Retinal ischemia can cause vision-threatening pathological neovascularization. The mechanisms of retinal ischemia are not fully understood, however. Here we have shown that leukocytes prune the retinal vasculature during normal development and obliterate it in disease. Beginning at postnatal day 5 (P5) in the normal rat, vascular pruning began centrally and extended peripherally, leaving behind a less dense, smaller-caliber vasculature. The pruning was correlated with retinal vascular expression of intercellular adhesion molecule-1 (ICAM-1) and coincided with an outward-moving wave of adherent leukocytes composed in part of cytotoxic T lymphocytes. The leukocytes adhered to the vasculature through CD18 and remodeled it through Fas ligand (FasL)-mediated endothelial cell apoptosis. In a model of oxygen-induced ischemic retinopathy, this process was exaggerated. Leukocytes used CD18 and FasL to obliterate the retinal vasculature, leaving behind large areas of ischemic retina. In vitro, T lymphocytes isolated from oxygen-exposed neonates induced a FasL-mediated apoptosis of hyperoxygenated endothelial cells. Targeting these pathways may prove useful in the treatment of retinal ischemia, a leading cause of vision loss and blindness.
Retinal ischemia is correlated with, and can cause, the pathological retinal neovascularization that occurs in retinopathy of prematurity. In the eyes, pathological neovascularization often results in blindness through hemorrhage, retinal detachment and glaucoma. At present, there is no available treatment to prevent retinal ischemia, and thus ophthalmologists ablate the ischemic (hypoxic) retina in an attempt to suppress the subsequent pathological neovascularization. Antiangiogenic agents may eventually prove helpful, but they would probably further aggravate the ongoing retinal ischemia, which would limit their ultimate efficacy. Although angiogenic mechanisms have been investigated intensively1,
2,
3,
4,
5, retinal ischemia remains poorly understood. As the prevention of retinal ischemia both protects the retina itself and suppresses pathological neovascularization, the elucidation of the cellular and molecular mechanisms underlying retinal ischemia is an important goal that is directly linked to the prevention of neonatal blindness.
The retinal vasculature grows and remodels during development and disease in response to local oxygen concentrations. In the rodent, the retinal vasculature develops postnatally and extends to fill the peripheral avascular retina. Shortly after its formation, the dense primary vasculature is remodeled into the secondary vasculature, with small avascular zones forming around the oxygenated retinal arterioles and the dense primary vasculature giving way to a narrower, smaller caliber secondary vasculature. When the retina is exposed to high concentrations of oxygen, as occurs in premature infants receiving supplemental oxygen, large areas of the newly formed retinal vasculature are obliterated, leading to retinal ischemia when the child breathes room air (with 21% oxygen). The ischemic retinopathy model manifests extensive retinal ischemia and is often used to gain insights into human diseases1,
2,
3,
4,
5 like retinopathy of prematurity.
Normal developmental vascular remodeling and pathological vaso-obliteration share similar features. Both processes are associated with endothelial cell apoptosis6,
7, which depends in part on the hyperoxia-induced downregulation of vascular endothelial growth factor (VEGF)6, a known endothelial cell survival factor6,
8,
9,
10. The periods of normal vascular pruning and pathological vaso-obliteration are restricted to a temporal 'plasticity window' that coincides with the formation of a partially developed, pericyte-poor vasculature11.
Some information is known about the Fas-FasL−mediated inhibition of vessel growth. In age-related macular degeneration, retinal pigment epithelial cells inhibit choroidal vessel growth through a Fas-FasL−mediated apoptosis of choroidal endothelial cells12. In murine carcinoma, FasL+ T lymphocytes suppress tumor vessel growth13.
Here we focused on the involvement of leukocyte-endothelial FasL-Fas interactions in the genesis of the two types of retinal vascular regression, physiological vascular remodeling and pathological vaso-obliteration. The results of these studies indicated cellular and molecular targets for the prophylactic treatment of retinal ischemia.
Leukocytes and retinal vascular development After removal of non-adherent leukocytes by perfusion, we injected FITC-labeled concanavalin A lectin (conA) into the heart to image the retinal vasculature and leukocytes, as described14. From P1 to P12, the developing retinal vasculature of the rats grew toward the retinal periphery at a rate of approximately 1 disc diameter per day (Fig. 1a). On P5, the number of adherent leukocytes increased considerably. As the vasculature grew toward the periphery, the zone containing the leukocytes, which was two to three disc diameters wide, migrated toward the periphery, staying two to three disc diameters behind the leading edge of vascular growth (Fig. 1a, arrowheads).
Figure 1. Temporal and spatial correlation of leukocyte adhesion with retinal vascular remodeling.
(a) Developing rat retina, P1−P12. The dense primary vasculature (arrowheads) with abundant leukocytes was remodeled outwardly. The initial pruning occurred centrally (posteriorly) on P5−P6, followed by peripheral waves of pruning from P6 to P12. (b) Magnification of white-outlined boxes in a. The vasculature was denser on P4−P5 than on P3, and then became much less dense on P6. The leukocytes (arrows) adhered to the dense vasculature (arrowheads) just before the pruning and disappeared after the pruning. Bottom panel, central (left) to peripheral (right) retina. (c) Green fluorescence from FITC-coupled conA (left) and red fluorescence from the antibody to CD45 (middle) identified conA-stained cells (arrows, left) within the P5 retinal vessels as being CD45-positive leukocytes (arrows, middle) when the images were superimposed (right). (d) The number of leukocytes adhering to the entire retinal vascular bed peaked on P5 and 'retreated' to a new plateau on P6−P11 (n = 6−15 for each postnatal day). (e) Both vascular density and the number of branch-surrounded spaces increased at P4 and peaked at P5 (n = 10−15 for each postnatal day). (f) The increase in vascular density on P4 was accompanied by an increase in adherent leukocytes on P5, followed by vascular pruning on P6 (n = 10−15 for each postnatal day). (g,h) Western blotting for ICAM-1 showed more ICAM-1 protein in P4−P6 rat retinas than in fully pruned adult retinas. Correction was made by CD34 signal for vascularity. Similar results were obtained in three separate experiments. (i) Green fluorescence from the FITC-coupled conA (top) and red fluorescence corresponding to the antibody to ICAM-1 (bottom) showed increased ICAM-1 expression (bottom) in the dense vasculature at P8 with many associated adherent leukocytes (arrows, top).d, e and f: Error bars indicate s.d.
As the zone of adherent leukocytes moved progressively toward the periphery, it left behind in its wake a pruned and remodeled secondary vasculature (Fig. 1b, P3−P6 and P8). Immunohistochemistry confirmed that the intravascular conA−stained blood elements were leukocytes, as they were positive for CD45 (leukocyte common antigen; Fig. 1c). The total number of adherent leukocytes in the rat retina peaked on P5 and reached a plateau from P6 to P11, as the vascular pruning moved peripherally (Fig. 1d). By P12, all pruning was completed and the leukocytes had all but disappeared (Fig. 1a). When the leukocyte and vascular changes were compared within a zone five disc diameters from the disc, it was apparent that the decrease in both vascular density and branch-surrounded spaces was preceded by an increase in leukocyte density, indicating a cause-effect relationship (Fig. 1e,f).
We next studied the mechanism of adhesion and the functions of the leukocyte integrin CD18 and its corresponding ligand, ICAM-1. Western blotting showed much more ICAM-1 in P4−P6 retinas actively undergoing remodeling than in normal adult rat retinas in which vascular remodeling was absent (Fig. 1g,h). ICAM-1 immunohistochemistry in P8 retinas localized the upregulation of ICAM-1 expression to the dense primary vasculature containing adherent leukocytes (Fig. 1i). In contrast, evidence of ICAM-1 expression was faint in P1 and P12 retinas, time points at which leukocyte adhesion and remodeling were suppressed (data not shown).
To investigate the causal involvement of leukocytes in vascular pruning, we studied the P4−P6 period, during which the biggest changes in leukocyte numbers and vascular density took place (Fig. 1b, P4−P6, and d). Rats received either a CD18-neutralizing antibody or a control antibody, and the leukocyte and vascular changes were quantified within the five−disc diameter zone surrounding the optic disc. CD18 blockade significantly inhibited leukocyte adhesion on P5 (Fig. 2a−c; 59.9 15.6 leukocytes versus 119.2 21.9 leukocytes, P < 0.01) and suppressed 76.9% of the vascular density change by P6 (Fig. 2d−f; 36.7 3.3% versus 27.3 2.7%, P < 0.01). The number of branch-surrounded spaces was also suppressed by 82.9% (Fig. 2g; 55.6 3.1% versus 38.7 2.9%, P < 0.01).
Figure 2. Suppression of leukocyte adhesion and vascular pruning with CD18 blockade.
(a,b) CD18 blockade in rats reduced the number of adherent leukocytes on P5 (a), whereas the control antibody had no effect (b). (c) Leukocyte adhesion was significantly inhibited in the rats treated with antibody to CD18 (n = 12, 59.9 15.6 leukocytes) compared with the control group (n = 10, 119.2 21.9 leukocytes; *, P < 0.01). (d,e) CD18 blockade led to the suppression of vascular pruning on P6 (d), whereas vascular pruning proceeded normally in rats receiving control antibody (e). (f) There was a significant difference in vascular density within five disc diameters of the optic disc between rats treated with antibody to CD18 (n = 10, 36.7 3.3%) and control rats (n = 8, 27.3 2.7%; *, P < 0.01). (g) There was a significant difference in the number of branch-surrounded spaces located within five disc diameters of the optic disc between rats treated with antibody to CD18 (n = 10, 55.6 3.1) and control rats (n = 8, 38.7 2.9; *, P < 0.01). (h,i) CD18-deficient mice (h) had delayed vascular pruning in the periphery at P9 compared with age-matched wild-type mice (i). (j) There was a significant difference in vascular density within three disc diameters from the periphery between CD18-deficient mice (n = 10, 40.9 4.5%) and wild-type mice (n = 10, 24.5 4.8%; *, P < 0.01). Ab, antibody; anti-CD18 Ab, antibody to CD18; CD18-KO, CD18-deficient.
To confirm that the vascular pruning was CD18 dependent, we compared the P9 retinal vasculature of CD18-deficient mice15 with that of wild-type mice. We quantified the peripheral vascular density in a concentric three−disc diameter zone extending to the vascular front. By P9, the CD18-deficient mice showed much less pruning of the retinal vasculature than did wild-type mice (Fig. 2h−j; 40.9 4.5% versus 24.5 4.8%, P < 0.01). Nevertheless, the adult CD18-deficient mice had a normally pruned retinal vasculature (data not shown), a finding consistent with the fact that the CD18-deficient strain was 'leaky', or only partially deficient in CD18 (ref. 15).
FasL is expressed mainly on T lymphocytes16, whereas Fas is ubiquitously expressed on a variety of cell types16. To study the involvement of FasL in vascular remodeling, we treated rats with a FasL-neutralizing antibody or an isotype control. FasL blockade led to more than 75% suppression of vascular pruning on P6 (Fig. 3a−d; 79.1%, 37.1 1.5% versus 29.5 3.4%, P < 0.01 for vascular density and 76.2%, 54.4 3.2% versus 39.5 3.0%, P < 0.01 for branch-surrounded space). We found no significant difference in leukocyte density on P5 between rats treated with FasL antibody and those treated with control antibody (115.6 25.8 leukocytes versus 113.2 23.2 leukocytes; P > 0.05). CD8 immunohistochemistry showed that many of the adherent leukocytes were cytotoxic T lymphocytes (Fig. 3e−g). To further confirm the involvement of T lymphocytes, we used a neutralizing antibody to CD2. CD2 is specific to T lymphocytes and is essential in the adhesion of T lymphocytes to target cells. CD2 blockade significantly suppressed vascular pruning (Fig. 3h−k; 75.1%, 36.4 2.9% versus 28.7 3.0%, P < 0.01 in vascular density and 78.6%, 53.7 3.2% versus 37.0 3.2%, P < 0.01 in branch-surrounded space). Staining for CD25 confirmed the activation of T lymphocytes, which were positive for CD25 (Fig. 3l−n).
Figure 3. Suppression of vascular pruning with FasL blockade.
(a,b) FasL blockade in rats led to a suppression of vascular pruning on P6 (a), whereas vascular pruning proceeded normally in the rats receiving the control antibody (b). (c) There was a significant difference in vascular density between the group treated with antibody to FasL (n = 13, 37.1 1.5 %) and the control group (n = 11, 29.5 3.4%; *, P < 0.01). (d) There was a significant difference in the number of branch-surrounded spaces between the group treated with antibody to FasL (n = 13, 54.4 3.2) and the control group (n = 11, 39.5 3.0; *, P < 0.01). Error bars indicate s.d. (e−g) Green fluorescence from the FITC-coupled conA (e) and red fluorescence from the antibody to CD8 (f) identified some of the conA−stained cells (arrowheads, e) within the P5 retinal vessels as being CD8-positive T lymphocytes (arrows, e) when the images were superimposed (g). (h,i) CD2 blockade in rats led to a suppression of vascular pruning on P6 (h), whereas vascular pruning proceeded normally in the rats receiving the control antibody (i). (j) There was a significant difference in vascular density between the group treated with antibody to CD2 (n = 12, 36.4 2.9 %) and the control group (n = 10, 28.7 3.0%; *, P < 0.01). (k) There was a significant difference in the number of branch-surrounded spaces between the group treated with antibody to CD2 (n = 12, 53.7 3.2) and the control group (n = 10, 37.0 3.2; *, P < 0.01). (l−n) Green fluorescence from the FITC-coupled conA (l) and red fluorescence from the antibody to CD25 (m) identified some of the conA-stained cells (arrowheads in l) within the P5 retinal vessels as being CD25-positive T lymphocytes (arrows in m) when the images were superimposed (n). Ab, antibody; anti-FasL Ab, antibody to FasL; anti-CD2 Ab, antibody to CD2 (CD2 blockade).
Leukocytes and vaso-obliteration We studied the involvement of leukocytes in oxygen-induced vaso-obliteration using the ischemic retinopathy model. Neonatal retinas had few leukocytes and a stable vascular density up to P3 (Fig. 1e). Exposure to 80% oxygen on P2 led to rapid and profound vaso- obliteration (Fig. 4a). The number of adherent leukocytes increased 6 h after systemic exposure to 80% oxygen (Fig. 4b), and the vascular density decreased shortly thereafter (Fig. 4c). In concert with vascular density changes, the number of branch-surrounded spaces also decreased (Fig. 4d). Western blotting for ICAM-1 showed a time-dependent increase in retinal ICAM-1 at 6 h, the time at which obliteration of the vasculature had begun (Fig. 4e,f).
Figure 4. Temporal and spatial correlation of adherent leukocytes with ischemic retinopathy.
(a) Exposure of rats to 80% oxygen on P2 led to rapidly progressive vaso-obliteration of the retinal vasculature over 12 h. (b) Magnification of the white-outlined boxes in a. Adherent leukocytes (arrows) accumulated within 6 h of exposure to 80% oxygen. (c) The number of adherent leukocytes peaked at 6 h and the vascular density decreased substantially after 9 h (n = 7−11 for each time point). (d) The number of branch-surrounded spaces correlated with the change of vascular density, as shown in c. (e,f) Western blotting showed a time-dependent increase in ICAM-1 protein that coincided with the leukocyte adhesion. Similar results were obtained in three separate experiments.
To investigate the causal involvement of leukocytes in the obliteration of the vasculature, we treated rats with CD18-neutralizing and control antibodies, as described above. We evaluated vascular density and branch-surrounded spaces 12 h after the rats began breathing 80% oxygen, a time at which the vasculature ordinarily was obliterated. CD18 blockade inhibited approximately 60% of the vaso-obliteration (Fig. 5a−d; 64.2%, 22.0 2.4% versus 13.2 2.5%, P < 0.01 for vascular density and 58.0%, 29.1 2.2% versus 15.2 2.0% for branch-surrounded spaces).
Figure 5. Suppression of vaso-obliteration with CD18 blockade.
(a,b) CD18 blockade in rats led to the suppression of vaso-obliteration at 12 h (a), whereas vaso-obliteration was unimpeded with the control antibody (b). (c) There was a significant difference in vascular density between rats treated with antibody to CD18 (n = 9, 22.0 2.4%) and control rats (n = 8, 13.2 2.5%; *, P < 0.01). (d) There was a significant difference in the number of branch-surrounded spaces between rats treated with antibody to CD18 (n = 9, 29.1 2.2) and control rats (n = 8, 15.2 2.0; *, P < 0.01). (e,f) CD18-deficient mice (e) showed less vaso-obliteration after a 48-h exposure to 80% oxygen than did wild-type mice (f). (g) There was a significant difference in residual vascular network between CD18-deficient mice (n = 11, 40.4 3.3%) and wild-type mice (n = 9, 27.3 3.9%; *, P < 0.01). Ab, antibody; anti-CD18 Ab, antibody to CD18; CD18-KO, CD18-deficient mice.
To confirm that the vaso-obliteration was CD18 dependent, we quantified the vasculature of both CD18-deficient mice and age-matched wild-type control mice after they had breathed 80% oxygen for 48 h, which began at P7. The CD18-deficient mice showed much less vaso-obliteration of retinal vasculature than did wild-type controls (Fig. 5e−g, 40.4 3.3% versus 27.3 3.9%, P < 0.01).
To determine if the vaso-obliteration was FasL dependent, we treated the rats with a FasL-neutralizing antibody or an isotype control antibody. FasL blockade significantly suppressed the vaso- obliteration at 12 h (Fig. 6a−d, 69.7%, 22.9 4.6% versus 15.9 3.8%, P < 0.01 for vascular density, and 64.6%, 30.9 2.3% versus 16.0 2.1% for branch-surrounded space, P < 0.01), confirming the involvement of FasL in the pathogenesis of retinal ischemia.
Figure 6. Suppression of vaso-obliteration and in vitro endothelial cell apoptosis with FasL blockade.
(a,b) FasL blockade (a) led to suppression of vaso-obliteration at 12 h, whereas vaso-obliteration was unimpeded with the control antibody (b). (c) There was a significant difference in vascular density between the group treated with antibody to FasL (n = 10, 22.9 4.6%) and the control group (n = 8, 15.9 3.8%; *, P < 0.01). (d) There was a significant difference in the number of branch-surrounded spaces between the group treated with antibody to FasL (n = 10, 30.9 2.3) and the control group (n = 8, 16.0 2.1; *, P < 0.01). (e−i) Rhodamine-labeled apoptotic cells in the cultures of leukocytes and endothelial cells were detected with TUNEL staining. T lymphocytes were labeled with carboxyfluorescein diacetate succinimidyl ester (green). Compared with normal T cells (e), T cells from hyperoxygenated neonates (f) considerably increased the number of TUNEL-positive hyperoxic endothelial cells. The process was inhibited substantially through FasL antibody blockade (h). Coincubation of normoxic endothelial cells with T cells from oxygenated neonates resulted in significantly fewer apoptotic endothelial cells (i) than co-incubation with hyperoxic endothelial cells (f). (g) FasL blockade led to significant suppression of cell death (n = 24−28 for each condition; *, P < 0.01). , normal T cells; and , oxygenated neonate T cells; , oxygenated neonate T cells plus antibody to FasL.
To confirm that T lymphocytes can cause the FasL-mediated apoptosis of hyperoxic endothelial cells in vitro, we evaluated cell death using a system of leukocytes and endothelial cells cultured together. Compared with T cells isolated from rats breathing 21% oxygen (Fig. 6e), T cells from neonates undergoing the process of hyperoxia-induced vaso-obliteration (Fig. 6f) increased the apoptosis of endothelial cells pre-stimulated with 80% oxygen (Fig. 6g; 0.34 0.30 cells/mm2 versus 6.40 2.34 cells/mm2, P < 0.01). Administration of a FasL neutralizing antibody (Fig. 6h) led to a 69.3% suppression of the apoptosis (Fig. 6g, 6.40 2.34 cells/mm2 versus 1.96 0.88 cells/mm2, P < 0.01) caused by the T cells from the neonates incubated with high oxygen concentrations. We found no decrease in cell death when a control nonimmune antibody was applied to the T cells from neonates incubated with high oxygen concentrations (6.40 2.34 cells/mm2 versus 6.26 2.41 cells/mm2, P > 0.05). Furthermore, the endothelial cell death induced by the T cells from neonates incubated with high oxygen concentrations was suppressed (Fig. 6g, 6.40 2.34 cells/mm2 versus 0.26 0.24 cells/mm2, P < 0.01) when nonstimulated (normoxic) endothelial cells were used (Fig. 6i), as compared with results obtained with endothelial cells exposed to 80% oxygen (Fig. 6f).
Discussion Our data have shown that leukocytes remodeled the retinal vasculature during development and obliterated it in disease. Both processes involved the CD18-dependent adhesion of leukocytes to the nascent vascular wall, followed by FasL-mediated endothelial cell death. The results were confirmed in two species and two models, using both neutralizing antibodies and genetically altered animals. The idea of the involvement of FasL was directly supported by the results of the in vitro experiments using endothelial cells and T cells cultured together.
A previous report17 showed that macrophages are associated with the complete regression of the pupillary membrane; this membrane is superfluous after the development of the eye. In contrast, our results indicated direct causal involvement of cytotoxic T lymphocytes in the physiological vascular remodeling that occurs during retinal development and the pathological vaso-obliteration that characterizes retinal disease. This is a distinctly different process in terms of the selective, not complete, regression of endothelial cells.
During the development of the retinal vasculature, vessels grow toward the avascular periphery and are guided by glial cells that secrete hypoxia-induced VEGF18. Our data here indicate that the dense primary vasculature may result in tissue hyperoxia, leading to the recruitment of leukocytes to readjust the vascular density to an optimal level. The adhesion of these leukocytes to specific areas of the vasculature is guided by the spatially coordinated expression of ICAM-1. Hyperoxia can upregulate the endothelial expression of ICAM-1, which interacts with the leukocyte 2 (CD18)- integrins19,
20,
21. The interaction of CD18 with ICAM-1 results in leukocyte activation22, leading to the FasL-mediated apoptosis of endothelial cells. Excessive oxygen can also result in the generation of reactive oxygen species, causing cytotoxicity and tissue damage. Conversely, antioxidants can suppress hyperoxia-induced ICAM-1 expression on endothelial cells19. Vascular pruning may represent, in part, a defense mechanism elaborately programmed to protect the retinal tissue from oxygen toxicity, in which activated cytotoxic T lymphocytes selectively prune the overly dense primary vasculature to an optimal level by targeting and attacking endothelial cells expressing ICAM-1. The involvement of T lymphocytes was further confirmed by the neutralization of CD2, a T-cell-specific adhesion molecule important for T-cell-mediated immunity. The idea of the functional involvement of cellular immunity in the process of vascular remodeling is also supported by the presence of adherent leukocytes positive for CD25, an interleukin-2 receptor that is expressed on activated T lymphocytes.
In the ischemic retinopathy model, exposure of P2 retina to 80% oxygen triggers leukocytes to adjust the vascular density to meet the tissue oxygen demand. The lung is another organ in which hyperoxia can lead to a related pathology20. When the rodent lung is exposed to hyperoxia, vascular endothelial cells upregulate expression of ICAM-1 and are damaged by leukocytes. To a certain degree, mechanisms seem to be shared among organs.
Although the less-than-complete inhibition of pruning can be explained by the 'leaky' nature of the CD18-deficient mouse15, it is also possible that compensatory mechanisms may have intervened. For example, vascular regression may be mediated not only by the FasL-Fas pathway but also by other apoptotic pathways, including the perforin-granzyme system. Another possible mechanism is that oxidative stress leads to endothelial cell apoptosis through FasL-Fas endothelial-endothelial interaction23,
24.
The hyperoxia-induced suppression of VEGF expression, which speeds endothelial cell apoptosis, was previously shown to be required for vaso-obliteration6. The downregulation of this survival factor probably synergizes the leukocyte-mediated pruning of the vasculature found here. The same process is probably exaggerated during the breathing of 80% oxygen that occurs in the ischemic retinopathy model. With systemic hyperoxia, the leukocyte-mediated retinal vascular obliteration is compensatory, realigning the vascular supply with the needs of the retina. When the oxygen concentration is reduced to 21% (room air), however, the retinal vascular density is suddenly inadequate, leading to pathological neovascularization.
Retinal ischemia−induced neovascularization causes blindness in various diseases such as diabetic retinopathy and retinopathy of prematurity. At present, there is no treatment available to prevent retinal ischemia. VEGF is an endothelial cell survival factor and VEGF administration was shown to rescue the retinal vessels in ischemic retinopathy6. VEGF is also potent mitogen of endothelial cells, however, and such treatment could also worsen ischemia-induced neovascularization. Our study has identified specific cellular and molecular targets for the potential prevention of retinal ischemia. More work is needed to establish the validity of this therapeutic approach in the vision-threatening ischemic retinopathies.
Methods Animals and the ischemic retinopathy model. All animal experiments followed the guidelines of the Association for Research in Vision and Ophthalmology and were approved by the Animal Care Committee of the Massachusetts Eye & Ear Infirmary. We used Long-Evans rats (Charles River Laboratories), CD18-deficient mice (C57Bl/6J-Itgb2tm1bay)15 and C57Bl/6J mice (Jackson Laboratories); the C57Bl/6J mice were wild-type controls. P2 rats with their nursing mothers were put in airtight incubators connected to an 80% oxygen tank for 3−12 h. P7 mice were placed in the same environment for 48 h. Oxygen levels were monitored with a GPR-20 Series % O2 Analyzer (Analytical Industries).
Lectin labeling of adherent retinal leukocytes. The retinal vasculature and adherent leukocytes were labeled with FITC-coupled conA (Vector Laboratories). A published perfusion-labeling technique14 was used, with slight modification. Animals were anesthetized with intramuscular xylazine hydrochloride and ketamine hydrochloride. The chest cavity was carefully opened and a perfusion needle was introduced into the left ventricle. The animals were perfused with 500 ml PBS per kg body weight to remove erythrocytes and non-adherent leukocytes. The perfusate was drained through the right atrium. Perfusion with FITC-coupled conA (40 g/ml in PBS; pH 7.4; 5 mg per kg body weight) was then done to label adherent leukocytes and vascular endothelial cells, followed by removal of the residual unbound lectin by PBS perfusion. The retinas were carefully removed and fixed with 1% paraformaldehyde. Images of the flat mounts were obtained using an epifluorescence microscope (DM RXA; Leica).
Systemic CD18, FasL and CD2 blockade. Animals were randomly assigned to receive intraperitoneal injections of 1 mg/kg of a mouse neutralizing antibody to rat CD18 (clone WT.3; Serotec) or 1 mg/kg of a mouse isotype control antibody (R&D Systems). In separate experiments, animals received 1 mg/kg of either a hamster antibody to mouse FasL (clone MFL4; BD PharMingen) or a hamster isotype control antibody (BD PharMingen). All blocking reagents were injected at a final concentration of 1 mg/ml in sterile PBS 1−2 d before evaluation. In the CD2 blockade experiments, animals received intraperitoneal injections of 5 mg/kg per day of a mouse neutralizing antibody to rat CD2 (ref 25; clone OX-34; Research Diagnostics) or a mouse isotype nonimmune antibody (R&D Systems).
ICAM-1 and CD34 western blotting. Animals were killed with an overdose of anesthesia and their retinas were immediately isolated. The retinas were homogenized in lysis buffer and centrifuged at 4 °C for 10 min. The supernatant was collected and mixed with sample buffer. Samples (each with 100 g total protein) were boiled for 3 min, separated by SDS-PAGE and electroblotted onto a polyvinylidene difluoride membrane (BioRad). Nonspecific binding was blocked with 5% normal goat serum, and the membranes were incubated at room temperature for 1 h with a rabbit polyclonal antibody to human ICAM-1 (1:200 dilution; Santa Cruz Biotechnologies) or a rabbit polyclonal antibody to human CD34 (1:200 dilution; Santa Cruz Biotechnologies), followed by incubation with a horseradish peroxidase- conjugated goat antibody to rabbit immunoglobulins. The signals were visualized using the Opti-4CNTM substrate kit (BioRad) according to the manufacturer's protocol.
CD45, CD8, CD25 and ICAM-1 immunofluorescence. Adherent leukocytes were labeled with FITC-coupled conA as described above. Flat mounts were permeabilized with 0.5% Triton X (Sigma) in PBS for 24 h and nonspecific binding was blocked with 5% normal goat serum. The retinas were then incubated overnight at 4 °C with a mouse monoclonal antibody to rat CD45 (1:500 dilution; clone OX-1; BD PharMingen), CD8 (1:100 dilution; MRC OX-8; Serotec), CD25 (1:100 dilution; OX-39; Research Diagnostics) or ICAM-1 (1:100 dilution; G-5; Santa Cruz Biotechnologies), followed by incubation with a Texas Red−conjugated goat antibody to mouse immunoglobulins. The flat mounts were prepared with a fluorescence mounting medium, and then images were obtained with an epifluorescence microscope.
Culture of endothelial cells and T lymphocytes. Before being killed, rat neonates were incubated for 12 h with 80% oxygen as described above. Blood was collected before the rats were killed, and the T lymphocytes were purified by magnetic cell sorting using MicroBeads conjugated to a monoclonal anti-rat pan T-cell antibody (clone OX-52; Miltenyi Biotec), according to the manufacturer's protocol. Control T cells were isolated from animals maintained in room air. Human microvascular endothelial cells (Cascade Biologics) at passages six to eight were seeded at a density of 2 105 cells per well and were stimulated for 6 h at 37 °C with 80% oxygen or 21% oxygen in a humidified air-tight chamber (Modular Incubator Chamber; Billups-Rothenberg). The isolated T cells were incubated for 15 min at 37 °C with 50 M carboxyfluorescein diacetate succinimidyl ester (Molecular Probes). The fluorescent cells were washed and then incubated (8 105 cells/ml, 100 l per well) with endothelial monolayers for 4 h, after which the lymphocytes were removed and the endothelial monolayer was washed. To determine whether the apoptosis is FasL mediated, an antibody to FasL (clone MFL4; BD PharMingen) or an isotype control antibody (BD PharMingen) was applied for 10 min at 37 °C at a concentration of 10 g/ml to the T-cell suspension before the cells were cultured together. Cell death was assayed as described below.
TUNEL assay. Free 3'-OH DNA termini were labeled using the TUNEL procedure according to the manufacturer's recommendations (Intergen). Adherent labeled lymphocytes and endothelial cells were fixed in 1% paraformaldehyde for 10 min at room temperature. DNA fragments were labeled with a digoxigenin-labeled nucleotide and then allowed to bind an antibody to digoxigenin conjugated to a rhodamine reporter molecule. Apoptotic cells were detected using a CD-330 charge-coupled device camera (Dage-MIT) attached to an epifluorescence microscope (MZ FLIII; Leica). A minimum of eight fields each in three separate experiments was analyzed per condition.
Morphometric and statistical analysis. All results were expressed as mean s.d. The leukocytes in each flat mount were counted independently by two investigators. The vascular density was evaluated using NIH Image. In the photographed flat mounts, the vascular area was measured in pixels. The vascular area was divided by the retinal area (in pixels) to arrive at the vascular density. To quantify branching, the spaces surrounded by branching vessels were counted. More than ten areas measuring 300 m 300 m were counted per specimen and results were averaged. The values were processed for statistical analyses (Mann-Whitney U test). Differences were considered statistically significant at P < 0.05.
Received 5 February 2003; Accepted 19 March 2003; Published online: 5 May 2003.
REFERENCES
Aiello, L.P. et al. Suppression of retinal neovascularization in vivo by inhibition of vascular endothelial growth factor (VEGF) using soluble VEGF-receptor chimeric proteins. Proc. Natl. Acad. Sci. USA92, 10457-10461 (1995). | PubMed | ChemPort |
Das, A., McLamore, A., Song, W. & McGuire, P.G. Retinal neovascularization is suppressed with a matrix metalloproteinase inhibitor. Arch. Ophthalmol.117, 498-503 (1999). | PubMed | ChemPort |
Friedlander, M. et al. Involvement of integrins v3 and v5 in ocular neovascular diseases. Proc. Natl. Acad. Sci. USA93, 9764-9749 (1996). | Article | PubMed | ChemPort |
Moravski, C.J. et al. Retinal neovascularization is prevented by blockade of the renin-angiotensin system. Hypertension36, 1099-1104 (2000). | PubMed | ChemPort |
Yoshida, A., Yoshida, S., Ishibashi, T., Kuwano, M. & Inomata, H. Suppression of retinal neovascularization by the NF-B inhibitor pyrrolidine dithiocarbamate in mice. Invest. Ophthalmol. Vis. Sci.40, 1624-1629 (1999). | PubMed | ChemPort |
Alon, T. et al. Vascular endothelial growth factor acts as a survival factor for newly formed retinal vessels and has implications for retinopathy of prematurity. Nat. Med.1, 1024-1028 (1995). | PubMed | ChemPort |
Hughes, S. & Chang-Ling, T. Roles of endothelial cell migration and apoptosis in vascular remodeling during development of the central nervous system. Microcirculation7, 317-333 (2000). | Article | PubMed | ChemPort |
Spyridopoulos, I. et al. Vascular endothelial growth factor inhibits endothelial cell apoptosis induced by tumor necrosis factor-: balance between growth and death signals. J. Mol. Cell. Cardiol.29, 1321-1330 (1997). | Article | PubMed | ChemPort |
Gerber, H.P., Dixit, V. & Ferrara, N. Vascular endothelial growth factor induces expression of the antiapoptotic proteins Bcl-2 and A1 in vascular endothelial cells. J. Biol. Chem.273, 13313-13316 (1998). | Article | PubMed | ChemPort |
Gupta, K. et al. VEGF prevents apoptosis of human microvascular endothelial cells via opposing effects on MAPK/ERK and SAPK/JNK signaling. Exp. Cell. Res.247, 495-504 (1999). | Article | PubMed | ChemPort |
Benjamin, L.E., Hemo, I. & Keshet, E. A plasticity window for blood vessel remodelling is defined by pericyte coverage of the preformed endothelial network and is regulated by PDGF- B and VEGF. Development125, 1591-1598 (1998). | PubMed | ChemPort |
Wigginton, J.M. et al. IFN- and Fas/FasL are required for the antitumor and antiangiogenic effects of IL-12/pulse IL-2 therapy. J. Clin. Invest.108, 51-62 (2001). | Article | PubMed | ChemPort |
Joussen, A.M. et al. Leukocyte-mediated endothelial cell injury and death in the diabetic retina. Am. J. Pathol.158, 147-152 (2001). | PubMed | ChemPort |
Wilson, R.W. et al. Gene targeting yields a CD18-mutant mouse for study of inflammation. J. Immunol.151, 1571-1578 (1993). | PubMed | ChemPort |
Nagata, S. & Golstein, P. The Fas death factor. Science267, 1449-1456 (1995). | PubMed | ChemPort |
Diez-Roux, G. & Lang, R.A. Macrophages induce apoptosis in normal cells in vivo. Development124, 3633-3638 (1997). | PubMed | ChemPort |
Stone, J. et al. Development of retinal vasculature is mediated by hypoxia-induced vascular endothelial growth factor (VEGF) expression by neuroglia. J. Neurosci.15, 4738-4747 (1995). | PubMed | ChemPort |
Aoki, T. et al. Effect of antioxidants on hyperoxia-induced ICAM-1 expression in human endothelial cells. Adv. Exp. Med. Biol.411, 503-511 (1997). | PubMed | ChemPort |
Nishio, K. et al. Differential contribution of various adhesion molecules to leukocyte kinetics in pulmonary microvessels of hyperoxia-exposed rat lungs. Am. J. Respir. Crit. Care Med.157, 599-609 (1998). | PubMed | ChemPort |
Willam, C., Schindler, R., Frei, U. & Eckardt, K.U. Increases in oxygen tension stimulate expression of ICAM-1 and VCAM-1 on human endothelial cells. Am. J. Physiol.276, H2044-H2052 (1999). | PubMed | ChemPort |
Springer, T.A. Adhesion receptors of the immune system. Nature346, 425-434 (1990). | PubMed | ChemPort |
Sata, M. & Walsh, K. Endothelial cell apoptosis induced by oxidized LDL is associated with the down-regulation of the cellular caspase inhibitor FLIP. J. Biol. Chem.273, 33103-33106 (1998). | Article | PubMed | ChemPort |
Sata, M. & Walsh, K. Oxidized LDL activates fas-mediated endothelial cell apoptosis. J. Clin. Invest.102, 1682-1689 (1998). | PubMed | ChemPort |
Hirahara, H. et al. Long-term survival of cardiac allografts in rats treated before and after surgery with monoclonal antibody to CD2. Transplantation59, 85-90 (1995). | PubMed | ChemPort |
Acknowledgments This work was funded by the Roberta W. Siegel Fund, NIH EY12611 and EY11627, the Juvenile Diabetes Foundation, the Falk Foundation and the Iaccoca Foundation.
15.6 leukocytes versus 119.2 
, normal T cells; 
and 
, oxygenated neonate T cells; 
, oxygenated neonate T cells plus antibody to FasL.
2 (CD18)- integrins
g/ml in PBS; pH 7.4; 5 mg per kg body weight) was then done to label adherent leukocytes and vascular endothelial cells, followed by removal of the residual unbound lectin by PBS perfusion. The retinas were carefully removed and fixed with 1% paraformaldehyde. Images of the flat mounts were obtained using an epifluorescence microscope (DM RXA; Leica).
105 cells per well and were stimulated for 6 h at 37 °C with 80% oxygen or 21% oxygen in a humidified air-tight chamber (Modular Incubator Chamber; Billups-Rothenberg). The isolated T cells were incubated for 15 min at 37 °C with 50 
v
B inhibitor pyrrolidine dithiocarbamate in mice. Invest. Ophthalmol. Vis. Sci. 40, 1624-1629 (1999). | 
and Fas/FasL are required for the antitumor and antiangiogenic effects of IL-12/pulse IL-2 therapy. J. Clin. Invest. 108, 51-62 (2001). | 
