Inhibition of Axl with a pharmacological inhibitor suppressed retinal angiogenesis in a mouse model of oxygen-induced retinopathy (6). To study if this finding was related to a clinical significance, we investigated if Gas6 was enriched in vitreous from patients with PDR by ELISA. In this assay, vitreous from the macular hole was used as a control. As shown in Figure 1A, in PDR vitreous, Gas6 was elevated compared with that in the control vitreous. Next, we evaluated if Axl was activated in fibrovascular membranes from patients with PDR by immunohistochemistry. The results (Figures 1B–E) showed that Axl was indeed activated in the PDR membranes, suggesting that Gas6 in the PDR vitreous is one of the possible agents inducing Axl activation in the PDR membranes.

We next assessed if Axl activation could be induced by PDR vitreous in cultured HRECs. As expected, activation of Axl and Akt was heightened in HRECs after stimulation by PDR vitreous whose effect on induction was better than normal vitreous (Figure 2A). These signaling events might transform to angiogenesis-related cellular responses such as proliferation, migration, and tube formation (7–10).

Data from proliferation assays with HRECs showed that there was a significant increase in both the total cell number and Ki-67 (a nuclear protein that is associated with proliferation) positive cell number when these cells were treated with PDR vitreous (Figures 2B,C), indicating that PDR vitreous enhances cell proliferation.

Subsequently, we employed a wound healing assay to assess the migratory ability of HRECs induced by PDR vitreous. The results (Figures 2D,E) showed that PDR vitreous promoted HRECs migration, which is an important feature in angiogenesis.

Subsequently, we evaluated the effect of PDR vitreous on angiogenesis in an in-vitro model of a tube formation assay (6). As shown in Figures 2F,G, PDR vitreous heightened the ability of HRECs in forming tubes, suggesting that the signaling pathway of Gas6/Axl might play an important role in these PDR vitreous-enhanced cellular events intrinsic to angiogenesis.

To investigate the role of Axl in PDR vitreous-induced angiogenesis in vitro, expression of Axl was depleted in HRECs by using a CRISPR/Cas9 approach. As shown in Figure 3A, western blot analysis demonstrated that Axl depletion resulted from SpCas9 editing of genomic DNA under the single guide RNA-1 (sgRNA-1) guidance. In addition, we found that depletion of Axl attenuated PDR vitreous-induced cell proliferation (Figure 3B) and decreased the migratory capability of HRECs (Figures 3C,D). Furthermore, Axl removal resulted in a reduction in PDR vitreous-induced tube formation of the HRECs (Figures 3E,F). These results demonstrate that Axl plays a central part in PDR vitreous-stimulated activation of Akt and angiogenesis-related cellular responses of HRECs.

We next sought to find if pharmacological inhibition of Axl could block PDR vitreous-induced Akt activation, so that a potential pharmacological inhibitor could be used for treating retinal pathological angiogenesis. R428, a small molecule inhibitor specific for Axl, inhibited Axl at 4 μM without showing obvious toxicity to HRECs. Therefore, we treated HRECs with PDR vitreous along with R428 and found that 4 μM R428 completely blocked PDR vitreous-induced Akt activation (Figure 4A). As expected, R428 at this concentration also inhibited vitreous-stimulated proliferation, migration, and tube formation of HRECs (Figures 4B–G), suggesting that Axl is a potential mediator of retinal angiogenesis.

The Internal Review Committee of Xiangya Hospital approved this study. Research protocols adhered to the Association for Research in Vision and Ophthalmology Statement on Human Subjects and the tenets of the Declaration of Helsinki. All the participants gave a written informed consent prior to surgery and inclusion in this study. Surgical samples were collected at the Department of Xiangya Hospital. Most of the antibodies and reagents in this study were listed in our previous study (6).

Human retinal microvascular endothelial cells (Cell Systems, Kirkland, Washington, USA) were grown in an endothelial growth medium (EGM)-2 (Lonza, Walkersville, Maryland, USA) supplemented with a kit of growth factors (Lonza, Walkersville, Maryland, USA). Human embryonic kidney (HEK) 293T cells were cultured in high-glucose (4.5 g/l) Dulbecco's Modified Eagle Medium supplemented with 10% Foetal Bovine Serum. All the cells were cultured at 37°C in a humidified 5% CO₂ atmosphere (12).

The transfer vector of lentivirus was constructed by replacing different sgRNA in BsmBI-digested lentiCRISPR v2 vector as previously described (6). The three protospacer sequences from Axl loci (NC_000004.12) were 5′-AAGGTTCCTTCACTATCAGG-3′ (A1), 5′-GGGAATATCACAGGTGCCCG-3′ (A2), and 5′-CTTCTACCGGGAAACTGACT-3′ (A3) and the control sgRNA sequence targeting LacZ from Escherichia coli was (5′-TGCGAATACGCCCACGCGATGGG-3′) (13). All the clones were confirmed by DNA sequencing using a primer 5′-GGACTATCATATGCTTACCG-3′ from the sequence of U6 promoter that drives expression of sgRNAs. Both the synthesis of primers and oligos and sequencing of PCR products and clones were performed at Sangon Biotech (Shanghai, China). All the plasmids used were purchased from Addgene (Cambridge, Massachusetts, USA).

The procedure for lentivirus production was described in detail in our previous publication (14). Lentiviruses were produced by triple transfection of HEK 293T cells with pLentiCRISPRv2, the lentiviral packaging plasmid which encod HIV-1 Gag, Pol, Tat and Rev proteins, and pVSV-G [catolog number 52,961, 12,260, and 8,454 from Addgene (Cambridge, Massachusetts, USA), respectively] using lipofectamine 3000. After harvest, viruses were concentrated by centrifuging in a JA17 rotor (Beckman Coulter, Brea, California) at 25,000 g for 90 min at 4°C. The concentrated virus was resuspended in 300 μl of sterile TNE (50 mM Tris, Ph 7.8, 130 mM sodium chloride, and 1 mM Ethylenediaminetetraacetic acid) with gentle rotation overnight at 4°C. Next, these dissolved retroviruses were tittered for infecting HRECs in combination with 8 μg/ml polybrene (Sigma-Aldrich Corporation, St Louis, Mosby, USA) or kept at −80°C (12, 15, 16). The infected cells were selected in media with puromycin (Sigma-Aldrich Corporation, St Louis, Mosby, USA) (0.5 μg/ml) and the resultant cells were examined by Western blot analysis (12, 15, 16).

Human retinal microvascular endothelial cells were seeded in a 24-well plate at 70% confluence and then starved for 6–8 h in media deprived from serum and growth factors. Subsequently, the starved cells were pretreated with PDR vitreous (200 μl/ml) for 30 min and the control group was pretreated with boiled PDR vitreous. In certain experiments, cells were treated with different inhibitors for 30 min. All the cells were lysed in Radioimmunoprecipitation assay in the presence of a protease inhibitor (MedChemExpress shanghai, China). The protein-transferred membranes were blocked by 5% fat-free milk dissolved in Tris Buffered Saline/0.05% Tween-20 for 30 min and incubated with first antibodies (1:1,000) overnight at 4°C. After thoroughly washing, the membranes were incubated with horse radish peroxidase-conjugated secondary antibodies (1:5,000) and visualized using LumiBest enhanced chemiluminescence. β-actin and total Akt were served for normalization across the samples. Experiments were repeated at least three times and blot signal intensity was determined by densitometry using the National Institutes of Health ImageJ software (12).

Human retinal microvascular endothelial cells were seeded at a density of 2 × 10⁴ cells/well in 24-well plates and counted in a cell number counter after 48 h of continuous treatment with endothelial cell growth media-2 or PDR vitreous (1:3 dilution in EGM-2). At least three independent experiments were performed as described previously (6). In addition, immunofluorescent staining for Ki67 (rabbit anti-Ki67; Proteintech, Rosemont, IL US) for marking proliferating cells was performed as previously described (17).

Migration was assessed with a scratch-wound assay (18). Once cells reached 80% confluence in 24-well plates, they were starved for 4 h. After the cell monolayer was scraped with a sterile pipette tip (200 μl), the cells were washed twice to remove detached cells. One scratch was generated per well and imaged on an Leica imaging system every 6 h for 48 h. Images were analyzed by measuring the number of pixels in the wound area using Adobe Photoshop (Adobe Systems, San Jose, California, USA) and analyzed using the ImageJ software (3).

This assay was performed as previously described (19). A total of 15,000 HRECs were placed onto wells precoated with basement membrane extract (R&D Systems, Minneapolis, MN, US), which was from the storage at −80°C and thawed overnight at ice. Cells were imaged 4–6 h after cell plating and tube formation was quantified for the total length using the Angiogenesis Analyzer plugin for the ImageJ software (National Institutes of Health) (20).

Results are presented as mean ± SD. The Student's t-test was performed for comparisons between two groups and the one-way ANOVA (Kruskal–Wallis test) was used for comparisons among the multiple groups. p < 0.05 was considered as statistically significant.