The Roles of Vitreal Macrophages and Circulating Leukocytes in Retinal Neovascularization

Abstract

Purpose.: To analyze the roles of vitreal macrophages and circulating leukocytes in retinal vascular growth.

Methods.: Bone marrow (BM) cells from green fluorescent protein (GFP) transgenic mice were transplanted into postnatal day (P)1 mice after irradiation. The mice were exposed to 76% to 78% oxygen (P7–P12), to initiate oxygen-induced retinopathy (OIR). The eyes were collected at P8, P17, and P30, to analyze the engraftment of GFP-positive cells in the retina. GFP-positive peritoneal macrophages, clodronate liposomes, or control liposomes were injected into the eyes at P5 or P12 to examine the effects at P8 or P17. The number of Iba1-positive vitreal macrophages was quantified from histologic sections at P12 and P17.

Results.: Few transplanted GFP-positive cells were found in the retina at P8 in both wild-type and OIR mice. However, their number increased at P17 during retinal neovascularization in OIR. Most GFP-positive cells were Iba1-positive microglia, which comprised a minority of the total retinal microglia. Intravitreal injection of peritoneal macrophages showed only incidental migration of these cells into the wild-type retinas (P8), whereas the engraftment was more robust, typically around the neovascularization, in OIR mice (P17). Furthermore, native macrophages in the vitreous cavity became fewer (37.7% reduction) during neovascularization in OIR at P17. The selective depletion of vitreal macrophages by clodronate liposomes at P12 reduced retinal neovascularization in OIR mice by 59.0% at P17.

Conclusions.: Vitreal macrophages are attracted to the site of pathologic angiogenesis triggered by retinal ischemia, where they actively participate in vascular development.

Mammalian vascular structures develop through two major mechanisms. ¹ Vasculogenesis, which entails the arrangement of patent vascular structures by local integration of a subset of circulating blood cells called vascular endothelial precursors at the developing vascular front, ^{1 –3} does not rely on the proliferation of preexisting endothelial cells. Angiogenesis is the proliferation and development of dividing vascular endothelial cells at the tip of the growing vessels in response to local stimulatory cues without reliance on an external source of endothelial supplies. ¹ Although both mechanisms might be important during organogenesis, vascular regeneration and pathologic angiogenesis in mature animals depend mostly on angiogenesis. ⁴  

Contributions of both angiogenesis and vasculogenesis in the retina have been reported. ^{5 –8} For example, in an adult retinal neovascularization model induced by combined ischemia and forced expression of vascular endothelial growth factor (VEGF), intravenous transplantation of the hematopoietic stem cells purified from adult bone marrow caused their selective incorporation into newly growing vessels. ⁵ Some of these cells formed tubular structures and were therefore presumed to be endothelial cells, suggesting the contribution of vasculogenesis. At birth, the retina is avascular; instead, the hyaloid vasculature initially supplies the retina in the mouse. ⁹ The primary vascular network of the retina develops from the optic nerve head toward the periphery, with subsequent formation of a deeper vascular plexus during the first 2 weeks after birth. ¹⁰ This process is accompanied by the simultaneous regression of the hyaloid vasculature. ⁹ Evidence suggests that the developing retinal vascular network extends mainly by angiogenesis and that circulating blood cells are not necessary for this process. This possibility is strongly supported by the observation that neonatal retinal explants can form gross retinal vascular structures in vitro. ⁷  

Aside from the roles of leukocytes as the source of endothelial cells, the modulatory roles of blood cells in retinal vascular development have been proposed but with variations, partly attributable to various experimental conditions. For example, the genetic ablation of CD18 adhesion molecules targeting leukocytes increased the vascular density in the developing murine retina. ¹¹ Similarly, the blockade of T-cell function by an anti-CD2 antibody promoted the formation of extraretinal neovascular (NV) tufts in rat oxygen-induced retinopathy (OIR). ¹² These results imply the role of leukocytes in vaso-obliteration under both physiological and pathologic conditions. ^11,12 Conversely, circulating white blood cell depletion by irradiation decreases intraretinal revascularization in murine OIR, suggesting the role of these cells in vascular development. ¹³  

In the retina, local macrophages called microglia maintain a unique morphology despite the similarity of cellular functions and marker expressions among these cells. ¹⁴ After organogenesis, circulating macrophages and monocytes seldom penetrate the blood–retinal barrier to enter the retinal parenchyma. ¹⁵ However, under considerable retinal stress, macrophages and monocytes in the bloodstream, ciliary body, optic nerve, and possibly the vitreous and choroid migrate rapidly to the injury site. ^15,16 Most evidence suggests the role of macrophages in retinal vascular growth to be proangiogenic, ^17,18 but the associated mechanisms remain largely unknown. Particularly, little is known about the relative contributions of cells from extraretinal sources. Meanwhile, the reported role of these macrophages in photoreceptor degeneration is bivalent, behaving in both neurotoxic ¹⁹ and neuroprotective ways. ^20,21  

The results of this study show that only nonsignificant quantities of circulating leukocytes adhere to the retinal vascular lumen or integrate into the physiological vasculature in wild-type (WT) mice and pathologic NV tufts in OIR mice, denying the major direct roles of these cells in retinal vascular growth. Conversely, vitreal macrophages appeared to promote aberrant angiogenesis through migration and association with NV tufts. 

Materials and Methods

Animals and the OIR Model

All experiments were performed in accordance with the guidelines of the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and those for the use of animals at Nagoya University School of Medicine. Mice ubiquitously expressing green fluorescent protein (GFP;C57/BL6 background) ²² and congenic WT mice were used. OIR was induced by exposing mice to 76% to 78% oxygen (Proox 110; Bio Spherix Ltd., Lacona, NY) from postnatal days (P)7 to P12, which resulted in extensive obstruction of the retinal vessels. Returning the mice to room air at P12 triggered an outgrowth of pathologic retinal neovascularization. 

Immunohistochemistry

Immunohistochemical analyses were performed as previously described, ^23,24 with slight modification. The enucleated eyes were first fixed in 4% paraformaldehyde (PFA) for 2 hours. For flat mount analyses, radial incisions were made to relax the eye cups; the vitreous was removed thoroughly. The samples were permeabilized for 2 hours, blocked with 5% goat serum in phosphate-buffered saline (PBS), and incubated with a primary antibody for 12 hours and with combinations of secondary antibodies and Griffonia simplicifolia lectin-1 (GS; a marker for vessels and monocytes/microglia/macrophages; Sigma-Aldrich Corp., St. Louis, MO or Molecular Probes Inc., Eugene, OR) for 9 hours. For analyses of histologic sections, the eyes were incubated in 30% sucrose overnight and frozen in OCT compound (Tissue-Tek; Sakura Finetechnical Co. Ltd., Tokyo, Japan), followed by cryosectioning (14 or 30 μm thick). The sections were permeabilized (5 minutes), blocked (15 minutes), and incubated for 1 hour with primary antibodies and for another hour with combinations of secondary antibody, GS, and diamino-2-phenyl-indol (DAPI; Molecular Probes Inc.). 

For visualization of vasa hyaloidea propria (VHP) and vitreal macrophages after intravitreal injection of liposomes, the vitreous–retina complex was carefully peeled away from the remaining ocular structures and was stained with GS as described above. The vitreous–retina complex was placed in a 3.5-mm dish (Falcon; BD Biosciences, Franklin Lakes, NJ) containing PBS, and the images were obtained from the floating samples. 

The primary antibodies against Iba1 and its isotype IgG (Wako Pure Chemical Industries Ltd., Osaka, Japan), NG2 (Millipore Corp., Billerica, MA), and PECAM-1 (BD Biosciences) were used. 

Bone Marrow Transplantation

Bone marrow (BM) transplantation was conducted as described previously, with modifications. ¹⁵ The BM cells were harvested from femurs and tibias of GFP mice (4–8 weeks old), and 5 × 10⁶ cells were transplanted intravenously from the superficial temporal vein ²⁵ into C57BL/6 mice after irradiation (2.5 Gy) at P1. The head (therefore the eyes) of the recipient was protected with lead shields. For collecting the eyes, the animals were anesthetized intraperitoneally with tribromoethanol (0.4 mg/g body weight) and perfused with 4% PFA (300 μL/g body weight) through the left ventricle, to remove erythrocytes and nonadherent leukocytes from the vessels. The perfusate was drained from the right atrium. 

Preparation and Intravitreal Injection of Peritoneal Macrophages

First, PBS was carefully injected into the peritoneal cavity of the GFP mice (4–8-weeks-old) after carbon dioxide euthanatization, and the fluid containing the peritoneal macrophages was collected. The samples were centrifuged and resuspended in PBS. Intraocular injection of 2 × 10⁴ or 2 × 10⁵ cells in 0.5 μL was conducted in one eye. 

Preparation and Intravitreal Injection of Cells Derived from the Brain

The cerebral hemispheres were dissected from young mice (P10) when the glial development is still not as extensive as in adults. The olfactory bulbs were removed from the brain in Dulbecco's modified Eagle's Medium (DMEM; Sigma-Aldrich Corp., St. Louis, MO). The cerebral hemispheres were further cut into small cubes, gently dissociated, and incubated for 10 minutes at 37°C in 0.25% trypsin (Sigma-Aldrich Corp.). After the addition of DMEM containing 10% fetal bovine serum, the cells were washed twice with PBS. Intraocular injection of 2 × 10⁴ cells in 0.5 μL was conducted. 

Depletion of Macrophages

Liposome-encapsulated clodronate and control liposomes containing PBS only were prepared as previously described. ²⁶ Clodronate was a gift of Roche Diagnostics GmbH (Mannheim, Germany). Clodronate liposomes (0.5 μL) were injected into one eye and PBS liposomes (0.5 μL) into the other at P5 in WT mice or at P12 in the OIR model. 

Statistical Analyses

All images obtained were masked and randomized before analyses. For the distribution analyses of cells in the retinal sections, the entire retina was divided into the avascular area, the NV area, and the vascularized area without NV, and the total retinal length along the vitreoretinal surface for each category was measured. Then, cells positive for both the Iba1 antigen and DAPI staining were categorized into one of the three groups. The cells were counted from four retinal sections (14 μm thick) per eye and adjusted to the number of cells per 1 mm of retina for each group. For the quantification of intravitreally injected cells, 2 × 10⁴ cells were administered. The number of GFP-positive cells in the retina flat mount specimens was counted. 

The proportion of GFP-positive cells among the peripheral leukocytes (engraftment rate) was determined with a fluorescence-activated cell sorter (FACS, FACSCalibur; BD Biosciences) at P5, P8, P17, and P30. The total number of GFP-positive cells in the entire retina was counted at P8, P17, and P30. The number of GFP-positive cells immunopositive for Iba1, NG2, and PECAM-1 antigens was counted in 16 retinal sections from each eye. The proportion (percentage) relative to the number of total GFP-positive cells was then recorded. 

The number of Iba1-positive cells in the vitreous (adjusted for the area of the eye cup) and the area of GS-positive VHP (relative to the area of the eye cup) was determined in five sections per eye and averaged. The Iba1-positive cells in retinal flat mounts were counted in areas (500 × 500 μm) at equal distance from the optic nerve and retinal margin. The number was determined from each quadrant and averaged. The areas of NV or avascular retina relative to the entire retina were determined from GS-stained flat mount specimens. All statistical data are presented as the mean ± SEM. The difference was considered significant at P < 0.05. 

Results

Distribution of Microglia in the Developing Retina and OIR

We first studied the distribution of local retinal microglia in the WT retina. The Iba-1-positive microglia were mostly distributed evenly throughout the retina. Whereas distributions were similar during active retinal vascular growth at P7 and after the completion of vascular network at P17 (Figs. 1A, 1B), a modest accumulation of Iba-1-positive cells at the retinal vasculature was noted at P17. Next, we studied the localization of retinal microglia in two phases of OIR (Fig. 1C). During exposure to high oxygen at P8, the retina showed extensive vessel dropout (Fig. 1D). At this vaso-obliterative phase, Iba1-positive macrophages appeared to be distributed evenly in both the avascular and vascularized retina. After the mice were returned to room air for 5 days, NV tufts developed in the retina (Figs. 1E, 1F). At this NV phase, macrophages were concentrated around these abnormal vessels and were fewer in other areas of the retina. Analyses of the histologic sections confirmed that the increased macrophages in OIR were mostly localized at the surface layers of the retina, typically at the NV tufts (Figs. 1H, 1I). Quantification of Iba1-positive macrophages showed increased density of cells by 4.6- and 1.8-fold in the vascularized retina, with and without NV, respectively, compared with the avascular part (Fig. 1G). 

Figure 1.

 

Distribution of microglia in the retina. (A, B) WT retinal quadrants stained with GS (red) and Iba1 (green) at P7 (A) and P17 (B), respectively. The GS-positive retinal vessels extended toward the periphery at P7 (, the vascular front). Iba1-positive microglia were distributed evenly throughout the retina at P7, and a modest concentration of cells at the retinal vessels was observed at P17. The specimens were subjected to five freeze–thaw cycles before staining, for better visualization of the fine vascular structures in (B). (C) Experimental design for generation of OIR mice. (D–F) Retinal quadrants from an OIR mouse stained with GS (red) and for Iba1 antigen (green) at P8 (D, vaso-obliterative phase) and P17 (E, F, NV phase), respectively. At P8, the retinal vessels were obstructed in response to high oxygen with no gross alteration in distribution of Iba1-positive microglia. (○) the border between the vascularized and avascular retina (D; inset shows a magnified image of Iba1-positive microglia). In P17 OIR, Iba1-positive cells were concentrated around the NV tufts (E) that are highlighted in yellow in (F). (G) Quantification of endogenous Iba1-positive cells in histologic sections from the retina with OIR (n = 6). NV and vascularized areas represent vascularized retina, with and without NV tufts, respectively. Immunopositive cells were most concentrated at the NV tufts, whereas fewer cells were found in the avascular retina. (H, I) Histologic sections of the P17 retinas. GS-positive NV tufts at the surface layers of the retina were present in OIR (I) but not in WT (H) retinas. The Iba1-positive cells accumulated around the NV tufts (arrows). Sections, 30 μm thick. ON, optic nerve; GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer. Bar: (A, B, D, E) 200 μm; (H, I) 50 μm.

Distribution of Transplanted Leukocytes in Developing Retina and OIR

GFP-positive BM cells were transplanted at P1 (BMT; Fig. 2A). Time course profiles of the engraftment rates of donor leukocytes in the peripheral blood were analyzed at P5, P8, P17, and P30 (Fig. 2B) in the OIR and control mice. At P5, 20.9% of the peripheral leukocytes were GFP-positive. The proportion of the donor cells increased steadily thereafter in the WT and OIR mice. No difference was noted in the engraftment rates at a given age between OIR mice and controls, which enabled us to compare the number of GFP-positive cells in the retina between OIR mice and age-matched controls. Only scattered round or oval GFP-positive cells were observed, almost exclusively at the vasculature, in the retinas of both groups at P8, when extensive vaso-obliteration was observed in the OIR mice. Consequently, the number of GFP-positive cells in the retina showed no difference between the WT mice and the OIR model (Figs. 2C, 2D). Meanwhile, retinal GFP-positive cells were increased by 9.3-fold during the NV phase in the OIR eyes compared with the control eyes at P17 (Figs. 2C, 2D). Some cells found in OIR were ramified or spindle-shaped and were not associated with the vessels; they differed from the round or oval cells found almost exclusively at the vasculature in the controls. However, the distribution of GFP-positive cells was uneven in OIR retinas; there were NV tufts with no GFP-positive cells, whereas selected tufts contained large quantities of BM-derived cells (Supplementary Fig. S1). The GFP-positive cells remained numerous at P30 when the NV tufts had already regressed. Immunohistochemical analyses of P17 OIR mice identified these GFP-positive cells as mostly Iba1-positive macrophages (82.3% ± 3.1%), with one or more processes, consistent with the morphologic features of differentiated microglia (Figs. 2E). A few transplanted cells in the retina were positive for PECAM-1, an endothelial cell marker (11.5% ± 1.9%). No immunoreaction was detected among the GFP-positive cells by the anti-NG2 antibody, which targets pericytes (data not shown). 

Figure 2.

 

Analyses of transplanted BM cells during physiological and pathologic vascular growth. (A) A protocol for BM transplantation and subsequent analyses. BM transplantation (BMT) was conducted at P1. (B) Engraftment rate of donor cells among peripheral leukocytes. The rate (percentage) was measured in WT mice at P5 (n = 7), P8 (n =5), P17 (n = 5), and P30 (n = 5) and in OIR mice at P8 (n = 6), P17 (n = 6), and P30 (n = 8). A steady increase occurred in the rate after BMT. (C) Number of GFP-positive cells in the retina. The data show the total number of GFP-positive cells in the entire retina. The eyes from the same group of mice analyzed in (B) were evaluated on P8, P17, and P30. Dramatic increases in BM cells were observed in the retinas at P17 and P30 in OIR mice compared with those of the WT mice. (D) Representative images of GFP-positive cells in the retina. Only scattered GFP-positive cells (arrows) were seen in WT retinas at P8, P17, and P30 and OIR at P8. A greater number of transplanted cells were seen at P17 and P30 in OIR. (E) Differentiation of GFP-positive BM cells into Iba1-positive microglia in the retina. Right: retina stained with isotype IgG. Sections were 14 μm thick. OPL, outer plexiform layer; ONL, outer nuclear layer. Statistical data are expressed as the mean ± SEM. NS: not significant, P > 0.05. Bar: (D) 100 μm; (E) 50 μm.

Migration of Intravitreally Transplanted Peritoneal Macrophages into the Retina

To determine whether macrophages present in the vitreous are attracted to the retina, we injected GFP-positive peritoneal macrophages into the murine vitreous cavity. First, we injected these cells into the WT vitreous at P5 and collected the eyes at P8 (Fig. 3A, left). The superficial retinal vessels extended toward the periphery at P5, but almost reached the peripheral retina by P8. At P8, round or oval GFP-positive macrophages were observed at the retinal vessels. Some were found adjacent to the vessel (Fig. 3B). Next, we injected GFP-positive peritoneal macrophages in the eyes of OIR mice at P12 and analyzed the eyes at P17 (Fig. 3A, right). A vastly greater number of GFP-positive cells with various degrees of ramification was found in the retina (Fig. 3C). Although these transplanted cells resided in both the vascular and avascular retina, many of them were scattered around the NV tufts (Fig. 3C). Next, we injected BM cells instead of peritoneal macrophages. The distribution of injected BM cells and the ramified morphology of the engrafted cells were similar to those of peritoneal macrophages (Fig. 3D). Meanwhile, intravitreal injection of cells derived from the brain resulted in an almost complete lack of engraftment, averaging less than one cell identified per retina (Fig. 3E). Analyses of histologic sections from the eyes injected with peritoneal macrophages identified most GFP-positive cells at the surface of the retina, frequently near NV tufts (Fig. 3F). Quantification of injected GFP-positive peritoneal macrophages showed a large number of cells in the vascularized area without NV and avascular retina, whereas cells were found less at the NV tufts themselves (Fig. 3G). This reflects the distribution of injected cells preferentially surrounding NV tufts that are present at the border of vascularized and avascular retina. 

Figure 3.

 

Intravitreal injection of PMs and their migration into the retina. (A) Time courses of the experimental protocols. PMs were injected into the vitreous at P5 in WT mice (n = 6) or P12 in OIR mice (n = 14). The eyes were analyzed at P8 (WT) or P17 (OIR). (B) Distribution of injected GFP-positive PMs in a WT retinal flat mount at P8. Note that some PMs (green) were visible in the retina, many of which were associated with the vessels (red). (C) Distribution of injected GFP-positive PMs in the OIR at P17. A higher number of PMs were visible in the retina, many of which were scattered around the NV tufts (arrows). Inset: a magnified view of typical GFP-positive PM-derived cells with ramification. (D, E) Intravitreal injection of 2 × 10⁵ GFP-positive cells derived from BM (D) or brain (E) and their distribution in the OIR retina at P17 (n = 8 each). (D, inset) Magnified view of typical GFP-positive BM cells with ramification. (F) Distribution of injected GFP-positive PMs in the retinal section. PMs in the vitreous cavity migrated mostly into the surface layers of the retina (arrowheads). (G) Quantification of GFP-positive PM cells (2 × 10⁴ cells injected) in flat mounts from eyes with OIR at P17 (n = 6). “NV” and “vascularized area” represent vascularized retina with and without NV tufts, respectively. Sections, 30 μm thick. Statistical data are expressed as the mean ± SEM. NS, P > 0.05. Bar, (B–E) 200 μm; (F) 50 μm.

Decreased Native Vitreal Macrophages during Retinal Neovascularization in OIR

Exogenously transplanted peritoneal macrophages in the vitreous migrated toward the ischemic retina, particularly toward the NV tufts, in the OIR mice. To examine the hypothesis that endogenous vitreal macrophages are also attracted to NV tufts, we analyzed the quantities of vitreal macrophages and the relative area of VHP in ocular histologic sections from OIR mice and WT controls. The quantifications of Iba1-positive cells in the vitreous (Fig. 4B) and the relative area of VHP (Fig. 4C) showed no difference between the OIR and control eyes at P12 during the vaso-obliterative phase of OIR. However, the vitreal macrophages positive for Iba1 were fewer in the OIR eyes than in the control eyes at P17 during the NV phase (Figs. 4D, 4F). Conversely, the areas of VHP were increased greatly at P17 (Figs. 4E, 4F), consistent with a previous report. ²⁷  

Figure 4.

 

Reduced number of native vitreal macrophages during the NV phase of OIR. (A) Time course of the experimental protocol. (B–E) Number of Iba1-positive vitreal macrophages and areas of VHP relative to that of the vitreous cavity at P12 (B, C) and P17 (D, E). No differences were detected in quantities of vitreal macrophages and relative area of VHP between the OIR mice (n = 6) and the WT mice (n = 5) at P12 (B, C). However, at P17, the quantity of Iba1-positive vitreal macrophages was reduced (D), whereas areas of GS-positive VHP were increased (E) in OIR mice (n = 6) compared with the WT controls (n = 6). (F) Representative images of histologic sections of the eyes from the WT and OIR mice at P17 used for quantification. The number of Iba1-positive vitreal macrophages (arrows) was reduced, whereas the relative areas of GS-positive VHP (highlighted in blue in the bottom images) were increased in OIR mice compared with the WT controls. Statistical data are expressed as the mean ± SEM. NS, P > 0.05. Sections: 30 μm thick. V, vitreous cavity; R, retina; S, sclera. Bars, (F) 200 μm.

Reduced NV by Selective Depletion of Native Vitreal Macrophages in OIR

Our results suggest that the ischemic retina (but not the WT retina), particularly the NV tufts or the surrounding environment, produce cues to attract vitreal macrophages that are decreased in number during active neovascularization. This notion is consistent with the hypothesis that vitreal macrophages migrate into the pathologic retina. To examine the functional role of vitreal macrophages in abnormal retinal vascular growth, we used clodronate liposomes that induce selective apoptosis of macrophages through phagocytosis-mediated intracellular liberation of the cytotoxic agent. ²⁸ First, to study the tissue selectivity of the agent, we injected liposomes containing clodronate or PBS into the vitreous cavity at P5 and examined the vitreous and retina at P8 (Fig. 5A). The vitreal GS-positive macrophages were depleted completely in the ocular whole mounts from eyes injected with clodronate liposomes, but not in those from PBS-liposome–treated eyes (Fig. 5B). The VHP showed no obvious difference (Fig. 5B). Meanwhile, the quantities of retinal Iba1-positive microglia showed no difference between the retinal flat mounts from eyes treated with either agent (Figs. 5C, 5D), which indicates that clodronate liposomes selectively deplete vitreal macrophages, but not retinal microglia. The area of intraretinal vascular development was unaffected by the depletion of vitreal macrophages (Fig. 5E). 

Figure 5.

 

Proangiogenic role of vitreal macrophages in NV. (A) Time course of the experimental protocol. (B) Depletion of vitreal macrophages with intraocular injection of clodronate liposomes in WT animals at P8 (n = 6). Numerous GS-positive macrophages (inset, arrows) were present in the PBS-liposome–injected vitreous, although they were eliminated with a clodronate-liposome injection. Only VHPs (inset, arrowheads) remained after clodronate-liposome treatment. (C) Representative images of Iba1-stained retinal flat mounts from eyes treated with liposomes containing PBS (left) and clodronate (right). (D) The number of Iba1-positive retinal microglia in the eyes treated with liposomes. Clodronate liposomes did not affect the quantities of retinal microglia. (E) Effects of clodronate liposome–mediated depletion of vitreal macrophages in retinal vascular development. No effect was detected. (F) Time course of the liposome experimental protocol. (G–I) Effects of clodronate-liposome–mediated depletion of vitreal macrophages during the NV phase in OIR mice at P17 (n = 6). Both retinal NV (H) and avascular area (I) were reduced in the eyes injected with clodronate liposomes compared to those treated with PBS liposomes. The avascular retina and NV tufts are highlighted in yellow and red, respectively, below the representative GS-stained flat mounts in (G). Statistical data are expressed as the mean ± SEM. NS, P > 0.05. Bars, (B, G) 200 μm; (C) 50 μm.

Next, we injected liposomes containing either clodronate or PBS at P12 and analyzed the retina at P17 in the NV phase of OIR (Fig. 5F). The eyes injected with clodronate liposomes showed reduced NV (59.0% reduction; Figs. 5H, 5G) and also decreased the avascular area (39.6% reduction; Figs. 5I, 5G) compared with those treated with PBS liposomes. 

Discussion

We examined the roles of leukocytes and macrophages in NV growth in the murine retina with particular emphasis on the source of NV-associated microglia. The results suggest the direct involvement of circulating leukocytes in the formation of the physiological vascular network to be minor, considering the low number of BM-derived cells found in the WT retina. Most BM-derived cells in the developing retina were round or oval. They were associated almost exclusively with the vasculature, raising the possibility that they are adhesive cells in the vascular lumens that may be involved in pruning of the vessels. ¹¹ During active neovascularization in OIR, greater quantities of BM-derived cells were observed in the retina, mostly concentrated at NV tufts. However, their contributions to pathologic vascular development through direct association appear to be of little importance, since many NV tufts lacked engraftment of peripheral leukocytes. This conclusion was further supported by the observation that mice with severe combined immunodeficiency (a defect in Prkdc gene results in the functional impairment of T and B lymphocytes) showed no difference in their development of physiological retinal vessels and pathologic NV tufts in OIR compared with the congenic wild-type controls (data not shown). Meanwhile, the possible influences of peripheral leukocytes on NV formation through indirect mechanisms including their release of factors that modulate angiogenesis, remain undetermined. The identities of most GFP-positive cells in OIR were consistent with Iba1-positive ramified microglia. However, they seemed to comprise only a minor population of NV-associated microglia. The presence of a few BM-derived spindle-shaped cells associated with the vasculature prompted us to examine immunoreactivity toward NG2 and PECAM-1, which are, respectively, markers for pericytes and endothelial cells. Although no NG2-positive cells were observed, some cells that were positive for PECAM-1 were observed. Nevertheless, whether these cells truly represent vascular endothelial cells or merely perivascular cells ¹⁸ remains unclear from the results of the present study. 

Only a few peritoneal macrophages injected into the mouse eyes migrated into the retina during vascular development, although far more numerous injected cells were identified in OIR. Of particular interest is the fact that exogenous macrophages preferentially targeted the retina surrounding NV tufts in both vascularized and avascular areas; this distribution is in contrast with that of the endogenous retinal macrophages that are more concentrated at the NV tufts themselves. We suggest two possibilities or their combination to explain this difference. First, the difference in distribution might merely reflect the reduced migration of exogenous macrophages that are attracted to NV tufts. Conversely, the local retinal macrophages might migrate more efficiently toward the NV tufts. Second, the lack of nutrients derived from the circulation, including oxygen, in the avascular retina may degrade the viability of retinal macrophages, engendering death and dropout of these cells in the avascular retina. Although evidence to support these theories is scarce, prior exposure of resident retinal macrophages to high oxygen levels and fundamental differences in phenotypes evidenced by their distinct morphologic features may be confounding factors. Furthermore, injected macrophages seldom migrated into the deeper retinal parenchyma, which implies the unique role of vitreal macrophages in retinal surface disease, including NV tufts that provide local cues to attract these cells to the retina. Accordingly, the endogenous macrophages in the vitreous decreased in the vaso-proliferative phase but not in the preceding vaso-obliterative phase in OIR mice; this finding is consistent with the relocation of vitreal macrophages to the retinal surface in response to local angiogenic signals. The functional involvement of the vitreal macrophages in NV development was confirmed further by showing that the specific depletion of vitreal macrophages by intraocular injection of clodronate liposomes decreased NV formation, as previously reported. ¹² However, we speculate that the contributions of vitreal macrophages may be marginal, because few macrophages are present in the vitreous. Therefore, even considering their high affinity for surface retinal structures, it is possible that resident retinal microglia, the probable main source of the NV-associated macrophages, is far more influential. The fact that treatments and genetic defects that target the differentiation of all ocular macrophages nonspecifically show potent antiangiogenic effects greater than those observed in this study supports this theory. ¹⁷  

In conclusion, the results demonstrate that intraocular macrophages, including those in the vitreous cavity, are attracted to the site of pathologic angiogenesis triggered by retinal ischemia, where they participate actively in vascular growth. The role of circulating leukocytes may be limited. These findings provide the rationale for developing antiangiogenic treatments that target macrophages in the eyes, with less emphasis on leukocytes in the blood. 

Supplementary Materials

Figure sf01, PDF - Figure sf01, PDF 

Footnotes

 Supported by Grant-in-Aid for Scientific Research B 21791676 from the Ministry of Education, Culture, Sports, Science, and Technology of Japan. The clodronate was a gift of Roche Diagnostics GmbH, Mannheim, Germany. The GFP mouse strain (RBRC00267) was provided by RIKEN BRC through the National Bio-Resource Project of the MEXT (Ministry of Education, Culture, Sports, Science, and Technology), Japan.

Footnotes

 Disclosure: K. Kataoka, None; K.M. Nishiguchi, None; H. Kaneko, None; N. van Rooijen, Roche Diagnostics GmbH (F); S. Kachi, None; H. Terasaki, None