Regulation of signaling events involved in the pathophysiology of neovascular AMD.

Neovascular age-related macular degeneration (AMD) is a complex disease in which an individual's genetic predisposition is affected by aging and environmental stresses, which trigger signaling pathways involving inflammation, oxidation, and/or angiogenesis in the RPE cells and choroidal endothelial cells (CECs), to lead to vision loss from choroidal neovascularization. Antiangiogenic therapies have greatly improved clinical outcomes in the last decade; however, vision improves in less than half of patients treated for neovascular AMD, and treatments remain inadequate for atrophic AMD. Many studies focus on genetic predisposition or the association of outcomes in trials of human neovascular AMD but are unable to evaluate the effects between different cell types involved in AMD and the signaling events that take place to cause pathologic biologic events. This manuscript complements other reviews in that it describes what is known generally in human AMD studies and clinical trials testing methods to inhibit vascular endothelial growth factor (VEGF inhibitors) and presents pathologic signaling events that develop in two important cell types, the RPE cells and the CECs, when stimulated by stresses or placed into conditions similar to what is currently understood to occur in neovascular AMD. This manuscript complements other reviews by discussing signaling events that are activated by cell-cell or cell-matrix interactions. These considerations are particularly important when considering growth factors, such as VEGF, which are important in physiologic and pathologic processes, or GTPases that are present but active only if GTP bound. In either case, it is essential to understand the role of signaling activation to distinguish what is pathologic from what is physiologic. Particularly important is the essential role of activated Rac1 in CEC transmigration of the RPE monolayer, an important step in blindness associated with neovascular AMD. Other concepts discussed include the importance of feed-forward loops that overwhelm mechanisms that seek to restore homeostasis in cells and the importance of regulating, instead of abolishing, signaling events in a chronic, complex disease, such as neovascular AMD. These concepts are important as we move to the next stages in developing treatments for neovascular AMD. A novel therapeutic strategy that will be discussed is activating an isoform of the GTPase, Rap1, which can regulate downstream signaling and a pathologic feed-forward loop leading to Rac1 activation and migration of CECs.

Introduction

Concepts and importance of causal analysis with limitations of models of human disease as related to AMD

Age-related macular degeneration (AMD) is a leading cause of blindness in the elderly throughout the world [1]. AMD has historically been characterized as “dry” or “wet” AMD. Dry AMD makes up approximately 90% of all forms of AMD identified and includes clinical/pathologic features of pigmentary changes, basal laminar and linear deposits, drusen, and the advanced form of dry AMD, geographic atrophy [2]. Wet AMD is the less common form but accounts for most of the cases of legal blindness that occur from choroidal neovascularization (CNV), sometimes polypoidal choroidal vasculopathy [3] (although there is controversy that this may represent a different pathophysiology) [4], retinal angiomatous proliferation [5], or retinal pigment epithelial detachment [6]. The current nomenclature of AMD has evolved somewhat. Now, the appearance of drusen and pigmentary changes without neovascularization or geographic atrophy is sometimes referred to as “early” AMD, and wet, or neovascular AMD, and geographic atrophy fit into the category of “advanced” AMD [7]. This change in nomenclature occurred partly because early AMD can lead to either advanced form (geographic atrophy or neovascular AMD), and both advanced forms can coexist in the same eye [8,9].

Early AMD often is asymptomatic and includes the clinically identified features of drusen and pigmentary changes. From histopathologic and proteomic analyses, there is increased deposition of many compounds (for example, enzymes (TIMP3 and SerpinA3), vitronectin [10], bisretinoid fluorophores, [11] complement, [12, 13] oxidized proteins [14], lipoprotein derived debris [15], cholesterol esters, 7-ketocholesterol [16], advanced glycation end products [17], and linoleic hydroperoxide [18]) beneath the retinal pigment epithelial (RPE) cells and within Bruch’s membrane. Recognition of drusen and RPE defects, believed to represent stressed and/or dying RPE cells, can be accomplished with slit-lamp biomicroscopy on clinical examination, but the thickened extracellular matrix that makes up basal linear or laminar deposits is difficult to detect even on retinal imaging studies, such as optical coherence tomography (OCT), infrared imaging, and fluorescein angiography (FA) [19].

Advanced forms of AMD include atrophic AMD and neovascular AMD. Both forms are often symptomatic with loss of contrast sensitivity, the presence of scotomata or blind spots, and distortion, for example. In the advanced dry form of AMD, i.e., atrophic AMD, or geography atrophy, there is atrophy of the RPE cells and choriocapillaris with later photoreceptor loss [20,21]. In neovascular AMD, endothelial cells from the choriocapillaris of the choroid migrate to and across the RPE monolayer and into the sensory retina. The choroidal endothelial cells (CECs) proliferate and develop into CNV at any location between the choriocapillaris and neural retina. Neovascular AMD is associated with the loss of choriocapillaris that is hypothesized to create a hypoxic stimulus for the overlying RPE cells [21] and initiate the development of later CNV. CNV that may remain beneath the RPE monolayer but does not invade the photoreceptors and neural retina is known as occult or type 1 CNV, whereas CNV that proliferate within the sensory retina is termed classic or type 2 CNV [22,23]. Other forms of neovascular AMD include fluid beneath RPE cells, called RPE detachments, or vascular lesions arising from the retinal vasculature in the deep retina, currently known as retinal angiomatous proliferation (RAP) [19,24], or type 3 neovascularization. Historically, CNV has been classified by its location and associated fluid determined with OCT and leakage on FA. With the advent of OCT angiography (OCTA), characterization of CNV based on location and flow may be possible [25] and provide information about how active or aggressive a CNV lesion is.

The most important risk factor of AMD is advanced age [26], and some studies suggest that the longer humans live, the greater the likelihood of developing advanced AMD. Increasingly, genetic variants have been found in association with increased risk of AMD, suggesting that these variants may increase the predisposition to advanced AMD [27]. Several genetic variants have been identified, including on chromosome 10, 10q26, ARMS2 (Gene ID: 387715; OMIM number: 611313)/HTRA1 (Gene ID: 5654; OMIM number: 602194), and those affecting factors in the complement system. The initial reports of the Y402H variant of complement factor H [13,28-30] led to several hypotheses surrounding the complement system in AMD, but recent reports suggest more complex and previously unsuspected mechanisms involving lipoprotein turnover in Bruch’s membrane and will be discussed in the following section [28]. Additional studies reported rare variant alleles with high impact [29]. In addition to the role genetics plays in AMD risk, environmental and/or external stresses further increase the risk of AMD in individual patients [30]. Despite greater understanding of gene and environment associations with increased risk of AMD, the pathophysiologic mechanistic links are still incompletely understood in AMD.

Environmental and external stresses most known to affect the risk and progression of AMD include smoking, obesity, and reduced dietary antioxidants [27]. Since causal analyses are difficult in human studies, experimental models often use mice modified to express or knock out genetic variants, and that are exposed to external stresses known to increase AMD risk, such as smoking [31,32] or fat-feeding [28,33], to seek causal roles. Studies are fraught with the lack of cigarette a robust animal model of AMD [33]. Rodents lack maculae, and drusen or changes in the RPE monolayer and extracellular matrix rarely occur in mice until after 1 or 2 years of age. Even though rodents do not have maculae, these animals are useful in various models to address specific hypotheses. The laser-induced CNV model in mice is a robust model of injury and inflammatory-induced angiogenesis and has similarities to human neovascular AMD [34]. Growth factors, particularly vascular endothelial growth factor (VEGF) and platelet-derived growth factor (PDGF), are involved, as well as inflammatory and oxidative compounds [35,36]. The laser-induced CNV model also involves wound-healing, which is a recognized process involved in human neovascular AMD [37]. The model has been criticized for being acute and short-term, because AMD is a chronic condition, but the short-term nature of the model provides the ability to address focused hypotheses through the use of genetic models and external stresses. Human studies are essential but provide limited mechanistic insight into complex events in AMD. Even exploratory studies of genetic variants, mRNA sequencing, and proteomic associations may not provide causal information of biologic events in AMD, because they do not distinguish whether the gene, mRNA, or protein is expressed and activated to affect a signaling pathway or what cells are involved in the pathophysiology. Therefore, it is important to integrate data from mechanistic studies, often using animal models and cell culture, with knowledge gained from human tissue and experiments involving human cells, and from genetic, transcriptional and proteomic studies to piece together hypotheses that related causal events in AMD.

Therefore, to study the mechanisms that surround AMD pathophysiology (Figure 1), it is helpful to interpret data from several different types of experimental designs, including analysis of human clinical and pathologic specimens [10,12,20,38,39], imaging and genetic studies to develop hypotheses related to biologic events in AMD followed by a combination of culture techniques, including relevant coculture models, and genetically modified mice exposed to external stresses to model outcomes representative of forms of AMD. As information regarding outcomes from interventions in clinical trials accrues, it is also helpful to use this information to refine hypotheses to explore, for example, why differences between treatment regimens and various antiangiogenic agents are seen. When interpreting human studies, it is important to take into account that clinical trials require a specific hypothesis and that recruitment criteria may be too narrow to address the hypothesis; thus, the outcomes are not generalizable to all groups of patients with a complex disease, such as AMD.

Figure 1

Pathologic events in CNV. A: In healthy cells, RPE cells have strong junctional complexes and form a monolayer on Bruch’s membrane separate from the choroidal endothelial cells (CECs). B: Optical coherence tomography (OCT) of a human eye without age-related macular degeneration (AMD). Note the normal architecture of retinal layers, lack of drusen, and good foveal contour. C: With increasing age and related to genetic predisposition, diet, smoking, oxidative stress, and inflammation, Bruch’s membrane and the RPE extracellular matrix change in appearance with the formation of drusen and in composition, with deposition of oxidized lipoprotein, debris, complement, and many other factors that incite inflammatory, oxidative, and angiogenic signaling. Microglia are activated and release cytokines (e.g., tumor necrosis factor alpha, TNF-α) that stimulate RPE cells to express vascular endothelial growth factor (VEGF). Concurrently, CECs activated to migrate toward the RPE monolayer and proliferate into type 1 choroidal neovascularization (CNV). RPE cell–CEC contact initiates events that lead to RPE barrier compromise, which permits cells and growth factors to move from basal to apical aspects. CECs are attracted to migrate across the RPE monolayer and into the sensory retina toward a VEGF gradient to proliferate into type 2 CNV. D: OCT of a human eye with neovascular AMD, showing loss of the architecture of the retina and the RPE monolayer, loss of the foveal depression, and cysts within the inner retinal layers.

Myriad processes involved in AMD

In addition to the roles of genetics and complement, abnormal cell events have been experimentally studied in relationship to AMD and include endoplasmic reticulum stress and the unfolded protein response [40], dysregulated autophagy [41], and effects through the mammalian target of rapamycin (mTOR)/Akt pathways [42], effects from oxidized compounds including oxidized lipids [43] and other forms of oxidative stress [44,45] on inflammatory cells and angiogenesis [46]. Inflammatory processes involve macrophages [47] and microglia [46] with release of inflammatory cytokines, including tumor necrosis factor alpha (TNF-α) [48], that trigger signaling in other cell types. Many ligand and receptor pairs and signaling compounds have also been implicated in neovascular AMD, including hepatocyte growth factor (HGF)/c-Met receptor, Wnt/LRP6 [49,50], platelet-derived growth factor (PDGF)/PDGFR [35,51], transforming growth factor β (TGF-β)/TGF-β receptor [37,52], fibroblast growth factors (FGFs), TNF-α/TNFR, and eotaxin/CCR3 [53], as examples [54].

The Wnt signaling pathway has been implicated in neovascular AMD and in models of laser-induced CNV [49,50]. Platelet-derived growth factor modulates pericytes and stabilizes developing blood vessels [35]. PDGF antagonists are currently being studied with anti-VEGF agents in neovascular AMD [51]. The TGF-β signaling pathway is involved in wound healing, which is implicated in neovascular AMD [37]. Inhibition of TGF-β has been found to maintain RPE cells in an epithelial phenotype and reduce the transition from epithelial to mesenchymal types of cells [52]. Thrombospondin-1 is a potent antiangiogenic and anti-inflammatory molecule that has been found to play an important role in CEC phenotype. TSP1 mice were susceptible to CNV following laser, in part through recruitment of macrophages. Choroidal endothelial cells from TSP1 mice had reduced capillary morphogenesis and greater generation of nitric oxide, suggesting a role of thrombospondin-1 in angiogenesis, inflammation, and oxidation [55].

Although not the focus of this review, the formation of the inflammasome has been implicated in atrophic AMD through many pathways [56,57], including related to increased expression of VEGF [58].

In many of these studies, the effects of ligand and receptor pairs have been based on studies in solo cultures, animal models, and increased expression of the factor or receptor in human specimens. In this review, we focus on the effect of cell–cell interactions on downstream signaling mechanisms, particularly related to VEGF, because anti-VEGF agents are the standard of care for CNV in human AMD. We also focus on two GTPases, Rac1 and Rap1, which can be involved in signaling cascades related to many of the factors described. GTPases are activated by guanine nucleotide exchange factors to be GTP-bound and are inactivated by GTPase activating proteins to be GDP-bound. In this way, GTPases act as biologic switches and represent a potential mechanism to regulate pathologic processes in complex, chronic diseases, like AMD.

Types 1 and 2 CNV will be discussed in neovascular AMD. With the availability of agents that interfere with the bioactivity of VEGF, improvement in visual acuity is now possible for this once blinding disease. However, anti-VEGF treatments are effective in only about 40% of patients, and there is a concern that broad inhibition of pathologic and beneficial effects of VEGF may reduce the health of the neuronal and glial retina [59,60]. In addition, VEGF is essential to the health of the choriocapillaris, photoreceptors, and RPE cells [61-63]. Several experimental studies in which VEGF released from RPE cells was reduced either by poisoning the RPE monolayer or by specific inhibition showed loss of the choriocapillaris [64,65]. Loss of RPE cells has been associated with geographic atrophy [20,21]. Therefore, recent studies have evaluated whether anti-VEGF agents affect the progression of geographic atrophy, in which RPE cells and choriocapillaris become atrophic.

Need to regulate but not abolish angiogenesis in human neovascular AMD

Unlike in cancer, chronic diseases may be managed best by regulating, instead of inhibiting, angiogenic signaling or biologic events. In neovascular AMD, strong inhibition of angiogenesis may affect physiologic choroidal vasculature. VEGF has beneficial effects on the choriocapillaris and on the neural retina, including in adults [61-63]. In addition, the course of the disease may vary in the individual patient, and there is no way to know how much neutralization of VEGF is ideal at a specific time for a given diseased eye. Therefore, it is unknown whether the dose needed changes over the course of the disease in the individual patient. Currently, we rely first on information from clinical trials that determines mean changes in parameters from many patients for guidance about patient management and ocular and systemic safety. Imaging of the individual patient’s macula is used to assess intraretinal and subretinal fluid and determine how frequently to treat a patient. There are now anti-VEGF agents with different potencies (e.g., ranibizumab and bevacizumab that bind VEGF compared to more potent aflibercept that can bind VEGF and another family member, placental growth factor). The clinician, therefore, can choose another agent if efficacy is not found following one treatment regimen. Other agents can be used in combination, including various steroid formulations [66], and photodynamic therapy [67].

Knowledge gained from clinical trials

Treatment is initiated with monthly injections of an anti-VEGF agent. The decision to treat monthly or at longer intervals is most often based on the visual acuity, the presence of subretinal or intraretinal fluid on OCT and the clinical appearance of features like hemorrhage or leakage determined with FA [68]. Different management strategies include the decision to treat if features of subretinal or intraretinal fluid are present or wait until a future visit, or by a “treat and extend” approach in which treatment is given even if no features are found on OCT, but the duration until the next visit is increased before retreatment. Several agents have been studied (ranibizumab, bevacizumab, and aflibercept), and all agents have been found to improve outcomes in neovascular AMD [69]. Pegaptanib, an aptamer to VEGF165 [70], was the first anti-VEGF agent and has less effect than the three other agents.

The effect of anti-VEGF agents on geographic atrophy is difficult to assess because the end stage of neovascular AMD is the formation of a fibrovascular scar with loss of the architecture of the RPE monolayer and the outer retina [22,23]. The Inhibit VEGF in Age-related Choroidal Neovascularisation (IVAN) study reported that new geographic atrophy was more common in eyes that received monthly anti-VEGF treatment compared to treatment based on features of disease regardless of the anti-VEGF agent used (ranibizumab versus bevacizumab) [71]. The Seven-Year Observational Update of Macular Degeneration Patients Post-MARINA/ANCHOR and HORIZON Trials (SEVEN-UP) analyzed 14 studies of subjects treated with ranibizumab followed for a mean of 7.3 years and found an increase in geographic atrophy after treatment with anti-VEGF [72]. The Comparative of AMD Treatment Trials (CATT) study, which reported equivalent outcomes following treatment with ranibizumab or bevacizumab for AMD [73], reviewed eyes treated with anti-VEGF agents and compared the outcomes of geographic atrophy in those that were successfully treated with anti-VEGF to eyes that did not respond and reported that geographic atrophy occurred in both groups with no difference in progression noted between treatment responses [74]. In a later study, the CATT group reported that ranibizumab use might be more often associated with geographic atrophy than bevacizumab [75]. These studies were retrospective evaluations of clinical trial outcomes, and limitations associated with retrospective studies must be considered. Additional study is needed to assess whether there is a difference in geographic atrophy from anti-VEGF use, and a prospective design may provide greater insight. Other safety concerns with anti-VEGF agents include increased intraocular pressure, but studies do not agree on the presence of this or on possible causes [76]. Thus, anti-VEGF agents reduce vision loss from AMD, but safety concerns exist. Other treatments are needed to improve efficacy and safety.

Transition from type 1 to type 2 CNV when CECs transmigrate the RPE monolayer

Neovascular AMD can be distinguished between occult or type 1 CNV, classic or type 2 CNV, and RAP, or type 3 neovascularization. Early studies found that occult CNV could be asymptomatic and associated with good vision [77], whereas vision loss often occurs in type 2 CNV [78]. Type 1 or occult CNV refers to CNV that remains beneath the RPE monolayer. In contrast, type 2 or classic CNV refers to CNV that has invaded the neural retina. Infrared imaging of the macula in humans can identify features of type 1 CNV even before “ill-defined leakage of undetermined origin that is associated with retinal thickening” manifests on FA [19]. Type 2 CNV presents as lacy hyperfluorescence on early frames of a fluorescein angiogram followed by late leakage of fluorescein dye. Now, OCTA can distinguish CNV lesions [79] by the location of the optical segment where particle motion is detected [19]. OCTA avoids dye injection but requires particle motion to detect CNV [80].

Type 1 CNV can be associated with stable visual acuity. It has been reported that more than 50% of the time, severe vision loss occurs when activated CECs migrate through Bruch’s membrane to contact the RPE cells and then migrate across the RPE monolayer into the sensory retina and become type 2 CNV [81]. Vision loss can also occur with RPE barrier compromise identified as leakage of fluorescein dye on angiography in type 1 CNV. These findings and reports suggest that CNV that remains contained beneath the sensory retina and without compromise of the RPE barrier integrity may not be detrimental. An extension is that CNV that is quiescent so as not to induce a fibrovascular response may potentially be beneficial by providing oxygenation and a means for removing accumulated debris. Retinal angiomatous proliferation vasculature for (RAP or type 3 neovascularization) can be more difficult to identify. It is seen on videoangiography with indocyanine green or fluorescein dyes as having a feeding retinal arteriole to an intraretinal angiomatous formation and draining venule [82,83]. OCT and OCTA can also detect RAP [84]. RAP was postulated to be associated with extensive outer retinal debris and degeneration [82] and has currently been shown to be associated with reticular pseudodrusen [85], which are associated with choroidal atrophy, thinning of the choroidal layers [86], and geographic atrophy [87]. RAP lesions have been shown to be associated with geographic atrophy in several clinical studies [84].

Importance of cell–cell interactions and signaling events in RPE cells and CECs

One way to study signaling events surrounding CEC activation and migration across the RPE monolayer is to use cocultures of human CECs and RPE cells. Solo cultures are limited, because cells change phenotype over time and do not always behave in culture as in their tissue microenvironment. Cocultures may be more representative by virtue of cell–cell interactions [35,88] and permit the ability to assess effects of one cell type on the other. It is important to use cells of low passage that retain cell markers and characteristics when in their tissue microenvironments.

To study the effects of CEC interactions with RPE cells, we developed a coculture model [89] that represents the step when CECs make contact with RPE cells after having migrated through Bruch’s membrane. Human adult RPE cells are grown on an inverted Transwell insert overnight until the RPE cells attach. The insert is then turned over, and the RPE cells are allowed to grow in culture until they form tight barrier properties determined by high transepithelial electrical resistance (>100 ohms/cm2). Human CECs are then grown within the insert. By varying the width of the pore size of the insert, migration of CECs into the RPE monolayer is controlled. In the coculture assay, CECs extend processes through pores to contact RPE cells but are restricted from migrating [89]. Contact between RPE cells and CECs triggers the activation of signaling pathways within CECs or RPE cells. The individual cell types can be distinguished with vital dyes and separated by scraping cells off either side of the insert. In addition, it is possible to isolate CECs by using CD31 coated-magnetic beads to attract CECs or by flow cytometry. Activation of signaling pathways in each cell type can be determined. This model allows one to compare the role of contact between RPE cells and CECs to the effects of coculture without contact (i.e., non-contacting coculture), in which RPE cells and CECs are grown separately without contact but share the same media. In each case, the conditioned media or each cell type, RPE cells or CECs, can be analyzed. The non-contacting coculture may represent the physiologic situation between RPE cells and CECs, in which semiporous Bruch’s membrane lies between RPE cells and CECs. However, neither coculture model insert aims to represent Bruch’s membrane. Once activated signaling effectors are identified in either cell type, mutations or shRNAs can be introduced into the individual cell type, and the effect on CEC transmigration of the RPE monolayer is determined in the transmigration assay. The transmigration assay has an insert with wider pores that permit CECs to migrate across the RPE monolayer. In the transmigration assay, CECs, vitally labeled to distinguish them from RPE cells, are counted.

Using the coculture model, there was an increase in the fold expression of VEGF189, a cell-associated splice variant of VEGFA, in RPE cells grown in contacting coculture with CECs. There was little change in the expression level of other splice variants of VEGF in the contacting coculture or of any splice variant of VEGF in RPE cells grown in non-contacting or solo culture. There was also no significant change in VEGF splice variant expression level in CECs grown in solo or either coculture condition [90]. However, increased age of donor eyes or exposure to hydrogen peroxide caused VEGF splice variant expression levels to be increased in RPE cells. In addition, stresses that increased VEGF189 in RPE cells also increased phosphorylation of VEGFR2, but not of VEGFR1, in cocultured CECs. Knockdown of VEGF189 in human RPE cells reduced phosphorylation of VEGFR2 in CECs and inhibited, but did not abolish, CEC transmigration across the RPE monolayer by approximately 40%, similar to reported outcomes in human clinical studies with anti-VEGF agents [52]. CECs grown in contact with RPE cells had activation of the rhoGTPase, Rac1. In contrast to VEGF189 knockdown, inhibition of Rac1 by expression of the Rac binding domain of Rac1 effector, POSH, abolished the increased CEC transmigration across the RPE monolayer [91]. These data provide strong evidence that Rac1 activation is necessary for CEC transmigration of the RPE monolayer.

Rac1 GTPase is a common effector of several signaling events. In addition to the role of VEGF-induced VEGFR2 activation, activation of Rac1 has also been proposed through the guanine nucleotide exchange factor (GEF), VaV2 [92]. We found activation of Rac1 in CECs stimulated with the chemokine/receptor, CCL11/CCR3 [93], or by activation of phospho-inositol 3 kinase (PI-3 kinase) [91]. PI-3 kinase was either downstream or parallel to VEGFR2-induced Rac1 activation. Others have reported that thrombospondin-1 regulation is affected by activation of Rac1 [94]. Inflammatory cytokines, such as interleukin-17, induce angiogenesis in CECs through activation of Rac1 [95]. Thus, Rac1, an essential component in CEC transmigration and activation, is downstream of multiple different pathways, with some evidence of crosstalk (Figure 2) and might be an effective target for neovascular AMD. However, knockout of Rac1 is lethal, and Rac1 inhibitors have been studied in cancer and angiogenesis but have not proven as effective or safe [53,54], as hoped [96]. Therefore, other ways to inhibit Rac1-mediated CEC transmigration may be important for neovascular AMD. It is also important to recognize the need to target a pathway at several junctures to reduce the effect of crosstalk with other signaling pathways, and this strategy may be safer and more effective than attempts to abolish upstream signaling.

Figure 2

Signaling pathways in a feed-forward loop involve inflammation, oxidative stress, and angiogenesis. Tumor necrosis factor alpha (TNF-α) released by activated microglia causes RPE cells to overexpress vascular endothelial growth factor (VEGF). VEGF attracts and activates choroidal endothelial cells (CECs) through activation of Rac1, which aggregates with other subunits to activate NADPH oxidase. NADPH oxidase–generated reactive oxygen species (ROS) trigger Rac1 activation through nuclear factor-kappa B (NF-ҡB) signaling, which can further activate NADPH oxidase, thus setting up a feed-forward loop. Activated CECs migrate and proliferate to form choroidal neovascularization (CNV). Activated Rap1a safely inhibits CNV by reducing NADPH oxidase–generated ROS. Permission was granted from Elsevier to use parts of Figure 18 from Prog Retin Eye Res. 2008 Jul;27(4):331-71, namely the OCT images used in panel B and D.

Need for a feed-forward loop to overwhelm homeostasis

Effects found in experimental models may not translate to human neovascular AMD for several reasons: There are species differences; limitations of models, as discussed earlier; difficulty in translating the effects of nutrition and activities of daily life to the cell microenvironment; and the possibility that experimental outcomes in models can be overcome in the human body through mechanisms to restore homeostasis. Generally, compounds that broadly inhibit a pathway (e.g., anti-VEGF agents or corticosteroids) and affect multiple downstream pathways may be effective but are also likely to have more side effects than a more targeted downstream inhibitor. However, the safer targeted inhibitor may not be as effective if upstream signaling pathways have crosstalk with other signaling pathways involved in the pathophysiology. In cancer treatment, broad therapy may be necessary to increase a patient’s survival against death, and the time for intervention is short. However, in chronic diseases, regulation instead of inhibition, is important, and many stresses that induce signaling must also be considered. It is important to address feed-forward loops that occur and multiple pathways involved based on also crosstalk between pathways.

Oxidation, inflammation, and angiogenesis in CECs and RPE cells

In addition to its role in angiogenesis, Rac1 is also a subunit for some NOX isoforms that aggregate to activate NADPH oxidase, a leading generator of superoxide and reactive oxygen species. It is now appreciated that reactive oxygen species (ROS) can be important signaling effectors in addition to being damaging to cell membranes. (NADPH oxidase is also a leading means by which leukocytes fight pathogens, which is essential to the health of the individual. Therefore, this represents another potential reason why inhibiting Rac1 may present safety concerns.) NADPH oxidase is a leading generator of ROS in endothelial cells, and oxidative signaling can cause angiogenesis [97,98]. Furthermore, antioxidants and zinc given to patients with early AMD slow the progression to neovascular AMD [99,100]. In CECs, VEGF activated Rac1 and increased CEC migration. Activated Rac1 also led to ROS generation in CECs, as measured by increased 2’,7’-dichlorofluorescein diacetate (DCFDA) [101] (Figure 2), and ROS were inhibited by NADPH oxidase inhibitor, diphenyleneiodonium, or antioxidant, apocynin, or n-acetyl cysteine. These agents all reduced partially, but significantly, VEGF-induced CEC migration. Together, these findings suggest that Rac1 works in at least two known ways to increase CEC migration, directly by increasing actin cytoskeletal events for cell migration and through NADPH oxidase–generated ROS. Furthermore, inhibiting NOX2-induced NADPH oxidase through the use of apocynin or p47phox knockout mice led to significant reduction in CNV induced by laser.

Inflammation is recognized as important in the pathophysiology of neovascular AMD. One concept proposed in human AMD is that of parainflammation [102], which is a tissue adaptive response to toxins or cellular dysfunction. Parainflammation can occur from stimuli, including oxidative stress, dead cells, and potential changes in the extracellular matrix. A characteristic of parainflammation is the activation of microglia. Activated microglia have been proposed in neovascular AMD and angiogenesis to release angiogenic and inflammatory cytokines, including TNF-α [46,103]. One potential event that activates microglia is the age-related and diet-affected deposition of oxidized lipoproteins within Bruch’s membrane, which trigger activation of microglia to release TNF-α. Therefore, we studied the effects of TNF-α on CECs and RPE cells. In CECs, TNF-α led to increased generation of ROS by NADPH oxidase, as shown by inhibition with apocynin or knockdown of subunit, p22phox [104]. TNF-α increased CEC migration to a similar degree as did VEGF. Apocynin also inhibited TNF-α-induced CEC migration and Rac1 activation. Knockdown of Rac1 inhibited TNF-α-induced CEC migration. TNF-α induced the nuclear factor-kappa B (NF-ҡB) subunit, p-p65, and this was reduced by apocynin but not by knockdown of Rac1. Together, these data provide evidence that TNF-α mediated Rac1 activation via ROS triggered NF-ҡB [104]. Activated Rac1 then caused CEC migration. Intravitreal administration of a TNF-α antibody inhibited CNV in the laser-induced model and reduced labeling of ROS with E06 and of activated Rac1 in the mouse. These data point to a feed-forward loop in which TNF-α induced NADPH oxidase–generated ROS that activated Rac1, which could then activate NADPH oxidase. It also shows crosstalk with VEGF, which activates Rac1 through the guanine nucleotide exchange factor, Vav2 [92], and leads to NADPH oxidase activation (Figure 2).

Rap1aGTP, a potential regulator of RPE cell junctions and Rac1-mediated CEC events

The RasGTPase, Rap1, is important in cell junctions and motility [105,106]. Rap1 has two isoforms, which are 94% similar but are transcribed on different chromosomes. Rap1a is important in barrier properties [105,107], and Rap1b in angiogenesis [108]. As a GTPase, Rap1 also acts as a biologic switch, being activated by GEFs, which causes Rap1 to be GTP bound, and inactivated by GTPase-activating proteins (GAPs) that lead to the inactive GDP-bound Rap1. Activation of Rap1 pharmacologically with 8CPT-2Me-cAMP or with an adenoviral vector that introduced constitutively active Rap1aGTP into cultured CECs reduced TNF-α-induced ROS and Rac1-mediated CEC migration. Activation of Rap1 by chemical 8CPT reduced CNV and ROS generation in RPE cell and choroidal lysates from laser-injured mice, commensurate with the inhibition that occurred from intravitreal TNF-α antibody. One mechanism in which active Rap1a worked was by binding to the NADPH oxidase subunit, p22phox, to reduce the generation of ROS [107,109] (Figure 2).

TNF-α can mediate VEGF expression in RPE cells through NOX4/NADPH oxidase activation. TNF-α-induced VEGF was not inhibited by NF-ҡB inhibitor, Bay117082, but was inhibited by β-catenin inhibitor, XAV939. Either apocynin or XAV939 reduced the interaction of β-catenin with nuclear tissue cell factor-1 (TCF-1)/LEF. Together, these data support the hypothesis that TNF-α leads to β-catenin activation, which translocates to the nucleus to bind with TCF1/LEF and increases VEGF transcription [110]. These findings suggest an interaction between inflammatory mediator, TNF-α, and VEGF expression in RPE cells, mediated through NADPH oxidase. Since VEGF has soluble as well as cell-associated properties, it can access CECs and activate Rac1. VEGF also can act as a chemoattractant for migrating CECs (Figure 2).

In addition to activation of CECs, another important step in type 2 CNV is the loss of integrity of the RPE barrier. The RPE monolayer provides the outer blood retinal barrier that regulates what substances have access to its apical and basal aspects. The concept of tight junctions has evolved from a static barrier to a dynamic situation in which junctions constantly break down and remodel under physiologic stresses. However, pathologic stresses that overwhelm homeostatic mechanisms lead to barrier dysfunction with movement of fluid and substances across the barrier to regions that are usually protected. In the case of AMD, cells and substances, such as VEGF, can access the photoreceptors in the sensory retina. Rap1a is important in baseline RPE barrier properties [105], and Rap1b may be important in reformation of RPE cell junctions following a stress [107], such as breakdown of RPE cell junctions following calcium EGTA. However, because of Rap1b’s role in angiogenesis, Rap1b may be a less favorable therapeutic target than Rap1a [108].

ROS have been shown to reduce RPE barrier integrity by phosphorylation of cadherin-β-catenin complexes. Activation of Rap1a reduced ROS-mediated RPE barrier compromise [109]. With reduced RPE barrier integrity, VEGF can access the neural retina and potentially attract activated CECs to migrate and form CNV. Mice given intravitreal Rap1 activator, 8CPT, had significantly reduced laser-induced CNV [107]. Rap1b knockout mice treated with 8CPT have activated Rap1a only and also showed significant reduction in laser-induced CNV [109].

Role of extracellular matrix

It is essential to consider the role of the extracellular matrix (ECM) in cell migration. The ECM is the matrix onto which cells attach and engage during migration. Many studies have reported changes in the ECM and Bruch’s membrane that occur with aging [38], cigarette smoking [111], and fat-feeding [41]. Using a murine model of Cfh−/− mice aged 2 years and then fat-fed, Bowes-Rickman et al. found that CFH played a role in lipoprotein turnover in Bruch’s membrane [28]. This finding aligned with clinicohistopathologic evidence by Curcio et al. [15,16] that increased deposition of cholesterol esters and oxidized cholesterol occurs in AMD. In addition to dietary sources for cholesterol to the choriocapillaris, there is evidence that cholesterol can be produced in the retina [112]. One predominant component of oxidized lipoprotein debris that accumulates in Bruch’s membrane and the extracellular matrix in neovascular AMD is 7-ketocholesterol, which can cause angiogenesis [113]. Oxidized lipoproteins activate microglia, which release cytokines, including TNF-α.

Taken together, one scenario for the complexity of events in neovascular AMD is proposed. With increasing age and potentially through diet, oxidized lipoprotein debris and 7-ketocholesterol accumulate in ECM/Bruch’s membrane in AMD, and inflammation is increased as 7-ketocholesterol activates microglia to release cytokines. This event may occur earlier in life in patients with genetic predisposition, such as with complement factor H (Y402H) or ARMS2/HTRA1. TNF-α mediates β-catenin-induced expression of VEGF in RPE cells, which is released preferentially from the basal aspects [114] of the RPE monolayer. Secreted VEGF can affect CEC migration, potentially by permitting engagement with integrins in ECM components [115]. Secreted VEGF also activates CEC Rac1 and initiates CEC activation and migration. CECs that contact RPE cells are further activated via RPE cell-associated VEGF189. VEGF and TNF-α crosstalk leads to further activation of CEC Rac1 in a feed-forward loop. Generated ROS also reduce RPE barrier integrity and permit secreted VEGF to enter the neural retina where it attracts CECs to migrate, proliferate, and form CNV (Figure 2). In addition, age-related increases in CCL11 in RPE cells and CCR3 in CECs exacerbate Rac1 activation in CECs [93].

Additional questions exist. The events that must occur within CECs to migrate require coordination of multiple intracellular signaling cascades. Likewise, numerous events within RPE cells must be coordinated. CECs must engage the extracellular matrix and be affected by changes in its composition over the course of the development of AMD.

Conclusions

Neovascular AMD remains a leading cause of vision loss worldwide despite the use of antiangiogenic agents in AMD. Recent advances in diagnosis and treatment include new imaging modalities, new anti-VEGF agents, and greater understanding of the roles of feed-forward signaling loops and of cell–cell and cell–extracellular matrix interaction. Future therapeutic directions may involve regulating instead of abolishing pathologic signaling causing biologic events. Examples include “quieting” CECs activated to migrate, strengthening the barrier integrity of the RPE monolayer, and addressing the accumulation of debris within the extracellular matrix in the RPE monolayer and Bruch’s membrane. Clinical imaging modalities require additional study as a means of evaluating the efficacy of anti-VEGF treatment and to address ways to reduce geographic atrophy.