Activator of G-protein signaling 8 is involved in VEGF-induced choroidal neovascularization

Choroidal neovascularization (CNV) is associated with age-related macular degeneration (AMD), a major cause of vision loss among elderly people. Vascular endothelial cell growth factor (VEGF) is essential for the development and progression of AMD, and VEGF signaling molecules are effective targets for the treatment of AMD. We recently reported that activator of G-protein signaling 8 (AGS8), a receptor-independent Gβγ regulator, is involved in VEGF-induced angiogenesis in cultured endothelial cells (EC); however, the role of AGS8 in CNV is not yet understood. This study aimed to explore the role of AGS8 in CNV in cultured cells, explanted choroid tissue, and laser-induced CNV in a mouse AMD model. AGS8 knockdown in cultured choroidal EC inhibited VEGF-induced VEGFR-2 phosphorylation, cell proliferation, and migration. AGS8 knockdown also downregulated cell sprouting from mouse choroidal tissue in ex vivo culture. A mouse model of laser-induced CNV, created to analyze the roles of AGS8 in vivo, demonstrated that AGS8 mRNA was significantly upregulated in choroidal lesions and AGS8 was specifically expressed in the neovasculature. Local AGS8 knockdown in intravitreal tissue significantly inhibited laser-induced AGS8 upregulation and suppressed CNV, suggesting that AGS8 knockdown in the choroid has therapeutic potential for AMD. Together, these results demonstrate that AGS8 plays critical roles in VEGF-induced CNV.

Age-related macular degeneration (AMD) is a major cause of blindness among elderly people in developed countries 1 . It can be categorized into "dry" and "wet" forms 2 . Wet AMD is characterized by the growth of abnormal vessels from the choroid into the subretinal space, known as choroidal neovascularization (CNV). It occurs less frequently than the dry form but is responsible for 90% of cases of acute blindness 3 . Newly formed blood vessels from the choroid usually break through Bruch's membrane and develop under the retinal pigmented epithelium (RPE) 4 . Development of wet AMD is associated with local production of VEGF, therefore inhibition of VEGF signaling with anti-VEGF agents (i.e., ranibizumab, pegaptanib sodium, and aflibercept) has revolutionized the treatment of AMD via an antiangiogenic potential 5,6 . Although VEGF has essential functions in the pathophysiology of the cardiovascular system, several adverse events have been reported after the administration of anti-VEGF agents, including thromboembolic events, myocardial infarction, stroke, hypertension, gastrointestinal perforations, and kidney disease [7][8][9] . Therefore, there is an urgent need for alternative approaches to the treatment of CNV.
Heterotrimeric G-protein signaling plays multiple important roles in physiological and pathophysiological conditions including angiogenesis 10 . Conventionally, heterotrimeric G-proteins are activated by G-protein-coupled receptors (GPCR) in response to ligand stimuli; however, it is recognized that they are directly regulated by accessory proteins, which regulate G-protein signaling independently of GPCR activation 11 . We previously identified activator of G-protein signaling (AGS) 8, also known as fibronectin type III domain containing 1, from a repetitive transient ischemia model of the rat heart. AGS8 interacts directly with Gβγ subunits 12 and plays a pivotal role in hypoxia-induced apoptosis of cardiomyocytes in association with Gβγ 13,14 . Since the rat heart in the transient ischemia model from which we cloned AGS8 has highly developed collaterals 12 , we hypothesized that AGS8 is involved in the regulation of vascular development. Therefore, we further investigated the roles of AGS8 in vascular formation and found that knockdown of AGS8 by small interfering RNA (siRNA) reduced cell surface localization with control siRNA or AGS8 siRNA#1. After 48 h, AGS8 mRNA expression was analyzed by real-time PCR assay. Data are expressed as means ± standard error of the mean (s.e.m) from four experiments performed in triplicate. ** P < 0.01 (unpaired t-test). (B, left panel) Tube formation analysis of AGS8 knockdown in RF/6A cells. Cells were transfected with control siRNA or AGS8 siRNA#1, and seeded on Matrigel coated plates. Cells were treated with VEGF at 10 ng/mL and maintained for 6 h, and images were taken. (B, right panel) Tube length was digitally quantified. Data are means ± s.e.m. from four independent triplicate experiments. ** P < 0.01 (two-way ANOVA with Tukey's correction). N.S., not significant (P ≥ 0.05; unpaired t-test). (C) Phosphorylation of VEGFR-2 in RF/6A cells. RF/6A cells were transfected with control siRNA or AGS8 siRNA#1, then cultured for 48 h, starved and stimulated with 25 ng/mL VEGF for 10 min. Cell lysates were immunoprecipitated with VEGFR-2 antibody and subjected to immunoblotting with phosphotyrosine or VEGFR-2 antibodies. Data are means ± s.e.m. from four independent experiments. ** P < 0.01 (two-way ANOVA with Tukey's correction). N.S., not significant (unpaired t-test). Full-size images of immunoblots are of AGS8 and inhibited VEGF-induced tube formation, cell growth and migration, as well as the phosphorylation status of signaling molecules including VEGFR-2, ERK1/2, and p38/MAPK. Furthermore, AGS8-Gβγ formed a protein complex with VEGFR-2, suggesting a potential role for AGS8-Gβγ in VEGF signal processing during angiogenesis 15 . In addition, we recently demonstrated that AGS8 is also involved in the regulation of VEGFC-induced lymphangiogenesis through VEGFR-3 regulation in lymphatic endothelial cells 16 but is not involved in FGF-or EGF-induced cellular events 15,16 , suggesting that AGS8 is a specific VEGF signaling regulator.
To date, however, the pathophysiological significance of AGS8 remains unclear in in vivo animal models. Here, to elucidate the potential involvement of AGS8 in CNV, the roles of AGS8 were examined with an in vitro choroidal endothelial cell culture, ex vivo choroid explant culture, and an in vivo murine CNV model, which is an established model that closely mimics the pathogenesis of human AMD. We demonstrate here for the first time that AGS8 is involved in the development of CNV and is a potential therapeutic target for AMD.

Inhibition of AGS8 attenuates VEGF-induced cellular events in RF/6A choroidal endothelial cells.
To examine the role of AGS8 in CNV, we first examined the effect of AGS8 knockdown in cultured choroidal endothelial cells, RF/6A cells, which originate from rhesus choroid/retina tissues and are frequently used for in vitro CNV analyses [17][18][19] . Transfection of RF/6A cells with siRNA successfully inhibited the expression of AGS8 mRNA (18.5 ± 3.2% versus control (mean ± s.e.m); ** P < 0.01; Fig. 1A). Since VEGF is a primary factor necessary for the initiation of sprouting angiogenesis by binding to VEGFR-2 20 and is known to play a crucial role in pathological angiogenesis, CNV and wet AMD, the effects of AGS8 knockdown in VEGF-induced cellular events in choroidal ECs were analyzed. VEGF-induced tube formation of RF/6A cells on Matrigel was significantly attenuated by AGS8 siRNA#1 (55.2 ± 2.9%; ** P < 0.01; Fig. 1B).
The influence of AGS8 knockdown on VEGF-mediated cellular events was further analyzed. VEGF-induced proliferation of cells was inhibited by AGS8 knockdown (52.6 ± 3.1% versus control; ** P < 0.01; Fig. 2A) determined by MTT assay. VEGF-induced migration of cells was also attenuated by AGS8 knockdown (21.6 ± 5.1% versus control; ** P < 0.01; Fig. 2B) determined by a transwell migration assay. Two other siRNAs targeting different segments of AGS8 reduced the expression of AGS8 mRNA and inhibited the VEGF-stimulated growth of RF/6A cells (Fig. S3A,B). These data suggest that AGS8 is involved in VEGF-mediated cellular events in choroidal ECs.
Inhibition of AGS8 attenuates cell sprouting in mouse choroid explant culture. Next, we examined the role of AGS8 in a choroid sprouting assay, which is an ex vivo experimental model of CNV. Sprouting of vascular ECs from the choroid explant reproduces the processes of microvascular angiogenesis, including cell proliferation, cell migration, and tube formation 21 . Mouse choroid was dissected from the retina, and the fragments were embedded in Matrigel and cultured for 4 days. The cells growing out of the explants were stained with the endothelial marker isolectin and AGS8 (Fig. 3A). Flow cytometric analysis indicated that almost 70% of cells spreading out from the explant were CD31-positive endothelial cells (70.1% ± 2.04, mean ± s.e.m, n = 4) (Fig. 3B), which was consistent with a previous report 21 . To analyze its role, AGS8 was knocked down by siRNA transfection of the explants at days 2 and 3 of culture, and the culture was continued up until day 4. Real-time polymerase chain reaction (PCR) showed that transfection of AGS8 siRNA attenuated the expression of AGS8 in the migrated cells (24.2 ± 4.1% versus control; Fig. 3C). Finally, the area occupied by migrated cells was digitally quantified; it was found that an area of cells sprouting out from the explant was significantly reduced by AGS8 knockdown (54.2 ± 5.7% versus control, ** P < 0.01; Fig. 3D). The reduction of the sprouting area by AGS8 siRNA probably reflects AGS8 knockdown in endothelial cells; however it may include additional effects of AGS8 knockdown in other cells such as pericytes and fibroblasts. Separate experiments indicated that a VEGFR-2 inhibitor significantly attenuated the outgrowth of cells from explants in a dose-dependent manner (Fig. 3E), suggesting that AGS8 is involved in the regulation of VEGFR-2 signaling, which is critical for the initiation of microvascular angiogenesis in the choroid.

Induction of AGS8 expression in a laser-induced mouse CNV model. The laser-induced CNV
model is widely used as model of AMD with pathological angiogenegsis 22,23 . Two days after argon laser-induced injury of Bruch's membrane, mouse eyes were dissected and the choroidal tissues associating with RPE were presented in Supplementary Fig. 1. (D) Effect of AGS8 knockdown on the VEGF-induced phosphorylation of p38/MAPK, AKT, and ERK(1/2) in RF/6A cells. At 48 h after transfection of control siRNA or AGS8 siRNA#1, the cells were stimulated with VEGF at 10 ng/mL, and cell lysates (10-20 μg) were subjected to immunoblotting. Developed digital images were quantified by densitometric analysis. The values were normalized by p38/MAPK, AKT, and ERK(1/2). * P < 0.05, ** P < 0.01 (two-way ANOVA with Tukey's correction). Data are the mean ± s.e.m from 4 independent experiments. Full-size images of immunoblots are presented in Supplementary    (191.2 ± 9.1% versus control, P < 0.01; Fig. 4B). These results show that AGS8 was expressed in newly formed vessels in a laser-induced AMD model mouse.
Suppression of CNV in AGS8 knockdown in a mouse CNV model. Next, we attempted to knockdown AGS8 in the choroidal tissue by direct injection of siRNA into the vitreous 24 . After laser induction into mouse eyes, 20 ng of AGS8 siRNA or control siRNA combined with transfection reagent were concentrated to 1 µL and delivered to the intravitreal body by injection. AGS8 mRNA expression was significantly upregulated in CNV lesions (laser+) compared with control tissue (laser−) with control siRNA on day 2 (181 ± 18.0%) and day 4 (221 ± 18.5%; † † < 0.01, † < 0.05; Fig. 5A) and significantly attenuated by AGS8 siRNA on day 2 (43.9 ± 7.2% versus control) and day 4 (48.2 ± 10.7% versus control; **P < 0.01). Either 4 or 7 days after laser-induction, RPE-choroid tissues were isolated and co-immunostained with CD31 and AGS8 antibodies, indicating that the AGS8 expression observed in CNV was reduced by AGS8 knockdown (Fig. 5B). Interestingly, the CD31-postive area in the CNV, which was co-localized with AGS8, was attenuated by AGS8 knockdown (Fig. 5B). Neovessel areas measured on day 7 indicated that AGS8 knockdown significantly prevented the formation of CNV (56.2 ± 3.9% versus control; ** P < 0.01; Fig. 5C). Together, these results demonstrate that AGS8 was involved in the formation of CNV in a mouse AMD model and AGS8 loss of function potentially inhibits AMD by preventing or ameliorating CNV.

Discussion
In this study, we demonstrated novel regulation of CNV via an accessory protein for G-proteins, AGS8. Knockdown of AGS8 in the cultured choroidal cells inhibited VEGF-induced tube formation on Matrigel and the phosphorylation of VEGFR-2, as well as VEGF-induced cell migration and cell proliferation. In an ex vivo mouse choroid explant culture model, AGS8 knockdown significantly inhibited endothelial cell sprouting. In the laser-induced mouse AMD model, AGS8 was induced in neovessels on days 2 and 4 after surgery. Interestingly, intravitreal AGS8 siRNA injections significantly inhibited CNV formation and the vascular budding area of the RPE-choroid complex. These findings complemented the in vitro study, which showed that the molecular mechanism of angiogenesis is mediated by AGS8 15 and demonstrated the regulation of angiogenesis in vivo by accessory proteins for G-protein. Our data also suggest the potential of AGS8 as a therapeutic target to control neovascularization in human diseases. The mechanisms of CNV on AMD are complicated and have not yet been clarified 25 . It is now well known that VEGF plays a crucial role in abnormal blood vessel development in CNV 26 and that the inhibition of VEGF signaling can effectively control angiogenesis. In fact, intravitreal injections of anti-VEGF agents pegaptanib and ranibizumab have currently been approved for AMD treatment, while off-label use of bevacizumab has also become common 26 . Since VEGFR-2 is essential in almost all VEGF-mediated responses in pathological angiogenesis [27][28][29] , apatinib, a VEGFR-2 inhibitor, also effectively inhibits CNV-at least in mice 30 . We previously demonstrated that AGS8 regulated VEGF signaling via VEGFR-2 regulation in vascular endothelial cells in vitro 15 . Therefore, we attempted to control angiogenic events in an in vivo animal model through the suppression of AGS8. AGS8 knockdown successfully exerted anti-VEGF effects by preventing VEGF-mediated signaling, which led to the suppression of CNV in a mouse AMD model. This observation provides additional information on how to control the development of CNV.
Anti-VEGF therapies targeting VEGF have become integral components of anticancer regimens for many tumor types 31 and ocular diseases such as diabetic retinopathy 32 and AMD. Intravitreal injection of anti-VEGF agents has revolutionized the treatment of AMD, and these agents have been reported as highly effective for improving visual function. However, because VEGF is involved in a wide variety of physiological process, anti-VEGF agents carry potential risks of adverse events. Repeated and long-term injections of anti-VEGF agents may increase the chance of the systemic complications of thromboembolic events, myocardial infarction, stroke, hypertension, gastrointestinal perforations, and kidney disease [7][8][9] . Thus, AGS8 targeting could be an alternative approach for the specific downregulation of VEGF signaling in CNV along with fewer adverse effects because its expression was induced after laser-induced CNV and was limited in the neovasculature in the laser-induced lesion, not in the pre-existing vasculature.
AGS8 is an accessory protein for heterotrimeric G-proteins isolated from a repetitive transient ischemia model of the rat heart with extensive collateral development 12 . AGS8 interacts with Gβγ subunits and regulates cellular function without the activation of GPCRs 12,13 . Our previous study demonstrated that AGS8 is involved in the trafficking of VEGFR-2 to the cell surface in endothelial cells. Furthermore, the expression of c-terminal AGS8 increases the expression of VEGFR-2 in cultured cells 15 , which might due to the stabilization of VEGFR-2 in the AGS8-Gβγ protein complex. Recently, it was reported that VEGF mRNA and protein expression are induced in RPE cells and the receptor VEGFR-2 is also present in choroidal ECs in a laser-induced murine AMD model 33 .
Here, AGS8 was induced into the choroidal ECs in the neovasculature and might have contributed to development of CNV in choroidal ECs by regulating the VEGF signaling pathway.
In summary, the information we provide here is novel. AGS8 expression was upregulated in the CNV no later than 2 days after laser induction, and the inhibitory effect of AGS8 knockdown was significant when the CNV began to form. Restriction of AGS8 expression to within the neovasculature suggested that targeting AGS8 may cells from choroidal explanted cultures. (D, right panel) The areas of cells sprouting on day 4 were quantified by ImageJ software. Data are means ± s.e.m. from five time-independent experiments. ** P < 0.01 (unpaired t-test). (E) The explants were treated with vehicle or VEGFR-2 inhibitor for 4 days. Images were taken and the area of cell sprouting was quantified by ImageJ software. Data are the mean ± s.e.m. from 12 explants in 2 timeindependent experiments with similar results. ** P < 0.01 (unpaired t-test). be advantageous because it would not cause off-target effects in pre-existing vessels. This new information contributes to our understanding of CNV in human AMD.  34 , which has been used in biological studies of endothelial cells [17][18][19] . VEGFR-2 inhibitor was purchased from R&D Systems (Minneapolis, MN).  Expression of AGS8 mRNA in CNV tissue. At 2 or 4 days after laser-induced CNV surgery and intravitreal siRNA injection, mouse eyes were dissected, choroid was isolated, total RNA was purified, and real-time PCR was performed to evaluate the effect of AGS8 siRNA#1-induced knockdown (AGS8KD). Data are means ± s.e.m from four time-independent triplicate experiments. † P < 0.05, † † P < 0.01 (unpaired t-test), **P < 0.01 (two-way ANOVA with Tukey's correction). (B) SiRNA was injected into the vitreous for AGS8 knockdown in the choroid following laserinduced CNV surgery. Newly formed vessels were visualized on days 4 and 7 by immunohistochemical staining with CD31 (red) and AGS8 (green) antibodies. Representable images were shown.

Laser-induced experimental CNV Model.
Eight-week-old C57BL6/N male mice were used for laser-induced CNV induction according to a previous report 36 . Briefly, after anesthesia was induced with intraperitoneal injection of 1% pentobarbital sodium (50 mg/kg; Kyoritsu Seiyaku, Tokyo, Japan), the pupils were dilated using 0.5% tropicamide and 0.5% phenylephrine (Santen Pharmaceuticals, Osaka, Japan), followed by topical anesthesia using 0.4% oxybuprocaine hydrochloride (Santen Pharmaceuticals). Bruch's membrane was injured under a slit-lamp microscope with an argon laser photo coagulator using the following parameters: 514 nm wavelength, 200 mW intensity, 0.1 s duration, and a 100 μm lesion diameter. Equal numbers of laser lesions were generated in the control eye and AGS8 knockdown eye (6-8 spots per eye) in every preparation. On specific days indicated, choroids were isolated and processed for the following experiments after laser induction. ReverTra Ace (Toyobo, Osaka, Japan), respectively. Real-time PCR was carried out with KAPA SYBR Fast qPCR kit (Nippon Genetics, Tokyo, Japan) and STEP one plus (Bio-Rad) according to the manufacturers' instructions. The primers for real-time PCR were as follows: mouse AGS8 forward, 5′-TCTGGACACACGACTTCTGC-3′; reverse, 5′-GCTGTCTCTCCCATTTGGCT-3′; primer sequences for rhesus AGS8, which agrees with human primer sequence designed previously 15 , were forward, 5′-TTCCGTAACCCTCTCTCCCG-3′; and reverse, 5′-AACCCACGATCAAGGTCCAC-3′.

Intravitreal injection of siRNA to knockdown gene expression in choroid.
Immunohistochemistry. The explants were cultured on glass coverslips (Matsunami Glass, Osaka, Japan) coated by matrigel. The explants with sprouting cells cultured for 4 days were fixed with 4% PFA for 5 min, and choroid explants were removed. Following washing with PBS, the cells sprouted were incubated with 5% normal donkey serum in PBS for 1 h. The cells were incubated with AGS8 antibody13 and fluorescein labeled isolectin B4 (Vector Laboratories, Burlingame, CA) for 18 h, washed with PBS, incubated by HRP-labeled anti-rabbit IgG antibody for 1 h, and followed by TSA signal amplification system (Perkin Elmer, Shelton, CT) for AGS8 detection. The coverslips were mounted on a slide glass (Matsunami Glass) with VECTASHIELD mounting medium (Vector Laboratories, Burlingame, CA, USA). Images were obtained under an LSM710 fluorescence microscope (Zeiss, Oberkochen, Germany). Immunohistochemistry of RPE-choroidal flat-mounts was performed as described previously 36 . Mouse eyes were enucleated and fixed in 4% PFA for 30 min on ice. After the cornea, lens, retina and sclera were removed from RPE-choroid in ice-cold PBS, 100% methanol was substituted for PBS for fixation and the tissues were preserved at −20 °C in a freezer. On the day of experiments, methanol was replaced by PBS and 0.5% Triton X (PBST) at room temperature, and PBST with 5% goat serum (Jackson ImmunoResearch, West Grove, PA) blocking buffer for 30 min, before incubation with anti-mouse CD31 antibody (rat, 1:500; Thermo Fisher), or anti-mouse/human AGS8 antibody 13 (rabbit, 1:200) overnight at 4 °C. The next day, secondary antibodies, Cy3-conjugated rat anti-rat (1:500; Jackson ImmunoResearch) or Alexa-488 conjugated goat anti-rabbit (1:500; Invitrogen) were incubated for 3 h at room temperature. RPE-choroid was washed with PBST, cut by a razor blade to make flat mounted tissue, and the tissues were covered with a thin layer of VectaMount (Vector Laboratories, Burlingame, CA) and a coverslip before examination with a fluorescence microscope, Keyence BZ-9000 (Osaka, Japan). statistical analysis and miscellaneous procedures. All experiments were performed in duplicate, triplicate, or greater and repeated independently at least four times. All statistical analyses were performed with IBM SPSS software. Data were analyzed with one-way or two-way ANOVA followed by post-hoc tests for comparisons of the means. In the case of comparisons in pairs, unpaired t-tests were used. Statistically significant differences between the control and test groups were evaluated if the P value was lower than 0.05 (*) or 0.01 (**) as indicated on the graphs. Error bars in the graphs denote standard errors of the mean (s.e.m). Other miscellaneous procedures have been previously described 37 .