Retinal Gene Expression after Central Retinal Artery Ligation: Effects of Ischemia and Reperfusion

Abstract

Purpose.: To investigate the morphologic and molecular consequences of 30- and 90-minute central retinal artery ligation (CRAL)–induced retinal ischemia, followed by 3 and 12 hours of reperfusion, and to identify potential targets for therapy.

Methods.: Retinal ischemia was induced for 30 and 90 minutes by ligating the rat central retinal artery, and corresponding effects were examined histologically, immunocytochemically, and molecularly at 3 hours and 12 hours of reperfusion. Patterns of gene expression revealed significantly upregulated and downregulated genes by gene array analyses and were verified by quantitative RT-PCR. Functional pathways were correlated using gene set enrichment analysis.

Results.: Substantial morphologic changes occurred from 3 hours to 7 days after CRAL in rats, resulting in a cellular loss in most retinal layers, particularly in inner nuclear and ganglion cell layers. After 30 minutes of CRAL and 3 hours of reperfusion, transcription-related genes such as ATF3, ID2, Klf4, BTG2, c-Fos, and c-Jun were activated. After 12 hours of reperfusion, the genes associated with kinase and caspase molecular pathways—including MAP kinases, Casp3 and Casp9—were upregulated. CRAL of 90 minutes and 3 hours of reperfusion induced glycolysis and gluconeogenesis-related genes such as G6PC. However, prolonged reperfusion of 12 hours was characterized by prominent activation of apoptosis-related genes, including Tp53, Akt1, Akt3, Pik3R1, Prkcb1, caspases (Casp3, Casp7, Casp9), and TNF.

Conclusions.: CRAL is a clinically relevant retinal ischemia model, and gene expression analysis can provide information regarding the molecular mechanisms underlying the pathophysiological processes during retinal ischemia. In addition, CRAL represents an effective experimental model with which to study retinal inflammation, development, aging, and, neurodegeneration.

Central retinal artery occlusion (CRAO) is a type of stroke that affects the eye. ¹ It causes ischemia in the retina and is responsible for sudden visual loss, which can be irreversible in cases without early reperfusion. ^2,3 CRAO occurs in 1 in 10,000 of the population and most commonly between the ages of 50 and 70 years. Nearly one-half (45%) of the patients also have carotid artery disease. ^4,5 It has been suggested that retinal ischemia is a major cause of visual impairment and the development of glaucoma, ⁶ a leading cause of blindness. Retinal ischemia has also been associated with a high risk of subsequent ischemic stroke. As an extended part of the brain, the retina has all the characteristics of the central nervous system, from cell biology and biochemistry to histology, including the blood-retina (brain) barrier. Pathophysiological changes and signaling pathways activated in ischemic retina resemble those in ischemic brain, ^{7 –15} making the case for the retinal ischemia model as an alternative approach to study brain ischemia. Because the pathologic changes and molecular pathways activated in ischemic retina are very similar to those in ischemic brain tissue, this retinal ischemia model is ideal for studying the mechanisms of cerebral ischemia. 

The continuous and selective loss of retinal ganglion cells as a result of apoptosis has been implicated in the development of glaucoma, which eventually leads to a loss of vision. Several incriminating factors have been proposed to explain the apoptotic characteristics of the retinal ganglion cells. One factor implicated in glaucoma and other retinal diseases, such as retinal artery occlusion, is hypoxia/ischemia. ^{16 –21} Stroke caused by cerebral ischemia is a disruption of blood flow to the brain, which initiates a deleterious cascade of events that result in the death of neurons and lead to behavioral deficits. As in the case of cerebral ischemia, retinal ischemia is a consequence of an insufficient supply of blood to the retinal cells caused by the occlusion of blood vessels. ^{22 –25} Subsequent to reduced blood supply, retinal cells undergo hypoxia, which causes a number of pathophysiological processes, including glutamate toxicity and free radical formation, ultimately ending in damage to retinal neurons and in cell death by apoptosis. ^{22,24 –28} The focus of current treatments for retinal ischemia has been to reestablish the blood supply to the retina before damage results in irreversible vision loss. This can be accomplished by a combination of therapeutic procedures, including laser treatment, ocular massage, application of a thrombolytic agent such as tissue plasminogen activator, and vasodilators. However, these treatments have limited success. Laser treatment must be implemented early enough to be effective, whereas ocular massage and application of a thrombolytic agent and vasodilators are only useful in patients with CRAO when applied in combination with other treatments. ²⁹  

The retinal ischemia-reperfusion animal model is used in the study of transient ischemia-related diseases, such as CRAO, angle-closure glaucoma, and carotid artery disease. Various methods have been used to create a model of retinal ischemia-reperfusion. The two representative methods for producing a model of retinal ischemia-reperfusion already reported are raising intraocular pressure (IOP) above systolic arterial blood pressure through cannulation of the eye ^{24,30 –39} and ligating the central retinal artery together with the optic nerve. ^{25,30,40 –42} However, both these methods have drawbacks and limitations, ^25,42,43 including a low-stability procedure that requires needle penetration through the cornea in addition to inflammation and mechanical damage to ocular structures. Ligation of the optic nerve unavoidably involves damage to the nerve and can induce retinal changes unrelated to ischemia. In this study, we describe the usefulness of a new and refined model of retinal ischemia that is reproducible and easily reversible and that involves only the vascular structures of the retina. The microarray-based studies of transcriptional changes in the CRAL model of retinal ischemia and reperfusion can provide important insight into the mechanisms and pathways involved in human ischemic, inflammatory, and neurodegenerative retinal diseases. 

Methods

Animals

All animals in this study were treated in compliance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Male adult Lewis rats (weight range, 220–280 g) were anesthetized by induction with 5% halothane and by intramuscular injection with a mixture of 80 mg/kg ketamine and 10 mg/kg xylazine. The pupils were dilated with topical 0.5% tropicamide. Ten rats were used for each experiment. 

Surgery and Tissue Preparation

Retinal ischemia was induced by central retinal artery ligation (CRAL). CRAL surgery was performed by a trained ophthalmologist; the procedure was observed and confirmed by an ophthalmologist-surgeon at the Department of Ophthalmology. After a lateral conjunctival dissection and removal of the lateral rectus muscle, the left optic nerve was exposed by blunt dissection. The dural sheath surrounding the optic nerve was longitudinally opened with fine scissors, and the central artery was exposed. Then the central retinal artery was directly ligated with 10–0 nylon suture. The suture was positioned after the trifurcation of the ophthalmic into the central retinal artery and two posterior ciliary arteries. ⁴⁴ The suture was tightened until blood flow stopped in all retinal vessels. Complete cessation of blood flow in the retina was confirmed under the surgical microscope with a coverslip throughout the ischemic period (Fig. 1). The optic nerve head blood supply is derived mostly from the posterior ciliary arteries ⁴⁴ and peripapillary choroid, ⁴⁵ ; thus, a recurrent reduction in blood supply caused by the occlusion of the central retinal artery ⁴⁴ or the recurrent branches of the retina ⁴⁶ could not be completely ruled out. The right eye was treated as a sham control, including all the steps described except for suture insertion. After specific ischemic durations (30 and 90 minutes), the sutures were released, and the reperfusion of retinal blood flow was confirmed by fundus examination. The fundus photographs after the retinal occlusion were visualized under the surgical microscope with a coverslip. The coverslip tends to cause reflection, which can be a challenge for obtaining high-quality images of the choroidal vasculature. Recently, Zhang et al. ⁴⁷ developed a CRAO model that induced occlusion of the central retinal artery by laser irradiation–induced photothrombosis. The following retinal ischemia images in their studies also showed effects similar to those of our CRAL model. However, any potential side effects on the ophthalmic artery blood flow, secondary to the acute and chronic effects of the surgical procedure, could not be completely excluded. 

Figure 1.

 

Postperfusion of retinal artery. The blood supply to the retina can be monitored in real time with funduscopic examination before, during, and after CRAL. Reperfusion durations after CRAL are as denoted.

The surgical field was covered by autoclaved drapes. To avoid possible corneal exposure damage, a fine moist cotton pad was applied to the surface of the cornea after CRAL. Body temperature of the animals was maintained at 37°C with a heating pad during the experimental period. Cases requiring more than 5 minutes for the establishment of normal blood flow to the retina were excluded from this study. The surgical wound was closed with 8–0 sutures. All animals were allowed to recover approximately 10 minutes after ligation release. Postoperative care included topical administration of ophthalmic ointment (Vetropolymicin; Pitman-Moore, Washington Crossing, PA) and the analgesic buprenorphine (0.05–0.1 mg/kg) by subcutaneous injection immediately at the end of the surgery. The fully recovered animals were then returned to their individual cages. The topical antibiotic (Vetropolymicin; Pitman-Moore) was applied once daily in all animals of each experimental group. Animals were observed twice daily for adverse health conditions. At the assigned time points (3 hours, 12 hours, and 48 hours), the animals were anesthetized with diethyl ether. The American Veterinary Medical Association Panel on euthanatization currently promotes the use of isoflurane as a replacement for diethyl ether because of explosion hazards and increased toxicity. The animals were fixed by intracardiac perfusion with 60 mL cold (4°C) 0.01 M sodium phosphate, pH 7.4, with 0.14 M NaCl (PBS) followed by 4% (wt/vol) paraformaldehyde solution in 0.1 M phosphate buffer, pH 7.4 at room temperature over a period of 5 minutes for histochemical experiments, or with DEPC-treated 0.1 M phosphate buffer saline for molecular biological experiments. Both eyes were enucleated immediately and frozen in liquid nitrogen (Tissue Tek OCT Compound; Sakura Finetek, Tokyo, Japan) for histochemical analyses, or dissected retina tissues were flash frozen in 1.5 mL Eppendorf tubes. Frozen sections of 10-μm thickness were cut in a cryostat and stored at −20°C for morphologic and immunohistochemical examination. 

Light Microscopy Study

Morphologic evaluations were performed with 10-μm sagittal sections stained with hematoxylin and eosin. Representative photographs were taken in the region between 0.5 mm and 1 mm from the optic nerve head with a digital camera (DC100; Leica Microsystems GmbH, Wetzlar, Germany) and were transferred to a computer. Retinal thickness, from the inner limiting membrane to the retinal pigment epithelium, was measured for each section with appropriate calibrations. 

TUNEL Evaluation

Sections were fixed with 4% paraformaldehyde in 0.1 M PBS for 10 minutes at 4°C followed by treatment with 3% H₂O₂ to block endogenous peroxidases. TdT-dUTP nick-end labeling (TUNEL) was performed by incubating the fixed sections with TdT(1 μg/mL) and dUTP in TdT buffer (30 mM Trizma base, pH 7.2, 140 mM sodium cacodylate, and 1 mM cobalt chloride) for 90 minutes at 37°C. After rinsing, the reaction product was visualized with the ABC staining kit (Vector Laboratories, Inc., Burlingame, CA) using 3,3′-diaminobenzidine followed by methyl green counterstaining. Omission of TdT during the incubation followed by the rest of the labeling procedure served as a negative control. 

RNA Isolation

In total, seven distinct samples were used, each pooled from 10 animals. Specifically, total RNA was isolated from four experimental groups representing 30 and 90 minutes of retinal artery occlusion followed by 3 hours and 12 hours of reperfusion periods. Additionally, total RNA was isolated from two sham-treated samples (3 hours and 12 hours after surgery) and one normal retina derived from untreated animals. Total RNA was isolated using reagent (Trizol; Invitrogen Canada, Burlington, ON, Canada) according to the manufacturer's description. The concentration and purity of total RNA were determined by spectrophotometric measurement at wavelengths of 260 nm and 280 nm. To ensure the absence of any genomic DNA contamination, each RNA sample was incubated with DNase I, followed by phenol/chloroform extraction and ethanol precipitation. The RNA was again analyzed spectrophotometrically at 260 and 280 nm to calculate the concentration and was examined (RNA 6000 Nano LabChips with a 2100 Bioanalyzer [Agilent Technologies, Palo Alto, CA]) to verify RNA integrity. 

Microarray Hybridization and Data Analysis

The bioarray kit (UniSet Rat I Bioarray 10K; Amersham Biosciences/GE Healthcare, Piscataway, NJ) used in these experiments contained an array of 9589 30-base long single-stranded oligonucleotide probes within a single reaction chamber on a single slide. The 9589 probes represent 9203 unique accession numbers (genes), corresponding to approximately 8935 well-annotated mRNA sequences found in the Rat UniGene set of unique clusters and 386 control probes. Each bioarray contained probes for an independent assay of all genes in the set. 

Five micrograms of total RNA was used to prepare the target samples according to manufacturer's protocol. cDNA synthesis, in vitro transcription, and the production of a biotin-labeled cRNA probes for oligonucleotide arrays were performed as described by the manufacturer. The hybridization mixture contained 10 μg fragmented sample cRNA together with defined amounts of control bacterial spikes. Processed slides were scanned (Axon GenePix Scanner; Molecular Devices, Sunnyvale, CA) with the laser set to 635 nm, the photomultiplier tube voltage to 600, and the scan resolution to 10 μm, and images for each slide were analyzed using the Codelink Expression Analysis Software (GE Health Sciences, Uppsala, Sweden). 

Significant Analysis of Microarray

Significant analysis of microarray (SAM) algorithm allows for findings of differentially expressed genes in a set of experiments and replicates. ⁴⁸ SAM was performed on the quantile ⁴⁹ normalized data for all genes using the free available SAM software (Stanford, http://www-stat.stanford.edu/∼tibs/SAM/). SAM analysis was performed to determine the population of differentially expressed genes between the sham and the ischemia samples using a false discovery rate (FDR) of ≤1% as a threshold. 

Gene Set Enrichment Analysis

In addition to the analysis on the single gene expression, we performed analyses on sets of genes connected on a particular functional pathway. This may provide valuable information about the complex cellular and molecular responses after an ischemic event. To achieve this goal, Gene Set Enrichment Analysis (GSEA), a computational method of supervised analysis was used. GSEA connects the activation of sets of genes with biology, ^50,51 such as genes sets typically underlying different cellular pathways, genes coexpressed, or genes colocalized on a same cytogenetic band. This analysis considers that modest coordinate changes in the activity of several functionally related genes could be more relevant for a molecular network rather than a strong change in expression of a single gene. In our study the extent to which a set of genes was enriched after different ischemia protocols was analyzed. 

GSEA software implements a particular algorithm to determine whether a ranked list of genes from a microarray data set (based on their correlation to the control phenotype) contains a priori established gene sets enriched with the highly ranked genes. Predetermined gene sets have been derived from an extensive list of sets with genes confirmed based on public sources to represent a pathway, with common upstream cis-regulatory elements, or localized on the same chromosomal location. The curated gene sets were screened against the microarray data set, and enrichment plots were generated for each gene set. The proportion of false positives was controlled by calculating the FDR. Gene sets with FDR <25% were considered to be of potential interest; this threshold was used to facilitate the exploration process rather than being a strict decision-making threshold. ⁵²  

Real-Time RT-PCR

Oligonucleotide primers for genes of interest and for endogenous control genes (Actin and GAPDH) were designed for real-time PCR (Primer Express 2.0 software; Applied Biosystems, Foster City, CA). All primers were purchased from Proligo (Boulder, CO). The concentration of all RNA used for PCR was initially determined spectrophotometrically. Reverse-transcription reactions were carried out for each RNA sample using reverse-transcription reagents (Superscript II; Invitrogen Canada, Burlington, ON, Canada). Each reaction tube contained 1 μg total RNA in a volume of 20 μL containing 0.5 mM dNTP, 25 ng oligo (dT)12–18 primers, 1× RT buffer, 5 mM MgCl₂, 10 μM dithiothreitol, 2 U RNase inhibitor, and 50 U reverse transcriptase (Superscript II; Invitrogen Canada). 

Real-time PCR was performed with a sequence detection system (ABI Prism 7300; Applied Biosystems). PCR master mix (SYBR Green; Applied Biosystems) was used on cDNA samples in 96-well optical plates. For each 25 μL PCR reaction, 2 μL cDNA (appropriate dilution), 0.75 μL of both 5′ and 3′ primers (10 μM), 12.5 μL PCR master mix (SYBR Green; Applied Biosystems), and 9.75 μL water were mixed together. The parameters were 95°C for 10 minutes, 1 cycle, then 60°C for 1 minute and 95° for 15 seconds for 40 cycles. To confirm the amplification of a specific cDNA, the dissociation temperature was examined and compared with the calculated melting temperature for each amplified product. The amplified products were also examined by agarose gel electrophoresis. 

Relative quantification of gene expression was performed using the fold change method as described previously ⁵³ and as recommended by the manufacturer of the sequence detection system (ABI Prism 7300; Applied Biosystems). Briefly, the normalized values for all amplification runs were calculated by subtracting threshold cycle (Ct) values of the β-actin as the endogenous control from those of target genes. The standard curve for each gene was constructed from the four serially diluted samples in duplicate, starting with a 1:20 dilution of cDNA and 1:2 diluted to 1:40, 1:80, and 1:160. Relative gene expression was represented by the difference between the normalized values of the experimental samples and that of corresponding sham controls (ΔΔCt); 2^-ΔΔCt for each product was used to calculate fold changes. 

Results

Assessment of the Ischemic Effect after Central Retinal Artery Occlusion

As shown in Figure 2, substantial morphologic changes occurred in most retinal layers, especially in the inner nuclear layer (INL) and the ganglion cell layer (GCL) from 3 hours to 7 days after reperfusion after the occlusion of the central retinal artery. After 12 hours after reperfusion, the ganglion cells were noticeably absent, and after 24 to 48 hours of reperfusion, there was a massive loss of cells in the INL layer. Significant retinal degeneration was confirmed by the reduced thickness of the retina at 7 days. The retinal edema in the CRAL eyes occurred mainly in the GCL at the same time point. As shown in Figures 2a and 2b, the total thickness of ischemic retina was significantly increased compared with that of control eyes (360 ± 20 μM vs. 315 ± 23 μM; P < 0.004) 3 hours after reperfusion. Further evidence of ischemic damage was revealed by TUNEL labeling, as shown in Figure 2c. Although no TUNEL-positive cells were seen in the control eyes, numerous TUNEL-positive cells were present 12 hours after reperfusion after CRAL. The TUNEL-positive cells were located primarily in the GCL and INL. TUNEL-positive cells were still observed in the INL and outer nuclear layer (ONL) 48 hours after reperfusion but disappeared later. Furthermore, TUNEL staining was concentrated in the nuclei of the cells (Fig. 2d). 

Figure 2.

 

(a) 90 minutes of ischemia, 3 hours of reperfusion: three hours after reperfusion in the CRAL model stained with hematoxylin and eosin. Images were taken in the region 0.5 mm to 1 mm from the optic nerve. Retinal edema occurred in the GCL (black arrow). Left: control eye; right: CRAL. Scale bar, 50 μm. (b) Light microphotographs of the rat retina after different survival intervals after reperfusion in 10-μm thick sagittal sections stained with hematoxylin and eosin. Retinal damage is observed in the INL and GCL. (c) Light microphotographs showing the TUNEL staining at each time point after CRAL. TUNEL-positive cells were observed 12 hours after reperfusion. TUNEL-positive cells were still present to a lesser extent in the INL and ONL by 48 hours. No TUNEL-positive cells were seen in normal eyes. Scale bar, 50 μm. (d) High magnification of photomicrographs of TUNEL staining 12 hours after reperfusion in the CRAL eye. TUNEL staining was concentrated in the nuclei of cells in the ganglion layer (G; arrows). Scale bar, 25 μm.

Microarray Analysis to Identify Differentially Expressed Genes

To help delineate the genes involved in the ischemic response of retina, we analyzed gene expression profiles after mild (30 minutes) and severe (90 minutes) ischemia of rat retina, each followed by a 3-hour or 12-hour reperfusion period. Bioarrays (Uniset Rat I; GE Healthcare) were used to determine gene expression patterns after periods of reperfusion after CRAL ischemia. The CRAL ischemia-derived profiles were compared with their corresponding sham controls to identify the ischemia-specific differentially expressed genes. We observed that the biological replicates within each experimental group were more closely related and had a correlation coefficient above 0.98 compared with the lower correlation between ischemia (30 and 90 minutes) and the corresponding sham-treated groups (0.94 and 0.96). Correlation plots of biologically replicated experiments for ischemia and sham groups at 3 hours after reperfusion are shown in Figure 3. ⁴⁹  

Figure 3.

 

Raw data were processed by acquiring the pixel values, followed by quantile normalization. ⁴⁹ Shown here is the correlation plot between the biological replicates within the same experimental group and between the ischemia and the sham groups. The biological replicates within each experimental group were better correlated, with a correlation coefficient above 0.98, than between ischemic and sham control groups which showed lower correlations.

SAM Analysis Identified Significantly Differentially Expressed Genes

To assess the reproducibility of the data, normalized hybridization signals were compared using the Significant Analysis of Microarray (SAM) algorithm ⁴⁸ (Stanford University, http://www-stat.stanford.edu/∼tibs/SAM/). SAM is used to identify significant genes based on reproducible or differential expression between sets of samples. SAM was performed on quantile normalized pixel values across genes and experimental conditions (ischemia vs. sham). In all paired SAM analyses (ischemia vs. sham), selection of differentially expressed genes with an FDR ≤1% resulted primarily in the identification of genes with at least 1.5-fold change. After 30 minutes of ischemia, 1459 genes were identified as differentially expressed at 3 hours after reperfusion. More were downregulated (n = 967) than upregulated (n = 492), as shown in Figure 4a. The same criteria identified 690 genes as differentially expressed at 3 hours after reperfusion after 90 minutes of ischemia, with 150 genes upregulated and 540 genes downregulated (Fig. 4b). However, similar analyses 12 hours after reperfusion after 30 minutes of ischemia resulted in 524 upregulated and 136 downregulated genes among 660 differentially expressed genes (Fig. 4c). Finally, the analyses 12 hours after reperfusion after 90 minutes of ischemia resulted in 363 upregulated and 228 downregulated genes among 591 differentially expressed genes (Fig. 4d). A noticeable trend was that after both 30 and 90 minutes of ischemia, more genes were downregulated 3 hours of reperfusion and more genes were upregulated 12 hours of reperfusion. A list of selected previously functionally characterized genes using the same SAM criteria (FDR ≤1%) at 3 hours and 12 hours after reperfusion after 30 and 90 minutes of retinal ischemia are provided in Supplementary Table S1. 

Figure 4.

 

SAM. Comparison of each post-reperfusion time point with its corresponding sham-treated sample. (a) 30-minute ischemia versus sham 3-hour reperfusion, (b) 90-minute ischemia versus sham 3-hour reperfusion, (c) 30-minute ischemia versus sham 12-hour reperfusion, and (d) 90-minute ischemia versus sham 12-hour reperfusion, respectively, with the median of FDR ≤1%.

The sham control for each experimental time point was used to minimize the effects of confounding variables, such as the individual animal-specific effects of anesthesia or analgesia or the effects induced by the surgery itself, thus allowing for the evaluation of ischemia-specific effects. In addition the sham controls for each experimental group, we examined the gene expression pattern from an additional untouched control group of animals and compared this with the expression pattern of each sham control to identify the genes differentially regulated by anesthesia, analgesia, or the surgical procedure itself. The number of genes expressed primarily as a consequence of sham treatment is illustrated as a Venn diagram in Figure 5, and a corresponding list of known genes is included as Supplementary Table S2. Interestingly, the 3-hour sham-reperfusion groups showed deviations in the expression levels of a large set of genes (more than 4500). This effect was not observed at 12 hours of sham reperfusion, as summarized with Venn diagrams shown in Figure 5. 

Figure 5.

 

Venn diagrams showing the intersections between ischemia versus sham and sham versus normal controls of upregulated and downregulated genes after 30 minutes and 90 minutes of transient retinal ischemia, with 3 hours and 12 hours reperfusion. The numbers of genes in respective differentially expressed gene sets and the number of overlapping genes commonly expressed in ischemia-treated and sham-operated animals are shown. Only significantly expressed genes are considered.

Gene Set Enrichment Analysis

Using the gene expression profiles after 30 and 90 minutes of ischemia and 3 and 12 hours of reperfusion, we carried out GSEA analysis to delineate the involved functional pathways. We observed that the mild form of retinal ischemia for 30 minutes, followed by 3 hours of reperfusion, induced a significant overrepresentation of gene sets related to the regulation of transcription and cell adhesion (Fig. 6a) among 180 significantly enriched gene sets with FDR <25%. Many of the gene set probes were highly ranked in the ordered data set of this experimental group; in other words, these probes were differentially upregulated compared with sham control, resulting in the high running enrichment scores for the transcription and cell adhesion gene sets. Among the genes driving the high enrichment scores for most significant gene sets were ATF3, ID2, Klf4, BTG2, and Jun, as confirmed by the leading edge analysis (Supplementary Fig. S1. 

Figure 6.

 

Gene set enrichment analysis of functional pathways induced after mild (30 minutes) or severe (90 minutes) CRAL-induced retinal ischemia-reperfusion injury. Enrichment plots generated among the top-rated gene sets are illustrated. Passerini transcription was among the top gene sets enriched after mild (30-minute CRAL, 3-hour reperfusion) (a), whereas at 12 hours reperfusion induced enrichment of the gene sets for protein kinase cascade (b). The severe form of CRAL-induced ischemia (90 minutes) after 3 hours of reperfusion induced strong activation of the gene set for glycolysis and gluconeogenesis (c). At 12 hours of reperfusion, a strong activation of gene sets for apoptosis was induced (d). Interestingly, sham treatment for 90 minutes followed by 3 hours of reperfusion enriched the homeostasis-related gene set compared with normal control (e). This effect was not observed after 12 hours of sham reperfusion.

The mild form of retinal ischemia for 30 minutes, followed by the longer 12-hour reperfusion, induced enrichment of the gene sets for protein kinase cascade (Fig. 6b) among 128 significantly enriched gene sets with FDR <25%. The genes driving the high enrichment scores were MAP kinases and caspases (Casp3 and Casp9), which were also validated by the leading edge analysis (Supplementary Fig. S2). The severe form of retinal ischemia performed for 90 minutes, followed by 3 hours of reperfusion, induced strong activation of a gene set for glycolysis/gluconeogenesis (Fig. 6c) among 15 significantly enriched gene sets with FDR <25%. One of the genes driving the high enrichment score was glucose-6-phosphatase (G6PC; Supplementary Fig. S3). 

The severe form of retinal ischemia performed for 90 minutes, followed by the longer 12-hour reperfusion, induced strong activation of gene sets for apoptosis (Fig. 6d) among 160 top-ranked gene sets with FDR <25%. Among the upregulated genes driving the enrichment score high were Akt1, Akt3, Pik3R1, Prkcb1, tumor suppressor gene Tp53, caspases (Casp3, Casp7, and Casp9) and tumor necrosis factor (TNF; Supplementary Fig. S4). Interestingly, we also observed that sham treatment for 90 minutes, followed by 3 hours of reperfusion, enriched the homeostasis-related gene sets compared with normal controls (Fig. 6e). Endothelin receptor type A (EDNRA) was one of the genes that drove the enrichment score high for the homeostatic pathway. 

In gene set analysis, there were no enriched gene sets with FDR scores <25% in the sham control relative to retinal ischemia (i.e., gene sets that were enriched with downregulated genes after ischemia). 

RT-PCR Validates the Bioarray-Derived Differentially Regulated Genes

To examine the accuracy of differentially expressed gene identification by the microarray analyses, we compared the array-derived expression levels with those determined by real-time quantitative RT-PCR (qRT-PCR). The same samples of RNA were used for both the array and qRT-PCR analyses. In total, 50 target genes from the list of differentially expressed genes, as determined by SAM, were selected randomly for qRT-PCR validation. Real-time PCR data were normalized with β-actin as the internal control. Figure 7 ⁵³ presents results comparing microarray and real-time PCR findings for a sample of randomly selected differentially expressed genes. The expression (downregulation or upregulation) for all genes examined showed a direction of expression that was in good agreement with the two methods. 

Figure 7.

 

Validation of microarray-derived candidates with real-time qRT-PCR Confirmation of bioarray expression results using qRT-PCR. Candidate target genes were randomly selected from the SAM-analyzed data, and their nucleotide sequences were used to design primers for real-time RT-PCR analysis on individual RNA samples. β-Actin was used as a normalizer for each experiment. ΔΔCt was used to determine the fold change in gene expression. ⁵³ Results from bioarray and real-time PCR are expressed as log₂-transformed fold changes between ischemia and corresponding sham-treated samples. Transcripts are indicated with gene symbols.

We also examined the time course of changes at several periods of reperfusion for a number of genes that were identified as differentially expressed with real-time qRT-PCR. As shown in Figure 8, we observed that the KLF4 gene is induced primarily after 90 minutes of ischemia and 3 hours of reperfusion only. GRO1 and HMOX1 genes were most inducible after 90 minutes of ischemia and 12 hours of reperfusion, whereas TIS11, U76717, CRYBA4, SSR3, and PPP3CA were most inducible after 30 minutes of ischemia and 48 hours of reperfusion. 

Figure 8.

 

Expression pattern after early and late durations of ischemia and reperfusion. Time course delineation at 3, 12, and 48 hours of reperfusion with real-time qRT-PCR of individual differentially expressed genes identified using SAM statistics. Gene-specific nucleotide sequences were used to design primers for real-time RT-PCR analysis on individual RNA samples, as denoted. β-Actin was used as a normalizer for each experiment. ΔΔCt was used to determine the fold change in gene expression, as described in Materials and Methods. All real-time qRT-PCR results are expressed as log₂ fold changes between ischemia and corresponding sham-treated samples. Transcripts are indicated with gene symbols or GenBank accession numbers.

Class Neighborhood Analysis

We performed class neighborhood analysis ⁵⁴ of the upregulated and downregulated genes after 30 minutes and 90 minutes of transient retinal ischemia and 3 hours and 12 hours of reperfusion. The top 50 upregulated and 50 downregulated genes that most closely resembled the profiles of our experimental groups were selected. The top 10 that best distinguished each class were illustrated using a heat map (Fig. 9). The feature of mild ischemia after 3 hours of reperfusion is the exclusive induction of genes involved in amino acid transport (Slc6a6 ^55,56 ), adhesion and plasticity (neurexin Nrxn, ⁵⁷ Ncam1, and Opcml), neuronal regeneration (e.g., reticulon family member Nogo/Rtn4), and the reduced expression of stress response (e.g., endothelin Edn2, adenosine transporter Slc28a2, ryanodine receptor Ryr3), and the ionotropic glutamate receptor subunit Gria4 (GluR4). The hallmark of mild ischemia after 12 hours of reperfusion included the exclusive upregulation of genes for the sodium bicarbonate cotransporter Slc4a7 and cerebral ischemia-related heat shock protein Hspa4 (Hsp70). The expression profile of severe ischemia after 3 hours of reperfusion was characterized by the induction of immediate-early genes and transcriptional factors Fos, Junb, Egr2, antiapoptotic gene Btg2, and activity-regulated gene Arc. The structural eye genes such as eye lens crystalline Crybb1 and the growth factor Ctgf were downregulated. The marker class for severe ischemia at 12 hours of reperfusion included the induction of extracellular matrix peptidase inhibitor Serpinb2, plasminogen activator Plau and its receptor Plaur, intermediate filament nestin (Nes), and neutral amino acid transporter Slc1a5. Genes such as metallothionein 3 (Mt3) and receptor tyrosine kinase oncogene Erbb2 were exclusively repressed in this group. 

Figure 9.

 

Neighborhood analysis of the upregulated and downregulated genes after 30 minutes and 90 minutes of transient retinal ischemia and 3 hours and 12 hours of reperfusion. Rows: individual genes; columns: different samples. Gene expression levels are normalized across all samples (mean, 0; SD, 1). Expression intensities above the mean are red, and those below the mean are blue. Colors are scaled to reflect the expression levels relative to the mean, expressed as the number of SDs above or below the mean. Upregulated genes (top half) and downregulated genes (bottom half) in an individual experimental condition.

Discussion

This is the first report to describe gene expression changes after ischemia induced by CRAL. In this study, we established a CRAL model of transient retinal ischemia/reperfusion similar to the clinical setting of CRAO ^58,59 and monocular amaurosis fugax, ⁶⁰ transient visual loss caused by ischemia or vascular insufficiency; the pathology was almost identical to that in human patients. This model is reproducible, easily reversible, and interrupts the blood supply only to the retina without affecting other tissues in the eye. The most commonly used retinal ischemia models that reproduce the retinal ischemia-reperfusion injury are optic nerve bundle ligation and IOP elevation. The former ligates the bundle containing both the optic nerve and all the surrounding vessels, which supply blood to the entire eye. The latter raises the IOP by injecting fluid into the anterior chamber to create a transient obstruction of the retinal artery. ^22,24,25  

In the present study, we have demonstrated a similar type of occlusion of the retinal blood flow, as evident in the results of the retinal funduscopic examinations. The important point is the pattern of the vascular obstruction. In the clinical field, central retinal arterial obstruction is observed, for example, because of thrombus formation in the vessels. In such a case, only the central retinal artery is obstructed. The method introduced in the present study also exhibited the obstruction of the central retinal artery without any damage to other tissues, such as the optic nerve. In the anatomic view, all the retinal nerve fibers finally join the optic nerve. Thus, retinal constitutions are closely linked to the optic nerve. Therefore, maintaining the anatomic structure of the optic nerve is crucial for experiments involving the retina. Ligation of the central retinal artery with the optic nerve unavoidably involves optic nerve injury. Such a downside adds merit to the method developed in the present study. Therefore, the purpose of introducing the CRAL model was to assess the retinal ischemia in isolation from the effects of optic nerve ligation or the increase of IOP. However, the optic nerve ligation and the high IOP remain invaluable models with which to study aspects of eye pathophysiology related to retinal ischemia caused by increased IOP or the neurodegeneration of the optic nerve with variable etiologies. 

Although most ischemic studies have been performed using global or focal brain ischemia models, the retinal ischemic model is a unique alternative. In addition to a true resemblance to the central nervous system, the CRAL model has several advantages over conventional brain ischemia models. The retina is the only part of the central nervous system that is noninvasively accessible and visible. The blood supply to the retina can be monitored in real-time with funduscopic examination before, during, and after CRAL. In addition, gene therapy can be readily and repeatedly performed in a controlled and reliable manner. The retina has a less complicated structure than the brain, and the retinal cell types are well defined, relatively homogeneous, and have fewer and better known cell types in each layer than in the brain. Thus, the impact of injury and treatment on the function of retinal cells after ischemia and functional recovery can be assessed in a reliable and quantitative manner. 

We have evaluated the characteristics of the CRAL model using histologic, immunohistochemical, and molecular approaches. Our histologic and immunohistochemical analysis demonstrate that 30 minutes ischemia followed by reperfusion did not cause permanent damage to the retina compared with 90 minutes of ischemia that resulted in a complete loss of the ganglion cells in a few days. Histologic examination after CRAL at several reperfusion periods revealed that CRAL caused retinal degeneration similar to that of other models of retinal ischemia, such as optic nerve ligation model. However, the retinal degeneration observed after CRAL was more gradual than that observed in the optic nerve ligation model; a longer reperfusion period was required than in the optic nerve ligation model, even though transient early edema occurred in the GCL. Furthermore, the TUNEL-positive cells were seen in the ganglion layer after CRAL with characteristic intracellular distribution of the TUNEL reactive products. Clinically, the primary changes after CRAO include widespread retinal edema in the GCL ⁵⁹ similar to the observations in the CRAL model. We also observed that the GCL showed no significant edema soon after retinal ischemia but, like optic nerve ligation, demonstrated substantial cell loss at later time points. TUNEL-positive cells were present between 12 and 48 hours after reperfusion, which is consistent with previous results in an optic nerve ligation model. ^{13,15,61 –63} A noticeable difference from the optic nerve ligation model was that more TUNEL-positive cells were found in the ganglion cell layer in our CRAL model at 12 and 48 hours after reperfusion. In addition, the TUNEL staining in the CRAL model was found only in the nuclei. This is an expected pattern of TUNEL staining because the reaction labels the free 3′ end of DNA in the nuclei of apoptotic cells, and the nuclear membrane usually remains intact during apoptosis. A ring-like pattern of TUNEL staining has also been observed after IOP elevation. ^13,61,62  

We have anticipated that the lack of permanent damage after 30 minutes of ischemia and the complete loss of the ganglion cells after 90 minutes of ischemia might be reflected in the molecular makeup of gene expression within the retinal cells. To elucidate the molecular attributes of retinal ischemia in the CRAL model, we investigated gene expression patterns after 30 minutes of mild and 90 minutes of severe ischemia, followed by periods of early (3 hours) and late (12 hours) reperfusion. Although a number of gene expression studies have attempted to identify the suite of genes expressed in the retina after the increase of IOP of the whole eye, ^{56,64 –67} none have yet compared the gene expression patterns after CRAL-induced retinal ischemia followed by different periods of reperfusion. 

In the present study, our gene expression analyses revealed that a greater number of genes (1459 genes) were differentially expressed after 30 minutes of CRAL and 3 hours of reperfusion than either at 12-hour reperfusion after both 30 (660 genes) and 90 (591 genes) minutes of CRAL or at 3-hour perfusion after 90 (690 genes) minutes of CRAL. Among the differentially expressed genes at the 3-hour reperfusion period, significantly more genes were downregulated than upregulated after both 30 minutes (492 vs. 967 genes) and 90 minutes (150 vs. 540 genes) of CRAL-induced ischemia. Significantly more genes were upregulated than downregulated at 12 hours of reperfusion after 30 minutes (524 vs. 136 genes) and 90 minutes of ischemia, but the difference was smaller (363 vs. 228 genes), pointing out an early transient inhibition of transcription after CRAL ischemia. Sham controls were used to identify the genes differentially affected by anesthesia, analgesia, or the surgical procedure itself. Sham treatment showed more effects in the gene expression patterns at 3 hours than at 12 hours of reperfusion compared with untreated control, suggesting that stress effects are lessened gradually after the surgical procedure. 

Using GSEA, we found that during the 3-hour reperfusion period after 30 minutes of CRAL-induced ischemia, there is prominent activation of sets of genes involved in the regulation of transcription—among them ATF3, ID2, Klf4, BTG2, and immediate-early genes such as Jun. Considering that more downregulated than upregulated genes were found, we believe this is likely a feedback response to activate transcription. As expected, more genes were upregulated at 12 hours reperfusion after 30 minutes of CRAL-induced ischemia, including MAP kinases and caspases. The severe 90-minute form of retinal ischemia induced strong activation of the gene set for glycolysis/gluconeogenesis at 3 hours of reperfusion, which was followed by a strong activation of gene sets for apoptosis at 12 hours of reperfusion. 

During retinal development distinct steps of neuronal and glial differentiation take place. The establishment of specific neuronal pathways, functional synapses formation, and retinal vasculature facilitate the onset of vision. The roles of many genes during retinal development have been established. ^64,68,69 These genes show distinct prenatal and postnatal expression patterns during development. Usually they peak during development and decline in adulthood. Many of the genes affected with the CRAL model are normally expressed during retinal development. ⁶⁸ The mild form of CRAL-induced ischemia (30 minutes) after 3 hours of reperfusion induces early embryonic or early postnatal genes such as ATP-binding cassette (Abcd3), kinesin family member (Kif1b), neural cell adhesion molecule 1 (Ncam1), nardilysin (Nrd1), solute carrier family 6 (Slc6a6), and calcium/calmodulin-dependent protein kinase kinase 2 (Camkk2). The same mild ischemia induces other developmentally regulated genes after 12 hours of reperfusion, such as the embryonic SRY-box containing transcriptional factor 11 (Sox11), postnatal RAS p21 protein activator 1 (Rasa1), vitronectin (Vtn), and P21-activated kinase 2 (Pak2). The severe form of CRAL-induced ischemia (90 minutes) also alters the expression pattern of the developmental genes. Although some are common with the mild ischemia (Kif1b, Sox11, and Rasa1), there are early embryonic and postnatal genes that are specifically altered after 90 minutes of ischemia. Examples are the protease calpain 2 (Capn2), pleomorphic adenoma gene-like 1 (Plagl1), low-density lipoprotein receptor-related protein 4 (Lrp4), solute carrier family 1 (neutral amino acid transporter), member 5 (Slc1a5), and ornithine decarboxylase 1 (Odc1). Activation of these developmentally regulated genes may suggest that some of the processes active only during early retinal development ^64,68,69 are activated during and after ischemia to facilitate the rewiring and reestablishment of the functional retinal circuitry. 

Activation of the intrinsic neuroprotective molecular pathways involved in neuroprotection after ischemic preconditioning (IPC) was attributed to the activation of genes that enable retinal cells to cope with the adverse conditions during ischemia. ^56,70 Although the mechanisms of IPC are not well known, HIF1 transcription factor activation, ^70,71 Hsp27 upregulation, ⁷² and the induction of nitric oxide synthase (iNOS) ⁷³ have been suggested to underlie the IPC-induced neuroprotection in the retina. Our results show that some of the genes induced by ischemic preconditioning ^56,70 are also induced with both mild and severe CRAL-induced ischemia, irrespective of the period of reperfusion. A group of IPC genes were present in all conditions, including immediate-early gene Fos (FBJ murine osteosarcoma viral oncogene homolog) and cell death-related Annexin A1 (Anxa1). Proinflammatory IPC-induced genes that peak 3 hours after CRAL reperfusion, such as Kruppel-like factor 4 (Klf4) and activating transcription factor 3 (Atf3), were also identifiable. Another IPC gene, the early growth response factor 1 (Egr1), was also expressed 12 hours after both mild and severe ischemia. The pattern of gene expression in the CRAL model of retinal ischemia suggests that IPC genes are activated across different conditions, implying that protective mechanisms are triggered with both mild and severe ischemia and that only the balance between adverse and neuroprotective processes likely determines neuronal fate. 

All layers of the retina age, which can contribute to loss of vision from age-related retinal diseases. The genes involved in cell growth and protein metabolism have been demonstrated to be preferentially expressed in younger retinas, whereas stress response genes are expressed at higher levels in older retinas. ⁷⁴ Similarly, we found that 3 hours after CRAL there was a reduction of growth hormone receptor (Ghr) regardless of the length of the ischemic period. Experimental deletion of Ghr has also been associated with the process of aging and increased lifespan. ⁷⁵ Interestingly, after 12 hours of reperfusion and 30 or 90 minutes of CRAL, there was an increase of the transcription for two additional genes related to aging, the plasminogen activator urokinase (Plau) and cyclin-dependent kinase inhibitor 1A (Cdkn1a). Mice with overexpressed Plau or deleted Cdkn1a have demonstrated delayed aging and increased lifespan without accelerated cancer formation. ^76,77 Thus, our results demonstrate the suitability of the CRAL model in identifying global patterns of retinal gene expression related to age-associated retinal diseases. Therapeutic targeting of age-involved genes identified with the CRAL model might be beneficial for the treatment of retinal ischemia and pathologic retinal aging including age-related macular degeneration. 

Identifying the genes required to maintain the structural integrity of the brain and the mutations that cause adult-onset neurodegeneration has been instrumental in developing our understanding of the etiology and progression of such diseases. It is believed that ischemia-induced injury plays an important role in chronic neurodegenerative diseases of the retina, such as glaucoma and retinal ischemia. Glaucoma is a complex and genetically heterogeneous disease, characterized by the progressive apoptotic death of retinal ganglion cells, ¹⁷ leading eventually to irreversible blindness. The complex processes of retinal ganglion cell (RGC) death in glaucoma include impaired neurotrophic support, excitotoxicity, oxidative stress, and neuroinflammation. However, the molecular basis of the disorder remains unknown in most cases. The profile of retinal gene expression changes after CRAL is similar to the molecular events that underlie RGC death in an IOP elevation model. ⁷⁸ Common genes were observed as differentially expressed between the CRAL and the IOP models, such as Stat3 (Signal transducer and activator of transcription 3), c-Fos, Anxa1, c-Jun, and Atf3. In contrast, we found that Sox11 was the only common gene between CRAL and optic nerve axotomy, ⁷⁸ suggesting that CRAL was more similar with the IOP model. However, many genes we have already addressed, such as Klf4, Slc6a6, and Egr4, were selectively expressed in the CRAL model. One of the genes, Pdpk1 (3-phosphoinositide dependent protein kinase-1), is in the signaling pathways activated by several growth factors and hormones and is crucial for the activation of the AKT/PI3 prosurvival pathway. ⁷⁹  

We conclude that early changes after mild ischemia of 30 minutes are driven by the activation of transcription-related genes followed by the upregulation of kinases, glycolysis/gluconeogenesis, and eventually apoptosis-related genes. The results described here have led us to consider CRAL as a clinically relevant model of retinal ischemia and an effective experimental model with which to study the retinal processes underlying ischemia, inflammation, development, aging, and neurodegeneration. 

Supplementary Materials

Supplementary Table S1 - (PDF) 

Supplementary Table S2 - (PDF) 

Supplementary Figure S1 - (PDF) 

Supplementary Figure S2 - (PDF) 

Supplementary Figure S3 - (PDF) 

Supplementary Figure S4 - (PDF) 

Footnotes

 Supported in part by Canadian Stroke Network of Excellence Research Grant 20R30893 and by Canadian Institute for Health Research Grant MOP62734.

Footnotes

 Disclosure: S.S. Prasad, None; L. Kojic, None; Y.H. Wen, None; Z. Chen, None; W. Xiong, None; W. Jia, None; M.S. Cynader, None