This study was conducted in accordance with the Declaration of Helsinki and the guidelines of the Ethics Committee of Aier Eye Hospital (Changsha, Hunan, China). This study was approved by the ethics committee of Aier Eye Hospital (AIER2018IRB21) and registered on the International Clinical Trials Registry Platform (ChiCTR1900025449). Consent was obtained from all the participants before collection. Blood samples were obtained from 21 DR patients aged 37–71 years. Negative control (NC) blood samples came from 11 pterygium patients without DM aged 53–69 years. All the samples were stored at −80°C for further experiments. Human retinas in this study were obtained from organ donors without DM or retinal diseases.

We predicted SIRT1 as a potential target of miR-29b-3p by using miRNA database (TargetScanHuman 7.2). Then the 3′-UTR of human SIRT1 containing the predicted binding sites [wild type (WT)] or mutated binding sites [mutant type (MUT)] was amplified and inserted into pmir-RB-Report^TM vector. The reporter plasmids and miR-29b-3p mimics or NC were co-transfected into HEK-293T cells using Lipofectamine 2000 (Invitrogen) to determine if SIRT1 is a direct target of miR-29b-3p. Firefly and Renilla luciferase activities were measured 48 h after transfection using the Dual-Glo^® Luciferase Assay System (Promega, Madison, WI, United States).

Primary HRMECs were isolated according to methods described previously (Fan et al., 2016); briefly, retinas were immersed in phosphate-buffered saline (PBS) with 5% penicillin–streptomycin (Gibco, Thermo Fisher Scientific, Waltham, MA, United States) for 5 min and then transferred into Dulbecco’s modified Eagle’s medium (DMEM) to remove vitreous. Retinas were then minced into small pieces and digested with type II collagenase (Sigma Aldrich, St. Louis, MO, United States) at 37°C for 30 min. Cell suspension was filtered through a 70-μm mesh and cultured in endothelial cell medium (ECM; ScienCell, San Diego, CA, United States). Culture medium was replaced every 2–3 days.

To explore the HRMEC damage in DR, we constructed an in vitro model of hyperglycemia and hypoxia conditions. HRMECs were cultured in 5.5 mmol/L of glucose (normal control), 5.5 mmol/L of glucose and 24.5 mmol/L of mannitol (osmotic pressure control), 30 mmol/L of glucose [hyperglycemia (HG)], 150 μmol/L of CoCl₂ (hypoxia), 30 mmol/L of glucose, and 150 μmol/L of CoCl₂ (HG-CoCl₂). Culture medium was refreshed every 24 h. SRT 1720 Hydrochloride (MedChemExpress, Monmouth Junction, NJ, United States) was used as an activator to upregulate the expression of SIRT1.

Immunofluorescence to platelet endothelial cell adhesion molecule-1 (PECAM-1/CD31) and von Willebrand factor (vWF) were used to determine the endothelial cell purity (Gao et al., 2013). Primary antibodies to CD31 (mouse anti-CD31 antibody, ab24590, 1:100, Abcam) and vWF (rabbit polyclonal to vWF antibody, ab6994, 1:100, Abcam) were used to detect CD31 and vWF, respectively. Goat anti-mouse IgG secondary antibody (Alexa Fluor 594) and goat anti-rabbit IgG secondary antibody (Alexa Fluor 488) were used to detect the primary antibodies separately. Nuclei were stained with DAPI (blue). Cells of passages between 3 and 5 and 95% positive for CD31 and vWF were used in this study.

Cells were seeded in 6-well and 96-well plates with a density of 2 × 10⁵/well and 4 × 10³/well. The miR-29b-3p mimics, inhibitors, and their NCs were purchased from RiboBio (Guangzhou, China) and transfected into cells using riboFECT^TM CP Reagent (Guangzhou, China) according to the manufacturer’s protocols. NC mimics labeled with Cy3 fluorescence (Guangzhou, China) were transfected to observe the transfect efficiency directly. After 30 h of transfection, the HRMECs were collected for terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) stain, cell counting kit-8 (CCK-8), quantitative real-time reverse transcriptase-PCR (RT-qPCR), and Western blot (WB) assay.

For apoptosis and viability assay, 4 × 10³ cells/well were seeded into 96-well plates and cultured at 37°C with 5% CO₂ in a humidified environment. The One Step TUNEL Apoptosis Assay Kit (Beyotime) was used for detecting apoptotic cells. Nuclei were stained with DAPI (blue). Fluorescent images were acquired by a fluorescence microscope (ECLIPSE Ts2R, Nikon). The quantification of TUNEL-positive cells was obtained by ImageJ software and calculated by GraphPad Prism version 5.0. Cell viability was determined by a CCK-8 assay (MedChemExpress, Monmouth Junction, NJ, United States). Seven replicates per group and a group without cells served as the blank. After being treated with different conditions, 100 μl of fresh culture medium with 10% CCK-8 solution was added to each well and incubated at 37°C for 1.5 h. The absorbance at 450 nm was observed by Synergy^TM HTX Multi-Mode Microplate Reader (Bio-Tek Technologies, Winooski, VT, United States). The relative viability of cells was calculated according to the manufacturer’s protocol.

MicroRNA was isolated with a microRNA kit (Omega Bio-Tek, Norcross, GA, United States) and reversed to cDNA with a reverse transcription kit (Roche, Basel, Switzerland); the stem-loop method was especially used for microRNA reverse transcription as described previously (Chen et al., 2005). The RT product was subjected to 45 cycles of qPCR reactions with ChamQ Universal SYBR qPCR Master Mix (Vazyme Biotech, Jiangsu, China) in a Roche LightCycler^® 96 System (Roche, Basel, Switzerland). U6 was used to normalize the expression of microRNA. The relative expression level of miRNA was calculated by the 2^{–Δ Δ CT} method. The specific primers for miR-29b-3p and U6 are listed in Table 1.

After being treated with different conditions, cells were washed twice with ice-cold PBS and lysed with radioimmunoprecipitation assay (RIPA) buffer (Beyotime) supplemented with protease inhibitor cocktail (Sigma-Aldrich). Lysates were then centrifuged at 12,000 rpm for 20 min at 4°C to collect the supernatant. Protein quantification was performed using BCA Protein Assay Kit (Solarbio) according to the company’s protocol. The supernatant proteins were concentrated with the method described previously (Zaiss et al., 2013). Briefly, supernatant, methanol, and chloroform were mixed thoroughly. The mixture was centrifuged at 10,000 rpm for 10 min at 4°C, and the supernatant was discarded carefully. Then another volume of methanol was added to the pellet and vortexed to mix thoroughly. The mixture was again centrifuged at 12,000 rpm for 10 min at 4°C, and the supernatant was discarded. After being air-dried for 5 min, the proteins were dissolved with the lysis buffer from a Caspase 3 Activity Assay Kit (Beyotime), and the quantification was performed using a Bradford Protein Assay Kit (Solarbio).

Protein was denatured using sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) loading buffer (Solarbio) by heating the samples at 98°C for 6 min. Electrophoresis was performed using 10% SDS-PAGE gel and transferred onto nitrocellulose membranes (Pall) and blocked with 5% non-fat milk containing Tween-20 for 1 h at room temperature, followed by incubation with primary antibodies overnight at 4°C. IRDye^® 800CW goat anti-rabbit/mouse secondary antibody (LI-COR) was used to detect primary antibody binding. The immunoblots were analyzed and quantified using ImageJ software. Antibodies to SIRT1(19A7AB4), Bax (E63), and Bcl-2 (E17) were obtained from Abcam. Caspase-3 and β-actin (8H10D10) antibodies were obtained from Cell Signaling Technology. Total protein stain was performed by using a REVERT Total Protein Stain kit (LI-COR). Relative quantification of cleaved caspase-3 in supernatant was achieved by normalizing each target to the value of total proteins.

Statistical Package for Social Science (SPSS) software version 19.0 and GraphPad Prism version 5.0 were used for descriptive analysis. The data were shown as mean ± standard deviation (SD). The results presented in the paper were representative of at least three different repetitions. Student’s t test was performed to assess differences between two means. A chi-square test for qualitative data was applied. One-way or two-way ANOVA followed by Bonferroni’s post hoc test was performed in multiple means comparison. Statistical significance was defined as p < 0.05.

The baseline data of clinical samples are shown in Table 2. To explore the expression pattern of miR-29b-3p and SIRT1 in DR patients, RT-qPCR and WB were performed. MiR-29b-3p RNA was upregulated to 3.2-fold (Figure 1A), and SIRT1 protein was downregulated to 65% (Figure 1B) in DR patients’ blood samples. With the miRNA online database (TargetScanHuman7.2 and miRBase), we found that SIRT1 is a direct target of miR-29b-3p (Figure 1C). Dual-luciferase reporter assay using HEK-293T cells showed the direct interaction of miR-29b-3p and the 3′UTR of SIRT1. After 48-h cotransfection, overexpressed miR-29b-3p reduced the luciferase activity of WT reporter but had no inhibition on the MUT reporter (Figure 1D). The results of this study showed that miR-29b-3p could inhibit the expression of SIRT1 by binding with the 3′-UTR of SIRT1, and SIRT1 might be the downstream target gene of miR-29b-3p.

Human retinal microvascular endothelial cell clusters began to form on the third day after plating. After 10-day culture, the cells showed an oval morphology and a contact-inhibited monolayer (Figures 2A,B). Immunofluorescence was performed to detect CD31 and vWF, which were well-known typical vascular endothelial cell markers. As a result, both CD31 (Figure 2C) and vWF (Figure 2D) were positive on the same cells (Figure 2E). These results verified the cell type and purification.

After treatment with the HG-CoCl₂ condition and different controls, obvious apoptosis was observed by TUNEL assay (Figures 3A,B), and cell viability was decreased compared with that of controls (Figure 3C). MiR-29b-3p was upregulated (Figure 3D) in the HG-CoCl₂ condition, whereas SIRT1 protein was downregulated (Figure 3E). HG-CoCl₂ upregulated Bax/Bcl-2 ratio and the expression of cleaved caspase-3 significantly (Figures 3E,F).

After 30-h transfection, annulus red fluorescence was observed around the nucleus (Figure 4A). HRMECs were transfected with miR-29b-3p mimics (miR-29b-3pm), inhibitors (miR-29b-3pi), and their NCs. RT-qPCR and WB were performed to verify the transfection effect. The mRNA expression level of miR-29b-3p in miR-29b-3pm was elevated (Figure 4B), whereas miR-29b-3pi decreased the expression of miR-29b-3p obviously (Figure 4C). Relative SIRT1 protein expression was downregulated and the ratio of Bax/Bcl-2 was upregulated in miR-29b-3pm to NC, whereas relative SIRT1 protein expression was upregulated and the ratio of Bax/Bcl-2 was downregulated in miR-29b-3pi to NC (Figure 4D).

MiR-29b-3p mimics were transfected into HRMEC with or without SRT1720. Upregulated miR-29b-3p increased apoptosis (Figures 5A,B) and decreased cell viability in HRMEC (Figure 5C), whereas apoptosis was decreased (Figures 5A,B) and cell viability was upregulated after the treatment of SRT1720 (Figure 5C). Relative SIRT1 protein expression was decreased and Bax/Bcl-2 ratio and cleaved caspase-3 were upregulated after the transfection of miR-29b-3p, whereas after the treatment of SRT1720, relative SIRT1 protein expression was upregulated and Bax/Bcl-2 ratio and cleaved caspase-3 were downregulated (Figures 5D,E).

MiR-29b-3pi and SRT1720 effectively alleviated HRMEC apoptosis induced by HG-CoCl₂ (Figures 6A,B) and improved cell viability (Figure 6C). Relative SIRT1 protein expression was increased and Bax/Bcl-2 ratio (Figure 6D) and cleaved caspase-3 were also downregulated obviously by miR-29b-3pi and SRT1720 (Figure 6E).