TNF-α mediates choroidal neovascularization by upregulating VEGF expression in RPE through ROS-dependent β-catenin activation

Animals

Five- to eight-week-old C57Bl/6 wild-type mice were used in these studies. All animal procedures were conducted in accordance with the University of Utah Guide for the Care and Use of Laboratory Animals and the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research.

Laser-induced CNV model and injections

The laser-induced CNV model was performed as previously described [10]. Intraperitoneal injections of 100 mg/kg ketamine (Hospira, Inc., Lake Forest, IL) and 20 mg/kg xylazine (AnaSed Lloyd Laboratories, Shenandoah, IA) were used for anesthesia. Following mydriasis with 1% tropicamide ophthalmic solution (Bausch & Lomb, Tampa, FL), the mouse was secured on a stage, and a coupling agent, GenTeal (Alcon, Fort Worth, TX), was applied to the cornea. Laser photocoagulation was performed with a 532 nm Phoenix Image-Guided Laser System (Phoenix Micron IV, Pleasanton, CA) using the settings of 400 mW intensity and 100 ms duration. Four to five laser spots per eye were applied approximately 2 disc diameters from the optic nerve, avoiding major vessels. Disruption of Bruch’s membrane was confirmed by the appearance of a cavitation bubble. Right after laser injury, using a MICROLITER syringe (Hamilton Company, Reno, NV), an intravitreal injection of total volume of 1 µl of neutralizing anti-mouse TNF-α antibody (TNF-α Ab, 50 ng/ml, R&D Systems, Minneapolis, MN) or control, isotype immunoglobulin G (IgG, 50 ng/ml, R & D Systems). Both eyes of each mouse were lasered and injected with the same type of injection. Seven days after the laser treatment, the mice were euthanized by CO₂ exposure, and one eye from each mouse was collected for RPE/choroidal flatmount analysis and the fellow eye for protein by immunostaining of the retinal cryosections or western blot analysis. Each group included both eyes of at least eight mice.

Analysis of CNV volume in RPE/choroidal flatmounts

Eyes were first fixed in 4% paraformaldehyde (Electron Microscopy Sciences [EMS], Hatfield, PA) for 1 h. After the cornea was removed, the lens, retina and vitreous, and posterior eyecups of the RPE/choroid/sclera were dissected. The eyecups were stained with Alexa Fluor 568-conjugated isolectin B4 (Invitrogen, Carlsbad, CA, Cat # I21412) overnight at 4 °C to label invading choroidal vessels. The eyecups were then flattened by cutting radial incisions and flatmounted on a microscope slide with Fluoromount G (Southern Biotech, Birmingham, AL) for confocal imaging. The flatmounts were imaged by taking optical Z-sections at3 µm increments with a confocal microscope (FV1000, Olympus, Melville, NY). The volume (µm³) of each lesion of CNV was measured by computerized image analysis using software provided with the microscope. Lesions with obvious hemorrhage or bridging CNV were excluded. Each CNV spot was considered an independent data point. For each condition, at least 30 spots were analyzed from eight mice.

Retinal and RPE/choroidal section preparation and immunostaining of OxPLs

Eyes were first fixed in 4% paraformaldehyde for 1 h, and then the corneas and lenses were removed. Eyecups were incubated in 30% sucrose overnight and frozen at −80 °C after immersion in optimal cutting temperature compound (OCT: Tissue Tek; EMS). The eyecups were cut into 12 µm cryosections. Sections were incubated with E06 monoclonal antibody (1:100, Avanti POLAR LIPIDS, Inc., Alabaster, Alabama, Cat # 330001) overnight at 4 °C. After rinsing, the sections were incubated for 1 h with fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse IgM (1:100, Sigma-Aldrich, St. Louis, MO, Cat # F9259), Alexa Fluor 568-conjugated antibody for isolectin B4 (lectin, 1:200, Invitrogen) to stain the vessels, and Topro 3 (1:1,000, Life Technologies, Grand Island, NY) to stain the nuclei. The sections stained with no primary antibody were used as controls. Labeling for all sections was performed during the same experimental session. Images were captured using confocal microscopy (Olympus IX81) at 20X magnification. Within the sections of lectin-stained CNV, E06 labeling to measure oxidized phospholipids (OxPLs) was determined with fluorescent density using Image J. At least ten CNV lesions per condition were analyzed (n=6 per condition).

Cell culture

Human primary RPE cells (hRPE; Lonza, Walkersville, MD) were grown in retinal pigment epithelial cell basal media (RtEBM, Lonza) and used from passages 3–5. In some experiments, the cells were pretreated with the nuclear factor kappa B (NF-ҡB) inhibitor, Bay 11–7082 (5 µM), or Wnt/beta catenin inhibitors, JW 67 (20 µM) or XAV939 (1 µM; Tocris Bioscience, Bristol, UK), and then incubated with human recombinant TNF-α (20 ng/ml, R&D Systems) or PBS (1X, 137.93 mM NaCl, 8.06 mM NaPO₄, 2.67 mM KCl, 1.47 mM KH₂PO₄, pH 7.4) for further analyses.

siRNA transfection in human RPE cells

To knock down p22phox, RPE cells were transfected using Lipofectamine 2000 according to the commercial protocol (Life Technologies) with siRNA targeting the human CYBA gene (Gene ID: 1535 and OMIM 608508; p22phox) or silencer selective negative control siRNA (both from Life Technologies). Forty-eight hours after transfection, the cells were incubated with human recombinant TNF-α (20 ng/ml, R&D Systems) or PBS for further analysis. All experimental conditions were plated in triplicate.

ROS generation assay in human RPE cells

RPE cells were seeded into 96-well plates and infected with adenoviral constructs. Forty-eight hours post infection, the cells were loaded with 5 µM 2’,7’-dichlorofluorescein diacetate (DCFDA, Invitrogen) in serum-free medium for 30 min at 37 °C. After two washes with PBS, the cells were pretreated with apocynin (100 µM), or NADPH oxidase inhibitor, VAS 2870 (20 µM, EMD Millipore), or control PBS or dimethyl sulfoxide (DMSO) for 30 min and then incubated with TNF-α or control PBS for another 30 min. ROS generation was measured in a fluorescent plate reader (excitation at 488 nm and emission at 520 nm). Cells incubated with 10 µM H₂O₂ were used as a positive control (n=9 per condition).

Immunoprecipitation and immunoblots

RPE cells were lysed in radioimmunoprecipitation assay buffer (RIPA) with protease inhibitor cocktail (Roche Diagnostics, Indianapolis, IN) and orthovanadate (Sigma-Aldrich). Lysates were clarified with centrifugation at 1.61x10⁴ ×g for 5 min at 4 °C. The protein concentration in the supernatant was quantified with bicinchoninic acid assay (BCA; Pierce, Rockford, IL). The total protein (500 µg in 500 µl) for each sample was incubated with antibody to T cell factor 1 (TCF1, 1:200, Cell Signaling Technology Inc., Danvers, MA) by gently rocking at 4 °C overnight. Ten microliters of Dynabeads protein G (Invitrogen) were added and incubated for an additional 1 h at 4 °C. The antibody/protein/agarose complex was washed three times with RIPA buffer and resuspended in 2X sample buffer. The protein complex bound to TCF1 or pan-cadherin was separated with NuPAGE 4% to 12% Bis-Tris Gels (Invitrogen) and transferred to a polyvinylidene fluoride (PVDF) membrane, and then incubated with antibody to β-catenin (1:1,000, Cell Signaling Technology Inc.) overnight at 4 °C. All membranes were reprobed with TCF1 or pan-cadherin to ensure equal protein loading.

For TNF-α, VEGF and β-catenin in the RPE/choroids or in the RPE cells, 20 µg of protein from each treatment was loaded into NuPAGE 4% to 12% Bis-Tris Gels (Invitrogen), transferred to a PVDF membrane, and incubated with antibodies to TNF-α (1:1,000, Cell Signaling Technology Inc.), VEGF (1:500, Santa Cruz Biotechnology, Santa Cruz, CA), or β-catenin at 4 °C overnight. Densitometry analysis was performed on exposed films with the use of the software UN-SCAN-IT version 6.1 (Silk Scientific, Orem, UT). All experimental conditions were plated in triplicate (n=6).

RNA Isolation and qPCR analysis

Total RNA of the RPE/choroidal tissues was extracted with TRI Reagent (Sigma-Aldrich). RNA was quantified using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA). cDNA was generated with the use of a high-capacity cDNA archive kit (Life Technologies). Quantitative PCR (qPCR) was performed on a Mastercycler ep realplex system (Eppendorf, Hauppauge, NY) with the use of SYBR Green Master Mix (Roche Diagnostics) and primers synthesized by the core research facility of the University of Utah. Expression levels for VEGF were normalized to the mean value of the internal control, GAPDH. The primers (forward and reverse, respectively) were mouse VEGF 5′-TGG TTC TTC ACT CCC TCA AAT C-3′ and 5′- GGT CTC TCT CTC TCT TCC TTG A-3′, and mouse GAPDH 5′-GGA GAA ACC TGC CAA GTA TGA-3′ and 5′-TCC TCA GTG TAG CCC AAG A-3′.

Statistical analysis

Significant differences between groups were determined with ANOVA with post hoc protection using the Bonferroni multiple comparison test. Results were the mean ± standard error of the mean (SEM). A minimum p value of less than 0.05 was considered statistically significant. For animal studies, at least 30 laser spots were analyzed for CNV, and ten retinal sections, p<0.01 for oxidized phospholipids (OxPLs) from different animals. Western blot or PCR analyses included six to eight retinas/condition. For analysis of densitometry of the immunolabeled retinas, at least three sections imaged at 60 μm intervals were examined/eye. For IHC and western blots, a total of at least three (IHC) or five (western blots) eyes were analyzed for each condition. For in vitro studies, each experimental condition included an n=6–9, and each experiment was performed three times.

TNF-α mediates CNV formation in association with increased VEGF expression in a murine l laser-induced CNV model

TNF-α (p<0.05) and VEGF (p<0.01) protein in RPE/choroid lysates were significantly increased 7 days after laser injury compared to the non-lasered controls (Figure 1A,B). Similar to previous studies [19], treatment with intravitreal neutralizing antibody to mouse TNF-α (TNF-α Ab) significantly reduced the volume of laser-induced CNV compared to the mice treated with the isotype IgG control (IgG; Figure 1C,D, p<0.01) To determine if reduced CNV volume was associated with reduced VEGF, we measured the VEGF protein in the RPE/choroids by western blot and found that in the mice treated with intravitreal TNF-α Ab, VEGF protein in the RPE/choroids was significantly decreased compared to that in the RPE/choroids from the mice treated with intravitreal IgG (Figure 1E,F, p<0.01). These results provide evidence that TNF-α induced CNV was associated with increased VEGF expression.

TNF-α induces NADPH oxidase-generated ROS at CNV lesions and in cultured RPE cells

We previously found that external sources of oxidative stress were implicated in laser-induced CNV [10,20]. Therefore, we determined the effect of TNF-α on ROS generation in laser-induced CNV. We measured OxPLs by immunostaining with the E06 antibody. As shown in Figure 2A, OxPL staining was located within CNV, and quantification of immunofluorescence confirmed that OxPL staining was significantly reduced in the TNF-α Ab-treated mice compared to the control IgG-treated mice (Figure 2B, p<0.05). Additionally, ROS generation was measured with 2',7'-dichlorofluorescein diacetate (DCFDA) fluorescence in cultured human RPE cells incubated with different doses of recombinant human TNF-α for 30 min. Compared to PBS, TNF-α significantly increased ROS generation in a dose-dependent pattern (Figure 3A). To determine whether TNF-α-mediated ROS generation involved activation of NADPH oxidase, ROS were measured in RPE cells pretreated with apocynin (Apo) or the NADPH oxidase inhibitor, VAS2870 (VAS), for 30 min before stimulation with TNF-α. As shown in Figure 3B, pretreatment with Apo or VAS prevented TNF-α-induced ROS generation (p<0.001). We previously reported that NOX2 was involved in CNV pathogenesis, as evidenced by reduced CNV in laser-treated p47phox knockout mice compared to the wild-type mice [10]. We also found that VEGF-activated NOX4, an isoform of NADPH oxidase, mediated retinal microvascular endothelial cell proliferation by interacting with VEGFR2 [21]. Therefore, we determined whether TNF-α activated NOX4 in RPE cells. The NOX4 protein level and activation determined by cobinding with p22phox were measured in RPE cells treated with TNF-α or PBS, and VEGF was used as a positive control. As shown in Figure 3, compared to PBS, overnight treatment with TNF-α increased the NOX4 protein level (Figure 3C) and activation (Figure 3D,E, p<0.05). NADPH oxidase-dependent ROS generation was further determined in the RPE cells transfected with siRNA targeting human p22phox, a subunit required for NADPH oxidase activation by NOX2 or NOX4. Compared to the control siRNA and PBS treatment, knockdown of p22phox prevented ROS generation in RPE cells treated with PBS p<0.05 or TNF-α (p<0.001; Figure 3F). These results provide evidence that activation of NADPH oxidase was likely a source of ROS generation in the RPE cells and that TNF-α-mediated ROS in the RPE cells was generated through activation of NADPH oxidase.

TNF-α upregulates VEGF expression in RPE cells via ROS-dependent activation of β-catenin

To determine whether TNF-α affects VEGF expression in RPE cells, VEGF was measured with western blot after treatment of the RPE cells with TNF-α for 0, 3, 6, or 12 h. As shown in Figure 4A, TNF-α increased VEGF protein level as early as 3 h after TNF-α treatment and remained upregulated at 12 h compared to control. To determine whether ROS signaling was involved in TNF-α-mediated VEGF upregulation, RPE cells were pretreated with apocynin for 30 min to quench ROS before incubation with TNF-α overnight and then analyzed for VEGF mRNA with quantitative PCR. TNF-α-induced VEGF mRNA was significantly inhibited by apocynin treatment (Figure 4B, p<0.05), suggesting that TNF-α upregulates VEGF expression in RPE cells via ROS-dependent signaling.

Activation of NF-ҡB is a classic inflammatory signaling pathway mediated by TNF-α. Therefore, we tested whether inhibition of NF-ҡB would inhibit TNF-α-mediated VEGF expression in RPE cells. In RPE cells, TNF-α activated NF-ҡB determined by increased p-p65, but inhibition of TNF-α-induced NF-ҡB activation by Bay 11–7082 (Figure 5A) did not affect TNF-α-induced VEGF upregulation (Figure 5B). This failed to show that the NF-ҡB signaling pathway was involved in TNFα-induced VEGF upregulation.

Wnt/β-catenin signaling is implicated in inflammation. Activated β-catenin translocates to the nucleus and regulates gene expression by forming a complex with tissue cell factor (TCF)/lymphoid enhancing factor (LEF) transcriptional factors. Therefore, we determined whether Wnt/β-catenin signaling was involved in TNF-α-mediated VEGF upregulation. RPE cells were pretreated with JW67 [22] or XAV939 [23], inhibitors of Wnt/β-catenin signaling, before incubation with TNF-α, and the β-catenin and VEGF protein levels were measured. As shown in Figure 6A, β-catenin was increased by TNF-α as early as 1 h after treatment and was inhibited by pretreatment with XAV939, but not JW67. In the same cell lysates, VEGF protein was increased by TNF-α as early as 1 h after treatment and remained upregulated at 6 h after treatment. Inhibition of Wnt/β-catenin signaling by XAV939 prevented VEGF upregulation by TNF-α (Figure 6B). To determine the mechanism, co-immunoprecipitation of beta-catenin and TCF1 was determined. As shown in Figure 6, TNF-α treatment induced greater coimmunoprecipitated β-catenin and TCF1, and this was inhibited by pretreatment with apocynin or XAV939 (Figure 6C) or by knockdown of p22phox (Figure 6D). These findings provide evidence that TNFα led to VEGF upregulation through β catenin/TCF1 transcription.

Altogether, the results in Figure 6 and Figure 7 provide evidence that TNF-α upregulates VEGF expression in RPE cells via ROS-dependent activation of Wnt/β-catenin signaling, independent of NF-ҡB. These data provide evidence that TNF-α-mediated NADPH oxidase-generated ROS trigger Wnt/β-catenin activation and β-catenin translocation and binding with TCF1 that led to increased VEGF expression.