak3	1.00 ± 0.13	1.12 ± 0.17	↔	1.00 ± 0.17	3.32 ± 0.22	↑↑**
Kcne2	1.00 ± 0.18	1.06 ± 0.15	↔	1.00 ± 0.27	0.71 ± 0.12	↔
Lgals3	1.00 ± 0.17	2.91 ± 0.56	↑↑**	1.00 ± 0.41	12.79 ± 2.20	↑↑↑↑**
Lgals3bp	1.00 ± 0.08	0.64 ± 0.09	↓	1.00 ± 0.16	2.14 ± 0.27	↑↑**
Litaf	1.00 ± 0.13	1.32 ± 0.26	↔	1.00 ± 0.12	3.14 ± 0.26	↑↑***
Mct1	1.00 ± 0.15	0.74 ± 0.04	↔**	1.00 ± 0.11	0.49 ± 0.06	↓↓**
Nppa	1.00 ± 0.06	0.69 ± 0.04	↓	1.00 ± 0.31	0.28 ± 0.03	↓↓*
Pcgf1	1.00 ± 0.17	0.83 ± 0.02	↔	1.00 ± 0.17	0.51 ± 0.06	↓*
Prkcb1	1.00 ± 0.30	0.94 ± 0.36	↔	1.00 ± 0.28	0.48 ± 0.16	↓↓*
Serping1	1.00 ± 0.22	0.63 ± 0.08	↓	1.00 ± 0.21	1.86 ± 0.19	↑*
Slc6a11	1.00 ± 0.14	0.65 ± 0.064	↓	1.00 ± 0.22	0.53 ± 0.08	↓
Spp1	1.00 ± 0.26	1.09 ± 0.24	↔	1.00 ± 0.28	3.02 ± 0.78	↑↑*
Stat1	1.00 ± 0.11	0.82 ± 0.12	↔	1.00 ± 0.11	1.63 ± 0.19	↑*
Stat3	1.00 ± 0.08	1.18 ± 0.08	↔	1.00 ± 0.20	1.73 ± 0.19	↑*
Syn	1.00 ± 0.10	0.76 ± 0.05	↔*	1.00 ± 0.25	0.39 ± 0.03	↓↓*
Syp	1.00 ± 0.11	0.79 ± 0.05	↔	1.00 ± 0.06	0.61 ± 0.05	↓**
Timp1	1.00 ± 0.55	1.25 ± 0.08	↔	1.00 ± 0.24	7.82 ± 1.14	↑↑↑**
Tnfrsf12a	1.00 ± 0.44	2.79 ± 0.32	↑↑*	1.00 ± 0.13	5.54 ± 0.80	↑↑↑**
Vamp2	1.00 ± 0.12	0.80 ± 0.04	↔	1.00 ± 0.21	0.36 ± 0.03	↓↓**
VegfA	1.00 ± 0.16	0.84 ± 0.06	↔	1.00 ± 0.10	0.44 ± 0.05	↓↓**

 Mean retinal mRNA levels were compared between IR and sham (S) retinas after 4 or 48 hours of reperfusion. Numerical data are expressed as the mean ± SEM. Arrows indicate the direction and magnitude of changes in mRNA level: ↔, 0.7 < IR/S < 1.4; ↑, 1.4 ≤ IR/S < 2.0; ↑↑, 2.0 ≤ IR/S < 4.0; ↑↑↑, 4.0 ≤ IR/S < 8.0; ↑↑↑↑, 8.0 ≤ IR/S < 16.0; ↓, 0.7 ≥ IR/S > 0.5; ↓↓, 0.5 ≥ IR/S > 0.25; ↓↓↓, 0.25 ≥ IR/S > 0.125; and ↓↓↓↓, 0.125 ≥ IR/S > 0.0625. Statistical comparisons between IR and S eyes were calculated by using mixed-effects, one-way ANOVA model:

*   P < 0.05,

**   P < 0.01, and

***   P < 0.001. Additional data are presented in the Supplementary Data.

Table 3.

 

Effect of IPC and Bevacizumab on IR-Responsive mRNA Expression

Gene	IPC				Bevacizumab
	CIR vs. CS	TS vs. CS	TIR vs. TS	TIR vs. CIR	CIR vs. CS	TS vs. CS	TIR vs. TS	TIR vs. CIR
Bbs2	↓***	↔	↔***	↔**	↔***	↔	↔	↔
C1S	↑↑***	↔	↑↑***	↔	↑↑***	↔	↑*	↔
Carhsp1	↑↑***	↔	↑↑***	↔	↑↑**	↔	↑*	↔
Cntf	↑↑***	↔	↑↑***	↔*	↑↑***	↔	↑	↓
Cx3cl1	↓↓***	↔	↓***	↑**	↓***	↔	↔*	↔
Dcamkl1	↓***	↔	↓***	↔	↔*	↔	↔	↔
Ddit3	↓↓***	↔	↓***	↔	↔**	↔	↔	↔
Edn2	↑↑***	↔	↑↑*	↔	↑↑↑***	↑↑	↑↑*	↔
Elovl4	↓***	↔	↓**	↔*	↓**	↔	↔	↔
Gbp2	↑↑***	↔	↑↑***	↔	↑↑**	↔	↑*	↔
Gfap	↑↑↑***	↔	↑↑↑***	↔	↑↑↑**	↔	↑↑	↔
Hspb1	↑↑***	↔	↑*	↔	↑*	↔	↑	↔
Igf2	↓↓***	↔	↓***	↑↑**	↓↓**	↓*	↔	↑
Jak3	↑↑**	↔	↑↑**	↔	↑↑***	↔	↑↑**	↔
Kcne2	↓↓***	↔	↓***	↔	↓**	↓*	↔	↔
Lgals3	↑↑↑↑***	↓	↑↑↑***	↓*	↑↑↑↑**	↑↑	↔	↓↓
Lgals3bp	↑↑***	↔	↑↑***	↔	↑**	↔	↑**	↔
Litaf	↑↑***	↔	↑↑**	↔	↑↑***	↔	↑↑***	↔
Mct1	↓↓***	↔	↓***	↔	↓**	↔	↔	↔
Nppa	↓***	↔	↔***	↔	↓**	↔	↔	↔
Pcgf1	↓***	↔	↓***	↔**	↓***	↔	↓**	↔
Prkcb1	↓↓***	↔	↓***	↑**	↓**	↔	↔	↔
Serping1	↑↑**	↔	↑↑***	↔	↑**	↔	↔	↔
Slc6a11	↓↓***	↔	↓***	↑**	↓*	↓*	↓	↔
Spp1	↑↑***	↔	↑	↓*	↑*	↔	↔	↓
Stat1	↑*	↔	↑**	↔	↑**	↔	↔	↔
Stat3	↑**	↔	↑**	↔	↑**	↔	↔	↔
Syn	↓***	↔	↓***	↔*	↓***	↔	↓	↔
Syp	↓***	↔	↔***	↔**	↓*	↔	↔	↔
Timp1	↑↑***	↓	↑↑***	↔	↑**	↓	↑↑*	↔
Tnfrsf12a	↑↑***	↔	↑↑***	↔*	↑↑***	↔	↑↑*	↔
Vamp2	↓***	↔	↓***	↔**	↓**	↔	↓**	↔
VegfA	↓↓***	↔	↓***	↑**	↓*	↓	↔	↔

 Mean retinal mRNA levels were compared between groups including control (not pretreated) IR with control sham (CIR vs. CS), treated sham with control sham (TS vs. CS), treated IR with treated sham (TIR vs. TS), and treated IR with control IR (TIR vs. CIR). Arrows indicate the direction and magnitude of the respective ratios of mean values for mRNA levels, and asterisks indicate the significance of differences between the mean values in the groups: ↔, 0.7 < ratio < 1.4; ↑, 1.4 ≤ ratio < 2.0; ↑↑, 2.0 ≤ ratio < 4.0; ↑↑↑, 4.0 ≤ ratio < 8.0; ↑↑↑↑, 8.0 ≤ ratio < 16.0; ↓, 0.7 ≥ ratio > 0.5; ↓↓, 0.5 ≥ ratio > 0.25; ↓↓↓, 0.25 ≥ ratio > 0.125; and ↓↓↓↓, 0.125 ≥ ratio > 0.0625. Statistical comparisons between groups were calculated by using the mixed-effects, two-way ANOVA model:

*   P < 0.05,

**   P < 0.01, and

***   P < 0.001. Numerical data are presented in the Supplementary Data.

Effect of Ischemic Preconditioning on Neurodegeneration and Permeability

The protective effects of IPC were examined by applying a 10-minute ischemic insult followed by 24 hours of reperfusion before IR. IPC diminished caspase activation from a 27% increase to a 20% increase 48 hours after the 45-minute ischemic event (Fig. 5A). However, the difference between mean caspase activity in retinas subjected to IR with and without preconditioning did not reach statistical significance. In contrast IPC, significantly reduced the increase in DNA fragmentation after IR from 3.0- to 1.7-fold, (Fig. 5B). The difference between mean DNA fragmentation measures for IR retinas with and without IPC was highly significant. To confirm the protective effects on DNA fragmentation, the experiment was repeated with a 30-minute ischemic insult. In this case, IPC completely abrogated the effects of IR on DNA fragmentation (Fig. 5C). IPC diminished the increase in vascular leakage caused by 45 minutes of ischemia from 4.9-fold to 1.9-fold (Fig. 5D). However, this reduction was mainly due to the increased basal permeability caused by IPC itself, and there was no significant difference in Evans blue dye leakage into IR retinas, with and without IPC. Thus, IPC itself caused some permeability, but did not prevent a further increase in permeability in response to IR. When the analysis was repeated with a 30-minute ischemic insult, IPC again increased basal permeability, but had no significant effect on dye accumulation in IR retinas (Fig. 5E). In fact, the mean leakage in IPC-IR retinas was greater than that in control-IR retinas. 

Figure 5.

 

IPC diminished neurodegeneration but not vascular leakage in response to IR. For preconditioning, both eyes were subjected to 10 minutes of retinal ischemia. After 24 hours of ischemia, one eye of each animal was subjected to retinal ischemia for 30 or 45 minutes and reperfused for 48 hours before being assayed for (A) caspase-3 activity with 45 minutes ischemia, (B) DNA fragmentation with 45-minute ischemia, (C) DNA fragmentation with 30-minute ischemia, (D) Evans blue dye leakage with 45-minute ischemia, and (E) Evans blue dye leakage with 30-minute ischemia. The contralateral eye was subjected to needle puncture and served as sham control. Results shown are the means and SE of means obtained from eight animals in each group. Comparisons between sham and IR eyes and between control-IR and IPC-IR eyes were calculated using the mixed-effects, two-way ANOVA model: *P < 0.05, **P < 0.01, and ***P < 0.001.

Bevacizumab's Effect on Permeability and Apoptosis

To determine whether VEGF function plays a neuroprotective role and/or contribute to vascular leakage after IR, bevacizumab (Avastin; Genentech), was intravitreally injected 48 hours before IR. This treatment had virtually no effect on caspase activation after 45 minutes of ischemia and 48 hours of reperfusion (Fig. 6A). Likewise, bevacizumab had no significant effect on internucleosomal DNA fragmentation after IR (Fig. 6B). It should be noted that the treatment had no effect, positive or negative, on basal or IR-induced retinal cell death. In contrast, vascular leakage was significantly decreased by bevacizumab from a 3.1-fold increase with IR in vehicle-treated eyes to 1.9-fold increase with IR in bevacizumab-treated eyes (Fig. 6C). 

Figure 6.

 

Intravitreal bevacizumab injection inhibited vascular leakage but not overall apoptosis in response to IR. Both eyes of each animal in the treatment group were injected with 50 mg bevacizumab, and the eyes of the control animals were injected with the same volume (2 μL) of saline. After 48 hours of ischemia, one eye of each animal was subjected to retinal ischemia for 45 minutes and reperfused for 48 hours before assay to determine (A) caspase-3 activity, (B) DNA fragmentation, and (C) Evans blue dye leakage. The contralateral eye was subjected to needle puncture and served as the sham control. Results shown are the means and SE of means obtained from eight animals in each group. Comparisons between sham and IR eyes and between control-IR and treatment-IR eyes were calculated using the mixed effects ANOVA model: *P < 0.05, **P < 0.01, and ***P < 0.001.

Etanercept's Effect on Neurodegeneration and Vascular Permeability after IR

Similarly, intravitreal injection of etanercept was used to determine whether TNFα function contributes to the retinal response to IR. Etanercept had no significant effects on caspase induction, DNA fragmentation, or vascular permeability (Fig. 7). The increase in retinal permeability caused by IR was slightly greater in etanercept-injected eyes (9.2-fold versus 6.3-fold in vehicle-treated eyes). However, the difference between mean vascular leakage in vehicle-treated IR retinas and etanercept-treated IR retinas was not significant. 

Figure 7.

 

Intravitreal etanercept injection had no effects on neurodegeneration and vascular leakage in response to IR. Both eyes of each animal in the treatment group were injected with 100 mg of etanercept, and the eyes of the control animals were injected with the same volume (2 μL) of saline. Forty-eight hours later, one eye of each animal was subjected to retinal ischemia for 45 minutes and reperfused for 48 hours before assay to determine (A) caspase-3 activity, (B) DNA fragmentation, and (C) Evans blue dye leakage. The contralateral eye was subjected to needle puncture and served as the sham control. Results are the mean and SE of means obtained from eight animals in each group. Comparisons between sham and IR eyes and between control-IR and treatment-IR were calculated using the mixed-effects, two-way ANOVA model: *P < 0.05, **P < 0.01, and ***P < 0.001.

Effects of IPC and Bevacizumab on the Expression of IR-Responsive mRNAs

The effects of IPC and bevacizumab on mRNA expression after IR were examined, to identify transcriptomic responses to IR that may be indicative of neurodegeneration or vascular function (Table 3; Supplementary Data). IPC, which had a marked effect on neurodegeneration after IR, significantly altered the responses of 15 of the 33 IR-responsive mRNAs (Table 3, TIR versus CIR columns). These included the genes Bbs2, Cntf, Cx3cl1, Elovl4, Igf2, Lgals3, Pcgf1, Prkcb1, Slc6a11, Spp1, Syn, Syp, Tnfrsf12a, Vamp2, and VegfA. The mRNA expression of 11 of these was decreased by IR, with IPC resulting in a smaller decrease in their expression in response to IR. It is also noteworthy that IPC alone had no significant effects on the basal expression of the 33 IR-responsive mRNAs in sham-treated retinas at 48 hours after the ischemia procedure, which corresponds to 72 hours after IPC (Table 3, TS versus CS columns). Bevacizumab, which inhibited the vascular permeability response without affecting neurodegeneration, did not significantly change the effect of IR on expression of any of the 33 mRNAs. Although the magnitude of IR effects on Cntf, Igf2, Lgal3, and Spp1 mRNAs were altered by bevacizumab, none of these changes were statistically significant (Table 3, TIR versus CIR columns). Bevacizumab treatment alone significantly altered the basal expression of 3 of the 33 IR-responsive mRNAs, including Igf2, Kcne2, and Slc6a11 (Table 3, TS versus CS column). Thus, IPC had a relatively pronounced effect on the expression of these mRNAs in response to IR without significantly affecting the basal mRNA levels, whereas bevacizumab affected the basal expression of a subset of these mRNAs without significantly interfering with the mRNA responses to IR. Given that IPC significantly diminished apoptosis after IR and bevacizumab significantly affected only the Evan's blue leakage, the 33 IR-responsive mRNA set developed seems indicative of the neurodegenerative response and not the vascular permeability response to IR. 

Discussion

In the present study, the retina responded to a relatively brief ischemic event with both neurodegeneration and vascular permeability. Further, these physiologic outcomes could be largely separated according to specific pretreatment. The effects of IR on retinal vascular permeability have been examined in only one other study. Using magnetic resonance imaging, Wilson et al. ⁴⁴ demonstrated that IR produced vascular leakage in rabbit retinas that slowly declined between 1 and 8 days of reperfusion. To our knowledge, no study has been undertaken to examine the effects of preconditioning on retinal vascular permeability in IR. In the present study, IPC was effective in preventing neurodegeneration, but had little effect on permeability after IR. Conversely, blocking VEGF function with bevacizumab inhibited the vascular response without affecting measures of apoptosis. Thus, intraocular pressure–induced IR can serve as a convenient in vivo model of VEGF-mediated retinal vascular permeability. Of importance, these results demonstrate that the neurodegenerative and vascular responses to IR are not functionally linked. In further support of this conclusion is the observation that IPC and bevacizumab treatment had very different effects on the gene expression responses to IR. 

Ocular pressure–induced retinal IR injury can serve as a useful model of the damage and responses that occur on a smaller scale during focal ischemic events caused by vessel drop out or vessel occlusion during diseases such as diabetic retinopathy. The model has been widely used for the study of retinal neurodegeneration and the study of IPC. ³⁰ Some investigators have used ERG and the examination of retinal layer thinning as endpoints of neuronal damage. ^{31 –33} In the present study, we used caspase-3 activity and DNA fragmentation as indicators of cell death. Although both increased after IR, these measures did not correlate perfectly. For example, whereas DNA fragmentation was similarly increased after both 4 and 48 hours of reperfusion, a significant increase in caspase-3 activity was not observed at the earlier time point. We observed a very significant increase in caspase-3 activity at 24 hours (data not shown), but the maximum increase in activity seemed to occur at 48 hours. Thus, although both assays are interpreted as indicators of cell death, the temporal nature or sensitivities of these two endpoints differ. 

The cell death observed 2 days after IR where largely restricted to the ONL. Whereas IR causes the eventual thinning of the IPL, TUNEL staining indicated that death occurred mainly in the ONL at 48 hours. Nishijima et al. ²⁵ found that death occurs in the GCL and INL soon after IR and then migrates to the ONL (with a virtually identical staining pattern as shown in our Fig. 2). It is therefore possible that the death of neurons in the inner layers is characterized by DNA fragmentation but not caspase-3 activation, whereas appreciable caspase-3 activity coincides with the death of photoreceptors. Singh et al. ⁴⁵ suggested that this is indeed the case. They observed that active caspase-3 staining became evident in the INL and was extensive in the ONL after IR, whereas early ganglion cell death was associated with active caspase-2, but not caspase-3, staining. 

A set of 33 mRNAs were identified and validated as being significantly and reproducibly altered by IR. The mRNAs were not chosen as representatives of specific biological processes, but rather as being significantly and reproducibly altered by IR. Kamphuis et al. ⁴⁶ also used whole genome arrays to examine the genetic response to IR in rat retinas, as well as the effects of IPC on these responses. Their analysis identified several of the same IR-responsive genes described herein, including C1s, Carhsp1, Dcamkl1, Edn2, Gfap, Gbp2, Hspb1, Lgals3, Litaf, Nppa, Serping1, Spp1, Stat1, Stat3, and VegfA. Hspb1 was also identified as being responsive to IPC in the rat brain. ⁴⁷ In several additional studies Gfap was identified as being upregulated during retinal IR ^48,49 as well as cerebral IR. ⁵⁰ In the retina, Gfap expression is indicative of Müller cell astrogliosis. ⁵¹  

Several previous studies of gene expression changes after IR have focused on inflammatory cytokines. ^{52 –54} Approximately one fourth of the mRNAs identified in the present analysis are closely associated with inflammation, including complement component C1s, the chemokine Cx3cl1 (fractalkine), Lgals3 (galectin-3), its associated binding protein Lgal3bp, Litaf, and members of the JAK/STAT pathway (Jak3, Stat1, and Stat3). In a recent proteomics study, the galectin-3 protein was identified as being upregulated in rat retinas at 2 days after a 2-hour ischemic insult. ⁵⁵ Few cytokines traditionally associated with inflammation were found in the array analysis or were included in the present set of 33 mRNAs altered by IR, and expression of Cx3cl1 was actually decreased at both 4 and 48 hours after ischemia. Because fractalkine is expressed by neurons, the reduction in Cx3cl1 may be an indication of neurodegeneration rather than an inflammatory condition. In contrast, Zheng et al. ³⁴ demonstrated increased expression of several inflammatory markers, including TNFα mRNA, after 2 and 7 days of reperfusion in the rat retina. Using GFAP-promoter expression of a dominant-negative IκB mutant transgene, Dvoriantchikova et al. ⁵³ demonstrated that upregulation of TNFα, as well as that of many other inflammatory genes during IR, is dependent on NF-κB activity in glial cells. Furthermore, histologic observation 1 week after ischemia showed that blocking inflammatory gene expression is highly neuroprotective. However, it should be noted that both these groups used periods of ischemia longer than we used. Furthermore, we did not specifically examine the expression of inflammatory markers and therefore cannot speculate on their role in neurodegeneration or vascular permeability after a relatively mild ischemic insult. 

In the present study, TNFα mRNA was not identified as being upregulated after IR, and we detected no significant difference in the TNFα protein level in whole retinal lysates at 4 or 24 hours of reperfusion after 45 minutes of ischemia (data not shown). Using immunoblot analysis with antibody to human IgG, we were able to detect etanercept in retinas at 4 and 48 hours after intravitreal injection (data not shown). The lack of effect of etanercept in the present study could be due to the lesser severity of the ischemic insult used, or it may indicate that TNFα has no direct role in neurodegeneration or vascular permeability during the first 48 hours of reperfusion after ischemia. Three studies have shown that TNFα expression is increased after more extended retinal ischemia. ^34,56,57 Vinores et al. ⁵⁸ used knockout mice to show that TNFα is essential for leukostasis during oxygen-induced retinopathy and after intravitreal injection of VEGF, IL-1β, and platelet-activating factor. However, they also found that TNFα was not necessary for vascular permeability in response to these factors. The role of TNFα in retinal neurodegeneration may be complex. Berger et al. ⁵⁷ demonstrated a causal role for TNFα in IR injury in mice when they found that TNF receptor knockout alleviated the detrimental effects of IR on ERG amplitudes. In contrast, Fontaine et al. ⁵⁹ found that TNFα-knockout mice demonstrated no significant histologic differences after IR. An interesting finding in studies of mice deficient in TNFR1 and TNFR2 established that TNFR1 plays a neuroprotective role, whereas TNFR2 promotes neurodegeneration after IR. Berger et al. ⁵⁷ also found that intravitreal injection of recombinant TNFα protein exacerbated the detrimental effects of IR on ERG amplitudes in rats, whereas intravitreal injection of neutralizing antibody to TNFα alleviated this response. They also demonstrated that intravitreal injection of a TNF-blocking antibody significantly decreased the number of TUNEL-positive cells in the retina at 24 hours after a 45-minute ischemic insult. Thus, although other studies have identified a role for TNFα in IR response, in the present study no change in TNFα expression or effect of TNFα inhibitor was observed. The reasons for these discrepancies are not presently known. 

IR caused significant increases in VEGF protein in whole retinal lysates. Surprisingly, VegfA mRNA was identified as an IR-responsive mRNA with decreased expression at 4 and 48 hours of reperfusion. Although VEGF expression is known to increase in animal models of ischemic retinopathy, such as oxygen-induced retinopathy ⁶⁰ and diabetic retinopathy, ⁴ the expression of VEGF during IR is not well characterized. VEGF protein levels are increased approximately twofold in guinea pig retinas 7 days after 90 minutes of ischemia, ⁵⁶ and brain VEGF mRNA and protein expression are increased after cerebral IR in rats. ^61,62 However, changes in VEGF may be restricted to focal regions. Using immunohistochemical analysis in a rat intraocular pressure–induced IR model, Ogata et al. ⁶³ showed increased VEGF antibody binding on vascular cells, despite a lack of overall increase in the entire retina. During sustained cerebral ischemia, VEGF protein was confined to microglial cells in the penumbra of the infarct region. ⁶⁴ Thus, although global VEGF transcription may decrease during IR, focal expression of VEGF protein may still contribute an important role in the retinal response. Alternatively, VEGF can be regulated at the translational level under ischemic conditions due to internal ribosome entry sites in the mRNA ⁶⁵ and the change in VEGF protein observed herein may result from such posttranscriptional regulation. The present demonstration that bevacizumab significantly blocked the increase in vascular permeability after IR supports a role for VEGF protein in this process. The results also suggest that the IR model may serve as a useful tool for studying VEGF-dependent retinal edema. 

Surprisingly little is known regarding the role of VEGF in the physiological response to IR in the retina. Inhibiting VEGF signaling has been shown to prevent permeability in cerebral models of IR. Blockade of the KDR VEGF receptor kinase inhibits edema after cerebral IR, ⁶⁶ and soluble VEGF receptor reduces edema as well as lesion volume in a mouse model of cerebral IR. ⁶⁷ A potential neuroprotective effect of VEGF has also been observed. Using a rat model of IR induced by optic sheath occlusion, Nishijima et al. ²⁵ demonstrated an increase in VEGF protein expression after 3 and 6 hours of reperfusion. They also found that intravitreal injection of recombinant VEGF protein or a VEGF receptor agonist prevented neuronal death after IR. Furthermore, the neuroprotective effects of IPC have been attributed to VEGF expression. Blocking VEGF with a soluble VEGF receptor or VEGF-neutralizing antibodies diminishes the protective effects of IPC in this same model. ²⁵ Likewise, adenoassociated virus–mediated expression of VEGF prevents neuronal damage in cerebral IR, ⁶⁸ and antisense targeting of VEGF or VEGF receptor diminishes the protection of IPC in a cerebral IR model. ²⁸ The present results show that inhibition of VEGF function during ischemia and after reperfusion does not exacerbate overall retinal cell death. Although this conclusion is consistent with the hypothesis that VEGF does not provide an essential neurotrophic function during IR, it cannot be confirmed without specifically examining the death of retinal neurons. 

Anti-VEGF therapies are increasingly used for the treatment of retinopathy. ^69,70 In recent trials, intravitreal ranibizumab and bevacizumab have each significantly improved vision in patients with diabetic macular edema—better than laser treatments. ^71,72 Likewise, preliminary results suggest that both ranibizumab and bevacizumab can improve macular edema caused by branch or central retinal vein occlusions. ^{73 –77} There is concern that inhibition of VEGF function could cause neurodegeneration by blocking its neurotrophic effects. Chronic application of VEGF antagonists caused the loss of retinal ganglion cells in mice and rats. ²⁵ However, expression of a transgene encoding a soluble VEGF receptor in the retina blocks VEGF-induced permeability and choroidal neovascularization, but does not significantly alter ERGs or retinal ganglion cell loss. ²⁹ In another study, the results showed that repeated intravitreal injection of bevacizumab has no detrimental effects in rats. ²⁸ In addition, clinical trials of bevacizumab have not revealed any detrimental effects on human retinal function. ²⁷  

The present results demonstrate that VEGF function contributes to vascular permeability after retinal IR and thus suggest that the IR model can serve as a useful tool in the study of VEGF-induced retinal edema. The observation that bevacizumab inhibits vascular permeability without increasing overall apoptosis after retinal IR further supports the conclusion that this treatment is not necessarily neurodegenerative. This result is highly relevant to concerns that treatments blocking VEGF function may cause neuronal degeneration by blocking neurotrophic effects. Separable effects of IPC and bevacizumab on neurodegeneration and vascular permeability demonstrate that these are independent responses to IR. If this independent response is true of other retinal diseases that include vascular dysfunction and neuronal death, such as diabetic retinopathy, then complementary therapeutics targeting both permeability and neurodegeneration may be needed to restore normal function. The IR model may prove useful for preclinical testing of treatments directed toward each of these pathologic endpoints. 

Supplementary Materials

Supplementary Data - (.xls) 

Footnotes

 Supported by a grant from the Juvenile Diabetes Research Foundation (SFA, DAA, Co-Principal Investigators) that is part of a JDRF Diabetic Retinopathy Center Grant (Thomas W. Gardner, MD, Director).

Footnotes

 Disclosure: S.F. Abcouwer, None; C. Lin, None; E.B. Wolpert, None; S. Shanmugam, None; E.W. Schaefer, None; W.M. Freeman, None; A.J. Barber, None; D.A. Antonetti, None

The authors thank Georgina V. Bixler and Robert M. Brucklacher (Penn State College of Medicine Functional Genomics Core Facility; Willard M. Freeman, PhD, Director) for excellent technical contributions; Wendy Dunton and Melissa Bridi (Penn State Hershey Diabetic Animal Models Core Facility; Sarah K. Bronson, PhD, Director) for aid with animal acquisition and husbandry; and Wade Edris for skilled technical assistance with retinal sectioning and staining.