AMG 386, a Selective Angiopoietin 1/2-Neutralizing Peptibody, Inhibits Angiogenesis in Models of Ocular Neovascular Diseases

Abstract

Purpose: To determine whether systemic treatment with AMG 386, a selective angiopoietin 1/2-neutralizing peptibody, inhibits neovascular processes in animal models of ocular disease.

Methods: AMG 386 was tested in a laser-induced choroidal neovascularization (CNV) model in monkeys using fluorescein angiography. The biodistribution of ¹²⁵I-AMG 386 was determined in cynomolgus monkeys by whole-body autoradiography and radioanalysis of ocular tissues. A murine retinopathy of prematurity (ROP) model was used to examine the effect of AMG 386 on established and newly formed retinal vessels, either as a single agent or when combined with VEGF inhibition.AMG 386 pharmacokinetics were evaluated in each model.

Results: In the CNV model, AMG 386 significantly decreased fluorescent angiographic leakage and reduced fibroplasia, indicating an impaired healing response consistent with angiogenesis blockade. Radiolabeled AMG 386 was widely distributed across ocular tissues, with highest concentrations in the choroid, cornea, retinal pigmented epithelium, iris/ciliary body, and sclera. In the ROP model, AMG 386 prevented pathologic retinal angiogenesis when administered from P8 to P16 but transiently impeded regression of these abnormal vessels when administered from P17 to P23. Combining AMG 386 with VEGF inhibition led to cooperative prevention of retinal angiogenesis in this model. No AMG 386–related ocular toxicities occurred, and no treatment-related clinical observations were made in any of the studies.

Conclusions: In this study, AMG 386 inhibited angiogenesis in animal models of CNV and ROP, supporting investigation of AMG 386 for the treatment of ocular neovascular diseases in the clinical setting.

Introduction

Although angiogenesis is essential to certain normal physiologic processes, such as ovarian follicular maturation and endochondral bone growth, it is also a key feature of neoplastic and numerous nonneoplastic disorders, including exudative or “wet” age-related macular degeneration (AMD), proliferative diabetic retinopathy (PDR), and retinopathy of prematurity (ROP). ¹ Exudative AMD is characterized by choroidal neovascularization (CNV), which results in retinal edema, fibrous scarring, and significant permanent central vision loss. ² AMD is the leading cause of blindness among the elderly in developed countries. ³ Both PDR and ROP are described by hypoxia-induced pathological vessel growth ⁴ and are characterized by fibrous scarring that may ultimately lead to retinal detachment and visual impairment or even blindness. ^5,6  

Conventional treatment options, including photocoagulation, cryosurgery, photodynamic therapy, and vitrectomy, preserve vision only in a subset of patients, and even fewer achieve visual improvement. ^7–11 VEGF, through activation of its cognate receptor VEGFR2 (Flk-1), is a key regulator of angiogenesis ¹² and has been demonstrated to stimulate neovascularization in AMD, PDR, and ROP. ¹³ Pegaptanib and ranibizumab are two approved therapies for these diseases targeting the VEGF/VEGFR pathway. ^14–17 Recently, the anti-VEGF antibody bevacizumab has also proven effective in the treatment of ROP ¹⁸ and AMD. ¹⁹ However, these agents are administered via intraocular injection, which not only can be painful but also has inherent risks. ^20–22  

Angiopoietin 1 (Ang1) and 2 (Ang2) and their high-affinity receptor Tie2^23,24 are another important signaling pathway that regulates the later stages of angiogenesis. ^1,25 Ang1 stimulates Tie2 phosphorylation, ²³ which mediates vessel stabilization. ²⁶ Ang2 expression appears to disrupt the Ang1/Tie2 interaction, effectively destabilizing vessels prior to angiogenesis. ²⁴ In the presence of VEGF, this destabilization can lead to neoangiogenesis, ²⁷ as suggested by Asahara and colleagues using a murine model of VEGF-induced corneal neovascularization. ²⁸  

Evidence suggests that the angiopoietin/Tie2 axis plays an important role in retinal neovascularization and CNV, thus representing a potential target for anti-angiogenic therapy. Upregulation of Ang2 occurs in regions of neovascularization during ischemic retinopathy, ^29,30 which is supported by a recent gene expression-profiling study in a murine model demonstrating that Ang2 expression is most upregulated during the time period of noticeable extraretinal neovascularization. ³¹ Conversely, suppression of the Tie2 pathway via injection of soluble Tie2 inhibits both retinal ^30,32 and choroidal angiogenesis. ³² Inhibition of Ang2 with neutralizing antibodies and peptibodies prevents VEGF-stimulated corneal angiogenesis, ³³ while Ang2-deficient mice studied in an ROP model demonstrated a lack of ischemia-induced retinal neovascularization. ³⁴  

AMG 386 is an investigational, recombinant Fc-peptide fusion protein (peptibody) that potently neutralizes the interaction between Tie2 and its ligands Ang1 and Ang2. In tumor xenograft models, systemic treatment with AMG 386 results in inhibition of tumor growth with subsequent disappearance of all measurable tumor in a subset of animals and concomitant reduction of tumor endothelial cell proliferation. ³³ In patients with advanced solid tumors, AMG 386 shows promising antitumor activity. Specifically, in a phase-2 study of patients with advanced ovarian carcinoma, treatment with AMG 386 plus paclitaxel prolongs progression-free survival over placebo, with a well-tolerated toxicity profile distinct from that typically observed with VEGF(R) inhibitors. ³⁵ The series of studies described herein was designed to determine whether AMG 386 reduces neovascular processes in animal models of ocular diseases. In the current study, we showed that AMG 386 inhibited ocular neovascularization in the ROP and CNV models, supporting the clinical investigation of this agent in ocular neovascular diseases. 

Materials and Methods

Cynomolgus Monkey Studies

Laser-induced CNV and whole-body autoradiography studies using cynomolgus monkeys ( Macaca fascicularis ) of Vietnamese origin were conducted by Covance Laboratories Inc. (Madison, WI), the latter in accordance with Wisconsin Department of Health and Family Services, Radiation Protection Section. All procedures were in compliance with Animal Welfare Act Regulations (9 CFR 3). Animal treatment conformed to the guidelines outlined in the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research, ³⁶ and all procedures were approved by the Amgen Animal Care and Use Committee. The facilities where experiments involving animals were conducted were approved by the Association for Assessment and Accreditation of Laboratory Animal Care. 

Laser-Induced CNV

Twenty-four female cynomolgus monkeys approximately 2 to 5.5 years old and weighing between 2.0 and 3.4 kg at initiation of treatment were randomly assigned to four treatment groups (n = 6 per group). Beginning on study day 1, animals received control (saline) or 1.1, 5.7, or 28.6 mg/kg/dose of AMG 386 in Dulbecco's phosphate-buffered saline (D-PBS; Gibco/Life Technologies, Rockville, MD) every other day via bolus intravenous (IV) administration in a saphenous vein. Previous studies have shown that 3 μg/mL of AMG 386 is the minimum serum trough concentration necessary to achieve optimal antitumor efficacy in a mouse xenograft model. ³³ The doses used in the present study were selected to bracket this exposure level. On study day 2, nine lesions were symmetrically placed on the macula of each eye. The incidence of clinically relevant CNV lesions was determined via fluorescein angiography performed once before initiation of treatment and on study days 16 and 23 (day 14 and 21 post laser, respectively). Analysis was performed by an investigator masked to treatment assignments. A detailed description of the procedures is included in the Supplementary Material. On study day 24, animals were euthanized with sodium pentobarbital. At necropsy, eyes were enucleated and fixed, first by intraocular injection of refrigerated phosphate-buffered 6% glutaraldehyde until turgid, followed by immersion in the same fixative for 48 to 72 hours (at 2°C–8°C), and then held in 10% neutral-buffered formalin solution (at 2°C–8°C) until processed. Tissue containing laser burn sites was embedded in glycol methacrylate, sectioned at 2 μm, and stained with hematoxylin and eosin (H&E) for microscopic evaluation. A total of 108 laser sites/group was evaluated (6 animals/group, 2 eyes/animal, 9 sites/eye). Details are described in the Supplementary Material. 

Prior to dosing on days 7, 13, and 19, and on day 24, blood samples were collected from all animals for measurement of AMG 386 concentrations in the serum using the ELISA described in the Supplementary Material; lower limit of quantification [LLOQ], 102 ng/mL). 

Radioanalysis and Whole-Body Autoradiography

Eight cynomolgus monkeys, four male and four female, approximately 2 to 5 years old and weighing between 2.2 and 3.5 kg each, received a single bolus IV injection of ¹²⁵I-AMG 386 (2.77 μCi/mg; Amgen Inc., Thousand Oaks, CA) via a saphenous vein. The dose was calculated on the basis of body weight on the day of dosing, with a target dose level of 25 mg/kg of AMG 386 (70 μCi/kg). Blood, eye tissues, urine, and feces were collected, and the radioactivity determined using a Packard COBRA II 5003 solid scintillation counter (Waltham, MA). Details of the sample collection schedule and sample preparation are outlined in the Supplementary Material. Quantitative whole-body autoradiography of the remaining carcass to examine tissue distribution of radioactivity was performed as described in the Supplementary Material. Autoradiographic standard image data were sampled using Analytical Imaging Station (AIS) software (Imaging Research, Inc., St. Catharines, ON, Canada) to create a calibrated standard curve. Tissue concentrations were interpolated from each standard curve as nCi/g and then converted to μg equivalents/g on the basis of the test article specific activity. 

Maximum concentration (C_max) in serum and the time to reach maximum concentration (t_max) were obtained by visual inspection of raw data. Half-life (t_1/2) and area under concentration versus time curve from time 0 to infinity (AUC_0-inf) were calculated with WinNonlin, Version 4.1 (Pharsight Corporation, Mountain View, CA). 

Murine ROP Study

Forty-nine newborn C57BL/6 mice (body weight, 4.5–5.5 g) and their mothers (Charles River Laboratories, Raleigh, NC) were placed in an oxygen chamber (A-Chamber; BioSpherix Ltd., Redfield, NY) on postnatal day 7 (P7) and were exposed to an atmosphere of 75% ± 0.5% oxygen for 5 days. Animals were returned to room air on P12. At various postnatal dosing intervals, pups received once-daily subcutaneous doses of one of the following treatments (n = 7 per group): AMG 386 (0.1, 1, 2, 10, or 100 mg/kg/dose) (Amgen Inc.) in D-PBS; 10 mg/kg normal goat IgG (R&D Systems, Inc., Minneapolis, MN) in PBS; 10 mg/kg anti-mouse VEGF antibody (R&D Systems) in PBS; or 100 mg/kg Fc control (generated at Amgen Inc.; data on file). Seven additional mice (normoxic control group) and their mother remained without treatment at room air throughout the study. Blood samples for AMG 386 concentration analysis were drawn, and both eyes were removed for sectioning. AMG 273 (15 mg/kg; Amgen Inc.) was administered daily by oral gavage from P12 to P15. Animal treatment conformed to the guidelines outlined in the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. ³⁶  

Measurement of Retinal Neovascularization

Eyes from five of the seven animals (the first five in the animal-number series) in each treatment group were analyzed for retinal neovascularization. Both eyes from each animal were immersion-fixed in zinc formalin (or, in later studies, in Davidson's fixative) for 48 hours and then processed into paraffin blocks. Step sections were cut parallel to the optical axis at 100-μm intervals. Blocks were completely through-sectioned, resulting in 15 or 16 sections per eye. All sections were stained with H&E. The middle 10 slides from the step-section series were used in the analysis, bracketing the optical axis. For each section, the number of vascular nuclei (endothelial and pericyte nuclei) on the vitreous side of the inner limiting membrane were counted in a masked fashion, without knowledge of treatment. Counts from all 10 slides were summed for each animal. 

Measurement of AMG 386 Concentrations

Blood was collected by cardiac puncture immediately after animals were euthanized on P17. Serum was prepared from each sample, and concentrations of AMG 386 were measured using an ELISA (Amgen Inc.) as described in the Supplementary Material. The LLOQ of this assay was 39.1 ng/mL. 

In Situ Hybridization

Isotopic in situ hybridization (ISH) was performed using ³³P-labeled antisense and sense riboprobes transcribed from cDNA templates for mouse Ang1, Ang2, VEGFR2, and VEGF. Briefly, 5-μm sections of Davidson's fixed and paraffin-embedded retinas from paired normoxic and hyperoxic animals from P8, 12, 15, 17, and 24 were deparaffinized, deproteinated, treated with Proteinase K, acetylated, and pre-incubated in hybridization buffer. Tissue sections were probed with 1 × 10⁶ cpm of ³³P-labeled probe and incubated in a humidified chamber at 56°C for 16 hours. The sections were then treated with RNase A, washed, air dried, coated with Kodak NTB emulsion (Eastman Kodak, Rochester, NY), and stored in the dark at 4°C with desiccant for 3 weeks. Slides were developed in Kodak D19 developer, fixed, and counterstained with H&E. A detailed description of the methods is provided in the Supplementary Material. 

Statistics

Mean and SD were calculated. In the murine ROP study, a one-way ANOVA followed by Fisher's post hoc test for comparison between groups (StatView software 5.0.1; SAS Institute, Cary, NC) was performed, and results were expressed as the mean ± SEM. A P value < 0.05 was considered statistically significant. 

Results

AMG 386 Treatment in a Laser-Induced CNV Model

Figure 1A shows representative angiograms of laser-induced lesions after saline and high-dose AMG 386 treatment. In the vehicle-treated eye, five of the nine lesions were hyperfluorescent and showed late leakage of contrast agent beyond the borders of the treated area, classifying them as grade-4 lesions. Grade-4 lesions are considered to most closely resemble the active forms of classic CNV seen in various human retinal disorders, including AMD. In comparison, only one of the nine lesions in the eye treated with high-dose AMG 386 was a grade-4 lesion. Overall, treatment with medium or high doses of AMG 386 resulted in a significantly lower median number of grade-4 lesions than treatment with control on study days 16 and 23 (days 14 and 21 post laser, respectively; P ≤ 0.04; Fig. 1B) in an analysis performed by a masked investigator. 

Figure 1.

 

Effect of AMG 386 treatment in a cynomolgus monkey model of choroidal neovascularization. (A) Representative angiograms of eyes from cynomolgus monkeys performed 5 minutes after IV fluorescein injection on study day 23 (day 21 post laser treatment). Grade-4 lesions are identified by bright hyperfluorescence and contrast agent leaking beyond the borders of the treated area. Left: angiogram of a control animal (vehicle treatment). Right: angiogram of an animal treated with a high dose of AMG 386 (28.6 mg/kg). (B) The median number of grade-4 lesions observed after treatment with low (1.1 mg/kg), medium (5.7 mg/kg), and high (28.6 mg/kg) doses of AMG 386 on study days 16 and 23 (day 14 and 21 post laser, respectively) as assessed by an investigator masked to the treatment assignments. Median rather than mean number of lesions is reported because of a single outlying data point. *P = 0.02, versus vehicle; **P = 0.04, versus vehicle. (C) Mean (±SD) AMG 386 concentrations (μg/mL) over time in the serum of cynomolgus monkeys after administration of various doses of AMG 386 every other day for 3 weeks.

Laser sites were microscopically evaluated for total lesion severity and healing response, which is dependent on blood vessel growth and thus, an inverse indicator of AMG 386 efficacy. Initial laser injury was remarkably similar across animals and sites (see Supplementary Material ). On average, treatment with high-dose AMG 386 (28.6 mg/kg) resulted in less retinal elevation compared with control animals. Retinal elevation, defined as a separation of the neuroretina from the retinal pigmented epithelium (RPE) and creation of a fluid-filled space, occurred peripheral to and often bridged between the laser sites. Many of the fluid-filled spaces contained red blood cells and were often sites with extensive tissue proliferative reactions. Retinal elevation was quite common in some animals and not present in others, regardless of treatment group, and it usually correlated with vascular leakage at the time of necropsy. Further, it was considered a requisite change for, and was generally proportional to, the healing response in individual animals. High-dose AMG 386 treatment was also associated with reduced fibroplasia, compared with controls. Representative images of the laser injury sites are shown in Figure 2. Where present, the proliferating tissue extended from the choroid into the globe, forming a variably sized mushroom-shaped mass of fibrovascular tissue between the RPE and the neuroretina. Retinal herniation (defined as the presence of neuroretinal tissue within the choroid) was not visible in the control group but occurred in AMG 386–treated animals, appearing microscopically as though the neuroretina had collapsed into the space created by the laser burn. Proliferation of RPE, another healing response, was common. RPE cells proliferated into subretinal spaces, forming clumps at the margins of the tissue response. Although it correlated with the severity of retinal elevation, the difference between controls and AMG 386–treated animals was not as visible as that seen for fibroplasia and herniation. Treatment with low-dose (1.1 mg/kg) or medium-dose (5.7 mg/kg) AMG 386 had little or no effect on the tissue around the laser sites. 

Figure 2.

 

Paired photomicrographs of laser injury sites illustrating the healing response in the absence and presence of AMG 386. (A) Untreated control. Laser injury sites have near-complete loss of neuroretinal elements except for the ganglion cell layer, and the morphology of the underlying choroid is also altered. Melanin is clumped and present in histiocyte-like cells at the center of the laser sites. The RPE and Bruch's membrane are absent centrally. A layer of new connective tissue (fibroplasia) is present between the remnant of the neuroretina and the choroid. The fibroplasia is thickest centrally and is partly covered by the RPE peripherally. (B) High-dose AMG 386. The laser-induced histopathology is similar to that depicted in the untreated control; however, the thickness of the new fibrous connective tissue layer is reduced. GCL, ganglion cell layer; INL, inner nuclear layer; ONL, outer nuclear layer; PRL, photo receptor layer.

Serum concentration analysis showed that after IV administration of AMG 386 to cynomolgus monkeys every other day for 3 weeks (day 1 through day 23), the mean trough levels reached the steady state by day 7 and appeared to increase approximately proportionally with dose. The mean trough concentrations increased 28- to 34-fold as the doses increased by 26-fold. Mean serum AMG 386 concentrations in the serum are shown in Figure 1C. 

The IV administration of AMG 386 was not associated with any discernible ocular toxicity at day 21, nor were any clinical observations noted that were considered to be the result of treatment with AMG 386. All animals survived to the scheduled sacrifice on day 24. 

Biodistribution of AMG 386

Concentrations of ¹²⁵I-AMG 386–derived radioactivity and trichloroacetic acid (TCA)-precipitable radioactivity in the serum of cynomolgus monkeys declined bi-exponentially, and C_max was observed at 1 hour post dose, the first collection time point (Table 1, Fig. 3). Most of the radioactivity associated with serum was TCA-precipitable at all time points. Serum C_max based on TCA-precipitable radioactivity was similar for male and female cynomolgus monkeys (154 and 148 μg equivalents ¹²⁵I-AMG 386/g, respectively); t_1/2 was 57.8 and 65.6 hours, respectively. The pharmacokinetic profile of AMG 386 as TCA-precipitable radioactivity was comparable to that measured by the ELISA method (data not shown). 

Table 1.

 

Pharmacokinetic Parameters for Radioactivity in Serum after Administration of a Single Intravenous Dose of ¹²⁵I-AMG 386 (25 mg/kg) to Male and Female Cynomolgus Monkeys

	Sacrifice Time (h)	Radioactivity in Serum			TCA-Precipitable Radioactivity in Serum
		Cmax (μg equiv 125I-AMG 386/g)	AUC0-inf (μg equiv 125I-AMG 386·h/g)	t1/2 (h)	Cmax (μg equiv 125I-AMG 386/g)	AUC0-inf (μg equiv 125I-AMG 386·h/g)	t1/2 (h)
Male (n = 1)*	168	218	2500	57.0	154	1750	57.8
Female (n = 1)*	168	212	2070	60.5	148	1460	65.6

 C_max, maximum observed concentration (concentration at 1 hour).

*  Pharmacokinetic data through 168 hours obtained from one animal/sex as two animals/time point (n = 1/sex) were sacrificed at 1, 24, and 72 hours.

Figure 3.

 

Mean (+SD) concentration-time profiles of ¹²⁵I-AMG 386 in serum from cynomolgus monkeys after administration of AMG 386. Radioactivity in serum (as total and TCA-precipitable μg equivalents of ¹²⁵I-AMG 386/g tissue) from cynomolgus monkeys at specified time points after a single bolus IV administration of 25 mg/kg of AMG 386. Serum, n = 2 to 8 per time point. Error bars indicate SD for n ≥ 3.

¹²⁵I-AMG 386–derived TCA-precipitable radioactivity was widely distributed across ocular tissues (Fig. 4). All ocular tissues reached C_max at the first collection time point (1 hour post dose), except for the lens from male cynomolgus monkeys (C_max at 24 hours post dose). In eye tissues, the highest concentrations of ¹²⁵I-AMG 386–derived TCA-precipitable radioactivity were found in the choroid, cornea, RPE, iris/ciliary body, and sclera. The percentage of tissue concentration compared with serum as TCA-precipitable radioactivity ranged from less than 1% to approximately 110%, with choroid demonstrating the highest TCA-precipitable radioactivity. In nonocular tissues, the highest concentrations of ¹²⁵I-AMG 386–derived radioactivity in both sexes were present in renal cortex, renal medulla, thyroid, bile, urine, blood, and in stomach contents and gall bladder in females (data not shown). Furthermore, quantifiable but low concentrations were also present in testis, cerebellum, cerebrum, medulla oblongata, and spinal cord (data not shown). 

Figure 4.

 

Mean (+SD) concentration-time profile of ¹²⁵I-AMG 386 in serum and various eye tissues from cynomolgus monkeys after administration of AMG 386. TCA-precipitable μg equivalents ¹²⁵I-AMG 386/g tissue in serum and eye tissues from cynomolgus monkeys at specified time points after a single bolus IV administration of 25 mg/kg of AMG 386. Serum, n = 2 to 8 per time point; eye tissue, n = 4 per time point. Error bars indicate SD for n ≥ 3.

The ¹²⁵I-AMG 386–derived radioactivity was excreted mainly via urine, accounting for 47.4% and 49.6% of dosed radioactivity in males and females, respectively, at 168 hours post dose. Daily cage rinse accounted for 20.4% and 22.4%, respectively, of dosed radioactivity, while feces accounted for <2.5% of total radioactivity in either sex. At 168 hours post dose, 74.5% and 76.6% of the administered dose had been recovered in males and females, respectively. TCA-precipitable radioactivity in urine represented approximately 6% of total dosed radioactivity, indicating that ¹²⁵I-AMG 386–derived radioactivity was excreted mainly as free iodide or small peptide fractions. 

AMG 386 Treatment in a Murine ROP Model

Systemic treatment with AMG 386 significantly decreased hyperoxia-induced retinal neovascularization in a dose-dependent manner, as shown by a reduction in the mean sum of vascular nuclei compared with mice treated with the Fc and IgG controls (Fig. 5A; P < 0.0005). Figure 5B shows representative retinal histologies from AMG 386–treated eyes and relevant controls. The normoxic and untreated tissues were free from neovascularization at the retinal inner limiting membrane (ILM). The Fc-treated hyperoxic retinas were heavily neovascularized, with frequent and often large angiogenic tufts protruding across the ILM into the vitreous space. In marked contrast, the hyperoxic retinas treated with AMG 386 (100 mg/kg) were found to be substantially free from aberrant neovascularization, with only occasional vascular sprouts breaching the ILM. After nine once-daily subcutaneous administrations, AMG 386 concentrations in the serum appeared to increase approximately in proportion to dose in the range of 0.1 mg/kg to 100 mg/kg (Fig. 5C); a 1000-fold increase in dose resulted in a 1530-fold increase in exposure. 

Figure 5.

 

Effect of AMG 386 treatment on hyperoxia-induced retinal neovascularization in mice. (A) Presence of vascular-related (endothelial cell and pericyte) nuclei on the vitreous side of the retinal inner limiting membrane in histologic step sections from normoxic or hyperoxic mice treated either with increasing doses of AMG 386 or various controls. Data are presented as the mean ± SE; n = 5 animals per group. *P < 0.0005, versus Fc control group. (B) Representative histologic step sections of the mouse retina at postnatal day 17 from an untreated animal maintained in a normoxic environment (left) and from hyperoxic animals treated either with Fc control 100 mg/kg (middle) or AMG 386 100 mg/kg (right). Arrows indicate the inner limiting membrane; arrowheads indicate neovascular tufts. (C) Mean (±SD) AMG 386 concentration in the serum across AMG 386 doses in mice with hyperoxia-induced retinal neovascularization. (D) Representative brightfield and darkfield microscope images of H&E staining and in situ hybridization labeling of the expression of Ang1 and Ang2 in mouse retinas under normoxic conditions at postnatal day 17 (left) and hyperoxic conditions (right). Arrows approximate the inner limiting membrane; arrowheads indicate neovascular tufts. H&E, hematoxylin and eosin; ISH, in situ hybridization; P, postnatal day.

In situ hybridization showed a relative increase in Ang2 expression along the ILM in the hyperoxic mouse retinas compared with normoxic controls at P17 (Fig. 5D). A time course study showed that the increased expression of Ang2 began at P14/P15, peaked at P17, and decreased again at P24 (Supplementary Fig. S1 ). Normoxic controls showed no expression of Ang2 in the ILM at these same time points (data not shown). This increase in expression appeared localized to the neovascular tufts along the ILM. Likewise, VEGFR2 expression at P17 appeared to be increased in that same region, compared with P17 normoxic controls (Supplementary Fig. S2 ). In contrast, expression of VEGF along the ILM remained unchanged between the normoxic and hyperoxic conditions at P17. VEGFR2 and VEGF expression was also observed in the inner nuclear layer (INL) in both hypoxic and normoxic retinas. Compared with normoxic controls, VEGF expression in the INL appeared to be slightly increased at P17, while VEGFR2 expression was slightly decreased. Over time, robust VEGF expression in the INL under hypoxic conditions appeared at P12 through P17 and declined at P24 (Supplementary Fig. S3 ). Expression of VEGF in the INL in normoxic controls remained high throughout the same time points (data not shown). No Ang1 expression was detectible by ISH in either the hyperoxic or normoxic retinas at P17 (Fig. 5D). No ISH signals were observed with the sense probes (negative controls) for Ang1, Ang2, VEGF, or VEGFR2. 

The Effects of AMG 386 on Retinal Vascular Regression in the ROP Model

To evaluate the effects of AMG 386 on established retinal vessels, AMG 386 was administered once the pathological vessels had already formed and were in the process of spontaneous regression. AMG 386 prevented retinal angiogenesis when administered from P8 to P16 but impeded regression of these vessels when administered from P17 to P23 (Fig. 6A). This impairment in regression was not permanent, as the number of vascular tufts returned nearly to control levels by P30 even in the continued presence of AMG 386. 

Figure 6.

 

The effect of delayed AMG 386 dosing and combination treatment with AMG 386 plus a VEGFR inhibitor on oxygen-induced retinopathy. (A) The sum of vascular-related nuclei on the vitreous side of the retinal inner limiting membrane in histologic step sections from normoxic or hyperoxic mice treated for various times with AMG 386 (10 mg/kg) or Fc control. Days of treatment duration are noted on the graph. Data are presented as the mean ± SE; n = 5 animals per group. *P < 0.0001 versus respective Fc control. (B) The sum of vascular-related nuclei after dosing with AMG 386 (2 mg/kg, P8 to P16), AMG 273 (a VEGFR2 inhibitor at 15 mg/kg, P12 to P15) or the combination of the two inhibitors at the same doses and schedules. Data are presented as the mean ± SE; n = 5 animals per group. *P < 0.001 versus Fc control; ^# P = 0.006 versus AMG 273 alone. P, postnatal day.

AMG 386 Combined with VEGF Pathway Inhibition in the ROP Model

We examined the effects of inhibiting the VEGF/VEGFR pathway using AMG 273, an investigational small molecule that inhibits the VEGF receptors VEGFR1, 2, and 3 and platelet-derived growth factor receptor (PDGFR). We have previously shown that daily oral treatment with 50 mg/kg of AMG 273 completely inhibits neovascularization in this model. ³⁷ We selected the suboptimal dose of 15 mg/kg to combine with AMG 386, also at a suboptimal dose of 2 mg/kg. Both single-agent treatments significantly inhibited the neovascular response (P < 0.002), and the combination of both agents produced a greater inhibition, although the difference only reached statistical significance when compared with AMG 273 alone (P = 0.006) (Fig. 6B). 

Discussion

The present study explored an alternative approach of blocking angiogenesis associated with ocular diseases by specifically inhibiting angiopoietin function. ^1,25 The main findings are that systemic treatment with AMG 386, an investigational selective Ang1/Ang2 inhibitor, prevented retinal neovacularization in a cynomolgus monkey model of CNV and in a murine model of ROP. Furthermore, AMG 386 delayed the spontaneous regression of established retinal vessels. Last, AMG 386 combined with VEGF pathway inhibition cooperatively prevented angiogenesis in the ROP model. 

The effect of suppression of the Tie2 pathway on CNV lesions has been previously studied in mice, showing that systemic expression of soluble Tie2 resulted in a significant decrease in the incidence and extent of CNV lesions by 52% and 36%, respectively. ³² The present study used a cynomolgus monkey model of laser-induced CNV to demonstrate that treatment with AMG 386 significantly inhibited the occurrence of grade-4 lesions relative to control treatment. This finding is clinically relevant because the appearance of grade-4 lesions provides a good approximation of CNV in humans. Of note, 9 of the 10 lesions reported at study day 23 in the group receiving high-dose AMG 386 were present in one animal (left eye, 5 lesions; right eye, 4 lesions), which had an AMG 386 concentration in the serum that was 50% to 59% below the group mean at all 4 time points. In all other animals withAMG 386 serum concentrations near the mean, grade-4 lesions were virtually eliminated at study day 23. These observations suggest that the systemic exposure to AMG 386, which is correlated with exposure at the target site, dictates the protection against laser-induced CNV. The serum concentration profiles of TCA-precipitable ¹²⁵I-AMG 386–derived radioactivity in cynomolgus monkeys was comparable to that obtained after a single IV dose of unlabeled AMG 386 at 25 mg/kg (data not shown), suggesting that ¹²⁵I-AMG 386 retains the general structure of AMG 386. 

The reduction in the number of grade-4 lesions with AMG 386 treatment, compared with control, was consistent with findings of reduced fibroplasia responses and the presence of retinal herniation at the laser sites. Microscopically, fibroplasia is the primary healing response in laser-treated eyes. Since it does not occur in the absence of blood vessel growth and is the major process of filling abnormal spaces generated by primary and secondary laser injury, it is a useful indicator of AMG 386 efficacy. Retinal herniation occurs in the absence of a proliferative healing response and, thus, can be used as a negative indicator of vascular growth. Therapies that can reduce fibrosis are advantageous because the disruption of the tissue architecture of the posterior segment caused by this process is part of a pathway leading to permanent vision loss, which is common for most sight-threatening diseases. ³⁸ In fact, fibrosis is a key factor in both diabetic retinopathy, the leading cause of vision loss in Americans <65 years of age, and in AMD, the leading cause of vision loss in Americans >65 years of age. ³⁹  

Limited results from experiments in animal models of ROP have previously suggested that Ang2 plays an important role in pathologic angiogenesis. ^34,40,41 In one approach, Ang2-deficient mice failed to induce retinal vascularization, ³⁴ while newborn mice treated with a Tie2 antagonist showed significant inhibition of retinal neovascularization. ⁴⁰ The present study supports a role for Ang2 in retinal angiogenesis by demonstrating increased Ang2 expression in a mouse ROP model. Results from the ISH experiments show that under hyperoxic conditions Ang1 expression was limited, but Ang2 appeared upregulated in the neovascular tufts above the ILM. We have functionally confirmed the dominant role of Ang2 in the ROP model through experiments using selective inhibitors of Ang1 and Ang2; in these studies, we observed that inhibition of Ang2, but not Ang1, completely inhibited the neovascular response. ⁴² Furthermore, addition of Ang1 inhibition to Ang2 inhibition in this model provided no incremental anti-angiogenic effects. Expression of VEGFR2 was also increased in the neovascular tufts, most likely originating from endothelial cells, since pericytes are not known to be associated with the neovascular tufts of the ILM. Under hyperoxic conditions over time, VEGF expression in both the INL and along the ILM was highest at P12 through P17 and then declined. The data show that both the angiopoietin/Tie2 and VEGF pathways are important in hyperoxia-induced retinal neovascularization and are consistent with earlier reports that examined expression of angiogenic factors in the ischemic retina using RT-PCR. ^29,40,43 According to Pierce and colleagues, the VEGF signal originates mostly from cell bodies in the INL, likely to be Müller cells. ⁴³ However, Ang2 was not detected in the ganglion cell layer as reported by others using hypoxic retinas. ⁴⁴  

The present study also provides direct evidence that interference of the angiopoietin/Tie2 pathway inhibits retinal neovascularization in a murine ROP model. Specifically, it demonstrates a significant, dose- and concentration-dependent inhibition of retinal neovascularization with AMG 386 treatment. Similarly, treatment with an anti-VEGF antibody or a small-molecule inhibitor of VEGFR 1, 2, and 3 significantly blocked the development of new blood vessels in hyperoxic mice, underlining the importance of both pathways in angiogenesis. Of note is that the optimal biological dose of AMG 386 in the ROP model was 100 mg/kg (serum trough concentration of 635 μg/mL), much higher than the optimal biological dose reported for AMG 386 in a murine tumor xenograft model. In that study, Oliner and colleagues showed that 0.6 mg/kg associated with serum trough concentration of 3.0 μg/mL provides the same level of tumor growth inhibition as 15 mg/kg. ³³ Although many differences between the two models exist, it is possible that in the ROP model higher systemic AMG 386 doses may be necessary to reach effective concentrations in the deep retinal capillary beds. ²⁹ Indeed, the biodistribution results in cynomolgus monkeys demonstrate that retinal levels of AMG 386 were 30- to 100-fold lower than those in serum at each time point measured. Consistently, the choroidal levels of AMG 386 were 30% to 70% of those in the serum during the 48-hour period post dose, corroborating a similar serum trough concentration (3.7 μg/mL) achieved at the optimal biological dose (5.7 mg/kg) in the CNV model as in the tumor xenograft model. The different dose requirements in the murine ROP (100 mg/kg) and cynomolgus monkey CNV (5.7 mg/kg) models may be due to a differential biodistribution of AMG 386 in various eye tissues as well as differences inAMG 386 clearance between the two species, assuming that the drug distribution to eye tissues is similar in mice and cynomolgus monkeys. Despite the higher doses necessary to reach the optimal biological dose, no overt treatment-related ocular or systemic toxicities were observed in animals receiving AMG 386. Rigorous, controlled pharmacology studies would be required to confirm this observation. 

Several reports in various vascular contexts have noted that Ang2 and VEGF collaborate to spur angiogenesis, but Ang2 in the absence of VEGF induces vessel regression. ^45–49 This also appears to be true in the newborn eye, where retinal Ang2 and VEGF induce retinal angiogenesis, and Ang2 without VEGF induces regression of the hyaloid vasculature. ⁵⁰ Consistent with this concept, in Ang2-null mice, the hyaloid vasculature fails to regress and the retinal vasculature fails to develop. ⁵⁰ In the ROP model, pathological retinal angiogenesis occurs during a period of relative hypoxia starting at P12, with angiogenesis peaking at P17, ⁵¹ a time course that parallels the induction of VEGF expression as demonstrated by Pierce and colleagues ⁵² and in the study described herein (Supplementary Fig. S3 ). These retinal vessels regress by P26, coincident with a reduction in VEGF expression to baseline levels. ⁵² In this model, AMG 386 prevents pathological retinal angiogenesis when administered from P8 to P16, but it impedes regression of these abnormal vessels when administered from P17 to P23. The suppression of regression amounts to no more than a delay, as the number of vascular tufts returns nearly to control levels by P30, even in the continued presence of AMG 386. These results are consistent with the dual roles of Ang2 in the production or regression of vessels, depending on levels of VEGF. 

Anti-VEGF agents currently approved for the treatment of ocular diseases and others being tested in this setting (such as bevacizumab ^18,19 ) are administered via intraocular injection and are associated with a risk for adverse events such as endophthalmitis, elevated intraocular pressure, subconjunctival or intraocular hemorrhage, cataract, and retinal detachment. ^20–22 The IV administration of AMG 386 may circumvent these types of adverse events. Ongoing clinical studies of AMG 386 for the treatment of various cancers have shown that AMG 386 appears to have a unique and favorable safety profile. ^35,53,54 Robust clinical studies will have to establish the safety and efficacy profile of AMG 386 in patients with ocular diseases. An alternative strategy would be to combine AMG 386 with clinically approved VEGF inhibitors in an ocular setting. We have shown that AMG 386 can act in concert with a VEGFR inhibitor to prevent angiogenesis in the ROP model, suggesting that combined angiopoietin and VEGF pathway antagonism could be a promising clinical strategy. Nonetheless, we have also shown that AMG 386 impedes the regression of established ROP vessels when VEGF levels are falling, and it is possible that further reducing VEGF activity with a VEGF antagonist might exacerbate the AMG 386–mediated suppression of vascular regression. Future experiments will test this hypothesis and its therapeutic implications. 

The leading cause of vision loss in macular degeneration and diabetic retinopathy is macular edema resulting from a breach of the inner or outer blood-retinal barrier. The current study has not addressed the role of angiopoietin inhibition in the development of edema, and experimental inquiry into this question may be informative. 

The present study demonstrates that treatment with the Ang1/2 inhibitor AMG 386 blocks pathologic angiogenesis in two distinct animal models of ocular neovascular disease. In both models, the collected pharmacokinetic profile and the absence of any emerging safety concerns during the study support the systemic administration of AMG 386 in this setting. Based on its unique mechanism of action and favorable toxicity profile, future investigation of AMG 386 is warranted in neovascular AMD, PDR, and ROP, alone and in combination with other biologics, such as VEGF inhibitors. 

Supplementary Materials

Supplementary Methods/Results 

Supplementary Figure S1 

Supplementary Figure S2 

Supplementary Figure S3 

Footnotes

 Disclosure: J.D. Oliner, Amgen Inc. (E,I,P); J. Bready, Amgen Inc. (E,I); J. Estrada, Amgen Inc. (E,I); H. Ma, Amgen Inc. (E,I); J. Pretorius, Amgen Inc. (E,I); W. Fanslow, Amgen Inc. (E,I); S. Kaufman, Amgen Inc. (E,I); A. Coxon, Amgen Inc. (E,I); L. Nguyen, Amgen Inc. (E,I); E. Hurh, Amgen Inc. (E,I); T.M. Nork, Amgen Inc. (C); R.A. Leedle, Amgen Inc. (C)