Using the Past to Inform the Future: Anti-VEGF Therapy as a Road Map to Develop Novel Therapies for Diabetic Retinopathy

Clinical success of anti-VEGF therapy is based on basic scientific research into the mechanisms of angiogenesis, neovascularization, and vascular permeability leading to a broad consensus from the scientific community on the significance and requirement of this growth factor to these defined processes. Exploration in tumor biology led to the hypothesis that diffusible factors provided angiogenic and permeabilizing signals to the tumor vasculature. This led to the seminal hypothesis by Dr. Judah Folkman that inhibition of angiogenesis may be a strategy to halt tumor growth (8). Protein purification and molecular cloning allowed two groups to discover the potent angiogenic and permeabilizing factor: one coining the term vascular permeability factor and the other VEGF (9,10). A review of the biology of VEGF and its receptors on angiogenesis, proliferation, migration, and vascular permeability was performed by Chung and Ferrara (11). Here, we provide a retrospective analysis of the significance of the seminal findings in the development of anti-VEGF therapies and propose a model to apply to the newest set of preclinical targets for DR.

Clear and compelling genetic studies revealed that VEGF contributes a critical and essential role in vascular biology (rev. in 11). Genetic loss-of-function experiments demonstrated that developmental expression of VEGF is required for vasculogenesis and angiogenesis, as single Vegf allele inactivation resulted in embryonic lethality with deficient vascularization of several organs (12,13). Moreover, gene targeting via the Cre-loxP system and administration of a soluble VEGF receptor chimeric protein led to significant increases in mortality and impaired organ development; however, this critical requirement for VEGF waned by 4 weeks of animal maturation (14). This was the first evidence that VEGF function may be altered in the adult without the detrimental effects observed in development.

VEGF clearly contributes to vascular homeostasis, and further research confirmed that excess VEGF and aberrant VEGF signaling induces pathological angiogenesis and permeability. Pharmacological targeting of VEGF began in the cancer field prior to its application to ocular disease. The first preclinical evidence of anti-VEGF therapy was performed using targeted monoclonal antibody (mAb) technology that successfully prevented the growth of tumors in animal models (15). The results were verified by targeted deletion of the VEGF-A gene, which suppresses tumor angiogenesis in a well-established pancreatic islet carcinoma cancer model (16). Along with numerous other preclinical studies, these data suggest that pharmacological manipulation of VEGF may be effective for the treatment of malignancies, and the U.S. Food and Drug Administration (FDA) granted approval for anti-VEGF therapy for several cancer indications after several successful clinical trials. Nonetheless, there is concern that systemic anti-VEGF treatment may lead to an increased occurrence of cardiovascular side effects (17).

Years prior to the discovery of VEGF, the oncology and ophthalmology fields were united when Dr. Isaac Michaelson proposed a bold hypothesis in 1948 that a diffusible “Factor X” produced by the retina was responsible for retinal and iris neovascularization that occurred in proliferative DR (18). VEGF was later identified as a likely candidate for this retinal “Factor X” and helped to connect the underlying pathological angiogenic and permeability responses in numerous retinopathies (19). Analysis of human vitreous from patients with severe DR demonstrated that VEGF concentrations were increased in proliferative DR (20,21). Clearly, DR demonstrates both severe angiogenic and permeabilizing features closely related to VEGF’s biological function. Initial animal studies using intraocular VEGF injections and various VEGF inhibitors demonstrated the causal role of VEGF as a mediator of ischemia-induced intraocular neovascularization and retinal permeability (22–24).

These preclinical experiments led to several clinical trials evaluating the efficacy of anti-VEGF therapy in the eye and exemplify how evaluation of normal physiology and pathology may establish a clinical target. Genetic analysis of the physiologic role of VEGF and its receptor led to hypothesis-driven studies examining the role of the growth factor in models of retinal permeability and angiogenesis. Meanwhile, clinical studies revealed a close correlation of VEGF expression in proliferative DR. It is also noteworthy that the models used provided dramatic responses of ischemia-induced angiogenesis and permeability. To our knowledge, only one study demonstrated blockade of VEGF effects after 1-week induction of experimental diabetes and vascular permeability (25). Diabetic rodents likely reflect an early or background retinopathy with moderate permeability rather than the severe response in humans targeted by clinicians, which includes robust edema or neovascularization—both features lacking in the current rodent diabetic models.

Generating novel therapies for the indication to treat DR is extremely difficult owing to the long and incremental organ failure associated with diabetes. The first ocular indication for anti-VEGF therapy (FDA approved in 2006) was for neovascular age-related macular degeneration (wet-AMD). The pathogenesis of wet-AMD involves significant choroidal neovascularization and vascular permeability, and strong preclinical data demonstrate that VEGF mediates these effects. The Minimally Classic/Occult Trial of the Anti-VEGF Antibody Ranibizumab in the Treatment of Neovascular Age-Related Macular De-generation (MARINA) trial showed that ranibizumab treatment led to a 95% rate of visual stabilization, and nearly 40% of treated patients had a ≥15 letter gain of visual acuity at 12 months. Importantly, these benefits were associated with significant decreases in foveal center point thickness in the ranibizumab-treated patients (26). Additionally, the ANCHOR Study (Anti-VEGF Antibody for the Treatment of Predominantly Classic Choroidal Neovascularization in Age-related Macular Degeneration) demonstrated superior efficacy in treating choroidal neovascularization over photodynamic therapy with verteporfin (27,28). Notably, ocular anti-VEGF therapy was well tolerated with low risk of ocular adverse events and no systemic side effects as seen with intravenous administration for cancer treatment (29). FDA approval in 2010 for macular edema after retinal vein occlusion was granted after the Branch Retinal Vein Occlusion (BRAVO) and Central Retinal Vein Occlusion (CRUISE) phase III trials that demonstrated that 50% of patients gained ≥15 letters in visual acuity after treatment with ranibizumab (30,31).

Initial studies in persons with diabetes began in 2005 with pegaptanib, an aptamer targeting VEGF. This phase II trial demonstrated greater visual acuity improvements after treatment with pegaptanib compared with sham injection patients (32). After this study, numerous clinical trials have demonstrated success of anti-VEGF therapies including ranibizumab, bevacizumab, and afliberacept, which are extensively reviewed elsewhere (33). Subsequent studies examining dosing regimens and dose responses all arrived at the same conclusion; namely, multiple independent pharmacological agents targeting VEGF exhibit robust clinical efficacy. After the RIDE/RISE results, ranibizumab gained approval on 10 August 2012 for DME after evidence of significant visual acuity gains of ≥15 letters for 35–50% of the patients treated with ranibizumab, which correlated with dramatic improvements in retinal edema as measured by optical coherence tomography (7). Although some patients fail to respond to anti-VEGF therapy, this medical treatment has revolutionized ophthalmic care. Clearly, increased vascular permeability and retinal edema contribute to the pathology of DR and loss of visual acuity (34), and various studies targeting VEGF with chemically diverse therapies demonstrate that blocking excess VEGF signaling prevents loss of vision in DR.

While empiricism has gained great advances in therapies with limited mechanistic understanding (e.g., aspirin, penicillin), we propose that the approach used in understanding the role of VEGF in vascular biology provides a learning model for future drug development for DR. An analysis of the development of anti-VEGF therapies may be distilled to a set of principles that may expedite the validation and development of preclinical drug targets for DR prior to large clinical trials in humans. This model has five principles:

• 1) Physiology: knowledge of the target’s role in normal physiology from biochemical and genetic manipulation.

• 2) Pathology: demonstration of the target’s causal link to pathological features as evident from biochemical and genetic gain-of-function or loss-of-function experiments.

• 3) Relevance: evidence demonstrating a correlation with disease progression or severity in humans.

• 4) Reversibility: pharmacological agents targeting the factor must be able to elicit a biologically significant response in vivo in preclinical screens that are clinically meaningful.

• 5) Repeatability: structurally distinct probes that regulate the target, or another target on the same pathway, demonstrate a similar effectiveness.

VEGF clearly obtained broad scientific consensus as a valid target for vascular angiogenesis and vascular permeability. VEGF was shown to contribute to pathological vascular permeability and retinal edema as well as angiogenesis in preclinical studies, while clinical studies demonstrated a clear correlation with disease. Furthermore, multiple pharmacological agents successfully targeting VEGF showed dramatic improvements in vascular permeability and retinal edema in a range of disease models and clinical studies. Approaches independent of this model have demonstrated success in identifying compounds with great potential for clinical efficacy such as the discovery of the use of fenofibrate for DR (35). Indeed, future studies may provide important mechanistic understanding of the use of these compounds but will not be discussed here. Rather, the model provided by anti-VEGF therapies will be discussed as providing a framework for preclinical research that may improve drug development success and outcomes.

Although significant success is apparent with anti-VEGF therapy, a number of limitations exist including the need for repeat intraocular injections, the potential for disease rebound upon discontinuing treatment, and the fact that not all patients respond to the treatment, collectively lead to the justification for the investigation of novel therapies for DR. In addition to VEGF, numerous factors have been identified as potential candidates mediating the vascular dysfunction observed in DR. The pathology of the diabetic retina is complex and includes regions lacking vascular perfusion and degeneration as well as regions of vascular growth and responsiveness to VEGF. Figure 1 reveals the cells associated with the retinal capillary and the proposed factors that may contribute to DR. The first category includes targets that are downstream of VEGF signaling. Factors in this category may provide additive benefits to current anti-VEGF strategies through improved pharmacokinetic or drug delivery options such as an orally available alternative to repeat intraocular injections of anti-VEGF mAbs. The second category includes novel proangiogenic factors distinct from VEGF with unique signaling mechanisms that contribute to blood-retinal barrier dysfunction and angiogenesis. Finally, extracellular inflammatory factors that contribute to vascular permeability or vascular degeneration are discussed. Importantly, drugs for DR may be considered for systemic or local delivery. Local delivery provides a number of potential advantages including the potential for reduced side effects and prolonged drug half-life with the current disadvantage of the need for intraocular injection. The route of drug delivery may have a profound impact on the success of drug development for DR. We apologize in advance to our colleagues with exciting new targets not discussed, as limited examples will be highlighted (Table 1).

VEGF requires classical protein kinase C (PKC) and, specifically, PKCβ for signal transduction in vascular endothelium. Both diabetes and VEGF increase retinal PKCβ activity, and an orally active PKCβ-specific inhibitor, ruboxistaurin (RBX), can block VEGF-induced permeability and angiogenesis in rodents (36). PKCβ contributes a critical role in mediating proliferation through retinoblastoma-associated protein phosphorylation (37) and permeability through phosphorylation-induced endocytosis of the tight junction protein occludin (38). Clinical trials to determine the effect of RBX on visual loss in patients with DR demonstrated that the drug was well tolerated and reduced the progression of macular edema, the need for laser, and the risk of sustained moderate visual loss by ~50% (39). However, currently, RBX has not received FDA approval for treatment of DR. PKCβ has been one of the most well characterized DR therapeutic targets since VEGF, and further clinical efficacy studies with appropriately targeted drugs may be warranted.

Recent evidence suggests that atypical PKC (aPKC) isoforms also contribute to VEGF-induced permeability. aPKC contributes to normal developmental physiology by mediating various processes including cellular polarity and barrier establishment (40). Targeting aPKC isoforms both genetically and pharmacologically with siRNA and novel small-molecule inhibitors demonstrates a requirement for aPKC activity in VEGF-induced retinal endothelial permeability. These small molecule inhibitors also prevent VEGF-induced permeability in the rodent retina (41). aPKC may provide an ideal target, as other inflammatory factors such as thrombin, tumor necrosis factor (TNF), and chemokine ligand 2 (CCL2), discussed in more detail below, also signal through aPKC to mediate their effects on endothelial permeability (42–44). Although additional genetic data are needed to validate this target, aPKC represents a common signaling node required for multiple permeabilizing factors and may provide a target that offers improved efficacy over current monotherapies. Signal transduction pathways involved in permeability are summarized in Fig. 2.

From the success of anti-VEGF therapy, other factors that regulate angiogenesis and permeability have been examined as potential therapeutic targets to treat DR. Angiopoietins (Ang) have a clearly defined role in vascular development and physiology. The angiopoietins consists of several ligands and receptors, but the focus has been primarily on Ang1 and Ang2 and the receptor Tie2. Importantly, Ang1 and Ang2 have antagonistic roles upon receptor binding, as Ang1 is an agonist for Tie2, while Ang2 acts as a competitive antagonist of the Tie2 receptor (45). The signaling axis of Ang-Tie2 is essential for angiogenesis as Ang2 promotes vessel destabilization and initiates angiogenesis. Alternatively, Ang1 stabilizes vessels following angiogenesis, and has been implicated in pathological angiogenesis (45). Importantly, double blockade of VEGF and Ang2 with a chimeric decoy receptor has added benefit over VEGF and Ang2 blockade alone in inhibition of tumor angiogenesis and vascular leakage (46). Clinically, Ang2 concentrations are elevated in the vitreous of patients with PDR (47), while neutralizing antibody that blocks Ang2 prevents ocular angiogenesis in animal models of oxygen-induced retinopathy (48). In addition, Ang1 is a retinal vasoprotectant and exhibits anti-inflammatory activity in animal models of diabetes (49). Strong preclinical genetic and biochemical data supporting a role for Ang-Tie2 axis in physiological and pathological angiogenesis and permeability support targeting this pathway to promote vascular stabilization and reduce vascular leak. Evidence for changes in this pathway in human disease progression would support further drug development. The human protein tyrosine phosphatase β (HPTPβ) has emerged as a regulator of Tie2 signaling and provides a potential point of therapeutic intervention (50).

Platelet-derived growth factor (PDGF) has been suggested as a potent proangiogenic molecule, which is also important for the establishment of the blood-retinal barrier. The literature regarding PDGF’s role in retinal angiogenesis and DR pathogenesis is complex and even contradictory. Whole-body genetic loss-of-function studies demonstrate that PDGF and PDGFRβ are required for vascular development (51). Importantly, for blood-retinal barrier development, PDGF deficiency leads to pericyte loss, microaneurysm formation, and blood-retinal barrier dysfunction mimicking the DR phenotype (52). This work suggests that reduced pericyte density is sufficient to cause a vascular retinopathy phenotype in mice. This observation was mechanistically verified in a diabetic mouse model in which hyperglycemia induces loss of PDGF signaling in pericytes through the activation of PKCδ/Src homology phosphatase-1, leading to pericyte apoptosis and vascular dysfunction (53). These data suggest that PDGF supplementation may be protective in DR. Conversely, studies have demonstrated elevated vitreous concentration of PDGFAB in diabetic patients (54). Indeed, in mouse retina, excessive PDGFA from neuronal cells leads to proliferative disease (55). PDGF-specific antibody can enhance anti-VEGF therapy in models of ocular neovascularization and cause vessel regression, as some neovessels become refractory to sustained VEGF deprivation (56). Together, these studies suggest that PDGF is a well-validated preclinical target according to our model; however, additional studies are needed to determine how to effectively modulate the PDGF axis to prevent the pathogenesis of DR, as both supplementation and inhibition of PDGF may improve the DR phenotype depending on the specific vascular microenvironment.

Several laboratories have suggested that the Wnt signaling axis contributes a fundamental role in both normal and disease physiology and therefore represents an exciting pharmacological target for diabetic eye disease. Wnts are a large family of secreted proteins that have a wide range of functions including the regulation of endothelial gene expression and angiogenesis. Upon Wnt ligand binding to the Frizzled receptor and recruitment of LDL receptor–related peptide (Lrp)5/6, β-catenin is stabilized and trans-locates to the nucleus, where it binds TCF/Lef transcription factors and induces specific gene transcription. Gain- and loss-of-function experiments clearly demonstrate that normal vascular development requires Wnt signaling, specifically in the retina, where several targeted mutations of this signaling axis have drastic retinal vascular defects (57). Wnt also contributes to the pathophysiology of ischemic retinopathies. Deletion of Lrp5 and downstream signaling molecule dishevelled2 decreases the formation of pathological neovascularization in the oxygen-induced retinopathy model (58). Several endogenous antiangiogenic inhibitors of Wnt signaling have become potential therapies to inhibit aberrant Wnt signaling. SERPINA3K binds to Lrp6 and antagonizes Wnt signaling preventing the vascular permeability and inflammation seen in diabetic rats (59). Inhibition with an mAb to Lrp6 also prevents vascular leakage and inflammation of the retina of diabetic animals (60). Evidence suggests that pigment epithelium–derived factor (PEDF) acts as an antiangiogenic factor in the retina of rodents (61). The effect of PEDF may be attributed to antagonism of Wnt signaling through binding of Lrp6 preventing Lrp6–Frizzled receptor dimerization and signaling (62). Restoring proper balance of Wnt signaling represents an exciting novel target in normalizing retinal vessels in diabetes.

Other proangiogenic growth factors such as growth hormone (GH) and IGF-1 are linked to DR pathogenesis. Clinical trials suggest a delay in time to progression to DR and trends to improve visual acuity with somatostatin or the analog octreotide, presumably by inhibiting GH secretion (63). Conversely, recent work suggests a protective role for ischemia-induced neovacularization by IGF-1–binding protein 3 (64). Additional studies are needed to identify the pathophysiological role of IGF-1/GH signaling and the relationship to disease progression. In addition to GH/IGF-1, erythropoietin (Epo) is elevated in the vitreous of patients with proliferative DR (47). Epo enhances pathological retinal angiogenesis in mice in the oxygen-induced retinopathy model (65), while small interfering RNA–mediated knockdown of Epo suppresses retinal neovascularization (66), suggesting a pathogenic nature for Epo. Conversely, intravitreal injection of Epo inhibits retinal barrier breakdown in diabetic animals (67). Clearly, there is a lack of consensus on the pathogenicity of Epo, and further preclinical studies are needed to validate these targets.

In addition to proangiogenic signals, inflammatory cytokines represent novel targets for DR. Experimental evidence indicates that cytokines, in addition to VEGF, also contribute to vascular permeability in DR. A series of elegant experiments demonstrated that leukostasis occurs during streptozotocin-induced diabetes in mouse and rat models and that deletion of the gene for the adhesion protein ICAM or its leukocyte-binding partner CD18 ameliorated leukostasis (68). Furthermore, in an acute (1 week) model of diabetes, vascular permeability, leukostasis, CD18, and ICAM expression, as well as nuclear factor-κB activation, were all normalized by high-dose aspirin or the cyclooxygenase-2 inhibitor meloxicam and by a soluble TNF receptor/Fc hybrid (entanercept), suggesting that TNF and cyclooxygenase contribute to DR (69). A recent report provides insight into how VEGF and TNF may both contribute to different phases of vascular dysfunction in DR. Using mice with TNF gene deletion, the investigators demonstrate that diabetes-induced permeability was unaffected at 1 month; at 3 months, changes in permeability are partially reduced, while at 6 months the permeability defect is ameliorated (70). Furthermore, in humans elevated interleukin (IL)-1β and TNF levels have been associated with proliferative DR (71,72) and TNF with nonproliferative DR (73). In addition, anti-TNF therapies led to significant improvement in visual acuity compared with placebo-treated eyes in a controlled trial involving patients refractory to other treatments (74). ILs and other inflammatory factors may also contribute to DR; monocyte chemotactic protein-1 (MCP)-1 (or CCL2) along with IL-6 and IL-8 were found elevated in patients with retinopathies associated with inflammation including DR (75), and MCP-1 contributes important roles in leukocyte infiltration and vascular inflammation after retinal detachment (76). The contribution of inflammatory cytokines and chemokines to DR is complex and early in the discovery phase. What factors are relevant to human DR and at what phase of the disease process remain to be elucidated, but targeting inflammatory cytokines and chemokines offers an inviting option to control DR. However, the effectiveness of monotherapies targeting specific inflammatory cytokines in the face of the wide array of factors changed in DR remains to be determined.

Aberrant protease signaling is a relatively new area of investigation for DR. Some of the factors recently identified are actually well-known molecules in human physiology. One promising target is the kinin-kallikrein system, which contributes an essential role in inflammation, blood pressure control, and pain through bradykinin activation. Importantly, complement component C1-inhibitor is an important physiological inhibitor of plasma kallikrein. Deficiency of C1-inhibitor leads to plasma kallikrein and bradykinin activation and robust vascular permeability responses (77). Gao et al. (78) detected carbonic anhydrase (CA-1) and multiple components of the complement and kinin-kallikrein system in the vitreous of patients with DR. Importantly, carbonic anhydrase leads to alkalization of the vitreous initiating bradykinin-kallikrein-system and retinal permeability that may be prevented by a systemically administered PK inhibitor (79). The bradykinin-kallikrein system fits our model of a well-validated target linked to human disease, as prior literature has implicated this system in both normal and disease physiology, and multiple pharmacological agents effectively reverse aspects of DR in relevant animal models. Current clinical trials are underway to determine efficacy of kallikrien inhibition in humans to treat DME and vascular hemorrhage.

In this Perspective, a model for preclinical target identification and drug development is proposed to validate clinical trials based on the discovery and success of anti-VEGF therapy for diabetic eye disease. This model encompasses five principles, amassed by extensive preclinical research and supported by numerous independent laboratories that collectively provide compelling evidence for a drug target in human disease. This model aims to help guide preclinical experiments for potential drug targets and reduce the failure rate of agents entering clinical trials. Knowingly, many of the principles discussed in this Perspective have been integral to drug development for some time, but reviewing how these approaches have aided in the success of the development of the anti-VEGF therapies may provide a blueprint for the development of novel approaches to treat DR. Indeed, there are different approaches that may lead to validation of a preclinical target; however, in our perspective, the elucidation of physiologically relevant factors and defining their contribution to the pathogenesis of DR are clearly worth the efforts.

The authors are supported by funding from the National Institutes of Health (EY-012021), the JDRF, and Research to Prevent Blindness (all held by D.A.A.).

No potential conflicts of interest relevant to this article were reported.

P.M.T. and D.A.A. wrote, reviewed, and edited the manuscript. D.A.A. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank Drs. Steve Abcouwer (University of Michigan) and Elia Duh (Johns Hopkins University) for their comments and views.