Protective Roles of the Fractalkine/CX3CL1-CX3CR1 Interactions in Alkali-Induced Corneal Neovascularization through Enhanced Antiangiogenic Factor Expression¹

Corneal neovascularization (CNV)³ can arise from various causes, including corneal infections, misuse of contact lens, chemical burn, and inflammation (1, 2, 3) and frequently leads to impaired vision. Moreover, survival rates of corneal grafts were markedly reduced when placed into vascularized recipient beds or when CNV develops in the transplanted cornea (4, 5, 6, 7, 8, 9). Thus, it is necessary to develop effective measures to treat and/or prevent CNV based on the understanding of its pathogenesis at cellular and molecular levels.

Neutrophils are presumed to be the predominant leukocyte population, which infiltrate a variety of injured organs and tissues in the early stages of inflammation. Following neutrophil accumulation, mononuclear cells such as monocytes/macrophages, lymphocytes, and plasma cells infiltrate the site of tissue injury. Normal cornea does not contain any hematopoietic cells, but in the early phase of corneal injury, preceding the occurrence of CNV, a large number of neutrophils infiltrate the cornea followed by the accumulation of mononuclear cells (10, 11, 12), as observed with the injuries of other organs. Infiltrating neutrophils are presumed to contribute to the development of CNV. However, we previously proved that alkali injury-induced CNV can occur independently of granulocyte accumulation (10).

The recruitment of leukocytes to inflammatory sites is regulated by chemokines and individual chemokines attract specific leukocyte populations through processes determined by ligand specificity and by the expression patterns of the corresponding receptors (13). Mouse monocytes/macrophages express a distinct set of chemokine receptors such as CCR2 and CCR5 on their cell surface. Infiltrating monocytes/macrophages are presumed to be a rich source of angiogenic factors and can contribute to the development of neovascularization in various tissues, including cornea (14, 15, 16, 17). We observed that a potent angiogenic factor, vascular endothelial growth factor (VEGF), was consistently detected mainly in infiltrating monocyte/macrophages but not granulocytes (10). Indeed, the lack of CCR2 and/or CCR5 gene, reduced macrophage accumulation as well as CNV formation, when chemical and mechanical denudation was applied to corneal and limbal epithelium (18, 19). In contrast, several groups have provided evidence of the antiangiogenic activities of infiltrating macrophages in choroidal neovascularization (20, 21).

Mouse monocytes/macrophages also express CX3CR1 on their surface (22, 23). Moreover, a ligand for CX3CR1, CX3CL1/fractalkine, is highly expressed on vascular endothelial cells and therefore is presumed to have a role in angiogenesis (24, 25). Furthermore, the dichotomy of monocytes/macrophages has been proposed based on their expression levels of CCR2 and CX3CR1 (26, 27, 28, 29, 30). Thus, it is tempting to speculate that CX3CR1-positive monocytes/macrophages may exhibit distinct activities in CNV, compared with CCR2-positive ones. Hence, to address this assumption, we compared the molecular pathological changes in wild-type (WT) and CX3CR1-deficient mice, involving a frequently used model of ocular corneal neovascularization, namely, alkali injury-induced CNV (18, 19).

Recombinant mouse CX3CL1 (571-MF/CF) and recombinant mouse CCL2 were obtained from R&D Systems. Rat anti-mouse F4/80 mAb (clone A3-1) and rat anti-mouse CD68 mAb (clone FA-11) were obtained from Serotec and rabbit anti-mouse CD31 polyclonal Abs (pAbs) were purchased from Abcam (ab28364). Rat anti-mouse-Ly-6G mAb (clone 1A8, catalog no. 551459) and rat anti-mouse pan-NK cell mAb (clone DX5) were obtained from BD Pharmingen. Goat anti-ADAMTS-1 (a disintegrin and metalloprotease with thrombospondin 1; L-16, sc-31080), goat anti-thrombospondin (TSP) 1 (N-20), and goat anti-CX3CL1 pAbs (M-18) were obtained from Santa Cruz Biotechnology and rabbit anti-ADAMTS-1 pAbs (ALX-210-555) were from Alexis Biochemicals. Alexa Fluor 488 donkey anti-goat IgG (H + L), donkey anti-rat IgG (H + L) as well as Alexa Fluor 594 donkey anti-rabbit IgG (H + L) were purchased from Invitrogen Life Technologies. Rabbit anti-CX3CR1 pAbs were provided by Dr. T. Imai (Kan Research Institute, Kyoto, Japan).

CX3CR1-deficient mice were prepared as described previously (25) and were a gift from Drs. P. M. Murphy and J.-L. Gao (National Institute of Allergy and Infections Diseases, National Institutes of Health, Bethesda, MD). The mice were backcrossed to BALB/c mice for more than eight generations. Pathogen-free BALB/c mice were obtained from Clea Japan and were designated as WT mice. Age- and sex-matched CX3CR1-deficient and WT mice were kept under specific pathogen-free conditions in groups of five and fed regular laboratory chow and water ad libitum. A 12-h day and night cycle was maintained during the whole course of the study. All animal experiments were performed at the Institute for Experimental Animals (Kanazawa University Advanced Science Research Center, Kanazawa, Japan) in accordance with the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research and complied with the standards set out in the Guidelines for the Care and Use of Laboratory Animals of Kanazawa University and were approved by the Committee on Animal Experimentation of Kanazawa University.

Mice were anesthetized with an i.p. injection of 1.8% (v/v) Avertin at a dose of 0.15 ml/10 g body weight. A 2-mm disc of filter paper saturated with 1 N NaOH was placed onto the right cornea of each mouse for 40 s, followed by rinsing extensively with 15 ml of PBS. The corneal epithelia were removed using a corneal knife in a rotary motion parallel to the limbus by gently scraping over the corneal surface without injuring the underlying corneal stroma (10). Erythromycin ophthalmic ointment was instilled immediately following epithelial denudation. At the indicated time intervals (days 0, 2, 4, and 7), mice were killed and the corneas were removed from both eyes. These corneas were placed immediately into RNALate (Qiagen) and kept at −86°C until total RNA extraction. In another series of experiments, mice were killed at the indicated time points (days 0, 2, 4, 14, and 28) after alkali treatment and both eyes were entirely removed from each animal. These eyes were fixed in 10% neutral formalin buffer for histological analysis. The left eye of each mouse was used as an untreated control. In some experiments, recombinant CX3CL1 protein was dissolved in 0.2% sodium hyaluronate (Sigma-Aldrich) at 2.5 μg/ml. Five microliters of CX3CL1 preparation or vehicle was applied topically to the alkali-treated eye twice a day for 7 days. Each experiment was repeated at least three times.

Eyes were examined under a Lecia MZ16 surgical microsystem 14 days after alkali injury. In brief, under anesthesia, photographs of the corneas were obtained using a Leica DFC350FX digital camera that was linked to an operating microscope. Microscopic assessment was done by two independent observers without prior knowledge of the experimental procedures.

The paraffin-embedded tissues were cut into 5-μm- thick slices, mounted on poly-l-lysine-coated slides, and subjected to H&E staining. For immunohistochemical staining, the deparaffinized sections or fixed cryosections (8-μm thick) were used. Endogenous peroxidases were quenched in 0.3% (v/v) hydrogen peroxide for 10 min. For Ag retrieval, sections were treated with 0.1% trypsin in 0.1% CaCl₂ for 20 min at 37°C or immersed in 10 mM sodium citrate buffer (pH 6.0) and heated for 20 min in a microwave oven. After washing with PBS, slides were incubated with blocking reagent for 20 min. For the detection of macrophages, the sections were incubated overnight at 4°C with rat anti-mouse F4/80 Ab (1 μg/ml) or rat anti-mouse CD68 Ab (1 μg/ml). Tissue sections were then incubated with biotin-conjugated anti-rat Ig Ab as secondary Abs. To identify neutrophils and NK cells, the sections were incubated overnight at 4°C with purified rat anti-mouse-Ly-6G (2.5 μg/ml) and purified rat anti-mouse pan-NK cell mAbs (2.5 μg/ml), respectively. The sections were further incubated with biotin-conjugated rabbit anti-rat Ig Ab or biotin-conjugated rabbit anti-rat IgM Ab as the secondary Abs. Other sets of sections were incubated overnight at 4°C with rabbit anti-CX3CR1 and goat anti-CX3CL1 pAbs to detect CX3CR1 and CX3CL1 expression, respectively. These slides were incubated with biotin-conjugated anti-rabbit or anti-goat Ig Ab as the secondary Abs. The immune complexes were detected by using an ABC kit and a DAB Substrate Kit from Vector Laboratories according to the manufacturer’s instructions. Slides were then counterstained with hematoxylin and mounted. The numbers of positive cells were counted on five randomly chosen fields of corneal sections in each animal at 200-fold magnification, by an examiner without any prior knowledge of the experimental procedures. The numbers of positive cells per mm² were calculated.

The deparaffinized sections (5-μm thick) or fixed cryosections (8-μm thick) were stained using anti-CD31 pAbs and the numbers and sizes of the CNV were determined as described previously (10, 31, 32) by an examiner with no knowledge of the experimental procedures. Briefly, images were captured with a digital camera and imported into Adobe Photoshop (version 7.0). Then, the number of neovascular tubes per mm² and the proportions of CNV in the hot spots were determined using NIH Image analysis software version 1.62 (National Institutes of Health, Bethesda, MD). Most sections were from the central region of the cornea. The numbers and areas of CNV were evaluated on at least two sections from each eye.

Total RNAs were extracted from the corneas or cultured peritoneal macrophages with the RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions. The RNA preparations were further treated with RNase-free DNase I (Invitrogen Life Technologies) to remove residual genomic DNA. Two micrograms of total RNA was reverse-transcribed at 42°C for 1 h in 20 μl of reaction mixture containing mouse Moloney leukemia virus reverse transcriptase and hexanucleotide random primers (Qiagen). Real-time PCR was performed on an Applied Biosystems StepOne Real-Time PCR System using the comparative threshold cycle (C_T) quantification method. TaqMan Gene Expression Assays (Applied Biosystems) containing specific primers (accession numbers: CX3CL1, Mm00436454_ml; CX3CR1, Mm02620111_s1; TSP-1, Mm00449032_gl; TSP-2, Mm00449036_m1; ADAMTS-1, Mm00477355_m1; VEGF, Mm00437304_m1; TGF-β1, Mm03024053_m1; GAPDH, Mm99999915_ g1), and TaqMan MGB probe (FAM dye labeled)/TaqMan Fast Universal PCR Master Mix were used with 10 ng of cDNA to detect and quantify the expression levels of CX3CL1, CX3CR1, TSP-1, TSP-2, ADAMTS-1, VEGF, and TGF-β1. Reactions were performed for 20 s at 95°C, then with 40 cycles of 1 s at 95°C and 20 s at 60°C. GAPDH was amplified as an internal control. C_T values of GAPDH were subtracted from C_T values of the target genes (ΔC_T). ΔC_T values of WT mice were compared with ΔC_T values of CX3CR1-deficient mice.

Mononuclear cells were isolated from corneas according to the procedure described previously by Ohshima et al. (33), with some modifications. At 2 days after the alkali injury, corneas were removed, teased away with scissors, and incubated at 37°C for 30 min with constant shaking in the presence of 0.5 mg/ml collagenase type D (Roche Diagnostics). Cell suspensions were then passed over a nylon filter with a 100-μm pore size. The resultant cells were further layered on Histopaque-1083 (Sigma-Aldrich) and were centrifuged to obtain mononuclear cells. Isolated cells were stained with FITC-conjugated rat anti-mouse F4/80 mAb and rabbit anti-CX3CR1 pAbs followed by staining with Alexa Fluor 647-conjugated goat anti-rabbit IgG. Fluorescence intensities were determined with the help of a FACSCalibur (BD Biosciences), along with staining of the samples with nonimmunized rabbit IgG (Sigma-Aldrich) as a negative control.

Sections were deparaffinized and treated with 0.1% trypsin for 15 min at 37°C for Ag retrieval. Thereafter, the sections were incubated with the combinations of rat anti-mouse F4/80 and rabbit anti-mouse CX3CR1, rat anti-F4/80 and rabbit anti-ADAMTS-1, goat anti-ADAMTS-1 and rabbit anti-mouse CX3CR1, or goat anti-TSP-1 and rabbit anti-mouse CX3CR1 Abs overnight at 4°C. After being rinsed with PBS, for a double-immunofluorescence analysis using anti-F4/80, the sections were then incubated with the combination of Alexa Fluor 488 donkey anti-rat IgG and Alexa Fluor 594 donkey anti-rabbit IgG (1/100) for 40 min at room temperature in the dark. For a double-color immunofluorescence analysis of CX3CR1 and antiangiogenic factors, the sections were further incubated with the combination of Alexa Fluor 488 donkey anti-goat IgG and Alexa Fluor 594 donkey anti-rabbit IgG (1/100) for 40 min at room temperature in the dark. Finally, the sections were washed with PBS and immunofluorescence was visualized in a dual-channel mode on an Olympus fluorescence microscope. Images were processed using Adobe Photoshop software version 7.0.

Specific pathogen-free 8- to 10-wk-old male WT or CX3CR1-deficient mice were injected i.p. with 2 ml of sterile 3% thioglycolate medium (Sigma-Aldrich), and i.p. macrophages were harvested 3 days later as described previously (34). The resultant cell preparation consists of >95% macrophages as verified by using flow cytometric analysis on the cell preparation immunostained with anti-F4/80 Ab. The cells were suspended in antibiotic-free RPMI 1640 medium containing 10% FBS, and incubated in a humidified incubator at 37°C in 5% CO₂ in 24-well cell culture plates. Two hours later, nonadherent cells were removed and the medium was replaced. The cells were then stimulated with the indicated concentrations of murine CX3CL1 or CCL2 for 12 h. Total RNAs were extracted from the cultured cells and subjected to quantitative RT-PCR as described above. For an immunocytochemical analysis of ADAMTS-1 expression, murine macrophages were seeded onto the wells of a Lab-Tec chamber slide with eight wells (Nalge Nunc) at 5 × 10⁴ cells/well. After adhesion, the cells were stimulated with the indicated concentration of murine CX3CL1 for 24 h in a 37°C incubator with 5% CO₂ and then subjected to immunocytochemical study as previous described (35).

The means and SEM were calculated on all parameters determined in the study. Data were analyzed statistically using one-way ANOVA or two-tailed Student’s t test. A value of p < 0.05 was accepted as statistically significant.

Normal corneas do not contain any leukocytes or vascular structures. In line with our previous observations (10), alkali injury markedly increased the numbers of F4/80-positive macrophages and Gr-1-positive neutrophils in the cornea, reaching maximal levels at 4 days after the injury and decreasing thereafter (Fig. 1). Most of the infiltrated mononuclear cells observed in H&E-stained tissue samples were identified as macrophages by morphological criteria (data not shown). Only a few NK cells were detected after the injury (day 2, undetectable; day 4, 8.8 ± 3.0/mm²; day 7, 4.3 ± 2.4/mm²). Because CX3CR1 is expressed by a proportion of macrophages, we examined the expression of CX3CR1 and its ligand CX3CL1 in corneas after alkali-induced injury with the use of quantitative RT-PCR. CX3CR1 mRNA was barely detected in untreated eyes, but its expression was enhanced in eyes 2 days after injury and further augmented 7 days after the injury (Fig. 2,A). The mRNA of a specific CX3CR1 ligand, CX3CL1, was constitutively expressed in the eyes before and after the injury. CX3CL1 proteins were consistently expressed by corneal epithelial and corneal endothelium, the cell that is localized in the inner layer of the cornea and is distinct from vascular endothelium and, to a lesser degree, by corneal stroma of untreated WT mice. At 2–4 days after the injury, the time when a large number of macrophages infiltrated intracorneally, CX3CL1 was additionally detected in the infiltrating cells (Fig. 2,B, lower panel). CX3CR1-positive cells were not detected in untreated WT mice but appeared in the cornea 2 or 4 days after the injury (Fig. 2, B, upper panel, and C). Moreover, a majority of mononuclear cells, present in cornea 2 days after the injury, expressed F4/80 (data not shown) and a substantial proportion of these F4/80-positive cells expressed CX3CR1 (Fig. 2 D). These observations would indicate that alkali injury induced intracorneal accumulation of F4/80-positive cells and some of them expressed CX3CR1.

Given the fact that CX3CR1 is expressed on macrophages (22, 23), we examined the effects of CX3CR1 deficiency on leukocyte accumulation in injured corneas. Ly-6G-positive neutrophils accumulated to a similar extent in the corneas of WT and CX3CR1-deficient mice after the injury (data not shown). On the contrary, the number of F4/80-positive cells was markedly depressed in CX3CR1-deficient mice compared with WT mice (Fig. 3). When another macrophage marker, CD68, was used to identify macrophages, similar results were obtained (Fig. 3). Thus, the CX3CL1-CX3CR1 interactions may regulate F4/80- and CD68-positive macrophage, but not neutrophil accumulation in damaged corneas.

We next explored CNV in CX3CR1-deficient and WT mice. Until 7 days after the injury, CNV was not detected microscopically (data not shown). By contrast, CNV was macroscopically evident in WT mice 2 wk after the injury, as we previously reported (10). At this time point, macroscopic CNV was more enhanced in CX3CR1-deficient mice than in WT mice (Fig. 4,A). An immunohistochemical analysis using anti-CD31 Ab further demonstrated that vascular areas were increased to a larger extent in CX3CR1-deficient mice than in WT mice (Fig. 4, B and C). Moreover, we also observed that macroscopic CNV persisted in CX3CR1-deficient mice 4 wk after the injury, when CNV was resolved in WT mice (Fig. 4, B and C). These observations indicate that CX3CR1 deficiency enhanced the alkali-induced CNV, along with reduced macrophage accumulation.

Monocytes/macrophages can be a rich source of both angiogenic and antiangiogenic factors (14, 16, 17, 20, 21, 36, 37, 38), and the balance between these two distinct sets of factors can determine the outcome of angiogenesis processes in various situations. Hence, we examined the mRNA expression of angiogenic and antiangiogenic factors in corneas after injury. The mRNA expression of two potent angiogenic factors, VEGF and TGF-β, was marginally augmented and to similar extents in WT and CX3CR1-deficient mice after the injury (Fig. 5, D and E). The intraocular mRNA expression of the antiangiogenic factors TSP-1 and ADAMTS-1 and, to a lesser extent, TSP-2, was enhanced in WT mice after the injury, but their enhanced expression was depressed in CX3CR1-deficient mice (Fig. 5, A–C). These observations imply that CX3CR1 deficiency reduced antiangiogenic factor expression and favored angiogenesis.

In line with the previous observation that CX3CR1 is expressed on macrophages (22, 23), we observed that a substantial proportion of F4/80-positive macrophages expressed CX3CR1 (Figs. 2,D and 6,A). ADAMTS-1 protein was detected in F4/80-positive macrophages (Fig. 6,B). Moreover, CXC3CR1-positive cells, present in injured cornea, expressed antiangiogenic factors such as ADAMTS-1 (Fig. 6,C) and TSP-1 (Fig. 6 D). These observations imply that CX3CR1-positive macrophages counteracted alkali-induced CNV by expressing antiangiogenic factors such as TSP-1 and ADAMTS-1.

Because the CX3CR1-mediated signal can enhance the functions of macrophages, we next examined the effects of exogenous CX3CL1 on antiangiogenic factor expression by murine peritoneal macrophages. CX3CL1 enhanced TSP-1 and ADAMTS-1, but not VEGF, mRNA expression by peritoneal macrophages (Fig. 7,A). CX3CL1-stimulated macrophages consistently exhibited increased ADAMTS-1 protein expression compared with untreated ones (Fig. 7, D and E). By contrast, another macrophage-tropic chemokine, CCL2, enhanced VEGF, but not TSP-1 and ADAMTS-1, mRNA expression by peritoneal macrophages obtained from WT or CX3CR1-deficient mice (Fig. 7 B). Thus, CX3CL1 can induce CX3CR1-positive macrophages to express antiangiogenic factors such as TSP-1 and ADAMTS-1, while another macrophage-tropic chemokine, CCL2, can induce CX3CR1-negative macrophages to express VEGF.

Finally, we examined the effects of topically applied CX3CL1 on alkali-induced CNV. Topical administration of CX3L1 alone caused little accumulation of F4/80-positive macrophages in corneas (Fig. 8, A and B; day 2, undetectable; day 4, 16.8 ± 8.3). By contrast, topical administration of CX3CL1 augmented alkali injury-induced increases in the numbers of intracorneal macrophages. Similarly, the administration of CX3CL1 after alkali injury enhanced mRNA expression of antiangiogenic factors, TSP-1 and ADAMTS-1, but not VEGF, compared with vehicle treatment (Fig. 8,C), whereas CX3CL1 alone failed to increase the mRNA expression of these genes (data not shown). Concomitantly, at 2 wk after the injury, the alkali-induced CNV was significantly attenuated by treatment with CX3CL1, but not with vehicle (Fig. 8, D–I). These results further suggest that CX3CL1 can be effective for prevention and/or treatment for alkali-induced CNV.

We previously proved that alkali-induced CNV can develop independently of neutrophil accumulation (10) in contrast to several observations, which indicate the crucial involvement of neutrophils in various types of neovascularization (11, 12, 39, 40, 41, 42, 43, 44). We observed that VEGF proteins were detected mainly in the accumulated mononuclear cells in this alkali-induced CNV (10). This may mirror the assumption that macrophages are a rich source of various angiogenic factors. Macrophages utilize various chemokine receptors such as CCR2 or CCR5 for their trafficking. Accordingly, CCR2- or CCR5-deficient mice exhibited reductions in macrophage accumulation and CNV after corneal injury (18, 19). Moreover, several lines of evidence would indicate that monocytes/macrophages can be divided into CCR2^highCX3CR1^low and CCR2^lowCX3CR1^high subpopulations (26, 27, 28, 29, 30). Indeed, a proportion of intracorneally accumulated macrophages expressed CX3CR1 in the absence of enhanced mRNA expression of its single ligand, CX3CL1, in the injured eyes. CX3CL1 is a membrane-bound chemokine and can be recycled between plasma membrane and a juxtanuclear compartment (45). Thus, it is tempting to speculate that alkali injury could increase the trafficking of CX3CL1 from a juxtanuclear compartment to the plasma membrane and eventually increase the amount of biologically active CX3CL1, which is localized in the plasma membrane. This, in turn, can induce intracorneal accumulation of CX3CR1-expressing macrophages.

These observations prompted us to investigate the roles of the CX3CL1-CX3CR1 interactions in alkali injury-induced CNV by comparing the molecular pathological changes between CX3CR1-deficient and WT mice. We observed that macrophage accumulation was substantially reduced in injured corneas of CX3CR1-deficient mice compared with WT mice, consistent with previous reports that Ab-mediated inhibition of CX3CL1 reduced macrophage migration into the affected organs in experimental allergic myositis and collagen-induced arthritis (46, 47). Thus, the CXC3CL1-CX3CR1 interactions may have an important role in macrophage trafficking. However, CX3CR1 deficiency failed to abrogate completely macrophage accumulation into injured cornea, suggesting a contribution of other chemokines and/or other chemotactic factors to macrophage accumulation. Indeed, we observed that alkali injury enhanced intraocular mRNA expression of CCL2 and CCL3 (data not shown), the chemokines that exhibit potent chemotactic activities for macrophages. Thus, it is likely that alkali-induced macrophage accumulation is regulated by the combined action of CX3CL1 with other macrophage-tropic chemokines such as CCL2 and CCL3.

Macrophages are presumed to be proangiogenic in various types of pathological conditions (14, 17, 36, 37). Selective macrophage depletion with clodronate liposomes inhibited laser photocoagulation-induced VEGF production and subsequent choroidal neovascularization (13). Furthermore, Ambati et al. (19, 20) reported that CNV and VEGF production was suppressed in mice lacking CCR2 or CCR5, the specific chemokine receptors for CCL2 and CCL3, respectively. Thus, CCR2- and/or CCR5-expressing macrophages can promote neovascularization, probably by producing VEGF. On the contrary, under our present experimental conditions, genetic depletion of CX3CR1 reduced alkali injury-induced macrophage accumulation but promoted CNV with little effects on intraocular VEGF expression. Similarly, Apte et al. (21) demonstrated that inhibition of macrophage accumulation promoted laser photocoagulation-induced choroidal neovascularization, whereas direct injection of macrophages suppressed it (21). Considering the proposed hypothesis on the dichotomy of macrophages based on the expression levels of CCR2 and CX3CR1 (26, 27, 28, 29, 30), CX3CR1-expressing macrophages may exhibit biological functions distinct from CCR2-expressing ones, in the course of angiogenesis.

Normal cornea lacks any vasculature under physiological conditions. This physiological corneal avascularity is presumed to be maintained by the balance between angiogenic and antiangiogenic factor expression (48, 49, 50, 51, 52). Soluble VEGF receptor 1 is presumed to be a major antiangiogenic factor responsible for the maintenance of this physiological corneal avascularity by trapping constitutively expressed VEGF (53). However, it still remains to be established whether soluble VEGF receptor 1 can counteract CNV induced by corneal injury. Although CX3CR1-deficient mice exhibited intracorneal VEGF mRNA expression to a similar extent as WT mice, they still developed a more severe form of CNV than WT mice. Thus, the VEGF/soluble VEGF receptor 1 interactions may not be responsible for enhanced CNV observed in CX3CR1-deficient mice.

Alternatively, antiangiogenic molecules, TSP-1 and TSP-2, which are also constitutively present in cornea, can preclude vasculature development (48, 54, 55). TSP-1 levels in Bruch’s membrane and choroidal vessels was decreased in age-related macular degeneration (AMD) and a reduction in TSP-1 may permit the formation of choroidal neovascularization (56). Moreover, TSP-1 deficiency and, to a lesser extent, TSP-2 deficiency resulted in enhanced CNV induced by suture placement (48). TSP-1 and TSP-2 are processed to release antiangiogenic polypeptides by the action of ADAMTS-1 (57). ADAMTS-1 at a high concentration can also inhibit endothelial migration induced by combined treatment with VEGF and basic fibroblast growth factor (38). Mirroring its antiangiogenic activities, ADAMTS-1-deficient mice exhibited delayed skin wound closure, along with exaggerated neovascularization (57). Because the enhanced expression of TSP-1, TSP-2, and ADAMTS-1 was attenuated in CX3CR1-deficient mice following alkali injury, the reduced expression of these factors may be responsible for enhanced alkali-induced CNV in CX3CR1-deficient mice.

Monocyte/macrophages are a major source of TSP-1 (58). ADAMTS-1 was also detected in infiltrating macrophages in the early phase of skin wound sites (38). TSP-1 and ADAMTS-1 proteins were consistently detected in F4/80-positive and CX3CR1-positive cell populations in this model of alkali-induced cornea injury. Moreover, CX3CL1 augmented TSP-1 and ADAMTS-1 expression by murine peritoneal macrophages. Thus, potent antiangiogenic molecules, TSP-1 and ADAMTS-1, may be expressed by CX3CR1-positve macrophages. This may account for protective roles of the CX3CL1-CX3CR1 interactions in alkali injury-induced CNV.

Chemical burns of the eyes can destroy surface epithelia and cause ischemic necrosis of eye tissues and, if not treated properly, CNV ensues and impairs vision. Alkali, once applied onto eyes, is difficult to be completely rinsed away due to its rapid penetration into the anterior chamber and, therefore, an effective maneuver(s) is required to prevent alkali injury-induced CNV. Glucocorticoids are widely used for alkali injury of eyes with limited therapeutic benefits (59, 60). In this study, we demonstrated that topical application of CX3CL1 markedly augmented the intraocular expression of antiangiogenic molecules, TSP-1 and ADAMTS-1, without affecting VEGF expression. Moreover, topical application of CX3CL1, even when started after the injury, was very effective in preventing alkali injury-induced CNV. These observations raise the possibility of CX3CL1 as a preventive drug for alkali injury-induced CNV.

Two single nucleotide polymorphisms within the CX3CR1 gene were reported to be associated with AMD and reduce CX3CR1 protein expression (61, 62). Because choroidal neovascularization is a characteristic pathological feature of AMD, depressed CX3CR1 expression may be responsible for neovascularization in AMD, as observed in alkali-induced CNV. If so, topical application of CX3CL1 may be effective for the prevention and treatment of AMD, which is the leading cause of visual impairment and blindness among people age 65 years and older.

We thank Drs. Philip M. Murphy and Ji-Liang Gao (National Institute of Allergy and Infectious Diseases, National Institutes of Health) for providing us with CX3CR1-deficient mice and Drs. Masanobu Oshima and Hiroko Oshima (Division of Genetics, Cancer Research Institute, Kanazawa University) and Dr. Toshikazu Kondo (Department of Forensic Medicine, Wakayama Medical University, Wakayama City, Japan) for their technical assistance. We express our thanks to Dr. Joost J. Oppenheim (National Cancer Institute, Frederick, MD) for his critical reading of this manuscript.

The authors have no financial conflict of interest.