Blockage of PI3K/mTOR Pathways Inhibits Laser-Induced Choroidal Neovascularization and Improves Outcomes Relative to VEGF-A Suppression Alone

Abstract

Purpose: Choroidal neovascularization (CNV) is a major cause of visual loss with age-related macular degeneration (AMD). We evaluated whether blockade of phosphatidyl-inositol-3-kinase (PI3K) and the mammalian target of rapamycin (mTOR), by impairing VEGF-A and other growth factor receptors like platelet-derived growth factor (PDGF), would reduce laser-induced CNV in mice.

Methods: Choroidal neovascularization lesions were induced in C57BL/6 mice. Two groups of mice received oral GSK2126458 (3 mg/kg) or vehicle for 14 days following laser, whereas three groups were treated with GSK2126458 (6 μg/eye), aflibercept (2 μL/eye), or vehicle intravitreally on days 0 and 7 after laser. Vascular leakage was measured by fluorescein angiography (FA) on day 14. Choroidal neovascularization membranes were evaluated on choroidal flat mounts following FITC–dextran perfusion, as well as ED1 and isolectin B4 (IB4) immunohistochemistry.

Results: Oral and intravitreal (IVT) GSK2126458 reduced leakage and area of CNV lesions. Greater probability of leaking lesions (∼60%; P < 0.05) was observed in both vehicle groups. Fluorescein isothiocyanate-dextran–labeled total CNV burden area (total lesion area/eye) was reduced ∼67% (P < 0.05) and 35% (P = 0.0528) after oral and IVT GSK2126458 administration. GSK2126458 treatment reduced lesion size by ∼80% (P < 0.05) and 50% (P < 0.05) for oral and IVT control groups. Aflibercept did not alter lesion size (∼27% reduction).

Conclusions: Phosphatidyl-inositol-3-kinase/mTOR is involved in laser-induced CNV angiogenic processes. GSK2126458 effectively reduces CNV size and leakage. Choroidal neovascularization size following IVT GSK2126458 was smaller than after oral administration. Therefore, inhibition of PI3K/mTOR pathways may be more effective due to blockade of action of multiple growth factors.

Choroidal neovascularization (CNV), a common cause of vision loss in individuals older than 60 years in industrialized countries,^1,2 affects approximately 50 million people worldwide.^1,2 Early age-related macular degeneration (AMD) is characterized by the presence of drusen, hyperplasia of retinal pigment epithelium (RPE), and Bruch's membrane thickening. Advanced AMD is heralded by loss of photoreceptor–RPE–choriocapillaris complex (geographic atrophy [GA]), formation of CNV (wet AMD), or both, leading to visual morbidity.³ Wet AMD accounts for 10% to 15% of all AMD cases, and treatments are currently based on local neutralization of VEGF-A as monotherapy or in combination with photodynamic therapy.^4–7 Resistance to monotherapy Resistance to monotherapy before is developing into a challenging issue, and there is a need for additional disease-modifying therapies. Recently, better outcomes with dual inhibition of growth factors have been demonstrated with combination dosing of Fovista with ranibizumab for neovascular AMD.⁸ In preclinical models, blockade of platelet-derived growth factor (PDGF) action on pericytes renders the lesions more sensitive to the antiangiogenic properties of VEGF-A inhibitors.⁹ Such combination treatment either results in regression of established lesions or reduces the rate of lesion expansion after initiation.⁹ 

The phosphatidyl-inositol-3-kinase (PI3K) signaling pathway and a class IV PI3K serine/threonine protein kinase, the mammalian target of rapamycin (mTOR), are critical regulators of cell growth transformation, migration, and angiogenic processes. Both signaling pathways are downstream of most tyrosine kinase growth factor receptors involved in angiogenesis (EGFR, cMET, FGFR, VEGFR, PDGFR, among others).¹⁰ The PI3K pathway is therefore a good local target for inhibiting downstream effects of multiple growth factor receptors involved in retinal angiogenesis¹¹ simultaneously. Targeting of downstream signaling pathways for growth factors has the advantage of potentially achieving the effects of combining multiple specific growth factor inhibitors using a single agent, that is, VEGF-A and PDGF inhibition.⁸ Preclinical models that use receptor tyrosine kinase inhibitors targeting both VEGFR2 and PDGFR show CNV lesion regression.¹² Moreover, the nature of the agents, small molecule instead of a biologic, could provide additional advantages for creating a sustained intraocular delivery route of administration. 

GSK2126458 is a highly selective inhibitor of class I PI3K and mTOR pathways.^13,14 Studies have shown that this agent targets the critical survival pathway associated with PI3K/mTOR, and has antiproliferative effects in various solid tumors.^15,16 We hypothesized that a selective inhibitor of the PI3K/mTOR pathway could also inhibit or retard proliferation of endothelial cells associated with CNV membrane development. In this study, we examined whether oral or intravitreal (IVT) GSK2126458 could inhibit leakage and growth of laser-induced CNV lesions in the mouse and whether such treatments improve outcomes when compared to VEGF-A monotherapy. 

Materials and Methods

Ethics Statement and Animals

Experiments were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Studies also followed the GlaxoSmithKline Policy on the Care, Welfare and Treatment of Laboratory Animals and were reviewed by the Institutional Animal Care and Use Committee at the Schepens Eye Research Institute. Female C57BL/6 mice (4 weeks old) were purchased from Charles River Laboratories (Cambridge, MA, USA) and maintained in the animal facility of Schepens Eye Research Institute. Mice were anesthetized with intramuscular injections of a ketamine/xylazine mix (5:1; 0.2 mL/kg 100 mg/mL ketamine and 20 mg/mL xylazine) and placed on a circulating heating pad (37°C) during each procedure. Eyes were collected after animals were euthanized at specific time points in a CO₂ chamber. 

Laser-Induced Model of Choroidal Neovacularization

Choroidal neovascularization was induced by laser-mediated rupture of the Bruch's membrane. Pupils were dilated with 1% tropicamide (Bausch + Lomb, Tampa, FL, USA). Genteal gel (Novartis, Basel, Switzerland) was used as the optical contact and lubricant (Alcon, Fort Worth, TX, USA). Retinas were imaged with the Micron III retina imaging system (Phoenix Research Labs, Pleasanton, CA, USA). Laser spots (5 spots/eye) were applied to both eyes using the Streampix5 laser system (Meridian AG, Zürich, Switzerland). The following parameters were used to generate the lesions: 532-nm wavelength, 50-μm diameter, 50-ms duration, and 550 mW of power. A bubble was created at each laser spot. In all experiments, laser burns were conducted on day 0. 

Oral Gavage Dosing and Glucose Measurements

Two groups (n = 10 mice/group) of mice were administered GSK2126458 (3 mg/mL) or vehicle daily by gavage beginning 3 days prior to laser injury and continuing until day 14. Gavage was performed through a feeding needle (20 gauge × 38 mm; Instech, Plymouth Meeting, PA, USA) in a volume of ∼60 μL/mouse daily. Baseline blood glucose was measured on day −3 before the first oral gavage using a standard commercial glucometer (ACCU-CHEK Aviva Plus; Roche, Indianapolis, IN, USA). Blood glucose was evaluated again on day 10 post laser. A dose of 3 mg/kg has been previously noted in other animal models to be a minimally efficacious dose without obvious adverse side effects. However, this dose is associated with a sustained hyperglycemia that is known to be an on-target effect. This observation was used as a measure of target engagement. 

IVT Dosing and Glucose Measurements

Three groups (n = 10 mice/group) of mice were dosed with GSK2126458 or vehicle IVT immediately after laser injury (day 0) and on day 7. For all IVT injections, two holes were created using a 27.5-gauge needle at the limbus. A sterile Q-tip was used to gently drain a small amount of vitreous fluid (approximately 2 μL) from the first hole. Two microliters of the specified agent were injected through the second hole into the vitreous using a glass micropipette connected with a Hamilton syringe. Mice (n = 10 per group) were dosed with GSK2126458 (3 mg/mL), aflibercept (40 mg/mL), or vehicle. Baseline blood glucose was measured using a glucometer as described in the previous paragraph on days 0 and 7. Measurements were conducted during the IVT injection when the animals were under ketamine/xylazine anesthesia. 

Fluorescein Angiography and Measurement of Vascular Leakage

Fundus color photography and fluorescein angiography (FA) were performed with the Micron III retinal imaging microscope and digital imaging hardware as previously described.¹⁷ Initially on day 14, FA was performed after intraperitoneal administration of fluorescein (0.1 mL 10% wt/vol; HUB Pharmaceuticals, Rancho Cucamonga, CA, USA). Subsequent angiographic images were taken at 6 minutes after intraperitoneal administration of fluorescein. ImageJ 1.46 (National Institutes of Health, Bethesda, MD, USA)^18,19 was used to analyze images. Vascular leakage was measured by measuring the number of leaking spots (by FA) that extended beyond the original laser spot as visualized on the color fundus images. 

Measurement of CNV Areas

Choroidal neovascularization lesion sizes were quantified on choroidal flat mounts. Mice were killed on day 16 by perfusion with high molecular weight fluorescein isothiocyanate (FITC)–dextran particles (5 mg/mL) mixed in 10% gelatin (Sigma-Aldrich Corp., St. Louis, MO, USA) as previously described.^20,21 Choroidal flat mounts were visualized with an epifluorescent compound dissecting microscope (Olympus SZX16; Olympus, Tokyo, Japan). Images of the CNV lesions were captured using this microscope via a SDF PLAPO 1.6DF objective lens with an additional ×3.2 magnification coupled to a personal computer with image capture and analysis software (CellSens Standard, 1.7; Olympus). Choroidal neovascularization areas (μm²) were measured using the ImageJ 1.46 software from FITC–dextran and immunohistochemical staining (see below). 

Immunohistochemical and Hematoxylin and Eosin Staining

Choroidal flat mounts were incubated overnight at 4°C with isolectin B4 (IB4) derived from Griffonia (Bandeiraea) simplicifolia agglutinin (Vector Laboratories, Burlingame, CA, USA) diluted in blocking buffer (Li-Cor, Odyssey, Lincoln, NE, USA) containing 0.5% Triton X-100. Briefly, the choroidal flat mounts were washed three times in 1× PBS for 30 minutes and then incubated in the blocking buffer (Li-Cor, Odyssey) at room temperature for 30 minutes. After overnight incubation with IB4, flat mounts were rinsed in PBS three times (10 minutes) and incubated with a secondary antibody (Cy3, 1:500; Life Technologies, Grand Island, NY, USA) solution at room temperature for 30 minutes. Similarly, some choroidal flat mounts were probed with ED1 (1:100, mouse IgG; Abcam, Cambridge, MA, USA) using the same procedure and secondary antibody. Complexes were evaluated under an inverted fluorescence microscope (Olympus 1X51). 

For hematoxylin/eosin (H&E) staining, slides of frozen sections were washed in PBS (1×) for 10 minutes and incubated with eosin (Fisher Scientific, Pittsburgh, PA, USA) for 30 seconds, rinsed with distilled H₂O twice, and counterstained with hematoxylin (Vector Laboratories) for 60 seconds. Slides were then rinsed with distilled H₂O and dehydrated by placing them in increasing concentrations of ethanol (75%, 90%, 95%, 100%) followed by xylene. Sections were covered with transparent mounting medium and imaged with Olympus IX51 inverted microscope (Tokyo, Japan). 

TUNEL staining used directions provided by the manufacturer (Roche). Frozen sections were washed in PBS (1×) and incubated with TUNEL reaction mixture at 37°C for 1 hour, briefly washed in PBS, and incubated in substrate solution for 30 minutes at room temperature. Slides were rinsed in PBS for 2 minutes, covered with transparent mounting medium, and imaged with a Nikon E800 microscope (Tokyo, Japan). 

Statistical Analysis

Data were analyzed using the R software for statistical computing (version 3.2.1²²) under RStudio. In the figures, data are presented as either the fitted means or the geometric means, and the whiskers denote the 95% confidence limits. The analysis of laser CNV model data follows that described by Lambert et al.²⁰ Data were fitted to various generalized mixed effects models under the binomial distribution for probability of having leaking lesions; Poisson distribution for pixel count (lesion CNV areas); or normal distribution for glucose data. Models included various random intercept terms, described in detail in the legends to the figures. Data were transformed prior to the analysis using the appropriate transformations (logarithm for the Poisson count data, logit for the binomial data, and finally the base 10 logarithm for the glucose data). Multiple comparisons were conducted using Dunnett's test (for comparisons versus a control group). A P < 0.05 was used as the minimum criterion for defining statistically significant differences. 

Results

On day 0, five laser spots (50 μm, see Fig. 1, upper left) were applied to each eye (Fig. 1, lower left). A white burn core similar in size to the original laser target beam was visible on the fundus image (red circles in all parts of Fig. 1). A shockwave range (white dash circle in Fig. 1, lower left) extends beyond the burn core. Generally, the shockwave range was quite similar among the laser burns (Fig. 1, lower left). On day 14, the fundus images show a chorioretinal scar area expanding from the burn core (Fig. 1, upper right). Fluorescein angiography images were digitized from each eye at 6 minutes (to assess late leakage) and were used to determine whether the lesions were leaking. 

Figure 1

 

Fundus images of laser-induced CNV on days 0 and 14 (left and right, respectively). At top left is a fundus image of the retina with a red target laser beam prior to photocoagulation. The insets (red boxes in both right images and left bottom image) are magnified images of original laser spots at 50 μm (red circles, 5 spots/eye). The image at bottom right is a fluorescein angiography (FA) image obtained on day 14. Red circles indicate the size of the laser spot at the time of photocoagulation. The white interrupted line circles identify the area of reaction to photocoagulation. At bottom right, the white interrupted line circles identify the leaking area detected during FA.

Quantification of Vascular Leakage

Vascular leakage on FAs was defined when a diffuse circle of fluorescein expanded beyond the original lesion area (discontinuous white circles in Fig. 1, bottom right; lesions 2, 3, and 5 in Fig. 1, bottom right, were not considered leaking). In the oral and IVT treatment groups, the probability of leaking lesions was reduced by GSK2126458 administration (Fig. 2, oral and IVT; P < 0.05). In contrast, IVT aflibercept did not affect the probability of having leaking lesions (NS, Fig. 2, IVT). Interestingly, all IVT-treated groups had significantly lower probabilities of having leaking lesions than the orally treated groups (Fig. 2, crosses; P < 0.05). It is important to mention that non-CNV retinal vessels within the retinal vascular plexi were not affected by GSK2126458 treatment. This indicates that the effect of the compound is selective to the immature vessels growing during CNV development. 

Figure 2

 

Oral and intravitreal (IVT) administration of GSK2126458 reduces the probability of having leaking lesions (n = 10 mice/group). Data are presented as the predicted probabilities ± 95% confidence limits. Analysis was conducted using a generalized mixed effects binomial model with a logit link function utilizing treatment as factor and random intercept terms for cage, subject, and eye within subject. Multiple comparisons were conducted by the Dunnett's test for comparisons versus each control group and the IVT aflibercept-treated group. Asterisks indicate statistically significant (P < 0.05) differences versus the respective vehicle controls (either oral or IVT). A cross denotes statistically significant (P < 0.05) differences versus the oral vehicle group.

Measurements of CNV Burden by FITC-Dextran

Total CNV burden area was reduced by oral GSK2126458 administration (Fig. 3, asterisks; P < 0.05). A trend toward reduction in CNV burden area was also observed in the IVT-injected groups (Fig. 3; P = 0.0528). Aflibercept (IVT) did not significantly alter CNV burden (Fig. 3). Choroidal neovascularization burden areas in all IVT groups were significantly lower than in the oral treatment groups (Fig. 3, single cross). This suggests that the vehicle itself, or a process related to the IVT injections, exerted some significant antiangiogenic effects. 

Figure 3

 

Total CNV burden area (sum of individual lesion areas per eye) as measured by FITC–dextran labeling is reduced by oral or intravitreal (IVT) administration of GSK2126458 (n = 10 mice/group). Data are presented as geometric means ± 95% confidence limits. Analysis was conducted using a generalized mixed effects Poisson model with treatment as factor and random intercept terms for cage, subject, and eye within subject followed by the Dunnett's test on the natural logarithm of the data. Asterisks indicate statistically significant (P < 0.05) differences versus their respective vehicle controls. A cross indicates statistically significant differences versus the gavage vehicle, whereas a double cross denotes a P = 0.0528 versus the IVT vehicle control group.

Measurements of CNV Burden by IB4 Staining of Choroidal Flat Mounts

Isolectin B4 total CNV burden per eye was significantly (P < 0.05) reduced by GSK2126458 treatment after oral and IVT administration (asterisks in Fig. 4). Aflibercept (IVT) induced a numerical reduction in the total CNV burden area; this difference was not statistically significant (Fig. 4). As observed in the FITC–dextran measurements, total CNV burden area was in general lower in the IVT-treated groups compared to the oral vehicle control groups (Fig. 4, single crosses). 

Figure 4

 

Total CNV burden area (sum of individual lesion areas per eye) as measured by IB4 immunostaining is reduced by oral or intravitreal (IVT) administration of GSK2126458 (n = 10 mice/group). Data are presented as geometric means ± 95% confidence limits. Analysis was conducted using a generalized mixed effects Poisson model with treatment as factor and random intercepts for subject and eye within subject followed by the Dunnett's test on the natural logarithm of the data. Asterisks indicate statistically significant (P < 0.05) differences versus their respective vehicle controls. A cross indicates statistically significant differences versus the gavage vehicle.

Measurements of CNV Area per Lesion by IB4 Staining of Choroidal Flat Mounts

GSK2126458 significantly (P < 0.05) reduced individual CNV lesion size after oral and IVT administration (Fig. 5). Aflibercept (IVT) reduced CNV lesion areas slightly; however, the effect was not significant (Fig. 5, double cross; P = 0.0921). Choroidal neovascularization areas were generally lower in the IVT-treated groups (Fig. 5, single crosses; P < 0.05). 

Figure 5

 

CNV lesion areas as measured by IB4 immunostaining is reduced by oral or intravitreal (IVT) administration of GSK2126458 (n = 10 mice/group). Data are presented as geometric means ± 95% confidence limits. Analysis was conducted using a generalized mixed effects Poisson model with treatment as factor and random intercepts for eye, subject, lesion, eye within subject, and lesion within eye within subject followed by the Dunnett's test on the natural logarithm of the data. Asterisks indicate statistically significant (P < 0.05) differences versus their respective vehicle controls. A cross indicates statistically significant differences versus the gavage vehicle, whereas a double cross denotes a P = 0.0918 versus the IVT vehicle control group.

Effects of GSK2126458 on Serum Glucose Levels

Oral GSK2126458 administration elevated serum glucose levels (P < 0.05, Fig. 6, oral). Interestingly, serum glucose levels were significantly increased in all IVT-injected groups (Fig. 6, IVT). The observation that the IVT vehicle- and aflibercept-treated groups showed changes in glucose indicated that a procedural element was responsible for these effects. Glucose measurements in IVT-treated groups were conducted under ketamine/xylazine anesthesia during the injection procedure, probably contributing the glucose level increase. 

Figure 6

 

GSK2126458 increases glucose levels after oral administration (n = 10 mice/group). Changes in glucose are observed in all the animals after intravitreal (IVT) administration, indicating that a procedural phenomenon (most likely ketamine anesthesia used for IVT) is involved in addition to potential effects of systemically absorbed GSK2126458 after IVT administration. Data are presented as the fitted geometric means ± 95% confidence limits. Analysis was conducted using a mixed effects model with treatment and time as factors and a random intercept (subject) followed by the Tukey test on the base 10 logarithm of the data. Asterisks indicate statistically significant (P < 0.05) differences versus their respective vehicle controls.

ED1 and IB4 Immunohistochemistry and H&E and TUNEL Staining

Flat mount areas immunopositive for ED1 or IB4 were larger in mice treated with vehicle than in GSK2126458-treated animals (Figs. 7, 8). Whole scan images of flat mounts are presented in the upper part of the figure for each treatment category at lower magnification (Figs. 7, 8). A representative CNV lesion for each category was imaged at higher magnification and is presented in the lower part of the figure. These observations, together with their respective quantification as shown in Figures 4 and 5, suggest that GSK2126458 is an effective inhibitor of laser-induced CNV by potentially inhibiting vessel development. 

Figure 7

 

Immunohistochemistry of ED1 representing macrophage cell surface antigen (microglial) after perfusion with fluorescein isothiocyanate (FITC)–dextran gelatin. Smaller, less vascularized lesions were observed in the GSK2126458- and aflibercept-treated groups. IVT, intravitreal. Scale bars in the top line of images correspond to 200 μm but in the second row of images indicate 50 μm.

Figure 8

 

Immunohistochemical labeling of IB4 after perfusion with fluorescein isothiocyanate (FITC)–dextran gelatin. Decreased infiltration of IB4-positive cells (endothelial) was observed in the GSK2126458- and aflibercept-treated groups. IVT, intravitreal; IB4, isolectin B4. White bars correspond to 200 and 50 μm in the second row of images.

Retinal cross sections were evaluated for changes in retinal anatomy and retinal apoptosis using H&E and TUNNEL staining, respectively. Figure 9 includes eye sections of animals treated with GSK2126458 and vehicle administered by gavage/IVT injection. Retinal anatomy was well preserved without visible retinal toxicity after oral or IVT GSK2126458 administration (Fig. 9). Isolated apoptotic cells (arrowheads) were seen in some sections, but no differences were seen among the active and the control groups. In contrast, when an eye in which laser burns had been recently applied was used, abundant TUNEL-positive cells were observed within the laser spot (Fig. 9, positive control). 

Figure 9

 

Representative H&E and TUNEL staining for detection of cellular toxicity and apoptosis in eyes treated with GSK2126468 or vehicle by gavage or IVT injection. Top row, H&E. There is not observable difference in the morphology of retinal cross sections among the study groups. Middle and bottom rows, TUNEL staining at ×60 magnification (middle row) shows a few positive cells (white arrowheads) in the outer nuclear layer (ONL) in vehicle delivered by gavage and in GSK2126458 delivered by IVT injection (bottom row, at ×100 magnification). Rare positive cells were also found in the ONL of mice receiving aflibercept (white arrowheads). No positive cells were detected in retinas of mice receiving GSK2126458 either by gavage or by IVT injection. As a positive control, an eye in which recent laser burns had been placed was used. TUNEL staining was able to identify abundant TUNEL-positive (i.e., apoptotic) cells located in the area of the laser burn.

Discussion

Anti-VEGF-A monotherapy for AMD, although profoundly effective in controlling the course of disease in some patients, is often ineffective in other patient groups. Clinical observations confirm an increased prevalence of resistance to anti-VEGF-A monotherapy²³ as shown by persistent subretinal fluid and retinal edema.²⁴ Other non-VEGF-A inhibitors are being evaluated as replacement or adjunctive therapies to approved VEGF-A inhibitors. An alternative approach to blockade of multiple growth factors is the use of small-molecule inhibitors of signaling cascades at a level sufficiently downstream that it allows for simultaneous multiple growth factor action inhibition. It may be possible, therefore, to suppress a number of important pathologic processes associated with neovascular AMD with a single therapeutic agent. 

A highly selective and potent PI3K/mTOR pathway inhibitor was evaluated on CNV lesion development in a mouse laser-induced CNV model. GSK2126458 may be an effective tool to prevent cell proliferation, and hence has been postulated as useful in cancer therapy.¹⁹ The PI3K pathway is important for embryonic vasculogenesis as embryos with kinase-dead p110 develop vascular abnormalities.²⁵ Phosphatidyl-inositol-3-kinase is activated through VEGF, EGF, and PDGF signaling and participates in processes that lead to both angiogenesis and vascular permeability.²⁶ Phosphatidyl-inositol-3-kinase–mediated angiogenic signaling is partly due to its effects on regulation of nitric oxide (NO) signaling and through induction of NO synthase (specifically, eNOS).²⁷ Phosphatidylinositol-3-kinase is downstream of angiopoietin-1, which promotes endothelial survival. In addition, as many tyrosine kinase receptors involved in angiogenic processes, including EGFR, cMet, FGFR, VEGFR, PDGFR, and angiopoietin-1, have proangiogenic properties.^10,11,28,29 Despite an ability of GSK2126458 to inhibit downstream signaling of angiopoietin-1, and the potential for blood–retina barrier disruption, no effects on nonlesion retinal vessels were observed in FA assessments. GSK2126458 is therefore an attractive target for simultaneous inhibition of proangiogenic signaling pathways. Phosphatidyl-inositol-3-kinase activity also signals through the mTOR pathway in two distinct forms, mTORC1 and mTORC2. The mTORC1 (sensitive to rapamycin) affects growth processes including ribosome generation, autophagy, and protein/lipid synthesis¹¹ (independent of AKT pathways). The mTORC2 activity is related to the AKT pathway as mTORC2 phosphorylates AKT, PKC isoforms, and glucocorticoid kinase,³⁰ and its effects are for the most part related to downstream effects of AKT. Inhibition of these pathways by GSK2126458 causes profound inhibitory effects of cell cycling, growth, migration, and other components of angiogenesis. 

Data in tissue culture and mouse models indicated that the PI3K pathway inhibitors' antiangiogenic effects parallel their antitumor efficacy.³¹ Genetic ablation of class IA PI3K, specifically in the vascular endothelium, resulted in impaired vessel integrity during development and tumor angiogenesis.³¹ GSK2126458 inhibited CNV vascular leakage when administered either orally or by IVT injection. GSK2126458-treated CNV lesions were significantly smaller than those observed in vehicle-treated groups. Specific immunohistochemical staining for microglial cells using the ED1 marker showed less microglial cell infiltration at the laser photocoagulation sites in GSK2126458-treated animals (both oral and IVT). A superior effect of IVT GSK2126458 over IVT aflibercept on vascular leakage of CNV was observed. The increased suppression of overall CNV leakage is likely due to a vascular permeability and overall CNV lesion size (burden) reduction. GSK2126458 effects are observed in the absence of apparent cellular toxicity to the retina as demonstrated by H&E and TUNEL staining. In addition, the compound does not appear to affect normal retinal vasculature as shown by no leakage in retinal vessels other than CNV vessels. Since GSK2126458 reduced total CNV burden whereas treatment with aflibercept did not, it is likely that the overall lower CNV leakage is due to CNV lesion size reductions in GSK2126458-treated animals. These findings are consistent with the concept that GSK2126458 suppresses both VEGF-A and PDGF activity, since blockade of both pathways leads to a reduction in vascular permeability, manifested through suppression of VEGF-A and reduced proliferation of CNV vessels as a result of VEGF-A and PDGF signaling disruption. GSK2126458 also has a broader capacity to suppress other growth factor signaling cascades, contributing to these effects. The observation that GSK2126458 leads to greater efficacy in such CNV models may indicate effectiveness in limiting lesion growth after lesion initiation, which has been shown with anti-VEGF-A and anti-PDGF combinations.⁹ Since GSK2126458 appears to block both VEGF-A and PDGF downstream signaling, it is unlikely that GSK2126458 combination therapy would lead to additive effects; however, both the potential for additive effects and impact on existing lesions and permeability will need to be explored in future studies. 

GSK2126458 treatment led to the elevation of serum glucose, especially in the mice that received oral GSK2126458. This effect is a known pharmacologic effect of the compound and represents the result of blocking systemic insulin signaling (diabetogenic effect). High glucose levels were also observed in all mice treated by the IVT route, independent of treatment. This observation indicates that another process, in addition to GSK2126458 administration, is operating in these animals. Careful evaluation of the protocol used in the IVT cohorts suggests that glucose changes in the IVT-treated animals may be due to anesthesia induction prior to conducting the glucose measurements. Hyperglycemia appears to be a well-known effect of ketamine/xylazine anesthesia. Saha et al.³² and Brown et al.³³ demonstrated sustained hyperglycemia after anesthesia with a ketamine/xylazine mixture. Although the effects of GSK2126458 on glucose homeostasis may be apparent after systemic distribution of the agent after IVT injection, the fact that glucose levels were similar to those observed in the IVT vehicle control group indicates a limited effect of IVT GSK2126458 on plasma glucose. We speculate that the limited amount of IVT GSK2126458 that escapes into the systemic circulation is probably too low to affect systemic glucose levels. However, because of the confounding effect of the anesthesia, this cannot be definitively determined. 

In summary, our findings confirm that GSK2126458, a selective inhibitor of PI3K/mTOR pathway, blocks vascular leakage and reduces CNV lesion size in a laser-induced model of neovascular AMD. Because of its potential to block downstream signaling of multiple growth factors involved in angiogenesis, GSK2126458 could be a good candidate drug for the treatment of wet AMD—and not only to induce symptomatic relief as shown in our studies, but potentially to exhibit improved disease-modifying properties in neovascular AMD patients and in particular in patients who show a degree of insensitivity to anti-VEGF approaches. Currently, proposed anti-VEGF and anti-PDGF combined therapies aim to achieve improved disease-modifying properties in the treatment of wet AMD. These studies suggest that the same outcomes may be possible using a highly selective PI3K/mTOR inhibitor as monotherapy. 

Acknowledgments

Supported by GlaxoSmithKline through an academic collaboration grant to Schepens Eye Research Institute, Massachusetts Eye and Ear (Boston, MA, USA). 

Disclosure: J. Ma, None; Y. Sun, None; F.J. López, None; P. Adamson, None; E. Kurali, None; K. Lashkari, None