Angiopoietin 2 Induces Pericyte Apoptosis via α3β1 Integrin Signaling in Diabetic Retinopathy

Diabetic retinopathy (DR) is the leading cause of visual loss in working-aged people and the most common microvascular complication in diabetic patients despite the recent improvement in the management of DR via glycemic control and photocoagulation (1). Macular edema (leakage) and neovascularization (angiogenesis) both cause severe vision loss in DR (2), and pericyte loss is one of the earliest and most characteristic changes of DR (3).

The pericyte plays two major important clinical roles in DR. First, the pericyte enwraps endothelial cells to keep the integrity of inner blood–retinal barrier (BRB) with the role of microvascular autoregulation (4). Thus, pericyte loss could weaken the inner BRB, even when endothelial cells are intact, and lead to capillary instability and vascular leakage in macular edema. Second, microaneurysm and neovascularization occur in proliferating endothelial cells at the site of pericyte loss (5). Because pericyte loss is an early diabetic change, preventing pericyte loss for the primary prevention of DR would be beneficial.

Although pericyte loss is important in early DR, the mechanism by which hyperglycemia leads to pericyte loss remains largely unknown. However, Ang2 plays a critical role in pericyte loss in DR. Hyperglycemia causes pericyte apoptosis and, ultimately, pericyte loss (6–8). Ang2 increases in the vitreous of patients with proliferative DR (9). In addition, Ang2 is upregulated by hyperglycemia in the diabetic retina and in endothelial cells (10–12). Ang2 induces pericyte loss in normal mice retina and in mice overexpressing Ang2 (10,13). Thus, we postulated that hyperglycemia increases Ang2, which in turn induces pericyte apoptosis in DR.

The role of Ang2 in pericyte loss has been studied (10,11,13–15). Apoptosis and migration are both suggested mechanisms of pericyte loss by Ang2; however, the precise mechanism by which Ang2 induces pericyte loss has not yet been fully elucidated. Ang2 has been known to bind to the endothelial-specific Tie-2 tyrosine receptor with similar affinity to Ang1 (16). Ang2 acts in an autocrine manner in angiogenesis. This endothelial cell–derived antagonistic ligand of vessel maturation and remodeling controls the Ang1–Tie-2 signaling axis (17). Although Ang2 has been postulated to naturally bind to the Tie-2 receptor in the pericyte as the Ang-Tie system (11,13), whether Tie-2 indeed serves as a receptor for Ang2-induced pericyte loss is not clear. Integrin was recently found to mediate platelet-derived growth factor-BB–induced pericyte loss in tumor vessels (18). Also, Ang2 binds to integrin and regulates angiogenesis through integrin signaling (19). Thus, we hypothesized that Ang2 induces pericyte apoptosis via integrin signaling.

In this study, we demonstrated that Ang2 induced pericyte apoptosis via the p53 pathway under high glucose. Interestingly, integrin, not Tie-2 receptor, was important for Ang2-induced pericyte apoptosis under high glucose. High glucose increased integrin α3β1 in pericytes. Furthermore, our results showed Ang2-induced pericyte apoptosis was effectively attenuated by blocking integrin α3β1 in vitro and in vivo. Taken together, Ang2 induced pericyte apoptosis via α3β1 integrin signaling in DR.

Human umbilical vein endothelial cells (HUVECs; Lonza), human retina microvascular endothelial cells (HRMECs; ACBRI), human brain astrocytes (ACBRI), and human pericytes (PromoCell) were maintained in EBM-2, M199 medium, DMEM with 20% FBS, and pericyte media containing growth factors (PromoCell), respectively. All cells were cultured at 37°C in an incubator with a humidified atmosphere of 95% O₂ and 5% CO₂.

Recombinant mouse and human Ang2, human integrin α3β1, and phycoerythrin (PE)-conjugated anti–Tie-2 and mouse IgG antibodies were purchased from R&D Systems. Other reagents and antibodies were anti–Bcl-2 family antibodies (Epitomics); anti–phospho-p53, anti–poly ADP ribose polymerase (PARP), anti-cleaved caspase-3, anti-integrin β1 antibodies (Cell Signaling Technology); anti-p53, anti-Tie1, peroxidase-conjugated secondary antibodies (Santa Cruz Biotechnology); anti-Tie2 antibody and H-Gly-Arg-Gly-Asp-Ser-OH (GRGDS) peptide (Millipore);MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; Sigma-Aldrich); fluorescein isothiocyanate–conjugated annexin V/propidium iodide assay kit (BD Biosciences); anti–neuron-glial antigen 2 (NG2), anti-integrin α3β1, anti-Ang2 antibodies (Abcam); and TUNEL fluorescein kit (Roche). Anti-integrin α1 (clone FB12, MAB1973), anti-integrin α3 (clone P1B5, MAB1952), anti-integrin β1 (clone 6S6, MAB2253), and α-integrin blocking and immunohistochemistry kit (α1-6, v) were purchased from Millipore and used for functional blocking. Small interfering (si)RNAs for p53 were purchased from Bioneer (Daejeon, Korea).

All animal experiments in this study were in strict agreement with the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research and the guidelines of the Seoul National University Animal Care and Use Committee. Eight-week-old, pathogen-free male C57BL/6J mice were purchased from Central Laboratory Animal Inc. After an 8-h fast, diabetes was induced by one intraperitoneal injection of freshly prepared streptozotocin (STZ; Sigma-Aldrich) at a concentration of 180 mg/kg body weight in 10 mmol/L citrate buffer (pH 4.5). Age-matched controls received citrate buffer only. Mice with blood glucose levels >300 mg/dL 4 days after STZ injection were deemed diabetic.

Diabetic and nondiabetic mice were killed 6 months after diabetes induction. Eyes were collected under deep anesthesia and immediately frozen at −80°C for ELISA or were fixed in 4% paraformaldehyde for retinal digestion. Glucose levels and body weight were monitored consecutively, and glycated hemoglobin was determined before mice were killed.

Vascular preparations of whole-mount retinas were performed using a trypsin digestion technique. Briefly, retinas were fixed for at least 24 h in 4% paraformaldehyde, incubated in water for 1 h, and then digested in 2.5% trypsin (Gibco) at 37°C for 1 h. After careful removal of the inner limiting membrane, the retinal vessels were isolated by careful irrigation with filtered water. The retinal digest samples were dried and stained with periodic acid Schiff base for 15 min and hematoxylin.

To determine numbers of retinal pericytes and acellular capillary, retinal digest preparations (n = 7–8) were analyzed. Pericytes were identified according to the morphology and relative location to capillaries. Total pericytes were counted in 10 randomly selected areas (original magnification ×400) in the middle one-third of the retinal capillary area. The number of pericytes and acellular capillaries were standardized to the capillary area (numbers of cells or acellular capillaries/mm² capillary area). The capillary area was calculated using the NIS-Elements AR 3.2 program (Nikon, Tokyo, Japan). Samples were evaluated in a masked fashion.

Normal 6-week-old male mice were used. An intravitreal injection was performed under deep anesthesia with 1 μL containing 100 ng Ang2 and 500 ng anti-integrin α3 (clone P1B5) or anti-integrin β1 (clone 6S6). Sterile PBS was injected for control. After 10 days, retinas underwent digestion preparation.

Retinal digestion preparations were incubated with anti-rabbit NG2 antibody (1:100) and a TUNEL fluorescein kit. Nuclei were counterstained with DAPI. TUNEL and NG2 double-positive cells were evaluated with a fluorescence microscope (Nikon).

In all experiments, 2.5 × 10⁴ cells were seeded into 96-well plates. After 24 h, cells were treated with Ang1 (300 ng/mL), Ang2 (300 ng/mL) under normal glucose (5 mmol/L glucose), high glucose (25 mmol/L glucose), and high mannitol (5 mmol/L glucose and 20 mmol/L mannitol) as an osmotic control for 48 h. Cell viability was determined by MTT assay according to the manufacturer’s instructions. Three independent experiments were performed for each experimental condition.

To evaluate apoptosis, 5 × 10⁵ cells were treated with Ang2 (300 ng/mL) under normal glucose, high mannitol, and high glucose at 37°C for 48 h. To determine the effect of integrin blocking, cells were treated with anti–integrin blocking antibodies (5 μg/mL) 1 h before the addition of Ang2. The cells were harvested and washed twice in PBS. Cells were stained with fluorescein isothiocyanate annexin-V and propidium iodide (PI) for 15 min and analyzed by flow cytometry. Annexin V–positive/PI-negative cells were determined to be apoptotic.

To evaluate Tie-2 expression, pericytes and HUVECs (1 × 10⁶) were suspended in the complete media (25 μL) for each experiment. PE-conjugated anti-Tie2 antibody was added to the each sample at 4°C for 1 h. PE-conjugated mouse IgG antibody was used as a control.

All RNA was collected and isolated from cells using the RNeasy Plus Mini kit (Qiagen). cDNAs were prepared from RNAs (1 μg) using 2.5 μmol/L oligo-dT primers, 1 mmol/L deoxyribonucleotide triphosphates, and murine leukemia virus RT. Quantitative (q)PCR assays were performed in qPCR Master Mix for SYBR Green PCR Master Mix (Applied Biosystems) using 7900HT real-time PCR (Applied Biosystems). Reaction conditions were 50 cycles of 95°C for 5 s and 60°C for 20 s for qPCR. Quantitative real-time PCR was performed using the following primers: Ang2 (forward: 5′-ACTGTGTCCTCTTCCACCAC-3′and reverse: 5′-GGATGTTTAGGGTCTTGCTTT-3′); Tie-2 (forward: 5′-GCTTGCTCCTTTCTGGAACTGT-3′ and reverse: 5′-CGCCACCCAGAGGCAAT-3′) (20); Tie-1 (forward: 5′-AGAACCTAGCCTCCAAGATT-3′ and reverse: 5′-ACTGTAGTTCAGGGACTCAA-3′); ITGA1 (forward: 5′-GGTTCCTACTTTGGCAGTATT-3′ and reverse: 5′- AACCTTGTCTGATTGAGAGCA-3′); ITGA3 (forward: 5′-AAGGGACCTTCAGGTGCA-3′ and reverse: 5′-TGTAGCCGGTGATTTACCAT-3′); ITGB1 (forward: 5′-GAAGGGTTGCCCTCCAGA-3′ and reverse: 5′-GCTTGAGCTTCTCTGCTGTT-3′); and β-actin (forward: 5′-GCCGCCAGCTCACCAT-3′ and reverse: 5′-TCGATGGGGTACTTCAGGGT-3′). A mean quantity was calculated from triplicate qPCR for each sample and normalized to the control gene.

Different primers were used for Tie-2 RT-PCR (forward: 5′-TGTTCCTGTGCCACAGGCTG-3′ and reverse: 5′-CACTGTCCCATCCGGCTTCA-3′). PCR products were separated on 1% agarose gels and visualized using SYBR Safe DNA Gel Stain (Invitrogen) under ultraviolet transillumination.

For immunoprecipitation, Ang2 (500 ng), integrin α3β1 (500 ng), and anti-integrin α3β1 antibodies (2 μg) were incubated with G-Sepharose beads at 4°C overnight (19). Immune complexes were collected by centrifugation and washed with buffer three times. For immunoblotting, cells were harvested and lysed in radioimmunoprecipitation assay buffer with a protease inhibitor cocktail. Protein lysates were resolved by SDS-PAGE and transferred onto nitrocellulose membrane. The membranes were incubated with primary antibodies (1:1,000) at 4°C overnight and secondary antibodies (1:5,000) at room temperature for 1 h. The membranes were incubated with enhanced chemiluminescent substrate (Pierce) and exposed to film.

Statistical analyses were performed using the standard two-tailed Student t test assuming unequal variances, and P < 0.01 was considered statistically significant. Quantitative data are given as mean ± SD. Data in figures are depicted as mean ± SE.

The numbers of pericytes and acellular capillaries in retinal digestion were compared between mice with 6-month STZ-induced diabetes and age-matched nondiabetic control mice (Fig. 1A). On one hand, the number of retinal capillary pericytes was significantly decreased in the STZ-induced diabetic mice retinas (932.6 ± 70.3) compared with that of the control group (1,341.8 ± 55.9, P < 0.001; Fig. 1B). On the other hand, the number of retinal acellular capillaries was significantly increased in the STZ-induced diabetic mice retinas (93.0 ± 22.5) compared with that of the control group (28.1 ± 11.5, P < 0.001; Fig. 1C). Table 1 reports the metabolic and physical parameters of the experimental groups.

The effect of hyperglycemia on Ang2 expression in STZ-induced diabetic retinas was evaluated by qRT-PCR. In diabetic retinas at 6 months, Angpt2 mRNA increased 1.87-fold compared with nondiabetic normal retinas (P = 0.004; Fig. 2A). Also, Angpt1 mRNA increased 2.54-fold (P = 0.004) and Vegfa mRNA increased 1.6-fold (P = 0.027; Supplementary Fig. 1A and B).

Next, to determine the source of Ang2 in the diabetic retinas, we examined the effects of high glucose on in vitro ANGPT2 mRNA transcription by qRT-PCR in three major components of the BRB: HRMECs, pericytes, and astrocytes. On one hand, high glucose increased ANGPT2 mRNA level in HRMECs more than 1.5-fold (1.52 ± 0.09, P = 0.003) compared with normal glucose (Fig. 2B). On the other hand, high mannitol, an osmotic control, did not increase ANGPT2 mRNA (1.21 ± 0.13, P = 0.094) in HRMECs (Fig. 2B). Also, high glucose did not increase ANGPT2 mRNA in pericytes (1.13 ± 0.22, P = 0.418) and astrocytes (1.13 ± 0.22, P = 0.418; Fig. 2B and C). These data demonstrate that Ang2 increases in the diabetic retina and that the source of Ang2 increase is retinal microvascular endothelial cells.

We determined the effect of Ang2 on the cell viability and apoptosis of pericytes under high glucose. Ang2 alone did not affect cell viability in pericytes (97.6 ± 3.6%, P = 0.221). High glucose reduced cell viability in pericytes (89.4 ± 7.9%, P = 0.020; Fig. 3A). Interestingly, Ang2 aggravated cell death under high glucose in pericytes (72.4 ± 2.9%, P = 0.002; Fig. 3A). Next, pericyte apoptosis was assessed by annexin-V/PI flow cytometric analysis. Of importance, the number of apoptotic pericytes was increased under high glucose (7.7 ± 0.3%, P < 0.001) compared that of the control group (2.4 ± 0.4%). Furthermore, Ang2 significantly aggravated high glucose–induced pericyte apoptosis (25.9 ± 0.4%, P < 0.001), whereas Ang2 alone did not induce apoptosis (2.7 ± 0.1%, P = 0.298) under normal glucose (Fig. 3B and Supplementary Fig. 2A).

Next, we determined the effect of Ang2 on the cell viability and apoptosis of HRMECs under high-glucose conditions. Ang1 and Ang2 increased cell viability in HRMECs under high glucose (114.4 ± 8.8% [P = 0.014] and 109.5 ± 2.3% [P = 0.040], respectively; Fig. 3C). As expected, Ang2 significantly decreased the apoptotic cell population in HRMECs (Fig. 3D, and Supplementary Fig. 2B). These data suggested that Ang2 plays a synergistic role in pericyte apoptosis under a high-glucose condition and has a protective effect on the endothelial cell.

Western blot studies confirmed that Ang2 induced the apoptosis pathway with an increase of Bax, cleaved PARP, and cleaved caspase-3 under high glucose but not under normal glucose (Fig. 4A). Next, we aimed to identify the mechanism that mediates Ang2-induced pericyte apoptosis under high glucose. We found that Ang2 induced p53 phosphorylation (Fig. 4B) and, subsequently, p53 accumulation (Fig. 4C) under high glucose. Then, to determine the role of the p53 pathway for the observed Ang2-mediated pericyte apoptosis, we treated pericytes with control siRNA or two different p53 siRNAs. The p53 siRNAs effectively downregulated p53 expression (Fig. 4D) and also attenuated Ang2-induced pericyte apoptosis under high glucose (Fig. 4E). Interestingly, Ang2 phosphorylated extracellular signal–related kinase (ERK), but not Akt (Fig. 4F). This ERK phosphorylation by Ang2 was inhibited by PD98059 (ERK inhibitor), and the ERK inhibitor attenuated Ang2-induced p53 phosphorylation (Fig. 4G). These data suggest that Ang2 induces pericyte apoptosis via the p53 pathway under high-glucose conditions.

To determine whether the Tie-2 receptor is related with Ang2-induced pericyte apoptosis, Western blot analysis (Fig. 5A) and RT-PCR (Fig. 5B) for Tie-2 and Tie-1 were performed on lysates obtained from HUVECs and pericytes. Interestingly, pericytes did not express Tie-2 or Tie-1, whereas HUVECs expressed Tie-2 and Tie-1. These data were confirmed by qRT-PCR. TIE2 mRNA and TIE1 mRNA levels were significantly lower in pericytes than in HUVECs (Fig. 5C). Compared with HUVECs with Tie-2 expression, pericytes did not express Tie-2 in FACS analysis (Fig. 5D).

To test whether integrins may serve as receptors for Ang2 in pericyte apoptosis under high glucose, we incubated pericytes under high glucose for 15 min with Ang2 and Gly-Arg-Gly-Asp-Ser (GRGDS) peptides, which can inhibit integrins that bind the Arg-Gly-Asp (RGD) sequence (21). GRGDS (0.5 mg/mL) attenuated Ang2-induced p53 phosphorylation (Fig. 5E). This result showed that integrin signaling is involved in Ang2-induced p53 phosphorylation. Of importance, these data suggested that integrin, not Tie-2, is important for Ang2-induced pericyte apoptosis under high glucose as an Ang2 receptor.

As shown in Fig. 3C and Fig. 5E, Ang2 induced pericyte apoptosis via the integrin receptor under high glucose but not under normal glucose. We hypothesized that high glucose preconditioned pericytes susceptible to Ang2 by changing the integrin pattern. We screened the integrin α subunits (α1–6 and αv) to determine which integrin subunit is responsible for Ang2-induced pericyte apoptosis. Integrin α1 and α3–blocking antibodies attenuated Ang2-induced p53 phosphorylation (Fig. 6A).

Next, to determine whether high glucose changes the integrin expression pattern, we performed qRT-PCR and Western blot studies for α1, α3, and β1. The choice was made because integrin α1 and α3 can form the heterodimer with only integrin β1 (22). High glucose increased mRNA levels at 24 and 48 h for ITGα1 (1.86-fold and 2.20-fold, P < 0.01; Fig. 6B), ITGα3 (1.50-fold and 1.83-fold, P < 0.01; Fig. 6C), and ITGβ1 (1.50-fold and 1.43-fold, P < 0.01; Fig. 6D), respectively. In addition, high glucose increased integrin α1, α3, and β1 expression (Fig. 6E). However, the integrin α1 was rarely expressed compared with integrin α3. In this regard, we performed a coimmunoprecipitation assay to show direct binding of Ang2 to integrin α3β1. Indeed, Ang2 directly bound to integrin α3β1 (Fig. 6F).

From the result of integrin expression in pericytes under high glucose, integrin α1β1 or α3β1 were supposed to be the possible receptor of Ang2. To determine whether Ang2 induced pericyte apoptosis through integrin α1β1 or α3β1, pericytes were incubated under 25 mmol/L high glucose with Ang2 and anti-integrin α1, α3, and β1 antibodies for 48 h. Interestingly, Ang2-induced pericyte apoptosis under high glucose was attenuated by anti–integrin α3 and β1 antibodies but not by the anti–integrin α1 antibody (Fig. 7A, and Supplementary Fig. 3A). These results were confirmed for p53 expression on Western blot analysis (Fig. 7B). Ang2 significantly increased p53 expression under high glucose (2.7 ± 0.2, P < 0.01) but was significantly attenuated by anti–integrin α1, α3, and β1 antibodies (1.5 ± 0.2, 1.0 ± 0.1, and 0.7 ± 0.1, respectively; P < 0.01).

On the basis of in vitro experiments, we next intravitreously injected normal mice with 100 ng Ang2 with and without 500 ng anti-α3 or 500 ng anti-β1 antibodies. Ten days after the intravitreal injection, the isolated retinas were digested with trypsin for pericyte evaluation (Fig. 7C). The in vivo intravitreal injection of Ang2 induced pericyte loss in the retina compared with that in PBS injected control mice (1,047 ± 52 and 1,401 ± 109 cells/mm², P < 0.001). Ang2-induced pericyte loss was significantly attenuated by anti-α3 or anti-β1 antibodies (1,354 ± 148 and 1,380 ± 66 cells/mm², P < 0.001; Fig. 7D). In addition, Ang2-induced TUNEL and NG2 double-positive pericytes were decreased in mice injected with anti-α3 or anti-β1 antibodies (Fig. 7E). These in vivo data suggest that Ang2 induced pericyte loss via an apoptotic mechanism in the retina. Overall, these data demonstrate that integrin α3 and β1 are important for Ang2-induced pericyte apoptosis via the p53 pathway.

In this study, we demonstrated that pericyte loss occurred with Ang2 increase in diabetic mice retinas (10,11,13). Previously, Tie-2 was known to be related with Ang-induced pericyte survival and recruitment (11); however, there was no direct evidence that Ang2 induced pericyte apoptosis via Tie-2. Unlike the previous study (11), pericytes did not express Tie-2 mRNA or protein in our study. Although Ang2 alone did not induce apoptosis in vitro (11), Ang2 induced pericyte apoptosis under high glucose via the p53 pathway (Fig. 4). Thus, we postulated that Ang2 would induce Tie-2–independent apoptosis in pericytes. Accumulating evidence supports the role of integrin in Ang2 activity in endothelial cells (19); however, little is known about the Ang2/integrin system in the pericyte. We thus focused on the integrin receptors on pericytes.

Integrins are cell adhesion molecules that are expressed on the surface of endothelial cells and pericytes. The contribution of endothelial and mural cell integrins to angiogenesis has been studied. Pericytes express integrins, including the collagen receptors α1β1 and α2β1; the laminin receptors α3β1, α6β1, α6β4, and α7β1; the fibronectin receptors α4β1 and α5β1; and the osteoponin receptors α8β1 and α9β1 (22). Cytokines and extracellular matrix change integrin subtypes in pericytes (23). On the basis of the differential response to Ang2 in the pericytes depending on the glucose concentration (Fig. 3B and 4A), we hypothesized that high glucose could change the integrin subtype more susceptible to Ang2. During qRT-PCR array for screening the integrin subtype change response to high glucose, only ITGα1, ITGα2, ITGα3, and ITGβ1 mRNA increased more than 1.5-fold among ITGα1, ITGα2, ITGα3, ITGα4, ITGα5, ITGα6, ITGα7, ITGαv, and ITGβ1 mRNA (data not shown). In addition, basal ITGα3 mRNA was more than six times higher than ITGα1 or ITGα2 mRNA in pericytes even under normal glucose conditions. Thus, we concluded that integrin α3 and β1 are both predominant and high-glucose inducible integrins in pericytes. The proportion of integrin α1 in pericytes is too minor to be responsible for Ang2-induced pericyte apoptosis (Fig. 6E and 7B). In addition, many integrins, including αvβ3, α5β1, αIIbβ3, αvβ6, and α3β1, recognize the tripeptide RGD in their ligands (24). However, integrin α1β1 recognizes a configuration of residues formed by arginine and aspartic acid residues, not RGD. Attenuation of p53 by GRGDS supported that Ang2-induced pericyte apoptosis was effectively attenuated by blocking integrin α3β1 (Fig. 5E and 7B). Integrin receptors induced by hyperglycemia in DR could make the pericyte more susceptible to the Ang2, and in turn, lead to the pericyte apoptosis by the Ang2/integrin signaling pathway.

The laminin receptor α3β1 integrin is expressed in vascular endothelial cells and acts to suppress pathological angiogenesis in the retina and other organs (25,26). However, a role for integrin α3β1 in the pericyte is yet to be determined (22). Inhibition of β1 integrin induced a rounded morphology of the pericytes, suggesting pericyte adhesive properties were affected or that these cells were undergoing apoptosis (27). In this study, we suggested that integrin α3β1 was important for Ang2-induced apoptosis. This is notably different from the classical concept that Tie-2 mediates the effect of Ang2 as an Ang2 receptor. Absence of Tie-2 in the human pericytes used in this study was confirmed by various experiments, including RT-PCR with two different primers, Western blot, and FACS analysis (Fig. 5A–D). In addition, Ang1 and Ang2 can both directly bind to some integrins and thereby signal in the absence of Tie-2 (28–31). Thus, we suggested that integrin mediated Ang2-induced apoptosis at least in Tie-2–negative pericytes.

Previous studies have shown that vascular endothelial cells are the primary source of Ang2 (32–35). High glucose increases Ang2 mRNA in HRMECs (36). The endothelial cell is regarded as potential source of Ang2 in the retina by transient rapid release from the Weibel-Palade body and chronic upregulation of Ang2 (32,37,38). Chronic hyperglycemia upregulates Ang2 in the diabetic retina (36,39), and Ang2 upregulation is causally involved in the pathogenesis of pericyte loss in DR (10). Hyperglycemia induces pericyte loss not only by increasing Ang2 in endothelial cells but also by changing the integrin subtype prone to Ang2 in pericytes. Our data support that glycemic control is important to prevent Ang2-induced pericyte loss in early DR. Glycemic control is the effective treatment to reduce the progression of DR (40–42).

We showed that Ang2 induced pericyte apoptosis under high glucose via the p53 pathway. Furthermore, Ang2-induced pericyte apoptosis was effectively attenuated in vitro and in vivo by blocking integrin α3β1. Integrin α3 and β1 blockers reduced p53 signaling by Ang2. Ser-15 of p53 is phosphorylated by a mitogen-activated protein kinase (ERK1/2)-dependent pathway, and this step is required for apoptosis to occur (43).

Others have suggested that hyperglycemia causes pericyte loss by apoptosis (6–8). In this study, we postulated that hyperglycemia-induced Ang2 overexpression in endothelial cells, in turn, caused apoptotic cell death of pericytes. In an animal model of DR, high glucose was induced by STZ injection as early as 3 days. Then, Ang2 had increased at 3 months with pericyte loss, and finally, acellular capillary at 6 months (10,44). The mechanism by which capillaries become acellular is largely unknown. Apoptosis preceding the formation of acellular capillaries in retinas from diabetic rats and humans is one mechanism by which endothelial cells could be eliminated from the diabetic capillary (8). Retinal capillary coverage with pericytes is crucial for the survival of endothelial cells. Loss of pericytes is related to the increase of acellular capillaries from hyperglycemic injury (5). Our data indicate Ang2 did not directly induce endothelial cell death (Fig. 3C and D); thus, we postulated that an acellular capillary could cause endothelial cell death secondary to pericyte loss by Ang2.

In conclusion, we demonstrated that pericyte loss occurred with Ang2 increase in the diabetic mice retinas. Ang2 induced pericyte apoptosis via the p53 pathway under high glucose. High glucose increased integrin α3β1 in pericytes. Interestingly, integrin was involved in Ang2-induced pericyte apoptosis. Furthermore, Ang2-induced pericyte apoptosis was effectively attenuated by blocking integrin α3β1 both in vitro and in vivo. Taken together, Ang2 induced p53-dependent pericyte apoptosis via α3β1 integrin signaling in DR. We suggest that glycemic control or blocking Ang2/integrin signaling could be a potential therapeutic target to prevent pericyte loss in early DR.

Funding. This study was supported by the Seoul National University Research Grant (800-20130338 to Je.H.K.), the Seoul National University Hospital Research Fund (03-2013-0070 to Je.H.K.), the Pioneer Research Center Program of the National Research Foundation for the Ministry of Education, Science and Technology (NRF/MEST; 2012-0009544 to Je.H.K.), the Bio-Signal Analysis Technology Innovation Program of NRF/MEST (2009-0090895 to Je.H.K.), the Global Research Laboratory Program of NRF/MEST (2011-0021874 to K.-W.K.) and the NRF of Korea grant funded by the Korea government (2012R1A2A2A01012400, 2011-0030739 to C.-H.C. and 2012-0006019 to Je.H.K.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.W.P. and J.-H.Y. performed the experiments and wrote the manuscript. Ji.H.K. and K.-W.K. analyzed the data and reviewed the manuscript. C.-H.C. and Je.H.K. designed the study and critically revised the manuscript. Je.H.K. is the guarantor of this work, and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.