Mouse retinal vascular endothelial cells were purchased from PriCells Biomedical Technology Co., Ltd. (Wuhan, China), and were cultured in DMEM supplemented with 10% heat-inactivated fetal bovine serum, 100 U/ml penicillin, and 100 mg/l streptomycin (all from Gibco Life Technologies, Carlsbad, CA, USA) under conditions with 5% CO₂, at 37°C, in a humidified incubator.

Cells in the exponential growth phase were seeded in 6-well plates (Corning Incorporated, Corning, NY, USA) at a density of 5×10⁴ cells/ml. The cells were transfected with miR-218 mimic, miR-218 inhibitor, and the negative control (all from GenePharma, Shanghai, China) using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) at a 5:1 volume/mass ratio of reagent to oligodeoxynucleotide in serum-free M199 for 6 h. After transfection, the cells were incubated in complete DMEM. Transfection of the miR-218 mimics, miR-218 inhibitor and the negative control was performed using 1.0 μg of DNA per transfection. All transfections were performed according to the manufacturer's instructions.

Robo1 siRNA (sense, 5′-CGGGAAAGGCCG AGGAACAAAGGCAGC-3′ and antisense, 5′-GCUGCCUUU GUUCCUCGGCCUUUCCCG-3′) was synthesized by Guangzhou RiboBio Co., Ltd. (Guangzhou, China).

miR-218 was amplified using the following primers: miR-218 forward, 5′-TTCTGAGGATCCGTGGAGGCACCTTTTCCATA-3′ and reverse, 5′-ATTCTAAGATCTTTCACAGCTAGTCACACAATGG-3′. Plasmid vector pCDH-CMV-miR-218 was constructed by inserting a pri-miR-218 PCR fragment into the plasmid vector pCDH-CMV-MCS-EF1-Puro (Spiral Biotech, Inc., Norwood, MA, USA) through EcoRI and NotI digestion. The pCDH-CMV-mock vector was used as the negative control.

Total RNA was extracted from cells and retinal tissues using TRIzol reagent (Invitrogen). cDNA was synthesized, according to the manufacturer's instructions, with a cDNA synthesis kit (Takara Bio, Inc., Otsu, Japan). qPCR was undertaken using SYBR-Green.

The primer sequences were as follows: miR-218, 5′-TTG TGCTTGATCTAACCATGT-3′; miR-218-1, 5′-CCATGGAA CGTCACGCAGC-3′; miR-218-2, 5′-GCGGAAAGCACCGTGCTC-3′; Slit2 forward, 5′-GCGCGTCTGGTGTGAAT GAA-3′ and reverse, 5′-CACAGTGGCACCAGGAGCAT-3′; Slit3 forward, 5′-TGGAAATACGCCTAGAACAG-3′ and reverse, 5′-ACCAGCGACGTGAGTGAT-3′; Robo1 forward, 5′-GAGGTAGCTATACTACGGGATGAC-3′ and reverse, 5′-CAGATGTAGTAGCCGACATCAGAC-3′; U6 forward, 5′-CGCTTCGGCAGCACATATAC-3′ and reverse, 5′-AAAATATGGAACGCTTCACGA-3′.

Cell migration ability was assessed using a scratch wound assay. Transfected cells were cultured in 6-well plates. When cells reached 90% confluence, a scratch wound was created using a pipette tip. Wound edges were photographed with a Nikon Eclipse TE20000-U (Nikon, Tokyo, Japan) and scratch widths were analyzed using ImageJ software (NIH). Each assay was completed in triplicate.

Neonatal CB57BL/6J mice were obtained from the Animal Institute of Chinese Academy of Medical Sciences. Neovascularization was induced as described by Kong et al (27). Briefly, at postnatal day 7 (P7), CB57BL/6J mice were exposed to hypoxic conditions (75% O₂) for 5 days (P12) and then returned to room air in order to induce RNV.

The present study was approved by the Ethics Committee of Tianjin Eye Hospital (Heping, China). All animal experiments were conducted in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and the Guide for the Care and Use of Laboratory Animals.

Plasmids were packed using Lipof ectamine 2000 (Invitrogen) in accordance with the manufacturer's instructions. Twelve-day-old mice were anesthetized by intraperitoneal injection of ketamine hydrochloride (30 mg/kg). Each animal received intravitreous injections of 0.4 μl (2 μg) pCDH-CMV-218 in one eye and 0.4 μl (2 μg) control plasmid [plasmid vector pCDH-CMV-MCS-EF1-Puro (Spiral Biotech, Inc.) without an insertion of miR-218] in the contra-lateral eye as the negative control. Intravitreal injections were performed at 11:00 a.m., and were administered to 1-mm posterior to the limbus of the eye using a 10-μl Hamilton syringe fitted with a 32G needle.

Total proteins of cells and retinas were isolated and separated on 10% SDS-PAGE gels (Bio-Rad Laboratories, Inc., Hercules, CA, USA). Western blotting was performed according to standard protocols. α-tubulin was used as a loading control. The antibodies anti-Robo1 (ab7279), anti-Slit2 (ab7665), anti-Slit3 (ab11018) were from Abcam, and anti-α-tubulin (T5168) was from Sigma-Aldrich (St. Louis, MO, USA). Bands were quantified using ImageJ software (NIH).

On P17, mice (n=6/group) were anesthetized and perfused with fluorescein via retro-orbital injection of 2.5 mg/50 μl of FITC-dextran (Sigma-Aldrich), as previously described (28). Eyes were enucleated and fixed with 4% paraformaldehyde in PBS for 1 h. Retinas were then separated from the eyecup. Four incisions were made, and the retina was flat-mounted on a gelatin-coated slide. The vasculature was then examined under a fluorescence microscope (Nikon Eclipse TE20000-U). Images were analyzed using Photoshop 8.0 software (Adobe Systems, San Jose, CA, USA). Neovascularization was calculated as a ratio: the number of pixels in neovascular area to the total number of pixels in the retina.

At p17, the eyes of the mice (n=6/group) were enucleated and fixed with 10% formaldehyde and embedded in paraffin. Sagittal sections (6-μm-thick) were made through the cornea parallel to the optic nerve, and then they were stained with hematoxylin and eosin (H&E). The nuclei of vascular cells on the vitreal side of the retina were counted under a light microscope. Ten non-continuous sections from each eye were examined, and numbers of cells were averaged in each group. The average number of pre-retinal vascular nuclei was compared.

Data are expressed as the means ± standard deviation (SD) and were analyzed using SPSS 11.5 (SPSS, Inc., Chicago, IL, USA). To compare multiple sets of data, one-way analysis of variance (ANOVA) was used. For paired data sets, the LSD t-test was used. Bivariate correlations were calculated by Spearman's rank correlation coefficients. A P-value <0.05 was considered to indicate a statistically significant difference.

The cell migration ability of each group is shown in Fig. 1: overexpression of miR-218 caused by miR-218 mimic significantly reduced endothelial cell (EC) migration, whereas inhibition of miR-218 expression using the miR-218 inhibitor markedly promoted EC migration (Fig. 1A and B).

Robo1 expression of each group is shown in Fig. 1C. The results indicate that upregulation of miR-218 expression not only reduced EC migration but also decreased Robo1 expression. Conversely, downregulation of miR-218 expression using the miR-218 inhibitor increased Robo1 expression and, as was noted previously, promoted EC mobility. However, as shown in Fig. 1A, miR-218 inhibitor did not promote the migratory ability of ECs after Robo1 knockdown by siRobo1. These observations suggest that miR-218 suppresses EC migration by inhibiting Robo1 expression.

The expression level of miR-218 in the retinas of OIR mice was detected. As shown in Fig. 2A, RT-qPCR results demonstrated that the expression level of miR-218 was significantly decreased (P<0.05) in retinas of mice with OIR at P17. We then compared mRNA and protein expression levels of Robo1 in retinas of mice with OIR with those of the control mice. The mRNA and protein levels of Robo1 were markedly upregulated at P17 in mice with OIR (Fig. 2).

To determine which miR-218 coding gene is involved in downregulation of miR-218 in mice with OIR, we evaluated miR-218-1, miR-218-2, miR-218, Slit2 mRNA and Slit3 mRNA expression in retinas of mice with OIR by RT-qPCR. As shown in Fig. 3B–E, statistical analysis revealed that a positive correlation existed between miR-218-1 and Slit2, miR-218-2 and Slit3, as well as miR-218-1 and miR-218. There was no correlation, however, between miR-218-2 and miR-218. This meant that in the retinas of the mice with OIR, miR-218-1 and miR-218-2 were co-transcribed with their host gene respectively and that downregulation of miR-218 was caused by decreased expression of miR-218-1, not miR-218-2. When mRNA levels of Slit2 and Slit3 in retinas from control mice and mice with OIR were compared, mRNA expression of Slit2 in retinas of mice with OIR was 0.48-fold less (P<0.05) than in the retinas of the normal control mice; mRNA expression of Slit3 in retinas from mice with OIR was 1.12-fold greater than in the control mice although this was not statistically significant (P>0.05) (Fig. 3A). Similar results were also observed in protein expression levels of Slit2 and Slit3 (Fig. 3F and G). The results suggest that decreased expression of miR-218-1 and its host gene Slit2 is involved in RNV in OIR.

A plasmid pCDH-CMV-miR-218 expressing miR-218 and a negative control plasmid pCDH-CMV-mock expressing mock sequence were constructed. Intravitreal injection of plasmids was performed at P12.

At P17, we analyzed mRNA levels of miR-218 and Robo1 after intravitreal injection to verify whether plasmids reached the retina. RT-qPCR results showed that pCDH-CMV-miR-218 significantly increased the expression of miR-218 and reduced that of Robo1 in retinas (Fig. 4A). Western blotting indicated that intravitreal injection of pCDH-CMV-miR-218 significantly decreased the protein level of Robo1 in the retinas of mice with OIR (Fig. 4B). However, pCDH-CMV-mock had almost no effect on the expression of miR-218 and Robo1.

To examine the effects of miR-218 on RNV, we evaluated the retinal vasculature at P17 by fluorescein angiography in the flat-mounted retinas. In retinas of normal mice we noted a mature capillary network that extended from the optic to the periphery (Fig. 5A1). The retinas of mice in the OIR and pCDH-CMV-mock groups displayed more neovascular tufts (Fig. 5A2 and A3). By contrast, the retinas of the pCDH-CMV-miR-218 group exhibited lesser neovascularization (Fig. A4). The RNV was quantified by measuring areas of new blood tufts in the whole mounted retina. The results showed that a large number of neovascular tufts appeared in the OIR group [neovascularization (NV), 38.9% of whole retina] and pCDH-CMV-mock group (NV, 36.8% of whole retina). However, compared to the OIR group, the neovascularized area was significantly decreased in the pCDH-CMV-miR-218 group (NV, 18.6% of whole retina) (Fig. 5B). No neovascularization was observed in the normal group. These results revealed that miR-218 exerts an anti-neovascularizing effect on RNV in mice with OIR.

To further study the inhibitory effect of miR-218 on angiogenesis, histological analysis was performed. We counted the vascular cell nuclei which broke through the inner limiting membrane (ILM), a marked feature of OIR (29). There were no neovascular nuclei in the normal group (Fig. 6A1), and the average number of vascular cell nuclei was significantly increased in the OIR group and pCDH-CMV-mock group, 61.48±6.92 and 58.98±6.48, respectively (Fig. 6A2 and A3). However, in the pCDH-CMV-miR-218 group we noted less neovascularization (Fig. 6A4) (12.64±1.42 pre-retinal cells/section), indicating that restoration of miR-218 played an inhibitory role in RNV in OIR (Fig. 6B).

Taken together, these results indicate that miR-218 inhibited retinal angiogenesis in OIR by suppressing Robo1 expression.