The Exosome-Transmitted lncRNA LOC100132249 Induces Endothelial Dysfunction in Diabetic Retinopathy Free

Diabetic retinopathy (DR), one of the most common microvascular complications in diabetes, is the leading cause of visual impairment and blindness in working-age adults globally (1). Although retinal abnormal neovascularization, the main pathogenetic basis of DR, is widely known to cause vitreous hemorrhage and tractional retinal detachment, its underlying pathogenesis is extremely complicated and multifactorial. To date, the pharmaceutical treatment of neovascularization for DR includes anti–vascular endothelial growth factor drugs or corticosteroids, which have been demonstrated to be effective in some cases (2,3). However, this treatment has several severe limitations, such as the short half-life of administered drugs, the repeated and uncomfortable injection process, and long-term adverse effects, all of which aggravate the patient’s economic and psychological burden. Hence, novel intervention biomarkers for DR diagnosis, treatment, and prevention are urgently needed.

Exosomes are small (30–150-nm diameter) membrane vesicles of endosomal origin that are continually secreted by a variety of cell types in response to physiological or pathological stimuli (4). They have been demonstrated to exist in almost all biological fluids and have been found to contain various types of cargoes (5). Furthermore, they can travel in the extracellular space as messengers carrying bioinformation in cell-to-cell communication and can be internalized by target cells via membrane fusion or endocytosis. After subsequent release of their cargo, exosomes mediate various biological and pathological processes.

Long noncoding RNAs (lncRNAs) were first reported in the late 1980s, defined as a class of regulatory noncoding RNAs without evident protein coding capability (6). They are present in either nucleus or cytoplasm and can interact with DNA, RNA, or proteins. miRNAs are the most known small noncoding RNAs, with a negative regulation role of mRNA gene expression through degradation induction or translation inhibition (7). Functionally, lncRNAs could act as competing endogenous RNAs or RNA sponges of target miRNAs, thus regulating gene expression at the posttranscriptional level. lncRNAs have been demonstrated to be widely involved in extensive pathological processes, such as oxidative stress, inflammation, and angiogenesis, all of which are critical in DR. Furthermore, exosome-transferred lncRNAs can bypass recipient cell transcriptional controls, thus providing a relatively direct means of regulation (8). In this study, we aimed to elucidate the roles of exosomes and their loaded lncRNAs in DR pathogenesis. Our findings may reveal novel mechanisms and intervention strategies for DR.

Vitreous samples for exosomal harvesting were collected from patients with proliferative DR (PDR) or patients with macular hole (MH) (as controls) from 1 March to 31 December 2019 at the First Affiliated Hospital of Nanjing Medical University. The procedures used in this study conformed to the tenets of the Declaration of Helsinki, and informed consent was obtained from each patient. This study was approved by the ethics committee of the Faculty of Medicine, Nanjing Medical University (2017-SR-283).

Exosomes from 25 patients with PDR and 40 control patients were used for exosomal sequencing (Supplementary Table 1). We first performed cataract surgery, if needed, and harvested 0.8 mL of vitreous humor for each eye before vitrectomy infusion, and then the samples were centrifuged at 10,000 rpm for 10 min to remove red blood cells and cell debris. In addition, 2 mL of plasma from the same patients was collected. All samples were stored at −80°C until analysis. Fibrovascular membranes (FVMs) from four patients with PDR and epiretinal membranes (ERMs) from four patients with idiopathic ERMs were harvested separately during surgery and fixed, dehydrated, and embedded for immunofluorescence (Supplementary Table 2). In this study, sex was not considered as a factor in the statistical analysis of the data.

All retinal cell lines, including human retinal vascular endothelial cells (HRVECs), human umbilical vein endothelial cells (HUVECs), Müller cells, 661W photoreceptor cells, retinal pigment epithelia, and human retinal pericytes were purchased from the American Type Culture Collection. Cells were cultured at 37°C with 5% CO₂ under three different conditions: normal glucose (NG) (5.5 mmol/L), high glucose (HG) (30 mmol/L), and hyperosmolarity (5.5 mmol/L glucose and 24.5 mmol/L mannitol) as an osmotic control. For in vivo experiments, primary mouse retinal vascular endothelial cells were cultured in DMEM with 10% FBS using the recommended protocol (9–12).

To mimic a hypoxic environment, we cultured HRVECs in 3% O₂ at 37°C for 48 h (13–15). To construct an oxidative stress model in vitro, we exposed HRVECs to H₂O₂ for 24 h at a concentration of 200 μmol/L (16,17). The inflammatory stimulus was lipopolysaccharide at 1 μg/mL administered to HRVECs for 24 h (18,19).

Vitreous humor, plasma, or conditioned cell culture medium was collected, and exosomes were isolated with a multistep centrifugation procedure as previously described (20,21). Finally, the precipitates were resuspended in PBS.

Exosomes were fixed in 2% paraformaldehyde and placed on a Formvar-coated copper grid for 30 min. After the grid was washed in PBS and incubated with 2% glutaraldehyde for 10 min, ultrathin sections of 100 nm in thickness were prepared and stained with uranyl acetate and lead citrate at room temperature. The morphology of exosomes was then viewed using transmission electron microscopy (TEM). The size distribution of exosomes was quantified with Zetasizer Nano ZS (Malvern Panalytical, Malvern, U.K.) according to the standard instructions.

In our pilot test, the total amount of exosomes in PDR vitreous humor was higher than that in MH with the same sample size. Normally, 4 mL of vitreous humor in PDR is a sufficient sample for exosomal RNA sequencing. Therefore, we pooled the vitreous humor from five patients with PDR (4 mL) as one sample and eight patients with MH (6 mL) as a control sample. In each group, the sample size was five. The vitreous humor of each group was mixed with Ribo Exosome Isolation Reagent (RiboBio, Guangzhou, China), and exosome isolation was performed according to the manufacturer’s instructions. Total RNA from exosomes was used for library preparation and sequencing according to the instructions provided with the NEBNext Ultra RNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA). Differentially expressed lncRNAs for subsequent analysis were assessed with the criteria of fold change >4 and P < 0.0083.

Isolated exosomes were labeled with PKH26, a red fluorescent cell linker dye (Sigma-Aldrich, St. Louis, MO). The labeled exosomes were cocultured with HRVECs at 37°C for 24 h. After incubation, cells were fixed with 4% paraformaldehyde and stained with DAPI. Uptake of exosomes was visualized with a laser confocal microscope (MIC00223 LSM5 Live; Zeiss).

Fluorescence in situ hybridization (FISH) assays were used to detect the distribution of LOC100132249 in HRVECs and HUVECs. A specific FISH probe for LOC100132249 was synthesized by RiboBio, and the assay was performed according to the recommendations of the Ribo FISH kit. Finally, images were captured with a confocal microscope (MIC00223 LSM5 Live).

The nuclear and cytoplasmic fractions of HRVECs or HUVECs were separated according to the instruction manual of the PARIS Kit (Invitrogen). The harvested cells were initially treated with cell fractionation buffer and incubated for 10 min on ice. The upper solution was collected by centrifugation as the cytoplasmic fraction, whereas the nuclear pellets at the bottom were lysed with cell disruption buffer. Quantitative RT-PCR (RT-qPCR) was used to detect the expression levels in the cytoplasm or nuclei of HRVECs and HUVECs.

Newborn C57BL/6J mice were used to establish an oxygen-induced retinopathy (OIR) model. The sex of newborn mice was difficult to distinguish and was not considered as a factor in the statistical analysis of the data. The juvenile mice and their mothers were exposed to a chamber with high oxygen concentration (75 ± 2%) at postnatal day 7 (P7) for 5 days and then were returned to room air conditions at P12 and administered an intravitreal injection of ∼1.5 μL of exosomes. Juvenile mice were sacrificed at P17, and their eyes were removed and fixed for 4 h at 4°C. The retinas were cut into four fragments and then blocked and permeabilized in 0.3% Triton X-100 plus 5% BSA overnight at 4°C. The whole-mount retinas were stained with isolectin B4 (IB4) (Vector Laboratories, Burlingame, CA) and anti-SNAI1 antibody (1:200, ab180714; Abcam). The number of retinas in each group was at least four. Images were acquired with a fluorescent microscope (Leica THUNDER DMi8). The avascular area and the neovascular tuft were manually outlined and calculated using ImageJ software (National Institutes of Health, Bethesda, MD) (22).

The statistical analyses of all experimental data were performed using SPSS version 19 software and presented as mean ± SD. A two-tailed unpaired Student t test was applied to compare the differences between two groups, whereas a one-way or two-way ANOVA was used for multiple group comparisons as appropriate. Differences were considered statistically significant at P < 0.05. All experiments were conducted in triplicate in at least three independent trials.

All data supporting the findings of this study are available from the corresponding authors upon reasonable request.

To evaluate the effects of exosomes derived from vitreous humor of patients with PDR (PDR-exo) and MH (MH-exo) on the angiogenic ability of HRVECs, we first extracted and characterized the PDR-exo and MH-exo. The exosomes all had a typical cup-shaped morphology and were 30–150 nm in size, as shown by TEM (Fig. 1B) and nanoparticle tracking analysis (Fig. 1D). Moreover, Western blot analysis verified several exosomal-specific markers, such as CD9, CD63, and TSG101, but not calnexin, an endoplasmic reticulum protein (Fig. 1C). After incubation with HRVECs, the PKH26-labeled exosomes were mainly distributed around the nuclei (Fig. 1E). After 24 h, compared with MH-exo, PDR-exo dramatically enhanced the cell proliferation, migration, and tube formation ability of HRVECs (Fig. 1F–I). These findings imply the ability of PDR-exo to aggravate HRVEC dysfunction.

To investigate the specific exosomal lncRNA candidates participating in angiogenesis, we compared the expression profiles of lncRNAs between PDR-exo and MH-exo using next-generation sequencing (Supplementary Table 3). A total of 899 lncRNAs were aberrantly expressed in PDR-exo, among which 547 were upregulated with P < 0.05 and log₂-fold change >2 (Fig. 2A). From the upregulated exosomal lncRNAs, we chose the top four for further RT-qPCR validation. LOC100132249 was the most significantly upregulated lncRNA in PDR-exo, with levels ∼10-fold higher than those in MH-exo (P < 0.0001), and was therefore selected for in subsequent analysis (Fig. 2B). For the vitreous exosomes, there might be contributions from the blood because of retinal hemorrhage. To demonstrate that the upregulated lncRNA LOC100132249 was enriched in vitreous humor, in blood, or in both, we also performed exosomal sequencing of blood from the same patients (Supplementary Table 4). We found that there was no difference of this lncRNA in blood between PDR or MH, indicating that lncRNA LOC100132249 was secreted from the retinal cells. Of note, LOC100132249 was much more upregulated in PDR-exo than PDR vitreous humor, ∼10-fold higher in exosomes, while only threefold higher in vitreous humor compared with their separate control (Fig. 2B and C), thus indicating that LOC100132249 was enriched in PDR-exo.

Chronic and sustained hyperglycemia triggers a series of events culminating in vascular dysfunction, and endothelial cells are recognized to be the earliest cellular targets with histological changes in PDR. To investigate the cell resources from which LOC100132249 is derived, we quantified the LOC100132249 expression level among several retinal cells under HG stimulus. The RT-qPCR results showed a clear elevation of LOC100132249 levels in HRVECs (Fig. 2D), then HUVECs (Fig. 2E). No significant difference was observed among other cell types (Supplementary Fig. 1A). Furthermore, HRVECs were exposed to oxidative stress (H₂O₂), an inflammatory stimulus (lipopolysaccharide), and hypoxia to mimic diabetic stress in vitro. Oxidative damage, inflammatory attack, and hypoxia did not affect the LOC100132249 expression in HRVECs (Supplementary Fig. 1B). Thus, we reasoned that HG-treated HRVECs might have been the main source of LOC100132249. Subsequently, FISH staining and subcellular fractionation indicated that LOC100132249 localized mainly in the nuclei of endothelial cells (Fig. 2F–I).

As described previously, LOC100132249 was highly expressed in HRVECs under HG. To further determine the HRVEC-derived exosomal LOC100132249 expression level, we isolated exosomes from HRVECs cultured under HG (HG-exo) or NG (NG-exo) conditions and performed further characterization (Supplementary Fig. 2A and B). LOC100132249 was significantly enriched in HG-exo, to levels approximately fourfold those of NG-exo (Fig. 3A), but was only 1.5-fold its corresponding cellular level (Fig. 2F). Next, PBS, NG-exo, or HG-exo were used to stimulate normal HRVECs, and the cellular LOC100132249 levels were quantified. The fluorescence-labeled exosomes clearly showed uptake by HRVECs (Fig. 3C). After 48 h, an increase in cellular LOC100132249 levels was observed in recipient HRVECs after incubation with HG-exo but not with PBS or NG-exo (Fig. 3B). Furthermore, the cell proliferation, migration, and tube formation abilities of HRVECs were significantly enhanced by HG-exo (Fig. 3D–H).

In the OIR mouse model, hematoxylin-eosin staining demonstrated that endothelial cell nuclei of sprouting vessels broke through the inner limiting membrane into the vitreous cavity, thus indicating retinal neovascularization (Fig. 3I). After the intravitreal injection of PBS, NG-exo, or HG-exo, exosomes were mostly internalized by retinal endothelial cells (Supplementary Fig. 2C), and the HG-exo group showed increased neovascular tuft area and fluorescence leakage, which represented the formation of immature angiogenesis in the retina. No significant difference in central avascular area was observed between groups (Fig. 3J and K).

To explore whether HG-exo mediated HRVEC angiogenesis is via exosomal LOC100132249, we suppressed or overexpressed this lncRNA in HRVECs under HG conditions (Fig. 4A). RT-qPCR indicated that the exosomal LOC100132249 levels were consistent with cellular LOC100132249 expression after suppression or overexpression (Fig. 4B). Next, after separately adding exosomes secreted by LOC100132249-interfered HRVECs (si-LOC100132249-exo) or exosomes secreted by LOC100132249-overexpressed HRVECs (LOC100132249 plasmid-exo) into normal HRVECs for 48 h, we observed that the cellular LOC100132249 expression was elevated in HRVECs cocultured with LOC100132249 plasmid-exo but was diminished in the si-LOC100132249-exo–treated group (Fig. 4C). Furthermore, knockdown of LOC100132249 in exosomes hindered, whereas upregulation of LOC100132249 in exosomes promoted, HRVEC proliferation, migration, and tube formation (Fig. 4D–G and J). In the OIR mouse model, whole-retina flat-mount staining with IB4 demonstrated that intravitreal administration of LOC100132249 plasmid-exo enhanced retinal vascular leakage, whereas si-LOC100132249-exo showed inhibitory effects (Fig. 4H, I, and K). These data indicate that exosomal LOC100132249 contributed to angiogenesis both in vitro and in vivo.

Accumulating evidence indicates that lncRNAs can function as sponges of miRNAs, thereby regulating mRNA expression levels. We thus identified potential miRNAs that interact with LOC100132249 by performing small RNA sequencing after transfecting LOC100132249 plasmid into HRVECs (Supplementary Table 5). As displayed in the volcano plot shown in Fig. 5A, 17 upregulated and 63 downregulated miRNAs were found to be differentially expressed. We validated 15 of the downregulated miRNAs binding LOC100132249 using RT-qPCR, among which miR-199a-5p and miR-1296-5p were verified to show similar change trends (Fig. 5B). In addition, we performed nuclear and cytoplasmic fractionation of HRVECs and observed that miR-1296-5p was predominantly localized in the cytoplasm, whereas miR-199a-5p was present in both the nuclei and cytoplasm (Fig. 5C). Given that LOC100132249 localized to nuclei, we assumed that LOC100132249 targeted miR-199a-5p in the nucleus. Therefore, we chose miR-199a-5p to perform subsequent investigation into the mechanism. Consistently, miR-199a-5p was also downregulated in exosomes derived from the vitreous humor of patients with PDR in our high-throughput sequencing (Fig. 5D).

To elucidate the effects of miR-199a-5p on HRVECs, we first transfected miR-199a-5p mimics or inhibitor into HRVECs (Fig. 5H). Subsequently, we demonstrated that miR-199a-5p mimics attenuated HRVEC proliferation, migration, and tube formation, whereas miR-199a-5p inhibitor had opposite effects (Supplementary Fig. 3A–E). Furthermore, the binding sites between LOC100132249 and miR-199a-5p were explored with starBase software and dual-luciferase reporting assays. As anticipated, the relative luciferase activity of wild-type LOC100132249 (LOC100132249) was clearly restrained in the presence of miR-199a-5p mimics, but this response was not detected in mutated LOC100132249 (Fig. 5E). In addition, the expression of miR-199a-5p was lower in HRVECs under HG (Fig. 5F) but was promoted by LOC100132249 knockdown (Fig. 5G). Moreover, the level of miR-199a-5p was upregulated after simultaneous transfection of LOC100132249 plasmid and miR-199a-5p mimics into HRVECs (Fig. 5I). More importantly, miR-199a-5p inhibition resulted in partial recovery of the decrease in proliferation, migration, and tube formation in HRVECs after LOC100132249 knockdown (Supplementary Fig. 4A–G). Collectively, these results illustrate the negative regulation between LOC100132249 and miR-199a-5p in the angiogenesis of HRVECs.

Because lncRNAs are known to block the regulatory effects of miRNAs on their target genes, we next predicted the potential target genes of miR-199a-5p through the intersection of four databases. Ultimately, 37 genes were eligible candidates, among which we focused on SNAI1, a crucial initiator involved in both physiological and pathological ocular neovascularization (Fig. 6A). Dual-luciferase reporter assays indicated binding sites for miR-199a-5p in the SNAI1 3′-untranslated region (Fig. 6B and C). Moreover, SNAI1 was upregulated in HRVECs after overexpression of LOC100132249 but was downregulated when LOC100132249 was interfered (Supplementary Fig. 5A). In contrast, cells transfected with miR-199a-5p inhibitor exhibited a high level of SNAI1 (Supplementary Fig. 5B). To interrogate whether miR-199a-5p effectively functioned as a bridge in LOC100132249-mediated regulation of SNAI1, we simultaneously transfected LOC100132249 plasmid and miR-199a-5p mimics into HRVECs and measured the level of SNAI1. As predicted, miR-199a-5p mimics abolished the LOC100132249-mediated restoration of SNAI1 levels (Fig. 6D). An opposite phenomenon was observed after treatment with si-LOC100132249 along with miR-199a-5p inhibitor (Fig. 6E). These data demonstrate that LOC100132249-mediated SNAI1 expression was dependent on the downregulation of miR-199a-5p.

To further assess the proangiogenic effect of SNAI1, we first indicated successful elevation of SNAI1 levels with the SNAI1 plasmid and depletion of SNAI1 levels with si-SNAI1-1 (Supplementary Fig. 5C). Next, we divided cells into three groups transfected with si-negative control (si-NC), si-LOC100132249, or si-LOC100132249 plus SNAI1 plasmid. The decline in SNAI1 in LOC100132249-depleted cells was dramatically upregulated with the SNAI1 plasmid (Fig. 6F). Likewise, SNAI1 increased with the overexpression of LOC100132249 and was blocked by cotransfection of the LOC100132249 plasmid and si-SNAI1 (Fig. 6G). In addition, SNAI1 interfering prevented LOC100132249 overexpression-induced angiogenesis, and vice versa, overexpression of SNAI1 reversed the si-LOC100132249–induced decrease of angiogenesis (Fig. 6H–L). In the OIR mouse model, we costained retinas with SNAI1 and IB4 (or CD31). In the LOC100132249 plasmid-exo–administered group, SNAI1 accumulated around the malformed vessels, with colocalization of highly expressed SNAI1 and IB4 (Fig. 7A and B). In frozen sections of retina in OIR mice, SNAI1 was highly expressed in the sprouting vessels, and the LOC100132249 plasmid-exo group showed the most abundant expression of CD31 as well as SNAI1 (Fig. 7C–E). The RT-qPCR and Western blot analysis of retinas from each group further confirmed our results (Fig. 7G–I). In FVMs harvested from patients with PDR, SNAI1 and CD31 were more abundant than they were in ERMs (Fig. 7F). Altogether, the genetic and functional data indicate that SNAI1 was responsible for LOC100132249-mediated angiogenesis through competition for miR-199a-5p binding.

In the development of angiogenesis, endothelial cells undergo a phenotypic switch from cobble-shaped to mesenchymal-like cells to obtain more invasive and migratory traits, a process sharing similarities with endothelial-mesenchymal transition (EndMT) (23). The morphology of HRVECs cultured under HG was elongated and spindle shaped (Supplementary Fig. 6A). HRVECs transfected with LOC100132249 plasmid showed diminished fluorescence intensity of the endothelial markers CD31 and VE-cadherin and elevated intensity of SNAI1 and the mesenchymal markers α-smooth muscle actin (α-SMA) and vimentin (Fig. 8A and Supplementary Fig. 6B).

Canonical Wnt/β-catenin signaling is a typical pathway reported to drive EndMT by triggering SNAI1 expression (24,25). Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis of exosomal vitreous humor sequencing also showed the involvement of the Wnt signaling pathway (Supplementary Fig. 6C). Furthermore, we observed considerably suppressed protein expression of β-catenin, SNAI1, and mesenchymal markers but intensified VE-cadherin and CD31 expression after interfering of LOC100132249 (Fig. 8B, C, and H). In addition, we stained β-catenin in cells stably expressing LOC100132249 and found that β-catenin was diminished in the cytoplasm but significantly elevated in the nucleus, thus implying that LOC100132249 might promote β-catenin transferring from the cytoplasm to the nucleus (Fig. 8A). Next, the upregulatory effects of LOC100132249 plasmid on β-catenin, SNAI1, and EndMT-associated proteins were found to be reversed by SNAI1 interfering or miR-199a-5p mimics in HRVECs (Fig. 8D, G, I, and J). In contrast, the endothelial biomarkers were downregulated, whereas the mesenchymal biomarkers, as well as SNAI1 and β-catenin, were upregulated in HRVECs cotransfected with si-LOC100132249 and miR-199a-5p inhibitor or si-LOC100132249 and SNAI1 plasmid (Fig. 8E, F, I, and J). Collectively, the data demonstrate that LOC100132249 regulated miR-199a-5p/SNAI1–mediated EndMT progression by activating Wnt/β-catenin signaling.

There is an urgent need to improve understanding of DR pathogenesis and to develop effective pharmacological strategies for retinal neovascularization. Retinal hyperglycemia plays a pivotal role in endothelial dysfunction in patients with DR and leads to compensatory pathological angiogenesis to reinstate metabolic equilibrium (26,27). The present study reports the first evidence, to our knowledge, that exosomal LOC100132249 is elevated in the vitreous humor of patients with PDR. Endothelial-derived exosomes accelerate pathological angiogenesis via enrichment in LOC100132249. Mechanistically, LOC100132249 modulates endothelial dysfunction through the miR-199a-5p/SNAI1 axis, which activates the Wnt/β-catenin signaling pathway involved in EndMT.

There have been massive efforts to uncover factors that might mitigate the occurrence and development of DR. During the past decade, emerging evidence has indicated the roles of noncoding RNAs as diagnostic biomarkers and therapeutic targets for DR. Noncoding RNAs, including lncRNAs, can be encapsulated by exosomes and remain stable until they are released to act in cell-to-cell communication when needed (28). Recently, lncRNAs have been found to be present in the vitreous humor, and their expression pattern may be altered by various pathogenic processes (29–31). Although several lncRNAs have been delineated, the roles of most lncRNAs in DR, particularly within exosomes, have not been comprehensively examined. Here, we first confirmed that exosomes derived from vitreous humor from patients with PDR were responsible for aberrant sprouting of vessels. The new exosomal lncRNA LOC100132249 was first detected via next-generation sequencing of PDR-exo. Furthermore, HG-stimulated HRVECs were found to be the origin of the highly expressed LOC100132249. This high expression further promoted proliferation, migration, and tube formation in normal endothelial cells through an miR-199a-5p/SNAI1 axis in EndMT. However, more clinical samples of other neovascularization-related retinal diseases are needed to further validate the role of exosomal LOC100132249.

EndMT is a type of epithelial-mesenchymal transition in which endothelial cells lose the integrity of adhesion and tight junctions between cells, thus manifesting a migratory, invasive, and proliferative phenotype with simultaneous morphological changes (32). EndMT was first observed in the development of the heart with respect to heart valve formation and, more importantly, angiogenesis (33). Recently, several lncRNAs have been found to be associated with EndMT (34,35). However, the roles of most lncRNAs in EndMT remain only preliminarily understood, and deeper exploration is needed. Recently, a partial EndMT process has been proposed during angiogenesis, in which endothelial cells remain in an intermediate stage and display a fundamental transition in both morphological aspects and functions but retain most endothelial characteristics. We found that LOC100132249/miR-199a-5p/SNAI1 loop regulated EndMT. Western blotting and immunofluorescence staining analysis showed coexpression of CD31/VE-cadherin and vimentin/α-SMA, thus suggesting that the cells might undergo partial EndMT. SNAI1, a prominent inducer of EndMT, has been found to be abundant in the retina and to play critical roles in ocular neovascularization (36). Moreover, SNAI1 is highly expressed in endothelial cells under abnormal ocular angiogenesis conditions (37), in agreement with our results indicating SNAI1 accumulation around leaky vessels in the OIR mouse model. Thus, as supported by the above evidence, SNAI1 may promote angiogenesis through modulation of EndMT.

The Wnt/β-catenin pathway is believed to be essential in controlling angiogenesis and EndMT (38,39). β-Catenin, the key molecule in this signaling pathway, alters not only the expression level but also the cellular localization when the pathway is activated by various stimuli. HG exposure, hypoxia, and inflammatory irritations have been found to be associated with induction of EndMT and activation of the Wnt/β-catenin pathway (40). In the present study, the newly demonstrated exosomal LOC100132249 and its downstream target gene were found to be involved in EndMT in DR through the Wnt/β-catenin pathway. Our findings are in line with those from previous reports indicating that SNAI1 positively regulates β-catenin, and the canonical Wnt signaling pathway mediates the translocation of β-catenin to the nucleus, where it triggers EndMT by inducing the expression of SNAI1 (24,41–43).

Of note, there are several limitations in our study. First, the use of different cell lines to detect the source of exosomal LOC100132249 could hardly represent the in vivo condition. Second, current results reveal that exosomal LOC100132249 regulated SNAI1 expression via sponging miR-199a-5p. However, the enrichment process of specific lncRNA LOC100132249 into exosomes is still largely unknown. Third, we have not studied the direct interaction between SNAI1 and β-catenin, and the mechanisms related to β-catenin nuclear translocation remain unclear, which calls for more investigations in our future study.

In conclusion, we demonstrate an interaction between HG-treated HRVECs and normal HRVECs via exosome-packaged lncRNAs. Our data provides new insight into the roles of exosomal LOC100132249 in EndMT and retinal neovascularization in PDR by targeting the miR-199a-5p/SNAI1/Wnt/β-catenin pathway. Our findings regarding exosomes derived from the vitreous humor in patients with PDR and their cargoes greatly enhance the understanding of the pathophysiological changes in PDR and may expedite the development of novel therapeutic targets in the future.

Acknowledgments. The authors thank the medical workers in the Department of Ophthalmology in Jiangsu Provincial People’s Hospital for specimen conservation.

Funding. This work was supported by the National Natural Science Foundation of China (8207097 to P.X., 81970821 to Q.L., 81900875 to Z.H., and 12027808 to C.T.), the Natural Science Foundation of Jiangsu Province (BK20191059 to Z.H.), and the Social Development Program of Jiangsu Province (BE2021744 to P.X.).

The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

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

Author Contributions. Z.H., J.W., T.P., and X.L. analyzed the exosomal sequencing data and together performed the animal experiments. Z.H. and J.W. wrote the manuscript. C.T., Y.W., and X.Wa. contributed to the discussion. Z.Z., Y.L., W.Z., C.X., and X.Wu. contributed to the analysis of the data. Q.G., Y.F., H.Q., A.M., and S.Y. collected the clinical samples. Q.L. and P.X. reviewed the manuscript. P.X. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.