Human retinal Müller glial cells (MIO-M1) (a generous gift from Prof. A. Limb, Institute of Ophthalmology, University College London, UK) were cultured in Dulbecco’s Minimal Essential Medium (DMEM) containing 1 g/L glucose with 10% (v/v) fetal bovine serum and 100 units/mL of Penicillin-Streptomycin solution. Confluent cells were starved overnight in serum-free DMEM to minimize the effects of serum. Subsequently, the cell cultures were either left untreated or stimulated for 24 h.

Human retinal microvascular endothelial cells (HRMECs) were purchased from Cell Systems Corporation (Kirkland, WA, USA) and maintained in complete serum-free media (Cat No SF-4Z0–500, Cell System Corporation) supplemented with “Rocket Fuel” (Cat No SF-4Z0-500, Cell System Corporation), “Culture Boost” (Cat No 4CB-500, Cell System Corporation) and antibiotics (Cat No 4Z0–643, Cell System Corporation) at 37°C in a humidified atmosphere with 5% CO₂. We used HRMECs up to passage 8 for all of our experiments. Cell cultures at about 80% confluency were starved in a minimal medium (medium supplemented with 0.25% “Rocket Fuel” and antibiotics) overnight to eliminate any residual effects of growth factors.

The following stimuli were used: diabetic mimetic conditions included treatment of Müller cells or HRMECs with 300 μM of the hypoxia mimetic agent cobalt chloride (CoCl₂) (Cat No A1425-L, Avonchem Limited, UK), or 25mM glucose (Cat No GL0125100, Scharlau S.L, Gato Prez, Spain). For high-glucose (HG) treatment, 25 mM mannitol (Cat No MA01490500, Scharlau S.L, Gato Prez, Spain) was used as a control.

Müller cells were treated with 10 μM of the ERR-α inhibitor XCT790 (Cat No X4752, Sigma) or 20 μM of the PGC-1α inhibitor SR-18292 (Cat No SML2146, Sigma) for 24 h to determine their inhibitory effect. Müller cells were treated with 20 μM of the PGC-1α inducer ZLN005 (Cat No SML0802, Sigma) for 48 h.

Next, the treatment of human Müller cells with diabetic mimetic conditions, 300 μM of the hypoxia mimetic agent CoCl₂ or 25mM glucose for 24 h was performed in the absence or presence of 1 h pretreatment with 10 μM of the ERR-α inhibitor XCT790, or 20 μM of the PGC-1α inhibitor SR-18292, or 10 μM of the HIF-1 α inhibitor YC-1 (Cat No Y102, Sigma).

After the treatment as detailed above, cell supernatants were collected and processed for enzyme-linked immunosorbent assay (ELISA) analysis. Harvested cells were lysed in a radioimmunoprecipitation assay (RIPA) lysis buffer (sc-24948, Santa Cruz Biotechnology, Inc.) for Western blot analysis.

As described previously (16, 17), adult male Wistar rats of 8–9 weeks of age (200–220 g) were fasted overnight, and a single-bolus dose of streptozotocin of 60 mg/kg in 10 mM sodium citrate buffer, pH 4.5, (Sigma, St. Louis, MO, USA) was injected intraperitoneally. Equal volumes of citrate buffer were injected in age-matched control rats. Rats were considered diabetic if their blood glucose levels were in excess of 250 mg/dL. The rat model of diabetic retinopathy demonstrates the early retinal changes of nonproliferative diabetic retinopathy that occur in humans, such as inflammation and breakdown of the blood-retinal barrier resulting in increased vascular permeability (18, 19). After 4 weeks of diabetes, the rats were euthanized, and retinas were isolated and frozen immediately in liquid nitrogen and stored at -80°C until analyzed. Similarly, retinas were obtained from age-matched nondiabetic control rats.

Undiluted vitreous fluid samples (0.3–0.6 ml) were obtained from 37 patients with PDR during pars plana vitrectomy, for the treatment of tractional retinal detachment, and/or nonclearing vitreous hemorrhage and processed as described previously (16, 17). We compared the samples from diabetic patients with those of a clinical control cohort. The control group consisted of 30 patients who had undergone vitrectomy for the treatment of rhegmatogenous retinal detachment with no proliferative vitreoretinopathy. Control subjects were clinically checked to be free from diabetes or other systemic disease. Epiretinal fibrovascular membranes were obtained from 13 patients with PDR during pars plana vitrectomy for the repair of tractional retinal detachment. The epiretinal membranes were processed as previously described (16, 17). Membranes were fixed for 2h in 10% formalin solution and embedded in paraffin. Patients with PDR were 25 (67.6%) males and 12 (32.4%) females, whose ages ranged from 35 to 75 years with a mean of 54.1 ± 10.8 years. The duration of diabetes ranged from 8 to 35 years with a mean of 19.3 ± 7.4 years. Twenty-two patients had insulin-dependent diabetes mellitus, and 15 patients had noninsulin-dependent diabetes mellitus. Treatment for hypertension was used by 18 patients, 6 patients had diabetic nephropathy, and 2 patients had cardiovascular disease. The control group included 20 (66.7%) males and 10 (33.3%) females, whose ages ranged from 29 to 75 years with a mean of 53.3 ± 15.9 years. There were no significant differences in the male to female ratio (p=0.938) and age (p=0.820) between patients with PDR and control patients.

ELISA kits for human vascular endothelial growth factor (VEGF) (Cat No DY293B), human monocyte chemoattractant protein-1 (MCP-1)/CCL2 (Cat No DY279), human angiopoietin 2 (Cat No DY623), and human matrix metalloproteinase-9 (MMP-9) (Cat No DY911) were purchased from R&D Systems, Minneapolis, MA, USA. An ELISA kit for human PGC-1α (Cat No MBS772836) was purchased from MY Biosource (San Diego, USA).

Levels of VEGF, angiopoietin 2, and PGC-1α in vitreous fluid; and VEGF, MCP-1, angiopoietin 2, and MMP-9 in cell culture medium were determined using the afore-mentioned ELISA kits according to the manufacturer’s instructions. The minimum detection limits for VEGF, MCP-1, angiopoietin-2, MMP-9, and PGC-1α ELISA kits were approximately 12 pg/ml, 9 pg/ml, 10 pg/ml, 10 pg/ml, 10 pg/ml, respectively.

Retina and cell lysates were homogenized in Western blot lysis buffer [30 mM Tris-HCl; pH 7.5, 5 mM EDTA, 1% Triton X-100, 250 mM sucrose, 1 mM sodium vanadate, and a complete protease inhibitor cocktail from Roche (Mannheim, Germany)]. After centrifugation of the homogenates (14,000 X g for 15 min, 4°C), protein concentrations were measured in the supernatants (Bradford protein assay kit; Bio-Rad Laboratories, Hercules, CA, USA). Equal amounts, either 30 µg or 50 µg of the protein extracts from lysates were subjected to SDS–PAGE and transferred onto nitrocellulose membranes.

To determine the presence of PGC-1α and ERR-α in the vitreous fluid samples, equal volumes (10 μL) of vitreous samples were boiled in Laemmli’s sample buffer (1:1, v/v) under reducing condition for 10 min.

Immunodetection was performed with the use of rabbit polyclonal anti-PGC-1α antibody (1:1000, ab54481, Abcam), mouse monoclonal anti-PGC-1α antibody (1:1000, MAB10784-SP, R&D Systems), rabbit polyclonal anti-PGC-1α antibody (1:1000, NBP104676, Novus Biologicals, LLC, Centennial, USA), mouse monoclonal anti-ERR-α antibody (1:1000, sc-65718, Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA), rabbit polyclonal anti-ERR-α (1:1000, Cat. no. ab137489, Abcam).

Nonspecific binding sites on the nitrocellulose membranes were blocked (1.5 h, room temperature) with 5% non-fat milk made in Tris-buffered saline containing 0.1% Tween-20 (TBS-T). Three TBS-T washings (5 min each) were performed before the secondary antibody treatment at room temperature for 1 h. The secondary antibodies included goat anti-rabbit immunoglobulin (SC-2004) and goat anti-mouse immuno-globulin (SC-2005) (1:2000, Santa Cruz Biotechnology Inc.). To verify equal loading, the nitrocellulose membranes were stripped and reprobed either with β-actin-specific antibody (1:2000, sc-47778, Santa Cruz Biotechnology Inc.) or β-tubulin-specific antibody (1:2000, ab21058, Abcam). Bands were visualized with the use of high-performance chemiluminescence (G: Box Chemi-XX9 from Syngene, Synoptic Ltd., Cambridge, UK), and the band intensities were quantified with the use of GeneTools software (Syngene by Synoptic Ltd.).

Reactive oxygen species (ROS) generation was measured in Müller glial cell monolayers using 2′-7′-dichlorofluorescin-diacetate (DCFH-DA). Briefly, cells were grown in standard cell culture media so that 3 x 10⁶–4 x 10⁶ cells were obtained the day before the experiment. Next, cells were harvested and seeded in a clear bottom 24-well microplate with 1 x 10⁵ cells per well. Cells were allowed to adhere overnight. Overnight starved cells were treated either with 10 mM hydrogen peroxide (H₂O₂) for 1 h or 25mM glucose for 24 h in the absence or presence of 48 h pretreatment with the PGC-1α inducer ZLN005 (20 μM). For high-glucose (HG) treatment, 25 mM mannitol was used as a control. After washing with 250 μL/well of PBS, the cells were stained by adding 200 μL/well of the 10 μM DCFH-DA (Invitrogen, CA) for 45 minutes at 37°C in the dark. After removing DCFH-DA solution, cell monolayers were washed with PBS and Cellular ROS production was measured immediately on a fluorescence plate reader (SpectraMax Gemini-XPS, Molecular Devices, CA, USA) with excitation and emission wavelengths of 488 nm and 525 nm, respectively. To normalize the fluorescence intensities with protein concentrations, cells were lysed and 1 μL of the supernatant transferred to a clear 96 well plate containing 100 μL of 1:5 diluted protein assay (Bradford assay) solution to measure the protein concentration.

The immunohistochemical and the sequential double immunohistochemical stainings were performed using the Leica Bond Max autostainer system (M496834 -Leica, Diegem, Belgium) with Bond Polymer Refine Red Detection kit (DS9390, Leica) and Bond Polymer Refine Detection kit (DS9800, Leica). The Bond Polymer Refine Red Detection kit is a biotin-free, polymeric alkaline phosphatase (AP)-linker antibody conjugate system for the detection of tissue-bound mouse and rabbit IgG and some mouse IgM primary antibodies. Bond Polymer Refine Red Detection utilizes a novel controlled polymerization technology to prepare polymeric AP-linker antibody conjugates. The Bond Polymer Refine Detection is a biotin-free, polymeric horseradish peroxidase (HRP)-linker antibody conjugate system for the detection of tissue-bound mouse and rabbit IgG and some mouse IgM primary antibodies. Bond Polymer Refine Detection utilizes a novel controlled polymerization technology to prepare polymeric HRP-linker antibody conjugates.

The Bond Polymer Refine Red Detection works as follows:

The Bond Polymer Refine Detection works as follows:

Sections were cut at 3-µm thick with a microtome (HistoCore MULTICUT-Semi-Automated Rotary Microtome; reference: 14051856372 - Leica). The used slides are from Dako (IHC Microscope slides – reference: K8020). We examined 75 slides. The paraffin-embedded tissue sections, the primary antibodies and the Bond Polymer Refine Red Detection kit (DS9390, Leica) were loaded onto the Bond Max autostainer and dewaxed (AR9222, Leica), followed by antigen retrieval. For CD31 and ERR-α, antigen retrieval was performed by boiling (99°C) the sections in citrate based buffer [pH 5.9–6.1] [ER1-AR9961, Bond Epitope Retrieval Solution 1; Leica] for 20 minutes. For CD45 and PGC-1α detection, antigen retrieval was performed by boiling (99°C) the sections in Tris/EDTA buffer [pH 9] [ER2-AR9640, Bond Epitope Retrieval Solution 2; Leica] for 20 minutes. Subsequently, the sections were incubated for 60 minutes with mouse monoclonal anti-CD31 (ready-to-use; clone JC70A; Dako, Glostrup, Denmark), mouse monoclonal anti-CD45 (ready-to-use; clones 2B11+PD7/26; Dako), rabbit polyclonal anti-PGC-1α antibody (1:100; ab54481, Abcam, Cambridge, UK) and rabbit polyclonal anti-ERR-α antibody (1:500; ab137489, Abcam). Optimal working conditions for the antibodies were determined in pilot experiments on kidney and heart sections. Then the sections were incubated with secondary antibodies, first with a post primary IgG for 20 mins, followed by a poly-AP IgG for 30 mins. The reaction product was visualized by incubation for 2 times 15 minutes with the Fast Red chromogen, resulting in bright-red immunoreactive sites. The slides were then faintly counterstained with hematoxylin for 5 mins. In between each step sections were washed with wash buffer (AR 9590, Leica).

To identify the phenotype of cells expressing PGC-1α and ERR-α, sequential double immunohistochemistry was performed. A Bond Polymer Refine Red Detection kit (DS9390, Leica) was used in combination with a Bond Polymer Refine Detection kit (DS9800, Leica). After dewaxing and antigen retrieval, the sections were incubated for 60 mins with the first primary antibody (anti-CD45) and subsequently treated with peroxidase-conjugated secondary antibody (post primary IgG linker reagent) for 8 mins to define the leukocytes. The resulting immune complexes were visualized by enzymatic reaction of the 3, 3’-diaminobenzidine tetrahydrochloride substrate for 5 mins. Incubation of the second primary antibodies (anti-PGC-1α and ERR-α) for 60 mins was followed by treatment with alkaline phosphatase-conjugated secondary antibody (poly-HRP IgG reagent) for 30 mins and fast red reactions for 2 times 15 mins. No counterstain was applied. Negative controls were by omission of the primary antibody from the staining protocol. Instead, the ready-to-use DAKO Real antibody Diluent (Agilent Technologies Product Code 52022) was applied.

The level of vascularization in epiretinal fibrovascular membranes was determined by immunodetection of the vascular endothelium marker CD31. Immunoreactive blood vessels and cells were counted in five representative fields, with the use of an eyepiece calibrated grid in combination with the 40x objectives. These representative fields were selected based on the presence of immunoreactive blood vessels and cells. With this magnification and calibration, immunoreactive blood vessels and cells present in an area of 0.33 mm × 0.22 mm were counted.

Data were collected, stored and managed in a spreadsheet using Microsoft Excel 2010^® software. Data were analyzed and figures prepared using SPSS^® version 21.0 (IBM Inc., Chicago, Illinois, USA). Age and gender of the human subjects from whom the samples were obtained were matched for cases and controls; independent t-test for age and Chi-square test for gender were used to test the differences between the two groups. Tests for normality were done using Shapiro-Wilk test and Q-Q plots. The data were normally distributed and hence reported as mean and standard deviation (SD) and illustrated with bar charts. Consequently, One-Way ANOVA and Independent t-tests (applying Bonferroni correction where necessary) were done to test the differences between the groups. Additionally, Pearson’s correlation analysis was carried out. Any output with a p-value below 0.05 was interpreted as an indicator of statistical significance.

Epiretinal fibrovascular membranes were studied by immunohistochemical analysis to examine the expression and tissue localization of PGC-1α (n=13) and ERR-α (n=10). No staining was observed in the negative control slides ( Figure 1A ). All membranes showed pathologic neovessels expressing the vascular endothelial cell marker CD31 ( Figure 1B ). Immunoreactivity for PGC-1α was detected in all membranes in endothelial cells lining pathologic neovessels and stromal cells ( Figure 1C ). Co-localization studies revealed that immunoreactivity for PGC-1α was detected in stromal leukocytes expressing CD45 ( Figure 1D ). In a way similar to PGC-1α, immunoreactivity for ERR-α was detected in vascular endothelial cells and stromal cells ( Figures 2A, B ). In line with its expected subcellular distribution, ERR-α was often visualized within cellular nuclei. Stromal cells were leukocytes co-expressing CD45 ( Figure 2C ). Significant positive correlations (Pearson’s correlation coefficient) were detected between the numbers of pathologic neovessels expressing CD31, reflecting the angiogenic activity, and the numbers of blood vessels (r=0.757; p=0.003) and stromal cells (r=0.662; p=0.014) ex-pressing PGC-1α ( Figure 3A ). Similarly, significant positive correlations were detected between the numbers of neovessels expressing CD31 and the numbers of blood vessels (r=0.961; p<0.001) and stromal cells (r=0.810; p=0.005) expressing ERR-α ( Figure 3B ).

Western blot analysis of equal volumes of vitreous fluid confirmed the presence of PGC-1α and ERR-α in the vitreous fluid. The presence of PGC-1α and ERR-α in the vitreous fluid reflects cell death accompanying the diabetic process as well as cell lysis induced by the freeze-thaw cycle. PGC-1α immunoreactivities were detected as three protein bands at approximately 120 kDa, 90 kDa and 40 kDa ( Figure 4A ). These corresponded to the full-length PGC-1α at about 90 kDa and 120 kDa. In all likelihood the predominant 40-kDa band represented the N-truncated PGC-1α isoform (20–23). This functional and biologically active proteoform of PGC-1α originates from alternative mRNA splicing introducing a premature stop codon (20–23). In Figure 4A , PGC-1α immunodetection was performed with the use of an antibody from Abcam. Additional control analyses with antibodies from R & D Systems and Novus biologicals, yielded similar immune reactivities. ERR-α immunoreactivities were expressed as 2 protein bands at approximately 90 kDa and 50 kDa ( Figure 4B ). Scanning analysis of immunoreactivities demonstrated increased levels of the 90-kDa and 40-kDa PGC-1α isoforms ( Figure 4A ) and both ERR-α forms ( Figure 4B ) in vitreous of PDR patients in comparison with controls.

To corroborate the above findings, we measured total immunoreactivities in vitreous samples. PGC-1α levels in vitreous samples from patients with PDR were significantly higher than the levels in nondiabetic controls (p<0.001) ( Table 1 ). To allow correlation analysis with ongoing angiogenesis in the ocular microenvironment of patients with PDR, we analyzed the levels of the proangiogenic factors VEGF and angiopoietin 2. VEGF and angiopoietin 2 levels were significantly enhanced in vitreous samples from patients with PDR compared to nondiabetic control samples (p=0.005 and p<0.001, respectively) ( Table 1 ). Significant positive correlations (Pearson’s correlation coefficient) were found between levels of PGC-1α and angiopoietin 2 (r=0.439; p<0.001) and between levels of VEGF and angiopoietin 2 (r=0.484; p<0.001).

Next we performed subgroup analyses to investigate the effects of gender and type of diabetes on the levels of PGC-1α, VEGF and angiopoietin 2 in vitreous samples from patients with PDR. As shown in Table 2 , there were no significant differences.

Next, we studied the regulation of retinal expression of PGC-1α and ERR-α in streptozotocin-induced diabetic rats. Whereas this animal model may not fully reflect all aspects of long-term diabetic retinopathy observed in the human samples illustrated above, it is accepted as a model to investigate short-term effects of acute inflammation associated with diabetic retinopathy. With the use of Western blot analysis of homogenized retinal tissue, we detected PGC-1α as 2 protein bands at approximately 110 kDa and 45 kDa. The predominant PGC-1α isoform was found at 45 kDa ( Figure 5A ). Densitometric analysis showed increased PGC-1α protein levels in the retina of rats after 4 weeks of STZ-induced diabetes ( Figure 5A ). ERR-α immunoreactivities were expressed as 3 protein bands at approximately 90 kDa, 60 kDa and 45 kDa ( Figure 5B ). Densitometric analysis demonstrated increased ERR-α protein levels in the retina of diabetic rats ( Figure 5B ).

To confirm the observed in vivo expression of PGC-1α and ERR-α by endothelial cells in epiretinal fibrovascular membranes from patients with PDR, we performed in vitro experiments on HRMECs. We showed by Western blot analysis that HRMECs constitutively express PGC-1α and ERR-α ( Figure 6 ). Treatment of HRMECs with high-glucose or the hypoxia-mimetic agent CoCl₂ did not affect the expression of PGC-1α ( Figure 6A ) or ERR-α ( Figure 6B ). Similarly, Western blot analysis demonstrated that Müller cells constitutively express PGC-1α and ERR-α ( Figure 7 ). Treatment of Müller cells with high-glucose or CoCl₂ did not affect the expression of PGC-1α ( Figure 7A ). However, treatment with high-glucose or CoCl₂ induced significant upregulation of the 50 kDa protein band of ERR-α and significant downregulation of the 90 kDa protein band of ERR-α ( Figure 7B ).

By the demonstration that the PGC-1α/ERR-α axis is upregulated in the ocular microenvironment of patients with PDR and in the retina of diabetic rats, we started to dissect the effects of inhibiting the PGC-1α/ERR-α pathway on diabetic conditions at the cellular level in vitro. Müller cells were cultured in the absence or presence of SR-18292 or XCT790. With the use of Western blot analysis of cell lysates, we demonstrated that treatment with SR-18292 induced significant down-regulation of the 55 kDa protein band of PGC-1α ( Figure 8A ). Similarly, treatment of Müller cells with XCT790 induced significant downregulation of the protein levels of ERR-α ( Figure 8B ). With the use of ELISA analysis, we demonstrated that the treatment of Müller cells with SR-18292 or XCT790 did not affect the expression of the proangiogenic factors VEGF and angiopoietin 2 in the culture medium as compared to untreated control (data not shown).

ELISA analysis revealed that treatment of Müller cells with the diabetic mimetic condition high-glucose (HG) induced significant upregulation of the proangiogenic factors VEGF, angiopoietin 2 and MMP-9 and the inflammatory chemokine MCP-1/CCL2 in the culture medium as compared to untreated control. On the one hand, pretreatment with SR-1829, XCT790 or the HIF-1α transcription factor inhibitor YC-1 significantly attenuated the levels of VEGF, angiopoietin 2 and MCP-1/CCL2 induced by HG. On the other hand, SR-18292, XCT790 or YC-1 did not affect the expression of MMP-9 induced by HG ( Figure 9 ).

ELISA analysis demonstrated that treatment of Müller cells with the hypoxia mimetic agent CoCl₂ induced significant upregulation of VEGF, angiopoietin 2, MMP-9 and MCP-1/CCL2 in the culture medium as compared to untreated control. Pretreatment of Müller cells with SR-18292, XCT-90 or YC-1 significantly attenuated the levels of VEGF, angiopoietin 2 and MCP-1/CCL2. SR-18292 and XCT790 did not affect the expression of MMP-9, however, YC-1 significantly attenuated the levels of MMP-9 induced by CoCl₂ ( Figure 10 ).

Collectively, these data indicated that the PGC-1α/ERR-α pathway is required for the diabetic mimetic conditions-induced upregulation of VEGF, angiopoietin 2 and MCP-1/CCL2.

With the use of Western blot analysis, we demonstrated that treatment of Müller cells with ZLN005 for 48 h induced significant upregulation of PGC-1α as compared to untreated control ( Figure 11A ). With the use of ELISA analysis, we demonstrated that treatment of Müller cells with ZLN005 induced significant upregulation of the proangiogenic factor VEGF, but not angiopoietin 2, MMP-9 and MCP-1/CCL2, in the culture medium as compared to untreated control ( Figure 11B ). These findings suggest that ZLN005 increased PGC-1α and its downstream transcription factor in Müller cells.

Figure 12 demonstrates the changes in the signal of DCF fluorescence, which is an indicator of reactive oxygen species (ROS), in Müller cells. High-glucose induced significant upregulation of DCF fluorescence as compared to untreated control. Pretreatment with ZLN005 significantly attenuated ROS generation induced by high-glucose ( Figure 12A ). Treatment of Müller cells with exogenous reactive oxygen species H₂O₂ induced ROS generation. Pretreatment with ZLN005 significantly downregulated H₂O₂-induced ROS generation ( Figure 12B ).