Prevention of Experimental Choroidal Neovascularization and Resolution of Active Lesions by VEGF Trap in Nonhuman Primates FREE

Age-related macular degeneration (AMD) is a leading cause of blindness whose incidence is likely to increase as the population ages.¹ The great majority of individuals with AMD have the dry form, which is characterized by atrophic degeneration of the retinal pigment epithelium with secondary (and often gradual) damage to the photoreceptors. However, 80% to 90% of patients with AMD who develop severe vision loss have the wet (neovascular) form,² which occurs when abnormal new blood vessels originating from the choroid grow through the Bruch membrane into the subretinal or intraretinal space. This choroidal neovascularization (CNV) was formerly treated with thermal laser photocoagulation according to protocols developed as part of the Macular Photocoagulation Study and related subsequent studies.^3- 8 Although the treatment was effective at slowing the progression of the disease, it seldom resulted in improved vision because the thermal laser also irreversibly damaged the overlying retina. The patients were often left with central scotomas from the treatment itself. Since then, drugs such as pegaptanib sodium (Macugen) and ranibizumab (Lucentis) have been developed for human use; these work by inhibiting vascular endothelial growth factor (VEGF).

VEGF Trap is a potent VEGF inhibitor comprising ligand-binding portions of human VEGF receptor 1 (VEGFR1) and VEGFR2 fused to the Fc segment of human IgG1 (Figure 1).⁹ VEGF Trap binds and neutralizes multiple isoforms of VEGF-A (dissociation constant of approximately 1pM) as well as the related angiogenic factor placental growth factor (PlGF) (dissociation constant of approximately 40pM). An intravenous formulation of VEGF Trap, generically known as aflibercept, is being developed for oncology; this formulation is hyperosmotic and diluted prior to intravenous infusion. VEGF Trap-Eye, known generically as aflibercept ophthalmic solution, is an iso-osmotic, ultrapurified formulation of VEGF Trap for intravitreous injection. Phase 3 studies of VEGF Trap-Eye in patients with neovascular AMD and retinal vein occlusion are currently in progress.

The purposes of this study were to evaluate the efficacy of systemic and intravitreous administration of VEGF Trap in a primate model of CNV and to evaluate histological changes associated with the angiographic improvements observed. This study was completed prior to initiating the human clinical trial program for VEGF Trap-Eye.

The effect of VEGF Trap treatment on laser-induced CNV was evaluated in cynomolgus monkeys (1.8-2.7 kg at initiation of dosing) using a modification¹⁰ of a model of CNV developed by Ryan¹¹ and Ohkuma and Ryan.¹² All of the experimental methods and techniques adhered to the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by our institutional animal care and use committee. Animals were anesthetized with ketamine hydrochloride and xylazine hydrochloride. A 532-nm diode laser (OcuLight GL; Iridex Corp, Mountain View, California) with a table-mounted slitlamp adapter was used to create small (75-μm diameter), intense laser spots of 0.1-second duration that were applied to 9 areas of the macula of each eye. Initially, the power setting was 500 mW for all spots except the one just temporal to the fovea, which was treated with 400 mW. If no hemorrhage occurred at a given spot, a second spot was placed adjacent to it using a laser intensity of 150 mW greater than the initial burn. The development of active CNV lesions was assessed by fluorescein angiography (FA), once before injury and 15, 20, and 29 days after laser injury. The CNV lesions were graded by a masked observer (T.M.N.) using the following scale: grade 1, no hyperfluorescence; grade 2, hyperfluorescence, without leakage; grade 3, hyperfluorescence early or midtransit, and late leakage; and grade 4, bright hyperfluorescence early or midtransit, with late leakage extending beyond the borders of the laser spot.

In the prevention studies, VEGF Trap was administered by intravenous injection (3 or 10 mg/kg of body weight weekly) or intravitreous injection (50, 250, or 500 μg/eye every 2 weeks) beginning approximately 1 week before laser injury. For intravitreous injection, VEGF Trap was formulated in 10mM sodium phosphate, 135mM sodium chloride, and 0.1% polyethylene glycol 3350 (pH 6.25) and injected through a 30-gauge sterile needle in a volume of 50 μL (500 or 50 μg) using a 1-mL tuberculin syringe or 25 μL (250 μg) using a 0.3-mL syringe. VEGF Trap for intravenous injection was formulated in 5mM sodium phosphate, 5mM sodium citrate, 100mM sodium chloride, 20% sucrose, and 0.1% polysorbate 20 (pH 6.0) and infused in a volume of 4 to 5 mL/kg of body weight over 30 minutes. Control animals received weekly intravenous infusions or biweekly intravitreous injections (50 μL) of placebo comprising the appropriate vehicle solutions according to the same schedule as for corresponding VEGF Trap–treated groups.

In the treatment study, a single intravitreous injection of VEGF Trap (500 μg) was given 15 days following laser injury, at which time active CNV had already formed. Each of the experimental and control groups comprised 6 animals, including 3 males and 3 females; both eyes were treated identically (Table 1).

Animals were anesthetized with ketamine and xylazine, and the eyes were instilled with 0.5% proparacaine hydrochloride, cleaned with 2.5% povidone-iodine, and rinsed with sterile saline. Immediately following each injection, a single topical dose of tobramycin and dexamethasone (Tobradex) ointment was applied to the eye. No systemic antibiotics were used. The left and right eyes of each animal received the same dose of either VEGF Trap or placebo (as opposed to a study design that used the fellow eye as the control) to eliminate the possibility of a systemic effect on the control eye.

Daily cage-side observations were performed on all animals to monitor for clinical signs of poor health, including any ocular abnormalities. Animals also underwent clinical ophthalmic examinations before the initiation of treatment and on postlaser days 7, 21, and 32 (intravitreous prevention groups) and days 9, 23, and 33 (intravenous prevention groups and intravitreous treatment group, excluding day 9). The anterior portion of each eye was viewed using a handheld slitlamp biomicroscope, and the ocular fundus was viewed with an indirect ophthalmoscope. Intraocular pressure was monitored. Fundus photographs were taken on the day of laser treatment (following laser injury) and approximately 4 weeks later, preceding the final FA.

Anterior chamber and vitreous cell scores were determined for right and left eyes using a slitlamp biomicroscope as follows: a score of 0 indicates no cells observed; a score of 0.5+, 1 to 5 cells per single field of focused beam; a score of 1+, 5 to 25 cells per single field of focused beam; a score of 2+, 25 to 50 cells per single field of focused beam; a score of 3+, 50 to 100 cells per single field of focused beam; and a score of 4+, more than 100 cells per single field of focused beam. Scores from both eyes were averaged per animal. Means and standard deviations are based on 6 animals per group. Figure 2 shows the timing of dosing, FA, ophthalmic examinations, and necropsy relative to the day of laser treatment for each of the 3 treatment arms.

For each of the 3 postlaser angiography intervals (days 15, 20, and 29), the proportions of grade 4 counts were dichotomized to 1 and 0 and the Cochran-Armitage trend test was applied to the intravenous prevention and intravitreous prevention groups separately. Fisher exact tests were also conducted for group comparisons between treated groups and the control group.

For the intravitreous treatment group and the intravitreous prevention placebo group, data from day 15 were treated as baseline data and were subtracted from the data on days 20 and 29. The difference was then analyzed for days 20 and 29 separately using Wilcoxon signed rank test.

All test results are exact because of the small sample sizes. All statistical tests were conducted at the 5% level.

Animals were killed on postlaser day 33 (intravitreous prevention groups) or day 35 (intravenous prevention groups and intravitreous treatment group) and the upper body was perfused through the aorta (descending clamped) with half-strength Karnovsky fixative. The eyes were removed, postfixed for 2 to 3 days in half-strength Karnovsky fixative, and then stored in formalin until processed.

One eye from each animal in the intravitreous placebo, VEGF Trap (500 μg) prevention, and VEGF Trap treatment groups was selected for histopathological evaluation. The selected eyes were representative and comprised approximately half of the grade 4 lesions for each of the groups. Strips of tissue containing 1 or 2 lesion sites were embedded in plastic. Sections 2 μm thick were taken at 30-μm steps through the middle of each lesion. The sections were stained with toluidine blue, and the sample with the most robust lesion was designated as the central cut. This section was then evaluated by an observer (R.R.D.) masked to the treatment condition.

A tissue proliferation score was calculated for each lesion based on 3 criteria: the size of the spindle cell proliferative lesion, the extent of new blood vessel proliferation in the subretinal space, and the elevation of the retina above the choriocapillaris (Figure 3). Each measure was graded from 0 to 3, with 0 indicating not present. The total tissue proliferation score comprises the sum of each of the described measures for each laser lesion site.

Intravitreous administration of the VEGF Trap placebo control article was well tolerated, with 0.5+ vitreous cells seen in 1 of 6 animals in this group. No anterior chamber cells were detected at the designated examination times in animals receiving intravitreous injections of placebo (Table 2).

Intravitreous administration of the VEGF Trap test material at all dose levels resulted in no (0) or mild (0.5+ to 1+) inflammatory cell scores in the anterior chamber or vitreous. During the course of the study, trace (0.5+) levels of anterior chamber cells were seen in 4 of 6 animals in the mid-dose group (250 μg/eye/dose) and 3 of 6 animals in the high-dose group (500 μg/eye/dose) in the multiple (biweekly) intravitreous dose prevention experiment and in 1 of 6 animals in the single intravitreous dose treatment study (500 μg/eye following CNV formation). Vitreous cell scores were also mild (0.5+ to 1+) in all of the groups that underwent intravitreous injection of VEGF Trap, but vitreous cells were more frequent and detected in all animals in these groups at some time during the study. This finding was not unexpected, because inflammatory cells are much slower to enter and clear from the more viscous vitreous gel than the aqueous humor. These results are summarized in Table 2. At no time or dose did the mean cell inflammatory score exceed 1+ in any eye. Ocular examinations were performed approximately 2 weeks following injections in the intravitreous prevention study, so early transient inflammation may have been missed. However, the animals in the intravitreous treatment group were examined 8 days after injection and only mild inflammation was observed (on study day 23) (Table 2). No animals showed gross evidence of ocular or systemic toxic effects based on daily cage-side inspections. There were no significant effects on intraocular pressure beyond a transient elevation in all groups immediately following intravitreous injection.

Intravenous administration of VEGF Trap placebo or VEGF Trap at a low or high dose produced no detectable anterior chamber or vitreous cells.

No evidence of a retinal inflammatory response (eg, perivascular sheathing, retinal thickening, optic nerve swelling, or retinal vascular leakage) was found on color fundus photography or FA in any of the animals.

Of the 4 grades assigned to the laser treatment spots, grade 4 (bright hyperfluorescence early or midtransit, with late leakage extending beyond the borders of the laser spot) corresponds to clinically significant leakage. Grade 4 lesions are thought to reflect the presence of new choroidal vessels that either have grown beyond the laser treatment spot or are leaking so intensely that the fluorescein dye has spread markedly away from the vessels. The results with respect to grade 4 leakage for all groups are shown in Table 3. The average number of grade 4 lesions in the intravitreous placebo group ranged from 26.9% to 32.4% during the times evaluated (postlaser days 15, 20, and 29), while 45.4% to 50.0% of the laser treatment areas show grade 4 leakage in the intravenous placebo group. The mean percentage of grade 4 lesions in the control groups was similar to that which has been reported by others using this animal model of CNV.^10- 11,13 By contrast, all of the VEGF Trap prevention groups showed marked reduction or complete absence of grade 4 lesions, irrespective of dose (Figure 4 and Figure 5). Table 4 shows the distribution of all lesion grades on day 29 for the prevention groups.

In the VEGF Trap treatment group (single intravitreous injection of VEGF Trap administered on postlaser day 15), 44.4% of laser treatment spots exhibited grade 4 leakage on day 15, similar to the percentage of grade 4 spots in the 2 placebo control groups. However, by postlaser day 20 (5 days following intravitreous administration of VEGF Trap), only 1.9% of the spots were grade 4; no spots were grade 4 at day 29 (Table 3, Figure 6, and Figure 7). When all lesion grades were compared, there was a marked shift from mostly grades 4 and 3 in the day 15 (pretreatment) angiograms to mostly grades 2 and 1 in the day 29 angiograms (Figure 8).

Consonant with the FA findings, histological evaluation revealed that intravitreous administration of VEGF Trap reduced proliferative responses of the retina to laser injury, particularly neovascular proliferation.

When VEGF Trap administration was begun prior to laser injury (prevention), choroidal fibroplasia and retinal elevation scores as well as CNV scores were all significantly lower in VEGF Trap–treated animals relative to placebo controls (Table 5 and Figure 9). When a single injection of VEGF Trap was given after grade 4 lesions had developed, there was also a trend toward decreased CNV, but mean scores for fibroplasia and retinal elevation were not significantly different from controls (Table 5 and Figure 10).

Important advances were made in the treatment of AMD by the application of drugs that act to destroy and/or prevent formation of the new blood vessels. The first of these to be approved for human use was photodynamic therapy using the photosensitizing dye verteporfin (Visudyne; Novartis, Basel, Switzerland) administered intravenously followed by exposure of the CNV to 689-nm low-energy laser. Photodynamic therapy greatly reduced direct retinal damage from prior thermal laser therapy. However, there were problems with recurrence, and patients continued to have a decline in vision over time.¹⁴

Following the development of photodynamic therapy, a new family of drugs that act to inhibit the cytokine VEGF-A was developed. VEGF-A has been implicated as a causal factor in the development of the wet form of AMD as well as other ocular vascular diseases characterized by pathological neovascularization and vascular leak and/or edema. A number of strategies are being developed to inhibit VEGF-A signaling in these conditions, including application of antibodies to VEGF-A or the VEGF receptors, VEGF-binding aptamers, and small interfering RNAs and treatment with kinase inhibitors. The first of these to be approved for human use was pegaptanib (Macugen), an RNA aptamer directed against the VEGF-A 165 isoform.^15- 16 Inhibition of VEGF-A 165 was shown to slow the progression of vision loss in wet AMD but did little to reverse vision loss. More recently, intravitreous administration of ranibizumab (Lucentis) has been approved for the treatment of AMD. Ranibizumab is a humanized monoclonal antibody Fab fragment that is directed against all isoforms of VEGF-A. It has largely replaced pegaptanib in clinical practice following 2 large, clinical, phase 3 trials (Minimally Classic/Occult Trial of the Anti-VEGF Antibody Ranibizumab in the Treatment of Neovascular AMD [MARINA]¹⁷ and Anti-VEGF Antibody for the Treatment of Predominantly Classic Choroidal Neovascularization in AMD [ANCHOR]^18- 19) showing that 94% to 96% of patients receiving 0.5 mg of ranibizumab monthly lost fewer than 15 letters of visual acuity and 34% to 40% actually gained 15 letters. The related drug bevacizumab (Avastin), a humanized whole IgG1 antibody approved for oncology, is also used off label by clinicians.²⁰ Despite these advances, the current treatment of choice for AMD (either ranibizumab or bevacizumab) requires repeated intravitreous injections on a monthly basis for an indeterminate period—possibly years—to maintain improvements in visual acuity.

Extensive literature demonstrates that VEGF-A is a critical factor contributing to the development of ocular neovascularization (for a review, see the article by Witmer et al²¹). In contrast to other agents that bind and neutralize only VEGF-A, VEGF Trap also binds and neutralizes PlGF.⁹ Placental growth factor is a member of the VEGF family of cytokines that is expressed prominently in the placenta, the tissue from which it was first isolated.²² It can promote angiogenesis directly or by enhancing VEGF-A activity.^23- 25 In contrast to VEGF-A, which also plays an indispensible role in normal vascular development, PlGF has been specifically implicated in promoting pathological neovascularization. While genetic deletion of even a single allele of VEGF-A results in profound impairments in vascular development, normal vascular development and function are not appreciably impaired in PlGF-null mice. However, genetic deletion or pharmacological inhibition of PlGF significantly reduces pathological neovascularization as well as the associated vascular leakage in numerous disease settings.²⁶ Like VEGF-A, PlGF appears to be involved in promoting ocular vascular disease in both humans and animals. For example, PlGF is present in CNV membranes excised from human eyes,²⁷ and experimental CNV is decreased in PlGF-null mice and mice treated with PlGF neutralizing antibodies relative to controls.²⁸

The proangiogenic and propermeability effects of VEGF-A are thought to be mediated primarily through VEGFR2 expressed on vascular endothelial cells. A structurally related receptor, VEGFR1, binds both VEGF-A and PlGF. In addition to being present on endothelial cells, where receptor ligation is also thought to promote angiogenesis and vascular permeability, albeit more weakly, VEGFR1 is expressed by many other cell types including leukocytes, pericytes, smooth muscle cells, and endothelial progenitor cells.²⁹ Thus, in addition to promoting angiogenesis and vascular permeability by acting directly on endothelial cells, VEGF and PlGF can also act via VEGFR1 on a variety of other cell types involved in blood vessel formation and stabilization. Moreover, VEGF and PlGF serve as potent chemoattractants and activators of leukocytes, particularly monocytes, in a variety of pathological conditions.^29- 30

VEGF Trap, administered either as serial subcutaneous injections or as a single intravitreous injection, has been shown to suppress laser-induced CNV in mice.³¹ Moreover, VEGF Trap given subcutaneously inhibits retinal neovascularization in transgenic mice that overexpress VEGF in photoreceptors. Furthermore, VEGF Trap was found to reduce breakdown of the blood-retinal barrier following intravitreous injection of VEGF and in transgenic mice that overproduce VEGF in the retina.³¹ Systemic administration of VEGF Trap also has been shown to suppress neovascularization and the associated inflammatory cell infiltrate following corneal injury³² and to delay corneal allograft rejection in mice.³³ More recently, VEGF Trap has been reported to prevent the development and promote the regression of recently formed CNV following subretinal injection of matrigel in rats.³⁴ Interestingly, VEGF Trap treatment also reduced CNV-associated fibrosis and inflammation in this model.

Although CNV can be induced in other species,^35- 37 only nonhuman primates have maculae similar to the human macula. Thus, the model by Ryan¹¹ of inducing CNV using intense, small laser spots applied to the macular retina to break the Bruch membrane has become a standard means of assessing the preclinical efficacy of pharmacological treatments for wet AMD (ie, CNV). For example, this model was used for preclinical evaluations of photodynamic therapy^38- 41 and Lucentis.¹⁰ Even so, the model has its limitations. The young nonhuman primates have otherwise healthy retinae (including retinal pigment epithelia) and the induced CNV, unlike CNV in elderly humans with AMD, is self-limiting, resolving in 6 to 8 weeks without treatment. Also, the model has considerable variability. Only about 40% of the treatment spots go on to develop grade 4 lesions¹¹ and 20% of the animals are nonresponders, with no CNV developing in either eye (T.M.N. and B.J.C., unpublished data, May 2008). Therefore, it is important to have an adequate number of subjects in each group.

In this animal model of CNV, VEGF Trap was highly effective at preventing the development of grade 4 leakage on FA regardless of dose or whether it was administered intravenously on a weekly schedule or intravitreously every 2 weeks (Table 3, Figure 4, and Figure 5). Histological assessment confirmed that choroidal new vessel formation, fibrotic changes, and retinal thickness also were markedly less in the treated eyes (Table 5).

Moreover, when a single intravitreous injection of VEGF Trap was given after grade 4 CNV had developed, leakage was stopped within 5 days in approximately 95% of previously active grade 4 lesions and within 14 days following treatment in 100% of the lesions (postlaser days 20 and 29, respectively) (Table 3, Figure 6, and Figure 7). Although the effect of a single intravitreous injection of placebo was not evaluated, grade 4 lesions persisted for the duration of the study in all animals receiving multiple intravitreous or intravenous injections of placebo (Table 3). Histological examination revealed a trend toward decreased CNV and fibrosis relative to controls, which was not statistically significant. VEGF is a powerful mediator of vascular permeability in addition to new vessel formation, so VEGF Trap may have blocked VEGF-induced leakage from choroidal neovessels. Alternatively, VEGF Trap may have reduced or stopped blood flow through the new vessels.

Intravitreous administration of VEGF Trap was well tolerated, with only a mild inflammatory response noted in the eyes that underwent intravitreous VEGF Trap treatment. Except for 1+ or fewer anterior chamber and vitreous cells in some eyes, no other ophthalmoscopic signs of inflammation were seen.

VEGF Trap is now in clinical trials (for a recent review, see the article by Dixon et al⁴²). A phase 1 trial of 25 patients with exudative AMD evaluated the tolerability and efficacy of intravenous administration of VEGF Trap at 3 different dose levels. Subjects had a significant decrease in retinal thickness as determined by optical coherence tomography,⁴³ although visual acuity was not significantly improved in this small safety study. However, 1 subject experienced grade 4 hypertension and 1 subject developed grade 2 proteinuria. Hypertension and proteinuria are now well-established class effects of systemic VEGF inhibition, and both patients exhibiting these adverse events in the study by Nguyen et al⁴³ had received the highest intravenous dose of VEGF Trap (3 mg/kg).

Another phase 1 study (Clinical Evaluation of Anti-angiogenesis in the Retina, CLEAR-IT 1) used intravitreous administration of VEGF Trap-Eye (aflibercept ophthalmic solution).⁴⁴ The first part of this study was a sequential cohort dose escalation (from 0.05 to 4.0 mg/eye) in 21 patients with exudative AMD. No serious systemic or ocular toxic effects were observed. However, a marked decrease in retinal thickness⁴⁴ and improvement in visual acuity⁴⁴ were noted. VEGF Trap-Eye also has been used in a small open-label safety study for treatment of diabetic macular edema.⁴⁵ A single dose of 4 mg was administered intravitreously to 5 patients who had undergone multiple prior treatments for diabetic macular edema. There was a median decrease in central macular thickness of 79 μm as well as some improvement in vision. A phase 2 trial in diabetic macular edema is in progress.

In a double-masked phase 2 trial (CLEAR-IT 2), VEGF Trap-Eye was evaluated in 157 patients with exudative AMD randomized to either monthly or quarterly intravitreous injections for 12 weeks at doses of 0.5 or 2 mg (monthly injections) and 0.5, 2, or 4 mg (quarterly). Following the 12-week fixed dosing period, patients continued to receive treatments on an as-needed basis at their originally assigned dosages. Reports of the 1-year results described a statistically significant improvement in vision, retinal thickness, and size of the CNV lesions,^46- 47 with few re-treatments required during the 40-week phase of as-needed treatment. Patients initially dosed on a schedule of 2.0 mg monthly received, on average, only 1.6 additional injections during the 40-week period of as-needed treatment, and those initially dosed on a schedule of 0.5 mg monthly received, on average, 2.5 injections. While as-needed dosing following a fixed quarterly dosing regimen (with dosing at baseline and week 12) also yielded improvements in visual acuity at week 52 as compared with baseline, the results generally were not as robust as those obtained with initial fixed monthly dosing. VEGF Trap-Eye was generally well tolerated and there were no drug-related serious adverse events. The most common adverse events were those typically associated with intravitreous injections.

Two phase 3 trials of 2 years' duration are under way to further investigate the efficacy and safety of VEGF Trap-Eye in wet AMD, VIEW 1 in the United States and Canada⁴⁸ and VIEW 2 in Europe, Japan, and Latin America.⁴⁹ For both trials, VEGF Trap-Eye is being administered intravitreously. In the first year of treatment, VEGF Trap-Eye was administered every 4 weeks at doses of either 0.5 or 2 mg. Another study arm used 3 initial monthly doses of 2 mg followed by 2-mg doses given at 8-week intervals. The active control arm comprised subjects receiving ranibizumab (0.5 mg) at 4-week intervals. The 1-year outcomes from these studies are pending publication.

Using an established primate model of CNV, administration of VEGF Trap in a prevention protocol markedly reduced vasoproliferative responses of the macaque retina to laser injury, substantially preventing the development of all components of CNV lesions as well as vascular leakage. When a single intravitreous VEGF Trap injection was given after grade 4 lesions had developed, there was resolution of vascular leakage. This also resulted in a trend toward lower histological scores for the neovascular components of the lesions, suggesting partial regression of newly formed vessels.