Attenuation of Choroidal Neovascularization by Histone Deacetylase Inhibitor

Choroidal neovascularization (CNV) is a blinding complication of age-related macular degeneration that manifests as the growth of immature choroidal blood vessels through Bruch’s membrane, where they can leak fluid or hemorrhage under the retina. Here, we demonstrate that the histone deacetylase inhibitor (HDACi) trichostatin A (TSA) can down-regulate the pro-angiogenic hypoxia-inducible factor-1α and vascular endothelial growth factor (VEGF), and up-regulate the anti-angiogenic and neuro-protective pigment epithelium derived factor in human retinal pigment epithelial (RPE) cells. Most strikingly, TSA markedly down-regulates the expression of VEGF receptor-2 in human vascular endothelial cells and, thus, can knock down pro-angiogenic cell signaling. Additionally, TSA suppresses CNV-associated wound healing response and RPE epithelial-mesenchymal transdifferentiation. In the laser-induced model of CNV using C57Bl/6 mice, systemic administration of TSA significantly reduces fluorescein leakage and the size of CNV lesions at post—laser days 7 and 14 as well as the immunohistochemical expression of VEGF, VEGFR2, and smooth muscle actin in CNV lesions at post-laser day 7. This report suggests that TSA, and possibly HDACi’s in general, should be further evaluated for their therapeutic potential for the treatment of CNV.


Cell culture
Human RPE cells were isolated in our laboratory from fetal human eyes of 18-20 weeks' gestation (Advanced Bioscience Resources, Inc, Alameda, CA) [49]. The cells were cultured in Dulbecco's modified Eagle's medium (DMEM, Fisher Scientific, Pittsburgh, PA) with 2 mM L-glutamine, 100 U/mL penicillin, 100 μg/mL streptomycin (Sigma-Aldrich, St. Louis, MO), and 10% heat-inactivated fetal bovine serum (FBS; Irvine Scientific, Santa Ana, CA). The culture method we used is a standard practice in our laboratory and regularly yields >95% cytokeratin-positive RPE cells. Cells used were from passages 2 to 4. Cell treatments were performed on chamber slides, 6-well plates, or 96-well plates (BD Falcon, San Jose, CA) and were initiated 24 h after cell plating or subculture. CECs were isolated from bovine eyes using magnetic beads bound to the specific endothelial marker Lycopersicon esculentum (Sigma-Aldrich), as previously described. [50] Human umbilical vein endothelial cells (HUVECs) were purchased from ATCC (Manassas, VA). Both BCECs and HUVECs were cultured in endothelial growth medium (EGM Bullet Kit, #CC-3124, Lonza, Switzerland), and both cell types were used from passages 2 to 8.

Cell Cycle and Cell Viability Analysis
For cell cycle analysis, RPE cells incubated in serum-free DMEM for 24 h and treated in serum-free DMEM with 0, 0.1, 0.5 or 1 μM TSA for 24 h were harvested, fixed in 1 mL of icecold 70% ethanol at room temperature for 10 min, and washed twice with ice-cold phosphatebuffered saline (PBS, pH 7.4). Each sample of 1×10 6 cells was pelleted and re-suspended in 1mL of 10 μg/mL propidium iodide (Life Technologies, Grand Island, NY), incubated at 37°C for 30 min, and analyzed using an EPICS XL-MCL flow cytometer (Beckman Coulter, Irvine, CA). Samples of 5×10 3 cells were used for analysis. The experiment was repeated three times. For cell viability analysis, RPE cells (5 × 10 4 ) were seeded in 96-well plates in DMEM with 10% FBS. After overnight incubation to allow for cell attachment, the medium was replaced with serum-free DMEM, and the cells were treated with 0, 0.05, 0.1, 0.3, 0.5, 0.7 or 1 μM TSA (Sigma-Aldrich) in 90 μL of serum-free DMEM per well for 24 h. At the end of TSA treatment, 10 μL of PrestoBlue reagent (Life Technologies) was added per well. After a 6-hour incubation absorbance was read at 570 nm, using a reference wavelength of 600 nm for normalization, with a multi-well plate reader (Benchmark Plus, Bio-Rad, Tokyo, Japan). The experiment was repeated three times.

Attachment Assay
Attachment was measured with a modified MTT [(3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide)] assay using 96-well plates coated with fibronectin (Life Technologies) (2 μg/cm 2 ). After treatment with TSA (0.05, 0.1, 0.3, 0.5 or 0.7 μM) for 24 hours, RPE cells were trypsinized and re-suspended in DMEM with 0.4% FBS. 100 μL of cell suspension (10 4 cells) were added to each well and allowed to attach for 5, 10, 15 or 30 min. The cells were washed gently with PBS twice, and 150 μL of fresh DMEM with 10% serum was added to each well with 20 μL of MTT (5 mg/mL; Sigma-Aldrich). After 4 hours of incubation at 37°C, the supernatants were decanted, the formazan precipitates were solubilized by the addition of 150 μL of 100% DMSO, and the plate was mixed on a plate shaker for 10 minutes. Absorbance at 550 nm was determined on a multiwell plate reader. The number of attached cells was proportional to the absorbance of MTT at 550 nm.

Migration Assay
Migration was measured using a modified Boyden chamber assay, as previously described [51]. Briefly, 5 × 10 4 RPE cells that were treated in serum-free DMEM with 0, 0.05, 0.1, 0.3, 0.5 or 0.7 μM TSA for 24 h were seeded in the upper part of a Boyden chamber together with the corresponding concentrations of TSA in 24-well plates. Inserts were coated with fibronectin (2 μg/ cm 2 ). The lower chamber was filled with 0.4% FBS-DMEM containing 10 ng/mL of recombinant PDGF-BB (R&D Systems Inc., Minneapolis, MN). After 5 hours of incubation, the inserts were washed three times with PBS, fixed with cold methanol (4°C) for 10 minutes, and counterstained with hematoxylin for 20 minutes. The number of migrated cells was counted by phase-contrast microscopy (×320). Four randomly chosen fields were counted per insert.
Real-time polymerase chain reaction RPE cells were pre-exposed to 0-0.5 μM TSA for 14 h. The cells were then treated with 150 μM cobalt chloride (CoCl 2 ) (Sigma-Aldrich, St. Louis, MO), with or without TSA, for 6 h. Additional RPE cells were exposed to TSA only for 20 h. HUVECs were treated with 0-0.7 μM TSA for 48 h. Total RNA was extracted from the cells with TriZol (Life Technologies), and reverse transcription was performed with 1 μg of total RNA, using the ImProm-II Reverse Transcription System according to the manufacturer's protocol (Promega, Madison, WI). Real-time PCR was performed in duplicate with a kit used according to the manufacturer's recommendation (Roche Diagnostics, Indianapolis, IN). The primer sequences used are listed in Table 1. The quantity of mRNA was calculated by normalizing the threshold cycle values of VEGF, PEDF or VEGFR2 to the threshold cycle value of the housekeeping gene RPL13A of the same RNA sample, according to the published formula [52].

Western blot analysis
Confluent human fetal RPE cells grown in 6-well plates were starved for 24 hours in serumfree DMEM, and then treated with 0, 0.05, 0.1, 0.3, 0.5 or 0.7 μM TSA only for 24 h, or with 0, 0.05, 0.1, 0.3, or 0.5 μM TSA for 18 h, and then co-treated with 150 μM CoCl 2 (Sigma-Aldrich) for 6 h, or pre-treated with 0, 0.1, 0.3, 0.5 or 0.7 μM TSA in DMEM with 0.1% FBS for 1h, and then co-treated with 20 ng/mL of human recombinant TGF-β1 (R&D Systems Inc.) for 72 hr. Bovine CECs (BCECs) or HUVECs grown in 6-well plates were starved for 4 h in endothelium basal medium (EBM, Lonza, Basel, Switzerland) with 1% FBS, and then treated with 0, 0.05, 0.1, 0.3, 0.5 or 0.7 μM TSA only for 24h or 48 h, or followed by stimulation with 20 ng/mL human recombinant VEGF (R&D Systems Inc Minneapolis, MN) for 10 min. Cells were harvested and lysed by RIPA buffer (Cell Signaling, Danvers, MA), and proteins were resolved on 4-15% Tris-HCl polyacrylamide gels (Bio-Rad) at 120 V. The proteins were transferred to polyvinylidene blotting membrane (Millipore, Bedford, MA). To assay HIF-1α, VEGF and PEDF, RPE cells that had been pre-exposed to TSA for 18 h were treated with CoCl 2 for 6 h. To assay-pAkt, p-p42/44 and caspase 3, BCECs were treated with TSA for 24 h. To assay p-  Tables 2, 3 and 4. Membranes were washed with TBS-Tween 20 (0.1%) (Bio-Rad) and then incubated with a horseradish peroxidase-conjugated secondary antibody (Vector Laboratories, Inc. Burlingame, CA) for 30 min at room temperature. To assay Akt, p42/44, or VEGFR2, the membranes were re-probed with the corresponding antibody, as listed in the tables. The subsequent procedures were performed as described above. Images were developed by adding enhanced chemiluminescence detection solution (Amersham Pharmacia Biotech, Cleveland, OH). Densitometry was performed using the Image J software (rsbweb.nih.gov/ij).

TUNEL assay
BCECs cultured in chamber slides were starved for 4 h in endothelium basal medium with 1% FBS and then treated with 500 μM hydrogen peroxide (Sigma-Aldrich) for 4 h or with 0.7 μM TSA for 24 h. Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay was performed using the In Situ Cell Death Detection Kit, TMR red (Roche Diagnostics, Mannheim, Germany), according to the manufacturer's instructions. TUNEL-positive cells were counted under a fluorescence microscope (EVOS, Advanced Microscopy Group, Bothel, MA). The number of positive cells was presented as the percentage of dead cells.

Tube formation assay
Confluent HUVECs were treated with 0, 0.3, 0.5, 0.7 or 1 μM TSA for 24 h. Two-dimensional tube formation was measured in Geltrex Reduced Growth Factor Basement Membrane Matrix gel (Life Technologies). The gel was placed in 96-well plates and incubated at 37°C for 30 min to reconstitute into basement membrane. Samples of 1 × 10 4 HUVECs were seeded on each well and incubated with EBM + 1% FBS with TSA and human recombinant VEGF (25 ng/ml) for 2 h. Tube formation was documented by photography with a phase-contrast microscope. The experiment was repeated three times. The amount of tube formation was quantified using the Image J software.

ChIP assay
ChIP assay was performed using the Imprint Chromatin Immunoprecipitation Kit (Sigma-Aldrich) according to the manufacturer's instructions, with some modifications: 5×10 5   with bedding material, in the vivaria at the Doheny Eye Institute. The mice were fed the standard laboratory chow in an air-conditioned room equipped with a 12-h light/12-h dark cycle. The mice were kept for 2 days before starting the experiment, and then randomly assigned to the control and experimental groups. Two groups of five mice each were included in two control groups that received intraperitoneal (IP) injections of PBS, one group for seven days of observation and the other for 14 days of observation. Two groups of five mice each were included in two experimental groups that received IP injections of TSA, one group for seven days of observation and the other for 14 days of observation. For all surgical procedures, the mice were anesthetized with an IP injection of 96 mg/kg of ketamine and 4.8 mg/kg of xylazine, and their pupils were dilated with topical 1% tropicamide (Alcon, Fort Worth, TX). Diode laser photocoagulation (75-μm spot size, 0.1-sec duration, 110 mW) was performed on both eyes of each mouse on day 0. Three laser photocoagulation burns were delivered to the retina lateral to the optic disc, through a slit lamp, with a coverslip used as a contact lens. Only lesions in which a subretinal bubble developed were used for experiments. Based on the fact that three lesions were induced per eye, and taken into considerations the variability in outcome, five mice per group are necessary to attain statistical significance. The TSA used in this study was obtained from Selleck Chemicals (Houston, TX). TSA was delivered to the mice by IP administration at 20 mg/kg per injection. Injections were made shortly after laser photocoagulation and every 48 h after laser treatment at about 1pm for 7 or 14 days, with the PBS control group injected first followed by the TSA group, and the mice were also checked for signs of discomfort. Control mice were injected with 0.5 mL of sterile PBS.

Fluorescein angiography and histological analysis
The effect of TSA treatment on the development of CNV was evaluated on days 7 and 14 by semiquantitative assessment of late-phase fluorescein angiograms using the Kowa Genesis 35 mm fundus camera, captured 3 min after IP injection of 0.1 mL of 2.5% fluorescein sodium (Akorn, Decatur, IL), as previously described [53]. Ten animals, 20 eyes, and 34 lesions were examined in the control group, and 10 animals, 20 eyes, and 35 lesions were examined in the TSA group. Leakage was defined as the presence of a hyperfluorescent lesion that increased in size with time in the late-phase angiogram. Angiography was graded in a masked fashion by two examiners using reference angiograms. Angiograms were graded as follows: 0, no leakage; 1, slight leakage; 2, moderate leakage; and 3, prominent leakage [52]. While under anesthesia, cervical dislocation was performed and the animals were checked for lack of breathing, lack of a heartbeat and the body becoming cold. The right eye from each mouse was used for histologic analysis while the left eye was used for CNV volume analysis (see next section). For histopathologic analysis, enucleated eyes were snap frozen. Sections (8 μm) from the center of the lesion were stained with hematoxylin and eosin (H&E), to assess the histology of the retina with TSA treatment, laser lesions, and subsequent CNV development. Measurement of CNV lesion area in the H&E-stained sections was carried out with the ImageScope software (Aperio, Vista, CA). Eight animals, eight eyes and eight lesions were examined in the control group, and nine animals, nine eyes and nine lesions were examined in the TSA group.

CNV volume analysis
Eyes were enucleated on days 7 and 14 after fluorescein angiography and euthanasia and fixed with 10% formalin overnight at 4°C. Eye cups were obtained by removing the anterior segments and neurosensory retina and washed three times in PBS. The remaining eye cups containing the RPE-choroid-sclera complex were incubated with blocking buffer (PBS containing 1% BSA and 0.5% Triton X-100) for 1 h at room temperature. The eye cups were stained with 10 μg/mL of FITC-isolectin B4 and then visualized using the 20× objective of a scanning confocal microscope (model LSM510; Carl Zeiss Meditec, Inc, Thornwood, NY). Fluorescence volume measurements were made by creating image stacks of optical slices within lesions. The image stacks were generated in the Z-plane, with the confocal microscope set to excite at 488 nm and to detect at 505 to 530 nm. Images were further processed using the microscope's system software (LSM; Carl Zeiss Meditec, Inc.), by closely circumscribing and digitally extracting the fluorescent lesion areas throughout the entire image stack. The extracted lesion was processed through the topography software to generate a digital topographic image representation of the lesion and an image volume. The topographic analysis program determines and displays the objects' surface contours by detecting fluorescent signal from the top of the image stack and then measures everything under the surface to yield a final volume (square micrometers ± SD) that reflects the CNV fluorescence volume [54]. Ten animals, 10 eyes and 20 lesions were examined in the control group, and 10 animals, 10 eyes and 28 lesions were examined in the TSA group.

Immunohistochemistry
Antibodies are listed in Table 6. Cryostat sections (8 μm) of snap frozen mouse eyes were prepared from animals treated with PBS or TSA from day 7 of the CNV model. The slides were fixed with methanol for 10 min, and then rinsed in PBS for 5 min. After fixation, the slides were incubated with 0.5% hydrogen peroxide for 5 min, followed by 5% normal goat serum for 15 min. Next, sections were incubated with primary antibodies overnight at 4°C. Binding of the primary antibody was visualized first by incubating with biotinylated anti-rabbit or mouse IgG antibody (Vector Laboratories) for 30 min, and then with horseradish peroxidase-conjugated streptavidin (Life Technologies) for 30 min. Sections were counter-stained with hematoxylin for 2 min, mounted in aqueous mounting medium (Vector Laboratories), and examined and quantified using the Aperio ScanScope CS, 20× lens with doubler for 40×, and the ImageScope software (Aperio).

Evaluation of TSA in vivo Toxicity
The toxicity of TSA on normal choroidal vasculature was evaluated in wild type male 6-8 week old C57Bl/6 mice after 14 days of systemic therapy. Three mice were assigned each to the TSA and PBS control groups, and they were intraperitoneally injected with 20 mg/kg TSA or the same volume of PBS in the same regiment as the mice in the CNV model. The animals were euthanized on day 14, and eye cups were isolated from enucleated eyes as described in "CNV Volume Analysis", and then the RPE layer was gently washed off. Afterwards, the choroid-sclera complex was fixed in 4% paraformaldehyde at room temperature for 20 min, and then washed with PBS at room temperature for 30 min. TUNEL assay was performed using the In Situ Cell Death Detection Kit, TMR red (Roche), according to the manufacturer's instructions, except that the permeabilization time was increased to 5 min. Positive controls consisted of flatmounts and sections treated with 2000 U/ml of rDNase 1 (Sigma). The flatmounts were incubated with 10 μg/mL of FITC-isolectin B4, mounted with DAPI and then visualized using the 10× and 40× objectives of the PerkinElmer Ultraviewer Spinning Disk Confocal Microscope (PerkinElmer, Waltham, MA). Three animals and six eyes were examined in the control groups, and three animals and six eyes were examined in the TSA group. TSA toxicity was also evaluated in the cryosections from the laser CNV mouse model by TUNEL assay using the In Situ Cell Death Detection Kit, TMR red, according to the manufacturer's instructions. The cryosections were then incubated with 10 μg/mL of FITC-isolectin B4, mounted with DAPI and then visualized with Spinning Disk Confocal Microscope PerkinElmer Ultraviewer. Three animals and three eyes were each examined in each of the PBS and TSA groups.

Statistical analysis
All in vitro experiments were repeated at least three times, the number of animals, number of eyes and number of lesions examined in the in vivo experiments were listed above, and were all analyzed by the Student's t test. Analysis of variance (one-way ANOVA) was performed to test the statistical significance for cell cycle analysis, PrestoBlue assay, attachment assay, all Western blots, all real-time PCR, migration assay, TUNEL assay, and tube formation assay.

TSA arrests RPE cell cycle progression
RPE cells were treated in serum-free DMEM with 0-1 μM TSA for 24 h. TSA caused a dosedependent increase in the percentage of RPE cells in the G1 phase and a reduction of cells in the S phase ( Fig. 1A). At 1 μM TSA, the percentage of RPE cells in G1 phase was at its highest (82.6%, t test: p<0.01; ANOVA: p<0.008) and that in S phase (7.2%, t test: p<0.01; ANOVA: p<0.004) was at the lowest level (Fig. 1B). This suggests that TSA suppresses RPE cell proliferation by inhibiting its cell cycle progression at the G1 phase. To determine whether the inhibition of cell cycle progression was associated with decreased cell number, we evaluated viable cell number by PrestoBlue assay. RPE cells were treated in serum-free DMEM with 0-1 μM TSA for 24 h, and then incubated with the PrestoBlue reagent for 6 h. The effect of TSA was small but statistically significant, which reduced the viable RPE cell number to 82.4% at 1 μM of TSA (Fig. 1C, t test: p<0.0001; ANOVA: p<0.0001). TSA inhibits the expression of TGF-β-induced α-smooth muscle actin. α-smooth muscle actin (α-SMA) is a marker for epithelial-mesenchymal transdifferentiation [14]. RPE cells were pre-treated in DMEM containing 0.1% FBS with 0-0.7 μM TSA for 1 h, and then co-treated with 20 ng/mL of TGF-β1 for 72 h. TGF-β up-regulated α-SMA expression, yet at 0.5 μM and 0.7 μM, TSA inhibited the TGF-β-induced expression of α-SMA, (Fig. 4; t test: p<0.05; ANOVA: p<0.01) suggesting that TSA can inhibit TGF-β-induced EMT in RPE cells. TSA down-regulates HIF-1α and VEGF and up-regulates PEDF One possible mechanism of the inhibition of angiogenesis by TSA is through the downregulation of HIF-1α and VEGF [47] and up-regulation of PEDF [55]. To stimulate VEGF expression we treated human RPE cells with a hypoxia-mimicking agent, CoCl 2 , for 6 h in serum-free DMEM after pre-treatment with 0-0.5 μM TSA for 14 h for real-time PCR, and 18 h of TSA pre-treatment for Western blot. As shown by real-time PCR, the levels of HIF-1α mRNA were not significantly changed; but at all concentrations of TSA used, VEGF mRNA level was down-regulated even in the presence of CoCl 2 ( Fig. 5A; t test: Ã p<0.05, ÃÃ p<0.01; ANOVA: VEGF: p<0.01, PEDF: p<0.01, HIF-1α: p>0.05). At the same time, the antiangiogenic and neuro-protective PEDF mRNA level was up-regulated when RPE cells were exposed to 0.5 μM of TSA used (Fig. 5A). To examine the protein expression of HIF-1α, VEGF and PEDF, Western blot was performed on RPE cells that had received the same treatment. TSA reduced the levels of HIF-1α and VEGF protein and raised the expression of PEDF protein in the presence of CoCl 2 (Fig. 5B and 5C; t test: p< 0.05; ANOVA: VEGF, PEDF and HIF-1α: p<0.05).

TSA treatment modifies transcriptional activity on VEGF and PEDF promoters
To examine whether the effect of TSA on the protein and mRNA levels of VEGF and PEDF were due to regulation at the promoters of these genes, ChIP assay was performed with lysed RPE cells. Chromatin from untreated RPE cells or RPE cells treated in serum-free DMEM with 150 μM of CoCl 2 for 6 h, or pre-treated with 0.3 μM TSA for 18 h and then co-treated with 150 μM of CoCl 2 for 6 h, was immunoprecipitated using antibodies targeting RNA polymerase II and acetyl-histone H3. The presence of these two proteins on the promoters indicates promoter opening at these genes and that these genes are actively being transcribed. CoCl 2 treatment resulted in more chromatin pulled down by both antibodies for both VEGF and PEDF, indicating that transcription of these two genes was increased by CoCl 2 . There was a statistically significant decrease in promoter opening in TSA and CoCl 2 -treated cells, compared to CoCl 2 -only-treated and untreated cells, for VEGF, while there was a statistically significant increase in promoter opening in TSA and CoCl 2 -treated cells, compared to CoCl 2 -only-treated and untreated cells, for PEDF ( Fig. 6; p<0.05), indicating VEGF and PEDF expression may be epigenetically regulated.

TSA increases apoptosis in BCECs
To visualize the effect of TSA on BCEC apoptosis, we performed TUNEL assay on BCECs treated with 0.7 μM TSA in EBM with 1% FBS for 24 h, the highest level of TSA used in the experiments. BCECs treated with 500 μM hydrogen peroxide for 4 h were used as positive control. TSA-treated BCECs exhibited some TUNEL-positive cells (7.2%, p<0.0002, Fig. 7) as compared to untreated control, but much less than the oxidant treated positive control (46.3%, t test: p<0.00007; ANOVA: p<0.0001).

TSA inhibits tube formation in HUVECs
To assess the efficacy of TSA in inhibiting angiogenesis, untreated HUVECs and HUVECs treated with 0.3, 0.5, 0.7 or 1 μM TSA in EBM with 1% FBS for 24 h were tested for tube formation. Untreated HUVECs showed a limited amount of tube formation. In the presence of 25 ng/mL of VEGF, extensive tube formation occurred with prominent branching. Tube formation induced by VEGF was diminished in a dose-dependent manner with increasing doses of TSA. (Fig. 9A and 9B; t test: p<0.05; ANOVA: p<0.05)

TSA down-regulates VEGFR2 in BCECs and HUVECs
To study whether TSA can inhibit VEGFR2 expression, BCECs were treated in EBM containing 1% FBS with 0-0.7 μM TSA for 24 h, and the cell lysates were analyzed by real-time PCR and Western blot. TSA down-regulated VEGFR2 mRNA and protein expression in a dosedependent manner; both isoforms were prominently down-regulated in the same decreasing pattern in Western blot. (Fig. 10A-C, left panels; t test: Ã : p<0.05, ÃÃ : p<0.01; ANOVA: realtime PCR: p<0.01, Western blot: p<0.0001) Interestingly, the mature, higher molecular weight VEGFR2 was the dominant isoform expressed in BCECs. (Fig. 10B) However, the supply of BCECs was limited, and thus the effect of TSA on VEGFR2 was also examined in HUVECs. HUVECs were treated in EBM containing 1% FBS with 0-0.7 μM TSA for 48 h, and the cell lysates were analyzed by real-time PCR and Western blot. A similar dose-dependently reduced expression in VEGFR2 mRNA and protein levels was also observed in HUVECs. (Fig. 10A-C, right panels; t test: p<0.05; ANOVA; real-time PCR: p<0.01, Western blot: p<0.01) To examine more in depth the effects of TSA on the signaling of VEGFR2, the phosphorylation of VEGFR2, indicating the activation of the receptor, was studied by treating HUVECs in EBM containing 1% FBS with 0-0.7 μM TSA for 48 h, and then stimulating them with 25 ng/mL of VEGF for 10 min. As expected, VEGF induced the phosphorylation of VEGFR2 (Fig. 10D); but at the highest dose of TSA used, the phosphorylation of VEGFR2 was reduced, while total VEGFR2 protein was also down-regulated. Densitometric analyses of phospho-VEGFR2 and total VEGFR2 levels normalized by GAPDH levels showed dose-dependent reductions by TSA ( Fig. 10D; t test: p<0.05; ANOVA: p<0.01); whereas densitometric analysis of phospho-VEGFR2 levels normalized by total VEGFR2 protein levels displayed no statistically significant changes among the different TSA concentrations used (data not shown), demonstrating that the reduced levels of activated VEGFR2 was caused by decreased total VEGFR2 protein levels. Our result indicates that TSA attenuates VEGF-mediated pro-angiogenic signaling via VEGFR2 in ECs.
TSA treatment modifies transcriptional activity on VEGFR2 promoter in HUVECs ChIP assay was employed, as described above, to determine whether promoter modulation was involved in the down-regulation of VEGFR2 in HUVECs. HUVECs were untreated or treated in EBM containing 1% FBS with 0.5 μM TSA for 48 h. Compared to untreated cells, TSA treatment led to less promoter opening on the VEGFR2 gene ( Fig. 10E; p<0.05), as the acetylhistone H3 and RNA polymerase II antibodies pulled down less chromatin in TSA-treated cells, indicating that transcription activity at this promoter was reduced in TSA-treated HUVECs. This suggests that the regulation of mRNA and protein expression of VEGFR2 in HUVECs is likely regulated by epigenetics.

TSA inhibits laser-induced CNV in mice
To study the effect of TSA on the development of laser-induced CNV, the formation of CNV was evaluated by three methods: fluorescein angiography (FA), histology and measurement of CNV volume. 5 drug-naïve mice per group were administered intraperitoneal injection of 20 mg/kg of TSA or the same volume of sterile PBS immediately after laser photocoagulation and every 48 h thereafter for 7 or 14 days. Late-phase FAs of both eyes in each mouse from the control and TSA-treated groups were evaluated according to the grading system described in the Materials and Methods section [53]. Ten animals, 20 eyes and 34 lesions were examined in the control group, and 10 animals, 20 eyes and 35 lesions were examined in the TSA group. Representative angiographic images in control and TSA-treated mice on days 7 and 14 are shown in  group, and 9 animals, 9 eyes and 9 lesions were examined in the TSA group. TSA treatment resulted in statistically significant reductions in lesion area on days 7 and 14 after laser photocoagulation (Fig. 12A-B; p<0.05). Choroidal flatmount with fluorescein-conjugated isolectin B4 staining was used to assess the volume of the CNV lesions on days 7 and 14 after laser photocoagulation. Ten animals, 10 eyes and 20 lesions were examined in the control group, and 10 animals, 10 eyes and 28 lesions were examined in the TSA group. Mice treated with TSA demonstrated a significantly smaller lesion size compared to those treated with PBS (Fig. 13A). Quantitative measurement of the volumes of the CNV lesions showed that TSA treatment resulted in an approximately 65.5% and 65.9% reduction, respectively, in choroidal vascular volume, compared with that of control mice, 1 week and 2 weeks after laser photocoagulation ( Fig. 13B; p<0.05).
TSA regulates VEGF, VEGFR2 and α-SMA expression in CNV mouse model Our in vitro experiment suggests that TSA inhibits angiogenesis by down-regulating VEGF and upregulating PEDF in RPE cells, as well as down-regulating VEGFR2 in HUVECs (Figs. 5A and B; 10A and B). TSA suppresses RPE activation by down-regulating TGF-β-induced α-SMA as well ( Fig. 4A and B). A similar pattern was observed by immunostaining in our CNV mouse model (Fig. 14). Three animals, 3 eyes, and 4 lesions were examined in the control group from day 7, and 3 animals, 4 eyes, and 5 lesions were examined in the TSA group from day 7. Lower levels of VEGF, VEGFR2 and α-SMA were expressed in the CNV lesions of TSA-treated mice, in  concomitant with a down-regulation of VEGFR2 total protein level. Densitometry data (right panel) for Western blot result of VEGFR2 and phospho-VEGFR2 normalized by GAPDH levels. (*: t test p<0.05) (E) ChIP assay was performed as described in Fig. 3 on HUVECs treated with 0.5 μM TSA for 48 h. Released chromatin was then amplified by PCR using primers targeting VEGFR2 encompassing the region from 200 bp upstream of the transcription start site to 200 bp downstream of the transcription start site. Amplified chromatin was then run on a 1% agarose gel. Less promoter opening was found in TSA-treated cells than in untreated cells. Densitometry of ChIP assay result normalized by input levels. (*: t test p<0.05).
doi:10.1371/journal.pone.0120587.g010 comparison to PBS-treated controls (Fig. 14A-G;). Since PEDF is rapidly secreted once it is synthesized in RPE cells, we were unable to capture its presence by immunostaining. The regulation of these proteins by TSA in our mouse model indicates that modulation of angiogenic and EMT gene expression is one route that TSA inhibits CNV formation in mice.

TSA does not induce apoptosis in normal choroidal vasculature
To investigate whether TSA has any toxicity on normal choroidal vasculature, we performed TUNEL assay on mice that had received either PBS or TSA for 14 days, in the absence of laser photocoagulation, as well as in cryosections from the laser CNV mice. Flatmounts from three mice and six eyes were examined in the control group and three mice and six eyes were examined in the TSA group. Cryosections from three animals and three eyes were each examined in each of the PBS and TSA groups. No TUNEL-positive signal was observed in the normal choroidal vasculature in any flatmount or cryosection from either the control or the TSA group Fig 13. CNV volume measurements after TSA treatment. Mouse eyes were fixed with 10% formalin and eyecups were obtained by removing the anterior poles and neurosensory retina. Eyecups containing the RPE-choroid-sclera complex were blocked with PBS containing 1% BSA and 0.5% triton X-100 and then incubated with 10 μg/mL of FITC-isolectin B4 overnight at 4°C. Fluorescent images captured using the 20× objective of a scanning confocal microscope were analyzed. On both days 7 and 14, the sizes of CNV volumes were much smaller in TSA-treated mice than in mice that received PBS only. (*: t test p<0.05, **: t test p<0.005; n = 10/group) (bar = 50 μm).

Discussion
Inhibition of VEGF is the focus of CNV therapies that include bevacizumab, ranibizumab and pegaptanib [13]. However, these drugs must be dosed intravitreally, given early in the course of the disease; they must be administered periodically; and can only partially restore vision [13]. Most of the preclinical approaches for therapy target only one aspect of the pathogenesis [13]. Here, we demonstrated that TSA has a multi-faceted anti-angiogenic effect in RPE and ECs. We propose that TSA or related agents should be considered for their potential as an efficacious treatment option for CNV.
Inhibition of angiogenesis by HDACi has been demonstrated in cancer biology and other physiological systems [41-43; 45-47]. TSA is the prototypic hydroxamic acid HDACi [51] that regulates the activity of Classes I and II HDACs [56]. The role of HDACi's in regulating CNV formation has been previously reported but not evaluated in sufficient mechanistic detail. The synthetic HDACi N-hydroxy-7-(2-naphthylthio) heptanomide (HNHA) inhibits laser-induced CNV formation in mice [48]. Kim et al. also showed that HNHA impedes VEGF-induced tube formation and suppresses the proliferation of HUVECs, but the effect of TSA on RPE's role in promoting CNV formation was not examined [48]. Deroanne et al. studied TSA's and SAHA's anti-angiogenic effect in various angiogenesis assays, and established that TSA down-regulated the expression of the VEGF receptors in HUVECs [57]. However, they did not explore the possible epigenetic modulation of these genes via ChIP assay or the effect of these HDACi's in VEGFR2 activation [57]. Iordache et al. investigated the role of HDAC in ECs. The authors stated that mature ECs have lower angiogenic potential; on the other hand, endothelial progenitor cells (EPC) may play a significant role in vascular healing and repair. They found that HDACs participate in EPC's differentiation into EC [58]. As a critical supportive cell type of the retina and the choroid, [22][23][24][25][26][27][28] we found that TSA inhibits RPE cells' proliferation, activation, and expression of VEGF and HIF-1α. We also showed that angiogenesis can be further diminished by an HDACi through the up-regulation of the anti-angiogenic and neuro-protective PEDF. We then showed that TSA reduces CEC's proliferation and survival, and induces cell death, by regulating cell signaling. Furthermore, TSA dramatically reduces the expression of a major angiogenesis-related cell surface receptor, VEGFR2, in CEC. We demonstrated that RPE cells' angiogenic potential diminished by TSA may lead to an anti-angiogenic effect in CEC, consequently resulting in inhibition of CNV.
CNV-associated wound healing requires the expression of various growth factors [14] that initiate EMT and RPE proliferation [59]. VEGF co-localizes with α-SMA-positive transdifferentiated RPE cells [5], and quiescent, monolayered RPE cells become multilayered in human CNV membranes [60]. We found that TSA arrested RPE cell cycle at G1 phase, likely inhibiting RPE proliferation by inducing the cell cycle kinase inhibitors p21 and p15 [61]. TGF-β, the inhibition of which reduces laser-induced CNV formation in rats [62], causes EMT in RPE cells [63] by inducing the major EMT-associated transcription factor SNAIL [15]. We showed that TSA suppressed the TGF-β-induced α-SMA, an EMT marker [15], in primary human RPE cells and in laser-induced mouse CNV cryosections. Xiao et al. also demonstrated that TSA inhibits RPE cell proliferation, suppresses RPE cell cycle progression at G1, and reduces the upregulation of α-SMA in response to TGF-β, but this was performed in an immortalized RPE cell line (ARPE-19) which may not reflect the physiological state of native RPE as well as the low passage primary RPE cells used in the present study [64]. PDGF is expressed in CNVassociated wound healing [14], the concurrent inhibition of which with VEGF attenuates CNV formation [65] and is shown to inhibit cell motility [66][67][68]. TSA impeded PDGF-induced RPE cell migration, while promoting RPE cell attachment to fibronectin. We propose that the pro-cell attachment effect of TSA on RPE cells opposes RPE cell migration induced by PDGF. The inhibition of RPE cell migration elicited by TSA is unlikely to be due to TSA's antiproliferative effect, as the duration of the migration assay lasted only 5 hrs.
The expression of VEGF is increased in CNV [69], while PEDF expression is reduced in AMD retinas [70]. We showed that TSA reduced HIF-1α and VEGF protein levels and upregulated PEDF in RPE cells. However, at 0.1 μM and 0.3 μM TSA, HIF-1α protein levels were even higher than that in cells treated with CoCl 2 only, while VEGF protein levels gradually decreased from 0.05 μM to 0.5 μM TSA. HIF-1α is regulated by acetylation/deacetylation and CoCl 2 . [71][72][73][74]. In hypoxia, the acetylation of HIF-1α at lysine 709 by p300, which can be blocked by HDAC1, prevents the polyubiquitination and subsequent proteasomal degradation of HIF-1α. [71] CoCl 2 mimics hypoxia by binding the oxygen-dependent degradation domain in HIF-1α which blocks its interaction with the von Hippel-Lindau protein [72]. HDAC6 mediates the association of HIF-1α with heat shock protein 90, resulting in HIF-1α's degradation via the proteasome [73]. HDACs activities may be differentially regulated at different TSA concentrations. HDAC1 and HDAC6 may be inhibited at lower TSA concentrations, causing CoCl 2 's potentiation of HIF-1α to associate with p300 that further preserves HIF-1α's stability. Other mechanisms may be activated at higher TSA concentrations. HDAC4 deacetylates HIF-1α at its five amino-terminal lysine residues in hypoxic and CoCl 2 -treated cells, which preserves HIF-1α's stability and increases its transcriptional activity. [74] TSA may target HDAC4 at higher concentrations and reduce HIF-1α protein level. TSA reduced active transcription at VEGF's promoter, suggesting that TSA's inhibition of VEGF may be a direct epigenetic effect, possibly by inducing compaction of chromatin around the hypoxia response element of the VEGF promoter [75]. Conversely, the reduced HIF-1α protein levels after TSA treatment may abrogate the induction of VEGF expression as well. TSA exerted a statistically significant enhancement of PEDF transcription, indicating that epigenetic mechanism may be a component in the effect of TSA on PEDF expression. Interestingly, plasminogen kringle 5 (K5) reduced the levels of HIF-1α and VEGF while PEDF was up-regulated in an ischemic retinal neovascularization rat model. HIF-1α failed to translocate to the nucleus in the presence of K5, suggesting the increased PEDF expression is related to suppression of HIF-1α target genes. [76] In bovine retinal capillary endothelial cells treated with VEGF, the expression and secretion of PEDF was significantly reduced. [77] TSA's up-regulation of PEDF may be caused by the reduced HIF-1α and VEGF protein levels. Neovascularization in CNV is comparable to tumor angiogenesis that can be epigenetically regulated [45][46][78][79]. The inhibitory effect of HDACi's on carcinogenic angiogenesis implicates a common theme of targeting the HIF-1α-VEGF-PEDF axis, which may point to a new therapeutic approach in treating CNV.
HDACi's can modulate EC proliferation and functions. We found that TSA inhibited BCEC survival as shown in the TUNEL assay, perhaps targeting cells that are multiplying and engaging in angiogenesis. TSA appeared to impede the proliferation and survival of BCECs by decreasing phosphorylated p42/44, which directs cell proliferation in the presence of growth stimuli [80]. Simultaneously, TSA reduced the amount of pro-survival [81] and cyclin D1-preserving [82] phosphorylated Akt. Histone acetyltransferase (HAT) activity may require or result in the modulation of signaling pathways. TNF activates and induces the nuclear translocation of Akt that in turn phosphorylates the HAT p300, resulting in the up-regulation of ICAM-1 [83]. In hypoxia, p42/44 promotes the transactivation potential of p300 and its association with HIF-1α [84], mediating HIF-1α-induced gene expression. Furthermore, p38 and Akt are required for recruiting and then degrading p300 in nucleotide excision repair [85]. Hence, regulation of cell signaling is involved in modulating HAT-mediated pathways.
TSA activated caspase 3 in EC at higher concentrations, echoing our results in the TUNEL assay. Active caspase 3 brings about the irreversible process of cell death [86], and HDACi's induce apoptosis through the intrinsic pathway [56]. TSA also inhibited in vitro angiogenesis in HUVECs in the presence of VEGF. Neovascularization requires ECs to proliferate and survive [12]. The inhibition of p42/44 and Akt, the activation of caspase 3, and the increased amount of TUNEL-positive BCECs caused by TSA hinder BECE's angiogenic potential, which could explain the result of our in vitro angiogenesis assay. Previous studies showed that TSA impeded angiogenesis in embryoid bodies and sprouting from rat aortic rings, and reduced the mRNA expression of VEGFR1, VEGFR2, Neuropilin-1, TIMP-1 and MMP-1 in HUVECs. Taken together, HDACi's exert a broad-spectrum anti-angiogenic effect on ECs [87]. Despite the induction of cell death in EC in vitro, 14 days of treatment with TSA in vivo elicited no TUNEL+ cell death in the quiescent, normal choroidal vasculature (results not shown).
To stimulate angiogenesis, VEGF binds and activates VEGFR2, VEGF's major proangiogenic receptor, to promote the survival, proliferation and migration of ECs [88]. Macrovascular EC may in some cases respond differently to stimuli when compared to microvascular EC [89,90]. It would have been best to perform all studies with BCECs, but their supply was limited and it was necessary that some experiments were performed using HUVECs. In this study, VEGFR2 was prominently down-regulated by TSA in both microvascular BCECs and macrovascular HUVECs, and previous studies have shown that VEGFR2 is phosphorylated in response to VEGF exposure in both HUVECs and BCECs [89,91]. The down-regulation of VEGFR2 by inhibiting histone deacetylation seems counterintuitive. HOXA9, a transcription factor that promotes angiogenic gene expression [92] and induces VEGFR2, is down-regulated by HDACi exposure, causing reduced VEGFR2 levels [93]. Furthermore, the gene-activating ALL-1 histone methyltransferase complex containing HDACs 1 and 2 is present on HOXA9's promoter. Nakamura et al. suggests that ALL-1's activity may require histone deacetylation [93]. Thus, the regulation of an upstream gene by epigenetics may explain our result on VEGFR2. On the other hand, HDACs 6 and 10 can induce the depletion of VEGFR2 via heat shock proteins [94]; TSA may suppress VEGFR2 by additional non-epigenetic mechanisms. Further, VEGF-induced phosphorylation of VEGFR2 and its downstream signaling was inhibited by TSA in HUVECs. Since VEGF mediates angiogenesis through its activation of VEGFR2-directed gene expression, our result suggests that this HDACi can inhibit the survival, mitogenicity and motogenicity of ECs in the pathogenesis of CNV. These results mirror other studies of the VEGF receptors. With either of two HDACi's, SAHA or LAQ824, both VEGFR1 and VEGFR2 were down-regulated in the human colon cancer cell line HCT116 [57]. Intriguingly, VEGFR1 and several other pro-angiogenic genes are up-regulated in fibroblast growth factor 2-and epidermal growth factor-treated mouse yolk sac endothelial cells (YSECs) and HUVECs in conjunction with an increase of histone H3 lysine 56 acetylation mediated by the histone chaperone, HIRA. However, this pro-angiogenic effect is probably caused by an HDAC-independent mechanism and likely stimulus-specific, because H3acK56 incorporation and VEGFR1 expression are not restored by exposing HIRA-knocked down YSECs to a HDACi. [95].
TSA significantly attenuated CNV formation in our laser-induced mouse model. CNV leakage, as shown in FA analysis, as well as CNV volume, was much less prominent in TSA-treated mice, compared to controls. As discussed above, TSA exhibits its anti-angiogenic activity in vitro in RPE cells, BCECs and HUVECs, which likely results in its inhibiting angiogenesis in vivo. Indeed, immunohistochemical staining of retinal sections from our mouse model demonstrated that TSA inhibited the expression of VEGF, VEGFR2 and SMA in CNV lesions. As shown by Crosson et al., TSA protected rats from ischemic retinal injury and reduced metalloproteinase (MMPs) secretion by blocking the effect of TNF [40]. This implies that TSA can reduce the hypoxic and inflammatory response in the eye and arrest CNV formation. TSA can also potentially minimize retinal injury and vision loss at the involution stage of CNV. Scarring that occurs in end stage CNV is caused by the pro-fibrotic TGF-β and the ECM-modifying tissue inhibitors of matrix metalloproteinases in activated RPE cells. [96] Yet photoreceptors would be blocked by the scar tissue from receiving nutrients from the choroid and die, exacerbating vision loss [97]. We showed that inhibiting HDACs subdues the wound-healing response, which likely suppresses fibrosis. HDAC inhibition also promotes MMP expression [98], and may thus lessen the loss of photoreceptors. Suppression of wound healing by a HDACi through the inhibition of TGFβ-1-induced SMA and fibronectin has also been illustrated in corneal fibroblasts [99].
Taken together, TSA exerts a pronounced anti-angiogenic effect that can attenuate CNV formation. Future studies should be undertaken to evaluate the potential efficacy of class-specific HDACi's to minimize unforeseen off-target effects of broad-spectrum HDACi's, as well as the clinically approved HDACi's such as SAHA as potential therapy of CNV.