EPCs were isolated from rat bone marrow and cultured as previously described (Zhang et al., 2014). Briefly, the bone marrow was separated from the femora and tibiae of SD rats (3 weeks old) under sterile conditions. Bone marrow mononuclear cells (BMMNCs) were further harvested by density gradient centrifugation (Histopaque 1,083, Sigma-Aldrich) at 400×g for 30 min at room temperature. The mononuclear cell layer was collected and washed with the complete EGM-2 medium (CC-3202, Lonza) three times. Then, EPCs were re-suspended with the complete EGM-2 medium and seeded in a six-well plate at a density of 8.5 × 10⁵ cell/cm². They were incubated at 37 °C with 5% CO₂ in humidified atmosphere. The culture medium was changed every 48 h. The non-adherent cells were removed after 3 days, and EPCs were used for the following experiments after a 7-days culture. EPCs were exposed to 30 mM d-glucose for 24 h to mimic a high-glucose condition. EPCs were treated with CoCl₂ (200 μmol/L) for 24 h to mimic hypoxic stress.

The identification of EPC phenotypes was done by using fluorescence microscopy and flow cytometry. For the detection by fluorescence microscopy, both double-positive staining and immunofluorescent staining were performed. After 7 days of culture, EPCs were incubated with Dil-conjugated acetylated low-density lipoprotein (Dil-ac-LDL, L3484, Invitrogen) and Ulex Europaeus Agglutinin I (UEA-I, L9006, Sigma-Aldrich). Double-positive staining for Dil-ac-LDL and UEA-l was considered to represent EPCs and counted at × 40 magnification under a fluorescence microscope (IX73P1F, Olympus). For extracellular labeling of CD133, CD34, and VEGFR2, EPCs were fixed in 4% PFA for 30 min; blocked with 5% bovine serum albumin; and labeled sequentially with the antibodies against CD133, CD34, and VEGFR2, respectively (1:200, Santa Cruz Biotechnology). They were subsequently washed three times in PBS and counterstained with diamidino-phenylindole for 5 min. Following washing, the images were captured by fluorescence microscopy. For the detection by flow cytometry, EPCs were digested with 0.25% trypsin and permeabilized with 0.1% Triton-X at 37°C for 30 min. Subsequently, these cells were incubated with antibodies against CD133, CD34, or VEGFR2 (1:50, Santa Cruz Biotechnology); detected by a CytoFLEX flow cytometer (Beckman Coulter); and analyzed with the CytExpert software.

Total RNAs were isolated from EPCs using the Small RNA Purification Kit (EZB-RN3, China) following the manufacturer’s instructions. The quality and quantity of total RNAs were measured using the NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific), and RNA integrity number (RIN) was generated using the Agilent 2,100 Bioanalyzer (Agilent Technologies). The samples with RIN >7.0 were subjected for small RNA sequencing.

About 1 μg of total RNAs from each sample was used to create small RNA library using the Small RNA Sample Preparation Kit (Illumina, San Diego, CA, United States) according to the manufacturer’s protocols. Briefly, the 3 and 5′ end of RNAs were ligated by the adapters. Then, the adapter ligated RNAs were reversely transcribed into cDNAs. Polymerase chain reactions (PCRs) were used to amplify the libraries, and polyacrylamide gel electrophoresis was performed for size selection. Finally, the amplified cDNAs were purified, and the library quality was evaluated on the Agilent 2,100 Bioanalyzer. All libraries were sequenced on the Illumina sequencing platform (HiSeqTM 2,500).

The redundant reads including the extraneous sequences, adaptor sequences, and low-quality sequences were removed from the raw data using Illumina CASAVA (version 1.8). The clean reads were obtained by discarding sequences shorter than 15 nt or longer than 35 nt using the cutadapt and fqtrim. Then, Bowtie software was used to index the genome. Then, the clean reads were aligned to the rat reference genome assembly using the miRDeep software. The remaining reads were used to identify the known miRNAs by comparing with the corresponding miRNA precursor sequences from miRBase. The read counts of the identified miRNA in each sample were generated, and only the ones with an overlap more than 50% of miRBase-defined miRNAs were considered significant. The analysis of differentially expressed miRNAs between the high glucose and control groups was performed using DESeq packages. The miRNAs with a nominal p value < 0.05 and a fold change >2 were identified as differentialy expressed.

To predict the possible biological function of differentially expressed miRNAs, the target genes were predicted by TargetScan (http://www.targetscan.org). The biological process (BP), cellular component (CC), molecular function (MF), and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway of the miRNAs were exhibited using the Gene Ontology (GO) project (http://www.geneontology.org) and the KEGG database (http://www.kegg.jp/), respectively. Fisher’s exact test was used to determine the significant GO terms and pathway categories (p < 0.05).

To verify the results of small sequencing, quantitative reverse transcription-polymerase chain reaction (qRT-PCR) was conducted to detect the expression pattern of 20 dysregulated miRNAs. The extracted RNAs were reversely transcribed into cDNAs using MicroRNA Reverse Transcription Kit (EZB-miRT4, China) with the specific stem-loop RT-primer, according to the manufacturer’s instructions. PCR products were amplified using the EZ-probe qPCR Master Mix for microRNA (EZB-miProbe-R2, China) on the PikoReal Real-Time PCR System (Thermo Fisher Scientific). The relative miRNA expression level was determined using U6 as an endogenous control and calculated with the 2^−ΔΔCt method. The primer sequence of miRNAs are shown in Supplementary Table S1.

EPCs were grown to approximately 70% confluence and transfected with the miR-375–3p mimics, inhibitor, or the corresponding negative controls using Lipo 6,000 Transfection Reagent (C0526, Beyotime, China) according to the manufacturer’s protocols. The sequences of miR-375–3p mimics, inhibitor, and the negative controls are shown below: miR-375–3p mimics sense: 5′-UUU​GUU​CGU​UCG​GCU​CGC​GUG​A-3′, antisense: 5′-ACG​CGA​GCC​GAA​CGA​ACA​AAU​U-3′; miR-375–3p inhibitor: 5′-UCA​CGC​GAG​CCG​AAC​GAA​CAA​A-3′; negative control mimics sense: 5′-UUC​UCC​GAA​CGU​GUC​ACG​UTT-3′, antisense: 5′-ACG​UGA​CAC​GUU​CGG​AGA​ATT-3′; negative control inhibitor: 5′-CAG​UAC​UUU​UGU​GUA​GUA​CAA-3′. The transfection efficacy was examined by qRT-PCRs.

The viability of EPCs was detected by MTT assay. Briefly, EPCs were seeded onto 96-well plates at the density of 1.5 × 10⁴ cells per well. After the required treatment, EPCs were incubated with MTT (5 mg/ml) in darkness at 37°C for 3 h. Then, the culture medium was removed, and DMSO (100 μl/well) was added into each well to dissolve the formazan crystals. Finally, the absorbance was determined at 595 nm by a microplate reader (FilterMax F5, Molecular Devices).

Cell proliferation was examined by CCK-8 assay (Dojindo, Kumamoto, Japan). Briefly, EPCs were plated onto a 96-well plate, and CCK-8 solution was added to each well. After incubation for 1 h at 37°C, the absorbance was determined at a 450-nm wavelength using a microplate reader (Molecular Devices).

The apoptosis of EPCs was determined using the Annexin V-FITC/PI Apoptosis Detection Kit (Cat: A211-01, Vazyme). After the required treatment, EPCs were digested with trypsin without EDTA and washed with PBS. These cells were re-suspended in 100 μl binding buffer. Under dark condition, 5 μl of Annexin V-FITC and PI Staining Solution were added to the suspension. Then, 400 μl binding buffer was added after 20 min. Finally, these cells were detected by the CytoFLEX flow cytometer.

Transwell assay was conducted to detect cell migration ability. After the required treatment, EPCs were seeded onto the top chamber with polycarbonate filters (8-μm pore size; Corning). They were cultured for 18 h at 37°C, and the cells that did not pass through the membrane pores were removed by cotton swabs. After fixation, the cells on the lower surface were stained with 0.5% crystal violet (C805211, Macklin) for 45 min. The migrating cells were photographed using a bright-field microscope, and five random fields were counted.

Tube formation assay was performed to detect the angiogenic activity of EPCs in vitro. Briefly, EPCs were re-suspended with the complete EGM-2 medium after the required treatment. They were then seeded onto a 24-well plate coated with Matrigel (356,234, Corning) at a density of 4 × 10⁴/ml and incubated for 8 h at 37°C in 5% CO₂. The tube formation ability of EPCs was examined under a light microscope (OLYMPUS, Japan), and the length of tube formation was quantified by Image J software.

DR patients were diagnosed according to the American Diabetes Association’s diagnostic criteria and fundus photography results. Patients with cancer, systemic diseases, and other ocular diseases were excluded. Non-diabetic cataract patients or non-diabetic patients with macular hole were included as the control group. Whole blood (5 ml) was collected from each patient. The blood was transferred to an anticoagulant tube and centrifuged at 3,000 × g for 10 min to collect the plasma fraction and cellular fraction. Aqueous humor samples were harvested from patients with DR and cataract during surgeries according to the Declaration of Helsinki. About 200 μl of aqueous humor was collected from each patient. Vitreous samples were collected from patients undergoing pars plana vitrectomy for the treatment of DR and macular hole for non-diabetic patients. Vitreous samples were collected, placed immediately on ice, centrifuged for 15 min to remove insoluble materials, and stored at 80°C until use. For miRNA detection, about 1 ml of plasma, 200 μl of aqueous humor, or 300 μl of vitreous samples were used for RNA isolation by the TRIzol^® Reagent (Invitrogen) according to the manufacturer’s instructions. First, the samples were centrifuged at 12,000 × g for 15 min at 4°C to obtain clear supernatant. Then, equal volumes of TRIzol reagents and glycogen were added to increase the extraction specificity. Subsequently, chloroform was added and incubated for 5 min. The mixture was centrifuged at 12,000 × g for 15 min at 4°C, and the upper aqueous phase was collected and mixed with ethanol. Finally, the supernatant was removed. The remaining RNA pellet was washed and dissolved in DEPC water.

All statistical analyses were performed using GraphPad Prism version 8.0. The difference was determined by Student t test (two-group comparison) or one-way ANOVA followed by the post hoc Bonferroni test (multi-group comparison). p < 0.05 was considered significant.

To identify the EPCs, double-positive staining, immunofluorescent staining, and flow cytometry were conducted. Double staining with the functional markers, UEA-l and Dil-Ac-LDL, indicated that the isolated cells were EPCs (Figure 1A). To further characterize EPCs, the expression of CD133, CD34, and VEGFR2 was examined using immunofluorescent staining and flow cytometry. As shown in Figure 1B, the majority of the isolated EPCs expressed surface markers, including CD133, CD34, and VEGFR2. Flow cytometry assays showed 98.92% of the isolated cells were CD133-positive, 98.64% of the isolated cells were CD34-positive, and 93.46% of the isolated cells were VEGFR2-positive, suggesting that the majority of the isolated cells were EPCs (Figure 1C).

To investigate the potential functions of miRNAs in EPCs, we investigated the miRNA expression profiles of EPCs derived from the bone marrow with or without high glucose stress. A box plot demonstrated that the distributions of miRNA expression datasets across different samples are not significantly different after normalization (Figure 2A), indicating that miRNA expression change was not random. A volcano plot was conducted to identify the differentially expressed miRNAs between the paired two groups (Figure 2B). A total of 72 miRNAs were differentially expressed with fold change greater than 2 and p < 0.05, including 53 up-regulated and 19 down-regulated miRNAs between normal EPCs and high glucose-treated EPCs (Supplementary Table S2). To demonstrate the expression patterns of significant miRNAs, we built up a hierarchically clustered heatmap using the top 10 up-regulated miRNAs and the top 10 down-regulated. The differentially expressed miRNAs were segregated into two groups (Figure 2C).

To validate the results of small RNA sequencing, qRT-PCRs were conducted to detect the expression pattern of 20 differentially expressed miRNAs, including 10 up-regulated miRNAs and 10 down-regulated miRNAs. We found a general consistency between qRT-PCR results and small RNA sequencing results (Table 1). Eight of the 10 up-regulated miRNAs and 8 of the 10 down-regulated miRNAs exhibited similar expression patterns as a result of small RNA sequencing.

To further probe into the functions of these differentially expressed miRNAs, GO enrichment and KEGG pathway analyses were conducted. GO analysis based on the putative target genes of miRNAs was divided into three subgroups: biological process, molecular function, and cellular component (Figures 3A–C). In terms of the biological process, the most significantly enriched GO term was “cell differentiation” and the most enriched GO term in the molecular function was “protein binding.” As to the cellular component, GO term was tightly associated with “cytoskeleton.” Pathway analysis was conducted to map the target genes of differentially expressed miRNAs to KEGG pathways, which displayed the top six signaling pathways. Among them, cGMP-PKG signaling was the most enriched signaling pathway (Figure 3D).

miR-375–3p was shown to be significantly up-regulated in EPCs following high-glucose stress. Moreover, rno-miR-375–3p has a homologous gene in human genome with 100% similarity of gene sequence, hsa-miR-375–3p (Supplementary Figure S1). Transfection of miR-375–3p mimics led to increased levels of miR-375–3p expression in EPCs (Figure 4A). MTT assays revealed that transfection of miR-375–3p mimics led to decreased viability of EPCs (Figure 4B). CCK-8 assays showed that miR-375–3p mimics significantly decreased the proliferation ability of EPCs (Figure 4C). Flow cytometry assays showed that transfection of miR-375–3p mimics accelerated high glucose-induced apoptosis of EPCs as shown by the increased number of apoptotic cells (Figure 4D). By contrast, transfection of miR-375–3p inhibitor led to increased cell viability, increased proliferation ability, and reduced apoptotic number of EPCs (Figures 4B–D). The migration ability of EPCs was then determined using the Transwell assay. Transfection of miR-375–3p mimics led to decreased migration ability of EPCs, whereas transfection of miR-375–3p inhibitor led to increased migration ability (Figures 4E,F). We also conducted tube formation assay to determine the role of miR-375–3p on the tube formation ability of EPCs. The results showed that transfection of miR-375–3p mimics significantly reduced the number of branch points and tube length. By contrast, transfection of miR-375–3p inhibitor contributed to tube formation of EPCs (Figures 4G,H). We also established the hypoxic stress model in vitro by exposing EPCs to CoCl₂ and exploring the role of miR-375–3p in EPC biology under hypoxic stress. Similarly, we also revealed that miR-375–3p inhibitor increased the viability and proliferation ability of EPCs, decreased CoCl₂-induced cell apoptosis, and increased the migration ability and angiogenic ability of EPCs (Supplementary Figure S2).

To explore the diagnostic value of miR-375–3p, we collected peripheral blood and aqueous humor from patients with DR and patients with cataract to detect the level of miR-375–3p expression. The results showed that the level of miR-375–3p expression was significantly up-regulated in the plasma fraction and cellular fraction of peripheral blood in DR patients (Figures 5A,B). Moreover, the level of miR-375–3p expression was significantly up-regulated in the aqueous humor of DR patients (Figure 5C). We also compared the expression pattern of miR-375–3p between non-diabetic patients with macular holes and patients with DR. qRT-PCRs showed that the level of miR-375–3p expression was significantly up-regulated in the vitreous samples of DR patients (Figure 5D). Collectively, these results suggest that miR-375–3p is a potential biomarker for the diagnosis of DR.