The microRNA-7-mediated reduction in EPAC-1 contributes to vascular endothelial permeability and eNOS uncoupling in murine experimental retinopathy

C57BL/6J mice and spontaneous diabetic heterozygous Ins2Akita^+/− mice (Jackson Laboratory, Charles River, Sulzfeld, Germany) were used throughout the study. Age-matched non-diabetic homozygous Ins2Akita^−/− mice served as control. All experimental procedures were performed according to the guidelines of the statement for animal experimentation issued by the Association for Research in Vision and Opthalmology and were approved by the local board for animal care (Medical Faculty Mannheim, Germany).

Oxygen-induced retinopathy was induced in C57BL/6J mice as described previously [22, 23]. In short, newborn mice (n = 6) at postnatal day (p) 7 were exposed to hyperoxia (75% oxygen) in an incubation chamber (Stuart Scientific, Redhill, UK) with their nursing mothers for 5 days and then returned to ambient air, creating a relative hypoxic environment. Control mice (n = 6) were kept at ambient air and used as a control group. Mice were killed at p12 (i.e., 6 h of relative hypoxia) and p13 (i.e., 24 h of relative hypoxia), and the retinas were isolated as described previously [23].

After 6 months of diabetes, mice (n = 6/group) were killed and the retinas were isolated as described previously [23].

Parallel cultures were maintained under normoxic (21% oxygen tension) and hypoxic conditions (2% oxygen tension) for 48 h prior to experiments. To establish hypoxia, cell culture medium was deoxygenated by bubbling gaseous N₂ through the medium at room temperature for 30 min. Cells were maintained in an hypoxic cell culture incubator at 37 °C containing 2% O₂, 93% N₂, and 5% CO₂.

Construction of 3′UTR reporter constructs and microRNA-7-5p transfection in COS7 and Endothelial cells. The 3′UTR fragments of EPAC-1 and EPAC-2 were isolated by conventional PCR amplification. 5′ SgfI- and 3′ NotI restriction sites (underlined) were incorporated in the primer sequences; EPAC-1 sense: 5′-CCGCCGGCGATCGCAGGAGTGGGTGGAGAGTGGA-3′ and antisense: 5′-CATGCGGCCGCGTGTCCCCACCCACGGCAAG-3′ and EPAC-2 sense: 5′-ATATATGCGATCGCACATTTCAAATGCCCAAAGC-3′ and antisense: 5′-GCAGCGGCCGCATTGAATGAACTATTTACAA-3′. Amplicons were isolated using the QIAquick Gel Extraction Kit (Qiagen Inc) according to manufacturer’s instructions, modified using SfgI and NotI restriction enzymes (Fermentas) and inserted in the psiCHECK-2 vector (Promega) using T4 ligase.

MicroRNA-7 mimics and scrambled sequences were obtained from ThermoFisher Scientific and co-transfected with 3′UTR reporter constructs into COS-7 cells using EndoFectin (GeneCopoeia, Rockville, MD). Co-transfection of miR-7 mimics and an 3′UTR-free psiCHECK-2 plasmid were used as controls.

Endothelial cells were transfected using microRNA-7 mimics and scrambled sequences using EndoFectin (GeneCopoeia, Rockville, MD).

Endothelial cells (1.0 × 10⁵/cm²) were cultured on polycarbonate cell culture inserts (pore size 0.4 µm, porosity 0.9·10⁸/cm²; Nunc, ThermoFisher, Waltham, CA) coated with 0.1% gelatin for 48 h. When appropriate, cultures were serum starved for 24 h and cells were treated with fenoterol (1 µM; Boehringer Ingelheim, Germany), forskolin (10 µM; Tocris, UK), 6-Bnz-cAMP (300 µM) or 8-pCPT-2′-O-Me-cAMP (100 µM) (Biolog Life Science Institute, Germany). Parallel cultures were maintained under normoxic (21% oxygen tension) and hypoxic conditions (2% oxygen tension) for 48 h prior to experiments. Permeability was assessed by the addition of 10 µg/ml FITC-dextran in the upper compartments, and fluorescence in the lower compartments was assessed on a spectrofluorescence reader at Ex485/Em519 after 30 min.

Endothelial cells were cultured in eight-well Lab-Tek^® chamber slides (Nunc, IL, USA) until 80% confluence was reached. Cells were incubated under normoxic or hypoxic conditions for 24 h. Samples were fixed in 2% paraformaldehyde in PBS for 20 min and permeabilized with 0.3% Triton X-100 for 10 min. Blocking of unspecific antibody activity was performed using in 2% BSA. Samples were incubated with rabbit polyclonal antibodies to human EPAC-1 (Abcam, #ab21236, 1:100) and eNOS (BD #61098, 1:200) diluted in 2%BSA/PBS overnight at 4 °C. Samples were washed and incubated with Alexa Fluor^® 594-conjugated goat anti-rabbit IgG antibodies (Molecular Probes, Invitrogen, OR, USA) diluted 1:500 in DAPI/PBS for 1 h at room temperature. Samples were mounted in Citifluor AF1 (Agar Scientific, UK), and cells were examined using a Leica TCS SP8 (Leica Microsystems, Germany) laser scanning fluorescence confocal microscope using a 63x/1.40 oil objective.

Total RNA was isolated using TRIzol reagent (Invitrogen, Waltham, CA) according to manufacturer’s instructions and quantified by spectrophotometry (NanoDrop Technologies, Waltham, MA). For gene expression analyses, 1 µg of total RNA was reversely transcribed into cDNA using the RevertAid™ First-Strand cDNA synthesis kit (Applied Biosystems, Carlsbad, CA) and amplified using species-specific primers (human primers: Suppl. Table 1; mouse primers Suppl. Table 3). For microRNA expression analyses, 20 ng total RNA was reversely transcribed using the TaqMan MicroRNA Reverse Transcription kit (Applied Biosystems) using specific stemloop templates for miRNA-7-5p (5′-GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACACAACAAA-3′) or RNU6 (5′-GTCGTATCCAGTGCAGGGTCCGAGGTATTCGCACTGGATACGACAAAAATATGG-3′) and amplified using sense 5′-TGCGGTTGGAAGACTAGTGAT-3′, antisense 5′-CCAGTGCAGGGTCCGAGGTCCG-3′ for miR-7-5p, and sense 5′-TGCGGCTGCGCAAGGATGA-3′, antisense 5′-CCAGTGCAGGGTCCGAGGTCCG-3′ for RNU6. Quantitative PCR expression analysis was performed on a reaction mixture containing 10 ng cDNA equivalent, 0.5 µM sense primers, and 0.5 µM antisense primers (all Biolegio, Leiden, The Netherlands) and FastStart SYBR Green (Roche, Almere, The Netherlands). Analyses were run on a Viia7 real-time PCR system (Applied Biosystems, Carlsbad, CA). Each reaction was performed in triplicate and gene expression was calculated using the ΔΔCt method. The data are expressed as fold change versus control.

Retinal digests and endothelial cells were lysed in RIPA buffer (ThermoScientific, Waltham, MA) and protein concentration determined by DC Protein Assays (BioRad, Hercules, CA) according to manufacturers’ instructions. 30 µg of protein/lane was loaded on a SDS-PAGE gel (8–12%) for electrophoresis and transferred to a nitrocellulose membrane. Membranes were incubated overnight with antibodies to EPAC-1 (Abcam, Cambridge, UK, #ab21236, 1:1000), VE-Cadherin (Cell Signaling, #2500, 1:1000), eNOS (BD, San Jose, CA, #61098, 1:2000), phopho-Ser1177 eNOS (BD, San Jose, CA, #612393, 1:1000), and β-actin (Cell Signaling, #4967, 1:5000) for 1 h at RT. IRDye^®-labeled antibodies (1:10.000, Li-COR Biosciences, Lincoln, NE) were used for detection. Bands were visualized using the Odissey^® Infrared Imaging System (Li-COR Biosciences, NE, USA). Densitometry was performed using Image J version 1.45 s (NIH, Bethesda, MD). Protein expression levels were normalized to β-actin.

Nitrite levels in the cell culture medium, a stabile indicator of NO production, were quantified using the Measure-iT™ High-Sensitivity Nitrite Assay Kit (Molecular Probes, Eugene, OR) according to the manufacturer’s protocol. Obtained nitrite concentrations were normalized against the total amount of cellular protein using the DC protein.

Reactive oxygen levels were determined by incubating the cells with 50 µM 2′,7′-dichlorofluorescin diacetate (DCFH-DA, Sigma-Aldrich, St. Louis, MO) for 10 min in dark. Cells were dissociated using Accutase™ solution (PAA Laboratories, Austria), pelleted by centrifugation, and suspended in PBS. The generation of intracellular ROS was determined by flow cytometry on the FACSCalibur and WinList version 6.0 software (both BD Biosciences, CA, USA).

Double DIG-labeled MicroRNA-7-5p and scrambled control probes, microRNA ISH buffers, and reagents were obtained Exiqon (Vedbaek, Denmark) and used according to the manufacturers’ protocol. Hybridization was performed for 16 h at 44 °C in a humidified chamber.

Data is expressed as mean ± SEM. Significant differences between two means were determined by Mann–Whitney two-tailed U test for unpaired data or by one-way analysis of variance (ANOVA) followed by Dunnett’s post hoc test, where appropriate. p values <0.05 were considered statistically significant.

Oxygen-induced retinopathy is associated with a marked decrease in EPAC-1 expression (Fig. 1a). After 6 h and 24 h of relative hypoxia, retinal EPAC-1 gene transcript levels were reduced (2.0- and 1.9-fold, respectively, p < 0.05, Fig. 1b Conversely, miR-7 expression was increased in the retina of OIR-mice (2.7- and 3.2-fold, p < 0.01) compared to normoxic control mice (Fig. 1c). The effect of hypoxia on EPAC-1 expression was confirmed in endothelial cell cultures exposed to 2% (hypoxia) or 20% (normoxia) oxygen, where EPAC-1 protein expression decreased (2.4-fold, p < 0.01, Fig. 1d, e) when cells were exposed to hypoxia for 24 h. These data indicate that the expression of EPAC-1 is oxygen sensitive and its expression is decreased during hypoxic stress.

Endothelial cells exposed to 2% oxygen for 24 h increased miR-7 expression by 9.1-fold (p < 0.01, Fig. 2a). MicroRNA-7 contains a seed sequence that has complementarity to the 3′Untranslated Region (UTR) of EPAC-1 and EPAC-2. EPAC-1 has five putative miR-7 binding sites, whereas EPAC-2 only has one putative miR-7 binding site (Fig. 2b). To confirm that EPAC-1 and EPAC-2 are genuine targets of miR-7, we produced reporter constructs wherein the expression of luciferase is under the control of the 3′UTR of EPAC-1 or EPAC-2. Co-transfection of COS7 cells with miR-7 mimics and the EPAC-1 reporter construct decreased luciferase activity (2.6-fold) compared to scrambled controls (p < 0.05), whereas the luciferase activity of the EPAC-2 reporter was unaffected (Fig. 2c). These data indicate that EPAC-1 is a specific target of miR-7. Corroboratively, in endothelial cells transfected with miR-7 mimics EPAC-1 protein expression was decreased 2.1-fold (p < 0.05, Fig. 2d).

Hypoxia causes endothelial hyperpermeability (Fig. 3a), which associates with only a minor decrease (1.2-fold, p < 0.01) in VE-cadherin expression (Fig. 3b). Stimulation of endothelial cells with the adenylyl cyclase activator forskolin (10 μM) significantly reduced the hypoxia-induced endothelial hyperpermeability (2.2-fold, p < 0.001, Fig. 3c). Similarly, the administration of the selective protein kinase A (PKA) agonist (6-Bnz-cAMP, 300 μM) and EPAC agonist (8-pCPT-2′-O-Me-cAMP, 100 μM), two downstream mediators of adenylyl cyclase activity, inhibited hypoxia-induced endothelial hyperpermeability [1.6-fold (p < 0.01) and 2.1-fold (p < 0.001), respectively] (Fig. 3c). Interestingly, treatment of endothelial cells with ESI-09, an antagonist to EPAC-1, induced endothelial hyperpermeability under normoxic conditions (Fig. 3d). Conversely, a β2-agonist, fenoterol (1 μM) did not reduce endothelial hyperpermeability (Fig. 3c). Also, none of the treatments altered endothelial permeability under normoxic conditions (not shown).

We next investigated if the addition of miR-7 mimics to endothelial cells would imitate the hypoxia-induced endothelial hyperpermeability. Indeed, supplementing endothelial cells with miR-7 mimics induced hyperpermeability (Fig. 3e) in the absence of hypoxia and without alterations to VE-cadherin expression (Fig. 3f). Remarkably, treating miR-7-expressing cells with 8-pCPT reduced endothelial hyperpermeability (1.7-fold, p < 0.05), indicating that activation of remnant EPAC-1 is sufficient to inhibit miR-7-induced endothelial hyperpermeability.

Endothelial NOS protein expression by cells exposed to hypoxia was reduced ~40% (p < 0.05, Fig. 4a, b) compared to normoxic controls, which resulted in decreased NO production (1.4-fold, p < 0.05, Fig. 4c) and the increased generation of reactive oxygen species (ROS) (1.6-fold, p < 0.01, Fig. 4d).

The addition of miR-7 mimics to endothelial cells did not cause a reduction in eNOS expression level (Fig. 4e). However, the addition of miR-7 mimics to endothelial cells did imitate the hypoxia-induced loss of eNOS activity as indicated by a reduction in eNOS phosphorylation at serine 1177 (p-eNOS/eNOS ratio; 1.7-fold decrease, p < 0.01; Fig. 4e).

Fenoterol (1 μM) tended to increase eNOS phosphorylation at serine 1177 (p-eNOS/eNOS ratio) in endothelial cells exposed to hypoxia (Fig. 4f), whereas forskolin (10 μM), 6-Bnz-cAMP (300 μM) and 8-pCPT-2′-O-Me-cAMP (100 μM) efficiently increased eNOS activation by 2.3, 2.8 and 2.2-fold respectively (all p < 0.05, Fig. 4f). Corroboratively, NO production in hypoxia-treated endothelial cells was increased (1.4- to 1.6-fold) by these activators of cAMP signaling (Fig. 4g), and ROS production was decreased to a similar extend (Fig. 4h).

In the Ins2Akita model for diabetic retinopathy (Table 1), retinal EPAC-1 levels were reduced (~5.7-fold, p < 0.01) compared to non-diabetic control mice (Fig. 5a), whereas gene expression levels of EPAC-2 remained unchanged (Fig. 5b). MiR-7-5p was detected by in situ hybridization in the retinae from diabetic Ins2Akita mice (Fig. 5c), where its expression of miR-7 was increased (3.2-fold, p < 0.01, Fig. 5d) compared to non-diabetic controls. In control C57BL/6 mice, miR-7-5p levels remained below the detection limit for in situ hybridization (Fig. 5c). EPAC-1 expression levels associated with miR-7-5p expression levels to a high extend (r ² = 0.464, p < 0.001, Fig. 5c) indicating that the hypoxia-induced expression of miR-7 might underlie the loss of EPAC-1 expression in diabetic retinopathy. In the Ins2Akita mice, endothelial hyperpermeability, as assessed by leakage of fluorescently labeled low molecular weight dextran, was not observed (data not shown). However, eNOS activation decreased 2.4-fold (p < 0.01, Fig. 5f) compared to control mice, suggestive of reduced EPAC signaling.

Besides EPAC-1 [21], miR-7 targets a number of additional gene transcripts (Suppl. Table 1). We analyzed the expression of 23 reported miR-7-5p target genes in endothelial cells that express miR-7-5p (Suppl. Table 2) and in retinal isolations from diabetic Ins2Akita mice (Suppl. Table 4) and found no other transcript which was decreased in both conditions.