IKK2 Inhibition Attenuates Laser-Induced Choroidal Neovascularization

Choroidal neovascularization (CNV) is aberrant angiogenesis associated with exudative age-related macular degeneration (AMD), a leading cause of blindness in the elderly. Inflammation has been suggested as a risk factor for AMD. The IKK2/NF-κB pathway plays a key role in the inflammatory response through regulation of the transcription of cytokines, chemokines, growth factors and angiogenic factors. We investigated the functional role of IKK2 in development of the laser-induced CNV using either Ikk2 conditional knockout mice or an IKK2 inhibitor. The retinal neuronal tissue and RPE deletion of IKK2 was generated by breeding Ikk2−/flox mice with Nestin-Cre mice. Deletion of Ikk2 in the retina caused no obvious defect in retinal development or function, but resulted in a significant reduction in laser-induced CNV. In addition, intravitreal or retrobulbar injection of an IKK2 specific chemical inhibitor, TPCA-1, also showed similar inhibition of CNV. Furthermore, in vitro inhibition of IKK2 in ARPE-19 cells significantly reduced heat shock-induced expression of NFKBIA, IL1B, CCL2, VEGFA, PDGFA, HIF1A, and MMP-2, suggesting that IKK2 may regulate multiple molecular pathways involved in laser-induced CNV. The in vivo laser-induced expression of VEGFA, and HIF1A in RPE and choroidal tissue was also blocked by TPCA-1 treatment. Thus, IKK2/NF-κB signaling appears responsible for production of pro-inflammatory and pro-angiogenic factors in laser-induced CNV, suggesting that this intracellular pathway may serve as an important therapeutic target for aberrant angiogenesis in exudative AMD.


Introduction
Age-related macular degeneration (AMD) is the most common cause of irreversible blindness in humans over 55, and affects 10 to 15 million people in the USA [1]. Many factors contribute to AMD development, including age, genetic predisposition and environmental factors, such as smoking and possibly diet [2]. There are two forms of AMD, i.e., dry (non-exudative) and wet (exudative). Dry AMD is most common accounting for 90% of the disease and marked by a slow progressive degeneration of both retinal pigment epithelium (RPE) and overlying photoreceptors within the macula. In contrast, wet AMD is characterized by aberrant angiogenesis within the subretinal space, referred to as choroidal neovascularization (CNV) or within the retina, referred to as retinal angiomatous proliferation (RAP), and usually causes severe and rapid vision loss [2]. FDA-approved drugs for the treatment of wet AMD include Lucentis (Ranibizumab), Eylea (Aflibercept), and Macugen (Pegaptanib), all of which target VEGF to slow the growth of abnormal blood vessels and reverse the increased vascular permeability associated with new vessel formation [3][4][5]. These drugs, as well as Avastin (Bevacizumab), have been shown to be effective in the treatment of many patients with wet AMD but they require frequent intravitreal injections and resistance to anti-VEGF monotherapy in wet AMD is also apparent. Therefore, development of new therapeutic approaches or combination therapies is needed.
The cause of CNV is still under intense investigation and inflammation appears to be an important component. Drusen, a clinical feature associated with AMD and choroidal neovascularization, contains C-reactive protein, complement factors (C3a and C5a), and matrix metalloproteases [6]. Macrophages, lymphocytes and other inflammatory cell types are found in subfoveal CNV membranes [7]. Genetic polymorphism in complement factor H, a complement-inactivating factor, as well as other complement genes has a strong association with AMD [8][9][10]. These observations suggest that inhibition of inflammation may be a logical approach to prevent or delay the onset of AMD.
Mounting evidence indicates that NF-kB signaling plays a key role in controlling innate and adaptive immunity [11]. The NF-kB family consists of 5 structurally related proteins including p100, p105, relA, relB and c-Rel. In resting cells, NF-kB dimers are retained in the cytoplasm through association with inhibitory IkB proteins, such as IkBa. IkB stability is controlled by the IKK complex. There are two IkB kinases in the complex, namely, IKK1 and IKK2, with IKK2 being the major control for NF-kB activation; lack of this kinase blocks NF-kB activity and, in turn, the pro-inflammatory response [12,13]. Activation of IKK2 causes IkB phosphorylation that leads to ubiquitin-dependent protein degradation, and thus, releases IkB-bounded NF-kB from the cytoplasm. Nuclear localization of NF-kB activates the transcription of many cytokine, chemokine, and growth factors genes [13].
We hypothesize that oxidative stress and hypoxia during aging induce sustained IKK-NF-kB activation and chronic inflammation in the RPE and choroid, leading to up-regulation of angiogenic factors and aberrant neovascularization (Fig.1). To assess the functional role of IKK2 in aberrant angiogenesis, we investigated the effect of Ikk2 gene deletion or chemical inhibition of IKK2 activity on development of laser-induced CNV, a murine model of wound healing accompanied by oxidative stress and inflammation [14,15]. Since total Ikk2 knockout mice die in utero between E12.5 and E14.5 due to liver apoptosis, we generated conditional knockout mice by breeding the Ikk2 2/flox mice with Nestin-Cre mice to create deletion of Ikk2 in the RPE and retinal neurons. Deletion of functional IKK2 in the retina did not affect normal retinal development or function, but significantly reduced laser-induced CNV. Similarly, intravitreal or retrobulbar injection of an IKK2 specific chemical inhibitor, 2-[(aminocarbonyl)amino]-5-(4-fluorophenyl)-3-thiophenecarboxamide (TPCA-1), caused a similar reduction in CNV. Both in vitro and in vivo experiments showed that TPCA-1 treatment inhibited injuryinduced expression of multiple genes, suggesting that the antiangiogenic effect of IKK2 inhibition involved multiple molecular pathways. Our data showed that NF-kB signaling, through regulation of pro-inflammatory and pro-angiogenic factors, played an important role in the development of laser-induced CNV. We propose that the IKK2/NF-kB pathway may serve as a novel therapeutic target in the prevention or treatment of wet AMD.

Ethics statement
All animal experiments (Protocol: 12010) were approved by Institutional Animal Care and Use Committee, the Animal Care Committee at the University of Louisville, and were conducted in accordance with the guidelines of Association for Research in Vision and Ophthalmology on the use of animals in research.

Animals
The Ikk2 +/2 mice and Ikk2 flox/flox mice were generated and described previously [12,16]_ENREF_32_ENREF_27. The Nestin-Cre + transgenic mice (stock # 003771) [17], the mT/mG (membrane-Tomato/membrane-Green) homozygous transgenic mice (stock # 007676) [18], and the wild type control C57BL6J mice (stock # 000664) were purchased from the Jackson Laboratory. Compound lines of the Ikk2 2/flox /Nestine-Cre +/2 and control Ikk2 +/flox /Nestine-Cre +/2 mice, were generated by crossing Ikk2 +/2 /Nestin-Cre +/+ with Ikk2 flox/flox mouse lines. The Nestin-Cre +/2 / mT/mG +/2 and control mT/mG +/2 mice were generated through the cross-breeding of the Nestin-Cre +/2 mice to the mT/mG +/+ homozygous transgenic mice. All animals were maintained and used in accordance with the guidelines of Association for Research in Vision and Ophthalmology on the use of animals in research and according to the guidelines of the Animal Care Committee at the University of Louisville.

Laser treatment
The laser-induced neovascularization was performed on mice at ages between 6-10 weeks. The pupils were dilated with a single drop mixture of 0.06% tropicamide and 0.3% phenylephrine hydrochloride. Two minutes later, the mice were anaesthetized with intraperitoneal injection of Ketamine-Xylazine solution (Ketamine at 15 mg/mL saline and Xylazine at 6 mg/mL saline) at 0.05 mL per 10 g body weight. Diode laser burns applied with a Novus Spectra ophthalmic laser (Lumenis, Inc., Santa Clara, CA) mounted on a slit lamp (Model SL-M; Zeiss, Inc., Tokyo, Japan) generated four lesions symmetrically surrounding the optic nerve of each eye. A coverslip coated with 2.5% hypromellose ophthalmic demulcent solution was held on the mouse cornea, which served to subtract the optics of the cornea and lens for optimal view of the retina and spots of the laser lesions [19]. The laser variables included a 50 mm spot size, 0.05 second-duration, and power at 250 mW. The power used was assessed by the ability to produce a blister indicating rupture of the Bruch's membrane. Laser spots were evaluated by fluorescein angiography and dextran green or isolectin IB4 staining for the presence of CNV at days 7 and 10 after laser treatment.

Retrobulbar injection
Under deep anesthesia, mouse was placed on its side, and a beveled G33 needle was inserted at an approximately 45u angle into the eye, from lateral to the medial canthus, through the conjunctiva membrane. The needle was positioned behind the globe of the eye in the retrobulbar sinus after piercing through the conjunctiva. The mice received unilateral retrobulbar injection of 20 mL of 10 mM IKK2 inhibitor TPCA1 in PBS containing 20% dimethyl sulfoxide (DMSO). Retrobulbar injection of 20 mL vehicle buffer (PBS/20% DMSO) served as control.

Intravitreal injection
Right after laser injury, the mouse eye was decompressed by inserting a 30G needle through the conjunctiva and sclera at Figure 1. Schematic diagram shows targeting strategy by inhibiting IKK2/NF-kB pathway to prevent choroidal neovascularization (CNV) in AMD. Chronic inflammation and oxidative stress caused by pathologic changes, tissue damage or local components activation in subretinal space induce expression and secretion of cytokines, growth factors and extracellular matrix molecules by RPE, which in turn initiate CNV. NF-kB is a key transcription factor necessary for inflammatory and stress responses. Inhibition of NF-kB activation by IKK2 inhibitor is postulated to prevent chronic inflammation-and oxidative stress-induced CNV formation. doi:10.1371/journal.pone.0087530.g001 IKK2 and Choroidal Neovascularization PLOS ONE | www.plosone.org 1 mm behind the limbus. Microinjector UMP3 equipped with Nanofil syringe 100 mL and 33G blunt needle (Word precision instruments, Sarasota, FL) was used for injection. The needle was positioned firmly when the operator felt a slightly resistance and then 2 mL of solution was slowly injected in 10 seconds. The topical Ocuflox drops were applied after procedure to prevent infection.

Optokinetic reflex (OKR) measurements
Visual function was assessed using a non-invasive OptoMotryß optokinetic testing system (CerebralMechanics, Lethbridge, AB, Canada), following the procedures described and validated previously [20,21]. Briefly, mice standing unrestrained on the central platform tracked a rotating grating with reflexive head movement behavior. Testing was initiated by projecting a grating of low spatial frequency [0.042 cycles/degree (c/d)], rotating at 12 degrees/second at maximum 100% contrast, i.e. the bars of the gratings are maximally black and the background is maximally white. Spatial frequency of the grating was increased until the animal no longer responded. The threshold of maximum spatial frequency that the mouse could track is a correlate of visual acuity. OKR could detect visual function for each eye separately. OKR response to the rotation of the gradient in a clockwise direction is dependent upon the left eye, whereas rotation in a counterclockwise direction is dependent upon the right eye.

Fluorescein angiography
In vivo fundus fluorescein angiography of CNV was conducted using a retinal camera (model TRC 50X; Topcon America Corp., Paramus, NJ) with Canon 5D digital imaging hardware. The pupil dilation and anesthesia were performed as describe above. Fluorescein fundus images were taken at 3-6 min after intraperitoneal injection of 0.3-0.5 mL of 100 mg/mL Fluorescein (10% w/v). Lesion sizes were graded on the spatial comparison between the fluorescein leakage area and the optic nerve disc area [22], 0: no leakage; 1: the area of fluorescein leakage was equal to disc area, 2: the area of fluorescein leakage was approximately double to the disc area.
Fluorescein-labeled dextran staining and isolectin IB4 staining to evaluate CNV area Under deep anesthesia, the mouse right atrium was cut and the blood immediately was washed for 3 min with 0.5 mL/min of PBS through the left ventricle with assistance of a perfusion pump, and followed by syringe injection of 1.5 mL PBS containing 50 mg/mL fluorescein-labeled dextran (FITC-dextran, 2 million average molecular weight, Sigma-Aldrich, St. Louis, MO), and then perfusion-fixed with 4% paraformaldehyde (PFA) for 6 min. The eyes were removed and immerse-fixed in 4% PFA at 4uC for overnight. After rinsed in PBS, cornea, lens, and the neurosensory retina were carefully removed from the eyecup. Five radial cuts were made from the edge of the eyecup to the equator; the RPE/ choroid/sclera complex was flat-mounted in 50% glycerol containing PBS with the sclera facing down on a glass slide. A laser spot with green vessels was scored as CNV-positive, and a laser spot lacking green vessels was scored CNV-negative. The flatmounted retina was imaged at 56 magnification using a Zeiss Axio Imager M2 system equipped with Apotome, and attached to both AxioCam and AxioVision cameras. The free hand selection tool was used to delineate the boundaries of CNV complexes and the total area of each lesion across all sections was measured and expressed in square micrometers (mm 2 ) to determine maximum CNV complex area. The sum of CNV area in each eye was compared among different animals and experimental conditions. For isolectin IB4 staining, laser injured eyes were enucleated and fixed at 4uC with ice-cold 4% paraformaldehyde in PBS for overnight. The anterior segment, lens, and retina were removed, and the remaining eye cups with attached RPE layer were washed with PBS and incubated for overnight at room temperature with1:100 dilution of 1 mg/mL solution of AlexaFluor 594 conjugated isolectin IB4 (Invitrogen, Oregon, USA) [23]. After incubation, the eye cups were washed with PBS, the sclerachoroid/RPE complexes with radial cut were flat mounted for fluorescence microscopy right after the staining. The area of fluorescence was quantified as described above.

Histological studies
The eyes were dissected from euthanized mice and fixed immediately in 4% PFA for overnight at 4uC, and then subjected to either paraffin embedding or OCT embedding for Cryostat sectioning. Seven mm of paraffin-embedded sections was cut and de-paraffinized prior to staining with hematoxylin and eosin (H&E). The mean thickness of the CNV membrane was obtained by examining the serial sections of whole eyes and using the maximum thickness measurements of each of the four lesion sites per eye. Frozen tissue sections from the Nestin-Cre +/2 /mT/mG +/2 and control mT/mG +/2 mice were examined for the switch from red fluorescence to green fluorescence using Zeiss Axio Imager M2 system equipped with ApoTome, and attached with AxioCam and AxioVision cameras. The cell layers of the inner nuclear layer and outer nuclear layer were counted by random sampling 10 sagittal sections from each eye.

RNA isolation and Quantitative (q)-PCR
The total RNA from the spleens, brains and retinas of Ikk2 2/flox / Nestine-Cre +/2 and WT control mice were extracted using TRIzol reagent (Invitrogen). The total RNA from the RPE/Choroid/Sclera tissues of laser injury eyes was extracted using TRIzol reagent (Invitrogen) after removing the anterior segment, vitreous, and retina. The A260/A280 ratio of all RNA samples was .2.0 as measured by Nanodrop. Double-stranded cDNA was reverse-transcribed using random primers and SuperScript VILO cDNA synthesis kit (Invitrogen). Real time qPCR by a SYBR green-based PCR method was performed on a MX3005p system (Agilent Technologies, Inc., Santa Clara, CA), with a program of a 10-minute initial hot-start activation of Taq polymerase at 95uC, followed by 40 cycles of amplification (95uC for 10 seconds, 56uC for 5 seconds, and 72uC for 10 seconds). The comparative CT (threshold cycle) method normalized to b-actin was used to analyze relative changes in gene expression. The primers for human NFKBIA (IkBa) (ID 10092619a1), IL1B (Il1b) (ID 27894305b1), CCL2 (ID 4506841a1), VEGFA (ID 30172564a1), PDGFA (ID 15208658a1), HIF1A (ID 4504385a1), MMP2 (11342666a2), mouse Hif1a (ID 7363433a1), and Hmox1 (6754212a1) were designed based on the online PrimerBank database (Harvard Medical School, Boston, MA). The qPCR primer sequences for human ACTB (beta-actin), mouse Actb(beta-actin) and mouse Ikbkb (Ikk2) have been published elsewhere [24,25]. The detailed primer information is listed in Table 1.

TNFa induced IkBa degradation and TPCA-1 treatment on ARPE-19 cells
ARPE-19 cell, a human RPE cell line, was cultured according to supplier's instruction (ATCC, catalog CRL-2302, Manassas, VA). When reaching 90% confluence in 3.5 cm plates, the cells were pretreated with TPCA-1 (Tocris Bioscience, Ellisville, Missouri) at 0, 2, 20 and 200 mM for 30 minutes, and then added with TNFa at 10 ng/mL for additional 15 or 30 minutes. Control samples are treated with TPCA-1 only. Protein lysates were subjected to Western blotting for detection of IkBa degradation.

Thermal injury experiments
The confluent ARPE-19 cells cultured on 3.5 cm plates were pretreated with or without 5 mM TPCA1 (Tocris Bioscience, Ellisville, Missouri) for 30 minutes prior to thermal injury. The thermal injury experiments were carried out by replacing the medium in cell plates with 10 mL HEPES-buffered saline prewarmed at 55uC for 5 seconds before changed back to normal culture medium at 37uC in the presence or absence of 5 mM TPCA1. Cells were harvested at 10, 30, 60, 120 minutes after heat shock for protein preparation and the total RNA for qPCR was isolated using TRIzol reagents at 4 hours after heat shock.

MTT cell viability assay
The HEK293T and ARPE-19 cells were cultured in poly-lysine coated and uncoated 96-well plates, respectively. The cells at 60-70% confluence were treated in the following 3 conditions, i.e., 50 mM TPCA1 alone, 10 ng/mL TNFa alone or in combination of both TPCA-1 and TNFa after a 30 minutes pretreatment of 50 mM TPCA1. Cell without any drug treatment served as controls. The cell viability after 12 hours of drug treatment was measured by a MTT assay. Briefly, 0.2 mL of 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide (MTT) at 5 mg/mL in PBS was added to each well and the cells were incubated at 37uC for 4 hours. After the medium was removed, the purple formazan precipitate was solubilized in 200 mL DMSO. Absorbance was measured at 560 nm with background subtraction at 670 nm. Measurements were made in triplicate and the error bars represent the standard deviation. The Student's two-tailed t-test was used to determine statistical significance.

Statistical analysis
The quantitative data were presented as means 6 standard deviation (SD). CNV spot number, area, and choroidal neovascular membranes between the mutant and WT groups or drug treated and vehicle-treated controls were compared by performing a two-tail Student's t-test on raw data. P-values of less than 0.05 were considered to be significant.

IKK2 is not essential for normal retinal development and function
To examine the functional role of IKK2 in development of laser-induced CNV, we generated conditional Ikk2 mutant mice by breeding Ikk2 2/flox mice carrying one Ikk2 mutant allele and the Ikk2-flox allele [27] with the mouse line expressing Cre recombinase under control of the Nestin promoter ( Fig. 2A). To directly visualize the neuronal and RPE deletion of the loxP-flanked sequence in the eye by the Nestin promoter-driven Cre recombinase, we first bred Nestin-Cre mice into mT/mG (membrane-Tomato/membrane-Green) dual color fluorescent Cre-reporter mice. GFP expression is blocked in these mice by a loxP flanked red fluorescent protein (tdTomato) expression cassette in non-Cre cells, but allowed in cells expressing Cre after the Cre-mediated deletion of the loxP flanked tdTomato expression cassette. This double-fluorescent system allows direct visualization of Cre positive (green) and Cre negative (red) cells in the retina. Our data clearly showed that Nestin-Cre caused a selective loxP flanked genomic recombination in retinal neurons and RPE cells but not in choroidal tissue (Fig. 2B). Very few red colored cells were detected in the neurosensory retina, indicating that the Nestin-Cre mediated recombination was highly efficient (Fig. 2C). Significant reduction of Ikk2 expression at the mRNA and protein level was detected in the retina and brain, but not skin, by RT-PCR and Western blotting after Nestin-Cre mediated deletion of the loxP-flanked Ikk2 allele (Fig. 2D). These results indicated that Ikk2 was efficiently knocked out in the RPE and retinal neurons in the Ikk2 2/flox /Nestin-Cre + mice.
After confirmation of the conditional knockout of the Ikk2 gene in retinal neurons and RPE cells, we first evaluated retinal development and found no overt morphological defect in the Ikk2 2/flox /Nestin-Cre + retina by light microscopy (Fig. 3A and 3B). Both WT and Ikk2 2/flox /Nestin-Cre + retinas showed approximately 9-10 rows of nuclei in the outer nuclear layer (ONL) and 5 rows of nuclei in the inner nuclear layer (INL). The average ONL cell layers for WT and mutant mice were 9.8+0.32 and 9.6+0.48, respectively (Fig. 3C). The average INL cell layers for WT and mutant mice were 4.6+0.48 and 4.5+0.5, respectively. There were no significant differences in OKR spatial frequency thresholds [38] between Ikk2 2/flox /Nestin-Cre + and WT control mice, suggesting normal visual acuity in Ikk2 2/flox /Nestin-Cre + mice (Fig. 3D). These results indicate that IKK2 is not essential for normal retinal development or function.

Loss of Ikk2 inhibits laser-induced CNV
To determine whether the absence of Ikk2 influences choroidal neovascularization in vivo, we adopted a laser-induced CNV model [19], and evaluated the incidence of fluorescein leakage on fluorescent angiograms (FA) performed on the Ikk2 2/flox /Nestine-Cre + (Ikk2 2/2 ) retina on day 7 after laser treatment, compared to laser-induced CNV in control WT and Ikk2 +/flox /Nestine-Cre + (Ikk2 +/2 ) mice ( Fig. 4A-C). Fluorescein fundus images were taken 3-6 minutes after intraperitoneal injection of 0.3 mL of 100 mg/ mL Fluorescein (10% w/v). During this period, the size of fluorescein leakage spots was relatively stable and the area of each fluorescent spot was measured and expressed as arbitrary unit in relevant to the area of the optic nerve disc ( Figure S1 in File S1). Newly formed microvessels with clear evidence of fluorescein leakage were observed in 63% of the laser-induced lesions in WT and Ikk2 +/flox /Nestine-Cre + mice, compared to 35% in Ikk2 2/flox / Nestine-Cre + mice. The average score of fluorescein leakage was reduced from 2.5 in WT mice and 2.4 in Ikk2 +/flox /Nestine-Cre + mice to 1.2 in Ikk2 2/flox /Nestine-Cre + mice (Fig. 4D).
CNV formation was further visualized by AlexaFluor 594labeled isolectin IB4 staining (Fig. 4E-G). Although there was no significant difference in CNV area between WT and heterozygous Ikk2 +/flox /Nestine-Cre + mice (Fig. 4H), complete knockout of both Ikk2 alleles in the retina of Ikk2 2/flox /Nestin-Cre + mice resulted in a 5-fold reduction in the size of CNV lesions compared to control mice (Fig. 4H). Furthermore, the thickness of the neovascular membrane was quantified by examining serial sections and using the maximum thickness measurements as indicated as in Fig.4I-J; an 80% reduction was observed in Ikk2 2/flox /Nestin-Cre + mice compared to Ikk2 +/flox /Nestine-Cre + mice ( Fig. 4I-K). These data showed that there was a significant reduction in the size and thickness of the CNV complex in Ikk2 2/flox /Nestin-Cre + mice, suggesting that IKK2 plays a critical role in aberrant angiogenesis.

IKK2 chemical inhibitors reduce laser-induced CNV
To explore the therapeutic potential of IKK2 inhibitors in treating CNV, we next tested the effects of a potent (IC 50 = 17.9 nM) and selective IKK-2 inhibitor, 2-[(aminocarbonyl)amino]-5-(4-fluorophenyl)-3-thiophenecarboxamide (TPCA-1) _ENREF_49 [28], on laser-induced CNV. One eye of each mouse was treated with laser burns, followed by a single retrobulbar or intravitreal injection of either TPCA-1 or the control vehicle. As shown in figure 5A, a single retrobulbar injection of 20 mL TPCA-1 at 1, 5, 10, or 20 mM, resulted in a dose-dependent reduction in laser-induced CNV, with 5 mM being the minimum concentration to show an inhibitory effect (Fig. 5A). The average area of CNV spots in the TPCA-1 treated eyes was significantly reduced, especially in both the 10 mM (56 mg) and 20 mM (112 mg) groups, being approximately one third of the CNV area in control eyes (Fig. 5A). Clinical examination of the retina and OKR (on days 6 and 21) post retrobulbar injection of 20 ml TPCA-1 was normal (Fig. 5B, Figure S2 in File S1). Importantly, mice injected with control solvent either retrobulbarly or intravitreally showed no significant difference in laser-induced CNV in WT mice, indicating that the surgical procedure itself did not affect the incidence of laser-induced CNV (Fig. 5C, vehicle vs. WT). We observed that retrobulbar injection achieved a similar inhibitory effect as intravitreal injection, although the retrobulbar route required a larger injection volume (20 ml of 10 mM TPCA-1 compared to 2 ml of 10 mM TPCA-1) (Fig. 5C). We also observed that the inhibitory effect of retrobulbar injected 10 mM TPCA-1 was comparable to intravitreal anti-mouse VEGF 164 polyclonal antibody (R&D System, Cat#AF-493-NA) with regard to CNV size (Fig. 5E).
We noticed that the laser burn causes a continuous influx of inflammatory cells into the injury site with the peak at 3 days postlaser, the same time interval required for the onset of neovascularization [29]. To test if TPCA-1 will inhibit CNV development after neovascularization has started, we gave TPCA-1 3 days after the laser burn. A similar inhibitory effect on CNV was observed with either retrobulbar or intravitreal injection (Fig. 5D). These data show that the IKK2 inhibitor, TPCA-1, can inhibit CNV whether administered prior to or after the onset of angiogenesis, suggesting that IKK2 is critical for the expression of trophic factor(s) important for aberrant neovascularization.

Local administration of TPCA-1 shows no systemic toxicity
It has been previously shown that systemic administration of the IKK2 inhibitor ML120B for four days caused depletion of bone marrow B cells and thymocytes due to increased TNF-a-mediated apoptosis of lymphocytes [30]. Similar toxicity was observed in mice in which lethally irradiated hosts were transplanted with either IKK2-or p65-deficient fetal liver cells [31,32]. In order to assess the potential toxicity of retrobulbar TPCA-1 we used flow cytometry to analyze systemic lymphoid tissues. A single retrobulbar injection of 56 mg TPCA-1 for 7 days did not alter the proportion of CD4 + /CD8 + double and single positive T cells in both spleen and thymus (Fig. 6A-B). TPCA-1 treated mice showed no obvious alteration in mature (IgD high IgM low ) or immature (IgD low IgM high ) B220 + B-cell populations in the spleen (Fig. 6C), or in mature recirculating (B220 + IgD high IgM int ), transitional immature (B220 + IgM high IgD int ), immature (B220 + IgM high IgD 2 ) or precursor (B220 + IgM 2 IgD 2 ) B cells in the bone marrow either  (Fig. 6D). There was also no significant difference noticed in the monocyte populations in the spleen of TPCA-1 treated mice -i.e. neutrophil (CD11b + Ly6G + /Ly6C int ), monocyte (CD11b + Ly6G 2 / Ly6C + ) and mature macrophage (CD11b + Ly6G 2 /Ly6C 2 ) cells (Fig. 6E). Staining of CD11b and CD4 positive cells with 7AAD showed no increase in cell death (Fig. 6F). Thus, our data suggested that periocular (i.e. retrobulbar) administration of the IKK2 inhibitor, TPCA-1, did not cause systemic toxicity to the immune system. In addition, histological examination showed no noticeable morphological changes or tissue toxicity in the liver and spleen of the TPCA-1 treated mice (Fig. 6G).
It is well documented that NF-kB inhibition enhances TNFainduced apoptosis in HEK293T cells. Similarly, TPCA-1 significantly reduced the viability of HEK293 cells, and the addition of TNFa increased cell death (Fig. 6H left panel), suggesting TPCA-1 sensitized HEK293T cells to TNFa induced apoptosis. We then asked whether TPCA-1 was also able to affect ARPE-19 cell viability. In contrast to HEK293T cell, TPCA-1 showed no toxicity for ARPE-19 cells (Fig. 6H, right panel), even at the higher concentration of 50 mM which specifically blocked NF-kB activation in both HEK293T and ARPE-19 cells (Fig. 7A-B).

TPCA-1 attenuates thermal injury-induced expression of angiogenic factors
Given that activation of IKK-NF-kB signaling induced expression of growth factors and chemokines associated with angiogenesis [33], we investigated whether TPCA-1 inhibited laser-induced CNV through down-regulation of other inflammatory and angiogenic factors that promoted neovascularization. It is well known that TNF-a activates the IKK-NF-kB signal transduction pathway, leading to transcriptional activation of downstream target genes. To determine an optimal inhibitory dose for TPCA-1 to block NF-kB activation, we first performed a dose-dependent inhibition assay in HEK293T cells using a commercially available kB luciferase assay system (Promega). The kB-luciferase activity induced by TNFa was significantly inhibited by TPCA-1 at 1 mM, and was dose dependent and reaching a peak at the highest dosage used in this assay (10 mM, approximately a 7-fold reduction) (Fig. 7A). We next examined the inhibitory effect of TPCA-1 on ARPE-19 cells by visualizing TNFa-induced IkBa degradation on Western blots. In general, TNFa activates IKK2 which phosphorylates its downstream IkBa serine residues at 32 and 36, leading to ubiquitin-dependent degradation of this inhibitory protein. Degradation of IkBa releases the NF-kB dimer into the nucleus to activate downstream transcriptional targets [34]. When ARPE-19 cells were treated with 10 ng/mL TNFa, endogenous IkBa protein was rapidly degraded within 15 minutes (Fig. 7B). In contrast, pre-incubation of the cells with 20 mM and 200 mM TPCA-1 completely blocked TNFa-induced IkBa degradation (Fig. 7B).
We next explored the inhibitory effect of TPCA-1 on NF-kB activation induced by heat shock to mimic the laser injury [35]. Human ARPE-19 cells were pretreated with or without 5 mM TPCA-1 for 30 minutes and were quickly shocked by heat at 55uC for 5 seconds; they were then cultured in the presence or absence of 5 mM TPCA-1 for various times. The phosphorylation of NF-kB p65 subunit at Ser536, which mediates IkB release and DNA binding, was analyzed by Western blotting. Heat shock induced a slight increase in p65 phosphorylation at Ser 536, starting at 30 minutes after heating; however, such heat shock-induced p65 phosphorylation was significantly reduced in the presence of TPCA-1, suggesting that TPCA-1 blocked the heat shock-induced NF-kB activation (Fig. 7C). Furthermore, to examine whether TPCA-1 affected heat shock-induced gene expression, we heatshocked ARPE-19 cells at 55uC for 5 seconds. Total RNA was isolated from those cells at 4 hrs and subjected to qPCR analysis of downstream target genes. As shown in figure 7D and 7E, heat shock significantly induced expression of NFKBIA, IL1B, CCL2, VEGF-A, PDGFA, HIF1A, and MMP-2, all of which have been shown to be regulated by NF-kB signaling and participate in regulation of angiogenesis. Notably, TPCA-1 significantly blocked the activation of these downstream genes ( Fig. 7D and 7E). Next, we confirmed that mRNA expression of Vegfa and Hif1a was induced by the laser injury and that their expression was suppressed by TPCA-1 in vivo (Fig. 7F). In addition, the induced heme oxygenase enzyme HO-1 isoform (Hmox1), a sensor of cellular oxidative stress, was also significantly induced by laser injury and its response was also blocked by TPCA-1, suggesting that NF-kB/IKK2 regulates laser induced CNV formation through multiple pathways including inflammation, hypoxia, and oxidative stress (Fig. 7F). Our results suggest that IKK2 inhibition by its specific inhibitor TPCA-1 is able to block laser-induced CNV through the IKK-NF-kB signaling pathway, and in turn, the transcription of its downstream target genes that promote neovascularization.

Discussion
Abnormal growth of choroidal blood vessels into the subretinal space (i.e. CNV) is a major cause of severe vision loss in AMD. The most effective current treatment for CNV is intravitreal injection of antibodies that block the action of VEGF [2]. Because over 60% of AMD patients with CNV do not have improved vision after anti-VEGF antibody treatment, other alternative treatments are being sought. Inflammation has been linked to the development of CNV and NF-kB is a master regulator of inflammation. Therefore, we explored the effect of inhibiting the NF-kB signaling pathway in an experimental model of CNV induced by laser injury to the retina. We observed in the present study that inhibition of IKK2/NF-kB activation by the pharmaceutical IKK2 inhibitor TPCA-1 or using Ikk2 conditional knockout mice dramatically inhibited laser-induced CNV.
Clinical evidence shows that in many patients VEGF is critical for endothelial cell proliferation and angiogenesis that leads to CNV. VEGF consists of several isomers (VEGF-A, VEGF-B, VEGF-C, and VEGF-D) that have a different ability to promote angiogenesis [36,37]. Increased levels of the VEGF protein has been detected in RPE cells of CNV membranes in AMD [7,[38][39][40][41]; in basal laminar deposits of surgically excised CNV membranes; in vitreous samples of patients with neovascular retinopathies [7,39,40]. In addition, VEGF is known as a vascular permeability factor that can increase blood vessel leakage by interrupting endothelial layer tight junctions. Multiple anti-VEGF drugs are beneficial in inhibiting CNV and restoring visual function. The production of VEGF and PDGF, and the activation of multiple pro-inflammatory gene products (e.g. CCL2, ICAM-1, IL1B and IL-8) are regulated by the transcription nuclear factor NF-kB [42][43][44][45][46]. Thus, inhibition of the NF-kB pathway by pharmaceutical intervention may open a novel approach to prevent CNV [22].
Mounting evidence indicates that NF-kB also functions as a regulator of angiogenesis promoting tumor growth by upregulating the expression of angiogenic factors, such as IL-8, CCL2 and VEGF [47,48]. It has been documented that NF-kB is involved in the upregulation of VEGF mRNA in multiple cancer cell lines [49]. Although the VEGF promoter contains a region with putative NF-kB binding sites, deletion of these binding sites had little effect on VEGF expression, suggesting that direct binding of NF-kB to the VEGF promoter may not be required for its expression [50]. However, it was recently proposed that NF-kB modulates VEGF transcription though HIF1A, a key transcription factor for VEGF during hypoxia [16,51], and a recent study showed that transcriptional activation of the Hif1a gene by IKK2activated NF-kB precedes HIF1A protein accumulation both ex vivo and in vivo [52]. Mice lacking Ikk2 subjected to hypoxia revealed a pronounced defect in Hif1a mRNA expression and protein accumulation, as well as defective induction of HIF1A targeted genes, including VEGF [52]. We hypothesize that injury or inflammation in the laser injured retina may induce aberrant angiogenesis through the release of angiogenic factors that are regulated directly by NF-kB similar to that demonstrated for tumor cells [52][53][54][55][56][57][58][59]. We observed that TPCA-1, a selective inhibitor of IKK2, led to significant suppression of several pro-angiogenic molecules including IL1B, CCL2, VEGFA, PDGFA, HIF1A, and MMP-2 (Fig. 7). Further-more, we also observed the inhibition of laser-induced CNV in genetic mouse mutants in which Ikk2 was selectively deleted in RPE cells and retinal neurons but not in the vascular endothelium.
Inhibition of the NF-kB activation pathway has gained increasing interest in the treatment of several diseases including asthma, chronic obstructive pulmonary disease, and inflammationassociated cancer [60,61] because of its anti-inflammatory effects. . 20 mL of IKK2 chemical inhibitor TPCA-1 at indicated concentration was retrobulbarly injected immediately after the laser injury. CNVs were visualized by FITC-dextran staining 7 days after laser injury and quantified as area size relative to CNV area in the control mice injected with vehicle solution. (B) Visual function was measured by OKR tests performed on the laser-injured and TPCA-1 treated mice on day-6 and -21 post laser injury. The mice received a retrobulbar injection of 20 mL TPCA-1 at the concentrations indicated under the X-axis of the graph in one eye right after laser injury. ''Untreated'' represents the results from the eyes without any injection. ''TPCA-1'' represents the visual acuity from the eyes retrobulbarly injected with 20 mL TPCA-1. ''c'' represents the combined visual acuity test results from both TPCA-1 treated and untreated eyes. (C) The size of CNV area developed in the laser-induced and TPCA-1 treated eyes was normalized to the size of CNV area in the laser-induced WT eyes that had not been given any injection after laser injury. The TPCA-1 treatment was given once to the mice via either retrobulbar (20 mL 10 mM TPCA-1 or control vehicle) or intravitreal (2 mL 10 mM TPCA-1 or control vehicle) injection immediately after laser injury and, seven days later, the FITC-dextran stained images were taken and calculated. (D) The size of CNV area developed in the laser-induced and TPCA-1 treated eyes was normalized to the size of CNV area in the laser-induced, and control solvent vehicle injected eyes. The TPCA-1 treatment was given once to the mice via either retrobulbar (20 mL 10 mM TPCA-1 or control vehicle) or intravitreal (2 mL 10 mM TPCA-1 or control vehicle) injection at 3 days after laser injury. The images were taken and calculated at 7 days after laser injury. (E) Comparison of the relative sizes of CNV area in the laser-injured mice that had been treated with either retrobulbar injection of 20 mL of 10 mM TPCA-1 or intravitreal injection of 2 mL of 2 mg/mL mouse VEGF 164 affinity purified polyclonal antibody. They were normalized to the size of CNV area in the laser-injured mice that had been treated with control solvent vehicle and IgG isotype control antibody, respectively. Error bars indicate standard deviation. Two-tail T-test: * p,0.05; ** p,0.01, *** p,0.005. ''n'' indicates the number of mice used at each condition. doi:10.1371/journal.pone.0087530.g005 IKK2, the proximal kinase leading to NF-kB activation, is an excellent target for inhibition of this pathway [13]. However, oral and systemic delivery of these inhibitors has been shown to be toxic to bone marrow and epithelial cells [30,[62][63][64]. Therefore, local delivery of this drug to the eye is an attractive therapeutic alternative. Our data showed that retrobulbar injection of 56 mg TPCA-1 inhibited laser-induced CNV in mice and had no detrimental effect to liver, spleen and bone marrow derived cells (Fig. 6A-G), as well as skin epithelium (data not shown). Consistent with a previous report [65], IKK2 inhibition by TPCA-1 did not cause ARPE-19 cell death even with TNFa induction (Fig. 6H). In addition, neither genetic deletion of Ikk2 gene nor periocular administration of TPCA-1 showed an adverse effect on retinal histology or visual function ( Fig. 3 and Fig. 5B).
Anti-VEGF therapy for CNV results in significant vision gain in approximately one-third of patients with wet AMD mostly due to its anti-permeability effect but not the inhibition of aberrant angiogenesis [66][67][68]. In search of better treatments, a multi-targeted therapeutic approach has been proposed and a recent clinical study using anti-PDGF and anti-VEGF combination therapy in wet AMD has reported improved visual outcome compared with anti-VEGF monotherapy [69_ENREF_69]. Targeting multiple signaling pathways involved in CNV may be a better option to improve therapeutic results. As a key regulator of several factors promoting inflammation and angiogenesis, IKK2/ NF-kB inhibition is a rational therapeutic strategy. Another advantage of IKK2/NF-kB inhibition therapy over anti-VEGF monotherapy is the targeting of an earlier phase of angiogenesis by inhibiting inflammation. In summary, we have demonstrated that periocular injection of TPCA-1, a selective IKK2 inhibitor, is able to reduce the development of laser-induced CNV by blocking activation of the NF-kB signal pathway that is important for expression of multiple inflammatory and angiogenic factors. Our results suggest that IKK2 inhibition could be a potential novel approach to prevent or reduce CNV formation.  inhibits the TNFa-induced kB-Luciferase activity in a dosage dependent manner. 293T cells containing kB-Luciferase reporter were pretreated with or without TPCA1 for 30 minutes, then stimulated with 10 ng/mL TNFa to induce kB-Luciferase expression in the presence of the indicated amount of TPCA-1 for 5 hours, and followed by cell lysis and luciferase assay. (B) TPCA-1 inhibits TNFa-induced IkBa degradation in a dosage dependent manner. ARPE-19 cells were pretreated with TPCA-1 at indicated concentration for 30 minutes prior to addition of 10 ng/mL of TNFa for 15 and 30 minutes in the presence of the same amount of TPCA-1. Protein lysates were analyzed for the levels of IkBa and tubulin (as a loading control) by Western blotting. (C) TPCA-1 treatment inhibits the thermal injury-induced p65 phosphorylation at serine 536 amino acid residue. The ARPE-19 cells at 80% confluence in the complete DMEM/F-20 medium were pretreated with 5 mM TPCA-1 for 30 minutes and followed by heat shock for 5 seconds with addition of 10 mL HEPES-buffered saline preheated at 55uC; and then the heating medium was immediately replaced with normal complete DMEM/F-20 medium containing 5 mM TPCA-1 and further cultured at 37uC for 10, 30, 60, and 120 minutes before the protein lysates were collected. Western blotting was used to detect the total NF-kB p65 protein and the activated form of NF-kB p65 phosphorylated at Serine 536. (D and E) TPCA-1 treatment reduces thermal injury-induced and NF-kB dependent transcription of signaling protein, cytokine and chemokine, such as IkBa, IL1B and CCL2 (D), as well as of those associated with angiogenesis, such as VEGF-A, PDGFA, HIF-1A, and MMP-2 (E) in ARPE-19 cells. Cells were treated similarly as described in C. Four hours after the thermal treatment, RNAs were isolated for qPCR analysis. Relative mRNA levels were normalized to b-actin first, and then the fold increase or decrease by heat shock. Cells were treated with heat shock at the presence or absence of TPCA-1 or treated with TPCA-1 without heat shock were compared to control cells without heat shock and TPCA treatment. Error bars indicate standard deviation. Two-tail T-test: * p,0.05, n = 3. The red * indicates the comparison between the heat shocked cells and control cells. The purple * indicates the comparison between the heat shocked cells at the presence or absence of TPCA-1. (F) Relative mRNA level of Vegfa, Hif1a, and Hmox1 was shown in the RPE/Choroid/ Sclera tissues collected from laser injured eyes treated with or without TPCA-1. Mice received the retrobulbar injection of 56 mg TPCA-1 or vehicle solution and the RNA samples were prepared from RPE/Choroid/Sclera tissues at day 1 and day 2 post injury. Each RNA sample was pooled from 4 eyes in the same condition and the control sample is from the eyes without laser injury. Red and green * show the significant difference between the indicated sample and control sample without laser injury. Purple and light blue *** show the significant difference between the laser injured samples at the presence or absence of TPCA-1. Error bars indicate standard deviation. Two-tail T-test: * p,0.05, *** p,0.005, n = 3. doi:10.1371/journal.pone.0087530.g007

Supporting Information
File S1 Contains: Figure S1 Fluorescein angiography of laser induced CNV and size grading for fluorescein leakage. (A) Lesion sizes were graded based on the relative comparison to the optic nerve disk area. (B-C) Lesion sizes in the same eye were graded at 3 and 6 minutes after intraperitoneal injection of fluorescein. Figure