Long Noncoding RNA PPT2-EGFL8 Regulates Pathological Retinal Neovascularization in PDR by Functioning as a Competing Endogenous RNA

Diabetic retinopathy (DR) is the most common microvascular complication of the population with diabetes and remains the leading cause of blindness among working-age adults in the developed world (1). DR is classified as nonproliferative DR (NPDR) or proliferative (PDR). Although sustained hyperglycemia is the major stimulus for NPDR development, retinal ischemia is a prerequisite for PDR and leads to increased angiogenic factor levels that accelerate retinal neovascularization (NV). If left untreated, retinal NV can result in retinal detachment, vitreous hemorrhage, glaucoma, and even irreversible loss of vision (2,3). In the last 4 decades, panretinal photocoagulation, a process in which the peripheral ischemic retina is burned with a laser to protect the patient’s central vision, has been the standard of care for PDR (4). Although effective in protecting central vision, panretinal photocoagulation leads to reduced peripheral and night vision. Moreover, PDR can progress in patients despite appropriate treatment. This issue emphasizes the importance of understanding the underlying mechanism(s) of retinal NV to identify therapeutic approaches for the prevention or treatment of PDR.

Long noncoding RNAs (lncRNAs) are commonly defined as transcribed RNAs of >200 nucleotides with no coding potential and are involved in numerous important biological processes (5,6). To date, only a limited number of lncRNAs have been well characterized, with a diverse array of mechanisms identified, including roles as guides, decoys, scaffolds, and signaling molecules (5,7). Recently, a class of lncRNAs, referred to as competing endogenous RNAs (ceRNAs), has been characterized (8). ceRNAs protect mRNAs by acting as molecular sponges for miRNAs that specifically repress target mRNAs. For instance, lncRNA MALAT1 may adsorb miR-320a or miR-150-5p as a molecular sponge and therefore modulate pathological angiogenesis in DR (9,10). Studies have demonstrated that many of these molecules target hypoxia-inducible factor/vascular endothelial growth factor (VEGF) signaling; however, not all patients respond well to anti-VEGF therapy. Moreover, for many patients who respond to anti-VEGF therapy, the response may only be temporary. The number of patients with diabetes with NPDR who continue to progress to PDR increases over time (11,12). This result suggests that other retinal angiogenic factors may also participate in the development of PDR in DR. We sought to explore the role of ischemia-driven lncRNAs and proangiogenic factors that directly contribute to retinal NV in patients with PDR.

In this study, we describe the function of a retinal angiogenesis-associated lncRNA (lnc-PPT2-EGFL8). PPT2-EGFL8 expression is downregulated during the pathogenesis of retinal NV and is required in ischemia-induced endothelial cell proliferation. Our study further reveals that PPT2-EGFL8 promotes retinal angiogenesis by absorbing miR-423-5p as a molecular sponge, which subsequently controls the activation of peroxisome proliferator-activated receptor-β/-δ (PPARD)/angiopoietin-like 4 (ANGPTL4) signaling, a critical signaling pathway involved in ischemic angiogenesis.

All human studies were conducted according to the Declaration of Helsinki principles and were approved by the Human Research Ethics Committee of the Affiliated Wuxi People's Hospital of Nanjing Medical University. All participants were age- and sex-matched and signed a consent form after reading information about the study. Serum samples were drawn from healthy individuals or individuals with diabetes mellitus without DR (DM-NDR), NPDR, or PDR immediately prior to intervention.

Extracellular vesicles (EVs) were isolated from serum. Total RNA in EVs was extracted using the Total Exosome RNA and Protein Isolation Kit (Invitrogen by Life Technologies). The amount and quality of lncRNA, mRNA, and small RNA in the total RNA was tested by RiboBio Co., Ltd. (Guangzhou, China). RNA library construction and sequencing were performed by RiboBio Co., Ltd. Next, the cDNA library of lncRNAs and mRNAs was sequenced on an Illumina Xten system (PE150). Then, the cDNA library of small RNA was sequenced on an Illumina HiSEq. 2500 (SE50). Raw reads were collected using related Illumina analysis software.

Male mice (8 weeks old) were purchased from the laboratory animal center of the Academy of Military Medical Sciences (Beijing, China), housed under standard conditions (22.5°C and 42.5% humidity, under a 12-h/12-h light–dark cycle, using heated wood chip litter as bedding material) in the specific pathogen free animal center of Wuxi People’s Hospital Affiliated with Nanjing Medical University, and permitted ad libitum consumption of water. Animals were ventilated after being anesthetized with a mixture of ketamine and xylazine, and the effectiveness of the anesthesia was monitored by observing slow breathing, loss of muscular tone, and no response to surgical manipulation. The retinas were then harvested for subsequent analyses. All studies were conducted in accordance with the Guide for the Care and Use of Laboratory Animals (National Institutes of Health, Bethesda, MD) and the Association for Research in Vision and Ophthalmology (ARVO) Statement for the Use of Animals in Ophthalmic and Vision Research. All animal experiments fulfilled the Nanjing Medical University requirements for humane animal care.

Intravitreal adeno-associated virus (AAV) injection was administered, under anesthesia, into the right eye after pupil dilation. A sclerotomy was created using a sharp 30-gauge needle tip at ∼0.5–1 mm posterior to the limbus, and 1–2 μL of saline containing 5 × 10¹² vector genomes (vg) of AAV2-ZsGreen-shRNA (AAV2-Null), AAV2-shPpt2-IRES-ZsGreen (AAV2-shPpt2), or AAV2-shEgfl8-IRES-ZsGreen (AAV2-shEgfl8) vectors supplied by Viral Therapy Technologies (Wuhan, China) was injected intravitreally using a 10-μL Hamilton syringe (701 RN; Hamilton Company, Reno, NV) attached to a 33-gauge blunt-end needle.

Diabetes was induced with an i.p. injection of streptozotocin (STZ) (55 mg/kg in 10 mmol/L citrate buffer at pH 4.5) for 5 consecutive days. The blood glucose levels were >300 mg/dL, indicating that diabetes had been established successfully. The STZ-treated mice were anesthetized with ketamine/xylazine, and an intravitreal injection of approximately 1 μl of the designated mixture was administered into the eyes using a 33-gauge needle.

In the neonatal mouse model of oxygen-induced retinopathy (OIR), 7-day-old mice were placed with their nursing dams in a 75% oxygen atmosphere for 5 days. Mice were subsequently returned to normal room air from postnatal day (P)12 to P17. At P3, randomly selected mice were intravitreally injected under anesthesia with 1 μL of saline containing the designated mixture using a 33-gauge needle under a surgical microscope. Eyes were dissected under deep anesthesia at P17 to assess the effect of OIR.

Human retinal microvascular endothelial cells (hRMECs) that were >90% positive for acetylated LDL were obtained from BeNa Culture Collection (Beina Chuanglian Biotechnology Institute, Beijing, China) and cultured in DMEM with 10% FBS (v/v) and 1% antimycotics and antibiotics (v/v). Mycoplasma contamination was not detected.

A Cy3-conjugated PPT2-EGFL8 probe was prepared from asymmetric PCR amplification, and the PCR products were gel purified (GenePharma, Shanghai, P.R. China). Briefly, hRMECs fixed with 4% paraformaldehyde (PFA) were dehydrated with 70–100% ethanol, air dried, and incubated at 37°C for 2 h with a denatured probe in formamide-containing buffer. The cells were then washed in hybridization buffer, followed by saline sodium citrate buffer, and mounted using Vectashield. The hybridized probes were visualized by fluorescein-Cy3–incorporated probe signals under a fluorescence microscope (Leica, Heidelberg, Germany).

siRNAs were designed and synthesized by RiboBio. The targeting sequences of siRNAs are listed in Supplementary Table 1. miRNA mimics and the matched negative controls were synthesized and purified by GenePharma (Shanghai, China), and the sequences are listed in Supplementary Table 2. RNA oligonucleotides were transfected by using Lipofectamine 2000 (Invitrogen, Carlsbad, CA), and the medium was replaced 6 h after transfection.

Tubule formation assays were performed using growth factor-reduced Matrigel (BD Biosciences; 356231). Matrigel (60–80 μL) was added to a 96-well plate and placed in a 37°C CO₂ incubator for 30 min. The hRMECs were counted and seeded at 2 × 10⁴ cells per well on the Matrigel into a 96-well plate. Images were captured 10 h later and analyzed using ImageJ software (National Institutes of Health).

Animals were anesthetized, and Cyclomydril (Alcon, Fort Worth, TX) was used to dilate their pupils. Spectral domain optical coherence tomography (OCT) images were obtained using the Micron IV image-guided OCT system (Phoenix Research Laboratories, Pleasanton, CA).

Mice were administered Evans blue (100 mg/kg) via the femoral vein and kept on a warm pad for 60 min. The eyes were collected and fixed in 4% PFA (w/v) for 2 h. The retinal flat mounts were acquired, mounted on glass slides, and examined under an Olympus BX-51 light microscope (Olympus, Tokyo, Japan).

Eyes were enucleated and fixed in 4% PFA (w/v) for 2 h at room temperature. Retinas were collected, equatorially bisected, and incubated with 3% trypsin (w/v) at 37°C for 2 h. They were then gently shaken to free the vessel network, washed, and mounted onto glass slides to dry. The retinal vasculature was stained with periodic acid–Schiff and hematoxylin. Digital images were examined using an Olympus BX-51 light microscope.

Mice were anesthetized, and a 14-gauge perfusion cannula was introduced into the left ventricle. Drainage was observed in the right atrium. The animals were perfused with PBS for 2 min. After PBS perfusion, fixation with 1% PFA (w/v) and 0.5% glutaraldehyde (w/v) was achieved over ∼3 min. Nonspecific binding was blocked with 1% albumin (w/v) in 5 mL of PBS, followed by perfusion with FITC-coupled concanavalin A lectin for 5 min (5 mg/kg body wt; Vector Laboratories, Burlingame, CA). Residual unbound lectin was removed with 1% albumin (w/v) in PBS perfusion for 1 min, followed by PBS perfusion for 4 min. The retinas were carefully removed, and flat mounts were prepared and imaged using a confocal microscope (Leica).

Standard immunofluorescence analysis was performed to indicate glial fibrillary acidic protein (GFAP; 1:1,000, Sigma-Aldrich) and IBA1 (1:1,000, Cell Signaling) expression, followed by goat anti-mouse/rabbit IgG (H + L) cross-adsorbed secondary antibody, Alexa Fluor 555 and IB4 (1:500; Life Technologies, Carlsbad, CA) in mouse retinas. Monoclonal anti–Ki-67 antibody (1:1,000; Abcam, Cambridge, MA) was used to mark the cell cycle in hRMECs.

Animals were anesthetized, and their eyes were dissected and fixed overnight in 4% PFA (w/v). The retinas and scleras were dehydrated in different concentrations of graded ethanol and embedded in paraffin. For hematoxylin-eosin (H-E) staining, 5-µm-thick sections were taken along the vertical meridian. Digital images of H-E staining were observed under an Olympus BX-51 light microscope.

RNA immunoprecipitation was performed using an RNA-Binding Protein Immunoprecipitation Kit (cat. no. 17-700; Millipore). Chromatin extract was prepared, and the protein-RNA complex (24 μg RNA) was immunoprecipitated with Ago2 antibody (4 μg) (cat. no. 030-110; Sigma-Aldrich). Each RNA immunoprecipitate was purified and subsequently analyzed by quantitative (q) real-time PCR.

HEK293T cells were cotransfected with the indicated mimics and reporter plasmids. After incubation for 24 h, relative luciferase activities were measured using the Dual-Luciferase Reporter Assay System (E1910, Promega) according to the manufacturer’s protocol.

Protein in cell lysates was analyzed via electrophoresis using an SDS-PAGE gel and transferred to polyvinylidene fluoride membranes (Millipore, Billerica, MA). Antibodies against PPARD (1:1,000), ANGPTL4 (1:1,000), and VEGFA (1:1,000) were obtained from Abcam. β-Actin antibody was used to confirm equal protein loading in the samples. The signal on the membranes was detected using an enhanced chemiluminescence system (West Pico kit; Pierce, Loughborough, U.K.). Band density was analyzed using ImageJ software. The antibodies are detailed in Supplementary Table 3.

RNA was extracted and analyzed using qRT-PCR. The data were analyzed using the 2^–ΔΔCT method and normalized to the endogenous control GAPDH mRNA (for humans), and the amount of target gene mRNA expression in each sample was expressed relative to that of the control. Primer sequences for real-time quantitative PCR (qPCR) were designed using Primer Express Software (Thermo Fisher Scientific, Waltham, MA) (Supplementary Table 4). miRNA detection was performed using an miRNA isolation kit (Vazyme, Nanjing, China). Reverse transcription of miRNAs was performed using looped miRNA-specific reverse transcription primers (RiboBio), with U6 as a reference. The relative expression levels of miRNAs were measured using the 2^–ΔΔCT method.

The level of ANGPTL4 in the serum of the healthy individuals or the patients with DM-NDR, NPDR, or PDR was measured using commercial ELISA kits (Cusabio, Wuhan, China).

The results are expressed as the mean ± SEM. ANOVA was used for multiple group comparisons, followed by the Tukey post hoc test. The Tukey post hoc test was run only if the F value achieved P < 0.05 and there was no significant variance in homogeneity. The data were analyzed using GraphPad Prism 5 statistical software (GraphPad Software, La Jolla, CA). A P value of <0.05 was considered statistically significant.

The data that support the findings of this study are available from the corresponding author, Xiaolu Wang, upon reasonable request. Some of these resources are available within the article or its Supplementary Materials, and others are available on request.

Following the exclusion of patients who did not meet the criteria for inclusion, 87 patients were enrolled in the study (Fig. 1A). Fundus photography and fundus fluorescein angiography were used to divide 77 individuals with diabetes were divided into five groups, including DM-NDR, mild NPDR, moderate NPDR, severe NPDR, and PDR patients (Fig. 1B). There were no statistically significant differences in the demographic baseline characteristics among these groups (Table 1). We assessed the angiogenic potential of serum and vitreous fluid of healthy individuals and individuals with DM-NDR, NPDR, or PDR in stimulating hRMEC tubule formation. Fluid from the healthy individuals or the patients with DM-NDR did not have a significant effect on tubule formation. In contrast, the serum or vitreous fluid from the patients with diabetes with PDR stimulated tubule formation (Fig. 2A–C). It has been reported that lncRNAs play a critical role in DR and that their expression profiles vary based on the cellular types and conditions (13). Deep sequencing identified 13,870 lncRNAs in the serum of the healthy individuals or the patients with DM-NDR, NPDR, or PDR. Among these, 606 annotated lncRNAs were altered with a log twofold change less than −2 or >2 (Fig. 2D). To analyze and classify the differentially expressed lncRNAs, we used series test of cluster to examine the corresponding changes in these underlying functional lncRNAs. In total, 35 profiles were generated, and each profile represented a cluster of multiple lncRNAs with coincident expression patterns (data not shown). Among the 35 patterns, 4 expression patterns, namely, profiles 15, 16, 17, and 31 showed statistical significance (P < 0.05) (Fig. 2E).

The biological processes from the Gene Ontology (GO) categories for the differentially expressed genes with downregulated expression were mainly involved in the regulation of signaling transduction, metabolic processes, and cell proliferation (Supplementary Fig. 1). In addition, target genes associated with the main biological pathways were identified by Kyoto Encyclopedia of Genes and Genomes (KEGG) by enrichment analysis, axon guidance, metabolic pathways, cell cycle, glutamatergic synapse, phosphoinositide 3-kinase–Akt signaling pathway, extracellular matrix-receptor interaction, and gap junction (Supplementary Fig. 2).

We narrowed the data set based on the expression level, conservation, and function of the coexpressed genes and focused on lncRNA ENSG00000232633.3, TEN1-CDK3, and PPT2-EGFL8. These results were confirmed by real-time PCR experiments, and PPT2-EGFL8 expression was significantly downregulated in individuals with PDR (Fig. 2F). Additionally, cell fractionation analysis indicated that PPT2-EGFL8 is mostly localized to the cytoplasm, which was consistent with the results of RNA fluorescence in situ hybridization (Fig. 2G and H). Full-length PPT2-EGFL8 (2728 nt) was amplified from the 5′ and 3′ ends (Supplementary Fig. 3).

PPT2-EGFL8 expression in hRMECs was significantly higher than that in Müller cells (Supplementary Fig. 4). Moreover, PPT2-EGFL8 significantly decreased in 30 mmol/L high glucose- or 1% O₂ (hypoxia)-treated hRMECs compared with the control cells (Fig. 3A and B). We overexpressed (OE) PPT2-EGFL8 in hRMECs and examined the role of PPT2-EGFL8 in hRMEC proliferation using tubule formation assays and the expression of the cell cycle marker Ki67 (Supplementary Fig. 5A and Fig. 3C–E). Tubule formation and Ki67 expression were dramatically decreased in the hypoxia-induced hRMECs and decreased in the PPT2-EGFL8-OE hRMECs (Fig. 3C–E). Moreover, hRMECs were transfected with siRNA targeting PPT2-EGFL8 (si-PPT2-EGFL8) or a nontargeting siRNA control (si-NC) for 48 h. The siRNA expression vectors efficiently reduced the level of PPT2-EGFL8 in hRMECs (Supplementary Fig. 5B). Accordingly, hypoxia-induced tubule formation and Ki67 expression were increased in the siRNA-transfected cells (Fig. 3F–H).

The conservation of the Egfl8 transcript in mice (mm10_percent: 92.72%) was higher than that of the Ppt2 transcript (mm10_percent: 51.68%) (Fig. 3I). The mouse retinas showed rich expression of Egfl8 (Supplementary Fig. 6A). We used the mouse model of OIR, which is often used in the study of ischemic retinal angiogenesis, and found that Egfl8 mRNA expression was decreased in the OIR model (Supplementary Fig. 6B). Mice were intravitreally injected with 1 × 10¹² vg/mouse AAV2-Ppt2, AAV2-Egfl8, or AAV2-null vectors and used to develop the OIR model (Fig. 3J). H-E staining indicated increased formation of pathological vessels in the retinas of the mice with OIR and fewer pathological neovascular tufts in the retinas of the AAV2-Egfl8–treated mice (Fig. 3K and L). There was no significant difference in the AAV2-Ppt2–treated retinas (Fig. 3K and L). Hyperoxia results in central vascular obliteration or degradation, and the subsequent relative hypoxia of room air leads to retinal NV in the OIR model. We therefore evaluated the potential role of PPT2-EGFL8 in OIR-induced retinal NV in the Evans blue-labeled retinal vessels of mice. The area of retinal NV tufts in the AAV2-Egfl8–treated mice was significantly attenuated compared with that of the AAV2-null–treated OIR mice (Fig. 3M and N). There was no significant difference in the AAV2-Ppt2–treated mice (Fig. 3M and N).

We further evaluated the effect of Egfl8 transcripts on microvascular dysfunction in the STZ-induced diabetic mice. The mRNA expression of Egfl8 was decreased in the retina of the STZ model (Supplementary Fig. 6C). The mice were intravitreally injected with 1 × 10¹² vg/mouse AAV2-Egfl8 or AAV2-null vectors, and retinal blood vessel density in horizontal and vertical direction was subsequently evaluated (Supplementary Fig. 7A and B). The retinal vessel density was altered in superficial vascular plexuses in the diabetic animal models, and the effect of Egfl8 was particularly striking in the superficial retinal layer of the peripheral-area retina (Supplementary Fig. 7C and D). Importantly, apart from the decreased inhibition of diabetes-induced retinal blood vessels, retinal OCT images also showed that Egfl8-OE with AAV2 transduction markedly reversed STZ-induced microvascular dysfunction, as reflected by fluorescein leakage and the number of hyperreflective dots in the superficial portion of peripheral retinas (Fig. 4A and B; arrow). STZ-induced microvascular leakage was also inhibited in the AAV2-Egfl8–treated mice (Fig. 4C–E). Moreover, diabetes-associated pericyte loss and capillary degeneration, as indicated by a decrease in the number of pericytes and an increase in the number of acellular vessels, were significantly attenuated by the overexpression of retinal Egfl8 in the animals challenged with STZ (Fig. 4F–H).

As a major characteristic of the chronic subclinical inflammatory response in DR, leukocytosis contributes to retinal capillary dropout, capillary closure, nonperfusion, and local ischemia (14,15). Five months after diabetes was successfully established, adherent leukocytes labeled in situ with FITC-linked concanavalin A in retinal flat mounts were counted in the blood vessels. Increased adherent leukocytes were observed in the AAV2-null–treated diabetic retinal capillaries compared with the controls (Fig. 4I and J). Moreover, there was a significant decrease in the number of adherent leukocytes in the AAV2-Egfl8–treated diabetic mice compared with the AAV2-null–treated diabetic mice (Fig. 4I and J). Retinal Müller cell gliosis and microglial cell activation are frequently increased in DR (16). There was a significant increase in the expression of GFAP, a key marker of retinal Müller cells, and ionized calcium-binding adaptor molecule (Iba1) in the AAV2-vehicle– or AAV2-null–treated diabetic mice, and this effect was diminished in the mice pretreated with AAV2-Egfl8 (Fig. 4K and L).

Based on bioinformatics studies, lncRNA PPT2-EGFL8 functions as a ceRNA for miR-423-5p, miR-6827-5p, and miR-6829-5p, subsequently targeting PPARD to form a ceRNA network (Fig. 5A). To confirm the pivotal role of miRNAs in this effect, we performed an RNA immunoprecipitation assay on Ago2, which is the critical component of the RNA-induced silencing complex. In parallel, downregulation of PPT2-EGFL8 expression in hRMECs significantly led to an increase in the recruitment of Ago2 to PPARD transcripts compared with control cells (Fig. 5B). Overexpression of PPT2-EGFL8 in hRMECs promoted the increased enrichment of Ago2 on PPT2-EGFL8 but subsequently decreased enrichment on PPARD transcripts (Fig. 5C).

Subsequent bioinformatic analysis by TargetScan, miRDB, and miRWalk revealed five putative miRNA response elements (MREs) of miR-423-5p, miR-6827-5p, and miR-6829-5p shared by lncPPT2-EGFL8 and the 3′ untranslated region (UTR) of PPARD (Fig. 5D–F; upper panel). Transfection of miR-423-5p mimics in HEK293T cells significantly inhibited the luciferase activity of psiCHECK2-PPT2-EGFL8 (LNC-wild-type [WT]), which contained PPT2-EGFL8 at the 3′ UTR of Rluc, while the psiCHECK2-PPT2-EGFL8 mutant (miR-423-5p, LNC-MUT) showed no response to miR-423-5p mimics (Fig. 5D–F; lower panel). Moreover, hypoxia (1% O₂) significantly induced the expression of miR-423-5p, which was potently inhibited by PPT2-EGFL8-OE and further increased by pretreatment with si-PPT2-EGFL8 (Fig. 5G and H). However, PPT2-EGFL8 had no significant influence on miR-6827-5p and miR-6829-5p expression (Fig. 5G and H). The expression of miR-423-5p detected by next-generation sequencing was markedly increased in the serum of patients with PDR (Supplementary Fig. 8). Moreover, comparable copy numbers of lncPPT2-EGFL8 and miR-423-5p were detected per hypoxic cell (Fig. 5I). Luciferase reporter gene assays were performed. Luc-WT-PPARD-3′ UTR (PPARD-WT) cotransfected with miR-423-5p mimics in HEK293T cells resulted in a significant decrease in luciferase activity and had a minimal effect on the Luc-mutant-PPARD-3′ UTR (PPARD-MUT) group (Fig. 5J and K). Therefore, PPARD was shown to be a target of miR-423-5p.

Hypoxia (1% O₂) induced a significant decrease in the PPARD mRNA level and an increase in the ANGPTL4 and VEGFA mRNA levels (Fig. 6A and B). Overexpression of PPT2-EGFL8 significantly attenuated hypoxia-induced changes in the mRNA levels of PPARD and ANGPTL4 (Fig. 6C). Moreover, hypoxia resulted in a significant decrease in the protein expression of PPARD and an increase in the expression of cANGPTL4; the effect was diminished in the cells pretreated with PPT2-EGFL8-OE (Fig. 6D–G). There was no significant difference in hypoxia-induced increased VEGF expression after PPT2-EGFL8-OE (Fig. 6C, H, and I). In addition, hypoxia-induced decreases in PPARD expression and increases in cANGPTL4 expression were further aggravated in the si-PPT2-EGFL8–transfected cells (Fig. 6J–N). There was no significant difference in hypoxia-induced increased VEGF expression after si-PPT2-EGFL8 transfection (Fig. 6J, O, and P).

HRMCEs were transfected with siRNA targeting PPARD (si-PPARD) or a nontargeting siRNA control (siNC) for 48 h. The siRNA expression vectors efficiently reduced the mRNA and protein level of PPARD in hRMECs (Fig. 6Q–S). Accordingly, the hypoxia-induced upregulation of ANGPTL4 expression further increased in the siRNA-transfected cells (Fig. 6T–V). Moreover, the activating PPARD agonist GW501516 enhanced the hypoxia-induced expression of cANGPTL4, and the inverse PPARD agonist DG172 strongly inhibited the hypoxia-induced expression of cANGPTL4 (Fig. 6W and X). Hypoxia-induced increases in hRMEC tubule formation were significantly inhibited by DG172 pretreatment; this effect was abolished by ANGPTL4 administration (Fig. 6Y and Z).

To further evaluate the effect of ANGPTL4 on blood vessels during sprouting angiogenesis, we injected mice with ANGPTL4 or saline solution in the vitreous humor at P3. Serial whole eye sections taken from P7 mice were analyzed. The number of hyaloid vessels residing in the vitreous increased in the ANGPTL4- and/or VEGFA-treated mice at P7. Moreover, ANGPTL4 induced retinal vessel breakage in the inner limiting membrane (ILM) and extension into the vitreous in the superficial portion of the inner retina, and VEGF mainly induced retinal edema (Fig. 7A). In the mouse model of OIR, H-E staining consistently indicated outpouching of ILM or retinal NV tufts growing into the core vitreous in the inner retinas of the ANGPTL4-treated mice, and VEGFA treatment increased the number of superficial retinal vessels (Fig. 7B and C). Meanwhile, ANGPTL4 was upregulated in the serum of individuals with PDR versus healthy individuals (Fig. 7D).

In the current study, we demonstrated that lncRNA PPT2-EGFL8 played an important role in hRMEC proliferation and retinal NV. We demonstrated the underlying mechanism of the effects of PPT2-EGFL8 in retinal angiogenesis by absorbing miR-423-5p as a molecular sponge, which subsequently controls the activation of PPARD/ANGPTL4 signaling (Fig. 7E).

Retinopathy of prematurity (ROP) and PDR are both regarded as retinal angiogenic disorders (17). ROP is characterized by an initial factor of relative ischemia, followed by abnormal NV that may culminate in the rapid outgrowth of pathological blood vessels from the retina to the vitreous humor and lens that may lead to retinal hemorrhage, detachment, and blindness (18). Both phases of ROP are reproduced by the mouse OIR model, which can be used to test therapeutic substances or other treatments (19). In our study, we observed that overexpression of Egfl8 could attenuate the deleterious effects of OIR-induced retinal angiogenesis driven by retinal hypoxia at P17, whereas the effects were aggravated in the animal group that received Egfl8.

Moreover, the STZ-induced diabetes model is one of the most commonly used models in diabetes research. STZ-induced diabetic mice display phenotypical and histopathological features consistent with those of PDR (19). In the STZ-induced diabetic mice, vascular leakage, increased acellular vessels, pericyte loss, and reduced visual function were found as early as 2–3 months after the STZ injection (20). The preproliferative stage of DR was detected in diabetic mice at 5–6 months and exhibited capillary dilation, varicose loop formation, and endothelial cell proliferation (21,22). This preproliferative stage of DR in STZ-induced diabetic animals corresponds to the intraretinal microvascular abnormality observed in human DR (23,24). In our study, the number of hyperreflective dots, the outpouching of the ILM, and the breach of the ILM in the superficial portion of the inner retina in OCT images were significantly increased in 5-month-old diabetic mice compared with control mice. Overexpression of Egfl8 inhibited the deleterious effects in the STZ-induced diabetic mice. Likewise, we observed that AAV2-Egfl8 treatment attenuated the occurrence of retinal NV and the chronic subclinical inflammatory response in DR.

Although several lncRNAs have been shown to increase in DR and have roles in microvascular dysfunction, little is known about the lncRNAs with downregulated expression during DR. In this study, we identified an RMEC-enriched lncRNA (named lnc-PPT2-EGFL8) that attenuated hRMEC proliferation and retinal NV in vitro and in vivo. Meanwhile, miR-423-5p expression increased significantly in the serum of individuals with DR compared with healthy individuals and was involved in the development of DR. On this basis, this study further provides comprehensive functional and mechanistic characterization of PPT2-EGFL8, functioning as a ceRNA. PPT2-EGFL8 blocks miR-423-5p to control PPARD protein levels in vitro and in vivo.

PPARs were previously shown to regulate differentiation and proliferation in different model systems (25). These systems include Ppard-deficient mice, which show an increased arteriovenous crossover and wider venous caliber in the retina. These findings demonstrate a fundamental role for PPARD in pathological angiogenesis and blood vessel remodeling in the retina (26). In addition, there is strong evidence that ANGPTL4 is directly regulated by ligand activation of PPARD based on the following analysis: 1) functional PPREs exist in the proximal end of ANGPTL4 gene, and 2) the promoter occupancy of PPARD on ANGPTL4 gene increased after ligand activation. In the absence of an exogenous ligand and the destruction of PPARD, the expression of ANGPTL4 increased. However, PPARD ligand activation leads to increased expression (27). Moreover, PPARD mediates transcriptional repression of ANGPTL4 in cancer cell invasion (28–30). Consistent with previous studies, we found that the downregulation of PPARD expression by siRNAs induced a significant increase in ANGPTL4 levels in hRMECs. Moreover, we made use of a recently developed PPARD agonist (GW501516) and subtype-specific PPARD inhibitor (DG172) (Fig. 7E and F), which regulate the expression of ANGPTL4 by transcriptional repression.

In the prevention or treatment of PDR, anti-VEGF therapy has attracted increased attention. A potent inhibition of PDR progression in some—but not all—patients receiving anti-VEGF therapy was observed, suggesting that other angiogenic factors may participate in PDR development (31). ANGPTL4, an ischemia-driven mediator, has been implicated in promoting retinal NV and vascular permeability in patients with diabetes (32,33). In some patients, the ANGPTL4 level remains elevated despite anti-VEGF therapy. However, ANGPTL4 inhibition significantly reduced the angiogenic effect in hypoxic retinal cells, which was additive with simultaneous VEGF inhibition (32). In our study, although both ANGPTL4 and VEGFA increased the number of hyaloid vessels residing in the vitreous, ANGPTL4 could induce retinal vessel breakage in the ILM and extension into the vitreous in the superficial portion of the inner retina at P7 in mice. Moreover, ANGPTL4 significantly induced the outpouching of ILM or retinal NV tufts growing into the core vitreous in the inner retinas, while VEGFA treatment mainly increased the number of superficial retinal vessels. These data suggest that ANGPTL4 plays a critical role in retinal NV.

In summary, the current study provides strong evidence that PPT2-EGFL8 promotes retinal angiogenesis by absorbing miR-423-5p as a molecular sponge, which subsequently controls the activation of PPARD/ANGPTL4 signaling. These results indicate that therapies targeting lncPPT2-EGFL8 may be an effective approach for the treatment of patients with PDR and that PPT2-EGFL8 could be used as a biomarker in patients receiving anti-VEGF therapy to help guide individualized therapy.

Funding. This study was supported by the Wuxi Taihu Lake Talent Plan, Supports for Leading Talents in Medical and Health Profession (grant nos.: 2020-THRCTD-1 and THRC-DJ-1), Top Talent Support Program for Young and Middle-Aged People of Wuxi Health Committee (grant nos. HB2020004 and HB2020022), the National Natural Science Foundation of China (grant no. 81800845), Medial Key Discipline Program of Wuxi Health Commission (ZDXK2021001) and Wuxi Translational Medicine Research Project (ZH201902).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Z.X., J.Y., H.Z., T.X., and Q.Y. performed in vitro and in vivo experiments and collected and analyzed the data. J.C., C.S., Y.Cao., M.W., Y.L., and Y.Cui. designed the experiments and revised the article critically for important intellectual content. Y.Y. and X.W. conceived and designed the study. All authors gave final approval of the version to be published. X.W. is the guarantor of this work and, as such, had full access to all the data in this study and takes responsibility for the integrity of the data and the accuracy of the data analysis.