Angiopoietin-2 Causes Pericyte Dropout in the Normal Retina: Evidence for Involvement in Diabetic Retinopathy Free

Experiments performed in this study adhered to the Association for Research in Vision and Opthalmology (ARVO) statement for the “Use of Animals in Ophthalmic and Vision Research.”

Six-week-old male Wistar rats (Charles River, Sulzfeld, Germany) were rendered diabetic by intravenous streptozotocin (STZ) injection of 65 mg/kg body wt (Boehringer Mannheim, Mannheim, Germany). Glucose levels and body weight were monitored consecutively, and glycated hemoglobin was determined by affinity chromatography (Glyc Affin; Isolab, Akron, OH).

Mice with the reporter gene LacZ under control of the Ang-2 promotor (Ang-2 LacZ knock-in) (22) have been recently developed, and were kindly provided by Regeneron Pharmaceuticals (Tarrytown, NJ). Wild-type litters were used as controls. Stable diabetes was induced in 6-week-old animals weighing 24 ± 3 g by STZ (150 mg/kg i.p.). Monitoring of glycemia was performed the same way as for diabetic and nondiabetic rats. Diabetic mice and age-matched controls were maintained for 6 weeks to analyze changes of Ang-2 expression under hyperglycemic conditions and for 26 weeks to analyze changes in pericyte numbers and formation of acellular capillaries (see below).

To determine the time of onset of pericyte loss, eyes were obtained from diabetic and nondiabetic rats (n = 5 for each group) at monthly intervals. One group of five nondiabetic animals was investigated at the beginning of the study. Retinae were obtained after enucleation of the eyes from the animals under deep anesthesia and immediately fixed in 4% buffered formalin. Retinal vascular preparations were performed using a pepsin-trypsin digestion technique as previously described (20,21). Briefly, a combined pepsin (5% pepsin in 0.2% hydrochloric acid for 1.5 h) trypsin (2.5% in 0.2 mol/l Tris for 15–30 min) digestion was used to isolate the retinal vasculature. Subsequently, the samples were stained with periodic acid Schiff base (PAS). The total number of pericytes was counted in 10 randomly selected fields of the retina using an image analyzing system (CUE-2; Olympus Opticals, Hamburg, Germany), and their numbers were normalized to the relative capillary density (numbers of cells per mm² of capillary area). Because significant pericyte loss was found to occur after 2 months of diabetes, eyes were obtained from rats after 3 weeks of diabetes, from age-matched nondiabetic controls, from diabetic rats after 3 months of diabetes, and from age-matched normal controls. Eyes were embedded in OCT (Miles, IL), and sections (6 μm) were cut on a cryostat at −20°C and placed on poly-l-lysine–coated slides.

Retinal digest preparations (n = 4) of 26-week-diabetic Ang-2 LacZ^+/− mice and nondiabetic controls were performed likewise, and pericyte numbers and acellular capillaries were analyzed in PAS-stained samples, as described above.

To identify hyperglycemia-mediated regulation of Ang-1 and Ang-2 in the retina, protein extracts from the four experimental groups were used for Western blot analysis. Briefly, retinal protein extracts were subjected to SDS-polyacrylamide-gel electrophoresis under reducing conditions, and after transfer to nitrocellulose filters, incubated with antibodies against Ang-1 (affinity-purified goat polyclonal antibody against the amino terminus of human Ang-1; 0.2 μg/ml, SC 6319; St. Cruz Biotechnology, Heidelberg, Germany) and Ang-2 (affinity-purified goat antibody against mouse Ang-2; 0.2 μg/ml; SC 7017; St. Cruz Biotechnology). Filters were developed using the enhanced chemiluminescence Western blotting detection system (Amersham, Braunschweig, Germany), and immunocomplexes were quantified using digital copies and quantitative analytical software (Analysis; Olympus Opticals, Hamburg, Germany). Filters were reprobed using a mouse monoclonal anti-actin antibody (0.1 μg/ml; C4, Boehringer Mannheim) and developed as described above. The data were used to normalize angiopoietin levels to protein loading.

Eyes from 6-week diabetic Ang-2 LacZ knock-in mice and age-matched nondiabetic controls (n = 5 retinae/group) were removed under deep anesthesia and stained for LacZ and lectin to visualize Ang-2 expression in relation to the retinal vasculature according to published protocols (23). Briefly, eyes from nondiabetic and diabetic Ang-2 LacZ knock-in mice were fixed in a solution containing 0.2% glutaraldehyde and 1.5% formaldehyde, and X-gal staining was done. Retinae were isolated from eyes and postfixed in 4% buffered formalin, permeabilized with 1% BSA and 0.5% Triton-X-100, followed by staining with TRITC-conjugated isolectin (Sigma L-5264, 1:50) at 4°C overnight. The numbers of Ang-2–expressing cells were assessed by counting LacZ-positive cells per microscopic field in the superficial and the deep layers of the retina. An average of 10 randomly selected microscopic fields was counted per retina at 40× magnification. The numbers of cells were recorded in the inner retinal layers (ganglion cell layer and nerve fiber layer), with microscopic focus on large vessels as a reference, and in the inner nuclear layer with the deep capillary layer in focus. Additionally, eyes stained for LacZ were obtained, and vertical cryostat sections were performed as described (22).

Normal male Wistar rats (age 6 weeks; Charles River, Sulzbach, Germany) were used. Ten microliters containing 0.330 μg (four animals) and 1 μg (four animals) of recombinant Ang-2 (R&D, Wiesbaden, Germany) were injected intravitreally into the left eye under deep anesthesia (Avertin). Sterile PBS was injected into the right eye for control. Eyes were obtained from the deeply anesthesized animals after 1 week, retinae were subjected to digestion preparation as described above, and the capillary network was evaluated for the numbers of pericytes per capillary area.

Diabetes was induced as described above. After 26 weeks of stable hyperglycemia, eyes were obtained from the following groups: nondiabetic wild-type mice (n = 6), diabetic wild-type mice (n = 6), nondiabetic Ang-2 LacZ knock-in mice (n = 4), and diabetic Ang-2 LacZ knock-in mice (n = 4). Eyes were obtained from the deeply anesthesized animals, retinae were subjected to digestion preparation as described above, and the capillary network was evaluated for the numbers of pericytes and acellular capillaries using previously described quantitation methods.

Results are given as means ± SD. For statistical analysis, ANOVA and the Bonferroni multiple comparison test were used (Instat; GraphPad, San Diago, CA).

We determined the onset and magnitude of pericyte loss in retinal capillaries over a period of 3 months of diabetes, using quantitative retinal morphometry. Table 1 shows body weight, blood glucose, and HbA₁ of the experimental groups. STZ-induced diabetic rats had a more than threefold increase of blood glucose and increased levels of glycated hemoglobin (threefold increase at 3 weeks, more than fourfold increase at 3 months). Diabetic animals gained much less weight (+10.3% over basal) than the nondiabetic controls (+23.5%). After 2 months of diabetes duration, the numbers of pericytes became significantly different between diabetic (2,342 ± 104 pericytes/mm² of capillary area) and nondiabetic animals (2,769 ± 99 pericytes/mm² of capillary area; P < 0.001) (Fig. 1), indicating an early onset of pericyte loss in the diabetic rat model. Based on these data, subsequent experiments were performed using retinae obtained at a time point before (i.e., 3 weeks after diabetes induction) and after (i.e., 3 months after diabetes induction) the onset of pericyte loss.

Next, the effect of hyperglycemia on Ang-1 and Ang-2 expression relative to the initiation of pericyte dropout was evaluated by immunoblotting. Ang-1 was present in nondiabetic retinae both at 3 weeks and 3 months. In diabetic retinae, Ang-1 increased 2.5-fold at 3 weeks and remained elevated over nondiabetic controls at 3 months, indicating a persistent upregulation by chronic hyperglycemia. Ang-2 protein expression in nondiabetic animals was very low at 3 weeks and low at 3 months. In contrast, in diabetic rats at 3 weeks, Ang-2 protein was upregulated 37-fold compared with nondiabetic levels and remained 2.5-fold elevated over nondiabetic controls at 3 months of diabetes (Fig. 2). These data demonstrate that Ang-2 is predominantly and strongly upregulated by hyperglycemia and that Ang-2 upregulation precedes the onset of pericyte loss in the diabetic retina.

We used a mouse model in which Ang-2 expression is visualized by a LacZ reporter construct to confirm the effect of hyperglycemia on Ang-2 expression. Previously, these mice had been shown to express Ang-2 in horizontal cells of the inner nuclear layer of the adult retina and in few cells of unclear allocation in the ganglion cell layer (24). The numbers of retinal cells with detectable LacZ expression were compared between mice with 6-week STZ-induced diabetes and age-matched nondiabetic control mice. The metabolic data of the two groups are given in Table 2. In nondiabetic animals (age 12 weeks), a total of 34 cells per microscopic field expressing Ang-2 was identified, of which 8% (2.7 ± 1.2 cells/area) were in the ganglion cell layer (Fig. 3), while the majority of cells were found in the outer capillary plexus, i.e., in the outer margin of the inner nuclear layer (31.25 ± 4.18 cells/area). There was a 2.8-fold increase of LacZ-positive cells in the ganglion cell layer (7.5 ± 1.61 cells/area; P < 0.0001 vs. nondiabetic group) and a 2.0-fold increase in the inner nuclear layer (60.87 ± 18.35 cells/area; P < 0.0001 vs. nondiabetic group) of the 6-week diabetic mice. Beside expression in the known locations of the retina, Ang-2 was also noticed in capillaries traversing from the ganglion cell to the inner nuclear layer of the retina, as depicted in Fig. 3. These data confirmed the data obtained in the STZ-diabetic rat model and are consistent with the hypothesis that chronic hyperglycemia upregulates Ang-2 in the retina.

Because both chronic hyperglycemia and pharmacological Ang-2 application to the eye induce pericyte dropout, we asked whether mice with reduced Ang-2 expression were protected from pericyte dropout and from subsequent consequences such as formation of acellular capillaries. Thus, we studied pericyte dropout and the formation of acellular capillaries in retinae of Ang-2 LacZ knock-in mice that had had diabetes for 26 weeks. Metabolic data of the mice are given in Table 3. Despite a >3.5-fold higher level of glycated hemoglobin and a 4.0-fold increase in glucose levels, pericyte dropout was absent in the diabetic Ang-2 LacZ knock-in mice, while a 25% loss of pericytes was observed in diabetic wild-type litters compared with nondiabetic wild-type controls (diabetic Ang-2 LacZ knock-in mice 1,762 ± 180 pericytes/mm² of capillary area, nondiabetic Ang-2 LacZ knock-in mice 1,835 ± 248 pericytes/mm²; NS) (Fig. 4A). A significant increase in the number of acellular capillaries was observed in diabetic Ang-2 LacZ knock-in mice despite the absence of pericyte loss (Fig. 4B). This increase was significantly less pronounced compared with the increase in formation of acellular capillaries in diabetic wild-type litters.

To investigate whether increased intraocular concentrations of Ang-2 can induce pericyte dropout, recombinant Ang-2 was injected intravitreally into 6-week-old nondiabetic rats. Retinae were isolated 1 week after the injection and subjected to retinal digest preparation and morphometry. Qualitatively, no signs of inflammation were observed in any of the treated animal eyes, suggesting that the pharmacological modulation of the cross-talk between capillary cells was not induced by inflammatory cells. The injection of 330 ng Ang-2 induced a 13% reduction and the injection of 1 μg Ang-2 a 22% reduction in pericyte numbers in the nondiabetic rat retinae after a week (control 2,871 ± 550 cells/mm² of capillary area; 330 ng Ang-2 2,490 ± 380 cells/mm² of capillary area; P < 0.01 vs. control; 1 μg Ang-2 2,260 ± 430 cells/mm² of capillary area; P < 0.001 vs. control; Fig. 5). These quantitative changes in pericyte coverage were not associated with endothelial cell proliferation, as assessed by quantitative morphometry (not shown). Importantly, Ang-2 injection did not affect the diameter or the regularity of retinal capillaries (capillary width of PBS-injected retinae: 7.29 ± 0.81 μm vs. Ang-2 1 μg 6.97 ± 0.69 μm; NS). Figure 5B and C illustrate the regularity of capillaries after Ang-2 injection. We did not observe qualitative changes; particularly, there was neither indication of sprouting and new vessel formation nor that of vascular regression.

Our experimental data suggest that chronic hyperglycemia upregulates Ang-2 in the diabetic retina and that Ang-2 upregulation is causally involved in the pathogenesis of pericyte loss in diabetic retinopathy. Thus, these data add the novel as yet unappreciated aspect that Ang-2 upregulation contributes to the pathogenesis of pericyte loss by a mechanism of active elimination.

Several mechanisms have been proposed to explain pericyte loss in the diabetic retina. One hypothesis refers to the accumulation of toxic intracellular products such as sorbitol or advanced glycation end products (AGEs) (25,26). The enzyme aldose reductase, which converts glucose to sorbitol, was thought to be only expressed in retinal pericytes, but other studies have shown expression in a variety of cell types including endothelial cells (27). Although this leaves the relative importance of the aldose reductase pathway for diabetic retinopathy undecided, its role in pericyte dropout is presumably not primary. AGEs are found to accumulate in pericytes of diabetic animals, but the time course of this accumulation is inconsistent with the time course of pericyte loss, suggesting that AGE accumulation may not be causal (26). AGEs can accumulate in pericytes when injected into nondiabetic animals. However, this may be more attributable to the documented propensity of pericytes for phagocytosis, rather than a specific mechanism of diabetes-induced damage (28). Further data suggesting a selective damage of pericytes in the diabetic retina include differential susceptibility to exposure to modified proteins, such as oxidized LDL (29), and the selective activation of proapoptotic pathways in response to glucose exposure (30). Rather than a passive destruction of pericytes, our data are consistent with the alternative hypothesis for diabetic pericyte loss, i.e., that pericyte loss in the diabetic retina is part of a process that is initiated by the active elimination of pericytes involving Ang-2.

There are three possible sources of Ang-2. In the normal retina Ang-2 is expressed in horizontal cells and in some undefined cell types of the ganglion cell layer (24). The expression of Ang-2 in microvascular endothelial cell has also been described (31). Moreover, in vitro data suggest that Müller cells are increasing Ang-2 expression under hyperglycemic conditions (32). Although we failed to assess the precise localization by immunological means of cells that contribute to the overall upregulation of Ang-2 in the diabetic retina in rats, our data suggest that cells providing Ang-2 become more numerous in diabetes. Moreover, our studies in the diabetic Ang-2 LacZ knock-in mouse indicate that vascular cells can become a source of Ang-2 under hyperglycemic conditions.

The (up)regulation of Ang-2 by hyperglycemia has not yet been reported. Factors described so far to be involved in the regulation of Ang-2 are hypoxia, tissue ischemia and growth factors such as VEGF (33–35). However, the specific mechanisms involved in the early upregulation of Ang-2 in the diabetic retina are unclear. Retinae from diabetic rats in which we found increased expression of Ang-2 protein do not yet develop significant numbers of acellular capillaries during the first 3 months, putting hypoxia as a possible inducer of Ang-2 expression into perspective. Recent data have shown that hyperglycemia-induced mitochondrial overproduction of reactive oxygen species leads to the inhibition of the glycolytic pathway, and the excess flow of substrates through biochemical pathways, which have fructose-6-phosphate and glyceraldehyde-3-phosphate as their starting points. Modifications of gene transcription and of protein production can be the result, possibly including increased transcription of Ang-2 (36–38).

Developmental studies have shown that the interaction of Ang-1 with Tie-2 leads to vessel maturation and stabilization via recruitment of pericytes and smooth muscle cells (11–15). Ang-1 is part of a system that keeps pericytes in place and renders vessels refractory to proliferative or regressive signals. Ang-2 has been invoked (no functional data as proof are available) as a natural antagonist of Ang-1 in cyclic physiologic remodelling via inhibition of Tie-2 phosphorylation (15). The vascular response to Ang-2 is context dependent, as Ang-2 can also induce rather than inhibit autophosphorylation of Tie-2 after prolonged exposure (22). Endogenous Ang-2, while expressed at low level or even absent in the adult retina, is strongly upregulated in experimental proliferative retinopathy, and exogenous administration of Ang-2 in a model of transient vessel formation (in which the endothelial cell remains VEGF dependent for survival), leads to endothelial cell proliferation and sprouting angiogenesis (39–41). Thus, Ang-2 can act as an angiogenic factor in the retina, provided that it encounters VEGF, and a sensitized retinal vasculature. Our study adds a new aspect to the complexity of the Ang-Tie system, i.e., that Ang-2 induces pericyte dropout in the mature retina without concomitant acute endothelial cell changes such as endothelial proliferation or irregular shaping of capillaries. Intraretinal sprouting angiogenesis is absent in our model systems of experimental diabetic retinopathy in rodents. Thus, Ang-2 upregulation in the early diabetic retina does not elicit an angiogenic response, despite loss of pericytes and presence of VEGF upregulation (10,42). Important information on the role of pericytes in general, and specifically in the retina, evolve from developmental studies. It is believed that the recruitment of pericytes and smooth muscle cells and the deposition of new extracellular matrix determines the transition from an immature to a mature capillary network (19). Previous data suggested that recruitment of pericytes to the retinal vessels marks the end of a window of plasticity, during which endothelial cell functions are modulated by the presence or absence of growth factors (43). New data, however, challenge this concept and demonstrate that pericytes are present in the vicinity of the sprouting tip (44). Which molecular signals mark the final maturation of retinal vessels remains unclear. More importantly, with regard to the pathogenesis of diabetic retinopathy, which develops almost exclusively in the mature retina, pericytes are crucially involved in promoting the survival of retinal capillaries. Active disruption of the cellular cross-talk between endothelial cells and pericytes leads to aberrant remodelling, while completion of vascular coverage with pericytes makes the endothelial cells refractory to the modulation by growth factors (8,9,45). It is not known which the relevant factors are, by which endothelial cells in the developing retina differ from those in the adult retina, but when a pericyte deficit persists into adulthood in retinal capillaries, a moderate but significant increase in acellular capillaries becomes apparent (10). Moreover, a dose-dependent effect of vascular responsiveness to pericyte deficiency exists according to a recent study showing responsive retinal neovascularization when the deficit in coverage of developing vessels by pericytes exceeds 50% of the retinal vessels (23).

Pericyte recruitment to capillaries in the developing retina is initially PDGF-receptor β dependent, while capillaries from older mice become resistant to PDGF-B depletion (9). From these data, it emerges that pericytes in the developing retina control sprouting and vessel remodelling, while in the adult/matured retina, they may serve as survival-supporting cells for endothelial cells. In this context, diabetes-induced Ang-2 upregulation may play a significant role in the vascular response to the hyperglycemic challenge. While the concomitant moderate upregulation of Ang-1 suggest an antipermeability and additional endothelial cell survival-promoting response, Ang-2 is considered as an initiator of pericyte elimination possibly facilitating repair of endothelial cells, but with the parallel casualty of the survival-promoting effect of pericytes.

The model of pericyte elimination by hyperglycemia-induced growth factor upregulation is supported by our data from diabetic heterozygous Ang-2 LacZ knock-in mice, in which Ang-2 gene dosage is reduced by 50%. We observed a complete prevention of diabetic pericyte loss in these animals, suggesting both a causal relationship between Ang-2 and diabetic pericyte dropout and a dose dependency of this effect. Despite the preservation of pericytes, the formation of acellular capillaries was significantly affected, indicating that additional hyperglycemia-associated mechanisms further contribute to diabetic endothelial damage. Evidence for the existance of such mechanisms has been provided by showing that endothelial cells, in contrast to smooth muscle cells, express Glut-1, which, in case of hyperglycemia, is not downregulated, leading to an increased flux of substrates through glycolysis and to increased production of mitochondrial reactive oxygen species (36). Mitochondrial overproduction of reactive oxygen species induces poly-ADP ribosylation through ROS-induced activation of poly(ADP-ribose)polymerase, leading to the inhibition of a critical enzyme of the glycolytic pathway. The resulting metabolic bloc results in the upbuilding of two important intermediates, fructose 6-phosphate and glyceraldehyde-3-phosphate, which give rise to the activation of different biochemical pathways involved in the pathogenesis of diabetic complications (46). The relative contribution of these pathways to endothelial damage in the absence of pericyte loss can be concluded from the experiments in diabetic Ang-2 LacZ mice.

In conclusion, our data suggest that Ang-2 modulates pericyte coverage of capillaries in the adult retina and that pericyte loss in early diabetic retinopathy is the result of an elimination process initiated by the upregulation of Ang-2. Despite pericyte preservation, acellular capillaries are not prevented, indicating that a complex interaction of vascular cells exist in which the survival of the endothelial lineage of the capillary has the highest priority.

This work was supported by grants from the Deutsche Forschungsgemeinschaft, the German Diabetes Association, and the Fakultätsschwerpunkt MIGROVAS (Faculty of Clinical Medicine Mannheim, University of Heidelberg). We thank Dr. Stan Wiegand for helpful comments on the manuscript.