The Role of Microglia and Peripheral Monocytes in Retinal Damage after Corneal Chemical Injury

Abstract

Eyes that have experienced alkali burn to the surface are excessively susceptible to subsequent severe glaucoma and retinal ganglion cell loss, despite maximal efforts to prevent or slow down the disease. Recently, we have shown, in mice and rabbits, that such retinal damage is neither mediated by the alkali itself reaching the retina nor by intraocular pressure elevation. Rather, it is caused by the up-regulation of tumor necrosis factor-α (TNF-α), which rapidly diffuses posteriorly, causing retinal ganglion cell apoptosis and CD45⁺ cell activation. Herein, we investigated the involvement of peripheral blood monocytes and microglia in retinal damage. Using CX3CR1^+/EGFP::CCR2^+/RFP reporter mice and bone marrow chimeras, we show that peripheral CX3CR1⁺CD45^hiCD11b⁺MHC-II⁺ monocytes infiltrate into the retina from the optic nerve at 24 hours after the burn and release further TNF-α. A secondary source of peripheral monocyte response originates from a rare population of patrolling myeloid CCR2⁺ cells of the retina that differentiate into CX3CR1⁺ macrophages within hours after the injury. As a result, CX3CR1⁺CD45^loCD11b⁺ microglia become reactive at 7 days, causing further TNF-α release. Prompt TNF-α inhibition after corneal burn suppresses monocyte infiltration and microglia activation, and protects the retina. This study may prove relevant to other injuries of the central nervous system.

Figure 1

Corneal alkali burn causes retinal damage in humans and animals. A: Digital image of a human cornea after alkali burn presented with extensive neovascularization and conjunctivalization. B: Fundus digital imaging of the optic nerve (ON) reveals end-stage glaucoma associated with advanced ON cupping and parlor. C:In vivo optical coherence tomography confirms significant loss of retinal nerve fiber layer (RNFL) thickness (dotted black line) compared with normal age-matched controls (green section). D: Digital image of a rabbit cornea after alkali burn with extensive neovascularization and conjunctivalization. E: Immunohistochemical analysis of rabbit retina using βIII-tubulin after corneal alkali burn (top panel) and of control eye (bottom panel) shows significant reduction in RNFL thickness in the burned eye compared with control retina (white arrowheads denote ganglion layer cells). F and J: P-phenylenediamine staining of rabbit and mouse ONs after corneal alkali burn (top panel) and of control eye (bottom panel) shows significant axonal degeneration in burned eyes (black arrows). G: Corresponding 40% of retinal ganglion cell (RGC) loss. H: Digital image of a mouse cornea after alkali burn with extensive neovascularization and conjunctivalization. I: Immunohistochemical analysis of flat mount mouse retinas using βIII-tubulin marker after corneal alkali burn (top panel) and of control eye (bottom panel) shows significant loss of RNFL density (white arrowheads). K: Corresponding 40% of RGC loss assessed by Brn3a marker. L: Microarray RNA analysis in mouse retinas 24 hours after the burn shows significant up-regulation (more than twofold log₂) of C-C and C-X-C motif genes involved in monocyte recruitment, colony stimulating factor 1 gene required for microglia survival, and inflammatory genes of the IL and tumor necrosis factor (TNF)-α families. G and K: Independent t-test was used. n = 3 (G, K, and L). ^∗∗P < 0.01, ^∗∗∗P < 0.001. Scale bars: 20 μm (E and I); 10 μm (F and J). Cnt., control; GCL, ganglion cell layer; INF, inferior (refers to different retinal segments); NA, not applicable; NAS, nasal; SUP, superior; TEMP, temporal.

Figure 2

Corneal alkali burn causes CX3CR1⁺ cell activation in the retina. A: Confocal images of flat-mounted retinas of CX3CR1^+/EGFP reporter mice 24 hours and 7 days after corneal alkali burn. Cells are color-depth coded on the basis of their position within the retina layers. Blue indicates cells in the ganglion cell layer (GCL; 0 μm), red indicates cells in the outer plexiform layer (OPL), and intermediate colors indicate cells between the GCL and OPL. Color bar translates color to μm. B: Morphometric analysis of CX3CR1⁺ cells was performed in three distinct retina layers: the GCL, inner nuclear layer (INL), and OPL. The longest process length measured from the edge of the cell body (in μm) was used as a morphometric descriptor of CX3CR1⁺ cell activation. A and B: At 24 hours after the burn, CX3CR1⁺ cells in the GCL start to become amoeboid, exhibiting significant reduction in processes length. CX3CR1⁺ cells in the INL and OPL are not affected. Anti–tumor necrosis factor (TNF)-α treatment preserves cell ramification in the GCL. C: At 7 days, CX3CR1⁺ cells accumulate around the outer segment of the optic nerve head (ONH) and align to the direction of the retina nerve fiber layer. Also, CX3CR1⁺ cells in all retinal layers became more amoeboid, with significant shortening of the cell processes. Anti–TNF-α treatment significantly reduces CX3CR1⁺ cell amoeboid transformation in all retinal layers and reduces cell accumulation around the ONH. B: Multiple comparisons using Holm-Sidak method. n = 3 per group (B). ^∗P < 0.05, ^∗∗P < 0.01. Scale bars = 100 μm (A). Original magnification, ×40 (C).

Figure 3

Activated retinal CX3CR1⁺ cells have microglia and macrophage signatures. A, B, and D: Biochemical assessment of CX3CR1⁺ cells using flow cytometry. CX3CR1⁺ CD11b⁺ cells divided into macrophages (CD45^hi) and microglia (CD45^lo). Major histocompatibility complex (MHC)-II expression was used as a metric of cell activation. Naive eyes express predominantly CD45^lo (90%), with only 4% of the CX3CR1⁺ cells expressing CD45^hi. None of the CD45 cell populations express MHC-II. Twenty-four hours after the burn, the percentage of CX3CR1⁺ CD11b⁺ CD45^hi cells in the retina increases with concomitant increase in MHC-II expression. Seven days after the burn, CX3CR1⁺ CD11b⁺ CD45^hi population decreases with concomitant reduction in MHC-II expression. At the same time, the percentage of CX3CR1⁺ CD11b⁺ CD45^lo remains unchanged, but the number of MHC-II⁺ cells increases. Treatment with monoclonal anti–tumor necrosis factor (TNF)-α antibody significantly reduces the number of CX3CR1⁺ CD11b⁺ CD45^hi MHC-II⁺ cells in the retina, but it has no effect on the number of CX3CR1⁺ CD11b⁺ CD45^lo MHC-II⁺ cells. At 7 days, anti–TNF-α treatment increases the percentage of CX3CR1⁺ CD45^lo CD11b⁺ and their corresponding MHC-II expression, but it has no effect on the number of CX3CR1⁺ CD45^hi CD11b⁺ MHC-II⁺ cells. Overall, anti–TNF-α treatment reduces the infiltration of CX3CR1⁺ CD11b⁺ CD45^hi cells in the retina and prevents MHC-II expression by these cells. C: Statistical comparisons of the number of CX3CR1⁺ CD11b⁺ CD45^hi MHC-II⁺ and CX3CR1⁺ CD11b⁺ CD45^lo MHC-II⁺ cells between the treatment groups using analysis of variance with Holm-Sidak method. Statistically significant values (≤0.05) are shown in bold. n = 3 in all groups. ANOVA, analysis of variance; FSC, forward scatter.

Figure 4

Monocyte infiltration into the retina after corneal burn. Image quantification of CX3CR1⁺ and CCR2⁺ cell infiltration in the retina after corneal alkali burn using CX3CR1::CCR2^EGFP/RFP bone marrow (BM) chimera. A, F, and G: In naive mice, only few ramified CX3CR1⁺ CCR2⁻ cells are present around the optic nerve head (ONH). These cells do not migrate beyond its boundary. However, scant CX3CR1⁻ CCR2⁺ cells are found across the retina, located in the ganglion cell layer (GCL). B: Twenty-four hours after corneal surface injury, amoeboid CX3CR1⁺ CCR2⁺ cells from the blood infiltrate the retina and locate themselves primarily in the GCL. An increased number of amoeboid CX3CR1⁺ CCR2⁺ cells is present around the ONH (white arrows) and along the retinal vessels (red arrows). Only a few cells are CX3CR1⁺ CCR2⁻. C, F, and G: At 24 hours, anti–tumor necrosis factor (TNF)-α treatment (Tx) reduces the infiltration of blood-derived CX3CR1⁺ CCR2⁺ cells in the retina. D, F, and G: At day 7, infiltrated blood-derived CX3CR1⁺ CCR2⁺ cells accumulate between the GCL and outer nuclear layer (ONL) and exhibit a phenotypic differentiation from amoeboid to dendritiform. The number of blood-derived CX3CR1⁺ CCR2⁺ cells in the GCL and inner plexiform layer remains elevated, whereas a significant number of CCR2⁺ cells are located around the outer segment of the ONH (white arrow). E–G: TNF-α inhibition reduces the number of peripheral CX3CR1⁺ CCR2⁺ cell infiltrates and promotes differentiation from amoeboid to dendritiform morphology and concomitant abolishment of CCR2 marker. CCR2 remains expressed by amoeboid cells. H: Flow cytometry in CX3CR1^+/EGFP BM chimeras confirms the reduction in CX3CR1⁺ cell infiltration after anti–TNF-α treatment. F and G: Independent group comparisons with naive mouse as reference using Holm-Sidak method. n = 6 per group (F and G); n = 3 per group (H). ^∗P < 0.05, ^∗∗∗∗P < 0.0001 versus naive. Scale bars = 100 μm (A–E). BMT, BM transfer; EGFP, enhanced green fluorescent protein.

Figure 5

Monocyte infiltration into the retina through the optic nerve head. Image quantification of CX3CR1⁺ CCR2⁺ cell infiltration in the optic nerve (ON) after corneal alkali burn using CX3CR1^+/EGFP::CCR2^+/RFP bone marrow (BM) chimera model. A–E: Confocal microscopy of ONs of BM-transferred (BMT) mice. A, F, and G: Blood-derived CX3CR1⁺ and CCR2⁺ cells physiologically populate the ON within 10 weeks after BMT. Most cells are either CX3CR1⁺CCR2⁻ or CX3CR1⁻CCR2⁺, with only few cells expressing both markers. CX3CR1⁺CCR2⁻ and CX3CR1⁺CCR2⁺ acquire a semiramified morphology, whereas CX3CR1⁻CCR2⁺ are mainly amoeboid in appearance. B, D, F, and G: Twenty-four hours after corneal alkali burn, CX3CR1/CCR2 cell numbers increase in the optic nerves, followed by gradual reduction at 7 days (P < 0.005). C and E–G: Anti–tumor necrosis factor (TNF)-α treatment (Tx) significantly reduces the number of CX3CR1/CCR2⁺ cells in the ON 24 hours and 7 days after the burn, and leads to reduction in the number of migrating cells into the retina. F and G: Independent group comparisons with naive mouse as reference using Holm-Sidak method. n = 3 per group (F and G). ^∗∗P < 0.01, ^∗∗∗∗P < 0.0001 versus naive. Scale bars = 75 μm (Α–Ε).

Figure 6

Activated neuroglia and blood monocytes promote neuroinflammation and degeneration. A and B: Tumor necrosis factor (TNF)-α staining in flat-mounted CX3CR1^+/EGFP mouse retinas. CX3CR1 and TNF-α markers are coexpressed in the ganglion cell layer 24 hours after the burn. B: Inset: Magnification of the dotted boxed area. C and D: Four weeks after the burn, CX3CR1⁺ cells (green) remain activated (amoeboid), and the retinal nerve fiber layer (βIII-tubulin; red) exhibits 40% tissue density loss (P < 0.05) compared with treatment with anti–TNF-α antibody that abrogates CX3CR1⁺ cell activation (less amoeboid) and inhibits retinal nerve fiber layer (RNFL) loss. Dotted boxed areas represent the section of the retina nerve fiber layer that was used for quantification. E and F: Quantification of RNFL thickness using vertical projection analysis of flat-mounted retinas 20 weeks after the burn shows 40% RNFL loss (P < 0.0001), compared with anti–TNF-α treated mice. The dotted line represents the location of the cross-sectional cut for quantification. G and H: Three-dimensional rendering of a retina flat mount shows amoeboid CX3CR1⁺ cells (semitransparent green) appearing to ensheathe (white arrows) β3-tubulin⁺ neuronal tissue (red) 4 weeks after corneal alkali burn. D and F: Independent group comparisons using Holm-Sidak method. n = 3 per group (D and F). ^∗P < 0.05, ^∗∗∗∗P < 0.0001. Scale bars = 50 μm (C and E). Original magnification: ×63 (A, B, and H); ×40 (G).

Figure 7

CX3CR1⁺ CCR2⁺ monocyte activation and tumor necrosis factor (TNF)-α expression in the retina. Flow cytometry of CX3CR1::CCR2^EGFP/RFP mice. A and B: Ocular injury increases the percentage of CD45^hi CCR2⁺ cells in the retina at 24 hours. TNF-α is expressed by CD45^lo CCR2⁻ (microglia) and CD45^hi CCR2⁺ (peripheral monocytes). C–E: At 24 hours after ocular surface injury, peripheral CX3CR1⁺ CCR2⁺ monocytes are shown to express high levels of activation marker major histocompatibility complex (MHC)-II^hi. The ocular burn has no significant effect on MHC-II expression by CX3CR1⁺ CCR2⁻ microglia, which remain at low levels (MHC-II^int). Independent group comparisons between naive and burn 24 hours using analysis of variance and Holm-Sidak method. n = 3 in all groups. ^∗P < 0.05 verus naive (unpaired t-test). FSC, forward scatter; GFP, green fluorescent protein; RFP, red fluorescent protein.

Figure 8

Proposed mechanism of retinal damage after corneal alkali burn. The effect of blood-derived monocytes and neuroglia in retinal degeneration after acute ocular surface trauma with alkali. A: Corneal alkali injury causes acute retinal inflammation and infiltration of CX3CR1⁺ and CCR2⁺ cells through the optic nerve head that align along the retinal vessels. B: Increased monocyte/macrophage trafficking causes neuroglial cell activation and subsequent elevation of inflammation. Activated CX3CR1⁺ cells enswathe retinal ganglion cells (RGCs) and nerve axons, a gliotic process that leads to retinal tissue damage. RNFL, retinal nerve fiber layer; TNF-α, tumor necrosis factor-α.

Supplemental Figure S1

Color-depth coding of CX3CR1⁺ and CCR2⁺ peripheral monocytes in the retina. Color-depth–coded confocal images of CX3CR1^+/EGFP::CCR2^+/RFP bone marrow–transferred (BMT) mice. Color analysis shows the position of the CX3CR1⁺ and CCR2⁺ cells within the retina tissue. Blue cells are located in the ganglion cell layer (GCL), and red in the outer plexiform layer (OPL). All intermediate colors are cells located between the GCL and OPL. Original magnification, ×63 (all images). TNF-α, tumor necrosis factor-α; Tx, treatment.