Enhanced ROBO4 is mediated by up‐regulation of HIF‐1α/SP1 or reduction in miR‐125b‐5p/miR‐146a‐5p in diabetic retinopathy

Abstract Retinal cell damage caused by diabetes leads to retinal microvascular injury. Roundabout 4 (ROBO4) is involved in angiogenesis, which varies with the development of diabetic retinopathy (DR). Here, we explored the transcriptional regulation and microRNA‐mediated modulation of ROBO4 expression and related retinal cell function in DR. A streptozotocin‐induced type I diabetic animal model was established to detect the expression of hypoxia inducible factor‐1α (HIF‐1α), specificity protein 1 (SP1) and ROBO4. Retinal pigment epithelium (RPE) cells were cultured under hyperglycaemia or hypoxia and used for mechanistic analysis. Furthermore, roles of miR‐125b‐5p and miR‐146a‐5p were evaluated, and their targets were identified using luciferase assays. The cell functions were evaluated by MTS assays, permeability analysis and migration assays. The development of DR increased the levels of HIF‐1α, SP1 and ROBO4 both in the DR model and in hyperglycaemic/hypoxic RPE cells. They were co‐expressed and up‐regulated in diabetic retinas and in RPE cells under hyperglycaemia/hypoxia. Knockdown of HIF‐1α significantly inhibited SP1 and ROBO4, whereas SP1 down‐regulation abolished ROBO4 expression in RPE cells under hyperglycaemia/hypoxia. miR‐125b‐5p and miR‐146a‐5p were down‐regulated by hyperglycaemia and/or hypoxia. Up‐regulation of miRNAs reversed these changes and resulted in recovery of target gene expression. Moreover, luciferase assays confirmed miR‐125b‐5p targeted SP1 and ROBO4, and miR‐146a‐5p targeted HIF‐1α and ROBO4 directly. The decreased cell viability, enhanced permeability, and increased cell migration under DR conditions were mitigated by knockdown of HIF‐1α/SP1/ROBO4 or up‐regulation of miR‐125b‐5p/miR‐146a‐5p. In general, our results identified a novel mechanism that miR‐125b‐5p/miR‐146a‐5p targeting HIF‐1α/SP1‐dependent ROBO4 expression could retard DR progression.


| INTRODUC TI ON
Diabetic retinopathy (DR), a microvascular complication of diabetes mellitus (DM), is the leading cause of blindness in adults worldwide. 1 DR is primarily caused by the long-term detrimental effects of high glucose, 2 leading to retinal microvascular defects and neuroretinal dysfunction and degeneration. 3,4 Notably, high glucose in early-stage DR can induce increased retinal vessel permeability, leakage of harmful substances and degeneration of capillaries, resulting in retinal hypoxia in the late stage of DR. 5,6 Hyperglycaemia and hypoxia may interact, contributing significantly to the progression of DR. 7 The blood-retinal barrier (BRB), which is comprised of the retinal vasculature and the retinal pigment epithelium, eliminates the neural elements of the retina as well as the cytotoxic products to protect the retina to modulate its extracellular chemical composition. 8 Endothelial cells harbouring tight junctions are responsible for maintaining the inner BRB, and their impairment results in enhanced vascular permeability. 9 The outer BRB, which is formed by retinal pigment epithelial (RPE) cells, plays a crucial role in fluid balance within the retina. 10 Breakdown of BRB resulted from the disruption of tight junctions is the main factor responsible for oedema and neovascularization. Therefore, protecting or reversing retinal microvascular dysfunction is fundamental for many studies of DR. However, extensive work has been carried out to identify factors involved in the disruption of the inner BRB during DR, the mechanisms implicated in outer BRB regulation have been poorly uncovered. As the molecular mechanisms underlying the pathogenesis of DR has not been fully understood, this study will focus on the role of RPE cells in DR.
Roundabout 4 (ROBO4) is specifically expressed in vascular endothelial cells and is involved in angiogenesis and the maintenance of blood vessel stability. 11,12 ROBO4 maintains the vascular integrity 13 and promotes pathological angiogenesis through various signalling pathways. 14,15 ROBO4 is also strongly overexpressed in the vessels of various types of tumours. 11,16 In retinal researches, ROBO4 expression and distribution have been studied in the fibrovascular membranes (FVMs) of patients with proliferative DR. ROBO4 is also expressed in the retinal pigment epithelium (RPE), playing important roles in RPE functions under hypoxia. 17 Thus, ROBO4 may have a role in the formation of FVMs and could exert physiologic effects on retinal cells. We previously showed that ROBO4 is co-expressed with hypoxia-inducible factor-1α (HIF-1α) in vessels of FVMs and is positively regulated by HIF-1α. 18 HIF-1α is an oxygen-sensitive transcription factor that is associated with angiogenesis during the progression of DR and FVM development. 19,20 Under conditions of low oxygen, hypoxia-induced proteins are up-regulated. 19,21,22 Thus, hundreds of proteins related to cell proliferation, survival, and angiogenesis can be activated by HIF-1α signalling pathways. 23 However, the modulatory effects of HIF-1α on ROBO4 expression are not direct.
Specificity protein 1 (SP1) and HIF-1α cooperate to promote tumour progression 24 and activate genes related to cell adaption for hypoxia. Transcriptional regulation of SP1 by HIF-1α was found to have protective functions in neurotoxicity. 25 Additionally, SP1 is necessary for full basal expression of ROBO4 in macrovascular endothelial cells. 26 DNA methylation of the proximal promoter of ROBO4 inhibits SP1 binding, inducing low ROBO4 expression in non-endothelial cells. 27 Thus, aberrant levels of ROBO4 induced by HIF-1α may be mediated via SP1 in DR.
MicroRNAs (miRNAs) are small non-coding RNAs that play important roles in the progression of DR. miRNAs modulate gene expression through transcriptional or post-transcriptional mechanisms, inducing mRNA degradation or protein regression by binding to the 3′-untranslated region (UTR) of target genes. 28,29 Here, we assessed the roles of miR-125b-5p and miR-146a-5p in HIF-1α/SP1-mediated ROBO4 expression in vivo in diabetic rats or in vitro in RPE cells under hyperglycaemia or hypoxia.

| Animal experiments
All animal experiments were conducted in accordance with the NIH were excised for preparation of retinal tissue sections (6 μm) to perform immunofluorescent staining. Retinal tissue for protein analysis and mRNA extraction was conducted in six eyes from each group. For hyperglycaemic studies, the cells were plated at 2,500 cells/cm 2 in 6-well plates (Corning, Acton, MA, USA) and treated with normal glucose (NG, 5.5 mmol/L D-glucose) or high glucose (HG, 25 mmol/L D-glucose) or with the mannitol osmotic control (MN, 19.5 mmol/L mannitol together with NG) for 1, 3, 5, or 7 days. For hypoxic experiments, cells were plated in 30-mm dishes (Corning); when the cell confluence reached 70%-80%, the medium was changed, and cells were plated in a sealed and anaerobic workstation (Ruskin Technologies, Pencoed, Wales, UK) with 1% O 2 , 5% CO 2 and 94% N 2 , and the same temperature (37°C) and humidity (90%) as that under hypoxic conditions for various times (4, 8, 16 and 24 hours).

| Cell culture and treatments
For hyperglycaemic and hypoxic study, cells were treated with NG or HG for 104 hours, and following by hypoxia and hyperglycaemia for 16 hours. All experiments were conducted at least three times.

| Quantitative analysis of mRNAs and miRNAs
For analysis of mRNA, 500 ng total RNA was reverse transcribed into cDNA using a Perfect Real Time RT reagent kit (Takara Bio, Dalian, China) in a 20-μL reaction volume. The qPCR mixture contained 1 μL cDNA, 10 μmol gene-specific primers (forward and reverse mixed together) and 10 μL of 2 × Fast SYBR Green Master Mix (Roche Diagnostics, Switzerland). Three replicates for each biological mixture were analysed on a LightCycler 480 (Roche Diagnostics). The data were normalized to the expression of housekeeping genes according to our previous study. 30 Specific primers were designed and verified in our previous studies. 30,31 Primer sequences are listed in Table S1.
In analysis of miRNAs or mRNAs, two negative controls were included with water instead of template. The relative expression levels of miRNAs or mRNAs were calculated by the 2 −ΔΔCt method, which was based on the ratio of gene expression between an experimental and control group.
The following siRNA sequences were synthesized by Bioneer For hypoxic studies, at 24 hours after passaging, cells were transfected with 50 nmol miRNA (miR-146a mimic or scramble) or 100 pmol siRNA (HIF-1α siRNA, Robo4 siRNA, or NC siRNA) using Lipofectamine RNAiMAX. Cells were treated with hypoxic conditions for 16 hours before mRNA or protein extraction. The sequence of the miR-146a-5p mimic was 5′-UGAGAACUGAAUUCCAUGGGUU-3′, and that of HIF-1α siRNA was 5′-CTGGACACAGTGTGTTTGA-3′. Firefly luciferase activity was normalized to that of Renilla luciferase for each sample.

| Monolayer permeability assay
For permeability assays, ARPE-19 cells treated with different conditions were seeded at 1 × 10 5 cells/well in the upper chamber (6.5mm diameter transwell with 0.4-μm pore polycarbonate membrane inserts; Corning) and cultured for 48 hours to reach confluence. The upper chamber was washed three times with PBS and treated with FITC-dextran (1 mg/mL; Sigma). The fluorescence intensity, equivalent to the relative amount of FITC-dextran in the lower chambers of the transwells, was measured over a 30-minutes incubation at 37°C and determined in triplicate using Varioskan Flash (excitation wavelength, 490 nm; emission wavelength, 520 nm; Thermo).

| Cell migration assay
The migratory ability of ARPE-19 cells under hyperglycaemic or hypoxic conditions was determined using the transwell system. A total of 5 × 10 3 cells from each group were seeded in the top chambers of 6.5-mm diameter transwells with 8.0-μm pore polycarbonate membrane inserts (3422; Corning) and cultured in 200 μL medium with 5% FBS. Bottom chambers were filled with 500 μL medium with 20% FBS. After incubation for 24 hours, cells on the top chamber were removed, and migrated cells were fixed with 4% paraformaldehyde and stained in 0.1% crystal violet solution. Images were captured, and cells were counted in five fields at 10 × magnification. Quantification was performed using Image J software.

| Statistical analysis
Data are presented as means ± standard deviations from at least three independent experiments. One-way analysis of variance was used for multiple comparisons and followed by the Student-Newman-Keuls post-hoc test to assess the statistical differences between groups. Comparisons between two groups were made using two-tailed Student's t tests (GraphPad Prism 6.0; GraphPad Prism, San Diego, CA, USA). Two-sided P < 0.05 were considered statistically significant.

| HIF-1α, SP1 and ROBO4 were up-regulated concomitantly in the retinas of diabetic animals
To confirm variations in HIF-1α, SP1 and ROBO4 expression in the development of early DR, we used an in vivo STZ-induced diabetic rat model. Blood glucose levels were markedly elevated in STZtreated rats during the first week after injection. These DM rats also exhibited significant weight loss that persisted over the course of DR progression as compared with age-matched control rats (Table   S2). Western blotting showed high levels of HIF-1α, SP1 and ROBO4 after 4 weeks of uncontrolled diabetes, with levels sustained at 6 and 8 weeks ( Figure 1A-D). HIF-1α and ROBO4 were up-regulated beginning at 4 weeks, whereas SP1 tended to increase at week 4 and was significantly overexpressed at week 6.
Increases in these three genes of more than 1.5-fold were observed in the hyperglycaemic group at 5-7 days. Changes in HIF-1α, SP1 and ROBO4 protein levels were consistent with corresponding mRNA levels ( Figure 2D-I). Thus, subsequent experiments of hyperglycaemia were performed using 5-day treatment with high glucose.

| HIF-1α, SP1 and ROBO4 were coexpressed and up-regulated in ARPE-19 cells under hyperglycaemic conditions
Double immunofluorescence staining of high-glucose-treated ARPE-19 cells showed that HIF-1α was weakly detected in the nucleus and cytoplasm of RPE cells under NG or MN and was strongly enhanced under HG ( Figure 3A). ROBO4 showed weak immunofluorescence in the cytoplasm and cytomembrane of RPE cells under NG or MN, and up-regulation was observed after treatment with HG.
SP1 was expressed in the nucleus of RPE cells, with up-regulation observed after culture in HG ( Figure 3B). Co-staining with ROBO4 showed strong fluorescence in the cytoplasm and cytomembrane of RPE cells in HG compared with that in NG or MN. These results confirmed the localization and up-regulation of HIF-1α, SP1 and ROBO4 in RPE cells exposed to high glucose.

| Hypoxia induced up-regulation of HIF-1α, SP1 and ROBO4 and down-regulation of miR-146a-5p in ARPE-19 cells
We then investigated the detailed changes in HIF-1α, SP1 and ROBO4 during different stage of DR using ARPE-19 cells under hypoxic conditions. As shown in Figure 4A, HIF-1α mRNA levels decreased during the first 8 hours of hypoxia and then increased.  Figure 4B and 4C). This discrepancy between HIF-1α mRNA and protein suggested that HIF-1α protein may be regulated by changes to protein stability and accumulation. 33  We then analysed the expression of another specific miRNA, miR-146a-5p, which may target the 3′-UTRs of HIF-1α and ROBO4 directly. Notably, miR-146a-5p has dual roles in nuclear factor-κB (NF-κB)-mediated inflammatory pathways in DR. 34 This miRNA is transactivated by NF-κB and exerts negative feedback on NF-κB activation. miR-146a-5p overexpression has protective effects on

F I G U R E 3 Abundance and localization of HIF-1α, SP1 and ROBO4 in ARPE-19 cells under hyperglycaemic conditions. Dual immunostaining of HIF-1α (red) and ROBO4 (green) A or SP1 (red) and ROBO4 (green) B, merged with DAPI (blue) in RPE cells under NG, HG and MN
high-glucose-treated endothelial cells and retinas of STZ-induced diabetic rats. 35 However, the role of miR-146a-5p in RPE cells under diabetic conditions has not been investigated. Our results showed that miR-146a-5p was down-regulated in RPE cells exposed to hypoxia for 16 hours (Figure 4D), and HIF-1α and ROBO4 were elevated, suggesting that miR-146a-5p and HIF-1α/ROBO4 may interact in ARPE-19 cells.
As high glucose and hypoxia are the two main initiators relating to DR, we performed parallel experiments on RPE cells exposed to a combined insult of high glucose and hypoxia. Then, the expression levels of miR-125b-5p and miR-146a-5p were detected. Not surprisingly, miR-125b-5p and miR-146a-5p were found decreased by 2fold in RPE cells induced by hyperglycaemia and hypoxia ( Figure S1).
These results confirmed that miR-125b-5p and miR-146a-5p were down-regulated in different stages of DR, and regulation on them may ameliorate the progression of DR.

| HIF-1α promoted ROBO4 expression by regulating SP1 in ARPE-19 cells under hyperglycaemic or hypoxic conditions
To explore the modulatory relationship of HIF-1α, SP1 and ROBO4 in DR, siRNA was used to silence the expression of these genes.
Notably, continuous HG treatment significantly up-regulated HIF-1α, SP1 and ROBO4 mRNAs, with no distinct differences between HG and the negative transfection group ( Figure 5A-C). The HG-induced > 1.5-fold increases in HIF-1α, SP1 and ROBO4 mRNAs were suppressed by transfection with specific siRNAs, showing decreases in more than 2-fold, lower than that in the NG group. Furthermore, SP1 was repressed with HIF-1α down-regulation, and ROBO4 was

| miR-125b-5p targeted SP1 and ROBO4 directly in ARPE-19 cells under hyperglycaemic conditions
We then explored the functional significance of hyperglycaemia-in-

| Down-regulation of ROBO4 through transcriptional repression or miRNA targeting improved cell functions in ARPE-19 cells under hyperglycaemic or hypoxic conditions
The C, D, Western blots and quantification of SP1 and ROBO4 in ARPE-19 cells under HG following transfection with miR-125b-5p mimic. β-Actin was used as a loading control. Scr, scramble control; miR-125b, miR-125b-5p mimic. All groups, n = 3; *P < 0.05; **P < 0.01; ***P < 0.001 versus NG; # P < 0.05; ## P < 0.01 versus scr. E, F, Base-pair comparison between mature miR-125b-5p and the WT or MUT putative target site in the 3′-UTR of SP1 or ROBO4 mRNA. The mutated binding site used for the luciferase assay is marked in red. Hsa-, human. Luciferase activity with various reporters was detected in the presence or absence of miR-125b-5p mimic in HEK293 cells (n = 3), *P < 0.05; ***P < 0.001 versus MUT plasmid or vector. WT, wild-type; MUT, mutant SP1 or ROBO4 siRNA or by miR-125b-5p mimic compared with that of mock transfection ( Figure S3). Furthermore, ZO-1, Occludin and Claudin-1 were down-regulated in RPE cells under hypoxic conditions. Knockdown of HIF-1α or ROBO4 or overexpression of miR-146a-5p up-regulated these tight junction-related proteins in RPE cells exposed to hypoxia ( Figure S4). Accordingly, these findings justified that modulation of ROBO4 by HIF-1α/SP1 knockdown or miRNA targeting had a protective role in RPE cell permeability under diabetic conditions.

| D ISCUSS I ON
Here, we found that there was a significant increase in ROBO4 ex-  29,34,43,44 In previous studies, miR-146a-5p was down-regulated and its overexpression played a protective role in DR through modulation of fibronectin and NF-κB, which are involved in inflammatory pathways. 35,47 Because one miRNA can target multiple genes, more genes associated with the progression of DR have been shown to be regulated by the same miRNA, supporting the potential therapeutic applications of miRNAs.
In this study, DR promoted retinal permeability in diabetic rats and RPE cells under hyperglycaemia or hypoxia. The levels of tight junction-related proteins were decreased during DR in diabetic retinas, consistent with a previous study, 48

ACK N OWLED G EM ENTS
This work was supported by grants from the National Natural Science Foundation of China (grant no. 81670871). The funder had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. This study was supported by all of the members of the Central Laboratory of the Second Hospital at Jilin University.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest associated with this manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.