Cytokine Levels in Experimental Branch Retinal Vein Occlusion Treated With Either Bevacizumab or Triamcinolone Acetonide

Abstract

Purpose: To compare gene expression changes following branch retinal vein occlusion (BRVO) in the pig with and without bevacizumab (BEV) and triamcinolone acetonide (TA).

Methods: Photothrombotic BRVOs were created in both eyes of four groups of nine pigs (2, 6, 10, and 20 days). In each group, six pigs received intravitreal injections of BEV in one eye and TA in the fellow eye, with three pigs serving as untreated BRVO controls. Three untreated pigs served as healthy controls. Expression of mRNA of vascular endothelial growth factor (VEGF), glial fibrillary acidic protein (GFAP), dystrophin (DMD), potassium inwardly rectifying channel subfamily J member 10 protein (K_ir4.1, KCNJ10), aquaporin-4 (AQP4), stromal cell-derived factor-1α (CXCL12), interleukin-6 (IL6), interleukin-8 (IL8), monocyte chemoattractant protein-1 (CCL2), intercellular adhesion molecule 1 (ICAM1), and heat shock factor 1 (HSF1) were analyzed by quantitative reverse-transcription polymerase chain reaction. Retinal VEGF protein levels were characterized by immunohistochemistry.

Results: In untreated eyes, BRVO significantly increased expression of GFAP, IL8, CCL2, ICAM1, HSF1, and AQP4. Expression of VEGF, KCNJ10, and CXCL12 was significantly reduced by 6 days post-BRVO, with expression recovering to healthy control levels by day 20. Treatment with BEV or TA significantly increased VEGF, DMD, and IL6 expression compared with untreated BRVO eyes and suppressed BRVO-induced CCL2 and AQP4 upregulation, as well as recovery of KCNJ10 expression, at 10 to 20 days post-BRVO.

Conclusions: Inflammation and cellular osmohomeostasis rather than VEGF suppression appear to play important roles in BRVO-induced retinal neurodegeneration, enhanced in both BEV- and TA-treated retinas.

Translational Relevance: Inner retinal neurodegeneration seen in this acute model of BRVO appears to be mediated by inflammation and alterations in osmohomeostasis rather than VEGF inhibition, which may have implications for more specific treatment modalities in the acute phase of BRVO.

Introduction

Retinal vein occlusion (RVO) is a common, variable, ischemic retinopathy that may affect either a branch retinal vein (BRVO) or the central retinal vein (CRVO). It is associated with retinal ischemia accompanied by the upregulation of vascular endothelial growth factor (VEGF), which is thought to play a major role in the pathogenesis of the disease.¹ Several other inflammatory cytokines are also upregulated,² resulting in the breakdown of the blood–retinal barrier (BRB) and the efflux of fluid and lipid exudates from permeable vessels. These changes lead to the development of macular edema (ME), which is the main cause of visual impairment following BRVO. 

Anti-VEGF agents, including ranibizumab, aflibercept, and bevacizumab (BEV), have mostly replaced other therapies as the main treatment modalities for RVOs. Although anti-VEGF agents have achieved significantly better levels of vision compared to those seen in the natural history arms, the vision of only a few recipients is restored to normal levels.³ Although they are very effective in the treatment of ME by reducing permeability for a limited time, they are less effective in reversing or ameliorating other changes that occur, such as the regulation of several inflammatory cytokines and factors and the consequent cell death that plays a critical role in the pathogenesis of BRVO.¹ 

We have previously demonstrated that apoptosis of retinal cells occurs rapidly following BRVO and is accompanied by changes in the regulation of inflammatory cytokines and factors and the breakdown of osmohomeostasis.⁴ In a later study, we further demonstrated increased neurodegeneration in pig BRVO retinas treated with the anti-VEGF agent BEV or the glucocorticoid triamcinolone acetonide (TA).⁵ In this current study, we further examined BRVO-induced alterations in the expression of a panel of 11 retinal genes that have various pathological effects associated with inflammation, angiogenesis, homeostasis, and cell death in drug-treated (BEV and TA) pigs and untreated pigs. 

Materials and Methods

Animals

Animal procedures were approved by the Animal Ethics Committee of the University of Western Australia in accordance with the National Health and Medical Research Council’s Australian Code of Practice for the Care and Use of Animals for Scientific Purposes (8th edition). They were adapted from the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and the policies in the “Guide to the Care and Use of Laboratory Animals” issued by the National Institutes of Health (Bethesda, MD). We have used the pig model in this and previous studies, as it has a retinal architecture similar to that of humans and replicates the features in BRVO.^4–8 

Surgery and Tissue Processing

Ten-week-old pigs (Sus scrofa; Landrace, Large White, Duroc) with a body weight of ∼30 kg were used in this study. Surgery was performed as described previously.⁵ Briefly, the pigs were anesthetized and then intubated, ventilated, and maintained with isoflurane (0.5%–2% ATTANE isoflurane, 1 mL/mL; Bayer Australia, Prymble, NSW, Australia) in oxygen. The eyes in all pigs underwent baseline assessment using binocular indirect ophthalmoscopy to exclude pre-existing retinal disorders. Three pigs (six eyes) were were used as the healthy control group (no interventions). There were four BRVO groups (2-, 6-, 10-, and 20-day time points) of nine pigs each. An intravenous injection was administered via an ear vein of 10 mg/kg of Rose Bengal dye (Sigma-Aldrich, St. Louis, MO), which is a dye with peak absorption of light close to the wavelength of the argon laser. A photothrombotic BRVO was induced in an inferior vein adjacent to the optic disc in both eyes of the four groups using an argon green laser, 532-nm wavelength (Ellex Medical Lasers, Adelaide, SA, Australia) as described previously.^4–8 Six pigs from each group received an intravitreal injection of TA (KENACORT-A 40; Aspen Pharmacare, St. Leonards, NSW, Australia) in one eye chosen randomly and BEV (Baxter Australia, Old Toongabbie, NSW, Australia) in the fellow eye immediately after a BRVO was created, with the remaining three pigs (six eyes) remaining untreated. 

A two-step technique was used for euthanasia as described previously.⁵ Immediately after euthanasia, the eyes were enucleated, and the anterior segments were removed. Two pieces on either side of the laser burn within the area of the BRVO were dissected and used for this study, as for the previous study.⁵ An 8-mm-diameter corneal trephine was used to cut out one of the pieces for quantitative reverse-transcription polymerase chain reaction (RT-qPCR). The other piece, approximately 2 × 3 mm in dimension, was bisected. One-half was fixed in 4% paraformaldehyde overnight at 4°C and processed in paraffin for immunohistochemistry, and the other half was fixed in 2.5% glutaraldehyde in 0.1-M phosphate buffer and processed in epoxy resin. An automated tissue processing machine (TP1020; Leica, Wetzlar, Germany) with an adapted set program⁹ was used to process tissue in paraffin, and the procedure was consistent for all samples. 

Quantitative Reverse-Transcription Polymerase Chain Reaction

RT-qPCR was used to analyze the gene expression of cytokines and factors. These included VEGF and stromal cell-derived factor-1α (SDF-1α) encoded by the CXCL12 gene, both of which are involved in angiogenesis; glial fibrillary acidic protein (GFAP); the specialized integral membrane-associated protein aquaporin-4 (AQP4) and the inwardly rectifying potassium channel K_ir4.1 (encoded by the KCNJ10 gene), closely associated with retinal edema; dystrophin protein (Dp; DMD gene), associated with BRB stability; the inflammatory cytokines interleukin-6 (IL-6) and interleukin-8 (IL-8); monocyte chemoattractant protein-1 (MCP1), encoded by CCL2, a potent chemoattractant to monocytes; intercellular adhesion molecule 1 (ICAM1), a promoter of microvascular endothelial injury; and heat shock transcription factor 1 (HSF1), a stress-responsive protein that preserves ganglion cells and their function. 

As in our previous study,⁴ the neural retinas were peeled from the discs and stored in RNAlater (QIAGEN, Hilden, Germany) at 4°C. After 1 week, they were transferred to a –20°C freezer. Total RNA was extracted using the RNeasy Plus Micro Kit and QIAshredder (QIAGEN). RNA purity was assessed with a NanoDrop 2000 Spectrophotometer (Thermo Fisher Scientific, Waltham, MA). First-strand cDNA was synthesized using 0.5 to 1 µg of total RNA and the RT2 First Strand Kit (QIAGEN) according to the supplier's protocol. Each RT-qPCR reaction was performed in a 25-µL mixture containing 1 µL (15–20 ng) of cDNA, a pair of porcine-specific primers (0.4 µM), and RT² SYBR Green qPCR Master Mix (QIAGEN). The QIAGEN primers used in RT-qPCR were VEGF (PPS00495A), CXCL12 (PPSO1281A), IL6 (PPS00991A), IL8 (PPS00237A), GFAP (PPS00012A), KCNJ10 (PPS07806A), AQP4 (PPS01854A), CCL2 (PPSOO734A), ICAM1 (PPSOO151A), HSF1 (PPSO9638A), and GAPDH (PPS00192A). DMD primers (pig DMD-F 5′ CCTCGGTTCAAGAGCTATGC 3′ and pig DMD-R 5′ TCCAACAATGAACTGCCAAA 3′; Integrated DNA Technologies, Coralville, IA) were also used. RT-qPCR was performed on the CFX Connect Real-Time PCR System (Bio-Rad, Hercules, CA), using the following thermocycling conditions: 95°C for 10 minutes, 40 cycles of 95°C for 10 seconds, and 60°C for 20 seconds. The PCR threshold cycle (C_t) values were determined using Bio-Rad CFX Manager 3.0 software. In all experiments, the relative gene expression levels were normalized against the pig housekeeping gene GAPDH and expressed as mean fold changes compared with the healthy control group using the ΔΔCt method.¹⁰ 

Immunohistochemistry

Sections 6 µm thick were sectioned on a Leica 2040 microtome. After deparaffinization, sections underwent antigen retrieval (permeabilized), and immunofluorescence staining for VEGF was performed as described previously.⁴ Briefly, the sections were blocked with serum followed by incubation with the primary antibody polyclonal rabbit anti-human VEGFA (Proteintech, Rosemont, IL) at 1:150 dilution. After 2 hours at room temperature or overnight incubation at 4°C, the sections were incubated again for 1 hour with the relevant fluorescence Alexa Fluor 546-conjugated secondary antibody (Thermo Fisher Scientific). After each incubation, the sections were washed with 3 × 3-minute phosphate-buffered saline (PBS). Slides were mounted with Hydromount Nonfluorescing Mounting Media (National Diagnostics, Charlotte, NC). In addition, sections with the omission of the primary antibody and with isotype control as negative controls were also performed. The images of the sections containing the distribution of the labeled protein were captured on an epifluorescent microscope equipped with fluorescence-relevant detection filter (excitation/emission maxima 556/573 nm) at a preset setting on the microscope at 20× magnification (Eclipse E800; Nikon, Tokyo, Japan). Semiquantitative analysis of VEGF protein expression was performed. 

Immunohistochemistry is a useful method to analyze protein expression and localization within tissues.¹¹ It has been extensively used for protein quantification in both diagnostic and clinical pathology and basic research.^12–14 Semiquantitative manual grading can introduce a level of subjectivity. Therefore, we performed semiquantitative analysis of VEGF immunohistochemistry using the semiautomatic image analysis system ImageJ (National Institutes of Health, Bethesda, MD) to determine the levels of VEGF protein, as bevacizumab selectively binds to free VEGF and consequently reduces the availability of VEGF to bind to the primary antibody used in immunohistochemical labeling. Images were captured at preset settings on the microscope that included sensitivity, light intensity, and position of all filters. Deconvolution (conversion of a color image to a gray image and set thresholding for all images) of the immunohistochemical images was followed by quantifying the final integrated density (product of area and mean gray value) after correction for background (area away from the retina) staining (Fig. 1). 

Figure 1.

 

Deconvolution of immunohistochemistry image of VEGF. (A) VEGF immunohistochemistry (IHC) of normal retina. (B) Image converted to grayscale. (C) Post threshold for analysis of integrated intensity.

Statistical Analysis

All data was analyzed using SigmaStat 3.5 (Systat Software, San Jose, CA). RT-qPCR results were analyzed using t-test between groups (normal, untreated, TA-treated, and BEV-treated). Six samples (six eyes from three pigs in the normal and untreated groups and six eyes from six pigs in the treated groups) were used in each group at each time point (2, 6, 10, and 20 days post-BRVO), with the exception of three eyes (two pigs) in the treated BRVO groups that were excluded from analysis due to low RNA quality (GAPDH C_t values > 26). Immunohistochemistry data for VEGF was also analyzed with t-test and Mann–Whitney rank-sum test between groups. Five samples (eyes) from three pigs in the normal and untreated and five samples from five pigs in each treated group at each time point were used. For each eye, the integrated intensities were measured from four randomly selected 20× magnification images, each showing an area of 0.01 mm², using Image J 1.45. P < 0.05 was considered statistically significant. Data are expressed as group means ± SEM. 

Results

Branch Retinal Vein Occlusion

As in our previous studies,^4–6 engorgement of the distal vein and scattered intraretinal hemorrhages were seen directly after photocoagulation of the pig eyes, showing that an occlusion had been created. On macroscopic examination of the enucleated eyecups, the laser burn was distinguishable as a white area in which a vein occlusion was induced. 

Changes in Retinal mRNA Expression

Effect of BRVO on Gene Expression

To investigate gene expression changes following BRVO, we chose a panel of 11 genes previously shown to respond to BRVO. We measured the mRNA expression levels of these genes in retinal regions immediately adjacent to the BRVO. For each gene, expression values were normalized to GAPDH mRNA levels and expressed as fold changes compared with healthy control eyes (day 0 in all graphs). 

Four genes showed significant (>10-fold) upregulation in response to BRVO (highly upregulated). Following BRVO, untreated BRVO retinas showed significant upregulation of IL8, GFAP, CCL2, and ICAM by day 10 post-BRVO (290-fold, 84-fold, 110-fold, and 16-fold, respectively; P < 0.05). For these genes, expression peaked at day 10 and remained high at day 20; however, only GFAP upregulation remained significant at day 20 (Fig. 2A). 

Figure 2.

 

Graph of RT-qPCR (normalized against the housekeeping gene GAPDH) showing the regulatory effect of BRVO on the mean value of the expression of pig retinal genes. (A) Highly elevated levels in IL-8, GFAP, AQP4, CCL2, ICAM1, and HSF1. (B) Moderately upregulated genes VEGFA, KCNJ10, and CXCL12. (C) Non-significant changes in IL6 and DMD at 2, 6,10, and 20 days post-occlusion. Significant difference are indicated by a black asterisk (*).

Two genes, HSF1 and AQP4, showed moderate (<threefold) upregulation following BRVO (moderately upregulated). Compared with healthy controls (day 0), untreated BRVO eyes showed a significant induction of HSF1 at day 2 (2.9-fold) and at day 10 (1.8-fold; P < 0.05), returning to healthy control levels by day 20. In contrast, AQP4 expression showed a non-significant reduction at early time points but was significantly increased at 20 days post-BRVO compared with healthy controls (2.3-fold; P < 0.05) (Fig. 2A). 

Three genes showed significant reductions in expression in response to BRVO (downregulated). Expression of VEGF, KCNJ10, and CXCL12 was significantly reduced by 6 days post-BRVO, with expression recovering to healthy control levels by day 20 (Fig. 2B). No significant changes in DMD or IL6 expression were detected following BRVO (Fig. 2C). 

Effect of Bevacizumab and Triamcinolone on Gene Expression Following BRVO

Highly Upregulated Genes

BRVO induced a large, rapid upregulation of IL8 expression in both treated and untreated groups, with significant elevation of IL8 expression in all groups at 10 days post-BRVO (P < 0.05) compared with healthy controls. In the BEV treatment group, IL8 expression was significantly upregulated at all time points compared with healthy controls. Compared with the untreated BRVO groups, the BEV treatment group showed significantly increased IL8 expression at 10 days post-BRVO (P < 0.05). In contrast, the TA treatment group showed an earlier, non-significant peak in IL8 upregulation at 6 days, followed by a significant reduction in IL8 expression at 10 days, compared with the untreated BRVO group (P < 0.05). IL8 expression was significantly reduced in TA-treated eyes at 20 days post-BRVO compared with BEV-treated eyes (P < 0.05) (Fig. 3A). 

Figure 3.

 

Graph of RT-qPCR (normalized against the housekeeping gene GAPDH) showing the mean value of the expression of IL8 (A) and GFAP (B) at the same time point, where # indicates a significant difference from untreated BRVO, $ indicates a significant difference from TA, and * indicates a significant difference from healthy controls.

GFAP expression was significantly increased in all BRVO groups at all time points compared with the healthy control group (P < 0.05), peaking at day 10 and remaining elevated at day 20 post-BRVO (Fig. 3B). Treatment with BEV or TA had no significant effect on GFAP upregulation following BRVO over this time period. 

Compared with healthy control eyes, untreated BRVO eyes showed significant upregulation of ICAM expression from days 2 to 10 post-BRVO (6- to 16-fold; P < 0.05). No significant differences in ICAM expression were found between treated and untreated BRVO eyes at any time point. Upregulation of ICAM expression was similar in the untreated and BEV-treated BRVO groups, with peak expression at 10 days post-BRVO. However, TA treatment altered the kinetics of ICAM induction following BRVO, as there was no significant increase at day 2 compared with healthy controls, and an earlier peak expression was observed at day 6. At days 10 and 20 post-BRVO, the TA-treated group had the lowest ICAM expression levels; however, these differences did not reach significance when compared with either the untreated or the BEV-treated groups (Fig. 4A). 

Figure 4.

 

Graph of RT-qPCR (normalized against the housekeeping gene GAPDH) showing the mean value of the expression of ICAM1 (A) and CCL2 (B) at the same time point, where # indicates a significant difference from untreated BRVO, $ indicates a significant difference from TA, and * indicates a significant difference from healthy controls.

CCL2 expression was increased in response to BRVO, with peak upregulation in the untreated group at 10 days post-BRVO (110-fold; P < 0.05). By day 20, CCL2 expression in the untreated group was not significantly different from the healthy control group. In contrast, BRVO-induced CCL2 upregulation was suppressed by TA treatment. Compared with untreated BRVO eyes, TA-treated BRVO eyes showed significant reductions in CCL2 expression at days 2 and 10 post-BRVO. The TA-treated group showed a significant 90% reduction in CCL2 expression at day 2 post-BRVO compared with the healthy control group (P < 0.05), which recovered to control levels by day 6. In contrast, BEV-treated eyes showed a significant upregulation of CCL2 at day 6 post-BRVO compared with healthy controls. BEV treatment did not significantly alter CCL2 upregulation compared with untreated BRVO eyes, although mean CCL2 expression values were lower in the BEV-treated group from days 2 to 10 post-BRVO (Fig. 4B). 

Moderately Upregulated Genes

Compared with healthy control eyes, untreated BRVO eyes showed a modest, non-significant reduction in AQP4 expression at 2 days, followed by a significant, 2.3-fold increase in expression by 20 days post-BRVO. In contrast, both BEV and TA treatments significantly reduced AQP4 expression at 2 days by 69% and 82%, respectively, compared with healthy controls (P < 0.05). AQP4 expression remained significantly reduced at day 20 compared with both healthy controls and untreated BRVO eyes (P < 0.05) (Fig. 5A). 

Figure 5.

 

Graph of RT-qPCR (normalized against the housekeeping gene GAPDH) showing the mean value of the expression of AQP4 (A) and HSF1 (B) at the same time point, where # indicates a significant difference from untreated BRVO, $ indicates a significant difference from TA, and * indicates a significant difference from healthy controls.

Following BRVO, untreated and TA-treated eyes showed an early, modest upregulation of HSF1 expression, peaking at day 2 with 2.9-fold and 2.6-fold increases, respectively, and returning to healthy control levels by day 20. In contrast, BEV treatment delayed BRVO-induced HSF1 upregulation. At 2 days post-BRVO, HSF1 expression was significantly reduced in BEV-treated eyes compared with untreated BRVO eyes (P < 0.05). At 6 days post-BRVO, HSF1 expression was significantly reduced in BEV-treated eyes compared with TA-treated eyes (P < 0.05). At 10 days, HSF1 expression was significantly increased in BEV-treated eyes (2.3-fold) compared with healthy control eyes but was not significantly different in untreated eyes (1.8-fold) or TA-treated eyes (1.8-fold; P < 0.05). Furthermore, HSF1 expression remained significantly elevated in BEV-treated eyes at day 20 (1.7-fold) compared with healthy controls (P < 0.05) (Fig. 5B). 

Downregulated Genes

In untreated BRVO eyes, we observed a significant decrease in retinal VEGF expression at 6 and 10 days post-BRVO (P < 0.05), which recovered to healthy control levels by day 20. In contrast, BEV-treated eyes showed significant upregulation of retinal VEGF expression at 2, 10, and 20 days post-BRVO (P < 0.05). TA treatment prevented BRVO-induced VEGF downregulation at days 2 and 6, and significantly increased VEGF expression at 10 days post-BRVO compared with healthy controls (P < 0.05), returning to control levels by day 20 (Fig. 6A). 

Figure 6.

 

Graph of RT-qPCR (normalized against the housekeeping gene GAPDH) showing the mean value of the expression of VEGFA (A) and KCNJ10 (B) at the same time point, where # indicates a significant difference from untreated BRVO, $ indicates a significant difference from TA, and * indicates a significant difference from healthy controls.

Following BRVO, untreated eyes showed a transient, early decrease in KCNJ10 expression, with a significant 43% reduction in expression at 6 days post-BRVO compared with healthy controls (P < 0.05). At 10 and 20 days, KCNJ10 expression recovered and was not significantly different from healthy controls. In contrast, both treatments deepened the downregulation of KCNJ10 expression at early time points and prevented recovery of expression at later time points. KCNJ10 expression in BEV- and TA-treated eyes was significantly reduced from days 2 to 20 post-BRVO compared with healthy controls (P < 0.05). Compared with untreated BRVO eyes, both BEV and TA treatments significantly reduced KCNJ10 expression at days 10 and 20 post-BRVO (P < 0.05) (Fig. 6B). 

Similar to KCNJ10, CXCL12 expression showed early downregulation in untreated eyes in response to BRVO (P < 0.05), followed by recovery to healthy control levels by days 10 to 20. Treatment with BEV or TA did not significantly alter CXCL12 expression compared with the untreated BRVO group (P > 0.05). Mean CXCL12 expression levels remained low in the TA-treated group and were significantly reduced at day 20 compared with BEV-treated eyes (Fig. 7). 

Figure 7.

 

Graph of RT-qPCR (normalized against the housekeeping gene GAPDH) showing the mean value of the expression of CXC12 at the same time point, where # indicates a significant difference from untreated BRVO, $ indicates a significant difference from TA, and * indicates a significant difference from healthy controls.

Following BRVO, untreated eyes showed no significant changes in DMD or IL6 expression compared with healthy controls. In contrast, both BEV and TA treatments significantly increased DMD and IL6 expression at all time points compared with untreated BRVO eyes (P < 0.05) (Figs. 8A, 8B). Although there were no significant differences in DMD or IL6 expression observed between the treatments, TA-treated eyes showed a large (274-fold), transient increase in IL6 expression at day 6, but BEV-treated eyes showed an earlier, moderate (35-fold) peak in IL6 upregulation at day 2 (Fig. 8A). 

Figure 8.

 

Graph of RT-qPCR (normalized against the housekeeping gene GAPDH) showing the mean value of the expression of IL6 (A) and DMD (B) at the same time point, where # indicates a significant difference from untreated BRVO, $ indicates a significant difference from TA, and * indicates a significant difference from healthy controls.

Immunohistochemical Localization of VEGF

There was an absence of specific staining of VEGF in immunostaining controls, in which the primary antibody was omitted or replaced with an isotype control antibody. Immunohistochemical localization of of VEGF protein demonstrated a significant reduction in BRVO-untreated eyes at 6 and 10 days, in BEV-treated eyes at 2 and 10 days, and in TA-treated eyes at all time points compared to normal controls. At day 6, the BEV-treated eye was significantly higher than the untreated and TA-treated BRVO (Fig. 9). 

Figure 9.

 

Graph showing the mean and ± SEM of the integrated intensity of VEGF immunolocalization in normal, untreated, and drug-treated retina. Statistical significance (P < 0.05) is indicated as follows: healthy control versus untreated BRVO (#); healthy control versus BEV-treated BRVO ($); healthy control versus TA-treated BRVO (*); untreated BRVO versus BEV-treated BRVO (†); and TA-treated versus BEV-treated BRVO (Ø).

Discussion

We have previously shown that inner retinal neural cells are susceptible to apoptosis in the acute stages of BRVO, raising the possibility that the administration of potential neuroprotective agents may offer some benefit in limiting these changes.^4,5 There appears to be potential for repairing cells that are in a pre-apoptotic state.¹⁵ Reversal of apoptotic processes by DNA repair has been demonstrated in diabetic mice where leukostasis, cell death, and vascular permeability were abrogated.¹⁶ Furthermore, it has been shown that in some circumstances apoptosis can be reversed at a later stage despite caspase activation.¹⁵ 

We previously have demonstrated the generation of a BRVO in the pig model.^4,6,8,17 This model uses intravenous Rose Bengal dye as a photothrombotic agent allowing an intravenous thrombosis to occur and limiting the amount of laser irradiation to the vein. There is some evidence that this dye may induce a number of protein modifications and interactions as may also occur with excess laser irradiation. For the purposes of this study, as both the control and treatment arms had a photothrombotic BRVO with very limited laser irradiation, we do not believe this has influenced the results.^18,19 This occurs as an acute obstruction, which is uncommon in humans where the occlusion occurs typically at the arteriovenous crossing and progresses over a variable period of time. This model has the advantage of enabling accurate measurement of the time course of intraretinal molecular changes against the exact time point of the venous obstruction. We have demonstrated that there is an early loss of inner retinal neural cells, a significant number of which were in a pre-apoptotic or apoptotic state.^4,5 There was also an initial downregulation of VEGF mRNA, which may be due to the loss of VEGF-expressing cells. Consequently, we speculated that if intraretinal VEGF, which has neuroprotective properties,²⁰ is reduced in the very early phases of the vascular occlusion, anti-VEGF therapy, which is the favored treatment modality for RVOs, may potentially further exacerbate neuronal cell death by apoptosis. Supportively, we have shown that treatment with both BEV and TA during this acute phase is associated with an increased level of both apoptosis and pyknosis in the inner nuclear layer and the ganglion cell layer during the 2- to 20-day time points.⁵ In the current study, we have investigated the changes in retinal gene expression following an acute BRVO and the effects of VEGF suppression with either BEV or TA treatments. 

VEGF (VEGFA) is a 40-kDa dimeric glycoprotein²¹ that presents in several isoforms. VEGF plays an important role in vasculogenesis, neoangiogensis, vascular permeability, vasodilation, recruitment of inflammatory cells, and apoptosis inhibition.²² VEGF is present in healthy retina in the absence of retinal injury.²³ Immunolocalization of VEGF protein has revealed its expression by several retinal cell types, including retinal pigment epithelial cells, neurons, astrocytes, endothelial cells, ganglion cells, Müller cells, and smooth muscle cells, as well as in the inner and outer plexiform layers.²⁴ VEGF is secreted intercellularly and among these cells, and Müller cells are the major source of VEGF in response to hypoxia.^21,25 

In a BRVO, there is a variable degree of impaired venous flow resulting in retinal ischemia and hypoxia.²⁶ Under hypoxic conditions, hypoxia-inducible factor 1 binds to the hypoxia response element present in the VEGF gene, inducing the transcription of VEGF protein and synthesis.²⁷ VEGF is the most important cytokine implicated in ME in RVO, having the highest VEGF concentrations in these eyes.²⁸ VEGF promotes the synthesis and release of nitric oxide from endothelial cells, resulting in vasodilation, the production of inflammatory cytokines, and enhanced permeability of the endothelial cells due to the increasing phosphorylation of tight junction proteins occludin and zonula occludens-1 (ZO-1).^29–31 BEV has been shown to inhibit VEGF-induced vasodilatation in isolated perfused pre-contracted pig retinal arterioles.³² Retinal blood flow and velocity have also been shown to be reduced to a variable and usually minor degree transiently after intravitreal anti-VEGF therapy in some clinical studies but not in others. This has not been associated with any effect on final outcomes, and how this effect may have influenced the outcomes in this study is uncertain.^33–35 

BEV is a humanized anti-VEGF monoclonal IgG1 antibody (U.S. Food and Drug Administration Final Labeling BL 125085 Supplement 2008; The European Medicines Agency, European Public Assessment Report, Bevacizumab Product Information H-C-582-11-23) that binds to circulating VEGF protein, inhibiting the binding of VEGF to its cell surface receptor VEGFR1.³⁶ On the other hand, TA, a corticosteroid, inhibits VEGF expression through binding to glucocorticoid receptors to allow its translocation into the nucleus to block activator protein 1 (AP-1) and nuclear factor kappa B (NF-κB), thus altering histone acetylation, methylation, and phosphorylation and resulting in repressed VEGF transcription, and by increased degradation of VEGF mRNA.^37–39 

In the BEV-treated retina, there is likely to be binding of BEV to VEGF intercellular protein, which may affect the amount of the intercellular protein available to attach to the primary antibody used in the immunohistochemistry assay. It is unlikely that BEV can enter the cell to have an effect on intracellular VEGF protein, although some studies have reported the internalization of BEV in cancer cells. Karpinska et al.⁴⁰ used single-molecule spectroscopy to demonstrate that the uptake of BEV in intracellular vesicles occurs by endocytosis but does not result in drug release into the cytosol in cervical and breast cancer cells, suggesting that attachment of the drug to VEGF protein within the cell is unlikely. RT-qPCR quantitates the VEGF mRNA levels within cells and it does not bind to BEV. 

We observed an early downregulation of retinal VEGF by both immunohistochemistry signal and mRNA expression after untreated BRVO in the pig with slow recovery toward healthy control levels by day 20 in this study, with similar changes seen in our previous study for the untreated BRVO.⁴ 

Reduced VEGF has also been seen in the aqueous and vitreous of patients and in other animal studies.^41–44 In the current study, VEGF mRNA was unexpectedly upregulated in the TA- and BEV-treated BRVO compared with both normal healthy eyes and also with untreated BRVO retina during the course of the acute phase. In this study, retinal VEGF mRNA levels were measured and were not an estimation of either intercellular or free VEGF within the aqueous and vitreous fluid, as seen in most other studies of VEGF expression following BRVO. BEV binds to circulating or free VEGF protein,³⁶ whereas TA reduces VEGF mRNA transcription.^45,46 The loss of a potential feedback mechanism through continued production of hypoxia-inducible factor 1 due to reduced intercellular levels of VEGF in the BEV eyes where some is bound to the antibody may explain these higher intracellular levels in the treated versus untreated BRVO. Although BEV does not appear to enter the cells, TA may have a delayed effect as seen in the reduced levels compared to BEV by day 20 (Fig. 6A). There is a complex arrangement between VEGF levels in the parenchyma and intracellular secretion mediated possibly by the cell membrane VEGF receptors.⁴⁵ 

Immunohistochemical analysis of VEGF protein localized both intracellularly and extracellularly and was in general reduced for both the untreated and treated BRVO eyes, with the exception of BEV at 6 days, where it was significantly higher than untreated and TA-treated BRVO but similar to normal healthy controls. The 6-day peak in BEV-treated BRVO VEGF protein may be driven by the earlier 2-day peak in mRNA expression. This peak in VEGF protein suggests that BEV is not interfering with the detection of VEGF by immunohistochemistry (Figs. 6A, 9). 

Complex interactions between cytokines can influence each other and induce changes in their regulation, thus disrupting homeostasis following a BRVO. VEGF induces the synthesis of ICAM1 and the production of MCP1 chemotaxis and margination.⁴⁷ In this study, ICAM upregulation following BRVO was not correlated with VEGF mRNA levels in the untreated group, suggesting a VEGF-independent mechanism of ICAM induction. 

MCP1 is a potent chemoattractant for monocytes, and their aggregation to endothelial cells causes microvascular endothelial damage resulting in vascular permeability. Dominguez et al. (Dominguez E, et al. IOVS. 2013;54:ARVO E-Abstract 100) demonstrated that BRVO induced MCP1 expression and retinal macrophage infiltration in a mouse model of laser induced BRVO. CCL2 mRNA expression (encoding MCP1) was strongly induced in the pig retina by BRVO, peaking with a 110-fold increase at 10 days and returning to control levels by day 20. Suppression was not correlated with VEGF expression in the early stages compared to VEGF for the untreated BRVO in this study. In general, TA was more efficient than BEV in reducing levels of CCL2, especially at the 10-day time point. BEV targets VEGF, whereas TA affects expression of a range of cytokines. 

ICAM1 mRNA was significantly upregulated in treated and untreated BRVO compared to normal retinas, although the upregulation was delayed in the TA group, reaching a peak at 6 days before reducing to lower levels compared to BEV and untreated BRVO. Tatar et al.⁴⁸ demonstrated that TA has no inhibitory effect on macrophage infiltration or ICAM1 in choroidal neovascular membranes in the early term. They attributed the increase in ICAM1 to enhanced IL-1 and TNF-α activity induced by amplified macrophage density enhancing ICAM1 production. ICAM1 is upregulated by hypoxia and also plays an important role in leucocyte recruitment, increased rolling and adhesion to the endothelium resulting in stagnation of blood flow.¹ 

SDF-1 (encoded by CXCL12) plays an important role in hematopoiesis. It has been suggested that SDF-1 is produced by Müller cells, astrocytes, microglial cells, and endothelial cells. SDF-1 has been shown to have neuroprotective properties⁴⁹ and to suppress apoptosis in cultured dendritic cells. Blockade of SDF-1 in retinal detachment increased photoreceptor cells loss. SDF-1 has been shown to play a major role in proliferative retinopathies and is upregulated in the vitreous in eyes with ocular neovascularization secondary to diabetic retinopathy and RVO disease.^2,50 However, elevated vitreous levels of SDF-1 were only found in those eyes with neovascularization, whereas the levels of SDF-1 were similar to those of control eyes in conditions without intraocular neovascularization. In the present study, SDF-1 mRNA was not upregulated in the untreated BRVO at any time point. There was no significant difference in the expression of SDF-1 mRNA in either treated group of BRVO with the exception at 20 days, when a significant increase was seen in the BEV- compared to TA-treated eyes. Overall, it would appear that this cytokine is more involved in eyes when neovascularization develops and prior to this has little role in the early stages of BRVO in the pig. 

IL-6 is a multifunctional cytokine playing an important role in development differentiation and regeneration of neurons; however, overproduction can lead to neuronal degradation.⁵¹ IL-6 induces the expression of MCP1 and VEGF.^52,53 It has been classified as an inducer of phagocytic and cytotoxic activity.⁵⁴ Noma et al.⁵⁵ demonstrated that there is a significant correlation of IL-6 and VEGF with severity of ME in BRVO. In addition to MCP1, IL-6 attracts leukocytes by increasing ICAM1 and vascular cell adhesion molecule (VCAM1)⁵⁶ to promote adhesion to the vascular endothelium and migration through the vessel wall. IL6 mRNA was elevated in both of the drug-treated BRVO groups compared to untreated BRVO eyes, where it was not upregulated, at all time points, thus raising the possibility of its involvement in the neurodegeneration seen in our previous study.⁵ IL-8 is induced by VEGF.⁵⁷ Cytokine levels of both IL-6 and IL-8 were significantly upregulated in aqueous humor and in the vitreous in patients at several months after the onset of BRVO.^43,58 Sohn et al.⁵⁹ demonstrated no significant changes in IL-6 or IL-8 in the aqueous humor of BRVO patients treated with BEV and no significant changes in the TA treated of IL-8, whereas IL-6 was significantly downregulated 4 weeks after drug treatment. Although this information is useful, the levels in intraocular fluids months after the occlusive event most likely bear little relevance to the intraretinal changes at the time of the occlusion itself, and levels of cytokines in the vitreous or aqueous humor may not reflect the levels in retinal tissue. Little is known of the permeability of these proteins and whether they are bound to any extent in the retina, limiting the rate of diffusion into the vitreous and aqueous fluids. It has been suggested that some cytokines can bind to extracellular matrix whereby their effects are localized to specific areas or stored in their matrix.⁶⁰ The inner limiting membrane can also act as a diffusion barrier to mediators such as VEGF.⁶¹ Both of these cytokines are implicated in endothelial permeability¹—IL-6 by inducing the formation of gap junctions⁶² and IL-8 by downregulation of tight junctions.⁶³ Chu et al.⁶⁴ demonstrated the elevation of IL-6 and IL-8 after intravitreal BEV in cultured RPE cells. IL8 mRNA was generally upregulated in the untreated BRVO retina with a peak of 280-fold at day 10 with little difference between this and the BEV-treated eyes compared to normal eyes. The TA-treated eyes showed a significant reduction in IL8 expression compared to untreated BRVO eyes in the later stages. It did not appear to be correlated with VEGF expression in this study, at least in these early stages. IL-8 is associated with the recruitment of inflammatory cells, including neutrophils and macrophages, through CXCR2 (interleukin-8 receptor β [IL-8Rβ]) and induces VEGF expression.^65–68 It has been suggested that IL-8 is the main inflammatory factor involved in ME associated with BRVO. IL-8 is neurotoxic, and its overproduction could lead to cell death. The higher levels of IL8 mRNA in the BEV-treated BRVO at the 20-day time-point may have contributed to the significantly higher levels of apoptosis compared to TA-treated BRVO. It appears that, although these cytokines, which combat injury when upregulated, could result in detrimental effects contributing to cell death as seen in our previous study.⁵ 

Müller cells show one of the earliest and most sensitive manifestation by reactive gliosis to retinal injury expressed by the upregulation of GFAP⁴¹ and the production of cytokines.⁶⁹ Gliosis is initially neuroprotective; however, excessive production can result in adverse effects on neuronal and photoreceptor cell viability.⁷⁰ Significantly increased mRNA levels were seen in both the treated and untreated BRVO retina compared to normal retina, and there was no significant difference between them, suggesting that TA and BEV have minimal effects on controlling gliosis in BRVO in the acute phase. However, we have previously shown that TA does downregulate GFAP in this model of BRVO by 11 weeks.⁶ 

Müller cells also play a vital role in maintaining retinal osmohomeostasis and are richly endowed with proteins such as dystrophin, K_ir4.1 and AQP4, which play a crucial role essential for the maintenance of the BRB.^71–73 The DMD gene (encoding dystrophin) is ubiquitously expressed and produces several alternative isoforms, of which Dp71 is the major isoform present in the central nervous system of humans and pigs. In the retina, dystrophin Dp71 is localized to the glial perivascular end feet of Müller cells with the same distribution pattern and clustering of K_ir4.1 and AQP4.^74,75 It is important for the preservation of retinal functional integrity supporting photoreceptor and ganglion cell survival. It has also been demonstrated that the inactivation of Dp71 in Dp71-null mice⁷⁶ and its reduction in an experimental mouse model of retinal detachment decreased the expression of AQP4 and induced the redistribution of K_ir4.1, leading to BRB disruption.^77–79 This would have dire consequences, as downregulation and mislocation of K_ir4.1 would lead to an inability to extrude K⁺, intracellular accumulation of K⁺, increased intracellular osmotic pressure, and an influx of water from the blood into the cells, culminating in glial cell swelling and impairment of fluid (metabolic water) clearance.⁷⁹ In our study, DMD mRNA was significantly elevated in the treated BRVO retinas at all time points compared to untreated and normal retinas. With the DMD mRNA being elevated in the treated retinas, it would be expected that K_ir4.1 mRNA (encoded by KCNJ10) levels would also be upregulated. These were, however, significantly downregulated at the 6-day time point during the acute phase in the untreated BRVO and in the drug-treated retinas at all time points compared to the normal retinas. There was also significant downregulation in the drug-treated BRVOs compared to the untreated BRVOs at 10 and 20 days post-occlusion. A possible explanation for this is that inflammatory cytokines may downregulate K_ir4.1. AQP4 mRNA levels in the untreated BRVOs were not significantly changed compared to the normal retinas with the exception of the 20-day time point, which showed significantly elevated levels. The treated BRVOs showed a general reduction in levels compared to both untreated and normal retinas. This may be due to those AQP4 pools, which are independent of dystrophin, not being affected by the preservation of dystrophin in the drug-treated BRVO retinas.⁷⁹ The interaction of AQP4 pools with proteins, such as the dystrophin-associated complexes, are not clearly understood. Puwarawuttipanit et al.⁸¹ suggested that, unlike K_ir4.1, the membrane anchoring protein syntrophin of AQP4 is insensitive to ischemia. Pannicke et al.⁷¹ also demonstrated that AQP4 was largely unaltered in the early stages of a rat model of retinal transient ischemic reperfusion, whereas Müller cells within hours significantly downregulated K_ir4.1 channels, swelled, and became reactive, resulting in ganglion cell death. 

Heat shock proteins are essential constituents of a complex defense mechanism in cell survival under adverse conditions, including hypoxia and ischemia, and heat shock factors are required for the expression of heat shock proteins.^82,83 In vertebrates, the HSF1 gene encodes the main factor responsible for stress induced protein regulation. Interestingly, it has been recently recognized as a regulator of Dp71 expression.⁸⁴ HSF1 has a dual role that can be either cytoprotective or lethal.^85–87 It is critical in maintaining protein homeostasis and refolding denatured proteins and can suppress apoptotic pathways; however, when tissue injury is severe or prolonged it induces cell death. Liu et al.⁸⁸ demonstrated that HSF1 is neuroprotective in retinal ischemic reperfusion in a mouse model, and boosting HSF1 may limit neurodegeneration. In this study, HSF1 mRNA was upregulated in the early stages for both untreated and TA-treated retinas, whereas the BEV-treated retinas showed a delayed upregulation. It would appear that from these results there is little influence on HSF1 levels from either drug. 

In our previous study, we demonstrated that treatment with BEV or TA increased retinal apoptosis following BRVO in the pig,⁴ suggesting that these drugs may exacerbate retinal damage in the acute phase. We identified significant effects of these drugs on gene expression following BRVO. Notably, both TA and BEV treatments led to apparent reductions in BRVO-induced CCL2 upregulation and prevented BRVO-induced VEGF downregulation. BEV treatment induced a moderate twofold upregulation of retinal VEGF mRNA expression at 2 days post-BRVO which remained elevated at 20 days. TA induced a smaller increase in VEGF mRNA expression at 10 days post-BRVO which returned to control levels by 20 days. However, in spite of these modest increases in VEGF mRNA expression in the drug-treated BRVO groups, retinal VEGF immunoreactivity was decreased in treated and untreated groups compared with healthy controls. 

Both BEV and TA treatments also prevented recovery of AQP4 and KCNJ10 expression, with each gene remaining significantly downregulated at 20 days post-BRVO. DMD expression, however, was increased by both BEV and TA treatments, which could potentially reflect a compensatory mechanism for reductions in AQP4 and K_ir4.1 protein expression, both of which co-localize with this dystrophin in Müller glial endfeet at the ILM.⁷⁶ The magnitude of peak BRVO-induced GFAP upregulation showed a non-significant trend toward higher levels of induction in the treated BRVO groups than the untreated groups. Moreover, IL6 expression was significantly upregulated by both BEV and TA treatments but not in the untreated group, suggesting exacerbation of the inflammatory response. Together, these observations suggest that early treatment with anti-VEGF agents worsens Müller glial and retinal responses to BRVO in the pig. In untreated BRVO, expression of several Müller glial markers showed a return to or toward healthy control levels by 20 days, suggesting an early, transient Müller glial response in the acute phase. BEV and TA may interfere with stabilization of Müller glial gene expression in the acute phase. 

In our previous study, both TA- and BEV-treated groups showed increased apoptosis and pyknosis in the inner retinal neural cells compared to the untreated BRVO group, although less so in the TA-treated group.⁵ Although we originally hypothesized that VEGF blockade may be responsible for this in the acute phase of BRVO, this would appear not to be the case. Rather, it appears from this study that both drug treatments may have this effect through upregulation of the inflammatory cytokine IL-6 and prevention of both the recovery of KCNJ10 expression and the upregulation of AQP4 following injury, with consequent effects on cellular osmohomeostasis. 

To the best of our knowledge, our research investigating neurodegeneration and cytokine regulation in retinal tissue in the pig model in acute BRVO as early as 2 days has not been performed previously. The findings from this study, where levels were examined from retinal samples at time points close to an acute ischemic event with administration of the drugs within a half-hour of the occlusion, would indicate that further research is indicated to elucidate the complex interactions between the induced cytokine and protein cascades and their effects on the health of retinal neural cells. 

This study does have significant limitations. The numbers in each time-point group are small, and the types of cytokines examined were not exhaustive. Cytokine dysregulation after ischemic retinal events is complicated, and it is likely that other interactions did occur that were not examined. The results, although of interest, do not bear much relationship to the clinical picture of BRVOs, where the occlusions typically tend to occur in a more gradual fashion in an older age group with underlying vascular anomalies. The results, especially those relative to VEGF suppression and neural retinal degeneration, may offer some reassurance. 

Acknowledgments

The authors thank Luke Jennings for technical assistance. 

Supported by a grant from the National Health and Medical Research Council (APP1173403 to D-YY) and by a National Health and Medical Research Council Investigator Grant (MRFF1142962 to FKC). 

Disclosure: I.L. McAllister, BAbbVie (C), Bayer (C); S. Vijayasekaran, None; S. McLenachan, None; R. Bhikoo, None; F.K. Chen, None; D. Zhang, None; E. Kanagalingam, None; D.-Y. Yu, None