Pharmacology of Antiangiogenic Agents
Peter A. Campochiaro
The eye has several avascular tissues including the cornea, lens, vitreous, and outer retina. Growth of blood vessels in these tissues compromises the optical qualities of the eye. Also, normal retinal functioning requires an intact blood–retinal barrier. These unique features of the eye make neovascularization (NV) a liability, whereas in most tissues throughout the body it is an essential component of wound repair. In fact, diseases complicated by NV are the most common causes of vision loss in developed countries.1,2 Choroidal neovascularization (CNV) occurs in diseases that compromise Bruch’s membrane and/or the retinal pigmented epithelium (RPE), including age-related macular degeneration (AMD), pathologic myopia, ocular histoplasmosis, angioid streaks, and several types of choroiditis. Retinal NV occurs in ischemic retinopathies such as diabetic retinopathy, central and branch retinal vein occlusion, and several types of vasculitis. Excessive vascular permeability and macular edema also occur in ischemic retinopathies.
Partial elucidation of the molecular pathogenesis of retinal and choroidal NV
Clinical observations often provide clues that assist in directing research regarding the pathogenesis of disease. Clinicians have known for many years that there is a strong correlation between retinal ischemia and retinal NV. This led to the hypothesis that ischemic retina releases a substance(s) that stimulates retinal NV.3,4,5 Investigators began searching for retina-derived factors that could stimulate proliferation and migration of endothelial cells in vitro and angiogenesis in vivo. Basic fibroblast growth factor (FGF2) has these characteristics, and when it was isolated from retina, FGF2 was felt to be a leading candidate as a primary stimulus for retinal NV6 until studies in genetically engineered mice demonstrated that it is not.7,8,9 Vascular endothelial growth factor (VEGF) also fit the profile, and when it was demonstrated that VEGF production is substantially elevated in ischemic tissue,10,11 many investigators postulated that it may be released from ischemic retina and stimulate retinal NV. Levels of VEGF are elevated in eyes with retinal NV and VEGF antagonists suppress retinal NV in animal models.12,13,14,15,16 Overexpression of VEGF in transgenic mice results in sprouting of NV from the deep capillary bed of the retina.17,18 However, increased production of VEGF in the RPE does not cause CNV19 and because ischemia is not clearly present in patients with CNV as it is in patients in retinal NV, it was unclear whether VEGF was a major stimulus for CNV until it was demonstrated that VEGF antagonists strongly suppress CNV in a mouse model.20 Thus, these studies in animal models strongly suggested that VEGF was a critical stimulus in both retinal and choroidal NV (for reviews, see references21,22,23,24).
Development of VEGF antagonists
Neutralizing monoclonal antibodies are extremely valuable research tools, because they provide a means to recognize a protein of interest in tissues and block their activity. A monoclonal antibody that binds with high affinity in a desired region of a protein can also be the basis of a therapeutic agent; it is humanized by mutating nonbinding regions and changing amino acids that differ between mouse and human IgGs to make as many as possible correspond to human sequence. Bevacizumab is a full-length humanized monoclonal antibody that binds all isoforms of VEGF-A25 developed by scientists at Genentech, a biotech company in south San Francisco. It was developed to administer by intravenous infusion to treat metastatic cancer. It was felt that the 150 kDa molecular weight of bevacizumab would limit its penetration through the retina after intraocular injection; therefore, ranibizumab, a 48 kDa Fab was developed for ocular indications. It is derived from another monoclonal antibody and was modified to increase its binding affinity for VEGF-A (all isoforms). In bioassays, ranibizumab was 5- to 20-fold more potent than bevacizumab on a molar basis.26 The increased potency, smaller size for enhanced penetration into the retina and choroid, and systemic half-life of a few hours compared to roughly 3 weeks for bevacizumab were considered advantages for intraocular delivery. The half-life after a single intraocular injection of ranibizumab in monkeys was 3 days, and serum levels were very low, approximately 1,000-fold lower than levels in the eye.27 The half-life after an intraocular injection of the full-length antibody, trastuzumab (148 kDa), which is comparable in size to bevacizumab, is 5.6 days.28 Thus, Genentech began clinical trials to test the effects of systemic infusions of bevacizumab in patients with cancer and trials to test the effects of intraocular injections of ranibizumab in patients with CNV due to AMD.
The specificity of antigen–antibody interactions is determined by the tertiary structures of each allowing them to “fit together” in such a way that hydrogen bonding occurs. RNA molecules have tertiary structure, and it is possible to generate a library of RNA molecules and screen for those that by chance happen to have a tertiary structure that causes them to specifically bind to a protein of interest. Such RNA molecules that specifically bind a protein are called aptamers and an aptamer that specifically binds the VEGF165 isoform was identified by scientists at NeXstar, a biotech company in Boulder, CO.29 It was licensed by a start-up company, EyeTech, and when given by intraocular injection every 6 weeks for a year, reduced severe visual loss in patients with AMD due to CNV30. The results with ranibizumab were far more impressive and mutually confirmatory in two trials investigating effects in slightly different patient populations. The different patient populations have to do with fluorescein angiographic characteristics of CNV by which it is categorized as occult, classic, or minimally classic. Classic CNV hyperfluoresces early, is usually fairly well defined, and tends to show substantial leakage during fluorescein angiography, whereas occult CNV generally becomes hyperfluorescent relatively late, is poorly defined, and usually does not show impressive leakage. The major significance of the distinction is that classic CNV tends to grow more rapidly and cause visual loss more quickly than is the case for occult. Lesions frequently contain both types; those that are more than 50% classic are referred to a predominantly classic, whereas those that contain less than 50% are referred to as minimally classic. Photodynamic therapy (PDT) with verteporfin caused a significant reduction in severe vision loss in AMD patients with predominantly classic, but not those with minimally classic.31 In the MARINA trial,32 monthly intravitreous injections of 0.3 or 0.5 mg of ranibizumab reduced the percentage of patients with moderate loss of vision (15 letters) over the course of a year from 38% in the sham injection group to about 5.5%. The percentage of patients who experienced substantial improvement in vision (15 letters) was increased from 5% in the sham injection group to 24.8% and 33.8% in the 0.3 and 0.5 mg ranibizumab groups, respectively. A useful endpoint regarding function is the percentage of patients who achieve 20/40 vision, because that level of vision is sufficient for driving and reading reasonably sized print. There were no differences among the groups at baseline (about 15%), but at 1 year approximately 40% of patients in the ranibizumab groups were 20/40 compared to 11% in the sham injection group. Another useful endpoint regarding function is the percentage of patients 20/200 or worse which is legal blindness; approximately 12% of patients in the ranibizumab groups were 20/200 or worse compared to 43% in the sham injection group. Thus by multiple different assessments of visual function, patients treated with 0.3 or 0.5 mg of ranibizumab did substantially better than those given sham injections, and the results at 2 years were nearly identical to those at 1 year. Measurement of the area of CNV showed no change over the course of a year in patients treated with ranibizumab compared to an increase of 2 disc areas in the control group. This indicates that although the ranibizumab did not cause the neovascularization to regress, it completely stopped it from growing.
In the ANCHOR trial, patients with predominantly classic CNV due to AMD were randomized to receive monthly intraocular injections of 0.3 or 0.5 mg of ranibizumab plus sham PDT every 3 months or monthly sham injections plus PDT with verteporfin every 3 months.33 At 12 months, 5.7% of patients receiving 0.3 mg and 3.6% of patients receiving 0.5 mg of ranibizumab lost 15 or more letters of VA compared to 35.7% in the PDT group. The percentage of patients who gained 15 or more letters was 5.6% in the PDT group compared to 34.7% and 40.3% in the 0.3 mg and 0.5 mg ranibizumab groups, respectively. These results and other VA outcomes were remarkably similar to those seen in the MARINA trial and together they indicate that ranibizumab is the first agent able to cause significant visual improvement in a substantial number of patients with CNV due to AMD. Although the studies were not powered to distinguish differences between the 0.3- and 0.5-mg doses, there was some suggestion that the 0.5-mg dose was superior and there was no evidence of significant toxicity with either dose; this is likely to be the rationale for FDA approval of the 0.5-mg dose for treatment of choroidal neovascularization due to AMD. The results demonstrate that monthly injections of ranibizumab are far superior to PDT or pegaptanib and therefore ranibizumab is now standard care.
Both the ANCHOR and MARINA trials suggested that one of the major anatomic effects of ranibizumab is to reduce leakage from CNV. They also showed that ranibizumab stops or markedly slows growth of CNV, but does not induce its regression. Although the total area of CNV was essentially unchanged, ranibizumab caused a reduction in the area of classic CNV. Thus, in addition to reducing growth, ranibizumab may promote conversion of classic to occult CNV; this is intriguing because occult CNV is a more quiescent form of neovascularization that tends to be less aggressive.
It is important to know if monthly injections of ranibizumab are needed or whether equally good results are achievable with less frequent intraocular injections. This question was addressed in the PIER trial in which patients were randomized to receive sham injections or 3 monthly injections of 0.5 mg of ranibizumab followed by injections at 6 and 9 months, with the primary endpoint at 12 months. Patients treated with ranibizumab showed a mean improvement in VA of 4.8 letters at month 3, but at month 12 there was a mean reduction of 0.2 letters compared to a reduction of 16.3 letters in the sham injection group. Compared to the sham injection group, the ranibizumab group did not show a significant increase in the percentage of patients with improvement of 15 or more letters of VA. These results are disappointing and suggest that a fixed schedule of injections every 3 months is not a good strategy.
A variable injection schedule based on OCT findings was tested in the PrONTO study.34 This was an open-label study in which 40 patients with choroidal neovascularization due to AMD and central 1.0 mm retinal thickness ≥ 300 μm were given three monthly injections of 0.5 mg of ranibizumab and then additional injections only if one of the following criteria were met: (1) loss of 5 letters of VA in conjunction with fluid in the macula detected by OCT, (2) increase in central retinal thickness ≥ 100 μm, (3) new onset classic choroidal neovascularization, (4) new hemorrhage in the macula, or (5) OCT showing persistent macular fluid 1 month after injection of ranibizumab. These criteria may be overly stringent, and one could make a case for replacing them with a single criterion, evidence of subretinal or intraretinal fluid on OCT. However, despite the somewhat stringent criteria, the results at 1 year were quite good. The mean change in VA was +9.3 letters, 95% of patients lost ≤ 15 letters, and 35% gained ≥ 15 letters, all outcomes that are similar to those seen in the MARINA Trial. The mean number of injections was 5.6 compared to 13 in the MARINA Trial. Although a small open-label nonrandomized trial is not comparable to a large randomized clinical trial, there is at least a suggestion that monthly injections are not necessary for optimal management of all patients.
Bevacizumab for choroidal neovascularization due to AMD
After systemically administered bevacizumab was approved for colorectal cancer, it was studied in a series of 18 patients with CNV due to AMD who showed improvement in median VA of 14 letters and substantial reduction in foveal thickness over a period of 24 weeks.35 There was a significant elevation in blood pressure that required adjustment and/or institution of antihypertensive medicines in several patients. Case studies have suggested that intraocular injection of 1.25 mg of bevacizumab may provide benefit in patients with neovascular AMD36,37 has subsequently been shown to provide benefit. It is unclear how bevacizumab and ranibizumab compare when given by intraocular injection, but a clinical trial sponsored by the NIH that will test this issue will shortly get underway.
VEGF antagonists for CNV due to diseases other than AMD
At the same time that systemically administered bevacizumab was being tested in AMD, it was tested in two patients with CNV due to pathologic myopia, and there was dramatic improvement.38 This prompted an uncontrolled, open-label trial in which 10 patients with CNV (4 with pathologic myopia, 2 with ocular histoplasmosis, 2 with punctate inner choroidopathy, 1 with birdshot chorioretinopathy, and 1 with angioid streaks) were given up to 4 infusions of 5 mg/kg of bevacizumab.39 Infusions were well tolerated, and there were no ocular or systemic adverse events, including no significant elevation in blood pressure. At baseline, median VA was 25.5 letters read at 4 m (20/80) and median central retinal thickness was 346 μm. At the primary endpoint, 24 weeks, median VA was 48.5 letters (20/32), representing 4 lines of improvement from baseline, and median central retinal thickness was 248 μm representing a 72% reduction in excess central retinal thickness. All patients except one showed a reduction in area of CNV with a median reduction of 43%. Intraocular injections of bevacizumab has also been shown to cause improvements in patients with NV due to pathologic myopia40,41,42 and in a patient with CNV due to angioid streaks.43 Based on results with bevacizumab, it is likely that ranibizumab is likely to provide benefit in patients with CNV due to diseases other than AMD.
VEGF and diabetic macular edema (DME)
Diabetic retinopathy is the most prevalent cause of vision loss in working-age individuals in developed countries.44 Severe vision loss occurs due to traction retinal detachments that complicate retinal NV, but the most common cause of moderate vision loss is macular edema. Macular edema occurs from leakage of plasma into the central retina causing it to thicken because of excess interstitial fluid (for review see reference 45). The excess interstitial fluid is likely to disrupt ion fluxes and the thickening of the macula results in stretching and distortion of neurons. There is reversible reduction in visual acuity, but over time the perturbed neurons die resulting in permanent visual loss.
Retinal hypoxia has been implicated in the pathogenesis of DME.46 Hypoxia causes increased expression of VEGF, which in addition to stimulating NV, is a potent inducer of vascular permeability that has been shown to cause leakage from retinal vessels.47,48 Oral administration of a nonspecific VEGF receptor kinase inhibitor for 3 months significantly reduced excess foveal thickness measured by OCT in patients with DME that reverted to baseline when the drug was stopped.49 Recently, 10 of 10 patients with DME treated with intraocular injections of 0.5 mg of ranibizumab showed improvement in visual acuity and foveal thickness measured by OCT.50 Mean values at baseline were 503 μm for foveal thickness and 28.1 letters read on an Early Treatment Diabetic Retinopathy Study (ETDRS) visual acuity chart (20/80). At 7 months (1 month after the fifth injection), the mean foveal thickness was 257 μm, a reduction of 246 μm (85% of the excess foveal thickness present at baseline), and mean visual acuity was 40.4 letters (20/40), an improvement of 12.3 letters. This suggests that VEGF is an important therapeutic target for DME and randomized, controlled trials testing ranibizumab and other VEGF antagonists are underway.
Ranibizimab for macular edema due to vein occlusions
Macular edema is a major cause of visual loss in patients with branch or central retinal vein occlusions. In these disease processes, just as in diabetic retinopathy, there is retinal ischemia and VEGF levels are elevated. Therefore, the effect of intraocular injections of ranibizumab is being tested in patients with macular edema due to vein occlusions and preliminary results suggest a substantial beneficial effect.51
Other VEGF antagonists
Another VEGF antagonist that is in phase III clinical trials is VEGF Trap. VEGF Trap is a 115-kDa recombinant fusion protein consisting of the VEGF binding domains of human VEGF receptors 1 and 2 fused to the Fc domain of human IgG1.52 VEGF Trap strongly inhibits choroidal neovascularization and VEGF-induced breakdown of the blood–retinal barrier.53 Compared to ranibizumab, animal studies have shown that VEGF Trap has a longer half-life in the eye after intraocular injection, a higher binding affinity to VEGF-A, and it binds other members of the VEGF family including Placental Growth Factors 1 and 2, which have been shown to contribute to retinal neovascularization and excessive vascular permeability.27,52,54 Systemic administration of VEGF Trap reduces leakage, subretinal fluid, and ERT in patients with choroidal neovascularization (CNV) secondary to age-related macular degeneration (AMD), but higher doses induced hypertension and proteinuria.38 In order to avoid systemic toxicity, VEGF Trap has been formulated for intravitreal injections; in early-phase trials, intravitreal injections of VEGF Trap have caused substantial reductions in excess foveal thickness in patients with CNV due to AMD (reported at the Association for Research in Vision and Ophthalmology Annual Meeting, May 2007, Fort Lauderdale, FL, abstract #2868). The ongoing phase III trials are testing for noninferiority of 0.5 and 2 mg of monthly injections of VEGF Trap compared to monthly injections of 0.5 mg of ranibizumab. They are also examining the effect of injections of 2 mg of VEGF Trap given every 8 weeks.
VEGF stimulates endothelial cells through binding to transmembrane tyrosine kinase receptors. Inhibition of the kinase activity of VEGF receptors is a particularly efficient way to block VEGF signaling and because this can be done with small molecules rather than with antibodies, more flexibility is provided with regard to delivery. Because there is substantial homology among certain tyrosine kinase receptors, specificity is relative. Although in general a high level of specificity for the target of interest is desirable, because off-target effects could potentially compromise safety, there is a silver lining to the lack of specificity cloud with regard to kinase inhibitors. That is because receptor families with greatest homology to VEGF receptors also play a role in neovascularization. For instance, there is substantial homology between VEGF and platelet-derived growth factor (PDGF) receptors, and it is common for kinase inhibitors to block both types of receptors. Production of high levels of either PDGF-BB or VEGF in the retina causes severe NV and retinal detachment.55,56 PDGF-BB is a survival factor for pericytes, one of the major sources of VEGF and other survival factors for endothelial cells. As a result, combined inhibition of VEGF and PDGF receptors has greater effects on ocular angiogenesis than either alone; in fact, kinase inhibitors that block both VEGF and PDGF receptors are particularly efficacious inhibitors of retinal or choroidal NV.15,16,20 TG100572 is an agent that inhibits VEGF and PDGF receptor kinases as well as the intracellular signaling molecule Src kinase, which mediates angiogenic and permeability effects of VEGF. TG100801 is an inactive prodrug that generates TG100572. Topical administration of TG100801 to the cornea results in substantial levels of TG100572 in the retina and choroid and suppresses ocular NV and excessive vascular permeability.57