Assessing a Novel Depot Delivery Strategy for Noninvasive Administration of VEGF/PDGF RTK Inhibitors for Ocular Neovascular Disease

Abstract

Purpose.: Two noninvasive delivery strategies for VEGF/PDGF receptor tyrosine kinase inhibitors (RTKI) were explored that exploited uveal retention as a means for establishing an ocular drug depot: a single oral “loading” dose and topical administration.

Methods.: Melanin binding was confirmed by centrifugation and mass spectrometry. Ocular retention was examined in pigmented and albino rats. Ocular release kinetics were measured 3 to 28 days postdosing in pigmented rats. Microautoradiography was used to demonstrate retention of RTKI in the uveal tract. A uveal drug depot of pazopanib was created by a single oral dose prior to induction of laser choroidal neovascularization (CNV). Choroid/retinal pigmented epithelium (RPE) retention of a related RTKI with enhanced topical bioavailability, GW771806, was confirmed by bioanalytics, and its ability to regress CNV compared with pazopanib.

Results.: Pazopanib and GW771806 directly bound melanin and were retained within the uveal tract of pigmented rats for weeks following a single oral dose. Pazopanib was undetectable systemically following a single oral administration prior to CNV induction, and reduced CNV as well as twice daily dosing. Topical ocular delivery of GW771806 at 5 mg/mL led to high choroidal/RPE exposure and significantly regressed CNV lesions; 2 mg/mL prevented lesion progression.

Conclusions.: Uveal retention of drugs such as pazopanib can be used to create a sustained-release depot. Topical GW771806 regressed CNV. These data indicate that topical or infrequent oral loading dose treatment with VEGF/PDGF RTKI retained in the choroid/RPE might allow noninvasive treatments for ocular neovascular disease.

Introduction

The discovery of the vascular endothelial growth factor A (VEGF-A) pathway as an effective therapeutic target for pathological ocular neovascularisation, particularly in choroidal neovascularization (CNV) in age-related macular degeneration (AMD) has revolutionized its management. Although anti–VEGF-A therapy is not a cure for CNV in AMD, it has been proven to improve vision and slow disease progression by reducing both excessive vascular permeability and angiogenesis induced by VEGF-A. ^1–3 Currently, all VEGF-A antagonists approved for clinical use are macromolecules, including pegaptanib (an RNA aptamer that binds to and neutralizes VEGF-A); ranibizumab (a humanized fragment of an anti–VEGF-A neutralizing antibody^2,4–6 ); and aflibercept (a fusion protein containing VEGF-A–binding regions of VEGFR-1/2 and the constant region of human IgG1 ^{7 ,8} ). These macromolecular VEGF-A antagonists are also being tested for treatment of other VEGF-driven neovascular diseases in the eye including diabetic retinopathy. ^9–13 The molecular size and biochemistry of current VEGF-A antagonists necessitates their delivery by invasive and frequent intraocular injection, ^2,14,15 adding inconvenience and a significant cumulative risk of sight-threatening complications. 

Consequently, novel sustained-release formulations and delivery devices for macromolecular VEGF-A antagonists are being developed. ^16–18 Small molecules that target the activity of multiple tyrosine kinase–containing receptors, such as VEGF receptors (VEGFRs) and platelet-derived growth factor receptors (PDGFRs), do not require invasive delivery and are being tested as treatments for neovascular diseases in the eye, with the potential to be more effective in regressing neovascular membranes. ^19–28  

The potential for receptor tyrosine kinase inhibitors (RTKIs) in the treatment of CNV without intravitreal delivery has been explored using pazopanib, ^29–33 a small-molecule low nanomolar inhibitor of VEGFR, ³⁴ PDGFR, and c-Kit tyrosine kinases. ^33–37 Pazopanib has been shown to be effective in inhibiting the development of CNV by >90%; it can also cause regression of established neovascular membranes following oral delivery. ³³ Here we investigate two distinct approaches for noninvasive delivery of pazopanib and a related RTKI, GW771806, which has similar cellular and enzyme profiles but has a more attractive solubility profile. We determined if uveal-bound pazopanib might act as a sustained-release depot, having biological activity in the absence of detectable plasma concentrations. In addition, we chose to explore an alternative to oral dosing, using topical delivery of GW771806 to achieve effective chorioretinal exposure. Our findings highlight a novel strategy for drug delivery using a melanin-binding strategy, and the potential for less invasive delivery of RTKIs for treatment of ocular neovascular disease. 

Methods

Animals and Welfare

Eight-week-old female wild-type mice (C57Bl/6 6JOla Hsd and C57Bl/6NCrl), female Dark Aguti (DA) rats, Long–Evans (LE) rats, and Sprague–Dawley (SD) rats were obtained from a commercial supplier (Harlan UK Ltd., Blackthorn, UK). Animals were anesthetized with a single intraperitoneal (IP) injection of medetomidine hydrochloride (Domitor; Pfizer Animal Health, New York, NY) and ketamine (Fort Dodge Animal Health Ltd., Southampton, UK) in sterile water. Anesthesia was reversed by a synthetic α2-adrenergic antagonist, atipamezole IP (Antisedan, 20%; Orion Pharma, Espoo, Finland). When necessary, the pupils were dilated with 1 drop each of phenylephrine hydrochloride 2.5% and tropicamide 1% (Chauvin Pharmaceuticals Ltd., Kingston-Upon-Thames, UK). All animal procedures were ethically reviewed and approved by the British Home Office Animals Scientific Procedures Act 1986 and performed in accordance with European Directive 86/609/EEC, the GlaxoSmithKline (GSK) Policy on the Care, Welfare and Treatment of Animals, and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Delivery of Bioactive Compounds (GSK, UK)

For mice, 0.2 mL of a suspension containing pazopanib or vehicle was delivered by oral gavage using a standard gavage needle. For rats, topical pazopanib (in 25 mM sodium phosphate, 7% Captisol, 20 mM sodium chloride, pH 5) and GW771806 (in 25 mM sodium phosphate, 15% Cavitron 82004, 10 mM sodium chloride, pH 7.5) were administered bilaterally twice a day (BID) as 30 μL drops to each cornea. Neutralizing anti-VEGF domain antibody (dAb) (DMS1529), 2 μL at 1 mg/mL was injected intravitreally into both eyes using a 30-G needle (Hamilton Company, Reno, NV) after experimental CNV induction (see Supplementary Material and Supplementary Tables S3 and S4). 

Melanin Binding Studies

Sepia officinalis melanin (1 mg/mL in Dulbecco's phosphate-buffered saline [DPBS] with 0.01% NaN₃) was mixed 1:100 with each test compound (1 μM in DMSO) in a deep-well plate (Eppendorf 96-well Protein LoBind Deepwell Plate; Eppendorf AG, Hamburg, Germany). DPBS without melanin was used as the control. The sealed plate was incubated at room temperature for 1 hour on a rotary shaker set at 500 rpm. Following centrifugation at 10,000 rpm for 10 minutes, 1 μL of supernatant was mixed 1:100 with 10% methanol in water and analyzed on a high-throughput mass spectrometry system (Agilent RapidFire system; Agilent Technologies, Inc., Santa Clara, CA), coupled with a mass spectrometer (MS) (API 4000 MS; AB SCIEX, Framingham, MA). Each compound of interest was tuned on the MS to provide the most sensitive detection. The percentage bound to melanin was calculated for each compound using the equation 100 × (1 − [F/T]), where F is the average signal from triplicate wells with melanin, representing free compound not bound to melanin, and T is the average signal from triplicate wells without melanin, representing total compound. 

Ocular Half-Life, Release Kinetics, and Pharmacokinetic Studies

Concentrations of GW771806 and pazopanib were determined in ocular homogenates following a single oral administration of 10, 30, or 100 mg/kg compound in 0.5% hydroxypropyl methylcellulose with 0.1% Tween 80 (pH nonadjusted) to LE or SD (albino) rats at 3 days postdosing. For study of release kinetics, LE pigmented rats were dosed with 100 mg/kg pazopanib or GW771806 and blood samples taken at 2 hours following administration. Ocular tissues were collected from six rats per time point at days 3, 7, 14, and 28. Ocular tissues from both eyes were collected and rinsed in ice-cold saline to remove any visible blood. Tissues from the right eye were weighed, homogenized, and extracted with 1:1 (v/v) methanol: 0.5 M HCl. The extraction volume was recorded and analyzed as described for plasma samples following protein precipitation and HPLC/MS/MS analysis. 

For pharmacokinetic studies, a single dose of pazopanib at 10 or 300 mg/kg was orally administered to mice (n = 15/group) as a suspension (2% hydroxypropylmethylcellulose, 0.1% Tween 80, pH 1.4). Blood samples were collected terminally from n = 3 mice/dose group at 2, 8, 24, 72, and 102 hours after administration. A further three mice served as a predose control. 

Plasma samples were analyzed for pazopanib and GW771806 using an analytical method based on protein precipitation, followed by HPLC/MS/MS analysis. The lower limit of quantification (LLQ) for pazopanib and GW771806 was 10.0 ng/mL with a higher limit of quantification (HLQ) of 10,000 ng/mL. Quality control samples, prepared at three different analyte concentrations and stored with study samples, were analyzed with each batch of samples against separately prepared calibration standards. 

For the ocular biodistribution study in the rat, neural retina and choroid/RPE samples were collected from animals that had been administered either topical pazopanib or GW771806 (30 μL in each eye for 7 days BID, doses 8 hours apart) 16 hours after the last dosing. Tissue samples were extracted in 50% methanol/0.5 M HCl and analyzed as described for plasma samples following protein precipitation and HPLC/MS/MS analysis. The LLQ for pazopanib and GW771806 was 2.00 ng/mL with an HLQ of 10,000 ng/mL. 

Tissue levels of pazopanib and GW771806 dosed by oral administration at 10 mg/kg were determined by quantitative whole body autoradiography using [¹⁴C]-labeled pazopanib (57.5 μCi/mg) and GW771806 (117 μCi/mg) following a single dose in pigmented female rats. 

Pharmacokinetic analysis was performed using mean tissue or plasma concentrations and a commercial analytical software program (WinNonLin, Enterprise v4.1 software; Pharsight Corp., Cary, NC). 

Experimental CNV

For the mouse, a slit-lamp–mounted diode laser system (wavelength 680 nm; Keeler, Windsor, UK) was used to injure Bruch's membrane in each eye (laser settings: 210 mW power, 100 ms duration, 100 μm spot diameter). CNV lesions were made in three areas approximately 3 to 4 disc diameters from the optic nerve head at the 10 o'clock, 2 o'clock, and 6 o'clock positions, avoiding any vessels. 

In the rat, experimental CNV was induced bilaterally using laser light photocoagulation (PC) with a diode-pumped argon laser (wavelength 647 nm; Coherent Novus OMNI PC Laser; Coherent, Inc., Santa Clara, CA) attached to a slit-lamp funduscope (laser settings: 150 mW, 0.2 second, 100 μm diameter). Eight to 10 laser lesions were radially distributed within a 1.5-mm radius of the optic disc, avoiding major retinal vessels (lesion interdistance > 200 μm). 

Only laser spots that resulted in the formation of a cavitation bubble, indicative of Bruch's membrane disruption, were included in the analysis. ^38,39 Eyes were also excluded if there was significant cataract or keratopathy formation. 

In Vivo Fluorescein Fluorescence Angiography

In mice, in vivo fluorescein fluorescence angiography (FFA) was performed at 1 week and 2 weeks after induction of laser CNV. Following IP fluorescein injection (2% fluorescein, 10 mL/kg body weight), images from the early (90 seconds postinjection) and late (7 minutes postinjection) phases were obtained using a small animal fundus camera (Kowa Genesis, Tokyo, Japan). The pixel area of CNV-associated hyperfluorescence was quantified for each lesion using image-analysis software (Image Pro Plus; Media Cybernetics, Silver Spring, MD). Where comparison of two eyes in the same animal was required, the sequence of imaging from right to left was alternated to avoid bias. 

For rats, a confocal scanning laser ophthalmoscope (cSLO, HRA2; Heidelberg Engineering, Heidelberg, Germany) was used to record laser-induced CNV. The focal plane was centered on the RPE, the field-of-view was 55°, and sequences of 100 consecutive frames were digitized as sequences of 8-bit, 1536 × 1536 pixel images, and averaged. FAs were made after IP injection of 0.5 mL fluorescein sodium (2%) at 4 ± 1 minutes (late-phase) in both eyes. Near-infrared images (70 mW at 790 nm) were recorded simultaneously to help identify lesions. CNV lesions were evaluated by quantitative assessment of late-phase FA. Prior to quantification, the gain and brightness of all images used in analysis were uniformly normalized and a coarse highpass filter was applied to remove illumination artifacts (MATLAB; The MathWorks, Inc., Natick, MA). The intensity and area of leakage in late-phase FA was derived semiautomatically by multiplying the area of CNV (μm²) by the mean pixel brightness value (0 to 1) in that area (NIH ImageJ software, developed by Wayne Rasband, National Institutes of Health, Bethesda, MD; available at http://rsbweb.nih.gov/ij/index.html). 

Statistical Analysis

CNV membranes were quantified using fluorescein angiographs, with each individual CNV lesion analyzed as individual data point (n). Data were analyzed using one-way ANOVA followed by Fisher's post hoc test or unpaired t-test to compare between different test groups and time points (SPSS v.17; SPSS, Inc., Chicago, IL). Values of P < 0.05 were considered statistically significant. Data are shown as means ± SEM unless otherwise noted. All data analysis was performed in a masked fashion. 

Results

Pazopanib and GW771806 Are Retained in the Uveal Tract

To examine ocular retention of pazopanib and GW771806 in pigmented tissues of the eye, levels of both compounds were measured in ocular homogenates 3 days after oral administration of each compound to pigmented rats and albino rats. The results confirmed the better ocular exposure for GW771806 compared with pazopanib (Fig. 1A). Significantly higher levels of both compounds were retained in the eyes of pigmented rats compared with albino rats 3 days postdosing (Fig. 1A), consistent with the high degree of direct melanin binding observed for both compounds (see Supplementary Material and Supplementary Table S1). Ocular clearance or release kinetics for both compounds confirmed the better ocular exposure for GW771806. The ocular half-lives for pazopanib and GW771806 were approximately 439 and 442 hours, respectively (Fig. 1B). Because mass spectrometry measures intact compounds, these data confirm that uveal-bound compounds are released in vivo. 

Figure 1

 

Ocular retention and release kinetics of pazopanib and GW771806 in pigmented and albino rats. (A) Concentrations of pazopanib and GW771806 were determined in ocular homogenates 3 days following a single oral dose of 10, 30, or 100 mg/kg compounds administered to pigmented LE rats, and 100 mg/kg only to albino SD rats. Better ocular exposure for GW771806 was confirmed, and both compounds were either nondetectable or near the lower limit of quantification in the ocular homogenate from albino rats, consistent with melanin binding–mediated retention of both compounds in the pigmented eyes (*P < 0.01; **P < 0.001; ***P < 0.0001, t-test; n = 3 rats per group; error bars: SD). (B) Ocular release kinetics of pazopanib and GW771806 following a single oral gavage dosing in LE rats. The ocular half-lives for pazopanib and GW771806, measured between 3 and 28 days postdosing of 100 mg/kg compound, were determined to be approximately 439 and 442 hours, respectively (n = 6 rats per group; error bars: SD).

Whole-body autoradiography of a rat following a single oral dose (10 mg/kg) of pazopanib and GW771806 was performed to determine compound distribution in ocular tissues. Following single oral doses of [¹⁴C]-labeled pazopanib, drug-related material was widely distributed throughout the body, with the exception of brain, and was not quantifiable in most tissues 3 days postdose (data not shown). However, pazopanib radioactivity was observed in melanin-containing tissues including pigmented skin, meninges, and the uveal tract (iris, ciliary body, and choroid) up to 35 days after a single oral delivery (Fig. 2A). Similar ocular distribution was observed for [¹⁴C]-labeled GW771806 (Fig. 2B). Next, plasma drug concentrations following a single oral administration of pazopanib (10 or 300 mg/kg) to pigmented mice were determined. Plasma levels of drug in the 300 mg/kg dose group approached limits of detection at 102 hours, and were undetectable in the 10 mg/kg dose group at 72 hours postdelivery (Fig. 2C). The systemic half-life (T _1/2) for pazopanib following a single oral administration was determined to be 4.6 to 8.6 hours, depending on the dosage (see Supplementary Material and Supplementary Table S2). 

Figure 2

 

Pazopanib and GW771806 were retained in melanin-containing tissues in the eye for an extended period of time after a single oral dose. (A, B) Whole body radiography of a rat at various time points up to 35 days following a single dose (10 mg/kg) of [¹⁴C]-labeled pazopanib or [¹⁴C]-labeled GW771806 administered by oral gavage. Pazopanib and GW771806 radioactivity was observed in melanin-containing tissues, including pigmented skin, meninges, and the uveal tract for up to 35 days after delivery, suggesting a selective association of pazopanib and GW771806 with melanin-containing tissues. (C) Plasma concentrations following a single oral administration of pazopanib (10 or 300 mg/kg) to 10-week-old mice (n = 15/dose group). Levels of drug in the 300 mg dose group approached the lower limits of detection at 102 hours and were undetectable in the 10 mg dose group >72 hours postdelivery. Error bars: SD.

Pazopanib Retained in the Uvea Suppresses CNV Development

Given the biodistribution data, we hypothesized that the uveal-bound material could act as a slow-release depot for pazopanib. To test this novel concept, we investigated if a single oral loading dose was sufficient to inhibit mouse CNV. Previously published pharmacokinetic data indicated that pazopanib has a predicted T _1/2 in mouse plasma of <4 hours. ³³ Based on this T _1/2, it was calculated that only 0.0001% of the original oral dose would remain after 3 days (18 half-lives). In our study, the T _1/2 for pazopanib was between 4 (10 mg/kg) and 8.6 (300 mg/kg) hours. Assuming a conservative T _1/2 of 9 hours, there would still be <0.5% of the original oral dose remaining after 3 days. Indeed, the plasma concentration of pazopanib in the 300 mg/kg group 3 days postdosing was 54 ng/mL (0.12 μM), which is nearly 600-fold lower than the C _max achieved by the 10 mg/kg dose (Fig. 2C and data not shown). We therefore used a single oral loading dose 3 days prior to the induction of CNV by diode laser. The results from this dose group were compared with a standard CNV dosage regimen, whereby animals were dosed twice daily with vehicle or an oral dose of 10 mg/kg pazopanib starting 5 days before laser and continued for the entirety of the CNV experiment. 

Results from fluorescein angiography indicated that pazopanib delivered at a single dose of at least 10 mg/kg 3 days prior to induction of CNV was capable of significantly (P < 0.05) reducing CNV lesion size 2 weeks postinduction of CNV, with the exception of the group receiving a dose of 30 mg/kg, in which a reduction was observed that did not achieve significance (P = 0.09) (Fig. 3C). The results from the 10 to 300 mg/kg dosing groups were similar to those from the animals receiving the continuous dosing regimen from day −5 to day 14, with an approximate 50% to 60% decrease observed in both (Figs. 3A, 3C). These results indicate that a therapeutic depot level of pazopanib in the uvea can be achieved with a single oral dose of approximately 10 mg/kg in the mouse CNV model. Interestingly, there was an apparent lack of dose–response beyond 10 mg/kg for pazopanib (single oral dosing), likely because maximum ocular exposure and therefore “loading” have been reached at approximately 10 mg/kg (Fig. 2C). This is consistent with the lack of dose–response for ocular pazopanib retention (Fig. 1A). 

Figure 3

 

Pazopanib uveal depot established 3 days prior to induction of CNV was pharmacologically active in suppressing CNV. Various doses of pazopanib or vehicle were delivered by oral gavage either continuously from day −5 to day 14 or a single dose 3 days prior to induction of CNV by diode laser. (A) Continuous oral dosing group, mean lesion size for each experimental group at 1 week and 2 weeks postinduction of CNV. Significant reductions (t-test; **P < 0.01) in mean lesion size of 54% at 1 week postinduction of CNV and 60% at 2 weeks postinduction of CNV compared with vehicle only (n = 69–76 lesions per group; error bars: SEM). (B) Timeline for the single oral dosing experiment. (C) Mean lesion size in each group, as determined by early phase fundus fluorescein angiography at 1 week and 2 weeks. Results indicate that pazopanib delivered at a single oral dose of 10 mg/kg and above, 3 days prior to induction of CNV, is capable of significantly reducing CNV lesion size 2 weeks postinduction of CNV (*P < 0.05, t-test; n = 32–36 lesions per group; error bars: SEM).

The durability of the depot effect of pazopanib was further tested by administering a single dose of 30 mg/kg pazopanib, 3 mg/kg pazopanib, or vehicle by oral gavage 5 days prior to induction of CNV by diode laser. A single 30 mg/kg oral dose of pazopanib suppressed CNV induced 5 days later by 33% at the 1 week postlaser time point, and 14% at the 2 week postlaser time point, although these differences did not reach statistical significance (Fig. 4) (t-test vehicle versus pazopanib-dosed groups; P = 0.05 and P = 0.07 for group receiving 30 mg/kg pazopanib at 1 week and 2 weeks, respectively). No effect was observed at the 3 mg/kg pazopanib dose, suggesting that a minimum peak oral dosage is critical for achieving an optimum depot effect. Indeed, mice dosed with very low oral pazopanib doses (4, 1.5, 0.5 mg/kg) twice daily starting 5 days before CNV induction and continued for 2 weeks postlasering failed to significantly suppress CNV development (data not shown). Together, these results show that pazopanib retained as a “depot” in the uvea after a single minimum oral dose (10 mg/kg) is available and pharmacologically active to suppress the development of angiogenesis in the laser-induced CNV model. 

Figure 4

 

Pazopanib delivered 5 days prior to CNV induction was pharmacologically active in suppressing CNV. (A) Timeline for experiment. Young adult wild-type mice were administered a single oral dose of 30 mg/kg pazopanib, 3 mg/kg pazopanib, or vehicle 5 days prior to induction of CNV by diode laser. (B) 30 mg/kg pazopanib 5 days before CNV induction resulted in detectable reduction of lesion size by a third at 1 week postinduction of CNV and possibly beyond, although results did not achieve significance (#P = 0.05 at 1 week with P = 0.08 at 2 weeks; t-test). (C) Representative images taken from vehicle-administered and 30 mg/kg pazopanib-dosed animals at 1 week postinduction of CNV. n = 75 to 83 lesions per group; error bars: SEM.

Biodistribution of Pazopanib and GW771806 after Topical Administration

Since pazopanib was retained in the uveal tract after systemic dosing and this depot was effective in reducing CNV, we determined if we could achieve therapeutically effective levels in the chorioretinal space using an eye drop formulation. We also compared topical pazopanib with the closely related but more soluble RTKI GW771806. First, neural retina and RPE/choroid tissue samples collected from animals treated with topical pazopanib or GW771806 at different concentrations were analyzed to determine the biodistribution of the drugs. Pharmacokinetic analysis of tissue samples showed 118-fold (pazopanib, 5 mg/mL eye drops) and 263-fold (GW771806, 5 mg/mL eye drops) higher presence of both RTKIs in the RPE/choroid complex compared with the neural retina in treated animals (Fig. 5A). Noticeably higher (2.7-fold) levels of GW771806 (5 mg/mL eye drops) were found in the RPE/choroid tissue samples compared with pazopanib at the highest concentration (Fig. 5B), indicating enhanced ability of GW771806 to reach or be retained in the target tissue, perhaps due to its slightly higher aqueous solubility (see Supplementary Material and Supplementary Table S1). 

Figure 5

 

Pazopanib and GW771806 are elevated in the RPE/choroid after topical administration. Pazopanib and GW771806 formulated for topical use were administered to SD rats at a dosage of 30 μL in each eye for 7 days BID, 8 hours apart, and tissues were collected 16 hours after the last eye drop dosing. (A) Comparison of drug levels in the retina and RPE/choroid following topical application at a concentration of 5 mg/mL (n = 6 eyes/group). (B) Comparison of drug levels in the RPE/choroid following topical application at various concentrations (n = 6 eyes/group, ***P < 0.001; one-way ANOVA). Error bars: SEM.

Topical GW771806 Suppresses CNV Development

We next examined the ability of pazopanib and GW771806 to regress existing CNV lesions. This is a more rigorous assessment of the topical RTKIs and more akin to the actual setting in which these would be used in patients. Lesions were induced on day 0 and the baseline CNV areas measured by FFA on day 7. Topical administration of each drug was started on day 10 and continued for 8 days. The final CNV areas were recorded by FFA on day 18, approximately 18 hours after the last topical dose (Fig. 6A). The baseline and endpoint FFA measurements allowed the determination of CNV growth or regression between days 7 and 18 after laser-CNV induction. Comparison of FFA showed a significant increase in CNV areas 18 days after lasering in the vehicle-treated eyes, from 44.33 ± 1.694 to 49.41 ± 1.442 units (mean ± SEM, P < 0.05; Figs. 6B, 6C, 7; see Supplementary Material and Supplementary Figs. S1A, S1B). 

Figure 6

 

Topical GW771806 suppressed CNV development, whereas topical pazopanib did not. Various dosages of topical RTKIs were administered alone or in combination with a single intravitreal injection of anti-VEGF dAb, starting at 10 days postinduction of CNV. Vehicle alone and a single intravitreal injection of anti-VEGF dAb were used as controls. (A) Timeline for experiment shown. (B) Topical administration of pazopanib did not reduce size of CNV lesions, whereas intravitreal anti-VEGF dAb significantly reduced CNV. (C) Topical GW771806 significantly attenuated lesion size. (D) Topical GW771806 was significantly better than topical pazopanib at reducing CNV lesion size (n = 96 lesions/group, *P < 0.05, **P < 0.01, ***P < 0.0001; one-way ANOVA). Error bars: SEM.

Figure 7

 

Representative late-phase FFA images for eyes treated with topical pazopanib or GW771806. Representative late-phase fluorescein angiograms showing the same lesions at days 7 and 18 postinduction of CNV by diode laser. Eyes were treated with (A) vehicle, (B, C) topical pazopanib, (D) intravitreal anti-VEGF dAb, or (E, F) topical GW771806. In eyes treated with GW771806 or anti-VEGF dAb, CNV lesions decreased in size, whereas the lesions increased in size in the vehicle and topical pazopanib-treated eyes.

Compared with vehicle-treated negative control, and in contrast to the neutralizing anti-VEGF domain antibody (dAb) positive control, topical administration of pazopanib did not result in any significant change to the CNV lesion area at the concentrations tested (Fig. 6B). However, topical GW771806 at 2 and 5 mg/mL significantly suppressed CNV development between days 7 and 18 in a dose-dependent manner (Fig. 6C). In fact, GW771806 at 5 mg/mL caused a significant regression of CNV between day 7 and day 18 (P < 0.05) and also a significant reduction of CNV compared with the vehicle control on day 18 (P < 0.0001) (Figs. 6C, 6D). The enhanced anti-CNV activity of topical GW771806 is consistent with the better RPE/choroid availability (Fig. 5B) and in vivo antiangiogenic effects in a matrigel plug assay (data not shown), compared with pazopanib. 

Discussion

Although anti-VEGF therapies delivered by intravitreal injection can improve outcomes in neovascular AMD, there is still room for improvement, particularly in the area of drug delivery. Improved delivery could involve less frequent intravitreal injection or, better still, noninvasive administration of anti-VEGF therapy to the posterior segment of the eye. To this end, we explored the hypothesis that melanin-binding drugs that target the VEGF-A pathway, such as pazopanib and GW771806, are retained in the pigmented uvea and function as a sustained-release drug depot at the back of the eye in pigmented animals. Indeed, administering mice a single oral “loading” dose of pazopanib at 3 days before the induction of CNV suppressed CNV development 7 and 14 days after CNV induction. Bioanalytical analysis confirmed that the plasma levels of pazopanib were below the limits of detection prior to initiation of CNV. The observed depot effect compared favorable with a standard twice-daily dosing regimen (10 mg/kg twice daily by oral gavage). 

Our results strongly suggest that the uvea-retained pazopanib depot was active upon release and suppressed CNV development locally. This novel tissue-targeted drug delivery strategy might allow future treatment of neovascular AMD via a single, noninvasive oral administration of relatively small amounts of drug. Obviously this depot delivery strategy relies on the specific melanin-binding properties of a compound, which must favor release and, importantly, maintenance of bioactivity upon release from cells at the back of the eye. ^40–43 These properties might not be intrinsic to a particular class of compound, but could potentially be engineered into a molecule to enable depot delivery. 

Oral dosing of multitargeted RTKIs, even if limited in frequency, poses increased systemic risks compared with local ocular administration. Topical delivery is another promising strategy for administration of this class of therapeutics, given the known transcleral route of distribution of most small molecules ^44,45 and the potential for uveal binding. Indeed, topical administration of pazopanib and GW771806 resulted in significant accumulation of the drugs in the RPE/choroid complex. Topical pazopanib did not suppress CNV at any of the concentrations tested, but topical GW771806 at 5 mg/mL significantly suppressed and, in fact, regressed CNV. The effect of 5 mg/mL topical pazopanib is in contrast to previous studies showing an effect in limiting lesion progression, ³¹ but in those studies topical pazopanib was dosed immediately after FA analysis (day 7), whereas in our studies the drug was not dosed until day 10. The efficacy differences between pazopanib and GW771806 cannot entirely be explained by the higher depot levels, since GW771806 and pazopanib topically administered at 2 mg/mL were found at similar levels in the RPE/choroid, yet only GW771806 significantly suppressed CNV. 

Our GW771806 data suggest that even transient local ocular exposure combined with better tissue penetration and retention in pigmented tissues could greatly improve the local delivery and efficacy of a drug for treating CNV. In fact, our topical GW771806 treatment was more effective than a single high-dose intravitreous injection of an anti-VEGF dAb in regressing CNV. This suggests that at least in this rat model of laser-induced CNV, continuous local topical administration of RTKIs with desirable tissue penetration and retention properties and a broad spectrum action on cellular receptors critical for angiogenesis, leads to a better therapeutic response. 

Together, our data highlight the need for further exploration of the utility of the melanin or uveal depot concept, especially when combined with topical delivery. This strategy is currently being explored in the clinical setting with RTKIs for treatment of patients with ocular neovascular conditions. 

Supplementary Materials

pdf S1 

pdf S2 

Acknowledgments

The authors thank Anne Goodwin for critical reading of the manuscript. 

Footnotes

 Supported by grants from the National Institute for Health Research Biomedical Research Centre for Ophthalmology (SJR, JWB, DTS). PA, AGK, DL, and CS are full-time employees of GlaxoSmithKline.

Footnotes

⁶  These authors contributed equally to the work presented here and should therefore be regarded as equivalent authors.

Footnotes

 Disclosure: S.J. Robbie, GlaxoSmithKline (F); P. Lundh von Leithner, None; M. Ju, None; C.A. Lange, GlaxoSmithKline (F); A.G. King, GlaxoSmithKline (E); P. Adamson, GlaxoSmithKline (E); D. Lee, GlaxoSmithKline (E); C. Sychterz, GlaxoSmithKline (E); P. Coffey, None; Y.-S. Ng, GlaxoSmithKline (F); J.W. Bainbridge, GlaxoSmithKline (F); D.T. Shima, GlaxoSmithKline (F)