Introduction

Diabetic retinopathy is the major cause of blindness in diabetic patients and induces progressive damage to the retinal microvasculature [1]. Blood–retinal barrier (BRB) breakdown and consequent vessel leakage in the retina are the main events in the pathogenesis of diabetic retinopathy [2, 3]. Pericytes sheathing retinal capillary endothelial cells play a role in regulating endothelial cell proliferation and survival [4, 5]. Pericyte loss is considered a hallmark of diabetic retinopathy, and is associated with the development of vessel leakage and leucocyte adhesion, which signify BRB breakdown [6–8]. Pericyte loss is thought to occur via activation of key apoptotic proteins in the diabetic retina [9]. However, the underlying molecular mechanisms are currently unclear.

The α-crystallins are a family of small heat-shock proteins that perform many physiological functions, including the maintenance of cell survival [10, 11]. α-Crystallins not only possess chaperone-like activity in vitro, but also remodel and protect the cytoskeleton, inhibit apoptosis and enhance the resistance of cells to stress [12]. α-Crystallins can be either of two gene products: αA- and αB-crystallin. αA-Crystallin is predominantly produced in the lens and retina, while αB-crystallin is widely distributed in non-lenticular tissues [13]. A previous study reported that αA-crystallin is more protective against apoptosis than αB-crystallin in cultured lens epithelial cells [14]. Moreover, overproduction of αA- or αB-crystallin enhances resistance to thermal, photochemical and other stress conditions [10, 14]. Overproduction of αB-crystallin significantly reduces the apoptosis triggered by dicarbonyl-modified fibronectin in retinal capillary pericytes [15]. In view of the protective effects of α-crystallins in preserving the integrity of mitochondria, restricting the release of cytochrome c, blocking the degradation of poly-ADP-ribose-polymerase (PARP) and repressing the activation of caspase-3 [16], these proteins could be novel targets in gene-based therapy for early diabetic retinopathy.

The chaperone activity of αA-crystallin in the retina decreases significantly with ageing. The decreased production of αA-crystallin and increased truncation at the C- and N-terminals with ageing, together with its oxidation, suggest a general decrease in chaperone activity in the retina [17, 18]. Consistent with this, streptozotocin (STZ)-induced diabetic rats with uncontrolled hyperglycaemia show substantial loss of α-crystallin chaperone function [19, 20]. Diabetes strongly reduces the chaperone function of α-crystallins by reducing their solubility and disrupting the normal interaction of α-crystallins with BAX [21]. Studies using αA- and/or αB-crystallin-knockout mice have revealed increased retinal cell death in endophthalmitis and uveitis models, suggesting that α-crystallin prevents retinal cell death during inflammation [22, 23]. It is also reported that the absence of αB-crystallin leads to dramatic attenuation of angiogenesis via modulation of vascular endothelial growth factor (VEGF) in models of intraocular disease [24]. These functions have been implicated in disease process, and therefore α-crystallins may be effective targets for disease therapy. However, their physiological significance in the context of the pathological vasculature in diabetes remains unknown.

To investigate changes in gene expression in the retina during early-stage diabetic retinopathy, we performed differential proteomic analysis using two-dimensional electrophoresis (2-DE), in combination with matrix-assisted laser desorption ionisation time-of-flight (MALDI-TOF) mass spectrometry. The presence of αA-crystallin in pericytes was analysed by immunoblotting and immunostaining methods. Decreased amounts of αA-crystallin in human pericytes under hyperglycaemic conditions in vitro were also confirmed by immunoblotting and immunostaining. Finally, a recombinant adenovirus (Ad-αAc-Gfp) expressing the gene encoding αA-crystallin (αAc, also known as Cryaa) was injected into the vitreous body to examine the effect of αA-crystallin on apoptotic cell death and vessel leakage in the diabetic retina.

Methods

Animals

Eight-week-old male C57BL/6 mice were obtained from KOATEC (Pyeongtaek, Korea). Mice were maintained on a standard rodent diet available, together with water, ad libitum. All experiments were approved by the Animal Care and Use Committee of Gyeongsang National University, Jinju, Korea. Diabetes was induced by intraperitoneal injection of 55 mg/kg STZ (Sigma-Aldrich, St Louis, MO, USA) in 0.05 mol/l citrate buffer (pH 4.5) on five consecutive days. Age-matched control mice received citrate buffer alone. Diabetes was defined as blood glucose levels >22.2 mmol/l 1 week after STZ injection.

Cell culture

Human retinal pericytes (ACBRI 183; Cell Systems, Kirkland, WA, USA) were cultured in complete classic medium supplemented with 10% (vol./vol.) fetal bovine serum, antibiotics and human recombinant growth factors (CultureBoost-R; Cell Systems) on attachment-factor-coated dishes. Cells were maintained in normal glucose (5 mmol/l) or high glucose (30 mmol/l) for 5 days. Mannitol was used as a control to rule out the effect of osmotic pressure.

Visualisation of retinal vasculature

Retinal vasculature was assessed by angiography using high-molecular-weight-FITC-conjugated dextran (FITC-dextran, MW 2 × 106 [Sigma-Aldrich]) and tetramethylrhodamine (TMR)-conjugated dextran (TMR-dextran, MW 2 × 106 [Invitrogen, Carlsbad, CA, USA]) as previously described [25, 26]. FITC-dextran and TMR-dextran were dissolved in 0.05 mol/l sodium citrate (pH 4.5) at final concentrations of 50 mg/ml and 10 mg/ml, respectively, and infused through the left ventricle of control and diabetic mice. After FITC-dextran or TMR-dextran injection, eyes were dissected and placed in 4% (wt./vol.) paraformaldehyde (PFA) for 6 h. Flat-mounted retinas were visualised using fluorescence microscopy.

Measurement of BRB breakdown

BRB breakdown was evaluated as described previously [27]. Both control and diabetic mice received injections into the left jugular vein of 45 mg/kg Evans Blue dye. At 2 h after infusion, blood was extracted through the left ventricle, and mice were perfused with PBS to completely remove the Evans Blue dye in blood vessels. Retinas were carefully dissected and the weight measured after thorough drying in a Speed-Vac (TL-100; Beckman Coulter, Fullerton, CA, USA). Next, retinas were incubated in 120 μl formamide for 18 h at 70°C to extract the Evans Blue dye, and the extract filtered through an ultra-filter (30,000 MW). Absorbance of the extract was measured with a spectrophotometer at 620 nm. The dye concentration in extracts was calculated using a standard curve of Evans Blue in formamide and normalised to the dried retinal weight.

2-DE and protein identification

The total protein from each retina (50 μg/strip) was resuspended in rehydration buffer (2 mol/l thiourea, 6 mol/l urea, 4% (wt/vol.) 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonic acid (CHAPS), 65 mmol/l DTT, 0.5% (vol./vol.) ampholytes, 0.002% (wt/vol.) bromophenol blue), loaded onto 7 cm immobilised pH gradient (IPG) strips (Immobiline DryStrip, pH 4–11 non-linear [Genomine, Pohang, Korea]). Isoelectric focusing was performed at 250 V for the first 15 min, followed by a gradient increase to 10,000 V for a total of 100 kVh. After isoelectric focusing and equilibration, IPG strips were subjected to 13.5% (wt/vol.) SDS-PAGE. Protein spots were visualised using silver staining according to the procedure of Kim et al [28]. Proteins were identified with peptide mass fingerprinting (PMF) analysis by Genomine using MALDI-TOF mass spectrometry (Ettan MALDI Pro; Amersham Bioscience, Piscataway, NJ, USA).

Antibodies

Mouse monoclonal antibodies against αA-crystallin, α-smooth muscle actin (α-SMA), NG2 chondroitin sulphate proteoglycan (NG2) and α-tubulin were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA), Chemicon (Tamecula, CA, USA) and Sigma-Aldrich, respectively. Rabbit polyclonal antibodies against αA-crystallin, active caspase-3 and green fluorescent protein (GFP) were obtained from Abcam (Cambridge, MA, USA) and the rabbit polyclonal PARP antibody from Cell Signaling Technology (Beverly, MA, USA). Secondary horseradish-peroxidase-conjugated anti-mouse and anti-rabbit IgG for western blotting, Alexa Fluor 350, 488 and 594 goat anti-mouse IgG, and Alexa Fluor 488 goat anti-rabbit IgG for immunofluorescent staining were purchased from Pierce Biotechnology (Rockford, IL, USA) and Invitrogen.

Immunoblotting

Protein extraction and immunoblotting were performed as described previously [28]. Total protein (30 μg) was subjected to SDS-PAGE and blotted onto a nitrocellulose membrane. Blots were incubated with primary and secondary antibodies, and visualised using an enhanced chemiluminescence kit (Amersham Biosciences, Pittsburgh, PA, USA). Data were quantitatively analysed with the Soft Imaging System and SigmaGel 1.0 software (Jandel Scientific, San Rafael, CA, USA).

Immunofluorescence staining

Retinal cryosections and chamber slides were incubated with blocking solution for 30 min, followed by a mixture of primary antibodies overnight with gentle rocking. After washing in PBS, sections were incubated with secondary antibodies and wet mounted using Kaiser’s solution (Merck, Darmstadt, Germany). Cell apoptosis was assessed using the TUNEL assay (In situ Cell Death Detection kit [Roche, Mannheim, Germany]), as described previously [29, 30]. To determine the number of pericytes surrounding retinal vessels, angiography with infusion of TMR-dextran (Invitrogen) and immunofluorescence staining with anti-α-SMA antibody to dye vessels and pericytes were performed. All images were obtained with an IX2-DSU disk-scanning biological microscope (Olympus, Hamburg, Germany) and a confocal microscope (Axioplan 2 Imaging; Carl Zeiss, Göttingen, Germany).

RT-PCR

To evaluate αA-crystallin mRNA expression in mouse retina, RT-PCR was performed. Total RNA was extracted from whole retina with TRI-reagent and RNeasy mini kit (Qiagen, Hilden, Germany), as described by the manufacturer. RNA (1 μg) was used to generate cDNA using SuperScript III reverse transcriptase (Invitrogen). Gapdh expression was used for normalisation. The PCR primers used in this study are listed in electronic supplementary material (ESM) Table 1.

Adenovirus injection

Ad-αAc-Gfp was constructed using the ViraPower Adenoviral Expression system (Invitrogen), and purified with an Adeno-X Virus purification kit (Clontech Laboratories, San Jose, CA, USA), according to the manufacturer’s instructions, with minor modifications. Ad-αAc-Gfp was deposited in the Korea Collection for Type Cultures with accession number KCTC 11844BP, as provided by the international depositary authority. Recombinant adenovirus expressing Gfp and αA-crystallin was injected with a 30-gauge needle into the left and right vitreous of mice, respectively, under an operating microscope. At 2 weeks after the adenovirus injection, diabetes was induced with STZ injection. The efficiency of the adenovirus infection was tested via GFP fluorescence imaging in retinal vessels.

Transmission electron microscopy

Whole retinas were fixed in 2.5% (wt/vol.) glutaraldehyde solution, postfixed in 1% (wt/vol.) osmium tetroxide and dehydrated in ethanol. After embedding in Epon, thin sections were stained with 5% (wt/vol.) uranyl acetate followed by 0.4% (wt/vol.) lead citrate, and were observed with transmission electron microscopy (TEM).

Determination of adherent leucocytes

Quantification of retinal leucostasis was performed as described previously [7]. After anaesthetising mice, the chest cavity was opened and the descending aorta was carefully clamped. To remove erythrocytes and non-adherent leucocytes, PBS was perfused into the left ventricle. Fixation was achieved by perfusion with 1% (wt/vol.) PFA under physiological pressure. Non-specific binding was blocked with 1% (wt/vol.) BSA in PBS, and this was followed by perfusion with TMR-isothiocyanate (TRITC)-coupled concanavalin A lectin (Vector Laboratories, Burlingame, CA, USA) to label adherent leucocytes and vascular endothelial cells. At 5 min after infusion, PBS and 1% (wt/vol.) BSA in PBS were sequentially perfused to remove residual unbound lectin. Whole retinas were carefully dissected and flat-mounted with ProLong Gold anti-fade reagent (Invitrogen). Retinas were then observed by fluorescence microscopy, and the total number of adherent leucocytes per retina determined.

Statistical analysis

Data are representative of four independent values and presented as means ± SEM. Data were considered significant at p < 0.05. Statistical analyses were performed using the Kruskal–Wallis H test and Mann–Whitney U test (SPSS software).

Results

Diabetes induces vessel leakage and pericyte apoptosis in retinas

Diabetic retinopathy leads to increasingly progressive alterations in the retinal vasculature. Accordingly, we evaluated the relative change in vascular permeability and pericyte apoptosis in mouse retina after 2 and 6 months of diabetes. Fluorescent angiography disclosed diabetes-induced vessel leakage in the peripheral retinas, but not in control retinas (Fig. 1a). Consistent with this finding, we observed a 1.63- and 5.86-fold increase in vessel leakage in diabetic retinas at 2 and 6 months, respectively, compared with controls, using an Evans Blue leakage assay (p = 0.047 and p = 0.022, respectively, Fig. 1b). Moreover, immunoblot analyses showed an increased amount of pro-apoptotic proteins such as cleaved PARP, active caspase-3 and BCL2-associated X protein (BAX; p = 0.001, p = 0.045 and p = 0.036, respectively), and decreased levels of the anti-apoptotic protein B cell leukaemia/lymphoma 2 (BCL2) (p = 0.022) in whole-retina extracts of diabetic mice (Fig. 1c). Immunofluorescence staining for TUNEL and the pericyte marker NG2 disclosed an increase in TUNEL-positive pericytes in diabetic retinas, compared with controls (Fig. 1d).

Fig. 1
figure 1

Diabetes induces vascular leakage and pericyte loss in retinas. (a) Flat mount of retinas infused with FITC-dextran in mice 2 and 6 months after induction of diabetes and controls. Arrows show microaneurysms, arrowheads represent neovascular tufts and asterisks depict the peripheral avascular area in diabetic retinas. Scale bars, 200 μm. C2M, control 2 months; C6M, control 6 months; D2M, diabetes 2 months; D6M, diabetes 6 months. (b) Quantitative determination of retinal blood vessel leakage by Evans Blue dye. Data show means ± SEM (n = 4) (black bars, diabetes; white bars, control). (c) Levels of cleaved PARP, active caspase-3, BAX and BCL2 in whole-retina extracts of control and diabetic mice were determined using immunoblotting. Data are presented as means ± SEM (n = 4); *p < 0.05 compared with control. (d) Double immunofluorescence staining for NG2 and TUNEL in retinal sections. Arrowheads indicate cells in the nerve fibre layer (NFL) that are α-SMA- and TUNEL-positive. Scale bars, 20 μm. CTL, control; DM, diabetes; WT, weight

Full size image

αA-Crystallin production is suppressed in diabetic retinas

To investigate different protein levels in retinas 2 and 6 months after the induction of diabetes, proteomic analyses were performed using total retinal extracts. 2-DE analysis disclosed two decreased spots in diabetic retinas compared with controls (Fig. 2a). These spots were identified as the two isotypes of αA-crystallin via PMF using MALDI-TOF mass spectrometry. Immunoblot analysis further confirmed the two bands (at approximately 17 and 20 kDa) of αA-crystallin. The total amount of both αA-crystallin isoforms was decreased by 59% in diabetic retinas compared with controls (p = 0.032, Fig. 2b). Consistent with αA-crystallin protein production, mRNA was decreased by 46% in diabetic retinas (p = 0.021, Fig. 2c). Double immunofluorescence staining for αA-crystallin and α-SMA revealed predominant levels of αA-crystallin in retinal pericytes in both control and diabetic mice (arrows and arrowheads in Fig. 2d).

Fig. 2
figure 2

Diabetes decreases production of αA-crystallin in retinal pericytes. (a) Total retinal extract from control and 2- and 6-month-old diabetic mice were separated by 2-DE and visualised by silver staining. Two spots (arrows) in diabetic retinas were identified as αA-crystallin with MALDI-TOF and were smaller than in controls. (b, c) Protein levels and mRNA expression of αA-crystallin in control and diabetic retinas were determined via (b) immunoblotting and (c) RT-PCR. Data are presented as means ± SEM (n = 4); *p < 0.05 compared with control. (d) Double immunofluorescence staining for αA-crystallin and α-SMA in retinal sections. Arrows and arrowheads indicate αA-crystallin- and α-SMA-positive cells in NFL, respectively. Scale bars, 25 μm. C2M, control 2 months; C6M, control 6 months; CTL, control; D2M, diabetes 2 months; D6M, diabetes 6 months; DM, diabetes; No., number; NFL, nerve fibre layer

Full size image

High glucose induces cell death in human retinal pericytes

To identify the effect of elevated glucose levels on αA-crystallin production, human retinal pericytes (HRPs) were exposed to high glucose for 5 days. Exposure of HRPs to a high concentration of glucose (30 mmol/l; high glucose) resulted in suppression of αA-crystallin production by 76% (p = 0.027) and enhancement of VEGF production by 1.9-fold (p = 0.014) compared with a low concentration of glucose (5 mmol/l; normal glucose). However, high glucose did not affect the levels of pericyte markers, such as α-SMA and NG2, in HRPs (Fig. 3a). Triple immunofluorescence staining for αA-crystallin, α-SMA and TUNEL showed that αA-crystallin production decreased while TUNEL-positive cells increased in high-glucose-exposed HRPs (Fig. 3b). These results suggest that αA-crystallin production is suppressed and apoptotic cell death induced in HRPs under high-glucose conditions.

Fig. 3
figure 3

Hyperglycaemia induces cell death in human retinal pericytes. (a) HRPs were exposed to normal (5 mmol/l) or high glucose (30 mmol/l) for 5 days. Levels of αA-crystallin, VEGF, α-SMA and NG2 were detected via immunoblotting and normalised to β-actin. Data are presented as means ± SEM (n = 5). *p < 0.05 compared with normal glucose. (b) Cells were fixed on chamber slides at 5 days after glucose treatment. Triple immunofluorescence staining for αA-crystallin (green), α-SMA (blue) and TUNEL (red) was performed. TUNEL staining was used as an apoptosis marker and α-SMA staining as a pericyte marker. Arrowheads show TUNEL-positive cells in high-glucose conditions. Scale bars, 20 μm. HG, high glucose; NG, normal glucose

Full size image

αA-Crystallin attenuates diabetes-induced retinal apoptotic cell death

To verify the effects of αA-crystallin on vascular pathology in the diabetic retina, we used adenovirus-mediated gene delivery systems. Schematic diagrams of the recombinant adenovirus expressing Gfp and αA-crystallin-Gfp (Ad-Gfp and Ad-αAc-Gfp) are presented in Fig. 4a. The injection schedule for adenovirus and STZ is shown in Fig. 4b. GFP fluorescence imaging in retinal vessels showed that adenovirus-mediated GFP production continued throughout the entire 10 weeks after adenovirus injections (arrowheads in Fig. 4c). 2-DE and immunoblotting experiments confirmed the overproduction of αA-crystallin in control and diabetic retinas after the injection of Ad-αAc-Gfp (arrows in ESM Fig. 1a, b). Adenovirus-mediated αA-crystallin production was increased in whole-retina extracts of diabetic mice (p = 0.046, Fig. 4d), but did not change in control mice. GFP production was similarly induced in all retinas after injection of recombinant adenovirus. Furthermore, levels of cleaved PARP, active caspase-3 and BAX in diabetic retina, which induce retinal cell death, were effectively reduced (p = 0.003, p = 0.013 and p = 0.039, respectively), while BCL2 production was increased (p = 0.025, Fig. 4e) with Ad-αAc-Gfp compared with Ad-Gfp treatment. These results support the theory that overproduction of αA-crystallin reduces apoptotic cell death in diabetic retinopathy.

Fig. 4
figure 4

Adenovirus-mediated production of αA-crystallin reduces apoptotic cell death in the diabetic retina. (a) Schematic diagram of adenovirus Ad-Gfp and Ad-αAc-Gfp used in this study. (b) Schematic diagram of the adenovirus and STZ injection strategy in the animal model. Ad-Gfp (1 × 109 pfu/μl, 2 μl) and Ad-αAc-Gfp (2 × 108 pfu/μl, 2 μl) were injected into the vitreous humour of both the control and diabetic mice. (c) Adenovirus-mediated gene expression was assessed by determining GFP production using fluorescence microscopy. GFP (arrowheads) was observed in retinal blood vessels at 4 and 10 weeks after adenovirus infection. Scale bars, 10 μm. (d, e) Production of αA-crystallin, cleaved PARP, active caspase-3, BAX and BCL2 in mouse retinas at 2 months from diabetes induction after adenovirus injection was detected via immunoblotting and normalised to α-tubulin production (black bars, diabetes; white bars, control). Data are presented as means ± SEM (n = 5). *p < 0.05 compared with Ad-Gfp-treated control. † p < 0.05 compared with Ad-αAc-Gfp-treated diabetic group. White bars, control; black bars, diabetes. CMV, cytomegalovirus promoter; CTL, control; DM, diabetes; ES, encapsidation signal; LITR, left inverted terminal repeats; RITR, right inverted terminal repeats

Full size image

αA-Crystallin prevents pericyte death and vascular leakage in the diabetic retina

To determine whether increased production of αA-crystallin has therapeutic effects on diabetic retinopathy, the number of pericytes and retinal permeability were measured after Ad-αAc-Gfp treatment. TUNEL-positive retinal pericytes decreased in Ad-αAc-Gfp-treated diabetic mice compared with those in Ad-Gfp-treated diabetic mice (Fig. 5a). In parallel, angiography with TMR-dextran infusion and immunofluorescence staining for α-SMA showed that the number of retinal pericytes per vessel (0.01 mm2) decreased after induction of diabetes compared with the control. Treatment with Ad-αAc-Gfp effectively blocked loss of pericytes in diabetic retinas (p = 0.024, Fig. 5b). Diabetes-induced vascular leakage was significantly reduced upon Ad-αAc-Gfp treatment (p = 0.046, Fig. 5c) In addition, TEM disclosed significantly thicker retinal capillary basement membrane in diabetic retinas compared with controls. Ad-αAc-Gfp attenuated pericyte loss and basement membrane thickening (Fig. 5d). TRITC-coupled, concanavalin-A-stained adherent leucocytes in vessels increased by 1.6-fold in diabetic retinas. However, Ad-αAc-Gfp did not significantly affect the number of adherent leucocytes compared with Ad-Gfp (Fig. 5e). Together, our results show that αA-crystallin is involved in diabetes-induced pericyte death and vascular leakage.

Fig. 5
figure 5

Adenovirus-mediated production of αA-crystallin prevents pericyte loss and vessel leakage in the diabetic retina. (a) Immunofluorescence staining for αA-crystallin and TUNEL was performed in retinal sections. Arrows represent cells in the NFL that are positive for αA-crystallin and TUNEL. Scale bars, 25 μm. (b) Angiography with TMR-dextran infusion and immunofluorescence staining for the pericyte marker α-SMA (green) was performed. The number of pericytes per vessel (0.01 mm2) was counted in mouse retinas 2 months after the induction of diabetes (black bars) and compared with control (white bars). Data are presented as means ± SEM (n = 6). (c) Quantitative determination of retinal blood vessel leakage by Evans Blue dye. Data are presented as means ± SEM: μl plasma (g retinal dry weight)−1 h−1 (n = 6). (d) Representative transverse sections of retinal capillaries. Arrowheads, basement membrane thickness. (e) After in situ labelling with TRITC-coupled concanavalin A lectin, adherent leucocytes (arrowheads) were photographed at ×400 magnification. Scale bars, 50 μm. Adherent leucocytes were counted in the whole retina (black bars, diabetes; white bars, control). Data are presented as means ± SEM (n = 5). *p < 0.05 compared with Ad-Gfp-treated control. † p < 0.05 compared with Ad-αAc-Gfp-treated diabetic group. CL, capillary lumen; CTL, control; DM, diabetes; E, endothelial cell; P, pericyte; RBC, red blood cell; WT, weight

Full size image

Discussion

We observed increased vessel leakage and production of pro-apoptotic proteins as well as pericyte apoptosis in mouse retina after induction of diabetes, which are characteristic features of pericyte loss and BRB breakdown [6, 7, 31]. Proteomic assessment of changes in gene expression confirmed decreases in two αA-crystallin subtypes in diabetic retinas compared with non-diabetic controls. Interestingly, αA-crystallin expression was specific to NG2-positive pericytes in the nerve fibre layer of both diabetic and control retinas. A previous study using gene microarrays showed that expression of many genes of the crystallin family are significantly downregulated in the retinas of mice with diabetic retinopathy. In addition, recent proteomic analysis of retina in oxygen-induced retinopathy mice showed that several crystallins, such as isoform 1 of αA-crystallin A, isoform 2 of αA-crystallin, αB-crystallin, γD-crystallin and β-A3/A1 crystallin, are downregulated [32]. Although several papers have reported that αA-crystallin is increased in the retinas of human diabetic patients and animal models of type 1 diabetes [21, 33–35], diabetes impairs the chaperone function of the α-crystallins by reducing their solubility. Increased production of αA-crystallin can protect retinal neurons from cell death through interaction with the pro-apoptotic proteins [21, 22].

Cellular apoptosis induced by hyperglycaemia occurs in many vascular cells and is crucial for the initiation of diabetic pathologies. High glucose induces pericyte apoptosis through activation of protein kinase C-δ in bovine retinal pericytes [36]. High glucose also induces mitochondrial dysfunction and apoptosis in bovine retinal pericytes. The detrimental effects of high glucose on mitochondrial function and cellular metabolism could play a role in the accelerated apoptosis associated with the retinal pericytes in diabetic retinopathy [37]. We showed, for the first time, that αA-crystallin decreased in human retinal pericytes cultured under high-glucose conditions, suggesting that decreased αA-crystallin in pericytes in response to hyperglycaemia leads to apoptotic cell death.

Advances in the field of ocular gene transfer are progressive, and viral-vector-mediated delivery of target genes that modulate endogenous protein levels in the retina may represent an effective gene therapy strategy to treat pathological angiogenesis in diabetic retinopathy [38, 39]. The viability of adenovirus as a gene therapy vector for ocular disease has recently been demonstrated in clinical trials for age-related macular degeneration and retinoblastoma [40, 41]. In the present study, we constructed a recombinant adenovirus as a vector to transfer αA-crystallin for gene therapy in diabetic retinas, and GFP fluorescence imaging showed that adenovirus-mediated gene expression was maintained in retinal vessels for up to 10 weeks after intravitreal injection. Ad-αAc-Gfp treatment significantly inhibited diabetes-induced αA-crystallin decrease, PARP cleavage and caspase activation. Thus, the recombinant adenovirus may provide a functional approach for further studies on αA-crystallin in diabetic retina.

As pericyte loss is directly associated with vessel leakage, we examined the effect of Ad-αAc-Gfp on vessel leakage in diabetic retinas. TUNEL-positive pericytes were reduced in the nerve fibre layer, but the number of pericytes in retinal capillaries was increased in Ad-αAc-Gfp-treated diabetic mice compared with Ad-Gfp-treated diabetic mice. The diabetes-mediated increase in vessel leakage was effectively blocked with Ad-αAc-Gfp administration. These results indicate that αA-crystallin effectively inhibited pericyte loss in the retina of diabetic mice, preventing vessel leakage.

In the retinal microcirculation, comparable numbers of endothelial cells and pericytes are present in the capillary wall [42]. Pericyte loss and thickening of the basement membrane are early histopathological changes that occur in the retinal microcirculation in diabetic retinopathy [43]. Clinical evidence has shown that basement membrane thickening is directly related to hyperglycaemia and can be reduced with good diabetic control [44]. Consistent with this, our TEM findings showed pericyte loss and thickening of the basement membrane in diabetic retinas at 2 months after STZ injection that were significantly less in mice receiving Ad-αAc-Gfp. Although the precise mechanisms underlying basement membrane thickening are not clear, our results suggest that αA-crystallin is protective against diabetic vascular damage, including basement membrane thickening.

Increased leucocyte adhesion may induce the formation of non-perfused capillaries, which is believed to be a major contributor to the increasing capillary permeability and angiogenesis that occur with progression of diabetic retinopathy [45–47]. Indeed, we found that adherent leucocytes were significantly increased in the entire retinas 2 months after induction of diabetes compared with controls, but leucocyte adhesion was not regulated by Ad-αAc-Gfp. Therefore, the protective effect of αA-crystallin on pericytes is independent of leucocyte adhesion, at least in our model.

Taken together, our data indicate that increased levels of αA-crystallin effectively prevents pericyte loss and BRB breakdown in diabetic retina. Therefore, we suggest that protection of pericytes by αA-crystallin may be a potent therapeutic target for capillary abnormalities in early diabetic retinopathy.