Retinal Microglial Activation Following Topical Application of Intracellular Toll-Like Receptor Ligands

Abstract

Purpose: We previously have reported that application of the intracellular toll-like receptor (TLR)-9 ligand CpG-ODN onto the injured corneal surface induces widespread inflammation within the eye, including the retina. We tested the hypothesis that topical application of two other intracellular TLR agonists, Poly I:C and R848, would cause retinal microglial activation and migration into the subretinal space.

Methods: C57BL/6J wild-type and Cx₃cr1^gfp/+ mice were anesthetized and received central corneal abrasions followed by topical application of Poly I:C (TLR3 agonist), R848 (TLR7/8 agonist), or CpG-ODN (TLR9 agonist). Eyes were imaged in vivo by using spectral-domain optical coherence tomography to assess and quantify vitreous cells and retinal edema. Tissues were processed for whole-mount immunofluorescence staining or gene expression studies. Microglial activation was determined by morphologic changes, major histocompatibility complex (MHC) class II reactivity, and migration to the subretinal space. Expression of proinflammatory cytokine gene IL-6, IL-1β, IFN-γ, and MCP-1 in retinal tissues were analyzed.

Results: At 24 hours, topical treatment with CpG-ODN and R848, but not Poly I:C, led to altered microglial morphology. One week after CpG-ODN and R848-treatment, eyes exhibited vitritis and mild inner retinal edema, increased number of subretinal Iba-1⁺ cells, and an increase in MHC II⁺ cells in the neural retina. Proinflammatory cytokine genes were upregulated after R848 treatment, whereas in the CpG-ODN group, only IL-1β and MCP-1 were significantly upregulated. Retinal microglial activation was not observed in the Poly I:C–treated group.

Conclusions: Topical application of CpG-ODN and R848, but not Poly I:C, to the damaged corneal surface can cause activation and migration of retinal microglia.

Microglia are the resident macrophage cell population of the central nervous system including the retina.^1,2 There is compelling evidence that while microglial cell bodies are relatively static, their extensive ramifications are highly motile, constantly surveilling the local tissue environment.³ Microglia respond rapidly to noxious sterile stimuli or neural injury by changing the polarity of cell processes and phagocytosis of cell debris.^4,5 Upon activation in response to light injury or photoreceptor degeneration, retinal microglia lose their ramified processes, assume a rounded morphology, and migrate toward the subretinal space, a region normally devoid of resident immune cells.^6–8 These well-described morphologic changes indicative of microglial activation are accompanied by upregulation of major histocompatibility complex (MHC) class II⁹ and production of proinflammatory cytokines including IL-1β, IFN-γ, TNF-α, and MCP-1.² 

In common with other myeloid-derived cells, innate immune surveillance by microglia is partly dependant on the expression of a range of pattern recognition receptors including toll-like receptors (TLRs).¹⁰ Toll-like receptors recognize molecular patterns unique to bacteria, viruses, fungi, parasites,¹¹ as well as endogenous damage signals such as heat shock proteins and mitochondrial DNA released from injured cells.^12,13 Currently, 13 TLRs have been identified in humans and mice, with TLRs 1 to 9 being highly conserved between the two species.¹⁴ Toll-like receptors are grouped by their specific abilities to recognize conserved structural motifs of microorganisms and can be further divided into two groups: those located at the cell surface and others where intracellular processing occurs. Of those that are expressed on the host cell surface, TLR2 recognizes lipoproteins and peptidoglycans of yeasts and gram-positive bacterial cell walls.¹⁵ TLR4 is required for the detection of lipopolysaccharide (LPS)¹⁶ and TLR5 is the receptor for bacterial flagellin.¹⁷ In contrast, TLR3, TLR7/8, and TLR9, located intracellularly within the endosomes, sense microbial nucleic acids.¹⁸ TLR9 recognizes unmethylated CpG motifs, which are abundant in the genomes of many bacteria and viruses but are not present in eukaryotic nuclear DNA.¹⁹ TLR7/8 detects single-stranded RNA (ssRNA) from the genomes of ssRNA viruses and TLR3 acts as sensor for viral double-stranded RNA.²⁰ 

Toll-like receptor activation in the retina is predictably implicated in the host response to infectious microorganisms such as Staphylococcus aureus²¹ and Bacillus cereus.²² However, it is becoming increasingly apparent that TLR activation of innate inflammatory pathways is involved in a wide range of sterile retinal pathologies including photoreceptor degeneration²³ and ischemia reperfusion injury.^7,24 TLR3 ligands can directly induce apoptosis of retinal pigment epithelial cells in vitro, and genetic studies suggest a protective role for specific genetic variants of TLR3 in geographic atrophy.²⁵ We previously have reported that topical application of the TLR9 ligand CpG-ODN type B (a synthetic homologue of bacterial and viral DNA) onto the injured corneal surface induces widespread inflammation within the eye, including the retina.²⁶ By contrast, LPS (TLR4 ligand), while inducing an intense corneal inflammatory response, fails to cause posterior segment inflammation. In the present study, we sought to further examine this phenomenon and establish the inflammatory potential of other intracellular TLR ligands applied to the debrided cornea to induce retinal microglial activation. The data demonstrated that only ligands normally recognized by the intracellularly localized TLR7/8 and TLR9 induce widespread TLR-mediated microglial activation in the retina and migration of microglia to the subretinal space, leading us to propose a novel mechanism by which DNA and ssRNA released during corneal infections may have the unsuspected consequence of causing microglial activation and migration in the nearby retina. 

Methods

Animals

C57BL/6J mice (Animal Resources Centre, Murdoch, Western Australia) and Cx₃cr1^gfp/+ transgenic mice (Animal Research Laboratory, Monash University, Clayton, Victoria, Australia) were maintained on 12-hour day/night cycles with free access to food and water. TLR7^−/− mice were provided by Regeneron (Rensselaer, NY, USA). Cx₃cr1^gfp/+ mice contain an enhanced green fluorescent protein (GFP)–encoding cassette in the Cx₃cr1 locus, allowing transcription of GFP at the expense of one copy of Cx₃cr1. Cells that typically express Cx₃cr1 will therefore also express GFP, enabling monocyte-derived cells to be monitored in situ.²⁷ Genotyping was performed to confirm the lack of the Crb1^rd8 mutation in these mice.²⁸ Mice were maintained under specific pathogen-free conditions and procedures were performed in accordance with the guidelines of the Animal Ethics Committee at Monash University, Australia, and the Florey Institute of Neuroscience and Mental Health. All animal procedures conformed to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Mouse Model of Corneal Inflammation

Mice were anesthetized by intraperitoneal injection of ketamine/xylazine and the epithelium of the central cornea was debrided as previously described.²⁹ Immediately following epithelial debridement, 20 μg of either phosphorothioate CpG oligonucleotide 1826 (type B, TLR9 ligand), R848 (Resiquimod, TLR7/8 ligand), Poly I:C (low-molecular-weight TLR3 ligand) or sterile saline was applied to each eye. In the absence of corneal epithelial damage or debridement there was no response to ligands. Similarly, in previous studies²⁶ we have shown that scrambled CpG (i.e., GpC-ODN) does not elicit a corneal inflammatory response. All TLR ligands were purchased from Invivogen (San Diego, CA, USA). At 24 hours or 1 week after treatment, animals were euthanized and eyes placed into either 4% paraformaldehyde (PFA; Sigma-Aldrich, Castle Hill, New South Wales, Australia), RNAlater (Life Technologies, Carlsbad, CA, USA) or snap-frozen in optimal cutting temperature compound (Tissue-Tek; ProSciTech, Kirwan, Queensland, Australia) for tissue processing. 

Spectral-Domain Optical Coherence Tomography (SD-OCT)

At 1 week post treatment, mice were anesthetized and dilating drops were applied to the cornea (Mydriacyl 0.5%; Alcon Laboratories, Frenchs Forest, NSW, Australia). Mice were placed on the animal imaging mount and the rodent alignment stage was attached to the Bioptigen SD-OCT imaging device (Bioptigen, Inc., Durham, NC, USA). Once the eye was aligned with the noncontact mouse retina lens, any excess mydriatic drops were absorbed by using a cotton tip and saline applied once to prevent corneal drying. Volumetric 1.4 mm × 1.4 mm scans containing optic nerve head were acquired at a rate of 1000 A-scans per 100 B-scans. 

Quantification of Vitreous Cells in SD-OCT Scans

Vitreous opacities in SD-OCT images have been histologically verified in previous studies to be infiltrating inflammatory cells.^26,30 Posterior segment images were acquired by using the Envisu R-2200 OCT (Bioptigen, Inc.) and analyzed by using FIJI,³¹ an open-sourced platform. To quantify the inflammatory response, a custom FIJI script was written to automate the following process. Firstly, all images underwent binary conversion by using Yen thresholding.³² A median filter was then applied to remove one pixel–wide speckle noise. Next, a 15 × 50 pixel–wide region of interest was created one disc diameter from the center of the optic nerve on each side. All cells with a radius of two or more pixels within the region of interest were automatically segmented and counted by using the Analyze Particles function on FIJI. 

Quantification of the Retinal Nerve Fiber Layer (RNFL) Thickness

Thicknesses of individual retinal layers at multiple cross-sectional images were determined by using the Bioptigen segmentation software InVivoVue Diver. Manual calipers were used to mark distinct boundaries of the retinal sublayers from the superior, nasal, inferior, and temporal quadrants approximately 225 μm from the optic nerve head. The thickness of the RNFL was defined by the distance between the inner limiting membrane and the ganglion cell layer (GCL).³³ 

Real-Time Reverse Transcription PCR

Total RNA was extracted from retinal tissue by using the PureLink RNA Mini Kit (Invitrogen, Carlsbad, CA, USA) and subjected to on-column digestion of DNA by using PureLink DNase Set (Invitrogen). DNAse-treated RNA was reverse transcribed to cDNA, with each reaction containing 500 ng RNA per 10-μL reaction volume (Transcriptor First Strand cDNA Synthesis Kit; Roche Applied Sciences, Mannheim, Germany). Mock cDNA reactions, in which the enzyme reverse transcriptase was omitted (No-RT controls), were also generated in order to ensure the removal of genomic DNA had been successful. Synthesised cDNA served as template in real-time PCR reactions using Fast SYBR Green Master Mix (Applied Biosystems, Mulgrave, VIC, Australia). No-RT controls and No-Template controls were included in each PCR run. Reactions were carried out in a ViiA 7 Real-Time PCR system (Applied Biosystems), using the following conditions: 95°C for 20 seconds, followed by 35 cycles at 95°C for 30 seconds, 60°C for 30 seconds, and 72°C for 20 seconds. Expression of proinflammatory genes was analyzed with the 2^−ΔΔCt method, using the ribosomal protein L32 as a reference gene. Primer sequences are listed in the Table. 

Table

 

Primer Sequences of Proinflammatory Genes Analyzed by Real-Time RT-PCR

 

Immunofluorescence Staining of Retinal and Choroidal Tissue

Retinas from Cx₃cr1^gfp/+ mice, as well as retinas and choroids from C57BL/6J wild-type (WT) mice, were dissected and divided into radial portions to be used in whole-mount immunofluorescence staining, as previously described.³⁴ Tissue whole-mounts were incubated in 20 mM ethylenediaminetetraacetic acid for 30 minutes at 37°C and blocked in 0.3% Triton X-100, 3% bovine serum albumin in PBS for 30 minutes at room temperature before antibody incubation. Tissues were then incubated with rabbit anti-mouse ionized calcium-binding adaptor molecule 1 (Iba-1, 1:400; Wako Pure Chemical Industries, Osaka, Japan) and rat anti-mouse MHC class II (M5/114.15.2, 1:400; BD Pharmingen, San Diego, CA, USA). Incubations were carried out for four hours at room temperature, followed by overnight incubation at 4°C. The following day, whole-mounts were washed three times in PBS and incubated with biotinylated goat anti-rat (1:400; Amersham Biosciences, Buckinghamshire, UK) diluted in blocking buffer, for two hours at room temperature. Following an additional three washes in PBS, whole-mounts were incubated with Alexa-fluor 647 goat anti-rabbit (1:300; Molecular Probes, Carlsbad, CA, USA) and Streptavidin Cy3 (1:300; Molecular Probes) diluted in PBS. In some experiments, whole-mounts were instead incubated with biotinylated goat anti-rabbit (1:300; Vector Laboratories, Burlingame, CA, USA) followed by Streptavidin Cy3, as described above for biotinylated goat anti-rat. Whole-mounts were finally incubated with Hoechst/PBS (1/500) for 10 minutes at room temperature and cover-slipped with the vitreous side up (Iba-1–stained Cx₃cr1^gfp/+ retinas and Iba-1/MHC class II–stained C57BL/6 retinas) or with the photoreceptors side up (Iba-1–stained C57BL/6 retinas), the latter being referred to as photoreceptor layer (PRL) in the remainder of the article. Choroidal whole-mounts were mounted with the retinal pigment epithelium (RPE) side up, thus making the apical surface of the RPE visible for quantitation of subretinal macrophages. Negative controls were generated by omitting the primary antibody from the immunostaining protocol or substituting the primary antibody with an isotype control antibody. 

Microglial stratification in retinal layers was analyzed by using cryosections of eyes from Cx₃cr1^gfp/+ mice. In brief, 5-μm sections were fixed for 30 minutes in 4% PFA and incubated with Hoechst for 10 minutes. 

Quantitative Analysis of the Number and Characterization of Retinal Microglia

To quantify and characterize retinal microglia, Iba-1–immunostained Cx₃cr1^gfp/+ retinas were analyzed by confocal microscopy. Previous studies have reported no difference in the number and phenotype of retinal microglia in Cx₃cr1^gfp/+ mice compared with WT mice.³⁵ Z-stacks of 1-μm optical sections were taken by using the 20× objective (SP5; Leica Microsystems, Mannheim, Germany). The total number of microglia was counted in z-series videos by using FIJI/ImageJ software (http://imagej.nih.gov/ij/; provided in the public domain by the National Institutes of Health, Bethesda, MD, USA) and numbers were converted to represent density/mm². In retinas from each treatment group (n = 5), three representative fields encompassing the central, paracentral, and peripheral regions were selected for counting. In each field, cell bodies were counted separately in the inner plexiform layer (IPL) and outer plexiform layer (OPL). Imaris v6.1.3 (Bitplane, Zurich, Switzerland) was used for the three-dimensional analysis of microglial processes. Surface rendering of GFP⁺ microglia from z-stacks of confocal microscope images was performed and the lengths of processes were determined by measuring the distance from the centroid of a cell body to the tip along the processes. Retinas from saline-treated mice and from one TLR ligand treatment group (CpG-ODN) were selected for initial statistical analysis. For each retina (n = 5), six randomly selected microglia from the IPL of each of the individual representative fields (central region, paracentral region, and peripheral region) were measured. From this analysis, we showed no regional variance within each retina or between each animal within a treatment group. Therefore, six randomly selected microglia in the IPL from the central region of each animal were measured. The mean and standard error from three to five animals per treatment group was calculated. 

Confocal microscopy was also used to quantify the number of cells expressing MHC class II in response to treatment. Confocal z-series of Iba-1/MHC class II–stained retinas from C57BL/6 WT mice were acquired with a 20× objective (SP5; Leica Microsystems). Z-stacks consisted of 3- to 4-μm optical sections, and two random fields were captured for each retinal whole-mount (each field representing 0.6006 mm²). Cells expressing MHC class II were manually counted by using FIJI/ImageJ. For any given animal, the number of MHC class II–expressing cells in each field were divided with the total area imaged, hence represented as cells/mm². 

Quantitative Analysis of the Number of Subretinal Microglia/Macrophages

To quantify the number of microglia/macrophages present in the subretinal space enclosed by the PRL and the RPE, MHC II⁺ cells were analyzed by epifluorescence microscopy. The surface of whole-mounts was imaged with a 20× objective (Olympus Provis AX70 microscope; Olympus, Tokyo, Japan) and captured with a Sis FVII digital camera. Three random fields were captured for each whole-mount (each field representing 0.1462 mm²). Cells expressing Iba-1 were manually counted by using FIJI/ImageJ and represented as cells/mm². 

Statistical Analyses

Statistical analyses were carried out by using unpaired, two-tailed Mann-Whitney test or unpaired, two-tailed Student's t-test (GraphPad Prism, v6; GraphPad Software, Inc., La Jolla, CA, USA). A P value < 0.05 was considered significant. 

Results

Microglial Activation in Response to TLR3, TLR7/8, and TLR9 Ligands at 24 Hours Post Treatment

To facilitate visualization of microglia morphology we used Cx₃cr1^gfp/+ mice in which most monocyte-derived cells are GFP⁺.²⁷ Cryosections of saline-, CpG-, R848- and Poly I:C–treated eyes demonstrated the localization of microglia in the GCL and plexiform layers of the mouse retina (Fig. 1A). Representative en face confocal images of Iba-1–stained retinal whole-mounts demonstrated the presence of Iba-1⁺ Cx₃cr1^gfp+ microglia in the IPL and outer plexiform layer (not shown). The presence of less ramified Iba1⁺ Cx₃cr1^gfp+ macrophages (arrows), indicating infiltrating macrophages, was most apparent after CpG-ODN treatment but not saline, R848, or Poly I:C treatment (Fig. 1A). Dendritic tracing performed on randomly selected microglia in the IPL in retinal whole-mounts from all experimental groups revealed that the average dendrite length, used as a measure of microglial activation, was significantly reduced in CpG-ODN– and R848-treated eyes but unchanged in saline- and Poly I:C–treated eyes (Figs. 1B, 1C). Despite differences in the dendrite length of microglia in CpG-ODN– and R848-treated eyes, the density of microglia was similar across treatment groups at 24 hours (Fig. 1D). 

Figure 1

 

GFP⁺ retinal microglial response to TLR3, TLR7/8, and TLR9 ligands 24 hours post application to the debrided cornea. (A) Confocal imaging of GFP⁺ microglia in retinal histologic sections (top panel) and whole-mounts (costained with Iba1; bottom panel) from Cx₃cr1^gfp/+ mice. Histologic cross-sections reveal normal microglial distribution following CpG, R848, and Poly I:C treatment. En face confocal images of the IPL show an increase in Iba1 expression in CpG- and R848-treated mice, with a subset of Iba-1⁺ GFP⁻ cells in CpG-treated eyes (bottom panel, arrows). (B) Analysis of dendrite length in retinal GFP⁺ microglia from Cx₃cr1^gfp/+ mice, located in the IPL. Treatment with CpG and R848 induces morphologic changes in retinal microglia, with cells appearing more amoeboid as opposed to the ramified shape displayed by Poly I:C–treated mice and saline controls. (C) Quantification shows that CpG and R848 treatment significantly decreases the dendrite length of these microglia (n = 3–5 per treatment group). (D) Quantification of microglial density confirms that there is no difference between treatment groups (n = 5 per treatment group). P values were generated by two-tailed Mann-Whitney test, in which *P < 0.05 was considered significant. Error bars depict mean ± SEM. Scale bars: 100 μm. Hoechst was used as nuclear stain (blue; [A]).

Moderate Vitritis and Microglial Activation in CpG-ODN–Treated Eyes at 1 Week

We have previously reported that topical application of CpG-ODN does indeed induce the infiltration of inflammatory cells into the vitreous at 24 hours.²⁶ To further examine the posterior segment inflammatory response after 1 week, SD-OCT images of the posterior segment were collected. Representative SD-OCT B-scans of the optic nerve head and retina at 1 week post treatment demonstrated a clear vitreous in saline-treated eyes (Fig. 2A), compared with the presence of numerous round inflammatory cells after CpG-ODN treatment (Figs. 2A, 2B), which conforms with descriptions of vitritis in previous studies.³⁰ As compared to saline-treated eyes, the RNFL thickness was increased by approximately 20% in CpG-ODN–treated eyes (Fig. 2C). Proinflammatory cytokine mRNA (IL-1β and MCP-1) was upregulated in CpG-ODN–treated eyes (Fig. 2D). Microglia appeared more numerous and less ramified in the IPL than in saline-treated controls (Fig. 2E). Iba-1⁺ macrophages in the subretinal space were a distinctive characteristic of CpG-ODN–treated eyes but were absent in saline-treated controls (Fig. 2F). Furthermore, in CpG-ODN–treated eyes there was a statistically significant increase in microglial density in the IPL (Fig. 2G), and Iba-1⁺ cells in the subretinal space, with evidence of cellular processes extending through the nuclear layers, which is not a feature of the distinctive laminar arrangement of microglia in normal eyes (Fig. 2H). In addition, the numbers of MHC class II⁺ cells throughout the entire thickness of the neural retina increased significantly (Fig. 2I). Representative images of MHC II⁺ cells in the inner retina are shown in saline- and CpG-ODN–treated eyes (Fig. 2J). 

Figure 2

 

Stimulation of corneal TLR9 by CpG induces a retinal microglial activation at 1 week post treatment. Retinal microglial activation and vitritis in WT C57BL/6 mice (A–D, F, G, I, J) or Cx₃cr1^gfp/+ mice (E, H) in response to topical CpG application. (A, B) The OCT imaging and quantification reveal an increased number of infiltrating vitreous cells (arrows) in the eyes of CpG-treated mice compared with saline treatment. Additional effects of CpG treatment include thickening of the RNFL (C) and upregulation of mRNA of proinflammatory genes IL-1β and MCP-1 (D). (E, G) Confocal imaging of GFP⁺ microglia in retinal histologic sections (top panel) and whole-mounts (costained with Iba1; bottom panel) from Cx₃cr1^gfp/+ mice. Histologic cross-sections (E) reveal occasional vertical orientation of GFP⁺ cells, indicating microglial migration. Optical sections of the IPL show an increase in Iba1 expression in CpG-treated mice. (F, H) Confocal imaging of the subretinal space—consisting of the PRL and the RPE on the choroidal surface—reveals the presence of Iba1⁺ macrophages in CpG-treated eyes, cells that are virtually absent in saline controls. (H) Quantification of GFP⁺ microglial density in the retina reveals a significant increase after CpG treatment. (I, J) Confocal imaging and quantitative analysis of MHC class II⁺ cells in retinal whole-mounts. The CpG treatment induces increased MHC class II expression in the retina. Each treatment group consisted of n = 6 unless stated otherwise. P values were generated by two-tailed Mann-Whitney test (B, C, G–I) or two-tailed Student's t-test (D). P values of <0.05 were considered significant, with asterisks defined as follows: *P < 0.05 and **P < 0.01. Bars depicted represent mean ± SEM. Scale bars: 100 μm. Hoechst was used as nuclear stain (blue; [E, F, J]).

Severe Vitritis and Microglial Activation in R848-Treated Eyes at 1 Week

Analysis of the posterior segment of R848-treated eyes revealed a significant number of inflammatory cells in the vitreous, compared with saline-treated eyes (Figs. 3A, 3B), which was approximately 3-fold greater than with CpG-ODN treatment (see Fig. 2B). The RNFL was significantly thicker after R848 treatment (Fig. 3C), and gene expression analysis revealed extensive and significant upregulation of all proinflammatory cytokines tested (Fig. 3D). Cx₃cr1^gfp+ microglia appeared to be less ramified and had larger cell bodies in R848-treated eyes than saline controls, which were visible on both cryosections and Iba-1⁺–stained retinal whole-mounts (Fig. 3E). Large numbers of Iba-1⁺ macrophages were present in the subretinal space after R848 treatment (Figs. 3F, 3G). The total number of Cx₃cr1^gfp+ microglia (Fig. 3H) and MHC II⁺ cells within the neural retina was significantly increased, compared to control eyes (Figs. 3I, 3J), and an order of magnitude greater than in CpG-treated eyes (see Fig. 2J). In TLR7^−/− mice treated with R848, few microglia were observed in the subretinal space, and rare MHC II–expressing microglia were observed in the neural retina. Microglia exhibited typical ramified morphology and were distributed normally in the retinal layers (Supplementary Fig. S1). 

Figure 3

 

Stimulation of corneal TLR7/8 by R848 induces retinal microglial activation at 1 week post treatment. Microglial activation and vitritis in the retina of WT C57BL/6 mice (A–D, F, G, I, J) or Cx₃cr1^gfp/+ mice (E, H) in response to topical R848 application. (A, B) The OCT imaging reveals the presence of significant number of vitreous cells (arrows) infiltrating the eyes of R848-treated mice compared with saline-treated mice. Additional effects of R848 treatment include thickening of the RNFL ([C]; n = 5–6) and a significant upregulation of the proinflammatory genes IFN-γ, IL-1β, MCP-1, and IL-6 (D). (E, G) Confocal imaging of GFP⁺ microglia in retinal histologic sections (top panel) and whole-mounts (costained with Iba1; bottom panel) from Cx₃cr1^gfp/+ mice. Histologic cross-sections reveal the presence of occasional, vertically arranged microglia, indicating migration. Optical sections of the IPL show an increase in Iba1 expression in R848-treated mice compared to saline-treated mice (n = 3 per treatment group). (F, H) Large numbers of Iba1⁺ cells in R848-treated mice, cells that are virtually absent in saline controls. (I, J) Confocal imaging of retinal whole-mounts after R848 application to the debrided cornea, which induces strong MHC class II expression in the retina. Each treatment group consisted of n = 6 unless stated otherwise. P values were generated by two-tailed Mann-Whitney test (B, C, G–I) or two-tailed Student's t-test (D). P values of <0.05 were considered significant, with asterisks defined as follows: *P < 0.05, **P < 0.01, and ***P < 0.001. Bars depicted represent mean ± SEM. Scale bars: 100 μm. Hoechst was used as nuclear stain (blue; [E, F, J]).

Absence of Vitritis and Microglial Activation Following Topically Applied TLR3 Agonist Poly I:C at 1 Week

Having previously reported that topically applied Poly I:C leads to infiltration of MHC II⁺ macrophages to the injured mouse cornea after 1 week,³⁶ we wished to assess whether this TLR3 agonist would elicit vitritis and microglial activation similar to the other intracellular TLR agonists used in the present study. In WT mice whose cornea had been exposed to TLR3 ligand 1 week earlier, SD-OCT imaging of the central retina demonstrated a normal vitreous that, similar to saline-treated controls, was devoid of inflammatory cells (Figs. 4A, 4B). There was no significant difference in the thickness of the RNFL (Fig. 4C) and mRNA expression levels of the cytokines IFN-γ, IL-1β, IL-6, and MCP-1 (Fig. 4D), compared to controls. Analysis of cryosections from treated Cx₃cr1^gfp/+ mice revealed normal morphology of microglia (Fig. 4E) and the absence of Iba-1⁺ macrophages in the subretinal space (Figs. 4F, 4G). There was no quantitative difference in the density of Cx₃cr1^gfp+ microglia (Fig. 4H) or MHC II⁺ cells in the neural retina (Figs. 4I, 4J) between control and Poly I:C–treated eyes. In summary, the TLR3 ligand Poly I:C and saline-treated controls failed to elicit any inflammatory changes in the posterior segment as a sequelae to topical application to the debrided cornea. 

Figure 4

 

Absence of retinal microglial activation 1 week after stimulating corneal TLR3 with Poly I:C. Analyses of inflammation in retinas from WT C57BL/6 mice (A–D, F, G, I, J) or Cx₃cr1^gfp/+ mice (E, H) in response to application of Poly I:C (TLR3) to the cornea. (A, B) The OCT imaging shows no signs of vitreous cells in the retina of Poly I:C–treated mice (n = 5–6). Similarly, Poly I:C–treated mice have a normal RNFL thickness ([C]; n = 5) and a retinal proinflammatory gene expression profile (D) similar to saline controls. (E, H) Confocal imaging of GFP⁺ microglia in retinal histologic sections (top panel) and whole-mounts (costained with Iba1; bottom panel) from Cx₃cr1^gfp/+ mice. Histologic cross-sections display normal microglial stratification in retinal layers. The IPL optical sections show levels of Iba1 expression and numbers of microglia comparable to saline controls (n = 3 per treatment group). (F, H) Rare/absent Iba1-expressing cells in Poly I:C–treated mice, as in saline controls. (I, J) Confocal imaging of retinal whole-mounts stained for MHC class II expression. Poly I:C application to the debrided cornea does not induce overt MHC class II expression in the retina. Each treatment group consisted of n = 6 unless stated otherwise. Data from Poly I:C experiments were subjected to two-tailed Mann-Whitney test (B, C, G–I) or two-tailed Student's t-test (D) but failed to attain statistical significance. Bars depicted represent mean ± SEM. Scale bars: 100 μm. Hoechst was used as nuclear stain (blue; [E, F, J]).

Discussion

Activation of retinal microglia has been implicated in the development and progression of retinal degenerative diseases such as diabetic retinopathy,³⁷ retinitis pigmentosa,³⁸ and age-related macular degeneration.³⁹ We have previously demonstrated that toll-like receptor–induced anterior segment inflammation and systemic viral infection can cause retinal inflammation in mice.^26,40 In the present study we sought to extend our previous report showing that topically applied TLR9 ligand can induce retinal inflammation in mice by elucidating whether other topically applied intracellular TLR ligands can cause activation of retinal microglia and induce their migration into the subretinal space. The SD-OCT imaging was valuable in quantifying vitreous infiltrates and measuring changes in retinal thickness; in addition, we took advantage of Cx₃cr1^gfp/+ reporter mice to perform image analysis of microglia morphology as evidence of activation. Our data revealed that another intracellularly recognized TLR ligand, R848, leads to marked activation and migration of retinal microglia, inner retinal edema, and upregulation of proinflammatory genes after topical application to the injured cornea. No signs of innate inflammation were observed in the retina of eyes in which ligands were topically applied to the intact corneal surface (data not shown), confirming that breach of the corneal epithelium, a normally effective barrier to microbial products,⁴¹ is required for activation of the innate immune response. 

R848, a synthetic mimic of naturally occurring single-stranded RNA, activates cells expressing TLR7 and TLR8 and is involved in mediating antiviral responses.⁴² Owing to its ability to modify immune responses, it has been used as a topical agent to treat precancerous skin lesions (actinic keratosis) and herpes simplex virus-2–associated lesions, and as an adjuvant in vaccines.⁴³ However, to date we are unaware of any use in the context of ocular infections. Murine TLR8 has been previously suggested to be nonfunctional, which is evidenced by a lack of responsiveness to R848 treatment in TLR7-deficient mice, even though TLR8 is present.^36,44 Our data demonstrating a lack of microglial activation in TLR7^−/− mice in response to topical R848 support these previous studies and confirm the specificity of R848 as a TLR7 ligand. 

TLR3 is another nucleic acid–sensing TLR, which recognizes double stranded RNA of viral origin and Poly I:C is a synthetic analog of viral dsRNA.²⁰ Addressing whether topically applied TLR3 ligand can induce intraocular inflammation is of particular interest in light of recent controversial studies implicating a role for nucleic acid–sensing TLR3 in the development of geographic atrophy in age-related macular degeneration.^25,45 Furthermore, TLR3 is activated by endogenous products of damaged RPE cells and is involved in retinal degenerative processes in mice with defective clearance of all-trans-retinal in the photoreceptors.²³ Exogenous delivery of 2 μg Poly I:C into the subretinal space also causes severe retinal inflammation and photoreceptor necrosis in mice.⁴⁶ Thus, while it is clear that direct exposure to TLR3 ligands in the mouse retina can trigger innate immune responses, our data demonstrated that this ligand does not cause retinal inflammation after topical corneal application. It is possible that, unlike the highly stable and low-molecular-weight CpG-ODN and R848, the stability and relatively larger size of Poly I:C limit its diffusion through the ocular media. However, in a recent study of ocular pharmacokinetics of fluorochrome-labeled macromolecules, RNA nanoparticles, but not dsRNA, were easily detectable in the posterior segment of the eye after subconjunctival delivery.⁴⁷ The inability of dsRNA to traverse the ocular media, despite being of similar size to the RNA nanoparticles, may explain why this TLR ligand failed to induce intraocular inflammation in our model.⁴⁷ 

Time-lapse intravital imaging studies of brain microglia have elegantly shown that upon activation by diverse stimuli, such as local laser lesions or injection of picogram quantities of LPS, ramified microglia respond by rapidly changing cell motility, polarity, and activity in an adenosine triphosphate–dependant manner.^4,5 As morphologic alterations in microglia are well recognized and widely accepted as a means of assessing functional status, we investigated microglial dendrite length as an indicator of their response to topical intracellular TLR ligands. We observed retraction of retinal microglia dendrites 24 hours after topical application of TLR7/8 and TLR9 ligands, but not the TLR3 ligand Poly I:C, to the injured cornea. In addition to alterations in microglial morphology, loss of microglial stratification and the appearance of MHC II⁺ cells in the retina and Iba-1⁺ cells in the subretinal space are other indications of microglial activation in response to a proinflammatory milieu in the retina. The presence of subretinal macrophages in particular has been reported in a range of rodent models of retinal conditions including acute elevated intraocular pressure,⁴⁸ light-induced degeneration,⁶ and ageing.^49,50 The presence of subretinal macrophages 1 week following TLR7/8 and TLR9 activation in the cornea parallels reports of macrophage migration to the subretinal space in rodent models of systemic viral and fungal infections,^40,51 and highlights the potential significance of these cellular responses in the development of retinal diseases. 

This study provided further evidence that anterior segment inflammation following topical application of TLR ligands to the injured ocular surface can lead to altered microglia homeostasis deeper in the retina. This is potentially significant when considering the well-described age-associated priming of microglia and the subsequent exaggerated and prolonged inflammatory responses following innate immune activation in aged animals.⁵² Thus, activation of retinal microglia during corneal microbial infection may play a role in early-onset retinal degenerative diseases, where chronic inflammation is implicated.⁵³ These data suggest that TLR7/8 and TLR9 receptor activation accompanying viral and bacterial infections of the cornea may have unsuspected consequences for the normally quiet immunologic environment of the neural retina. 

Acknowledgments

The authors thank Janet Choi (University of Melbourne) who provided technical help with animal studies, tissue processing, and image analysis and Jeremiah Lim (University of Melbourne) who helped with image analysis of vitreous cells. 

Supported by funding from the National Health and Medical Research Council of Australia (Project Grants 1026301 and 1042612) (HRC, JVF, and PGM). 

Disclosure: H.R. Chinnery, None; C. Naranjo Golborne, None; C.M. Leong, None; W. Chen, None; J.V. Forrester, None; P.G. McMenamin, None