Abstract
Purpose.:
To evaluate the effects of a vascular endothelial growth factor trap (VEGF Trap) on retinal vascular development and pathologic neovascularization (NV) in the canine model of oxygen-induced retinopathy (OIR).
Methods.:
Newborn dogs (postnatal day [P]1) were exposed to 100% O2 and then returned to room air on P5. VEGF Trap (5, 25, or 250 μg) was injected intravitreally in one eye and human FC (hFc) injected in the fellow eye of air control and oxygen-treated dogs on P8. The retinal vasculature and NV were evaluated on P21. Other oxygen-exposed animals received 5 μg of VEGF Trap or hFc on P22 after confirmation of retinopathy of prematurity (ROP)–like pathology and were evaluated at P45.
Results.:
In air controls, both the vascularized area of the retina and the density of superficial capillaries were reduced in 250 or 25 μg VEGF Trap–injected eyes, and deep capillaries were absent. Eyes that received the 5 μg dose were indistinguishable from controls. In oxygen-treated animals, all eyes injected with VEGF Trap exhibited markedly less intravitreal NV than that of hFc-injected fellow eyes, irrespective of dose. Retinal vascular area in OIR animals was significantly reduced in eyes injected with 250 or 25 μg of VEGF Trap, but the 5 μg dose did not inhibit retinal revascularization. Eyes with existing NV that received 5 μg VEGF Trap at P22 exhibited substantial resolution of OIR pathology at P45.
Conclusions.:
The VEGF Trap inhibited the formation of NV, but higher doses also inhibited revascularization of retina when injected at P8. In contrast, the lowest dose tested effectively blocked NV and caused regression of existing NV, without appreciably affecting vasculogenesis or retinal revascularization. These findings suggest that dose selection is an important variable when considering the use of VEGF-targeting agents for the treatment of ROP.
The primary or superficial retinal vasculature in neonatal dogs and fetal humans forms centrifugally by vasculogenesis, de novo formation of blood vessels by differentiation, and coalescence of vascular precursors or angioblasts.
1,2 Angioblasts expressing CXCR4 and CD39 differentiate and migrate through cell-free spaces created by Muller cell processes, assemble into cords, and then primordial capillaries.
1,2 During formation of the dog primary retinal vasculature, vasoproliferative activity is low and most of the cells in mitosis are ablumenal in position and appear to be astrocytes, supporting the view that primary retinal vessel assemblage occurs initially by the process of vasculogenesis.
1 –3 The secondary or deep capillary network develops by angiogenesis, proliferation, and migration of endothelial cells from previously formed superficial retinal blood vessels.
Human retinopathy of prematurity (ROP) is the major cause of blindness in children. Infants born prematurely have incompletely vascularized retinas because the peripheral retina is avascular. ROP occurs with oxidative stress, including exposure of the developing retinal vasculature to hyperoxia. The infants with the most immature retinal vasculature have the greatest risk of ROP.
4 The accepted therapies for ROP are ablation of the peripheral avascular retina, the source of angiogenic growth factors, with cryotherapy or laser.
4,5 Although these therapies are sometimes effective, a new therapeutic approach is desirable not only to improve control of neovascularization (NV) but also because peripheral retina is destroyed by these ablative therapies.
Although an ROP-like retinopathy can be induced in many species by exposing neonatal animals to hyperoxia, the vasculopathy in dog most closely resembles that seen in human ROP.
6,7 Exposure of 1-day-old dogs to hyperoxia for 4 days results in cessation of vasculogenesis as well as vaso-obliteration or destruction of portions of the developing retinal vasculature.
8 When the animals are returned to room air, the vasoproliferative phase of oxygen-induced retinopathy (OIR) is initiated in response to the relative hypoxic state of the poorly vascularized inner retina.
8 Three days after return to room air (postnatal day [P]8), there is extensive proliferation of cells in the retinal vasculature, suggesting that subsequent revascularization in the dog model of ROP occurs principally by angiogenesis.
3 By P21, dilated and tortuous retinal vessels are present in the posterior pole, vascularization of peripheral retina remains incomplete, and vitreous hemorrhage and florid intravitreal (ITV) NV are present, which may be accompanied by ITV hemorrhage.
7 At P45, persistent ITV NV causes tractional retinal folds, tented ITV vascularized membranes, and severe vitreous synchysis in canine OIR eyes. Immunohistochemical analysis revealed inner retinal astrogliosis occurs at the edge of the vasculature and in avascular retina.
9 These results demonstrate that end-stage OIR in dog shares many features with chronic human ROP.
7
It is now apparent that vascular endothelial growth factor (VEGF) plays a critical role both in retinal vascular development
10,11 and in pathologic angiogenesis in ischemic retinopathies
12 –14 and other forms of ocular NV.
15 During development of the superficial vascular network in rat, VEGF is produced by astrocytes in advance of the centrifugal expansion of the vasculature to ora serrata.
11 However, in human and dog, the astrocytes do not precede the forming inner vasculature.
9,16 VEGF production in inner retina has been observed in humans but the cells producing it were not defined.
10 During formation of the deep capillary network, Muller cell production of VEGF is increased in response to the physiologic hypoxia induced by neuronal maturation.
17 Neutralizing VEGF inhibits superficial retinal vasculature development,
18 perhaps by inhibiting endothelial cell migration and the formation of filopodia used for guidance in forming the retinal vasculature.
19
Similarly, when the retinal vasculature is attenuated by occlusive events such as diabetic or sickle cell retinopathy, hypoxia stimulates VEGF expression, resulting in pathologic NV in the retina and at the interface of vitreous and retina.
19 –21 Elevated levels of VEGF have been found in both human ROP
22 and animal OIR models.
23
Recently, reagents targeting VEGF have been developed and have been approved for the treatment of the neovascular or “wet” form of age-related macular degeneration, whereas others are in late-stage clinical trials for this indication.
24 –26 Results to date have been quite promising and the side effects appear limited
24 –26 at least in part because, in contrast to developing blood vessels, the blood vessels of the adult retina and choroid are fully differentiated
27 and maintenance of the structural and functional integrity of the vasculature in the adult eye may no longer be dependent on VEGF.
Experimental interventions that block VEGF signaling have been shown to inhibit formation of ITV NV in rodent OIR models.
28 –30 However, the extent to which these interventions might also inhibit retinal revascularization after vaso-obliteration has not been resolved.
31 Nevertheless, several uncontrolled clinical studies have been conducted on premature infants with stage 4 and stage 5 ROP using ITV administration of bevacizumab, an antibody that binds and neutralizes all VEGF-A isoforms. In these limited studies, not only was neovascular regression observed but also regression of the hyaloid vasculature in vitreous.
32,33
The purpose of this study was to evaluate one of the VEGF-targeted reagents presently being tested in clinical trials, the VEGF Trap, in the canine model of OIR. The VEGF Trap blocks all isoforms of VEGF-A, as well as the related molecule, placental growth factor (PlGF). This study demonstrates that selective inhibition of pathologic ITV NV can be achieved at doses that do not appreciably affect normal retinal vasculogenesis or revascularization of the retina after hyperoxic insult.
This study demonstrated that single ITV injections of VEGF Trap can prevent the development of pathologic ITV NV in the canine model of OIR, and also produce regression of established neovascular membranes. These effects were observed at all doses of VEGF Trap tested: 250, 25, and 5 μg/eye.
An abundance of evidence now demonstrates that VEGF plays a critical role in both normal vascular development in tissues throughout the body and pathologic NV in various disease states
36 –39 ; the present findings are in agreement with this body of literature. With respect to the retina, previously published studies have consistently reported that administration of various pharmacologic inhibitors of VEGF-A, including antibodies, aptamers, and soluble receptors, are capable of ameliorating pathologic NV in murine and rat models of OIR.
29,30,40 Although it is also generally accepted that VEGF plays an indispensable role in vascular development in the retina
29 –31,40 few studies have specifically evaluated the effects of VEGF inhibitors on either retinal vascular development or revascularization after exposure to hyperoxia.
Ishida et al.
31 reported that administration of a VEGF
164-specific aptamer inhibited only pathologic NV, whereas a VEGFR-1 fusion protein also attenuated normal retinal vascular development in mice, as well as retinal revascularization in the OIR model. Similarly, we have previously reported that ITV administration of an antibody against VEGFR-2 (cp1c11) inhibited retinal vascular development as well as NV in the canine OIR model.
41 However, Geisen et al.
29 reported that an anti-VEGF antibody did not appear to interfere with normative revascularization of the retina post hyperoxia in the rat OIR model. Budd et al.
42 recently observed the same result with an inhibitor of VEGFR-2 signaling (SU5416). Although the reasons for these seemingly discrepant results are presently unclear, it is important to note that the relative potency of the above-cited agents is unknown, and the effects of varying the dose were not explored.
In the present study, we evaluated the effects of VEGF Trap on normal retinal vascular development and revascularization over a 50-fold range of doses, from 5 to 250 μg/eye. In contrast to the effect on pathologic NV, the effects of VEGF Trap on both retinal vascular development and revascularization after exposure to hyperoxia, were clearly dose dependent. Specifically, in air-raised animals, the area of retina covered by the superficial vascular plexus was significantly reduced in the 250 and 25 μg/eye doses of VEGF Trap and the development of the newly forming deep capillary plexus was also inhibited. In animals exposed to hyperoxia, the 250 and 25 μg doses also reduced the extent of retinal revascularization in a dose-dependent fashion. However, ITV administration of 5 μg VEGF trap did not affect any of the above-mentioned parameters, in either oxygen or air-raised pups. This dose effectively prevented the development of NV but the retinal vasculature of pups treated at this dose was indistinguishable from that of similarly raised hFc-treated controls. Thus, selective inhibition of pathologic NV in OIR may be achieved with a potent, pan-VEGF inhibitor by appropriately titering the dose.
We have recently demonstrated that VEGFR-2 (also known as kinase insert domain receptor) is upregulated in the canine model of OIR.
41 In 15-day-old oxygen-exposed animals, where both established ITV NV and immature forming NV exist, we observed higher levels of VEGFR-2 immunostaining in the immature NV, as well as the reforming retinal vasculature. This explains why at high doses the VEGF Trap and VEGFR-2 antibody block both of these angiogenic events. Other groups have demonstrated that VEGF and its receptors are elevated in oxygen-induced retinopathy models in other species.
14,43,44
There are several possible reasons that low doses of the VEGF Trap were able to block pathologic NV without interfering with normative retinal revascularization. For example, the drug may have been occupied by VEGF in vitreous, and was not available to bind retinal VEGF. However, the low dose of VEGF Trap did not alter normal retinal vascular development in air-raised animals, where ITV levels of VEGF were undoubtedly much lower, making this explanation seem implausible. However, VEGFR-2 immunoreactivity was far greater on preretinal forming vessels in OIR than that in vessels developing by vasculogenesis in the dog retina.
41 An alternative explanation is that pathologic angiogenesis is inherently more dependent on the presence of VEGF-A than are vessels developing normally by vasculogenesis. Shih and associates
45 argue that PlGF may be the critical VEGF family member and VEGFR-1 the important receptor for development and maintenance of the superficial vasculature in mouse; however, the VEGF Trap neutralizes PlGF and all isoforms of VEGF.
Some recent nonclinical studies have suggested that VEGF also serves a neurotrophic function.
46 –48 However, evaluation of retinal function in rabbit
49 and in mouse
50 after bevacizumab injection revealed no change in electroretinography (ERG). Similarly, studies in nonhuman primates have not detected any adverse effect of repeated ITV administration of VEGF Trap on the mature retinal vasculature, retinal morphology, or neural retinal function as assessed by ERG (Zimmer E, et al.
IOVS 2006;47:ARVO E-Abstract 1751). Moreover, in the present study, there were no observable morphologic changes in the neural retina of eyes receiving VEGF Trap at any dose. Importantly, the now extensive use of pan-VEGF blockers in humans, first in carefully controlled clinical studies and subsequently in clinical practice, has not been associated with impairments in neural retina structure or function.
The canine OIR model provides a useful forum for preclinical evaluation of angiogenesis inhibitors. The resultant pathology closely resembles human ROP, in an eye approximately equivalent in size and vitreous volume to that of a 25-week gestation human. Furthermore, in the canine model of OIR, the pathologic NV persists at least until P45 and does not spontaneously regress, unlike in rodent models. We have evaluated the effects of a potent VEGF inhibitor, VEGF Trap, using this model, and determined that a single injection of as little as 5 μg effectively prevented the development of ITV NV, and also caused regression of established NV and relief of vascular tortuosity in eyes with proliferative retinopathy. However, as expected from findings obtained with other VEGF blockers, high doses of VEGF Trap also inhibited normal retinal vascular development. These results suggest that careful dose titration of VEGF inhibitors may be necessary to produce an optimal and selective inhibition of NV in the developing retina.
Supported in part by National Eye Institute Grants EY09357 (GAL) and EY01765 (Wilmer Ophthalmological Institute), by the Brownstein Family Foundation, and by Regeneron Pharmaceuticals, Inc.
Disclosure:
G.A. Lutty, Regeneron Pharmaceuticals, Inc. (F);
D.S. McLeod, Regeneron Pharmaceuticals, Inc. (F);
I. Bhutto, Regeneron Pharmaceuticals, Inc. (F);
S.J. Wiegand, Regeneron Pharmaceuticals, Inc. (E, F)
The authors thank Sam D'Anna for assistance with fundus photography and Susan Croll for assistance with statistics.