Regulation of Long Noncoding RNA NEAT1/miR-320a/HIF-1α Competitive Endogenous RNA Regulatory Network in Diabetic Retinopathy

Abstract

Purpose: To determine the mechanism that long noncoding RNA NEAT1 (lncNEAT1)/miR-320a competitive endogenous RNA (ceRNA) network regulates hypoxia-inducible factor–1α (HIF-1α) in ARPE-19 cells and its potential role in diabetic retinopathy (DR).

Methods: ARPE-19 cells were cultured in a normal or high-glucose (HG) medium, and cell migration, invasion, and permeability were detected by scratch, transwell, and FITC-dextran staining assays. LncNEAT1, HIF-1α, ZO-1, occludin, N-cadherin, and vimentin levels were tested. The binding of lncNEAT1 to miR-320a was verified by dual-luciferase reporter assay, and the binding of miR-320a to HIF-1α by RIP assay. ARPE-19 cells were treated with lncNEAT1 or HIF-1α shRNA or miR-320a agomir to determine the activation of ANGPTL4/p-STAT3 pathway. The effect of lncNEAT1 in DR and its regulations on miR-320a and HIF-1α were determined in a rat model of DR.

Results: HG treatment promoted the migration, invasion, and permeability of ARPE-19 cells. After lncNEAT1 silencing, HIF-1α, N-cadherin, and vimentin levels were downregulated, ZO-1 and occludin levels were upregulated, and the migration, permeability, and invasion of HG-treated ARPE-19 cells were inhibited. However, HIF-1α overexpression increased N-cadherin and vimentin expression, reduced ZO-1 and occludin expression, and promoted the migration, permeability, and invasion of ARPE-19 cells. The binding of miR-320a with both lncNEAT1 and HIF-1α was predicted and confirmed. In a diabetic rat model, silencing lncNEAT1 inhibited HIF-1α/ANGPTL4/p-STAT3 pathway activation and alleviated retinopathy.

Conclusions: The lncNETA1/miR-320a/HIF-1α ceRNA network activates the ANGPTL4/p-STAT3 pathway and promotes HG-induced ARPE-19 cell invasion and migration.

Diabetes is often complicated by microvascular disease, called diabetic retinopathy (DR), which accounts for part of visual disability and loss in working-age adults.¹ DR has a nonproliferative stage, featured by vascular tortuosity, microaneurysms, retinal bleeding, and lipid exudates, and a proliferative stage, featured by the occurrence of fragile new aberrant vessels.² Retinal pigment epithelium (RPE) is a single layer of pigmented cells at the interface between photoreceptors and choroidal vasculature and forming a selective barrier that regulates the transport of nutrients from blood to the outer retina.³ Moreover, RPE secretes several growth factors that regulate vascular homeostasis, such as pigment epithelium-derived factor and somatostatin, two antiangiogenic factors.⁴ Therefore the integrity of RPE is crucial to normal visual function and prevention of aberrant neovasculogenesis. However, under pathological circumstances such as DR, RPE cells become dedifferentiated and exhibit mesenchymal-like characteristics, namely increased capacity to proliferate and migrate; this process is called epithelial-mesenchymal transition (EMT), which can be caused by impaired tight junctions.^5,6 A growing number of studies have been devoted to the gene regulatory networks and genomic variations behind DR to prevent and treat this complicated disease, yet no ideal targets have been determined.^7,8 Molecules regulating EMT and permeability of RPE cells may be novel competitive targets in DR therapy. 

Multiple differentially expressed long-noncoding RNAs (lncRNAs) have been detected in DR.⁹ Among the dysregulated lncRNAs, lncNEAT1 drives high glucose (HG)-induced RPE cell EMT by upregulating SOX4 expression by binding to microRNA (miR)-204.¹⁰ Moreover, lncNEAT1 activates TGF-β1 and VEGF pathways and promotes apoptosis, oxidative stress, and inflammatory responses in HG-insulted human retinal endothelial cells.¹¹ LncNEAT1 stimulates HG-induced angiogenesis in retinal endothelial cells partially by regulating miR-125b-5p/SOX7 axis.¹² LncRNAs post-transcriptionally modulates gene expression based on the competitive endogenous RNA (ceRNA) mechanism, that is to compete with target mRNAs for binding shared miRNAs.^13–15 These previous findings suggest that lncNEAT1 regulates different types of retinal cells in DR through different ceRNA networks. This study focused on the role of lncNEAT1 in RPE and explored other action mechanisms of lncNEAT1 in DR. 

Relevant results have revealed that hypoxia inducible factor 1α (HIF-1α) affects angiogenesis and inflammatory response in diabetic retinas.^16,17 Reportedly, lncNEAT1 stimulates epithelial cell invasion and migration by inducing HIF-1α activation.¹⁸ On this basis, the present study predicted shared miRNAs of lncNEAT1 and HIF-1α and found that both lncNEAT1 and HIF-1α have binding sites to miR-320a. miR-320a expression is downregulated in islets from individuals with type 2 diabetes and inversely relates to glucagon secretion.¹⁹ miR-320a could also improve the viability of Müller cells which are responsible for neural retinal structure maintenance, retinal neuron survival, and information processing.²⁰ We thereby speculated that lncNEAT1/miR-320a/HIF-1α might act as a ceRNA network to affect the invasion, migration, and permeability of RPE cells. HIF-1α has been shown to promote the permeability, invasion, and migration of ARPE-19 cells by activating ANGPTL4 to upregulate the phosphorylation level of STAT3 and reduce ZO-1 and occludin levels.²¹ In this study, we also investigated whether lncNEAT1/miR-320a/HIF-1α had effects on the ANGPTL4/phosphorylated (p)-STAT3 pathway in ARPE-19 cells. Overall, in view of the complexity of DR pathogenesis, this study validated the hypothesized mechanism behind the regulation of lncNEAT1 in a streptozocin (STZ)-induced rat diabetes model and cultured human RPE cells in vitro to provide new ideas for the future clinical treatment of DR. 

Material and Methods

ARPE-19 Cell Culture and Transfection

Human RPE cells (ARPE-19) were purchased from the ATCC (Manassas, VA, USA) and cultured in DMEM (Gibco, Grand Island, NY, USA) with 10% FBS and 1% penicillin/streptomycin at 37°C under 5% CO₂. To establish an in vitro model of DR, ARPE-19 cells were treated with 30 mM glucose for 48 hours (HG group), and control cells were treated with 5.5 mM glucose (CON group). 

The lncNEAT1 knockdown vector (lncNEAT1 shRNA), miR-320a agonist (miR-320a agomir), miR-320a antagonist (miR-320a antagomir), HIF-1α knockdown or overexpression vector (HIF-1α shRNA or pcDNA3.1-HIF-1α), and negative controls (agomir NC, antagomir NC, pcDNA3.1-NC, and NC shRNA) were purchased from GenePharma (Shanghai, China). Transfection was performed with Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA, USA) for 48 hours before subsequent experiments. 

QRT-PCR

Total RNA, extracted with TRIzol (Invitrogen), was reverse-transcribed with a PrimeScript RT kit (Takara, Tokyo, Japan), and miRNA cDNA was acquired using a miRcute Plus miRNA first strand cDNA synthesis kit (Tiangen Biotech, Beijing, China). Expression of genes was detected using a LightCycler 480 PCR instrument (Roche, Indianapolis, IN, USA) under the following reaction conditions provided by a SYBR Green Mix kit (Roche): 95°C (10 seconds), 45 cycles of 95°C (five seconds), 60°C (10 seconds), and 72°C (10 seconds), and 72°C (5 minutes). Three replicates per sample were designed for qPCR, with β-Actin or U6 as the internal reference and data analyzed with the 2^−ΔΔCt method. The primers for amplification and their sequences are detailed in Table. 

Table.

 

Primer Sequences

 

Western Blot

Cells were lysed by RIPA lysis solution (Boster, Wuhan, China) to obtain protein samples. After measurement of protein concentration using BCA kits (Beyotime Institute of Biotechnology, Jiangsu, China), protein was added into sample buffer (Beyotime Institute of Biotechnology) and then heated for three minutes for denaturation, and then protein was electrophoresed at 80 V for 0.5 hour, and after bromophenol blue entered the separation gel, at 120 V for one to two hours. A membrane was loaded with the protein in an ice bath (300 mA, one hour), followed by rinsing (one to two minutes) and blocking (60 minutes) or at 4°C overnight. Primary antibodies including β-actin (4970S, 1:1000; Cell Signaling Technology, Danvers, MA, USA), HIF-1α (14179S, 1:1000; Cell Signaling Technology), ANGPTL4 (ab196746, 1:1000; Abcam, Cambridge, MA, USA), N-cadherin (14215S, 1:1000; Cell Signaling Technology), vimentin (5741S, 1:1000; Cell Signaling Technology), STAT3 (9139S, 1:1000; Cell Signaling Technology), p-STAT3 (9145S, 1:2000; Cell Signaling Technology), ZO-1 (sc-33725, 1:500; Santa Cruz Biotechnology, Dallas, TX, USA), and occludin (91131S, 1:1000; Cell Signaling Technology) were incubated on a shaker for two hours, and then the membrane was washed thrice for 10 minutes each time. The membrane was transferred to a secondary antibody solution, incubated for one hour at room temperature, and washed three times for 10 minutes each time. Developer was dropped onto the membrane for chemiluminescent imaging of blots (Bio-Rad Life Science, Hercules, CA, USA). 

Scratch Assay

Cells (2 × 10⁶ cells/well) were seeded in six-well plates, with three replicate wells in each group, and cultured at 37°C and 5% CO₂ for 24 hours until cell monolayers were plated. The monolayer cells were scratched with a sterile pipette tip (200 µL), and after washing with PBS, cultured for another 24 hours. Cells were observed under a microscope, with photos taken and the scratch distance recorded after incubation for 0 and 24 hours, respectively. Cell mobility = (Scratch distance _{0 h} − Scratch distance _{24 hours})/Scratch distance _{0 hour}. 

Transwell Invasion Assay

Transfected ARPE-19 cells (5 × 10⁴ per well) were seeded, and 10% FBS-DMEM was added into transwell plates (Corning Inc., Corning, NY, USA) precoated with Matrigel (0.1 mL/well, 200 µg/mL; BD Bioscience, Franklin Lakes, NJ, USA). After culture for 24 hours (5% CO₂, 37°C), invading cells in each group were fixed with 4% paraformaldehyde for 10 minutes, stained with 0.5% crystal violet for 10 minutes at room temperature, and counted in five randomly selected areas under a light microscope. 

FITC-Dextran Assay

ARPE-19 cells (1 × 10⁴) were cultured in transwell chambers (Corning) for48 h. After removal of the medium, 0.01% FITC-dextran (Sigma-Aldrich, St. Louis, MO, USA) was added to the upper chamber, and medium in the lower chamber after 60 minutes was taken for fluorescence intensity measurement by a fluorescence plate reader (Molecular Devices, San Jose, CA, USA), with the intensity normalized to the control group. 

RNA Immunoprecipitation (RIP)

A RIP kit (Millipore, Burlington, MA, USA) was purchased to assess the binding to Ago2. ARPE-19 cells were washed with cold PBS and lysed with an equal volume of RIPA lysis buffer (Beyotime Institute of Biotechnology) on ice for five minutes. Cell lysate was spun in a centrifuge at 14000 rpm and 4°C for 10 minutes to obtain the supernatant. The supernatant was separated into two fractions for input and antibody incubation. For immunoprecipitation, 50 µL magnetic beads resuspended in 100 µL RIP Wash Buffer were incubated with 5 µg Ago2 antibody (ab186733, 1:50; Abcam) or IgG antibody (negative control; ab172730, 1:100; Abcam) for 30 min; 900 µL RIP Wash Buffer containing rinsed bead-antibody complexes was mixed with 100 µL cell supernatant at 4°C overnight; bead-antibody-antigen complexes were collected using a magnetic rack. The collected complexes and the input were digested with proteinase K for RNA extraction and qRT-PCR analyses. 

Dual-Luciferase Reporter Assay

The sites of miR-320a binding to HIF-1α and miR-320a binding to lncNEAT1 were predicted by online software starBase (http://starbase.sysu.edu.cn/), followed by mutation designing and sequence synthesis. The wild or mutated sequence of the binding site (WT-HIF-1α or Mut-HIF-1α; WT-NEAT1 or Mut-NEAT1) was inserted into the luciferase reporter gene vector (pGL3-Promoter) and then cotransfected into HEK-293T cells (Shanghai Sixin Biotechnology, Shanghai, China) with 30 nM miR-320a agomir or agomir NC, followed by measurement of relative luciferase activity (i.e., firefly luciferase activity/Renilla luciferase activity). 

Animal Experiment

Forty male Sprague-Dawley rats (200 ± 20 g; Shanghai Laboratory Animal Center, Chinese Academy of Sciences, Shanghai, China) were housed under constant temperature (22°C–26°C) and humidity (55% ± 5%) conditions in a sterile laminar flow room of specific pathogen-free grade. The 40 rats were randomized into CON, DR, DR-sh-NC, and DR-sh-lncNEAT1 groups (n = 10/group). The use of animals complied with regulations and procedures for laboratory animal management and relevant ethical requirements, with ratification by the ethics committee of the First Affiliated Hospital of Harbin Medical University. Diabetes was induced in rats in the DR, DR-sh-NC, and DR-sh-lncNEAT1 groups by two intraperitoneal injections (on day 1 and day 4) of citric acid buffer (pH 4.5, 0.01 M) containing STZ (S0130, 65 mg/kg; Sigma-Aldrich) in a fasting state (fasting for 12 hours before injection), with rats in the control group receiving equal injections of citric acid buffer. Blood glucose was measured every two days after STZ injection, and rats were considered to have diabetes when blood glucose was greater than 16.7 mmol/L. After successful induction of diabetes, rats in the DR-sh-NC and DR-sh-lncNEAT1 groups were injected with lentivirus-packaged NC shRNA or lncNEAT1 shRNA (1 µL, 10⁹ PFU/mL) into their bilateral vitreous cavities on day 7 for intervention of lncNEAT1,^22,23 and an equal volume of saline solution was administered to rats in the DR group. Four weeks later, the rats were euthanized. The retinal basal tissues of the left eye were collected for hematoxylin and eosin (H&E) staining, and those of the right eye for qRT-PCR or Western blot detection. 

H&E Staining of Retinal Tissue

The removed rat eyeballs were fixed with 4% paraformaldehyde overnight at 4°C. Retinas and scleras were dehydrated with graded ethanol solution, embedded in paraffin, and then sliced into 5-µm sections with a microtome. After dewaxing and rehydration, the sections were dyed with hematoxylin for three to five minutes and rinsed with deionized water before treatment with 1% hydrochloric acid ethanol for 20 seconds, and 1% ammonia for 30 seconds. After rinsing with deionized water, the sections were counterstained with 1% eosin solution for five minutes, followed by dehydration (75%, 90%, 95%, and absolute ethanol, five minutes each), clearing (xylene, 10 minutes × 2 times), and microscopy (Olympus, Tokyo, Japan). 

Statistical Analysis

Statistical analysis was conducted with GraphPad Prism 8 software, with all data shown as mean ± SD. T-testing, one-way analysis of variance, and Tukey's multiple comparisons test were used for two-group, multi-group, and post-hoc multiple comparisons, respectively, with P < 0.05 as statistically significant. 

Results

Silencing lncNEAT1 Inhibited HIF-1α Expression and Reduced Permeability and Migration of ARPE-19 Cells

CeRNA regulatory networks that might play a role in DR were analyzed in previous studies, and relevant results revealed that lncNEAT1 and HIF-1α are correlated with DR.^16,17,24 An HG environment promotes the permeability and migration ability of human RPE ARPE-19 cells.⁶ Therefore ARPE-19 was selected to study DR at the cellular setting. LncNEAT1 and HIF-1α expression in ARPE-19 cells in an HG environment was tested by qRT-PCR, and the results exhibited that lncNEAT1 and HIF-1α were definitely elevated in ARPE-19 cells after HG treatment (Fig. 1A, P < 0.01). ARPE-19 cells transfected with lncNEAT1 shRNA or NC shRNA were cultured in an HG environment and named as HG, HG + sh-NEAT1, and HG + sh-NC groups. For screening the most effective lnNEAT1 shRNA, three shRNAs were designed and sh-NEAT1 (2) had the best effect, so the experiment was subsequently performed using this shRNA (Fig. 1B). The qRT-PCR unveiled no significant differences in lncNEAT1 and HIF-1α expression between the HG + sh-NC and HG groups (Fig. 1C) but remarkable decreases in lncNEAT1 and HIF-1αin HG-treated ARPE-19 after sh-NEAT1 transfection (Fig. 1C, P < 0.05), illustrating that lncNEAT1 was involved in the regulation of HIF-1α expression. Furthermore, the invasion (Fig. 1D, P < 0.01), migration (Fig. 1E, P < 0.05), and permeability (Fig. 1F, P < 0.05) of ARPE-19 cells were definitely enhanced by HG and restrained after lncNEAT1 knockdown, but no significant differences were observed between the HG + sh-NC and HG groups. Western blot results exhibited that ZO-1 and occludin (permeability-associated proteins) levels were definitely lower and those of N-cadherin and vimentin (migration-associated proteins) were definitely higher in HG-stimulated ARPE-19 cells than in control cells (Fig. 1G, P < 0.05). However, ZO-1 and occludin levels were markedly increased and HIF-1α, N-cadherin, and vimentin were markedly decreased in HG-treated ARPE-19 after lncNEAT1 knockdown (Fig. 1G, P < 0.05). The above observation revealed that HG upregulated lncNEAT1 and HIF-1α expression and enhanced the invasion, migration, and permeability of ARPE-19 cells, while silencing lncNEAT1 inhibited HIF-1α expression and reduced the permeability and migration of ARPE-19 cells in an HG environment. 

Figure 1.

 

Silencing lncNEAT1 suppressed HIF-1α expression and reduced the permeability and migration of ARPE-19 cells. (A) LncNEAT1 and HIF-1α RNA levels in ARPE-19 cells were detected by qRT-PCR; (B) after ARPE-19 cells were transfected with lncNEAT1 shRNA or NC shRNA, the most effective shRNA for knocking down lncNEAT1 was screened out by qRT-PCR; (C) NEAT1 and HIF-1α expression levels in ARPE-19 cells were detected by qRT-PCR; (D) cell invasion was evaluated by transwell assay; (E) cell migration was determined by scratch assay; (F) cell permeability was analyzed by FITC-dextran staining; (G) the expression levels of ZO-1, occludin, N-cadherin, and vimentin were detected by Western blot. n.s represents no significant difference, *P < 0.05, **P < 0.01, and ***P < 0.001. The experiment was repeated three times. CON group: cells were treated with 5.5 mM glucose as a control; HG group: cells were treated with 30 mM glucose; HG + sh-NEAT1 group: cells were transfected with lncNEAT1 shRNA in an HG environment; HG + sh-NC group: cells were transfected with negative control of sh-NEAT1 (NC shRNA) in an HG environment.

LncNEAT1 and HIF-1α Competitively Bound miR-320a

Increasing evidence has supported that diverse noncoding RNAs, including small noncoding RNAs, pseudogenes, lncRNAs, and circRNAs, have ceRNA activity, among which lncRNAs have become competitive platforms for miRNAs and mRNAs.²⁵ Therefore, after validating the association between lncNEAT1 and HIF-1α, we speculated that lncNEAT1 regulated HIF-1α expression through ceRNA. StarBase website prediction showed a binding site for miR-320a in HIF-1α and a targeting site between lncNEAT1 and miR-320a (Fig. 2A), so we further speculated that lncNEAT1/miR-320a might act through a ceRNA network to jointly regulate HIF-1α expression and ARPE-19 cells. qRT-PCR exhibited that miR-320a in the HG group was definitely reduced versus that of the CON group (Fig. 2B, P < 0.01). RIP assay exhibited that lncNEAT1, miR-320a, and HIF-1α were enriched by Ago2 antibody, whereas lncNEAT1, miR-320a, and HIF-1α were hardly detected with IgG antibody (Fig. 2C, P < 0.001). To further validate the crosstalk of lncNEAT1, miR-320a, and HIF-1α, dual-luciferase assays were performed, and the results exhibited no difference in luciferase activity between the Mut-HIF-1α or Mut-NEAT1 groups (Figs. 2D, 2E), whereas the luciferase activity in HEK-293T transfected with WT-HIF-1α or WT-NEAT1 was definitely lowered by miR-Agomir co-transfection (Figs. 2D, 2E; P < 0.05). Shortly, both lncNEAT1 and HIF-1α could bind to miR-320a. 

Figure 2.

 

LncNEAT1 and HIF-1α competitively bound miR-320a. (A) StarBase predicted the binding of lncNEAT1 with miR-320a and the binding of miR-320a with HIF-1α; (B) miR-320a expression in ARPE-19 cells was measured by qRT-PCR; (C) the binding of lncNEAT1 and HIF-1α with miR-320a was tested by RIP assay; (D) the binding of lncNEAT1 with miR-320a was determined by dual-luciferase reporter assay; (E) the binding of miR-320a to HIF-1α was determined by dual-luciferase reporter assay; (F) after transfection of NEAT1 shRNA and miR-320a agomir alone or together in ARPE-19 cells, qRT-PCR was used to detect lncNEAT1 and miR-320a expression; (G) after transfection of miR-320a agomir and pcDNA3.1-HIF-1α alone or together in ARPE-19 cells, qRT-PCR was used to detect lncNEAT1 and miR-320a expression. n.s represents no significant difference, *P < 0.05, **P < 0.01, ***P < 0.001, and the experiment was repeated three times. CON group: cells were treated with 5.5 mM glucose as a control; HG group: cells were treated with 30 mM glucose; HG + sh-NEAT1 + miR-Antagomir group: cells were co-transfected with lncNEAT1 shRNA and miR-320a antagomir in an HG environment; HG + sh-NEAT1 + Antagomir NC group: cells were co-transfected with lncNEAT1 shRNA and negative control of miR-320a antagomir (antagomir NC) in an HG environment; HG + sh-NC + Antagomir NC group: cells were co-transfected with negative control of sh-NEAT1 (sh-NC) and negative control of miR-320a antagomir (antagomir NC) in an HG environment.

NEAT1 shRNA was transfected alone or together with miR-320a antagomir into ARPE-19 cells to further demonstrate the regulatory relationship between lncNEAT1/miR-320a and HIF-1α. The qRT-PCR demonstrated no significant difference in lncNEAT1 and miR-320a expression between the HG + sh-NC + Antagomir NC and HG groups (Fig. 2F); however, miR-320a antagomir reversed the decrease in lncNEAT1 expression (Fig. 2F; P < 0.01) and the elevation in miR-320a expression (Fig. 2F; P < 0.01) in sh-NEAT1-transfected ARPE-19 cells under HG conditions. The above results further illustrated that miR-320a was a target gene of lncNEAT1 and that lncNEAT1 negatively regulated miR-320a expression. 

miR-320a agomir and pcDNA3.1-HIF-1α were transfected alone or together into ARPE-19 cells. The qRT-PCR revealed no significant difference in miR-320a and HIF-1α expression between the HG and HG + Agomir NC+pcDNA3.1 group (Fig. 2G); however, miR-320a was definitely increased and HIF-1α was decreased in HG-stimulated ARPE-19 after miR-320a agomir treatment (Fig. 2G; P < 0.01); pcDNA3.1-HIF-1α transfection resulted in no significant difference in miR-320a expression (Fig. 2G), but it definitely elevated HIF-1α expression in the presence of miR-320a agomir (Fig. 2G; P < 0.05). Those results indicated that HIF-1α was a target gene of miR-320a and negatively regulated by miR-320a. Altogether, lncNEAT1 and HIF-1α could competitively bind to miR-320a, and lncNEAT1 could upregulate HIF-1α by inhibiting miR-320a. 

Overexpression of HIF-1α Activated the ANGPTL4/p-STAT3 Pathway to Promote Permeability and Migration of ARPE-19 Cells

HIF-1α promotes the permeability, invasion, and migration of ARPE-19 cells by activating ANGPTL4 to upregulate the phosphorylation level of STAT3 and reduce ZO-1/occludin protein levels.²¹ For verifying the activation of the pathway, ANGPTL4 expression in ARPE-19 cells under normal or HG conditions was determined using qRT-PCR. ANGPTL4 expression in the HG group was markedly higher than in the CON group (Fig. 3A; P < 0.01). Western blot presented that HIF-1α and ANGPTL4 expression was definitely increased, STAT3 was not significantly different, and p-STAT3 was definitely increased in the HG group versus the CON group (Fig. 3B; P < 0.05). Those findings illustrated the activation of the ANGPTL4/p-STAT3 pathway. After transfection with HIF-1α shRNA or NC shRNA, ARPE-19 cells were cultured in an HG environment. Additionally, in the designed three HIF-1α shRNAs, sh-HIF-1α (1) had the best effect, and subsequent experiments were performed using this shRNA (Fig. 3C). Levels of HIF-1α, ANGPTL4, and p-STAT3 proteins were definitely decreased after sh-HIF-1α transfection, with insignificant changes in STAT3 expression (Figs. 3D, 3E; P < 0.05). The scratch, transwell, and permeability assays showed the invasion (Fig. 3F; P < 0.01), migration (Fig. 3G; P < 0.05), and permeability (Fig. 3H; P < 0.05) of HG-treated ARPE-19 cells were markedly reduced after HIF-1α knockdown. Moreover, ZO-1 and occludin were markedly elevated, and N-cadherin and vimentin were markedly decreased after HIF-1α knockdown (Fig. 3I, P < 0.05). Sh-NC transfection resulted in no significant change in HG-stimulated ARPE-19 (Figs. 3D–I). These results indicated that HG promoted the permeability and migration of ARPE-19 cells by upregulating HIF-1α expression and activating the ANGPTL4/p-STAT3 pathway. 

Figure 3.

 

Overexpression of HIF-1α activated the ANGPTL4/p-STAT3 pathway to promote permeability and migration of ARPE-19 cells. (A) ANGPTL4 expression in ARPE-19 cells was detected by qRT-PCR; (B) HIF-1α, ANGPTL4, STAT3, and p-STAT3 levels were detected by Western blot; (C) HIF-1α expression was detected by qRT-PCR to screen the most effective shRNA sequence for HIF-1α. Next, ARPE-19 cells were cultured in an HG environment after transfection with HIF-1α shRNA or NC shRNA. (D) HIF-1α and ANGPTL4 levels were detected by qRT-PCR; E: HIF-1α, ANGPTL4, STAT3, and p-STAT3 levels were detected by Western blot; (F) cell invasion was detected by transwell assay; (G) cell migration was detected by scratch assay; (H) cell permeability was analyzed by FITC-dextran staining; I: ZO-1, occludin, N-cadherin, and vimentin levels were detected by Western blot. n.s represents no significant difference, *P < 0.05, and **P < 0.01. The experiment was repeated three times. CON group: cells were treated with 5.5 mM glucose as a control; HG group: cells were treated with 30 mM glucose; HG + sh-HIF-1α group: cells were transfected with HIF-1α shRNA in an HG environment; HG + sh-NC group: cells were transfected with negative control of sh-HIF-1α (NC shRNA) in an HG environment.

miR-320a Overexpression Downregulated HIF-1α to Restrain ANGPTL4/P-STAT3 Pathway Activation and ARPE-19 Cell Migration

ARPE-19 cells were treated with miR-320a agomir, pcDNA3.1-HIF-1α, or corresponding NC vectors in an HG environment. The qRT-PCR and Western blot results presented no significant difference in lncNEAT1, miR-320a, HIF-1α, and ANGPTL4, nor in STAT3 and p-STAT3 between the HG + Agomir NC + pcDNA3.1 and HG group (Figs. 4A, 4B); lLncNEAT1, HIF-1α, ANGPTL4, and p-STAT3 were modestly expressed and miR-320a was definitely increased after miR-320a agomir transfection, whereas STAT3 was not significantly different (Figs. 4A, 4B; P < 0.05). PcDNA3.1-HIF-1α cotransfection counteracted miR-320a agomir-induced inhibition of HIF-1α, ANGPTL4, and p-STAT3, but it did not affect lncNEAT1, miR-320a, and STAT3 (Figs. 4A, 4B; P < 0.05). 

Figure 4.

 

Overexpression of miR-320a downregulated HIF-1α and inhibited the migration ability of ARPE-19 cells. After ARPE-19 cells were transfected with miR-320a agomir, pcDNA3.1-HIF-1α, or the corresponding negative control vector in an HG environment, (A) lncNEAT1, miR-320a, HIF-1α, and ANGPTL4 expression levels in ARPE-19 cells were detected by qRT-PCR; (B) HIF-1α, ANGPTL4, STAT3, and p-STAT3 proteins were detected by Western blot; (C) cell invasion was measured by transwell assay; (D) cell migration was measured by scratch assay; (E) cell permeability was analyzed by FITC-dextran staining; (F) ZO-1, occludin, N-cadherin, and vimentin protein levels were detected by Western blot. n.s represents no significant difference, *P < 0.05, and **P < 0.01. The experiment was repeated three times. HG group: cells were treated with 30 mM glucose; HG + miR-Agomir + pcDNA3.1-HIF-1α group: cells were co-transfected with miR-320a agomir and pcDNA3.1-HIF-1α in an HG environment; HG + miR-Agomir + pcDNA3.1 group: cells were co-transfected with miR-320a agomir and negative control of pcDNA3.1-HIF-1α (NC pcDNA3.1) in an HG environment; HG + Agomir NC + pcDNA3.1 group: cells were co-transfected with negative control of miR-320a agomir (agomir NC) and negative control of pcDNA3.1-HIF-1α (NC pcDNA3.1) in an HG environment.

The scratch, transwell, and permeability assays demonstrated reductions in the invasion, migration, and permeability of ARPE-19 cells overexpressing miR-320a, which were counteracted by HIF-1α co-overexpression (Figs. 4C–E; all P < 0.05). Western blot showed ZO-1 and occludin levels were obviously elevated and N-cadherin and vimentin were decreasingly expressed in ARPE-19 overexpressing miR-320a; these alterations were nullified by HIF-1α co-overexpression (Fig. 4F, P < 0.05). Overall, miR-320a overexpression could downregulate HIF-1α expression and inhibit ANGPTL4/p-STAT3 pathway activation and migration ability of ARPE-19 cells in an HG environment. 

LncNEAT1/miR-320a/HIF-1α Axis Activated the ANGPTL4/P-STAT3 Pathway To Promote the Migration Ability of ARPE-19 Cells

To verify whether lncNEAT1 could inhibit miR-320a and upregulate HIF-1α expression to activate the ANGPTL4/p-STAT3 pathway, we transfected ARPE-19 cells with lncNEAT1 shRNA, miR-320a antagomir, or the corresponding NC vector in an HG environment. qRT-PCR and Western blot revealed lncNEAT1, HIF-1α, ANGPTL4, and p-STAT3 levels were definitely decreased and miR-320a was increased after sh-NEAT1 transfection, whereas STAT3 was not significantly different (Figs. 5A, 5B; P < 0.05); miR-320a was markedly decreased, HIF-1α, lncNEAT1, and ANGPTL4 were expressed highly, and p-STAT3 protein was markedly increased in ARPE-19 co-transfected with sh-NEAT1 and miR-320a antagomir, with STAT3 not significantly affected (Figs. 5A, 5B; P < 0.05). 

Figure 5.

 

LncNEAT1/miR-320a/HIF-1α axis activated the ANGPTL4/p-STAT3 pathway to promote the migration ability of ARPE-19 cells. After ARPE-19 cells were transfected with lncNEAT1 shRNA, miR-320a antagomir, or the corresponding negative control vector in an HG environment, (A) lncNEAT1, miR-320a, HIF-1α, and ANGPTL4 levels in ARPE-19 cells were detected by qRT-PCR; (B) HIF-1α, ANGPTL4, STAT3, and p-STAT3 protein levels were detected by Western blot; (C) cell invasion was measured by transwell assay; (D) cell migration was measured by scratch assay; (E) cell permeability was analyzed by FITC-dextran staining; (F) ZO-1, occludin, N-cadherin, and vimentin levels were detected by Western blot. n.s represents no significant difference, *P < 0.05, and **P < 0.01. The experiment was repeated three times. CON group: cells were treated with 5.5 mM glucose as a control; HG group: cells were treated with 30 mM glucose; HG + sh-NEAT1 + miR-Antagomir group: cells were co-transfected with lncNEAT1 shRNA and miR-320a antagomir; HG + sh-NEAT1 + Antagomir NC group: cells were co-transfected with lncNEAT1 shRNA and negative control of miR-320a antagomir (antagomir NC) in an HG environment; HG + sh-NC + Antagomir NC group: cells were co-transfected with negative control of sh-NEAT1 (sh-NC) and negative control of miR-320a antagomir (antagomir NC) in an HG environment.

The scratch, transwell, and permeability assays exhibited that the invasion, migration, and permeability of ARPE-19 cells were definitely reduced after lncNEAT1 knockdown, which was reversed by miR-320a silencing (Figs. 5C–E, all P < 0.05). Western blot exhibited that the levels of ZO-1 and occludin were definitely elevated and N-cadherin and vimentin were diminished in ARPE-19 cells underexpressing lncNEAT1; these protein expression changes were nullified in ARPE-19 cells with both silenced lncNEAT1 and miR-320a (Fig. 5F; P < 0.05). No significant difference was found between the HG + sh-NC + Antagomir NC and HG groups (Figs. 5A–F). In brief, the lncNEAT1/miR-320a/HIF-1α axis promoted the invasion, migration, and permeability of HG-insulted ARPE-19 cells by activating the ANGPTL4/p-STAT3 pathway. 

Silencing lncNEAT1 Upregulated miR-320a and Inhibited HIF-1α Expression To Alleviate Retinopathy in Diabetic Rats

Afterward, we intended to investigate whether the lncNEAT1/miR-320a/HIF-1α axis could also work in a rat model of DR. LncNEAT1 was intervened by injecting lentiviral lncNEAT1 shRNA or NC shRNA into the vitreous cavities of model rats to verify whether diabetes-induced retinopathy could be reversed or treated to some extent. The H&E staining demonstrated that the retinal morphology of the CON group was normal; in the DR and DR + sh-NC groups, the retinal structures were disordered, the outer retinal layer showed a distorted morphology and obviously decreased thickness, and the outer plexiform layer, outer nuclear layer, and photoreceptor inner and outer segment (IS + OS) were thinned remarkably; however, after sh-NEAT1 treatment, the retinal pathologies were alleviated and the retinal structures became relatively neat, characterized by improved morphology and increased layer thickness (Fig. 6A). 

Figure 6.

 

Silencing lncNEAT1 upregulated miR-320a and inhibited HIF-1α expression to alleviate retinopathy in diabetic rats. (A) H&E staining of rat retinal sections; (B) lncNEAT1, miR-320a, HIF-1α, and ANGPTL4 expression levels in rat retinal tissues were detected by qRT-PCR; (C) the protein levels of HIF-1α, ANGPTL4, STAT3, p-STAT3, ZO-1, occludin, N-cadherin, and vimentin in rat retinal tissues were measured by Western blot. n.s represents no significant difference, *P < 0.05, and **P < 0.01. N = 10. NFL, nerve fiber layer; GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer; IS + OS, photoreceptor inner and outer segment. DR group: STZ was injected to induce diabetes; CON group: the same amount of citrate buffer was injected as control; DR-sh-lncNEAT1 group: lentiviral lncNEAT1 shRNA was injected into the bilateral vitreous cavities of diabetic model rats; DR-sh-NC group: lentiviral negative control shRNA was injected into the bilateral vitreous cavities of diabetic model rats.

The qRT-PCR results revealed that lncNEAT1, HIF-1α, and ANGPTL4 were definitely increased and miR-320a was decreased in the DR group versus the CON group (Fig. 6B; P < 0.01); lncNEAT1, HIF-1α, and ANGPTL4 were definitely decreased and miR-320a was definitely increased after sh-NEAT1 treatment (Fig. 6B; P < 0.01). Western blot data illustrated no significant difference in STAT3, but HIF-1α, ANGPTL4, p-STAT3, N-cadherin, and vimentin levels were definitely increased and ZO-1 and occludin were decreased in the DR group in comparison with the CON group (Fig. 6C; P < 0.01); no significant difference was seen in STAT3, but HIF-1α, ANGPTL4, p-STAT3, N-cadherin, and vimentin were definitely decreased and ZO-1 and occludin were increasingly expressed after lncNEAT1 knockdown (Fig. 6C; P < 0.01). No significant difference was found between the DR + sh-NC and DR groups (Figs. 6B, 6C). Collectively, silencing lncNEAT1 in rats could upregulate miR-320a and inhibit HIF-1α activation to alleviate DR. 

Discussion

DR, as a common eye disease caused by diabetes mellitus, is a predominant cause of preventable blindness.²⁶ The ceRNA network has been largely addressed and explored in the course of DR.^8,15 This study disclosed that lncNEAT1 sponged miR-320a to restrain miR-320a-mediated inhibition on HIF-1α, which activated ANGPTL4/p-STAT3 signaling and triggered RPE cell invasion and permeability in DR. 

Persistent hyperglycemia damages mitochondrial function in various types of retinal cells, contributing to DR.²⁷ Mitochondrial DNA (mtDNA) is epigenetically modified under hyperglycemic conditions, resulting in mitochondrial damage.²⁸ Apart from alterations in mtDNA, levels of various nuclear genome-encoded lncRNAs, such as NEAT1, MALAT1, and HOTTIP, are elevated in DR.^11,29,30 Although mitochondria have their separate set of genomes, namely mtDNA, nuclear-encoded lncRNAs can translocate into mitochondria. HG increases mitochondrial translocation of lncNEAT1 in human retinal endothelial cells, thereby damaging mitochondrial structural and genomic integrity.²⁴ Moreover, lncNEAT1 can regulate mitochondrial homeostasis through sequestration of multiple mitochondrion-related mRNAs in nuclear paraspeckles.³¹ Mitochondria produce the majority of intracellular reactive oxygen species, and the production is greatly augmented in response to stress, causing oxidative mitochondrial damage.³² Mitochondrial oxidative stress can activate several inflammatory pathways and enhance vascular permeability.³³ LncNEAT1 promotes DR by stimulating oxidative stress, VEGF, and inflammatory responses in retinal endothelial cells.¹¹ The regulation of lncNEAT1 on mitochondrial dynamics may have an impact on endothelial oxidative stress and inflammation in DR. This study focused on another important cells, RPE cells, in DR and obtained findings that lncNEAT1 promoted permeability and EMT of HG-insulted RPE cells. Considering the previous findings, the potential role of lncNEAT1 in mitochondrial homeostasis, oxidative stress, and inflammation in RPE cells should be addressed in future studies. 

HIF-1α is involved in the progressive course of DR.^16,17,34 Here, we found that HG upregulated the expression of lncNEAT1 and HIF-1α and enhanced the invasion, migration, and permeability of ARPE-19 cells, whereas silencing lncNEAT1 or inhibiting HIF-1α expression reduced the permeability and migration ability of ARPE-19 cells in an HG environment. StarBase website prediction showed miR-320a could bind to HIF-1α and lncNEAT1. Previous evidence reported that miR-320a might act as a novel biomarker for DR.³⁵ HIF-1α was one of the downstream targets of miR-320.³⁶ More recent data revealed that HIF-1α could upregulate HECT domain E3 ubiquitin-protein ligase 2, a target gene of miR-320a, in renal cells.³⁷ Preliminary results in our study indicated that lncNEAT1 and HIF-1α could competitively bind to miR-320a, and lncNEAT1 could upregulate HIF-1α expression by inhibiting miR-320a. Therefore, we further speculated that lncNEAT1/miR-320a might act as a ceRNA network to jointly regulate HIF-1α and affect the invasion, migration and permeability of ARPE-19 cells. 

Yang et al.²¹ previously revealed that HIF-1α could promote the invasion and migration of ARPE-19 cells by activating ANGPTL4 and upregulating the phosphorylation level of STAT3. Our study illustrated that overexpression of HIF-1α activated the ANGPTL4/p-STAT3 pathway to promote permeability and migration of ARPE-19 cells. Notably, overexpression of miR-320a could downregulate HIF-1α expression and inhibit ANGPTL4/p-STAT3 pathway activation and migration ability of ARPE-19 cells. Consistently, a previous study demonstrated that miR-320a effectively inhibited cell proliferation and metastasis by regulating STAT3 signals in lung cancer.³⁸ Zhou et al.³⁹ also found that miR-320/KLF5/HIF-1α axis was involved in EMT, invasion and migration of gastric cancer cells. Our further experiments observed that the lncNEAT1/miR-320a/HIF-1α axis promoted the invasion, migration, and permeability of ARPE-19 cells by activating the ANGPTL4/p-STAT3 pathway. After we explored lncNEAT1/miR-320a as a ceRNA regulatory network of HIF-1α at the cellular level, we finally verified the mechanism in a rat model. Rat experiments found that silencing lncNEAT1 could upregulate miR-320a and inhibit HIF-1α activation to alleviate DR. 

Conclusion

In conclusion, our study found that the lncNETA1/miR-320a ceRNA network could regulate HIF-1α expression to activate the ANGPTL4/p-STAT3 pathway and promote ARPE-19 cell invasion and migration in DR (Fig. 7), which may provide a novel mechanism and biomarkers for DR. However, our study still has some limitations. DR is a complex disease, and many miRNAs and lncRNAs may be involved in DR progression. In this study, we only investigated the lncNETA1/miR-320a ceRNA network, and other potential regulatory factors could not be excluded. Additionally, whether the lncNEAT1/miR-320a/HIF-1α axis has effects on other pathways also needs to be determined. Importantly, we have not yet explored whether lncNEAT1 affects DR by affecting mitochondrial dynamics, which is also a direction for further research in the future. Taken together, this study lays theoretical foundations for targeting lncNETA1/miR-320a ceRNA network in DR therapy, and further studies will be performed to address those limitations to provide more promising biomarkers and molecular mechanisms for DR. 

Figure 7.

 

High lncNEAT1 expression under HG conditions inhibited miR-320a expression and upregulated downstream target gene HIF-1α expression, ultimately increasing retinal cell permeability in diabetic rats.

Acknowledgments

The authors acknowledge and appreciate their colleagues for valuable efforts and comments on this paper. 

Disclosure: X. Zhu, None; Y. Wang, None; L. Cheng, None; H. Kuang, None